Mitochondrial Disease and Type 2 Diabetes
Muscle and neurologic disorders that were clinically classified as probable genetic syndromes with maternal inheritance have eventually been shown to be associated with a variety of deletion, duplication, and mutational alterations of specific mitochondrial genes.122 However, it recently became apparent that some of these neuromuscular mitochondrial syndromes are also associated with type 2 diabetes mellitus, in some cases even without neuromuscular findings.123
Mitochondria are critically important for cellular energy production, and their number in individual cells tends to reflect the requirement for oxidative phosphorylation.122 Thus, some cell types may have up to 1000 per cell, whereas others, such as erythrocytes, have none. Within the mitochondrial matrix there are between 2 and 10 molecules of double-stranded, circular DNA. The entire sequence of the 16,569 base pairs of human mitochondrial DNA is known. Importantly, mitochondrial DNA is largely maternally inherited because essentially all of the cytoplasm and mitochondria are derived from the ovum, and only the spermatozoal head enters to provide nuclear DNA. In addition, mitochondria may replicate independently of the cell cycle. The precise mechanisms for this are not understood. Further, mitochondrial DNA is more susceptible to mutation than nuclear DNA. This is probably because mitochondrial DNA is not protected by histones, and is exposed to a high level of free radicals produced during oxidative phosphorylation. In addition, only one of the five eukaryotic DNA polymerases, gamma, is present in mitochondria, and this polymerase is less efficient in dealing with mismatches than some of the corresponding nuclear enzymes. Because there are a number of mitochondria per cell, when a mutation occurs, the cell will have two populations of mitochondrial DNA: wild-type and mutant. This phenomenon is termed heteroplasmy. Because mitochondrial DNA may segregate unevenly to daughter cells during division, heteroplasmy ratios may vary between daughter cells. If the mutation confers a replication advantage, it may become the dominant form of mitochondrial DNA within a cell, in fact, replacing the original wild-type sequence completely (called homoplasmy). If this occurs in female germ cells, the sequence will be passed on to the offspring. If the novel mitochondrial DNA mutant is highly deleterious, it is likely to be eliminated from the mitochondrial population and homoplasmy of wild-type restored. Presumably, most mutations associated with human disease are relatively benign for replication, leading to persistent heteroplasmy. However, it is generally true that rapidly dividing cells have low levels of heteroplasmy, while slowly dividing cells will have relatively high levels of heteroplasmy, suggesting segregation of mutated mitochondrial DNA in some way. The mitochondrial genome is compact, with no introns and no large noncoding sequences. Thirteen polypeptides, which are all subunits of enzymes of oxidation phosphorylation, are encoded together with 22 transfer RNAs and 2 RNAs required for their intramitochondrial synthesis. Nuclear DNA codes for all of the other subunits of oxidative phosphorylation (the majority). Mitochondria are usually present in excess in cells, and therefore clinical features of phosphorylation deficiency only arise if the levels fall below a threshold. Thus, disordered function may not arise until the level of heteroplasmy is sufficient to interfere with normal function, which may take a considerable period of the patient's life span. Because different tissues have different requirements for oxidative phosphorylation, symptoms will vary greatly depending upon their need for this function.124
The first case of mitochondrial disease with diabetes, but not muscle or nerve dysfunction, except for sensorineural hearing loss, was reported in 1992,123 with an increasing number being reported since that original report. Enough cases of mitochondrial disease with diabetes have been reported that some general comments about the clinical characteristics of these patients and their genetic and clinical heterogeneity can be made.122,125–128 The general characteristic appears to be an early, selective impairment of insulin secretion. Later, it appears as a global metabolic dysfunction similar to classical type 2 diabetes with insulin resistance and fasting hyperglycemia. The diabetes often progresses to absolute insulin deficiency and the need for insulin treatment. In some cases, an autoantibody to glutamic acid decarboxylase (GAD) 129 or islet cells (ICA)130 is present. Because these markers are used to assist in the diagnosis of type 1 diabetes, there can be confusion regarding classification. In a detailed analysis of subjects with a known mitochondrial defect from families with associated diabetes, those with normal glucose tolerance or impaired glucose tolerance appear to have either a failure to prime the insulin secretory response by a prolonged glucose infusion or an inability to entrain ultradian insulin secretory oscillations.128,131 As the disease progresses, clear-cut impaired insulin secretory rates to standard glycemic challenge become obvious, and insulin resistance may develop.132,133 Because it has been shown that pancreatic islet β-cells are dependent on mitochondrial metabolism for function,134 and are characterized by high rates of blood flow to the organ and relatively high rates of oxygen consumption, it is assumed that they are a critical tissue, along with nerve and muscle, which depend on mitochondrial oxidative phosphorylation. Despite the general similarities that have been described, there are great differences among individuals in clinical presentation, age of onset, rate of progression, and eventual dependence on insulin. Furthermore, individuals with the same molecularly determined genetic lesion might have different clinical manifestations. Thus, several quite different clinical syndromes may have the same underlying molecular defect. Some of the heterogeneity can be explained by the heteroplasmic nature of the cellular basis for the disease and some by varying genetic backgrounds of the subjects. Thus, reasonable segregation into distinct clinical syndromes is not apparent from the studies that have been performed so far because there is not a good correlation between molecular and clinical findings in many patients.
Specific Mitochondrial Syndromes
Sensorineural Hearing Loss and Diabetes; Maternally Inherited Diabetes and Deafness (MIDD).
In 1992, Ballinger et al. reported a single, large pedigree with maternally inherited diabetes and nerve deafness associated with a 10.4-kb deletion of mitochondrial DNA.123 Further studies of this pedigree have shown the presence of a variety of mitochondrial rearrangements. It is likely that the primary defect is mitochondrial DNA duplication.125 Shortly afterward, van den Ouweland et al. described a pedigree with maternally inherited diabetes and deafness with a heteroplasmic A to G mutation at 3243 in the tRNAleu(UUR) portion of the mitochondrial genome.135 This was rather surprising because the same mutation had been previously reported to be associated with the MELAS syndrome (mitochondrial encephalopathy, lactic acidosis, and stroke-like episodes)136 (see below). In fact, it had been noted for many years that diabetes was sometimes associated with the MELAS syndrome. However, the van den Ouwland case and pedigree, and others reported since, are only affected with diabetes and/or deafness. The clinical diagnosis of diabetes usually, but not always, precedes the detection of clinical sensorineural hearing loss, but the age of presentation in a rather large Japanese population has varied from 11 to 68 years.137 The diabetes often progressed to sulfonylurea failure (on average, approximately 8 years after the initiation of treatment). Some cases were initially considered to represent the slow onset of type 1 diabetes because of the finding of GAD antibodies, and rather rapid progression to insulin requirement. In this population, approximately 10 percent of such clinically defined slow onset “type 1” patients selected for follow-up were in fact found to be affected by the MIDD syndrome.137 This syndrome has been estimated to account for approximately 1.3 percent of Japanese patients137 and 1.5 percent of unselected European cases of type 2 diabetes.126 However, careful random population screens have not been performed. There are reports of an associated high frequency of abnormal brain MRIs and single photon emission computed tomography (SPECT) imaging studies in such patients, suggesting very subtle neurologic defects in addition to the hearing loss.129 Occasional cases of cardiomyopathy have been reported with biopsy-proven heteroplasmy for the 3243 A to G mutation.138 The identical clinical picture has also been associated with several other less-common mitochondrial mutations. The most common of these, the tRNA(lys) 8296 A to G mutation, was estimated to explain 1 percent of Japanese type 2 diabetes cases,139 and the 3271 T to C tRNAleu(UUR) mutation was estimated to explain one-tenth as many.140 An explanation for the same molecular lesion in MELAS and MIDD is not clear, nor is there a very good correlation between the degree of heteroplasmy in peripheral tissues and the severity of the diabetes mellitus, suggesting that other genes or environmental factors are important.
Mitochondrial Encephalopathy, Lactic Acidosis, and Stroke-Like Episodes (MELAS).
This disorder may result in hemiparesis, hemianopia, or cortical blindness. Patients may present with focal seizure. CT scanning reveals low-density areas affecting both the gray and white matter, not necessarily correlating with the clinical features. The neurologic deficit may resolve, or the patient may have recurrent attacks leading to progressive encephalopathy, disability, and death. The 3243 A to G mutation is most common, but others136,140 have been reported. In those patients who survive long enough, diabetes is a common finding.122,136
Kearns-Sayre Syndrome (KS).
KS describes patients who present with chronic, progressive, external ophthalmoplegia and pigmentary retinopathy under age 20, together with one or more of the following: ataxia, complete heart block, and cerebral spinal fluid protein concentration of greater than 1 g/liter. The majority of the patients have a mitochondrial DNA deletion on muscle biopsy. Diabetes is frequently associated.122,141
Maturity Onset Diabetes of the Young (MODY)
In the great majority of patients with type 2 diabetes, a diagnosis is made in middle age. However, a subclass of this syndrome includes families in whom diabetes can be recognized in children, adolescents, and young adults, which has been termed maturity onset diabetes of the young (MODY).142 Autosomal dominant inheritance has been established in these families, and specific genetic mutations have been identified. It is the only form of diabetes in which a definite mode of inheritance has been determined. At least five specific mutations have been described, and this does not include all of the families in which a dominant genetic pattern with young family members has been observed. Therefore, more mutations are yet to be discovered. At times, these patients have been confused with type 1 diabetes patients because they are young, some variants lead to relatively severe forms of insulin-deficient diabetes, and by and large most patients with this syndrome are lean. However, most of the patients are ketosis resistant, and even those who need insulin for control of glycemia are not usually ketosis prone. Furthermore, because the more common variety of type 2 diabetes can also be present among these families, it is possible for an individual family member to have both syndromes, and therefore relative insulin deficiency may become more evident in the presence of insulin resistance, obesity, and, during periods of stress, infections or trauma. In some MODY patients, there may be easily diagnosable clinical hyperglycemia, but in other members of the family with the same mutation, much milder degrees of hyperglycemia or glucose intolerance may be present. In the younger members diagnosed by genetic screening, some may have normal glucose tolerance, although when studied carefully, subtle abnormalities of insulin secretion may be detectable. MODY can be suspected and recognized if type 2 diabetes occurs in three or more generations and the pattern of inheritance conforms to autosomal dominant. Families have been diagnosed in Europe, the United States, Japan, and other Asian countries. Progression of hyperglycemia tends to be relatively slow, but this depends somewhat on the specific genetic defect. Glucose tolerance may fluctuate depending on other factors, and variability among family members may be related to other genetic and environmental factors, which can fluctuate over time. In the absence of a specific molecular diagnosis, there may, therefore, be confusion between classical late onset type 2 diabetes, type 1 diabetes of slow onset, and MODY diabetes. The prevalence of MODY is still unknown, and estimates have varied widely from 2–5 percent of patients with type 2 diabetes to 10–20 percent of families with type 2 diabetes. Studies of the insulin secretory defects early in the course of the disease indicate different physiological abnormalities that distinguish among the various genetic mutations.
The R-W pedigree, which includes more than 360 members spanning 6 generations and 74 members with diabetes, including those with MODY, has been studied prospectively since 1958.57,143 Linkage studies had shown that MODY-1 was linked tightly to markers in chromosome 20, band 20q-13.1, eventually describing a 13-cM interval between markers D20S169 and D20S176.144 The demonstration in 1996 that mutations in the HNF-1α gene are the cause of MODY-3145 led the same group to screen the HNF-4α gene for mutations in the R-W pedigree because HNF-4α had been shown to be located between the same critical markers.146 Furthermore, it had been shown that the HNF-4α protein was a member of the steroid/thyroid hormone receptor superfamily and an upstream regulator of HNF-1α expression.147 In the R-W pedigree with MODY-1 there was a C → T substitution in codon 130 leading to a threonine to isoleucine substitution and a C → T substitution in codon 268, which generated a nonsense mutation CAG (Gln) → TAG (AM) (Q268X). Both the isoleucine 130 and the amber mutation at codon 268 were present on the same allele.145
The clinical data indicate that the nonsense mutation in the HNF-4α gene is highly, but not completely, penetrant. There are some subjects who have normal glucose tolerance and some who have impaired glucose tolerance, but most have clinical hyperglycemia. However, the age of diabetes onset is variable.145 The family is interesting because one member has type 1 diabetes and one branch has typical type 2 diabetes without the HNF-4α gene mutation. HNF-4α is most highly expressed in liver, kidney, and intestine, but is also present in pancreatic islets and insulinoma cells. It is called a hepatic nuclear factor because it is a key regulator of hepatic gene expression and a major regulator of HNF-1α, which, in turn, activates the expression of a large number of liver-specific genes. Homozygous loss of functional protein in mice leads to embryonic lethality and defects in gastrulation.148 The phenotype of the heterozygous animals has not been described. It has been hypothesized that the mutation in the R-W family would lead to a protein unable to dimerize or bind to the recognition site and activate transcription, and that impaired function is a result of a deficiency of HNF-4α rather than a dominant negative mechanism.145 There are no obvious hepatic, renal, or gastrointestinal defects in family members, but detailed investigations of liver glucose metabolism have not been made. Study of marker-positive family members with normal or just impaired fasting glucose shows reduced insulin secretory rates above 7 mM glucose at steady state and an impaired ability to prime β-cell function after a 42-h intravenous glucose infusion57 (Fig. 67-8). This pattern is different from MODY-2 and MODY-3, despite the relationship between HNF-4α and HNF-1α, the genetic mutation in MODY-3.56,58 Approximately one-third of affected individuals require insulin treatment, and patients with this form of diabetes may have microvascular complications.
Insulin secretion rates during graded intravenous glucose infusions administered to 6 marker-negative and 10 marker-positive subjects from the R-W MODY-1 family after an overnight fast (baseline) and after a 42-h intravenous infusion of glucose (postglucose) at a rate of 4 to 6 mg · kg−1 · min−1. Data expressed as means ± SE. (Reproduced from Froguel.154 Used with permission.)
Recently, a second and a third mutation have been associated with MODY-1 or early onset type 2 diabetes. One, a glutamine for glutamic acid substitution in exon 7 (E276Q),149 was present in all of the diabetic members of the pedigree, plus two unaffected subjects, but was not detected in 75 normal control subjects or 95 U.K. Caucasian subjects with late onset type 2 diabetes. The other is an Arg127/Trp (R127W) mutation located in the T-box region that may play a role in dimerization and DNA binding.150 None of 53 unrelated nondiabetic Japanese, 53 Chinese, 51 Caucasian, or 57 African-American subjects were positive. The mutation was present in three other members of a family with a dominant pattern of diabetes inheritance, with some members being diagnosed as young as 11 years, consistent with MODY diabetes. However, two family members with early onset disease do not have the mutation and one member with the mutation was only diagnosed at age 90 (Fig. 67-9).
The distribution of HNF-4α gene mutations in MODY-1 families. When there is more than one family with a mutation, the number of families is shown in parentheses. (Reproduced from Hattersley.176 Used with permission.)
Functional significance of mutations in HNF-4α has been supported by a study of a nonsense mutation (Q268X) in exon 7, the defect in the original family with MODY-1. Mutant and wild-type HNF-4α were expressed, and DNA binding and transcriptional activation studied. The HNF-4α mutant did not bind to the usual HNF-4α binding sites as a homodimer or heterodimer, and exhibited no transcriptional transactivation activity. Furthermore, it does not interact with the wild-type protein in a dominant negative fashion. In embryonic stem cells, the glucose transporter 2 (GLUT-2), the glycolytic enzymes aldolase B and glyceraldehyde-3-phosphate dehydrogenase, and liver pyruvate kinase were down-regulated, as well as expression of the fatty acid-binding proteins and cellular retinal binding protein, indicating the likely interference with glucose metabolism in the subjects with the HNF-4α mutation to a critical level in the islet.147
Linkage between this region has been observed in late onset type 2 diabetes in American Caucasian sib-pairs,151 and in the same region in French families, diagnosed under age 45, but not in African-American families.151 Screening the French families found one to contain a mutation of HNF-4α (V391I) which cosegregated with diabetes and deficient insulin secretion.152 However, this one family does not explain the regional linkage and an additional locus is assumed to be present in type 2 diabetes. A Val/Met 255 mutation was found in HNF-α in 4 of 477 Danish type 2 diabetes patients and in none of 217 glucose-tolerant controls.153 A Thr/Ile130 variant was present in 4.7 percent of 509 type 2 diabetes patients compared to 1.9 percent in 230 controls. However, in a population of 942 Swedish men, the allele frequency of this variant was similar in IGT, type 2 diabetes, and controls, as it was in a population of 369 young healthy Danes. Furthermore, there is no effect of this mutation on insulin levels in either of these other two populations.153
The first candidate gene linkage in the MODY syndrome was described from a study of 16 French families with three or more generations of impaired glucose tolerance or type 2 diabetes, and was found to link to a polymorphic marker adjacent to glucokinase at chromosomal locus 7p.102 Eventually, a large number of families with 42 different mutations involving all 10 exons in populations distributed worldwide, including all racial types (Caucasian, Black, and Asian) were described154 (Fig. 67-10). Twenty-eight of these mutations alter the protein sequence by changing one amino acid; six transform the sequence at the site of RNA splicing of an intron-exon or exon-intron junction, resulting in the expression of an abnormal species of messenger RNA; and eight are responsible for the synthesis of a truncated protein by creating a premature termination codon by point mutation or deletion. Most are present in only single families. The enzyme is present at critical levels in the liver and the pancreas, playing an important role in hepatic glucose storage by phosphorylation of glucose after absorption and in the endocrine pancreas by coupling the first step of glucose metabolism to insulin secretion through the generation of ATP and regulation of islet potassium conductance and calcium levels.155 The missense mutations fall into three main categories: those of conserved active site residues, which generally have a drastic effect on catalytic activity, those predicted to distort enzyme structure, which also tend to show reduced activity, and mutations of surface residues, which may reduce the stability of the structure or affect the conformational change observed on binding of glucose.156–158 Mutations may also alter the interaction of glucokinase with other proteins such as the glucokinase regulatory protein.159 Islet glucokinase and hepatic glucokinase are identical from exon 2 through 10, but have a variable exon 1. Most of the mutations are localized in their shared areas. In general, the degree of hyperglycemia is relatively mild, although it can be detected in children. Fewer than 50 percent of subjects present with overt diabetes, but most of the rest demonstrate impaired fasting glucose and/or glucose intolerance. Most of the individuals who carry the mutation are affected before puberty. Long-term complications are relatively unusual with this form of diabetes. Hyperglycemia remains stable, whether treated or not, for many years, and does not progress very fast as in typical type 2 late onset diabetes. Treatment with oral agents is usually satisfactory, and can lead to reasonable control for many years142,154,160,161 (Fig. 67-11). Amyloid accumulation in the pancreas was not present in the one case examined with this form of diabetes.162 It is not known whether this is because of the intrinsic nature of the lesion or the mildness of the hyperglycemia. Obesity is not usually present and hypertension and dyslipidemia are not common. Macrovascular disease is also not common. The major observed functional defect is impaired insulin secretion, but this may not be evident unless sophisticated testing is performed. Twenty-four-hour profiles of affected subjects show persistent, round-the-clock hyperglycemia, but insulin levels are only slightly reduced after meals and are not different at baseline and between meals. However, when glucose levels are matched during continuous infusions, they are reduced by at least 50 percent. At maximal hyperglycemia (i.e., >12 to 15 mM), they are markedly reduced compared to healthy controls56,161,163 (Fig. 67-12). Islet priming by a 42-h glucose infusion shows a response, but of smaller magnitude than controls, and is quite different than MODY-1 or MODY-3.57,58,154 After an overnight fast, the entire curve is shifted down and to the right, whereas in MODY-1 the curve is flat throughout, just as it is in MODY-3. In MODY-2, glucose potentiation of arginine-induced insulin secretion is reduced, and glucose suppression of arginine-induced glucagon secretion is impaired.164 Insulin sensitivity is usually normal, but may be reduced in individuals who are obese or in whom hyperglycemia is severe.154,165 The average fasting hepatic glycogen content was similar in glucokinase-deficient subjects as compared to control subjects; it increases in both groups after meals, with a continuous pattern throughout the day. However, the net increment in hepatic glycogen content after each meal was 30 to 60 percent lower in glucokinase-deficient subjects than in control subjects. In the 4 h postbreakfast, flux through the gluconeogenesis pathway relative to the direct pathway of glycogen synthesis was higher in the glucokinase-deficient subjects than in control subjects. Thus, glucokinase-deficient subjects have decreased net accumulation of hepatic glycogen, and relatively augmented hepatic gluconeogenesis after meals.154,166,167 This may implicate mutations in hepatic glucokinase in the pathophysiology of the syndrome. While occasional cases of late onset type 2 diabetes have been reported with mutations in glucokinase,168 the pattern of inheritance could not be discriminated from that of MODY diabetes, and it is possible that some mutations produce a syndrome so mild that only obesity in middle age or aging-associated insulin resistance will be associated with clinical diagnosis.
The exon-intron organization of the human glucokinase gene and mutations found in subjects with MODY-2 diabetes. Amino acid residues are numbered as in the β-cell form of human glucokinase. The two amino acid polymorphisms in the unique NH2-terminal portion of β-cell glucokinase that is encoded by exon 1a are indicated (Asn/Gln-4 [D/N4] and Ala/Thr-11). (Reproduced from Froguel.154 Used with permission.)
The relationship between age and fasting plasma glucose values in MODY-2 patients with glucokinase mutations. The fasting plasma glucose values increase very gradually with age (r = 0.41; p = 0.002). (Reproduced from Tanizawa and Chiu.261 Used with permission.)
Insulin secretion rates during graded intravenous glucose infusions administered to 6 MODY-2 subjects with glucokinase (GCK) mutations and 6 control subjects after an overnight fast (baseline) and a 42-h intravenous infusion of glucose (postglucose) at a rate of 4 to 6 mg · kg−1 · min−1. Administration of the glucose infusion enhanced the insulin secretory response to glucose in both control and glucokinase-deficient subjects. Data are expressed as means ± SE. (Reproduced from Froguel.154 Used with permission.)
In 1995, a genome-wide segregation analysis of highly informative microsatellite markers in 12 French families with MODY diabetes reported the localization of a gene for MODY to the long arm of chromosome 12 in 6 of 12 families.169 Subsequently, using the identified markers D12S86 and D12S342, four more families from Denmark, the U.S., Germany, and Japan also showed linkage.170 In 1996, Yamagata et al. showed that MODY-3 was the gene encoding the hepatocyte nuclear factor-1α, a transcription factor involved in tissue-specific regulation of liver genes that is also expressed in pancreatic islets and other tissues.145,171 Seven different mutations in HNF-1α were found to cosegregate with MODY-3 diabetes. More than 50 different mutations have now been identified in MODY-3 families of various populations172–176 (Fig. 67-13). An insulin secretory defect in the absence of insulin resistance was observed in diabetic and nondiabetic carriers of MODY-3 mutations with a pattern that is distinct from MODY-1 and MODY-2 during sequential glucose infusions.58,154 In marker-positive NGT subjects, the defect was not observed at glucose levels less than 8 mM, but became evident as glucose rose above this level (Fig. 67-14). Priming was still present in nondiabetic marker-positive subjects, but not in those with clinical hyperglycemia (Fig. 67-15). This is a clinically more severe form of MODY, characterized by a high prevalence of overt diabetes, frequently treated with oral hypoglycemic agents and often requiring insulin.177 This type of MODY diabetes resembles late onset type 2 diabetes in its natural history with patients progressing rapidly from impaired glucose tolerance to overt hyperglycemia with severe deterioration of insulin secretion. Complications, such as proliferative retinopathy, have been observed frequently and at rates comparable to late onset type 2 diabetes. However, there is a low prevalence of obesity, dyslipidemia, and arterial hypertension, and, unlike MODY-2, clinical disease with hyperglycemia usually develops after puberty.177 HNF-4α is a transcription factor for HNF-1α. There is a mutation in one family with MODY-3 diabetes in the conserved region of the promoter that disrupts the binding site for the transcription factor HNF-4α.178 This finding suggests that a 50 percent reduction in HNF-1α activity can lead to β-cell dysfunction and hyperglycemia. During screening of 110 Japanese subjects with type 2 diabetes, one patient was found with a mutation at G191D who was diagnosed with hyperglycemia at 64 years of age and who had a sib who was diagnosed with hyperglycemia at 54 years of age. The older sib was known to have to have a normal oral glucose tolerance at age 62.179 This mutation was not present in 50 unrelated subjects with normal glucose tolerance, and has not been reported in association with MODY diabetes elsewhere. While one group screened early onset Japanese subjects for mutations of this gene and found none,180 another group reported 7 of 83 such cases (8 percent) so surveyed.179 Thus, the true frequency of genetic variation in HNF-1α remains to be determined.
The distribution of HNF-1α gene mutations in MODY-3 diabetes. When there is more than one family with a mutation, the number of families is shown in parentheses. (Reproduced from Hattersley.176 Used with permission.)
Relationship between average plasma glucose concentrations and insulin secretion rates during the stepped glucose infusion studies in seven diabetic MODY-3 subjects (□), six nondiabetic MODY-3 subjects (▴), and six control subjects (○). The lowest glucose levels and insulin secretion rates were measured under basal conditions, and subsequent levels were obtained during glucose infusion rates of 1, 2, 3, 4, 6, and 8 mg · kg−1 · min−1, respectively. (Reproduced from Byrne et al.58 Used with permission.)
Graded intravenous glucose infusions were administered to six control subjects (A), six nondiabetic MODY-3 subjects (B), and seven diabetic MODY-3 subjects (C) after an overnight fast (baseline, ▴) and after a 42-h intravenous infusion of glucose (postglucose, □) at a rate of 4 to 6 mg · kg−1 · min−1. (Reproduced from Byrne et al.58 Used with permission.)
Hepatocyte Nuclear Factor-1β (HNF-1β).
HNF-1β is a member of a complex transcriptional regulatory network that includes HNF-1α and HNF-4α. It is a homeodomain-containing factor that is structurally related to HNF-1α, and functions as a homodimer or a heterodimer with HNF-1α. They are stabilized by DcoH/pterin-4α-carbinolamine dehydratase (DCoH), which binds to the dimerization domains of HNF-1α and HNF-1β, forming a heterotetrameric complex that enhances transcriptional activity. Both HNF-1β and DCoH are expressed in mouse pancreatic islets. Fifty-seven unrelated Japanese subjects with MODY were screened for mutations in these genes, and one family with a nonsense mutation in codon 177 (R177X) of HNF-1β was found, which, in this family, was associated with diabetes.181 This family is complex in that the other parent has what appears to be late onset type 2 diabetes with no mutations identified. The truncated protein when expressed does not stimulate transcription of a rat albumin promoter-linked reporter gene or inhibit the activity of wild-type HNF-1β. Further screening has been reported in additional families with early onset type 2 diabetes and a frameshift mutation, with a GG insert in exon 3 of HNF-β for Ala 263, designated A263fsinsGG, found.182 This MODY family has diabetes in three generations with renal dysfunction consisting of renal cysts, proteinuria and/or elevated creatinine. The renal lesion may appear first. The other family also had some renal dysfunction and, therefore, this may be a feature of the HNF-1β mutation. Screening of 11 MODY families left from 58 families screened for all known MODY genes in the U.K. was negative for HNF-1β mutations. Four had sufficient numbers to eliminate linkage.183
One family with a mutation in the insulin promoter factor-1 (IPF-1) gene was recently reported.184 IPF-1, also known as IDX-1, STF-1, and PDX-1, regulates both early pancreatic development and the expression of key endocrine β-cell-specific genes, most notably insulin. Targeted homozygous destruction in mice results in pancreatic agenesis.185 Family studies stimulated by a case with a homozygous inactive mutation in exon 1 of the coding sequence with pancreatic agenesis and neonatal diabetes186 found a high prevalence of diabetes mellitus with age-dependent autosomal dominant transmission in six generations.184 Average age of onset of hyperglycemia is 35 years (range: 17 to 67 years). The hyperglycemia is mild, is treated with diet or oral hypoglycemic agents, and clinically resembles MODY diabetes. The association between the mutation and the clinical syndrome was so strong, that the authors suggested the term MODY-4. Screening of 67 Japanese subjects with type 2 diabetes for polymorphisms of exon 2 revealed no mutations.187
Genetic Studies of Late Onset Type 2 Diabetes Mellitus
While the pathophysiology of hyperglycemia and type 2 diabetes mellitus is clear, understanding the etiology of the underlying defects at a molecular level and the identification of the genetic defects associated with typical late onset type 2 diabetes have been difficult. This is a complex metabolic disorder with both insulin resistance and impaired β-cell function,188,189 and both of these defects have primary (i.e., genetic), environmental (e.g., dietary), and secondary (i.e., hyperglycemia-induced) causes. Thus, in the full-blown clinical syndrome, all patients have both findings, and both contribute in varying degrees to the severity of the hyperglycemia. A variety of metabolic studies have demonstrated impaired β-cell function and insulin resistance in family members of subjects with type 2 diabetes,188,189 and in identical twins of type 2 diabetic patients with normal glucose tolerance, who have a high likelihood of developing the syndrome.190 It is still not clear in any individual patient, which of these two disorders began first, how much is related to the environment and to genetic dysfunction, and how much of the full-blown syndrome is secondary to the hyperglycemia. There is evidence that both β-cell function191 and insulin sensitivity192 are familial traits in certain populations, and abnormalities of insulin secretion and action have been reported to predate the onset of glucose intolerance.188,189 Nevertheless, there remain many uncertainties regarding progression from normal to impaired glucose tolerance and eventually to fasting hyperglycemia and clinical type 2 diabetes mellitus. For this reason, it has been difficult to identify the genetic defects that are presumed to underlie much of the risk of type 2 diabetes. The most convincing evidence of a genetic factor is derived from twin studies.193–197 Although there is some variation in reported risk when comparing monozygotic and dizygotic twins, concordance is always higher for monozygotic twins than dizygotic twins. Initial studies reported almost 100 percent concordance in monozygotic twins and roughly 50 percent in dizygotic twins, but ascertainment bias was believed to be a serious problem in the generation of this data. More recent studies suggest that concordance is more likely to be in the 60 to 70 percent range in monozygotic twins and in the 10 to 20 percent range in dizygotic twins. Thus, the relative risk of developing type 2 diabetes in the sib of an identical twin with the disease was increased 2 to 3.4 times compared with nonidentical twins, and this increases with age. Further evidence for genetic factors is the familial aggregation of type 2 diabetes in a variety of ethnic groups and a variation of prevalence among these groups, ranging from as high as 80 percent in subjects such as the Pima Indians whose parents both had early onset diabetes, to the estimates of approximately 5 percent in Caucasian-Americans, 10 percent in African-Americans, 25 percent in Mexican-Americans, and 35 percent in Pima Indians overall.198 The history of metabolic studies of insulin resistance and insulin secretion in relatives of patients with type 2 diabetes is complex. Suffice it to say that one can quote a 1989 study that found impaired insulin secretion and insulin resistance to be present in impaired glucose tolerance, but that insulin resistance was the only abnormality present in the normal glucose-tolerant first-degree relatives of patients with type 2 diabetes, although this was not universal.199 Similarly, long-term follow-up of 155 normal glucose-tolerant offspring of two parents with type 2 diabetes, compared with 186 normal control subjects followed for an average of 13 years, demonstrated low glucose clearance and high serum-insulin levels (a marker for insulin resistance) as independent risk factors for the development of diabetes among these offspring.200 A 1992 study of Mexican-American children of type 2 diabetic parents with normal glucose tolerance demonstrated basal hyperinsulinemia and hyperinsulinemic responses to intravenous glucose challenge during a hyperinsulinemic glucose clamp, consistent with insulin resistance prior to the development of hyperglycemia.201 However, in 1995, studies of identical Caucasian twin pairs from Denmark who were discordant for type 2 diabetes demonstrated a significant impairment of insulin secretion and insulin action of roughly equal severity in normal glucose-tolerant twins, with both defects more severe in the twins with overt hyperglycemia.190 And in 1995, a test of 100 subjects of European ancestry with normal glucose tolerance with or without a first-degree type 2 relative, demonstrated a high proportion of first-degree affected relatives with impairment of either first, second, or both phases of insulin secretion, and only rare subjects with impaired insulin action.202 A similar study of lean first-degree relatives drawn from newspaper advertisements in New Haven, Connecticut, also supported both possibilities with the suggestion that insulin resistance was present in many, but that impaired insulin secretion was present in those with the greatest degrees of insulin resistance.203 In a subsequent study, glucose-tolerant Caucasian first-degree relatives were compared with normal glucose-tolerant control subjects using a glucose clamp and oral glucose tolerance test (OGTT).204 First-phase glucose-induced insulin secretion was not different, but second-phase was reduced. The insulin sensitivity index was not different. However, a different study in Caucasians found hyperinsulinemia independent of obesity in first-degree relatives during a 75 g OGTT.205 In the same year, another study of offspring of type 2 diabetics was reported.206 Twenty offspring from parents with low basal C-peptide levels were compared with 18 offspring of high-basal C-peptide parents and 14 controls without a family history. The frequency of IGT was 50 percent in both offspring groups. Acute insulin responses were low in the offspring with low-basal C-peptide (insulin-sensitive) parents and insulin resistance was found-during a euglycemic clamp in the offspring from the high-basal C-peptide (insulin-resistant) parents. The authors conclude that both insulin resistance and impaired β-cell function are independent heritable contributors to type 2 diabetes risk. Thus, the studies continue to show first-degree relatives and monozygotic twins to be at risk and to have abnormalities in insulin secretion and sensitivity, but the type of abnormality and the degree of abnormality appears to vary from study to study, perhaps depending on the population studied, including its age, sex, and racial composition.
For this reason, genetic screening has used diverse phenotypes including clinical diabetes mellitus and specific physiological variables for the testing of candidate genes in both the insulin sensitivity and the insulin secretion pathways as potential contributors to the syndrome. This requires careful and complete phenotyping, which may be difficult and expensive. In addition, positional cloning has been attempted after locating susceptibility genes from either a genome scan using random markers, or discrete regions of a chromosome that are strongly associated with the disease, or one of the physiological markers believed to contribute to the etiology of the disease. The generally disappointing results have now led to the common belief that type 2 diabetes is a polygenic disorder which probably results from several combined gene defects that interact with environmental factors to produce the eventual clinical syndrome. This being the case, it's perhaps not surprising that none of these genetic approaches has been particularly successful in unraveling the important, underlying etiologic factors in late onset type 2 diabetes mellitus. Nevertheless, success using the MODY phenotype has encouraged continuing efforts using more powerful techniques. To date, the study of candidate genes is the most commonly used approach. The reasons for candidacy are numerous. We discuss only those related to insulin secretion. All of the MODY genes discovered so far have been found to result from heterozygous mutations that impair insulin secretion. Two possible approaches are available.207,208 One relies on sib-pair analysis with the expectation that sharing among sibs is greater than chance for genetic markers associated with diabetes, insulin resistance, or islet dysfunction. As only affected sibs are analyzed, no assumptions are necessary about the mode of inheritance, and defining the nonaffected phenotype is avoided. However, a high degree of heterozygosity is required to optimize sib-pair analysis and identification of those affected is arbitrary, being based on a continuous physiological variable. Identity by descent or identity by state can be utilized. Considering the age of onset of type 2 diabetes, identity by state is the most frequent method, but this requires determination of the frequency of the alleles in an appropriately matched population, and may require a large number of pairs. Until recently, candidate genes have usually been evaluated in populations of unrelated individuals. These studies require careful clinical and metabolic phenotyping in well-matched case and control populations, especially regarding ethnicity and differences in genetic background. Today, many genes have been cloned and sequenced using molecular scanning methods designed to detect changes in nucleotide sequence in fragments of DNA, such as single-strand conformation polymorphism (SSCP) or heteroduplex analysis. An association between a mutation or polymorphism with β-cell dysfunction, impaired glucose tolerance, or diabetes can then be analyzed.
All of the known MODY genes have been considered as possible candidates for gene defects in late onset type 2 diabetes mellitus. Positive associations between type 2 diabetes and particular glucokinase polymorphisms were originally observed in American Blacks and Mauritian Creoles,209,210 suggesting that the glucokinase locus might be implicated in diabetes in these populations. However, mutations in the coding region were not found. Variants were observed in the liver promoter at −258 and the islet β-cell promoter at −30. A block transversion mutation near this latter region in the rat promoter resulted in a 22 percent reduction in promoter activity in vitro,211 and the region of the −30 base is known to be completely homologous between rat and human. However, the functional importance of this region in humans is not certain. This variant promoter was associated with an increased frequency of impaired glucose tolerance in a group of middle-aged Japanese men who were surveyed in a population under study in Seattle, and confirmed when this population was resurveyed 5 years later.212 Further analysis of the population demonstrated an association between this variant and a 30 percent reduction in early insulin responses to oral glucose challenge. This was true whether or not the individual had normal or impaired glucose tolerance. Thus, it was concluded that the −30 promoter variant may be an example of a common genetic variation that increases the risk of impaired glucose tolerance and contributes to the increased frequency of type 2 diabetes in this population, but that the pathophysiology of impaired glucose tolerance is heterogeneous even in this population, and, therefore, multiple factors are probably interacting to produce clinical hyperglycemia. Such subtle changes may be very difficult to detect using standard genetic approaches. It is likely that regulatory regions of genes will have to be scanned, and intermediate phenotypes, such as β-cell function and insulin sensitivity, will be important. For this phenotyping, it is necessary to recognize that insulin sensitivity alters β-cell responses to glucose, and therefore the assessment of β-cell function can be complex. This may explain why populations do not show linkage between the glucokinase locus when clinical diabetes is used as the phenotype,191,213–216 but may show association with glucose intolerance.212,217 However, a recent study of β-cell function in Danes showed no association with the promoter variant.218
The High-Affinity β-Cell Sulfonylurea Receptor (SUR-1).
This molecule couples with an inwardly-rectifying potassium channel (Kir6.2) and acts as a regulatory subunit for coupling glucose metabolism through its nucleotide binding sites for ATP and ADP to regulate the potassium channel and membrane conductance for control of calcium influx and insulin secretion.219 Its initial cloning in 1995 and localization to 11p15.1220 suggested the possibility that mutation might be associated with familial persistent hyperinsulinemic hyperglycemia of infancy (PHHI), an autosomal recessive disorder characterized by unregulated insulin secretion and congenital hypoglycemia known to localize to this region. Two separate SUR-1 gene splice-site mutations, which segregated with the disease phenotype, were identified in affected individuals from nine different families.221 Both mutations resulted in aberrant processing of the RNA sequence and disruption of the putative second nucleotide-binding domain of the SUR protein. A third mutation was subsequently described in another series of families involving a deletion of codon F1388.222 This particular mutation was expressed and shown to abolish ATP-sensitive potassium channel activity in intact cells. Subsequently, a nonsense mutation in the inward rectifier potassium channel gene was found to be associated with the same phenotype.223 A nonsense mutation, a tyrosine → stop at codon 12 (Y12X), was also observed in the homozygous state in a single proband.223 Only one of 78 PHHI probands was positive for this mutation. Subsequently, a series of missense mutations in the SUR-1 molecule was described: H125Q, N188S, F591L, T1139M, R1215Q, G1382S, and R1394H.224 Most of these generated functional channels in the absence of ATP. However, all but N188S and H125Q had a reduced response to stimulation by MgADP.
Because of the importance of the K channel protein and the β-cell sulfonylurea (SU) binding protein SUR-1 to the KATP channel, both have been screened for their contribution to type 2 diabetes. In 346 Mexican-American type 2 affected sib-pairs derived from 176 families and an additional 110 ethnically and geographically matched control subjects, no evidence of linkage based on allele sharing for SUR-1 was found.225 There was no evidence for association based on allele frequencies in patients and control subjects for any microsatellite marker. It was suggested that the SUR gene was not likely to play a major role in this population. One year later, SSCP was used to identify sequence changes in the two nuclear binding site regions of the SUR-1 gene. Eight variants were found and three were evaluated in two northern European Caucasian populations from Utah and the U.K.226 A silent variation in exon 22, ACC → ACT at nucleotide 761, was more common in patients than in controls with an odds ratio of 3, and an intronic T → C change at position −3 of the exon 24 splice-acceptor site was also more common with an odds ratio of 1.9. The combination occurred in 8.9 percent of patients and in 0.5 percent of control subjects. It was suggested that this locus might be a contributor to type 2 diabetes in northern European Caucasians. A subsequent study partly confirmed an association in type 2 diabetic subjects between morbid obesity and the exon 22 variant at codon 761, although not with the intron 24 variant, except that individuals homozygous for this variant had a more deleterious form of obesity.227 Sib-pair linkage with type 2 diabetes in French Caucasian families gave no evidence for linkage to the SUR locus, except in one set of obese families. In 1998, 100 Japanese type 2 diabetic patients had the entire coding region examined with the identification of 5 amino acid substitutions and 17 silent mutations.228 No mutation altered the properties of expressed channels, and it was eventually concluded that SUR-1 mutations did not appear to be major determinants of type 2 susceptibility in Japanese. Finally, 63 type 2 diabetic patients were completely screened from a Danish population.229 Two missense variants were identified, five silent variants in the coding region, and four intron variants. These variants were examined for association with type 2 diabetes and for a possible influence on insulin and C-peptide secretion after intravenous glucose and tolbutamide in a random sample of unrelated, healthy Danish Caucasians. One variant, a C/T in the silent exon 18 at position 775, and an intron −3c/−3t were found together more frequently in type 2 diabetics than in control subjects. Ten of 300 unrelated, healthy young Danish Caucasians had the combined at-risk genotype associated with a 50 percent reduction in serum C-peptide and a 40 percent reduction in serum insulin responses to tolbutamide, but not glucose injection. In summary, while the sulfonylurea receptor mutants are clearly important to PHHI, so far there is only a possible association with type 2 diabetes mellitus in some populations. Similar screening of the associated potassium channel Kir6.2(Bir) in a small group of 35 Caucasian type 2 subjects detected 5 variants, but subsequent analysis of 306 affected people and 175 controls detected no type 2 diabetes disease association.230
GLUT2 is a high Km (low affinity) glucose transporter expressed in liver, pancreatic islet β-cells, kidney, and intestine. Coupled with glucokinase, GLUT2 has been proposed to act as part of the glucose-sensing mechanism in β-cells.231 In animal models of type 2 diabetes, GLUT2 is suppressed, and it has therefore been considered a candidate for a type 2 diabetes risk factor. However, in animals the levels of GLUT2 expressed are usually such that glucose transport is not rate limiting, requiring an almost complete absence of GLUT2 in order to impair glucose-induced insulin secretion. Under certain circumstances, where GLUT2 is suppressed in animals or in vitro, there has been retention of glucose-induced insulin secretion.232,233 In addition, it has been suggested that GLUT2 is not a major β-cell glucose transporter in the human islet as opposed to rodents and many other species.234 Nevertheless, considerable screening activity has been done. Initial analyses of the GLUT2 locus by population-association studies and linkage analyses (reviewed in Matsubara235) failed to show any evidence that it contributes to susceptibility to type 2 diabetes, except for one association study in which molecular scanning of the GLUT2 gene in 48 former gestational diabetic African-American women with overt type 2 diabetes revealed two amino acid substitutions—one in exon 3 and the other in exon 4B.236 The exon 3 mutation, a C → T Thr 110 Ile, was not more frequent in type 2 patients, but the exon 4 mutation was present in one type 2 diabetic patient, a Val 197 Ile substitution, and was not present in 52 control subjects. Subsequent expression studies demonstrated that the C110T mutation had no effect on glucose transport, but that the G197A mutation abolished transport activity of GLUT2 in Xenopus oocytes.237 No clinical data are available on this patient and no further cases or follow-up have been reported. Subsequently, sib-pair analysis in Pima Indians demonstrated significant linkage between GLUT2 and acute insulin response, but no linkage with type 2 diabetes.191 Studies in a Japanese group of pedigrees with affected members showed a possible association between a polymorphic marker near the GLUT2 locus, but follow-up in 60 subjects with complete gene scans found three Thr110Ile mutations and one Pro68Leu mutation, which were not associated with disease.235 In summary, one rare patient has a defective mutant gene, but common association with disease is unlikely.
In 1995, a screening study of the glucagon receptor gene was reported in a population of Sardinian and French families.238 Eighty multiplex French type 2 diabetic families were investigated first. The SSCP method revealed six polymorphisms: four in introns, and one silent change in an exon. In exon 2, a single heterozygous missense mutation GGT40 → AGT40 (Gly40Ser) was identified in 5 of the 80 tested. Subsequently, a further 136 unrelated patients from additional families and 96 unrelated patients from Sardinia were screened. Five of the French subjects were positive, as were 8 of the 96 Sardinian patients. The overall frequency was 4.6 percent in the French and 8.3 percent in Sardinians. Three control groups were tested. Of 159 middle-aged normoglycemic French Caucasians with no family history of diabetes, none was found to carry the mutation; of 183 presumed nondiabetic, nonhypertensive French subjects with no known familial diabetic background, 1 individual was positive; and of 780 randomly chosen individuals, ages 25 to 70, who considered themselves to be nondiabetic upon questioning, 9 (1.2 percent) were carriers. Then, 96 randomly selected individuals from Sardinia were screened. Three were positive. The trend of the two populations was similar. When tested together there was a highly significant association with type 2 diabetes. Linkage was tested using the transmission/disequilibrium (TD) test. The estimated transmission frequency was 0.727, which was significantly higher than the expected value of 0.5. This required subjects with impaired glucose tolerance (IGT), impaired fasting glucose (IFG), or type 2 diabetes to be considered affected, although IGT and IFG combined were also significant. Overt type 2 diabetes assessed alone was not. This receptor variant was expressed in BHK cells and transfected clones tested for binding. Compared to wild-type receptor, labeled glucagon had a threefold lower affinity for the mutant.
Unfortunately, subsequent studies were quite variable and often not confirmatory. Analysis of 383 type 2 diabetic patients in Japan found no one with the Gly40Ser mutation,239 and similar findings were reported among diabetic patients in the Netherlands.240 A study of diabetes in the United Kingdom found this mutation to be present in 15 of 1691 patients (2.2 percent), but only in 1 of 425 (0.2 percent) geographically matched controls. However, an increased frequency of the mutation was also found in probands of type 1 diabetes, and it was suggested that the mutation might represent some type of founder effect rather than true association.241 A further study of 104 German patients found no one with the mutation,242 and two more Japanese populations were screened with no mutations found in any of the diabetic subjects.242–244 To further confuse the issue, a separate study in hypertension found an increase in the frequency of this same mutation in a population of individuals with hypertension compared to appropriate controls unrelated to diabetes.245 Confirming authors postulated that the renal glucagon receptor might be an explanation for this interaction.246 However, no such relationship was found in hypertensive Sardinian type 2 diabetic subjects, compared to nonhypertensive Sardinian type 2 diabetic subjects.247 However, a more detailed look at the Sardinian population revealed significant geographic variation within Sardinia in the frequency of this mutation, varying from 1.4 percent in central Sardinia to 7.6 percent in southern Sardinia. Moreover, it was again noticed in a nondiabetic population that there was no relationship between the mutation and hypertension. At this time, 3.6 percent of type 2 diabetic subjects were found to have the mutation, and 4.2 percent of nondiabetic subjects, suggesting that the original report might have been skewed by variable geographic backgrounds in the original study. However, 10 subjects with the mutation were carefully matched for diabetic state, body mass index (BMI), age, sex, and geographic origin, with 10 patients without the mutation, and a glucagon infusion test was performed to study the ability of glucagon to cause hyperglycemia.248 Carriers had a significantly lower increment in plasma glucose than wild-type subjects, which was consistent with the reported decreased binding affinity of the mutant in vitro.249
Lastly, a Russian group screened its population of 150 unrelated type 2 diabetic patients, and compared it with 109 nondiabetic subjects for the mutation.250 None of the type 2 diabetics showed the mutation, whereas two of the control subjects were carriers. Thus, it seems unlikely that this mutation is associated with clinical hyperglycemia. It should be remembered, however, that the original report from Sardinia was based largely on the definition of affected as impaired fasting glucose and/or overt diabetes and was not statistically significant when the overt diabetic population was tested alone. Similar association data have been reported with the −30 β-cell glucokinase promoter (see above). This suggests that there may be genetic factors that predispose to impaired glucose tolerance whose impact on blood glucose is modest. Because they are common variants, this effect does not become noticeable when surveying the entire population and comparing overt diabetes while potentially including impaired glucose tolerance and impaired fasting hyperglycemia among the nonaffected populations. Thus, detailed phenotypic characterization may be critical if such variation is a significant contributor to the eventual development of risk for type 2 diabetes in a population.
A missense mutation of the IAPP or amylin gene (S20G) was reported in 12 of 294 Japanese type 2 diabetes patients screened.78 None of the 187 controls or 59 type 1 patients was affected. Eight of the 12 were diagnosed under age 35, had severe hyperglycemia requiring insulin treatment, and all had strong family histories of type 2 diabetes. The majority of circulating hormone (84 percent) was mutant by HPLC. The parent who transmitted the mutation was thought to be nondiabetic by history, but had IGT by testing in the three families studied. The authors concluded that the mutation probably only produced IGT unless associated with other unknown susceptibility factors where it led to early onset and increased severity of hyperglycemia or impaired β-cell function. No other studies of this mutation have been reported.
As mentioned in the MODY diabetes section above, several of the MODY genes have been associated with late onset type 2 diabetes, including the mutations usually associated with the MODY syndrome. This includes a Val → Met 255 mutation in HNF-4α153 and a valine → isoleucine mutation at codon 393.152 Similarly, an Arg583Gln mutation in the HNF-1α was found in 2 of 245 type 2 diabetic subjects in a Danish population, but in none of the appropriate controls.251
A general screen of 19 genes was performed using associated polymorphic markers, including 10 expressed in pancreatic β-cells. Included were two G-coupled inwardly rectifying potassium channels (GIRK-1) (KCNJ-3) and (GIRK-2)(KCNJ-7), plus glucagon; the glucokinase regulatory protein; the glucagon-like protein-1 receptor (GLP-1R); the insulin gene transactivator islet-1 (ISL-1) and the candal-type homeodomain-3 gene (CDX-3); the proprotein convertase-2 (PCSK-2); hexokinase-1 (HK-1) and the mitochondrial isoenzyme (FAD-glycerophosphate dehydrogenase (GPD-2); as well as the isoform of the pyruvate kinase (PKN), which is expressed in β-cells.252 The screening was performed in 172 French families selected to have at least two affected sibs. All available sibs, as well as parents, were included. The study population included 751 subjects of whom 573 were considered affected. Individuals were considered affected if they had overt type 2 diabetes mellitus according to the World Health Organization (WHO) criteria, IGT, or impaired fasting glucose with fasting plasma glucose levels greater than 6.1 (110 mg/dL) and a 2-h postload value of >7.8 mM (140 mg/dL). No evidence for linkage was found for any of the genes screened, but suggestive evidence of linkage was observed for the cholecystokinin-B (CCK-B) receptor gene, which is localized at 11p-15.4, in individuals diagnosed with overt hyperglycemia under the age of 45 years. No mutational screening was performed at the time of this publication.
In summary, candidate genes relating to the β-cell have been screened in a variety of populations, but variable results, often conflicting, have been reported, partly attributable to factors including inadequate sample size, heterogeneous populations, ascertainment bias, genetic heterogeneity, inadequate phenotypic characterization, and methodologic problems. Total genome scans have been equally disappointing to date, although not nearly as extensive. This is in marked contrast to MODY diabetes in which this clearly genetically distinct group, which is identified by its clinical genetic characteristics, has begun to reveal its secrets, although even in this much-easier-to-divide-and-segregate subgroup, all of the genes contributing have yet to be discovered, suggesting that there is much work yet to be done.
There are five groups that have assembled the necessary resources, including identification of a suitable population, clinical phenotyping, metabolic studies, and genetic analytic techniques, which are pursuing this approach. However, because the populations vary and the selection criteria are different, it is likely that the outcomes will be different. The first group reported on a population of 217 individuals from 26 families screened from 4000 individuals in a population isolated in western Finland (Botnia), an area of Swedes living in Finland.253 The total population of the region is 60,000. Over a 5-year period, families with a positive diabetes history at four health centers were recruited to participate in clinical characterization studies with an overall participation rate of 90 percent. In total, 1180 type 2 diabetic patients and 3005 family members were phenotyped. Families containing 3 or more affected individuals, with at least 1 subject having disease onset before age 60 years and another before age 65 years, were selected for analysis. A 10-cM sequence length polymorphic map was made on the 217 individuals and a nonparametric method used for evaluation. This analysis revealed no significant or even suggestive evidence for susceptibility loci predisposing to type 2 diabetes. However, to further phenotype individuals, they selected the early insulin response during the OGTT (30-min insulin level) among affected individuals as an indicator of insulin secretion to rank families. Families were partitioned into four quartiles and the data reanalyzed. In the quartile with the lowest insulin levels, a region near D12S366 on chromosome 12q was highly significant and, with additional markers in this region, evidence for linkage was observed with an approximate LOD score of 3.65. This is the region of the MODY-3 gene, but the lowest quartile families do not fit the usual phenotypic definition of MODY-3 families. Subsequent screening of 86 of these subjects failed to demonstrate a mutation in the HNF-1α gene, suggesting that a nearby gene may be involved.
At about the same time, another sib-pair analysis was performed on a group of Mexican-American families from Starr County, Texas.254 This group consisted of 408 Mexican-Americans with 330 affected sib-pairs from 170 sibships, located along the Rio Grande River, bordering Mexico. Native-American genes are highly represented among this population. The average distance of the polymorphic markers was 9 cM. Parents were not typed. Allele frequencies were estimated from the family sample. Evidence for linkage was evaluated using two different two-point linkage test statistics. The maximum LOD score was computed using the estimated identity-by-descent probabilities and the possible triangle using an identity-by-state method. One marker at D2S125 showed suggestive evidence for linkage with a LOD score >2.6, P < 0.0005. MLS values were increased, but not significantly for other markers in the region. In the identity-by-state tests, this marker had the best evidence for linkage as well. The affected were diagnosed by interview if they were currently taking blood-glucose-lowering medications or had taken such in the past for more than 1 year and were diagnosed over age 30 years. No metabolic measurements were made. Follow-up studies were made in another group of Mexican-Americans consisting of 166 individuals from 76 sibships for a total of 110 affected sib-pairs, and were compared with a group of Japanese affected sib-pairs consisting of 213 individuals from 97 sibships for a total of 140 affected sib-pairs using WHO criteria and an oral glucose tolerance test to determine the affected or current oral hypoglycemic agent or insulin use. Several markers showed some evidence for linkage with type 2 diabetes in this second sample. When combined, only D2S125 met the genome-wide criteria with significant evidence for linkage with identity by state (IBS) chi-square and MLS values in the combined sample of 30.25 (p < 1 × 10−6) and 4.10 (p < 0.0001). There was no evidence for linkage in the Japanese population or in a group of heterogeneous non-Hispanic Caucasians who were also screened. All of the known MODY genes, and many of the other candidate genes previously discussed, were also screened, but there was no evidence for linkage in this Mexican-American population. D2S125 is located at 2Q37. No obvious candidate genes are located in this region. A subsequent report of a British affected sib-pair population of 109 type 2 diabetics found no evidence for linkage.255 A recent genome-scan report in a Finnish population, reported by the FUSION group, excluded linkage with this region.256 A recent abstract262 suggests a common single-nucleotide polymorphism UCSNP-43 in the calpain gene (a non-lysosomal cysteine protease) near DS125 showed linkage and association with diabetes. The authors further describe other polymorphisms in the gene which significantly modify risks. None of these polymorphisms were in the coding region of the gene and one haplotype with three polymorphisms (UCSNP-43, -19, and -63) were minimally required to distinguish the high-risk haplotype.
A similar sib-pair analysis of 579 individuals in 32 families of low-income Mexican-American diabetics identified in an epidemiological survey carried out in San Antonio, Texas, eventually yielded 103 sibships containing 349 individuals, 83 of whom were diabetic, yielding 545 sib-pairs. A LOD score >2 was observed for a marker on chromosome 11 and chromosome 6 for 2-h glucose and 1-h glucose, respectively.257 There was no evidence for linkage at D2S125.
Most recently, an autosomal genome scan for loci linked to prediabetes phenotypes in Pima Indians was reported.258 Three hundred and sixty-three nondiabetic Pima Indians were genotyped at 516 polymorphic microsatellite markers on all 22 autosomes. Each was chosen because he or she had a full sib who also participated in the studies. A subset of 231 individuals with fasting and 2-h glucose concentrations <7.8 mM during a 75 g oral glucose tolerance test was identified for analyses related to insulin secretion. An intravenous glucose tolerance test and a two-step hyperinsulinemic euglycemic glucose clamp were also performed on all subjects in addition to a large number of other measurements. Seven markers had a positive LOD score of 1.5 or greater for glucose tolerance and four additional markers had LOD scores >2 for fasting and 2-h insulin concentration. There were no LOD scores over 2 for fasting or 2-h glucose concentrations, or for insulin secretion, but there were two LOD scores >2 for insulin action. None of these scores was considered statistically significant for a full genome scan, which has been suggested to require a LOD score of 3.6 for a 5 percent probability. Unfortunately, there was no evidence that any of the physiological markers were linked to D2S125 on distal 2q, which had been identified in Mexican-Americans from Starr County, Texas (a genetically similar group). In addition, they failed to replicate their own data previously reported on linkage of acute insulin response to an area on 1p or linkage to insulin action at 4q. There was no linkage to markers located at the MODY-1, -2, or -3 genes, and there was no linkage to 2-h or fasting glucose concentrations on chromosomes 6 and 11, as reported for Mexican-Americans from San Antonio, Texas. The q-terminal region D22S683-D22S270 mapping to 22q12-13 provided the most significant evidence of linkage to a prediabetic phenotype in the entire genome scan (for fasting plasma glucose concentration). Known genes in this region encode the somatostatin receptor 3 and the structural locus for casein kinase-1 epsilon, neither of which is known to be involved in carbohydrate metabolism directly.
The last major group to propose such work reported its initial research design and metabolic findings in 1998.259 The FUSION project is a collaborative study between the National Human Genome Research Institute at the NIH and the Department of Epidemiology and Health Promotion at the National Public Health Institute in Helsinki, Finland, in addition to a variety of other academic colleagues in both countries. Based on the genetic isolation of the Finnish population, they have ascertained 495 families with confirmed type 2 diabetes in at least two sibs and no history of type 1 diabetes among first-degree relatives. In a subset, a spouse and offspring were sampled, and participated in an intravenous glucose tolerance test to measure insulin secretion and insulin action. This testing was performed in two nondiabetic offspring in 156 families, with a confirmed nondiabetic spouse. Type 2 diabetes was confirmed by oral glucose tolerance test if no diabetes medication was being given, using WHO criteria. A family was eligible if a proband was diagnosed between age 35 and 60 years and at least one parent was apparently nondiabetic. The study is expected to present data very shortly. A second group for confirming any findings is planned.
In summary, whole genome searches using the affected sib-pair method have been undertaken by a variety of groups to complement the ongoing work on candidate genes. The results to date have been modest, largely because of the difficulties of the effort. It is pointed out by the FUSION group that 400 pairs typed on a 10-cM map will require a 1.8-fold excess risk to siblings of affected individuals at a LOD score of 3.0 with 82 percent power. However, this drops to a LOD score of 1.0 for a 1.4-fold excess risk. It is also pointed out that it would require 800 such pairs to have 70 percent power to detect a locus with an impact estimated by the Starr County investigators for their chromosome 2 finding to produce a LOD score of 3.0, and even this would be reduced to a LOD score of 1.59 if the locus only confers a 1.2-fold excess risk. Thus, it is becoming obvious that detection of genes for type 2 diabetes is going to be extremely complex and difficult if the impact of these genes is relatively modest and their frequencies relatively low or if they interact in some way, which is likely. This probably explains why even the initial efforts have not replicated each other, and no truly significant LOD scores have been observed. This will almost certainly require pooling of data among investigative groups. Fortunately, such efforts have apparently begun.