In vivo magnetic resonance spectroscopy of musculoskeletal disorders

MISSING IMAGE

Material Information

Title:
In vivo magnetic resonance spectroscopy of musculoskeletal disorders
Physical Description:
xvii, 208 leaves : ill., photos ; 29 cm.
Language:
English
Creator:
Ballinger, James Ray
Publication Date:

Subjects

Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1994.
Bibliography:
Includes bibliographical references (leaves 188-207).
Statement of Responsibility:
by James Ray Ballinger.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002019642
notis - AKK7091
oclc - 32801344
System ID:
AA00003615:00001

Full Text













IN VIVO MAGNETIC RESONANCE SPECTROSCOPY
OF MUSCULOSKELETAL DISORDERS















By JAMES RAY BALLINGER
















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


1994
















ACKNOWLEDGEMENTS


I would like to express most sincere thanks for the following

people who made this educational experience possible for me:

Katherine Scott, Ph.D., Chairperson of my committee, for her

inspiration, skilled guidance, patience, and wisdom; Raymond

Andrew, Ph.D., for his thoughtful comments on my publications and

thesis; Jeffrey Fitzsimmons, Ph.D., for discussions about coils,

NMR in general, and his insight into people; Richard Briggs,

Ph.D., for discussions about pulse sequence techniques and pH

measurements; Thomas Mareci, Ph.D., for discussions about lactate

editing and localization techniques; and Haejin Kang, Ph.D., for

teaching me how to use the animal spectrometer, implementing

phosphorus spectroscopy of implanted tumors in mice, acquiring.

data, and constructing two of the phosphorus coils used in my

studies. He has been an "adopted brother" to me and I will always

remember my time with him.

I would also like to thank Byron Croker, M.D., for help with

culturing tumor cells, planning mouse studies, and reviewing

results of the tumor culture and mouse experiments; William Brey,

Randy Duensing, Andrew Mitchell, fellow graduate students, who

have shared time and thoughts with me; Carol Sweeney, for the cell

culture work, mouse husbandry, collection and fitting of mouse

data, and being a friend; Jim Scott, for construction of the in-

magnet ergometer and phantoms, helping to collect exercise data,









volunteering to spend time in the magnet, and discussions ranging

from RF coils to Shiitake mushrooms; Barbara Beck, for instruction

on coil construction; Teresa Lyles, M.S., for her secretarial and

statistical expertise, and her moral support; Christine Stopka,

Ph.D., for her wonderful personality and encouragement to study

claudication patients with magnetic resonance spectroscopy; Lori

K. Marburger, M.S., who trained the peripheral vascular disease

patients, performed the exercise tests and analyzed the data from

the initial batch of patients; William Breshue, Ph.D., for

referral of congestive heart failure and heart transplant patients

and for stimulating discussions about phosphorus metabolism and

exercise physiology; Alka Velenik, Ph.D., for setting up the

transfer of exercise data from the Seimens scanner to a

workstation for analysis; Michael Welsh, for coordinating

spectroscopy sessions for the congestive heart failure patients

and help in analyzing data; Michael Ingenio, M.D., for helping to

acquire data on more recent claudication patients; Auhre-~ Sobczak,

for help in fitting spectra and entering data; and Susan Kohler,

Ph.D., for her helpful suggestions on implementing magnetic

resonance spectroscopy on out patients and for writing a macro for

manipulation of chemical shift imaging data.

Finally, I would like to thank Suzanne Spanier, M.D., for

helpful discussions and providing pathological correlation for the

musculoskeletal tumor cases; Juan Vasquez, for helping to work out

"bugs" in the hardware of the whole body scanner; Drs. Mark

Scarborough and Tom Nelson for referral of musculoskeletal tumor

patients; Nancy Dixon, RN, for helping to coordinate the

spectroscopy sessions with the tumor patients' chemotherapy









hospitalizations; Drs. Rudy Gurtner and James Seeger for referral

of our peripheral vascular disease patients; Vincent Group6,

Ph.D., for encouraging me to return to Graduate School for my

Ph.D., and for teaching me to ascribe any success I may have to

God, and not to my own questionable superiority; Joanne Ballinger,

PharmD, my wife, who has been patient with my odd hours through

graduate school; and last but not least, Christopher and Jennifer,

my son and daughter, for helping me to keep my sense of humor, to

live one day at a time, and to enjoy my second childhood.

















TABLE OF CONTENTS


ACKNOWLEDGEMENTS ii

LIST OF TABLES vii

LIST OF FIGURES ix

KEY TO SYMBOLS AND ABBREVIATIONS . xiv

ABSTRACT xvi

CHAPTERS

1 INTRODUCTION 1

Research Hypotheses 3
Specific Objectives 6
Collaborations. 8

2 TECHNIQUES 10

Localization Techniques. . 10
Water Suppression Techniques . 14
Lactate Editing 18
Quantitation of molar Concentrations of Metabolites
In Vivo 20

3 P-31 SPECTROSCOPY OF HUMAN OSTEOSARCOMA IN A MOUSE MODEL
AND IN HUMANS 26

Review of the Literature . 26
Treated and Untreated Mice. . 32
Sensitive and Resistant Mice . 47
P-31 Spectroscopy of Musculoskeletal Tumors in
Humans 57
Anesthetic and Temperature Dependence of pH and P-31
Metabolites in Implanted Human Osteosarcoma in
Nude Mice 72

4 PROTON SPECTROSCOPY OF MUSCULOSKELETAL TUMORS .87

Review of the Literature . .87
Mouse Study. 92
Human Study. 102











5 PHOSPHORUS SPECTROSCOPY OF SKELETAL MUSCLE IN
PERIPHERAL VASCULAR DISEASE AND CONGESTIVE HEART
FAILURE 135

P-31 MRS Monitoring of Low Intensity, Pain Free
Exercise Therapy for Individuals with
Intermittent Claudication due to Peripheral
Vascular Disease 135
Comparison of Cellular Metabolism in PVD Patients,
CHF Patients, Heart Transplant Patients and
Normal Volunteers using P-31 MRS 153

6 SUMMARY AND CONCLUSIONS . 166

APPENDICES

A SAMPLE SIZE DETERMINATION . 172

B STUDY SUBJECT DEMOGRAPHICS . 173

C INFORMED CONSENT FORM FOR MUSCULOSKELETAL TUMOR PATIENTS
. 174

D SUPERVISORY COMMITTEE MEMBERS . 178

E EQUIPMENT LIST 179

F EXPERIMENTAL PARAMETERS FOR MRS . 181

G CHEMOTHERAPY PROTOCOL . 185

H B, DISTRIBUTION PLOTS FOR RF COILS .186

REFERENCES 188

BIOGRAPHICAL SKETCH 208

















LIST OF TABLES


Table 3-1
Results of Four-day Growth Assay .51

Table 3-2
P-31 T1 Relaxation Times in Normal Skeletal Muscle 63

Table 3-3
Variation in Normalized and Unnormalized P-31 Metabolites
over Three Months in Skeletal Muscle of Volunteers 67

Table 3-4
Normalized Metabolite Signal for Tumor and Skeletal Muscle
72

Table 3-5
Difference Between Rectal and Skin Temperature in Mice .81

Table 4-1
Mouse Metabolite Results Normalized to Unsuppressed Water
Signal 102

Table 4-2
Summary of Tumor Patients . 106

Table 4-3
Peak Positions of Metabolites of Interest .. 121

Table 4-4
Relaxation Measurements from Four Volunteers and One Patient
121

Table 4-5
Summary of Cho/Cr Ratio Data .. 125

Table 4-6
Per Cent Change (decrease) in the Cho/Cr Ratio
in Tumors with Follow up . 128

Table 5-1
Mean and Standard Deviation of Resting Metabolites and pH
163

Table 5-2
Mean and Standard Deviation of the Metabolites
and pH at their Extremes . 163

vii









Table 5-3
Mean and Standard Deviation of the Recovery Rates of
Metabolites and pH 164


Table 5-4
MVC and Time to Fatigue during Exercise Test


viii


. 165

















LIST OF FIGURES


Figure 3-1

Figure 3-2







Figure 3-3



Figure 3-4



Figure 3-5


Figure 3-6




Figure 3-7



Figure 3-8







Figure 3-9



Figure 3-10


RF coil positioned over mouse tumor .


36


Representative spectra from an untreated tumor at
five days (Bottom) and twenty-three days (top)
post implantation. The PME peak is labeled with a
long vertical arrow, the PCr peak with a short
horizontal arrow. Note the increase in size of the
PME peak and the decrease in size of the PCr peak
on day twenty-three. . 40

Graph of the mean volume of treated and untreated
tumors as a function of time. Arrow marks chemo


day.


41


Graph of the mean PME of the treated and untreated
tumors as a function of time. Arrow marks chemo
day. 41

Graph of the mean PCr/Pi ratio for the treated and
untreated tumors as a function of time. 43

ROC Curve for detecting untreated tumors with
PCr/Pi ratio. The true positive rate (sensitivity)
is plotted as a function of the false positive
rate (1-specificity) 45

Mean change in volume of sensitive and resistant
tumors as a function of time. All of the mice
received chemotherapy on day zero. 53

Mean change in PCr/Pi for sensitive and resistant
mice as a function of time. All of the mice
received chemotherapy on day 0. Change in the
PCr/Pi ratio is seen as early as day 1 post
chemotherapy. This change occurs before any change
in the volume of the tumors (compare with Figure
3-9). 54

Graph of the mean PME one standard deviation vs
time for sensitive and resistant mice. All of the
mice received chemotherapy on day zero. 55

ROC Curve for detecting resistant tumors with the
change in slope of the PCr/Pi ratio
postchemotherapy. The true positive rate












Figure 3-11


Figure 3-12








Figure 3-13



Figure 3-14



Figure 3-15



Figure 3-16




Figure 3-17





Figure 3-18


Figure 3-19



Figure 3-20


(sensitivity) is plotted as a function of the
false positive rate (1-specificity). 56

Photograph of the 1.5 T GE Signa whole-body
imager. 59

Photograph of CSI data superimposed on an axial MR
image of tumor. The arrow points to the femur.
Surrounding the black cortex of the femur on the
left, right and top, is the malignant tumor
appearing grey. The white areas below the femur
are mostly fat. The small light grey object at the
bottom of the photograph is the external standard.
. 64

Pre (bottom) and post (top) baseline-corrected P-
31 spectra from the skeletal muscle of a normal
volunteer. 66

Graph demonstrating linearity of the PO4 signal
plotted as a function of the CSI voxel volume.
. 68

P-31 spectrum from malignant tumor in a patient.
Note lower signal to noise when compared to a
normal volunteer (Figure 3-13). 70

An example of a fitted tumor spectrum from the
same patient as Figure 3-15. The spectrum is from
a different part of the tumor than that in Figure
3-15. 71

Rectal temperature of unheated anesthetized and
unheated, unanesthetized mice vs time post
anesthesia. One anesthetized mouse showed a
spontaneous increase in temperature after about
1.25 hours. 78

Rectal temperature of the heated, anesthetized
mice vs time postanesthesia. The slight dip in
temperature before 30 mins occurred while the mice
and the coil were being positioned. 80

Graph of the pH of the six heated mice vs
experiment number. Note lack of significant
change over time (experiment number). 82

Graph of the pH of the six unheated,
anesthetized mice vs rectal temperature.
Regression lines have been fitted to the data of
all but one of the mice. See the text for an
explanation. 83









Figure 3-21



Figure 4-3




Figure 4-4






Figure 4-5






Figure 4-6










Figure 4-7



Figure 4-8



Figure 4-9






Figure 4-10


Graph of peak areas of the PCr, Pi, and ATP as a
function of time. Experiment numbers are at
approximately 15 min intervals. 84

H-1 spectra from mouse tumor acquired on 2 T
spectrometer. The top spectrum is with water
suppression; the bottom spectrum is without water
suppression. 97

Water-suppressed H-I spectrum from mouse tumor
acquired on 1.5 T whole-body MR scanner. The short
arrow marks the partially-suppressed water peak
and the long arrow marks the Cho/Tau peak. The
large unmarked peak is the lipid signal. The
number of averages was 768. 98

T1 and T2 weighted images of tumor in mouse. a)
Tl-weighted SE image (TR: 0.5 sec, TE: 30 ms). b)
T2-weighted SE image (TR: 2 sec, TE: 80 ms). The
tumor is grey in shade with an arrow pointing to
the tumor/muscle interface. The white areas around
the tumor are subcutaneous fat. 100

Tl-weighted images, pre and post gadolinium-DTPA
enhancement, a) Pre contrast image. b) Post
contrast image. Note improved delineation of
tumor/muscle boundary (arrows) on post contrast
image. The skin is seen as a thin rim of light
grey between the bright subcutaneous fat and the
black air. Along the far right side of the
tumor/muscle interface in the postcontrast image,
there is evidence of muscle invasion by the tumor,
not appreciated on the precontrast image.. 101

Spectra from phantoms containing
Glutamine/Glutamate, Creatine/Creatinine, Taurine,
and Choline. 111

H-1 spectrum from the skeletal muscle of a normal
volunteer with the metabolite peaks of interest
labeled. 112

Spectra from lactate phantom. The top two spectra
were obtained with different mixing times as
indicated. Subtraction of two spectra resulted in
preservation of the lactate signal and elimination
of the residual water signal as seen in the bottom
spectrum. 113

Lactate editing of normal skeletal muscle. Note
the elimination of most of the lipid and residual
water signal. No significant lactate is seen in
normal muscle spectra. . 114









Figure 4-11




Figure 4-12






Figure 4-13



Figure 4-14






Figure 4-15





Figure 4-16




Figure 4-17a-c


Lactate editing of malignant tumor. Residual
signal seen on the difference spectrum probably
represents both lactate and unsubtracted lipid.
116

MR images showing graphic selection of
spectroscopy voxel using the STEAM sequence.
a) Sagittal spin echo image of a volunteer's brain
with desired voxel outlined. b) Localized image
acquired with the STEAM sequence from area
outlined on sagittal image. 117

Graph demonstrating linearity of the water signal
with the voxel volume varying in size from about
one cc up to at least 27 cc. 118

Tumor spectrum simulation with combination of
spectra from phantoms. The tumor spectrum is on
the top, showing a large Cho peak in the middle.
The combination of spectra from phantoms is on the
bottom (20 mM Cho, 20 mM Cr, 40 mM Tau, 20 mM
Glu). 120

Spectra from the same tumor in different locations
showing the considerable spectral heterogeneity in
the tumor. The bottom spectrum is from viable
tumor. The top spectum is from necrosis. Note the
large lipid peak in the necrotic spectrum. 123

ROC curve for the accuracy of the Cho/Cr ratio in
distinguishing malignant from benign tissue. The
area under the curve was estimated to be
0.95600.0519. 127

Spectra from three of the tumors with followup. a)
Chondrosarcoma. (L.W.) Note the initial decrease
in size of the Cho peak followed by an increase.
129

b) Osteosarcoma. Note the slight decrease in the
Cho peak after three days of chemotherapy. At
three weeks and six weeks, the Cho signal has
disappeared. 130

c) Spectra from patient with MFH. The bottom
spectrum is preradiation therapy; the middle
spectrum is after four weeks of radiation therapy;
the top spectrum is after eight weeks of radiation
therapy. Note that the initially large Cho peak
(arrow) disappears by eight weeks. 131









Figure 4-18


Figure 4-19




Figure 5-1




Figure 5-2


Graphs of the change in the Cho/Cr ratio for four
patients and volume of tumor for three patients
with follow up.
a) Patient L.W., Chondrosarcoma; b) Patient C.M.,
Osteosarcoma. 132

c) Patient C.P., Malignant fibrous histiocytoma;
d) Patient L.D., Ewing's Sarcoma. 133

Graph showing the change of the normalized H-l
metabolites in the four tumor patients with follow
up post therapy. The concentration units are
arbitrary. 134

Set of spectra from one of the exercise patients,
pre- and postexercise test. The bottom spectrum is
preexercise, the remainder are postexercise in
order of time posttest. . 146

Metabolite levels and pH as a function of time for
two IC patients, pre- and posttraining. The top
is from a patient with moderate disease, the
bottom from a patient with severe disease.
a) PCr. 149


b) Pi.

c) PME.

d) pH.


Figure 5-3


Figure 5-4





Figure 5-5


150

151

152


Selected spectra from exercise testing of normal
subjects and patients at 85% of their MVC. From
left to right are the normal subjects, heart
transplants, CHF, and PVD patients. 160

Graph of metabolite concentrations and pH as a
function of time for a normal volunteer. Each
number represents a 32 sec time increment. The
arrows indicate the starting and the stopping time
for exercise. 161

Graph of metabolite concentrations and pH as a
function of time for a IC patient. Each number
represents a 32 sec time increment. The arrows
indicate the starting and the stopping time for
exercise. 162


xiii












KEY TO SYMBOLS AND ABBREVIATIONS


ADP Adenosine Diphosphate

ATP Adenosine Triphosphate

CHESS CHEmical Shift Selective

CHF Congestive Heart Failure

Cho Choline

CI Confidence Interval

CIS Cisplatin (cis-diamminedichloroplatinum)

Cr Creatine/Creatinine

CT Computed Tomography

DANTE Delays Alternating with Nutations for Tailored

Excitation

FID Free Induction Decay

FOV Field of View

Glu Glutamate/Glutamine

HCCTP hexachlorocyclotriphosphazene

IC Intermittent Claudication

ISIS Image-Selected In vivo Spectroscopy

Lac Lactate

MR Magnetic Resonance

MRI Magnetic Resonance Imaging

MRS Magnetic Resonance Spectroscopy

NMR Nuclear Magnetic Resonance

PCr Phosphocreatine

Pi Inorganic phosphate

PME Phosphomonoesters


xiv









PDE Phosphodiesters

ppm parts per million

PRESS Point RESolved Spectroscopy

PVD Peripheral Vascular Disease

ROC Receiver Operating Characteristics

RF Radio Frequency

SE Spin Echo

S/N Signal-to-Noise

SPARS SPAtially Resolved Spectroscopy

STEAM STimulated Echo Acquisition Mode

T Tesla

TE Echo Time

TM Mixing Time

TR Repetition Time

VOI Volume of Interest
















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

IN VIVO MAGNETIC RESONANCE SPECTROSCOPY
OF MUSCULOSKELETAL DISORDERS

By

James Ray Ballinger

August, 1994

Chairman: Katherine N. Scott
Major Department: Nuclear Engineering Sciences

P-31 MRS is used to do the following: detect chemotherapy-

resistant human osteosarcomas implanted into a nude mouse model;

determine anesthetic and temperature effects on spectra in the

nude mouse model; monitor the metabolic changes in skeletal muscle

that occur in peripheral vascular disease (PVD) patients with

claudication; and compare metabolic changes in skeletal muscle of

PVD and congestive heart failure (CHF) patients with age-matched

normal volunteers. Development of P-31 MRS in the therapy

monitoring of patients with musculoskeletal tumors is also

presented. Water suppressed, proton magnetic resonance

spectroscopy is developed for use in diagnosis and monitoring of

the therapy response of musculoskeletal neoplasms including

osteosarcoma.

Significant changes in the phosphocreatine to inorganic

phosphate ratio are found in treated and sensitive osteosarcomas

in mice before change in the size of the tumors. This ratio does


xvi









not change significantly in either untreated or resistant tumors.

Minimal direct anesthetic effects are seen on P-31 metabolites and

pH in implanted osteosarcoma in mice. A profound effect of

anesthesia is seen on the thermoregulatory ability of the mice.

The changes in pH with change in temperature match those described

in other animal models and humans.

Proton spectroscopy is successfully implemented in eleven

musculoskeletal patients and statistically significant differences

in the choline to creatine (Cho/Cr) ratio between malignant tumors

and necrosis, edema, benign tumors, and normal muscle are found.

Four of the patients with malignant tumors have follow-up studies.

Follow up studies show a statistically significant drop in Cho/Cr

ratio with treatment. In one patient, a change is seen within

three days of the start of chemotherapy.

Significant increase in the post-exercise rate of recovery of

the phosphorylated sugar peaks and pH is seen using P-31 MRS

following low-intensity training of PVD patients. Significant

differences in the rate of recovery of high energy metabolites and

pH in skeletal muscle are seen between normal volunteers and PVD,

CHF, and heart transplant patients.


xvii
















CHAPTER 1
INTRODUCTION


The objective of this dissertation was to develop and

implement magnetic resonance spectroscopy (MRS) techniques in the

diagnosis, treatment monitoring, and basic understanding of the

disease process of musculoskeletal disorders. The techniques used

were in vivo phosphorus and proton MRS applied to a mouse model

and to humans. A 2.0 Tesla small-bore and a 1.5 Tesla whole-body

magnet were used for these studies. Volume-selective and surface

coil techniques were used for localization. A treadmill and an in-

magnet ergometer were used for the exercise studies. Bone and soft

tissue tumors were examined as well as skeletal muscle in patients

with congestive heart failure, peripheral vascular disease, and

heart transplant.

The diagnosis and treatment monitoring of disease by

clinicians utilizes the history or complaints of the patient,

physical examination, and various tests or procedures. The tests

range from rather invasive and sometimes risky procedures such as

surgery and biopsy to relatively risk-free test such as body fluid

analyses, EKG's, x-rays, and MR imaging and spectroscopy. The

purpose of this dissertation is to explore several unique

applications of MR spectroscopic techniques in the diagnosis,

treatment monitoring, and basic understanding of musculoskeletal

disorders.











Magnetic resonance spectroscopy (MRS) of intact biological

tissues was first reported by two groups: Moon and Richards using

P-31 MRS to examine intact red blood cells in 1973 (1), and Hoult

et al. using P-31 MRS to examine excised leg muscle from the rat

in 1974 (2). Since then MRS has been applied to almost every organ

of the body including brain (3-7), heart (8-11), liver (8, 12-14),

kidney (15-17), prostate (18-20), and extremities (21-23). MRS is

useful for looking at disorders of metabolism, tumors and certain

inflammatory and ischemic diseases. Most of the work with in vivo

MRS in humans has been in the brain. Abnormalities have been seen,

sometimes with earlier detection than for any other diagnostic

procedure short of biopsy, in primary brain tumors (6, 24-31),

infections such as AIDS (32, 33), demyelinating disorders such as

multiple sclerosis (34), epilepsy (35), and stroke (36, 37).

Spectroscopic changes are documented in a variety of enzyme

deficiencies, mitochondrial abnormalities, dystrophies,

inflammatory myopathies, and thyroid disease. In muscle these

diseases include phosphofructokinase deficiency, amyloglucosidase

deficiency, Duchenne muscular dystrophy, Becker muscular

dystrophy, dermatomyositis, polymyositis, inclusion body myositis,

hypothyroidism, and congestive heart failure (CHF) (38-43).

The research presented in this dissertation deals with

application of MRS to disorders of the extremities, specifically,

the musculoskeletal system. These applications have been

relatively neglected except unlocalized P-31 MRS of mitochondrial

and enzyme abnormalities. Diseases of the musculoskeletal system

for which MRS may have some value include metabolic (including

ischemic) diseases and neoplasms. MRS may be useful for diagnosis,










treatment monitoring, or understanding of the basic mechanism of

these diseases.

Changes in muscle function and P-31 metabolism of patients

are reported with both peripheral vascular disease (44-46) and CHF

(47-49). P-31 MRS data are presented in this dissertation from a

unique treatment for peripheral vascular disease (PVD) patients

introduced a few years ago (44). This treatment was implemented

recently on PVD patients locally.

MRS of neoplastic disease involving the musculoskeletal

system has not been evaluated extensively. This is probably a

result of the rare occurrence of musculoskeletal tumors and the

demanding technical problems. For this dissertation, several types

of musculoskeletal tumors in humans and an osteosarcoma model in

mice are studied with P-31 and proton (H-l) MRS. I hope that this

technique will eventually allow detection of chemotherapy and

radiation therapy nonresponders in humans earlier than

conventional imaging procedures. This would result in reducing the

delay for surgery and reducing unnecessary chemotherapy and

radiation therapy costs. Our spectroscopy research group plans to

extend the animal research to evaluating new drugs for reversing

chemotherapy resistance. This will be applied to humans later if

successful in the animal model.

The effects of temperature and anesthesia on the P-31

metabolites and pH will be examined in mice. If present, these

effects could alter the results of the chemotherapy studies.

Research Hypotheses

The following are the hypotheses that will be tested in this

dissertation:











Mouse P-31 Tumor Studies

The presence or absence of statistically significant

differences in the rates of change of the

phosphocreatine/inorganic phosphate ratio (PCr/Pi) and

phosphomonoesters (PME) between a) untreated and treated mice; and

b) chemotherapy-treated sensitive and resistant mice will be

determined. This information will be used to calculate the

sensitivity and specificity for detecting the untreated mice and

the resistant mice.

Human P-31 Tumor Studies

The PCr/Pi ratio and the PME signal will be examined in

spontaneous musculoskeletal tumors in humans, using localized P-31

MRS, in hopes of using this information to detect tumors that fail

to respond to therapy.

Anesthetic and Temperature Dependence of P-31 Metabolites

The presence or absence of the following phenomena will be

determined:

1) Anesthesia with Innovar-Vet causes loss of thermoregulation in

nude mice at doses required for immobilization for MRS studies.

2) Intracellular pH in implanted osteosarcoma in the nude mouse

exhibits a temperature dependency similar to that reported in

other tissues and in other models, including humans.

3) Anesthesia effects may be seen in the pH and P-31 metabolites

of osteosarcoma.

Animal H-l Tumor Study

An osteosarcoma mouse model will be used to determine the

following:











1) Localized H-1 MR spectra, using water suppression techniques,

will be successfully acquired from osteosarcomas implanted in nude

mice.

2) Proton spectra from mouse tumors will show abnormalities in the

choline/creatine (Cho/Cr) ratio.

Human H-1 Tumor Study

The H-1 MRS of mice will be complemented with examination of

musculoskeletal tumors in humans to evaluate the following

hypotheses:

1) Localized H-1 MR spectra, using water suppression techniques,

will be successfully acquired from spontaneous musculoskeletal

tumors, including osteosarcomas, in humans.

2) Viable tumor tissue will be distinguished from normal muscle,

edematous tissue, and necrosis by H-1 MRS in humans.

3) The response to chemotherapy of musculoskeletal tumors in

humans will be detected and quantified by changes in the H-i

spectrum.

4) The localized H-1 spectral data from humans will be correlated

with magnetic resonance imaging (MRI), computed tomography (CT),

and histopathology of the tumors.

5) Changes in Cho/Cr ratio will be used to differentiate

chemotherapy non-responders from responders and will do so earlier

than with MRI or CT.

Exercise Studies

We will use P-31 MRS of leg muscle in conjunction with

exercise testing to test the following:











1) Improvement in the post-exercise recovery rates of phosphorus

metabolites and pH will be seen after low-intensity exercise

training.

2) Significant differences will be seen in the immediate post-

exercise metabolite levels and pH in the skeletal muscle of PVD

patients, CHF patients, heart transplant patients, and normal

volunteers.

3) Significant differences will be seen in post-exercise recovery

rate of phosphorus metabolites and pH in skeletal muscle of PVD

patients, CHF patients, heart transplant patients, and normal

volunteers following exercise.

Specific Objectives

Mouse P-31 Tumor Studies

Phosphorus MRS data will be collected from an osteosarcoma

mouse model using a 2 Tesla (T) MR spectrometer. Data will be

obtained and analyzed to show any statistically significant

differences in metabolite levels between untreated and treated

mice and between treated sensitive and resistant tumor mice.

Human P-31 Tumor Study

1) Existing localization techniques on a 1.5 T whole body magnetic

resonance (MR) scanner (Signa, General Electric Company,

Milwaukee) will be evaluated and (if necessary) modified to obtain

well-localized P-31 spectra. Chemical shift imaging (CSI)

techniques will be evaluated with phantoms, human volunteers, and

patients. Spectral data will be obtained with signal-to-noise and

spectral resolution adequate to show metabolite concentration

differences between normal muscle, treated tumors, and untreated

tumors.











2) Localized spectra of the tumor, necrosis, and surrounding

normal and edematous tissue will be compared to the MRI in the

humans.

Anesthetic and Temperature Dependence of P-31 Metabolites

1) Nude mice with implanted tumors will be anesthetized with

Innovar-Vet and their rectal and skin temperature measured over a

two hour period to detect any loss of thermoregulation.

2) Serial P-31 spectra will be obtained from anesthetized mice

that become hypothermic and from mice that are heated to maintain

normal body temperature over a two-hour period. Peak areas and pH

will be measured and correlated with the rectal temperature and

the time from injection of anesthesia.

Animal H-l Tumor Study

Surface coils and a Faraday shield will be used initially for

localization in mice in a 2 T spectrometer/imager. Water

suppression techniques will be evaluated with phantoms and mice to

obtain 200+ fold suppression of water. Spectral data will have

spectral resolution sufficient to resolve the Cho and Cr peaks

(-0.2 ppm).

Human H-l Study

1) Existing localization techniques on the whole-body MR scanner

will be evaluated and (if necessary) modified to obtain well-

localized H-l spectra. Various water suppression techniques will

be evaluated with phantoms, human volunteers, and patients to

obtain 200+ fold suppression of water. Shimming and adjustment of

voxel size will be done to obtain a spectral resolution sufficient

to resolve the Cho and Cr peaks (-0.2 ppm).










2) Localized spectra of the tumor, necrosis, and surrounding

normal and edematous tissue will be compared to the MRI in the

tumor patients.

3) The prebiopsy H-1 spectral data will be correlated with MRI and

CT, and the histopathology of the biopsy material in humans.

4) Changes in spectral data during and after therapy will be

documented. These changes will be compared with imaging data,

gross and microscopic histopathology, and tumor response to

therapy. Estimates of the ability of H-1 MRS to distinguish benign

from malignant tissues and therapy responders from nonresponders

will be made.

Exercise Studies

1) Recovery rates will be calculated for pH and P-31 metabolite

levels obtained in the whole body MR scanner following a treadmill

exercise stress test of PVD patients before and after low-

intensity exercise training. The data will be tested for any

significant changes from pre- to posttraining.

2) P-31 metabolite levels and pH will be measured from PVD

patients, CHF patients, heart transplant patients, and normal

volunteers in the whole body MR scanner during and following

exercise with an in-magnet ergometer. Absolute metabolite

concentrations and recovery rates will be calculated. Differences

in these parameters among the four groups of subjects will be

determined.

Collaborations

Part of the work presented in this dissertation was done in

collaboration with others. The treated and untreated mouse section

and the sensitive and resistant mouse section were a joint effort











of J.R. Ballinger, H. Kang, and C.A. Sweeney. Specific details of

each person's contribution are presented later. The sections on P-

31 spectroscopy of musculoskeletal tumors in humans and anesthetic

and temperature dependence of pH and P-31 metabolites in mice, and

the chapter on H-1 spectroscopy of musculoskeletal tumors was the

work of J.R. Ballinger alone. The spectroscopy results in the

chapter on P-31 spectroscopy of skeletal muscle in peripheral

vascular disease and congestive heart failure was the work of J.R.

Ballinger, with the ancillary exercise testing performed by C.

Stopka, W. Breshue, and colleagues. A. Velenik, a postdoctorate

fellow at the time, set up and performed the transfer of some

exercise data from the Siemens scanner to a workstation for

analysis.
















CHAPTER 2
TECHNIQUES


Localization Techniques


Radio Frequency Localization

In vivo MRS generally requires some degree of localization.

With implanted tumors in mice, it is desirable to obtain the

signal from the tumor only and not from underlying muscle. Muscle

and fat are found surrounding most musculoskeletal tumors in

humans. Muscle may cause signal contamination in both unlocalized

P-31 and H-l tumor spectra, and fat may cause undesirable lipid

contamination in H-l tumor spectra. Localization requirements are

less strenuous in P-31 MRS of calf muscle for metabolic studies.

Here too, potential problems exist without some degree of

localization. Different muscle groups are used to different

degrees in any given exercise (for example the soleus and the

gastrocnemius in the calf). These two muscle groups also have

different proportion of muscle fiber types (50). Both these

factors may introduce variability in spectral data.

Surface Coil Localization

The simplest localization technique is with a surface coil.

Localization with this technique relies on the limited extent of

the B, field to define the volume of interest (51). The distance

from the coil at which a 900 tip angle occurs (providing the

maximum signal) may be varied by adjusting the amplitude or length











of the radio frequency (RF) pulse. This technique has been used

for localizing implanted tumors in nude mice by various authors,

including ourselves (18, 52, 53). Double-tuned coils to P-31 and

H-l provide proton imaging and swimming capability along with

phosphorus spectroscopy (54). A Faraday shield around the base of

the tumor may be used to remove additional unwanted signal from

surrounding muscle and other tissues (55).

Selective Presaturation

Another localization technique involves selective saturation

of the spins outside the volume of interest (VOI). Excitation and

dephasing of the spins outside the VOI are immediately followed by

acquisition of a free induction decay (FID), spin echo (SE), or

stimulated echo from the tissue inside the VOI. This technique may

be combined with selective excitation localization techniques. The

VOI may be defined by saturating orthogonal slabs or cylindrical

volumes (56-58).

Selective Excitation

Several localization techniques use selective excitaticn and

detection of spins to detect the signal from within a VOI. One of

the original volume excitation localization techniques is VSE

(volume selective excitation) (59). More accurate localization

techniques with less power deposition are used more commonly now.

ISIS (Image-selected in vivo spectroscopy) localization (60) has

been used frequently with P-31 MRS and has been reported at least

once with H-l MRS of the human brain (61). This technique uses

three 180 inversion RF pulses with gradients to select orthogonal

slabs, defining the VOI at the intersection. This is followed by a

900 excitation pulse and signal acquisition. An eight-step phase











cycling routine is used to eliminate the residual signal from

outside the desired voxel. Additional spatial presaturation pulses

may be necessary to eliminate additional extraneous signal from

outside the VOI as the result of subtraction errors during the

ISIS experiment. A multivolume ISIS technique has been described

using multiple-line frequency selective pulses (62). For n

volumes, 2" experiments must be done. The number of experiments

required for a small number of voxels with in vivo spectroscopy

would not be a problem as we need about 256 averages for adequate

signal-to-noise.

Commonly used selective excitation localization techniques

for H-l MRS include STEAM (STimulated Echo Acquisition Mode) (63,

64), PRESS (Point Resolved Spectroscopy) (65, 66), and SPARS

(SPAtially Resolved Spectroscopy) (67, 68). STEAM selectively

excites three orthogonal planes with two 900 RF pulses and one 180

RF pulse followed by acquisition of a stimulated echo. A localized

image may be obtained to confirm proper positioning and size of

the VOI. This sequence collects only one-half of the original

signal but allows a short echo time (TE) (20 ms or less).

If more signal is necessary because of a small VOI or low

levels of metabolites, the entire signal can be collected as a

spin echo using the PRESS technique. PRESS consists of gradient

localized 90-180-1800 RF pulses. The SPARS technique consists of

three sets of 900-180-90 RF pulses applied in the presence of

gradients. SPARS localization results in significant contamination

from outside the VOI unless care is taken in adjusting the 1800

pulses. This problem is not seen with STEAM (22). SPARS has a

greater RF load on the patient than STEAM or PRESS (67). There is











a sacrifice of a longer echo time in the PRESS and particularly

the SPARS techniques resulting in T2 decay of the signal. This is

rarely a problem with H-l MRS because of the long T2 relaxation

times of observable metabolites (69-71).

Chemical Shift Imaging

Chemical Shift Imaging (CSI) involves the use of stepped

gradients to phase encode the spectra in one to three dimensions

(72, 73). CSI has one major advantage over single volume

localization methods. Spectra from multiple locations within the

VOI may be sampled simultaneously (60, 74, 75). Sampling multiple

areas within a tumor is desirable because of the heterogeneity of

many tumors. Spectra from adjacent normal tissue are sometimes

desirable.

There are two significant disadvantages of these methods.

First, 45% or more of the signal from a voxel comes from bleeding

from other voxels (76). This occurs because CSI uses a finite

number of phase encoding steps to generate a limited series of

sine and cosine functions to encode discrete and discontinuous

voxels. The bleeding can be reduced with the use of selective

Fourier transform localization (76, 77). With this technique, the

k-space data set is weighted by a function that maximizes the

signal from acquisitions with small gradients and minimizes the

signal from acquisitions with strong gradients. Maximum SNR per

unit time is accomplished with this technique by varying the

number of averages at each phase-encoded step. Similar results may

be obtained with post processing by multiplying the k-space data

matrix in the spatial directions by an apodizing function, such as

a Hanning, sine, Gaussian, or Fermi function. Use of a filter









14

causes a small drop in SNR, yielding approximately 2% less signal

than a CSI acquisition without a filter. The filter decreases the

contamination from outside the voxel to about 11% (76). The width

of the response function in each spatial dimension increases by

1.6 times that without weighing the data. To compensate for the

increase in width of the response function, the phase encoded

steps may be increased by 1.6 or the field of view (FOV) may be

decreased by the same factor. A modification of the selective

Fourier transform CSI experiment was developed using variable flip

angles at different phase encoding steps to reduce bleed and

improved SNR per unit time (78). The efficiency of these

techniques has been compared recently (79).


Water Suppression Techniques


The adult human is approximately 60% water by weight;

skeletal muscle is approximately 79% water (80). This results in a

concentration of water in skeletal muscle of about 44 M. The

concentration of metabolites of interest in H-l MRS is between 5

mM and 30 mM. This 1400+ fold difference in concentration results

in a dynamic range problem for the receiver and the analog-to-

digital converter of most spectroscopy systems and causes

difficulty in quantitating accurately nearby peaks. It is

therefore desirable to reduce or eliminate the water signal for

detection and accurate quantification of metabolites.

Desirable features of a solvent suppression technique include

uniform excitation of the nonsuppressed peaks of interest. Hore

has summarized other desirable features for solvent suppression

pulse sequences:













(a) insensitivity to Bo inhomogeneity and small
errors in the choice of transmitter frequency; (b)
wide band excitation, preferably on both sides of the
solvent; (c) insensitivity to pulse imperfections (B,
inhomogeneity, nonideal pulse shapes, off resonance
effects, phase shift errors); (d) only linear phase
correction required; (e) simple modification to
obtain a 180 pulse; and (f) easy to program and use.
(81) pp. 285-286

One of the first proposed methods for reducing the water

signal in H-l MRS was Redfield's long, weak 90 degree RF pulse

(82). The pulse results in a narrow bandwidth of excitation

centered off resonance from water and on the frequency of the

metabolite resonance of interest. This method is sensitive to

magnetic field inhomogeneity resulting in incomplete suppression

of the water peak, or nonuniform excitation of the desired

metabolite, or both. The length of the pulse results in T2

weighting to the signal and a rolling baseline artifact. Redfield

later introduced a "2-1-4 pulse" sequence that uses alternating

180 degree phase shifts of a constant amplitude RF pulse in a

timing pattern of 2-1-4-1-2 (83). This results in a sinusoidal

excitation spectrum where the water resonance is at a null point.

This sequence results in a 100-fold decrease in the water signal.

Disadvantages of the 2-1-4 sequence include nonuniform excitation

of resonances of different chemical shifts and incomplete water

suppression.

Plateau and Gueron introduced a sequence of strong pulses for

water suppression (84). These pulses are simpler to generate and

less sensitive to errors in pulse amplitude than Redfield's soft

pulses. This sequence consists of two pulses separated by a delay

T (900, T, -900, acquire) with the pulses centered on the water











frequency. In the rotating frame of reference, the water

magnetization vector is flipped on to the +x axis. After a period

T, the water vector has remained unchanged in position, while

other resonances have undergone precession in the x-y plane. The

second pulse returns the water vector to the z axis resulting in

no component in the x-y plane and therefore no signal. The off

resonant vectors will have a component of varying degree remaining

in the x-y plane resulting in a signal. The amplitude of the

signal from the other spins follows a sinusoidal pattern in the

frequency domain, with maximum signal from those spins resonating

at 1/(4T) Hz, 3/(4T) Hz, etc. relative to the frequency of

water. In a modification of this technique by Bleich and Wilde

(85) (900, T, +900, acquire) the RF pulse is centered Af Hz away

from the water resonance such that the peak of interest is located

Af/2 from water. The water vector is allowed to rotate 180 in time

T at which time the second pulse returns it to the z axis. The

peak of interest will remain in the x-y plane. These two hard-

pulse sequences result in a 300-fold decrease in the water signal.

Disadvantages of these sequences include the high sensitivity of

the water suppression to phase shifts of the RF pulses and to

amplitude balance in the RF channels (81).

Hore first described the binomial pulse sequences in 1983 in

two papers (81, 86). The binomial pulse sequences were developed

to have the following properties: 1) a broad null to accommodate

widening of the water peak due to magnetic field inhomogeneity and

to allow for minor errors in the setting of the transmitter

frequency; 2) a wide band of excitation in the remainder of the












spectrum of interest; 3) insensitivity to imperfections in the RF

pulse shape and amplitude (as may be caused by finite rise and

fall times, inhomogeneous RF fields from surface coils, and off

resonance effects due to the long pulse lengths needed by whole

body coils); and 4) a short sequence length to allow for T1

measurements if desired.

The Fourier transformations of the binomial pulse sequences

approximate the desirable features of the excitation spectrum at

small flip angles. Hore chose a sinusoidal function for an

excitation spectrum: S(w)=sinn(cT/2). The inverse Fourier

transformation of this function gives a series consisting of

equally spaced delta functions with alternating signs given by the

binomial coefficients (81). This series can be approximated by a

series of equally spaced short pulses whose amplitudes are given

by the binomial series, e.g., 1,2,1; 1,3,3,1; 1,4,6,4,1. The RF

pulses alternate 180' phase shifts. The longer sequences have

broader null regions and would therefore be more efficient in case

of a broad water peak. The disadvantage of the longer sequences is

that a large frequency dependent phase shift is introduced into

the spectrum (87). The symmetric sequences (1,1; 1,3,3,1) are

insensitive to minor errors in the flip angle compared to the

asymmetric sequences. This is because each pulse in a symmetric

sequence has an oppositely phased pulse of the same amplitude.

Signal errors resulting from imperfect 1800 phase shifts can be

corrected with phase cycling (88). An approximately 1000-fold

decrease in the water signal can be obtained with the most

commonly implemented binomial sequence: 1,3,3,1. Additional

sequences that may be used are 1,1,8,8,1,1 and 1,5,20,20,5,1 (87).











Water suppressed spectroscopy can also be performed using

presaturation with a narrow bandwidth, frequency selective RF

pulse, the so-called CHESS (CHEmical Shift Selective) technique

(63, 89-93). Frahm et al. describe a combination of two CHESS

pulses followed by STEAM localization (94). Greater suppression

can be obtained using three CHESS pulses (95, 96), a technique we

have available on our whole body scanner. Addition or substitution

of a binomial sequence (81, 86), or a DANTE (Delays Alternating

with Nutations for Tailored Excitation) sequence (97) are

alternatives. The water suppression from T2 decay with the long TE

PRESS localization technique can be supplemented with preceding

CHESS pulses or an inversion pulse (98).


Lactate Editing


Lactate editing is necessary in in vivo experiments for at

least two reasons. First, the peak position (1.33 ppm) overlcs

that of the lipid signals (1.10-1.48 ppm) and second, the lactate

signal intensity may be several times smaller than the lipid

signal.

Several spectroscopic techniques may be used for lactate

editing including homonuclear double-resonance difference (99,

100), double quantum coherence transfer (101-105), and zero

quantum coherence (106-109). A zero quantum technique for lactate

editing has been described using stimulated echo localization

(106). This sequence is the only technique available to us to use

on the whole-body imager at this time.

Lactate methyl protons are J-coupled to the adjacent

methylene protons, resulting in phase modulation that is a











function of the TE and the mixing time (TM). The lipid protons

that resonate at a similar frequency show only mild J-coupling and

do not exhibit significant phase modulation (110). By acquiring

two spectra with a different TM and subtracting, the lipid signal

can be removed, leaving only the lactate signal. The lactate peak

phase modulation alternates minima and maxima at TE = 1/2J where J

is the coupling constant for lactate and equals 7.35 Hz. At 1/J =

136 ms, the lactate peak modulates with the TM with a period of 1/

Af where f is the chemical shift between the methyl and methylene

peaks of lactate. At 1.5 T, f is equal to 178.5 Hz. Best results

are obtained by keeping the TE constant at 136 ms and varying the

TM.

The homonuclear double-resonance difference technique

requires two transmitters and subtraction of spectra. This works

well with in vitro and ex vivo work in spectrometers but most

whole-body imagers do not have two transmitters. Most of the

double quantum coherence transfer techniques result in a 50% loss

of the lactate signal and require phase cycling to help reduce the

water signal. Double quantum coherence transfer techniques have

been reported that require no phase cycling and might be more

desirable for in vivo experiments but have not yet been

implemented on whole body imagers (111, 112).

An alternative to zero and double quantum editing techniques

is to take advantage of the short T1 of lipids compared to lactate

by the use of a 180 inversion pulse to null the lipid signal. As a

result of using this technique, 20%-30% of the lactate signal is

lost. The inversion pulse has been used with a binomial spin echo

sequence to obtain both water and lipid suppression (113). This











inversion pulse has been combined with two CHESS pulses for both

water and lipid suppression (114) and is available now on our

whole body spectrometer.

The differences in T2 between lipids and lactate were used in

a mouse model to eliminate most of the lipid signal (58). This is

effective when only a small amount of lipid is present and

required a TE of at least 270 ms.


Quantitation of molar Concentrations of Metabolites In Vivo


Tofts and Wray have published a comprehensive review of

quantitation methods (115). They include a discussion of the

assumptions and problems of each technique. Buchli and Boesiger

have published recently an evaluation of the accuracy and

reproducibility of several techniques for quantitation of P-31

spectra (116).

Twelve factors to be considered in quantitating in vivo MR

spectroscopy follow:

1) The minimum detection limits for metabolites are about 0.5 mM

for P-31 and 0.1 mM for H-l.

2) "NMR invisibility" sometimes occurs when a metabolite is not

mobile enough (i.e., rapidly rotating or translating) to give a

narrow peak. Much of the choline and phosphomonoester, for

example, is bound in cell membranes as phospholipids and is not

visible in normal tissue.

3) Normal tissues and to a greater degree, tumors, are

heterogeneous. The heterogeneity may be macroscopic (fat vs.

muscle or tumor vs. necrosis) or microscopic (different tumor

subtypes or intracellular vs. extracellular metabolites).












4) The transmitter (B1) homogeneity and receiver sensitivity must

be considered. Surface coils have quite an inhomogeneous B1 field.

Volume coils are better in terms of B1 homogeneity.

5) Coil loading must be considered where using external standards

for reference.

6) Peak ratios avoid some transmitter, receiver and coil problems.

The problem with peak ratios is that changes in individual

metabolites are not determined; changes may occur in both the

numerator and the denominator. This is less of a problem in normal

tissues where certain metabolites have relatively constant

concentration (such as adenosine triphosphate (ATP) in muscle or

water in brain and muscle), unlike the situation found in tumors.

7) Tissue extracts have been used to help quantitate MRS data,

however, the concentrations depend on what technique is used and

how efficient the technique is in extracting all of the

metabolite.

8) External standards may be used for quantitation. These should

experience the same B, field and receiver sensitivity as the tissue

or corrections should be made for differences. When the external

standard is used during a separate experiment from the tissue,

coil loading must be considered.

9) Internal standards may be used as alluded to under (6) above.

Both ATP and internal water are frequently used. In tumors or

metabolic diseases, they may not be as constant as in normal

tissue, adding possible error and variability to results.

Exogenous standards given internally have been used primarily in

animals. Their invasive and possible toxic effects are drawbacks

in humans.











10) Concentration measurements require some sort of volume

determination unless the volume of the tissue and the standard are

the same.

11) Relaxation effects need to be considered when TR < -5 T1 for

the metabolites of interest or the standard. Also, if there is a

delay of acquisition as occurs with spin echoes and stimulated

echoes, T2 decay must be considered unless the TE << T2 for the

metabolites and standard. Variation of the water T1 and T2 times

has been found in implanted mammary adenocarcinoma in mice as a

function of the age of tumor (117).

12) Various methods have been used to measure the signal magnitude

in NMR spectra. These include manual measurements of peak height,

triangulated peak areas, and cutting out and weighing the paper

that the spectrum is drawn on. Automated methods include

integrating the spectrum between two points, fitting Lorentzian or

Gaussian shaped curves to the frequency-domain data, or fitting

damped sinusoids to the time-domain data (118, 119). The manual

methods are time consuming, and peak height and triangulation are

not very accurate unless the peak line width and shape remain the

same. Integration works well on well separated but not overlapping

peaks. Some authors claim that fitting of the time domain data is

more reproducible than fitting of the frequency domain data.

Problems with time domain fitting include the need for very good

SNR and the more computer intensive processing than is required

for fitting frequency data.

The significant factors to consider in measuring metabolite

concentrations from in vivo experiments using MRS have been











expressed in an equation by Bottomley and Hardy (120). The

equation is as follows:




S V F (T1l,a) E (T2, Td)
sv I VF (T1 ,a C)E (T2, Td)


Where:

[M] is the metabolite concentration

[s] is the standard concentration

subscripts m and s denote metabolite and standard, respectively

S is the NMR signal

V is the sample volume

4 is the detection coil sensitivity

F is a function accounting for T1 saturation effects

a is the flip angle

E is a function accounting for T2 decay occurring during the delay

time Td

Two specific methods proposed for quantitation of in vivo MR

spectra include the following: Michaelis et al. suggest the use of

control spectra obtained from standard solutions in separate

experiments but under identical experimental conditions including

identical coil loading (121). This paper and two other abstracts

report the use of the water signal present in the brain as a

control or calibration standard for the quantitation of proton

spectra (122, 123). This method was originally proposed by

Thulborn and Ackerman (124). The water concentration in the brain

is relatively constant (=75-85%) (122).











Recently a combination technique using both an internal

standard and an external standard was proposed for proton

spectroscopy (125). Alger et al. use tissue water as an internal

standard and a reference tube of distilled water as an external

standard. The metabolite signals are referenced to the internal

water signal in the same volume of tissue. This in turn is

referenced to the ratio of water signal intensities of the same

tissue volume and the reference standard obtained from a short TE

MRI image. The concentration of pure water in the reference

standard is known, and with estimates of the relaxation factors of

the metabolites, tissue water and external reference standard, the

metabolite concentration can be calculated. The metabolite

concentration would be computed by the following equation:




[M]= [W] *- -*-*C *
W N N
s p



Where:

[M] is the metabolite concentration

M is the metabolite NMR signal

W is the unsuppressed tissue water signal

N, is the number of acquisitions used to obtain the unsuppressed

water signal

N, is the number of acquisitions used to obtain the water

suppressed spectra

Np is the number of equivalent protons producing the metabolite's

signal











[W,] is the tissue water concentration derived from the equation

below:



W
[W]y=55Mk*--W*C1*C2*C5




Where:

55 M is the concentration of pure water in the standard

W, is the signal intensity of water in the region of interest from

the MRI

W, is the signal intensity of the pure water reference from the MRI

Cj.4 are the correction factors for T1 and T2 relaxation

C, is a correction factor for sensitivity differences in the

receiver coil for the standard and volume of interest.

This method is quite attractive for use in in vivo

experiments with humans. Images are obtained to measure volume

changes in the tumors or tissue of interest and to provide a

localization guide for gradient localization techniques. Little

additional time would be required in using this quantitation

technique.
















CHAPTER 3
P-31 SPECTROSCOPY OF HUMAN OSTEOSARCOMA IN A MOUSE MODEL
AND IN HUMANS


Extensive literature exists describing abnormal P-31 spectra

of animal and human tumors. There are fewer reports of alteration

of the P-31 spectra following therapy. For a concise review see

Negendank (126).


Review of the Literature


General Problem

Osteosarcoma is a relatively rare malignancy of bone

occurring predominantly in teenagers and young adults with an

annual incidence in the United States of approximately 2100 new

cases (127, 128). Bone tumors account for about 5% of all

childhood malignancies. Osteosarcomas comprise about 60% of

malignant childhood bone tumors (129). The tumor consists of

malignant osteoid-forming cells, i.e., those cells that form the

organic matrix in which bone ossification occurs (130). These

tumors occur primarily in the metaphyseal or growth regions of

long bones, especially near the knee in the distal femur and

proximal tibia. They have a propensity with time to extend beyond

the bone into the adjacent soft tissues and joints.

Less common forms of osteosarcoma occur in older adults

(usually > 50 yrs. old) associated with either a benign disease

called Paget's disease or with previous radiation therapy (130).












Earlier in this century, there were two groups of patients, that

we no longer see, that developed osteosarcomas. Patients with

tuberculosis of the spine were once treated with irradiation; and

watch-dial painters often placed their brushes containing radium-

226 doped paint in their mouths. Both groups had an increased

incidence of osteosarcoma (131). These less common forms have a

poorer prognosis than the primary osteosarcoma in children.

Osteosarcomas in children have not been directly associated with

any external or environmental factors, but are usually associated

with the rapid growth that occurs in this age group (129). In some

cases, osteosarcomas are associated with potentially inherited and

acquired genetic defects (132-135).

History of Treatment

Before 1972, the 5-year survival rate of osteosarcoma was

20%, amputation being the primary treatment method (136). Post-

operative adjuvant therapy was introduced by Jaffe in 1972,

reporting a 30% to 40% response rate (137). In 19--, Jaffe

introduced the idea of pre-operative or neoadjuvant chemotherapy

with a primary response rate of 60. (138). The 5-year survival

rate for osteosarcoma in humans from three recent randomized

trials, reviewed by Eiber and Rosen in 1989, was 70% (range 66- to

7".) (130). These three clinical studies used combination

chemotherapy including doxorubicin, methotrexate, and cisplatin.

The 30% mortality rate in this review article is from drug-

resistant tumors.

Surgical management of osteosarcoma following chemotherapy

currently includes amputation, disarticulation, and limb salvage

procedures. Limb salvage procedures involve the replacement of the










effected segment of bone with normal donor bone. This preserves

normal function of the patient's limb, unlike amputation and

disarticulation. This may be possible if tumor involvement of the

soft tissues and bone marrow is not extensive and if the adjacent

joint is not traversed.

Monitoring of Therapy Response with Imaging Techniques

Response to chemotherapy is followed with MRI and CT;

however, these modalities depend primarily on evaluating gross

anatomical features such as change in tumor size. Detection of a

significant response to chemotherapy with conventional clinical

and imaging techniques requires a 4-6 week period. This results in

a delay of either definitive surgery or change in the chemotherapy

regimen if the desired response fails to occur.

A significant diagnostic problem is distinguishing soft

tissue extension of tumor from surrounding edema and inflammation.

Edema was present in six out of 21 osteosarcomas reported in a

recent paper (139). Neither CT nor MRI (with or without

enhancement) can accurately detect the difference between tumor

and edema in osteosarcoma (139), chondrosarcoma (140), and soft-

tissue tumors (141) and inflammation in soft-tissue tumors 142).

A 90% necrosis of the tumor on histopathological examination

has a better prognostic value for patient survival than tumor

size, site, and classification (143). The survival rate was 91

for patients with greater than 90% necrosis, compared to 14%

survival for patients with less than 90% necrosis. Unfortunately,

current imaging techniques including standard MRI fail to

accurately determine the amount of necrosis in osteosarcomas and

other malignant musculoskeletal tumors (142, 144-146). Two recent










papers report the use of dynamic Gd-DTPA enhanced MRI to predict

the amount of necrosis following chemotherapy. The first paper was

able to distinguish >90% necrosis from <90% necrosis (144). The

second paper mapped six osteosarcomas into 11 to 56 regions each

(147). The fraction of regions with rapid signal intensity change

with time predicts the histopathological grade of the tumor, i.e.,

<50% necrosis, 50%-90% necrosis, >90% necrosis, and 100% necrosis.

Individual regions were occasionally false positive for tumor when

vascular fibrosis was present. Two recent abstracts discriminate

chemotherapy responders from nonresponders based on the rate on

contrast enhancement after a bolus injection of gadopentetate

dimeglumine (148, 149).

P-31 Magnetic Resonance Spectroscopy

P-31 spectra of tumors and malignant cells consist mainly of

peaks for phosphomonoester (PME), inorganic phosphate (Pi),

phosphodiester (PDE), phosphocreatine (PCr), nucleoside

triphosphate (Primarily ATP) and diphosphodiester (DPDE). The PME

peak contains various sugar phosphates, phosphoryl ethanolamine

(PE), and phosphoryl choline (PC). The latter two are precursors

of phospholipids and are produced by choline and ethanolamine

kinases (150). Phospholipids are important constituents of cell

membranes and organelles involved in the synthesis of proteins and

generation of energy (151). The PDE peak consists primarily of

glycerolphosphoryl choline (GPC) and glycerolphosphoryl

ethanolamine (GPE), which are membrane breakdown products (150).

Thus, the P-31 spectrum gives information on phospholipid

synthesis (PC and PE) and degradation (GPC & GPE), cell energetic

(PCr, ATP, & Pi), pH (Pi and PCr positions) (152) and











glycoprotein/glycolipid synthesis (DPDE). Actively growing cells

(as in malignant tumors) have high rates of energy consumption and

protein/membrane synthesis and breakdown, contributing to P-31

spectral abnormalities.

P-31 spectra of bone tumors, including osteosarcoma in

humans, show increased PME, PDE, and Pi, and a decrease in PCr

(153-160) relative to normal skeletal muscle. The concentration of

ATP may increase, decrease, or remain unchanged, depending on the

study. The pH of the tumors is also variable, depending on the

type and the stage of the tumor. An elevated pH is seen in two

studies (154, 155), a normal pH is seen in four studies (153, 156,

159, 161), both slightly elevated and normal values in one study

(160).

Treatment Response

P-31 MR spectral changes may be seen in human osteosarcoma

patients on the second day after beginning chemotherapy (153,. A

decrease in PME in human osteosarcomas as a response to

chemotherapy is reported (160, 162, 163). An increase in the

PCr/Pi ratio is reported, following treatment, in human

osteosarcoma (154, 160), soft tissue sarcomas (163), and in murine

osteosarcoma (164), mammary adenocarcinoma (165), RIF-1 tumors

(166), and 9L gliosarcoma (167).

Subtle changes may be present in osteosarcomas and other

tumors within one hour of the start of chemotherapy (154, 168).

The prognostic value of these changes is uncertain. Following

chemotherapy in humans, long term decrease in PDE is associated

with >90% necrosis (155). An increase in the PCr/Pi ratio is

correlated with a decreased transverse tumor diameter by Semmler









31

et al. (154). Ross et al. (153) observe a decrease in PME and ATP

relative to PCr. Pi showed biphasic behavior with respect to time

in his study, initially decreasing, then increasing. Wehrle et al.

(166) observe biphasic behavior of Pi with respect to the

chemotherapy dose in RIF-1 tumors implanted in mice. A decrease in

Pi relative to a-NTP nucleosidee triphosphate) is seen at moderate

doses of cyclophosphamide (150 mg/kg i.p., LDI0 or less). At high

doses (200 mg/kg, approximately LD., dose), they observe an

increase in Pi relative to a-NTP. Wehrle also observes a decrease

in PME and an alkaline shift in pH. Koutcher showed a significant

increase in PDE/PME in malignant fibrous histiocytomas that

responded to chemotherapy compared to a slight decrease or no

change with nonresponders (163). The responders also had a

significantly lower ?DE/PME before chemotherapy compared to

nonresponders. Redmond et al. showed a significant decrease in PME

in human osteosarcomas (160). A recent abstract found that the

PCr/Pi and PCr/ATP ratios discriminate between chemotherac-'

responders and nonresponders (169).

pH Changes

Alkaline shifts in tumor pH following treatment have been

seen for a variety of tumors including non-Hodgkin's lymphoma

(170). The pH in three osteosarcomas following chemotherapy failed

to change significantly (160).

Advantages of P-31 MRS

1) The high energy phosphates so important to tumor

metabolism, i.e., PCr and ATP, are readily seen at 1.5 T. ATP is

important as the immediate source of energy for most metabolic











processes. Decreased levels are found in tumors with reversal

following effective treatment. ATP is identified and quantified by

H-1 MRS in human skeletal muscle at 4.1 T (171) but not at 1.5 T.

Total creatine can be resolved with H-1 MRS, however PCr and Cr

cannot be separated at 1.5 T.

(2) The different phosphoesters, PME and PDE are easily

resolved with P-31 MRS. Although the PME and PDE peaks show

similar changes in implanted osteosarcoma in mice (158), with

decoupling techniques the various components of the PME peak (PE

and PC) can be resolved and appear to have individual significance

(172). The elevated PME in most malignant tumors is related to

increase in PC; however, elevated PE and decreased PC have been

reported in human colon cancer (173). The PE/PC ratio may be a

more sensitive indicator of malignancy and of tumor response to

therapy than PME alone (172).

(3) The P-31 spectrum has no solvent signal to interfere with

resolution of metabolite peaks, unlike the water signal seen with

in vivo H-l MRS.


Treated and Untreated Mice


Our goal is to develop better techniques to monitor the

chemotherapy response of tumors. In the present study, we used

phosphorus NMR spectroscopy to compare the changes in

phosphomonoester(PME) signal and the phosphocreatine/inorganic

phosphate (PCr/Pi) ratio from implanted human osteosarcoma in the

nude mouse with and without chemotherapy (treated and untreated).

We evaluated the accuracy of the change in PME and PCr/Pi in

detecting the untreated mouse. This is a logical precursor to











studying resistant and sensitive tumors in mice. In the latter

case we would wish to predict which tumors are resistant.

The study protocol, data acquisition and preliminary analysis

was performed by Haejin Kang, Ph.D. as part of his graduate work.

The study is included in this dissertation because extensive work

has been done with the raw data beyond that presented previously.

Specifically, the following work was done by or under my

direction: a) spectra were refit, b) statistical analysis was

performed on the PME changes with treatment, c) the possible

explanation of volume changes accounting solely for the PME

changes was addressed, and d) changes in PCr/Pi were investigated

in terms of predicting treatment response independent of volume

change. In addition, experiments were performed by myself to check

the reproducibility of metabolite measurements using repeated

spectral acquisitions on three mice.

Materials and Methods

Animal preparation

The nude, athymic mouse is a well characterized and accepted

model for the study of the properties of human tumor cells (155,

174-176). Human tumors studied in nude mice include small cell

carcinoma of the lung (155, 164, 174), mammary carcinoma (164),

ovarian carcinoma (164), neuroblastoma (177) and prostate

carcinoma (178). Human osteosarcoma in mice has not been studied

with MRS to my knowledge; however, the murine Dunn osteosarcoma

has been (164). Our experience has shown that human osteosarcomas,

unlike some other human tumors, grow relatively well in nude mice.

Our spontaneous regression rate of implanted osteosarcomas in nude

mice is 1.5% (n=180) with none occurring in study mice. The mice











that showed regression were from earlier preliminary studies.

These mice were approximately 14 weeks old at time of implantation

and were not irradiated. Nude mice older than 12 weeks have been

shown to gradually develop T-cell activity that is apparently the

cause for the observed tumor regression (179, 180). This tumor

cell line makes bone in the mouse, repeating its human origin. We

have also shown that the cell line responds to the same

chemotherapeutic agents as do native osteosarcomas in humans. The

kinetics of chemotherapeutic agents in nude mice has been studied

and compared to the kinetics in humans (177, 181-184).

Twenty-two female Balb-c mice, weighing between 25-30 grams,

were quarantined and acclimated to environmental conditions of

270C20 and 40-50 humidity for 5-7 days before implantation. The

mice were fed and watered ad libitum.

At 10 weeks of age, the mice were irradiated in a Cs'" gamma

irradiator (Gammacell 40, Atomic Energy of Canada Limited, Ottawa,

Canada) at a dose of approximately 500 Rads to suppress any early

T cell activity (185, 186). The next day, a suspension of 6 X 10

trypsinized cells from a standard tissue culture (187) of the

human osteosarcoma cell line 791T (Zoma Corp., Berkeley, CA) was

implanted subcutaneously over the gluteus maximus of the

anesthetized mice. The mice were anesthetized for both

implantation and spectroscopy. An intraperitoneal (i.p.) injection

of 0.04-0.05 ml of Innovar-Vet (fentanyl citrate 0.4 mg/ml and

droperidol 20 mg/ml, Pittman Moore, Washington Crossing, NJ),

diluted to 10% v/v with normal saline, was used for anesthesia.

This dose provides adequate anesthesia for one to two hours. The

mice were irradiated and implanted with tumor cell suspension by









35

Carol Sweeney, laboratory technician. The tissue culture work that

she did was under the direction of Byron Croker, M.D., Director of

Pathology at the VA Medical Center, Gainesville, FL.

Cisplatin (7 mg/kg, Bristol Laboratories, Evansville, IN),

dissolved in 0.9% sodium chloride, was administered in 11 mice via

a tail vein on day nine post implantation. This dose of cisplatin

is a dose that is pharmacokinetically equivalent to the clinical

dose in humans (Rational Dose) (177, 184). The tumors were

followed with MRS for at least three weeks post implantation.

Spectroscopy

Spectroscopy was performed on a Spectroscopy Imaging Systems

Corporation Model VIS 85/310 imaging spectrometer with a 310 mm

diameter horizontal bore Oxford Instruments magnet operating at 2

T (34.61 MHz for P-31). Haejin Kang Ph.D., built the RF coil, and

acquired the data. The spectrometer was operated from and the

spectra analyzed on a Sun 3/110 work station (Sun Microsystems,

Inc., Mountain View, CA). A home-made, 3-turn solenoid coil with

an internal diameter of 13 mm and a depth of 6 mm (volume of the

coil: 0.80 cc), double-tuned to H-l and P-31 (54), was positioned

over the tumor as shown in Fig. 3-1. A fenestrated Faraday shield

was positioned around the base of the tumor for further

localization by excluding signal from adjacent muscle (not shown)

(55). On histological evaluation, the skin of the nude mouse is

quite thin and frequently has no muscle associated with it. It

probably contributes little to the NMR signal. This is unlike the

rat where significant subcutaneous muscle is present. There is

probably a significant contribution to the signal from muscle











Y

























Figure 3-1 RF coil positioned over mouse tumor.



underlying the tumor that is not shielded by the Faraaay shield.

This is discussed later under the Results and Discussion section.

The magnet was shimmed on the water peak on each mouse to a

line width of between 0.2 and 0.4 ppm. P-31 spectroscopy was

performed with a non-selective 3-lobed, 12 Usec, 90, sinc-shaped

RF pulse followed 30 usec later by acquisition of the FID signal.

The 900 RF pulse power was set by maximizing the signal from the

tumor. The acquisition parameters were: 2000 acquisition points,

TR=1.5 sec, spectral width=2000 Hz, 1024 averages, and a 26 min

acquisition time. The total exam time was about 50 min.

Spectroscopy was performed one day before implantation, then twice

a week starting when the tumors were 0.23 cc 0.07 in volume (nine

days post implantation).











The volume of the tumors was initially calculated with the

formula for an ellipsoidal volume (n/6)*L*W*D from measurements

made with calipers where L, W, and D are the length, width, and

depth of the tumor respectively (156). Later, the tumor volume was

calculated as the average of the formulas for an ellipsoid and a

prolate spheroid (formed by rotation of an ellipse about its major

axis, L): (n/6)*L*W*D and (n/6)*L*W2. We have found that this

method of volume calculation is the best estimate of tumor volume.

The average of these two formulas had a better correlation

coefficient with the volume of water displaced by excised tumors

of different sizes (compared to the two formulas separately, as

shown in the Results and Discussion section). Imaging of the tumor

to determine size was tried but failed to give adequate signal-to-

noise (S/N) to distinguish the tumor, muscle interface in a 15 min

period. The filling factor for the coil during the acquisitions

immediately before and after chemotherapy ranged from 0.25 to

0.625. Reproducibility of spectra and metabolite signals was

checked by J.R. Ballinger by obtaining 6-7 sequential spectra each

in three additional mice with tumors (See Results and Discussion).

After 10 Hz line broadening of the FID and Fourier

transformation of the data, the spectra were fit using the Fitspec

software provided by SISCO. Zero and first order phase correction

were applied to the frequency domain data. Fitting of spectra was

performed initially by Carol Sweeney and Haejin Kang and later by

Carol Sweeney and J.R. Ballinger.











Analysis of data

Data were statistically analyzed by J.R. Ballinger. The PME

area and PCr/Pi ratio were used in statistical analysis. The

change in each was compared to the volume changes of the tumors.

An unpaired, two-tailed Student's t-test was used to test the

difference between the treated and untreated mice in the PME

change from the day after chemotherapy to the following session

(four days later). The change in the slope of the PCr/Pi curve

from pre-chemotherapy to the first study post-chemotherapy was

tested for significance in both the treated and untreated mice

with a paired, two-tailed Student's t-test. A receiver operating

characteristic (ROC) curve was used as an indicator of the

predictive value of the change in the slope of PCr/Pi in detecting

tumor treatment.

Results and Discussion

Results

Two treated mice died as the result of the anesthesia and a

third was sacrificed because of an eye infection. This compares

favorably with the rat mortality rate from moderate to high doses

of Innovar-Vet (6%-18%) even though the dose per kg in mice is

higher (188-190). Smaller amounts of Innovar-Vet were inadequate

in sedating the mice.

Using six to seven consecutive spectra, reproducibility of

the spectra and metabolite areas was checked in each of three mice

by calculating the per cent standard error:



(Standard)eviation) 100%
(MeanArea)











The range of the per cent standard error for PME was 7.8% to

14.1%, for Pi 9.0% to 11.1%, and for PCr 3.5% to 9.9%. The range

of the standard error for the PCr/Pi ratio was 8.6% to 12.1%.

Thus, the reproducibility of the spectral acquisition and peak

fitting appears quite reasonable.

We had a final number of 10 untreated mice and 9 treated

mice. Representative spectra from an untreated tumor at five days

and twelve days post implantation are shown in Figure 3-2. Figure

3-3 is a graph of the time course of the average tumor volume of

the treated and untreated mice. Comparison of the three techniques

of calculating the volume of the tumors was made by using a paired

t-test to test the significance of the difference between the

three formulas and the volume measured by water displacement of 36

excised tumors. The ellipsoidal volume formula (n/6)*L*W*D

underestimated the volume by an average of 0.32 cc (31% of the

actual volume), which was statistically significant (p<0.0001).

The prolate spheroid formula (n/6)*L*W2overestimated tumor volume

by an average of 0.21 cc (20%), which was statistically

significant (p=0.04). The mean of these two formulas

underestimated the tumor volume by 0.06 cc (5.6%), which was not

statistically significant (p=0.39).

A graph of the time course of the average amount of PME is

shown in Figure 3-4 for the treated and the untreated mice. PME

levels of the treated mice decrease from day 10 to day 14 post

implantation (day 1 to day 5 post chemotherapy).




























is 1S -5 -l. -1 -0 -s -30


IL

to5 -s


Figure 3-2


-to tS -1 -s r-n


Representative spectra from an untreated tumor at
five days (Bottom) and twenty-three days (top)
post implantation. The PME peak is labeled with a
long vertical arrow, the PCr peak with a short
horizontal arrow. Note the increase in size of the
PME peak and the decrease in size of the PCr peak
on day twenty-three.









Vol in cc


1.6

1.4 -

1.2

1

0.8 -

0.6 -

0.4

0.2


0


Figure 3-3


5 10 15 20 25 30
Day Post Implantation
Graph of the mean volume of treated and untreated
tumors as a function of time. Arrow marks :hemo day.


PME Area


110

102
94

86

78
70

62
54
46


30
0


Figure 3-4


-Treated -Untreated Std. Dev.


4 8 12 16 20 24 28
Day After Implantation
Graph of the mean PME of the treated and untreated
tumors as a function of time. Arrow marks chemo day.


-----i---;rC


--T












The changes in Figure 3-4 roughly mirror the changes in

volume of the tumors seen in Figure 3-3. PME levels of untreated

mice increased or remained unchanged during this period.

Figure 3-5 shows a graph of the PCr/Pi ratio for the treated

and untreated mice over time. The posttreatment slope of the

PCr/Pi curve for the treated mice shows a significant change

(increase) from the pretreatment slope (p=0.0249) with a 954

confidence interval of (+0.0669, +0.755). The posttreatment slope

of the PCr/Pi curve for the untreated mice does not show a

significant change in direction from the pretreatment slope

(p=0.4448, confidence interval: -0.253, +0.530). During the time

that the PCr/Pi slope of the treated tumors changes, the slope of

the volume curves does not change (Figure 3-3). Note that the

above statistical analysis was not testing directly for

differences between the change in PCr/Pi slopes between the

treated and untreated tumors. Testing the differences in slope for

the same mouse allows use of a paired t-test. The paired t-test,

allows use of a smaller number of subjects than an unpaired

test. The posttreatment slope we have used is from day 7 to day 10

postimplantation, or from two days before treatment to one day

post treatment. Ideally, we would want to measure the

posttreatment slope from the day of treatment onward. This is not

possible for technical reasons and because of the fragility of the

mice. An alternative is to project the pretreatment slope to the

day of treatment, and use that point as the initial point for the

posttreatment slope calculation. This procedure results in little

change in the results. The new p value is 0.0303 with an area




















PCr/Pi Ratio


0 5 10 15 20 25
Day Post Implantation


Figure 3-5


Graph of the mean PCr/Pi ratio for the treated and
untreated tumors as a function of time.











under the ROC curve equal to 0.7230.131 (see below for original

ROC area).

Sensitivity in this study is the number of detected untreated

mice divided by the actual number of untreated mice for a given

decision threshold for PCr/Pi slope change. The specificity is the

number of treated mice called treated divided by the total number

of treated mice. The complementary nature of sensitivity and

specificity can be represented with a ROC curve (191-194) as shown

in Fig. 3-6.

The sensitivity (true positive fraction) is plotted as a

function of 1-specificity (false positive fraction) for detecting

an untreated tumor at various levels of PCr/Pi slope change post

treatment. The points for our ROC curve were obtained from Charles

E. Metz's LABROC1 program that fits binomial curves to the PCr/Pi

slope data for the treated mice and from the untreated mice by the

maximum-likelihood algorithm (195, 196). A point on the ROC curve

is obtained by calculating the area under the binomial curves

located to the right of a particular threshold or PCr/Pi value.

The calculated areas correspond to the true positive fraction and

to the false positive fraction. This process is repeated for

several thresholds to obtain the entire curve. A straight diagonal

line from the bottom left corner to the top right corner would

indicate no predictive value of the test (area=0.5). The larger

the area under the curve, up to one, the greater the predictive

value or efficacy of the test. In this study we obtained an area

of 0.640.12. The area under a ROC curve is independent of the

position of the decision threshold, unlike accuracy, and may

therefore be a better measure of test performance (192).
















true positive rate


0.2!


0 0 0 0
0 0.1 0.2 0.3


0.4 0.5 0.6 0.7 0.8 0.9 1


false positive rate













Figure 3-6 ROC Curve for detecting untreated tumors with PCr/Pi
ratio. The true positive rate (sensitivity) is plotted
as a function of the false positive rate (1-
specificity).











Discussion

Our data show a difference in the change of the slope of the

PME curves between treated and untreated osteosarcomas implanted

into nude mice; however, this PME change occurs simultaneously

with the change of the volume of the tumors. The PME changes could

be related to differences in the amount of skeletal muscle being

volume-averaged from under the tumor rather than actual

concentration changes within the tumor. We are currently acquiring

data using a slice selective localization technique that will

eliminate contamination of our tumor spectrum by underlying

muscle.

The PCr/Pi slope changes (A(PCr/Pi)/At) that we see are

statistically significant and occur prior to changes in volume of

the tumor. There is overlap of the data from these two groups as

shown in the figures. We would want to have a high specificity for

detecting an untreated tumor (or in humans, a nonresponder to

chemotherapy) to avoid discontinuing a treatment that is wor-:ing.

Using a threshold of change in the slope of PCr/Pi eqcal to -0.63,

we obtained an 86% specificity and a 29- sensitivity. Using a

decision threshold of 0.40, we can obtain a 60% specificity and a

60% sensitivity. These results will undoubtedly improve with

better localization as we will have less of a dilutional effect

from skeletal muscle contamination.











Sensitive and Resistant Mice


Materials and Methods

The objective of this experiment was to determine if changes

in PCr/Pi and PME can be used to predict lack of tumor response to

chemotherapy in a murine model of a chemotherapy-resistant human

osteosarcoma.

Introduction

In a previous section, the use of phosphorus NMR spectroscopy

(MRS) to compare the changes in phosphocreatine to inorganic

phosphate ratio (PCr/Pi) and phosphomonoester (PME) signal from

implanted human osteosarcoma in the nude mouse, with and without

chemotherapy was described. The accuracy of using post treatment

PCr/Pi changes to detect untreated mice was evaluated, and a

statistically significant difference between the treated and

untreated mice was found. It is hypothesized that a similar

difference in the change in PCr/Pi between treated cisplatin-

resistant and cisplatin-sensitive human osteosarcomas implanted in

the nude mouse will be found.

Tumor cell line characteristics and preparation

The 791T osteosarcoma cell line (Zoma Corp., Berkeley, CA) is

a high-grade, non-metastatic tumor cell in humans. The doubling

time in vitro is twelve and one half hours. The growth-rate in

nude mice depends on the number of cells implanted: 0.4 cc tumors

in 40-45 days for 4 x 107 cells and 6-8 days for 8 x 107 cells. In

the nude mouse, the tumors can grow to ca. 6 cc, but become very

necrotic and ulcerative at this stage. In our study, the mice are

sacrificed before the tumors reach 2 cc in size.











Carol Sweeney, Biological Laboratory Technician, cloned the

791T-E10 (E10) cell line from 791T by using limiting dilution

cloning, to produce a uniform cell population. Both cell lines

grow as monolayers in Dulbecco's modified Eagle medium (DMEM)

supplemented with 10% bovine calf serum, 15 mM HEPES buffer, and 2

mM L-glutamine. Drug resistant sublines were derived from E10

cells by intermittent treatment of the cells with increasing

concentrations of cisplatin (E10-CIS subline). Cisplatin (cis-

diamminedichloroplatinum) was obtained as Platinol (Bristol-Myers

U.S. Pharmaceutical and Nutritional Group, Evansville, IN). Cells

were seeded in T25 flasks at a concentration of 1 x 10" to 3 x 10'

cells/ml; twenty-four hours later they were treated with

concentrations of cisplatin in DMEM ranging from 10-22 ug/mi for

one hour. The cells were rinsed twice with DMEM and then cultured

in fresh DMEM until their numbers reached that of the untreated

cells. The concentration of cisplatin chosen for later treatments

depended on the length of time it took for cells to recover from

the previous treatment.

A modified version of the chromium (Cr-51) release assay,

developed by Brunner et al. (197), was used by Carol Sweeney and

Jim Scott, Senior Chemist, to measure the short term (<24 hours)

cytotoxicity of cisplatin (0 mg/ml 100 mg/ml) to E10 drug-

sensitive and E10-CIS drug-resistant cell lines.

A four-day growth assay was then used to compare the level of

resistance of E10 (drug-sensitive) to E10-CIS (drug-resistant)

cells beyond 24 hours. 1.8 x 105 cells were seeded on 60 x 15 mm

dishes; twenty-four hours later, the cells were treated with

varying concentrations of cisplatin in DMEM (0-32 pg/ml). Each











concentration in each cell line and controls was run in

triplicate. After one hour, cells were washed once with DMEM, then

fresh DMEM was added to all plates. Four days later, cells were

trypsinized and counted using a Coulter counter. The concentration

of the drug that resulted in 50% cell inhibition after four days

(ICso) was determined from a dose response curve.

E10 and E10-CIS were frozen in multiple vials for use in

mouse studies so that all mice would receive cells of

approximately the same "age" (passage number) and level of

resistance. For further information on the cell culture technique

see the paper by Blommaert et al. (187).

Animal model

Twenty-one female BALB/c-nu/nu mice, weighing between 20-30

grams, were used for this study. Mice were treated similarly to

those used in the treated and untreated study.

When the mice were 7-10 weeks old, and one day after

irradiation, a suspension of 6 x 10' E10 or E10-CIS cells in DMEM

was injected subcutaneously over the gluteus maximus by Carol

Sweeney. The mice were anesthetized with an i.p. injection of

Innovar (fentanyl citrate 0.05 mg/ml and droperidol 2.5 mg/ml,

Pittman Moore, Washington Crossing, NJ) for implantation (0.05 ml)

and for MRS (0.35-0.38 ml) of the tumor site. Innovar is

approximately 1/10 the concentration of Innovar-Vet, the later

being used for the first mouse study. The change in anesthetic

was only because of the lack of availability of Innovar-Vet at the

time of the second study. The dose of Innovar was adjusted

appropriately to account for differences in concentration.











Each mouse was administered a single injection of 7 mg/kg

cisplatin in a tail vein on approximately day 12 post implantation

by Carol Sweeney, when the tumor volume reached about 1/2 cc

(0.502 cc 0.190 cc). The tumors were followed with MRS for at

least three weeks post implantation.

Spectroscopy

Spectroscopy data collection and fitting were performed by

Haejin Kang, with assistance from Carol Sweeney and J.R.

Ballinger, as described in the previous section. In addition, an

external standard containing 0.018 cc of 1 M

hexachlorocyclotriphosphazene (HCCTP) in benzene was located above

the tumor, in the plane of the top of the coil. MRS was performed

starting six days before chemotherapy, then twice weekly for three

weeks.

Analysis of data

Analysis of the data was performed by J.R. Ballinger. The

peak areas were normalized to the external standard to eliminate

variability in the data from changes in instrumental sensitivity.

The PCr/Pi data were found not to be in a normal distribution by a

Shapiro-Wilks test; therefore, a Student's t-test could not be

used. Instead, a two-tailed Wilcoxon Rank Sum Test was used to

test the difference between the sensitive and resistant mice in

the PCr/Pi change from the day before chemotherapy to two days

following chemotherapy (three days later). The PME data were

tested with both a two-tailed Student's t-test and the Wilcoxon

Rank Sum Test. A ROC curve was drawn to show the predictive value

of the change in the PCr/Pi slope in detecting the resistant

tumors.










Results and Discussion

Tumor cell line

The chromium release assay showed spontaneous release of

chromium (2 hours to 24 hours = 5% to 35%, respectively, of total

activity) did not differ significantly with drug concentration or

cell line. This suggests that the cytotoxic effects of cisplatin

at the concentrations tested occurs beyond 24 hours.

Drug sensitivities of the parent and resistant cell lines

were compared based on their respective IC,0's determined from the

four-day growth assay that was run six times for each cell line

(See Table 3-1). The E10-CIS subline was five times more resistant

than the E10 parent cell line.


Table 3-1
Results of Four-day Growth Assay

Osteosarcoma Cell E-10 E-10 CIS
Line

Cisplatin 4.490.26 pg/ml 22.566.57 pg/ml
Concentration

Note. Concentration of cisplatin at which the viable cell number
is 50% inhibited compared to the cell number of untreated controls
(IC0o).


Animal study

Two of the cisplatin-sensitive tumors failed to grow well

before chemotherapy and were excluded from the analysis. One mouse

was sacrificed due to eye infection and weight loss. We had a

final number of eight cisplatin-resistant and ten cisplatin-

sensitive mice.

Spectra were similar in appearance and quality to those shown

in the experiment comparing treated and untreated mice. The mean








52

volume of the sensitive and resistant tumors as a function of time

is shown in Figure 3-7. The graph of the mean PCr/Pi vs time

(Figure 3-8) shows a divergence in the slopes after treatment of

the tumors. This change in the slope of the PCr/Pi occurs before

significant changes in tumor volume. The two-tailed, Wilcoxon Rank

Sum Test shows statistical significance at the a=0.05 level in the

difference between the post treatment slopes for the sensitive and

the resistant tumors. The change in the slope of the sensitive

tumor from pre- to post-chemotherapy was tested and had a p value

between 0.05 and 0.1. The change in the slope of the resistant

tumor pre- to post-chemotherapy was not significant with a p value

greater than 0.1. Calculating the post treatment slope from the

projected value of PCr/Pi at time of treatment resulted in no

significant change in these results.

The graph of the mean PME levels vs time (Figure 3-9) showed

a transient and statistically insignificant decrease in the slope

from day one to day five post chemotherapy in the sensitive

tumors, paralleling roughly the change in tumor volume. The

resistant tumors showed no change in the slope after chemotherapy.

The ability of the data to detect the resistant tumor may be

expressed with a ROC curve as discussed earlier (191-194). The

points on the ROC curve (Figure 3-10) were generated using Charles

Metz's LABROC1 program that fits a binomial curve to the PCr/Pi

data (195, 196). The area under the curve reflects the predictive

value of the test with an area of 0.5 indicating no value and an

area of one indicating a "perfect" test. Using the post-

chemotherapy slopes for the ROC analysis, we obtained an area of

0.66740.1249.
















Volume (cc)


1


0.8


0.6


0.4


0.2


-*Sensitive Resistant I Std. Dev.


0'
-10 -5 0 5 10 15
Day Pre and Post Chemotherapy















Figure 3-7 Mean change in volume of sensitive and resistant
tumors as a function of time. All of the mice
received chemotherapy on day zero.

















PCr/Pi Ratio

-- Sensitive Resistant i Std. Dev.


-5 0 5 10


Day Pre and Post Chemotherapy


Figure 3-8


Mean change in PCr/Pi for sensitive and resistant
mice as a function of time. All of the mice
received chemotherapy on day 0. Change in the
PCr/Pi ratio is seen as early as day 1 post
chemotherapy. This change occurs before any
change in the volume of the tumors (compare with
Figure 3-9).


9


-10


I ,
















PME


I Std. Dev. Sensitive Resistant


1 0 -.. .




0
-10 -5 0 5 10 1


Day Pre and Post Chemotherapy


Figure 3-9 Graph of the mean PME one standard deviation vs
time for sensitive and resistant mice. All of the
mice received chemotherapy on day zero.















true positive rate


0.8 "


0.2!


0.1 0.2 0.3


0.4 0.5 0.6 0.7 0.8 0.9 1


false positive rate


Figure 3-10


ROC Curve for detecting resistant tumors with
the change in slope of the PCr/Pi ratio
postchemotherapy. The true positive rate
(sensitivity) is plotted as a function of the
false positive rate (1-specificity).


U '


0











Discussion

Our data show a slight but statistically significant

difference in the change in PCr/Pi after treatment between

sensitive and resistant osteosarcomas implanted into nude mice.

The PCr/Pi change occurs before change in volume of the tumor. We

would want to have a high specificity for detecting a resistant or

nonresponder to chemotherapy to avoid discontinuing a treatment

that is working. Using the post treatment PCr/Pi slopes to detect

the resistant tumor, we can select a threshold that will give us a

70% specificity and a 54% sensitivity. This is similar to our data

with treated and untreated sensitive tumors. The minor difference

between these two studies is to be expected since there was only a

5-fold difference in drug sensitivity of the two cell lines in

vitro. We are currently acquiring data using a slice selective

localization technique that may improve our ability to distinguish

between sensitive and resistant tumors by diminishing

contamination of our tumor spectra by underlying muscle.


P-31 Spectroscopy of Musculoskeletal Tumors in Humans


Materials and Methods

Extensive developmental work has been done with phantoms and

volunteers in vivo MRS. Because of the small number of

osteosarcoma patients being referred to us and the low signal-to-

noise seen in the one tumor patient that had P-31 MRS, we chose to

delay acquisition of P-31 spectra from additional patients until

after construction of a double-tuned, quadrature H-1/P-31 coil and

possibly implementation of proton decoupling of spectra. We were

obtaining H-l spectra from patients that had significantly better











signal-to-noise than the P-31 spectra, and so chose to concentrate

on H-l MRS.

Instrumentation

The 1.5 T GE Signa whole-body magnet was used for phantoms,

volunteers, and the patient (Figure 3-11). A half-saddle coil

double-tuned to H-l and P-31 was used for most of the study (54).

The coil was constructed by Jim Scott.

Localization development

In the background section, the considerable heterogeneity of

tumors was noted. Therefore, we chose to use 2D-CSI for our

experiments. This technique gives better spatial resolution than

1D-CSI and is not as time consuming as 3D-CSI. Two problems had to

be overcome, the intervoxel bleeding that is inherently present

with use of the technique, and registration of the CSI spectra

with the image and volume of interest.

The numerical weighting, post processing technique of CSI

bleed reduction was used (76) as described in the Techniques

chapter. Before FFT, the k-space domain data matrix was multiplied

in both directions by the following Hanning equation:




W(g)=0.4 6[l+cos(ni) ]
G



Where g is the value of the phase-encoding gradient, whose values

range from -G to +G. The amount of signal bleeding before and

after use of the Hanning function was evaluated qualitatively and

quantitatively.

































































Figure 3-11 Photograph of the 1.5 T General Electric Signa whole-
body imager.











CSI and image registration was necessary to know from what

part of the tumor or adjacent soft tissues a given spectrum came

from. The SA/GE software allows the grid of spectra to be

superimposed onto an image for this purpose. Because of

differences in the way the images and the CSI data are acquired,

flip and translation corrections were needed for an exact match. A

left-to-right and top-to-bottom flip were required to match the

image and CSI matrix in terms of orientation. A half CSI voxel

shift was necessary in both spatial directions because the image

localizing and CSI localizing gradients have different symmetry

about the x and y axes. The flip and translation corrections were

checked with phantoms and volunteers to ensure proper positioning.

Occasionally, a CSI voxel did not fall exactly on the area of

interest. The voxel can be shifted by using a first order phase

correction on the k-space data in the gradient (not time)

dimensions. Susan Kohler, Ph.D., formerly of GE Medical Systems,

was kind enough to write a macro for us to do this data

manipulation. This macro was also tested on phantoms and

volunteers.

Relaxation measurements

T1 relaxation measurements were made on volunteers since we

were using a TR on the order of one T1 time of some of our

metabolites. Partial saturation experiments were used, varying the

TR from 1.6 sec to 11 secs. Data from these experiments were fit

with a monoexponential curve using Statistica software (StatSoft,

Tulsa, OK).












Shimming

Whole volume swimming on the water peak with H-1 MRS using a

hard pulse was initially used in phantoms and in volunteers. The

homogeneity obtained was only fair, so a slice selective swimming

technique was subsequently used.

Spectral analysis

The CSI experiments use phase encoding gradients turned on

after the acquisition 90 RF pulse is applied. In our case, this

results in a loss of eight dwell periods (2 ms) of data from the

FID. As a result, a large first order phase correction is

necessary, resulting in a prominent baseline roll.

This is corrected using the since deconvolution technique

mentioned in the Techniques chapter. When we first started, this

technique had not been automated. I took eight point sine waves

with relative frequencies matching those of the major metabolites,

Fourier transformed them and added them to the distorted P-31

spectrum. The relative magnitudes of the added functions matched

approximately the peak heights of the P-31 metabolites. The

absolute magnitudes had to be adjusted empirically. Recently, the

since deconvolution technique has been automated and included in

the SA/GE software.

Reproducibility and linearity

The reproducibility of the signal measurements from phantoms

and from volunteers was tested over the short term (-2 hours) and

a long term (2-3 months).

The linearity of the signal to the voxel size was checked

with phantoms using the CSI technique. We expected to use











different size voxels depending on the size of the tumor and

location.

Normalization of data

An external standard containing 2 cc of 0.5 M or 1 M HCCTP

was positioned at the base of the coil and used to normalize the

data and control for instrumental variation. The CSI volume

included the standard in one of the voxels. When the standard

overlapped voxels, the shifting technique mentioned above was used

to position the signal from the standard in the middle of one

voxel. Because we used a half-saddle coil to acquire our data, the

sensitivity of the coil to different voxels varied. This

positional dependence was corrected for by a factor arrived from

measurements of the signal from a large phantom containing a

solution of PO, in water.

Tumor patient experiment

The single patient that we did was a 75-year old woman with a

pleomorphic sarcoma, most consistent with a dedifferentiated

liposarcoma, in her leg. We used the double-tuned half-saddle coil

described above with the external standard. A localizing T1

weighted SE pulse sequence was used to obtain images through the

tumor. At the widest part of the tumor, swimming was performed on

an axial slice, 3.0 cm in thickness. Two-dimension CSI was then

performed at this slice with a 16 x 16 matrix, 30 cm FOV, 2 sec

TR, and two signal averages resulting in an acquisition time of

about 17 min. The voxel volume is equal to

thickness[(FOV/matrix)1.6]2 where 1.6 is a correction factor for

use of the Hanning filter. In this patient's case, this gives us a

volume of 27 cc. There were six voxels containing mostly tumor.












Spectra from these voxels were analyzed in the fashion described

above. PME was normalized to the external standard with a

correction for coil sensitivity differences. The PCr/Pi ratio was

calculated.

Results and Conclusions

Localization development

The amount of intervoxel bleed was reduced significantly with

the use of a Hanning window. The signal from voxels not containing

the phantom represented 31% of the total signal from the standard

without the Hanning window and 4.9% with the Hanning window. This

was best demonstrated with a small, -20 cc bottle of 500 mM HCCTP

solution.

Good registration of the image and CSI data was obtained. The

CSI shifting macro worked well. Figure 3-12 shows a photograph of

the CSI data superimposed on an image of a human tumor that we

examined with P-31 MRS.

Relaxation measurements

Table 3-2 shows the mean and standard deviation of

measurements of T1 time for the P-31 metabolites obtained from

calf skeletal muscle of three volunteers:


Table 3-2
P-31 T1 Relaxation Times in Normal Skeletal Muscle


This compares to values from the literature for human

skeletal muscle of 4.0170.90 sec for Pi, 5.520.20 sec for PCr




























































Figure 3-12


Photograph of CSI data superimposed on an axial MR
image of tumor. The arrow points to the femur.
Surrounding the black cortex of the femur on the
left, right and top, is the malignant tumor
appearing grey. The white areas below the femur
are mostly fat. The small light grey object at the
bottom of the photograph is the external standard.











and 4.310.60 sec for beta ATP. The values are in good agreement

except for the PCr where we get a smaller T1 time. This may be

related in part to the fact we went up to a TR of only eleven

seconds for the measurements. Eleven seconds was the maximum TR

that we could obtain on the whole-body scanner at the time. We

expect the tumor T1 times to be different from that of normal

muscle. In human brain, tumors have Tl's that are about 50% larger

than that of normal brain (5). The difference between T1 times in

human musculoskeletal tumors has not yet been studied

systematically because of the duration of the data acquisition

required.

Shimming

With whole volume swimming on the water peak, we obtained

line widths of up to 35 Hz from volunteers using the half-saddle

coil. With slice swimming the line width ranged from 10-20 Hz in

volunteers.

Spectral analysis

The since deconvolution of the P-31 spectra was successful in

obtaining a flat baseline. Manually, this process took about half

an hour per spectrum. The automated method was very quick, though

manual adjustments were occasionally necessary, especially with

noisy spectra. Figure 3-13 shows a pre and post baseline-corrected

spectrum from skeletal muscle of a volunteer performed with 2D-

CSI.

Reproducibility and linearity

Short term reproducibility was tested by obtaining six

spectra over a two-hour period from a phantom containing 500 mM
























































976.6 976.8 977


Figure 3-13 Pre (bottom) and post (top) baseline-corrected P-31
spectra from the skeletal muscle of a normal
volunteer.











phosphate solution. The standard error was 5.0%, probably

reflecting instrumental instability.

Long term reproducibility was tested on three human

volunteers by obtaining three 2D-CSI acquisitions of calf skeletal

muscle spaced over a three-month period. This data will reflect

both instrumental variations and physiological variations. The

variation was similar for each volunteer and the mean standard

error is given in the Table 3-3 for normalized and unnormalized

data. The normalized data was obtained by dividing the metabolite

signal by the external reference signal, and multiplying by a

correction factor for the B, differences between the two.


Table 3-3
Variation in Normalized and Unnormalized P-31 Metabolites
over Three Months in Skeletal Muscle of Volunteers

PME Pi PDE PCr y ATP a ATP 3 ATP Mean

Unnormal. 54% 40% 594 59% 45% 45: 45k 50-

Normalized 33% 13% 36% 9.3% 16% 18% 20% 21%


The normalized spectral data had less variation over nime

than did the unnormalized data. The smaller Pi peak and the broad

PME, PDE, and ATP peaks had greater variation than the tall narrow

peak of PCr. The noise contributes to a proportionately larger

fraction of the metabolite signal of small, broad peaks, than to

larger, narrow peaks. Also, fitting of small or overlapping peaks

is more variable.

Linearity was checked by measuring the signal from a

phosphate phantom using the CSI technique. Figure 3-14 shows a

graph of the PO, signal plotted as a function of voxel volume. The















Signal (area)


0 5 10 15 20


25 30 35 40 45 50


Volume in cc











Figure 3-14 Graph demonstrating linearity of the P04 signal
plotted as a function of the CSI voxel volume.











rity is reasonable considering that we are using a phase

ing technique. The volumes that we use on our patients is

ally on the order of 16 to 30 cc. The regression line fitted

e data intersects the positive y-axis just above the origin

r than through it. This finding may be a result of the

nce of noise in the spectra (estimated to be about 7% of the

signal), as well as the 4.9% bleed from other voxels causing

f the points to be shifted in a positive direction.

patient experiment

The spectra from the single malignant tumor examined (Figure

had worst signal-to-noise than the spectra from our normal

teers (Compare with figure 3-13). Other in vivo

roscopists have reported a similar marked decrease in P-31

olite signal from an osteosarcoma in a human (Personal

nication from William T. Evanochko, Ph.D. from the University

abama at Birmingham). The etiology of this signal abnormality

known. The tumor spectra varied in relative metabolite signal

different locations to a greater extent than spectra from

i1 muscle. This finding was expected due to tumor

*ogeneity. An example of a fitted tumor spectrum is seen in

-e 3-16.

A significant difference was seen between the volunteers and

tumor in the following: PME, PCr/Pi, pH, and beta ATP. The

ige and standard deviation of these values over six voxels

lining tumor in this patient and the values from four

iteers are given in Table 3-4.















































500 0 -500


Figure 3-15 P-31 spectrum from malignant tumor in a patient.
Note lower signal to noise when compared to a
normal volunteer (Figure 3-13).
























































Figure 3-16 An example of a fitted tumor spectrum from the
same patient as Figure 3-15. The spectrum is
from a different part of the tumor than that in
the Figure 3-15.











Table 3-4
Normalized Metabolite Signal for Tumor and Skeletal Muscle

PME PCr/Pi pH p ATP

Tumor 23.198.53 3.221.20 6.990.10 31.416.18

Volunteer 10.694.80 6.270.58 7.070.04 41.919.74
muscle


A significant difference in the PME levels(p=0.0107), PCr/Pi

(p=0.0002), pH (p=0.0442), and P ATP (p=0.0498) is seen between

the tumor and normal muscle. Statistically significant differences

were not seen in the PCr, Pi, and ADP levels because of large

variances. It is premature to draw any conclusions from this one

patient, except that the elevated PME and depressed PCr/Pi are

consistent with the mouse tumor work we have done.


Anesthetic and Temperature Dependence of pH and P-31 Metabolites
in Implanted Human Osteosarcoma in Nude Mice


Introduction and Review of the Literature

Innovar-Vet (fentanyl 0.4 mg/ml and droperidol 20 mg/ml,

Pittman Moore, Washington Crossing, NJ) is used successfully in

this laboratory to sedate the mice for both tumor implantation and

for NMR studies. Other anesthetic agents used in mice include

pentobarbital, gaseous anesthetics, and ketamine in combination

with various sedatives (198, 199). Authors have reported the

presence of time dependent temperature and anesthetic effects of

pentobarbital and ketamine on P-31 MRS of various tumors including

KHT and RIF-1 fibrosarcomas. The anesthetic effects included

initial suppression of high energy phosphate levels (PCr and ATP)

and are more marked with the use of pentobarbital compared to












ketamine. Reductions of PCr and ATP and elevation of Pi in early

spectra of RIF-1 tumors in mice following ketamine anesthesia were

observed (166, 200). Similar effects in mice have been seen in

RIF-1 tumors (198) and in C3H murine fibrosarcoma and spontaneous

mammary carcinoma (201) with sodium pentobarbital anesthesia. A

significant upward trend was seen in the pH during recovery from

pentobarbital anesthesia in RIF-1 tumors in mice (202). This

effect was inconsistently seen in spontaneous fibrosarcomas and

mammary carcinomas in mice by other authors (201). Innovar has

been shown to cause a 9% increase in proton T2 values in the brain

and in proton T1 and T2 of rat kidney (203). No significant change

was seen in the relaxation times of muscle, liver, fat, and other

organs. A mixture of fentanyl and fluanisome has been used to

anesthetize mice with KHT and RIF-1 tumors without change in P-31

metabolite concentrations or pH (204).

The physiological changes associated with Innovar-Vet have

been studied extensively in rats and compared with other

anesthetics and analgesics pentobarbitall, ketamine-xylazine, and

ketamine-diazepam) (188, 189, 205). Innovar-Vet is not a true

anesthetic but a combination of a potent analgesic and a major

tranquilizer. At low doses Innovar has a low mortality rate, long

duration of effect (108 min.), a moderate decrease in mean

arterial blood pressure (18%), and the least effect on blood pH,

pO, and thermoregulation. At high doses of Innovar there is an 18%

mortality rate, longer duration of effect (170 min), a 37? dr:p in

blood pressure, and a marked drop in both blood pH and pO, (189).

We wanted to determine whether anesthesia with Innovar-Vet

would have any effects on temperature regulation or on tumor









74

metabolite levels in nude mice during NMR studies. Our hypotheses

are as follows:

1) That anesthesia with Innovar-Vet causes loss of

thermoregulation in nude mice at doses required for immobilization

for NMR studies.

2) That a significant difference between the rectal temperature

and the skin temperature of mice will be found.

3) That pH in implanted osteosarcoma in the nude mouse exhibits

similar temperature dependency as that reported in skeletal muscle

of the rat and frog and in heart and brain of humans.

4) That anesthesia effects may be seen in the pH and P-31

metabolites of osteosarcoma.

Methods and Materials

Animal preparation

Twelve female Balb/c nude mice, weighing between 25-30 grams,

were quarantined and acclimated to environmental conditions of

271 C and 40-50% humidity for 5-7 days before tumor cell

implantation. Laminar flow ventilation and individual air-filter

covers over the cages help to maintain an aseptic environment. The

mice were provided food and water ad libitum.

A suspension of 6 x 107 cells of the human osteosarcoma cell

line 791T (Zoma Corp., Berkeley, CA) was implanted subcutaneously

over the gluteus maximus of anesthetized, seven week old mice by

Carol Sweeney. The tumors were studied after approximately two

weeks. The size of the tumors at the time of spectroscopy,

measured with calipers and calculated as the volume of an

ellipsoid (pi/6)*L*H*W, was 0.5160.155 cc.











Anesthesia

The mice were anesthetized with an intraperitoneal injection

of 0.04-0.05 ml of Innovar-Vet that was diluted to 10% with normal

saline (7 mg/kg). This dose provides adequate anesthesia for one

to two hours.

NMR experiments

In the first set of experiments, six mice were injected with

anesthetic and positioned in the magnet with no external heating.

The temperature in the bore of the magnet was 20.4 3.0 'C. In the

second set of experiments, another six mice were anesthetized,

positioned in the magnet, and warmed with 39 C air circulated

through the bore. The mice had a preanesthetic rectal temperature

of 36.60.5C. For both groups, spectra were obtained at 15 min

intervals starting approximately 30 minutes after injection of the

mouse. The 30 minutes were used for sedation, positioning the

mouse and swimming. Spectra were obtained for two to three hours

after injection with Innovar-Vet.

A third set of experiments was performed with four mice to

determine if single mice, placed in the bore of the relatively

cold magnet, dropped their temperature without the use of

anesthetic. The temperature of the mice was measured at time zero,

at 30 min, and at 60 min. The mice were removed from the magnet

for about two minutes to measure their rectal temperature at the

30 min and 60 min time points.

The rectal temperature of the mouse was monitored with a

fiberoptic probe connected to a Fluoroptic Thermometry System

(Luxtron 750), covered with a modified 5 french teflon angiography

dilator (Cook, Bloomington, IN). The probe was inserted between












1.5 and 2.0 cm into the rectum. The temperature was measured just

before sedation, at the beginning and end of each acquisition, and

every 5-10 minutes throughout the spectroscopy session. Besides

rectal temperature, skin temperature overlying the tumor was

measured with a second probe on three heated and three unheated

mice. To measure skin temperature, the tip of the probe was placed

so that it slightly indented the skin of the mouse overlying the

tumor. The probe was then taped to the coil support. The

preanesthetic tumor temperature was 35.20.36 C.

Spectroscopy was performed on a Spectroscopy Imaging Systems

Corporation Model VIS 85/310 imaging spectrometer with a 310-mm

diameter horizontal bore Oxford Instruments magnet operating at 2T

(34.61 MHz for P-31). The spectrometer was operated from and the

spectra analyzed on a Sun 3/110 work station (Sun Microsystems,

Inc., Mountain View, CA). A home-made, balanced-matched, 2-turn

solenoid coil double-tuned to H-l and P-31 (54), with an internal

diameter of 12 mm (constructed by Haejin Kang, Ph.D.) was

positioned over the tumor with a fenestrated Faraday shield

positioned around the base of the tumor to exclude signal from

adjacent muscle (55). The magnet was shimmed on the water peak on

each mouse to a line width of 0.2-0.4 ppm. P-31 spectroscopy was

performed with a non-selective 12 psec, 90' three-lobed sinc-shaped

RF pulse followed 30 psecs later by acquisition of the signal. The

900 RF pulse power was set by maximizing the signal from the tumor.

The acquisition parameters were: number of points=2000, TR=2.5

sec, 256 averages, 11 min acquisition time.

Each FID was apodized with 3 Hz line broadening, zero-filled

to 4000 points and Fourier transformed. Spectra were fit using the












Fitspec software provided by SISCO following zero-order phase

correction. pH was calculated from the shift of the Pi peak from

the PCr peak using the method of Kost. This method accounts for

the temperature at time of measurement (206). For purposes of pH

calculation, an average of the temperature at start and end of the

spectrum was used. For measurement of peak areas, each FID was

apodized to obtain 10 Hz line broadening to improve the signal-to-

noise prior to fitting.

Statistical analysis

In the unheated mice, the pH was plotted as a function of

the temperature and fitted with a least-squares linear regression

line. In the heated mice, the pH was plotted as a function of the

acquisition number and therefore as a function of time. The slope

for each group of mice was tested with a t-test to determine if it

was significantly different from zero (i.e., that there was a

significant dependence on either temperature or time). The Pi,

PCr, and ATP peak areas were tested as a function of time in a

similar fashion as the pH in the heated mice. In addition, a

paired t-test was performed on the first and last peak area for

each metabolite for both the heated and unheated mice. A paired t-

test was performed on the temperature data from the unanesthetized

mice comparing the values at time zero with those at 30 mins and

60 mins.

Results

A marked loss of thermoregulatory ability of all anesthetized

mice was seen. Unheated mice experienced a marked drop in

temperature (Figure 3-17). One mouse showed a spontaneous reversal

of the hypothermia. The unanesthetized, unheated mice showed no















Temperature (degree C)


Anesthetized -- Unanesthetized Mice


,x

=


0:00 0:30 1:00 1:30 2:00 2:30 3:00

Time post Anesthesia (hours)


Figure 3-17


Rectal temperature of unheated anesthetized and
unheated, unanesthetized mice vs time post
anesthesia. One anesthetized mouse showed a
spontaneous increase in temperature after about
1.25 hours.











significant decrease temperature at 30 mins (p=0.742, Confidence

Interval (CI): (-2.06, 1.73) and at 60 mins (p=0.449, CI: (-4.30,

2.77))(Figure 3-17). The heated mice showed a fairly constant

temperature after about 30 mins postinjection of anesthesia. The

slight drop in temperature seen before this time (Figure 3-18)

occurred while the mice and the coil were being positioned. A heat

lamp was used during a part of the initial 30 mins to minimize the

heat loss by the mice. The rectal and skin temperatures paralleled

each other in both the heated and unheated mice. The mean,

standard deviation, and range in temperature difference between

the rectal and skin temperatures for the unheated and the heated

mice are shown in Table 3-5. The difference between the skin and

rectal temperatures was significantly different from zero

(p=0.0053) though this difference was small (1.32C, range 0.4-

3.2). The purpose of measuring both the rectal and the skin

temperature was to estimate the accuracy of the rectal temperature

in predicting the temperature of the tumor itself, where the

spectra are obtained. We expected that the temperature of the

tumor to be between that of the rectum and skin, since the rectal

temperature is close to the core, or highest temperature in the

mouse and the skin temperature is the coolest part of the mouse.

Since the rectal and skin temperatures were so similar, it is

concluded that the rectal temperature is a reasonable estimate of

the tumor temperature for purposes of pH measurement.















Temperature (degree C)


0:00 0:30 1:00 1:30 2:00

Time post Anesthesia (hours)


Figure 3-18


Rectal temperature of the heated, anesthetized mice
vs time postanesthesia. The slight dip in temperature
before 30 mins occurred while the mice and the coil
were being positioned.











Table 3-5
Difference Between Rectal and Skin Temperature in Mice


Mouse Mean Difference Standard Range of
in oC Deviation of Differences
Difference
Unheated 1 2.56 0.44 1.5-3.0
Unheated 2 1.72 0.24 0.4-1.3
Unheated 3 0.81 0.18 0.5-1.0
Heated 1 0.91 0.30 0.4-1.3
Heated 2 1.47 0.95 0.4-3.2

Heated 3 1.43 0.77 0.6-2.9


An insignificant increase in the pH during recovery from

anesthesia was seen in the heated mice (Figure 3-19). A

significant correlation (p=0.0021) between the pH and temperature

was seen in the unheated mice (Figure 3-20). The regression lines

fitted to the data project to a pH between 6.9 and 7.2 at 363, like

the heated mice, except for one mouse. The reason for this

exception is unknown; however, the rectal temperature measurement

is highly dependent on the position of the fiberoptic temperature

probe in the rectum. Movement of the probe during the experiment

might explain this finding. Alternatively, the pH of that

particular mouse's tumor might really be 6.5 as regression of the

points would suggest. The mice showed an average change in pH of

-0.0250.011 Units/oC (95% CI: (-0.03598, -0.01391)).

Analysis of the peak areas showed no significant change over

time in the PCr, Pi, and ATP peak areas (Figure 3-21).

Discussion

Over three years, we have used 86 mice for three to twelve

spectroscopy acquisitions each with eight anesthetic fatalities












































" 4.


N -


.4.


1 2 3 4 5 6 7 8 9 10

Experiment #


Figure 3-19


Graph of the pH of the six heated mice vs
experiment number. Note lack of significant
change over time (experiment number).


7.6 h -


7.4 -


7.2 -


6.8
0
















pH
7.6_
6 Unheated Mice Mouse 3




7.4





7.2-





7





6.8 j
23 24 25 26 27 28 29 30 31 32 33 34 35 36

Temperature (degree C)







Figure 3-20 Graph of the pH of the six unheated,
anesthetized mice vs rectal temperature.
Regression lines have been fitted to the data of
all but one of the mice. See the text for an
explanation.