Genotype-Structure-Phenotype Correlations of Disease-Associated IGF1R Variants and Similarities to Those of INSR Variants

Insulin-like growth factor 1 receptor (IGF1R) mediates the mitogenic and metabolic actions of IGF-1. IGF1R is a homolog of the insulin receptor (INSR); both receptors have the same domain structure, consisting of an α2/β2 heterotetramer (1). IGF1R is synthesized as a proreceptor in the endoplasmic reticulum and transported to the Golgi complex, where a protease cleaves it into α- and β- subunits, as with INSR. The mature receptor is then expressed as an α-β/α-β tetramer on the cell surface. The recently reported crystal structures of the IGF1R ectodomain (2) showed a receptor consisting of a first leucine-rich repeat (L1) domain, a cysteine-rich region (CR), and a second leucine-rich repeat (L2) domain, followed by three consecutive fibronectin type III domains (FnIII-1, FnIII-2, and FnIII-3), which includes an insert domain containing both the carboxy-terminal region of the α-chain (αCT), important for binding of IGF-1, and the α-β cleavage site (2).

While INSR variants can lead to severe insulin resistance (3–5), IGF1R variants can cause intrauterine and postnatal growth retardation. In addition, these variants may be found responsible for mental retardation, likely reflecting the role of the GH/IGF-1 axis in cognitive function (6). Since the year 2003, when Abuzzahab et al. (7) reported disease-associated IGF1R variants in patients with growth retardation, to date, >40 IGF1R variants have been identified in patients with growth retardation (8–22), most of whom were heterozygous carriers of these variants. We reported the first case of intrauterine growth restriction and short stature in a Japanese patient heterozygous for a missense variant (p.R739Q) of IGF1R (23).

Little has been reported on the genotype-phenotype relationship in IGF1R variants. However, of the heterozygous missense variants of IGF1R, those in the α-subunit often affect stature to a lesser extent than those in the β-subunit, which may be due in part to the dominant-negative effect, as transphosphorylation in the β-subunit is required to achieve an active conformation. On the other hand, it was recently reported that carriers of compound heterozygous variants (located in functionally important domains, such as the L1 or FnIII domains of IGF1R) exhibited more severe phenotypes, characterized by symptoms such as speech and developmental delays and mental retardation, compared with carriers of heterozygous variants (21). We previously performed in silico structural analyses of 12 missense variants located in the FnIII domains of INSR (each located in the FnIII-2 or -3 domains) to clarify their relationship with the phenotypic severity of insulin resistance (24). These analyses revealed that the missense variants of INSR, predicted to cause a severe impairment of the hydrophobic core formation and stability of the FnIII domains, did lead to the most severe phenotype of insulin resistance (25), while those predicted to cause local destabilization, but not to affect the folding of the FnIII domains, led to a less severe phenotype (26), suggesting genotype-structure-phenotype correlations in these variants (24).

In the current study, we hypothesized that there exists a close concordance between the structural modification of the FnIII domains in IGF1R variants and the phenotype of patients carrying these variants, given the structural similarity between IGF1R and INSR. Using previously reported missense variants of these domains in IGF1R, we performed a functional study of Chinese hamster ovary (CHO) cells to assess the impact of these variants on IGF1R expression and activity. In addition, we conducted an in silico structural analysis of these variants using a recently reported structural model of the ectodomain of IGF1R (2,27) to provide a structural explanation for the severity of growth retardation in patients with missense variants in IGF1R.

This was a retrospective study carried out using the clinical and genetic findings obtained from a case series of patients carrying variants of IGF1R registered in the Human Gene Mutation Database (HGMD) and the National Center for Biotechnology Information (NCBI) ClinVar database. The numbering of the amino acid residues in IGF1R was consistent with that used in the UniProt database (accession no. P08069), as previously reported (22). We selected heterozygous missense variants causing growth retardation previously identified in the FnIII-2 or -3 domains of IGF1R (p.R739Q [23], p.V629E [17], p.Y865C [11], and p.R739W and p.K720E [22]), using the above-mentioned databases. For this analysis, we included heterozygous missense variants but excluded homozygous or compound heterozygous variants causing growth retardation. In addition, we focused on missense variants of the FnIII domains in IGF1R, not rare in the general population (allele frequency >1% in the 1000 Genomes Project), and found that the variant p.N857S (rs45611935) met these criteria.

This study was approved by the ethics committee of the University of Tokyo (approval nos. G10077 and G3414) and was performed according to the approved guidelines.

DDK-tagged-pCMV-human IGF1R cDNA (OriGene Technologies, Rockville, MD) was used. We constructed mutant IGF1R expression vectors with the variant we previously reported [c.2216G>A (p.R739Q)] and those located in the FnIII domains except for the insert domain [c.1886T>A (p.V629E), c.2594A>G (p.Y865C), and c.2570A>G (p.N857S)], using the In-Fusion HD Cloning Kit (Clontech Laboratories, Mountain View, CA) following the manufacturer’s instructions. All variants were confirmed to be present 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 atmosphere containing 5% CO₂ and 95% air. Wild-type (WT) or mutant constructs with FnIII variants were transfected with Lipofectamine 3000 (Invitrogen, Thermo Fisher Scientific, Carlsbad, CA). For cotransfection, WT IGF1R/MOCK, WT/V629E, WT/K720E, WT/R739Q, WT/R739W, and WT/Y865C IGF1R plasmids were added in equal ratio as previously described (18). After 24 h, the cells were washed with PBS and cultured overnight and were stimulated with IGF-1 (0 or 10 nmol/L) (Wako Pure Chemical Industries, Ltd.) for 15 min at 37°C prior to the phosphorylation assay; they were rinsed with ice-cold PBS, and proteins were purified with the M-PER Mammalian Protein Extraction Reagent (Thermo Fisher Scientific, Rockford, IL).

Gel electrophoresis was performed with NuPAGE Novex 3–8% Tris-Acetate protein gels (Life Technologies, Carlsbad, CA), and proteins were transferred to a polyvinylidene fluoride membrane (Thermo Fisher Scientific). Blotted membranes were blocked with 5 mL of 1× iBind solution (Thermo Fisher Scientific) for 5 min at room temperature, probed with the iBind device (Thermo Fisher Scientific), and detected with the ECL Prime Western Blotting detection reagent. Antibodies specific for the β-subunit of human IGF1R (cat. no. 3018) and phosphorylated IGF1R (Tyr1135/1136) (cat. no. 3024) were purchased from Cell Signaling Technology Japan. An anti-rabbit IgG, horseradish peroxidase–linked antibody (cat. no. 7074) was also obtained as a second antibody from Cell Signaling Technology Japan. Images were captured with an iBright 1500 (Thermo Fisher Scientific).

A functional annotation of the FnIII domain variants of IGF1R was carried out using online prediction tools, such as SIFT, PolyPhen-2, and CADD via wANNOVAR (https://wannovar.wglab.org/) and in reference to the allele frequencies shown in the 1000 Genomes Project (28). Moreover, FoldX algorithm of the SNPeffect tool (29) was used to quantitatively estimate the impact of these variants on the stability of the proteins at the molecular level, based on ddG values as a measure of the free energy change from the WT to the mutant protein (30). The SNPs3D software was also used, in which the Support Vector Machine and data from structure- or alignment-based parameters were used to assess the impact of the variants (31).

Coordinate data for the X-ray crystal structure of the human IGF1R ectodomain in apo form (Protein Data Bank [PDB] identifier 5U8R) (2) and the cryo-electron microscopy (cryo-EM) structure of the head region of the IGF1R in complex with IGF-2 (PDB identifier 6VWI) (27) were obtained from the PDB (https://www.rcsb.org). The atomic coordinates for Pro735-Thr773 (including the proteolytic processing site involving α-β cleavages) in the disordered insert domain of 5U8R were modeled on model S1 of human INSR (32) with SWISS-MODEL (https://swissmodel.expasy.org/) (33). Structural models of IGF1R mutants (V629E, K720E, R739Q, R739W, N857S, and Y865C) were built with Swiss-PdbViewer (34). The mutant models were built on the basis of 5U8R, except that for K720E, which was built based on 6VWI. Comparisons were made between the structural models of WT IGF1R and each mutant using Waals (Altif Laboratories, Inc., Tokyo, Japan). Calculations were made for surface structure construction with eF-surf (https://pdbj.org/eF-surf/) and the resulting data visualized with Waals. The residues forming the folding nucleus of FnIII-2 and FnIII-3 were detected through comparison with the third FnIII domain of human tenascin (TNfn3) (PDB identifier 1TEN) (35) as previously described (24,36).

The data sets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

To date, several patients have been described as harboring heterozygous disease-associated variants (p.V629E, p.K720E, p.R739Q, p.R739W, and p.Y865C) located in FnIII-2 or -3 of IGF1R in the HGMD and NCBI ClinVar database. The clinical profiles of the patients with the above-mentioned disease-associated IGF1R variants are shown in Table 1.

Each patient showed growth retardation. In addition, most of the patients showed microcephaly, developmental delay, and elevated IGF-1 levels (Table 1). Included among these was the case of a heterozygous missense variant (p.R739Q) of IGF1R reported in our patient when she was 6 years old (23). This patient showed intrauterine and postnatal growth retardation due to this variant inherited from the proband’s affected mother, and her growth curve is shown in Supplementary Fig. 1. While, given the proband’s height at prepubertal age (approximately −2 SD), she had not received GH treatment, her height was measured as 143.4 cm (−2.6 SD) at the age of 14 years and remained profoundly less than −2.0 SD thereafter (Supplementary Fig. 1).

While the disease-associated variants of FnIII in IGF1R extracted in this study were not found in the 1000 Genomes database (Table 2), the SIFT, PolyPhen-2, and SNPs3D scores suggested these variants to be damaging. In addition, they showed a CADD score of ∼30 or more (Table 2). For analyses using the FoldX algorithm of the SNPeffect tool, the coordinate data were entered for the X-ray crystal structure of the human IGF1R ectodomain in apo form (PDB identifier 5U8R), which lacked the atomic coordinates for part of the insert domain. Thus, we focused attention on the variants of the FnIII domains, except for the insert domain, which provided positive ddG for disease-associated variants (p.V629E and p.Y865C), indicating that these variants destabilize structure.

The IGF1R variant p.N857S (rs45611935) was not shown to be rare among East Asians in the 1000 Genomes Project, with a minor allele frequency of 2.1%. The IGF1R locus was associated with human height in a recent large-scale genome-wide association study in a Japanese population (37); however, the variant rs45611935 showed no nominal association with human height or was not in linkage disequilibrium with the intronic lead variant (rs62024476) in the IGF1R locus (EAS: r² < 0.01). The variant rs45611935 was marked “benign (non-pathogenic)” in NCBI ClinVar, and the SIFT and PolyPhen-2 scores suggested that this variant was not damaging, as reported by the previous study (38). In addition, the variant had a CADD score of 11.7, much lower than those in disease-associated variants in the same domain (Table 2). The analysis involving the FoldX algorithm of the SNPeffect tool and SNPs3D score also suggested the influence of this variant to be insignificant.

For assessment of the impact of the variants in the FnIII domains of IGF1R, CHO cells were transfected with WT or mutant forms of IGF1R and cell lysates were analyzed by Western blotting using anti-IGF1R antibodies (Fig. 1A and B). It was found that the cells harboring variants causative of growth retardation (p.V629E [17], p.R739Q [23], p.R739W [22], and p.Y865C [11]), except for the variant p.K720E located in the αCT (22), had substantially lower expression of the mature IGF1R β-subunit than those with the WT receptor or the p.N857S variant (the likely benign variant) (Fig. 1A and C). In addition, cells expressing the disease-associated variants of IGF1R, including p.K720E (22), showed significantly decreased IGF-1–induced autophosphorylation of IGF1R compared with that in cells expressing the WT form or the variant p.N857S (Fig. 1B and D). Note that phosphorylation of the proreceptors with or without ligand stimulation was observed (Fig. 1B), just as in previous studies (13,39). Cotransfection of WT and disease-associated IGF1R variants that impaired the receptor processing, including those located outside the processing site (V629E and Y865C), substantially reduced autophosphorylation stimulated by 10 nmol/L IGF-1 (Supplementary Fig. 2); e.g., cotransfection of WT/V629E and WT/Y865C reduced autophosphorylation to 27.5 ± 7.2% and 8.9 ± 3.9% of the WT, respectively (Supplementary Fig. 2).

First, we detected the residues constituting the protein folding nucleus and the hydrophobic core of FnIII-2 and FnIII-3 in IGF1R by comparison with TNfn3 (PDB identifier 1TEN) (35), as described for our previous study of INSR (24). The FnIII-2 and FnIII-3 of IGF1R are shown to have the same topology as the other proteins in the FnIII family, i.e., seven β-strands composing two β-sheets, with the first consisting of β-strands A, B, and E and the second consisting of β-strands C′, C, F, and G. β-Strands B, C, E, and F form the common hydrophobic core of the FnIII domains (Supplementary Fig. 3). The folding nucleus of FnIII (in layer 3 of the four strands B, C, E, and F) is essential for formation of its topology, while the packing of the hydrophobic residues in layers 2 and 4 of these strands onto the folding nucleus contributes to significantly greater stability of the transition state for folding (36,40).

The structure of FnIII-2 in IGF1R was superimposed on that of TNfn3 with a root-mean-square deviation (RMSD) of 1.37Å over the structurally equivalent positions (76 residues). The TNfn3 residues I20, Y36, I59, and V70, which constitute the folding nucleus, were shown to overlap with V629, W648, I790, and I801 of FnIII-2, with an RMSD of only 0.47Å (Supplementary Fig. 3A). These latter four residues of FnIII-2 were shown to form the folding nucleus, with the residues W648 and I801 in one β sheet and the residues I790 and V629 in the other sheet forming the folding nucleus of FnIII-2 through hydrophobic interactions (Supplementary Fig. 4). Likewise, FnIII-3 was superimposed on the FnIII of TNfn3 with an RMSD of 1.44Å over the structurally equivalent positions (69 residues). The residues L850, Y869, L893, and A903 in FnIII-3 were detected as those constituting the protein folding nucleus, with an RMSD of only 0.36Å (Supplementary Fig. 3B). In FnIII-2 and FnIII-3 of IGF1R, the hydrophobic core consisting of the residues in layers 2–4 of strands B, C, E, and F was also shown to be crucial for the stabilization of the FnIII domains (Supplementary Fig. 3).

The structure of the dimeric ectodomain of IGF1R is shown in Fig. 2A. Of the six missense variants evaluated, V629E, Y865C, and N857S were shown to be located in the FnIII domains, except for the insert domain (Supplementary Fig. 5). Of the other variants present in the insert domain, K720E was shown to be located in the αCT, while R739Q and R739W were shown to be located in the proteolytic processing site (Fig. 2B).

V629 was identified as the residue constituting the folding nucleus of FnIII-2, which was crucial for FnIII folding (Supplementary Fig. 3A). WT V629 was shown to be in contact with I801 of strand F to form the folding nucleus, as well as with W631, T788, and I803 to form the hydrophobic core (Fig. 3A). The substitution of V629 with glutamic acid with a longer side chain led to a steric clash with I801 (Fig. 3B), thus resulting in a folding defect in FnIII-2. A comparison with our previous study of INSR showed that this variant was consistent with the definition of a group 1a variant responsible for the most severe insulin resistance (Table 3) and thus directly affected the folding nucleus (24).

Y865 was shown to be located on strand C of FnIII-3, with its side-chain aromatic ring shown to be in contact with the surrounding residues (I834, W852, I862, Y885, I905, and A907), while W852 and I905 were shown to form the hydrophobic core critical for the stabilization of the FnIII-3 domain (Fig. 3C and Supplementary Fig. 3B). Substitution of Y865 by cysteine with a smaller side chain resulted in a loss of contact with these residues (Fig. 3D). Thus, the Y865C mutant was predicted to result in a loss of contact with the surrounding residues, including the hydrophobic core packed onto the folding nucleus, thus significantly destabilizing the FnIII-3 domain in the folding process. This variant was consistent with a group 1b variant in the previous classification of INSR (Table 3) and thus affected the hydrophobic core of the FnIII domain (24).

N857 was shown to be located on the BC-loop (the loop between B and C strands) and distant from the folding nucleus of FnIII-3. With its side chain exposed, it did not interact with any residues in IGF1R (Fig. 3E). Substitution of N857S caused no structural changes in the residues neighboring N857 (Fig. 3F) and was predicted to have no significant impact on the structure of FnIII-3.

Proteolytic processing of a proreceptor leads to the generation of α and β subunits in IGF1R. R739 is situated within the α-β proteolytic processing site ⁷³⁷RKRR⁷⁴⁰ in the IGF1R proreceptor. The R739Q and R739W mutants occurred in the α-β cleavage site, thus changing the amino acid sequence within the cleavage site from R-K-R-R to R-K-Q-R and R-K-W-R, respectively (Fig. 4). In addition, substitution of R739 with a positively charged side chain by glutamine or tryptophan with a neutral side chain led to loss of the positive charge in the α-β cleavage site. The processing site of IGF1R is cleaved by the protease furin, which cleaves substrates containing the motif R-X-[K/R]-R, preferring positively charged residues at the cleavage site. Therefore, the R739Q and R739W mutants may not have been readily recognized by furin, resulting in defective processing. Moreover, because the side chain of tryptophan is rigid and bulkier than that for arginine (Fig. 4), introduction of the W739 variant may cause steric hindrance in the binding interface. Thus, the R739W mutant has greater potential to interrupt the interaction with furin. This could be one reason why the processing of R739W was more impaired than that of R739Q (Fig. 1A). Indeed, a similar R762S mutant has also been found within the processing site of INSR (41).

K720 is shown to be located in the αCT. Recently, single-particle cryo-EM analysis revealed an IGF1R dimer complex with IGF-1 or IGF-2 in the active state (27,42). For K720E, we built a mutant model using the cryo-EM structure of IGF1R and IGF-2 complex in the active state (PDB identifier 6VWI) to verify structural changes caused by K720E in its active conformation. In an IGF1R dimer in its active state, the IGF-1 binding site is shown to be formed by the L1 and CR domains of one IGF1R protomer and the αCT and FnIII-1 domains of the other (27,42). The side chain of K720 on the αCT was estimated to form a salt bridge with D519 in the FnIII-1 (27) (Supplementary Fig. 6). This salt bridge was assumed to be responsible for keeping the spatial αCT position and stabilizing the IGF-1 binding site in the active conformation.

Substitution of K720 by glutamic acid with a negatively charged side chain caused not only disruption of this salt bridge but also electrostatic repulsion with D519, resulting in interruption of the interaction between the αCT and FnIII-1 domain. Therefore, the K720E mutant in its active conformation was predicted to reduce the stability of the IGF-1 binding site, consequently affecting IGF-1 binding, which would result in reduced autophosphorylation, as observed in Fig. 1B. In contrast to the active form, no structural change was predicted around E720 as consequence of the K720E variant in an apo dimer of IGF1R (PDB identifier 5U8R), suggesting no influence on the structure of the K720E mutant in the inactive state, which would lead to normal processing of the variant receptor. These results are in agreement with the findings of the functional analysis (Fig. 1). Noteworthy, it was reported that the D734A mutant located on the αCT domain of INSR is normally processed and that the variant distorted the insulin binding site in vitro (43).

In the current study, we assembled disease-associated missense variants, as well as a variant classified as benign, located in the FnIII-2 and -3 domains of IGF1R to perform a functional study of CHO cells, in which we found impaired receptor processing or autophosphorylation in cells expressing the previously reported disease-associated variants of FnIII but not in those expressing the WT form or the variant classified as benign. For disease-associated variants (except for K720E located in the αCT), autophosphorylation signals seemed to correlate with the signals of the corresponding protein expression; however, extremely low expression of mature β-subunits makes it difficult to clearly assess this (Fig. 1). In addition, we conducted an in silico structural analysis of the missense variants in the FnIII domains of IGF1R, which revealed that the disease-associated variants were predicted to lead to severely impaired hydrophobic core formation and stability of the FnIII domains or affect α-β cleavage sites. The likely benign variant was predicted not to impair the folding of the FnIII domains or destabilize the domain structure, consistent with the results of the above-mentioned functional study of CHO cells. These results were also validated by our analyses using the FoldX algorithm (Table 2) conducted to calculate the free energy of WT and mutant proteins based on their three-dimensional structures and to predict the effect of variants on protein stability, while taking into account several factors, such as hydrogen bonding, van der Waals, and atomic clashes.

In our previous study, an in vitro functional experiment suggested that the R739 variant in the processing site could have a dominant-negative effect (23). In the current study, cotransfection of WT and variants outside the processing site, which impaired the α-β cleavage (p.V629E and p.Y865C), substantially reduced autophosphorylation stimulated by IGF-1 (Supplementary Fig. 2), suggesting that these variants could also have a dominant-negative effect. As previously reported (23), one possibility is that the unprocessed IGF1R binds IGF-1, even with low affinity, similar to the unprocessed INSR bound to insulin (44), which prevents IGF-1 from binding to the normally processed receptor. Although a theoretical possibility cannot be denied, that patients who appeared to be heterozygous for a nonsynonymous variant may have a second disease-associated variant outside the coding sequence, as reported in a previous analysis of INSR (45), substantially impaired receptor function in the presence of disease-associated variants (Fig. 1 and Supplementary Fig. 2) suggests that the heterozygous disease-associated variants here analyzed were responsible for the growth retardation in those patients.

FnIII domains play important roles in receptor functioning, such as ligand binding and processing, composed of a β-sandwich with a Greek key motif and containing a hydrophobic core constructed through the packing of two antiparallel β-sheets (46,47). A previous study presented the locations of amino acids in the variant forms of IGF1R in a crystallographic structure (21), although a complete structural explanation for the severity of disease in patients carrying these variants was not provided. In the current study, using the recently presented crystal structures of the ectodomain of IGF1R, we conducted a structural bioinformatics analysis of the missense variants in the FnIII domains of IGF1R and provided detailed structural explanations for the effect of these variants, suggesting that in silico structural analysis of IGF1R variants may prove useful for predicting the clinical severity of variant-associated disease based on structural modifications in the structure of the receptor. In addition, we conducted an in vitro functional study of these variants and suggested a pathogenic role for disease-associated variants (p.V629E, p.K720E, p.R739Q, p.R739W, and p.Y865C) reported in previous studies (11,17,22,23), but not for a likely benign variant (p.N857S) (Fig. 1), thus indicating the effectiveness of the in silico structural analysis performed in this study.

Relationships between the structural characteristics of IGF1R variants and their pathogenicity appeared more or less similar to the above-mentioned structure-phenotype correlations in the FnIII variants of INSR reported in our previous study (24), in which the FnIII variants of INSR were divided into groups according to their structural characteristics. Likewise, in the current study, we divided the FnIII variants of IGF1R into groups to allow comparisons with the FnIII variants of INSR (24) (Table 3). Intriguingly, each disease-associated FnIII variant of IGF1R was shown to have structural characteristics similar to those of the INSR variants in some groups (Table 3), suggesting a similarity in the molecular mechanism of structural modification between disease-associated FnIII variants of IGF1R and INSR.

Recently, a number of studies have been conducted to perform comprehensive genetic analyses in patients with short stature or growth retardation, using whole-exome sequencing (21,48,49) or next-generation sequencingbased gene panel analysis containing a number of genes including IGF1R (22). While such comprehensive genetic studies may help further elucidate the genetic architecture of growth retardation and have considerable potential to establish more definitive genotype-phenotype correlations in disease-associated IGF1R variants, it remains no less important to evaluate newly identified variants for their pathogenicity. The current study suggested that structural bioinformatics analysis of candidate variants of IGF1R may have a role to play as a primary screening method before performance of functional biological experiments that remain resource intensive and time intensive, as suggested in the previous structural analysis of variants in other genes (50,51). Indeed, these analyses may help provide supporting evidence for the pathogenicity of relevant variants, according to the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG-AMP) guideline (52). Successful identification of disease-associated variants of IGF1R in patients with growth retardation is clinically important. For example, lower efficacy of GH treatment could be anticipated in advance in patients with growth retardation who carry these variants compared with those who do not.

In conclusion, through both in silico structural and in vitro functional analyses, we demonstrated that the disease-associated mutants are predicted to severely impair FnIII domain stability and hydrophobic core formation or affect the processing site, while the variant classified as benign is predicted not to affect FnIII domain folding or lead to protein folding defects or significant destabilization of the domain structure. Genotype-structure-phenotype correlations in the FnIII variants of IGF1R were shown to be consistent with those of INSR, highlighting the functional importance of FnIII in both IGF1R and INSR. Disease-variant correlations, such as those suggested in this study, should facilitate early diagnosis in patients with disease-associated IGF1R and INSR variants and provide valuable insights into the disease-causing mechanisms of these receptors.

Acknowledgments. The authors thank M. Fujisawa, E. Izumi, and I. Kaisaki (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 received a Grant-in-Aid for scientific research from the Ministry of Education, Culture, Sports, Science and Technology of Japan (grant 19K16534 to J.H.).

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

Author Contributions. J.H., F.Mi., H.K., and N.S. designed the study and wrote the manuscript. J.H., F.Mi., T.Kat., F.Ma., and S.-I.A. conducted the research and analyzed the data. Y.K.S. and T.Kat. contributed to preparation of the manuscript. All authors 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.

Prior Presentation. Parts of this study were presented in abstract form at the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22–26 June 2018.