Directed differentiation of patient-derived inducible hepatocytes. To identify cell-intrinsic determinants of insulin resistance in the liver, iPS cells were derived from a cohort of 16 individuals that included 8 patients with T2D and 8 age-matched controls, with equal numbers of males and females. As previously described, the iPS cells were created using Sendai virus reprogramming (Figure 1A). Compared with the control donors, the T2D donor group had a higher average BMI (29.7 ± 0.9 versus 26.8 ± 0.6 kg/m2), serum glucose levels (7.5 ± 0.4 versus 5.3 ± 0.1 mM), and hemoglobin A1c (5.7 ± 0.4% versus 4.5 ± 0.1%) (24). The iPS cells were differentiated into hepatocytes (iHeps) over 21 days using the 4-stage growth factor protocol described in Methods (Figure 1A) (27). The differentiated iHeps from both the controls and T2D donors exhibited an almost complete loss of the pluripotency markers OCT4 and NANOG accompanied by simultaneous 103-fold to 105-fold increases in the hepatocyte marker proteins asialoglycoprotein receptor 1 (ASGR1), albumin (ALB), transthyretin (TTR), apolipoprotein A2 (APOA2), alpha-fetoprotein (AFP), and cytochrome P450 and C2C9 (CYP2C9) as determined by quantitative reverse transcription PCR (qRT-PCR) (Figure 1, B and C, and Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI183513DS1). Consistent with normal differentiation, at days 8 and 13, there were also transient increases in expression of the hepatocyte stage–specific transcription factors GATA4 and HNF4, which mark hepatic endoderm and immature hepatocytes (Supplemental Figure 1B). Expression of albumin at the protein level was confirmed by immunoblotting and showed a greater than 100-fold increase in iHeps compared with iPS cells, with over 70% of cells strongly positive for albumin, as determined by flow cytometry. This was also confirmed by immunofluorescence (Supplemental Figure 1, C–E).

Directed differentiation of iHeps. (A) Schematic of iPS cell differentiation into iHeps using a 4-stage growth factor protocol. The cells were obtained from 8 control individuals and 8 patients with T2D. In the B–E, the cells from male donors are represented by squares and cells from female donors by circles. (B and C) Gene expression levels of OCT4, NANOG, ALB, and ASGR1 were determined using RT-qPCR and are plotted on a log10 scale. *P < 0.05, **P < 0.01, and ***P < 0.001, for day 1 versus day 21, by paired, 2-tailed t test. (D) Relative gene expression levels of PCK1 and FASN were determined by RT-qPCR. n = 6–8. *P < 0.05 and **P < 0.01, for basal versus insulin, by paired, 2-tailed t test and unpaired, 2-tailed t test for control plus insulin versus T2D plus insulin. (E) Representative immunoblot analysis of phosphorylation of proteins in cells from 4 control and 4 T2D male donors. Quantification of the data in the bar graphs was normalized to total protein on the right. All data indicate the mean ± SEM. CTL, control; BAS, basal; INS, insulin.

T2D iHeps show selective insulin resistance and cell-intrinsic insulin signaling defects. In addition to gene expression features of normal hepatocytes, control iHeps showed robust gene expression responses to insulin stimulation. Thus, following 3 hours of stimulation with insulin, we detected a 75% decrease in mRNA levels for the key gluconeogenic enzyme phosphoenolpyruvate carboxykinase (PCK1) and a 5-fold increase in the levels of mRNA for fatty acid synthase (FASN), a key lipogenic enzyme (Figure 1D). Interestingly, in T2D iHeps, insulin-mediated suppression of PCK1 was lost, while both basal and insulin-stimulated expression of FASN was increased almost 2-fold compared with the controls (Figure 1D). Thus, T2D iHeps showed evidence of selective insulin resistance in gene expression, such that insulin did not suppress gluconeogenesis normally but had an exaggerated response on de novo lipogenesis.

To identify potential signaling defects in T2D, fully differentiated iHeps from patients with T2D and from controls were stimulated with 100 nM insulin for 10 minutes. Since glucocorticoids can induce insulin resistance, dexamethasone and hydrocortisone were removed from the growth media 48 hours before insulin stimulation. Western blot analysis showed a classic insulin response in control iHeps with an approximately 90-fold increase in phosphorylation of the IR at tyrosine 1150 (pIRY1150), and this was reduced by approximately 50% in the T2D cells (P < 0.05) (Figure 1E). Similarly, AKTT308 showed a greater than 9-fold stimulation by insulin in control cells, which was reduced in T2D iHeps by approximately 30% (P < 0.05). T2D iHeps also showed lower insulin-stimulated phosphorylation of GSK3αS21/GSK3βS9 and FOXO1T24/FOXO3aT32. These changes in phosphorylation occurred with no change in the total protein levels of any of these signaling proteins (Figure 1E). Thus, control iHeps showed robust insulin responses in early insulin signaling, and these events were significantly decreased in cells from patients with T2D in vitro. This occurred in the absence of extrinsic factors that might contribute to insulin resistance, i.e., the changes were cell intrinsic.

Defining the insulin-regulated phosphoproteome of iHeps. To more fully define the spectrum of signaling changes, we performed global phosphoproteomics analysis of control and T2D iHeps with and without insulin stimulation (100 nM, 10 minutes) using LC-MS/MS. This identified 46,422 phosphosites, of which 21,863 were considered class I, i.e., had a localization probability of 75% or higher (in these samples, the average was 97%). The fold change of each phosphosite between basal and insulin-stimulated conditions (insulin/basal) in control iHeps is shown as a volcano plot in Figure 2A and by hierarchical clustering (Supplemental Figure 2A). This revealed 292 phosphosites on 226 proteins that increased phosphorylation in response to insulin and 114 phosphosites on 98 proteins that decreased phosphorylation by insulin stimulation (Supplemental Figure 2A and Supplemental Table 1).

Insulin-regulated phosphosites in control and T2D iHeps. (A) Volcano plot showing the phosphopeptides increased or decreased in phosphorylation upon insulin stimulation in control iHeps. (B) Hierarchical clustering of phosphosites identified in control and T2D iHeps shows increased and decreased phosphorylation following stimulation with 100 nM insulin for 10 minutes. Rows represent z scores of log2-transformed intensity of phosphosites for each sample labeled in each column. Classes 1A and 2A show equal increases in both control and T2D cells; classes 1B and 2B show impaired signaling in T2D cells; classes 1C and 2C show emergent signaling in T2D iHeps. (C and E) Enrichment analysis of overrepresented Reactome pathways analysis of proteins that showed increased or decreased phosphorylation in both control and T2D cells. (D and F) Quantification of selected phosphosites representing some of the significantly enriched pathways is shown with data from males and females combined. In the panels, the cells from male donors are represented by squares, and female donors are represented by circles, with colors indicating individual donors. n = 8 control donors; n = 6 T2D donors. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by paired, 2-tailed t test analysis of raw intensities between groups for basal versus insulin and unpaired and 2-tailed t test between control versus T2D with or without insulin. Data are presented as the mean ± SEM.

Reactome pathway analysis of proteins that increased phosphorylation following insulin stimulation showed enrichment of phosphoproteins related to classical insulin signaling pathways (growth factor signal transduction, IRS-mediated signaling, and mTOR-mediated signaling), as well as programmed cell death, cell junction organization, and RNA metabolism (Supplemental Figure 2B). These phosphosites included many previously described in insulin signaling, including IRS2S639, AKT2S474, GSK3αS21, GSK3βS9, ACLYS455, pP70S6KS447, EIF4BS422, MAPK1Y187, and MAPK3Y204 (Supplemental Figure 2C and Supplemental Table 1), along with some less-studied phosphorylation events on classical signaling proteins, including IRS2T350,S346,T520, FOXO1S319,S329, FOXO3S59,T307,S315, and TSC2S492,S534 (Supplemental Table 1). This insulin-stimulated cluster was also enriched in proteins involved in the Rho GTPase cycle, a pathway well known to be insulin stimulated in muscle and adipose tissue (28), but less studied in hepatocytes. This included insulin-stimulated phosphorylation of multiple Rho guanine nucleotide exchange factors (GEFs), such as ARHGEF12S1288, which catalyze the exchange of GDP to GTP, as well as phosphorylation on several Rho GTPase-activating proteins (GAPs), such as RALGAPA2T532, which act to enhance GTP hydrolysis to GDP, thus maintaining the GTPases in an inactive state (28) (Supplemental Figure 2D and Supplemental Table 1). Among the insulin-regulated phosphorylations in control iHeps, those with the highest fold stimulation were PFKFB2S466,S483, SENP2S40, PDE3BS495, and AFDNS1806 (Supplemental Table 1). Previous studies have shown that AKT mediates phosphorylation of PFKFB2S466,S483 and plays a vital role in insulin’s effect to increase glycolysis (29, 30). Insulin-stimulated phosphorylation of actin/Rap1-binding protein afadin (AFDNS1806) has also been observed in iPS-derived myoblasts (24). Although the functional effect of this phosphorylation is unknown, AKT-mediated phosphorylation of AFDNS1718 is thought to play a role in nuclear localization and negative feedback on insulin action by recruitment of histone deacetylase 6 (HDAC6) (31, 32).

Many proteins also showed decreased phosphorylation in response to insulin, as exemplified by acetyl-CoA carboxylase α (ACACAS25), spectrin α, nonerythrocytic 1 (SPTAN1S1029), tumor protein P53 (TP53S155), and poly(ADP-ribose) polymerase 1 (PARP1T88) (Supplemental Figure 2, E and F, and Supplemental Table 1). Reactome pathway analysis of these proteins showed enrichment of Rho GTPase cycle effectors (AKAP12S347), as well as other proteins involved in vesicle transport (EPN1S419,S489), membrane trafficking (RABEP1S410, TBC1D4S588), SUMOylation (PARP1T88, TP53S155), and RNA modification (RPL34S12), including core components of the spliceosome and several serine/arginine–rich (SR-rich) proteins linked to mRNA splicing and miRNA metabolism (SF3B2Y379, SRRM1S211,T411,S409, and SRRM2S919,T1453,T316) (Supplemental Table 1). Multiple sites on NHERF1S269,S280,S290,T293 were also dephosphorylated upon insulin stimulation. NHERF1 (also known as SLC9A3R1) is a sodium/hydrogen exchange regulatory cofactor that can interact with phosphatase and tensin (PTEN) and has been suggested to function as a negative regulator in the InsR/PI3K/AKT1 pathway (33–35). Thus, there was a strong enrichment of classical insulin signaling proteins and many novel proteins whose phosphorylation was either positively or negatively regulated by insulin in control iHeps (Supplemental Table 1).

Multiple layers of dysregulated insulin signaling in T2D. To identify the nature of the cell-intrinsic insulin signaling defects in the liver in T2D, we compared the phosphoproteome of control and T2D iHeps with and without insulin stimulation. Principal component analysis (PCA) indicated a clear separation of samples based on insulin stimulation, the presence or absence of T2D, and donor sex (Supplemental Figure 3A). Considering only sex-common sites that exhibited an increase or decrease in phosphorylation upon insulin stimulation in control or T2D iHeps with a P value of less than 0.05 and a ±1.5-fold change, 1,007 phosphosites exhibited increases in phosphorylation and 703 sites exhibited decreases in phosphorylation upon insulin stimulation in either the control iHeps, T2D iHeps, or both (Supplemental Figure 2B and Supplemental Table 2). We termed the upregulated phosphosites class 1 and the downregulated phosphosites class 2. Both the insulin-upregulated and insulin-downregulated classes could then be subdivided into sites for which stimulation was similar in control and T2D cells (classes 1A and 2A, representing 362 and 169 sites, respectively). These were (a) sites for which the effect of stimulation was significantly (P < 0.05) reduced in T2D cells compared with control cells, i.e., exhibiting “impaired” insulin signaling (classes 1B and 2B), representing 196 and 128 sites, respectively, and (b) sites for which insulin stimulation was significantly (P < 0.05) greater in iHeps from patients with T2D than in those from control individuals, which we termed “emergent” signaling (classes 1C and 2C), representing 429 and 426 sites, respectively (Figure 2B and Supplemental Table 2), of which 249 and 302 met the criteria of both showing ±1.5-fold stimulation and a P value of less than 0.05 in the insulin-stimulated state.

Focusing on class 1A phosphosites, Reactome pathway analysis showed enrichment of proteins involved in signal transduction by growth factors, mTOR signaling, IRS-mediated signaling, Rho GTPase cycle, and metabolism of RNA that were not significantly altered by T2D, despite the presence of insulin resistance. These included insulin stimulation of PRKAA1S496, TSC2S492, and RPS6S240 (Figure 2, C and D, and Supplemental Table 2). On the other hand, class 2A phosphosites, i.e., those decreased by insulin to a similar extent in both control and T2D cells, were enriched in proteins involved in the Rho GTPase cycle, mRNA splicing pathway, SUMOylation, and the apoptosis pathway. The latter included phosphorylation of DIS3Y735, TP53S155, and PLECS4396 (Figure 2, E and F). DIS3 is a 3′ to 5′ exoribonuclease and catalytic subunit of the exosome, a protein complex involved in the degradation and processing of both nuclear and cytoplasmic RNA species, and PLEC encodes plectin, a cytoskeletal protein that maintains tissue integrity.

Impaired insulin signaling in T2D iHeps. More important in the context of insulin resistance are the phosphosites altered by T2D. Class 1B phosphosites showed reduced phosphorylation following insulin stimulation in iHeps from patients with T2D compared with those from control individuals, indicating “impaired” insulin signaling (Figure 3, A and B, and Supplemental Table 3). This group included many important components of the IR/IGF1R and Rho-GTPase pathways (Figure 3, C and E). Among these, IRS1S527, a site ascribed to S6 kinase phosphorylation, showed impaired signaling in T2D cells. Interestingly, this may be a liver-specific alteration, since this site shows no alteration in phosphorylation in skeletal muscle or iPS cell–derived myoblasts from patients with T2D as compared with controls (24, 36). In addition, ACOX1S26 and NDRG2S332 show impaired signaling in T2D. Acyl-CoA oxidase 1 (ACOX1) is involved in FA metabolism by peroxisomes, and phosphorylation at ACOX1S26 protects from polyubiquitination and proteasomal degradation, thereby increasing activity. NDRG2S332 phosphorylation, on the other hand, can be mediated by AKT and novel protein kinase Cs (PKCs) and is also decreased in lipid-induced insulin resistance in C2C12 cells (37). This is thought to play a role in the inactivation of PTEN (38). MINDYS441 (also known as FAM63A) also shows reduced insulin-stimulated phosphorylation in T2D. MINDY is a Lys-48 deubiquitinase, and intronic and missense variants of MINDY1 have been associated with T2D, suggesting that normal MINDY function is protective of T2D (39). In the present study, the signal transduction proteins EIF4B, RPS6, RPS6KB1, and TSC2 also showed impaired phosphorylation in T2D, as did ERBIN, MARK3, and several proteins involved in Rho GTPase signaling, including the Rho GAPs ARHGAP32, Rho GEFs (ARHGEF7, ARHGEF11), and the Rho effector FARP1 (Supplemental Figure 3, B and C, and Supplemental Table 3).

Impaired insulin signaling in T2D iHeps. (A) Hierarchical clustering showing impaired insulin-regulated phosphosites in control and T2D iHeps. Rows represent normalized log intensity z scores for intensities of the phosphosites for each sample. (B) Schematic showing the number of phosphosites with impaired signaling in T2D iHeps, i.e., phosphorylation events that were lost or had significantly reduced regulation in T2D compared with control cells. Data were separated into phosphosites normally increased by insulin stimulation (left) and sites normally decreased by insulin (right), using data shown in the heatmap in Figure 2B. (C and D) Reactome pathway enrichment of “impaired sites” in T2D iHeps. (E and F) Quantification of exemplary impaired phosphosites normally increased phosphorylation in E or decreased phosphorylation in F. In the panels, the cells from male donors are represented by squares, and those from female donors are represented by circles, with colors indicating individual donors. n = 8 controls; n = 6 T2D. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by paired, 2-tailed t test analysis of raw intensities between groups for basal versus insulin, and unpaired, 2-tailed t test between control versus T2D with or without insulin. Data are presented as the mean ± SEM.

Class 2B phosphosites, on the other hand, represent sites whose phosphorylation was decreased by insulin stimulation in controls, and this response was impaired in T2D iHeps. This subclass was enriched in proteins not classically associated with insulin signaling but involved in the Notch/HLH transcription pathway, antiviral mechanisms of IFN-stimulated genes, RNA metabolism, and the Rho-GTPase cycle (Figure 3, D and F). For example, insulin stimulation reduced the phosphorylation of NCOR1S1980 in control iHeps, and this effect was lost in T2D (Supplemental Figure 3, D and E). Although the role of this phosphorylation is unknown, phosphorylation of NCOR1S1460 decreases its ability to interact with the liver X receptor (LXR), thus increasing the transcription of lipogenic LXR in the liver (40). Also in the class 2B subclass were phosphorylation of AUP1S385, BBS2S365, and Serpin B9S366, all of which were decreased by insulin stimulation and showed reduced insulin regulation in T2D. Ancient ubiquitous protein 1 (AUP1) plays an important role in lipid droplet formation, degradation on INSIG1, and SREBF1 and ApoB stability (41–43). BBS2 is a ciliary protein involved in the pathogenesis of Bardet-Biedl syndrome (BBS), a genetically heterogeneous disease characterized by obesity, diabetes, and hyperphagia. Genetic knockout of BBS2 in mice results in hepatic insulin resistance (44–46). Other phosphorylations normally downregulated by insulin that were impaired in the T2D cells included the proteins in the Rho GTPase cycle, albeit these involved different phosphosites and proteins than did those falling into class 1B (Supplemental Table 3).

Emergent insulin signaling in T2D iHeps. The most interesting classes of altered protein phosphorylations in T2D are classes 1C and 2C. These phosphosites represent subsets of insulin-regulated phosphorylations that showed little or no response to insulin in control cells but markedly increased or decreased their phosphorylation in T2D iHeps. Thus, these sites represent emergent or new insulin-regulated phosphorylations in the context of insulin resistance (Figure 4, A and B).

Emergent insulin signaling in T2D iHeps upon insulin stimulation. (A) Hierarchical clustering showing insulin-regulated phosphosites in control and T2D iHeps. Rows represent normalized z scores corresponding phosphosite intensities for each sample. (B) Schematic showing the number of phosphosites with “emergent” signaling in T2D cells, i.e., phosphorylation events that were increased or had significantly reduced regulation in T2D compared with control cells using data from the heatmap in Figure 2B. (C and D) Reactome pathway enrichment of “emergent sites” in T2D iHeps. (E and F) Quantification of exemplary emergent phosphosites normally increased in phosphorylation in C or decreased in phosphorylation in D. In the panels, the cells from male donors are represented by squares and those from female donors by circles, and colors indicate individual donors. Data are presented as the mean ± SEM. n = 8 control; n = 6 T2D. *P < 0.05, **P < 0.01, and ***P < 0.001, by paired, 2-tailed t test of raw intensities between groups for basal versus insulin, and unpaired, 2-tailed t test between control versus T2D with or without insulin.

Class 1C phosphosites, i.e., those that were increased by insulin stimulation in T2D iHeps by greater than 1.5-fold and with a P value of less than 0.05 but with little or no change in control cells, were enriched for phosphosites on a third subset of proteins involved in Rho-GTPase signaling but different from those in classes 1B and 2B, which showed impaired or defective phosphorylation, as well as enrichment for proteins involved in vesicle-mediated transport, metabolism of RNA, mTOR signaling, and regulation of PTEN (Figure 4, C and E). These phosphosites included increased phosphorylation of AKT1S1S202, which increased its association with the scaffold protein 14-3-3, thereby decreasing the pool of free AKT1S1 and lowering the activation of mTORC1 (47, 48). Interestingly, phosphorylation on IRS2Y675, a direct substrate of the IR and a residue in a PI3K-binding motif (49), was enhanced in T2D iHeps. RPS6KB1S452, a protein involved in mTOR signaling and whose phosphorylation is increased by insulin stimulation, also showed increased phosphorylation in T2D cells, as did phosphorylation of FOXK1S468, a transcription factor involved in insulin action that can physically interact with IRs (50) and play a role in the regulation of the gluconeogenic genes PCK1 and G6Pc in the liver (51). Other proteins involved in vesicle-mediated transport also showed emergent signaling in T2D and are summarized in Supplemental Figure 4, A and B, and Supplemental Table 4.

Finally, class 2C represents emergent phosphosites for which there was little response to insulin in control cells but whose phosphorylation was significantly reduced upon insulin stimulation in T2D cells (Figure 4, D and F). This included another subset of proteins in the membrane trafficking pathway and proteins in pathways of retrograde Golgi transport, such as ARID1AS1754,S1755, which were dephosphorylated in response to insulin in T2D iHeps but not in control cells (Supplemental Table 4). Deletion of ARID1A in liver has been shown to increase susceptibility to the development of hepatic steatosis and insulin resistance by impairing FA oxidation (52). Dephosphorylation of SEC16AS314, a protein involved in ER-to-Golgi transport (53, 54) in T2D, as well as the mineralocorticoid receptor NR3C2S283,S387 are sites known to affect receptor stability (55). NR3C2 has also been found to affect glucose regulation in skeletal muscle (56). FKBP5S13, which plays a major role in regulating AKT1activity, also showed emergent dephosphorylation. Finally, several proteins involved in chromatin modification, such as EP300, EP400, and KMT2D, showed insulin-stimulated decreases in phosphorylation in T2D iHeps, which was not seen in controls (Supplemental Figure 4, C and D). The role of these chromatin-modifying proteins in insulin signaling and T2D has yet to be explored.

Basal phosphorylation defects in T2D. In addition to differences in insulin-regulated phosphorylations, PCA of the phosphoproteomics data revealed alterations in basal phosphorylation between control and T2D iHeps (Figure 5A). Pathway analysis of these proteins showed enrichment in many of the same pathways exhibiting alterations in the insulin-regulated state. Thus, we observed increased basal phosphorylation in proteins involved in the Rho GTPase cycle, vesicle-mediated transport, cell-cell communication, diseases of signal transduction, and processing of mRNA (Supplemental Figure 5A and Supplemental Table 5), whereas proteins showing lower basal phosphorylation were enriched in pathways involved in the Rho-GTPase cycle, mRNA splicing, and IFN induced with helicase C domain 1–mediated (IFIH1-mediated) induction of IFN-α (Supplemental Figure 5B and Supplemental Table 5). Many proteins in the insulin signaling pathway, including IRS2S391,T527,T350,S1100, FOXO1S287,S329, GSK3BT390, and PIK3R1Y467, also exhibited significantly lower phosphorylation in the basal state in T2D iHeps (Supplemental Figure 5C). Proteins that exhibited higher basal phosphorylation in T2D iHeps were enriched in the Rho-GTPase cycle, including CKAP4S460, RANBP2S2626, and PPP1R12AS422 (Supplemental Figure 6A). By contrast, basal phosphorylation of CREBBPS2364, PCBP2S188, and TNAIP2S645, which are involved in IFIH1-mediated induction of IFN-α/-β antiinflammatory proteins, was lower in T2D cells (Supplemental Figure 6B). Protein kinase N1 (PKN1T774) also showed significantly increased basal phosphorylation, a site linked to PKN1 activation, in T2D iHeps (Supplemental Figure 6C). PNK1 has been implicated in adipocyte differentiation, insulin-induced actin cytoskeleton reorganization, and GLUT4 translocation (57) and is also identified as an insulin-regulated kinase in iHeps in the kinome analysis described below.

Sexual dimorphism in the basal and insulin-stimulated phosphoproteome. (A) PCA plot showing separation of the phosphoproteome after adjusting for the surrogate variable by phenotype (control, open circles; T2D, closed circles) and sex (blue, males; red, females) in the basal state. (B) Hierarchical clustering of phosphopeptides after adjusting for the surrogate variable that shows sexual dimorphism (F-test FDR <5%). Rows represent z scores of log2-transformed intensity of phosphosites for each sample in each column. (C and D) Quantification of basal and insulin-stimulated phosphorylation levels of exemplary proteins showing sexual dimorphism. In the panels, the cells from male donors are represented by squares, and those from female donors are represented by circles, with colors indicating individual donors. (E) Signaling map representation of some of the most enriched biological pathways and related proteins that exhibited significantly higher phosphorylation ratios in males (blue) versus females (pink). Data are presented as the mean ± SEM, ***P < 0.001, based on the F-test for sex from unadjusted intensities.

Identification of sex-specific phosphoproteomic changes in iHeps. A patient’s sex can be a modifier of disease pathogenesis, and both T2D and MASLD occur more frequently in men than women (58). In previous studies, we observed that the sex of the donor modifies the phosphoproteome of iPS cell–derived myoblasts (iMyos) (25). This was also true in iHeps. Indeed, of the 21,863 sites quantitated, 2,680 sites (12.2%) showed sexual dimorphism, with 1,553 sites exhibiting female dominance and 1,127 showing male dominance using an FDR of less than 0.1 (Figure 5, A and B, and Supplemental Table 6). Among the female-dominant sites were those on the amino acid transporter collectrin (CLTRNS125) and on the sphingosine kinase SPHK2T198 (Figure 5C). Consistent with this, previous studies have shown that knockout of SPHK2 improves insulin sensitivity in male mice but not female mice (59). Male-dominant phosphosites, on the other hand, included protein kinase C γ (PRKCGT635) and Sushi domain–containing 5 protein (SUSD5S105) (Figure 5D). The latter has been implicated in insulin resistance in adipose tissue (60). We also observed sexual dimorphism in both control and T2D iHeps upon insulin stimulation (Supplemental Figure 6D). Pathway analysis revealed male-dominant insulin-stimulated phosphorylation differences related to RNA metabolism, autophagy, and signal transduction proteins, whereas female-dominant phosphorylations occurred in pathways related to the cell cycle, cellular response to stress, and sphingolipid metabolism. A network diagram with representative proteins showing sex differential phosphorylation is shown in Figure 5E.

Defining the kinases that contribute to normal and altered insulin signaling in iHeps. Defining the potential protein kinases involved in the altered insulin action of T2D is a challenge, since the exact kinases responsible for many altered phosphorylation sites are unknown, and many kinase-substrate relationships remain undefined. Recently, Johnson et al. developed a tool to deconvolute these relationships using synthetic peptide libraries to profile the substrate sequence specificity of 303 mammalian Ser/Thr kinases, representing more than 84% of the Ser/Thr kinases present in human cells (61). These include many kinases not previously linked to insulin action. Probing the specific amino acid sequences surrounding the phosphoserines and phosphothreonines in our phosphoproteomics data against this database, we sought to identify the kinases involved in insulin signaling in iHeps and those that are altered in activity in T2D by ranking the top S/T kinases predicted to target that phosphorylation site and creating a percentile score, as previously described (61). In agreement with the known insulin signaling cascade, when this algorithm was applied to identify the kinases contributing to insulin regulation of phosphorylation in control iHeps, the top kinases predicted to be involved (at >95% confidence) were members of the AKT family (AKT1, AKT2); the S6 kinase family (P70S6K, RSK3, p90RSK); serum glucocorticoid kinase 3 (SGK3); and the phosphorylase kinases PHKG1 and PHKG2. This analysis also identified some less-well-studied kinases, including protein kinase N (PKN1, -2, and -3), which has been linked to insulin stimulation of GLUT4 translocation in adipocytes (57), and the doublecortin kinases DCAML1 and DCAML2, which have also recently been linked to insulin action (62) (Supplemental Figure 7A and Supplemental Table 7).

More interestingly, the kinome analysis allowed a direct comparison of the kinases predicted to contribute to the altered serine/threonine phosphorylation in T2D in the insulin-stimulated states. This is shown as a bubble plot in Figure 6A comparing kinases activated in control versus T2D cells. Consistent with the above data and with previous studies, kinome analysis predicted that reduced activity of the kinase family member AKT2 was contributing to the reduced insulin-stimulated phosphorylations in T2D. The kinome analysis also predicted reduced activity of several other kinases, including PKCθ, YANK3 (STK32c), AMPKA2, BRSK2, MAPKAPK2, CHK2, PHKG2, TTBK2, and WNK3, as contributors to the impaired insulin-stimulated phosphorylations observed in T2D iHeps (Figure 6A, highlighted with black arrows). By contrast, for the substrates showing increased phosphorylation in T2D iHeps, kinome analysis best mapped to increased activity of the Rho-associated kinases ROCK1 and -2, as well as BCKDK and MST4 (Figure 6A, highlighted with red arrows). A summary of the altered substrate phosphorylations predicted to be the result of these increases or decreases in kinase activity is provided in Supplemental Table 8, and the top altered impaired or emergent kinase-substrate pairs in T2D are shown in Supplemental Figure 7B.

Kinome profiling of predicted kinases related to insulin stimulation and altered in T2D. (A) Bubble plot showing kinases predicted to be activated in iHeps upon insulin stimulation for control and T2D cells. In the bubble maps, the size and color represent the adjusted P values and frequency factor (FF), respectively, for the significant kinases (adjusted P ≤ 0.1). The black arrows indicate kinases with reduced activity in T2D following insulin stimulation, and the red arrows show the kinase predicted to have increased activity in T2D upon insulin stimulation. (B) Representative immunoblot analysis of phosphorylation of proteins in 4 T2D male donors with and without 100 nM insulin and 10 nM ripasudil. Quantifying the data in bar graphs normalized to a total protein is on the right. In the panels, the cells from male donors are represented by squares, and color indicates individual donors. Data are presented as the mean ± SEM. n = 6–8. *P < 0.05 and ***P < 0.001, by 2-tailed, paired, 2-tailed t test for basal versus insulin, insulin versus insulin+inhibitor, and unpaired, 2-tailed t test for insulin versus inhibitor.

For those kinases detected in the LC-MS/MS proteomics analysis, none of the kinases with predicted reduced or increased activity showed a difference in abundance between control and T2D iHeps at the protein level, except CHK2, which was higher at the protein level, but predicted to be lower in activity in T2D compared with control iHeps, suggesting that the changes in activity were not likely due to changes in kinase abundance (Supplemental Figure 8). On the other hand, several of the predicted alterations were consistent with previous studies. Thus, ROCK1 protein levels and activity are increased in the liver of mice on a high-fat diet (HFD) (63), BCAA metabolism has been linked to insulin resistance and T2D (20), and the STE20-type kinases MST3 and MST4 have been linked to the progression of hepatocellular carcinoma (64). To directly test whether increased ROCK activity contributes to the altered insulin signaling in T2D, iHeps from control individuals and patients with T2D were treated with a selective inhibitor of ROCK1 and -2, ripasudil (63, 65). Proximal insulin signaling assessed by immunoblotting showed significant improvement by this treatment (Figure 6B and Supplemental Figure 9), indicating that the overactivity of ROCK kinases likely contributes to the intrinsic alterations in insulin signaling observed in T2D iHeps.

An integrated picture of the insulin-regulated phosphoproteome in normal and T2D iHeps. Putting all of the above data together, we can construct a composite insulin-signaling map highlighting some of the important alterations in protein phosphorylation in T2D iHeps, including impaired and emergent changes (Figure 7). Note that in the proximal insulin-signaling pathway, insulin-dependent phosphorylation of AKT2S474 was preserved in T2D, whereas phosphorylation of AKT2S478 was reduced. AKT2S478 is less well explored in insulin signaling. It has been reported that Cdk2/cyclin A can phosphorylate AKT2S478 and synergize with S474 phosphorylation at the HM site to activate this kinase allosterically (66). Interestingly, although serine phosphorylation of IRS1 is usually regarded as being negatively regulatory for insulin signaling (36), multiple sites of serine phosphorylation of IRS1 were reduced in T2D iHeps compared with controls, suggesting that these are part of a normal feedback mechanism and that this mechanism may be lost in T2D. In addition, we observed reduced phosphorylation of mTORS2473 and reduced phosphorylation of multiple proteins involved in Rho signaling, including ARHGAP12S176, ARHGEF7S572,S608 and ARGAP17S702.

Insulin signaling map showing some critical nodes of phosphoproteome alterations in T2D iHeps. The signaling map shows exemplary phosphosites detected in the phosphoproteomic analysis. The effect of insulin stimulation is shown in color, with red representing increased phosphorylation and blue representing decreased phosphorylation (a significant difference at P < 0.05). For each site, the stimulation (up or down) in the control iHeps is indicated in the left half of the circle, while that for the T2D iHeps is indicated in the right half of the circle. Each of the phosphosites is color coded according to the effects of insulin on phosphorylation. The pathways shown are representative of those enriched in the Reactome pathways analysis.

The emergent sites of Ser/Thr phosphorylation in T2D iHeps were equally prominent. These include TSC1S504,S1074, 4E-BP1S52,S92, and multiple sites on the transcription elongation factor IWS1. TSC1 is an important negative regulator of mTOR signaling, and loss of TSC1 results in mTOR-dependent increased phosphorylation of the ribosomal protein S6, p70S6K, and 4E-BP1, leading to cell growth and proliferation. IWS1, an RNA processing regulator recruits the histone methyltransferase SETD2 to RNAPII and regulates the splicing of FGFR2-related genes (67). Some proteins had multiple alterations. The nuclear corepressor NCOR1 showed losses in the regulation of sites that in iHeps from control patients showed both increased and decreased phosphorylation following insulin stimulation. This could potentially modify its ability to interact differentially with and alter the activity of its different transcription factor partners, including LXR, ERR, and PPAR. Likewise, closely related proteins sometimes showed changes in opposite directions. For example, in T2D iHeps, IRS1 were found to have many sites that lost insulin-stimulated serine phosphorylation, whereas IRS2 had sites that gained phosphorylation. Likewise, phosphorylation of TBC1D4S588 was reduced in T2D iHeps, while phosphorylation of TBC1D1S108 was increased.

Finally, it is worth noting that multiple alterations in phosphorylation were observed for proteins in the nucleus, including proteins involved in transcription (IWS1, NCOR1, and NDRG1) and RNA splicing (PCBP2, PRPF4B, SF3B2, and SRRM2). IWS1 showed multiple sites of emergent phosphorylation in T2D cells (IWS1S248,S313,S362,S720), whereas phosphorylation of NDRG1S362,S367 after insulin stimulation was lost. Likewise, among insulin downregulated sites, NCOR1S1756,1977 phosphosites were lost in T2D cells. Interestingly, 3 proteins involved in RNA splicing lost insulin regulation of sites of tyrosine phosphorylation normally downregulated following insulin stimulation, including PCBP2Y46, PRPF4BY140, and SF3B2Y379, suggesting loss of action of a nuclear tyrosine kinase or gain of action of a nuclear tyrosine phosphatase in T2D cells. SRRM2, on the other hand, lost 3 insulin-downregulated phosphosites (SRRM2S455,S2453,S1458) and gained 2 insulin-upregulated sites (SRRM2S440,T569). SRRM2 organizes splicing condensates to regulate alternative splicing, however, the effects of these phosphorylations on this activity remain to be determined.