Diabetes mellitus John A. Todd O x f o r d University, Oxford, UK Developments on four fronts have contributed to an exciting year for the study of diabetes. These include advances in molecular genetic mapping, analysis of animal models of disease, understanding of disease phenotype, and the extension of statistical methods to the study of complex, non-Mendelian traits. Current Opinion in Genetics and Development 1992, 2:474-478
Introduction Clinical diabetes mellitus is defined as defective control of glucose homeostasis as a result of complete or partial insulin deficiency and can be broadly divided into three main types: type 1 or insulin-dependent diabetes mellitus (IDDM); type 2 or non-insulin-dependent diabetes mellims (NIDDM); and maturity-onset diabetes of the young (MODY). Type 1 diabetes is a T lymphocyte-dependent autoimmune disease of the insulin-producing [B-cells of the pancreas and is characterized by absolute insulin-dependency, childhood-onset (mostly less than 18 years) and the association of disease with the presence of certain alleles of the human leukocyte antigen (HLA) class II genes, HLA-DQ and HLA-DRWIthin the major histocompatibility complex (MHC). Type 2 diabetes affects up to 5% of global populations and is ten times more common than type 1 diabetes. The disease is not associated with autoimmunity or HLA markers, has a late 'age of onset' (mostly over 50 years of age) and is caused both by defects in the secretion of insulin by 13-cells and by defects in the ability of tissues to respond to insulin (insulin resistance). MODY is a form of NIDDM that usually occurs before the age of 25, accounts for only 5% of all diabetes, and has an autosomal dominant mode of inheritance that is in contrast to the other two types, the inheritance of which is not clear. In 1991, the first MODY gene was mapped [ 1 - ] using PCR technology and a variable number of tandemly repeated sequence or microsatellite marker locus on human chromosome 20. A large number of microsatellites have been developed as polymorphic markers in the mouse genome and used to map several type 1 diabetes loci outside the MHC [2.°]. The accumulation of these highly polymorphic markers in humans promises to accelerate the localization of further diabetes genes. In addition to requiring a highly informative map with markers at every 5 cM, gene mapping of complex dis-
ease also requires large numbers of nuclear pedigrees [3 °°] or, in the case of MODY, very large multigeneration pedigrees. The key papers [ 1 - - - 3 o,] highlight the exciting discoveries that await us over the next few years. In tl~is review, recent events are described in the context of advances in molecular genetic mapping, analysis of animal models of disease, understanding of disease phenotype, and the extension of statistical methods to the study of complex, non-Mendelian traits.
Advances in gene mapping The ability to map genes for complex traits depends mainly on the resolution of the map of the genome, that is the spacing and informativeness of the marker loci, the number and types of pedigree available and the amount of locus heterogeneity, that is the number of unlinked loci that can cause the disease and the degree to which they can be classified as essential. The development of microsatellites, which occur at least every 50kb in the human genome  and display size polymorphism that is the result of variation in the number of simple sequence repeats, first identified in human  and then in mouse , provides us with the ability to construct 5cM genome maps. Risch  demonstrated that cost-effective linkage studies using affected relative pairs and identity by descent (IBD) required that marker loci possess polymorphism information content (PIC) values greater than 0.7, equivalent to a marker locus with four alleles with equal population frequencies. At least 30% of microsatellites have PIC > 0.7 . Bell et al. [1 o-] tested the linkage of 79 distinct DNA markers, which included at least one marker on every autosome except chromosome 18, to MODY in one very large pedigree. One of these marker loci, a tetranucleotide microsatellite repeat located in an Alu element in the adenosine deaminase gene on chromosome 20,
Abbreviations HI.A--human leukocyte antigen; IBD--identity by descent; IDDM--insulin-dependent diabetes mellitus; MHC--major histocompatibility complex; MODY--maturity-onset diabetes of the young; NIDDM--non-insulin-dependent diabetes mellitus; NOD~non-obese diabetic; PIC--polymorphism information content.
(~ Current Biology Ltd ISSN 0959-437X
Diabetes mellitusTodd was tightly linked to MODY in this family. The study demonstrates the successful use of exclusion mapping in scanning the genome for a gene causing a monogenic trait. It also highlighted the care that must be taken in clinical diagnosis and inclusion criteria of affected individuals because this MODY family contained both IDDM and NIDDM cases, and inclusion of these would have lowered the lod scores obtained. (The lod score is determined by calculating the probability of obtaining the phenotypic data observed in a pedigree as a function of the parameter(s) of interest.) The definition of MODY is not clear cut, and perhaps in the presence of modifying genes, NIDDM may look like MODY and vice versa. The demonstration of this linkage is an important step in the genetics of diabetes. The next questions include how many other MODY pedigrees are affected as a result of mutation in this chromosome 20 gene, and what is the function of the gene product in [3-cell function and carbohydrate metabolism? Even if this gene contributes little to determining the overall frequenw of MODY, its characterization and the mapping of other loci in other MODY pedigrees will provide considerable insights into the aetiology of diabetes.
Development of linkage methods for complex traits The most important recent publications on this topic have been reviewed by Risch , but the paper by Hyer el aL [3"] is specifically relevant to the genetics of type 1 diabetes. These authors [3 °°] tested for linkage to human chromosome l l q based on the observations that chromosome 9 is weakly linked to IDDM in the non-obese diabetic (NOD) mouse, and that mouse chromosome 9 is homologous to human chromosome 11q. A set of 17 different polymorphic marker loci located at an average distance of 20 cM along the region of homology of chromosome 11q with mouse chromosome 9 were used in multipoint linkage analysis using the affected sib-pair method (where the affected pairs of sibs are exanlined for IBD at the marker locus). Importantly, they showed that the power to detect linkage (lod score > 3) depends on the IBD probabilities for allele sharing at the marker locus. The maximum lod score was < 0.5 for all marker loci tested. The question then asked was to what extent can the possibility that an IDDM susceptibility gene resides in the region be excluded? For all genetic models in which the probability of sibs sharing two alleles IBD at the presumed locus is P2 > 0.5, more than 90% of the region between the two most extreme markers can be excluded at a lod score < - 2 . Finally, theoretical affected sib-pair sample sizes were calculated to detect or exclude linkage. For example, the detection of a susceptibility gene with an effect on IBD probabilities equivalent to that of HLA loci would require a mean of 25-50 families with fully informative markers at 20 cM spacing. A sample of 200 affected sib-pairs may be adequate to exclude linkage for genetic models under IBD probability for P2 > 0.3. These analyses show that with a highly in-
formative map and 200 affected sib-pairs, it should be possible to detect IDDM susceptibility genes that exert 2-3-fold weaker effects than HI.A. These conclusions are consistent with previous theoretical estimations of the number of families required to detect gene effects that are weaker than or equivalent to HLA effects . Unfortunately, very few marker systems, with the exception of a few variable number of repeat sequences and the HLA gene complex itself, are full), informative. To date, there are over 150 microsatellites published with PIC >0.7 , but most of these are PIC <0.8. In our hands, marker loci with PIC = 0.7, only make about 50% of sib-pairs fully infomlative. Clusters of physically linked microsatellites at a density spacing of greater than 5 cM will be required to fully utilize valuable family material and to detect susceptibility genes with effects as weak as the insulin gene region susceptibility determinant(s) on human chromosome 11p15 [11,12]. Risch [13 o] has shown that contrary to recent claims, testing of multiple markers in a genome search for susceptibility loci actually decreases the false-positive rate among significant tests. In contrast, testing multiple genetic models by varying either the mode of inheritance parameters or the diagnostic criteria can increase the posterior falsepositive rate. In the latter case, Risch suggests subtracting logl0t from the maximum lod score, where t is the number of different genetic models and/or diagnostic models tested. In the stud), of complex disease, exclusion mapping must be combined with candidate gene sequence analysis. This concept is clearly defined by Sobell et al.  who advocate the sequencing of candidate genes in a systematic way. This approach is complementatT to linkage analyses and will be greatly facilitated by the large scale sequencing of cDNAs in world-wide genome projects. The candidate gene approach has been very successfully applied in the case of diabetes, leading to the identification of both the HI.A.and insulin gene regions as susceptibility loci for type 1 diabetes, and the insulin receptor as a cause of insulin resistance in type 2 diabetes . In human type 1 diabetes, there is increasing evidence for gene interactions between insulin, HLA and disease [11,16], and also between disease, the T-lymphocyte receptor [3-chain locus (chromosome 7q35) and the immunoglobulin heavychain region genes [17, 18].
Identification of subphenotypes in disease One of the greatest fears that investigators of complex disease have is that the genetics will be so complex and the disorder so heterogeneous that resolution is almost impossible unless one is lucky enough to select a candidate gene polymorphism that is associated with the disease. For IDDM and NIDDM, segregation analyses have not revealed any clear mode of inheritance. One solution to this problem is tO continue to analyze the biochemistry and pathology of the disease, and thus attempt to identify a characteristic that can be accurately and con-
Mammaliangenetics venienfly measured in patients and relatives, and which might be inherited in a less complex way. Mosthaf et al. [19"] describe a tissue-specific alteration of the insulin receptor RNA splice pattern in NIDDM patients. The insulin receptor exists in two isoforms, which are generated by alternative splicing and differ by a 12 amino acid insertion sequence in the ~x-subunit. Expression of the isoforms varies according to tissue type and each has a different affinity for insulin. In normal individuals, skeletal muscle biopsies show predominant expression of the high affinity isoform RNA as revealed using PCR, whereas 10/10 NIDDM patient biopsies showed approximately equal levels of RNA for the two isoforms. The authors speculate that this may be related to the insulin resistance evident in diabetic skeletal muscle. This characteristic, if confirmed, may provide a way to dissect diabetes by genetic analysis of a subphenotype. The NOD mouse spontaneously develops an insulindependent autoimmune diabetes with many similarities to the human disease. The NOD mouse thus offers experimental routes via which subphenotypes or components of the disease may be identified and analyzed. One of the most interesting recent examples of this utility is the demonstration by Forsgren et al. [20,] that in allophenic chimeras the presence of NOD thymic epithelium is necessary for the development of insulitis. Insulitis, the infiltration of the pancreatic islets by cells of the immune system, precedes and is necessary for the development of overt diabetes. Allophenic chimeric mice are engineered by pairing eight-cell stage mouse embryos in culture followed by the transfer of the embryos into the uteri of pseudopregnant foster mothers. Progeny with coat-colour chimerism are subsequendy selected. Mice with thymuses that contained a mixture of at least 50% NOD and diabetes-resistant epithelia cells developed insulitis. These results suggest that the NOD MHC class II inmmne-response molecules, which are likely to be primary determinants of insulitis and disease susceptibility and are expressed in the thymic epithelium, may exert their effect by positively selecting for T lymphocytes that are autoreactive and cause ]3-cell death.
Genetic analysis of diabetes in rodents In both the human and mouse, the inheritance of diabetes is complex and under the control of several unlinked loci, which in both cases include the MHC. Todd et al. [2-,] constructed a genetic map of the mouse genome, which is composed of over 50 polymorphic PCR-analyzed microsatellites, and then typed these marker loci in a large number of the affected and nondiabetic progeny from a first backcross generation (NOD outcrossed to C57BI./10). Two susceptibility loci that are located outside the MHC on chromosome 17 were mapped to chromosomes 3 and 11. The chromosome 3 gene, designated ldd-3, controls both overt diabetes and insulitis. Several other chromosomes showed weaker evidence of linkage to diabetes, suggesting that at least 9
unlinked loci, including the MHC, may contribute to disease susceptibility in this particular backcross. It is likely that crosses of NOD with different diabetes-resistant strains will reveal additional susceptibility loci. For example, Prochazka et aL  have reported evidence from C57BL/KsJ x NOD breeding experiments for a susceptibility locus on chromosome 2, which was not apparent in the NOD x C57BL/10 cross used by Todd et al. [2-]. Since then, strong evidence has been provided for at least one locus on mouse chromosome 1 affecting the development of diabetes and insulitis [22.], and sialitis [23"], the infiltration of the salivary glands by cells of the immune system. Comparison of the data from these two papers suggests that in the proximal region of mouse chromosome 1 there may be two distinct loci determining predisposition to autoimnmnity. Mapping of mouse genes associated with autoimmunity and diabetes opens up four avenues of research. First, the characterization of the mouse susceptibility genes will help accelerate understanding of the biochemistry of the disease and pinpoint the metabolic and immunesystem pathways that are important for lB-cell function and breakdown in the natural state of inmmne nonresponsiveness to self. Second, as has been found in the MHC for type 1 diabetes , a diabetogenic mutation may be conserved between species, or alternatively different mutations occurring in the same gene in mouse and man may result in a common defect. Both of these considerations benefit from the extensive homology observed between the mouse and human genomes in that it is possible to predict with reasonable confidence where a gene homologue will map in one species given its location in the other species . As such, the chromosomal locations of mouse susceptibility loci suggest the map positions of the human homologues [22",23.]. Whether or not diabetogenic alleles of the human homologues exist is still an open question but one that can be investigated experimentally. Third, chromosomal regions in the mouse that are linked to diabetes may contain known genes which can be considered as candidate genes. For example, the proximal region of mouse chromosome 1 contains several highly feasible candidate genes: including Cd-28, a T-cell activation molecule; II-lrl, the receptor for interleukin-1, which is a key cofactor in T-cell activation; the Lsh/lty/Bcg locus, which controls susceptibility to infectious disease by regulating macrophage activity; and Bcl-2, the enforced expression of which in B cells of transgenic mice prolongs antibody responses and causes autoimmune disease . Rapid PCR-assisted sequencing of these candidates from NOD and diabetes-resistant mouse strains may provide initial clues as to the identity of the chromosome 1 susceptibility locus or loci. Finally, dissection of the genetics of autoimmune destruction in mice will provide some understanding of the complexities of the disease. How many unlinked loci are involved, how many of these are sufficient to cause insulitis or diabetes and in what ways do they interact?
Diabetes mellitus Todd
Conclusion Advances in the genetics of diabetes promise success for the application of basic research to the identification of high risk individuals in the population and ultimately, by using rationally designed preventative or therapeutic treatments, to the manipulation of pathways that lead to insulin deficiencies. Furthermore, analysis of type 1 diabetes is developing into one of the best worked examples of how to study the genetics of a complex trait.
RISCH N: Linkage Strategies for Genetically C o m p l e x Traits. III. T h e Effect of Marker Polymorphism o n Analysis of Affected Relative Pairs. Am J H u m Genet 1990, 46:242-253.
HEARNECM, GHOSH S, TODD JA: Microsatellites for Linkage Analysis of Genetic Traits. Trends Genet 1992, in press.
RISCH N: Developments in Gene Mapping with Linkage Methods. C ur t Opin Genet Dev 1991, 1:93--98.
RISCHN: Linkage Strategies for Genetically C o m p l e x Traits. If. The Power of Affected Relative Pairs. A m J H u m Genet 1990, 46:229-241.
JULIERC, HYER RN, DAVIESJ, MERLIN F, SOULARUE P, BRL~rr L, CATHELINEAUG, DESCHAMPS l, ROTTERJl, FROGUALP, ET At+: Insulin-IGF2 Region on C h r o m o s o m e 1 l p Encodes a Gene Implicated in HLA-DR4-dependent Diabetes Susceptibility. Nature 1991, 354:155-159.
RlSCH N: Assessing the Role of HLA-linked and Unlinked Determinants of Disease. Am J H u m Genet 1987, 40:1-14.
References and recommended reading Papers of particular interest, published within the annual period of review, have been highlighted as: • of special interest •. of outstanding interest 1. e.
BELL GI, XIANG K-S, NEWblAN MV, WU S-l-l, WRIGHT LG, FAJANS SS, SPIEI.bIAN RS, Cox NJ: Gene for Non-insulin-dependent Diabetes Mellitus (Maturity-onset Diabetes of the Young Subtype) is Linked to DNA Polymorphism on H u m a n Chrom o s o m e 20q. Proc Nail Acad Sci USA 1991, 88:1484-1488. This paper describes the first susceptibility gene in diabetes to be localized by exclusion mapping. In addition, the marker locus closest to the disease locus is a highly polymorphic tmldemly repeated simple sequence or microsatellite, and demonstrates the utility of this marker system. TODD JA, kaTMAN TJ, CORNALLRJ, GHOSH S, l-ltda. JR.S, H :IbXRNE CM. KNIGHT AM, [.OVI~ JM, MCALEER MR, PRiNS J-B, ET" A/.: Genetic Analysis of A u t o i m m u n e Type I Diabetes Mellitus in Mice. Nature 1991, 351:542-547. The paper describes both the first sTstematic exclusion map of the mot, se genome, which was constructed using PCR-analyzed microsatellites, and the first example of a dissection of a spontaneously developing multifactorial disease. Unequivocal evidence for linkage of chromosomes 3 and 11 to autoimmune type 1 diabetes is obtained. The methods used are applicable to other diseases and demonstrate the utility of microsatellite marker loci mad the power of analysis provided by inbred strains and mouse genetics.
~SCH N: A Note on Multiple Testing Procedures in Linkage Analysis. Am J H u m Genet 1991, 48:1058-1064. This paper addresses the controversial issue of the effect of mulUple tesUng on the statistical analysis of complex disease. Risch shows that the testing of multiple marker loci in an exclusion mapping study actually decreases the false-positive rate amongst significant tests. In contrast, and as a word of warning to researchers in the field, he finds that testing multiple genetic models or using different diagnosUc criteria increases the posterior false-positive rate. 14.
SOBELLJL, HESTON EL, SOMMERSS: Delineation of Genetic Predisposition to Multifactorial Disease: A General Approach on the Threshold of Feasibility. Genomics 1991, 12:1-6.
FlARiNG HU: The Insulin Receptor: Signalling Mechanism and Contribution to the Pathogenesis of Insulin Resistance. Diabetologia 1991, 34:848-861.
O\VERBACHD, GUNN S, GABBAYKH: Multigenic Basis for Type 1 Diabetes: Association of HRAS1 Polymorphism with HLADR3,DQw2/DR4,DQw8. Diabetes 1990, 39:1504-1509.
FIELDLL, STEPHURE DK, MCARTHURRG: Interaction b e t w e e n T CeU Receptor Beta Chain and lmmunoglobuUn Heavy Chain Region Genes in Susceptibility to Insulin-dependent Diabetes Mellitus. Am J H um Genet 1991, 49:627-634.
PROPERT DN, TAIT BD, HARRISON LC: Interaction of Immunoglobulin AUotypes (GM and KM), HLA, and Sex in Insulin-dependent (Type 1) Diabetes. Dis Markers 1991, 9:43-45.
HYER RJN,JUUER C, BUCKIJ'.'YJD, TRUCCO M, RO'I-I'ERJ, SPtEt>l~ R, BARNF'Iq"A, BMN S, BOITARD C, DESCHAMPS I, ET AI.: High Resolution Mapping for Susceptibility Genes in H u m a n Polygenic Disease; Insulin-dependent Diabetes Mellitus and C h r o m o s o m e l l q . Am J H u m Genet 1991, 48:243-259. An elegant and graphical analysis of the number of families that are required to enable the detection or exclusion of claromosom',d regions inw)lved in susceptibility to a complex, multifactorial disease. The authors provide a practical example by excluding linkage of 17 markers on human chromosome l l q to type 1 diabetes, and also use the established HLA linkage as a internal standard for this multipoint linkage approach. The power to detect linkage by affected sib-pair analysis is shown to be a function of IBD probabilities, and gives some indication of the range of gene effects, relative to the HLA, that might be detectable. 4.
STALIANGSRL, FORD AF, NELSON D, TORNEY DC, HILDEBR.PuND CE, MOYZlS RK: Evolution and Distribution of (GT)n Repetitive Sequences in Mammalian Genomes. Genomics 1991, 10:807-815.
WEBERJL, POLYMEROPOUI.OSMH, MAY PE, KWITEKAE, XlAO H, MCPHERSON JD, WASMU'm JJ: Mapping of H u m a n Chromos o m e 5 Microsatellite DNA Polymorphisms. Genomics 1991, 11:695-700.
HEam~ECM, McAt.EER MA, LOVE JM, AITMAN TJ, CORNAU+ RJ, GHOSH S, KNIGHT AM, PRINS J-B, TODD JA: Additional Microsatellite Markers for Mouse G e n o m e Mapping. M a m m Genome 1991, 1:273-282.
MOSTHAFL, VOGT B, HARING HU, ULLRICHA: Altered Expression of Insulin Receptor Types A and B in the Skeletal Muscle of Non-insulin-dependent Diabetes Mellitus Patients. Proc Natl Ac ad Sci USA 1991, 88:4728--4730. Type 2 diabetes, NIDDM, is a genetically complex, heterogeneous disease. One route for studying the genetics of the disorder is to identify subphenoWpes that might have a more simple mode of inheritance and that are reproducibly typed in patients. Tests of defects associated with N1DDM, for example the oral glucose tolerance test, are notoriously Jrreproducible. Mosthaf et al. describe a PCR assay, of insulin-receptor RNA that is derived from the skeletal muscle of patients and control individuals and which detects alternatively spliced forms of the receptor. The two different receptor isoforms have different affinities for insulin and the authors find an increase in the level of the lower affinit3, form in skeletal muscle biopsies from patients with NIDDM. This observa. tion might be related to the insulin-resistant state observed in diabetic skeletal muscle. 20.
FORSGRENS, DAHL U, SODERSTROMA, HOLMBERGD, MATSUNAGA
T: The Phenotype of Lymphoid CeUs and Thymic Epithe-
lium Correlates with the D e v e l o p m e n t of A u t o i m m u n e Insulitis in NOD ,--* C57BL/6 Allophenic Chimeras. Ptx~c Nail Ac ad Sci USA 1991, 88:9335--9339. A good demonstration of the dissection of disease aetiology using a combinaUon of embryology and immunological studies in the NOD mouse.
Mammalian genetics 21.
PROCHAZKAM, SERREZEDV, FRANKELXX/N, LEITER EH: NOR/Lt Mice: MHC-matched Diabetes-resistant Control Strain for NOD Mice. Diabetes 1991, 41:98-106.
CORNALLRJ, PP,INS J-B, TODD JA, PRE~EY A, DELARATO NH, WICKERIS, PETEILSONLB: Type 1 Diabetes in Mice is Linked to the lnterleukin-1 Receptor and Lsh/lry/Bcg Genes on Chromosome 1. Nature 1991, 353: 262-265. Expanding the work presented in [2"'], the authors describe an insulitis and diabetes gene located on mouse chromosome 1. Includes a demonstraUon of how linkage analysis can lead us to consider certain genes as candidate genes for disease susceptibility loci because they fail in the area of linkage to disease.
significant evidence that this gene also affects periinsulitis, the study is the first published use of an F2 outcross in the analysis of NOD autoimmunity, in which dominant suscepUbility genes can be detected. Backcrosses to NOD mice (see [2.% 22-]) cannot be used to detect fully dominant gene effects.
GARCHONH-J, BEDOSSA P, ELOY L, BACH J-F: Identification and Mapping to Chromosome 1 of a Susceptibility Locus for Periinsulitis in Non-obese Diabetic Mice. Nature 1991, 353:260-262. Using the microsateRites described in [2.,], Garchon and colleagues report evidence for a locus on chromosome 1 that is near Bcl.2 and which controls the development of sialiUs, the inffltraUon of the salivary #ands by cells of the immune system. Although they did not present
TODDJA, BELLJl, McDEvlTT HO: HLA-D[3 Gene Contributes to Susceptibility and Resistance to Insulin-dependent Diabetes MelUtus. Nature 1987, 329:599--604.
NADEAUJH: Maps of Linkage and Synteny Homologies Between Mouse and Man. Trends Genet 1989, 5:82--86.
STRASSERA, WHITTINGHAM S, VAUX DL, BATH ML, ADAMS JM, CORY S, HARMS AW: Enforced BCL2 Expression in ~-Iymphoid Cells Prolongs Antibody Responses and Elicits Autoimmune Disease. Proc Natl Acad Sci USA 1991, 88:8661-8665.
JA Todd, Nuffleld Department of Surgery, Oxford University, John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.