Structural Basis and Genotype–Phenotype Correlations of INSR Mutations Causing Severe Insulin Resistance

Mutations of the insulin receptor (INSR) gene result in extreme insulin resistance and dysglycemia (1), leading to several syndromes with various abnormal phenotypes that depend on the severity of INSR dysfunction. Patients with Donohue syndrome (DS), formerly known as leprechaunism, have the most severe insulin resistance (2,3) and patients with type A insulin resistance syndrome (type A-IR) display somewhat less severe manifestations (4,5), whereas Rabson-Mendenhall syndrome (RMS) represents an intermediate condition (6,7). Patients with type A-IR can live beyond middle age and present with hypertrichosis, acanthosis nigricans, and female hyperandrogenism. Patients with RMS generally survive into childhood or early adulthood and their characteristic symptoms are hypertrichosis, dysplastic dentition, and coarse and dysmorphic facial features. Patients with DS seldom live beyond infancy. They have dysmorphic facial features (so-called elfin appearance) and little subcutaneous fat.

INSR is a gene consisting of 22 exons and 21 introns. The proreceptor undergoes glycosylation and dimerization, followed by translocation to the Golgi apparatus and then processing of the dimer to yield a heterotetramer composed of two α-subunits and two β-subunits (8). Although there are no clear genotype–phenotype correlations for INSR mutations causing severe insulin resistance, it has been suggested that homozygous or compound heterozygous mutations of the α-subunit cause more severe syndromes (DS and RMS), whereas heterozygous β-subunit mutations lead to milder insulin resistance (9,10). Longo et al. (11) reported that missense mutations causing the most severe manifestations affected the extracellular portion of INSR and markedly reduced binding of insulin.

Some researchers have performed structural analysis of mutations of various proteins other than INSR to predict clinical manifestations and establish structure–phenotype correlations (12–14), and a structural bioinformatics approach should be useful for predicting the diverse phenotypes caused by monogenic mutations. However, there is no clear evidence of structure–phenotype correlations in patients with severe insulin resistance due to INSR mutations. McKern et al. (15) presented data on the structure of the extracellular portion of INSR, reporting that the extracellular portion of the monomer consists of a leucine-rich repeat domain (L1), a cysteine-rich region (CR), a second leucine-rich repeat domain (L2), and three fibronectin type III (FnIII) domains (FnIII-1 to FnIII-3). Insulin binds to two sites on INSR, and the FnIII domains contain parts of the primary and secondary insulin-binding sites (15,16). FnIII-2 contains the insert domain within which there is the α-β cleavage site and the carboxy-terminal region of the α-chain (αCT) involved in the primary insulin-binding site.

In this study, we examined four unrelated families with severe insulin resistance, and we identified five novel mutations of INSR and a gross deletion that removed exon 2. To assess the impact of mutations causing DS on INSR expression, INSR activity, and downstream signaling, we conducted a functional study in Chinese hamster ovary (CHO) cells. Using mutation data from the National Center for Biotechnology Information ClinVar database, Human Gene Mutation Database (HGMD), and UniProt database, we analyzed the distribution of INSR missense mutations in patients with severe insulin resistance to investigate the relationship between the mutation location and the severity of insulin resistance. We also performed in silico structural analysis of pathogenic missense mutations, with the aim of establishing structure–phenotype correlations.

We studied four patients with suspected insulin receptor abnormalities who were referred to our hospital (Table 1). Two patients had RMS (RMS-1 and RMS-2), one patient had DS (DS-1), and one patient had type A-IR (TypeA-IR-1). Detailed clinical information is provided in the Supplementary Data. This research was approved by the ethics committee of The University of Tokyo (approval numbers G3414 and G10077) and was implemented according to the approved guidelines. Parents gave written informed consent for genetic testing of their children. Genomic DNA was extracted from peripheral blood samples.

The 22 exons of INSR and its intron–exon junctions were amplified by PCR using the 21 pairs of primers listed in Supplementary Table 1. Then the PCR products were purified and directly sequenced.

A 60-mer oligonucleotide-based 4 × 44 K comparative genomic hybridization (CGH) microarray (INSR array) was custom-designed using the Agilent SureDesign (Agilent Technologies, Santa Clara, CA) web-based application (https://earray.chem.agilent.com/suredesign/). The INSR array contained 40,335 probes covering the entire INSR gene. The median probe spacing was 193 bp and the array focused on the 14.1 Mb genomic region encompassing INSR in 19p13.2. Normal male human reference DNA provided by Agilent in the SureTag Complete DNA Labeling Kit was the control for CGH analysis. After digestion with AluI and RsaI, genomic DNA from the DS-1 patient and his parents was labeled with Cy5-dUTP, and normal male human reference DNA was labeled with Cy3-dUTP. Purification of labeled products, array hybridization, washing, and scanning were conducted according to the CGH Enzymatic Labeling kit protocol v.7.1 (Agilent Technologies). Data analysis was performed using Agilent CytoGenomics 3.0.1.1. Copy number aberration calls were based on a minimum regional absolute average log₂ ratio of 0.25 and minimum contiguous probe count of 3. For break point analysis, a pair of primers was used to amplify the segment across the break point junction (Supplementary Table 1). Amplified junction fragments were directly sequenced.

GFP tagged-pCMV-human INSR cDNA (Origene, Rockville, MD) was used. A mutant INSR expression vector (p.V657F) with the point mutation (NM_000208.2:c.1969G>T) was constructed by using the GeneArt Site-Directed Mutagenesis System (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. In the same way, mutant INSR expression vectors with the following mutations of the FnIII domains (except for the insert domain) were constructed: c.2504G>T (p.S835I), c.2525C>T (p.A842V), c.2453A>G (p.Y818C), c.1904C>T (p.S635L), c.2465T>C (p.L822P), c.2776C>T (p.R926W), c.2810C>T (p.T937M), c.2621C>T (p.P874L), c.1975T>C (p.W659R), c.2633A>G (p.N878S), and c.2774T>C (p.I925T). Presence of the mutations was verified by Sanger sequencing.

CHO cells were maintained at 37°C in Nutrient F-12 Mixture (HAM) medium (Invitrogen) supplemented with 10% FCS in a humidified incubator with 5% CO₂/95% air. Transfections of either wild-type (WT) constructs or mutant constructs with FnIII mutations was performed with Lipofectamine 3000 (Invitrogen). After 72 h, cells were starved of serum for 4 h and stimulated with insulin (0, 10, or 100 nmol/L) (Sigma-Aldrich, St. Louis, MO) for 5 min at 37°C before the phosphorylation assay. Cells were rinsed with ice-cold PBS and proteins were purified using M-PER Mammalian Protein Extraction Reagent (Thermo Fisher Scientific, Hudson, NH).

Protein samples were mixed with NuPAGE sample buffer with or without NuPAGE sample reducing agent (Invitrogen). The final concentration of DTT in samples with reducing agent was 50 mmol/L. Electrophoresis was performed using NuPAGE Novex 3–8% Tris-Acetate Protein Gels (Life Technologies, Carlsbad, CA), and proteins were transferred to a Hybond P PVDF membrane (GE Healthcare, Milwaukee, WI). After blocking with 5% skim milk in TBS-T, membranes were probed overnight at 4°C with primary antibodies diluted in TBS-T, followed by secondary antibodies for detection using ECL Prime Western Blotting Detection Reagent. An antibody specific for the β-subunit of human INSR was purchased from Santa Cruz Technologies (sc-711), and antibodies targeting Akt (9272), phospho-INSR (Tyr1150/1151) (3024), and phospho-Akt (Thr308) (9275) were from Cell Signaling Technology Japan. As the secondary antibody, goat anti-rabbit IgG-HRP (sc-2004) was obtained from Santa Cruz Technologies. Images were captured with an LAS-3000 (Fujifilm, Tokyo, Japan) and were quantified using ImageJ software (National Institutes of Health, Bethesda, MD).

We analyzed the distribution of INSR mutations in the four patients. We counted missense mutations within the FnIII domains or the other domains of INSR that were identified in our study or were registered in the HGMD, ClinVar, and UniProt databases. Each mutation was assigned as a cause of DS, RMS, or type A-IR based on source articles, and the diagnosis was checked against information from Online Mendelian Inheritance in Man (OMIM) (http://www.ncbi.nlm.nih.gov/omim). Next, we used the Fisher exact test to investigate whether mutations classified as causing DS or RMS were more frequent in the FnIII domains than mutations classified as causing the milder type A-IR, with reference to Guo et al. (17). We also similarly examined whether mutations causing type A-IR were more frequent in the tyrosine kinase (TK) domain than mutations causing DS or RMS.

The X-ray crystal structure of INSR was obtained from Protein Data Bank (PDB) entry 4ZXB (18). Croll et al. (18) presented 4ZXB, which only consists of the relatively well-ordered protein and glycan residues, but they also presented an extended model of human INSR (model S1) including the insert domain, which was subjected to energy minimization to obtain a physically reasonable configuration. Model S1 has also been used for in silico structural analysis, but it lacked the atomic coordinates for R759-S763 (including the α-β proteolytic processing site). To obtain these missing residues, loop modeling was performed with the SWISS-MODEL homology modeling server (http://swissmodel.expasy.org/) (19). Structural models of mutant INSR were built using Swiss-PdbViewer (20). Each amino acid residue was substituted and energy minimization was performed to avoid steric hindrance. Then each mutant model and WT structure was compared using Waals (Altif Laboratories, Inc., Tokyo, Japan). Calculations for surface structure construction and electrostatic potential mapping were performed by using eF-surf (http://ef-site.hgc.jp/eF-surf/), and the resulting data were visualized with Waals. To identify the folding nucleus of FnIII-2 and FnIII-3, nucleation positions were detected by comparison with the third FnIII domain of human tenascin (TNfn3), as described previously (21). TNfn3 was the first β-sandwich protein to be characterized in detail by Φ-value analysis and was deposited as PDB entry 1TEN (22).

A Fisher exact test was performed for comparisons and P < 0.05 was considered statistically significant. All analyses were done with R software.

Sanger sequencing of INSR in the patients and their families revealed the mutations shown in Fig. 1A. Numbering of the amino acid residues in INSR is based on the UniProtKB/Swiss-Prot file P06213, as previously reported (23). Patient RMS-1 was a compound heterozygote for two novel mutations, c.2504G>T (p.S835I) and c.2525C>T (p.A842V), both of which were within FnIII. Patient RMS-2 was a compound heterozygote for a novel mutation, c.2997T>G (p.Y999*), and c.766C>T (p.R256C), which was previously identified in a heterozygous patient with RMS (24). Tyr999 is located near Pro997, which was affected by an INSR missense mutation previously detected in RMS (11). Patient TypeA-IR-1 was a compound heterozygote for a novel mutation, c.1465A>G (p.N489D), and c.3160G>A (p.V1054M), which has not previously been identified in type A-IR, although the same mutation was previously reported in a patient with DS (compound heterozygous with p.Trp659Arg in the α-subunit) (25). With regard to p.N489D, another mutation, c.1466A>G (p.N489S), at the same amino acid position was previously described in a patient with type A-IR (26). The mutant alleles were confirmed to have been inherited from the parents of each patient. Patient DS-1 was heterozygous for a novel mutation, c.1969G>T (p.V657F), in FnIII inherited from his mother. The location of V657 is near W659, which was affected by a missense mutation previously detected in another patient with DS (25). No other candidate mutations of INSR were detected in the four patients.

To investigate the existence of a mutant allele not detected by Sanger sequencing, we performed CGH microarray analysis in the DS-1 patient and his parents. We found that the patient and his father were heterozygous for a deletion mutation of INSR that removed a sequence containing exon 2 (Fig. 1B), while there was no such deletion in his mother. To determine the break point, we created primers for the flanking sites and performed PCR, obtaining a fragment of about 800 bp in the patient and his father, but not in his mother (Fig. 1C). We determined the break point junction by Sanger sequencing (Fig. 1D), revealing a 24,792 bp deletion (Chr19:7,266,055–7,290,846). The junction sequences were a long interspersed element and a short interspersed element, with only two base pairs of microhomology at the break point (Supplementary Fig. 1).

To assess the impact of the p.V657F mutation in the FnIII domains of INSR, CHO cells were transfected with WT or mutant forms of INSR and cell lysates were analyzed by Western blotting under reducing and nonreducing conditions using anti-INSR antibodies. Under reducing conditions, there was an increase of the proreceptor and a decrease of the mature receptor in mutant cells (Fig. 2A). To evaluate whether this mutation affected INSR activity and downstream signaling, insulin-induced phosphorylation was assayed in vitro. In cells expressing p.V657F, insulin-induced autophosphorylation of INSR was significantly decreased compared with autophosphorylation in cells expressing the WT form (Fig. 2B). Posttranslational receptor processing involves multiple steps, including dimerization of the precursor form (proreceptor) and proteolytic cleavage of the dimeric form to yield the α2β2 tetramer. To investigate which step of posttranslational processing was impaired, Western blot analysis was conducted under nonreducing conditions, revealing a predominance of high molecular–weight oligomeric forms with both the WT and mutant receptors. These results showed that the mutant insulin proreceptor also underwent dimerization (Fig. 2C). Furthermore, phosphorylation of Akt protein downstream target of the signaling pathway was also reduced in cells expressing the INSR p.V657F mutant compared with cells expressing WT INSR, but unphosphorylated Akt levels were not different (Fig. 2D). We also investigated the other missense mutations in the FnIII domains (Supplementary Tables 2 and 3). It was found that mutations causing both DS and RMS showed a substantially lower amount of the mature insulin receptor (IR) β-subunit expression than in cells with the WT receptor, whereas amount of the mature IR β-subunit was higher with FnIII mutations causing RMS than with mutations causing DS (Supplementary Fig. 2).

Among our four patients with extreme insulin resistance, the patient with DS (DS-1) and one of the two patients with RMS (RMS-1) had FnIII domain mutations, but the other RMS patient (RMS-2) and the patient with type A-IR (TypeA-IR-1) did not. Structural analysis of the extracellular portion of INSR (15) has provided insight into its domain structure, which is shown in Fig. 3A. Because the α-β cleavage site and some residues of both the primary and secondary insulin-binding sites are located within the FnIII domains (Fig. 3A), we suspected that FnIII mutations might be associated with more severe phenotypes. Therefore, we analyzed the distribution of INSR mutations to identify the domains preferentially affected by mutations causing more severe insulin resistance. We only assessed missense mutations, as previously reported (27), because it is more difficult to determine the impact of mutations and localize important functional regions by analyzing nonsense mutations as well as rearrangements and insertions/deletions compared with missense mutations (17). We analyzed 82 INSR missense mutations that were detected in our study or registered as pathogenic in databases (Fig. 3B and Supplementary Table 4), and we found that the frequency of mutations affecting the FnIII domains was significantly higher in patients with DS (28.1%) than in patients with type A-IR (3.7%) (P = 0.016, Fisher exact test). The frequency of FnIII mutations was also higher in patients with RMS (17.4%) than in patients with type A-IR (3.7%), but the difference was not significant (P = 0.17). In addition, mutations of the TK domain showed a significantly higher frequency in patients with type A-IR (59.3%) than in patients with DS (12.5%) (P = 0.00025) or patients with RMS (26.1%) (P = 0.024).

Mutations causing DS were preferentially associated with the FnIII domains of INSR, although there were also some mutations causing RMS or type A-IR. Therefore, we performed structural analysis of missense mutations located in the FnIII domains to elucidate the relations with phenotypic severity.

The X-ray crystal structures derived from PDB entry 4ZXB and its extended model (model S1) were used for structural analysis. The structure of the dimeric extracellular portion of INSR (model S1) (18) is shown in Fig. 3C. Because the model lacked residues encoding the α-β cleavage site, we complemented it by using SWISS-MODEL (Fig. 3D). FnIII-2 has the same topology as other proteins in the FnIII family, i.e., seven β-strands composing two β-sheets. The first sheet consists of the A, B, and E β-strands, and the second sheet consists of the C′, C, F, and G β-strands. The B, C, E, and F β-strands form the common hydrophobic core of the FnIII domains (Fig. 4A). The folding nucleus of TNfn3 (in layer 3 of the B, C, E, and F strands), another protein belonging to the FnIII family, is essential for forming its topology, and the residues in layers 2 and 4 of the strands that pack onto the folding nucleus contribute significantly toward stabilizing the transition state for folding (21,28). We compared folding of FnIII-2 in INSR with that of TNfn3. When the structure of FnIII-2 in INSR was superimposed on the structure of TNfn3, a root-mean-square deviation (RMSD) of only 1.06 Å was observed over the structurally equivalent positions (72 residues). TNfn3 residues I20, Y36, I59, and V70, which form the folding nucleus, overlapped with L640, W659, I809, and I820 of FnIII-2, and the RMSD was only 0.40 Å (Fig. 4B and C). These four residues of FnIII-2 form the folding nucleus. That is, residues W659 and I820 in one β-sheet and residues I809 and L640 in the opposite sheet form the folding nucleus of FnIII-2 through hydrophobic interactions (Supplementary Fig. 3). In the same way, the structure of FnIII-3 was superimposed on that of TNfn3, and the RMSD was only 1.29 Å over the structurally equivalent positions (62 residues). The TNfn3 residues forming the folding nucleus overlapped with L869, Y888, L913, and V923 of FnIII-3, and the RMSD was only 0.43 Å. These four residues were considered to form the folding nucleus of FnIII-3 (Supplementary Figs. 3 and 4). In FnIII-2 and FnIII-3 of INSR, the hydrophobic core corresponding to the residues in layers 2–4 of the B, C, E, and F strands are also critical for stabilization of the FnIII domains.

WT V657 is structurally close to the folding nucleus of FnIII-2 and forms the hydrophobic core. This residue is in contact with L640, W659, I809, and I820, which form the folding nucleus. Using the Swiss-PdbViewer program, we substituted V657 with phenylalanine in INSR. Insertion of a bulky phenylalanine caused steric clashes with residues L640, I809, and I820 of the neighboring folding nucleus, which destabilized the folding process and led to defective protein folding (Fig. 4D–G). We found that S835I and A842V in FnIII-2 caused RMS, and S835I created steric clashes with two neighboring residues (P625 and A824), whereas A842V produced a steric clash with the neighboring residue S630 (Supplementary Fig. 5). However, these mutations are distant from the folding nucleus and structural changes may be small, resulting in only localized structural destabilization.

We analyzed 12 missense mutations located in FnIII, except for the insert domain, identified in our study or registered as pathogenic in databases (Fig. 4H and I), and we divided these mutations into three groups (Supplementary Table 2 and Supplementary Data). Group 1a mutations directly affect the folding nucleus that comprises critical residues for folding of the FnIII domains (Fig. 4G and Supplementary Fig. 6A–C), resulting in defective folding. Group 1b mutations affect the hydrophobic core residues packed onto the folding nucleus (Supplementary Fig. 6D–F) and also significantly destabilize the FnIII domains. Group 1 (1a and 1b) mutations cause DS. The group 2 mutation causes loss of the hydrophobic interaction contributing to stabilization of the domain structure (Supplementary Fig. 6G) and thus destabilizes the domain to some degree. It has been registered in HGMD as causing DS, and Grasso et al. (29) reported an intermediate phenotype between DS and RMS in the mutation data source article. Group 3 mutations are located away from the folding nucleus and lead to small structural changes (Supplementary Figs. 5 and 6H and I), only producing localized destabilization. These mutations cause RMS. The extent of the structural changes caused by mutations in groups 1–3 is consistent with the clinical phenotype. On the other hand, two other missense mutations of the insert domain contained in FnIII have been reported (D734A and R762S [30,31]), neither of which influence the domain structure, but they affect the functional region or processing site (Supplementary Fig. 6J and K). The influence of these mutations is determined by factors other than structural defects, leading to a range of phenotypes.

We studied four unrelated families with severe insulin resistance and identified five novel mutations and a deletion that removed exon 2 of INSR. The patient with DS and one of two RMS patients had FnIII mutations. Using online databases, we demonstrated that missense mutations causing DS were significantly more frequent in the FnIII domains than mutations causing type A-IR. This finding supports the importance of the FnIII domains, which contain the α-β cleavage site and part of the insulin-binding site of INSR. We also found that missense mutations causing type A-IR were significantly more frequent in the TK domain than those causing DS or RMS. Patients with DS or RMS (extreme conditions presenting in infancy) have biallelic INSR defects that almost always display recessive inheritance, whereas patients with type A-IR (less severe and usually diagnosed around puberty) may have a heterozygous mutation of the TK domain that causes insulin resistance by a dominant negative effect, unlike mutations of other INSR domains (32–34). Heterozygosity of the mutation should lead to the formation of some fully functional wt/wt receptors, which would result in a less severe phenotype.

INSR contains three tandem FnIII domains in its extracellular juxtamembrane region (35). Each FnIII domain is a β-sandwich protein with a Greek key motif and is formed by packing two antiparallel β-sheets to construct a hydrophobic core (36,37). TNfn3 also belongs to the FnIII family, and previous studies have shown that the folding mechanism is similar throughout this family. For structural analysis, we compared folding of FnIII-2 in INSR with folding of TNfn3 and identified the folding nucleus of FnIII-2. We substituted V657 in FnIII-2 of patient DS-1 with phenylalanine, which was predicted to cause steric clashes with the folding nucleus that destabilized the folding process and led to defective folding. Although S835I and A842V (identified in FnIII-2 of patient RMS-1 as causing RMS) also caused steric clashes with neighboring residues, they were located away from the folding nucleus and unlikely to significantly destabilize the domain structure.

Review of online databases showed that missense mutations of the FnIII domains predicted to result in protein folding defects or significant destabilization of the domain structure all caused DS (Supplementary Table 2). One mutation (D734A) causing DS was not predicted to destabilize the hydrophobic core of the FnIII domains, but it is located in αCT (part of the insulin-binding domain) and would cause distortion of the insulin-binding site (30). Mutations causing the less severe RMS probably did not have a large effect on FnIII folding or stability. These results indicate that prediction of the phenotypic expression of INSR mutations might be improved by adopting a structural bioinformatics approach in addition to biochemical data, particularly assessment of the reduction of insulin binding that Longo et al. (11) proposed as corresponding to clinical severity.

Transfection of CHO cells with V657F mutation of INSR led to impaired receptor processing and autophosphorylation, as well as reduced phosphorylation of Akt. Under nonreducing conditions, the high molecular–weight form of INSR (oligomeric form) was predominant when both WT and mutant receptors were analyzed. Therefore, the mutant proreceptor undergoes dimerization before excision of the subunit processing site, as does the native proreceptor (38). Thus, the mutation probably impairs proreceptor processing by disturbing the three-dimensional structure of FnIII-2 containing the α-β cleavage site. Several mutations outside the cleavage site that disturb α-β cleavage have been reported, e.g., p.H236R (39) and p.N42K (40) are not within the cleavage site but retard several posttranslational processing steps, including proteolytic α-β cleavage of INSR. In our patients, reduced expression of the mature receptor probably contributed to impaired intramolecular signal transduction, as reported previously (41,42). Furthermore, we also conducted a functional assessment of INSR proteins with the other missense mutations in the FnIII domains (except for the insert domain). Patient RMS-1 was compound heterozygous for p.S835I and p.A842V. Though the level of the mature IR β-subunit in cells expressing the p.S835I mutation was much lower than in cells expressing the WT receptor, proreceptor processing was less impaired in cells expressing p.A842V, resulting in a less severe phenotype in patient RMS-1 (Supplementary Fig. 2). The level of the mature IR β-subunit was substantially lower in cells expressing mutations causing DS and RMS than in cells expressing the WT receptor, whereas the mature IR β-subunit was not as low in cells expressing the other FnIII mutations causing RMS (p.S635L and p.N878S) as in cells expressing mutations causing DS (Supplementary Fig. 2).

It is desirable to consider the properties of mutations causing DS or RMS in the context of two mutant alleles. The severity of insulin resistance would be related to the functional impact of the two mutations. The patients identified with 12 missense mutations of the FnIII domains (except for the insert domain) showed either homozygous or compound heterozygous mutations (Supplementary Table 3). Of the FnIII mutations, those causing DS were either homozygous or compound heterozygous with protein truncating mutations (nonsense and frameshift mutations) or deletions, as well as missense mutations (p.A119V and p.A1055V) situated in functionally important sites. As for the missense mutations, it was reported that p.A119V markedly impaired insulin binding when the mutant receptor was expressed in vitro (11), whereas p.V1054M is located in the consensus sequence for ATP binding and a patient with severe insulin resistance was reported to be heterozygous for another mutation p.A1055V at the neighboring amino acid position (43). Taking into account results from functional assessment of INSR proteins with FnIII mutations in this study, we speculated that severe FnIII domains mutations that are compound heterozygous with other mutations causing severe functional impairment result in deleterious functional impact, leading to DS. However, severe mutations, such as FnIII domains mutations, that are compound heterozygous with other mutations causing less severe functional impairment result in less severe form of RMS. Detailed functional analysis (at the cellular and molecular levels) of mutations causing severe insulin resistance combined with accumulation of clinical data would further delineate the properties of each mutation.

In patient DS-1, we identified an in-frame deletion in the region covering exon 2 of INSR. EBV-transformed lymphocytes from a patient with type A-IR due to an in-frame deletion of exon 2 of INSR mRNA were reported to show impairment of insulin binding (44). The L1 domain of INSR coded by exon 2 interacts extensively with αCT, which in turn interacts directly with insulin (16). Deletion of exon 2 is thought to cause destabilization of αCT that binds to insulin in FnIII, leading to impaired insulin binding by INSR mutants.

The syndromes caused by INSR mutations (DS, RMS, and type A-IR) seem to represent a broad spectrum of disease due to considerable variation in the severity of receptor dysfunction, rather than each one being a distinct entity. As each of the syndromes caused by INSR mutations (DS, RMS, and type A-IR) is rare, diagnosis of the syndrome remains challenging. Ongoing efforts to apply genomics to health care on a larger scale should allow collaboration in identifying patients with severe insulin resistance and the causal mutations, leading to refinement of the diagnostic criteria for these syndromes.

In conclusion, we identified five novel mutations of INSR and a deletion that removed exon 2 in four patients with extreme insulin resistance. Missense mutations causing DS were significantly more frequently located in the FnIII domains than those causing the milder type A-IR. According to structural analysis, DS was caused by all of the missense mutations that were predicted to severely impair formation of the hydrophobic core and stability of the FnIII domains, whereas RMS was caused by all of the mutations predicted to produce localized destabilization and not affect folding of the FnIII domains. Thus, the genotype–phenotype and structure–phenotype correlations of INSR mutations identified in this study provide insights into the molecular mechanisms of severe insulin resistance, would assist with early diagnosis of these syndromes, and could lead to new treatment approaches.

Acknowledgments. The authors thank K. Ishinohachi (Department of Diabetes and Metabolic Diseases, Graduate School of Medicine, The University of Tokyo) for providing excellent technical support during this study.

Funding. This study was supported by a grant-in-aid for scientific research in priority areas (C) from the Ministry of Education, Culture, Sports, Science and Technology of Japan (MEXT) (grant 15K09409 to N.S.). This work was also partly supported by a grant-in-aid for scientific research from MEXT (grant 16K07211 to F.M.) and by grants from Core Research for Evolutional Science and Technology, Japan Science and Technology Agency (to F.M. and T.T.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.H., H.K., F.M., N.S., and T.Kad. designed the study and wrote the manuscript. J.H., F.M., K.H., M.Tan., K.Su., M.Tak., and N.S. conducted the experimental research and analyzed the data. H.K., K.A., T.Kaw., I.M., K.Sa., T.I., K.A.B., T.T., and T.Y. contributed to data analysis and preparation of the manuscript. H.I., S.T., and all coauthors read the manuscript and contributed to the final version of the manuscript. T.Kad. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Data Availability. The INSR mutations identified in this study have been deposited in National Center for Biotechnology Information ClinVar with accession numbers SCV000503034, SCV000503035, SCV000503036, and SCV000503037.

Prior Presentation. Parts of this study were presented at the 77th Scientific Sessions of the American Diabetes Association, San Diego, CA, 9–13 June 2017.