The insulin receptor adaptor IRS2 is an APC/C substrate that promotes cell cycle protein expression and a robust spindle assembly checkpoint

Insulin receptor substrate 2 (IRS2) is an essential adaptor that mediates signaling downstream of the insulin receptor and other receptor tyrosine kinases. Transduction through IRS2-dependent pathways is important for coordinating metabolic homeostasis, and dysregulation of IRS2 causes systemic insulin signaling defects. Despite the importance of maintaining proper IRS2 abundance, little is known about what factors mediate its protein stability. We conducted an unbiased proteomic screen to uncover novel substrates of the Anaphase Promoting Complex/Cyclosome (APC/C), a ubiquitin ligase that controls the abundance of key cell cycle regulators. Surprisingly, we found that IRS2 levels are regulated by APC/C activity and that IRS2 is a direct APC/C target in G1. Consistent with the APC/C’s role in degrading cell cycle regulators, we find that IRS2-null cells are deficient in proteins involved in cell cycle progression and display spindle assembly checkpoint defects during M-phase. Together, these findings reveal a new pathway for IRS2 turnover and indicate that IRS2 is a critical component of the cell cycle control system in addition to acting as an essential metabolic regulator.


Introduction 32
The insulin and insulin-like growth factor 1 receptors (IR/IGF1R) are receptor 33 tyrosine kinases that control metabolism, differentiation, and growth. Upon ligand binding 34 at the cell surface, the activated IR/IGF1R undergoes a conformational change that allows 35 it to auto-phosphorylate tyrosine residues on its cytoplasmic subunits (Haeusler et al., 36 2017). This facilitates the recruitment and phosphorylation of insulin receptor substrate 37 (IRS) proteins, which serve as scaffolds to initiate downstream signaling (Copps and 38 White, 2012). Two major pathways that are stimulated by this cascade are the PI3K-AKT 39 and Ras-Raf-MAPK pathways, which coordinate metabolic homeostasis and growth, 40 among other functions (Haeusler et al., 2017). 41 The most physiologically important and ubiquitously expressed IRS proteins are 42 IRS1 and IRS2. Though IRS1 and IRS2 share similar structural and functional features, 43 they have complementary roles and expression patterns that depend on tissue type and 44 physiological state (Haeusler et al., 2017). These differences are illustrated by divergent 45 phenotypes in knockout mice: whereas IRS1 knockout mice exhibit insulin resistance that 46 is compensated by increased pancreatic β cell mass, IRS2 knockout mice exhibit β cell 47 failure and resultant diabetes (Lavin et al., 2016). Distinct roles for IRS1 and IRS2 can 48 also be observed within the same tissue. For example, in skeletal muscle, IRS1 is 49 required for glucose uptake and metabolism, whereas IRS2 is important for lipid uptake 50 and metabolism (Bouzakri et al., 2006;Long et al., 2011). Furthermore, recent work has 51 shown that the ratio of IRS1 to IRS2 is important for hepatic glucose metabolism (Besse-52 inhibition. As an internal control, we detected a significant increase (p = 3.2 x 10 -5 ) in the 124 abundance of peptides derived from the N-terminal 110 amino acids of geminin (GMNN). 125 These residues are shared with the mAG1-geminin (1-110) reporter expressed in this cell 126 line, confirming earlier fluorescence-based validation of our experimental system. 127 While the majority of the previously reported APC/C Cdh1 substrates that were 128 quantified in our G1 proteomic experiment were stabilized following APC/C inhibition, 129 some remained constant. There are several possible explanations for this result. First, for 130 proteins that were identified based on a small number of peptides, inadequate 131 quantification may have resulted in inaccurate abundance assignments. Second, some 132 substrates may be APC/C Cdh1 -accessible only under conditions or in tissue types that 133 were not well modeled by the experimental parameters that we used. Third, some 134 proteins (e.g. FBXW5, ZC3HC1) (Klitzing et al., 2011;Puklowski et al., 2011) were 135 proposed to be APC/C Cdh1 substrates based on results obtained in Cdh1 overexpression 136 systems, indicating that APC/C Cdh1 activity may be sufficient but not necessary to control 137 their levels. 138 Of the 38 previously reported APC/C substrates that we identified, the median fold 139 change under APC/C inhibition compared to DMSO was 1.147. Based on this, to identify 140 new APC/C substrates, we screened for proteins that: (1) had a fold change ≥1.147 under 141 APC/C inhibition, (2) were identified and quantified based on >1 peptide, and (3) had a p-142 value < 0.05 across the three biological replicates measured in this experiment. This 143 narrowed our analysis to a subset of 204 proteins (Supplementary Table S3). Because 144 physical restriction that APC/C activity occurs within the cell. Based on these thresholds, 155 our analysis identified 26 proteins as potential D-and KEN-box containing APC/C Cdh1 156 substrates (Table 1, Figure 1D). Of these 26 proteins, 11 have previously been reported 157 as direct APC/C substrates, validating internally that this analysis was useful for 158 identifying APC/C substrates. 159 160 IRS2 levels are controlled by Cdh1 in a proteasome-dependent manner 161 Examining our 26-protein putative substrate list, we focused our attention on 162 IRS2-one of two major adaptors that promotes signaling through the insulin and insulin-163 like growth factor 1 receptors (IR/IGF1R). Using conditions identical to those under which 164 the proteomics experiment was conducted, we validated that IRS2 was upregulated at 165 the protein level under APC/C inhibition in G1-arrested RPE1 cells by immunoblot ( Figure  166 2A). Seeking to further validate this result in a distinct physiological context, we asked 167 whether APC/C inhibition in terminally differentiated C2C12 myotubes also increases IRS2 168 protein abundance. C2C12 myoblasts easily differentiate into multinucleated myotubes 169 following serum withdrawal and supplementation with growth factors (Figures S2A-S2B). 170 To validate that the APC/C is active in this system, we transfected C2C12 myoblasts with 171 a model APC/C substrate (N-terminal fragment of cyclin B1 fused to EGFP; NT-CycB-172 GFP), allowed cells to differentiate into myotubes, and found that APC/C inhibition 173 stabilized NT-CycB-GFP ( Figure S2C). Similarly, we found that acute APC/C inhibition in 174 myotubes also resulted in an accumulation of IRS2 protein (Figure 2B), thereby validating 175 this finding from our G1 experiment in RPE1 cells in an independent system. 176 To exclude the possibility that the change in IRS2 abundance that we observed 177 following APC/C inhibition was due to off-target effects of the small molecule APC/C 178 inhibitors, we depleted Cdh1 using RNAi to block APC/C Cdh1 activity in RPE1, C2C12, and 179 HeLa cells. In all three cell lines, we found that Cdh1 knockdown caused an accumulation 180 of endogenous IRS2 compared to control-transfected cells (Figure 2C-2E). 181 We next sought to confirm that the increase in IRS2 protein observed under APC/C 182 inhibition was due to impaired targeting of IRS2 to the proteasome. To test this, we 183 arrested RPE1 cells in G1 using palbociclib and acutely treated them with APC/C inhibitors 184 and/or a proteasome inhibitor (MG132) for 8 hours. This experiment revealed that APC/C

Cdh1 control of IRS2 degradation depends on an IRS2 D-box motif 217
Based on its SLiMSearch prediction, IRS2 contains four minimal D-box motifs 218 (RxxL), one extended D-box motif (RxxLxxxxN) and no KEN-box motifs. Of the four 219 minimal D-box motifs, none bear strong consensus similarity to previously validated D-220 box motifs, and one exists in a highly structured region of the protein (Krystkowiak and 221 Davey, 2017). Because of its high SLiMSearch parameter scores (Table 1), we focused 222 our efforts on determining whether the extended D-box motif located in the C-terminal 223 third of IRS2 is required for its APC/C Cdh1 dependent stability. IRS2's extended D-box 224 (amino acids 972-980 in human IRS2) is highly conserved in placental mammals despite 225 overall divergence in much of the C-terminus (Figure 4A), suggesting that this sequence 226 likely has a conserved function. 227 To test whether IRS2's full D-box is relevant for its Cdh1-dependent degradation, 228 we generated a mutant IRS2 construct bearing an R972A mutation (∆D), which was 229 expected to abrogate its function as a D-box (Glotzer et al., 1991). Using RPE1 cells 230 stably expressing C-terminally HA-tagged IRS2-WT or IRS2-∆D, we found that APC/C 231 inhibition following G1 arrest caused accumulation of IRS2-WT but not IRS2-∆D ( Figure  232 4B). The degree of accumulation of the WT protein depended on the dose of APC/C 233 inhibitors used ( Figure S4A). We were moreover able to repeat this result in terminally 234 differentiated C2C12 myotubes that stably expressed doxycycline-inducible, C-terminally 235 HA-tagged IRS2-WT or IRS2-∆D constructs that were treated with APC/C inhibitors 236 ( Figure 4C). 237 To further validate the Cdh1-dependence of IRS2's D-box motif, we asked whether 238 Cdh1 knockdown by siRNA could stabilize the IRS2-∆D protein. Using asynchronous 239 RPE1 cells stably expressing C-terminally HA-tagged IRS2-WT and IRS2-∆D, we found 240 that Cdh1 knockdown by siRNA caused an accumulation of IRS2-WT relative to control-241 transfected cells but not IRS2-∆D ( Figure 4D). This result was repeated in HeLa cells 242 stably expressing N-terminally FLAG-HA-tagged IRS2-WT and IRS2-∆D constructs 243 subject to the same conditions ( Figure 4E). 244 The stable cell lines described above express tagged IRS2 variants at low levels protein for the IGF1R and IR) shares 75% sequence homology with IRS2's N-terminus 248 and 35% homology with its C-terminus (Sun et al., 1995) but does not share the D-box 249 motif found in IRS2's C-terminus ( Figure 4F). In keeping with our hypothesis that Cdh1-250 mediated control of IRS2 is D-box dependent, IRS1 levels did not increase in G1-arrested 251 RPE1 cells treated with APC/C inhibitors as measured by either mass spectrometry 252 ( Figure 4G) or immunoblot ( Figure 4H). Furthermore, while it did display a change in 253 electrophoretic mobility compatible with mitotic phosphorylation, unlike IRS2, it did not 254 decrease in abundance at mitotic exit in RPE1 cells ( Figure 4I). Taken together, the 255 findings described above indicate that APC/C Cdh1 controls IRS2 levels in manner that is 256 dependent upon its C-terminal D-box motif. 257

IRS2 is required for normal expression of many proteins involved in mitosis 259
Many reported APC/C Cdh1 substrates (including several of those identified in our 260 initial proteomics screen) are required for normal cell cycle progression. Because 261 IR/IGF1R transduction promotes a variety of transcriptional programs (Copps and White, 262 2012), we hypothesized that IRS2 might promote the expression of proteins involved in 263 cell cycle control. To investigate this, we generated two IRS2 knockout RPE1 cell lines 264 using CRISPR/Cas9 (Figure 5A), henceforth referred to as ∆IRS2-A and ∆IRS2-B. Using 265 these cells, we again employed TMT-coupled quantitative proteomics. The proteomes of 266 wild-type, ∆IRS2-A, and ∆IRS2-B cell lines were analyzed in biological triplicate, and 267 relative abundances were ascertained based on TMT reporter ion signal-to-noise values. 268 Hierarchical clustering indicated that the proteomes of the two knockout cell lines 269 analyzed were more similar to each other than either knockout cell line was to wild-type 270 ( Figure S5A), indicating that deletion of IRS2 produced similar effects in both cell lines. 271 In order to exclude aberrancies that may have accrued during the CRISPR process or as 272 a result of clonal expansion, we focused the scope of our analysis to proteins that 273 changed significantly (p < 0.05) by more than 20% in both IRS2 knockout clones relative 274 to the WT cell line (Figures 5B-5C). We found 239 proteins that decreased by >20% in 275 both IRS2 knockout lines relative to the wild type line and 300 proteins that increased by We conducted gene enrichment analysis of the proteins that increased ( Figure  278 S5C) or decreased ( Figure 5D) by >20% in both knockout cell lines relative to wild type 279 cells. Of the 239 proteins that were depleted by >20% in both knockout cell lines, we 280 found a statistical over-representation of proteins participating in metabolic processes 281 characteristic of IR signal transduction. Notably, we also found an over-representation of 282 proteins involved in mitotic cell cycle regulation in this subset ( Figure 5D). This suite of 283 proteins included regulators of mitotic entry and exit as well as several factors involved in 284 spindle assembly ( Figure 5E). Consistent with the fact that strong depletion of most 285 critical cell cycle regulators renders cells inviable, most of the observed changes in cell 286 cycle-related genes were relatively modest ( Figure S5D). Based on these data, we 287 conclude that IRS2 is important for promoting the expression of a suite of proteins 288 involved in orchestrating the mitotic cell cycle, and deletion of IRS2 stunts their expression 289 in RPE1 cells. 290

IRS2 expression promotes a functional spindle assembly checkpoint 292
Because many of the factors that were depleted in IRS2 knockout cell lines are 293 known to be involved in regulating spindle assembly and mitotic exit, we asked whether 294 IRS2 knockout cell lines display phenotypic differences from wild-type cells in terms of 295 spindle assembly checkpoint function. Using a high content nuclear imaging assay to 296 measure mitotic fraction based on DAPI staining intensity (Sackton et al., 2014), we asked 297 whether IRS2 knockout cell lines display mitotic arrest differences compared to wild-type 298 cells when treated with spindle poisons. Wild-type cells treated with nocodazole (a 299 microtubule destabilizing agent) arrested in mitosis in a dose-dependent manner, 300 whereas both IRS2 knockout cell lines displayed depressed mitotic arrest ( Figure 6A, 301 Figure S6A). This was also true to a lesser extent in the presence of S-trityl-L-cysteine 302 (STLC), an Eg5 inhibitor ( Figure 6A). 303 Using time lapse video microscopy, we found that untreated IRS2 knockout cell 304 lines had no significant alteration in mitotic duration compared to wild-type cells (Figure an APC/C Cdc20 substrate that is stabilized by the spindle assembly checkpoint (Clijsters 309 et al., 2013)-display an accumulation of mAG1 fluorescence early in mitotic arrest, 310 consistent with checkpoint-mediated blockade of APC/C Cdc20 . In contrast, both IRS2 311 knockout cell lines display depressed mAG1 accumulation, consistent with higher 312 APC/C Cdc20 activity due to a weakened checkpoint ( Figure 6C). This phenotype, along 313 with the shorter mitotic duration and lower mitotic fraction in the presence of spindle 314 poisons, is consistent with cells bearing a defective mitotic spindle assembly checkpoint. 315 Based on these data, we conclude that IRS2 promotes a functional spindle assembly 316 checkpoint in RPE1 cells. Based on the results of an unbiased proteomic screen, we provide evidence that 320 IRS2, a critical mediator of IR/IGF1R signaling, is a direct APC/C Cdh1 substrate. We 321 demonstrate that IRS2 is stabilized by APC/C inhibition and Cdh1 knockdown in multiple 322 cell types and that this depends on IRS2's C-terminal D-box motif. In contrast, we find 323 that IRS1, a closely related IRS2 paralog that lacks a D-box, is not subject to regulation 324 by the APC/C. Taken together, these results show that APC/C activity directly controls 325 IRS2 levels in a D-box dependent manner. 326 We identified a high-mobility form of IRS2 that accumulates under APC/C 327 inhibition, likely corresponding to a difference in phosphorylation given that IRS2 has 328 ~150 annotated threonine, serine, and tyrosine phosphorylation sites (Hornbeck et al., 329 2015). This suggests that IRS2's APC/C-dependent stability could be regulated by 330 phosphorylation, possibly at sites near or within the D-box, which is an intriguing topic for 331 future study. Consistent with this, IRS2 phosphorylation is known to impact its stability in 332 other contexts, including following prolonged exposure to insulin or following mTOR Based on the data presented here, we propose a model (Figure 7) in which IRS2's 372 APC/C-mediated degradation in G1 serves to limit IRS2-dependent signaling during G1. 373 Upon APC/C inactivation at the G1/S boundary, IRS2 is able to accumulate and stimulate 374 signaling required for normal progression through the latter stages of the cell cycle,    separated by SDS-PAGE using either 4-12% Bis Tris gels or 3-8% Tris acetate gels 617 (Thermo Fisher Scientific). All IRS2 immunoblots were separated on 3-8% Tris acetate 618 gels with the exception of those shown in Figures 5A and S4B, which were separated on 619 4-12% Bis Tris gels. Proteins were transferred to polyvinylidene difluoride (PVDF) Membranes were blocked in 5% non-fat dry milk in Tris-buffered saline with 0.1% Tween 623 (TBS-T) before incubating with primary antibodies overnight at 4ºC with agitation. 624 Membranes were probed with secondary antibodies dissolved in 5% milk in TBS-T for 1-625 2 hours at room temperature before developing with an Amersham 600RGB imaging 626 system. Quantification of immunoblots was done using ImageJ (Schneider et al., 2012). 627 628 Antibodies 629 The following commercially available primary antibodies were used for immunoblotting: with summed signal-to-noise less than 100 were excluded from final result. Lastly, each 795 protein was scaled such that the summed signal-to-noise for that protein across all 796 channels equals 100, thereby generating a relative abundance (RA) measurement. 797 Triton X-100 in DPBS. Plates were sealed with aluminum tape (Nunc, 276014) and were 808 incubated for 45 minutes room temperature in the dark before imaging. All experimental 809 conditions were represented in triplicate on the same plate. Plates were imaged using an 810 ImageXpress Micro high-content microscope (Molecular Devices) equipped with a 10x 811 objective lens. Four images were acquired per well, yielding a total of 12 images per 812 conditions. Images were processed automatically in ImageJ to identify and count nuclei 813 as well as measure their maximum fluorescence intensity. ImageJ output files were 814

Materials Availability 832
All mass spectrometry raw files will be available through the PRIDE archive upon 833 publication. All other data are available in the associated supplementary data files. 834 Further information and requests for resources and reagents should be directed to the