X-Chromosome Gene Dosage and the Risk of Diabetes in Turner Syndrome

Abstract

Background: Turner syndrome (TS) is caused by the absence or fragmentation of the second sex chromosome. An increased risk of diabetes mellitus (DM) has consistently been noted, but the specific phenotype and genetic etiology of this trait are unknown.

Methods: In a prospective study, we examined the prevalence of DM in adult participants in an intramural National Institutes of Health (NIH) TS study. Results were analyzed with respect to karyotype, age, body mass index (BMI), and autoimmune indices. Insulin sensitivity and secretion were compared in age- and BMI-matched euglycemic women with TS and healthy female controls. We compared gene expression profiles in lymphocytes from differentially affected TS groups.

Results: Type 2 DM was present in 56 of 224 (25%) of the women with TS; type 1 DM was found in only one woman (<0.5%). DM was more prevalent among women with an isoXq chromosome compared to X monosomy (40.0 vs. 17.3%; P = 0.004). Euglycemic women with TS (n = 72; age, 33 ± 12 yr; BMI, 23 ± 3 kg/m²) had significantly higher glycemic and lower insulin responses to OGTT, with insulin sensitivity similar to controls. Gene expression profiles comparing 46,X,i(X)q vs. 45,X groups showed a significant increase in Xq transcripts and in potentially diabetogenic autosomal transcripts in the isoXq group.

Conclusion: Type 2 DM associated with deficient insulin release is significantly increased among women with monosomy for the X-chromosome but is increased even more among women with monosomy for Xp coupled with trisomy for Xq. These data suggest that haploinsufficiency for unknown Xp genes increases risk for DM and that excess dosage of Xq genes compounds the risk.

In Turner syndrome, haploinsufficiency for unknown Xp chromosome genes increases the risk for diabetes, and excess dosage of Xq genes compounds that risk.

Turner syndrome (TS) is due to loss of all or a significant part of one sex chromosome in a phenotypic female, usually presenting with short stature and premature ovarian failure. An increased frequency of diabetes mellitus (DM) in TS was first noted almost 50 yr ago by Ann Forbes and Eric Engel (1).

A high rate of DM was also noted in men with Klinefelter syndrome (47,XXY) (2), leading to the suggestion that DM in parents caused sex chromosome anomaly—as well as an increased risk for DM in offspring (3). However, extensive documentation of family history for a large group of girls and women with TS revealed that the prevalence of DM in first-degree relatives was similar to the general population (4).

The specific DM phenotype in TS remains unclear. A Danish registry study found an 11-fold increase in type 1 DM among TS patients and a 3- to 4-fold increase in type 2 DM (5). However, studies based on adult endocrine clinic populations describe a phenotype of gradually progressive, adult-onset glucose intolerance more typical of type 2 DM (6,7,8,9). A recent epidemiological study of mortality in TS found an 11-fold increased risk of death related to diabetes without specifying type 1 or 2 (10). To investigate the possibility that excess DM in TS could be explained by increased adiposity or altered body composition associated with premature ovarian failure, we previously compared nonobese, young women with TS, with no prior history of impaired glucose homeostasis, to a group of age- and body mass index (BMI)-matched women with 46,XX premature ovarian failure (11). Thirty-six percent of the TS group but no controls demonstrated impaired glucose tolerance (IGT) due to inadequate insulin secretion in response to iv or oral glucose, whereas insulin sensitivity was normal and similar in both groups (11).

None of the previous studies have had sufficient numbers or depth of clinical, genetic, and laboratory information to clarify the phenotype or analyze the contribution of specific X-chromosome dosage to the diabetic phenotype in TS. In the present study, we have attempted to clarify the clinical features and genetic origins of DM in TS in a large group of women with well-characterized 50-cell karyotypes.

Subjects and Methods

Study subjects were 224 consecutive adult participants (age range, 18–67 yr) in the National Institute of Child Health and Human Development (NICHD) natural history study on TS. The study includes phenotypic females with a 50-cell peripheral karyotype in which more than 70% of cells demonstrate loss of all or part of the second sex chromosome, without prespecified criteria regarding metabolic disorders. The recruitment was through the NIH website (http://turners.nichd.nih.gov/). The study was approved by the NICHD institutional review board, and all participants gave written informed consent. Controls were 30 women recruited through the NIH normal volunteer office (mean age, 32.7 yr), without a prior diagnosis of diabetes, who were not taking medications known to influence glycemic control, had normal puberty and menstrual cycles, were not obese, and were healthy overall.

Each participant had a comprehensive medical history, family history, and physical exam, which included standard measurements of height and weight using an SR scale (SR Scales, Tonawanda, NY) with a height rod.

Genotyping

Karyotype was determined by G-banding on 50 peripheral white blood cells for all women with TS. Fluorescence in situ hybridization using X- and Y-specific α satellite DNA probes was employed to characterize marker-chromosomes and ring-chromosomes. To compare the diabetic phenotype distribution between different karyotype groups, we defined the following groups: 1) 45,X, all 50 cells had karyotype 45,X; 2) del(X)(p), included 46,Xdel(X)(p) or 46,Xdel(X)(p)/45,X; 3) del(X)(q), included 46,Xdel(X)(q) or 46,Xdel(X)(q)/45,X; and 4) i(X)(q), included 46, Xi(X)(q) or 46, Xi(X)(q)/45,X.

In each of the last three groups, the fragmented X-chromosome was present in 20 to 100% of the white blood cells.

Gene expression profiling

Blood samples were collected in BD CPT tubes (BD, Franklin Lakes, NJ) and centrifuged at 1800 × g for 30 min. The mononuclear cell layer was separated and frozen at −70 C. Total RNA was isolated using the RNeasy mini kit (QIAGEN, Valencia, CA). Target production and hybridization using the Affymetrix protocol, scanning by the GeneChip Scanner 3000, and data extraction using the GeneChip Operating System v. 1.2 (Affymetrix, Santa Clara, CA) were all performed according to standard protocols at a core facility. CEL files were further analyzed using ChipInspector 1.3 (Genomatix Software GmbH, Munich, Germany) for total intensity normalization, SAM statistical analysis, probe set identification and quantification (at least three probe sets had to be positive for each transcript), with the false discovery rate set close to zero (12). The data set was searched for biological function information using the DAVID online program (13) with the PANTHER biological function tool. In addition, we used Genomatix Bibliosphere Pathway Edition (Genomatix Software GmbH) software to investigate anatomical associations in the altered gene set (14). This text-mining tool is able to find anatomical connections based on co-citations of gene names and specific tissues. We employed the most stringent co-citation filter: “gene. .tissue. .gene” (GFG level B3).

Metabolic profiling

Testing for serum anti-glutamic acid decarboxylase (GAD) antibodies (GAD 65 AB) and anti-islet cell antibodies was performed by RIA at Mayo Laboratories (Rochester, MN).

Three-hour oral glucose tolerance test (oGTT) was administered to all women with TS who did not have a prior history of DM and to 30 controls. After 3 d of ad libitum carbohydrate diet (>250 g/d) and an overnight fast, the participants were given 1.75 g/kg body weight (maximum 75 g) oral dextrose solution. Serum glucose and insulin were measured at 0, 30, 60, 120, and 180 min using the Synchron LX System with oxygen electrode (Beckman Coulter, Fullerton, CA) and immunochemiluminescent assay (Immulite 2000 analyzer; Diagnostic Products Corporation, Los Angeles, CA).

Based on the history, the fasting glycemia, and the glycemia at 2-h of the oGTT, we defined the following categories of glycemic control: 1) DM—the subject had been diagnosed with DM before the study and was taking glucose-lowering medications, or the results from the oGTT indicated DM according to World Health Organization’s 1985 criteria; 2) IGT—2-h blood glucose during oGTT of at least 140 mg/dl but less than 200 mg/dl, and fasting glycemia less than 126 mg/dl; and 3) normal glycemic control—no history of DM, fasting glycemia less than 100 mg/dl, and 2-h glycemia during oGTT less than 140 mg/dl.

Using serum glucose and insulin levels from the oGTT, we calculated the following parameters: 1) QUICKI (quantitative insulin-sensitivity check index) = 1/[log(I₀) + log(G₀)] (15), representing fasting insulin sensitivity; 2) insulin sensitivity index (ISI_M) as described by Matsuda and DeFronzo (16) = 10,000/√(G₀∗I₀∗I _{mean 0–180 min}∗G _{mean 0–180 min}), representing global, i.e. fasting plus postglucose load insulin sensitivity; 3) insulin secretory index (ΔI₃₀/ΔG₃₀) = (I_{30 min} − I_{0 min})/(G_{30 min} − G_{0 min}), representing insulin release relative to rise in glycemia; and 4) glucose area under the curve (AUC_G) during 180 min of oGTT, calculated using the trapezoid rule.

IGF2 levels were measured at the NIH Clinical Center using chemiluminescence immunoassay on Siemens Immulite 2500 analyzer (Siemens Healthcare Diagnostics Inc., Deerfield, IL). C-Reactive protein (CRP) levels were measured with Allied Biotech’s Human Cardiovascular Microarray Kit (Allied Biotech Inc., Vallejo, CA).

Statistics

Continuous data are presented as mean ± sd. Nominal data are presented as numbers, percentage and 95% confidence intervals (CI), where appropriate.

Comparison between group means was performed by ANOVA or analysis of covariance, with age and BMI as covariates. Comparison between proportions was performed by Z-test for proportions. We used a logistic regression model to evaluate the association of DM (nominal dependent variable) with karyotype [45,X; del(X)(q); del(X)(p); and i(X)(q)] while correcting for age and BMI.

Statistically significant test results were considered those that produced P values < 0.05 or less than a calculated cutoff value using Bonferroni correction for multiple comparisons where appropriate. JMP software was used (SAS Institute, Inc., Cary, NC).

Results

Baseline characteristics of the study population

All 224 women with TS were free from acute health problems. The mean age was 35.4 ± 11.3 yr, and mean BMI was 28.9 ± 7.7 kg/m². Most (91%) study participants were white, and the rest were Asian, African-American, and Hispanic in approximately equal proportions. The great majority (91.5%) had abnormal karyotypes in 100% of their peripheral white blood cells, with only 8.5% having a low-grade mosaicism for a normal 46,XX cell line (2–14% cells with a normal 46,XX karyotype out of 50 cells analyzed). The detailed karyotype distribution is presented in Table 1. A diagnosis of hypothyroidism reported 38.5% (95% CI, 32.2–45.2) of the participants, and overall thyroid autoimmune diseases (previous hypothyroidism, newly diagnosed hypothyroidism, or positive thyroid antibodies) were present in 62.5% (95% CI, 56.0–68.6) of the participants.

Table 1.

Karyotype distribution for 224 women with TS

Karyotype		n (%)
45,X	All 50 cells	110 (49)
IsoXq	46,X,i(X)(q); 45,X/46,X,i(X)(q); 45,X/46,X,idic(X)(p11.2)	49 (22)
Mosaic	45,X/46,XX; 45,X/46,XX/47,XXX	19 (8.5)
DelXp	46,X,del(X)(p); 45,X/46,X,del(X)(p)	13 (5.8)
DelXq	46,Xdel(X)(q); 45,X/46,Xdel(X)(q)	11 (4.9)
RingX	45,X/46,X,r(X)	11 (4.9)
Y	45,X/46,XY or 45,X/46,X fragment Y	6 (2.7)
Other	Mosaic for 45,X and ″marker″ X, or idic(X)(p22)	5 (2.2)

Prevalence of DM

Twenty-seven women enrolled in the study with preexisting DM diagnosed at an average age of 40.5 ± 12.0 yr and with average duration of 4.9 ± 6.2 yr. Twenty-six of them had type 2 diabetes and were treated with diet alone (n = 9), oral medications (n = 13), oral medications and insulin (n = 2), or insulin alone (n = 2). Only one patient carried a diagnosis of type 1 DM, and this was an atypical case, presenting at the age of 32 and initially treated with metformin. Later, 10–15 U glargine insulin was added when it was determined that fasting C-peptide was low and anti-GAD antibodies were elevated. Most patients with a prior diagnosis of DM were well controlled, with the average hemoglobin A1c less than 6.5%.

Thirty women were diagnosed with DM during the study, based on elevated fasting blood glucose or the results of the oGTT. Thus, the overall prevalence of DM in our study population was 25% (57 of 224) (95% CI, 20.2–31.5). This is 3- to 4-fold greater than the Centers for Disease Control (CDC) estimated prevalence of diagnosed and undiagnosed DM in the adult U.S. population for 2007 (www.cdc.gov/diabetes/pubs/pdf/ndfs_2007.pdf). In contrast, the prevalence of type 1 DM among women with TS (1 of 224 or 0.45%) is similar to that of the general U.S. population (CDC, ibid.).

Prevalence of IGT

We found IGT in 23.2% (52 of 224) of the patients. One had combined impaired fasting glucose and IGT; the rest had isolated IGT. None had isolated impaired fasting glucose. Approximately half of the women with TS (109 of 224 or 48.7%; 95% CI, 42.2–55.2) demonstrated abnormal glycemic control (existing DM, newly diagnosed DM, or IGT).

Evolution of impaired glucose homeostasis in TS

To investigate the pathophysiology underlying the high rate of DM in TS, we studied glucose homeostasis in a group of nonobese young women with TS who had normal glucose tolerance in comparison to a group of age- and BMI-matched control women (Table 2). Fasting glucose and insulin sensitivity were very similar in both groups. However, women with TS had higher glycemic response to oral glucose load (AUC_G + 20%; P < 0.001) and a significantly lower insulin secretory response (dI₃₀/dG₃₀, 51%; P = 0.016).

Table 2.

Reduced insulin response to glucose in euglycemic women with TS

	Controls (n = 30)	TS (n = 72)	P
QUICKI	0.40 ± 0.04	0.39 ± 0.04	0.25
ISIM	8.8 ± 4.9	8.1 ± 3.8	0.46
ΔI30/ΔG30(μU/mg)	1.34 ± 1.23	0.89 ± 0.59	0.016
AUCG (g/dl)	17.0 ± 3.1	21.0 ± 2.6	<0.0001

All women had normal fasting glucose and a normal response to the oGTT (i.e. 2-h glucose <140 mg/dl). The mean age was 33.3 ± 12 yr for the TS group and 32.7 ± 12 yr for controls. Mean BMI was 23.3 ± 3.3 kg/m²for the TS group and 23.3 ± 2.9 kg/m² for controls. P was derived from ANOVA. 

A multiple regression analysis showed that IGT and DM were positively associated with increasing age and BMI, and with declining insulin sensitivity and insulin secretion (Table 3).

Table 3.

Progression of impaired glycemic control in TS

	NGT (n = 115)	IGT (n = 52)	DM (n = 57)	Δ(DM − NGT)	PTREND
Age (yr)	32.6 ± 11.3	34.0 ± 9.6	42.4 ± 10.0	+30%	<0.0001
BMI (kg/m2)	26.4 ± 6.4	30.6 ± 7.6	32.3 ± 8.7	+22.3%	<0.0001
QUICKI	0.38 ± 0.04	0.37 ± 0.04	0.35 ± 0.04	−8%	0.0007
ISIM	7.2 ± 3.8	5.3 ± 3.3	4.3 ± 2.5	−40%	<0.0001
ΔI30/ΔG30(μ U/mg)	0.89 ± 0.6	0.71 ± 0.7	0.41 ± 0.5	−54%	<0.0001

NGT, Normal glucose tolerance. 

A family history for DM was not predictive for the phenotype.

Autoimmune thyroid disease was not associated with DM or IGT after correction for age and BMI (P = 0.243 from a multiple logistic regression analysis).

X-chromosome dosage and diabetes prevalence

To define the X-chromosome loci implicated in the TS diabetes phenotype, we divided our study population into four informative karyotype groups, namely: 1) 45,X; 2) delXq; 3) delXp; and 4) isochromosome Xq (iXq). We predicted that if DM in TS was due to haploinsufficiency for gene(s) located on Xp, then DM frequency should be increased in all the karyotype groups characterized by Xp monosomy (45,X; delXp; iXq) and normal in delXq where both Xp arms are retained. Conversely, if haploinsufficiency for Xq genes increased risk for DM, then the prevalence should be high in 45,X and delXq groups, but not in delXp or iXq groups (which have either the normal two copies of Xq or, in the case of iXq, three copies of Xq).

We found that DM prevalence was significantly increased in the “pure” 45,X group and also in the delXp group, but not in the delXq group (Fig. 1), consistent with haploinsufficiency for Xp causing an increased risk for diabetes. It was also apparent that the rate of DM was significantly higher in the iXq vs. the 45,X group (40 vs. 17.3%; P = 0.003 by Z-test for proportions).

Figure 1.

Prevalence of DM in different TS karyotype groups. The numerator in each bar indicates the number of subjects with DM; the denominator indicates the number of subjects in each karyotype group. The karyotype groups were similar in age, BMI, and adiposity. The expected DM prevalence (6.7%) is presented by the shaded area and was calculated from the age-adjusted DM prevalence for the U.S. population in age groups 20–39, 40–59, and 60–74 yr (19). DM prevalence was 17.8% (95% CI, 11.3–25.4%) in the 45,X group and 40% (95% CI, 26–55.4%) in the iXq group (P = 0.003). Odds ratio for DM (iXq vs. 45X) was 3.1 (95% CI, 1.4–7.1); delXp and delXq groups were not included in the statistical comparison because of small sample size. These data suggest that haploinsufficiency for Xp genes (45X, delXp, and iXq groups) confers a significantly increased risk for DM that is amplified in the isoXq group by the presence of additional doses of Xq genes.

Gene expression profiling

The increased risk for DM in women with i(X)(q) could be triggered by overexpression of Xq genes that escape inactivation. To investigate this further, we compared gene expression profiles from lymphocytes of 45,X (n = 10) vs. 46,X,i(X)(q) (n = 5) groups. These groups were similar in age, and all had normal hemoglobin A1c levels. More than 2000 unique transcripts were altered in this comparison (Supplementary Table 1, published as supplemental data on The Endocrine Society’s Journals Online web site at http://jcem.endojournals.org). The most pronounced difference (>16-fold) between these two groups was in the level of XIST expression. XIST (X inactivation-specific transcript; Xq13.2) is a noncoding RNA involved in X-chromosome inactivation, so little expression is expected in monosomic subjects, whereas more abundant expression is predicted for those with additional Xq copies. One hundred other Xq genes demonstrated altered expression levels (Supplementary Table 1). RPS4X (Xq13), which encodes ribosomal protein 4 known to escape X-inactivation (17) was also significantly elevated in the 46,X,i(X)(q) group. Several overexpressed Xq genes are involved in gene expression regulation, e.g. PIN4, VGLL1, and SMARCA1. A few Xq genes were down-regulated in the 46,X,i(X)(q) group. In particular, the antiinflammatory transcription factor gene TSC22D3 (Xq22.3), also known as glucocorticoid-induced leucine zipper, is suppressed, as is XIAP(Xq25), which encodes the X-linked inhibitor of apoptosis that protects β-cells from cytokine- and ER stress-induced programmed cell death (18).

The DAVID bioinformatics resource (13) identified the following functional categories overrepresented by more than 10% (P < 1⁻¹⁰) in the 46,X,i(X)(q) vs. 45,X gene set: “signal transduction; mRNA transcription; mRNA transcription regulation; ion transport; G protein signaling; cation transport; proteolysis and cell surface receptor signaling.” We used the Bibliosphere text-mining program to search literature references to the altered genes in association with specific tissues and found that the top three were “islets of Langerhans,” “pancreas,” and “insulin-producing cells” (Z-scores were 57, 47, 34, respectively). Table 4 summarizes alterations in genes with known roles in β-cell biology and survival, autoimmunity, and insulin processing and signaling. The arrays also demonstrated an increase in transcript levels for CRP—a marker of systemwide inflammation, and the IGF2. Measurement of IGF2 and CRP in the serum of iXq vs. 45,X subjects confirmed gene expression array results (serum IGF2 was 739 ± 34 ng/ml in iXq vs. 557 ± 27 ng/ml in 45,X; P < 0.05; and serum CRP was 2.46 ± 0.79 mg/liter in iXq vs. 1.33 ± 1.1 mg/liter in 45,X; P < 0.001).

Table 4.

Altered autosomal gene expression in 46,Xi(X)(q) vs. 45,X groups

Gene	Entrez gene ID	Change	Relevance
NR2F2	7026	↑	(COUP-TF11) involved with HNF4 in regulating insulin
HNF4	3172	↑	TF involved in β-cell development and MODY-1
FOXA2	3170	↑	TF in β-cell development, and insulin actions (35)
SOX9	6662	↑	TF involved in β-cell development (35)
KLF11	8462	↑	TF involved in β-cell development and insulin secretion (36)
GLIS3	169792	↑	TF involved in β-cell development (37)
MAF	4094	↑	β-cell TF regulates insulin production (38)
DDIT3	1649	↑	(CHOP) β zip protein promotes β-cell apoptosis (18)
MAPK8	5599	↑	(JNK1) kinase involved in β-cell apoptosis (39)
MAP3K1	4214	↑	(MEKK-1) kinase promotes β-cell apoptosis (39)
GAD2	2572	↑	Pancreatic autoantigen
ICA1	3382	↑	Islet cell autoantigen
IGF2	3481	↑	IGF2
CRP	1401	↑	CRP
XBP1	7494	↓	β-cell TF modulates ER stress (40)
IRS2	8660	↓	IRS2
PIK3R1	5295	↓	Phosphoinositide 3-kinase regulatory subunit

Selected autosomal gene transcripts altered in the 46,X,i(Xq) vs. 45,X groups. The gene symbol is in the first column, the Entrez gene ID is in the second column, the direction of change is in the third column, and information concerning published relation to diabetes mellitus is in the fourth column. The complete list of significantly altered transcripts is in Supplementary Table 1. TF, Transcription factor; ER, endoplasmic reticulum; MODY, maturity-onset diabetes of the young. 

Autoimmunity and abnormal glycemic control in TS

A total of 113 women were tested for antiislet cell and anti-GAD-65 antibodies. None tested positive for antiislet antibodies. Fourteen had positive anti-GAD titers (12.4%; 95% CI, 7.5–19.7). Of those with positive anti-GAD antibodies, one had type 1 DM, two had type 2 DM, four had IGT, and five were normal. The prevalence of anti-GAD antibodies was increased in the iXq group (21.4%; 95% CI, 10–39%) compared with the 45X group (5%; 95% CI, 1.8–14.4%; P = 0.03). The one patient with type 1 DM had a 46,X,i(X)(q) karyotype.

Discussion

This study has revealed a prevalence of type 2 DM of 25% among unselected (i.e. nonendocrine clinic or nonhospitalized) adults with TS. In a recent study, Cowie et al. (19) estimated the prevalence of diagnosed and undiagnosed type 2 DM combined for men and women and all ethnic and racial groups at 3.1% among 20 to 39 yr olds, at 12.4% among 40 to 59 yr olds, and 29.9% among 60 to 74 yr olds. Our study population had a median age of 34 (range, 18–67 yr), with an expected DM prevalence of less than 6.7%. Our finding of a 3- to 4-fold increase in type 2 DM in TS is consistent with the Danish registry report (5). In contrast to that report, we did not find an excess of type 1 DM (1 of 224, or <0.5%). We also did not find any cases of type 1 DM among the 125 girls with TS we have previously evaluated (20). It is possible that the lower prevalence of type 1 DM in our study subjects is a result of a selection bias in a way that individuals with type 1 diabetes are underrepresented because of lower referral rate. The authors of a recent study suggest an increase of the incidence of type 1 DM among 5220 girls with TS who were participants in the National Cooperative Growth study in the United States, compared with the general population; however, the standardized incidence ratio (SIR) from that study, similar to ours, did not reach statistical significance (SIR, 0.92–4.18) (21). Conversely, individuals already diagnosed with type 2 DM might be overrepresented in our population because of higher motivation to participate. However, such a bias, if indeed present, is unlikely to be of substantial magnitude because most of the participants were self-referred, community-dwelling, active individuals who had various motivations to participate in the TS study. The high prevalence of type 2 DM in TS requires a special awareness by the primary care providers (22).

It appears that decreased insulin secretory response to glucose is intrinsic to TS and is at the core of the high risk for DM. This abnormality is apparent early on in nonobese girls and young women with TS that have normal insulin sensitivity (6,11,23,24,25). Development of obesity and aging leads to insulin resistance, further decline in insulin response, and progression to IGT and overt DM (Table 3). Two small studies that compared overweight subjects with TS to normal-weight volunteers suggested that insulin resistance was a primary feature of TS (7,26), but studies that have compared BMI- or adiposity-matched groups have not found insulin resistance out of proportion to adiposity (6,11,27). Despite the limitations of using oGTT-derived parameters to evaluate insulin sensitivity and insulin secretion, our findings strongly support the concept of abnormal insulin secretory response as the main mechanism of Turner diabetic phenotype.

The TS diabetic phenotype seems to be similar to that seen in maturity-onset diabetes of the young syndromes that are caused by haploinsufficiency for genes involved in β-cell glucose sensing (glucokinase) or function (hepatic nuclear factors) (28). The age of onset is typically in the 30s or 40s, and the diabetes is usually mild. Only a few of the 56 diabetic patients in this study had a hemoglobin A1c greater than 7%.

An important facet of this study was our attempt to associate the metabolic defect(s) with a particular segment of the X-chromosome. We found a DM rate of approximately18% in 45,X, approximately 23% in delXp, but only 9% (similar to the general population rate) in the delXq group (Fig. 1). This finding is consistent with the hypothesis than haploinsufficiency for Xp gene(s) causes an increased risk for diabetes. The major X-chromosome pseudoautosomal region (PAR1) is located on the distal Xp. This region is identical on X- and Y-chromosomes and does not undergo X-inactivation. Haploinsufficiency for genes located in this region is thought to cause the TS phenotype. For example, haploinsufficiency for SHOX gene causes short stature in TS (29). PAR1 genes encode several types of receptors, phospholipases, protein phosphatases, GTP binding proteins, ATP transporter, and transcription factors. Therefore, haploinsufficiency for transcription factor genes could be involved in the abnormal insulin secretory response characteristic of TS.

We found a higher prevalence of DM (∼43%) among subjects with isochromosome Xq. Our data are corroborated by a recent U.K. epidemiological study of mortality in TS noting that the mortality from DM was particularly increased in those with an isochromosome Xq (10). Our data suggest that in TS, haploinsufficiency for unknown Xp gene(s) constitutes a “first hit” in susceptibility for DM, and that trisomy for Xq provides a “second hit,” greatly increasing the likelihood of developing DM. Supernumerary copies of Xq may increase risk for diabetes even in the absence of X-monosomy, e.g. among men with Klinefelter syndrome (47,XXY) and 48,XXYY (2,30,31).

The inactivation of the second and any additional X-chromosome is supposed to prevent overdosage of X-linked genes. However, recent studies in human cell lines have suggested that not all genes on the inactivated X are completely silenced (32). The extent to which X-linked genes escape from inactivation in humans is not yet known. To investigate the possibility of Xq “escaping” genes occurring in women with an isochromosome-Xq and potentially promoting the diabetic phenotype, we compared gene expression profiles in groups of 45,X vs. 46,Xi(Xq) subjects. These data support the notion that supernumerary Xq may trigger systemic changes via the expression of Xq genes that escape inactivation. Several Xq encoded transcription factors were overexpressed and may provide a link to the large number of autosomal genes that demonstrated altered expression in the two groups, including some involved in diabetes (Table 4). We found a highly significant overrepresentation of references to our set of altered gene expression in papers involving pancreatic islets and β-cell function. Alterations in transcript levels measured in lymphocytes may reflect system-wide effects as suggested by our finding that increased transcripts of IGF2, CRP, and GAD were paralleled by increased IGF2, CRP levels in the circulation as well as higher prevalence of anti-GAD antibodies in the iXq group. These observations suggest a proinflammatory state in individuals with an isochromosome Xq.

The clinical phenotype of 45,X and 46,X,i(X)(q) groups is very similar: both groups have short stature, premature ovarian failure, lymphedema, and congenital renal and cardiac defects to a similar extent. The only feature previously noted as exceptional in patients with isochromosome Xq was a higher prevalence of autoimmune thyroid disease (33). Thus, a hypothesis worth considering is the possibility that the excess DM found in the 46,X,i(X)(q) group could be due to a β-cell autoimmune component. Genes encoding two diabetes-inciting antigens were overexpressed in the iXq group. Although anti-GAD antibodies were detected in only a few, they were more common in the iXq group. Thus, there may be a “smoldering” β-cell autoimmunity, similar to latent autoimmune diabetes in adults (34), exacerbating the diabetic diathesis in women with i(X)(q).

In summary, this study demonstrates a high risk for type 2 DM among women with TS, especially those with an isochromosome Xq. We propose that haploinsufficiency for Xp gene(s) causes the basic deficit in β-cell function seen in 45,X patients and that excess dosage of Xq genes exacerbates the deficit, perhaps by altering other genes involved in β-cell development and function or survival, and/or by stimulating low-grade chronic autoimmunity that injures but does not obliterate the β-cells. Our findings illustrate the complexity of tracing genotype-phenotype associations in this disorder and suggest that clinicians need to be aware of the high diabetes risk.