Linkage analysis and development of congenic strains for the genetic dissection of insulin dependent diabetes in the NOD...


Material Information

Linkage analysis and development of congenic strains for the genetic dissection of insulin dependent diabetes in the NOD mouse
Physical Description:
ix, 112 leaves : ill. ; 29 cm.
Yui, Mary An-yuan
Publication Date:


Subjects / Keywords:
Linkage (Genetics)   ( mesh )
Diabetes Mellitus, Type I -- genetics   ( mesh )
Diabetes Mellitus, Type I -- veterinary   ( mesh )
Mice, Inbred NOD   ( mesh )
Mice, Inbred Strains   ( mesh )
Crosses, Genetic   ( mesh )
Mice, Congenic   ( mesh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1995.
Includes bibliographical references (leaves 96-111).
Statement of Responsibility:
by Mary An-yuan Yui.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
oclc - 50179932
System ID:

Full Text







This dissertation is dedicated
to my daughter
Julia Meihua Longmate.


I would especially like to thank the chairman of my committee,

Dr. Ward Wakeland, for his unflagging support and encouragement and

for many valuable discussions. I am also grateful to my committee

members, Drs. Noel Maclaren, Bill Winter, Maureen Goodenow, Mark

Atkinson, and Mary Brown, for their guidance in various phases of

my research.

I also thank, Drs. Ivan Cheng and K. Muralidharan, for

initiating this project and helpful advice. Kye Chesnut, George

Perrin, and Bertha Moreno-Altamirano were especially helpful with

the congenic mouse project. The mouse breeding would have been much

more difficult without the excellent mouse care provided by Fred

Grant and Roberto Luchetta. I also appreciate help with histology

from Drs. Winter and Lauwers, and access to a microscope provided

by Dr. Khan.

I have enjoyed my association with other members of the

Wakeland lab, especially Laurence Morel, Karen Jackson, Kye Chesnut,

Joe Wu, and Rick McIndoe and thank them for help and encouragement.

Thanks go to Christy Myrick for proofreading. Jayoo Gokhale


provided invaluable help with histology and photography. She and

Maria Elena Bottazzi provided support, friendship and great meals!

I want to thank my parents for their encouragement for my

endeavors through the years. Lastly, I am grateful to my husband,

Jeff Longmate, for his help with this research, including many

statistical analyses and interesting discussions, and for his love

and support throughout the years.


ACKNOWLEDGEMENTS ................. .iii

ABSTRACT ... .. ............... .vii


1 INTRODUCTION ................ .. 1

Insulin Dependent (Type I) Diabetes (IDD) 1
The Non-obese Diabetic (NOD) Mouse . 2
Characterization of Infiltrating Cells in NOD
Insulitis . . 4
T Cells are Required for IDD in NOD Mice 5
Immunoregulation and Thl vs. Th2 responses 8
The Role of Macrophages in NOD IDD . 10
Immunostimulation and IDD ... .. .12
Bone Marrow Transfer Studies and Pancreatic
Vulnerability .. . 12
The Role of the NOD Thymus in IDD ... .14
IDD Autoantigens . .. 15
Adhesion Molecules and Leukocyte Homing to the
Pancreas . . .. 19
The Role of the Major Histocompatibility Complex (MHC)
in IDD . ....... .21
Analysis of the Polygenic Inheritance of IDD
Susceptibility .... ... 25
Rationale for this study ... .. .. .. 27


Introduction .................. .32
Materials and Methods .... .. 35
Results . . .. 38
Discussion ........... .. .... 47


Introduction .................. 65
Methods .................. .. 69
Results and Discussion ... . 70


Introduction .... . 78
Materials and Methods .... . 81
Results and Discussion ... ...... 82

5 SUMMARY . . 92


BIOGRAPHICAL SKETCH ................ ... .112

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



Mary An-yuan Yui

August, 1995

Chairperson: Edward K. Wakeland, Ph.D.
Major Department: Pathology and Laboratory Medicine

Insulin dependent (Type 1) diabetes (IDD) is the result of

chronic autoimmune destruction of the insulin-producing pancreatic

beta cells in the islets of Langerhans, via a process termed

insulitis. Although some key immunological events in IDD

pathogenesis are known, the genetic factors involved in the disease

are not well-understood. IDD in humans and in a spontaneous animal

model, the nonobese diabetic (NOD) mouse, is inherited as a

multifactorial polygenic trait. Genetic analysis of IDD in the NOD

mouse will help in our understanding of the inheritance of complex

traits and will facilitate identification of the genes involved in


Microsatellite markers on all nineteen autosomal chromosomes

were used to analyze the inheritance of diabetes in an experimental

mouse cross: (NOD X C57BL/6)F1 X NOD. Six linkage groups,

containing at least eight susceptibility alleles, were found to

correlate with diabetes and/or insulitis. Inheritance was found to

be genotypically heterogeneous and the penetrance of IDD increased

with increasing numbers of susceptibility alleles, consistent with

a threshold liability trait.

Congenic mouse strains were developed to genetically dissect

the effects of individual susceptibility alleles. Six NOD-derived

genetic intervals containing IDD susceptibility alleles were bred

onto a resistant, C57BL/6 (B6), background by six successive

backcrosses, followed by intercross to establish the homozygous

congenic strains. At six months of age several of the congenic

strains exhibited some pancreatic periductal/perivascular (PD/PV)

infiltrates when examined histologically. B6.NODc6 did not have

pancreatic infiltrates. B6.NODc3 congenic mice, which have at least

two IDD susceptibility alleles, showed the largest and most frequent

PD/PV infiltrates, with a low penetrance of peri-islet infiltrates.

B6.NODcl and cl7 mice exhibited some PD/PV infiltrates, while

B6.NODcll mice had only very small infiltrates. These congenic mice

can recreate the earliest events in diabetogenesis and will be

useful to identify the roles of each allele in disease, allow fine-

mapping and cloning of the genes involved, and can be combined into


polycongenic mice to reproduce the stages of insulitis and



Insulin Dependent (Tvoe I) Diabetes

Insulin dependent (Type I) diabetes (IDD) is an autoimmune

disease caused by destruction of insulin-producing pancreatic beta

cells in the islets of Langerhans. Progressive loss of islet beta

cell mass by autoreactive immune cells culminates in insufficient

insulin production, resulting in frank diabetes characterized by

hyperglycemia, ketonuria, polyuria, glycosuria, polydypsia and

weight loss. The inability to regulate blood glucose levels

eventually results in death unless insulin is administered. In

addition, even with daily administration of insulin, many secondary

vascular and other effects result in increased morbidity and

decreased life span for those with this disease.

IDD in both humans and rodent models is a multifactorial

disease which results from both genetic and environmental factors.

Evidence for genetic factors includes a concordance rate of

approximately 33% among monozygotic twins [1,2], with the risk of

disease for an identical twin of an individual with IDD more than

80 times the risk for the general population (reviewed by Atkinson

and Maclaren [3]). Risk for first degree relatives is also high.

The large discordance between monozygotic twins suggests that the

environment (e.g., diet, infectious agents) plays a very important

role as well as genetics.

IDD is inherited as a complex polygenic trait, that is, a

number of genes contribute to IDD susceptibility. The strongest

susceptilibity genes for IDD are in the major histocompatibility

complex (MHC) class II, HLA-DR and -DQ, region of human chromosome

6 (reviewed by Nepom and Erlich [4]). Genetic analyses have shown

that up to 20 independent regions of the genome are associated with

increased susceptibility to IDD in humans [5]. Different

susceptibility loci are also associated with different study

populations making IDD a genetically heterogeneous disease in man.

These characteristics make identification of factors contributing

to disease, both genetic and environmental, difficult to study in


The Non-obese Diabetic (NOD) Mouse

The NOD mouse is a spontaneous animal model which closely

mimics human IDD. The mouse was derived from a single diabetic

female that spontaneously arose from a noninbred cataract-prone

mouse line which was established as a homozygous strain by over 20

generations of inbreeding [6]. NOD females display a high

penetrance for diabetes (typically 70-90% by 30 weeks of age), while

males usually have a lower penetrance of disease. As with humans,

inheritance is multifactorial, with both genetic and environmental

factors influencing susceptibility. In addition, inheritance is

polygenic, and, at least in part, recessive, as F1 hybrids between

NOD and diabetes-resistant mice do not display insulitis or diabetes

(reviewed by Leiter [7]).

Environmental factors can profoundly influence the incidence

of disease. Food and caging conditions can dramatically affect the

penetrance of disease, from as much as 20-100% between separate

colonies of genetically identical animals (reviewed by Leiter et al.

[8], and Pozzilli et al. [91). Interestingly, IDD susceptibility

is highest in colonies maintained in specific pathogen-free

conditions, while experimental infections decrease IDD

susceptibility. Possible reasons for this will be discussed later.

This mouse model has proven to be extremely useful in the

study of the pathogenesis and inheritance of IDD. Experimental

manipulations such as preventative therapies, cell transfers,

genetic crosses, insertion of transgenes, experimental infections,

hormone manipulations, etc., which are not possible on humans, can

be performed on these animals. Several therapies currently being

used clinically or undergoing human clinical trials were first

established in the NOD mouse [10].


Characterization of Infiltrating Cells in NOD Insulitis

Mononuclear cells are first seen as perivascular/periductal

infiltrates as early as 3-4 weeks of age in the NOD mouse [11-13].

By histological examination, Fujita et al. [14] found that the

infiltration initiated on the side of an islet near periductal

connective tissue with the interlobular exocrine duct, arterioles,

venules, capillaries and small lymphatics. The infiltrates then

surrounded the islets in a thick ring (peri-islet infiltrates). The

peri-insular capsule, consisting of basement membrane, collagen and

fibroblasts, then disappeared and massive cell infiltration in the

islets was commonly seen by 5-6 weeks [6,14]. These infiltrates

were mostly small lymphocytes and reticulocytes. This infiltration,

termed insulitis by von Meyenburg [15], resulted in loss of insulin

containing beta cell mass while glucagon-positive alpha cells and

somatostatin-positive delta cells remained intact. Electron

microscopy revealed beta cell death near infiltrating cells.

Finally, after disappearance of the beta cells, the infiltrating

lymphocytes also disappeared leaving only delta and alpha cells.

Diabetes ensued after about 13 weeks of age when a large proportion

of the islets had been destroyed.

Immunohistological studies show that macrophages and B- and T-

cells are all present in the pancreatic infiltrates. T-cells are

a major component of the infiltrates [11,13,16]. Both CD4 (T

helper) and CD8(T cytotoxic) positive T-cells were found to be

present. Miyazaki et al. [13] and Shimizu et al. [16] found that the

T lymphocytes were localized closest to the islets and proposed that

they were responsible for beta cell destruction. B-cells, though

abundant, were observed more peripherally [13] and islet cell

specific antibodies were either not detected [13,16] or detected

late in the disease process, at 12-18 weeks [11] More recently

however, Shieh et al. found that a small percentage of NOD beta

cells had surface bound autoantibodies prior to detectable insulitis

[17]. They and Tisch et al. [18] also detected serum autoantibodies

very early.

Using flow cytometric analysis of mononuclear cells from

collagenase-dispersed isolated islets, Jarpe et al. found CD8-

positive T-cells and MHC class II-positive macrophages at 5-6 weeks

of age [19]. At 8-10 weeks a brief wave of CD4-positive T-cells was

observed which was then followed by B-lymphocytes and CD8-positive

T-cells. They proposed that early beta cell damage could be a

result of antibody-dependent cellular cytotoxicity (ADCC) and later

damage by cytotoxic CD8 positive T-cells.

T Cells are Reauired for IDD in NOD Mice

A number of experiments have demonstrated that IDD in the NOD

mouse is primarily a T-cell-mediated disease. Neonatal thymectomy,

which prevents T-cell development, prevents diabetes in the NOD

mouse [20]. In addition, athymic nude mice, NODnu""/ do not develop

insulitis or diabetes [21]. Wicker et al. [22] demonstrated that

diabetes could be transferred into young NOD mice by adoptive

transfer of splenocytes from diabetic NOD mice. Using this system,

Bendelac et al. [23] and Miller et al. [24] showed that CD4-positive

and CD8-positive T-cells were required for disease transfer but B-

cells were not. In addition, Wang et al. [25] prevented diabetes

in NOD mice with injection of anti-CD4 antibodies. Using adoptive

transfer experiments with immunodeficient NOD"sid/"id recipient mice,

Christianson et al. [26] found that CD4-positive cells from

prediabetic NOD mice could enter the pancreas without CD8 cells, but

CD8-positive cells required CD4 cells to enter the pancreas, and

that both populations were required for diabetes. On the other

hand, CD4-positive cells from older, diabetic mice, could initiate

disease without CD8 cells, suggesting that prior chronic exposure

to beta cell antigens results in CD4-positive effector cells.

Haskins and coworkers have isolated islet-specific T-cell

clones from NOD insulitic lesions. Some of these clones have been

shown to cause diabetes when injected into young NOD mice [27,28].

The clones have heterogeneous receptor sequences indicating that

diabetes is oligoclonal, many autoantigens may be involved in

pathogenesis, and that T-cell receptor-specific immunotherapies will

not work to prevent IDD [29]. This is in agreement with studies

that have demonstrated that there is no dominant V0 or JR T-cell

receptor usage pattern in IDD [30-32]. Other T-cell clones have

been isolated from NOD islet isografts placed in diabetic NOD mice

[33]. This method was efficient for isolating large numbers of

islet-specific T-cell clones that transfer diabetes to NOD/LtSz-

scid/scid recipients.

One pathogenic CD4-positive clone which was further

characterized did not require CD8-positive T-cells to be

diabetogenic and recognized an antigen in the beta cell granule

membrane [34,35]. Transgenic mice expressing the rearranged a and

p chain genes of the T-cell receptor from this pathogenic clone

caused diabetes in a normally non-susceptible mouse strain [36].

All of these experiments together show that CD4-positive T-cells are

required for pancreatic infiltration and that in the normal course

of insulitis, which requires CD8 cells, end-stage CD4 effector cells

are generated which can, by themselves, cause diabetes. However,

the normal progression of diabetes pathogenesis which results in the

production of those end-stage pathogenic T-cells may require or be

accelerated by many different factors affecting CD8 cell activation,

thymic or peripheral tolerance induction, antigen processing and

presentation, cell trafficking, pancreatic vulnerability, cytokine

secretion or responsiveness, and/or autoantibody production, as well

as CD8 cells.

CD8- and CD4-positive T-cell clones have been isolated from

NOD islets which can prevent diabetes in NOD mice, indicating that

active suppression is also ongoing in insulitis [37,38]. Because

not all animals with insulitis progress to diabetes, the possibility

of "protective insulitis," due to regulatory CD4-positive and/or

CD8-positive T-cells, has been proposed [39-42]. Whether these

represent differences in Thl and Th2 responses is currently the

subject of intense interest.

Immunoreculation and Thl vs. Th2 responses

The Thl/Th2 dichotomy was first described by Mosmann and

Coffman [43] based on cytokine secretion patterns in CD4-positive

T-cells. Thl cells predominantly secrete IL-2, IFN-y, and TNF,

cytokines involved in cell mediated immunity, while Th2 cells

produce IL-4, -5, -6, and -10, cytokines involved in humoral

responses, anti-parasitic immunity, and allergy. It has been

proposed that Thl cells promote IDD pathogenesis while Th2 cells

protect against disease, and that the balance between the two cell

types determines the disease outcome [44,45]. Based on this

hypothesis, induction of a predominantly Th2 response would produce

"protective insulitis".

The majority of studies support this hypothesis of the roles

of Thl vs. Th2 cells in IDD. Several studies have suggested that

IFN-y, a Thl cytokine, is important in autoimmune beta cell

destruction. Transgenic mice expressing IFN-y under control of the

insulin promoter exhibit insulitis and develop diabetes [46]. Also,

anti-IFN-y antibodies inhibit the development of diabetes in NOD

mice [47,48].

In addition, injection of recombinant IL-10, which is produced

by Th2 cells and inhibits Thl responses, prevented diabetes and

insulitis in NOD mice as did the injection of anti-IFN-y, but not

anti-IL-4 -5 or -10, suggesting that diabetes is due to a Thl

response [49]. Moreover, injection of IL-12, a cytokine that

stimulates Thi cells, into NOD mice was found to accelerate diabetes

[50]. Pancreatic infiltrates in these IL-12-treated mice were found

to express high levels of IFN-y and low levels of IL-4. Rabinovitch

et al. [51] also found both IFN-y and IL-4 in NOD insulitic lesions

by RT-PCR. Complete Freund's adjuvant (CFA) treatment, which

prevents diabetes, resulted in a large decrease in IFN-y production

but no change in IL-4, indicating that IFN-y is the important

cytokine in determining the outcome of insulitis.

Two recent studies with transgenic T-cell receptors support

the idea that Thl-type responses are pathogenic. A pathogenic T-

cell receptor was found to induce diabetes when expressed as a

transgene on a C57B1 mouse background, which is prone to develop Thl

responses, but not on a Balb/c background, which is known to be more

likely to produce Th2 responses [52]. In a recent study, Katz et

al. [53] established Thl and Th2 cells from the same pathogenic T-


cell clone and found that the Thl-like cells promoted diabetes in

young NOD mice, while the Th2-like cells did not. The Th2-like

cells also failed to protect against pathogenesis by the Thl-like


On the other hand, there is some evidence that Th2 responses

may be involved in pathogenesis. When IL-10 is expressed in NOD

islet beta cells by a transgene with an insulin promoter, insulitis

and diabetes were accelerated, suggesting that diabetes is due to

a Th2 response [54]. Even expression of IL-10 in a diabetes

resistant mouse strain resulted in pancreatic inflammation without

islet infiltration or diabetes [55]. In addition, using RT-PCR to

detect cytokines in freshly isolated islets, Anderson et al. found

a preponderance of Th2 cytokines in pancreata from NOD mice with

insulitis [56].

The Role of Macrophages in NOD IDD

Macrophages play a crucial role in the initiation of an immune

response by processing and presenting antigens to T-cells, and as

a source of pro-inflammatory cytokines and factors, such as IL-1,

TNF, IL-6, prostaglandins, oxygen radicals, etc. In addition,

macrophages may be involved in tolerance induction both in the

thymus and in the periphery. Macrophages have been shown to be

present in pancreatic infiltrates, and may even predominate early

in insulitis [57]. Macrophages were also the first cells observed


in pancreata from adoptive transfer studies with splenocytes from

diabetic mice [58]. Early administration of silica particles,

which are selectively toxic to macrophages, resulted in a reduction

of insulitis and diabetes in the NOD mouse, whether or not induced

with cyclophosphamide [57,59,60]. These results indicate a crucial

role for macrophages in diabetes pathogenesis.

It is unknown whether a defect intrinsic in NOD macrophages is

involved in diabetogenesis. Studies by Serreze et al. [61,62] have

shown that the cytokine-stimulated development and MHC class I

regulation of NOD macrophages are defective and thereby may

influence diabetes progression. They have also found that antigen-

presenting cells expressing an MHC other that NOD H-297 can block

the development of autoreactive T-cells from NOD bone marrow [63].

Since NOD mice cannot generate normal mature myeloid cells and

express low levels of class I upon IFN-y stimulation, they postulate

that that these characteristics, in combination with the

diabetogenic H-27, result in an inability to activate tolerogenic

mechanisms while retaining the ability to activate "low level"

effector responses such as peripheral T-cell proliferation. If

these macrophage defects are of primary importance in diabetogensis,

they should map to a genetic interval associated with IDD



Immunostimulation and IDD

Although there is some evidence that infectious agents may

trigger autoimmunity and diabetes by molecular mimicry or other

mechanisms, many studies have demonstrated that immunostimulation

is protective from diabetes (reviewed by Yoon [64]) As mentioned

earlier, NOD mouse colonies maintained specific pathogen-free have

a higher penetrance of IDD. Experimental infections with

lymphocytic choriomeningitis virus (LCMV) [65,66], lactate

dehydrogenase virus [67], or encephalomyocarditis virus [68] can

prevent diabetes in NOD mice. In addition, stimulation with

complete Freund's adjuvant (CFA) which contains bacterial products

and is highly immunostimulatory, or a synthetic double-stranded

polyribonucleotide (poly I:C) which mimics viral infection and

induces interferon production, can also prevent diabetes [69-71].

These observations have led to the hypothesis that immunostimulation

increases release of pleiotropic cytokines in a way that promotes

immunoregulation of diabetogenic autoimmune pathways [69,72]. One

proposed mechanism is through upregulation of defective NOD

macrophage antigen presenting function to allow for normal tolerance

induction [73].

Bone Marrow Transfer Studies and Pancreatic Vulnerability

Bone marrow transfer studies have clearly demonstrated that a

major component of NOD susceptibility to IDD resides in the


hematopoietic bone marrow cells. Wicker et al. [74] found that

irradiated (NOD X B10)F1 mice reconstituted with NOD bone marrow

exhibited insulitis and diabetes within 9 months after transfer.

In addition, NOD and B10 neonatal pancreata grafted under the kidney

capsules of these chimeras (NOD bone marrow into irradiated Fl mice)

resulted in insulitis and beta cell destruction of both types of

grafts. This result indicates that B10 pancreas is just as

vulnerable to autoimmune destruction as NOD pancreas. Although these

experiments suggest that the NOD bone marrow alone expresses the

genes involved in IDD, these experiments do not rule out that the

NOD pancreas is more susceptible to disease or that the genes

impacting on pancreatic vulnerability are dominant or partially

dominant and therefore present in the Fl. In support of this, the

percentage of chimeric mice with insulitis (100% of NOD->NOD vs. 67%

NOD->F1) and diabetes (100% into NOD vs. 21% into Fl) is lower in

F1 as compared with NOD into NOD. As the authors point out, however,

the differences may be due to differences in levels of MHC

expression in the pancreas, thymus, and/or antigen presenting cells

between the homozygotes and heterozygotes. Similar results were

found by Serreze et al. using nonobese, nondiabetic (NON) mice [75].

To test whether or not MHC class IDb, which is shared by NOD,

B10, and NON mice in the previous bone marrow transfer studies, is

required for beta cell destruction, Serreze et al. created chimeras

with NOD bone marrow into (NOD/Lt X CBA/J)Fl mice (CBA/J is H-2k


which is unrelated to MHC from these other mouse strains) [76].

They found that both CBA/J and NOD/Lt islet grafts were rejected in

hyperglycemic but not normoglycemic chimeric mice, indicating that

class I restricted effectors were not required, at least in the late

stages of islet destruction.

LaFace and Peck reported that insulitis and a low penetrance

of diabetes can be induced by adoptive transfer of NOD bone marrow

cells into diabetes resistant recipients [77]. All of these results

provide compelling evidence that the NOD bone marrow cells are

intrinsically diabetogenic, although defects residing in other cells

may be required for full expression of diabetes, e.g. beta cells,

thymic epithelium, or vascular endothelium.

The Role of the NOD Thvmus in IDD

The thymus is an organ required for T-cell development and

central tolerance to self antigens. It plays a crucial role in

self/nonself discrimination by positive selection for T-cells that

recognize self-MHC molecules, and negative selection against self-

reactive T-cells. The thymic epithelium is believed to be the site

of positive and negative selection. The process by which self

antigens are expressed in the thymus is unknown, although Jolicoeur

et al. detected low levels of mRNA from insulin and other pancreas-

specific genes expressed in the thymus [78]. Migrating macrophages

may sample tissues and bring antigens to the thymus for use in

selection. A defect in the thymic selection process has been

postulated to be a cause of autoimmunity.

A number of studies have demonstrated that injections of

islets or islet antigens (such as GAD) into the NOD thymus can

prevent diabetes [18,59,79]. These results indicate that central

tolerance to islet beta cell antigens in the NOD mouse is


Neonatal NOD thymus, irradiated or 2-deoxyguanosine-treated to

kill hematopoietic cells, was transplanted into athymic nude mice

to test for the role of the NOD thymus in IDD [80]. Insulitis, but

not diabetes, was found in Balb/c-nu/nu, but not CBA-nu/nu or

C57BL/6-nu/nu mice, leading the authors to suggest that there is a

primary defect in the NOD thymus leading to the development of

autoreactive T-cells. They also concluded that autoreactivity is

restricted to the H-2Kd allele shared by NOD and BALB/c mice. It

had previously been shown that NOD thymic epithelial cell clones

express higher levels of class I MHC than did CBA cells [81].

IDD Autoanticens

A key event in diabetes pathogenesis must be the recognition

of an islet cell antigen, or set of antigens, in the context of the

NOD MHC, by a non-tolerized CD4-positive T cell(s). Therefore,


there is tremendous interest in the nature of the eliciting

autoantigen(s). If a single initiating antigen is identified, those

individuals identified to be at risk for the development of IDD can

be tolerized to that specific antigen. In addition, immunotherapies

such as targeting of specific T-cell receptors or classes of

receptors (e.g. specific VPs) or use of peptide antagonists, can be

initiated. All of these approaches have been reported to be

successful in the induced murine model of experimental allergic

encephalomyelitis (EAE) (reviewed by Utz and McFarland [82]).

However, EAE is a disease model induced by injection of a known

antigen, myelin basic protein, plus pertussis toxin, and is known

to have a T-cell restricted response. Spontaneous disease models

such as NOD IDD may have several initiating antigens, an

unrestricted T cell response, and several possible pathways to beta

cell damage.

The majority of identified autoantigens are probably a result

of determinant spreading [83], that is intra- or inter-molecular

spreading of the antigens eliciting immune responses from the first

antigen or peptide to others, due to release of sequestered

antigens. Continuation of the pathogenic anti-islet immune response

and resultant diabetes may be dependent upon exposure of and T-cell

responses to these new autoantigens.

Autoantigens can be identified as targets of T-cells and/or B-

cell autoantibody production. Although several islet-specific


autoantibodies have been identified, the role of these antibodies

in IDD pathogenesis is unclear since B-cell-free T-cell transfers

can elicit disease. However, antibodies may contribute to the

progression of spontaneous disease. B-cells and antibodies are

present in the NOD pancreas early in insulitis and could be involved

in ADCC [17]. Although autoantigens recognized by T-cells may be

more relevant to IDD initiation, antigens recognized by antibodies

are more readily identified. Several islet-specific autoantibodies

have been found to be in common between humans and mice, including

antibodies specific for insulin, glutamate decarboxylase (GAD), heat

shock proteins (Hsps), and carboxypeptidase H (CPH) (reviewed by

Atkinson and Maclaren [84]).

The two most well-characterized autoantigens for B- and T-cell

responses are GAD and insulin. GAD was first identified as a 64K

islet cell antigen recognized by autoantibodies in human, BB rat,

and NOD mouse serum [85-88]. Two recent studies found that immune

responses to GAD were the first to be observed in NOD mice. Kaufman

et al. [89] found that GAD T-cell responses were the earliest, first

detected at 4 weeks of age concurrent with the onset of insulitis.

This anti-GAD response increased for 4 weeks and declined

thereafter. T-cell responses to Hsp and CPH were first observed at

6 and 8 weeks respectively. Insulin-specific T-cells were not found

until week 15. They also found that the GAD-specific T-cells

secreted IFN-y in response to antigen, indicative of a Thl type of

response. NOD mice were given a single tolerizing intravenous

injection of 50 pg of recombinant purified GAD that resulted in

virtual elimination of GAD-reactive T-cells, no insulitis in 75% of

the mice, and protection from the development of diabetes.

The second study by Tisch et al. [18] also found that GAD-

specific CD4-positive T-cells appeared earlier than those specific

for other islet antigens, peripherin, CPH, and Hsp60. Anti-GAD

antibodies also first appeared at 4 weeks, before antibodies to

other islet antigens were seen. Tolerogenic intrathymic injections

of GAD reduced T-cell responses to GAD, and prevented diabetes. The

results of these studies indicate that GAD may be a very important

initiating antigen in NOD diabetes.

Other antigens may also be involved in the initiation of IDD.

Gelber et al. [90] found islet cell extract-specific peripheral T-

cells in young female NOD mice before detectable insulitis. These

cells were not specific for recognized autoantigens: GAD, Hsp65,

insulin, ICA69, CPH, peripherin. In addition, the antigens for

previously isolated pathogenic T-cell clones are as yet unidentified

[35] .

Prevention or delay in NOD diabetes has also been reported for

Hsp (Elias et al. [91,92]) and insulin given subcutaneously

(Atkinson et al. [93]) or orally (Zhang et al. [94]). Probably

because these are not the initiating antigens, treatments reduce the

severity of, but do not prevent, insulitis.

Adhesion Molecules and Leukocyte Homing to the Pancreas

Adhesion molecules can play several important roles in immune

responsiveness: mediating interactions between leukocytes and

between leukocytes and vascular endothelial cells, and affecting

trafficking, homing, and entry of leukocytes into tissues (reviewed

by Pober and Cotran [95] and Carlos and Harlan [96]).

One of the initiating events in the development of IDD may be

abnormal trafficking and homing of leukocytes to the pancreas.

Adhesion of the leukocytes to the endothelium in the pancreas,

allowing for entry into the tissue, may be an initiating event in

pathogenesis, and involves several homing receptors. The adhesion

molecules that have been investigated in the NOD mouse so far

include lymphocyte homing receptors involved in binding to

peripheral lymph nodes (L-selectin) or mucosal lymphoid tissue

(LPAM-1, oa47-integrin) high-endothelial venules, as well has HEV

ligands, peripheral vascular addressin (PNAd) and mucosal vascular

addressin (MadCAM-1). Hanninen et al. [97] reported that a4p7-

integrin was expressed on most infiltrating cells in the pancreas

and that L-selectin expression is increased in islets at later

stages of insulitis. Mucosal vascular addressin (MadCAM-1) and PNAd

also became detectable during inflammation. They concluded that

mucosal (MadCAM-l/a4P7-integrin) and peripheral (PNAd/L-selectin)

recognition systems are involved in pancreatic inflammation. Yang

et al. [98] found that blocking L-selectin and very late antigen 4

(VLA-4) with specific monoclonal antibodies inhibits insulitis and

prevents diabetes in the NOD mouse. Administration of VLA-4-

specific monoclonal antibodies can also prevent insulitis and

diabetes in an adoptive transfer model of disease in NOD mice [99].

In an adoptive transfer model in the NOD mouse, administration

of monoclonal antibodies to a4-integrin and one of its ligands,

vascular cell adhesion molecule 1 (VCAM-1), led to a delay in the

onset of diabetes and a decrease in the overall incidence [100].

Insulitis was also reduced in anti-a4-integrin treated adoptively

transferred mice. VCAM expression was found to be elevated in the

islets and on pancreatic vascular endothelium. In this study,

antibodies to intercellular adhesion molecule 1 (ICAM-1) had little

effect on IDD. However, other studies have shown that monoclonal

antibodies to ICAM-1 and its ligand, leukocyte function-associated

antigen-1 (LFA-1), can prevent diabetes in NOD mice and in adoptive

transfer studies [101].

Inhibition of macrophage migration into the pancreas may also

prevent diabetes. In adoptive transfer studies, Hutchings et al.

[40] showed that antibodies against an adhesion-promoting type 3

complement receptor on macrophages prevented insulitis. In

addition, antibodies to the phagocyte dimer, CD11/CD18 (02-

integrins), could inhibit IDD in NOD mice [102].


The Role of the Maior Histocompatibility Complex (MHC) in IDD

MHC class I and II molecules play a crucial role in immune

responsiveness by presenting foreign (and self) antigens to T-cells

and by shaping the T-cell repertoire by positive and negative

selection of T-cells in the thymus. The MHC correlates with disease

in several autoimmune diseases, including IDD, in man and rodent

models [103]. Genetic studies in the NOD mouse have clearly

demonstrated the importance of the NOD MHC, designated H-20', in

diabetes susceptibility [104-107]. Also, NOD.H-2b mice, which

differ from NOD only at the MHC locus, exhibit inflammation in the

pancreas and submandibular gland, and autoantibody production, but

no insulitis or diabetes was observed [108].

Based on MHC congenic studies, the NOD H-291 was thought to be

recessive for diabetes, and codominant for insulitis and

cyclophosphamide-induced diabetes in crosses with NOD.Thy",B10-H-2b,

and NOD.B10-H-2 mice [108,109]. However, Wicker et al. found a low

penetrance of diabetes (3%) among 115 MHC heterozygotes between

backcross generations 2 and 9 which was not due to recombination

within the MHC [110]. The entrance of insulitis among NOD MHC

heterozygotes was 50%. This led Wicker et al. to suggest that the

NOD MHC is "dominant with low penetrance" rather than reccessive

[111] .


The NOD MHC class II expresses a unique I-A molecule [112] and

no I-E [104]. The NOD I-Aa is identical in sequence to I-Ad, but

the I-AP lacks a common ammino acid, aspartic acid, at position 57

in the peptide binding groove [113]. This feature is particularly

interesting because it is shared with human HLA-DQP alleles which

have a positive IDD association.

The lack of I-E expression in NOD mice is due to a promoter

defect in the I-Ea locus [104]. Several investigators have found

that introduction of an I-E molecule, by breeding or transgenic

insertions, into NOD mice resulted in resistance to diabetes

[108,114-116]. Transgenes may themselves have undesired effects.

McDevitt and colleagues reported generalized B-cell defects in some

MHC transgenic mice. However, the source of the I-E molecule may

be important as Wicker et al. showed that certain I-E molecules bred

onto NOD were permissive for a low penetrance of diabetes while

others were completely resistant [116]. This pattern of inheritance

suggests that presence of non-NOD MHC class II molecules is

protective. The mechanism of this protection may be negative

selection of autoreactive T-cells, antigen competition, induction

of peripheral energy through autoantigen presentation, or a

decreased density of I-A97 surface molecules.

Two diabetes resistant sister strains were derived from the

same noninbred ICR strain from which NOD arose, the non-obese non-

diabetic (NON) and the cataract Shionogi (CTS) mouse strains (Makino

et al. [117]). Ikegama and Makino [118] reported that CTS mice are

intra-MHC recombinants sharing MHC class II genes with NOD but not

class I genes. They also bred NOD.CTS-H-2 congenic mice and found

that these mice were susceptible to insulitis and diabetes, thereby

supporting class II alleles as being involved in diabetes (Ikegami

and Makino [6]).

There is some evidence suggesting that MHC-linked loci other

than class II alleles may be involved in NOD susceptibility to IDD.

The penetrance of IDD in NOD.CTS-H-2 mice was found to be lower than

that for mice homozygous for NOD H-29 derived from the same cross

[6]. MHC class I and/or class III genes may also be involved in IDD

susceptibility. The class I H-2K gene is 5' of the class II region.

The MHC class III region is located between the class II genes and

class I H-2D. The MHC therefore includes a large number of genes.

These genes include several molecules that have been implicated in

IDD: peptide processing and transport genes (TaD and Lmp), TNFg,

three heat shock proteins, and several complement genes, as well as

a number of unidentified genes. Ikegami et al. [119] and Lund et

al. [120] compared the class III regions from NOD and CTS mice and

found by restriction fragment length polymorphisms (RFLP) and

microstellite typing that they shared the 5' region of class III

from class II to Hsp70 and differed 3' of Hs70B through H-2D.

Therefore, if there is another susceptibility gene in NOD, it is

either 5' from class II (including H-2K and Tia and Lm" genes) or


in the class III region 3' from EHs70 or H-2D. Faustman et al.

[121] suggested that defects in ITa genes might lead to a defect in

class I expression on NOD splenocytes. However, two other studies

did not confirm this [122,123].

No structural defects have been reported in MHC class I genes,

however, MHC class I is required for spontaneous NOD diabetogenesis.

Treatment of NOD mice with anti-class I antibodies prevents

spontaneous and cyclophosphamide-induced diabetes [124]. Although

Faustman et al. (1991) reported that MHC class I deficient BALB/c-

B2-microglobulinn"n/"nu mice become hyperglycemic after 18 months of

age, this was not confirmed with other class I deficient mice.

Furthermore, NOD-B2-microglobulinnln/nun mice are completely

resistant to insulitis and diabetes [20,125,126]. A probable

explanation for this effect is that class I deficient mice also lack

CDB-positive T-cells, and CD8 cells are required for spontaneous

diabetes. The lack of insulitis is not due to loss of islet cell

class I molecules, as splenocytes from diabetic NOD adoptively

transferred into NOD class I null mice can induce delayed insulitis

and diabetes [126].

There are several reported defects in class I expression in

NOD mice. NOD islets express high constitutive levels of MHC class

I protein [127,128], and the cytokine-induced regulation of class

I expression is aberrant in bone marrow-derived macrophages from NOD

mice, due to transcriptional factor defects [61,62].

Analysis of the Polycenic Inheritance of IDD Susceptibility

Inheritance of IDD and insulitis susceptibility is polygenic

in the NOD mouse. In various crosses between NOD mice and non-

susceptible mouse strains, F1 mice do not develop IDD, suggesting

that inheritance is recessive. Analysis of various backcrosses with

NOD resulted in estimates of three to five susceptibility alleles

depending on the strain combinations. NON and C3H crosses gave the

highest rates of diabetes among the backcrosses with 10%, while

C57BL/10 was 4.5%, C57BL/6 was 1.3 and C57BL/KsJ was only 0.9%

[7,105,106,129]. Rates for insulitis were higher than diabetes in

all crosses, therefore fewer genes were postulated to be involved

in insulitis susceptibility. The first diabetes susceptibility

locus to be identified was the NOD MHC (Iddl) as discussed

previously [104-107]. A second diabetes susceptibility gene was

identified in a cross between NOD and the related but non-diabetic

strain, NON, which was linked to Thy-1 on chromosome 9 (Idd2 [105]).

Genome-wide mapping studies were initially limited by the poor

availability and spacing of polymorphic markers which were also

time-consuming as they were primarily based on RFLP (Southern blot)

techniques [105,130]. The development of simple sequence repeat

(SSR) polymorphic loci for rapid genome-wide analysis has made

mapping studies of disease susceptibility genes, especially for

polygenic traits, feasible [131,132]. The technique is based on the

large numbers of di-, tri-, and tetra-nucleotide repeats present

throughout the genomes of mammals which vary in length between

genetically divergent individuals. Currently, more than 4,000 mouse

SSRs are available [133] and they have been used to map genes

involved in cancer [134], systemic lupus erythematosus (SLE) [135],

and antibody response genes (Wu,J., Longmate,J., Adamus,G.,

Hargreaves,P., Wakeland,E., submitted) in addition to IDD.

Todd et al. [136] were the first to use SSRs to map non-MHC

IDD susceptibility loci. Using (NOD X C57BL/10-NOD.H-2g7)F1 X NOD

backcross mice, 3 recessive NOD IDD susceptibility genetic intervals

were identified on chromosomes 3 (Idd3), 6 (Idd6), and 11 (Idd4),

and 2 dominant susceptibility intervals from the non-diabetic

strain, on chromosomes 7 (Idd7) and 14 (Idd) were detected

[136,137]. Chromosome 1 (Idd5) was also found to be weakly linked

to diabetes in the same cross [138]. In addition, studies on

insulitis and sialitis in a cross between NOD and C57BL/6 and NZW

identified two genes on chromosome 1 related to these traits [139].

Chromosome 3 was later found to have a second peak locus, Iddlo,

contributing to diabetes susceptibility [140,141]. Chromosomes 4

(Idd9) [142] and 2 (Iddl3) [143] were also found to correlate with

diabetes in crosses between NOD and B10 and NOR (non-obese

resistant). Two other NOD-derived genetic intervals on chromosomes

4 (Idd1l) and 14 (Iddl2) were found in crosses between NOD and

C57BL/6 or SJL [144]. All together 12 NOD-derived and 2 non-NOD-

derived intervals have been identified in various crosses with NOD

mice. Different strain combinations can reveal different genetic

intervals. For instance, Idd2, the ThIl-linked gene found to be

strongly associated with diabetes in the NOD and NON cross, was not

detected in the NOD and C57BL/10-NOD-H-2g7 [137] or C57BL/6


Candidate genes proposed to be the diabetes susceptibility

genes include Interleukin 2 (IL-2), a T-cell cytokine, for Idd3

[145], Fc receptor (FcR) for IddlO [141], and both Bcl-2, a gene

involved in cell apoptosis [139], and Lsh/Ity/Bca, a gene affecting

macrophage function [138], for Ldd5. Although these genes differ

in sequence, expression or function between NOD and some diabetes

resistant strains, it has yet to be determined if these are, in

fact, the susceptibility genes. Different genetic approaches, such

as the use of congenic, transgenic, and/or knockout mice will be

required to determine the role of these and other candidate genes

in IDD susceptibility.

Rationale for this Study

By the time IDD is diagnosed in humans, the beta cells are

already irreversibly destroyed and endogenous insulin production can

only be restored by pancreas or islet transplantation. Because of

this, one of the goals of diabetes research is to identify

individuals at risk for the development of diabetes so that

preventative therapies such as immunosupression, insulin oral

tolerance, etc. can be initiatedbefore loss of the beta cells is

complete. If genes contributing to the risk of IDD could be

identified, people with could be screened and those with

susceptibility genes could be started on preventative therapies.

Because the inheritance of IDD is complex, identification of

the genes involved in diabetes susceptibility in outbred populations

like humans will be very difficult [5]. Animal models, such as the

NOD mouse, allow for the use of specific crosses to identify the

genetic intervals, and ultimately the susceptibility genes. As an

initial screen for diabetes susceptibility genes, the syntenic

regions between haman and mouse chromosomes can be studied.

Syntenic regions are groups of genes which have remained linked

together through evolutionary time, such that they are found

together in mice and humans. This approach was successful for the

identification of human IDDM7 which is syntenic with mouse JIdd5on

chromosome 1 [146]. Even if some or all of the diabetes

susceptibility genes differ between humans and mice, identification

of murine susceptibility genes will identify the pathways involved

in diabetes pathogenesis which may be affected in humans as well as

in mice.

Although IDD pathogenesis in the NOD mouse has been studied

intensively in recent years, the only susceptibility gene which has

been identified is A3 of the MHC, and it is still not known if it

is the only susceptibility allele in that genetic interval.

Identification of other genes could reveal alleles that are involved

in tolerance induction, antigen processing or presentation, T-cell

development, cytokine secretion or responsiveness, beta cell

vulnerability, or lymphocyte trafficking.

Successes in the the identification of genes involved in

disease have so far been restricted to single-gene Mendelian traits

wherein identification is fairly straight forward: linkage analysis,

positional cloning, and identification of mutations [147]. Many

diseases, however, are inherited as complex polygenic traits,

including systemic lupus erythematosus (SLE), hypertension, cancer,

antigen-specific antibody production, developmental abnormalities

such as cleft palate, many phenotypic characteristics such as height

and weight, and behavior. In addition, even traits linked to a

single strong susceptibility gene can be tremendously influenced by

background genes. An example is the ipr-fas gene which, on the MRL

mouse strain, results in a severe lymphadenopathy and early onset

SLE, but on other genetic backgrounds exhibits only mild

lymphadenopathy and no SLE [148].

This research focuses on approaches that can be used to study

complex traits such as diabetes. The strategy that is being used

for the genetic dissection of IDD pathogenesis is shown in Figure

1-1. A linkage analysis was first performed on a backcross, (NOD

X C57BL/6)F1 X NOD, to find genetic intervals correlating with IDD


and insulitis susceptibility (Chapter 2) Then, to dissect the

different genes contributing to diabetes, we made congenic mouse

strains each containing one of the IDD susceptibility intervals from

NOD bred onto a resistant C57BL/6 background (Chapter 3). These

congenic lines were then characterized histologically for pancreatic

pathology (Chapter 4).

Select parental strains
Determine mode of inheritance

Produce informative backcross or intercross progeny

Sweep genome of progeny with polymorphic markers
Identify genomic intervals affecting traits

Identify candidate loci within interval Produce congenic strains
Characterize polymorphisms

Phenotypic dissection Fine map

Functional characterization Identify genes

Figure 1-1. Flow diagram of the strategy being used for
the genetic dissection of IDD pathogenesis.



Insulin dependent (Type 1) diabetes (IDD) results from the

autoimmune destruction of insulin-producing pancreatic beta cells

by the immune system. Autoreactive T lymphocytes and macrophages

infiltrate pancreatic islets and form inflammatory lesions (termed

insulitis), resulting in the progressive destruction of the insulin-

producing beta cells, and culminating in the loss of insulin

production and the development of IDD (reviewed by Castafio and

Eisenbarth [149]). Although these key events in the pathogenesis

of IDD are well documented, the immunological mechanisms responsible

for their occurance are only poorly understood.

The pathogenesis of IDD has been studied extensively using the

nonobese diabetic (NOD) mouse strain as an experimental animal

model. The NOD mouse strain is extremely diabetes-prone, with NOD

females developing highly penetrant, spontaneous IDD sharing most

of the pathologic features of human IDD [117,150,151].

Interestingly, environmental factors appear to dramatically impact

the frequency of spontaneous IDD observed in NOD mice. Separate

colonies of genetically-identical NOD mice often exhibit extreme

variations in IDD penetrance (from 20% to 100% affected females),

with IDD susceptibility highest in colonies maintained in specific-

pathogen-free conditions [9]. The reasons for this correlation are

not understood, but the impact of the environment on the penetrance

of IDD is clearly evident.

The mode of inheritance of IDD susceptibility in crosses with

NOD is extremely complex, due predominantly to the involvement of

multiple genetic and environmental factors [7,8,10]. Todd and co-

workers reported interval mapping data in a cross of IDD

susceptibility using (NOD X C57BL/10-NOD.H-2D7)Fi X NOD backcross

progeny (BC1) [136,137]. A total of 8 recessive IDD susceptibility

alleles derived from NOD and 2 dominant IDD susceptibility alleles

derived from the non-diabetic C57BL/10-NOD.H-297 strain were

detected in this cross. These results illustrate the polygenic

inheritance of IDD susceptibility from NOD and indicate that non-

diabetic standard inbred strains contain some alleles capable of

potentiating diabetes susceptibility when combined with diabetes-

prone genotypes.

Although these studies have identified the approximate genomic

locations of several IDD susceptibility genes in the mouse genome,

the identities of the genes responsible remain a mystery. The

strongest association with IDD susceptibility in NOD is with the MHC

class II genes [104-107] and Ab97 in NOD has been shown to share

unique structural features with HLA-DQB alleles that correlate with

IDD susceptibility in humans [113]. The identities of the non-MHC

genes contributing to IDD susceptibility in NOD are currently

unknown, although some interesting candidate genes have been

recently identified [137,145].

Susceptibility to IDD in humans is also inherited in a

multifactorial fashion, consistent with the involvement of multiple

genes and environmental factors [3]. Recently, several laboratories

have used sib-pair analysis to identify human genomic intervals

potentially containing IDD susceptibility modifiers [5,152,153].

The mode of inheritance of IDD susceptibility detected was complex

and the effects of some susceptibility alleles were dependent upon

the presence of specific MHC alleles [152]. Furthermore, the

association of marker loci with IDD susceptibility was inconsistent

between panels of families, suggesting that the genetic basis for

increased IDD susceptibility differed between populations. These

results indicate that the inheritance of IDD susceptibility in human

populations will be extremely complex.

In this paper we report our analysis of the inheritance of IDD

and insulitis susceptibility in an (NOD X C57BL/6)F1 X NOD backcross

(BC1). We demonstrate that IDD susceptibility is inherited in this

model system as a genetically heterogeneous trait, in which several

distinct genetic pathways can lead to IDD susceptibility. Based on

these results, we propose that IDD susceptibility is inherited as


a threshold liability in which both environmental and genetic

factors contribute to the susceptibility of individuals to diabetes.

Materials and Methods

Animals. Mice were bred and maintained in the Department of

Pathology Mouse Colony at the University of Florida. The origins

of NOD/Uf have been described previously [154] and breeding pairs

for C57BL/6 were originally obtained from Jackson Laboratory.

Diagnosis of diabetes. All mice were observed daily for

evidence of illness, weight loss, or excessively wet bedding as

evidence of polyuria. Mice with evidence of illness and all mice

in cages with wet bedding were tested for whole blood glucose

(Chemstrip bG, Boehringer Mannheim, Indianapolis, In) from tail

bleeds. Diabetes was diagnosed on the basis of glucose values > 240

mg/dL that persisted over 3 or more days.

Ouantitation of insulitis. Insulitis scores were measured on

6 month old mice by Drs. Cheng and Winter as described previously

[130,154]. Briefly, pancreata were formalin fixed, paraffin

embedded, sectioned and stained with hematoxylin and eosin.

Multiple pancreatic levels were read, islet by islet, and individual

islets were scored as follows: zero: no peri-islet insulitis, 1:

peri-islet insulitis, 2: intra-islet insulitis covering <50% of the

total islet area, and 3: intra-islet insulitis covering Z50% of the

total islet area. More than 30 islets per pancreas were scored for

insulitis severity by two independent observers. The insulitis score

for each pancreas was obtained by calculating the mean of all the

individual islet scores.

Genotypic analysis. Genomic DNA was prepared from liver or

tail samples using standard techniques. Marker loci used in this

study were PCR-based simple sequence repeat (SSR) loci. SSR primers

were synthesized according to previously published sequences [131]

or purchased as Mappairs (Research Genetics, Huntsville, Al).

Genomic DNA (100 ng) was amplified in 20 ul reactions with

200 nM of each primer, 0.2 mM dNTPs (Pharmacia), 0.75 U of Taq DNA

polymerase, in a standard incubation buffer containing 50 mM MgCl2.

Amplifications were carried out in Perkin-Elmer/Cetus 9600 thermal

cycler under the following conditions: 1 cycle at 94*C for 2 min;

35 cycles at 94C for 30 sec, 50-62*C for 30 sec, and 72'C for 1

min; 1 cycle at 72'C for 2 min. The optimal annealing temperature

was determined for each primer pair. PCR products were visualized

on 5% agarose gels stained with ethidium bromide and each individual

was scored as NOD homozygous or NOD/B6 heterozygous for each locus.

Linkage analysis. The segregation patterns of 98 marker loci

were analyzed among 167 (NOD/Uf X C57BL/6)F1 X NOD/Uf (BC1) progeny

(48 diabetic and 119 non-diabetic mice) with the MAPMAKER computer

package [155] Linkage maps for all 19 mouse autosomes were

established based on our cross and oriented relative to a composite

mouse genomic map using the positions of anchor loci in the GBASE

using the MIT maps. Distances between linked marker loci were

computed using Haldane's formula.

Identification of Genomic intervals containing IDD and

insulitis susceptibility alleles. Genomic intervals containing

alleles associated with insulitis susceptibility were detected by

screening individual markers for association with insulitis severity

using a non-parametric rank sum test and by analysis of insulitis

scores as a quantitative trait using MAPMAKER-QTL [155]. Genomic

intervals containing alleles associated with IDD susceptibility were

identified by screening markers individually for evidence of

association with diabetes using Pearson's chi-square statistic.

Simultaneous evaluation of markers was performed using cumulative

logistic regression (Agresti[156]).

Analysis of penetrance. The impact of non-H-2 susceptibility

modifiers on the penetrance of IDD among diabetic BC1 progeny was

modeled using logistic regression (using non-diabetics as controls)

and log-linear modeling (comparing to an expected 50:50 genotypic

ratio) by Dr. Longmate (manuscript in preparation). Penetrance

estimates were corrected for selective typing of diabetics using

Bayes theorem (Breslow and Day [157]).



Phenotypic analysis of insulitis and IDD in NOD and BC1

roageny. Drs. Cheng and Wakeland initiated a (NOD/Uf X C57BL/6)F1

X NOD/Uf backcross (BC1) in 1988 to identify genomic intervals

containing IDD and insulitis susceptibility alleles in NOD. BC1

progeny were monitored for IDD until 52 weeks of age. Diabetic BC1

were collected from a continuously maintained panel of about 200

female BC1 progeny through April, 1994. This resulted in the

collection of 43 diabetic BC1 progeny from a total of 1418 BC1

females screened. This collection was supplemented with 5

additional diabetic BC1 females obtained by Dr. L. Lotte-Harberg

(Diabetes Research Institute, Dusseldorf, Germany) in a similar

fashion to provide a total of 48 diabetic BC1 females from 1624 BC1

mice (2.95%).

Genetic susceptibility to insulitis was analyzed separately

from IDD susceptibility, so that the roles of the genes identified

in the etiology of each phenotype could be assessed independently.

Therefore, Drs. Cheng and Winter measured insulitis by histological

examination of pancreata from 119 non-diabetic BC1 progeny [130].

Fifty-two BC1 progeny (43%) failed to develop detectable insulitis.

Among the 67 BC1 progeny that did develop insulitis, insulitis

scores were highly skewed with the majority of the BC1 progeny

showing relatively low levels of insulitis.


Genotypic analysis of BC1 procenv. The 119 non-diabetic BC1

progeny screened for insulitis together with the 48 diabetics

resulted in a data set of 167 mice. Ninety-eight non-diabetic and

28 diabetic mice were originally genotyped for RFLP markers by Dr.

Cheng [130]. After Dr. Cheng completed his study, an additional 19

diabetics were obtained which were necessary to improve the sample

size for statistical analysis. In addition, a large number of PCR-

based SSR markers had become available. Because most of Dr. Cheng's

markers were RFLP-based and therefore expensive and time-consuming,

and large portions of the genome had not been swept by his study

because of the paucity of polymorphic markers in some chromosomal

regions, Dr. Muralidharan and I replaced most of Dr. Cheng's loci

with SSR markers and retyped the original mice plus the 19

additional diabetics and, for some loci, another 20 non-diabetic

mice. A total of 98 polymorphic marker loci were then included in

the final linkage map. Of these polymorphic markers, I genotyped

all or some of the mice for approximately 50 SSR loci. Figure 2-1

presents the linkage maps for all 19 autosomes that were generated

using MAPMAKER-EXP and identifies the regions of the genome covered

by these markers. Based on the current mouse microsatellite linkage

map, the 1236 cM contained within these linkage groups spans

approximately 90% of the autosomal genome.

Identification of cenomic intervals associated with IDD and

insulitis susceptibility. Genomic intervals containing loci


contributing to IDD susceptibility were sought by chi-square

analysis and results for all markers used are presented in Table 1.

Two approaches are possible for evaluating the association of marker

loci with diabetes. The ratio of homozygotes to heterozygotes among

diabetic mice can be compared to an assumed 50:50 ratio, or to the

segregation ratio observed in non-diabetic controls. Use of

observed segregation ratios controls for the presence of segregation

distortions which might occur due to deleterious alleles segregating

within the cross. Apparent distortions from the predicted 50:50

ratios were detected in markers from two regions associated with

diabetes susceptibility. H-2 had only 48 homozygotes out of 120

non-selected mice (p=0.035) and D6MIT14 had only 40 homozygotes of

99 non-selected mice (p=0.07). Consequently, chi-square results

calculated using the values from the non-selected mice are presented

for these two regions. For the H-2 region on chromosome 17, the

relationship to diabetes is overwhelming and thus the distortion has

little effect on the detection of linkage. However, the continuity

corrected chi-square test for D6MIT14 on chromosome 6 leads to a p-

value of 0.018 (as does the exact test) while the binomial

comparison to a rate of 0.5 gives p=0.11. The binomial comparison

does detect association between IDD and markers on chromosome 6,

however, the peak association using this test is 32 cM centromeric

from D6MIT14 at D6MITI0.

As shown in Table 2-1, linkage groups on chromosomes 1, 2, 3,

6, 11, and 17 were significantly associated with IDD susceptibility

in this experimental cross. Fourteen IDD susceptibility loci have

been defined thusfar based on several different crosses

[105,130,136-140,142-144], and the intervals segregating in our

cross coincide closely with those reported to contain Iddl, Idd3,

Idd4, Idd5, Idd6, IddlQ, and Idd13. The ratios of diabetic

homozygotes and heterozygotes at the peak loci detected by this

study and that of Ghosh et al. [137], together with their IDD gene

designations and a combined chi-square computation, are presented

in Table 2-2. Both studies independently detected the same

intervals on chromosomes 1, 3, 6, and 11, although there were

significant variations in the strengths of the associations detected

and in peak loci within the intervals. Combining the data from both

studies increases the strength of the statistical associations

defining Idd3, Idd4, Idd5, Idd6, and Iddl0. Our data set also has

an excess of heterozygotes for Plau among diabetics, supporting the

detection of Idd8 on chromosome 14, although the association in our

data set is not statistically significant. Similarly, their data set

supports the presence of IddI on chromosome 2, although the

association in their study is not significant. However, our results

differ with respect to the detection of Idd7 on chromosome 7, in

that we fail to detect an excess of heterozygotes in this region.

Finally, neither of these data sets detects an association between

IDD susceptibility and the central region of chromosome 4, or an

NOD-derived recessive allele in the centromeric region of chromosome

14, as recently reported in a cross of NOD with C57BL/6 and SJL by

Morahan and coworkers [144]. These results indicate that most of

the loci associated with IDD susceptibility in NOD are consistently

detected by independently-derived data sets, but that some

variations do occur among weakly associated loci.

Intervals associated with susceptibility to insulitis were

sought among 119 non-diabetic BC1 using a rank sum test comparing

the insulitis scores of homozygotes and heterozygotes, and using

MAPMAKER-QTL (data not shown). Both analyses detected the same

regions, however, the relative strengths of the associations

detected by these two techniques varied considerably among loci.

Z scores from rank sum test and significance levels for the

association of all marker loci with insulitis scores are presented

in Table 2-1. A total of 5 genomic intervals were identified as

significant by both statistical tests. The strongest association

with the rank sum test was with Bcl-2 on chromosome 1 (Z=3.56,

LOD=2.30), in agreement with the findings of Garchon et al. [139] .

A second interval defined by D1MIT4, close to the centromere on

chromosome 1, also strongly correlated with insulitis in our data

set (Z=2.77, LOD=1.58). This locus is near the region reported by

Cornall et al. [138] to contain 1dd5.

The strongest association detected by MAPMAKER-QTL was with

D17MIT10 (Z=3.54, LOD=6.54), a locus 3 cM telomeric to H-2 on

chromosome 17. This result supports the importance of the H-2

region in insulitis susceptibility but is surprising in that the

strongest association is not with H-2. However, these results do

not provide statistical evidence for a separation of the insulitis

susceptibility locus from H-2. Two more loci associated with

insulitis are located on chromosome 3 in close proximity to Idd

(D3MIT5, Z=2.58. LOD=2.48) and IddlQ (D3MIT10, Z=2.96, LOD=3.51).

The linkage map (Figure 2-1) summarizes the locations of the genetic

intervals found to correlate with insulitis and/or diabetes in this


Penetrance of IDD among BC1 progeny with different cenotypes

for recessive modifier allelles The frequency of 2.95% diabetics

among the BC1 progeny would predict the detection of 5

susceptibility loci derived from NOD, if each locus was fully

recessive and homozygosity at each was required for IDD

susceptibility. However, H-2 was the only IDD susceptibility locus

for which homozygosity was required in our cross (Table 2-1). All

other IDD susceptibility loci were homozygous in the majority of the

diabetic BC1 progeny, but for each interval identified, several

diabetic BC1 progeny were heterozygous throughout the entire 95%

support interval (data not shown). Therefore, these IDD

susceptibility loci appear to be inherited as recessive modifier

genes in which homozygosity contributes to, but is not essential

for, the development of diabetes.

To estimate the relative impact of each modifier on the

penetrance of IDD in this cross, we measured IDD penetrance in each

of 32 segregant genotypes formed by the 5 strongest recessive

modifier alleles we detected, assuming homozygosity for chromosome

17. These were 1dd3, associated with 11-2 on chromosome 3; Iddl,

associated with IThb on chromosome 3; Idd4, associated with D11MIT42

on chromosome 11; Iddl3, associated with D2MIT47 on chromosome 2;

and Idd6, associated with D6MIT14 on chromosome 6. The genotypes

of these peak marker loci should reflect the genotypes of the IDD

susceptibility loci they represent 90% of the time (assuming that

the susceptibility locus is within 10 cM on either side of the peak

marker). The proportion of BC1 progeny in each segregant class,

from which the diabetic mice were selected, was estimated from the

allelic frequencies of the marker loci among the 119 non-diabetic

BC1 progeny genotyped for the analysis of insulitis. Because of the

poor estimates of penetrance based on observed diabetics due to the

low number of diabetics (48) divided among 32 genotypic classes, the

penetrance of IDD within each genotype was modeled using a method

that combines the estimated frequencies of these genotypes with the

overall probability of diabetes, via Bayes' theorem (see Materials

and Methods). These results are presented as smoothed penetrance

values in Table 2-3.

Several interesting insights into the mode of inheritance of

IDD susceptibility in this cross are revealed by this analysis.

First, 15 of the estimated 26 BC1 progeny homozygous for all 5

modifier alleles were diabetic. This result yields a penetrance

estimate of 56.8% (or 51.6% for the smoothed penetrance) of IDD

susceptibility within this genotypic class. IDD penetrance was

about 65-75% among NOD/Uf females during the course of this study

indicating that about 70-80% of the genetic variance in IDD

susceptibility detected in this cross is accounted for by the

combined effect of homozygosity for all 5 modifiers plus H-27. The

remaining 20-30% of the genetic variance probably reflects the

cumulative effects of the modifiers Idd, Idd8, and JIdd, all of

which showed mild distortions in their frequencies among the BC1

diabetics in our cross (see Table 2-1), but were not included in the

analysis of penetrance.

Genetic heterogeneity amona diabetic BC1 procenv. The data in

Table 2-3 indicate that diabetic mice occurred in 15 different

genotypic combinations of IDD modifier alleles. Diabetics were

detected in all combinations homozygous for 4 modifiers and in 5 of

10 combinations homozygous for 3 modifiers. These results indicate

that several distinct genotypes can result in animals with

significant susceptibility for the development of diabetes. For

example, in the case of genotypic classes homozygous for 4

modifiers, which have estimated penetrances for IDD in range of 10%


to 40%, diabetic BC1 progeny only share genotypes at 3 of the 5

recessive modifiers (Table 2-3). These results indicate that

several genetic pathways can lead to IDD susceptibility among the

BC1 progeny.

Only 5 diabetics were homozygous for fewer than 3 IDD

susceptibility modifiers. This result is consistent with the

observed low frequency of diabetes in this cross (7.4% of H-297

homozygotes were diabetic), and the inheritance of IDD

susceptibility as a polygenic recessive trait. However, the

detection of a diabetic BC1 animal homozygous for only one

susceptibility modifier (D2MIT47 homozygote in Table 3) and 4

diabetics homozygous for only two susceptibility modifiers is

inconsistent with this simple interpretation. These outliers might

reflect the cumulative effects of the weaker IDD susceptibility

alleles or recombinational events separating some of the marker loci

from their associated susceptibility alleles.

To address the genetic basis of these diabetic outliers among

the BC1 progeny, we assessed their genotypes in more detail. The

diabetic animal homozygous for a single susceptibility modifier had

recombinational breakpoints adjacent to the marker loci for Idd4 and

Idd6, indicating that Idd4 and/or Idd6 might be homozygous in this

animal. In addition, the genetic interval encompassing Idd9 was

homozygous. Similar observations were made concerning the four

diabetic mice found in classes with only 2 susceptibility alleles,


suggesting that these animals may be outliers due to rare

recombinational events between the marker loci and the

susceptibility genes and/or the cumulative effects of the weaker IDD

susceptibility alleles segregating in this cross.

IDD penetrance increases with the total number of IDD

modifiers. The data in Table 2-3 reveal that IDD penetrance

increases with the number of modifier alleles present within

individual BC1 progeny (p<0.0001 by logistic regression). These

results indicate that susceptibility to IDD depends upon the number

of modifiers present in individual genotypes (Figure 2-2). This

result is consistent with the inheritance of IDD susceptibility as

a threshold liability, as first proposed by Sewell Wright for the

inheritance of polydactyly in guinea pigs [158].

Interactions between IDD modifiers. Based upon multiple

logistic regression models the relative strengths for each locus is

given in Figure 2-2. No one locus is much stronger than the others

and some evidence for an interaction between D6MIT14 and D11MIT42

was found in the model that best fit the data.


This study presents the first detailed analysis of the

inheritance of diabetes and insulitis susceptibility in a large

outcross between NOD/Uf and C57BL/6. We have identified a total of

8 intervals in 6 linkage groups containing recessive alleles

contributing to susceptibility to insulitis and/or diabetes in this

cross. These intervals are generally similar or identical to

intervals identified in previous studies (Iddl, Idd3. Idd4. Idd5.

Idd6. Idd10. Iddl3), although the positions and relative strengths

often varied considerably from those detected previously

[136-140,142,143]. These variations may represent differences in

the non-diabetic partner strains utilized, or may reflect variations

intrinsic to interval mapping of polygenic traits. In this regard,

our results were most similar to those reported by Ghosh et al.

[137] in their analysis of NOD crossed with B10.H-29'. This is not

surprising since C57BL/6 and C57BL/10 are closely-related

substrains. The variations detected between our studies and theirs

(involving the detection of Idd7, Idda, Idd9, and Iddl3) may reflect

variations in environmental factors, minor variations between these

sub-strains, or inconsistencies in interval mapping analyses.

Genetic variations distinguishing C57BL/6 and C57BL/10 have been

detected on chromosomes c2, c4, c7, and c14 (E. Leiter, personal


This study was designed to provide a simultaneous analysis of

the inheritance of susceptibility to IDD and insulitis. Since the

BC1 progeny analyzed for insulitis were not diabetic, this result

allowed an identification of loci contributing to insulitis

independently from their potential role in diabetes susceptibility.

Our findings suggest that insulitis susceptibility is associated

with a subset of the loci correlating with susceptibility to IDD,

and possibly with some additional loci. Idd3 and Iddl0 on

chromosome 3 both strongly correlated with susceptibility to IDD and

insulitis, suggesting that their contribution to IDD pathogenesis

involves mechanisms potentiating inflammation of the pancreas.

However, Idd5 on chromosome 1 mapped to a position between the two

peaks for insulitis susceptibility detected in our cross and by

Garchon et al. [139]. Furthermore, in that study Idd5 most

strongly correlates with cyclophosphamide-induced diabetes rather

than spontaneous diabetes. This result may indicate that both

insulitis susceptibility alleles on chromosome 1 are needed to

increase IDD susceptibility, resulting in the spurious mapping of

Idd5 between these 2 loci. Alternatively, IDD susceptibility may

be mediated by a locus separate from the two insulitis

susceptibility loci. Further work will be required to resolve this


A role for the H-2 region in insulitis was also supported by

our analysis, however, the location of the locus contributing to

insulitis mapped telomeric to the position of Iddl. Although

interval mapping analyses are incapable of resolving intervals

smaller than about 10-15 cM and thus cannot distinguish between 1

versus 2 contributing loci in this interval, these results may

indicate that more than one gene within the H-2 interval contribute

to diabetes and insulitis. The rest of the IDD suspectibility loci


identified in our cross did not correlate with insulitis, suggesting

that their roles in IDD pathogenesis do not involve this process.

Clearly inheritance of IDD and insulitis susceptibility is

more complex than was postulated by backcross analyses and is not

a result of inheritance of 5 required and recessive genes [7]. Only

one gene, that on chromosome 17 linked to the MHC, is required to

be homozygous for diabetes in this and other studies. However, we

and others have found that homozygosity for this locus is not

required for insulitis [108,110]. No other loci are required to be

homozygous for IDD or insulitis. There are two likely explanations

for this. One is that loci are not fully recessive but rather they

are partially dominant, that is the heterozygote exerts some

susceptibility between the two homozygotes (i.e. bb < bn < nn in

terms of relative susceptibility). This backcross analysis could

only distinguish between the homozygote (nn) and the heterozygote

(bn). An F2 analysis could sort out these effects, however with

such a large number of susceptibility genes segregating and the low

penetrance of disease even in the backcross, such an approach would

require extremely large numbers of animals. The other possibility

is that loci are fully recessive but that two or more genes overlap

in function such that only one is sufficient, but that perhaps two

are more effective at producing disease. Both types of inheritance

may be operating with different susceptibility genes. Sorting these

possibilities out will require the use of congenic mouse strains to

separate the effects of one locus from the others and in specific

combinations. We have recently completed construction of such mouse

strains (Chapter 3).

Inheritance of IDD susceptibility as a threshold liability.

Further analysis of data has revealed several interesting features

of the mode of inheritance of IDD susceptibility in our cross. The

genotypes of susceptible animals were heterogeneous, exhibited

variable levels of IDD penetrance, and susceptibility increased with

an increase in the number of recessive modifiers present in

individual animals. Although the relative strengths of the

modifiers may vary and the genes may interact, as we found for

chromosomes 11 and 6, these features are consistent with the

inheritance of IDD susceptibility as a threshold liability. Similar

results were found with an analysis of polygenic susceptibility to

murine systemic lupus erythematosus [135].

Multifactorial inheritance via a threshold liability, first

proposed by Sewell Wright to account for the inheritance of

variations in toe number among strains of guinea pigs [158], refers

to the notion that complex phenotypes will be expressed in

individuals based on an overall "liability." The liability of an

individual to express the trait is effected by three elements: 1)

their genetic susceptibility, which results from the number of

susceptibility alleles they inherit; 2) the penetrance of these


alleles, and 3) the impact of environmental factors on the

expression of the trait.

In our analysis, environmental factors were held relatively

constant, and thus susceptibility of the BC1 progeny to IDD resulted

from the number and penetrance of the susceptibility modifiers

present in each genome. Non-genetic variations in the penetrance

of diabetes is clearly a factor, given the fact that only up to 75%

of NOD/Uf females develop IDD when maintained in a highly controlled

environment. This incomplete penetrance may arise from random

variations occurring during embryonic development and the ontogeny

of the immune system. For example, although the B and T cell

antigen receptor repertoires of individual animals within an inbred

strain are quite similar, each animal expresses an antigen receptor

repertoire with distinct features in its fine specificity for

antigen. These variations probably result from the stocastic

processes involved in the generation of antigen receptor repertoires

via gene rearrangement and a plethora of other factors effecting

immune responsiveness. These variations in the immune system,

together with stocastic variations occurring during embryogenesis in

general, all cumulatively impact on the expression of a

multifactorial phenotype such as diabetes.

The interplay of these penetrance factors and the genetically-

defined predisposition to IDD in the inheritance of IDD

susceptibility as a threshold liability is illustrated in Figure 2-


3. In this illustration, liability to diabetes increases along the

X axis. The top panel depicts the situation with the parental

strains and their Fl hybrid. The liability of individuals within

each of these strains would vary due to the penetrance factors

discussed above, and thus each strain is depicted as a normally

distributed population along the liability axis. The mean liability

of each strain is positioned along the liability axis based on its

intrinsic genetic susceptibility. The vertical dashed line

represents the liability threshold for the development of diabetes.

Thus, NOD/Uf mice are positioned on the liability axis so that 75%

of the individuals within this strain will be susceptibile to IDD.

The genomes of C57BL/6 and the Fl hybrid are intrinsically much less

permissive for the development of IDD and consequently are positions

further away from the liability threshold. The lower panel

illustrates the inheritance of IDD among the BC1 progeny produced

in our cross, based on their content of susceptibility modifiers

(from Table 3). Each incremental increase in susceptibility alleles

in individual animals moves them closer to the liability threshold,

and when a population of animals with a similar content of

predisposing modifier alleles are analyzed, a characteristic

penetrance will be observed.

This illustration, which was adapted from Sewell Wright's

original paper on threshold inheritance, is only used to clarify

the concept of a liability threshold. Our modeling of the

inheritance of IDD susceptibility among the BC1 progeny in our cross

is consistent with the inheritance of diabetes as a threshold

liability, however, definitive proof will require the analysis of

IDD penetrance in a collection of congenic and polycongenic strains

carrying variable numbers of susceptibility modifiers. These

studies are currently underway.

The inheritance of IDD as a threshold liability predicts

several interesting features concerning its analysis in outbred

human populations. First, susceptibility will be genetically

heterogeneous, and as a result, different genes will be segregating

in individual families. As a result, mapping studies, which require

that the data from many families are pooled for analysis, will often

yield conflicting results. Furthermore, the mode of inheritance

(dominant versus recessive) will vary depending upon the specific

genes) segregating within individual families. Finally, since many

different genotypes are liable to diabetes (each with a specific

penetrance), no single IDD susceptible genotype will be shared by

all diabetics. These features will undoubtedly complicate the

analysis of such traits in outbred populations.

TABLE 2-1. Linkage of marker loci with diabetes and insulitis in
the (NOD X B6)F1 X NOD BC1 mice.

hom:het X2 p Z-scoreb p
1 D1MIT4 26:18 2.77 0.0056
6 D1Nds4 27:17 1.76 0.079
30 D1MITS 31:17 1.66 0.096
11 D1MIT8 26:22 3.15 0.0017
3 Bcl-2 24:24 3.56 0.00038
5 RP154B 14:14 2.76 0.00572
5 D1MIT31 24:20 2.42 0.015
17 Ly22 11:16
0 PolE 12:15

3 D1MIT15 18:30
20 D1MIT17 20:24
2 D2MIT7 24:20

36 D2Ndsl 26:18
14 D2MIT17 33:15
20 D2MIT47 36:12 6.7 0.01
1 D2MIT26 36:12 6.7 0.01
9 D2MIT48 27:17
15 D2MIT229 25:23
3 D3MIT1 40:8 16.2 0.00006 2.85 0.0044
4 Glut2 25:3 15.6 0.00008 1.94 0.052
4 IL2 42:6 21.4 <0.00001 2.49 0.013
2 D3MIT5 42:6 18.8 0.00001 2.58 0.0099
4 D3Ndsl 42:6 16.1 0.00006 2.42 0.016
18 D3MIT10 40:8 14.1 0.0002 2.96 0.0031
1 Tshb 41:7 15.4 0.00009 2.94 0.0032
2 RP154A 24:4 8.3 0.004 2.75 0.0059
17 Adh 33:15 2.58 0.0098
15 D3MIT19 27:17
4 D4MIT1

33 D4MIT17 15:33 6.7 (het)' 0.009
45 D4MIT12 25:19
15 D4MIT33 29:19
D4MIT59 30:18
5 D5Ndsl 22:22

4 D5MIT81 19:25
47 D5MIT30 19:22
32 D5MIT191 19:22

Table 2-1 -- continued

hom:het X2 p Z-scoreb p
6 37 Mtv8 23:16

0.1 Ly3 17:10
41 D6MIT10 32:16
11 D6MIT13 31:17
16 D6MIT14 30:18 5.8 0.016
0.6 Kras 17:11

7 CKMM 26:22
7 D7MIT55 24:20
32 c/D7MIT30 27:21
29 D7MIT105 26:22

8 D8MIT94 30:18

4 D8MIT4 30:18
25 D8MIT104
6 MT2 23:21
33 D8MIT140 25:22

9 Ld1r 13:10

5 Mets 13:11
3 D9MIT2 27:21
4 Thyl 15:9 -2.49(het) 0.013
3 RP154C 16:8 -2.44(het) 0.015
1 Ncam 18:10
19 D9MIT8 24:20
20 Gnai2 9:15

10 D10MIT42 23:21

53 D10MIT2 32:16
5 D10MIT28 27:19
11 Glns 26:18

14 Csfgm 30:14
6 PolB 18:10
13 DllNdsl 33:11
17 Gfap 37:11 11.5 0.0007
5 D11MIT42 38:10 14.0 0.0002
12 Ode 25:19

28 D12MIT141 20:24
18 Mtv9 13:14

13 D13Ndsl 22:22
7 D13MIT39 25:19
27 D13MIT74 25:21

Table 2-1--continued

hom:het X2 p Z-scoreb p
14 D14MIT11 16:26
3 Plau 18:26
18 D14MIT18 19:29
15 Hpg 19:29
25 D14MIT95 21:27
15 D15MITS5 28:20
21 D15MIT29 27:17
18 D15MIT108 30:15
7 D15MIT35 26:18
16 D16Nds2 13:11
3 D16MIT4 26:18
20 D16MIT19 24:20
17 D17MIT18 44:4 37.2 <0.00001
1 Sod2 27:0 33.6 <0.00001
16 H2 48:0 48.6 <0.00001 3.48 0.00051
3 D17MIT10 46:2 37.3 <0.00001 3.54 0.00040
26 D17MIT42 35:9 11.8 <0.00001 2.02 0.043
18 D18MIT4 18:10
>50 D18MIT17 29:15
19 MtvB 18:9
7 D19MIT59 26:22
30 PolA 18:10
20 D19MIT1 24:20

Diabetes X2 comparing homozyogote (hom) to heterozygote (het) ratios among
diabetics to those among typed non-diabetics

b Z-scores >2.0 for a Rank sum test comparing homozygotes to heterozygotes using
insulitis scores as described in the text

c het excess of heterozygotes over homozygotes

Table 2-2. Statistical analysis of Idd loci contributing to
diabetes susceptibility by combining data from this study and that
of Ghosh et al. (1993) [137]

IDD Genetic Ch Diabetics y2
Locus Marker This study Ghosh et al.
Hom : Hetb Hom : Het

Idd-1 H-2 17 48 : 0 48.0

Idd-2 D9MIT2/Thyla 9 15 : 9 60 : 43 4.2

Idd-3 D3NdS1 3 42 : 6 88 : 18 73.0

Idd-4 D11MIT42/DIlNdsl 11 38 : 10 73 : 33 25.6

Idd-5 DIMIT5 1 31 : 17 64 : 42 8.4

Idd-6 D6MIT14/D6MIT10 6 32 : 16 69 : 37 12.6

Idd-7 Ckmm 7 26 : 22 33 : 70 7.2

Idd-8 Plau 14 18 : 26 31 : 74 17.3

Idd-9 D4MIT59/D4Ndsl6 4 30 : 18 23 : 8 9.2

Idd-10 Tshb 3 41 : 7 81 : 25 52.6

Idd-13 D2MIT26 2 36 : 12 58 : 42 10.8

a If the two studies did not use the same markers, the nearest peak
markers are given (our study/Ghosh et al.)

b Hom:Het is the ratio of homozygotes to heterozygotes in the

I I I 59 I I

0. 0 0 H 0 0 00 ,1 f rC l H 0 O H 0 N CO Oj 0

0) II I I


o I I I I I
.0) o I o .I I

a w l m m l m m |m l i
4 I
1 1 I I
*I 1 I I I I
0 H 0 o 0 0 H H a 0 0 0 0 0

I I I m m m I


a m l I I II r- i ri m 1 2 1
X m a m m i i m z m Mi

0 I

m I m m m I z m z m m z m m I
-,A 4J I I I I

I 1 1 1

m ml I z
0 00

^ 0)1t 10. 106.1p

En u




tu r



*H 44






*3 a
*d H

H A)


H 9


0l0 I a 0 m Cl I A

A m H l r i L r1 r0jl m| l

m z z I I m z z I I z




B R a I a a l I n a Il


m a a a z a mi I a z a z a


S I a I
A A LA I'. a 1I)I

a l I I

ii I


m A
44 ua)
14 01a)










0 H


H .0

0 0

u a)

O rl

4 0H

u "
*H f0
0)0 0

m a'
a .0
8H H
.0 O



0d ~
OH ?.'0

14 H
0 I

0- ^
0 '0 a)
a) 1

|a) |
a) a)I
a)4 44^
E-i 8 -
0oJ a)

9- II

ll s I


ii i !

| I ^ | 1
N? N

*^-j^. ^
t r" 9

,~ NN N N

0t c~~

p 9

N. N

_ __l 4- U
i^ C nM *H .-l
4 u ,
H0 .

4j 0
5o 5 9 u ac rl
__ __ lC.) 4J -H H

-- UlP U fl) Hi
0 0 N

s s a n n -r
Jct S S m cr
C I* ^ 0 ra

M0 0.
U- 0 -l
.0 3

x? r m a

0 04w
B w u
Q 010

ea ?< e ) fu
4j H~
r= 4~

r r 4
0:4 0 d (

0 4 ) u

m 0. 0

C, 9H Ido
"4 4J 5 4
^ 0 0 w m
Nm N
.0C PAi
OH 0A -
a P Pr 01" 4
NOP N >.
r3 PA '
o u-
^_______T ____ TW PAS
________ 0'0.0e~

^^ ^P ^ Nn a 4JP.O

U4-) a) M -r -
Hr Ea 0 l
-H 4.) 0 4)

0 o ,Q to
O 4 ) o o)

1 -1 .O O>
) 0b r-1 IH

H 10 0 0
0 0 u

a 0 0 0i

tO OM 44 O
S)0) 0" o-

41 t -H a

S'o RH 0
4404- t

f o 4J
1u 00

4 1 ( a
0 0

a i) 4 ED 0

i-'- (0 m 0 0 .
. f 0y1 E- -H
N ^ r-t ) 1) (





Q 0
I -



9' 0'0 ?'0 0"0


*C C57BU6 F1 NOD


.C57BL/6 F NOD

0 1 2 3 4 5 6


Figure 2-3. An illustration of inheritance of IDD susceptibility as
a threshold liability in which increasing numbers of susceptibility
modifier alleles increase the penetrance of disease (see text for



The recent availability of PCR-based microsatellite markers

for genome-wide genetic screening has revolutionized the mapping of

genetic intervals involved in simple Mendelian and complex traits

in humans and animal models [131-133]. Complex traits are typically

multifactorial (influenced by environmental and genetic factors) and

polygenic (many genes are involved). Complex traits that have been

mapped in animal models include hypertension in rats [159], and

insulin dependent diabetes (IDD) [136,137], cancer susceptibility

[134], and systemic lupus erythematosus (SLE) [135] in mice. One

of the greatest difficulties in mapping genes involved in complex

traits is that they do not follow a simple Mendelian mode of

inheritance and cosegregation is not perfect between markers and the

trait [160]. For this reason the intervals cannot be narrowed below

approximately 10-20 CM with reasonable numbers of backcross or

intercross animals because each recombination is not necessarily



One approach to identification of the genes in identified

genetic intervals is the evaluation of candidate genes for

mutations, aberrant expression and altered function. The large

interval sizes, the low density of cloned genes in most regions of

the mouse genome, and the unknown functions of each of several

susceptibility genes make this method unlikely to succeed. Even if

candidate genes differ in sequence or expression and function

between diseased and normal mice, demonstration of involvement in

the complex trait is more difficult, and will still require a

genetic approach, such as the use of mice with gene knockouts or


An alternative approach to dissection of complex traits is

through construction of congenic mouse strains. Congenic mouse

strains have been bred and used to understand the functions of

individual genes separate from genetic background, since Snell

produced a series of congenic lines that differed from one another

at the MHC [161]. These mice were crucial for demonstrating the

role of the MHC in immune responsiveness [162]. Congenic lines are

produced by the repeated backcrossing of mice with an allele of

interest to mice of another genetic background then intercrossing

to obtain a homozygote line, such that the two lines should differ

only (or nearly so) at the selected locus and unselected flanking

loci (reviewed by Flaherty [163]).

Congenic mouse strains constructed with genetic intervals can

facilitate the investigation of the contribution of each

unidentified gene, alone and in combinations, to the trait. In

addition, the interval can be fine mapped allowing the exclusion of

some candidate genes and making positional cloning feasible. For

these reasons the approach we used to genetically dissect the

inheritance of insulin dependent (Type 1) diabetes (IDD) was to

produce a series of congenic mouse strains, each with a genetic

interval found in our backcross to contain as yet unidentified IDD

susceptibility genes) (Chapter 2).

IDD is an autoimmune disease with complex inheritance. The

non-obese diabetic (NOD) mouse develops spontaneous autoimmune IDD

which closely mimics human disease (Makino et al. [117]). Based on

mapping studies with the NOD mouse and various diabetes resistant

strains of mice, 14 genetic intervals have been mapped which

contribute to diabetes in various crosses [105,130,136-140,142-144].

Our cross identified eight NOD-derived susceptibility genes in six

linkage groups. Only one of the intervals, on chromosome 17, was

required for disease while the other intervals were neither required

nor sufficient to cause diabetes. Because of the complex mode of

inheritance and because at least 2 genes are present on chromosomes

1 and 3, the genetic intervals for identification of the

diabetogenic genes cannot be unambiguously narrowed using backcross

animals. Therefore, to dissect the inheritance of diabetes, we

produced a set of congenic mouse strains containing the NOD-derived

diabetogenic genetic intervals on a resistant, C57B1/6, background.

Congenic mouse strains can be made in two ways, either by

breeding resistant intervals onto NOD and looking for loss of

phenotype, or by breeding the diabetogenic intervals onto a

resistant strain and looking for a gain in phenotype. The former

approach was used by Prochazka et al. [105] with ThyJ--linked Idd2

on chromosome 9, by Wicker and co-workers [140,141] with chromosome

3 containing I4d3 or Iddl0 or both, and by Serreze et al. [143] with

Iddl3 on chromosome 2. In all cases the result of the loss of the

NOD-derived intervals was a decrease in penetrance, but not complete

elimination, of diabetes. Although it is a slower approach, we

chose to breed NOD-derived intervals onto resistant (C57B1/6) mice

for several reasons. With these congenics we should be better able

to determine the functions and pathogenic effects of each locus on

the development of insulitis and diabetes. Interactions between the

genes can be studied by combining intervals in polycongenics. Based

on observed phenotypes we may be able to narrow the range of

potential candidate genes for a particular genetic interval. In

addition, these congenic mouse strains will allow us to study the

mode of inheritance of diabetes. If diabetes is indeed a threshold

trait (see Chapter 2), then increased numbers of susceptibility loci

should lead to diabetes with increasing levels of penetrance.

Finally, fine-mapping each interval should be easier with

observation of a gain, rather than a partial loss, of phenotype.


Mice were bred and maintained in the Department of Pathology

mouse colony at the University of Florida, Gainesville. The origins

of NOD/Uf have been described previously [154] and C57BL/6 mice were

obtained from The Jackson Laboratory, Bar Harbor, ME.

NOD-derived diabetogenic genetic intervals identified in our

previous linkage analysis were transferred onto a resistant

(C57Bl/6) background by six successive backcrosses followed by

intercross (Figure 3-1). Polymorphic PCR-based microsatellite

primers were obtained from Research Genetics (Huntsville, AL) or

synthesized at the University of Florida DNA synthesis core based

upon published sequences [131]. Tail DNA was extracted from tail

biopsies by standard methods. The PCR amplification was as follows:

100 ng tail DNA, 200 nM of each primer, 0.2 mM dNTPs, and 0.75 units

of Taq polymerase (Boehringer-Mannheim, Indianapolis, IN) in a

standard buffer with 50mM MgCl2. The optimal annealing

temperatures were determined for each primer pair and varied from

52-600C. PCR products were visualized on 5% agarose gels.

Large intervals, based on 95% confidence interval estimates

using MAPMAKER [155] were selected to ensure that we did not lose

the disease locus or loci. In cases where peak loci differed


between our study and those of studies by Todd and co-workers using

B10.H-2g7 mice [136,137], both peaks were included.

Results and Discussion

Seven congenic mouse strains were produced, each containing an

NOD-derived IDD susceptibility genetic interval (Table 1, Figure 3-

2). Chromosome 1 was divided into 2 large overlapping intervals

because we found 2 peaks for insulitis (D1MIT4 and Bcl2) and a

separate peak for diabetes (D1MIT5) in between those peaks (Chapter

2). The congenic mouse strain with the more centromeric interval,

B6.NODclc, encompasses the first insulitis peak and the diabetes

peak, while the strain with the more telomeric interval, B6.NODclt,

contains the diabetes peak and the second insulitis peak. B6.NODc3

includes both identified peaks affecting both insulitis and

diabetes, IL2 and Tshb. Because of differences in peak map location

for IDD susceptibility for chromosome 6 due to a segregation

distortion, both D6MIT10 and D6MIT14 were included in the interval

selected for B6.NODc6. The interval for B6.NODcll is large because

the peak reported by Todd and co-workers was DllNdsll while ours was

more telomeric (D11MIT42). For the chromosome 17 congenic strain,

B6.NODcl7, we included both H-2 and our peak for insulitis,


After six backcrosses (N7) the probability of an unlinked non-

selected NOD-derived marker being carried is 0.78%. To increase the


probability of loss of unwanted NOD genomic intervals, we also used

microsatellite markers to screen the backgrounds of each congenic

mouse strain at backcross 3, 4, and/or 5. First, mice with other

identified diabetes susceptibility intervals were excluded. Then,

2-4 other polymorphic markers on each autosomal chromosome were

tested (Figure 3-2) and mice with unwanted NOD-derived markers were

excluded from breeding in subsequent generations where possible. In

addition, the ends of the congenic intervals were mapped with

flanking polymorphic SSR markers (Figure 3-3).

Breeding large genetic intervals rather than a single locus

proved to be more difficult than breeding for a single allele

because of recombination within the interval at each generation.

In particular, at intercross, the heterozygote interval can

recombine in each parent, resulting in a very low percentage of

offspring that are homozygous for the entire NOD-derived interval.

If selecting for a single locus, 25% of the progeny should be

homozygous for the desired allele, however, for example, with the

chromosome 3 interval that is 43 cM long, the probability of progeny

with the desired homozygous interval is only about 6%. In addition,

both a male and a female are required for establishment of a

congenic line. As a result, establishment of congenic lines for

the larger intervals required many breeding cages and typing of

large numbers of offspring.

In addition, some lines were harder to establish than others

for other reasons. There was an expected proportion (6%) of males

from the B6.NODclt intercross that were homozygous for the NOD-

derived interval. However we had a lower than expected proportion

of females (40%), as well as a low percentage of females homozygous

for the NOD interval (3%). In addition, several of the homozygotes

died early or were infertile, suggesting the presence of a

deleterious locus within the NOD interval differentially affecting

females. That the NOD mouse does not show this effect indicates the

possibility of coevolved genomes, in which the effects of

potentially deleterious genes might be masked by other genes.

These congenic mice constructed with diabetes susceptibility

genetic intervals should be very useful in dissecting the

pathogenesis of diabetes. Some B6.NODclc, clt, c3, and c17 congenic

mice already have some periductal/perivascular and periislet

infiltrates that indicate that loci within these intervals are

sufficient to reproduce the very earliest stages of insulitis

(Chapter 4). By combining various intervals we should be able to

recapitulate insulitis and diabetes and assess the role of each gene

in producing disease. Each interval combined with chromosome 17

(MHC), which is required for diabetes and is correlated with

insulitis, and will allow for immunological studies. Determination

of which polycongenics result in insulitis and diabetes will provide

insight into the mode of inheritance of diabetes. We can test the


threshold model of diabetes inheritance wherein addition of

diabetogenic genes increases the penetrance of disease. Finally,

congenic mice with phenotypes can be used to fine map the

diabetogenic genes. Narrower intervals will allow better

exclusion/inclusion of candidate genes and positional cloning of

the genes. This approach should be useful for the genetic

dissection of diabetes as well as other traits with complex

inheritance and is also currently being used in our lab for the

analysis of the inheritance of systemic lupus erythematosus [135].


Table 3-1. Congenic mouse strains created, Idd designations,
markers used to select the ends of the intervals, and the interval
sizes selected.



B6.NODclc Idd-5 DIMIT3-Bcl2 44

B6.NODclt Idd-5 DIMIT5-DIMIT15 47

B6.NODc2 Idd-13 D2MITI7-agouti 27

B6.NODc3 Idd-3, 10 DIMIT132-Tshb 43

B6.NODc6 Idd-6 D6MIT54-D6MIT14 24

B6.NODc11 Idd-4 Csfgm-DllMIT42 42

B6.NODC17 Idd-1 D17MIT21-DI7MITIO 19



N1 (50% B6)

N2 (75% B6)

N3 (87.5% B6)

N4 (93.8% B6)

N7 (99.2% B6)



nb bb C 57BI/6

nb bb



nn nn nb nb bb

Congenic Strain

Figure 3-1. Illustration of the breeding strategy used in the
construction of congenic mouse lines ( N = breeding generation;
n = NOD allele; b = B6 allele). In parentheses is the average
percentage of the mouse genome that should be B6-derived.

SI I f



~! I

~ L

I [

m ammm
l s s

-J -~sBsssBsi-i- ;


0 4 0



.0 ( 0
4-1 H

(4 0
0 .) -1,
4- 1
0 1

01 1 4J


a u V
O mH
Ek *Hl -


cl1 c17





- DnDS4

- DfMIT212





- DIT2
- 11-2


- Thhl



- D6MIT13





Figure 3-3. Diagram of genomic intervals selected for the congenic
strains (gray boxes), unselected flanking intervals known to be
included (black boxes), and regions of recombination in which the
congenic interval ends (lines). Markers between which the interval
ends are in bold.

- D17MIT3S6
- H-2
- D17MIT10



The immunopathology and genetics of human IDDM is closely

mimicked by a spontaneous diabetes mouse model, the non-obese

diabetic (NOD) mouse [117,149], making it an ideal model for

detailed study of the factors involved in diabetes. Our strategy

for the study of inheritance of diabetes susceptibility was to first

map the genetic intervals containing genes involved in NOD diabetes

susceptibility by linkage analysis in a backcross between NOD and

a resistant mouse strain, C57BL/6, using polymorphic simple sequence

repeat (SSR) markers located throughout the mouse genome (see

Chapter 2). To analyze the roles of individual susceptibility

genetic intervals identified in our cross, a series of congenic

mouse strains were produced, each carrying a diabetes susceptibility

interval on a resistant genetic background (see Chapter 3).

Other studies using congenic mouse strains to study the

inheritance of IDD have used the opposite approach, that of breeding

resistance alleles onto the NOD background. With that approach, a

decrease in the frequency of diabetes can demonstrate the importance

of the allele in diabetogenesis. However this approach does not

allow further characterization of the pathogenic role played by the

unidentified genes in IDD. In addition, for purposes of eventual

mapping of the susceptibility genes, partial loss of phenotype is

not as good as a gain of phenotype.

Wicker et al. [108,110] and Prochazka et al. [109] bred the MHC

from B10 and NON mice respectively onto the NOD background, and none

of the NOD.H-2b or NOD.H-2"1 congenic mice became diabetic, thereby

demonstrating that the unique NOD H-20 is required for diabetes.

Although H-297 had been postulated to be recessive for IDD, a few

(approximately 3% of females) H-2'1/b heterozygotes became diabetic

[108]. Wicker et al. [111] termed this inheritance "dominance with

low penetrance". In addition, NOD.H-2b and NOD.H-2"' congenic mice

had only perivascular/periductal and periinsular infiltrates, no

insulitis, demonstrating that NOD.H-29" is required for beta cell

autoimmunity. The NOD H-2 is dominant for insulitis as

approximately 50% of H-20'/b heterozygotes exhibited insulitis and

25% of females became diabetic after treatment with

cyclophosphamide. Prochazka et al. [109] also bred an NON-derived

IDD resistance locus onto the NOD background using a single linked

marker, Thvl. Initially this locus did not appear to effect the

entrance of diabetes, however, subsequently it was found that the

allele is weakly codominant for diabetes susceptibility (unpublished

data, discussed by Wicker at al. [111]).

Wicker and co-workers also used congenic mouse strains to

demonstrate the importance of the two IDD susceptibility loci on

chromosome 3, Idd3 and Iddl1 [140,141]. Idd3 and Iddio were

identified to be very strong, linked loci in crosses between NOD and

B10.H297 [137] and C57BL/6 (Chapter 2 and Cheng [130]). Congenic

mice with both reccessive alleles from B6 (NOD.B61-T-"M)or B10

(NOD.B6L2-"~m) had female incidences of diabetes of 1.2% and 2.8%

respectively while the entrance of diabetes among NOD females in

the colony is 77.8%. Congenic mice with B6-derived resistant

alleles of Idd3 (NOD.B6') and IddlQ (NOD.B6T~) had female

incidences of diabetes of 28.4% and 32.9% respectively. The authors

concluded that the non-NOD-derived Idd3 and Iddil genes act

synergistically in conferring diabetes resistance. Insulitis was

still present in NOD.B62"-""T although at a much lower rate.

NOD.B6B2 congenic mice exhibited a lower incidence of insulitis than

NOD but higher NOD.B6'2-Tl" mice. The incidence of insulitis in the

NOD.B6"S congenics was not very different from that of NOD mice,

indicating that the allele may not contribute strongly to insulitis.

Serreze et al. [143] also used congenic mice to demonstrate the

presence of a chromosome 2 diabetes susceptibility gene.

Replacement of that interval with one from C57BL/KsJ decreased the

penetrance of diabetes from 100% to 33%.


In this study, the congenic mice constructed as described in

Chapter 3, were examined to determine whether any genetic interval

by itself could produce pancreatic pathology that might suggest its

role in diabetogenesis. These B6 congenic mice with NOD-derived

susceptibility loci show that genes influencing increased

trafficking of leukocytes into the perivascular/periductal areas of

the pancreas are found in the NOD genetic intervals on chromosomes

1, 3 and 17.

Materials and Methods

Mice were bred and housed in the Department of Pathology mouse

colony at the University of Florida. Derivations of the congenic

strains, B6.NODclc, clt, 3, 6, 11, and 17, as well as NOD/Uf and

C57BL/6 mice, are given in Chapter 3 and Winter et al. [154].

For histological examination, pancreata were removed from

congenic mice sacrificed at 6 months of age or older and fixed in

10% neutral buffered saline for 24 hours for paraffin embedding and

sectioning. At least 3 non-contiguous sections were cut and stained

with hemotoxylin and eosin for microscopic examination. Pancreata

were scored for leukocytic infiltrates as follows: 0 = no

infiltrates; 1 = few, small perivascular/periductal monocytic

infiltrates; 2 = more extensive or frequent perivascular/periductal

infiltrates; 3 = perislet infiltrates; 4 = intraislet infiltrates.

At least 15 islets were checked for each mouse.

Results from histological examination of pancreata from 6

month old B6.NODcl, 3, 6, 11 and 17 congenic mice are shown in

Figure 4-1 and summarized in Table 4-1. B6.NODc6 did not show any

pancreatic infiltrates at 6 months of age (Figure 4-la). B6.NODcll

mice exhibited few (usually only 1) very small monocytic infiltrates

at a level similar to what is seen in control 8 month old B6

parental mice (Figure 4-1b). With B6.NODcl7, clc, and clt mice,

about half had a few small to large focal infiltrates, none of which

focused on the islets (Figure 4-1c,d,e). Inheritance of these

intervals may be codominant as heterozygotes also exhibited some

infiltrates. B6.NODc3 mice, which have at least two genes involved

in diabetes and insulitis, exhibited small to large and more

frequent periductal/perivascular infiltrates in about half of the

6 month old mice. In addition, two mice were found with single

periislet infiltrates (Figure 4-1f). The genes) apear to be

recessive as heterozygotes did not show any infiltration (Table 4-


When older congenic mice (8-16 months) from some of the

congenic strains were examined (Table 4-1), the perivascular/

periductal infiltrates were more extensive but still no intraislet

infiltration, true insulitis, was observed. Although the


infiltration can become quite extensive in some old B6.NODc3 mice,

it still typically remains perivascular/periductal (Figure 4-lg).

The same was true for B6.NODclc mice (Table 4-2). Older B6.NODc6

mice exhibited some very small and infrequent

perivascular/periductal infiltrates similar to those seen in B6


These results indicate that several of the diabetogenic

intervals on chromosomes 1, 3, and perhaps 17 affect non-specific

infiltration into the pancreas. This fits with the results from the

backcross mapping study (Chapter 2), wherein chromosomes 1, 3 and

17 correlated with insulitis while chromosomes 6 and 11 did not.

Insulitis was never observed and only in the case of B6.NODc3 was

there an occasional animal with periislet inflammation, suggesting

that there is no beta cell-specific autoimmunity being generated in

these mice to focus the inflammation on the islets.

A proposed model for the roles of individual genes in IDD

pathogenesis is given in Figure 4-2. Two main factors that should

influence insulitis and thereby diabetes are (1) the mice do not

delete or otherwise inactivate beta cell-specific T-cells and (2)

these cells have the opportunity to come in contact with their

antigen in the pancreas. A gene (or genes) on chromosomes 1, 3,

and perhaps 17 may be involved in abnormal lymphocyte trafficking

and/or recruitment which allows more contact between the immune

system, low frequency autoreactive T cells in particular, and

pancreatic antigens. However, without the other NOD diabetogenic

genes, autoreactive T-cells have probably been eliminated through

normal tolerance mechanisms. Because the NOD MHC on chromosome 17

is so important for insulitis and diabetes, and because of the

antigen processing, presentation, and tolerance induction functions

of MHC molecules, it is likely that beta cell-specific autoreactive

T cells require addition of at least the NOD H-297. Results with

NOD.H-2b and NOD.H-2" congenic mice which get perivascular,

periductal and periislet infiltrates but rarely insulitis, support

this idea [108-110]. Combining chromosome 17 with chromosome 1

and/or 3 may result in an immune response to beta cell antigens.

These hypotheses can be tested by the construction of polycongenic

mice with two or more intervals which is currently underway. In

addition, the IDD genetic intervals on chromosomes 1, 3 and 17

probably contain 2 or more susceptibility loci (reviewed in Chapter

1). For that reason, congenic mice with smaller intervals for

chromosome 1 and 3 are being constructed to separate the different

susceptibility loci, to determine which loci are responsible for the

increased pancreatic infiltration, and to eventually map the genes.

These results with the B6.NOD congenic mice show that they

will be useful in reproducing the various stages in the pathogenesis

of diabetes and will allow for the dissection of the roles of each

gene. They also should permit mapping and eventual identification

of the IDD susceptibility genes.

S)i 0W X
U> 0 U

4 H -1 r-H


Sa nQ .

P 4
"o .. S J
fto m



; 01

-% ,..,

.a -

''..t.. a "a' ,

2 'r :/ -'
... i. ,. :
i~1 ', ._- ..

'~ p- ....,, -



L. A'rII
* g
r 1'3.$d

? -

Pr1 ^2^ -:.~ *' '1'*"*

': sff'^ '':^
-T WIT,..



C1, 3, AND/OI





R 17?



Figure 4-2. Proposed model for diabetes pathogenesis.


Table 4-1. Leukocytic infiltrates in pancreata from B6.NOD congenic
(NN) and (B6.NOD X B6)F1 (BN) mice at 6 months of age

Congenic Locus Genotype Phenotypes
line NN BN observed,

B6.NODclc IddS 2/3b 3/5 0,1,2

B6.NODclt Idd5? 8/13 0/2 0,1,2

B6.NODc3 Idd3,10 8/19 0/10 0,1,2,3

B6.NODc6 Idd6 0/15 0/4 0

B6.NODc1l Idd4 3/13 0,1

B6.NODc17 Iddl 13/26 2/3 0,1

a Phenotypes based on pancreatic histology scored as follows: 0=no
infiltrates; 1=few, small leukocytic perivascular/periductal
infiltrates; 2=more extensive or frequent infiltrates; 3=periislet
infiltrates; 4=intraislet infiltrates.

b Number positive/total number of mice scored.


Table 4-2. Leukocytic infiltrates in pancreata from older B6.NOD
congenic mice at >8 months of age.

Mouse Locus # positive/ Phenotype
strain total # observed

B6.NODclc Idd5 3/6 0,1,2

B6.NODc3 Idd3,10 11/16 0,1,2,3

B6.NODc6 Idd6 4/17 0,1

B6 3/8 0,1

SPhenotypes based on pancreatic histology scored as follows: 0=no
infiltrates; l=few, small leukocytic perivascular/periductal
infiltrates; 2=more extensive or frequent infiltrates; 3=periislet
infiltrates; 4=intraislet infiltrates.