Mitotic regulators and the SHP2-MAPK pathway promote IR endocytosis and feedback regulation of insulin signaling

Insulin controls glucose homeostasis and cell growth through bifurcated signaling pathways. Dysregulation of insulin signaling is linked to diabetes and cancer. The spindle checkpoint controls the fidelity of chromosome segregation during mitosis. Here, we show that insulin receptor substrate 1 and 2 (IRS1/2) cooperate with spindle checkpoint proteins to promote insulin receptor (IR) endocytosis through recruiting the clathrin adaptor complex AP2 to IR. A phosphorylation switch of IRS1/2 orchestrated by extracellular signal-regulated kinase 1 and 2 (ERK1/2) and Src homology phosphatase 2 (SHP2) ensures selective internalization of activated IR. SHP2 inhibition blocks this feedback regulation and growth-promoting IR signaling, prolongs insulin action on metabolism, and improves insulin sensitivity in mice. We propose that mitotic regulators and SHP2 promote feedback inhibition of IR, thereby limiting the duration of insulin signaling. Targeting this feedback inhibition can improve insulin sensitivity.

T he pancreatic hormone insulin controls glucose homeostasis and promotes cell growth and proliferation. Dysregulation of insulin signaling is linked to human metabolic syndromes and cancer 1 . Insulin binds to the insulin receptor (IR) on the plasma membrane (PM), and triggers phosphorylation-mediated activation of crucial enzymes that regulate glucose and lipid metabolism, and cell growth and division 2,3 . Activated IR phosphorylates itself and the IR substrate (IRS) proteins on tyrosines. Phosphorylated IRS proteins bind to multiple downstream effectors and adaptors, and activate two major branches of insulin signaling: the phosphatidylinositol 3-kinase (PI3K)-protein kinase B (AKT) and mitogen-activated protein kinase (MAPK) pathways. The PI3K-AKT pathway mainly governs metabolic homeostasis, whereas the MAPK pathway controls cell growth and proliferation. Src homology phosphatase 2 (SHP2, also known as PTPN11) binds to the C-terminal phosphotyrosine sites of IRS1/2 and promotes the activation of the MAPK pathway 4,5 . Mutations of IR cause severe inherited insulin resistance syndromes 6 , but the molecular mechanisms underlying insulin resistance in type 2 diabetes are complex and multifactorial 1 . One common theme is that insulin resistance in diabetic animals or patients causes ectopic accumulation of diacylglycerol and abnormal activation of novel protein kinase Cs, which suppress insulin signaling at the level of IRS1 and possibly IR 1,7 .
We have recently discovered a critical role of MAD2, BUBR1, and p31 comet in insulin signaling during interphase (Fig. 1a) 18,19 . MAD2 and BUBR1 are required for clathrin-mediated endocytosis of IR. MAD2 directly binds to the C-terminal MAD2interacting motif (MIM) of IR, and recruits the clathrin adaptor AP2 to IR through BUBR1. p31 comet prevents spontaneous IR endocytosis through blocking the interaction of BUBR1-AP2 with IR-bound MAD2. Adult liver-specific p31 comet knockout  Representative images of liver specimens from human non-diabetic and diabetic patients stained with DAPI (blue) and anti-IR (red) and anti-ZO-1 (green) antibodies. Scale bars, 40 µm. c Quantification of the percentage of cells with positive IR PM signals in liver specimens in b (nondiabetes, n = 51; diabetes, n = 19; mean ± s.d.; ****p < 0.0001; Mann-Whitney test, a non-parametric test). The liver specimens from insulin-treated patients were presented as orange triangles. d Western blot analysis of lysates of HepG2 cells stably expressing IR-GFP WT that had been transfected with the indicated siRNAs. e Cells in d were serum starved, treated without or with 100 nM insulin for 5 min, and stained with anti-GFP antibodies. Scale bar, 5 μm. f Quantification of the ratios of PM and IC IR-GFP signals of cells in e (siLuc, n = 39; siLuc + Ins, n = 14; siBUBR1, n = 18; siBUBR1 + Ins, n = 29; siCDC20, n = 17; siCDC20 + Ins, n = 34; siBUB1, n = 23; siBUB1 + Ins, n = 12; siBUB3, n = 15; siBUB3 + Ins, n = 10; siMAD1, n = 19; siMAD1 + Ins, n = 18; mean ± s.d.; ****p < 0.0001; two-tailed unpaired t test). Source data are provided as a Source Data file (KO) mice exhibit premature IR endocytosis in the liver and whole-body insulin resistance. Conversely, BUBR1 deficiency delays IR endocytosis and enhances insulin sensitivity in mice. These findings implicate dysregulation of IR endocytosis as a potential mechanism of insulin resistance.
In this study, we show that IRS1/2 cooperate with several spindle checkpoint proteins to promote insulin-stimulated IR endocytosis. In addition to MAD2 and BUBR1, CDC20 is also required for IR endocytosis. We further identify feedback regulation of IR endocytosis through a phosphorylation switch on IRS1/2 that is dependent on SHP2 and the MAPK pathway. Finally, we present evidence to suggest that targeting SHP2 might be a viable strategy to increase insulin sensitivity and treat diabetes.

Results
Reduced IR PM localization in diabetes patients. Previous biochemical studies have shown that insulin binding to liver PMs is reduced in obese and diabetic mice and humans 20,21 , suggesting that the level of IR at the PM might be reduced in people with diabetes. To confirm these findings, we examined the IR PM levels in liver samples from human patients using immunofluorescence staining. Because of the challenges of collecting liver sections from normal healthy individuals, we used surgical resection samples from patients with hepatocellular carcinoma that contained normal (non-malignant) and malignant tissues, and analyzed only normal hepatocytes. We performed immunofluorescence with anti-IR and anti-ZO-1 (as a PM marker) antibodies on 51 non-diabetic and 19 type 2 diabetes patient samples, and analyzed IR PM levels. The percentage of cells with IR PM signals in the liver sections from type 2 diabetes patients was significantly lower than that in non-diabetic patients (Fig. 1b, c). There was no correlation between insulin treatment and IR PM levels. These findings suggest that reduced IR PM levels might be a contributing factor to insulin resistance in human patients. CDC20 is required for proper IR endocytosis. We have previously shown that MAD2 and BUBR1 promote IR endocytosis through recruiting AP2 (Fig. 1a), and p31 comet inhibits this process through antagonizing the MAD2-BUBR1 interaction 18 . We further tested whether other well-known mitotic regulators also have a role in IR endocytosis. Depletion of CDC20, BUB1, BUB3, or MAD1 did not alter the ratio of PM and intracellular (IC) IR signals in the absence of insulin (Fig. 1d-f). Depletion of CDC20 or BUBR1, but not BUB1, BUB3, or MAD1, increased the PM/IC ratio of IR in the presence of insulin and delayed insulin-activated IR endocytosis. Normalization of PM/IC ratios of insulin-treated samples to those of untreated samples produced similar results (Supplementary Fig. 1a). Thus, like BUBR1, CDC20 is required for insulin-activated IR endocytosis. CDC20 binds to both MAD2 and BUBR1 and nucleates the formation of MCC that consists of BUBR1, BUB3, MAD2, and CDC20 11,13,22,23 . BUB3 is not absolutely required for the integrity and function MCC, but enhances the efficiency of MCC formation by targeting BUBR1 to kinetochores. Structures of MCC nicely explain the binding cooperativity among MAD2, CDC20, and BUBR1, and highlight the importance of the BUBR1 KEN1 motif in MCC assembly 22 . We hypothesized that the assembly of an MCC-like complex on IR might strengthen the weak MAD2-BUBR1 interaction and help to recruit AP2 to IR.
To test this hypothesis, we first examined the region of BUBR1 responsible for IR endocytosis. The C-terminal domain of BUBR1 binds to the N-terminal trunk domain of AP2B1 (Fig. 2a-c and Supplementary Fig. 1b), consistent with a previous report 24 .
Next, we generated human 293FT cell lines stably expressing IR tagged with green fluorescent protein (GFP) at its C terminus ( Supplementary Fig. 1c) and examined the localization of IR-GFP after transfection with the indicated small interfering RNAs (siRNAs) or siRNA-resistant Myc-BUBR1 constructs. Expression of wild-type (WT) BUBR1, but none of BUBR1 truncation mutants, restored IR endocytosis in cells depleted of BUBR1 (Fig. 2d, e and Supplementary Fig. 1d, e). BUBR1 interacts with CDC20 through multiple binding motifs 11,13,25 . We constructed BUBR1 mutants targeting KEN1 (ΔKEN1) or KEN2 (ΔKEN2) boxes in the N-terminal region, and Phe (ΔPhe, also known as ABBA3) and D boxes (ΔD2) in the middle region (Fig. 2a). The ΔKEN2, ΔPhe, and ΔD2 mutants restored insulinactivated IR endocytosis in cells depleted of BUBR1 (Fig. 2d, e and Supplementary Fig. 1d, e). The ΔPhe + ΔD2 mutant was, however, slightly defective. Consistent with this cellular phenotype, the middle region of BUBR1 containing Phe and D2 boxes showed weak binding to AP2B1, and mutation of both Phe and D boxes in full-length BUBR1 (ΔPhe + ΔD2) decreased the interaction, suggesting that BUBR1 binds to AP2B1 through multiple binding sites (Fig. 2b). Strikingly, BUBR1 ΔKEN1 could not restore IR endocytosis in BUBR1-depleted cells (Fig. 2d, e and Supplementary Fig. 1d, e). Given the critical role of BUBR1 KEN1 in MCC assembly, these data suggest that IR uses an MCC-like complex to recruit AP2, except that in this context MAD2 binds to the MIM on IR, not the MIM on CDC20.
IRS1/2 promote IR endocytosis. Insulin binding to IR promotes clathrin-mediated endocytosis of the insulin-IR complex, which regulates the intensity and duration of signaling. Two sequence motifs in IR--the NPEY 960 and di-leucine (L 986 L) motifs--have previously been implicated in AP2 binding and IR endocytosis [26][27][28][29] (Fig. 3a). We have shown that the MIM (R 1333 ILTL) of IR binds to MAD2, which in turn recruits AP2 with the help of BUBR1 and CDC20 18 .
To study the relative contributions of these IR motifs to endocytosis, we generated HepG2 cell lines stably expressing IR-GFP WT, the MIM mutant (4A), Y960F, or L986A/L987A (AA), and examined the subcellular localization of these IR-GFP proteins (Fig. 3b, c). Without insulin treatment, IR WT, 4A, and Y960F localized to the PM, but IR AA was enriched in IC compartment. A large fraction of IR AA co-localized with the late endosome marker RAB7, indicating that IR AA underwent unscheduled endocytosis and accumulated in late endosomes (Fig. 3d). The di-leucine motif is located in a β strand of the N-terminal lobe of the IR kinase domain (Fig. 3e) 30 . Mutation of this motif is expected to alter the structural integrity or activity of the IR kinase domain.
The levels of ectopically expressed IR-GFP proteins were higher than that of the endogenous IR ( Supplementary Fig. 2a, b). Furthermore, partial depletion of endogenous IR did not change the subcellular localization of these IR mutants. Thus, the behaviors of these mutants were not dependent on stoichiometric heterodimerization with the endogenous IR. On the other hand, because the IR depletion was incomplete, it remains possible that the residual endogenous IR might modify and regulate the ectopically expressed IR in a catalytic manner.
IR Y960F, 4A, and Y960F/4A mutants were less efficiently internalized after insulin stimulation ( Fig. 3f and Supplementary  Fig. 2c). As Y960 is phosphorylated in the activated IR 31 , defective endocytosis of IR Y960F suggests that phosphorylation of Y960 (pY960) might be required for timely IR internalization. The phosphotyrosine-binding domain of IRS1/2 directly binds to phosphorylated NPEY 960 motif in activated IR [32][33][34][35] . Co-depletion of IRS1/2 blocked IR endocytosis induced by insulin, whereas depletion of either had no effect (Fig. 4a, b and Supplementary  Fig. 3a, b). Expression of RNAi-resistant IRS1 restored IR endocytosis in cells depleted of both IRS1/2. Thus, IRS1/2 act redundantly to promote IR endocytosis likely through binding to the phospho-NPEY 960 motif. These results can explain why the activated IR is preferentially internalized.
These results suggest that IRS1 and IRS2 directly bind to AP2M1 through multiple YXXΦ motifs.
RNAi-resistant IRS1 3YA mutants could not restore IR endocytosis in 293FT or HepG2 cells depleted of IRS1/2 (Fig. 4d, e and Supplementary Fig. 4c-f). Failure of these mutants to functionally complement indicates that the IRS1/2-AP2M1 interaction is required for insulin-activated IR endocytosis. We could not detect IRS1 in late endosomes during IR endocytosis, suggesting that IRS1 might dissociate from the IR complex prior to the fusion of the endocytic vesicles with the late endosome. Endogenous IRS1 interacted with the AP2 complex in 293FT cells and in primary mouse hepatocytes stimulated with insulin, but not in untreated cells (Fig. 4f, g). Thus, IRS1/2 bind to AP2 through canonical YXXΦ motifs in vitro and in mammalian cells, and promote insulin-activated IR endocytosis.
The SHP2-MAPK pathway promotes IR endocytosis through IRS1/2. The tyrosine residues in YXXΦ motifs on IRS1/2 can be phosphorylated by the activated IR and be dephosphorylated by the tyrosine phosphatase SHP2 2,39 . Strikingly, the serine residues immediately following the YXXΦ motifs are well conserved ( Fig. 4c and Supplementary Fig. 3f), and can be phosphorylated by ERK1/2 [40][41][42] . ERK1/2-dependent phosphorylation has been proposed to reduce IRS1 tyrosine phosphorylation through negative feedback [40][41][42] , but the mechanism and function of this feedback regulation are unknown. We hypothesized that the MAPK pathway and SHP2 might regulate the IRS1/2-AP2 interaction and IR endocytosis through modulating IRS1/2 phosphorylation patterns.
To test this hypothesis, we examined the effects of inhibiting SHP2 or the MAPK pathway on insulin-activated IR endocytosis (Fig. 5a, b and Supplementary Fig. 5). The IR inhibitor (BMS536924) expectedly blocked IR endocytosis. Strikingly, the MEK inhibitors (U0126 and PD0325901) and the SHP2 inhibitor (SHP099) also inhibited IR endocytosis. By contrast, the AKT inhibitor (AKTi, VIII) did not affect IR endocytosis, indicating a specific requirement for the MAPK pathway and SHP2. Inhibitors of MPS1 (Reversine) and PLK1 (BI2546) did not appreciably inhibit IR endocytosis, ruling out the involvement of these mitotic kinases in this process.
The tyrosine residues in YXXΦ motifs on IRS1/2 also bind PI3K 43,44 . The failure of IRS1 3YA mutant to restore IR endocytosis in cells depleted of IRS1/2 could be an indirect consequence of reduced PI3K activity. To test this possibility, we checked the effect of IRS1 3YA on insulin signaling ( Supplementary Fig. 4g). IRS1/2 depletion was sufficient to inhibit IR endocytosis, but was insufficient to inhibit insulin signaling at insulin concentrations used in our IR endocytosis assays. Expression of IRS1 WT or the 3YA mutant did not appreciably alter insulin signaling. Furthermore, the PI3K inhibitor (GDC0941) did not affect insulin-activated IR endocytosis (Fig. 5a, b and Supplementary Fig. 5). This result argues against reduced PI3K activity as the underlying reason for the observed defects of insulin-activated IR endocytosis in IRS1/2depleted cells, and further confirms that the PI3K-AKT pathway is not involved in this process.
We did not observe well-defined electron density for pS616 in IRS1, despite its ability to enhance the IRS1-AP2 interaction. pS616 is located in the vicinity of a positively charged patch on AP2M1 formed by residues K405, H416, and K420 (Fig. 6c), suggesting that the phospho-serine might engage in favorable electrostatic interactions with this basic patch. Mutations of H416 and K420 did not, however, reduce IRS1 binding ( Supplementary  Fig. 6b-d). Mutation of K405 destabilized the AP2M1 protein and reduced its binding to both the phosphorylated (pS616) and unphosphorylated IRS1 peptides. Thus, consistent with the lack of electron density, pS616 does not make defined electrostatic interactions with specific acceptor residues, and interacts with the positively charged patch as one structural entity.
SHP2 promotes IR endocytosis in mice. Liver-specific SHP2 KO mice show increased insulin sensitivity 45,46 , suggesting that SHP2 attenuates certain aspects of insulin signaling in the liver. We examined whether SHP2 inhibition could improve insulin sensitivity. The allosteric SHP2 inhibitor, SHP099, stabilizes the inactive conformation of SHP2, thus inhibiting its phosphatase activity 4 . WT mice maintained on high-fat diet (HFD) for - 5 weeks were treated with SHP099 (60 mg per kg body weight) by daily oral gavage for 6 days, and then tested for glucose and insulin tolerance. Strikingly, SHP099 administration markedly increased glucose tolerance and insulin sensitivity in HFD-fed mice (Fig. 7a, b). SHP099 did not change the body weight of mice fed HFD (Fig. 7c). Insulin stimulation caused IR endocytosis and reduced the IR staining at the PM in mouse liver sections (Fig. 7d, e). SHP099 delayed insulin-activated IR endocytosis. Finally, SHP099 inhibited the insulin-stimulated IRS1-AP2 interaction in primary hepatocytes (Fig. 4g).
To further confirm the requirement of SHP2 in promoting IR endocytosis in vivo, we introduced adeno-associated viruses (AAV) encoding control (Ctrl) or SHP2 short-hairpin RNAs (shRNA) into mice fed HFD via tail-vein injection. The SHP2 protein level in the liver from mice treated with AAV-SHP2 shRNA was reduced by~70% as compared to that in mice treated with Ctrl shRNA (Supplementary Fig. 7a). By contrast, the SHP2 protein level in white adipose tissue (WAT) and skeletal muscle was not effectively depleted by SHP2 shRNA. Thus, as expected, SHP2 shRNA delivered by AAV was most effective in the liver. Consistent with the phenotypes of SHP099 administration, AAV-SHP2 shRNA treatment markedly increased glucose tolerance and insulin sensitivity in mice fed with HFD ( Supplementary Fig. 7b, c). AAV-SHP2 shRNA treatment did not change body weight ( Supplementary Fig. 7d). Importantly, AAV-mediated SHP2 silencing inhibited insulinactivated IR endocytosis in the liver ( Supplementary Fig. 7e, f). These results confirm a role of SHP2 in promoting IR endocytosis and in metabolic homeostasis in vivo.
In addition to being a major organ for insulin signaling, liver is a major site for insulin clearance and defects in this process can cause hyperinsulinemia 47,48 . Hepatic insulin clearance is mainly mediated by IR endocytosis, as liver-specific IR KO mice and mice deleted of the carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1), a key regulator of IR endocytosis, develop severe hyperinsulinemia 47,48 . We thus examined the effects of SHP099 on insulin clearance. To estimate insulin clearance, we examined the levels of insulin and Cpeptide, a cleavage product of proinsulin, and determined the ratio of the serum levels of C-peptide and insulin. In HFD-fed mice, SHP099 slightly increased the fasting insulin level, but this increase was not significant (Fig. 7f). The serum levels of Cpeptide were similar to those in vehicle-treated groups in both conditions (Fig. 7g), suggesting that SHP099 did not alter insulin secretion. As a result, the C-peptide:insulin ratio in mice fed HFD after SHP099 administration showed a mild reduction as compared to the control group (Fig. 7h). Therefore, despite causing defective IR endocytosis in the liver, SHP2 inhibition led to a mild defect in insulin clearance, but not severe hyperinsulinemia as observed in the CEACAM1 KO mice.
As shown in Fig. 5b, the inhibitory effect of SHP099 on IR endocytosis was not as complete as that of the IR inhibitor at 20 min after insulin stimulation, suggesting that there might be SHP2-independent IR endocytosis mechanisms. We have confirmed a requirement for CEACAM1 in IR endocytosis in our cellular assays (Fig. 7i, j). It is possible that CEACAM1 is required for all mechanisms of IR endocytosis, whereas SHP2 only regulates the IRS1/2 branch (Fig. 7k). In addition, other protein tyrosine phosphatases can contribute to IRS1/2 dephosphorylation, thereby regulating IR endocytosis. Future studies are needed to define the relationships and relative contributions of the various regulators of IR endocytosis.
SHP2 inhibition enhances insulin-activated AKT pathway. We next examined the effect of SHP2 inhibition on insulin signaling in the liver, epididymal WAT, and skeletal muscle from mice fed with HFD. We monitored the insulin-induced phosphorylation of IRS1 (pY608 and pS612; equivalent to pY612 and pS616 of human IRS1, respectively), IR (pY1152/1153 and pY962; equivalent to pY1150/1151 and pY960 of human IR, respectively), AKT (pT308), and ERK1/2 (pT202/Y204, pERK1/2). Indeed, SHP2 inhibition enhanced and prolonged the IRS1 pY608 signal in the liver, indicating that SHP2 is a relevant phosphatase for this phosphorylation in vivo (Fig. 8a, b). Consistent with a previous report 4 , SHP099 inhibited the activation of the MAPK pathway by insulin in the liver. SHP2 inhibition attenuated insulin-induced IRS1 pS612 signal, consistent with the fact that ERK1/2 mediate this phosphorylation in vivo. By contrast, insulin-triggered activating phosphorylation of AKT was elevated and prolonged in the liver from SHP099-treated mice. Consistent with an inhibition of IR endocytosis, the total IR levels in both groups of mice were increased by SHP099 treatment, whereas the ratios of phospho-IR to total IR were not altered. These results suggest that targeting SHP2 can block the feedback regulation of IR endocytosis by selectively inhibiting the MAPK pathway. Suppressed IR endocytosis prolongs signaling through the PI3K-AKT pathway, which regulates metabolism and does not depend on SHP2.

Discussion
We have elucidated a feedback regulatory mechanism of IR endocytosis and identified SHP2 as a molecular target whose inhibition delays insulin-activated IR endocytosis (Fig. 8c). Targeting this feedback regulation of IR endocytosis prolongs the metabolic branch of insulin signaling and improves insulin sensitivity in mice. The mechanism of IR endocytosis was extensively studied for decades. Our recent discovery that mitotic checkpoint regulators, including MAD2 and BUBR1, are required for IR endocytosis promoted us to re-examine the mechanism of IR endocytosis. In addition to the recently discovered MIM, two other sequence motifs in IR, the NPXY 960 motif and the di-leucine motif (L 986 / L 987 ), had previously been implicated in AP2 binding and IR endocytosis [26][27][28][29] . In this study, we have shown that the di-leucine motif is not required for IR endocytosis. The NPXY motif is indeed required for IR endocytosis. Instead of directly binding to AP2, the phosphorylated NPXY motif binds to IRS1/2, which in turn recruits AP2 through multiple YXXΦ motifs. We propose that IRS1/2 bound to pY999 of IR provide one binding site for AP2, and the MAD2-CDC20-BUBR1 complex bound to the MIM of IR provides another AP2-binding site. These two modules collaborate to recruit AP2 to IR, triggering IR endocytosis. Alternatively, as IRS1/2 contain multiple functional YXXΦ motifs, each of these motifs and MAD2-CDC20-BUBR1 may mediate the recruitment of one AP2 complex, leading to the clustering of multiple AP2 molecules on IR and efficient assembly of the clathrin coat.
Our findings presented herein suggest that an MCC-like mitotic regulator module is assembled onto IR to control its endocytosis in interphase. These results raise the intriguing possibility that IR might reciprocally regulate MCC assembly and spindle checkpoint signaling during mitosis. It will be interesting to examine the effect of the connection between IR and mitotic Fig. 5 The SHP2-MAPK pathway promotes insulin-activated IR endocytosis. a HepG2 cells expressing IR-GFP WT were starved, treated with the indicated inhibitors for 2 h, treated without or with 100 nM insulin for 20 min, and stained with anti-GFP (IR; green) and DAPI (blue). Scale bar, 5 µm. b Quantification of the ratios of PM and IC IR-GFP signals of cells in a (DMSO, n = 47; DMSO + Ins, n = 56; IRi, n = 50; IRi + Ins, n = 65; U0126, n = 96; U0126 + Ins, n = 111; PD0325901, n = 26; PD0325901 + Ins, n = 54; SHP099, n = 43; SHP099 + Ins, n = 49; GDC0941, n = 57; GC0941 + Ins, n = 48; AKTi, n = 65; AKTi + Ins, n = 96; BI2536, n = 26; BI2536 + Ins, n = 24; Reversine, n = 37; Reversine + Ins, n = 27; mean ± s.d.; ****p < 0.0001; two-tailed unpaired t test). c Binding of IRS1 peptides to AP2M1 (residues 160-435). Input and proteins bound to IRS1 peptide beads were analyzed by SDS-PAGE and stained with Coomassie (CBB). The relative band intensities are shown below (mean ± s.d.; n = 5 independent experiments). d Isothermal titration calorimetry (ITC) analysis of binding between IRS1 peptides and AP2M1 (residue 160-435), with K d indicated. e The IRS1 peptides were incubated with active SHP2 for the indicated durations, spotted onto membranes, and detected with the anti-pY612-IRS1 antibody. f Quantification of the relative SHP2 activity in e (mean ± s.d.; n = 4 independent experiments; ****p < 0.0001; two-tailed unpaired t test). g Model of the regulation of insulin-activated IR endocytosis by a phosphorylation switch on IRS1/2. Insulin-bound IR phosphorylates itself and IRS1/2, and activates the PI3K-AKT and MAPK pathways. SHP2 acts upstream of RAS-RAF and promotes the activation of MAPK pathway. p31 comet binds to the IR-bound MAD2 and blocks IR-AP2 association to prevent premature IR endocytosis. In feedback regulation, activated ERK1/2 phosphorylate S616 and other sites on IRS1. SHP2 binds to the C-terminal phosphotyrosine site on IRS1 and dephosphorylates pY612 of the doubly phosphorylated IRS1 (pY612/pS616), thus promoting IRS1-AP2M1 association. p31 comet is released from MAD2 by an unknown mechanism, allowing the assembly of an MCC-like complex on IR. MAD2-and IRS1/2-dependent AP2 recruitment and clustering trigger clathrin-mediated IR endocytosis. Source data are provided as a Source Data file  The crystal structure of AP2M1 bound to pS616-IRS1. a Ribbon diagram of the crystal structure of AP2M1 (residues 160-435) bound to pS-IRS1. pS-IRS1 is shown as sticks. b Surface drawing of AP2M1, with pS-IRS1 shown as sticks. c A close-up view of the surface drawing of AP2M1 colored by its electrostatic potential (blue, positive; red, negative; white, neutral). pS-IRS1 is shown as sticks checkpoint regulators on aneuploidy and spindle checkpoint activities. The levels of CDC20 and BUBR1 vary throughout the cell cycle and during aging 49,50 . BUBR1 insufficiency causes premature aging and enhances insulin sensitivity in mice 18,49 . IR/IGF1R signaling is associated with aging and longevity 51 . Future studies are needed to test if the alteration of the levels of mitotic regulators directly contributes to aging through the regulation of insulin signaling.
Endocytosis of a cell surface receptor normally occurs after the receptor has been activated and has transduced its signals downstream. The mechanism ensuring that IR endocytosis only occurs after downstream signaling is unknown. In this study, we have shown that the MAPK pathway is required for IR endocytosis. Activated ERK1/2 phosphorylate the YXXΦS motifs in IRS1/2 and promote dephosphorylation of the tyrosine in these motifs by SHP2 in vitro and in vivo. The YXXΦS motifs containing phospho-serine and unphosphorylated tyrosine are optimal ligands for AP2. Because ERK1/2 activation is a critical downstream event of insulin signaling, the requirement for an active MAPK pathway in IR endocytosis constitutes important feedback regulation that ensures proper insulin signaling. Chemical inhibition of SHP2 is an effective way to disrupt this feedback regulation, as it not only directly prevents the removal of the negative tyrosine phosphorylation but also inhibits the MAPK pathway and blocks the installation of the positive serine phosphorylation.
Our study suggests that inhibition of SHP2 delays IR endocytosis, prolongs IR signaling through the PI3K-AKT pathway, and improves insulin sensitivity in the mouse. A recent study showed that the SHP2 inhibitor 11-a attenuated the activation of PI3K-AKT pathway upon hepatocyte growth factor stimulation in melanoma cell lines 52 . Thus, SHP2 inhibition selectively enhances the PI3K-AKT pathway upon insulin stimulation, but not other stimuli. SHP2 inhibition prolongs phosphorylation of YXXΦ motifs on IRS1/2. Because these phosphotyrosine motifs also interact with PI3K, the enhanced tyrosine phosphorylation of these motifs caused by SHP2 inhibition may directly promote PI3K activation. This direct effect may contribute to the effectiveness of SHP2 inhibitors in counteracting insulin resistance.
Although we have shown the reduction of IR on the cell surface in liver samples of human type 2 diabetes patients, it can be a consequence of insulin resistance, rather than the cause. It is also possible that the reduced IR localization at the PM is due to a decrease in overall IR expression. Despite these caveats, our immunofluorescence results are consistent with earlier findings, and suggest that mechanisms regulating IR PM levels, including endocytosis, might be defective in people with diabetes. The circulating insulin level is elevated with insulin resistance to prevent hyperglycemia. There is also evidence that insulin hypersecretion can cause insulin resistance 53,54 . Genetic suppression of insulin secretion improves insulin sensitivity in aging mice and extends lifespan 51 . It will be interesting to test if the hyperinsulinemia-mediated insulin resistance is associated with the reduction of IR on the cell surface.
Our current study monitored IR endocytosis and insulin signaling at fixed time points, but not in real time. Moreover, although the depletion of endogenous IR did not change the cellular localization of IR mutants, we cannot rule out contributions from heterodimers formed between the endogenous IR and the ectopically expressed IR mutants to the localization and endocytosis of IR mutants. Targeting and mutating the endogenous IR loci are needed to examine the real-time kinetics of IR endocytosis and the effects of various perturbations in cells and in vivo.
Future studies are required to clarify the potential role of premature IR endocytosis in the pathogenesis of human insulin resistance syndromes. Mutations of IR are known to cause inherited severe insulin resistance syndromes 6 , but the mechanisms by which these mutations affect IR function have not been systematically explored. It will be interesting to test whether these mutations cause premature IR endocytosis and whether SHP2 inhibitors can recover the IR PM levels and insulin sensitivity.
Hyperactivation of signaling through receptor tyrosine kinases (RTKs) drives cancer progression. Chemical inhibitors of RTKs are widely used in the clinic to treat human cancers. A common signaling pathway downstream of RTKs is the RAS-RAF-MAPK pathway, which promotes cell proliferation and survival. SHP2 acts upstream of RAS-RAF to activate the MAPK pathway 55,56 . Thus, intense efforts have been devoted to develop chemical inhibitors of SHP2, which are expected to have therapeutic potential in cancer chemotherapy. SHP099, a specific allosteric inhibitor of SHP2, has been shown to have efficacy in targeting RTK-driven cancers in animal models 4 . Although phenotypes of SHP2 KO mice suggest potential adverse effects of SHP2 inhibition 57-59 , SHP099 administration into mice did not show significant toxic effects in that previous study 4 and in the present study. Our study shows that short-term SHP2 inhibition improves systemic insulin sensitivity in mice. Thus, SHP2 inhibitors, such as SHP099, can be potentially repurposed to treat type 2 diabetes. Obesity increases the risks of both diabetes and certain types of cancers in humans 60 . The prevalence of all three conditions (obesity, diabetes, and cancer) has increased in recent years. SHP2 inhibitors may be particularly beneficial to patients who suffer from both diabetes and cancer.

Methods
Mice. Animal work described in this manuscript has been approved and conducted under the oversight of the University of Texas Southwestern Institutional Animal Care and Use Committee. All animals were maintained in a specific antigen-free barrier facility with 12 h light/dark cycles (6 a.m. on and 6 p.m. off). Mice were fed a standard rodent chow (2016 Teklad Global 16% protein rodent diet, Harlan Laboratories). For inducing insulin resistance, C57BL/6J (Stock No. 000664, Jackson laboratory) were fed a HFD (60%) (OpenSource Diets, Cat. No. D12492).
For the expression of IRS1 WT or mutants in HepG2 cells, cDNA encoding IRS1 WT or mutants were cloned into the pBabe-mCherry-puro vector. The vectors were co-transfected with viral packaging vectors into 293FT cells, and the viral supernatants were collected. The concentrated viruses were added to HepG2 cells stably expressing IR-GFP WT. The cells were transfected with the indicated siRNA at 1 day after viral infection. Analysis was performed at 4 days after infection. The cells were serum starved for 14 h before insulin treatment.
Primary hepatocytes were isolated from 2-month-old male mice with a standard two-step collagenase perfusion procedure. Cells were plated on collagen-coated dishes and incubated in attachment medium [Williams' E Medium supplemented with 5% (v/v) FBS, 10 nM insulin, 10 nM dexamethasone, and 1% (v/v) penicillin/ streptomycin]. After 2-4 h, the medium was changed to low-glucose DMEM supplemented with 5% (v/v) FBS, 10 nM dexamethasone, 10 nM insulin, 100 nM triiodothyronine, and 1% (v/v) penicillin/streptomycin. After 1 day, the cells were serum starved for 14 h and treated with DMSO or SHP099 for 2 h.
Tissue histology and immunohistochemistry. Mouse tissues were fixed in buffered 10% Neutral Buffered Formalin (NBF) for hematoxylin and eosin staining by the Molecular Pathology Core at University of Texas Southwestern Medical Center. For human patient sample analysis, the deparaffinized sections were subjected to antigen retrieval with 10 mM sodium citrate (pH 6.0), incubated with 0.3% H 2 O 2 , blocked with 0.3% bovine serum albumin (BSA), and then incubated first with anti-IR (CT-3, Millipore, 1:100) and anti-ZO-1 (Thermo Scientific, 1:200) antibodies and then with secondary antibodies (Alexa Fluor 568 goat anti-mouse and Alex Fluor 488 goat anti-rabbit; Molecular Probes). The slides were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). Five to nine images (depends on the percentage of normal hepatocytes) were randomly taken under ×40 magnification. The total cell numbers and numbers of IR PM-positive cells were counted at least twice for individual images. Total cell numbers are counted with images stained with the ZO-1 membrane marker. Over 100 cells were analyzed for each patient samples. All immunohistochemistry and scoring were performed blinded to the diabetes status. Human tissue collection and analysis were performed under the supervision of the Institutional Review Board (STU 062013-043).
Metabolic profiling. Blood samples were collected from overnight-fasted mice using submandibular bleeding methods. For serum preparation, blood was allowed to form clots at room temperature for 30 min, centrifuged at 3000 × g for 15 min at 4°C, and stored at −80°C. Serum insulin and C-peptide levels were determined with the STELLUX Chemic Rodent Insulin ELISA kit (Alpco) and Mouse C-peptide ELISA kit (Alpco), respectively. ELISA analysis for insulin and C-peptide was performed by Metabolic Phenotyping Core at UT Southwestern Medical Center.
Immunoprecipitation and quantitative western blots. Cells were incubated with the cell lysis buffer [50 mM HEPES (pH 7.4), 150 mM NaCl, 10% (v/v) glycerol, 1% (v/v) Triton X-100, 1 mM EDTA, 100 mM sodium fluoride, 2 mM sodium orthovanadate, 20 mM sodium pyrophosphate, 0.5 mM dithiothreitol (DTT), 2 mM phenylmethylsulfonyl fluoride (PMSF)] supplemented with protease inhibitors (Roche) and PhosSTOP (Roche) on ice for 1 h. The cell lysates were cleared by centrifugation and incubated with antibody-conjugated beads. The beads were washed, and the bound proteins were eluted with the SDS sample buffer and analyzed by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and western blotting. For quantitative western blots, anti-rabbit immunoglobulin G (IgG) (H + L) (Dylight 800 or 680 conjugates) and anti-mouse IgG (H + L) (Dylight 800 or 680 conjugates) (Cell Signaling) were used as secondary antibodies. The membranes were scanned with the Odyssey Infrared Imaging System (LI-COR, Lincoln, NE). Uncropped and unprocessed scans of the most important blots are provided in the Source Data file.
For analyzing the ratios of phospho/total protein signals, the band intensities of phospho-proteins and total proteins were measured at the indicated time points. The values were divided by the basal intensities of phospho-proteins and total proteins, respectively. The relative intensities of phospho-proteins were divided by the relative intensities of total proteins in each condition to yield the final ratios.
Immunofluorescence. Indirect immunofluorescence microscopy was performed on cells grown on coverslips and fixed in cold methanol at −20°C for 10 min. The fixed cells were incubated with phosphate-buffered saline (PBS) for 30 min and 3% BSA in 0.1% PBST for 1 h, and then treated with diluted antibodies in 0.3% BSA in 0.1% PBST at 4°C overnight. After being washed, cells were incubated with fluorescent secondary antibodies and mounted on microscope slides in ProLong Gold Antifade reagent with DAPI (Invitrogen). Images of fixed cells were acquired as a series of 0.4 μm stacks with a DeltaVision system (Applied Precision, Issaquah, WA). Raw images were deconvolved using the iterative algorithm implemented in the softWoRx software (Applied precision, Issaquash, WA). The central section of a 0.4 µm z-stack containing three contiguous focal planes was used for quantification. The cell edges were defined with Image J. The whole cell signal intensity (WC) and IC signal intensity (IC) were measured. The PM signal intensity (PM) was calculated by subtracting IC from WC. Identical exposure times and magnifications were used for all comparative analyses.
The AP2M1 fragment (residues 160-435) was cloned into a modified pET28a that introduced an N-terminal His 6 -tag followed by a thrombin cleavage site. The plasmid was transformed into BL21 (DE3). Protein expression was induced by 0.2 mM IPTG at 20°C overnight. The harvested pellets were lysed in the lysis buffer II [20 mM Tris-HCl, pH 7.5, 500 mM NaCl, 20 mM imidazole, and 1 mM PMSF]. After sonication, lysates were cleared by centrifugation at 4°C. The supernatants were filtered by 0.45 μm filter and incubated with pre-equilibrated Ni2 + -NTA beads (Qiagen). Protein-bound beads were washed with 150 ml of wash buffer I [20 mM Tris-HCl, pH 7.5, 1 M NaCl, and 20 mM imidazole] and with 50 ml of wash buffer II [20 mM Tris-HCl, pH 7.5, 100 mM NaCl, and 20 mM imidazole]. The proteins were then eluted with the elution buffer [20 mM Tris-HCl, pH 7.5, 100 mM NaCl, and 150 mM imidazole] and incubated with thrombin (Sigma) at 4°C overnight. The protein was further purified with a Superdex 200 size exclusion column (GE Healthcare). The relevant protein fractions were pooled, aliquoted, and snap-frozen for future experiments.
Crystallization of the AP2M1-pIRS1 complex. The purified AP2M1 (residues 160-435) was mixed with the pS-IRS1 peptide (CHTDDGYMPMpSPGVA, residues 607-620) at a molar ratio of 1:5 and then crystalized with the hanging-drop vapor diffusion method. The crystals of the AP2M1-pS-IRS1 complex grew within few days after the protein solution was mixed with the reservoir solution [1.0 M sodium malonate, pH 5.0, 0.1 M sodium acetate tri-hydrate, pH 4.5, and 2% (w/v) PEG 20k]. All crystals were cryoprotected with the reservoir solution including 15% (w/v) glycerol for data collection.
Data collection and structure determination. X-ray diffraction datasets were collected at the Advanced Photon Source beamline Sector 19-ID at a wavelength of 0.97914 Å and at 100 K. HKL3000 was used to process the datasets 62 . The crystal of the AP2M1-pS-IRS1 complex diffracted to a minimum Bragg spacing of 3.2 Å and exhibited the symmetry of space group P6 4 with cell dimensions of a = b = 125.33 Å, c = 74.82 Å. There are two molecules in the asymmetric unit, with a 53.4% solvent content.
The structure was determined by molecular replacement with PHASER-MR 63 , using the structure of the AP2M-IGN38 complex (PDB ID: 1BXX) as the search model. Structure refinement was performed with COOT and PHENIX [64][65][66] . The final R work and R free were 20.3% and 23.6%, respectively. The Ramachandran statistics indicated that 91%, 8.4%, and 0.4% of the residues were in the favored, the allowed, and outliers regions, respectively. Data collection and refinement statistics are provided in Table 1. The model quality was validated with Molprobity 67 . All structural figures were generated with the program PyMOL (http://www.pymol. org/) with the same color and labeling schemes.
Protein-binding assays. For GST pull-down assays of IVT IRS1 or IRS2 proteins, beads bound to GST-AP2M1 or GST were incubated with IVT products diluted in the cell lysis buffer at 4°C for 2 h. After incubation and washing, proteins bound to beads were eluted with the SDS loading buffer, resolved with SDS-PAGE, and detected with Coomassie staining or immunoblotting with the appropriate antibodies.
For peptide pull-down assays, 2 mg peptides were coupled to beads with the SulfoLink Immobilization Kit (Thermo Scientific). Beads were resuspended in Buffer A [50 mM Tris, pH 7.5, 0.2 M NaCl, 0.1% NP-40, 10 mM sodium fluoride, 5 mM beta-glycerophosphate, 1.5 mM sodium orthovanadate, 10% glycerol, and 1 mM EDTA] as a 1:1 slurry. Ten microliter beads were incubated with 1 μg purified AP2M1 protein (residues 160-435) on a rotator at 4°C for overnight. The beads were washed five times with Buffer A. The bound proteins were eluted with SDS sample buffer and analyzed by SDS-PAGE and Coomassie Blue staining.
The ITC assays were performed with a MicroCal Omega ITC200 titration calorimeter (GE Life Sciences) at 20°C with minor modifications 68 . Briefly, the recombinant AP2M1 protein (residues 160-435) and peptides were dialyzed into the HEPES buffer [25 mM HEPES, pH 7.5, and 50 mM NaCl]. For each titration, 300 μl of AP2M1 (50 μM) were added to the calorimeter cell. IRS1 peptides (YMPMS, 528.9 μM or YMPMpS, 507.4 μM) were injected with an injection syringe in nineteen 2.0-μl portions. Raw data were processed and fitted with the NITPIC software package 69 .
Phosphatase assays. Active SHP2 (2.9 μM, SignalChem) diluted in the phosphatase dilution buffer [50 mM imidazole, pH 7.2, 0.2% 2-mercaptoethanol, and 65 ng μl −1 BSA] was incubated with IRS1 peptides (2.6 mM) at 37°C for the indicated time points. Two microliters of reaction products were spotted onto 0.45 μm nitrocellulose membrane (BioRad) and dried completely. The membrane was blocked with 5% nonfat milk in Tris buffered saline (TBS) for 1 h, and washed once with TBS-T (0.02% Tween 20). The membrane was incubated with anti-IRS1-pY612 antibodies diluted in TBS-T at 4°C overnight. After washing with TBS-T, the anti-rabbit IgG (H + L) Dylight 800 conjugates (Cell Signaling) were applied as secondary antibodies. The membranes were scanned with the Odyssey Infrared Imaging System (LI-COR, Lincoln, NE) for quantification.
In vivo insulin signaling and IR endocytosis assays. HFD-fed mice were administered vehicle or SHP099 for 5 days. The mice were fasted overnight and administered vehicle or SHP099 once more. At 2 h after the last administration, the mice were injected with or without 1 U insulin (3.19 µM, Eli Lilly) via inferior vena cava. For hepatic knockdown of SHP2, mice fed HFD were injected with AAV-GFP or SHP shRNA. At 17 days after injection, the mice were fasted overnight and injected with or without 1 U insulin (Eli Lilly) via inferior vena cava. The livers were collected at the indicated time points. WAT and skeletal muscles were collected at 2 and 3 min after the indicated time points, respectively. Thirty micrograms of tissue was mixed with the cell lysis buffer [50 mM HEPES (pH 7.4), 150 mM NaCl, 10% (v/v) glycerol, 1% (v/v) Triton X-100, 1 mM EDTA, 100 mM sodium fluoride, 2 mM sodium orthovanadate, 20 mM sodium pyrophosphate, 0.5 mM DTT, and 2 mM PMSF] supplemented with protease inhibitors (Roche), PhosSTOP (Roche), 25 U ml −1 turbo nuclease (Accelagen), and 100 μM cytochalasin B (Sigma), homogenized with Minilys (Bertin Technologies), and then incubated on ice for 1 h. After centrifugation at 20,817 × g at 4°C for 1 h, the lysates were analyzed by quantitative western blotting. For IR endocytosis assays, the livers were fixed in 10% NBF and embedded in paraffin blocks. Sections were deparaffinized, subjected to immunohistochemistry as described in the Tissue histology and immunohistochemistry section. Images of sections were acquired as a series of 0.4 μm stacks with a DeltaVision system (Applied Precision, Issaquah, WA). Images were analyzed as described in the Immunofluorescence section.
Statistical analyses. Prism was used for the generation of all curves and graphs and for statistical analyses. Results are presented as mean ± s.e.m. or mean ± s.d. Two-tailed unpaired t tests were used for pairwise significance analysis. Mann-Whitney test, a non-parametric test, was performed for human sample analysis. Sample sizes were determined on the basis of the maximum number of mice that could be bred in similar ages to maintain wellmatched controls. Power calculations for sample sizes were not performed. We monitored weight and health conditions of mice, and excluded mice from experiments if the animal was unhealthy and the body weight was more than two standard deviations from the mean. Randomization and blinding methods were not used, and data were analyzed after the completion of all data collection in each experiment.

Data availability
The coordinates of the AP2M1-pIRS1 structure have been deposited to the Protein Data Bank (PDB ID: 6BNT). All other data are available from the authors upon request. The source data underlying Figs. 1c, d, f, 2b, c, e, 3c, f, 4d, e, f, g, 5b, c, e, f, 7a-c, e-j, and 8a, b and Supplementary Figs 1a, b, d, e, 2b, c, 3a, b, d, e, 4a-c, e-g, 5, 6c, d, 7a-d, f and 8a are provided as a Source Data file.