Mitotic Regulators and the SHP2-MAPK Pathway Promote Insulin Receptor 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 extracellularly 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.


Introduction
The 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][4] . Activated IR phosphorylates itself and the insulin receptor 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 phospho-tyrosine sites of IRS1/2 and promotes the activation of the MAPK pathway 5,6 . Mutations of IR cause severe inherited insulin resistance syndromes 7 , 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 (PKCs), which suppress insulin signaling at the level of IRS1 and possibly IR 1,8 .
We have recently discovered a critical role of MAD2, BUBR1, and p31 comet in insulin signaling during interphase (Fig. 1a) 28 . Specifically, MAD2 and BUBR1 are required for clathrinmediated endocytosis of IR. MAD2 directly binds to the C-terminal MAD2-interacting 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 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 novel 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
Decreased IR plasma membrane localization in human diabetics. Previous biochemical studies have shown that insulin binding to liver plasma membranes is reduced in obese and diabetic mice and humans 29,30 , suggesting that the level of insulin receptor at the plasma membrane (PM) 5 might be reduced in diabetics. 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-ZO1 (as a PM marker) antibodies on 51 non-diabetic and 19 type 2 diabetes patient samples, and analyzed IR PM levels. IR PM signals in the liver sections from type 2 diabetes patients were significantly weaker than those 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.
We note, however, that the reduction of IR on the cell surface in liver samples of human type 2 diabetes patients can be a consequence of insulin resistance, rather than the cause. It is also possible that the reduced IR localization at the plasma membrane 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 plasma membrane levels, including endocytosis, might be defective in diabetics. CDC20 is required for proper IR endocytosis. We have previously shown that MAD2 and BUBR1 promote IR endocytosis through recruiting AP2, and p31 comet inhibits this process through antagonizing the MAD2-BUBR1 interaction (Fig. 1a) 28 . We further tested whether other wellknown mitotic regulators also have a role in IR endocytosis. Depletion of CDC20, but not BUB1, BUB3, or MAD1, delayed insulin-activated IR endocytosis (Fig. 1d- Fig. 1g and Supplementary Fig. 1a,b), consistent with a previous report 32 . Expression of wildtype (WT) BUBR1, but none of BUBR1 truncation mutants, restored IR endocytosis in cells depleted of BUBR1 (Fig. 1h,i and Supplementary Fig. 1c). BUBR1 interacts with CDC20 through multiple binding motifs 22, [33][34][35] . 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. 1g). The ΔKEN2, ΔPhe, ΔD2 and ΔPhe+ΔD2 restored insulin-activated IR endocytosis in cells depleted of BUBR1 (Fig. 1h,i and Supplementary Fig.   1c). However, BUBR1 ΔKEN1 could not restore IR endocytosis in BUBR1-depleted cells. 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 7 motifs in IR--the NPEY 960 and di-leucine (L 986 L) motifs--have previously been implicated in AP2 binding and IR endocytosis [36][37][38][39] (Fig. 2a). We have shown that the MAD2-interacting motif (MIM, R 1333 ILTL) of IR binds to MAD2, which in turn recruits AP2 with the help of BUBR1 and CDC20.
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. 2b,c). Without insulin treatment, IR WT, 4A, and Y960F localized to the plasma membrane (PM), but IR AA was enriched in intracellular compartment (IC). 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. 2d). Thus, the di-leucine motif is not required for IR endocytosis and actually prevents it through unknown mechanisms. The di-leucine motif is located in a β strand of the N-terminal lobe of the IR kinase domain (Fig. 2e). Mutation of this motif is expected to alter the structural integrity or activity of the IR kinase domain. IR Y960F, 4A, and Y960F/4A mutants were less efficiently internalized after insulin stimulation (Fig. 2f). The IR Y960F/4A double mutant was not significantly more defective than the single mutants (Fig. 2f). As Y960 is phosphorylated in the activated IR 40 , defective endocytosis of IR Y960F suggests that phosphorylation of Y960 (pY960) might be required for timely IR internalization.
Expression of RNAi-resistant IRS1 restored IR endocytosis in cells depleted of both IRS1/2. Thus, 8 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. IRS1 interacts with the adaptor related protein complex 1 Mu 1 subunit (AP1M1) and AP2M1 through multiple YXXΦ (X, any amino acids; Φ, bulky hydrophobic residues) motifs 46,47 .
We tested whether mutations of YXXΦ motifs in IRS1 and IRS2 disrupted AP2M1 binding.
RNAi-resistant IRS1 Y612A and 3YA mutants could not restore IR endocytosis in 293FT or HepG2 cells depleted of IRS1/2 (Fig. 3d,e and Supplementary Fig. 3d,e). 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 during the uncoating of the clathrin coat and prior to the fusion of the endocytic vesicles with the 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. 3f,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 regulation. The tyrosine residues in YXXΦ motifs on IRS1/2 can be phosphorylated by the activated IR and be dephosphorylated by the tyrosine phosphatase SHP2 2,49 . Strikingly, the serine residues immediately following the YXXΦ motifs are well conserved ( Fig. 3c and Supplementary Fig. 2e), and can be phosphorylated by ERK1/2 [50][51][52] . ERK1/2-dependent phosphorylation has been proposed to reduce IRS1 tyrosine phosphorylation through negative feedback [50][51][52] , 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. 4a,b). 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 53,54 . It is formally possible that the failure of IRS1 3YA mutant to restore IR endocytosis in cells depleted of IRS1/2 was an indirect consequence of reduced PI3K activity. To test this possibility, we checked the effect of IRS1 3YA on insulin signaling ( Supplementary Fig. 3e). 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 (GDC-0941) did not affect insulinactivated IR endocytosis (Fig. 4a,b). This result argues against reduced PI3K activity as the underlying reason for the observed defects of insulin-activated IR endocytosis in IRS1/2-depleted cells, and further confirms that the PI3K-AKT pathway is not involved in this process.
Taken together, our results support the following mechanism for insulin-activated IR endocytosis (Fig. 4g). The activated IR phosphorylates the tyrosine residues in YXXΦS motifs and the C-terminal SHP2-docking sites of IRS1/2, and stimulates the PI3K-AKT and MAPK pathways. In a negative feedback mechanism, activated ERK1/2 phosphorylate the serines in YXXΦS motifs on IRS1/2 and assist SHP2 to dephosphorylate IRS1/2. The IRS1/2 YXXΦS motifs with the serine phosphorylated and tyrosine dephosphorylated bind to AP2 with optimal affinities, promoting clathrin-mediated endocytosis of IR.
Structural basis of the phospho-regulation of IR endocytosis. We next determined the crystal structure of AP2M1 (residues 160-435) bound to the serine-phosphorylated YXXΦS motif from Table 1). The overall structure of the AP2M1-pS-IRS1 complex was similar to those of previously determined structures of AP2M1 bound to YXXΦ motifs. AP2M1 contained two interlinked β-sandwich subdomains: subdomain 1 (β1-6 and 17-19) and subdomain 2 (β7-16) (Fig. 4h). The pS-IRS1 peptide binds at the edges of strands β18 and β17 in subdomain 1, and interacts with residues from strands β1, β17, and β18 (Fig. 4h,i). In particular, Y612 and M615 make extensive hydrophobic interactions with AP2M1. The RNAi-resistant IRS1 3YF mutant with tyrosines in the YXXΦS motifs replaced by phenylalanines could not fully restore IR endocytosis in 293FT cells depleted of IRS1/2 (Fig. 3d,e). The hydroxyl group of Y612 forms a hydrogen bond with D176 in β1, providing an explanation for why phenylalanines cannot functionally substitute for tyrosines. Phosphorylation of Y612 is expected to introduce both static hindrance and unfavorable electrostatic interactions with D176, explaining why tyrosine phosphorylation of YXXΦS motifs disrupts the IRS1-AP2 interaction.
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. 4j), suggesting that the phosphoserine might engage in favorable electrostatic interactions with this basic patch. Mutations of H416 and K420 did not, however, reduce IRS1 binding ( Supplementary Fig. 3f,g). 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 knockout (KO) mice show increased insulin sensitivity 55,56 , 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 5 . Wild type mice maintained on high-fat diet (HFD) for 5 weeks were treated with SHP099 (60 mg/kg body weight) by daily oral gavage for 6 days, and then tested 13 for glucose and insulin tolerance. Strikingly, SHP099 administration markedly increased glucose tolerance and insulin sensitivity in HFD-fed mice (Fig. 5a,b). SHP099 did not change the body weight of mice fed HFD (Fig. 5c). Insulin stimulation caused IR endocytosis and reduced the IR staining at the PM in mouse liver sections (Fig. 5d,e). SHP099 delayed insulin-activated IR endocytosis. Finally, SHP099 inhibited the insulin-stimulated IRS1-AP2 interaction in primary hepatocytes (Fig. 3g).
To further confirm the requirement of SHP2 in promoting IR endocytosis in vivo, we introduced adeno-associated viruses 8 (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. 4a). By contrast, the SHP2 protein level in 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 ( Fig. 5a,b). AAV-SHP2 shRNA treatment did not change body weight (Fig. 5c).
Importantly, AAV-mediated SHP2 silencing inhibited insulin-activated IR endocytosis in the liver (Fig. 5d,e). These results confirm a role of SHP2 in promoting IR endocytosis and in metabolic homeostasis in vivo.
Liver is a major site for insulin clearance and defects in this process can cause hyperinsulinemia [57][58][59] . Hepatic insulin clearance is mainly mediated by IR endocytosis, as liverspecific 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 57,59 .
We thus examined the effects of SHP099 on insulin clearance. To estimate insulin clearance, we examined the levels of insulin and C-peptide, a cleavage product of proinsulin, and determined the ratio of the serum levels of C-peptide and insulin. The fasting insulin level in mice fed normal chow was not altered by SHP099 administration, despite the inhibition of IR endocytosis in the liver (Fig. 5f). In HFD-fed mice, SHP099 slightly increased the fasting insulin level. The serum levels of C-peptide were similar to those in vehicle-treated groups in both conditions (Fig. 5g), 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. 5h). 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. 4b, 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 SHP2independent IR endocytosis mechanisms. We have confirmed a requirement for CEACAM1 in IR endocytosis in our cellular assays (Fig. 5i,j). It is possible that CEACAM1 is required for all mechanisms of IR endocytosis whereas SHP2 only regulates the IRS1/2 branch (Fig. 5k). 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 in mice.
We next examined the effect of SHP2 inhibition on insulin signaling in the liver, epididymal white adipose tissue (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), 15 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. 6a,b). Consistent with a previous report 5 , SHP099 inhibited the activation of the MAPK pathway by insulin in the liver. SHP2 inhibition attenuated insulininduced IRS1 pS612 signal, consistent with the fact that ERK1/2 mediate this phosphorylation in vivo. By contrast, insulin-triggered activating phosphorylation of AKT was significantly increased 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.
SHP099 greatly inhibited the activation of the MAPK pathway and the serine phosphorylation on IRS1 (pS612) upon insulin stimulation in skeletal muscle ( Supplementary Fig.   5a,b). However, the IRS1 pY608 level in skeletal muscle from SHP099-treaed mice exhibited only a slight increase, and the levels of activating phosphorylation of AKT were similar to those in vehicle-treated mice. SHP099 did not appreciably alter the insulin-stimulated activation of the PI3K-AKT or MAPK pathways in WAT ( Supplementary Fig. 5c,d). These data suggest that SHP2 inhibition has distinct effects on insulin signaling and possibly IR endocytosis in different tissues.
The underlying reasons for these variations among tissues are unknown at present. Because SHP099 only increases AKT signaling in the liver, but not in skeletal muscle or WAT, our findings suggest that the liver is a major site of action for the insulin-sensitizing effect of the SHP2 inhibitor in vivo. We cannot, however, rule out the contributions of other tissues to the observed phenotypes.

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. 6c). 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 MAD2-interacting motif (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 [36][37][38][39] . In this study, we have shown that the dileucine 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 implicate 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 checkpoint regulators on aneuploidy and spindle checkpoint activities.
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 phosphorylates the YXXΦS motifs in 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. SHP2 inhibition prolongs phosphorylation of YXXΦ motifs on IRS1/2. Because these phospho-tyrosine motifs also interact with PI3K, the enhanced tyrosine phosphorylation of these motifs as a result of SHP2 inhibition may directly promote PI3K activation. This direct effect may contribute to the effectiveness of SHP2 inhibitors in counteracting insulin resistance. Genetic suppression of IR endocytosis in other ways, without the inhibition of SHP2, will better define the role of IR endocytosis during insulin signaling.
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 7 , 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. 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 5 . Although phenotypes of SHP2 KO mice suggest potential adverse effects of SHP2 inhibition 62-66 , SHP099 administration into mice did not show significant toxic effects in that previous study 5 and in the present study. Our study shows that 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 67 . 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. Glucose and insulin tolerance tests, and metabolic analysis were performed as described previously 28 . For in vivo pharmacological assays, 6-8-week-old male mice were fed high-fat diet (HFD) for 5 weeks. Two days before drug administration, mice were switched to normal chow. SHP099 (MedChem Express) was dissolved in DMSO and diluted into a 0.5% hypromellose and 0.1% Tween-80 solution. 60 mg/kg of SHP099 was administered by daily oral gavage for 6 days.  293FT or HepG2 cells expressing IR-GFP WT, or mutants were generated as described previously 28 . Briefly, cDNAs encoding IR mutants were cloned into the pBabe-GFP-puro vector.

Mice. Animal work described in this manuscript has been approved and conducted under
The vectors were co-transfected with viral packaging vectors into 293FT cells, and the viral supernatants were collected at 2 days and 3 days after transfection. The concentrated viruses were added to 293FT and HepG2 cells with 4 µg/ml of polybrene. Cells were selected with puromycin

Declaration of interests
The authors declare no competing interests.    The mice were administered vehicle or SHP099 for 6 days. i HepG2 cells stably expressing IR-GFP were transfected with CEACAM1 siRNAs, serum starved, treated without or with 100 nM insulin form 5 min, and stained with anti-GFP and DAPI. Quantification of the ratios of PM and IC IR-GFP signals of cells was shown (mean ± SD). j Western blot analysis of cell lysates in (i).

Figure Legends
k Model of the regulation of insulin-activated IR endocytosis by CEACAM1, the MAD2-CDC20-BUBR1 module, and the SHP2-IRS1/2 module. shown as sticks. The potential acceptor residues for IRS1 pS616 are labeled. g Binding of the pS-IRS1 peptide to WT and mutants of AP2M1 (residues 160-435). Input and proteins bound to pS-IRS1 peptides were analyzed by SDS-PAGE and stained with Coomassie (CBB). h Quantification of the relative band intensities in (g). Mean ± SD; n=3 independent experiments.