Nucleotide substitutions in CD101, the human homolog of a diabetes susceptibility gene in non-obese diabetic mouse, in patients with type 1 diabetes.

AIMS/
INTRODUCTION: Although genome-wide association studies have identified more than 50 susceptibility genes for type 1 diabetes, low-frequency risk variants could remain unrecognized. The present study aimed to identify novel type 1 diabetes susceptibility genes by newly established methods.
MATERIALS AND METHODS: We carried out whole-exome sequencing and genome-wide copy-number analysis for a Japanese family consisting of two patients with type 1 diabetes and three unaffected relatives. Further mutation screening was carried out for 127 sporadic cases. The functional consequences of identified substitutions were evaluated by in silico analyses and fluorescence-activated cell sorting of blood samples.
RESULTS: Whole-exome sequencing and genome-wide copy-number analysis of familial cases showed co-segregation of the p.V863L substitution in CD101, the human homolog of an autoimmune diabetes gene in the non-obese diabetic mouse, with type 1 diabetes. Mutation screening of CD101 in 127 sporadic cases detected the p.V678L and p.T944R substitutions in two patients. The p.V863L, p.V678L and p.T944R substitutions were absent or extremely rare in the general population, and were assessed as 'probably/possibly damaging' by in silico analyses. CD101 expression on monocytes, granulocytes and myeloid dendritic cells of mutation-positive patients was weaker than that of control individuals.
CONCLUSIONS: These results raise the possibility that CD101 is a susceptibility gene for type 1 diabetes.

Introduction

Type 1 diabetes mellitus is a multifactorial disease in which pancreatic β‐cells are destroyed primarily by a T cell‐mediated autoimmune reaction1. Autoimmunity in type 1 diabetes is facilitated by various cells, including dendritic cells (DCs)1, 2, 3. A subset of patients with type 1 diabetes shows familial aggregation, suggesting a significant contribution of genetic factors to the etiology of the disease1. Type 1 diabetes represents a polygenic disorder, whereas other forms of diabetes mellitus, such as maturity‐onset diabetes of the young, neonatal diabetes and syndromic diabetes, often result from monogenic mutations or chromosomal abnormalities1, 2. Genome‐wide association studies (GWAS), together with candidate gene approaches, have identified more than 50 susceptibility genes for type 1 diabetes2. However, these genes account for just ~80% of the genetic heritability of type 1 diabetes, indicating that additional disease‐associated genes remain to be identified2, 4. In particular, susceptibility variants of low frequency might be undetected by GWAS, as this method focuses primarily on relatively common polymorphisms5.

CD101 (also known as V7 and IGSF2) is a transmembrane glycoprotein expressed on various types of immune cells6, 7, 8, 9. Cd101 has been reported as an ‘autoimmune diabetes gene in the non‐obese diabetic (NOD) mouse’10, 11, 12. Murine CD101 is known to modulate the function of regulatory T cells and antigen‐presenting cells, with genotype‐dependent Cd101 expression determining the risk of autoimmune diabetes in NOD mice11, 12. In addition, monoclonal antibodies against CD101 inhibit allogeneic T cell responses9. Human CD101 has also been implicated in immune regulation7, 8, 13, 14, 15, 16. Previous studies have suggested that human CD101 plays a costimulatory role in T cell activation mediated by T cell receptor/CD3 or skin DCs7, 8. However, CD101 mutations have not been associated with diabetes in humans. Here, we report the identification of three CD101 substitutions in patients with type 1 diabetes. These substitutions were detected through whole‐exome sequencing of familial cases and mutation screening of sporadic cases.

Materials and Methods

Whole‐exome sequencing and genome‐wide copy‐number analysis of a family with type 1 diabetes

The present study was approved by the institutional review board committee at the National Center for Child Health and Development, and was carried out after obtaining written informed consent. We carried out molecular analyses of a Japanese family (family A) consisting of two patients with type 1 diabetes and three unaffected relatives. The male proband (case 1) and his mother (case 2) developed diabetes at the ages 2.6 and 18 years, respectively (Table 1). At disease onset, case 1 was positive for the insulin autoantibody, whereas case 2 was positive for the islet cell surface antibody. Human leukocyte antigen (HLA) typing showed known risk alleles of the Japanese population17, DRB1*04:05 and DQB1*04:01, in cases 1 and 2 and two unaffected family members, and DQB1*03:02 in case 1 and his unaffected father (Table 1). The unaffected grandmother of case 1 (the mother of case 2) carried two risk alleles, DRB1*04:05 and DQB1*04:01, together with a protective DQB1*03:01 allele. Cases 1 and 2 showed no additional clinical features. No family history of other autoimmune diseases was recorded in this family.

Table 1

Clinical and molecular findings of CD101‐substitution‐positive patients and unaffected family members

Familial/Sporadic	Family A						Sporadic	Sporadic
Cases	Case 1 (proband)	Case 2 (mother)	Grandmother of case 1	Father of case 1	Older brother of case 1	Older sister of case 1	Case 3	Case 4
Sex	Male	Female	Female	Male	Male	Female	Male	Female
Clinical findings at the onset of the disease
Age at onset (years)	2.6	18	–	–	–	–	10.5	13.2
Body mass index (kg/m2)	14.8 (14.4–17.2)	22 (16.3–23.5)	–	–	–	–	17.1 (15.0–21.5)	16.6 (16.3–23.5)
Blood glucose (mmol/L)	18.9 (4.1–5.2)	8.6 (4.4–5.2)	–	–	–	–	51 (4.5–5.2)	13 (4.4–5.2)
HbA1c (mmol/mol)	95 (26–36)	53 (26–36)	–	–	–	–	89 (26–36)	92 (26–36)
HbA1c (%)	10.8 (4.5–5.4)	7 (4.5–5.4)	–	–	–	–	10.3 (4.5–5.4)	10.6 (4.5–5.4)
Blood C‐peptide (nmol/L)	0.2 (0.3–0.6)	NA	–	–	–	–	0.2 (0.4–0.8)	NA
Urine C‐peptide (μg/day)	<9 (12–34)	15 (29–167)	–	–	–	–	NA	90 (29–167)
Anti‐GAD antibody (U/mL)	Negative (<1.5)	NA	–	–	–	–	3.8 (<1.5)	81 (<1.5)
Insulin autoantibody (%)	20.8 (<10)	NA	–	–	–	–	NA	NA
Islet cell surface antibody	NA	Positive † (negative)	–	–	–	–	NA	NA
Molecular findings
HLA‐DRB1	*4:05	*4:05	*4:05	*11:01	*4:05	*4:05	*4:05	*9:01
	*11:01	*4:05	*12:02	*13:02	*13:02	*13:02	*9:01	*9:01
HLA‐DQB1	*3:02	*4:01	*3:01 ‡	*3:02	*4:01	*4:01	*3:03	*3:03
	*4:01	*4:01	*4:01	*6:09	*6:09	*6:09	*4:01	*3:03
CD101 substitution	p.V863L	p.V863L	p.V863L	None	None	None	p.V678L	p.T944R

Reference ranges are shown in parentheses. Values above or below the reference range are boldfaced. Glycated hemoglobin (HbA1c) is measured by a National Glycohemoglobin Standardization Program‐certified method (%), and converted to International Federation of Clinical Chemistry value (mmol/mol). Human leukocyte antigen (HLA) class II genotypes known as risk alleles of diabetes in Japanese are underlined. †Actual value unknown. ‡Known as a protective allele of diabetes in Japanese. NA, not analyzed.

We carried out whole‐exome sequencing using genomic DNA samples obtained from cases 1 and 2 and three unaffected family members (the father and two older siblings of the proband). DNA libraries were constructed using a SureSelect Kit (51 Mb version 4; Agilent Technologies, Santa Clara, CA, USA), and sequenced using a Hiseq 1000 sequencer (Illumina, San Diego, CA, USA). Nucleotide alterations were called by Avadis NGS 1.3.1 (DNA Chip Research, Yokohama, Japan) or SAMtools 0.1.17 software (https://sourceforge.net/projects/samtools/files/samtools/). We searched for nucleotide alterations shared by cases 1 and 2, but absent from the three unaffected relatives. We focused on exonic mutations that alter protein sequences and intronic substitutions located within a 5‐bp region from an exon–intron boundary. Known polymorphisms with an allele frequency of more than 1.0% in the general population (NCBI Browser, http://www.ncbi.nlm.nih.gov/), and mutations whose functional outcomes were predicted as ‘benign’ by in silico analysis using PolyPhen‐2 (http://genetics.bwh.harvard.edu/pph2/) were excluded from further analysis. We referred to the OMIM database (https://www.ncbi.nlm.nih.gov/omim) to examine whether the genes identified in the present study were associated with any human disorders. We also searched the PubMed database (http://www.ncbi.nlm.nih.gov/pubmed/) for previous reports on these genes. Nucleotide alterations in CD101 (NM_004258) and GAD2 (NM_000818.2) were confirmed by Sanger sequencing. Primer sequences are available on request. In addition, we analyzed the parental origin of a CD101 mutation identified in case 2. Copy‐number alterations in case 1 were analyzed by array‐based comparative genomic hybridization (SurePrint G3 Human Microarray, 2 × 400 k format; Agilent Technologies). We referred to the Database of Genomic Variants (http://projects.tcag.ca/variation/) to exclude known benign variants.

Mutation screening of CD101 in sporadic cases with type 1 diabetes

We carried out mutation screening of CD101 in 127 sporadic cases with type 1 diabetes (49 males and 78 females, aged 2.0–18.1 years). All patients were of Japanese origin, and developed diabetes between the age of 0.9 and 15.7 years. The patients were positive for anti‐GAD and/or islet antigen 2 antibodies. Patients with syndromic diabetes were excluded from the study.

Nucleotide alterations in CD101 coding exons and their flanking regions were examined by amplicon‐sequencing using Nextera Kits (FC‐121‐1031 and FC‐121‐1012; Agilent Technologies) and a Miseq next‐generation sequencer (Illumina). All CD101 nucleotide alterations, except for common polymorphisms found in the NCBI Browser, were confirmed by Sanger sequencing. To assess the presence or absence of the pathogenicity of CD101 substitutions, we attempted to obtain parental samples of mutation‐positive patients.

Functional assessment of CD101 substitutions

We examined whether the CD101 substitutions identified in the patients are present in the general population. First, we analyzed 185 DNA samples obtained from healthy Japanese controls (Human Science Research Resources Bank, Tokyo, Japan; present distributer, National Institute of Biomedical Innovation, Osaka, Japan). The presence or absence of the CD101 substitutions were examined by single‐nucleotide polymorphism genotyping (custom‐made TaqMan SNP Genotyping Assays; Life Technologies, Carlsbad, CA, USA) or by Sanger direct sequencing. Second, we examined frequencies of the CD101 substitutions in databases. We referred to the Human Genetic Variation Database (http://www.genome.med.kyoto-u.ac.jp/SnpDB/) and the Exome Aggregation Consortium (ExAC) Browser (http://exac.broadinstitute.org/). We also examined the position and evolutionary conservation of affected amino acids in the CD101 protein.

Expression analysis of CD101 in unaffected individuals and substitution‐positive patients

To analyze CD101 expression on hematopoietic cells, we carried out multicolor fluorescence‐activated cell sorting using LSRFortessa (BD Biosciences, San Jose, CA, USA). Fresh peripheral blood was obtained from five healthy individuals and three CD101 substitution‐positive patients (cases 1, 2 and 4). Fresh blood samples of case 3 were unavailable. Populations of lymphocytes, monocytes and granulocytes were differentiated using forward scatter and side scatter of flow cytometry. For analysis of CD101 expression on DCs, cells expressing human CD45 were gated on HLA‐DR+ and lineage−, and divided into two subpopulations in terms of expression of CD11c and CD123; that is, CD11c+/CD123− cells and CD11c−/CD123+ cells, which were referred to as myeloid and plasmacytoid DCs, respectively (Figure 1a,b). Antibodies used were FITC‐conjugated human CD45, APC‐Cy7‐conjugated HLA‐DR, APC‐conjugated CD123, PerCP‐Cy5.5‐conjugated CD11c, PE‐conjugated CD101, and brilliant violet 421‐conjugated CD3, CD19 and CD14 for lineage markers. The CD101 antibody was purchased from AbD Serotec (Kidlington, Oxford, UK), and other antibodies were purchased from BioLegend (San Diego, CA, USA).

Figure 1

CD101 expression on peripheral blood cells. (a) Major subpopulations of peripheral blood cells, such as lymphocytes (Lymp), monocytes (Mono) and neutrophils (Neutro), were differentiated based on forward scatter (FSC)/side scatter (SSC) of flow cytometry. (b) Cells expressing human CD45 were gated on human leukocyte antigen‐DR positive (HLA‐DR+) and lineage−, and divided into two subpopulations; that is, CD11c+/CD123− myeloid dendritic cells (mDCs) and CD11c−/CD123+ plasmacytoid DCs (pDCs). (c) Results of cases 1, 2 and 4 are shown in blue lines. Representative results of control individuals and the unstained control population of cells are shown with red and gray lines, respectively. Samples of case 3 were not available for this analysis.

Results

Molecular analyses of family A

Whole‐exome sequencing of family A showed that nine hitherto unreported nucleotide changes in nine genes and 15 rare polymorphisms in 14 genes co‐segregated with type 1 diabetes (Table 2; Figure 2a,b). Seven of the 24 nucleotide changes were intronic substitutions or an inframe insertion of unknown pathogenicity. Of the 23 mutated genes, CD101 represented the human homolog of a known diabetes‐causative gene in NOD mice11, 12, and GAD2 encoded a major autoantigen for type 1 diabetes3. The p.V863L substitution in CD101 was hitherto unreported, whereas the p.I228T substitution in GAD2 was found in 262 of 121,380 alleles in the ExAC Browser. The remaining 21 genes have not been associated with type 1 diabetes or other autoimmune disorders; eight of these genes were associated with some genetic disorders other than diabetes. The p.V863L substitution in CD101 was shared by the unaffected mother of case 2 (the grandmother of case 1). Mutations in known diabetes‐associated genes, including INS, PTPN22, IL2RA and CTLA4, were not found in this family. Genome‐wide copy‐number analysis in case 1 detected no deletions or duplications, except for common copy‐number variations.

Table 2

Nucleotide substitutions identified by whole‐exome sequencing of family A

Gene	Nucleotide change	Amino acid change	Allele frequency in the general population		In silico functional prediction‡	Human disease associated with the gene§
			HGVD	ExAC
Rare polymorphism
COL3A1 †	IVS25 + 5	–	0.004	74/121,400	Unknown	Ehlers–Danlos syndrome
NEDD4L	IVS14 + 5	–	0.002	13/103,616	Unknown	Unknown
GK	IVS17 + 4	–	0.008	0	Unknown	Glycerol kinase deficiency
SMARCAD1	IVS1‐2	–	0	1/119,520	Unknown	Adermatoglyphia
GAD2	T683C	I228T	0.005	262/121,380	Possibly damaging	Unknown
COL3A1 †	G3133A	A1045T	0.003	71/121,296	Probably damaging	Ehlers–Danlos syndrome
OR52D1	G604A	G202R	0.004	878/121,376	Probably damaging	Unknown
OBSCN	T8774G	L2925R	0.002	6/119,982	Probably damaging	Unknown
LYST	T9898C	Y3299C	0.002	3/21,498	Probably damaging	Chediak–Higashi syndrome
VCPIP1	G2513C	P837A	0.001	0	Probably damaging	Unknown
MYF5	A431G	N144S	0.001	0	Possibly damaging	Unknown
ZNF469	C7381T	R2461W	0.001	0	Probably damaging	Brittle cornea syndrome
FOXS1	G878C	P292A	0.007	15/118,480	Possibly damaging	Unknown
C1QTNF8	C140T	D46N	0.001	1/109,732	Possibly damaging	Unknown
PDE11A	G481C	S160X	0.002	0	No data	Pigmented nodular adrenocortical disease
CREB3L1	IVS11 + 1	–	0	0	Unknown	Unknown
GOLGA8S	IVS16‐5	–	0	0	Unknown	Unknown
TPRN	1650insCCG	550_551insA	0	0	Unknown	Hearing loss
CD101	G2587T	V863L	0	0	Possibly damaging	Unknown
PREX2	T2699C	I900T	0	0	Possibly damaging	Oncogene
LOH12CR1	G451C	E151Q	0	0	Probably damaging	Unknown
ZNF532	C449G	P150R	0	0	Probably damaging	Unknown
SDF2L1	C185A	S62Y	0	0	Probably damaging	Unknown
SHROOM2	C3890T	S1297F	0	0	Probably damaging	Ocular albinism

ExAC, the Exome Aggregation Consortium Browser; HGVD, Human Genetic Variation Database. †Two polymorphisms were identified in COL3A1. ‡Based on the data of PolyPhen‐2 (http://genetics.bwh.harvard.edu/pph2/). §Based on the data of OMIM (http://www.ncbi.nlm.nih.gov/omim).

Figure 2

Nucleotide substitutions of identified in the present study. (a) The structure of wild‐type DNA and CD101 protein, and the position of three substitutions. The black and white boxes on genomic deoxyribonucleic acid denote the coding regions and the untranslated regions, respectively. The p.V678L and p.V863L substitutions resided within the immunoglobulin‐like domains (blue boxes), while the p.T944R substitution was located in a connecting region between an immunoglobulin‐like domain and the transmembrane domain (orange box). The signal peptide (red box) was not affected by these substitutions. (b) Three substitutions in . These substitutions were identified by next‐generation sequencing and confirmed by Sanger sequencing (left panel). The black arrows indicate mutated nucleotides. Representative results of an in silico analysis (PolyPhen‐2) are shown (middle panel). Evolutionary conservation of each substitution is shown (right panel).

Mutation screening of CD101 of 127 sporadic patients identified 10 rare nucleotide substitutions (Table S1). These substitutions included p.V678L and p.T944R, which were absent or extremely rare in the general populations, and were assessed as ‘probably damaging’ by in silico analysis (Table S1; Figure 2a,b). Case 3 carrying the p.V678L substitutions and case 4 carrying the p.T944R substitutions developed diabetes at 10.5 and 13.2 years‐of‐age, respectively (Table 1). At disease onset, these cases were positive for anti‐GAD antibodies. Both cases 3 and 4 carried risk HLA alleles of the Japanese population. Blood samples from family members of cases 3 and 4 were unavailable for genetic testing.

The p.V863L, p.V678L and p.T944R substitutions were absent in 185 healthy Japanese individuals. Furthermore, these substitutions have not been registered in the Human Genetic Variation Database. Likewise, the p.V678L and p.V863L substitutions were absent from 121,412 alleles in the ExAC Browser, and the p.T944R substitution has been identified in one of 121,370 alleles.

The p.V678L and p.V863L substitutions resided within the immunoglobulin‐like domains of CD101, whereas the p.T944R substitution was located in a connecting region between an immunoglobulin‐like domain and the transmembrane domain (Figure 2a). The signal peptide was not affected by these substitutions. The three substitutions involved nucleotides that were evolutionary conserved among humans, rhesus, dogs and elephants (Figure 2b).

In keeping with previous reports7, 8, 9, 13, 14, CD101 was highly expressed on monocytes and granulocytes of control individuals, whereas lymphocytes including T cells, B cells and natural killer cells of these individuals barely expressed it (red lines in Figure 1c). Interestingly, the expression of CD101 on the monocytes and granulocytes of cases 1, 2 and 4 was weaker than that of the control individuals. The reduction of CD101 expression was also observed in myeloid DCs of these cases (Figure 1b and blue lines in Figure 1c). Expression of CD101 was weak or absent on plasmacytoid DCs of both patients and control individuals. Collectively, the CD101 mutations likely affected protein expression primarily on myeloid lineage cells.

Discussion

Whole‐exome sequencing of family A showed that a p.V863L substitution in CD101, the human homolog of an autoimmune diabetes gene in NOD mice10, 11, 12, co‐segregated with type 1 diabetes. Notably, type 1 diabetes was seen in cases 1 and 2, but not in the father or the older siblings of case 1, although HLA risk alleles were shared by all of these individuals. Whereas 23 other nucleotide substitutions were also linked to the disease phenotype of this family, there were no data supporting an association between these substitutions and diabetes. In fact, many of the 23 substitutions were of unknown pathogenicity, or resided within genes that underlie human disorders other than diabetes. Although cases 1 and 2 carried a p.I228T substitution (rs143186590) in GAD2 encoding a major autoantigen of type 1 diabetes, a relatively high frequency of GAD2 p.I228T in the general population argued against the pathogenicity of this mutation. Johnson et al.18 suggested that genetic variations in GAD2 are unlikely to underlie type 1 diabetes. Actually, GAD2 variations have been associated with obesity and anxiety disorders, rather than type 1 diabetes19, 20. Mutations in known diabetes‐causative genes, including INS, PTPN22, IL2RA and CTLA4 were excluded in family A. Furthermore, although several submicroscopic deletions in the genome have been associated with type 1 diabetes21, such defects were absent in case 1. Subsequently, we carried out CD101 mutation screening of 127 sporadic patients with type 1 diabetes and identified additional missense substitutions. The p.V863L, p.V678L and p.T944R substitutions in CD101 were absent or extremely rare in the general population, and were scored as possibly or probably damaging. Furthermore, the three substitutions were well‐conserved among most placental mammals. Collectively, the present findings, in conjunction with prior observations that CD101 plays a significant role in T cell regulation6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 22, 23, imply an association between CD101 mutations and type 1 diabetes. CD101 mutations might increase genetic predisposition to diabetes only in individuals at risk of the disease, because cases 1–4 invariably carried one or more of the HLA risk alleles. While the mother of case 2 was free from the disease despite having the same CD101 substitution as cases 1 and 2, this normal phenotype might reflect the presence of the protective DQB1*03:01 allele. Considering the small number of participants in the present study, our findings need to be validated in future studies.

Fluorescence‐activated cell sorting analysis detected CD101 expression on hematopoietic cells in both control individuals and mutation‐positive patients. The percentage of CD101‐positive cells among monocytes, neutrophils and myeloid DCs was somewhat lower in cases 1, 2 and 4 compared with that in five control individuals. The underlying mechanism of the reduced CD101 expression in these cases remains unknown. Actually, the signal peptide and the transmembrane domain are not affected by the p.V678L or p.T944R substitutions. Furthermore, it is unclear whether these minor differences in the expression levels are of clinical significance. We cannot exclude the possibility that CD101 mutations identified in the present study are functionally neutral variants. Nevertheless, altered CD101 expression has been associated with various autoimmune disorders. Jovanovic et al.16 showed that the fraction of CD101‐positive cells among the CD8+ T cell population was reduced in patients with rheumatoid arthritis. Ǻkesson et al.23 found that CD101 expression in regulatory T cells was moderately increased in children with celiac disease. In addition, reduced surface expression of murine CD101 was associated with the risk of infection‐induced liver autoimmunity22. Altered expression of CD101 on DCs could influence the progression of pancreatic insulitis, because DCs are known to play a critical role in the recruitment of lymphocytes in insulitis of the NOD mouse24. Further studies, such as viral vector‐mediated transduction of the mutant CD101 to immune cells, will clarify the functional consequences of the p.V678L, p.V863L and p.T944R substitutions.

Previous GWAS did not suggest any association between CD101 and diabetes. This can be explained by the rarity of CD101 substitutions in the general population. It is known that disease‐associated variants with an allele frequency of less than 0.5% in the general population are barely detectable by GWAS, unless the variants underlie monogenic Mendelian disorders5. As we identified CD101 substitutions in just two of the 127 sporadic patients, such substitutions appear to be rare, even in patient cohorts. The present results show that next‐generation sequencing of familial cases is useful for identifying rare risk variants that have been missed by GWAS.

In conclusion, we identified rare CD101 mutations in familial and sporadic cases of type 1 diabetes. The present findings, in conjunction with the results of previous studies6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 22, 23, 24, raise the possibility that CD101 is a susceptibility gene for type 1 diabetes.

Disclosure

The authors declare no conflict of interest.

Table S1| Frequency of single nucleotide polymorphisms in CD101 in our sporadic patients and the Japanese general population.

Click here for additional data file.