|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 MULTIVARIATE DATA ANALYSIS STRATEGIES FOR BIOLOGICAL APPLICATIONS OF MATRIX ASSISTED LASER DESORPTION/IONIZATION MASS SPECTROMETRIC IMAGING By ROBERT FRANCIS MENGER A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 Robert Francis Menger
3 To my friends and family
4 ACKNOWLEDGMENTS First and foremost, I wish to thank my advisor, Dr. Richard A. Yost. Dr. Yost truly exemplifies the passion of a lifelong scientist, and I believe a bit of this passion transfers to each and every graduate student he mentors. Next, I would like to thank all of the current and past members of the Yost group. I would especially like to thank Dan Magparangalan for taking me under his wing during my first year in the lab. In addi to thank Kyle Mino Whitney Stutts Wei Tsai Noelle Elliott Mike Costanzo, Raquel Hendershot, and Dave Pirman for all of their support in and out of the lab. As soon as I decided to join this group, I knew that it was a family impersonating a lab. I would also like to acknowledge all of my collaborators. Without their gracious donation of samples and valuable insight, this dissertation would not have been possible. I would especially like to thank Dr. David A. Ford Dr. Dhanalakshmi S. Anbukumar, a nd Dr. Brad Wacker at Saint Louis University Dr. John Bowden at the Medi cal University of South Carolina, and Dr. Art Edison and Chaevien Clendinen at the University of Florida for the donation of samples and their valuable suggestions throughout my resea rch career at Florida. During my time in graduate school, I had the privilege to study at the FOM Institute also like to acknowledge the Heeren group members, especia lly Andras Kiss, Don Smith, Nadine Mascini, Julia Jungmann, and Gert Eijkel for all of the help during my time in the Netherlands.
5 the scenes people in the UF like to thank Dr. Ben Smith, Lori Clark, and Antoinette Knight. E ach and every y a smile and a friendly word. Finally, I would like to thank all of my friends and family. Ever yone has always been extremely supportive, giving me the means and encouragement to pursue this career. Without you guys, I would not be here today.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 10 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 A BRIEF INTRODUCTION TO MASS SPECTROMETRIC IMAGING AND MULTIVARIATE DATA ANALYSIS ................................ ................................ ......... 17 Mass Spectrometric Imaging ................................ ................................ .................. 17 Concept ................................ ................................ ................................ ............ 17 Microprobe and microscope mode imaging ................................ ...................... 17 Ionization sources ................................ ................................ ............................ 18 MALDI ................................ ................................ ................................ ........ 18 SIMS ................................ ................................ ................................ .......... 21 DESI ................................ ................................ ................................ .......... 23 Emerging ionization sources ................................ ................................ ...... 23 Mass analyzers ................................ ................................ ................................ 26 Ion traps ................................ ................................ ................................ ..... 26 Time of flight mass analyzers ................................ ................................ .... 31 Fourier transform mass analyzers ................................ .............................. 33 Compound identification: accurate mass ................................ .......................... 36 Compound identification: tandem mass spectrometry ................................ ...... 37 High energy vs low energy CID ................................ ................................ 37 Tandem in space vs tandem in time MS 2 ................................ ................... 38 MS n on the Thermo LTQ XL ................................ ................................ ....... 38 Data processing ................................ ................................ ............................... 40 Image generation ................................ ................................ ....................... 40 Normalization ................................ ................................ ............................. 41 Multivariate Data Analysis ................................ ................................ ....................... 42 Principal component analysis ................................ ................................ ........... 43 Partial least squares discriminant analysis ................................ ....................... 46 Applying multivariate data analysis to MSI ................................ ....................... 46 Scope of the Dissertation ................................ ................................ ........................ 47 2 MALDI MASS SPECTROMETRIC IMAGING OF CARDIAC TISSUE FOLLOWING CORONARY ARTERY LIGATION INDU CED MYOCARDIAL INFARCTION ................................ ................................ ................................ .......... 55 Introduction ................................ ................................ ................................ ............. 55
7 Experimental ................................ ................................ ................................ ........... 58 Chemicals and reagents ................................ ................................ ................... 58 Biological sample preparation ................................ ................................ .......... 59 Instrumentation ................................ ................................ ................................ 60 Standard analysis ................................ ................................ ............................. 61 Tissue analysi s ................................ ................................ ................................ 61 Statistical analysis ................................ ................................ ............................ 62 Enzymatic digestion with PLA 2 on tissue ................................ .......................... 64 Results ................................ ................................ ................................ .................... 64 Identification of infarcted myocardium ................................ .............................. 64 TTC staining ................................ ................................ ............................... 64 Creatine ................................ ................................ ................................ ..... 65 Phospholipase A 2 ................................ ................................ ....................... 66 Identification of At Risk Myocardium: Triacylglycerides ................................ .... 68 Guided multivariate data analysis and identification of tissue specific markers ................................ ................................ ................................ ......... 69 Infarcted and perfused myocardium ................................ ........................... 69 At risk myocardium and perfused myocardium ................................ .......... 77 At risk myocardium and infarcted myocardium ................................ .......... 78 PCA of all three regions of interest ................................ ............................ 78 Comparing multiple biological samples ................................ ...................... 79 Unbiased multivariate data analysis and tandem mass spectrometry .............. 81 Conclusions ................................ ................................ ................................ ............ 83 3 DEVELOPMENT OF 9 AMINOACRIDINE AS A DUAL MODE MALDI MATRIX FOR SMALL MOLECULE MASS SPECTROMETRIC IMAGING STUDIES OF MYOCARDIAL INFARCTION ................................ ................................ ............... 122 Introduction ................................ ................................ ................................ ........... 122 Experimental ................................ ................................ ................................ ......... 124 Chemicals and reagents ................................ ................................ ................. 124 Extraction of 9 AA free base ................................ ................................ ........... 125 UV/Vis absorption spectrophotometry ................................ ............................ 125 Biological sample preparation ................................ ................................ ........ 126 Mass spectrometry and imaging ................................ ................................ ..... 127 Statistical analysis ................................ ................................ .......................... 127 Results ................................ ................................ ................................ .................. 128 Standard characterizatio n ................................ ................................ ............... 128 Phosphatidylcholine (PC) ................................ ................................ ......... 12 9 Phosphatidylethanolamine (PE) ................................ ............................... 130 Phosphatidylserine (PS) ................................ ................................ .......... 131 Phosphatidylinositol (PI) ................................ ................................ .......... 132 NAD + a nd NADH ................................ ................................ ...................... 132 Comparison of pure 9 AA free base and 9 AA extract ................................ ... 134 Positive mode MS imaging of a rat coronary artery ligation model of myocardial infarction ................................ ................................ ................... 135 Detected analytes in positive mode ................................ ......................... 135
8 Principal component analysis of pe rfused and infarcted myocardium ...... 136 Dicarboxylacylcarnitines as potential blood borne biomarkers of myocardial infarction ................................ ................................ ............. 138 Negative mode MS imaging of a rat coronary artery ligation model for myocardial infarction ................................ ................................ ................... 140 Principal component analysis ................................ ................................ ... 140 MS n identification and imaging ................................ ................................ 142 Phospholipids ................................ ................................ ........................... 142 Nucleotides and nucleotide sugars ................................ .......................... 147 Conclusions ................................ ................................ ................................ .......... 149 4 MALDI MS METABOLIC PROFILING OF CAENORHABDITIS ELEGANS .......... 192 Introduction ................................ ................................ ................................ ........... 192 Experimental ................................ ................................ ................................ ......... 193 Chemicals and reagents ................................ ................................ ................. 193 MALD I MS of nematodes ................................ ................................ ............... 194 Data processing and analysis ................................ ................................ ......... 195 MALDI MSI and tandem MS ................................ ................................ ........... 195 Results ................................ ................................ ................................ .................. 196 MALDI MS profiling of nematode cuticles ................................ ....................... 196 Principal component analysis ................................ ................................ ......... 198 MALDI MSI and tandem MS ................................ ................................ ........... 199 Conclusions ................................ ................................ ................................ .......... 202 5 SUMMARY AND FUTURE WORK ................................ ................................ ....... 218 Summary ................................ ................................ ................................ .............. 218 Future Work ................................ ................................ ................................ .......... 220 Time course and repe rfusion studies of myocardial infarction ........................ 220 Correlation of MSI and LC MS data ................................ ............................... 222 Improving MALDI MSI spatial resolution on the MALDI Thermo LTQ XL via reduction of the laser spot size ................................ ................................ .... 222 Outlook ................................ ................................ ................................ ................. 225 LIST OF REFERENCES ................................ ................................ ............................. 229 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 238
9 LIST OF TABLES Table page 2 1 Statistical significance testing for various ions in in farcted and perfused myocardium ................................ ................................ ................................ ........ 95 2 2 MS 2 imaging experiments performed ion the [M+Na] + ions from intact PC and PE ions containing stearic acid in the sn 1 po sition of the glycerol backbone .. 100 4 1 List of compounds that correlate strongly with the wild type N2 (positive loadings coefficient) or fat6;fat7 (negative loadings coefficient) nematodes ..... 216
10 LIST OF FIGURES Figure page 1 1 Representation of the MALDI process ................................ ................................ 49 1 2 Positive portion of the first stability region for a quadrupole mass filter .............. 50 1 3 Sample stability diagram expressed in terms of a z and q z for a QIT ................... 51 1 4 Schematic of the Thermo Scientific LTQ XL with a MALDI ionization source ..... 52 1 5 Thermo Scientific LTQ XL tandem MS n scan function ................................ ........ 53 1 6 Example of PC A applied to 2 variable dataset ................................ ................... 54 2 1 MSI analysis of cardiac ti ssue following ligation surgery ................................ .... 88 2 2 MS image of m/z 132 normalized to the TIC from control cardiac tissue ............ 89 2 3 Mass spectra of the lipid region collected in di fferent regions of myocardium .... 90 2 4 MALDI MS spectra acquir ed from sham surgery myocardium ............................ 91 2 5 Example of TAG identification and localization ................................ ................... 92 2 6 Multivariate data analysis of in farcted and perfused myocardium ...................... 93 2 7 PCA loadings plots dictating separation between in farcted and perfused myocardium ................................ ................................ ................................ ........ 94 2 8 MS 2 spectrum using CID of m/z 52 0 from infarcted cardiac tissue ..................... 96 2 9 Example of MS n ident ification and imaging of an LPC ................................ ........ 97 2 10 MS 2 spectrum of m/z 848 using PQ D from infarcted cardiac tissue ................... 98 2 11 MS images normalized to the TIC from a heart followi ng LAD coronary artery ligation ................................ ................................ ................................ ................ 99 2 12 MS 2 PC fragment ion intensity (loss of trimethylamine) for four common intact PCs in infarcted (blue ) and perfused (red) myocardium ................................ ... 101 2 13 MS 2 PE fragment ion intensity (loss of ethenamine) for four common intact PEs in infarcted (blue ) and perfused (red) myocardium ................................ ... 102 2 14 MS 2 images of select abundant ceramides in cardiac tissue fol lowing coronary artery ligation ................................ ................................ ..................... 103
11 2 15 MALDI MS 2 spectrum of m/z 348 obtained from perfused myocardium following LAD ligation surgery ................................ ................................ .......... 104 2 16 MALDI MS 2 spectrum of m/z 162 obtained from perfused myocardium following LAD ligation surgery ................................ ................................ .......... 105 2 17 PCA scores plot demonstrating the separation between at risk (red) and perfused (green) myocardium ................................ ................................ ........... 106 2 18 Principal component 1 loadings plot dictating the separation between perfused and at risk myocardium ................................ ................................ ..... 107 2 19 PCA scores plot demonstrating the separation between at risk (red) a nd infarcted (green) myocardium ................................ ................................ ........... 108 2 20 Principal component 1 loadings plot dictating the separation between pe rfused and infarcted myocardium ................................ ................................ 109 2 21 PCA scores plot generated from sampling the infarcted (green), perfused (blue), and at risk myocardium (red) fol lowing coronary artery ligation ............ 110 2 22 Principal component 1 loadings plot following PCA analysis of all three regions of myocardium ................................ ................................ ..................... 111 2 23 Principal component 2 loadings plot following PCA analysis of all three regions of myocardium ................................ ................................ ..................... 112 2 24 PCA scores plot generated from analy sis of three biological samples ............. 113 2 25 PCA scores plot generated from analysis of multiple biological samples ......... 114 2 26 Loadings plot of principal component 1 generated from PCA of multiple biological samples including three diffe rent regions of myocardium ................. 115 2 27 Loadings plot of principal component 2 generated from PCA of multiple biological samples including three different regions of myocardium ................. 116 2 28 Graphical representation of loadings from PCA of multiple biological samples with the inclus ion of all three tissue regions ................................ ..................... 117 2 29 Principal component scores im ages from unbiased PCA analysis ................... 118 2 30 Loadings p lots from unbiased PCA analysis ................................ .................... 119 2 31 Loadings plot of principal component 19 (PC 19) from unbiased PCA analysis ................................ ................................ ................................ ............ 120 2 32 MS 2 spectrum of m/z 400 collected from cardiac tissue following experimental coron ary artery ligation surgery ................................ ................... 121
12 3 1 Potential negative mode MALDI mat rices for MSI of cardi ac tissue ................. 153 3 2 MS 2 spectrum of m/z 494 from an LPC 17:0 standard collected in negative mo de using 9 AA as a MALDI matrix ................................ ................................ 154 3 3 Negative mode MALDI MS 2 spectrum of m/z 714 from a sy nthetic PE (16:0/18:2) standard ................................ ................................ ......................... 155 3 4 Example of MS n in negative mode for a PS ion ................................ ................ 156 3 5 MALDI MS spectrum of NADH in negative mo de using 9 AA as a MALDI matrix ................................ ................................ ................................ ................ 157 3 6 Negative mode MALDI MS 2 spectrum of m/z 664 obtained from a n NADH standard ................................ ................................ ................................ ........... 157 3 7 Negative mode MALDI MS spectrum of NAD+ standa rd using 9 AA as a MALDI matrix ................................ ................................ ................................ .... 158 3 8 Negative mode MALDI MS 2 spectrum of m/z 540 obtained from an NAD standard u tilizing 9 AA as a MALDI matrix ................................ ....................... 158 3 9 MALDI MS spectra collected in positive ionization mode from the 9 AA extract (Top) and the commercia lly available freebase (Botto m) ...................... 159 3 10 MALDI MS spectra collected in negative ionization mode from the 9 AA extract (Top) and the commercia lly available freebase (Bottom) ...................... 160 3 11 Averaged MS spectra from cardia c tissue following LAD ligation ..................... 161 3 12 UV Vis absorption spectrum of 2,5 dihydroxybenzoic acid dissolved in 70:30 MeOH:H 2 O (v/v) ................................ ................................ ............................... 162 3 13 UV Vis absorption spectrum of 9 aminoacridine dissolved in 70:30 EtOH:H 2 O (v/v) ................................ ................................ ................................ .................. 162 3 14 PCA scores plots separating infarcted myocardium (red triangles) and perf used myocardium (green crosses) ................................ ............................. 163 3 15 PCA loadings plots from analysis of in farcted and perfused myocardium ........ 164 3 16 MS image s from infarcted cardiac tissue ................................ .......................... 165 3 17 MS images generated for low molecular weight ions from ligated hearts ......... 166 3 18 MS 2 spectrum of m/z 204 obtained from perfused cardiac tissue coa ted with 9 AA as a MALDI matrix ................................ ................................ ................... 167
13 3 19 MS 2 spectrum of m/z 248 obtained from perfused cardiac tissue coa ted with 9 AA as a MALDI matrix ................................ ................................ ................... 168 3 20 MS 2 spectrum of m/z 262 obtained from perfused cardiac tissue coa ted with 9 AA as a MALDI matrix ................................ ................................ ................... 169 3 21 Images of a heart fol lowing coronary artery ligation ................................ ......... 170 3 22 Negative mode MALDI MS spectra ................................ ................................ .. 171 3 23 MS images of m/z 540 and 1207 ................................ ................................ ...... 172 3 24 PCA scores plot from negative mode MALDI MS analysis of infarcted (red triangles) and perf used myocardium (green crosses) ................................ ....... 173 3 25 1 dimensional loadings plot from principal component 1 (PC 1) ....................... 174 3 26 Negative mode MS images of various ions ................................ ...................... 175 3 27 PCA scores plot from negative mode MALDI MS analysis of infarcted (red triangles) and perfused myocardium (green crosses) ................................ ....... 176 3 28 1 dimensional PCA loadings plot of principal component 1 (PC 1) generated from analysis of infarcted, at risk, and perfu sed myocardium ........................... 177 3 29 1 dimensional PCA loadings plot of principal component 2 (PC 2) generated from analysis of infarcted, a t risk, and perfused myocardium ........................... 178 3 30 MS images of various ions loading either on principal component 2, which demonstrates po sitive correlat ion with the at risk myocardium ........................ 179 3 31 Negative mode MALDI MS 2 spectra of various high mass ions ........................ 180 3 32 Fragmentation pattern for TLCL ................................ ................................ ....... 181 3 33 Summed MS image of m/z 1447, 1469, and 1485 norm alized to the total ion current ................................ ................................ ................................ .............. 182 3 34 MS 2 sp ectrum of m/z 945, identified as the [M+Na 2H] of dilysocardiolipin ..... 183 3 35 MS 2 analysis of dilysocardiolipin ................................ ................................ ....... 184 3 36 MALDI MS 2 spectrum of m/z 599, identified as the [M H] ion of LPI 18:0 ........ 185 3 37 Negative mode MALDI MS 2 spectrum of m/z 480 ................................ ............ 186 3 38 Negative mode MALDI MS 2 spectrum of m/z 508, identified as the [M CH 3 ] of LPC 18:0 ................................ ................................ ................................ ....... 187
14 3 39 Negative mode MALDI MS 2 spectrum of m/z 426 ................................ ............ 188 3 40 Negative mode MALDI MS n identification and imaging of UDP from myocardium ................................ ................................ ................................ ...... 189 3 41 Negative mode MALDI MS 2 spectrum of m/z 565 ................................ ............ 190 3 42 MALDI MS 2 spectrum of m/z 606 detected from myocardium .......................... 191 4 1 Representative MS spectra from C. elegans utilizing DHB as a MALDI matrix 205 4 2 Representative MS spectra from C. elegans u tilizing 9 AA as a MALDI m atrix 206 4 3 PCA scores plot describing the separation between wild type N2 (green crosses) and fat6;fat7 double mutant (red triangles) C. elegans ...................... 207 4 4 PCA loadings plot from principal component 1 dictating the separation between the N2 (positive) and fat6;fat7 (negative) C. elegans strains .............. 208 4 5 Principal component analysis scores plot (PC1 vs. PC4) detailing the separation between the wild t ype N2 and daf 22 gene knockout ..................... 209 4 6 PCA loadings plot from principal component 4 dictating the separation between the N2 (positive) and daf 22 (negative) C. elegans strains ................ 210 4 7 MS images of wild type (Left) and mutants (Right) for various ions (structures shown) det ected from the nematode cuticle ................................ ..................... 211 4 8 MALDI MS 2 spectrum obtained from a wild type N2 nematode ........................ 212 4 9 MALDI MS n structural elucidation of m/z 580, identified as the [M+K] + of LPC 20:5 ................................ ................................ ................................ .................. 213 4 10 MALDI MS 3 spectrum of m/z ed from a wild type N2 nematode ................................ ................................ ................................ ......... 214 4 11 MALDI MS 2 spectrum of m/z 258 collected from a mutant fat6;fat7 nematode 215 4 12 MS 2 elucidation of AMP ................................ ................................ .................... 217 5 1 Standard optical Configuration of the MALDI source interfaced with the Thermo Scientific LTQ XL ................................ ................................ ................ 226 5 2 Representat ion of a Galilean beam expander ................................ .................. 227 5 3 Schematic of LTQ optical configuration with Galilean beam expander implemented on the optical rail ................................ ................................ ......... 228
15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy MULTIVARIATE DATA ANALYSIS STRATEGIES FOR BIOLOGICAL APPLICATIONS OF MATRIX ASSISTED LASER DESORPTION/IONIZATION MASS SPECTROMETRIC IMAGING By Robert Francis Menger M ay 2013 Chair: Richard A. Yost Major: Chemistry Matrix assisted laser desorption/ionization ma ss spectrometric imaging (MALDI MSI) is a label free imaging technique utilized to determine the spatial distribution of biomolec ules in tissue. Although MALDI MSI is a proven technique for targeted analysis, global analyses such a eluded researchers. Initially, the difficulty in performing global analyses stemmed from instrumentation; however, modern instrumentation allows for the simultaneous detection of thousands to millions o f analytes from a single interrogated position. Although datasets that contain approximately 2,000,000 mass features can be collected quite rapidly (~1 10 hours for a 1 cm 2 sample), processing this dataset from a univariate perspective takes orders of mag nitude longer. Thus, data processing is currently the major hindrance to performing global MSI analyses. (e.g., LC MS) provide efficient methods to reduce the dimensionality of complex datasets. One such technique, principal component analysis (PCA), performs data reduction by determining the axes of greatest variance within a dataset. The result is
16 that in the ideal scenario, variability between a case and control, or diff erent regions of tissue for MSI, can be differentiated using a single principal component. Following rotation of the multivariate dataset, the original samples are projected onto principal component axes, and the separation of samples in principal compone nt space can be visualized using a 2 or 3 dimensional scores plot. The mathematical formula for the principal component axes is a summation of the original variables multiplied by their respective weighting factors or loadings Thus, the loadings can b e inspected to This dissertation describes PCA methodologies that strive to reduce analysis time, while simultaneously extracting the most relevant information from an MSI dataset. Following a brief introduction to MSI and multivariate data analysis, the develo ped methodologies are utilized for three biological applications: 1) positive mode MSI analysis of a model system for myocardial infarction, 2) positive and negative mode MSI analysis of myocardial infarction using 9 aminoacridine as a MALDI matrix, and 3) the identification of biomolecules from Caenorhabditis elegans and related species.
17 CHAPTER 1 A BRIEF INTRODUCTION TO MASS SPECTROMETRIC IMAGING AND MULTIVARIATE DATA ANALYSIS Mass Spectrometric Imaging Concept Mass spectrometric imaging (MSI) is an analytical technique utilized to generate chemically selec tive images by directly interrogating a sample surface. 1 To date, two imaging modalities exist: microprobe mode imaging and microscope mode imaging. Despite the imaging modality, the general concept remains the same. In brief, an ionization source is utilized to generate ions from a discre t e position (or area) on a sample surface. The ions are then transported to a mass analyzer for immediate detection or alternatively, isolation, fragmentation, and subsequent detection. The resulting mass spectrum is stored and a relative X,Y position is associated with the mass spectrum. Following detection, an X,Y stage is utilized to move the sample by a specified raster step size, and the process is repeated until the entire tissue has been interrogated. An image can then be generated by extracting the ion intensity for a specified mass to charge ratio ( m/z ), and plotting the intensity versus the X,Y position on the sample Microprobe and microscope mode imaging As previously stated, two modalities exist for MSI. The most common imaging modality is microprobe mode imaging, wherein an ionization source is utilized to interrogate a relatively small area of the target sample. 2 The generated ions are the n mass analyzed, and all of the ions are treated as being from the same position on the target sample. Following collection of the first mass spectrum, the sample is rastered with respect to the stationary ionization source, and the ionization and mass an alysis
18 process is repeated. In contrast, microscope mode imaging utilizes an ionization source to interrogate a larger area of the target sample. As before, the ions are transferred to the mass analyzer following ionization; however, in this modality, th e spatial orientation of the ions is maintained from ionization to the detector for mass analysis 2 Both modalities have a number of inherent advantages and pitfalls. First, in microprobe techniques, the spatial resolution of the imaging experiment is often limited by the ionization source. With microscope techniques, the spatial resolution is often limited by the ion optics and the resolut ion of the position detector. Alt ho ugh better spatial resolution can be achieved with microscope techniques, the instrument cost and sensitivity may suffer. Thus, f or the purposes of this dissertation, all MSI from this point forward will refer to microprobe mode imaging. Ionization source s The initial step in the MSI process is that of ionization, wherein neutral molecules from the sample surface are ionized and subsequently transferred to the mass analyzer. A number of ionization sources are currently commercially available with imaging capable mass spectrometers. Although c hoice of a suitable ionization source is largely dependent upon the sam ple and analytes to be detected, t he three most common ionization sources for MSI are matrix assisted laser desorption/ionization (MALDI), seconda ry ion mass spectrometry (SIMS), and desorption electrospray ionization (DESI). Each of these ionization sources present unique advantages, and will be discussed further below. MALDI Matrix assisted laser desorption/ionization (MALDI) first reported by K aras et al. 3, 4 and Tanaka et al. 5 is currently the most common ionization source utilized in MSI.
19 fragmentation) that utilizes a UV l aser to induce both desorption and ionization of analytes The most common lasers utilized for UV MALDI are a N 2 gas laser ( =337 nm) or a frequency tripled Nd:YAG solid state laser ( =355). Typically, N 2 lasers offer pulse rates between 10 and 100 Hz, w hereas Nd:YAG lasers offer pulse rates around 1 kHz. This pulse rate often dictates the mass analyzer utilized with the respective MALDI source. For instance, N 2 lasers are often paired with relatively slow mass analyzers (e.g., ion traps), whereas Nd:YA G lasers are often paired with time of flight (ToF) mass analyzers. Gas lasers, including a N 2 laser, are also relatively inexpensive compared to solid state lasers; 6 thus, when the repetition rate is not needed to accelerate experimentation, gas lasers are often chosen. The target sample is coated w ith a UV absorbing matrix to aid in both desorption and ionization of the analytes from the sample surface. Alt hough there are various MALDI matrices commercially available, most are small organic acids (for positive mode) or bases (for negative mode) tha t absorb strongly in the UV. Choice of a MALDI matrix is largely dependent upon the analyte of interest. For instance, sinapinic acid is commonly utilized for large intact proteins, whereas cyano 4 hydroxycinnamic acid is commonly utilized for smaller peptides. 7 The matrix is typically saturated in a solvent system that can readily extract analytes of interest from the target Upon evaporation of the matrix solvent, matrix and analyte cocrystals are formed. Irradiation of these cocrystals with a UV laser generates singly charged ions over a wide m/z range (Figure 1 1)
2 0 The ionization mechanism for MALDI is still widely debated throughout the MS community; however, two models have been proposed: the lucky survivor model 8 and the gas phase protonation model 9 The lucky survivor model proposes that analytes are incorporated into the matrix and analyte cocrystals as charged species. Upon desorption, most of the highly charged matrix/ analyte clusters become neutralized or undergo charge reduction by counter ions of opposite charge prese nt in the plume ; however, a small portion of singly charged 8 This mechanism was further refined later to also account for anion formation, in addition to cation formation. 10 A lternately, the gas phase protonation model suggests that analytes are not present as charged species in the cocrystals. Instead, matrix radical cations are proposed to be generated following multiphoton ionization Following formation of the matrix radi cal cation a proton can be efficiently transferred to form a protonated matrix ion. Subsequently, transfer of a proton from the protonated matrix ion to the neutral analyte molecule will be dependent upon on the proton affinities for both species, and is thought to be an thermodynamically favorable process in many cases 9 Recently, there has even been some evidence for a unified theory that combines both the lucky survivor and gas phase protonation models. 11 Regardless of the ion ization mechanism in MALDI, the matrix application process greatly influences the quality of spectra obtained. In general, the ideal application process will maximize the extraction of a nalytes from the sample, while minimizing the latera l diffusion of those analytes for MSI The most conventional application technique utilized for coating is pneumatic spraying, including the use of nebulizers and TLC sprayers. 12 Pneumatic spraying p rovides a rapid method that produces relatively small
21 crystal sizes; however, the majority of devices utilized for delivering matrix in this manner are manually controlled, possibly introducing inhomogeneity during coa ting. A number of commercially available matrix coating apparatuses have been developed to attempt to reduce inhomogeneity including the Bruker ImagePrep (pneumatic spraying) and the Shimadzu Chemical Inkjet Printer, or ChIP, (microspotting). 7 Although fully automated, these systems prove to be rather expensive. A promising compromise between cost, automation, a nd crystal size appears to be the use of commercial office inkjet printing technologies for coating MALDI matrix. A modified inkjet printer developed in the Yost lab produced smaller crystal sizes and a more homogenous matrix coating than user controlled pneumatic spraying devices. 12 Additionally, the price of a standard inkjet printer is inexpensive relative to the ImagePrep or the ChIP. The biggest limitation of commercial inkjet printers appears to be solvent com patibility; standard inkjet printers currently utilize plastic cartridges and print heads that are incompatible with certain solvents (e.g., chloroform). The most recent matrix application technique to gain interest in the MSI community is sublimation. 13 This technique produce s a smaller crystal size relative to pneumatic spraying at the cost o f reduced time for matrix analyte inter action and analyte extraction. 7 SIMS Secondary ion mass spectrometry, or SIMS, is the oldest ionization technique commonly utilized in MSI. In SIMS, reagent ions serving as a focused primary ion beam are directed at a sample surface. The primary ions serve to eject secondary ions from the sample surface for subsequent detection by the mass analyzer Generation of the primary ion beam can be performed with a number of sources including liquid metal ion guns (e.g ., Ga + In + or Au + ), solid state guns (e.g., Cs + ) or C 60 + cluster sources 14
22 Depending upon the ion dose delivere d to the surface, SIMS can be performed in either static or dynamic modes. For the purposes of this discussion, static SIMS can be considered ion doses lower than 10 12 ions/cm 2 14 When operated in this mode, the primary ions do not penetrate farther than the outermost monolayer of the sample surface. 14 In contrast, dynamic SIMS uses a relatively high ion dose resulting in ejection of the secondary ions further than the outermost monolayer. Relative to MALDI, SIMS is considered a h arsh ionization technique The relatively high energy interaction between the primary ion bea m and the sample surface induces in a large degree of source fragmentation, resulting in reduced selectivity 2 To illustrate this point, diverse compound classes such as phosphatidylcholine lipid species are fragmented to one or two fragment ions related to the common headgroup 15 and yield little to no information regarding the intact molecular weight (MW) of the species. Furthermore, u nder typical opera ting conditions, the practical upper mass limit for SIMS is ~ m/z 1000. 2 T he addition of a matrix molecule to the surface (similar t o MALDI) can serve to reduce the amount of source fragmenta tion thereby increasing the effective upper mass limit of the technique 16 Despite the extensive source fragmentation and relatively low mass limit, SIMS offers a number of advantages over MALDI. Perhaps the most significant advantage is enhanced spatial resolution. In theory, the spatial resolution in MALDI is diffraction limited; however, practical considerations with the optical configuration usually limit practical spatial resolution is limited by the focusing of the primary ion beam and the lateral interaction of the primary ion beam with the surface ; modeling has suggested that molecules are
23 desorbed 5 10 nm from primary ion beam 17 Currently practical spatial resolutions in SIMS are routinely reported as being less than 50 nm 15 DESI Desorption electrospray ionization (DESI ) is an ambient condition ionization technique for MSI that was first reported by 18 In this technique, an electrospray needle is utilized to spray charged droplets at a sample surface. Upon striking the surface, t hese charged droplets cause simultaneous desorption and ionization of analyte molecules from the sample surface. Interestingly, the ionization mechanis m of DESI appears to vary depending on the analyte. Certain analytes (e.g., pept ides and proteins) yield multiply charged ions suggesting a solution phase ionization mechanism similar to electrospray ionization (ESI), whereas others yield singly charged ions, suggesting a charge transfer ionization mechanism similar to atmospheric pr essure chemical ionization (APCI). 19 Relative to the other traditional MSI ionization tec hniques, DESI appears to be the most amenable for ambient sampling. There is no need to place the sample under any kind of vacuum, and there is little to no sample preparation (excluding sectioning and mounting) that need be performed. Despite these adva ntages, DESI demonstrates a spatial resolution of approximately 200 m, thereby limiting DESI MSI applications. Emerging ionization sources Recently, a number of ionization sources have been reported in the literature that may prove valuable for MSI resear ch. These ionization sources include laserspray ionization (LSI) laser ablation electrospray ionization (LAESI) and the liquid microjunction Alt hough a detailed description of these sources is beyond the scope of this dissertation, each of the emergin g sources will be described briefly.
24 LSI, first reported by Sarah Trimpin and coworkers in 2010, 20 utilizes a UV laser to irradiate the back side of a thin sample mounted on either a UV transparent cover slip or mi croscope slide This configuration is commonly referred to as the transmission geometry, and was used on early microprobe instruments such as the LAMMA 500. 21 Similar to atmospheric pressure (AP) MALDI, 22 t he mounted sample is coated with a MALDI matrix and held at atmospheric pressure If the sample is sufficiently thin, the analyte/matrix cocrystals will be ablated by the laser, and the subsequent ions will be transferred to the mass analyzer by means of a heated capillary and ion optics Alt hough the only deviation from standard AP MALDI experiments is the transmission geometry, vastly different spectra are produced with LSI. In i nitial studies, lysozyme, a large protein, was analyzed with 2,5 dihydroxybenzoic acid (DHB) as the MALDI matrix using the transmission geometry and a high laser fluence, producing highly charged ions similar to those obtained with ESI. 20 In contrast, the same sample analyzed in the typical reflective geometry and a low laser fluence produce s predominantly singly charged ions. Later, it was determined that neither the geometry nor the laser fluence are relevant fo r generating multiply charged ions. 23 Instead, it was determined that a minimal or zero potential difference between the target and the sample inlet and a heated region prior to mass analysis are both required for the observation of these multiply charged ions. 23 Based on the se observations, an ionization mechanism similar to ESI has been proposed, in which charged matrix/analyte clusters are partially desolvated releasing matrix molecules until a highly charged analyte droplet is obtained 23
25 Regardless of the ionization mechanism LSI offers a couple of interesting advantages over traditional ionization sources. First, in the transmission geometry, the optics for the laser can be brought sufficiently close to the back side of the target slide or cover slip. The implication here is tha t the principles of near field optics can be applied, potentially allowing the MALDI spot size to approach or even sur pass, the diffraction limit, suggesting the potential for submicron imaging Second, the ability to produce multiply charged ions permit s the use of ion traps for a wider range of analytes. Thus, intact hig h molecular weight proteins, which normally require digestion prior to analysis on io n trap mass spectrometers, may be brought into the effective m/z range for these mass analyzers. LAESI combines the use of a laser to ablate analytes from a sample surface and an orthogonal ESI source for post ionization. 24 This technique offers a number of interesting advantages similar to L SI. The utilization of postionization by ESI offers the ability to generate multiply charged ions, permitting the analysis of high molecular weight species. Furthermore, as a mid IR laser is often used for ablation, water be analyzed without sample prep aration other than sectioning. The final family of sources that merits consideration for MSI is that of the liquid microjunction variety. In these sources, a solvent stream that contacts the surface of a sample serves to simultaneously extract and transport the analytes to the ionization source, typically ESI. 25 Similar to LAESI and LS I, liquid microjunction sources allow for multiple charging of analytes when combined with ESI, making them suitable for pairing
26 with ion traps. However, the major disadvantage associated with these types of sources is spatial resolution, which is current ly on the order of 500 m. 25 Mass analyzers Following ionization, the generated ions are transferred to the mass anal yzer by differential pumping and a series of ion optics. At present, a number of mass analyzers are available for integration with sources amenable to MSI. Historically, t he most common mass analyzers for MSI have been ion traps and time of flight ma ss a nalyzers; however, there has been a recent surge of interest in the MSI community in Fourier transform mass analyzers. The operational principles and the relative merits of each of these mass analyzer classes will be discussed in this section. Ion traps T he two most common ion traps used for MSI are the three dimensional quadrupole ion trap (QIT) and the two dimensional linear ion trap (LIT). The concept of trapping ions in a quadrupolar field was initially proposed in patent describing quadrupole mass filters. 26 Co nsequently, t he operation of a QIT derives from a quadrupole mass filter, in which four parallel hyperbolic (or cylindrical) rod electrodes are utilized in a square array. 27 A combination of radio frequency ( RF ) and direct current (DC) potentials is applied to opposite pairs of rods to generate the quadrupolar field. 27 Depending on the magnitude of the potentials applied and the mass to charge ( m/z ) of the ion, the trajectory of the ion will either be stable, passing through the quadrupole mass filter, or unstable, causing the ion to collide with a rod or become ejected from the mass filter. The stability of an ion within a quadrupolar field is determined by a s et of non linear equations known as the Mathieu equations. As applied to a quadrupole mass
27 filter a Mathieu stability diagram can be plotted in terms of the parameters a and q of which t he formulas for these two parameters are listed below (Equation 1 1 and Equation 1 2) The parameter, a is influenced by the mass of an analyte ( m ) the electronic charge ( e ), the radius of an electrode ( r 0 ) the drive frequency ( ) and the applied DC potential ( U ) (1 1) In contrast, q is influenced by the mass the electronic charge the radius, the drive frequency, and the applied RF potential ( V ) (1 2) The positive portion of the first stability region for a quadrupole ma ss filter is shown in Figure 1 2 with areas of stability shaded in grey. Mass discrimination in the quadrupole mass filter is achieved using the apex of the stability diagram. If one applies the appropriate RF and DC potentials, a n operating line is created where only a small range of mass t o charge ( m/z ) values falls within the uppermost portion of the stability diagram. Furthermore by increasing the DC and RF potentials while maintaining the ratio of DC to RF potential, a mass spectrum can be scanned in order of increasing m/z A lthough ma ny of the fundamentals are similar, the QIT has a much different configuration than a quadrupole mass filter The QIT has three electrodes as opposed to the traditional four in the quadrupole mass filter Two of the QIT electrodes have an outer hyperbolo idal geometry, serving as end caps, and the remaining electrode has an inner hyperboloidal geometry, serving as a ring electrode. 27 Furthermore, the optimal theoretical radius for the ring electrode is only half that of the distance between the two
28 end caps 27 in contrast to the quadrupole mass filter, in which all four electrodes are equidistant. As a result, the parameters a and q for the Mathieu stability di agram must be considered in both the end cap separation ( axial direction ) and the ring electrode radius ( radial direction ) Thus, for a QIT the parameter a is now a function of mass ( m ), the electronic charge ( e ) the DC potential ( U ), the radius of the ring electrode ( r 0 ), the distance between the ring electrode and either end cap ( z 0 ), and the drive frequency ( ). The equation for a in the axial direction ( a z ) is shown below: ( 1 3) Due to the relative radii between the end caps and ring electrode, a in the radial direction ( a r ) is simply: ( 1 4) Similarly, q in the axial direction ( q z ) is defined as: ( 1 5) Finally, q in the radial direction ( q r ) is defined as: ( 1 6) With an understanding of a and q in both the axial and radial directions, the regions of stability can be defined. Formerly s ymmetrical about the q axis the Mathieu stability diagram for a QIT becomes stretched due to the necessity for both radial and axial stability in the ion trap A typical stability diagram (expressed in terms of a z and q z ) for a QIT is shown in Figure 1 3 Assuming an ion has a stable ion trajectory within an ion trap, the ion will oscill ate at a specific fundamental frequency. This fundamental
29 frequency, known as the secular frequency ( ) is influenced by a secondary trapping parameter ( u ) and (Equation 1 7) ( 1 7) The Dehmelt approximation defines u for valu es of q less than 0.4 and is given as: ( 1 8 ) For larger values of q the Dehmelt approximation is no longer an effective approximation for u and the precise value of u is found by solving a continuous fraction, 27 which is beyond the scope of this work. Substituting u and subsequently a u and q u into E quation 1 7, one finds that secular frequency is inversely proportional to m/z Revisiting the stability diagram in Figure 1 3 wh en 0 DC voltage is applied (i.e., a z = 0), the stability diagram intersects with the q z axis at a value of 0.908, yielding the theoretical low mass cutoff (LMCO) for a QIT. Consequently, the LMCO is directly proportional to the applied RF voltage, and ion s can be selectively ejected in increasing m/z by simply ramping RF voltage. Alternatively ions can be ejected by applying a waveform with the appropriate resonant frequency matching the secular frequency of the ion to be ejected. As we will find later, a combination of these two ejection methods more commonly known as mass selective instability, is the most practical approach. The LIT consists of three hyperbolic quadrupole rod arrays in sequence. The outer two sets of rod arrays are typically shorter in length than the middle set of rods. DC potentials are applied to all three sets of rods to trap the ions within the center section. 28 On the Thermo Scientific LTQ XL, which w ill be used extensively in the work
30 of this dissertation three discre t e axial DC potentials are applied. The schematic for the MALDI LTQ XL is shown in Figure 1 4 During ion trapping in positive mode, voltages of 9 V, 12 V, and 7 V are utilized for the front, center, and rear rod arrays, respectively. 29 For negative mode, DC voltages of the same magnitude, but opposite polarity are applied. As the design of an LIT closely resembles that of a quadrupole mass filter, Equation 1 1 and Equat ion 1 2 can be utilized to calculate the parameters a and q respectively. Similarly, the Mathieu stability diagram is the same for an LIT and a quadrupole mass filter; however, i t should be noted that during mass analysis, the LTQ XL is oper ated in RF on ly mode (i.e., a = 0 and only RF voltage is applied ) similar to the QIT ; thus, the y axis on the stability diagram can be largely ignored. Alt hough the operating p rinciples for an LIT are a hybrid of the quadrupole mass filter and the QIT, the LIT offers a variety of pr actical advantages over the QIT, including increased ion storage volume, ease of ion injection, higher trapping efficiency, and lower mass discrimination. 28, 30 Perhaps the most striking advantage is the increase d ion storage volume of the LIT; t he practical storage volume of the LIT is dictated by the length of the center section, whereas the practical storage volume of a QIT is dictated by r 0 Lengthening r 0 in the QIT requires higher RF voltages for efficient trapping; 30 however, axial confinement in the LIT is b rought about by DC voltage. This abilit y to store more ions in the LIT results in reduced space charge effects; t he space charge limit can be thought of as the maximum number of ions trapp ed while maintaining efficient activation, isolation, mass resolution, or mass accuracy. 28 Thus, the more ions
31 that can effectively be trapped results in greater sensitivity and increased dynamic range resulting in a higher performance mass analyzer. When considered as a whole ion traps are relatively simple mass analyzers that require little upkeep while providing superior MS n capabilities. 31 Alt hough stand alone ion traps are valuable in MSI, these mass analyzers are lacking in performance relative to more complex and expensive mass analyzers assuming only one stage of ma ss analysis The three most prominent disadvantages of ion traps are mass resolution, mass range, and scan speed. Under typical operating conditions, most ion traps demonstrate unit resolution across the usable mass range. Despite the use of resonance e jection voltages and slower scan speeds, this process is highly influenced by space charge effects (discussed above). Furthermore, the mass range of commercial ion traps limit s analyses to ions less than 4,5 00 amu ; 2 however, this upper mass limit has been circumvented in a number of experimental configurations 32 Finally, the scan speed of a conventional ion trap is approxi mately 1 scan/sec Alt hough this can be improved, an increase of scan speed is often accompanied by a loss in resolution. Thus, this scan speed/resolution tradeoff presents a concerning barrier for rapid high resolution imaging data acquisition. Despite the aforementioned disadvantages, the ease of implementing two or more stages of mass spectrometry (MS 2 and MS n ) should not be overlooked. The importance of this capability fo r MSI will be highlighted throughout this dissertation. Time of flight mass analyzers Time of flight (ToF) mass analyzers are also well suited for integration with pulsed ionization sources such as MALDI. In MALDI ToF MS, an ion packet generated by the io nization source is accelerated into a flight tube Following acceleration, each ion in
32 the initial ion packet should now have the same kinetic energy Prior to acceleration an ion has a potential energy ( E p ) related to its charge ( z ), the charge of an e lectron ( e ), and the applied voltage ( V ) (Equation 1 9) (1 9) Following acceleration, the ion has a kinetic energy ( E k ) that is simply related to the m ) and its velocity ( v ) (Equation 1 10). (1 10) Assuming all of the potential energy i s converted to kinetic energy, E quation 1 9 and Equation 1 10 can be equated to form E quation 1 11. (1 11) Thus, if E quation 1 11 is rearranged, we find that the m/z ratio of an ion is in versely proportional to the square root of its velocity. Furthermore, the time it takes an ion to travel down a drift tube ( t ) is solely dependent on the velocity of the ion and the length of the drift tube ( L ) (Equation 1 12). (1 12) By combining E quation 1 11 and Equation 1 12 and solving for time, we find that the drift time of an ion is proportional to m/z (Equation 1 13). (1 13) Thus, ToF mass analyzers provide a simple method for mass analysis ; ions with a greater m/z ratio travel more slowly down the drift tube and reach the detector later than ions with a lower m/z
33 When compared to ion traps, linear ToF mass analyzers exhibit a much higher upper mass limit (~150 kDa). 2 Although linear ToF geometries can analyze higher molecular weight species, the lower mass region of the MALDI MS spectrum (<600 D a) is typically obscured by matrix related ions. Furthermore, the mass resolution in the linear geometry is fairly poor. The addition of an electrostatic mirror, commonly referred to as the reflectron geometry, compensates for variations in the velocity of isobaric ions, resulting in a much improved mass resolution, far surpassing that of conventional ion traps. 2 Thus, ToFs provide a compromise between mass range and resolut ion that is well suited for a number of MSI applications. The major disadvantage for ToF mass analyzers is that they are typically limited to one or two stages of mass analysis when placed in tandem with a quadrupole or second ToF analyzer Fourier transf orm mass analyzers F ourier transform mass analyzers offer an alternative to the destructive methods of ion detection used in other mass analyzers. T he conventional process for ion detection involves ions striking a surface (e.g., a conversion dynode or electron multiplier ), wherein the collision of the ion with a surface generates a secondary signal that can be measured (e.g., the generation of secondary electrons that are subsequently multiplied and measured as current ). In contrast, Fourier transf orm mass analyzers generate a detectable signal by analyzing the image current generated from trapped ions. Two types of Fourier transform mass analyzers, Fourier transfo rm ion cyclotron resonance and o rbitrap mass analyzers, will be discussed in this sec tion. In Fourier transform ion cyclotron resonance mass spectrometry (FT ICR MS), ions are confined by a combination of magnetic and electrostatic fields Confinement is achieved by placing an FT ICR cell with applied electrostatic potentials inside the
34 central bore of a superconducting magnet. The FT ICR cel l is typically segmented into a number of sections: some of which are utilized for excitation and the remaining sections for detection. Initially, the ions within the trapping area exhibit an incoh erent motion. 33 Thus, following ion injection, the random distribution of ions throughout the ion trap results in a charge balance between the opposing detection plates. 33 By applying an radio frequency ( RF ) waveform to the excitation plates that continually increas es (or decreases) in frequency, ions of similar m/z that are in resonance with the excitation field begin to increase in cyclotron radius within the trapping area, forming coherent ion packets. As the coherent ion packets travel close to the detection plates, the charge of the ion packet induces an image curr ent, which can then be measured A transient is generated by plotting the magnitude of the image current over time. By performing fast Fourier transform on the transient, the data can be converte d from the time domain to the frequency domain, yielding a plot of magnitude vs. cyclotron frequency ( c ) Under ideal conditions, the cyclotron frequency exhibits a dependence on the m/z of the ion and the magnetic field strength ( B 0 ) detailed in Equatio n 1 14 34 ( 1 14 ) The major advantage of this manner of detection lies in mass resolution and mass accuracy offering anywhere between a 10 100 fold improvement over traditional mass analyzers. 35 Furthermore, the mass resolving power of the mass analyzer is directly prop ortional to the magnetic field; 34 thus, as progressively stronger superconducting magnets are fabricated, the resolving power of the technique will continually improve. Currently, there are two major di sadvantages to the use of FT ICR MS instruments: cost and upkeep, both of which stem from the superconducting magnet. Typical high
35 performance FT ICR MS instruments cost at least 1 million USD. 36 Furthermore, the superconducting magnet must be cryogenically cooled, resulting in the consumption of both liquid helium and nitrogen. Orbitrap mass analyzers, first introduced b y Alexander Makarov in 2000, 37 provide an alternative to the relatively expensive FT ICR MS for high performance mass le 37 In a purely electrostatic field, the position of an ion ( z(t) ) at any given point in time along the axis of the electrodes (i.e., the z axis) can be thought of as a simple harmonic oscillator, expressed in Equa tion 1 15. 37 (1 15) In this equation, k represents the field curvature, is the frequency of axial oscillation, z 0 is the initial position of the ion, t is time, and E z is the energy characteristic of the ion, as defined by Equation 1 16. 37 (1 16) Under ideal conditions, also exhibits a dependence upon mass to charge ( m/q ), as expressed in Equation 1 17. 37 (1 17) Thus, the frequency of oscillation of an ion along the z ( m/z ) 1/2 Similar to FT ICR MS, ions of the same m/z must exhibit coherent motion along the z axis for an image current to be generated. Coherence is achieved by rapidly pulsing the ions into the orbitrap at a position offset from the equator, defined as z =0. 38 Assuming the spread of flight times into the trap is considerably small (on the
36 order of nanoseconds), the ions will exhibit coherent motion along the z axis without an excitation w aveform. 38 In this fashion, image current is detected by segmenting the outer electrode into two independent sections about the equator. From this point onward, detection and processing proceed parallel to that of FT ICR MS. In regards to performance, orbitrap mass analyzers have been reported to have mass resolution and mass accuracy rivaling, or in some cases exceeding, that of FT ICR m ass analyzers. 39, 40 Furthermore, improving the performance characteristics of the orbitrap can be achieved via increasing the electrostatic field, eith er by increasing the applied voltages or by altering the orbitrap geometry. 40 As these performa nce characteristics can be achieved with a purely electrostatic field, the orbitrap eliminates the need for a superconducting magnet, and consequently, the disadvantages that arise from a superconducting magnet. Compound identification: accurate mass Once a mass spectrum is generated, compound identification is conducted for unknown ions with relevant distributions within the sample. Although MALDI MS often provides the molecular weight of biomolecules, the nominal molecular weight alone is not suitab le for compound identification; a number of different compounds can produce ions at the same nominal m/z For further identification, there are two mass spectrometric strategies: 1) obtaining accurate mass measurements and 2) performing tandem mass spectr ometry (MS n ). Accurate mass measurements, assuming the data are collected on a well calibrated instrument, provide information about the collective mass defect of the ion. Based on the mass defect, an empirical formula for the ion in question can be calc ulated. This strategy is often employed on high resolution ToF mass analyzers and Fourier transform mass analyzers Accurate mass measurements,
37 however, cannot provide structural information to differentiate isomeric compounds (i.e., compounds that have same empirical formula, but differ in structural configuration). Compound identification: tandem mass spectrometry Tandem mass spectrometry ( MS n ) provides complementary information to accurate mass studies. In this strategy, an ion of interest is isolate d, and energy is imparted upon the ion. Eventually, the internal energy of the ion becomes great enough that labile bonds begin to break, resulting in lower mass fragment ions compared to the initial precursor ion. Based on the fragmentation pattern, one can deduce the initial structure of the precursor ion. MS n can be conducted through a number of dissociation methods, including collision induced dissociation (CID), electron transfer dissociation (ETD), electron capture dissociation (ECD), photodissociat ion (PD), and in source fragmentation. Of these dissociation methods, CID has become the method of choice for dissociation of small molecules. In this method, energy is imparted by accelerating the precursor ion resulting in energetic collisions with an inert collision gas. The collisions result in the kinetic energy into internal energy. Provided enough kinetic energy is converted to internal energy the precursor ion will fragment. Once a labile bond i s broken, one side of the original precursor ion will often retain the charge, generating the fragment ion. The remaining side is commonly referred to as a neutral loss (NL), and is not directly detected. High energy vs l ow energy CID Alt hough the fragmen t ions detected are largely compound dependent, the kinetic energy of the precursor ion also determines the generated fragments. High energy CID is defined as a precursor ion having a kinetic energy of 1 keV or greater. 6 In this
38 instance, the high energy collisions with the relatively immobile colli sion gas result in a rapid increase in the internal energy of the ion. Low energy CID is defined as a precursor ion that has a k inetic energy between 1 eV and a few hundred eV. 6 In general, low energy CID is utilized with triple quadrupole and ion trap mass analyzers, whereas high energy CID has tra ditionally been utilized on tandem sector instruments. 6 Tandem in s pace vs t andem in t ime MS 2 MS 2 can be performed either tandem in space or tandem in time, depending on the mass analyzer. Tandem in space MS 2 is typically performed on either a triple quadrupole mass analyzer ( QQQ ) or other linear mass analyzer arrangement such as a quadrupole time of flight mass analyzer (QToF) or tandem time of flight mass analyzer (ToF ToF) In a QQQ, a precursor ion is mass selected in the first quadrupole, fragmented in the second quadrupole utilizing collisions with a neutral collision gas and the fragment ions are mass analyzed in the third quadrupole. Tandem in time MS 2 is typically performed within ion traps or other mass analyzers where the ions are stored within the analyzer region. I n th is scenario, selection of the precursor ion, fragmentation of the precursor ion, mass analysis of the resulting product ions all occur within the same trapping region with the added benefit that fragmentation can be extended beyond the initial precurs or ion. Thus, higher stages of mass spectrometry (i.e., MS 3 MS 4 ed for additional mass analyzers MS n on the Thermo LTQ XL This section will discuss the theory and practical aspects of performing MS n on a commercial linear ion trap mass analyzer the Thermo LTQ XL (instrument schematic depicted in Figure 1 4) In general, MS n on the linear ion trap consists of precursor ion isolation, collisional cooling of the precursor ion, ion activation (resulting in
39 fragmentatio n), and finally mass analysis and detection of the resulting fragment ions. Each of these events is detailed in Figure 1 5 and will be discussed in terms of the voltages and waveforms applied. Ion isolation in a commercial Thermo LTQ XL is a two step pro cess. The first step in isolation is to shift the ion to be isolated to a q of 0.83 Recall that the LMCO of an ion trap is directly proportional to the RF voltage applied; thus, as a consequence of the first step of isolation, a number of low mass ions fall at a q with an unstable trajectory that is at a q value past the LMCO at q =0.908 To eject the remaining ions, excluding the precursor ion, a sum of sines ion isolation waveform is applied that contains the resonant frequencies of all ions within the trap, except the precursor ion of interest. ion population. In do ing so, the RF voltage is decreased until the precursor ion is held at q = 0.25. Furthermore, minor collisions with the helium damping gas remove kinetic energy from the ions, moving the ion population to ward the center of the ion trap and ensuring that the precursor ion population has a stable trajectory within the trap. Activation (fragmentation) of the ion population is then induced by applying a supplementary resonant excitation voltage across the exit rods. The overall magnitude of this excitation volta ge is both compound and mass dependent; however, the Thermo LTQ XL uses a normalized collision energy (NCE%), which is a linear function that increases with increasing m/z When applied, t his voltage increases the precursor ion radial direction, resulting in an increase in kinetic energy. As the kinetic energy of the population increases, energetic collisions will occur between the precursor ion and the helium damp ing gas. With enough time for energetic
40 collisions, the precurso r ions will gain enough internal energy to fragment typically along the lowest energy pathway. The resulting lower mass fr agment ions will not undergo further fragmentation as the excitation voltage is in resonance with the precursor ion population. It should be noted that broad band excitation can be performed on the Thermo LTQ XL, which would yield further fragmentation of the fragment ions. Isolation and fragmentation can theoretically be repeated indefinitely assuming the fragment ions have stable trajectories within the trap and are amenable to fragmentation by CID Once all of the desired stages of MS n have been performed, the fragment ions are ejected from the trap in order of increasing m/z for detection. To eject the ions, a combination of the main RF voltage and the supplementary resonant ejection voltage is utilized. Although the ions could theoretically be ejected with solely the main RF voltage, the supplementary resonant ejection vol tage has been shown to improve the m ass re solution in both the LIT and QI T 28, 41 On the LTQ XL, the resonant ejection wa veform is in resonance when the ions to be ejected fall at q =0.88. Data processing Image generation Mass spectrometric images are generated by extracting the intensity of a specified MS or MS n m/z range, and subsequently plotting that intensity vs. the X,Y position on tissue. There are a number of free software packages available for image generation. The most notable of these software packages is BioMap, a package originally developed for magnetic resonance imaging in 1996 and later adapted for MSI by Ma r kus Stoeckli at Novartis 42 BioM ap offers image generation with a number of statistical features, and supports a number of imaging dataset formats, including .msi
41 and .imzML. These two file forma ts are sufficient for most instrument platforms (except Thermo Scientific imaging instruments), as there are free converters for both .msi and .imzML available. At present, a converter exists for converting Thermo Scientific .raw files to .imzML files; however, the converter is functional only for datasets that conform to specific constraints (i.e., images collected with rectangular dimensions) Thermo Scientific developed Thermo ImageQuest to support imaging experiments con ducted on all Thermo instruments. This software uses two file formats in concurrence, the .raw file containing the mass spectra and a .MALDIpos file that associates the relative position with each mass spectrum. Although sufficient for basic image genera tion, ImageQuest lacks functionality for more compl ex statistical analysis. Normalization The MALDI ionization process is thought to be the largest source of signal variability in a MALDI MSI experiment Unfortunately, a number of sources of variability e xist in this process. For instance, heterogeneous application and crystallization of the MALDI matrix will contribute to fluctuation in ionization yields. 43 Also, the local sample topography can also introduce variability, as relatively small changes in height can cause variations in laser fluence delivered to the sample. Final ly, the shot to shot variability in the laser can also contribute to fluctu ation in ionization yields. 43 To account for this variability, normalization is often utilized for each individual pixel. To date, a number of strategies have been proposed al though normalization still remains a hotly debated topic in the MSI community MSI normalization involves dividing the ion intensity of interest by either the total ion current (TIC), an internal standard, a matrix ion, or a target specific ion that reflects the relative intensity at that pixel. An internal standard appears to be th e optimal method for normalization; 43 45
42 however, endogenous signal at the s ame nominal mass of the internal standard in single stage MS experiments may confound analysis. Furthermore, isotopically labeled internal standards can be costly, and an internal standard may not demonstrate the same ionization efficiency as all analytes of interest in the target sample. Similarly, normalization to a matrix ion appears impractical, as the ionization efficiency of the matrix ion may not reflect the ionization efficiency of the analytes of interest. 46 Finally, normalization to a target specific ion assumes that there are no isobaric ions, and that the ion concentration does not vary throughout tissue. Currently, the simplest effective method for normalization appears to be dividing by the TIC. This method, however, is not without drawbacks, as the relative abundance of m atrix and analyte ions may vary from pixel to pixel. Recentl y, researchers have proposed a novel normalization method that utilizes the dataset. 46 This method eliminates many of the aforementioned problems; however, rigorous statistica l analysis is required to normalization Multivariate Data A nalysis Over the past 10 years, there have been an increasing number of instruments capable of performing high mass resolution MSI. This increase in mass resolution has allowed researchers to analyze increasingly larger number s of biological compounds by MSI than ever before. In particular, Fourier transform mass analyzers (e.g ., orbitrap mass analyzers or Fourier tra nsform ion cyclotron resonance cells), in which resolving power is theoretically proportional to the transient time, 39 present a promising opportunity for discriminating and imaging nominal isobaric compounds. Unfortunately
43 an increase in mass resolution is invariably linked to an increase in data set size, and consequently, analysis time sider t wo isobaric phosphatidylcholine (PC ) ions commonly detected in positive mode MALDI MS studies of mammalian tissue: the [M+H] + of PC (16:0/20:4 ) and the [M+Na] + of PC (16:0/18:1). Nominally, both ions appear at m/z 782; however, the two ions have ex act masses of m/z 782.5694 and 782.5676 for PC (16:0/20:4 ) and PC (16:0/18:1), respectively. To resolve these two lipid ions the mass analyzer would need a resol ution of approximately 500,000, and would similarly need to collect data at 0.001 amu intervals If data of this nature were collected in the mass range of 150 2000 amu a common mass range used for lipid and metabolite MALDI mass spectrometry a total of 1,850,000 possible images could be generated. Assuming a researcher could effectiv e ly analyze a single image each second, it would take the researcher approximately three months to thoroughly analyze the entire dataset. Th e time consuming nature of fully interpreting a high resolution MSI dataset stems from its inherent multidimensional ity; the number of dimensions in the dataset is equal to the number of variables or mass features in this test case, 1,850,0000 Thus, a more efficient method of interpreting highly dimensional datasets would be beneficial to the MSI community. Principal component analysis Multivariate data analysis techniques, such as principal component analysis (PCA) are methods utilized to reduce the dimensionality of datasets. 47 These data reduction techniques serve to calculate the axis of greatest variation also known as the first component, through the multid imensional space of a dataset. The second component is then orthogonal to the first, and the process can be repeated for as many principal
44 components as there are variables. The result ing component s consist of linear combination s of weig hted variables Variables that contribute greatest to the variance within a dataset are given relatively large weighting factors (loadings), and variables that have minor contributions are given relat ively small weighting factors is represented by the s ummation of the products of a variable and its corresponding loading coefficient The samples in the dataset can then be projected onto component axes by plotting the scores in 2 or 3 dimensions (also known as a scores plot) Theoretically, the scores plot can be extended to as many components as there are variables; however, 2 or 3 dimensions are all that can be effectively visualized by humans Furthermore, if the dataset contains a sufficiently high number of variables, a nd significant correlations exist within the dataset, the first few principal components will account for a large percentage of th e variation 47 Alt hough the concept of PCA appears simple the data manipulation required and the mathematical principles that underlie PCA are more complicated. Prior to pe rforming PCA, the sample data must first be transformed so that all variables are centered about the origin, commonly known as mean centering. Next, although not required, it is recommended that the data are normalized so that unit varianc e is obtained. Once the data are transformed into a suitable format, one can begin to perfo rm PCA. The first step in this process is to calculate the c ovariance matrix ( ) of the variables or alternatively the correlation matrix depending on the normalization of the v ariables Once is calculated, the mathematical formula of the k th principal component is simply given by Equation 1 15. (1 15)
45 In this equation, k is simply an eigenvector of corresponding to the k th largest eigenvalue of the variance of z k and x is a vector of the standardized variables 47 To maximize the variance of z 1 1 1 should be maximized; however, it should be noted that this maximum will only be found if a normalization constraint is applied, the most common being that the sums of squ ares of the elements within k are equal to 1. 47 This process is then repeated for z 2 using the same constraints, except that z 2 must be uncorrelated with z 1 The process is then extended for as many principal components as there are variables in the dataset. A test data set for this type of ana lysis is displayed in Figure 1 6 A The data set contains ten school students in two sample groupings: 1 st grade students and 12 th grade Initially, the averages for all vari ables are calculated, and individual data points are mean centered (Figure 1 6 B) Following mean centering a 2 dimensional plot can be generated to vis ualize the sample groupings in Euclidean space (Figure 1 6 C ). From this plot, it is clear to see that the direction of greatest variance within the dataset is a line passing through the center of both sample groupings. This direction, depicted in the Figure 1 6 C as a green line, is the first principal component in PCA and demonstrates variability between the sample groupings The second principal component, by definition, is uncorrelated (read orthogonal in a graphical sense) to the first principal component, and is represented by a red line in Figure 1 6 C This principal component largely demonstrates variability within a sample grouping. By rotating the 2 dimensional space so that the principal components now fall on the x and y axes, the data are projected into a 2 dimensional scores plot (Figure 1 6 D ). The loadings
46 represented by k for the mathe matical example above, that contribute to this separation can then be examined to determine the features that drive the separation in the data set. Alt hough this is a trivial exercise with just 2 or 3 variables, extending this e xercise to further variable s yields complex calculation s and a data space that humans are incapable of visualizing Fortunately, modern computers can perform complex PCA calculations and rotations with in a multidimensional data set containing thousands of variables within seconds allowing for separation of sample groupings to be visualized in 2 or 3 dimensional principal component axes Furthermore, a number of open source algorithms have been written to make these calculations readily available to the scientific community. Parti al least squares discriminant analysis Similar to PCA, partial least squares discriminant analysis (PLS DA) also determines the direction of max imum va riance within a dataset. The major difference between PCA and PLS DA is that PLS DA determines the axis of greatest variation based on the covariance as determined by the data and sample groupings 48, 49 This type of multivariate analysis is cons idered a supervised analysis, and is often used to develop a classification model. Furthermore, PLS DA serves to eliminate variance that is unrelated to sample groupings (e.g., sample preparation or unrelated biological variability). Applying multivariate data analysis to MS I Alt hough PCA has primarily been utilized in chromatographic applications of mass spectrometry, there has been recent interest to utilize PCA with MSI applications. 48, 50 In general, there are two different approaches for adapting MSI datasets for use with PCA. The simpler approach is a guided methodology, wherein individual regions of
47 interest (ROIs) within the dataset are averaged for PCA. This approach allows the researcher a measure of co ntrol over the dataset, as unwanted pixels or regions of pixels can be omitted. Alt hough this approach may simplify PCA interpretation, there is the potential for researchers to overlook valuable information. In contrast, the second approach is an unbiase d method, wherein each pixel, or alternatively a specified number of averaged pixels, is treated as an individual sample. In this method, the only excluded pixels are those corresponding to areas off of the target sample (background pixels). The second a pproach has the advantage of the ability to plot scores images, wherein the score of a particular principal component can be plotted in lieu of intensity for each pixel. Furthermore, ion distributions that may have been overlooked using the guided approac h may be found with this unbiased approach. The main disadvantage of the unbiased approach is that the inclusion of pixels not corresponding to ROIs may confound PCA analysis and interpretation. If possible, a combination of both the above approaches is recommended when interrogating an MSI dataset. At present, the main hindrance for this depth of analysis is the lack of suitable methodologies and available software, both of which will be addressed during this work. Scope of the Dissertation The remainde r of this dissertation provide s methodologies for integrating MSI and multivariate data analysis as applied to two biological applications. Chapter 2 will detail the use of PCA and PLS DA as applied to positive mode MALDI MSI datasets collected from a mod el system for myocardial infarction. In this work, the ability of both PCA and PLS DA to efficiently determine region specific markers will be highlighted. Furthermore, the relative merits of two PCA methodologies (guided and unbiased) will be explained. Chapter 3 will discuss the use of 9 aminoacridine as a matrix for both
48 positive and negative mode imaging of the same biological system as introduced in Chapter 2. This chapter will highlight the ability of a basic matrix to expand the breadth of detec table analytes by MALDI MS. Furthermore, the methodologies developed in Chapter 2 will be used to determine novel tissue specific markers for myocardial infarction. The final research chapter, Chapter 4, explore s the utility of MALDI MSI to analyze whole organisms without any prior sample preparation other than matrix coating. The methodologies developed in Chapter 2 will then be utilized to validate this approach through the analysis of genetic mutants. Finally, Chapter 5 provide s a brief summary of the research performed, the future work that will be condu cted, and the outlook for MALDI MSI as coupled to multivariate data analysis.
49 Figure 1 1. Representation of the MALDI process (Adapted from Chughtai et al.) 2 Dimensions are not drawn to scale.
50 Figure 1 2 Positive portion of the first stability region for a quadrupole mass filter. Regions of stability within the quadrupole mass filter are shaded in grey. A theoretical operating line for a quadrupole mass filter is also shown (red) with ions of varying m/z in green. (Adapted from Douglas et al ) 51
51 Figur e 1 3 Sample stability diagram expressed in te rms of a z and q z for a QIT. Regions of stability are shaded in grey. Furthermore, the LMCO for a conventional QIT operated in RF only mode is designated. (Adapted from March) 52
52 Figure 1 4 Schematic of the Thermo Scientific LTQ XL with a MALDI ionization source. (Adapted from Garrett et al.) 31
53 Figure 1 5 The r mo Scientific LTQ XL tandem MS n scan function T he three applied waveforms (main RF, resonant excitation, and ion isolation) for precursor ion isolation (red), precursor ion collisional coolin g (blue), precursor ion activation (orange), and fragment ion ejection (green) are displayed
54 Figure 1 6 Example of PCA applied to 2 variable dataset. The data are expressed in terms of A) raw v alues and B) mean centered values. Additionally, the data are plotted in relatio n to C) the two variables and D) the first two principal components generated by PCA.
55 CHAPTER 2 MALDI MASS SPECTROMETRIC IMAGING OF CARDIAC T ISSUE FOLLOWING CORONARY ARTERY LIGA TION INDUCED MYOCARD IAL INFARCTION In troduction Coronary heart disease (CHD) has remained the number one cause of death in the United States over the past four decades. 53, 54 Many deaths occurring in patients with CHD arise from acute myocardial infarction (MI), more commonly known as a heart attack. The onset of MI is most often the result of atheromatous occlusion of coronary arteries restrictin g blood supply to cardiac tissue. 55 The lack of oxygenated bloo d to cardiac tissue results in severe and often irreversible damage that may ultimately lead to heart failure. Although MI affects millions of people each year, the blood borne and tissue specific biochemical changes that occur following MI are not fully characterized, and few of the known biochemical markers are utilized in a clinical setting. To date, increases in two sets of blood borne protein markers, creatine kinase (CK MB) and troponins, are commonly used for the detection of myocardial necrosis (t issue death) following MI; however, these biomarker concentrations return to normal levels within days. 56 The discovery of a robust biomarker would improve both clinical diagnosis capabilities and our understanding of this condition. Furthermore, identifying biochemical changes in infarcted myocardium may provide mechanistic insig hts for new therapeutic intervention targets designed to limit myocardial tissue loss. Thus, there is a significant need to study the biochemical changes resulting from both CHD and MI. Portions of this chapter were reprinted with permission from Menger, R. F.; Stutts, W. L.; Anbukumar, D. S.; Bowden, J. A.; Ford, D. A.; Yost, R. A. Anal. Chem. 2012 84 1117 1125. Copyright 2012, American Chemical Society.
56 Mass spectrometry (MS) is an analytical tool with the potential to pr ovide mechanistic revelation for CHD characterization, biomarker discovery, and clinical diagnosis. 57 MS was util ized as early as 1996 to characterize protein epitopes for MI 58 as well as alterations in fatty acid metabolism in ischemic myocardium. 59 Since these early studies, the focus of CHD MS research has shifted to proteomic and metabolomic analyses of plasma, serum, and urine revealing a number of proteins 60 62 and metabolites 63, 64 as potential biomarkers and/or predictors for MI resulting from CHD. In addition, an extensive list of potential protein markers in serum was recently compiled, with the most abundant 25% projected to be amenable to tandem MS platforms. 65 Although there has been extensive study on biological fluids, MS analysis of intact cardiac tissue from CHD a nd MI positive specimens has not been thoroughly explored Thus, an analytical method that characterizes small molecule biochemical changes in intact cardiac tissue following MI may yield valuable insight. Furthermore, if this method can provide localiza tion of these relevant biomolecules, a deeper understanding of the biochemistry underlying MI may be obtained. Mass spectrometric imaging (MSI) is a microprobe technique that generates chemically selective images from thin tissue sections 1, 66 The most common ionization technique for MSI, matrix assisted laser desorption/ionization (MALDI) is a soft ionization technique well suited for the analysis of small and large biomolecules. 3, 5 The coupling of MALDI and MSI allows for tissue imaging by rastering a sample with respect to a stationary laser beam, col lecting mass spectra at discrete positions. An image is then generated by plotting the intensity of a s elected mass to charge ( m/z ) versus the X,Y position, thereby providing both chemical and spatial information. To date, MALDI
57 MSI analyses of intact tissue have identified the spatial distribution of endogenous and exogenous compounds including proteins, 1, 67 peptides, 22, 68 lipids, 31, 69, 70 and drugs. 71 73 One advantage of this technique is the ability to elucidate relative intensity changes and spa tial distributions resulting from external stimuli such as administration of an exogenous drug 71 or injury models. 74 Thus, MALDI MSI should be an ideal tool to delineate local alterations in lipids and metabolites in infarct ed myocardium following a model such as a left anterior descending (LAD) coronary artery ligation. This model permits comparison of perfused (non affected) and affected zones of tissue within a single section, eliminating much of the variability inherent in tissue to tissue comparisons. Although MALDI MSI tissue experiments can generate a wealth of data, interpretation of multi dimensional data sets can be time consuming, tedious, and/or and has rec ently be come of interest in the MSI community. 75 77 Common multivariate techniq ues such as principal component analysis (PCA) and partial least squares discriminant analysis (PLS DA) can be utilized for data reduction in multi dimensional data sets. Briefly, these techniques gr oup samples, or regions of tissue for MSI, based on linear combinations of weighted variables, known as components. Although PCA and PLS DA both determine the direction of maximum variance within a data set, the differs. PCA is an unsupervised technique that finds directions maximizing the total variance within a data set, whereas PLS DA is a supervised technique that separates samples based on the covariance determined by both the data set and membership groupi ngs. 49 For visualization, samples are
58 represented on two dimensional or three dimensional scores plots, and the weighting factors for each variable are displayed on loadings plots. Variables with the largest weighting factors contribute significantly to the grouping separation; thus, data sets containing large numbers of variables are reduced to a specified number of variables, simplifying data interpretation. This work re ports the combined use of MALDI MSI and two multivariate data analysis methodologies to study lipids and metabolite s in cardiac tissue following LAD coronary artery ligation. Initially, a guided multivariate data analysis methodology that interrogates user selected regions of interest from the MSI dataset is introduced to identify tissue specific markers of these regions following MI. In addition, an unbiased multivariate da ta analysis methodology is introduced for more exploratory analysis of MSI datasets, and the relative merits of each methodology are discussed. Experimental Chemicals and r eagents 2,5 dihydroxybenzoic acid (DHB) and phospholipase A 2 (PLA 2 ) from bovine panc reas were purchased from Sigma Aldrich (St. Louis, MO). Sodium acetate (NaOAc), anhydrous creatine, HPLC grade water (H 2 O), and HPLC grade methanol (MeOH) were purchased from Fisher Scientific ( Fair Lawn, NJ ). 1 heptadecanoyl 2 hydroxy sn glycero 3 phosp hocholine (LPC 17:0) was purchased from Avanti Polar Lipids (Alabaster, AL) 100% ethyl alcohol (EtOH) was purchased from Decon Labs (King of Prussia, PA). A creatine standard was prepared in H 2 O to a final concentration of 100 ppm. Additionally, the LP C 17:0 standard was dissolved and diluted in EtOH to a final concentration of 100 ppm. PLA 2 was dissolved in H 2 O and diluted to a final concentration of 500 ppm. A MALDI matrix solution consisting of 40 mg/mL DHB in
59 70:30 MeOH:H 2 O (v/v) was prepared for t he analysis of creatine (m atrix 1). A second matrix solution was prepared as above, except sodium acetate (NaOAc) was added to a final concentration of 10 mM for the analysis of lipids (m atrix 2). Biological sample p reparation All animal procedures were conducted in accordance with guidelines published in the Guide for the Care and Use of Laboratory Animals (National Research Council, National Academy Press, Washington, DC, 2010) and were approved by the Animal Care Committee of Saint Louis University. L igation of the LAD was performed as previously described 78 In brief, male Sprague Dawley rats (250 300 g body weight) were injected with ketamine/xylazine (55 mg/m L 7 mg/ mL ; 0.1 mL /100 g, i.p.). R ats were subsequently intubated and injected with Buprenex ( 0.05 mg/ mL ; 0.1 mL /100 g, i.p.) Animals were then ventilated with air at a tidal volume of 3 4 mL and a rate of 50 60 breaths/min (Harvard Apparatus ). A left lateral thoracotomy was then performed. The thoracic cage was exposed and the intercostal space between ribs 4 and 5 was separated with a retractor. The left atrial appendage was retracted and a 6 0 suture was placed around the proximal LAD. This LAD suture was tied tightly or loosely to produce an infarction (6 animals) o r sham surgery (6 animals) respectively, and then the thoracic incision was closed with a 5 0 suture. After recovery from surgery, rats were weighed and individually housed Twenty four hours following LAD occlusion rats were euthanized with pentobarbi tal (~800 mg/kg, i.p.) and subsequently hearts were removed and flash frozen in liquid nitrogen. All organs were stored at preparation. Hea r t tissue was bisected along a transverse plane that passed through the left and right ventricl es. The upper half of the heart was subjected to 2,3,5
60 triphenyltetrazolium chloride (TTC) staining to distinguish perfused and damaged tissue. 79 The remaining lower half of the heart was utilized for MALDI MSI analysis. Cardiac organs were sectioned using a Microm HM 505E cryostat (Waldorf, Germany). Optimal cutting temperature polymer (OCT) was not used for tissue mounting as OCT is reported to produce abundant ion signals in MALDI analysis of thin tissue sections, resulting in analyte ion suppression. 80 Instead, organs were held atop a drop of HPLC grade H 2 O on the cutting stage and placed into the cryostat chamber held at thereby freezi ng the drop of water and fusing the organ to the cutting stage. Subsequent 1 0 m thick sections were thaw mounted atop glass microscope slides and stored at desiccator for approxima tely 45 minutes to remove excess H 2 O. Serial cardiac se ctions were coated with either matrix 1 or m atrix 2. MALDI matrix was spray coated atop tissue sections using a glass Type A Meinhard n ebulizer (Golden, CO). Nitrogen was used as the nebulizing gas a t a pressure of 30 PSI and the matrix solution was delivered at a flow rate of 3 mL/min. In this method, approximately six passes were conducted over each slide followed by 15 seconds of waiting time in order to avoid excessive wetting of the tissue sectio ns. The process was repeated until a homogenous layer of matrix crystals was obtained over the entire tissue. Approximately 8 mL of MALDI matrix solution (320 mg of DHB) were used per microscope slide to obtain a suitable matrix layer on the tissue secti ons. Instrumentation E xperiments were performed using either a Thermo Scientific LTQ XL linear ion trap mass spectrometer (San Jose, CA) or a Waters Synapt (Milford, MA ) All experiments were performed on the LTQ XL unless otherwise noted. The LTQ XL was
61 equipped with a MALDI ionization source, consisting of a Lasertechnik Berlin MNL 106 LD N 2 is laser has a repetition rate of 60 Hz and The Waters Synapt was equipped with a MALDI ionization source consisting of a frequency tripled Nd:YAG laser ( =355 nm). This laser has a repetition rate of 200 H z and produces a laser spot diameter of appr oximately 75 m. Standard a nalysis S tandards were deposited atop a 96 well stainless steel MALDI plate using a modified dried droplet method. 3 M atrix 1 (for creatine) or M atrix 2 (for lipids) were pipetted onto the well plate. Solvent was evaporated from the mixture using gentle heat, leaving behind MALDI matrix and analyte co crystals. MS and MS 2 spectra were acquired using a laser energy of 5 J and 3 laser shots per laser stop, and 100 spectra were averaged to produce one standard MS or MS 2 spectrum. Tissue a nalysis Following analysis of standards, MSI was performed on pr epared cardiac tissue sections. The tissue sections were rastered with respect to the laser at a horizontal and vertical step size of 100 m. MS spectra ( m/z 100 250) were collected in profile mode over tissue sections coated with Matrix 1. Additionally, MS spectra ( m/z 200 2000) were collected in centroid mode (to minimize the size of the data files) over tissue coated with Matrix 2. A laser energy of 4 J and 3 laser shots per laser stop were utilized for all MS and MS 2 imaging experiments. Images were generated using Thermo ImageQuest v1.0.1 software. All MS images were normalized to the total ion current (TIC), but normalization was not conducted for MS 2 images.
62 Compound identification for lipid species from cardiac tissue was performed using tandem MS with collision induced dissociation (CID) or pulsed Q collision induced dissociation (PQD). For MS 2 experiments, an isolation width of 1.2 amu and a coll ision energy of 35 AU (arbitrary units normalized to m/z 400) was utilized. In instances where the mass cut off hindered ion identification, PQD was utilized with an isolation width of 1.2 amu and a collision energy of 25 AU. 81 In instances where MS 3 was necessary to identify ions, the same settings were utilized for the second stage of MS, and the third stage of MS was performed with an isolation width of 1.5 amu and a collision energy of 35 AU. Statistical a nalysis Multivariate data analysis and significance testing were performed to determine significant m/z values (features) differing between viable and infarcted tissue within the same tissue section. In doing so, two different methodologies were applied: 1) a guided multivariate data analysis methodology and 2) a supervised multivariate data analysis methodology. For the guided methodology, PCA and PLS DA were performed using Metaboanalyst web server. 49, 82 Five samples, each consisting of an average of 25 mass spectra, were selected from each tissue region Lists of m ass to charge (from centroid data) and intensity were exported from QualBrowser into Microsoft Excel and saved as .csv files. Since Metaboanalyst does not support mean centering without further normalization (e.g., normalization to the range or standard deviation of the mean), the data were processed in two different methods: the first method mean centered the intensity at each m/z value before it was imported into Metaboanalyst and the second method left the data unprocessed before it was imported. After the
63 process ed m/z and intensity lists were imported, a mass tolerance of 0. 7 5 m/z was utilized to counteract possible mass shifts due to space charge effects. Next, m/z values known to arise from MALDI matrix ions were removed from all samples. In this case, the in tense matrix ions observed in the mass spectra were m/z 273 and 274, representing the [2M 2H 2 O+H] + ion and its 13 C isotope, respectively. Following MALDI matrix peak removal, signal intensities were normalized to the TIC within each sample to account for signal variability inherent in MALDI MS tissue analysis. Additionally, the utility of scaling techniques (e.g., autoscaling) was investigated. After multivariate data test was performed on selected features using Microsoft Excel. For the unbiased multivariate data analysis methodology, PCA was performed using the ChemomeTricks tool box developed at FOM Institute AMOLF As opposed to separate sample For files collected on the Thermo LTQ XL, the .RAW file was first converted to .NetCDF using the Roadmap File Converter provided with Xcalibur. The data in the .NetCDF file were e xtracted with in house software, and saved as a .MAT file containing pert inent spatial and spectral information. Due to the large amount of data present in this dataset, peak selection and alignment were performed. Fo r this process, a summed spectrum of all pixels was created. Next, baseline subtraction was performed to enhance the signal to noise ratio of relevant mass features. To further reduce the size of the dataset, a user selected peak threshold (typically 0.1 15% of the base peak area ) was a pplied, and all mass features falling below this threshold were excluded. The remaining data were then saved as a .mat file. Fo llowing peak picking and alignment, a first iteration of PCA was performed. Pixels corresponding to areas off
64 tissue, and ions with high loadings coefficients in these pixels (matrix ions), were then removed. A second iteration of PCA was then performed, and scores images were generated for the first 20 principal components. Loadings plots were then generated for all principal components that demon strated relevant localization. Finally, clustering analysis using a user selected number of principal components was performed. Enzymatic d igestion with PLA 2 on t issue T he in vitro action of PLA 2 was also explored on sham surgery card iac tissue. In this method, 5 of 500 ppm PLA 2 in water was spotted on freshly sectioned tissue (before desiccation). The volume and concentration of PLA 2 was chosen so that approximately half of the tissue contained completely digested phospholipids. The tissue was then incubated at room temperature for one hour. A f terwards, the tissue was placed in a vacuum desiccator for one hour. As a control, the same procedure was conducted with the exception that HPLC grade H 2 O was spotted atop tissue in lieu of PLA 2 Following desiccation, the tissue was coated with m atrix 2 as described above. Coated tissue was then introduced into the mass spectrometer. Representative mass spectra, consisting of an average of 100 scans, were collected from both tissue sam ples. MS imaging experiments were also conducted over both tissues. Results Identification of infarcted m yocardium TTC s taining Prior to preparation for MSI analysis, the upper half of selected cardiac organs was submitted to TTC staining. This staining protocol is traditionally utilized to identify regions of tissue damage. In its original state, TTC is colorless; however, in th e presence of dehydrogenases from mitochondria of healthy tissue, enzymatic reduction
65 alters TTC to a formazan, producing a brick red color (Figure 2 1A) 79 Infarcted tissues lack the functioning mitochondria necessary for enzymatic reduction; thus, the tiss ue remains unstained. 83 TTC st aining of the heart (Figure 2 1B) illustrated infarcted myocardium in the affected area of the left ventricle, which was previously sup plied arterial blood by the LAD Creatine Biochemically, creatine kinase (CK) consumes one molecule of adenosine triphosp hate (ATP) to catalyze the conversion of one molecule of creatine to one molecule of phosphocreatine, which serves as a high energy phosphate buffer for ATP in muscle tissue. Three to six hours following a heart attack, CK MB, a creatine kinase selective to cardiac tissue, is released into the interstitial fluid, and consequently the bloodstream, resulting in an elevated plasma level of CK MB. 56 Although CK MB can be detected using biological assays, the molecular weight of CK MB (~86 kDa) and the limited mass range of a linear ion trap hinder direct detection of this enzyme. Conv ion trap. One might expect decreased levels of creatine within infarcted or damaged tissues if the damaged tissue also leaks water soluble metabolites such as the substr ate of CK. Prior to tissue analysis, a creatine standard was characterized in MS and MS 2 modes. MS analysis yields analyte ions at m/z 132 and m/z 154 corresponding to the [M+H] + and [M+Na] + ions, respectively. The ion at m/z 154 is isobaric with the [M] + for DHB; thus, the addition of sodium acetate to the MALDI matrix would not be beneficial for MS studies of creatine, as it may drive creatine ion signal to [M+Na] + The MS 2 spectrum of m/z 132 demonstrates a single abundant fragment ion at m/z 90
66 (protonated n methylglycine, more commonly known as sarcosine) resulting from the loss of CH 2 N 2 Following authentic creatine characterization, MS and MS 2 were applied to assess the spatial localization of creatine in cross sectional tissue sections from h earts sliced at the mid ventricle level. Both control rats and rats subjected to 24h of regional ischemia as a result of LAD coronary artery occlusion were analyzed. The MS image of m/z 132 normalized to the TIC (Figure 2 1 C ) demonstrates a lower signal in the infarcted region of tissue, positively correlating with the TTC image (Figure 2 1 B ). Similarly, the MS 2 image mapping the transition from m/z 132 90 also demonstrates a lower signal in the infarcted region of tissue (Figure 2 1 D ). In contrast, c ontrol tissues demonstrated a constant signal throughout cardiac tissue sections (Figure 2 2 ). The lower creatine signal may reflect the state of the plasma membrane of cells within infarcted myocardium. Upon rupture of these plasma membranes, water solu ble enzymes (e.g., CK) and possibly small metabolite substrates such as creatine may leach out from infarcted myocardium. Similar to what is observed with CKMB following MI, 84 a concurrent increase in blood borne creatine concentration would be expected. Phospholipase A 2 Phospholipase A 2 (PLA 2 ) is an enzyme that hydrolyzes the sn 2 acyl bond of intact phospholipids (PLs) including phosphatidylcholines (PCs) and phosphatidylethanolamines (PEs), yielding free fatt y acids and lysophospholipids (lyso PL s ) including lysophosphatidylcholines (LPCs) and lysophosphatidylethanolamines (LPEs). Multiple investigations have suggested an important role of increased PLA 2 activity as an enzymic mediator of the
67 pathophysiological sequelae o f myocardial ischemia. 85 87 Although LPCs and LPEs have been considered important arrhythmogenic lipids generated by PLA 2 88 the spatial localization of these lysoPLs in ischemic myocardium has not been investigated Accordingly, MSI experiments were performed on tissue sections of hearts subjected to LAD coronary artery ligati on. Representative mass spectra (an average of 50 scans) from the lipid region ( m/z 450 900) for infarcted and perfused tissue are shown in Figure 2 3 An increase in relative intensity is ob served for various ions in the lyso PL region ( m/z 450 600) of t he infarcted zones of these sections; however, few other major differences were elucidated from the averaged mass spectra from the two zones. To confirm that the above ions in the lysoPL region resulted from the enzymatic action of PLA 2 in vitro digestion of control myocardium (sham surgery) was performed. In doing so, freshly sectioned myocardium was exposed to approximately 5 L of 500 ppm PLA 2 for 1 hour prior to desiccation. The tissue section was the n desiccated, prepared, and an alyzed as above. Representative mass spectra from the lipid region of the mass spectrum are displayed for both the digested and control myocardium in Figure 2 4. A manual comparison between the lysoPL regions in Figures 2 3 and 2 4 reveals a striking res emblance between the infarcted myocardium from the ligation model (Figure 2 3A) and the digested myocardium from the sham model (Figure 2 4A), suggesting that these markers of infarcted ( necrotic ) tissue are formed via the action of PLA 2 Similarly, the c ontrol myocardium in both models (perfused myocardium in the ligation model and myocardium without enzymatic modification in the sham model) do not demonstrate appreciable ion signal in the lysoPL region; however, abundant ion signal is observed in the int act PL region for both models.
68 Identification of At Risk Myocardium: Triacylglycerides Triacylglycerides (TAGs) have also been shown to be increased in response to MI, which likely reflects inefficient mitochondrial beta oxidation of fatty acids during isc hemia. 89 Oil red O st aining conducted on hearts following myocardial infarction has identified lipid droplets in the at risk myocardium (i.e., the region surrounding infarcted tissue). 89 It has been hypothesized that these lipid droplets are largely composed of TAGs and cholesteryl esters. 85 In order to verify the hypothesis that these lipid droplets contain elevated concentrations of TAGs, MS 2 imaging experiments were conducted over infarcted cardiac tissue. It was recently reported that high concentrations of PCs hinder the detection o f other classes of lipids by MALDI MS; 90 thus, MS 2 is often necessary f or the detection of low abundance TAGs. MS 2 analyses of alkali metal adducts (e.g., [M+Na] + ) of TAGs re veal abundant fragmentation relating to the three fatty acid tails. A representative MS 2 spectrum of m/z 881 from cardiac tissue is shown in Figure 2 5A Abundant fragment ions are observed at m/z 625, 601, 599, and 597, arising from NLs of 256, 280, 282, and 284 Da, respectively. These NLs indicate two isobaric TAG ions at m/z 881, the [M+Na] + of TAG (16:0/18:1/18:1) and the [M+Na] + of TAG (16:0/18:0/18:2). The two most abundant fragments, the l oss of the 18:1 tail at m/z 599 (NL of 282) and the loss of the 16:0 tail at m/z 625 (NL of 256), are approximately twice as intense as the loss of the 18:0 and 18:2 tails at m/z 597 and 601, respectively, reflecting the relative abundance of the fatty aci d tails within these isobaric TAGs. MS 2 imaging experiments over infarcted cardiac tissue sections were conducted on commonly observed TAGs, namely TAG 50:2, TAG 52:3, TAG 52:2, and TAG 54:3 present as sodiated species at m/z 853, 879, 881, and 907, respec tively. Images were
69 generated for the expected neutral losses from each of the TAG species. The images for two of the fragment ions from m/z 881 the loss of the 16:0 tail at m/z 625 ( NL of 256 ) and the loss of the 18:1 tail at m/z 599 (NL of 282 ), are s hown in Figure s 2 5B and 2 5 C Intense ion signal is observed surrounding the area of infarcted tissue, correlating well with the lipid droplets observed in electron microscopy. The localization of TAGs in the area surrounding infarcted tissue supports the hypothesis that lipid droplets following MI contain elevated levels of TAGs. Guided multivariate data a nalysis and identification of tissue specific m arkers Infarcted and perfused m yocardium PCA and PLS DA were conducted on both tissue zones of a singl e infarcted section. A single tissue section was chosen to minimize any variance that might occur from section to section or specimen to specimen. Multiple samples were generated from a single cardiac section by choosing five horizontal lines, each line consisting of 25 mass spectra, within a selected tissue region. Scores plots from PCA and PLS DA, shown in Figure 2 6 were generated using Metaboanalyst. Separation between perfused and infarcted myocardium was observed in both multivariate data analysi s techniques. In PCA of the mean centered and TIC normalized data, principal component 1 (PC 1) and principal component 2 (PC 2) carried 85.8% and 7.2% of the total variance respectively ( Figure 2 6A ). The separation between infarcted and perfused tissu e was largely dictated by PC 1, whereas PC 2 accounted for the variance within the sample groupings. The autoscaled PCA plots demonstrated a similar separation (data not shown). In c ontrast to PCA, PLS DA is a supervised technique that separates samples b ased on the largest covariance in the data set (i.e., the sample groupings are known).
70 The PLS DA scores plot for the TIC normalized peak list is shown in Figure 2 6B Similar to PCA, component 1 carries 85.8% of the covariance, and sufficiently separate s the two tissue zones. Although the PCA and PLS DA scores plots show similar separation along the first component, PLS DA demonstrated a tighter cluster within each zone along the second component. PLS DA was also conducted on the autoscaled data (data not shown). The generated scores plot also shows sufficient separation between the perfused and infarcted zones along component 1. The loadings plots for these four separation methods were then analyzed for significant m/z values that influenced separatio n (Figure 2 7 ). In all four methods, the separation was dictated by PC 1 (for PCA) or c omponent 1 (for PLS DA). Thus, analysis of the loadings plots can be simplified to just a single PC or component for identifying significant m/z values. When analyzin g the loadings for the mean centered, TIC normalized data without autoscaling ( Figure 2 7A ), m/z values in the lyso PL region were positively correlated with infarcted tissue (e.g., m/z 496, 524, 546, and 562). In contrast, some m/z values present in the i ntact PL region were positively correlated with the unaffected perfused tissue (e.g., m/z 810 and 848). Analysis of the loadings for the autoscaled data proved to be more difficult as many m/z values proved to have similar loadings values ( Figure 2 7B ). In principle, any variable that differs significantly between sample sets should influence the separation; however, a major drawback for autoscaling is that low intensity signals, which may have large relative variances, are weighted equally with high in tensity variables. Upon analysis of the autoscaled loadings values, the majority of the m/z values with high loadings values appeared to be of low
71 intensity; therefore, autoscaling was not utilized for feature identification. Loadings plots from the PLS DA data produced many of the same m/z values as PCA. Significance testing was then performed on selected m/z values identified by PCA and PLS DA from the mean centered, TIC normalized data sets (without autoscaling). Additionally, significance testing was performed on a MALDI matrix ion ( m/z 273) for comparison. A one test was performed on the signal intensities at three significance levels (95%, 98%, and 99%). The results ( T able 2 1 ) demonstrate a significant difference for many of th e features ident ified by PCA and PLS DA in the lyso PL region, including m/z 496, 518. 524, 544, and 546, each positively correlated with infarcted tissue. Additionally, two m/z values in the intact PL region, m/z 832 and 848, correlated negatively with in farcted tissue and passed the t test at a significance level of 95% or greater. The opposite trends for these two lipid classes support increased PLA 2 activity in the infarcted myocardium Targeted MS n studies for ion identification were conducted on sign ificant features within the region of interest utilizing CID and PQD. MS 2 experiments revealed the major contribution of all significant features to be intact PC, intact PE, LPC, or LPE species. Previous work has demonstrated that MS 2 differentiates alka li metal adducts of PEs, alkali metal adducts of PCs, and protonated PCs; a neutral loss (NL) of 43 or 59 indicates the presence of a n alkali metal adducts of a PE or PC, respectively, 91, 92 wher e as a fragment ion at m/z 184 indicates a protonated PC. 31 Protonated PCs and LPCs yield relatively uninformative fragmentation; however, the mass of the fatty acid tail can be inferred f or LPCs, as only one fatty acid tail is prese nt. Fragmentation of alkali metal adducts of PCs and LPCs produces more relevant structural information.
72 MS 3 fragmenting the daughter ion resulting from a NL of 59, was used to determine the presence of sodiu m or potassium in cationization. In this instance, a NL of 146 indicates the presence of a sodiated species, whereas a NL of 162 indicates the presence of a potassiated species. 92 For LPCs, the remaining mass can be accounted for by the glyc erol backbone, a hydroxyl group in the sn 2 position, and the lone fatty acid tail in the sn 1 position. Low abundance neutral losses corresponding to the mass of the fatty acid tails, which are crucial for structural elucidation of intact PCs containing two fatty acid tails, may also be observed in alkali cationized PC fragmentation. The tentative identification for all of the significant features is listed in Table 2 1 At m/z 520 and m/z 544 MS 2 revealed isobaric lipids, specifically an LPC and LPE a t m/z 524 and two LPCs at m/z 544 Representative MS 2 spectra for m/z 520, 546, and 848 are shown in Figures 2 8 2 9, and 2 10 respectively. MS 2 analysis using CID averaged over cardiac tissue following LAD coronary artery ligation demonstrated a number isobaric lipids at m/z 520 ( Figure 2 8 ). The most intense fragment ions occur at m/z 502 and 184, indicating the presence of a protonated LPC, specifically LPC 18:2. Since the unsaturat ed fatty acid tails are typically found at the sn 2 position of the glycerol backbone, it is unlikely that this LPC formed via the action of PLA 2 The next most intense fragments occur from NLs of 43 and 61 at m/z 477 and 459, respectively. These fragmen ts would indicate the presence of a cationized PE at this nominal m/z likely the [M+K] + of LPE 18:0. Unl ike LPC 18:2, LPE 18:0 is a likely end product of hydrolysis by PLA 2 as the saturated fatty acid tails are primarily found at the sn 1 position of th e glycerol backbone. Thus, one might anticipate that the major contribution of signal in infarcted tissue may arise from the LPE. Finally,
73 a NL of 59 is present. This NL could nominally result from an alkyl linked LPC, more commonly known as a lysoplasm enylcholine (O LPC). The structure of this ion is unidentified. In contrast to m/z 520, m/z 546 produced fragments relating to a single abundant LPC MS 2 analysis ( Figure 2 9A ) demonstrated an abundant NL of 59 (trimethylamine from the PC headgroup), indicating a cationized PC. MS 3 of m/z 546 487 ( Figure 2 9B ) produced abundant NLs of 124 and 146 and an ion at m/z 147. The NL of 124 represents the fragmentation of the remainde r of the PC headgroup, and the NL of 146 represents the loss of neutral sodiated cyclophosphane. If the NL occurs from the other half of the lipid, sodiated cyclophosphane manifests itself as an ion at m/z 147. These three fragment ions would indicate th e presence of a sodiated LPC. From the remaining mass present in the lipid, m/z 546 was identified as the [M+Na] + of LPC 18:0. For intact PCs PQD was use d for compound identification in order to circumvent the low mass cutoff (LMCO) normally present for CID experiments on the linear ion trap. 81 For all PQD experiments, the LMCO was reduced to 50 Da. In a typical CID experiment isolating and fragmenting m/z 848, the LMCO, which is determi ned by the q isolation (0.25 on the Thermo LTQ XL) would have been 230 Da. This LMCO would preclude characteristic fragment ions for all protonated, sodiated, and potassiated ions ( m/z 184, 147, and 163, respectively). An MS 2 spectrum utilizing PQD from cardiac tissue is shown for m/z 848 in Figure 2 10 In this instance, the dominant fragment occurs from a NL of 59, indicating a cationized PC. Additionally, fragment ions are observed from NLs of 183 and 221. The 38 Da difference between these two fragments would indicate the presence of potassium. Fu rthermore, the fragment at m/z 163 is
74 indicative of potassiated cyclophosphane. MS 3 using CID of m/z 848 789 demonstrated a minor NL of 284, commonly corresponding to oleic acid. Therefore, the ion was identified as the [M+K] + of PC (18:0/20:4). Mass spe ctrometric images were generated from infarcted cardiac sections for each of the identified LPCs and PCs. Representative images for three lysoPLs ( Figure 2 11A, 2 11B, and 2 11C ) and one intact PC ( Figure 2 11D ) are shown. For each LPC or LPE, an intense ion signal is observed in the core of infarcted myocardium, and an area of less intense ion signal is observed moving outward from this area. In contrast to the lysoPL s, pertinent intact PCs, such as PC (18:0/20:4), demonstrate the reverse trend. An are a of relatively low signal is present near the core of infarcted myocardium, and an area of more intense signal is observed moving outward from this area, suggesting that intact PCs are being converted to LPCs via the action of PLA 2 within the area of infa rction. To verify the major contribution to infarcted tissue of m/z values containing isobaric lipids, MS 2 imaging experiments were conducted over cardiac tissue following LAD coronary artery ligation. An example of this strategy is shown as an inset in F igure 2 8 In this instance m/z 184 was chosen as the characteristic fragment for LPC 18:2 ; m/z 502 (NL of 18) was not chosen as it is typically assigned as a loss of water, which is considered a non specific neutral loss that can occur from a number of compounds naturally present in tissue or the MALDI matrix. For LPE 18:0, m/z 477 (NL of 43) was chosen as the characteristic fragment, although m/z 459 (NL 61) could have been chosen as well. The characteristic fragment for the LPE demonstrates a noticeably high ion signal in infarcted tissue relative to the LPC; thus, it was determined that the LPE is
75 the major contribution to the signal in infarcted tissue at m/z 520. This is an ex pected finding, as it was hypothesized that PLA 2 would be more active in producing LPE 18:0 than LPC 18:2 as previously described in this chapter A systematic study was performed to determine the PLA 2 activity as a function of the sn 2 fatty acid tail. In doing so, MS 2 imaging experiments were performed for common intact PCs and PEs containing stearic acid (18:0) in the sn 1 position of the glycerol backbone For all species, the partial head group loss (NL 43 for PEs and NL 59 for PCs) from the [M+Na] + was monitored. A full list of the experiments performed and the ions monitored is detailed in Table 2 2 In all eight of the ions sampled, the fragment ion intensity decreased within the infarcted myocardium (Figure s 2 12 and 2 13 ). Furthermore, both P Cs and PEs demonstrated the same pattern in terms of % signal decrease; PLA 2 demonstrated the greatest activity towards docosahexanoic acid, followed by arachidonic acid, oleic acid, and finally linoleic acid. Interestingly, these experiments also suggest that PLA 2 may have a greater activity following MI towards intact PEs than PCs. Multivariate data analysis and MS n also shed light upon a number of sphingolipids and their localization following LAD coronary artery ligation. In particular, m/z 725 was identified by PCA as positively correlating with the infarcted myocardium. Based on targeted MS n experiments, t his ion was identifie d as a sphingomyelin specifically the [M+Na] + of SM(d18:1/16:0). In addition to m/z 725, m/z 264 demonstrated increased s ignal in infarcted myocardium. Th is ion at m/z 264 is hypothesized to be the d18:1 sphingosine base, which can be formed from enzymatic cleavage or source fragmentation from ceramides, sphingomyelins, or sphingosine 1 phosphate. Alt hough
76 PCA and PLS DA d id not identify any ceramides as colocalizing with the d18:1 sphingosine base or sphingomyelins, the most intense fragment ion for ceramides containing the d18:1 sphingosine base ( m/z 264) 93 was observed in the MS 2 spectrum of m/z 520. This fragment demonstrated localization in the infarcted myocardium, yielding a similar localization to both the LPCs and the LPEs. As discussed earlier, the primary ion at m/z 520 within infarcted myocardium was determined to be an LPE; however, the sphingosine fragment ion suggests the presence of a dehydrated ceramide containing both the d18:1 sphingosine base and palmitic acid that is the [M+H H 2 O] + of Cer (d18:1/16:0). In addition to the ceramide containing a palmitic acid tail, the most abundant ceramides in myocardium as determined by Knapp et al. 94 were all submitted to MS 2 imaging. For these experiments, the [M+H H 2 O] + was fragmented, and the fragment ion corresponding to th e d18:1 sphingosine base at m/z 264 was mapped (Figure 2 14 ). In all instances, this fragment ion demonstrated localization in the infarcted myocardium, substantiating the previously determined localization of sphingomyelins and the sphingosine base. For illustrative purposes, an isobaric ion to Cer (d18:1/22:0) at m/z 604 demonstrating localization in the perfused myocardium is also shown. In addition to the lipids discussed above, a number of other metabolites were found to be affected as a result of th e LAD coronary artery ligation. PCA analysis demonstrated a number of low mass ions positively correlating with the perfused myocardium, including m/z 132, 162, and 348. Like the ion at m/z 132 (creatine), m/z 162 and 348 are hypothesized to be water sol uble metabolites that leak into the interstitial fluid following r upture of the plasma membrane. Targeted MS 2 experiments
77 were conducted in perfused myocardium to identify these metabolites. Fragmentation of the ion at m/z 348 demonstrated a single abundant ion at m/z 136 (NL 212), corresponding to protonated adenine (Figure 2 15 ) Thus, the ion at m/z 348 was identified as the [M+H] + of adenosine monophosphate (AMP). MS 2 analysis of m/z 162 resulted in two fragment ions of similar abundance, m/z 103 and m/z 60, suggesting that m/z 162 is the [M+H] + of carnitine (Figure 2 16). The fragment ion at m/z 103 (NL 59) corresponds to a loss of trimethylamine, and the fragment ion at m/z 60 (NL 102) corresponds to a loss of 3 hyd roxybut 3 enoic acid. At risk m yocardium and perfused m yocardium PCA was also conducted on an MSI dataset collected from cardiac tissue following LAD ligation to determine further markers for at risk myocardium, in addition to TAGs. In doing so, represent ative mass spectra from at risk myocardium and perfused myocardium were extracted as detailed above. The 2 dimensional scores plot constructed from PC 1 and PC 2 following PCA demonstrates separation between the two sample groupings (Figure 2 17 ) with PC 1 accounting for 61.4% of the variance. Furthermore, PC 1 dictated separation between at risk and perfused myocardium. Analysis of the sample groupings demonstrated a tighter cluster for the perfused myocardium, suggesting a larger amount of biological variation within at risk myocardium. A 1 dimensional loadings plot was created for PC 1 to determine markers for at risk myocardium (Figure 2 1 8 ) As expected, ions corresponding to TAGs (e.g., m/z 879 and 903) positively correlated with the at risk myocardium, and a number of intact PCs (e.g., m/z 806, 820, and 848) positively correlated with perfused myocardium. Furthermore, carnitine at m/z 162 and AMP at m/z 348 also positively correlated with
78 perfused myocardium. A number of unexpected findings were gleaned from analysis of the loadings plot. Namely, three intact PCs, m/z 780, 808, and 832, were found to load with the at risk tissue. All of these PCs were identified by MS 2 as being sodium adducts of PCs that contain either linoleic acid ( m/z 7 80 and 808) or arachidonic acid ( m/z 832). Furthermore, the loadings plot demonstrated an ion at m/z 337, presumably a source fragment from linoleic acid containing lipids, also positively correlated with at risk myocardium. Finally, the two TAGs with th e highest loadings coefficients, m/z 879 and 903, were putatively identified as linoleic acid containing TAGs. Thus, linoleic acid may physiological response to myocardial ischemia. At risk myocardium and infarcted m yocardi um PCA was also conducted on spectra extracted from the at risk and infarcted myocardium. The scores plot, displayed in Figure 2 19 demonstrates separation between these two regions along the first principal component, which accounted for 77.2% of the va riance. The second principal component accounted for a 10 fold smaller amount of the variance (6.7%) and largely accounted for the biological variation within the sample groupings, the majority of which occurred in the at risk myocardium. A 1 dimensional loadings plot was then created to analyze the spectra features that drive the separation along the first principal component (Figure 2 20 ). Unfortunately, little new information was gleaned from this analysis as the separation was largely dictated by the differences between the lysoPL and intact PL content of the two tissue regions. PCA of all three regions of i nterest In addition to the PCA analysis conducted above, PCA was also performed to analyze all three regions (i.e., infarcted, perfused, and at ri sk myocardium) in a single analysis. The scores plot from this analysis is displa yed in Figure 2 21 in which t he
79 three regions of myocardium are well separated on the first and second principal components. The first principal component demonstrated sepa ration between the infarcted and perfused myocardium. This result was not unexpected, as the first principal component by definition carries the largest amount of variance (59.3% in this case), and the most significant histological and chemical difference s are observed between these two regions. Accounting for 23.7% of the variance, the second principal component largely accounted for distinction between at risk myocardium and perfused myocardium. Alt hough sufficient separation was achieved in the scores plot, the loadings plots were examined to determine if the markers determined by this analysis agree with the two region comparisons discussed above Initially, the loadings for the first principal component were examined (Figure 2 22 ). Similar to the two region comparison, the majority of the variation on the first principal component was found in the intact PLs mainly loading with perfused myocardium, and the lyso PLs mainly loading with the infarcted myocardium. As exhibited in F igure 2 23 p rincipal component 2 also yielded similar features for the two and three region comparisons. Most importantly, a number of TAGs loaded with the at risk myocardium. Furthermore, various sodium adducts of intact PLs ( m/z 766, 780, 790, 808, a nd 832) loaded with the at risk myocardium, further suggesting that greater concentrations of sodium and possibly linoleic acid, exist in the at risk myocardium. Comparing multiple biological s amples The feasibility of applying the developed multivariate data analysis method to independent biological samples was also investigated. To greatly minimize any variation that might occur in sample preparation, all biological samples were thaw
80 mounted onto a single microscope slide. In the first attempt, a total of 3 tissues (2 infarct and 1 sham surgery) were analyzed. Upon availability of more biological specimens, the procedure was repeated with a total of 5 tissues (2 infarct and 3 sham surgeries) with similar results The results from the first comparison a re shown in Figure 2 24 Similar to PCA of the infarcted and perfused myocardium from a single biological sample, separation between the infarcted myocardium and the perfused myocardium was observed on the first principal component. Furthermore, the sham clustered with the perfused myocardium from the ligation surgeries. The tight clustering of the sham surgeries and perfused myocardium suggests that the perfused myocardium is either unaffected by the local blood deprivat ion in the left ventricle or that the chemical alterations occurring cannot be detected with this analysis. The second principal component, accounting for only 10.7% of the variance, dictates the sample to sample variation. It should be noted that the pre vious analysis was conducted with the infa r cted and perfused myocardium from ligation surgeries and control myocardium from sham surgeries; however, at risk myocardium was omitted from the analysis. At risk areas from the two ligation surgeries were subse quently added to further evaluate the potential to use this multivariate methodology on multiple biological samples. The results from this analysis are displayed in Figure 2 25 The inclusion of all three tissue regions and multiple biological samples se rved to complicate interpretation of the PCA scores plot; however, the first principal component (accounting for 58% of the variance) still dictated separation between the infarcted myocardium and the perfused
81 myocardium. The second principal component (a ccounting for 15.5% of the variance, which previously dictated separation between at risk myocardium and perfused myocardium when only one biological sample was included, still demonstrates this separation, but also splits the biological sam ples for infarc ted myocardium. Accordingly, interpretation of the loadings plots is complicated by the splitting of the infarcted myocardium along principal component 2. A lthough a 1 dimensional loadings plot of principal component 1 (Figure 2 26 ) can be used in this sc enario to determine localization for ions in the infarcted and perfused myocardium, a 1 dimensional loadings plot of principal component 2 (Figure 2 27 ) is insufficient to fully interpret ions localizing within the at risk myocardium. In this instance, a 2 dimensional loadings plot (Figure 2 28 A ) or alternatively a biplot (Figure 2 28 B) would be much more informative as the at risk myocardium should positively correlate with ions that have negative loadings coefficients on principal component 1 and posit ive loadings coefficients on principal component 2. Regardless, this multivariate data analysis method has demonstrated the ability to distinguish condition from control despite the presence of biological variation. Un biased multivariate data a nalysis and tandem mass s pectrometry In addition to the methodology discussed above, an un biased PCA approach was conducted similar to that conducted by van Hove et al. 50 In this approach, each individual pixel from the dataset was deemed an independ ent sample (excluding pixels off tissue). Initially, spectral and spatia l data were extracted from the using an in house data converter. Following data extraction, peak picking and alignment were performed using an algorithm developed at the FOM Institut e AMOLF to reduce the size of the dataset and shorten computing time. Then the data was normalized to the TIC, mean
82 centered, and normalized to the standard deviation of the sample means for unit variance. Following normalization, a first iteration of P CA was performed. PCA scores images were generated for the first 4 principal components. Invariably, the first principal component scores image demonstrated separation of pixels off tissue and pixels on tissue. Pixels that demonstrated a score higher th an a user defined threshold were then removed. It should be noted that this threshold is adjusted so as not to exclude pixels corresponding to areas on tissue. Finally, mass features corresponding to known matrix and background analytes were removed. Aft er the above data processing a second iteration of PCA was performed on the remaining data Initially, scores plots were generated for the f irst 20 principal components. Alt hough the maximum number of principal components is equal to the number of varia bles, principal components with a low percentage of the variance rarely demonstrated relevant localization. To this point, the first 20 principal components appear to be a reasonable cutoff f or datasets with approximately 1000 variables (i.e., 2 % of the p rincipal components). Once a suitable number of principal components were displayed, loadings plots were generated for principal components that demonstrated r elevant localization in the regions of interest Upon analysis of cardiac tissue following coronary artery ligation, the first principal component demonstrated separation between infarcted and perfused myocardium and the third principal component demonstrated separation between the at risk myocardium a nd the remain der of the tissue (Figure 2 29 ) Loadings plots from these principal components correlated well with the guided method (Figure 2 30 ) For example, the first principal component in both methodologies demonstrated high loadings coefficients
83 for intact PC and PE s correlating with the per fused myocardium, and similarly with L PCs and L PEs correlating with infarcted myocardium. Similarly, principal component 2 in the guided method and principal component 3 in the unbiased method displayed TAGs correlating with at risk myocardium. In additio n to the three ion distributions discussed above, a fourth ion distribution was detected in the negative portion of principal component 19 despite the relatively low contribution (0.5 %) to the total variance (Figure 2 31 ). This ion distribution, though occupying just a few pixels of the at risk myocardium demonstrates high loadings coefficients for ions at m/z 400, 424, 426, and 428 (Figure 2 31 ). Targeted and whole tissue imaging MS 2 experiments were conducted to ident ify the structure of these ions. Initial analysis yielded fragment ions resulting from a neutral loss of 59 presumably the loss of trimethylamine, for m/z 400 and 428 Furthermore, a neutral loss of 161, attributed to the loss of free carnitine, was observed for all four ions as well. Thus, the four ions were identified as palmitoyl carnitine, linoleoyl carnitine, oleoyl carnitine, and stearoyl carnitine for m/z 400, 424, 426, and 428 respectively. The MS 2 spectrum fragmentation pattern and MS 2 images of two characteristic fragments for palmitoyl carnitine are shown in Figure 2 32 as an example. Interestingly, the two unsaturated acylcarnitines (linoleoyl and oleoyl) exhibit an abundant neutral loss of 179 Alt hough this fragment presumably results from successive losses of free ca rnitine followed by water from the dehydrated fatty acid tail, it is currently unclear as to why unsaturated and saturated acylcarnitines exhibit different fragmentation patterns. Conclusions This research is the first presentation of MSI displaying potent ial markers for MI within intact cardiac tissue. First, creatine, a metabolite involved in the creatine kinase
84 metabolic pathway, demonstrated decreased ion signal within areas of infarcted myocardium agreeing well with TTC stained hearts following LAD c oronary artery ligation. The spatial localization of creatine proved to be a valuable indicator for infarcted myocardium; however, creatine did not distinguish between the affected and at risk regions of tissue. Following the targeted study for creatine a combination of multivariate data analysis and tandem mass spectrometry was conducted on an MSI data set to identify PC, LPC, and LPE markers of MI. These studies indicated that many LPCs and select LPEs demonstrated increased ion signal and conversel y, select intact PCs and PEs demonstrated decreased ion signal within the area of infarction. Presumably, the lysoPLs are being formed via enzymatic hydrolysis by PLA 2 It is also interesting that two intact PCs found to decrease in infarcted tissue presumably contain arachidonic acid in the sn 2 position as i t has been reported that PLA 2 and arachidonic acid may play a role in the cellular response to myocardial infar ction, specifically involving protection against ischemic cell death. 95 The complementary localization of LPC 18:0 at m/z 546 and its precursor, PC (18:0/20:4) at m/z 848, support this hypothesis. Further research w as conducted concerning the fatty acid specificity of PLA 2 and it was concluded that for stearic acid containing PLs, PLA 2 appears the most active for PLs containing docosahexanoic acid followed by arachidonic acid. This trend was observed using targeted MS 2 experiments for intact PCs and PEs. The PL findings of this study suggest that PLA 2 activity increased in infarcted myoc ardium following LAD ligation. In addition to the glycerophospholipids discussed above, a number of sphingolipids were found to lo calize within the infarcted myocardium. Specifically, the
85 d18:1 sphingosine base, SM (d18:1/16:0), and nearly all of the abundant ceramides were found to localize in areas of necrosis. At present, the reason for the localization of sphingomyelin is uncle ar; however, arachidonic acid liberated by PLA 2 during ischemia has been hypothesized to stimulate sphingomyelinase (SMase) activity. 96, 97 SMase is an enzy me that serves to hydrolyze intact sphingolmyelins, forming ceramides, 98 possibly offering an exp lanation for the increased content of ceramides in infarcted myocardium. Furthermore, Nakane et al has speculated that an increase in both gangliosides and sphingosine accompany this increase in ceramide concentration following ischemia. 96 Alternatively, the d18:1 sphingosine base may result from fragmentation of the dehydrated ceramides within the MALDI ionization source. The at risk myocardium, which provides the most interesting biological relevance concerning MI, was also characterized using targeted tandem MSI and multivariate data analysis techniques. This tissue type, although considered ischemic, has the potential to become viable if reperfusion (i.e., the restoration of blood flow) is performed, thereby decreasing the overall size of the infarction and limiting cell death. The targeted tandem MSI studies demonstrated that TAGs were localized with in the at risk are as of infarcted hearts following LAD ligation Although the current spatial resolution of this method does not permit a direct comparison of microscopic and MSI images, it is believed that lipid droplets are responsible for TAG localization in at risk zon es. Additionally, a number of linoleic acid containing intact PLs and TAGs were found to be positively correlated with the at risk myocardium following PCA. Finally, a number of acylcarnitines were found to localize in limited areas of the at risk myocar dium. Previous research has suggested that acylcarnitines accumulate during acute
86 myocardial ischemia, demonstrating approximately a 10 fold increase in extracts following a rabbit model of ischemia. 59 This study indicated that the likely cause of this accumulation of acylcarnitines was inhibiti on of fatty acid oxidation in the mitochondria. 59, 85 As inhibition of fatty acid oxidation was also implicated in the accumulation of TAGs following MI, it is not surprising that TAGs and acylc arnitines colocalize in the at risk myocardium. The implication of these findings is that inhibition of fatty acid oxidation may play a key role in myocardial protection under ischemic conditions. It should be noted that tissue specific markers do not imm ediately result in blood or plasma markers of MI. Nevertheless, it may be beneficial to investigate these tissue specific markers in biological fluids using targeted, sensitive methods (e.g., selected reaction monitoring with triple quadrupole MS 2 coupled to liquid chromatography). In particular, small metabolites such as creatine, free carnitine, and AMP that may leak into the interstitial fluid following infarction may demonstrate an increase in blood borne concentration. The discovery of one or a numb er of small molecule marker s would offer a valuable alternative to the traditional protein markers utilized in a clinical setting. MALDI MSI offers advantages over traditional myocardial infarction characterization techniques such as histological staining and biological assays. In particular, the ability to detect, identify, and localize small molecular species allows for identification of putative markers for MI with a single analytical technique. Furthermore, a wealth of data was collected relatively qu ickly (e.g., approximately one hour for a single cardiac section), making these experiments prime for multivariate data analysis This research has presented two
87 PCA methodologies (guided and unbiased). W hen used in a complementary fashion these techniques provide an extremely powerful basis for unprecedented exploratory MSI research. Thus MALDI MSI, coupled with multivariate data analysis has the potential to be applied to a wide variety of tissue appl ications, making it a powerful tool for diseased state characterization and biomarker discovery.
88 Figure 2 1. MSI analysis of cardiac tissue following ligation surgery. A ) Enzymatic reduction of TTC to TPF, B) Photograph of the upper half of a heart stained with TTC following LAD coronary artery ligation. Unstained tissue, outlined in blue, indicates tissue dama ge from myocardial necrosis. C ) MS image of m/z 132 normalized to the TIC and D ) MS 2 image of m/z 13 2 90 from the lower half of the same heart depicted in B Lower ion signal was observed in the infarcted region near the right side of the tissue section.
89 Figure 2 2 MS image of m/z 132 normalized to the TIC from control cardiac tissue.
90 Figure 2 3 Mass spectra of t he lipid region collected in different regions of myocardium. Spectra were acquired from either A) infarcted or B ) perfused zones of cardiac tissue.
91 Figure 2 4. MALDI MS spectra acquired from sham surge ry myocardium The spectra were acquired from locations with either A) in vitro digestion by PLA 2 or B) no enzymatic alteration.
92 Fi gure 2 5. Example of TAG identification and localization. A) MS 2 spectrum of m/z 881 from the at risk region of cardiac tissue following LAD coronary artery ligation and MS 2 images of B) m/z 881 599 and C) 881 625 from the same tissue section. The primary ion at m/z 881 is TAG 52:2.
93 Figure 2 6. Multivariate data analysis of infa rcted and perfused myocardium. A ) PCA scores plot and B ) PLS DA scores plot for mean centered, TIC normalized data. Each sample consists of 25 spectra taken along a horizontal line within either infarcted (green crosses) or healthy tissue (red triangles) from a single tissue section following an LAD coronary artery ligation. The ovals indicate the 95% confidence interval for the sample groupings.
94 Figure 2 7 PCA loadings pl ots dictating separation between infarcted and perfused myocardium. The data were processed using 2 methods: either A) mean centered data or B ) autoscaled data.
95 Table 2 1. Statistical significance testing for various ions in infarcted and perfused myocardiu m. m/z Infarcted Signal/TIC Perfused Signal/TIC Pooled Standard Deviation Direction t MS n Identification Ion 273 0.01490 0.01330 0.001 8 1.12 DHB [2M 2H 2 O+H] + 459 0.00088 0.00037 0.000 2 3.26 LPC 16:0 [M C 3 H 9 N+Na] + 487 0.00163 0.00051 0.000 5 2.80 LPC 18:0 [M C 3 H 9 N+Na] + 496 0.00236 0.00085 0.000 8 3.84 LPC 16:0 [M+H] + 518 0.00297 0.00083 0.0005 4.97 LPC 16:0 [M+Na] + 520 0.00261 0.00069 0.0008 2.96 LPE 18:0 LPC 18:2 [M+K] + [M+H] + 524 0.00748 0.00196 0.002 1 3.26 LPC 18:0 [M+H] + 534 0.00130 0.00072 0.000 4 1.91 LPC 16:0 [M+K] + 544 0.00317 0.00241 0.000 3 3.29 LPC 18:1 LPC 20:4 [M+Na] + [M+H] + 546 0.00619 0.00150 0.0013 4.29 LPC 18:0 [M+Na] + 782 0.00571 0.00600 0.000 9 0.41 PC (16:0/18:1) [M+Na] + 810 0.00651 0.00612 0.000 6 0.85 PC (18:0/18:1) [M+Na] + 832 0.00767 0.01190 0.00 20 2.64 PC (18:0/20:4) [M+Na] + 848 0.00350 0.00818 0.000 3 19.50 PC (18:0/20:4) [M+K] +
96 Figure 2 8 MS 2 spectrum using CID of m/z 520 from infarcted cardiac tissue. Fragment ions are identified from two isobaric lipid ions: the [M+K] + of LPE 18:0 (squares) and the [M+H] + of LPC 18:2 (stars). Furthermore, the fragment (i.e., m/z 18:0 and m/z from cardiac tissue following LAD coronary artery ligation are shown as an inset. The LPE contributes more ion signal within the infarcted zone than the LPC.
97 Figure 2 9. Example of MS n identification and imaging of an LPC. A ) MS 2 spectrum using CID of m/z 546 and B ) MS 3 spectrum using CID of m/z 546 487 from infarcted cardiac tissue. The major fragment in A is a NL of 59, indicating a c ationized PC. Furthermore, in B the 22 Da difference between m/z 363 and 341 indicate a sodiated PC. Also, the fragment ion at m/z 147 is indicative of sodiated cyclophosphane, a characteristic ion of sodiated PCs. The ion was identified as the [M+Na] + of LPC 18:0. The structure and an MS 2 image of m/z 546 487 from cardiac tissue following LAD coronary artery ligation is shown as an inset in A
98 Figure 2 10 MS 2 spectrum of m/z 848 using PQD from infarcted cardiac tissue. A NL of 59 indicates a cationized PC. The 38 Da difference between m/z 655 and 62 7 and the fragment ion at m/z 163 (potassiated cyclophosphane) indicate a potassiated PC. MS 3 demonstrated a minor NL of 284, indicating a stearic fatty acid tail in the sn 1 position of the glycerol backbone (data not shown). Therefore, the ion was iden tified as the [M+K] + of PC (18:0/20:4). The structure and an MS 2 image of m/z 848 789 from cardiac tissue following LAD ligation are shown as an inset.
99 Figure 2 11 MS im ages normalized to the TIC from a heart following LAD coronary artery ligation. A ) the [M+K] + of LPE 18:0 at m/z 520, B ) the [M+Na] + of LPC 18:0 at m/z 546, C ) the [M+Na] + of LPC 16:0 at m/z 518, and D ) the [M+K] + of PC (18:0/20:4) at m/z 848.
100 Table 2 2 MS 2 imaging experiments performed ion the [M+Na] + ions from intact PC and PE ions containing stearic acid in the sn 1 position of the glycerol backbone. Intact PCs Intact PEs Species MS 2 Transition Species MS 2 Transition PC (18:0/18:1) m/z PE (18:0/18:1) m/z PC (18:0/18:2) m/z PE (18:0/18:2) m/z PC (18:0/20:4) m/z PE (18:0/20:4) m/z PC (18:0/22:6) m/z PE (18:0/22:6) m/z
101 Figure 2 12 MS 2 PC fragment ion intensity (loss of trimethylamine) for four common intact PCs in infarcted (blue) and perfused (red) myocardium. All four of these ions contain stearic acid in the sn 1 position. Error bars represent the standard deviation of the mean (n =30).
102 Figure 2 13 MS 2 PE fragment ion intensity (loss of ethenamine) for four common intact PEs in infarcted (blue) and perfused (red) myocardium. All four ions contain stearic acid in the sn 1 position but differ in sn 2 fatty acid tail Error bars represent the standard deviation of the mean ( n=30 ).
103 Figure 2 14 MS 2 images of select abundant ceramides in cardiac tissue following coronary artery ligation. All MS 2 experiments were performed on the [M+H H 2 O] + For contrast, an uniden tified ion that demonstrates complementary localization and is nomin ally isobaric to the ceramide a t m/z 604 is also displayed.
104 Figure 2 15 MALDI MS 2 spectrum of m/z 348 obtained from perfused myocardium following LAD ligation surgery. The MS image (right) and fragmentation pattern of AMP (center) which is proposed as the dominant species at m/z 348, is also displayed.
105 Figure 2 16 MALDI MS 2 spectrum of m/z 162 obtained from perfused myocardium following LAD ligation surgery. The MS image (top) and fragmentation pattern of free carnitine (bottom), which is proposed as the dominant species at m/z 162, is also displayed.
106 Figure 2 17 PCA scores plot demo nstrating the separation between at risk (red) and perfused (green) myocardium. Each sample represents an average of approximately 10 mass spectra.
107 Figure 2 18 Principal component 1 loadings plot dictating the separation between perfused and at risk myocardium. Additionally, the MS images of two ions, indicative of the two tissue regions, are displayed. The ion at m/z 348, identified as AMP is indicative of perfused myocardium, and the ion at m/z 879, identified as TAG (16:0/18:1/18:2), is indicati ve of the at risk myocardium.
108 Figure 2 19 PCA scores plot demonstrating the separation between at risk (red) and infarcted (green) myocardium. Each sample represents an average of approximately 10 mass spectra.
109 Figure 2 20 Principal component 1 loadings plot dictating the separation between perfused and infarcted myocardium.
110 Figure 2 21. PCA scores plot generated from sampling the infarcted (green), perfused (blue), and at risk myocardium (red) following coronary artery ligation. The ellipses represent the 95% confidence interval of the sample groupings.
111 Figure 2 22. Principal component 1 loadings plot following PCA analysis of all three regions of myocardium. This principal component, accounting for 59.3% of the variance, dictates separation between the infarcted (positive portion) and perfused myocardium (negative portion).
112 Figure 2 23. Principal component 2 loadings plot following PCA analysis of all three regions of myocardium. This principal component, accounting for 23.7% of the variance, dictates separation between the at risk (positive portion) and perfused myocardium (negative portion).
113 Figure 2 24. PCA scores plot generated from analysis of three biological samples. Principal component one demonstrates separation between infarcted myocardium from ligation surgeries (negative portion) and perfused myocardium from both sham and ligation surgeries (positive portion).
114 Figure 2 25 PCA scores plot generated from analysis of multiple biological samples. Furthermore, all three tissue regions (infarcted, perfused, and at risk myocardium) were included in the analysis.
115 Figure 2 26 Loadings plot of principal component 1 generated from PCA of multiple biological samples including three different regions of myocardium. This principal component largely dictates the separation between infarcted myocardium and the remaining tissue regions.
116 Figure 2 27 Loadings plot of principal component 2 generated from PCA of multiple biological samples including thr ee different regions of myocardium.
117 Figure 2 28 Graphical representation of loadings from PCA of multiple biological samples with the inclusion of all three tissue regions. Two di fferent methods are displayed: A) a 2 dimensional lo adings plot and B) a biplot.
118 Figure 2 29 Principal component scores images from unbiased PCA analysis. Principal component 1 accounting for 13.2% of the total variance, demonstrates separation between the perfused (negative) and infarcted myo cardium (positive). Principal component 3 accounting for 5.3% of the total variance, demonstrates separation between the at risk (negative) and perfused myocardium (positive).
119 Figure 2 30 L oadings plots from unbiased PCA analysis. Principal component 1 (top) demonstrates separation between infarcted and perfused myocardium and principal component 3 (bottom) demonstrates separation between perfused and at risk myocardium.
120 Figure 2 31 Loadings plot of principal component 19 (PC 1 9) from unbiased PCA analysis. The scores images for the negative portions of PC 19 and principal component 3 (PC 3) are also displayed.
121 Figure 2 32 MS 2 spectrum of m/z 400 collected from cardiac tissue following experimental coronary artery ligation surgery. The fragmentation pattern of the identified ion (palmitoyl carnitine) and MS 2 images of two characteristic ions displaying increased signal in the at risk myocardi um are also displayed.
122 CHAPTER 3 DEVELOPMENT OF 9 AMINOACRIDINE AS A DUAL MODE MALDI MATRIX FOR SMALL MOLECULE MASS SPECTROMETRIC IMAGING STUDIES OF MYOCARDIAL INFARCTION Introduction Matrix assisted laser desorption/ionization (MALDI) is a soft ionization mechanism suitable for the analysis of small and large biomolecules. Since the two breakthrough MALDI MS publication s by Tanaka et al 5 and Karas et al 3 a number of small organi c compounds (matrices) have become well established for the analysis of lipids, 90, 99 proteins, 1, 3, 100 and peptides 101 in positive ionization mode. In general, this group of compounds consists of w eak acids that exhibit strong absorption in the ultraviolet portion of the electromagnetic spectrum affording the ability to protonate analytes to form positive ions As MALDI is still a relatively young ionization method, an all purpose negative mode ma trix has primarily eluded researchers. Recently, a number of small organic compounds have been proposed as potential negative mode MALDI matrices. As opposed to the positive mode matrices, which typically exhibit weak acidity via functional groups such as carboxylic acid, these compounds are generally basic in nature. The most promising set of these potential matrices are aromatic compounds containing at least one amino group. The structures of three of these potential matrices 9 aminoacridine (9 AA) 1 ,5 diaminonaphthalene (DAN) and 1,8 bis(dimethyl amino)naphthalene (DMAN), are displayed in Figure 3 1. exhibiting a pK b of approximately 1.9 102 Although DMAN exhibits superbasic properties, recent research has revealed that DMAN produces an unstable sign al under the high vacuum source conditions employed by most MALDI MS instruments 103, 104
123 Th is signal instability is likely a result of the relatively high volatility of DMAN causing matrix crystals to sublime from the ta rget surface over time. DAN is a relatively weak base (pK b ~ 9.56 ), and has recently gained interest in the MSI community 103 This matrix gangliosides. 105 Although there have been no reports of problems with signa l stability, the relative ly weak basicity of this matrix raises concerns about the breadth of analytes that can be analyzed in negative mode. The most promising of these three c andidates, 9 AA is a fairly strong base (pK b ~4 ) and was first reported as a suitable negative mode MALDI matrix by Vermillion Salsbury et al. in 2002. 106 Since this initial stu dy 9 AA has been re ported as an effective negative mode MALDI matrix for low molecular weight acids, 107 endogenous metabolites, 108, 109 and a variety of lipid classes. 109 111 There has yet to be a report describing signal stability issues under high vacuum, and the basicity appears to be suitable to ionize analytes of interest that are ineffectively ionized in positive mode by acidic matrices such as DHB Furthermore, there have been a number of publications describing th e use of 9 AA as a matrix for MALDI metabolomics 108, 109 and MALDI mass spectrometric imaging (MSI) applications. 111 113 Previous research has demonstrated the benefits of multivariate data analysis, namely principal component analysis (PCA), for the analysis of MALDI MSI datasets collected in positive mode with DHB 48 The present research utilized a rat model for myocardial infarc tion (MI), simulating a heart attack resulting in significant tissue death This model, known as a left anterior descending (LAD) coronary artery ligation, deprives oxygenated blood to the left ventricular myocardium leaving the right ventricular
124 myocard ium largely unaffected. The result of this ligation model is that both the injured and control samples can be interrogated within a single tissue section. Using MSI and PCA, a number of tissue specific alterations in lipid and metabolite profiles were ob served. 48 Although this endeavor interrogated a number of compound classes in positive mode, a number of analytes were excluded from the study due to inefficient ionization in positive mode. This chapter will demonstrate the utility of 9 AA as a suitable dual mode MALDI matrix (positive and negative mode) for the analysis of MI with MSI and PCA. Experimental Chemicals and r eagents HPLC grade water (H 2 O) an d HPLC grade methanol (MeOH) were purchased from Fisher Scientific ( Fair Lawn, NJ ). 100% ethanol (EtO H) was purchased from Decon Labs (King of Prussia, PA). 9 amino acridine (9 AA) and 9 aminoacridine hydrochloride hemihydrate were purchased from MP Biomedicals (Solon, OH). 2,5 dihydroxybenzoic acid (DHB) and nicotinamide adenine dinucleotide, reduced (NADH) were purchased from Sigma Aldrich (St. Louis, MO). nicotinamide adenine dinucleotide in the non reduced form (NAD + ) was purchased from Acros Organics (Geel, Belgium) Lipid standards, namely 1 heptadecanoyl 2 hydroxy sn glycero 3 phosphocholine ( LPC 17:0), 1 hexadecanoyl 2 (9Z,12Z octadecadienoyl) sn glycero 3 phosphoethanolamine (PE (16:0/18:2)), L phosphatidylserine from porcine brain, and L phosphat idylinositol from bovine liver were purchased from Avanti Polar Lipids ( Alabaster, AL). 9 AA was dissolved in 70:30 EtOH:H 2 O (v/v) to a final concentration of 6 mg/mL to serve as a MALDI matrix. For comparison, DHB was dissolved in 70:30 MeOH:H 2 O (v/v) to a final concentration of 40 mg/mL. NAD + and NADH were dissolved
125 in H 2 O and 0.01M NaOH, res pectively to a final concentration of 1 mg/mL to serve as nucleotide standards. The lipid standards were either dissolved (in the case of a powder) or diluted (in the case of a solution) to 1 mg/mL in EtOH for all standard analyses. Extraction of 9 AA free b ase As the free base of 9 AA is relatively expensive (~1g for $70) and the hydrochloride salt is much less so (~50g for $70), an acid base extraction was perfo rmed, similar to that previously reported. 107 Approximately 400 mg of the hydrochloride salt was dissolved in 25 mL of boiling H 2 O. Following removal from heat, sodium carbonate was added for a lkalinization. The free base was then extracted with CHCl 3 The CHCl 3 layer was then obtained and the solvent was evaporated. The resulting solid was reconstituted in 70:30 EtOH:H 2 O (v/v). The suspension was centrifuged, and the supernatant was utilized as a MALDI matrix. UV/ Vis absorption s pectrophotometry All experiments were conducted on a double bea m Hewlett Packard 8450A UV/Vis s pectrophotometer equipped with a diode array dete ctor. Matched quartz cuvettes were used for the blank and sample solutions. Blank solutions were prepared according to the respective solvent systems for the MALDI matrices: 70:30 EtOH:H 2 O (v/v) for 9 AA and 70:30 MeOH:H 2 O for DHB. The wavelength range on the instrument was set to 200 900 nm. Prior to analysis of each sample, the baseline (or balance) was measured by analyzing the blank solution in both the sample and reference cuvettes. The blank solution in the reference cuvette was then replace d wit h the MALDI matrix, and UV/V is absorption spectra were obtained.
126 Biological sample p reparation All animal procedures were conducted in accordance with guidelines published in the Guide for the Care and Use of Laboratory Animals (National Research Council, National Academy Press, Washington, DC, 2010) and were approved by the Animal Care Committee of Saint Louis University. A 24h LAD ligation (6 animals) was performed as previously described to induce myocardial ischemia 48, 78 In addition to the ligation surgery case group two separate control groups sham (6 animals) and nave (3 animals) were prepared. For the sham control, all surgical steps were conducted for the LAD ligation, except the LAD suture was tied loosely around the artery, retaining normal blood flow to the left ventricle. For the nave control, no surgery was performed pri or to euthanization. 24 hours following either LAD ligation or sham surgery the rats were euthanized with pentobarbital (~800 mg/kg, i.p.). The hearts were then excised, flash frozen in liquid nitrogen, and stored at 80 C until further use. Hearts fro m all sample groupings (infarction, sham, and nave) were bisected along a transverse plane at the midventricular level. The upper half of the heart was subjected to 2,3,5 triphenyltetrazolium chloride (TTC) staining as previously described. 79 The lower hal f of the heart was sectioned on a Microm HM 505E cryostat (Waldorf, Germany) at 25 C to 10 m thickness along the same transverse plane as the bisection. 48 Following sectioning, tissues were thaw mounted atop a microscope slide. In order to ensure the ability for efficient tiss ue to tissue comparison, tissues from different sample groupings were thaw mounted atop the same microscope slide. Furthermore, serial sections were obtained on two separate microscope slides to compare the two prepared MALDI matrices (9 AA and DHB) Tis sue sections were then stored at 80 C until further use.
127 Prior to matrix application, tissues were dehydrated in a vacuum desiccator for approximately 45 minutes. Immediately following desiccation, tissue sections were coated with MALDI matrix (9 AA or DHB ) using a Type A3 Glass Me inhard Nebulizer (Golden, CO). Care was taken to coat serial sections with 9 AA and DHB for the best comparison. Approximately 150 mg 9 AA (25 mL of solution) or 400 mg DHB (10 mL of solution) was deposited atop the microscop e slide before adequate crystallization was observed atop the tissue sections. It should be noted that excessive tissue wetting was avoided to minimize analyte migration Mass spectrometry and i maging All experiments were performed on a Thermo Scientific LTQ XL (San Jose, CA ) equipped with a Thermo MALDI ionization source. The MALDI ionization source consisted of a N 2 laser ( =337 nm) with a repetition rate of 60 Hz and an observed laser spot diameter of approximately 100 m. The source regio n of the LTQ was maintained at intermediate pressure (~75 mTorr) for all imaging experiments unless otherwise noted. For all experiments with 9 AA in negative ionization mode a laser energy of 7 .5 and 3 laser shots per laser stop were utilized. For a ll experiments with DHB and 9 AA in positive ionization mode a laser energy of 4 and 3 laser shots per laser stop were utilized. To preserve the number of laser shots per laser stop, automatic gain control (AGC) was toggled off. MS 2 experiments were conducted with an isolation width of 1.2 amu and a collision energy between 30 and 45 A.U. A raster step size of 150 was utilized for all imaging experiments unless otherwise noted Statistical a nalysis Following MS imaging, PCA was conducted to deter mine tissue specific markers of MI. Briefly, representative samples from each region of tissue were extracted from
128 the centroided imaging dataset. Each sample consisted of 5 15 spectra averaged from a horizontal line within the region of interest. Each individual sample was saved as a .CSV file and all of the samples were grouped according to the region of interest. The resulting groups were then compressed into a .ZIP file and imported into the Metaboanalyst Web Server. Once imported, a mass tolerance of 0. 75 amu was utilized to account for any mass shifts due to space charge effects. Major matrix peaks, resulting from either 9 AA or DHB were then removed. These matrix peaks were determined by manual inspection of regions surrounding the tissue secti ons. All intensities were then normalized by the total ion current (TIC). The data were then mean centered or alternatively autoscaled for each m/z value. PCA was then performed, and 2 dimensional scores plots were generated to visualize the sample grou pings in principal component space Depending on the principal component that dictated the separation between sample groupings, a 1 dimensional loadings plot was generated to determine m/z values of interest. Results Standard c haracterization Prior to use with cardiac tissue following coronary artery ligation, a number of standards were characterized using 9 AA as a negative mode MALDI matrix. Initial experiments examined two previously published solvent systems for 9 AA: 1) 60:40 IPA:ACN (v/v ) and 2) 70: 30 EtOH:H 2 O (v/v) 9 AA proved to be fully soluble in the first solvent system at a concentration of 10 mg/mL. The same concentration prepared in the second solvent system generated a suspension. The suspension prepared in the second solvent system was spun down using a centrifuge, and the supernatant was collected to serve as the MALDI matrix. Both matrices were spotted on a 384 well
129 MALDI sample plate for characterization. Upon visual inspection of the spotted matrices, 9 AA prepared in solvent syste m 2 provided more homogenous crystallization than the first solvent system. It was later determined that the solubility of 9 AA was approximately 6 7 mg/mL in the second solvent system. Consequently, a MALDI matrix of 6 mg/mL 9 AA in 70:30 EtOH:H 2 O (v/v ) was used for standard characterization and tissue analysis Phosphatidylcho line (PC ) Previous experiments with MALDI ToF MS at low source pressure produce d [M CH 3 ] ions from PCs utilizing 9 AA as a MALDI matrix 114 To determine if these ions are produced at intermediate pressure, such as the source pressure on a Thermo MALDI LTQ XL, a 1 mg/mL lysophospatidylcholine (LPC) standard was prepared. The prepared standard LPC 17:0 (MW=509 Da) was mixed in a 1:1 ratio with the MALDI matrix, and spotted using the traditional dried droplet method. MS analysis in negative mode produced the s ame ions previously reported, with the [M CH 3 ] occurring at m/z 494. MS 2 was then conducted to characterize the fragmentation pattern of this ion The MS 2 spectrum of m/z 494 yielded an abundant ion at m/z 269, corresponding to the [M H] of heptadecanoic acid (Figure 3 2 ) Less abundant ions were observed at m/z 224 and 242 corresponding to the [M CH 3 H 2 O] and [M CH 3 ] respectively, of glycerophosphocholine. The same matrix/analyte spot was also analyzed in positive mode. Abundant i on s were observed at m/z 510, corresponding to the [M+H] + and, m/z 1019, corresponding to the [2M+H] + of LPC 17:0. Although abundant protonated monomer and dimer ions were observed, n o appreciable alkali metal adducts were observed in the positive mode analysis of this lipid standard with 9 AA. MS 2 fragmentation of the protonated monomer
130 at m/z 510 yielded the typical PC fragment at m/z 184 corresponding to the phosphocholine head group. Fragmen tation of the protonated dimer at m/z 1019 yielded ions at m/z 510 (the protonated monomer), m/z 496 (the demethylated protonated monomer), and m/z 524 (the methylated protonated monomer) Phosphatidylethanolamine (PE ) A major disadvantage reported in the literature for 9 AA as a negative mode MALDI matrix is the inability to distinguish PCs and PEs. 114 The characteristic ion of a PC in negative mode, the [M CH 3 ] and the characteristic ion of a PE with two less carbons contained within the fatty acid substituents the [M H] are isomeric Although this is a p roblem for instruments with only one stage of mass analysis, tandem MS can differentiate these isomers. A 1 mg/mL synthetic standard of PE (16:0/18:2 ) (MW = 715 Da ) was mixed in a 1:1 ratio with the prepared 9 AA matrix solution. The mixture was then spotted onto a MALDI target plate and allowed to dry with the aid of gentle heating. Negative mode MALDI MS analysis of the spotted analyte/matrix mixture yielded an ion at m/z 714, corresponding to the [M H] CID of m/z 714 (Figure 3 3 ) yields a number of abundant fragment ions, with the two most abundant fragment ions, m/z 279 and 255, corresponding to the carboxylate anion s of linoleic acid (18:2) and palmitic acid (16:0), respectively. Furthermore, the ions at m/z 452 and 476 correspond to the loss of dehydrated linoleic acid and palmitic acid, respectively. Once the two fatty acid tails are identified the remainder of the mass can be attributed to the glycerol back bone and the head group. In contrast, for the isomeric PC (18:0/18:2) one would expect fragment ions at m/z 279 and 284, corresponding to the carboxylate anions of linoleic acid (18:2) and stearic acid (18:0), respectively.
131 Phosphatidylserine (PS ) Further characterization of lipid standards using 9 AA as a MALDI matrix in negative mode was performed on an l ph osphatidylserine (PS) extract from porcine brain. MS analysis of this extract produced two abundant ions at m/z 701 and 788 (Figure 3 4 ). Furthermore, low abundant ions were observed at m/z 747 and 834 Interestingly, both pairs of ions were separated by 87 amu The lower mass ion in both pairs of ions is thought to be a source fragment resulting from the loss of serine. Furthermore, the intensity ratio of m/z 701 to 788 and m/z 747 to 834 are similar, suggesting that the degree of source fragmentatio n is maintained for both species The suspected intact PS ions were then subjected to MS 2 analysis, and the fragmentation pattern for PS ions in negative mode was determined. The MS 2 spectrum of m/z 788 is displayed in Figure 3 4 A Unlike the other phosp holipids examined, PS ions in negative mode exhibit a neutral loss related to the head group. More specifically, this neutral loss results from the cleavag e of serine (NL 87), further suggesting that the ions at m/z 701 and 747 in the MS spectrum are sour ce fragments. MS 3 of m/z 3 4 B). The ions at m/z 283 and 281 correspond to the sn 1 fatty acid tails of stearic acid (18:0) and oleic acid (18:1), respectively identifying m/z 788 as the [M H] of PS (18:0/18:1) Additionally, neutral losses on both sides of the ester linkage to the glycerol backbone are observed for both of these tails. Similar MS 2 fragmentation was observed for m/z 834; however, MS 3 fragmentation of m/z s uggested fatty acid tails of stearic acid (18:0) and docosahexanoic acid (22:6), identifying the precursor ion as the [M H] of PS (18:0/22:6). Thus, when one considers the PS fragmentation pathway, the
1 32 MS 2 spectrum confirms the presence of a PS and MS 3 yields information of the fatty acids bound to the glycerol backbone. Phosphatidylinositol (PI ) An extract of l phosphatidylinositol (PI) from bovine liver was also analyzed in the manner di scussed above. PIs have been sh own to readily ionize in negative mode with a variety of MALDI matrices, including 9 AA. 111, 115 Furthermore, the negative mode fragmentation pathways of PIs have been reported in depth; 116 therefore, the results of this study will only be briefly reported in this section. Analysis of the extract produced an abundant ion at m/z 863. Additionally, a number of low intensity ions were obser ved throughout the lipid region of the mass spectrum. Similar to PEs, MS 2 fragmentation of these ions produced abundant carboxylate anions indicative of the fatty acid substituents esterified to the glycerol backbone. For instance, MS 2 analysis of the pr ecursor ion at m/z 863 yielded ions at m/z 283 and 281, corresponding to the [M H] for stearic and oleic acid, respectively. Additionally, abundant neutral losses of the same fatty acids were observed at m/z 581 and 579. NAD + and NADH In addition to neg atively charged phospholipids, 9 AA has been reported to effectively ionize nucleotide species containing at least one phosphate group. 108, 109 Prior to tissue analysis, nicotinamide adenine dinucleotide (NADH), a water soluble nucleotide that has previously been observed in myocardial extracts, 109 was analyzed using the protocol discussed above. MS analysis of the spotted standard demonstrated an abundant ion at m/z 664, corresponding to the [M H] of NADH (Figure 3 5 ) Following MS analysis, CID was conducted on this ion to characterize the fragmentation pattern of NADH (Figure 3 6 ). The most abundant ions resulted from cleavages along
133 the two phosphate functional groups bridging nicotinamide and adenine ; the ions at m/z 408 and 397 result from cleavage of nicotinamide riboside and adenosine, respectively Additionally, cleavages are observed between the two phosphate groups, resulting in NLs of nicotinamide ribose monophosphate and adenosine monophosphate at m/z 346 and 335, respectively. Returning to the MS spectrum in Figure 3 5 the second most m/z 540. As this ion does not appear in the MS 2 spectrum of m/z 664, it is unlikely that m/z 540 is a source fragment. In contrast, ions corresponding to the most abundant NLs in the MS 2 spectrum of m/z 664 (e.g., m/z 408 and 346) are observed and are hypothesized to result from source fragmentation In addition to the reduced form of nicotinamide adenine dinuc leotide, the oxidized form of the same di nucleotide (NAD + ) was also characterized utilizing 9 AA as a MALDI matrix. An increase in the NADH/NAD + ratio has been shown to demonstrate a positive correlation with MI; 117 thus, distinguishing the two forms of this nucleotide may prove beneficial. A MALDI MS spectrum of NAD + collected in negative mode is displayed in Figure 3 7 Interestingly, th e [M H] ion of NAD + did not appear as the most intense non matrix ion. Instead, two ions at m/z 540 and 558, the former ion being more intense, represented the most intense analyte ions After analyzing the structure of NAD + it was concluded that the i on at m/z 540 likely resulted from the loss of nicotinamide. The ion at m/z 55 8 could result from contamination of ADP Ribose during synthesis of NAD + To confirm the identity of the proposed ion (i.e., m/z 540 ), MS 2 was performed (Figure 3 8) Following CID of m/z 540, abundant fragments were observed a t m/z 426
134 408, 346, and 328, likely corresponding to the [M H] of ADP, the [M H H 2 O] of ADP, the [M H] of AMP, and the [M H H 2 O] of AMP, respectively. Furthermore, the ion at m/z 273 resu lts from a NL of 267, the same as the nominal mass of adenosine. These five fragments present strong evidence for the presence of two phosphate groups bound in series to an adenosine nucleotide. Finally, the ion at m/z 480 (NL of 60) is a co mmon cross ri ng cleavage that occurs in the presence of sugars. Thus, taking into account the remaining mass, a ribose ring must also be bound to the two phosphate groups, substantiating the structure proposed in Figure 3 8 Comparison of pure 9 AA free base and 9 AA extract The supernatant of the extraction detailed above was tested for feasibility as a MALDI matrix. In doing so, the supernatant was either spotted on a MALDI target plate, or mixed with an LPC standard and spotted using the dried droplet method detail ed above. The resulting spots were then analyzed in both positive and negative mode by MALDI MS. A comparison of the pure MALDI matrix and the extract in positive mode is shown in Figure 3 9 A similar comparison for negative mode is shown in Figure 3 1 0 The spectra from the extract and commercial ly purchased free base are largely comparable. The sole striking difference appears in negative mode, wherein the ratio of the [M H] ( m/z 193) to the [M ] ( m/z 194) appears lower in the extract though the reason for this phenomenon is currently unclear The initial HCl salt, dissolved in 70:30 EtOH:H 2 O (v/v) was also tested, producing similar m/z values to that of the extract and pure free base. Further experiments were conducted analyzing the LPC standar d In both positive and negative ionization modes, MALDI MS analysis utilizing the extract as a matrix provided the expected [M+H] + or [M CH 3 ] ion for this standard with similar abundance to the commercially available freebase.
135 Positive mode MS imaging o f a rat coronary artery ligation model of myocardial infarction To evaluate the utility of 9 AA as a positive mode matrix for MALDI MSI, imaging experiments were conducted on transverse sections of cardiac tissue from the rat model for MI. For comparison, serial sections were coated with either 9 AA or DHB. The serial sections were then analyzed sequentially to eliminate any day to day variation. Detected analytes in positive mode Although 9 AA is primarily utilized for negative mode analyses a number of analytes can be ionized in positive ion mode with 9 AA as a matrix In particular, analytes containing a preformed charge (e.g., choline containing analytes) are still efficiently ionized. Figure 3 11 displays the averaged mass spectra from cardiac tiss ue coated with either 9 AA (Figu re 3 11A) or DHB (Figure 3 11 B). A number of common ions (e.g., m/z matrix All of the common ions were identified as either PCs, sphingomyelins (SMs), or acylcarnitines. It should be noted that each of these compound classes contains a choline functional group that retains a preformed positive charge on the quaternary amine. The relative abundances of these ions, however, do not appear conserved between 9 AA and DHB coated cardiac tissue. For example, the three ions at m/z 810, 832, and 848 are thought to correspond to the [M+H] + [M+Na] + and [M+K] + respectively, of PC (18 :0/20:4). Analogous to the analysis of the LPC standard, the formation of the [M+H] + from tissue appears favorable when 9 AA is used wher e as the formation of the [M+Na] + appears favorable when DHB is used As alkali adducts were not observed in the LPC standard analysis with 9 AA, the observed adduction is thought to result from the naturally occurring sodium and potassium in the tissue.
136 Although 9 AA effectively ionized compounds with a preformed positive charge, a number of analytes are absent from the 9 AA spectrum that were previously detected with DHB. For example, positive ions from nucleotides such as adenosine monophosphate and adenosine diphosphate occurring at m/z 348 and 428, respectively are noticeably absent from the 9 AA coated tissue. F urthermore, the positive mode peaks for cholesterol at m/z 369 and heme B at m/z 616 are also absent from the 9 AA spectrum Finally, a greater degree of source fragmentation is observed for positive ions with 9 AA than DHB. This phenomenon can be observe d via m/z 666 in Figure 3 11 A and m/z 725 in Figure 3 11 B. Presumably, the ion at m/z 725 corresponds to the [M+Na] + of SM (d18:1/16:0) and m/z 666 corresponds to the [M+Na C 3 H 9 N] + of the same SM. Alt hough many factors contribute to source fragmentation, laser fluence source pressure, extraction v oltage (restricted to MALDI ToF MS), and the choice of the MALDI matrix are among the most cited 118, 119 As the laser fluence and source pressure were maintained between analyses the absorptivity of the matrix at 337 nm was thought responsible. The UV V is absorption spectra for the DHB and 9 AA matrix solutions are shown in Figure 3 12 and Figure 3 13 respectively Unlike DHB, a relative minimum exists at 337 nm for 9 AA, suggesting that the weak absorption at this wavelength contributes to the source fragmentation. Principal c o mponent analysis of perfused and infarcted m yocardium The utility of 9 AA as a positive mode MALDI matrix for identifying tissue specific markers of MI was explored. Transverse cardiac sections, coated with either 9 AA or DHB, from a rat LAD ligation model of MI were analyzed in positive ionization mode, and an established multivariate data analysis methodology was applied to determine
137 markers for the infarcted (damaged) and perfused (healthy) myocardium. 48 In previous studies conducted in this lab, a number of LPCs and PCs were identified as markers for infarcted and perfused myocardiu m, respectively. 48 P hosph olipase A 2 (PLA 2 ), an enzyme that hydroly zes the sn 2 acyl bonds of PCs was thought responsible for these markers. Separation of infarcted and perfused myocardium was observed al on g the first principal component for both 9 AA and DHB coated tissue (Figure 3 14 ) The first principal component accounted for a larger percentage of the variance in the 9 AA tissue relative to the DHB tissue (95.8% vs 85.8% respectively ). This difference can be explained by the analytes detected with each matrix. For instanc e, PCs comprise th e vast majority of the positive mode ion current with 9 AA, whereas PCs comprise a smaller percentage of the ion current with DHB. As most intact PCs contain an unsaturated fatty acid tail in the sn 2 position of the glycerol backbone, a nd therefore can be acted upon by PLA 2 it is not surprising that the spectral variation between the two regions of tissue is greater from the tissue coated with 9 AA relative to DHB. A one dimensional loadings plot for the first principal component is displayed in Figure 3 15 for both 9 AA and DHB coated tissue. Based on a quick inspection, the two loadings plots demonstrate so me similar trends; the intact PC s ( m/z 750 900) load with the perfused myocardium and the LPCs ( m/z 450 600) load with the infarcted myocardium. Furthermore, SM (d18:1/16:0) at m/z 725 and its source fragment at m/z 666 load with the infarcted myocardium following analysis of both tissues The images for a number of these lipids and the TTC stained tissue indicating the different tissue regions are displayed in Figure 3 16 For all of these ions, 9 AA produced images of a
138 quality rivaling DHB. Despite these similarities, a number of small metabolites found to be significant with DHB do not load with either tissue region when 9 AA is utilized. As previously mentioned, the basic nature of the matrix is thought to prevent efficient ionization of these compounds. Dicarboxyla cylcarnitines as potential blood borne biomarkers of myocardial i nfarct ion Alt hough there are a number of metabolites not detected with 9 AA, the region between m/z 200 and 450 has an appreciably lower matrix background than DHB After a closer inspection of the lower mass region for 9 AA, a number of low molecular weight io ns (e.g., m/z 204, 222, 248, and 262) were found to load with the perfused tissue. A comparison of the images obtained for the four labeled compounds in the lower mass region of the 9 AA loadings plot is shown in Figure 3 17 For each of these ions the tissue coated with 9 AA provided better image contra st between the infarcted and perfused myocardium; two ions ( m/z 222 and 262) did not display any noticeable image contrast between these tissue regions within the DHB coated tissue. Each of these ions is hypothesized to leak from infarcted myocardium following rupture of the plasma membrane, as previously rep orted. 48 To identify these ions, MS 2 was performed. The lowest molecular weight ion m/z 204, demonstrated the highest relative abundance of the four ions. Prior to fragment ation studies, the ion was putatively identified to be the [M+H] + ion of acetylcarnitine (C2 carnitine) MS 2 of this ion (Figure 3 18 ) produced abundant fragment ions at m/z 60, 85, and 145. The second fragment ( m/z is characteristic of all proto nated carnitine ions resulting from successive loss of the fatty acyl chain ( in this case acetic acid ) and trimethylamine. The ion at m/z 145, resulting
139 from a neutral loss (NL) of 59, corresponds to loss of trimethylamine. Furthermore, the ion at m/z 60 corresponds to protonated trimethylamine. Finally, a low abundance fragment ion is observed at m/z 144 (NL 60), resulting from the loss of acetic acid. A s imilar fragmentation pattern was observed for m/z 248; however, structures with standard fatty a cyl chains could not be assigned C4 carnitine would exhibit an [M+H] + ion at m/z 232 and C6 carnitine would exhibit an [M+H] + ion at m/z 260. Instead, the mass to charge ratio of this ion suggests an acyl chain amounting to 104 Da Based on this mass, ma lonic acid, a dicarboxylic acid, was hypothesized to be esterified to carnitine. The MS 2 spectrum in Figure 3 19 supports this hypothesis. Alt hough the most abundant fragments (i.e., m/z 85 and 189) yield little information concerning the fatty acid tail fragment ions also appear at m/z 87 and 105, corresponding to the [M+H] + of dehydrated and intact malonic acid, respectively. In addition, the fragment ion at m/z 144 results from the loss of malonic acid. Similarly, the ion at m/z 262 was hypothesized to contain a fatty acyl chain with a mass of 118 Da. Two possibilities exist for this mass: succinic acid and methylmalonic acid, both of which are present endogenously. The MS 2 spectrum for this ion is displayed in Figure 3 20 To date, the exact fatty acy l chain has not been identified; however, for illustrative purposes, th e ion is drawn with methylmalonic acid esterified to carnitine Once again, fragment ions corresponding to the [M+H] + of the dehydrated and the intact fatty a cyl chain are observed. Although the exact fatty acid structure for m/z 262 has not yet been identified, both C3 and C4 dicarboxylacylcarnitines have been detected from the rat heart. 120
140 Negative mode MS imaging of a rat coronary artery ligation model for myocard ial infarction Following positive mode analysis negative mode MALDI MSI was performed on tissue sections from the rat LAD ligation model for MI. A comparison of the upper half of the heart, stained with TTC, and the MALDI MSI total ion current (TIC) is s hown in Figure 3 2 1 TTC exhibits a color change (from colorless to red) in the presence of mitochondria from healthy tissue. This color change stems from enzymatic reduction of the TTC by dehydrogenases to produce a red formazan. The histological staining in Figure 3 21 A demonstrates a fairly large area of infarction (areas of white tissue) near the left vent ricle. Based on the asymmetry of the left ventricle, representative mass spectra were averaged from the infarcted myocardium and the perfused (healthy) myocardium. The averaged mass spectra from these two regions are displayed in Figure 3 22 A number of differences between the two tissue regions can be ascertained by direct comparison of these two spectra. For instance, m/z 540 putatively i dentified as NAD + appears to show a relatively high intensity in perfused myocardium; however, the same ion does not appear above the background in the infarcted myocardium. In contrast, m/z 1207 demonstrates a higher intensity in infarcted myocardium th an perfused myocardium. The images of these two ions are illustrated in Figure 3 23 Principal component a nalysis To further analyze these two regions of tissue, the aforementioned multivariate data analysis methodology was applied to spectra extracted fr om the perfused and infarcted myocardium. Separation between the perfused and infarcted samples was observed along principal component 1, which accounted for 85.3% of the total variance
141 (Figure 3 24 ). Principal component 2, which accounted for 5.5% of th e total variance, dictated the variation within a tissue region. A 1 dimensional loadings plot for principal component 1 was generated to determine ions positively correlating with either of the two regions (Figure 3 25 ). Ions that load positively on principal component 1 demonstrate a positive correlation with perfused myocardium. Conversely, ions that load negatively on principal component 1 demonstrate a positive correlation with infarcted myocardium. Images for a number of select ions are shown in F igure 3 26 Analysis of the NADH standard suggests that m/z 664 which localizes in the perfused myocardium, is the [M H] of NADH; however, MS 2 analysis is necessary to confirm this identification as is discussed later in the results Once markers for the perfused and infarcted myocardium were selected, spectra were also extracted from regions suspected of being at risk myocardium. These samples were then included in PCA analysis to determine endogenous markers of this tissue region. The scores pl ot generated from the inclusion of all three tissue regions (infarcted, perfused, and at risk myocardium) is depicted in Figure 3 27 Separation is observed from all three tissue regions, with principal component 1 (66.6% of the variance) dictating separa tion between infarcted myocardium and the remaining two tissue regions, and principal component 2 (19.8% of the variance) dictating separation between perfused and at risk myocardium. The loadings plot from principal component 1 (Figure 3 28 ) appears large ly similar whether the at risk myocard ium samples are included or not, the major difference being that m/z 1469 loads with the infarcted myocardium when the at risk myocardium is excluded (Figure 3 25 ) and conversely loads with the perfused/at risk myocard ium
142 when the at risk myocardium is included. Principal component 2, however, demonstrates a number of features that demonstrate high correlation with the at risk myocardium. These features, loaded on the negative portion of principal component 2 (Figure 3 29) include a number of low mass ions (e.g., m/z 244, 403, 565, and 606) in addition to m/z 1469. The MS images for a few of these ions is displayed in Figure 3 30 MS n identification and i maging Ions that demonstrated both a significant loadings coeff icient in PCA analysis and localization in the region of interest were subjected to MS n for identification. Initially, targeted experiments within the region of interest were performed. Following identification of the ion, MS n imaging experiments were pe rformed, and the characteristic fragment ion of the identified compound was mapped over the entire tissue for localization confirmation. In general, identified compounds were found to belong to one of two compound classes: 1) phospholipids or 2) nucleotide s / nucleotide sugars Both of these classes will be discussed below. Phospholipids As previously discussed in this dissertation, positive mode analysis of a ligation model demonstrated that phospholipase A 2 (PLA 2 ) cleaved the sn 2 fatty acid tail of intact phospholipids t o generate lysophospholipids. Alt hough PLA 2 is thought to have a greater activity towards intact PCs and PEs, other lipid classes that pr e fe re ntially ionize in negative mode (e.g., phosphatidyl inositols and cardiolipins) are susceptible to this enzyme as well. Analysis of the loadings plot in Figure 3 28 yields three ions at m/z 1447, 1469, and 1485. The 22 amu difference between m/z 1447 and 1469 suggests that the two
143 ions differ by replacement of a hydrogen with a sodium at om Similarly, the 38 amu difference between m/z 1447 and 1485 suggest that the two ions differ by replacement of a hydrogen with a potassium atom MS 2 analysis identified these ions as the [M H] the [M+Na 2H ] and the [M+K 2H ] of tetralinoleoyl cardiolipin (TLCL). The MS 2 spectrum for m/z 1447 is shown in Figure 3 31A The ion at m/z 695 (NL 752) corresponds to phosphatidic acid (18:2/18:2), and the ion at m/z 831 (NL 616) results from the loss of diacylglycerol (18:2/18:2). Furthermore, a low abundan ce ion at m/z 1167 (NL 262) is observed, corresp onding to the loss of dehydrated linoleic acid. The sodium adduct at m/z 1469 demonstrates similar, though not identical fragmentation to the protonated ion (Figure 3 31B ). In this instance, the mos t abundant fragment corresponds to the loss of the diacylglycerol ( m/z 853) instead of the ion corresponding phosphatidic acid. Furthermore, three low abundant ions at m/z 1207, 1189, and 1167, corresponding to the losses of dehydrated linoleic acid, lin oleic acid, and sodiated linoleic acid, respectively are observed Similarly, the potassium adduct of TLCL also demonstrates an abundant loss of the diacylglycerol ( m/z 869), and three lower abundant ions at m/z 1223, 1205, and 1167 corresponding to the loss of dehydrated linoleic acid, linoleic acid, and potassiated linoleic acid (Figure 3 31C) The apparent differences in the fragmentation pathways for these three ions (mainly the absence of m/z 695 and 751) a re not unexpected, as protonated phospholipid ions and alkyl adducts of phospholipids have been shown to generate vastly different fragment ions under low energy CID conditions. 121 Analogous to PC ions, fragmentation of the carbon oxygen bond in the bridging glycerol backbone appears to be inhibited due to the stabilization of either phos phate group by the metal cation. 122 For further
144 clari fication, the fragmentation pattern for all three of these ions is displayed in Figure 3 32 As previously noted in Chapter 2 sodium and potassium adducts of phospholipids demonstrate different localization following MI. This phenomenon also occurs with the intact cardiolipins; the protonated ion and potassium adduct localize in the perfu sed myocardium, and the sodium adduct localizes in both the infarcted and at risk myocardium. As neithe r sodium nor potassium was supplemented in the matrix solution or the tissue, these localizations are thought to reflect the natural sodium and potassium content within myocardium. To analyze the overall intact cardiolipin content within the tissue, a summed image of the three cardiolipin ions was generated (Figure 3 33 ). As can be seen in this image, the overall cardiolipin content appears to decrease despite the apparent increase in sodium content in infarcted myocardium. Similar to intact PCs and PEs, PLA 2 is t hought to be responsible for this decreased signal of in tact cardiolipins in infarcted myocardium. Further verification of the enzymatic action of PLA 2 on intact cardiolipins was found with the presence of both mono and dilysocardiolipins. Unlike most phospholipids, cardiolipins contain two glycerol backbones and four fatty acid tails ; thus, there are two sn 2 fatty acid tails available for hydrolysis. To this point, t hree ions were identified from the loadings plots related to mono and dilysocardiolipins, all of which positively correlated with the infarcted myocardium. Two of these ions, m/z 1185 and 1207, were identified as the [M H] and [M+Na 2H] respectively, of monolysocardiolipin. Interestingly, the [M+K 2H] did not demonstrate a significant loadings coefficient with either infarcted or perfused myocardium. The third ion at m/z
145 945 was identified as the [M+Na 2H ] of dilysocardiolipin, and also demonstrate d a positive correlation with infarcted myocardium. Targeted and imaging tandem MS experiments were conducted to verify the identity and localization of the aforementioned lysocardiolipins. The MS 2 spectrum of dilysocard i olipin at m/z 945 is displayed in Figure 3 34 Generally, the lysocardiolipins demonstrated analogous fragmentation to the intact cardiolipins, producing abundant fragment ions resulting from cleavage of P O bonds exemplified by the ions at m/z 591, 507 and 415. Additionally, fragment i ons corresponding to either the NL ( m/z 665) of or carboxylate anion ( m/z 279) of linoleic acid were also observed, though less abundant than cleavages about the phosphate groups. MS 2 images mapping the two most abundant transitions for dilysocardiolipin are also shown in Figure 3 34 Both fragment ions demonstrate localization within the infarcted myocardium. Alt hough the [M H] of dilysocardiolipin did not appear above the baseline in the principal component 1 loadings plot targeted MS 2 imaging experi m ents were conducted on m/z 923 to decouple the effect of salt adduction and localization. That is, if PLA 2 is truly acting upon intact cardiolipins, the [M H] ion of dilysocardiolipin should also localize within the infarcted myocardium. To test this hypothesis, an MS 2 imaging experiment was conducted on m/z 923. The most abundant fragment occurred at m/z 433, resulting from cleavage at the sn 1 position of the bridging glycerol backbone (Figure 3 35B The MS 2 image for this transition (Figure 3 35A ) demonstrates a distinct localization with the infarcted myocardium despite the relatively low signal generated. This localization lends credence to the idea that PLA 2 is acting upon intact cardiolipins to form lysocardiolipins.
146 Alt hough PLA 2 has also be en reported to act upon phosphatidylinositols (PIs) the most abundant intact PI, PI (18:0/20:4) at m/z 885, did not demonstrate localization in the perfused myocardium. In fact, m/z 885 demonstrated a relatively large loadings coefficient with the intact my ocardium. Interestingly, the L PI that results from hydrolysis by PLA 2 ( m/z 599) also positively correlated with the infarcted myocardium. It is not currently clear as to why the intact PI demonstrates higher signal in infarcted myocardium; however, th e generation of the LPI in infarcted myocardium suggests that PLA 2 is enzymatically cleaving fatty acids from intact PIs. MS 2 experiments fragmenting m/z 599 were conducted to confirm this putative identification. The MS 2 spectrum of this ion, displayed in Figure 3 36 demonstrates four abundant fragment ions at m/z 419, 315, 283, and 241. The fragment ion at m/z 419 arises from an NL of 180, presumably due to the loss of inositol, indicative of the PI headgroup. Further more, fragment ions are observed at m/z 283 and 315, corresponding to cleavage at the sn 1 position of the glycerol backbone; the fragment ion at m/z 283 is the carboxylate ion for stearic acid and the ion at m/z 315 corresponds to stearic acid as an NL, c onfirming this ion as the [M H] of LPI 18:0. The most abundant fragment of this precursor ion (i.e., the carboxylate anion at m/z 283) demonstrates localization in the infarcted myocardium, as was indicated by the loadings plot from principal component 1 Referring back to Figure 3 25 and Figure 3 28 two low abundant masses in the lysoPL region, m/z 480 and 508, consistently load with the infarcted myocardium. These masses are consistent with 3 different lysoPLs, LPC 16:0, LPC 18:0, and LPE 18:0, that w ere previously identified in positive mode as being generated through enzymatic action of PLA 2 Due to the [M CH 3 ] ion formed in negative mode for PCs,
147 MS analysis is insufficient to determine whether m/z 480 is LPC 16:0 or LPE 18:0. Accordingly, targeted and imaging MS 2 experiments were conducted within the infarcted myocardium. The MS 2 spectrum for m/z 480 is displayed in Figure 3 37 Not surprisingly, fragment ions characteristic of both isobaric species are observed; the carboxylate anions for LPC 16:0 and LPE 18:0 are observed at m/z 255 and 283, respectively. The MS 2 images for these two fragment ions demonstrate localization for the characteristic LPC 16:0 fragment ( m/z 255) within infarcted myocardi um. In contrast, the LPE 18:0 fragment ( m/z 283 ) only displays a slight increase in abundance within the infarcted myocardium. MS 2 analysis of m/z 508 (Figure 3 38 ) produced a n abundant fragment ion at m/z 283 and a less abundant fragment at m/z 224, co nfirming the ion as the [M CH 3 ] of LPC 18:0. Similar to LPC 16:0, t he MS 2 image of m/z confirmed that LPC 18:0 was localized within the infarcted myocardium. Nucleotides and nucleotide s ugars A num ber of m/z values identified by PCA to have relevant localization within the regions interest were identified by MS 2 as nucleotides. In nearly all instances, the nucleotides were found to contain at least one phosphate group (i.e., nucleoside mono di or triphosphates) that likely aided in the negative mode ionization. The most abundant detected nucleotides and nucleotide sugars belonged to the family of adenosine phosphates. The nucleotide phosphates (without an additional bound sugar residue) were observed in negati ve mode in the form of [M H] ions at m/z 346, 426, and 506 for adenosine monophosphate (AMP), adenosine diphosphate (ADP), and adenosine triphosphate (ATP), respectively. The MS 2 spectrum for ADP at m/z 426 is displayed in Figure 3 39 For both ATP and ADP, NLs of 98 at 408 and 328, respectively, were observed suggesting the loss of phosphoric acid in both instances.
148 Furthermore, in the case of ADP, a fragment ion was observed at m/z 134, suggesting the presence of deprotonated adenine. AMP did not re adily lose phosphoric acid, but instead exhibited a NL of 135, corresponding to the loss of adenine to form the product ion at m/z 211. The fragmentation for these three ions indicates that the charge is localized on the phosphate group nearest to adenosine. MS imaging and PCA of all three of these ions demonstrated a marked decrease in signal in the infarcted myocardium, suggesting that these nucleotides are leaching into the bloodstr eam following myocardial ischemia. Also within the adenosine family, NADH and NAD + were putatively identified based nominal m/z Furthermore, both of these ions m/z 66 4 and 540 for NADH and NAD + respectively, exhibited large loading s coefficients indicating localization in the perfused myocardium. MS 2 analysis of the precursor ion at m/z 540, yielded four abundant fragment ions at m/z 426, 408, 346, and 328, confirming this ion as the same species present in the NAD + standard. Sim ilarly, analysis of m/z 664 produced to abundant fragment ions at m/z 397 and 408, confirming this ion as the [M H] of NADH. In addition to adenosine containing nucleotides several uridine containing nucleotide and nucleotide sugars almost exclusively i n the form of diphosphates, were detected with significant abundance. The [M H] of uridine diphosphate (UDP) was identified at m/z 403 (Figure 3 40) MS 2 of this ion yielded an ion resulting from a NL of 98 at m/z 305. MS 3 was conducted on the m/z ion is related to UDP. This experiment yielded a single abundant fragment at m/z 111 (NL 194), corresponding to the [M H] of uracil, further validating the assignment of m/z
149 403 as the [M H] of UDP. Additionally, a fragment ion was observed at m/z 267 (NL of 136) that is presumably unrelated to UDP. An MS 2 imaging experiment validated this hypothesis, as the two fragments demonstrate different localizations; the fragment ion occurring at m/z 267 demonstrates local ization lar gely in the perfused myocardium, whereas the fragment ion occurring at m/z 305 demonstrated progressively decreasing signal in the at risk, perfused, and infarcted myocardium. The increased abundance of UDP in the at risk myocardium was substan tiated by the identification of two UDP sugars with similar localization. These two sugars, detected at m/z 565 and 606, were identified as UDP glucose (UDP Glc) and UDP N acetylglucosamine (UDP GlcNAc), respectively. The MS 2 spectra of UDP Glc and UDP GlcNAc are displayed in Figure 3 41 and Figure 3 42 respectively. CID of both ions yields a fragment ion at m/z 323 resulting from cleavage between the two phosphate groups. Interestingly, UDP GlcNAc demonstrates much more ex tensive fragmentation than UDP Glc. In addition to the fragment ion at m/z 323, UDP GlcNAc yields abundant ions at m/z 403, 385, 362, and 282 with the fragment ion at m/z 403 likely resulting from the loss of dehydrated N acetylglucosamine Conclusion s T his work has demonstrated the utility of 9 AA as a dual mode matrix for MALDI MSI. 9 AA exhibited more efficient ionization in negative mode relative to DHB. Furthermore, a number of analytes were still detected in positive mode with a signal to noise ra tio rivaling that of DHB. Perhaps one of the more understated advantages of 9 AA is that the observed crystal size produced is orders of magnitude smaller than DHB despite the use of a less volatile solvent system (70:30 EtOH:H 2 O vs. 70:30 MeOH:H 2 O
150 for 9 AA and DHB, respectively). This crystal siz e may prove beneficial as MALDI MSI experiments are continually moving towards higher spatial resolution. Alt hough 9 AA has proved a valuable matrix for both n egative and positive mode MALDI MS analyses, the price of the pure free base may detract users from utilizing 9 AA for tissue analysis. This work has demonstrated that a relatively simple and fast acid/base extraction can be used to obtain the free base from the relatively inexpensive h ydrochloride hydrate salt of 9 AA. It should be noted that this extraction technique has not yet been optimized; however, any reasonable extraction yields will drive the cost per analysis to rival standard MALDI matrices. Furthermore, it appears that the hydrochloride salt, prepared in the same solvent system as described in this work, may also provide a suitable MALDI matrix for negative mode analyses. This alternative is the optimal solution, as there is no requirement for a liquid/liquid extraction an d the price per analysis is approximately 50x lower than using the commercially available freebase. The one caveat identified to this point is the artificial addition of chloride ions to the sample that may ultimately confound analysis of endogenous chlor ide containing analytes. Finally, the analysis of the rat coronary artery ligation model with 9 AA has been extremely valuable in regards to advancing the body of knowledge surrounding lipid chemistry in MI Firstly, the enzyme PLA 2 was found to act up on both intact cardiolipins, intact PIs, and intact PCs in tissue, yielding lyso PLs in areas of myocardial ischemia. this is the first instance in which PLA 2 has been shown to act upon cardiolipins and intact PIs in situ following MI. Cardiolipins provided a particularly elegant substrate for PLA 2 as two sn 2 fatty acyl substituents were available
151 for hydrolysis, evidenced by the formation of dilysocardiolipin in infarcted myocardium Fi nally, short chain dicarboxyl acylcarnitines detected in positive mode, were also identified as demonstrating a decreased signal in infarcted myocardium. These same analytes were previously overlooked in Chapter 2 using a similar multivariate data analysis methodology with DHB as a MALDI matrix. The relatively low matrix background of 9 AA was instrumental in detecting t he changes in these low abundance, low m/z analytes. Interestingly, increased blood borne concentrations of dicarboxylacylcarnitines were just recently identified as predictors fo r future MI events. 123 Thus, we may have only brushed the surface of the role that these low abundant analytes play in MI. Additionally, a number of nucleotides and nucleotide sugar were found to show dramatically decreased signal in infarcted myocardium. As with the adenosine containing nucleotides detected in positive mode (Chapter 2) these metabolites are thought to leak from tissue following r upture of the plasma membrane. An unexpected finding within this compound class however, was that uridine containing nucleotides (e.g., UDP) were found to localize within the at risk myocardium. Interestingly, UDP and UTP are known as a positive inotropes (i.e., stimulating muscle contraction). 124 Furthermore, UTP, which has shown to be released into the blood stream following myocardial ischemia, also stimulates vasodilation. 125 Based on this study, it would appear that uridine phosphates are upre gulated under ischemic conditions. Following necrotic cell death, increased levels of these nucleotides may be released into the blood stream, triggering the cardio protective aspects of UTP and UDP. 124, 126
152 The biological imp lication of these findings is three fold. First, this methodology offers an alternative for post mortem identification of MI via a number of tissue specific markers of tissue death (necrosis) Second, and more importantly, this methodology has identified a host of s mall molecule markers of MI, the most valuable of which are water soluble metabolites that potentially diffuse into the bloodstream. Instead of the typical immunoassays that analyze protein biomarkers that decay within days (or even hours) of the onset of MI, a targeted metabolomics approach utilizing the identified small molecule biomarkers may be useful to determine if a living patient has been afflicted with acute MI. As the accuracy of prediction scales with the number of markers, a large number of lo w molecular weight markers may prove more accurate than current diagnostic tests using protein biomarkers Finally, this methodology has provided support for the hypothesis that pyrimidine nucleotides are cardioprotective.
153 Figure 3 1. Potential negat ive mode MALDI matrices for MSI of cardiac tissue.
154 Figure 3 2 MS 2 spectrum of m/z 494 from an LPC 17:0 standard collected in negative mode using 9 AA as a MALDI matrix. The structure and fragmentation pattern of the ion are also shown.
155 Figure 3 3. Negative mode MALDI MS 2 spectrum of m/z 714 from a synthetic PE fragments are also indicated.
156 Figure 3 4 Example of MS n in negative mode for a PS ion. A) MS 2 spectrum o f m/z 788 and B) MS 3 spectrum of m/z 788 701 from an l phosphatidylserine extract from porcine brain. The structure of the identified ion, PS (18:0/18:1), and the fragmentation pattern is also shown.
157 Figure 3 5 MALDI MS spectrum of NADH in negative mode using 9 AA as a MALDI matrix. Source fragments resulting from the deprotonated molecular ion of NADH ( m/z 664) are denoted with stars. Figure 3 6 Negative mode MALDI MS 2 spectrum of m/z 664 obtained from an NADH standa rd. The fragmenta tion pattern of NADH is also shown.
158 Figure 3 7 Negative mode MALDI MS spectrum of NAD+ standard using 9 AA as a MALDI matrix. Figure 3 8 Negative mode MALDI MS 2 spectrum of m/z 540 obtained from an NAD standard utilizing 9 AA as a MALDI matrix. The proposed structure and fragmentation pattern for the ion at m/z 540 is also shown.
159 Figure 3 9 MALDI MS spectra collected in positive ionization mode from the 9 AA extract (Top) and the commercially available freebase (Bottom).
160 Figure 3 10 MALDI MS spectra collected in negative ionization mode from the 9 AA extract (Top) and the commercially available freebase (Bottom).
161 Figure 3 11 Averaged MS spectra from cardiac tissue following L AD ligation Tissue was coated with either A) 9 AA or B) DHB.
162 Figure 3 12 UV V is absorption spectrum of 2,5 dihydroxybenzoic acid dissolved in 70:30 MeOH:H 2 O (v/v). Figure 3 13 UV V is absorption spectrum of 9 aminoacridine dis solved in 70:30 EtOH:H 2 O (v/v).
163 Figure 3 14 PCA scores plots separating infarcted myocardium (red triangles) and perfused myocardium (green crosses) T issue was coated with either A) 9 AA or B) DHB. The circles represent the 95% confidence in tervals of the sample groupings.
164 Figure 3 15 PCA loadings plots from analysis of infarcted and perfused myoc ardium T issue was coated with either A) 9 AA or B) DHB. The positive portion of the first principal component (PC 1) corresponds to ions demonstrating localization in the perfused myocardium and the negative portion of PC 1 corresponds to ions demonstrating localization in infarcted myocardium.
165 Figure 3 16 MS images from infarcted cardiac tissue Tissue was coated with either A) 9 AA or B) DHB Furthermore, C displays the upper half of the heart stained with TTC to visualize the different regions of tissue.
166 Figure 3 17 MS images generated for low molecular weight ions from lig ated hearts Tissue was coated with either A) 9 AA or B) DHB.
167 Figure 3 18 MS 2 spectrum of m/z 204 obtained from perfused cardiac tissue coated with 9 AA as a MALDI matrix. The structure and fragmentation of the identified ion, acetylcarnitine, is also shown.
168 Figure 3 19 MS 2 spectrum of m/z 248 obtained from perfused cardiac tissue coated with 9 AA as a MALDI matrix. The structure and fragmentation of the identified ion, m alon ylcarnitine, is also shown.
169 Figure 3 20 MS 2 spectrum of m/z 262 obtained from perfused cardiac tissue coated with 9 AA as a MALDI matrix. The structure of methylmalonylcarnitine, one of the possible isomers at m/z 262, is also shown.
170 Figure 3 21. Images of a heart following coronary artery ligation. A) Uppe r half of the heart stained with TTC and B) total ion current obtained from an MS imaging experiment.
171 Figure 3 22. Negative mode M ALDI MS spectra The spectra were collected from the A) perfused myocardium and B) infarcted myocardium. The red boxes indicate two ions that are present in higher abundance in either tissue region.
172 Figure 3 23. MS images of m/z 540 and 1207. The images demonstrate complementary localizations in cardiac tissue following LAD ligation.
173 Figure 3 24 PCA scores plot from negative mode MALDI MS analysis of infarcted (red triangles) and perfused myocardium (green crosses). The circles represent the 95% confidence interval of the sample groupings.
174 Figure 3 25 1 dimensional loadings plot from principal component 1 (PC 1). Ions loaded positively on PC 1 demonstrate a positive correlation with perfused myocardium, and ions loaded negatively on PC 1 demonstrate a positive correlation with infarcted myocardium.
175 Figure 3 26 Negative mode MS images of various ions Images consist of ions loading either negatively (Top) or positively (Bottom) on principal component 1, dictating separation between the perfused and infarcted myocardium.
176 Figure 3 27 PCA scores plot fr om negative mode MALDI MS analysis of infarcted (red triangles) and perfused myocardium (green crosses). The circles represent the 95% confidence interval of the sample groupings.
177 Figure 3 28 1 dimensional PCA loadings plot of principal component 1 ( PC 1) generated from analysis of infarcted, at risk, and perfused myocardium. Ions loaded positively on PC 1 demonstrate a positive correlation with infarcted myocardium.
178 Figure 3 29 1 dimensional PCA loadings plot of principal component 2 (PC 2) ge nerated from analysis of infarcted, at risk, and perfused myocardium. Ions loaded positively on PC 2 demonstrate a positive correlation with perfused myocardium, and ions loaded positively on PC 2 demonstrate a positive correlation with at risk myocardium
179 Figure 3 30 MS images of various ions loading either on principal component 2, which demonstrates positive correlation with the at risk myocardium.
180 Figure 3 31 Negative mode MALDI MS 2 spectra of various high mass ions. The three ions, A) m/z 1447, B) m/z 1469, and C) m/z 1485 were collected from myocardium coated with 9 AA following LAD ligati on.
181 Figure 3 32 Fragmentation pattern for TLCL. The pertinent NLs for the [M H] [M+Na 2H] and [M+K 2H] are displayed.
182 Figure 3 33. Summed MS image of m/z 1447, 1469, and 1485 normalized to the total ion current.
183 Figure 3 34 MS 2 spectrum of m/z 945 identified as the [M+Na 2H] of dilysocardiolipin. The proposed fragment ation pattern and MS 2 images of the two most abundant fragment ions are also displayed.
184 Figure 3 35 MS 2 analysis of dilysocardiolipin. A) MS 2 image of m/z 923 433, the most abundant transition for the [M H] ion of dilysocardiolipin and B) the structure of this ion demonstrating the pertinent cleavage.
185 Figure 3 36 MALDI MS 2 spectrum of m/z 599 identified as the [M H] ion of LPI 18:0. In addition to the MS 2 image of the most abundant fragment ion ( m/z proposed structure and fragmentation of the ion are also shown.
186 Figure 3 37 Negative mode MALDI MS 2 spectrum of m/z 480 The MS 2 images of the two most abundant fragments are shown as an inset.
187 Figure 3 38 Negative mode MALDI MS 2 spectrum of m/z 508, identified as the [M CH 3 ] of LPC 18:0. The MS 2 image of the most abundant fragment is also displayed.
188 Figure 3 39. Negative mode MALDI MS 2 spectrum of m/z 426 The proposed fragmentation pattern of the identified ion, ADP and the MS 2 image o f the most intense fragment is also shown
189 Figure 3 40 Negative mode MALDI MS n identification and imaging of UDP from myocardium A) MS 2 spectrum of m/z 403 images of two intense fragments, and the structur e of the identified ion, UDP. B) MS 3 spectrum of m/z
190 Figure 3 41 Negative mode MALDI MS 2 spectrum of m/z 565. The proposed fragmentation pattern of the identified ion, UDP Glc, is also displayed.
191 Figure 3 42 MALDI MS 2 spectrum of m/z 606 detected from myocardium. The p roposed fragmentation pattern of the identified ion, UDP GlcNAc, is also displayed.
192 CHAPTER 4 MALDI MS METABOLIC PROFILING OF CAENORHABDITIS ELEGANS Introduction Matrix assisted laser desorption/ionization mass spectrometric imaging (MALDI MSI) is an ideal method to reveal the distribution of biomolecules in tissue. 1, 48 Unlike other mass spectrometry based surface analysis techniques such as secondary ion mass spectrometry (SIMS) or la ser ablation inductively coupled plasma mass spectrometry (LA ICP MS), MALDI MS provides a method for soft ionization of biomolecules (e.g., lipids, peptides, and proteins) 3 Traditional MALDI MSI methodologies have investigated thin, microtomed tissue sections; however, MAL DI MSI has the potential for small organism surface analysis without prior sectioning, 127 assuming the organism satisfies the size constraints of typical instruments. Caenorhabditis elegans ( C. elegans ) is a fre e living nematode found in soil and compost, and is one of the most widely studied organisms in biological and biomedical sciences. Since the initial methodology describing the mapping of genetic mutants by Sydney Brenner over 40 years ago, 128 numerous studies have established C. elegans as an important model organism. C. elegans is a self fertilizing nematode with 959 somatic cells. 129 Its small size (~1 mm in length), sho rt life cycle (3.5 days), and simple body plan all contribute to the ease of laboratory manipulations, which now have enabled a rich area of research into C. elegans chemical biolo gy. These studies depending on the analyte concentration and analytical methodology, require on the order of 10,000 4,000,000 developmentally synchronized worms grown in liquid culture. 130 For example, in one recently reported investigation, exudates (i.e the exometabolome) from approximately 4,000,000 wild type and mutant nematodes were
193 collected with subsequent charac terization by NMR and LC MS. 130 In addition to the exometabolome, a number of investigations have been conducted using wild type C. elegans and re adily available gene knockouts, characterizing the fatty acid content from fractionated nematode extracts using techniques such as GC FID or GC MS. 131 135 Although a number of significant alterations in fatty acid catabolism were reported, other small molecules that are inefficiently extracted from the nematode surface may not b e observed using such methods. Thus, a method that efficiently probes the metabolome of individual C. elegans genotypes in s itu biology. In this report an a lternative approach using MALDI MSI for C. elegans chemical biology studies that uses small numbers of individual animals rather than pooled samples is presented There are ma ny advantages to this approach, including minimal sample preparation, a minimal number of specimens needed, and the ability to eventually apply this method to parasitic species that cannot be cultured. To demonstrate the potential for this approach, multi variate analysis was utilized to compare MALDI MSI data from several wild type and mutant strains Experimental Chemicals and r eagents 2,5 dihydroxybenzoic acid (DHB) was purchased from Acros Organics (Geel, Belgium). 9 aminoacridine (9 AA) was purchased from MP Biomedicals (Solon, OH). HPLC grade methanol (MeOH), HPLC grade water (H 2 O), and sodium acetate (NaOAc) were purchased from Fisher Scientific ( Fair Lawn, NJ ). 100% ethanol (EtOH) was purchased from Decon Labs (King of Prussia, PA). DHB was disso lved in 70:30 MeOH:H 2 O (v/v) to a final concentration of 40 mg/mL with a final concentration of 10
194 mM NaOAc added to serve as a MALDI matrix. Similarly, 9 AA was dissolved in 70:30 EtOH:H 2 O (v/v) to a final concentration of 6 mg/mL; however, NaOAc was not added to the 9 AA matrix solution. MALDI MS of n ematodes Synchronized young adult C. elegans strains (wild type N2, mutant daf 22 or mutant fat 6;fat 7 ) were washed three times in distilled H 2 O and then pipetted onto a microscope slide and allowed to dry. C. elegans were coated with the prepared MALDI matrix (either DHB or 9 AA) by pneumatic spraying with a Type A glass Meinhard nebulizer (Golden, CO). Matrix solution was delivered using a fl ow rate of 3.0 mL/min and nitrogen was used as a nebulization gas at 30 psi. Three passes were conducted over the microscope slide at a height of approximately 10 cm before allowing a drying time of approximately 15 seconds. The process was repeated unti l 8 mL (320 mg or 48 mg for DHB or 9AA, respectively) of MALDI matrix solution was deposited atop the microscope slide. All MALDI MS experiments were conducted utilizing a Thermo Scientific LTQ XL linear ion trap mass spectrometer (San Jose, CA) equipped w ith a Thermo MALDI ionization source. The MALDI ionization source consisted of a Lasertechnik Berlin MNL 106 LD N 2 laser (Berlin, Germany). The 337 nm laser was operated at a repetition rate in diameter. The source region of the instrument was maintained at a pressure of 75 mTorr. A laser Automatic gain control (AGC) was toggled off during analysis in o rder to maintain a fixed number of laser shots at each position along the nematode.
195 Data processing and analysis MALDI MS spectra were collected over the length of each nematode. In doing so, ten random positions were sampled from head to tail to generate one sample MS spectrum Within each sample grouping (N2, daf 22 and fat6;fat7 ) approximately 10 nematodes sampled, and each individual nematode was analyzed in triplicate. Furthermore, mass spectra from areas off the nematodes were collected in tripli cate during each experiment to determine background ions resulting from DHB or 9 AA. Mass to charge ( m/z ) and intensity lists were exported from Qualbrowser and saved as .CSV files Major matrix ions as determined by experiments collected off the nemato de surface, were then removed The remaining mass features were then normalized to the total ion current (TIC), and each mass feature was mean centered (i.e., row wise normalization). T he processed data were then imported into Metaboanalyst for multivari ate analysis utilizing PCA as previously reported. 48 2 dimensional scores plots were generated to visualize the separation of the sample grouping s Loadings which are the mass intensities that underlie the relationships between the mass spectra given by the scores were then analyzed to determine important mass features MALDI MSI and tandem MS Following multivariate data analysis, MS imaging was conducted over select ed nematodes to determine the localization of mass features with high loadings coefficients on the princi pal component of interest In doing so an oversampling raster step size of 25 m wa s utilized. Assuming the mass feature was localized on the nematode cuticle compound identification for ions with high loading s values was conducted using tandem MS with collision induced dissociation (CID). In these experiments, MS 2 imaging experiments were conducted over the entire nematode. Similar to MS imaging, MS 2
196 spectra were collected on the cuticle using a raster step size of 25 m (oversampling). In all MS an d tandem MS experiments, a laser energy of 4.5 J and 3 laser shots per spot were utilized. Additionally, the collision energy was maintained at 35 (normalized collision energy %) Following data collection, MS 2 images were generated for intense fragment ions. Only those ions that were found to localize on the nematodes were used for analyte identification. Results In this study, MALDI MS was utilized to characterize the surface metabolic profiles of individual whole C. elegans To validate the approach, three different genotypes were compared : wild type (N2) and mutant strains daf 22 and fat6;fat7. The daf 22 mutant lacks a gene that encodes an SCPx homolog, resulting in disruption of oxidation. 136 Thus, daf 22 mutants may accumulate long chain fatty acid lipids on the cuticle Furthermore, daf 22 mutants have been shown t o be deficient in ascarosides, signaling pheromone s that have also been implicated in d auer development. 130, 136, 137 The fat 6;fat 7 mutants lack two and thus demonstrate a decreased overall fatty acid content due to increased fatty acid catabolism. 131, 138 Mutations in fat 6 and fat 7 result in a deficiency of 20 carbon polyunsaturated fatty acids as well as oleic acid (18:1) and its derivatives. 131 MALDI MS p rofiling of nematode cuticles MALDI MS was utilized to characterize the compounds naturally occurring on both wild type N2 and mutant cuticles. Care was taken to obtain average mass spectra along the entire length of the nematode to eliminate spectral var iation that may occur due to the area of the cuticle sampled. On average, each nematode was approximately 1 mm in length; thus, 10 mass spectra were required to sample the entire length of the
197 nematode assuming a laser spot diameter of 100 m. DHB was in itially tested as a MALDI matrix for profiling for all three genotypes Representative MS spectra from the wild type nematode (N2) and the gene knockout nematodes ( daf 22 and fat6;fat7 ) are shown in Figure 4 1. Analysis of the nematode cuticles coated wit h DHB demonstrated a number of MALDI matrix related ions (e.g., m/z 154, 231, 273, and 313), the most intense of which are labeled with an asterisk Although matrix related ions are generally uninformative, the wild type N2 exhibited a relatively high abundance of m/z 231, putatively identified as the [M+2K H] + of DHB when compared to both the fat6;fat7 and daf 22 mutants As potassium salt was not added to the matrix solution, this matrix ion may reflect the potassium content of the nematode. Addit ionally, a number of suspected phosphatidylcholine (PC) lipid ions were observed on the nematode namely m/z 184, 542, and 580, putatively identified as the [M+H] + of phosphocholine the [M+H] + of lysophosphatidylcholine (LPC) 20:5, and the [M+K] + of LPC 2 0:5. In general, the daf 22 demonstrated a relatively high abundance for both the protonated ion and the potassium adduct of LPC 20:5. Furthermore, the fat6;fat7 nematodes exhibited much lower signal for all LPCs identified in the wild type N2 and daf 22 gene knockout. As the previous chapter demonstrated, 9 AA can efficiently ionize PCs in positive mode with comparable sensitivity to DHB. Accordingly, 9 AA was applied to the fat6;fat7 and N2 nematode cuticles to interrogate the PC content. As compared to DHB, the use of 9 AA (Figure 4 2 ) as a MALDI matrix produced strikingly different mass spectra from the nematode cuticle. In particular, 9 AA demonstrated a significantly lower matrix background in positive mode, with only one appreciable matrix ion at m/z
198 195. The basicity of 9 AA (pK b of approximately 4 ) is thought to be responsible for this low background; however, the basicity also limits the number of compounds that can be efficiently ionized in positive mode; analytes with a preformed positive charge (e.g., choline and carnitine containing analytes) are among the exceptions. A comparison of the mass spectra from the wild type N2 and fat 6;fat 7 mutant cuticles is displayed in Figure 4 2A and 4 2B The wild type N2 spectrum produced a number of abundant lyso and intact PCs, with the most abundant at m/ z 808, putatively identified as the [M+Na] + of PC (18:1/18:1). In contrast, the fa t 6 ; fat7 mutants demonstrated low signal from both the lyso and intact PCs, and demonstrated relatively high ion signal from m/z 258, 280, and 296, putatively identified as t he [M+H] + [M+Na] + and [M+K] + respectively, of glycerophosphocholine (GPC). Principal component analysis PCA revealed separation between the N2 wild type and fat6;fat7 mutant nematodes along the first principal component, as evidenced by the scores plot in Figure 4 3 The loadings plot shown in Figure 4 4 reveal s the mass s pectral variation between the two C. elegans genotypes The fat6;fat7 mutants are deficient in C18 and C20 mono and polyunsaturated fatty acids, 131 and thus masses associated with such lipids, most notably, PC (18:1/20:5) at m/z 828 and PC (18:1/18:1) at m/z 808 exhibit an appreciable loadings coefficient with the N2 wild type nematodes. Masses associated with glycerophosphochol ine (i.e., m/z 258, 280, and 296) lacking both the sn 1 and sn 2 fatty acid substituents, loaded heavily with the fat6;fat7 mutants. This finding is supported by p revious metabolomics studies of mutant aqueous extracts that have reported increased concen tration of glycerophosphocholine in fat mutant strains. 139
199 PCA was also applied to data collected from the N2 wild type and daf 22 mutant strains. Although separation was observed between the two genotypes, typically the within sample grouping variation was dictated by the first principal component, and a princip al component that accounted for a relatively low amoun t of the variance (typically <5 %) dictated separation between the sample groupings (Figure 4 5) Despite the relatively low variance, the loadings plot from the first principal component that dictated separation between the sample groupings was analyzed (Figure 4 6) In general, the daf 22 mutants demonstrated an increase in lysophospholipids (e.g ., m/z 542) ; however, it is currently unclear why this phenomenon occurs. MALDI MSI and tandem MS Following PCA, an MS imaging experiment was conducted to verify the presence of pertinent m/z values (with high loadings coefficients) on the cuticle. Due to the dimensions of the nematodes, approximately 10 MS spectra can be collected across the nematode using a traditional sampling approach (i.e., rastering the sample with a step size equal to the laser spot diameter). To overcome this sampling issue, an oversampling approach similar to that previously reported 140 was taken, in which approximately 75% of the previously sampled area was interrogated in each additional scan. In doing so, the number of MS spectra acquired over the nematode cuticle was increased by a factor of 16x. In these experiments, the applied laser energy and number of laser shots were sufficiently l ow as to not ablate the entire interrogated a rea, permitting multiple experiments (i.e., MS followed by MS 2 and MS 3 analyses) over a single nematode. Ions that demonstrated localization on the cuticle and an appreciable loadings coefficient on the principal component of interest were then submitted to MS n imaging
200 analysis for compound identification. In general, two classes of ions were identified by MS n : 1) choline containing phospholipids and metabo lites and 2) ade nosine containing nucleotides. Choline containing lipids and metabolites have well documented fragmentation patterns. The most abundant choline containing phospholipid, phosphatidylcholine (PC), exhibits fragmentation characteristic of th e PC headgroup; protonated PCs exhibit a characteristic fragment ion at m/z 184, and alkali adducts of PCs exhibit a characteristic neutral loss (NL) of 59. 48, 91 Adenosine containing metabolites demonstrate an abundant fragment ion at m/z 136. Furthermore, adenosine phosphates exhibit fragmentation related to the phosphate group, typically observed as NL of 80 or 98. Representative images from both the wild typ e and mutant nematodes coated with either DHB or 9 AA are shown in Figure 4 7 Imaging analysis of the N2 and daf 22 cuticles coated with DHB demonstrated higher abundance for m/z 542 in the daf 22 mutant (Figure 4 7 A). MS 2 analysis of m/z 542 displayed in Figure 4 8 yielded an abundant fragment ion at m/z 184, suggesting the presence of a protonated LPC with a single fatty acid chain, eicosapentaenoic acid ( EPA, 20:5 ), esterified to the glycerol backbone. EPA is not an abundant fatty acid observed in phospholipids (PLs) from mammalian tissues; however, previous research has identified EPA as the major fatty acid substituent in PCs for C. elegans 131, 141 The pota ssium adduct of the same LPC was identified by MS 2 and MS 3 at m/z 580 (Figure 4 9 ). Fragmentation of this ion yielded a neutral loss (NL) of 59, resulting from cleavage of trimethylamine from the phosphocholine headgroup. Subsequent fragmentation of the resulting ion at m/z 521 yielded ion signal at m/z 397 (NL of 124), resulting from cleavage of the remainder of
201 the phosphocholine headgroup. Furthermore, the fragment ion at m/z 163 has been identified as potassiated cyclophosphane, confirming this ion a s a potassium adduct. With this knowledge, the remaining mass of the precursor ion can be attributed to the fatty acid substituent, the glycerol backbone, and the hydroxyl group in the sn 2 position of the glycerol backbone. Therefore, the ion was identi fied as the [M+K] + of LPC 20:5. Other known abundant fatty acids in C. elegans 131 were als o found to be present in LPCs using a sim ilar identification strategy Imaging analysis of the N2 and fat6;fat7 cuticles coated with 9 AA demonstrated a higher abundance for both intact and lyso PCs in the wild type N2 nematodes (Figure 4 7 B). Specifically, lyso and intact PCs that contain oleic acid (18:1) in either the sn 1 or sn 2 position of the glycerol backbone were found to be depleted in the fat6;fat7 mutants, as evidenced by the images for m/z 808 (putatively identified as the [M+Na] + of PC (18:1/18:1)). Accordingly, MS n experiments wer e conducted to verify the putative assignments of ions that dictated separation between the mutant and wild type nematodes. Similar to the other alkali adducts of PCs interrogated, m/z 808 demonstrated an abundant NL of 59 following CID. MS 3 of m/z (Figure 4 10) demonstrated an abundant neutral loss of 124, once again corresponding to the loss of the remainder of the phosphocholine headgroup. Lower in abundance, the NL of 146 suggested the presence of sodium, rather than potassium, acting a s the alkali metal adducting with this lipid. Finally, the ion at m/z 467, corresponding to a NL of 282, presumably resulted from cleavage of oleic acid from the sn 1 position, and to a lesser extent the sn 2 position of the glycerol backbone. A summary of the major ions loading with either the wild type N2 or fat6;fat7 double mutants is presented in Table 4 1.
202 In contrast to the intact and lyso PCs the fat6;fat7 nematodes demonstrated a higher abundance for free glycerophosphocholine (GPC) on a single w orm basis (Figure 4 7B) This metabolite has hydroxyl groups in lieu of fatty acids in both the sn 1 and sn 2 positions of the glycerol backbone. MS 2 analysis of m/z 258 (the [M+H] + of glycerophosphocholine) confirmed this assignment, as a single abundan t fragment ion was observed at m/z 104, corresponding to protonated choline (Figure 4 11) In addition to the choline containing lipids and metabolites mentioned above, a series of ions at m/z 348, 428, and 508 were observed with relatively high signal on the DHB coated nematode cuticle (Figure 4 12 A and Figure 4 12 B). The 80 Da increase in mass with each successive ion indicated an increasing number of phosphate groups bound to a central moie ty. After further investigation, these ions were found to be adenosine monophosphate (AMP) in its linear form, adenosine diphosphate (ADP), and adenosine triphosphate (ATP), in order of increasing m/z All of these ions exhibited a characteristic fragmen t ion at m/z 136 upon CID that localized on the nematode cuticle ( Figure 4 12 C and Figure 4 12 D). This fragment ion is stably stored in the linear ion trap when fragmenting both AMP and ADP; however, m/z 136 is not stored under normal experimental conditi ons ( q activation =0.25 for m/z 508) for the higher mass ATP ion at m/z 508 To observe this fragment ion for ATP, the q of activation for m/z 508 was decreased to 0.23. In addition to the characteristic fragment ion at m/z 136, ADP and ATP exhibited fragm entation relating to phosphate groups, with characteristic NLs of 80 and 98 from both ions Conclusions This work has demonstrated the detection of endogenous compounds directly from the nematode cuticle using positive mode MA LDI MSI. In doing so, a number of
203 low molecular weight PLs and metabolites were identified. These analytes were detected directly from intact nematodes without prior extraction or alteration other than matrix application The detected PLs are likely resultin g from the lipid rich epicuticle, the outermost layer of C. elegans The fatty acid configuration in both the lysoPLs and the intact PLs agree well with previous reports detailing the analysis of fatty acid methyl esters from nematode extracts. 132 Furthermore, d istinction between wild type and gene knockouts was achieved by utilizing MALDI MSI in conjunction with PCA. The loadings plot of the first principal component, which differentiated wild type and fat6;fat7 double mutants substantiates previous reports that there is higher fatty acid content in the wild type as compared to the fat double mutant s 139 PCA also demonstrated that the fat6;fat7 mutants were deficient in many PC species containing unsaturated fatty acids ( e g 18:1) and that the mutants also exhibited a higher content of glycerophosphocholine. A similar phenomenon has b een observed in metabolomics analysis of aqueous and lipid extracts by both NMR and LC MS 139 Thus, t his research has demonstrated the utility of MALDI MS to perform metabolic profiling of whole organisms, such as C. elegans without the need for prior sectioning As opposed to previous profiling techniques that utilize on average 4 5 orders of magnitude more biological specimens this methodology has determined biologically relevant chemical alterations resulting from genetic mutations with as few as 10 biological replicates. In addition, the time required for both sample preparation and analysis was drastically reduced as compa red to LC MS profiling of the exudates. Although this research was performed using commercially available mutants that are
204 readily cultured, the work presented demonstrates the potential of MALDI MS to determine metabolic profiles for organisms that are e ither rare or impossible to culture.
205 Figure 4 1. Representative MS spectra from C. elegans utilizing DHB as a MALDI matrix Three genotypes are displayed, including A) the wild type N2 B) the fat6;fat7 and C) daf 22 mutant cuticles Ions known to r esult from the MALDI matrix are labeled with an asterisk.
206 Figure 4 2. Representative MS spectra from C. elegans utilizing 9 AA as a MALDI matrix Two genotypes are displayed, including A) the wild type N2 and B) the fat6;fat7 mutant cuticles Ions known to result from the MALDI matrix are labeled with an asterisk.
207 Figure 4 3 PCA scores plot describing the separation between wild type N2 (green crosses) and fat6;fat7 double mutant (red triangles) C. elegans
208 Figure 4 4 PCA loadings plot from principal component 1 dictating the separation between the N2 (positive) and fat6;fat7 (negative) C. elegans strains.
209 Figure 4 5. Principal component analysis scores plot (PC1 vs. PC4) detailing the separation between the wild type N2 and daf 22 gene knockout. DHB was utilized as a MALDI matrix. The ovals represent the 95% confidence interval of the sample groupings.
210 Figure 4 6. PCA loadings plot from principal component 4 dictating the separation between the N2 (positive) and daf 22 (negative) C. elegans strains.
211 Figure 4 7 MS images of wild type (Left) and mutants (Right) for various ions (structures shown) detected from the nematode cuticle. A displays a comparison of N2 and daf 22 using DHB as a MALDI matrix and B display s a comparison of N2 and fat6;fat7 using 9 AA as a MALDI matrix.
212 Figure 4 8 MALDI MS 2 spectrum obtained from a wild type N2 nematode The proposed structure of the ion, LPC 20:5, and the fragmentation pattern is also shown.
213 Figure 4 9 MALDI MS n structural elucidation of m/z 580, identified as the [M+K] + of LPC 20:5. A) MS 2 spectrum of m/z 580 and B) MS 3 spectrum of m/z type N2 nematode.
214 Figure 4 10 MALDI MS 3 spec trum of m/z collected from a wild type N2 nematode. The structure of the identified ion, the [M+Na] + of PC (18:1/18:1), and the fragmentation pattern is also displayed.
215 Figure 4 11 MALDI MS 2 spectrum of m/z 258 collected from a mutant fat6 ;fat7 nematode. The structure and fragmentation of the proposed ion, the [M+H] + of glycerophosphocholine, is also displayed.
216 Table 4 1. List of compounds that correlate strongly with the wild type N2 (positive loadings coefficient) or fat6;fat7 (negative loadings coefficient) nematodes m/z Identification Ion Loadings 162 l carnitine [M+H] + 0.036 184 Phosphocholine [M+H] + 0.313 258 Glycerophosphocholine [M+H] + 0.651 280 Glycerophosphocholine [M+Na] + 0.371 296 Glycerophosphocholine [M+K] + 0.362 522 LPC (18:1) [M+H] + 0.066 544 LPC (18:1) [M+Na] + 0.064 560 LPC (18:1) [M+K] + 0.025 786 PC (18:1/18:1) [M+H] + 0.068 806 PC (18:1/20:5) [M+H] + 0.093 808 PC (18:1/18:1) [M+Na] + 0.153 824 PC (18:1/18:1) [M+K] + 0.065 828 PC (18:1/20:5) [M+Na] + 0.143 844 PC (18:1/20:5) [M+K] + 0.072
21 7 Figure 4 12 MS 2 elucidation of AMP. A) MS image of m/z 348 normalized to the TIC, B) optical image, and C) MS 2 image of m/z type N2 nematode. The MS 2 spectrum of m/z 348 and the structure of the identified ion (AMP) is shown in D
218 CHAPTER 5 SUMMARY AND FUTURE WORK Summary This dissertation has presented multivariate data analysis strategies for two biological applications of MALDI mass spectrometric imaging datasets. Initially, a guided multivariate data analysis methodology was developed using a model system simulating myocardial infarction. Transverse cardiac sections from this model, also known as a 24 hour coronary artery lig ation, proved to be extremely valuable for method development, as a number of spatially segmented and spectrally distinct regions of interest were present. The methodology utilized dimensionality reduction techniques, namely principal component analysis ( PCA) and partial least squares discriminant analysis (PLS DA), to efficiently sift through large imaging datasets and extract relative biochemical markers for each region of interest. Application of this technique to the model system produced results consi stent with the preexisting body of knowledge for myocardial infarction. Mainly, the activity of phospholipase A 2 was confirmed in areas of infarcted myocardium. Furthermore, the presence of lipid droplets in the at risk myocardium was confirmed via the p resence of elevated triglyceride concentration. In addition, a number of novel water soluble metabolites were identified as potential blood borne biomarkers for myocardial infarction. Alt hough a number of lipidomic and metabolomic markers for myocardial i nfarction were identified with the established methodologies, experiments in Chapter 2 of this dissertation were limited to positive ionization mode with the matrix 2,5 dihydroxybenzoic acid. Chapter 3 of this dissertation describes a body of research
219 pro posing the use of 9 aminoacridine (9 AA) as a dual purpose (positive and negative mode) MALDI matrix for the study of lipids and metabolites in the same model system. Alt hough predominantly a negative mode matrix, 9 AA proved to be an effective MALDI matr ix for the ionization of phosphatidylcholines and acylcarnitines in positive ionization mode A low matrix background, coupled with the relative basicity of the matrix, served to suppress chemical noise often associated with MSI experiments on tissue. By applying the methodologies developed in Chapter 2, dicarboxyl acylcarnitines were identified as potential blood borne biomarkers of myocardial infarction. Furthermore, serial sections, coated with DHB for comparison, exhibited similar phosphatidylcholine localization to that of 9 AA coated sections, thereby validating this matrix for positive mode ionization. The same methodology was also applied to the ligation model in negative ionization mode using 9 AA as a MALDI matrix. The matrix proved to be effect ive for the ionization of a broad range of analytes, including anionic lipids (e.g., cardiolipins and phosphatidylinositols), nucleotides (e.g., adenosine monophosphate and NADH), and even select cationic lipids (e.g., lysophosphatidylethanolamin e s and lys ophosphatidylcholines). Multivariate data analysis conducted on the various regions of myocardium confirmed the ability of phospholipase A 2 to act on both cardiolipins and phosphatidylinositols. Furthermore, a number of nucleotides were identified as potential blood borne marke rs for myocardial infarction. Alt hough most nucleotides demonstrated localization in the perfused myocardium, the uridine nucleotides appeared with increased abundance in the at risk myocardium. This localization has never befo re been reported; however, various clinical trials have identified uridine
220 nucleotides as potential pharmacological agents to limit tissue loss. Thus, we hypothesize that the presence of uridine nucleotides in the at risk myocardium may have mechanistic i The final research chapter of this dissertation described a methodology for determining the metabolic profiles from Caenorhabditis elegans ( C. elegans ). In this work, three genotypes, a wild ty pe and two genetic mutants, were analyzed. Once again, a combination of MALDI MSI and multivariate data analysis was utilized to identify metabolic differences arising from the different genotypes. To increase the number of samples collected from a nemat ode, an oversampling approach was conducted. Utilizing this approach, biologically relevant biochemical distinctions between the genotypes were identified. Furthermore, these results were obtained from the analysis of individual nematodes, as opposed to thousands, or even millions of nematodes, as previously reported. Future Work Time course and reperfusion studies of myocardial infarction Due to the limited availability of biological samples, MSI experiments were performed on only one time point (24 h) f ollowing administration of the LAD coronary artery ligation surgery. A more detailed study analyzing multiple time points, both before and after the 24 h time point, may yield information concerning the progression of myocardial infarction and the myocard and long term response to the condition. In particular, understanding how myocardium responds to the loss of oxygenated blood over time via at risk myocardium markers may result in the development of pharmaceuticals to limit myocardial tissue loss. Additionally, studies that map the area of infarction at various time points immediately following ligation
221 surgery may lead to a greater understanding of optimal pharmaceutical delivery time and the expected tissue loss as a result of delay betwee n onset and administration. In addition to studying the progression of MI and tissue response to MI, it would be valuable to study the effect of reperfusion (i.e., the restoration of oxygenated blood to previously ischemic areas of myocardium) on the viabi lity and long term recovery of myocardium Currently, the initial therapeutic response applied to a patient that has experienced MI is to perform some form of reperfusion, such as thrombolysis, angioplasty, or in severe cases, bypass surgery. 142 The restoration of blood flow to the previously occluded artery results in a smaller fina l infarct size, and consequently a lower mortality rate for a given population. 143 The administration of reperfusion is unfortunately not without risk. The immediate restoration of oxygenated blood to myocardium may result in cell death in myocardium that was reversibly inj ured prior to reperfusion. 143 This process is widely known throughout the medical field as hough the beneficial effects of reperfusion appear to outwei gh the negative effects, many cardiologists believe that it is possible to minimize ischemia/reperfusion injury via intervention However, b efore an optimal intervention strategy can be formulated (e.g., postconditioning or pharmacological intervention) a better understanding of the biological processes occurring following reperfusion is needed. The methodologies developed in this work offer a promising opportunity to study the biochemical mechanism of ischemia/reperfusion injury. Furthermore, these met hodologies provide an effective means to evaluate intervention methods for ischemia/reperfusion injury.
222 Correlation of MSI and LC MS d ata Future work will also focus on correlating MSI and LC MS results from the ligation model. At present, quantitation by MALDI MSI remains difficult; however, there are many established methods for quantitation of tissue extracts and plasma samples by LC MS. A comparison of the results from both methods would serve as an effective validation for the methodologies presented in this dissertation. In particular, the quantitative analysis of tissue extracts would yield confirmation on the tissue specific biological pathways mentioned in this report. Furthermore, the ability to analyze plasma from the same biological specimen, as compared to a control specimen, would offer confirmation for the potential blood borne markers of MI proposed in this work. Improving MALDI MSI spatial r esolution on the MALDI Thermo LTQ XL via reduction of the laser spot s ize The spatial resolution of MALDI MSI experiments is theoretically limited by three factors: 1) the raster step size of the sample stage, 2) the size of the matrix/analyte cocrystals, and 3) the spot diameter of the ionization source. Of the three limitations, the raster step size of the sample proves to be insignificant on modern instruments. Furthermore, recent MALDI matrix coating techniques have produced crystal sizes much smaller than the laser spot diameter of the source utilized in this research Thus, the laser spot diamet er is currently the major hindrance for performing high spatial resolution MALDI MSI experiments. There are two prevailing strategies for reducing the effective spot diameter of a UV laser. The first strategy involves using an aperture to allow a small po rtion of the laser beam to reach the sample surface, and the second strategy involves using optics to reduce the beam diameter. Alt hough the simpler approach, t he use of an aperture
223 reduces the total energy incident on the surface by a factor of the radi us squared; however, the laser fluence (J/cm 2 ) is retained. This decrease in total energy may result in significant losses in sensitivity. In contrast, reducing the spot diameter of the laser optically retains the total energy incident on the surface, bu t with an increase in laser fluence that may lead to a greater degree of source fragmentation. Preliminary efforts have attempted to implement a Galilean beam expander 144 on the Thermo LTQ XL, of which the standard optical configuration is displayed in Figure 5 1 A Galilean beam expander, detailed in Figure 5 2, consists of two lenses : a planoconcave lens and a planoconvex lens The planoconcave and planoconvex lenses are arra nged so that the true focal point of the planoconvex lens and the virtual focal point of the planoconcave lens coincide in space. The result is that t he incident beam is continually expanded by the planoconcave lens until contacting the planoconvex lens, wherein the diameter is conserved, following the same angle as the incident beam. The expansion, or magnification ratio, of the configuration is deter mined by the ratio of the two focal points. The magnified beam can then be focused onto the sample target by a convex focusing lens. Assuming the focusing lens is larger than the incident beam, the incident beam diameter is proportional to the numerical aperture ( NA ), as the beam diameter directly determines the half angle between the outermost portion of the beam and the focal point ( Equation 5 1 ) ( 5 1) Furthermore, the numerical aperture is inversely proportional to lateral spatial resolution ( x ) and directly proportional to the wavelength of irradiation ( ), as detailed in Equation 5 2.
224 ( 5 2) B y substituting Equation 5 1 into Equation 5 2, one can see that by increasing progressively smaller lateral resolutions can be achieved. Thus, an increase in the beam diameter prior to the focusing lens will produce a concurrent decrease in the spot size on the MALDI target plate. The most logical implementation of this configuration on the LTQ XL is th e optical rail between the UV and dichroic mirrors. The optical rail provides a stable horizontal axis that allows fine adjustment of the interlens distance. Furthermore, the rail forces the two lens faces to be aligned within the plane of the mirrors An added advantage is that this region of the MALDI source is not held under vacuum, and can easily be accessed for implementation or removal of the lenses. Alt hough the lenses have been obtained, slight imperfections in the machining of the laser optics m odule have resulted in an optical rail that is not absolutely level. Furthermore, the current LTQ optical configuration allows few degrees of freedom of alignment; the UV mirror can only be adjusted in the vertical dimension and the dichroic mirror can on ly be adjusted in the horizontal dimension. Due to the aforementioned flaws, implementation of the beam expander resulted in an enlarged beam that was not centered in the convex focusing lens T he focus ed beam would then miss the hole in the transfer qua drupole (Q00), and not reach the target plate. We hypothesize that the beam is vertically off center, as the entire range of the horizontal dichroic mirror was tested. Unfortunately, the LTQ does not permit coarse vertical tuning of the bea m following th e optical rail, and any skewing of the beam by the UV mirror would result in a
225 misaligned beam travelling through the beam expander. Thus, an alternative method must be implemented to use this beam expander. Outlook MSI is an analytical technique that is rapidly gaining utilization for a breadth of applications. Specifically, MSI has inherent utility for characterization of diseases that produce spatially localized biochemical alterations within tissue. Over the course of this research, the MSI community has achieved great strides in not only multivariate data analysis, but als o quantitative analysis and high spatial and spectral resolution instrumentation. T he ability to combine multivariate data analysis with high resolution MSI will eventually p ermit the rapid analysis of multidimensional datasets at previously uncharted spatial and spectral resolutions
226 Figure 5 1. Standard o ptical Configuration of the MALDI source interfaced with the Thermo Scientific LTQ XL.
227 Figure 5 2. Representation of a Galilean beam expander.
228 Figure 5 3. Schematic of LTQ optical configuration with Galilean beam expander implemented on the optical rail. The spacing between the planoconcave and planoconvex lenses is 10 cm.
229 LIST OF REFERENCES (1) Caprioli, R. M.; Farmer, T. B.; Gile, J. Anal. Chem. 1997 69 4751 4760. (2) Chughtai, K.; Heeren, R. M. A. Chem. Rev. (Washington, DC, U. S.) 2010 110 3237 3277. (3) Karas, M.; Hillenkamp, F. Anal. Chem. 1988 60 2299 2301. (4) Karas, M.; Bachmann, D.; Bahr, U.; Hillenkamp, F. Int. J. Mass Spectrom. Ion Processes 1987 78 53 68. (5) Tanaka, K.; Waki, H.; Ido, Y.; Akita, S.; Yoshida, Y.; Yoshida, T.; Matsuo, T. Rapid Commun. Mass Spectrom. 1988 2 151 153. (6) de Hoffmann, E.; S troobant, V. Mass Spectrometry: Principles and Applications 3rd ed.; Wiley: New York, 2007. (7) Kaletas, B. K.; van der Wiel, I. M.; Stauber, J.; Dekker, L. J.; Guzel, C.; Kros, J. M.; Luider, T. M.; Heeren, R. M. A. Proteomics 2009 9 2622 2633. (8) Kar as, M.; Gluckmann, M.; Schafer, J. J. Mass Spectrom. 2000 35 1 12. (9) Zenobi, R.; Knochenmuss, R. Mass Spectrom. Rev. 1998 17 337 366. (10) Karas, M.; Kruger, R. Chem. Rev. (Washington, DC, U. S.) 2003 103 427 439. (11) Jaskolla, T. W.; Karas, M. J. Am. Soc. Mass Spectrom. 2011 22 976 988. (12) Baluya, D. L.; Garrett, T. J.; Yost, R. A. Anal. Chem. 2007 79 6862 6867. (13) Hankin, J. A.; Barkley, R. M.; Murphy, R. C. J. Am. Soc. Mass Spectrom. 2007 18 1646 1652. (14) Amstalden van Hove, E. R.; S mith, D. F.; Heeren, R. M. A. J. Chromatogr., A 2010 1217 3946 3954. (15) Lanni, E. J.; Rubakhin, S. S.; Sweedler, J. V. J. Proteomics 2012 75 5036 5051. (16) Wu, K. J.; Odom, R. W. Anal. Chem. 1996 68 873 882. (17) Winograd, N. Appl. Surf. Sci. 2003 203 13 19. (18) Takats, Z.; Wiseman, J. M.; Gologan, B.; Cooks, R. G. Science 2004 306 471 473.
230 (19) Takats, Z.; Wiseman, J. M.; Cooks, R. G. J. Mass Spectrom. 2005 40 1261 1275. (20) Trimpin, S.; Inutan, E. D.; Herath, T. N.; McEwen, C. N. Mol. Cel l. Proteomics 2010 9 362 367. (21) Simons, D. S. Appl. Surf. Sci. 1988 31 103 117. (22) Guenther, S.; Rmpp, A.; Kummer, W.; Spengler, B. Int. J. Mass Spectrom. 2011 305 228 237. (23) McEwen, C. N.; Trimpin, S. Int. J. Mass Spectrom. 2011 300 167 172. (24) Nemes, P.; Vertes, A. Anal. Chem. 2007 79 8098 8106. (25) Van Berkel, G. J.; Pasilis, S. P.; Ovchinnikova, O. J. Mass Spectrom. 2008 43 1161 1180. (26) Paul, W.; Steinwedel, H. Apparatus for Separating Charged Particles of Different Speci fic Charges. U.S. Patent 2,939,952, June 7, 1960. (27) March, R. E.; Todd, J. F. J. Quadrupole Ion Trap Mass Spectrometry 2nd ed.; Wiley: Hoboken, New Jersey, 2005. (28) Schwartz, J. C.; Senko, M. W.; Syka, J. E. P. J. Am. Soc. Mass Spectrom. 2002 13 65 9 669. (29) LTQ ; Thermo Electron Corporation: San Jose, CA, 2006. (30) Syka, J. E. P.; Flies, W. J. J. Fourier Transform Quadrupole Mass Spectrometer and Method. U.S. Patent 4,755,670, July 5, 1988. (31) Garrett, T. J.; Prieto Conaway, M. C.; Kovtoun, V.; Bui, H.; Izgarian, N.; Stafford, G.; Yost, R. A. Int. J. Mass Spectrom. 2007 260 166 176. (32) Magparangalan, D. P.; Garrett, T. J.; Drexler, D. M.; Yost, R. A. Anal. Chem. 2010 82 930 934. (33) Marsh all, A. G.; Hendrickson, C. L. Int. J. Mass Spectrom. 2002 215 59 75. (34) Marshall, A. G.; Hendrickson, C. L.; Jackson, G. S. Mass Spectrom. Rev. 1998 17 1 35. (35) Marshall, A. G. Int. J. Mass Spectrom. 2000 200 331 356. (36) Glish, G. L.; Vachet, R. W. Nat. Rev. Drug Discovery 2003 2 140 150.
231 (37) Makarov, A. Anal. Chem. 2000 72 1156 1162. (38) Hu, Q. Z.; Noll, R. J.; Li, H. Y.; Makarov, A.; Hardman, M.; Cooks, R. G. J. Mass Spectrom. 2005 40 430 443. (39) Denisov, E.; D amoc, E.; Lange, O.; Makarov, A. Int. J. Mass Spectrom. 2012 325 80 85. (40) Makarov, A.; Denisov, E.; Lange, O. J. Am. Soc. Mass Spectrom. 2009 20 1391 1396. (41) Schwartz, J. C.; Syka, J. E. P.; Jardine, I. J. Am. Soc. Mass Spectrom. 1991 2 198 204 (42) Stoeckli, M.; Staab, D.; Staufenbiel, M.; Wiederhold, K. H.; Signor, L. Anal. Biochem. 2002 311 33 39. (43) Pirman, D. A.; Reich, R. F.; Kiss, A.; Heeren, R. M. A.; Yost, R. A. Anal. Chem. 2013 85 1081 1089. (44) Pirman, D. A.; Yost, R. A. Anal. Chem. 2011 83 8575 8581. (45) Pirman, D. A.; Kiss, A.; Heeren, R. M. A.; Yost, R. A. Anal. Chem. 2013 85 1090 1096. (46) Fonville, J. M.; Carter, C.; Cloarec, O.; Nicholson, J. K.; Lindon, J. C.; Bunch, J.; Holmes, E. Anal. Chem. 2012 84 1310 1319. (47) Jolliffe, I. T. Principal Component Analysis ; Springer Verlag: New York, 1986. (48) Menger, R. F.; Stutts, W. L.; Anbukumar, D. S.; Bowden, J. A.; Ford, D. A.; Yost, R. A. Anal. Chem. 2012 84 1117 1125. (49) Xia, J.; Psychogios, N.; Young, N.; Wisha rt, D. S. Nucleic Acids Res. 2009 37 W652 W660. (50) van Hove, E. R. A.; Blackwell, T. R.; Klinkert, I.; Eijkel, G. B.; Heeren, R. M. A.; Glunde, K. Cancer Res. 2010 70 9012 9021. (51) Douglas, D. J.; Frank, A. J.; Mao, D. M. Mass Spectrom. Rev. 2005 24 1 29. (52) March, R. E. Int. J. Mass Spectrom. 2000 200 285 312. (53) Xu, J.; Kochanek, K. D.; Murphy, S. L.; Tejada Vera, B. Natl. Vital Stat. Rep. 2010 58 1 135.
232 (54) Jemal, A.; Ward, E.; Hao, Y. P.; Thun, M. JAMA, J. Am. Med. Assoc. 2005 294 1 255 1259. (55) Hansson, G. K. N. Engl. J. Med. 2005 352 1685 1695. (56) Morrow, D. A.; Cannon, C. P.; Jesse, R. L.; Newby, L. K.; Ravkilde, J.; Storrow, A. B.; Wu, A. H. B.; Christenson, R. H.; Apple, F. S.; Francis, G.; Tang, W.; Members, N. W. G.; Memb ers, N. C. Clin. Chem. (Washington, DC, U. S.) 2007 53 552 574. (57) Gerszten, R. E.; Wang, T. J. Nature 2008 451 949 952. (58) Macht, M.; Fiedler, W.; Kurzinger, K.; Przybylski, M. Biochemistry 1996 35 15633 15639. (59) Ford, D. A.; Han, X.; Horner, C. C.; Gross, R. W. Biochemistry 1996 35 7903 7909. (60) Marshall, J.; Kupchak, P.; Zhu, W. M.; Yantha, J.; Vrees, T.; Furesz, S.; Jacks, K.; Smith, C.; Kireeva, I.; Zhang, R.; Takahashi, M.; Stanton, E.; Jackowski, G. J. Proteome Res. 2003 2 361 372. (61) Delles, C.; Schiffer, E.; von zur Muhlen, C.; Peter, K.; Rossing, P.; Parving, H. H.; Dymott, J. A.; Neisius, U.; Zimmerli, L. U.; Snell Bergeon, J. K.; Maahs, D. M.; Schmieder, R. E.; Mischak, H.; Dominiczak, A. F. J. Hypertens. 2010 28 2316 2322. (62) von zur Muhlen, C.; Schiffer, E.; Zuerbig, P.; Kellmann, M.; Brasse, M.; Meert, N.; Vanholder, R. C.; Dominiczak, A. F.; Chen, Y. C.; Mischak, H.; Bode, C.; Peter, K. J. Proteome Res. 2009 8 335 345. (63) Zhang, H. Y. ; Chen, X.; Hu, P.; Liang, Q. L.; Liang, X. P.; Wang, Y. M.; Luo, G. A. Talanta 2009 79 254 259. (64) Sabatine, M. S.; Liu, E.; Morrow, D. A.; Heller, E.; McCarroll, R.; Wiegand, R.; Berriz, G. F.; Roth, F. P.; Gerszten, R. E. Circulation 2005 112 3868 3875. (65) Anderson, L. J. Physiol. London 2005 563 23 60. (66) Garrett, T. J.; Yost, R. A. In Practical Aspects of Trapped Ion Mass Spectrometry, Volume V ; CRC Press: Boca Raton, Florida, 2009, pp 417 438. (67) Chaurand, P.; Schwartz, S. A.; Caprioli, R. M. Anal. Chem. 2004 76 86A 93A.
233 (68) Taban, I. M.; Altelaar, A. F. M.; van der Burgt, Y. E. M.; McDonnell, L. A.; Heeren, R. M. A.; Fuchser, J.; Baykut, G. J. Am. Soc. Mass Spectrom. 2007 18 145 151. (69) Landgraf, R. R.; Conaway, M. C. P.; Garrett, T. J.; Stacpoole, P. W.; Yost, R. A. Anal. Chem. 2009 81 8488 8495. (70) Garrett, T.; Menger, R.; Dawson, W.; Yost, R. Anal. Bioanal. Chem. 2011 1 11. (71) Hsieh, Y.; Casale, R.; Fukuda, E.; Chen, J. W.; Knem eyer, I.; Wingate, J.; Morrison, R.; Korfmacher, W. Rapid Commun. Mass Spectrom. 2006 20 965 972. (72) Hopfgartner, G.; Varesio, E.; Stoeckli, M. Rapid Commun. Mass Spectrom. 2009 23 733 736. (73) Reich, R. F.; Magparangalan, D. P.; Garrett, T. J.; Yos t, R. A. In Mass Spectrometry in Drug Metabolism and Disposition ; John Wiley & Sons, Inc.: Hoboken, New Jersey, 2011, pp 449 482. (74) Hankin, J. A.; Farias, S. E.; Barkley, R.; Heidenreich, K.; Frey, L. C.; Hamazaki, K.; Kim, H. Y.; Murphy, R. C. J. Am. S oc. Mass Spectrom. 2011 22 1014 1021. (75) Trim, P. J.; Atkinson, S. J.; Princivalle, A. P.; Marshall, P. S.; West, A.; Clench, M. R. Rapid Commun. Mass Spectrom. 2008 22 1503 1509. (76) McCombie, G.; Staab, D.; Stoeckli, M.; Knochenmuss, R. Anal. Chem 2005 77 6118 6124. (77) Bonnel, D.; Longuespee, R.; Franck, J.; Roudbaraki, M.; Gosset, P.; Day, R.; Salzet, M.; Fournier, I. Anal. Bioanal. Chem. 2011 401 149 165. (78) Fishbein, M. C.; Maclean, D.; Maroko, P. R. Am. J. Pathol. 1978 90 57 70. (79) Benedek, A.; Moricz, K.; Juranyi, Z.; Gigler, G.; Levay, G.; Harsing, L. G.; Matyus, P.; Szenasi, G.; Albert, M. Brain Res. 2006 1116 159 165. (80) Schwartz, S. A.; Reyzer, M. L.; Caprioli, R. M. J. Mass Spectrom. 2003 38 699 708. (81) Armenta, J. M.; Hoeschele, I.; Lazar, I. M. J. Am. Soc. Mass Spectrom. 2009 20 1287 1302. (82) Xia, J.; Wishart, D. S. Nat. Protoc. 2011 6 743 760.
234 (83) Goldlust, E. J.; Paczynski, R. P.; He, Y. Y.; Hsu, C. Y.; Goldberg, M. P. Stroke 1996 27 1657 166 2. (84) Thygesen, K.; Alpert, J. S.; White, H. D.; Force, E. A. A. W. T. Circulation 2007 116 2634 2653. (85) Ford, D. A. Prog. Lipid Res. 2002 41 6 26. (86) Ford, D. A.; Hazen, S. L.; Saffitz, J. E.; Gross, R. W. J. Clin. Invest. 1991 88 331 335. (8 7) Mancuso, D. J.; Abendschein, D. R.; Jenkins, C. M.; Han, X. L.; Saffitz, J. E.; Schuessler, R. B.; Gross, R. W. J. Biol. Chem. 2003 278 22231 22236. (88) Sobel, B. E.; Corr, P. B.; Robison, A. K.; Goldstein, R. A.; Witkowski, F. X.; Klein, M. S. J. Cl in. Invest. 1978 62 546 553. (89) Straeter Knowlen, I. M.; Evanochko, W. T.; Den Hollander, J. A.; Wolkowicz, P. E.; Balschi, J. A.; Caulfield, J. B.; Ku, D. D.; Pohost, G. M. Circulation 1996 93 1464 1470. (90) Petkovic, M.; Schiller, J.; Muller, M.; Benard, S.; Reichl, S.; Arnold, K.; Arnhold, J. Anal. Biochem. 2001 289 202 216. (91) Han, X.; Gross, R. W. J. Am. Soc. Mass Spectrom. 1995 6 1202 1210. (92) Garrett, T. J.; Yost, R. A. Anal. Chem. 2006 78 2465 2469. (93) Sullards, M. C. In Methods Enzymol. ; Academic Press: San Diego, California, 2000; Vol. 312, pp 32 45. (94) Piotrowska, M.; Kurek, K.; Blachnio Lipids 2012 47 847 853. (95) Starkopf, J.; Andreasen, T. V.; Bugge, E.; Ytrehus, K. Cardiovasc. Res. 1998 37 66 75. (96) Nakane, M.; Kubota, M.; Nakagomi, T.; Tamura, A.; Hisaki, H.; Shimasaki, H.; Ueta, N. Neurosci. Lett. 2000 296 89 92. (97) Robinson, B. S.; Hii, C. S. T.; Poulos, A.; Ferrante, A. Immunology 1 997 91 274 280. (98) Hannun, Y. A. J. Biol. Chem. 1994 269 3125 3128. (99) Schiller, J.; Arnhold, J.; Benard, S.; Muller, M.; Reichl, S.; Arnold, K. Anal. Biochem. 1999 267 46 56.
235 (100) Strupat, K.; Karas, M.; Hillenkamp, F. Int. J. Mass Spectrom. Io n Processes 1991 111 89 102. (101) Gobom, J.; Schuerenberg, M.; Mueller, M.; Theiss, D.; Lehrach, H.; Nordhoff, E. Anal. Chem. 2001 73 434 438. (102) Staab, H. A.; Saupe, T. Angewandte Chemie International Edition in English 1988 27 865 879. (103) Thomas, A.; Charbonneau, J. L.; Fournaise, E.; Chaurand, P. Anal. Chem. 2012 84 2048 2054. (104) Eibisch, M.; S, R.; Schiller, J. Rapid Commun. Mass Spectrom. 2012 26 1573 1576. (105) Juhasz, P.; Costello, C. E. J. Am. Soc. Mass Spectrom. 1992 3 78 5 796. (106) Vermillion Salsbury, R. L.; Hercules, D. M. Rapid Commun. Mass Spectrom. 2002 16 1575 1581. (107) Rapid Commun. Mass Spectrom. 2007 21 3295 3300. (108) Edwards, J. L.; Kennedy, R. T. Anal. Chem. 2005 77 2 201 2209. (109) Sun, G.; Yang, K.; Zhao, Z.; Guan, S.; Han, X.; Gross, R. W. Anal. Chem. 2007 79 6629 6640. (110) Cheng, H.; Sun, G.; Yang, K.; Gross, R. W.; Han, X. J. Lipid Res. 2010 51 1599 1609. (111) Cerruti, C. D.; Benabdellah, F.; Laprvote, O.; Touboul, D.; Brunelle, A. Anal. Chem. 2012 84 2164 2171. (112) Burrell, M.; Earnshaw, C.; Clench, M. J. Exp. Bot. 2007 58 757 763. (113) Marsching, C.; Eckhardt, M.; Groene, H. J.; Sandhoff, R.; Hopf, C. Anal. Bioanal. Chem. 2011 401 53 64. (114) Fuchs, B.; Bischoff, A.; Suss, R.; Teuber, K.; Schuerenberg, M.; Suckau, D.; Schiller, J. Anal. Bioanal. Chem. 2009 395 2479 2487. (115) Wang, H. Y. J.; Post, S. N. J. J.; Woods, A. S. Int. J. Mass Spectrom. 2008 278 143 149. (116) Hsu, F. F.; Turk, J. J. Am. Soc. Mass Spectrom. 2000 11 986 999. (117) Zima, A. V.; Copello, J. A.; Blatter, L. A. J. Physiol. London 2004 555 727 741.
236 (118) Gabelica, V.; Schulz, E.; Karas, M. J. Mass Spectrom. 2004 39 579 593. (119) Hillenkamp, F.; Karas, M. In MALDI MS ; Wiley VCH Verlag GmbH & Co. KGaA: Weinheim, Germany, 2007, pp 1 28. (120) Ellis, J. M.; Mentock, S. M.; DePetrillo, M. A.; Koves, T. R.; Sen, S.; Watkins, S. M.; Muoio, D. M.; Cline, G. W.; Taegtmeyer, H.; Shulman, G. I.; Willis, M. S.; Coleman, R. A. Mol. Cell. Biol. 2011 31 1252 1262. (121) Hsu, F. F.; Turk, J. J. Am. Soc. Mass Spectrom. 2003 14 352 363. (122) Domingues, P.; Domingues, M. R. M.; Amado, F. M. L.; Ferrer Correia, A. J. Rapid Commun. Mass Spectrom. 2001 15 799 804. (123) Sha h, S. H.; Sun, J. L.; Stevens, R. D.; Bain, J. R.; Muehlbauer, M. J.; Pieper, K. S.; Haynes, C.; Hauser, E. R.; Kraus, W. E.; Granger, C. B.; Newgard, C. B.; Califf, R. M.; Newby, L. K. Am. Heart J. 2012 163 844 850.e841. (124) Wihlborg, A. K.; Balogh, J .; Wang, L.; Borna, C.; Dou, Y.; Joshi, B. V.; Lazarowski, E.; Jacobson, K. A.; Arner, A.; Erlinge, D. Circ. Res. 2006 98 970 976. (125) Erlinge, D.; Harnek, J.; van Heusden, C.; Olivecrona, G. r.; Jern, S.; Lazarowski, E. Int. J. Cardiol. 2005 100 427 433. (126) Yitzhaki, S.; Shneyvays, V.; Jacobson, K. A.; Shainberg, A. Biochem. Pharmacol. 2005 69 1215 1223. (127) Yew, J. Y.; Soltwisch, J.; Pirkl, A.; Dreisewerd, K. J. Am. Soc. Mass Spectrom. 2011 22 1273 1284. (128) Brenner, S. Genetics 1974 77 71 94. (129) Sulston, J. E.; Horvitz, H. R. Dev. Biol. 1977 56 110 156. (130) Srinivasan, J.; Kaplan, F.; Ajredini, R.; Zachariah, C.; Alborn, H. T.; Teal, P. E. A.; Malik, R. U.; Edison, A. S.; Sternberg, P. W.; Schroeder, F. C. Nature 2008 454 11 15 U1146. (131) Brock, T. J.; Browse, J.; Watts, J. L. Genetics 2007 176 865 875. (132) Satouchi, K.; Hirano, K.; Sakaguchi, M.; Takehara, H.; Matsuura, F. Lipids 1993 28 837 840. (133) Watts, J. L.; Browse, J. Proc. Natl. Acad. Sci. U. S. A. 2002 99 5854 5859. (134) Brock, T. J.; Browse, J.; Watts, J. L. PLoS Genet. 2006 2 997 1005.
237 (135) Shmookler Reis, R. J.; Xu, L.; Lee, H.; Chae, M.; Thaden, J. J.; Bharill, P.; Tazearslan, C.; Siegel, E.; Alla, R.; Zimniak, P.; Ayyadevara, S. Aging 2011 3 125 147. (136) Butcher, R. A.; Ragains, J. R.; Li, W.; Ruvkun, G.; Clardy, J.; Mak, H. Y. Proc. Natl. Acad. Sci. U. S. A. 2009 106 1875 1879. (137) Butcher, R. A.; Fujita, M.; Schroeder, F. C.; Clardy, J. Nat. Chem. Biol. 2007 3 420 422. (138) Watts, J L.; Browse, J. Biochem. Biophys. Res. Commun. 2000 272 263 269. (139) Castro, C.; Sar, F.; Shaw, W. R.; Mishima, M.; Miska, E. A.; Griffin, J. L. BMC Genomics 2012 13 1 13. (140) Jurchen, J.; Rubakhin, S.; Sweedler, J. J. Am. Soc. Mass Spectrom. 2005 16 1654 1659. (141) Tanaka, T.; Ikita, K.; Ashida, T.; Motoyama, Y.; Yamaguchi, Y.; Satouchi, K. Lipids 1996 31 1173 1178. (142) Dirksen, M. T.; Laarman, G. J.; Simoons, M. L.; Duncker, D. J. G. M. Cardiovasc. Res. 2007 74 343 355. (143) Ibanez, B.; Fuster, V.; Jimenez Borreguero, J.; Badimon, J. J. Int. J. Cardiol. 2011 151 3 11. (144) Weichel, H. Laser beam propagation in the atmosphere ; SPIE Optical Engineering Press: Bellingham, Washington, 1990.
238 BIOGRAPHICAL SKETCH Robert Francis Menger was born in 1986 to Elaine and Arthur Menger the second of two chil dren He lived in East Islip, NY prior to attending Wake Forest University for college. At W ake For est, Robert received a Bachelor of Science in c hemistry and a Minor in m athematics. While studying capillary electrophoresis under Dr. Christa Colyer, Robert received the Eastern Analytical Symposium award for undergraduate research, and ultimately decided to pursue a career in analytical chemistry research. Following completion of his undergraduate studies, Robert joined the Richard A. Yost research group at the University of Florida t o work in mass spectrometric imaging Under the direction of Dr. Richard A. Yost, Robert received the Roger and Jo Bates fellowship. During his studies, Robert has twice traveled to the FOM Institute for Atomic and Molecular Physics in Amsterdam, the Netherlands to study cutting edge mass spectrometric imaging and statistical methods under the supervision of Profess or Ron Heeren Throughout his research career, Robert has developed a passion for studying biological systems on a molecular level, and he looks forward to new and exciting opportunities to pursue this passion following graduate school.