Crucial Role of the SH2B1 PH Domain for the Control of Energy Balance

Hyperphagia, severe obesity, insulin resistance, and neurobehavioral abnormalities have been reported in individuals with rare coding variants in the gene encoding SH2B1 (SH2-B, PSM) (1,2). Consistently, mice null for Sh2b1 exhibit obesity, impaired glucose homeostasis, and often, aggressive behavior (3–5). Transgenic expression of the β-isoform of SH2B1 (SH2B1β) in the brain largely corrects the obesity and glucose intolerance of otherwise Sh2b1-null mice (6), suggesting the importance of brain SH2B1 for the control of energy balance and glucose homeostasis.

At the cellular level, SH2B1 is an intracellular adaptor protein that is recruited to phosphorylated tyrosine residues on specific membrane receptor tyrosine kinases (e.g., receptors for brain-derived neurotrophic factor [BDNF], nerve growth factor [NGF], insulin) and cytokine receptor/Janus kinase (JAK) complexes (e.g., leptin receptor/JAK2) and enhances the function of these receptors (7–13). The exact mechanism(s) by which it does so is unclear, although a variety of mechanisms have been proposed. These include enhanced dimerization causing increased activation of the kinase itself (14), stabilization of the active state of the kinase (15), decreased dephosphorylation or increased complex formation of insulin receptor substrate (IRS) proteins bound to receptors or receptor/JAK2 (16,17), regulation of the actin cytoskeleton (18), and activation of specific pathways, including extracellular signal–regulated kinases (ERKs), Akt, and/or phospholipase Cγ (10,19). Some of these receptors, including the leptin, BDNF, and insulin receptors, play important roles in the central control of energy expenditure and/or glucose homeostasis (20). SH2B1β has been shown to enhance BDNF- and NGF-stimulated neurite outgrowth in PC12 cells (13,21).

The four isoforms of SH2B1 (α, β, γ, δ), which differ only in their COOH termini, share 631 NH₂-terminal amino acids. These amino acids possess a dimerization domain, Pleckstrin homology (PH) domain, src-homology 2 (SH2) domain, nuclear localization sequence (NLS), and nuclear export sequence (NES) (22–24) (Fig. 1A). The SH2 domain enables SH2B1 recruitment to specific phosphorylated tyrosine residues in activated tyrosine kinases (25). The NLS and NES are essential for SH2B1 to shuttle among the nucleus, the cytosol, and the plasma membrane (22,23). The NLS combined with the dimerization domain enables SH2B1 to associate with the plasma membrane (26). However, the function and importance of the SH2B1 PH domain remains largely unknown. Four human obesity-associated variants lie in the SH2B1 PH domain (Fig. 1A), suggesting the importance of the PH domain in SH2B1 function. The PH domains of some proteins bind inositol phospholipids to mediate membrane localization (27,28). However, 90–95% of all human PH domains do not bind strongly to phosphoinositides and presumably mediate other functions (29). Indeed, the PH domain of SH2B1 neither localizes to the plasma membrane nor is required to localize SH2B1β to the plasma membrane (23,30). Here, we tested the importance of the PH domain of SH2B1 in vivo by generating and studying mice containing human obesity-associated (P322S) or engineered (in-frame deletion of P317 and R318 [ΔPR]) mutations in the SH2B1 PH domain. Our results demonstrate that the SH2B1 PH domain plays multiple crucial roles in vivo, including for the control of energy balance and glucose homeostasis, and in in vitro studies, changes the subcellular distribution of SH2B1β and enhances NGF-stimulated neurite outgrowth in PC12 cells.

The Genetics of Obesity Study (GOOS) is a cohort of >7,000 individuals with severe obesity with age of onset of <10 years (31,32). Severe obesity is defined as a BMI (kg/m²) SD score >3 (U.K. reference population). Whole-exome sequencing and targeted resequencing were performed as in Hendricks et al. (33). All variants were confirmed by Sanger sequencing (1). HOMA of insulin resistance (HOMA-IR) was calculated using the equation HOMA-IR score = [(fasting insulin in μU/mL) × (fasting glucose in mg/dL)] / 405, which estimates steady-state β-cell function and insulin sensitivity (34,35). All human studies were approved by the Cambridge local research ethics committee. Each subject (or parent for those <16 years of age) provided written informed consent; minors provided oral consent.

Animal procedures were approved by the University of Michigan Committee on the Use and Care of Animals in accordance with Association for Assessment and Accreditation of Laboratory Animal Care and National Institutes of Health guidelines. Mice were bred at the University of Michigan and housed in ventilated cages at 23°C on a 12-h light (0600–1800 h)/12-h dark cycle with ad libitum access to food and tap water except as noted. Mice were fed standard chow (20% protein, 9% fat [PicoLab Mouse Diet 20 5058, #0007689]) or, as described in Fig. 2H–M and Supplementary Fig. 2F–K, a high-fat diet (HFD) (20% protein, 20% carbohydrate, 60% fat [D12492; Research Diets]).

CRISPR/Cas9 genome editing was used to insert the P322S mutation into mice. The reverse complement of the genomic Sh2b1 sequence in C57BL/6J mice (accession number NC_000073, GRC m38) was used to design the reagents for CRISPR. The guides were designed using the website described in Ran et al. (36). The mutations in the donor are summarized in Fig. 1C (details in the Supplementary Data). After testing, each guide/donor combination was injected into C57BL/6J oocytes by the University of Michigan Transgenic Animal Model Core. P322S and ΔPR founders were backcrossed to C57BL/6J mice. The mice were genotyped as described in Truett et al. (37) using primers listed in Supplementary Table 1. The P322S and ΔPR PCR products were digested with XbaI or purified and sequenced. Sh2b1 knockout (KO) mice were obtained from Dr. Liangyou Rui (University of Michigan) and genotyped according to Duan et al. (3). C57BL/6J mice used to invigorate our C57BL/6J colony came from The Jackson Laboratory. Relative levels of Sh2b1 gene expression were determined using RT-PCR (details in the Supplementary Data).

Mice were individually housed, and body weight and food consumption were assessed weekly.

Mice were fasted 0900–1300 h for glucose tolerance tests (GTTs) or 0800–1400 h for insulin tolerance tests (ITTs). Glucose or human insulin was injected intraperitoneally, and blood was collected from the tail vein. Blood glucose levels were assessed using a Bayer Contour glucometer. For plasma insulin levels, mice were fasted 0800–1400 h, and tail blood was assayed using an Ultra Sensitive Mouse Insulin ELISA kit (#90080; Crystal Chem). Tail blood (0900–1000 h) (Figs. 2F and 4G) or trunk blood after sacrifice (1000–1300 h) (Fig. 6B) from fed mice was tested for leptin using a Mouse Leptin ELISA kit (#90030; Crystal Chem).

Body composition was measured at room temperature in the morning or evening (Fig. 6A) using a Minispec LF90 II Bruker Optics nuclear magnetic resonance (NMR) analyzer (University of Michigan Animal Phenotyping Core). To assess metabolic state, mice were single-housed for 3 days and then tested for 72 h using a Comprehensive Lab Animal Monitoring System (Columbus Instruments). O₂ consumption (VO₂), CO₂ production (VCO₂), X activity, and Z activity were collected in 20-min bins. The final 24 h of recordings are presented. Mice were sacrificed (1000–1300 h) using decapitation under isoflurane. Trunk blood was collected and the serum stored at −80°C. Tissues were collected, weighed, cryopreserved in liquid nitrogen, and stored at −80°C.

Frozen tissues were lysed in L-RIPA lysis buffer (50 mM Tris, 150 mM NaCl, 2 mM EGTA, 0.1% Triton X-100, pH 7.2 containing 1 mM Na₃V0₄, 1 mM PMSF, 10 μg/mL aprotinin, 1 μg/mL leupeptin). Equal amounts of protein were immunoblotted with antibody to SH2B1 (αSH2B1) (sc-136065, RRID:AB_2301871; Santa Cruz Biotechnology) (1:1,000 dilution) or β-tubulin (sc-55529, RRID:AB_2210962; Santa Cruz Biotechnology) (1:1,000 dilution) as described in Joe et al. (19). For immunoprecipitations, tissue lysates containing equal amounts of protein were incubated with αSH2B1 (1:100) and immunoprecipitated and immunoblotted as in Joe et al. PC12 cells (ATCC) were cultured and treated as in Joe et al. Briefly, the cells were grown in PC12 medium A (RPMI medium, 5% FBS, 10% heparan sulfate) in 10-cm dishes coated with rat tail type I collagen (#354236; Corning). Cells were transfected and, 24 h later, incubated overnight in deprivation medium (RPMI medium, 2% heparan sulfate, 1% FBS) before being lysed and immunoblotted with αSH2B1.

The indicated construct was transiently transfected into 293T cells or PC12 cells. Cells were treated and live cell images captured by confocal microscopy using an Olympus FV500 laser scanning microscope and FluoView version 5.0 software, as in Joe et al. (19).

For Fig. 3D, PC12 cells were plated in six-well collagen-coated dishes, transiently transfected as indicated for 24 h, and incubated overnight in deprivation medium. Cells were treated, and neurite outgrowth was determined as in Joe et al. (19). For Fig. 1B, PC12 cells were treated as in Joe et al., with modifications described in the Supplementary Data.

A structural model for human SH2B1 was created by overlaying the PH domain sequence of human SH2B1 onto the mouse NMR structure of APS (Protein Data Bank ID 1V5M) using the PyMOL Molecular Graphics System version 2.2.3. ClustalW alignments were performed using LaserGene version 14.0.0 (DNASTAR, Madison, WI). Functional homology was defined as residues that match the consensus within 1 distance unit using the PAM250 mutation probability matrix.

All nonhuman analyses were carried out using GraphPad Prism software. Body weight, GTTs, and ITTs were analyzed by two-way ANOVA followed by Fisher least significant difference posttest. Food intake was analyzed by linear regression. Significance of the deviation of birth rate from the expected Mendelian ratio was assessed using χ² test. For other physiological parameters, experimental animals were compared with their wild-type (WT) littermates by two-tailed Student t test. Neurite outgrowth was analyzed by a two-tailed Student t test. For all comparisons, P < 0.05 was considered significant.

Any raw data sets generated during the current study are available from the corresponding author on reasonable request, with all reagent and analytical details included in the published article (and its Supplementary Data). The mouse models generated and analyzed during the current study are available from the corresponding author upon reasonable request.

Using exome sequencing, targeted resequencing, and Sanger sequencing of 3,000 individuals exhibiting severe obesity before the age of 10 years (33), we identified 15 rare variants in SH2B1 in 16 unrelated individuals (Table 1 and Fig. 1A). Eleven variants are newly identified, while four (R227C, R270W, E299G, V209I) have been previously reported in other obese individuals but not well characterized (4,38,39). Fourteen of these variants are in the first 631 amino acids shared by all isoforms of SH2B1. The mean ± SD BMI SD score of variant carriers was 4.0 ± 0.6. The 15th variant causes the G638R mutation in the COOH-terminal tail unique to the β-isoform of SH2B1. A number of the SH2B1 variant carriers had HOMA-IR (34,35) scores of >1.9, indicating insulin resistance and increased risk of type 2 diabetes (40). Some of the HOMA-IR scores, including those for people carrying three variants in or near the PH domain (G238C, R270Q, and M388V), were particularly high. A spectrum of neurobehavioral abnormalities, including learning difficulties, dyspraxia, hyperactivity/inattention, aggression/emotional lability, anxiety, and autistic traits, were detected in all the individuals for whom behavioral information was available (Table 1). In the neurite outgrowth assay, 7 of these 15 rare variants impaired the ability of SH2B1β to stimulate NGF-induced neurite outgrowth (Fig. 1B), suggesting that many of the variants negatively affect the neuronal function of SH2B1. Because the variants are found in multiple domains in SH2B1 and throughout the SH2B1 sequence, it is not surprising that individuals with different SH2B1 variants have different phenotypes. These newly characterized variants add support to SH2B1 being an important regulator of human body weight, insulin sensitivity, and behavior. Interestingly, four of the human obesity-associated SH2B1 variants lie in the PH domain of SH2B1, suggesting the importance of the PH domain in the ability of SH2B1 to regulate energy balance, glucose metabolism, and behavior.

To gain insight into the role of the PH domain of SH2B1 in energy balance and glucose metabolism, we studied the effect of the P322S human obesity-associated SH2B1 PH domain variant in mice. We chose the P322S variant because of its strong association with obesity in the proband family (1), the conservation of P322 across mammals and with the SH2B1 family member SH2B2/APS, its predicted disruptive effect on PH domain function by Provean (RRID:SCR_002182) and PolyPhen (RRID:SCR_013189) analysis, and P322S-dependent deficiencies in SH2B1 function observed in cultured cells (1). We used CRISPR/Cas9-based genome editing to introduce the P322S variant into Sh2b1 in C57BL/6J mice (Fig. 1C). DNA sequencing confirmed germline transmission of the P322S edit (Supplementary Fig. 1A). The P322S mutation neither affects the mRNA levels for any of the Sh2b1 isoforms in the examined tissues (brain, liver, and heart) (Supplementary Fig. 1B and C) nor alters SH2B1 protein levels or isoform selection in brain tissue (Fig. 1D and Supplementary Fig. 1D). However, we found that homozygous (P322S/P322S) mice are born at much less than the expected Mendelian ratio (Fig. 1E), suggesting that the P322S mutation disrupts SH2B1 PH domain function in a manner that interferes with embryo implantation and/or development. Consistent with this, preliminary data from timed pregnancies reveal that at embryonic day 17, the homozygous embryos (P322S/P322S) are also present at less than the expected Mendelian ratio.

In addition to the difficulty of producing sufficient P322S/P322S mice for study, the high rate of embryonic lethality in P322S/P322S mice suggested that the surviving P322S/P322S mice might have underlying poor health, which could interfere with the analysis of their metabolic phenotype. For these reasons, and because human obesity is linked with heterozygosity for P322S (1), we studied energy balance and glucose homeostasis in heterozygous (P322S/+) male (Fig. 2) and female (Supplementary Fig. 2) mice. We found no difference in food intake, body weight, or adiposity between WT and P322S/+ mice fed standard chow (9% fat) or an HFD (60% fat). However, in contrast to their WT littermates, 28-week-old HFD-fed P322S/+ male and female mice displayed glucose intolerance in an intraperitoneal GTT. Neither insulin concentrations nor the response to an ITT were altered in the P322S/+ animals compared with littermate controls, however. These findings suggest that the PH domain of SH2B1 is important for SH2B1 function, including for the control of glucose homeostasis, but that the resultant metabolic phenotype is less penetrant in the heterozygous state in mice than it is in humans. We thus sought to study mice homozygous for mutations in the SH2B1 PH domain.

Because of the early lethality of P322S/P322S mice, we examined the function of another SH2B1 mutation containing a two-amino acid deletion (ΔPR) within the PH domain of SH2B1 (Fig. 1C and Supplementary Fig. 1E). This mutation arose as a separate line during the generation of the P322S mice.

When transiently expressed as green fluorescent protein (GFP) fusion proteins in PC12 cells, SH2B1β and SH2B1β ΔPR demonstrated similar expression levels (Fig. 3A), suggesting that ΔPR does not destabilize the protein. However, while GFP-SH2B1β localizes primarily to the plasma membrane and cytoplasm in 293T and PC12 cells (as previously shown [22,26]), SH2B1β ΔPR localizes primarily to the nucleus (Fig. 3B and C). The nuclear localization of SH2B1β ΔPR suggests that the ΔPR mutation alters SH2B1 nuclear cycling to favor retention in the nucleus. We predicted that this altered localization would change the cellular function of SH2B1β ΔPR. Indeed, SH2B1β-dependent NGF-stimulated neurite outgrowth in PC12 cells was decreased in cells expressing SH2B1β ΔPR (Fig. 3D). Thus, disruption of the PH domain by the ΔPR mutation alters the subcellular distribution of SH2B1β and impairs the ability of SH2B1β to enhance neurotrophic factor–induced neurite outgrowth.

We examined the phenotype of the mice containing the ΔPR mutation with the hope that this mutation might produce a less dramatic reproductive phenotype than that observed with P322S in the homozygous state, allowing us to examine the effects of the ΔPR mutation on energy balance and glucose homeostasis in homozygous mice. As with the P322S mutation, ΔPR did not affect the mRNA levels for any of the Sh2b1 isoforms in the tissues tested (brain or heart) (Fig. 4A). At the protein level, the ΔPR mutation did not alter the relative levels of the different isoforms in brain tissue, although levels of SH2B1 protein were somewhat reduced (Fig. 4B). Importantly, in contrast to P322S/P322S mice, ΔPR/ΔPR homozygous mice were born and survived at the expected Mendelian frequency (Supplementary Fig. 1F), permitting us to examine the effect of this SH2B1 PH domain mutation in the homozygous state.

ΔPR/ΔPR mice fed standard chow exhibit significantly increased body weight compared with their WT littermates (Fig. 4C and D). By 20 weeks of age (Fig. 4C), ΔPR/ΔPR male mice were 15 g (>40%) heavier than their WT littermates, while female ΔPR/ΔPR mice were ∼9 g (∼35%) heavier than their WT littermates. It should be noted that we do not believe that the reduced levels of SH2B1 protein in the ΔPR/ΔPR mice can account for the increased obesity detected in ΔPR/ΔPR mice because heterozygote Sh2b1^−/+ mice are not obese (5). Body length was not significantly different in preliminary studies (Supplementary Fig. 1G and H). Overall adiposity (Fig. 4E and F) as well as circulating leptin concentrations (Fig. 4G) were increased in ΔPR/ΔPR homozygotes but not lean body mass (Fig. 4H). The heterozygous (ΔPR/+) male and female mice showed no significant increase in adiposity (Fig. 4F). However, ΔPR/+ males had a slight increase in circulating leptin levels (Fig. 4G), suggesting that in males, even a single copy of the ΔPR mutation may be sufficient to produce a minor effect on energy balance.

Increased food intake (assessed at 18–20 weeks) is observed in ΔPR/ΔPR male and female mice compared with their WT and ΔPR/+ littermates (Fig. 5A), while VO₂ (at 11–12 weeks of age) (Fig. 5B), respiratory exchange ratio (data not shown), and locomotor activity (data not shown) were not altered. On the basis of these data and the previous finding that Sh2b1 KO mice are obese primarily as a consequence of increased food intake (5), we believe it most likely that the ΔPR mutation caused obesity in the mice primarily as a consequence of increasing food intake rather than decreasing energy expenditure.

We initially examined parameters of glycemic control in ΔPR mice at 18–19 weeks of age. In homozygous ΔPR/ΔPR mice, hyperglycemia at baseline was evident (Fig. 5C) as well as impaired glucose tolerance (male and female mice) and insulin resistance (male mice) in intraperitoneal GTT and ITT, respectively (Fig. 5D and E). Male heterozygous ΔPR/+ mice (like P322S/+ mice) also displayed impaired glucose tolerance (Fig. 5D), although other parameters of glucose homeostasis were not different from WT littermates.

Because the disruption of glucose homeostasis in the aged ΔPR/ΔPR mice presumably resulted (at least in part) from their increased adiposity, we examined glucose homeostasis in young preobese mice to define any adiposity-independent effects of SH2B1 ΔPR on glucose homeostasis. We examined the adiposity of younger ΔPR/ΔPR mice to determine an age at which we might examine glucose homeostasis without it being confounded by increased adiposity. At 11–12 weeks of age, adiposity was already increased in male ΔPR/ΔPR mice but not detectably increased in female ΔPR/ΔPR mice (Fig. 6A). By 7 weeks, leptin levels were increased in male, but not female, ΔPR/ΔPR mice (Fig. 6B). We thus examined glucose homeostasis in ΔPR/ΔPR mice at 8 weeks of age, revealing hyperinsulinemia and glucose intolerance (with unchanged insulin tolerance) in both male and female ΔPR/ΔPR mice (Fig. 6C–F). The hyperinsulinemia and glucose intolerance in the presence of unchanged leptin and adiposity in young preobese females suggest that the ΔPR mutation interferes with glucose homeostasis independently of adiposity. Taken together, our results suggest that in obese ΔPR mice, the ΔPR mutation likely interferes with glucose homeostasis both independently of adiposity and secondary to the effects of adiposity on energy balance.

The identification of four human obesity-associated variants in the PH domain-encoding region of SH2B1, the fact that the three individuals with PH domain variants whose behavior has been documented all displayed behavioral abnormalities (1,2, and the present study), and the fact that the three individuals with variants in or near the PH domain had HOMA-IR scores suggesting severe insulin resistance and risk of type 2 diabetes, highlight the importance of the PH domain in SH2B1 function. The lethality of the human obesity-associated P322S mutation in the PH domain of SH2B1 in homozygous P322S/P322S mice demonstrates the importance of this mutation for SH2B1 function in vivo. Similarly, the obesity and diabetes observed in ΔPR/ΔPR mice highlight the importance of the SH2B1 PH domain for SH2B1-mediated metabolic control. The adiposity-independent glucose intolerance of young ΔPR/ΔPR female mice before the onset of obesity as well as of P322S/+ and ΔPR/+ mice also reveals the importance of SH2B1 and its PH domain for the control of glucose homeostasis independent of body weight, as previously suggested from the phenotype of humans bearing mutations in SH2B1 (1,2).

On the basis of the increased food intake of ΔPR/ΔPR mice and the findings in the Sh2b1 KO mice (5), we believe it likely that the increased body weight of the ΔPR/ΔPR mice is due to impaired function of SH2B1 in the hypothalamus. However, SH2B1 is also expressed in the periphery. The increased insulin concentrations with glucose intolerance in young, nonobese female ΔPR/ΔPR mice suggest alterations in tissues that control glucose uptake. However, the presence of glucose intolerance despite the increased insulin levels is consistent with the islets of ΔPR/ΔPR mice having an impaired ability to fully compensate for those alterations (41,42).

While humans heterozygous for P322S exhibit severe obesity, P322S/+ mice display mild glucose intolerance only in aged, HFD-fed animals. Because the region surrounding P322 is conserved between mice and humans (Fig. 7), it is unlikely that the more modest phenotype of P322S/+ mice compared with humans reflects species differences that result in structural changes in the SH2B1 PH domain, per se, but rather that PH domain binding partners may have different tolerances for P322S in mice and humans and/or that human physiology adapts more poorly to the resultant alterations in SH2B1 function. Consistent with the importance of the PH domain for the function of SH2B family members, at least nine point mutations (E208Q/E, A215V, G220V/R, A223V, G229S, D234N, F287S) have been identified in the PH domain of the SH2B1 ortholog SH2B3/Lnk in patients with myeloproliferative neoplasms (43–48) (Fig. 7).

To gain insight into how the SH2B1 P322S mutation or deletion of residues P317 and R318 in SH2B1 might regulate the function of the PH domain in SH2B1, we performed ClustalW analysis of SH2B family members and analyzed a model of SH2B1 that was based on the NMR structure of the PH domain of the SH2B1 family member SH2B2/APS (49). ClustalW analysis of the PH domains of the SH2B family members reveals that the PH domains are highly conserved (Fig. 7). In the model, residues P317, R318, and P322S in SH2B1 are on an exterior surface of the PH domain (Video 1). This surface is presumably a binding interface that interacts with either another region in SH2B1 or another protein. Another human obesity-associated variant in SH2B1 (E299G) as well as five of the human myeloproliferative neoplasm-associated variants in SH2B3/Lnk (G220V/R, A223V, G229S, and D234N) are in proximity to P317, R318, and P322 in SH2B1. In addition, eight of the human variants (E208Q, A215V, G220 V/R, G229S, D234N in SH2B3/Lnk and E299G, P322S in SH2B1) as well as P317 and R318 in SH2B1 are on or in proximity to this putative protein-binding interface (Fig. 7 and Video 1).

The number of human variants in SH2B1 and SH2B3 in this region of the PH domain suggests that small structural changes in this region as a result of mutation or other modification have the potential to produce substantial functional consequences. Because the residues corresponding to P317 and R318 in SH2B1 are on the surface of the PH domain and do not substantially change the direction of the loop, the P317, R318 deletion in SH2B1 would shorten the loop but not severely damage the overall structure. However, the deletion would be expected to diminish stabilization of the turn provided by the predicted π-π stacking between residues P317 and F309 in SH2B1. In addition, the deletion would be expected to alter the shape and electrostatics of the interface surface in the region of P317 and R318 in SH2B1.

Because SH2B1 from humans and mice share 95% sequence identity, with only one conservative difference (S325T) near P322 (Fig. 7), the structures in mouse and human are expected to be nearly identical. Therefore, the more modest phenotypes of P322S/+ mice compared with humans may be due to differences in the affinity of PH domain binding partners. The mouse binding partners may be able to accommodate the P322S mutation better than human binding partners, and/or human physiology adapts more poorly to the resultant alterations in SH2B1 function. Given the different phenotypes produced by the P322S and ΔPR mutations in mice, we postulate that the two mutations alter the structure of the SH2B1 PH domain in different ways to produce distinct changes in cell physiology. That relatively small changes in the PH domain, predicted to have only minor effects on PH domain structure, cause a rather profound effect on SH2B1β localization at the cellular level and energy balance and glucose homeostasis at the whole-animal level provides some of the first real evidence of the importance of the PH domain in SH2B1 function. While SH2B1β has been shown to cycle through the nucleus, it is generally found at the plasma membrane and in the cytoplasm (22,23,26). The accumulation of SH2B1β ΔPR in the nucleus indicates that the ΔPR deletion greatly alters the ratio between nuclear import and nuclear export of SH2B1β. Consistently, the ΔPR mutation as well as many of the other human obesity-associated SH2B1 variants impair the ability of SH2B1β to promote neurotrophic factor–induced neurite outgrowth of PC12 cells. Because neurite outgrowth in PC12 cells shares many properties with the formation of axons and/or dendrites (50), and because the Sh2b1 KO mice have impaired leptin signaling, it will be important in the future to examine the impact of SH2B1 PH domain changes on the structure of neurons that control energy balance.

Acknowledgments. The authors thank Drs. Malcolm Low, Miriam Meisler, Lei Yin, Xin Tong, Liangyou Rui, and Stephanie Bielas (University of Michigan) for helpful discussions and Dr. Rui (University of Michigan) for the gift of the Sh2b1 KO strain. The authors acknowledge the Wellcome-MRC Institute of Metabolic Science Translational Research Facility and Imaging Core Facility, both supported by a Wellcome Strategic Award (100574/Z/12/Z); the University of Michigan DNA Sequencing Core for DNA sequencing; and Dr. Thomas Saunders, Galina Gavrilina, and Dr. Wanda Filipiak of the University of Michigan Transgenic Animal Model Core as well as the Michigan Diabetes Research Center Molecular Genetics Core for help making the mouse models. The authors are indebted to the patients and their families for their participation and to the physicians involved in the Genetics of Obesity Study (www.goos.org.uk).

Funding. This work was supported by National Institutes of Health (NIH) grants R01-DK-54222 and R01-DK-107730 (to C.C.-S.). A.F. was supported by predoctoral fellowships from the Horace H. Rackham School of Graduate Studies, University of Michigan (Rackham Merit Fellowship); the Systems and Integrative Biology Training Program (NIH T32-GM-8322); and the Howard Hughes Medical Institute (Gilliam Fellowship for Advanced Study). Mouse body composition measurements were partially supported by the NIH-funded Michigan Diabetes Research Center (P30-DK-020572), Michigan Nutrition Obesity Research Center (P30-DK-089503), and Michigan Mouse Metabolic Phenotyping Center (U2C-DK-110678). Generation of the CRISPR mice was partially supported by the Molecular Genetics Core of the Michigan Diabetes Research Center (P30-DK-020572). Studies in humans were supported by the Wellcome Trust (207462/Z/17/Z to I.S.F. and WT206194 to I.B.); National Institute for Health Research Cambridge Biomedical Research Centre (to I.S.F.); and Bernard Wolfe Health Neuroscience Endowment (to I.S.F.).

The views expressed are those of the authors and not necessarily those of the NHS, National Institute for Health Research, or NIH.

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

Author Contributions. A.F. directed and conducted experiments, analyzed data, and prepared the manuscript. A.F. and L.S.A. designed and generated the mice. A.F., L.S.A., I.S.F, M.G.M., and C.C.-S. developed the concept, designed experiments, and interpreted the data. L.S.A. and J.S. analyzed the model of the PH domain (Fig. 7 and Video 1). L.S.A., I.S.F., M.G.M., and C.C.-S. made revisions to the manuscript. L.K.J.S. and E.M.d.O. characterized the human mutations in cells (Fig. 1B). A.E.M. helped to regenotype the mice. A.E.M., L.C.D., G.C., and Y.H. helped to measure body weight and food intake (Figs. 2A, B, H, and I, 4C, and 5A and Supplementary Fig. 2A, B, F, and G). P.B.V. conducted neurite outgrowth experiments (Fig. 3D) and helped with experiments for Fig. 3A and C. R.M.J. conducted preliminary experiments for Fig. 3B. J.M.C. made the GFP-SH2B1β WT and GFP-SH2B1β ΔPR constructs (Fig. 3). J.M.K., E.H., and I.S.F. performed the clinical studies in mutation carriers (Fig. 1A and Table 1). I.B. and I.S.F. performed the genetic studies (Fig. 1A and Table 1). E.S.C. maintained mouse colonies and helped to genotype mice and collect blood samples (Figs. 4G and 6B). All authors approved the final content. I.S.F. and C.C.-S. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this work were presented in poster form or as short oral presentations at the 2015 Neurotrophic Factors Gordon Research Conference, Newport, RI, 31 May–5 June 2015; 2016 Keystone Symposium on Molecular and Cellular Biology—Axons: From Cell Biology to Pathology, Santa Fe, NM, 24–27 January 2016; 2017 Keystone Symposium on Molecular and Cellular Biology: Neuronal Control of Appetite, Metabolism and Weight, Copenhagen, Denmark, 9–13 May 2017; 2017 Society for Advancement of Chicanos/Hispanics and Native Americans in Science (SACNAS) National Diversity in STEM Conference, Salt Lake City, UT, 19–21 October 2017; 2018 Molecular and Cellular Neurobiology Gordon Research Conference, Hong Kong, China, 1–6 July 2018; 2018 SACNAS National Diversity in STEM Conference, San Antonio, TX, 11–13 October 2018; Experimental Biology 2018, San Diego, CA, 21–25 April 2018; and 2019 Keystone Symposium on Molecular and Cellular Biology: Functional Neurocircuitry of Feeding Disorders, Banff, AB, Canada, 10–14 February 2019.