Integrated analysis of the adipocyte plasma membrane proteome reveals KCC1 and PIT2 as novel insulin-responsive transporters

TL;DR

This study identifies two new transporters, KCC1 and PIT2, that move to the cell surface in response to insulin in fat cells.

Contribution

The study discovers KCC1 and PIT2 as novel insulin-responsive transporters in adipocyte plasma membranes.

Findings

KCC1 and PIT2 translocate to the plasma membrane in response to insulin.

Knockdown of KCC1 or PIT2 impairs insulin-stimulated glucose transport.

Insulin resistance impairs the translocation of KCC1 and PIT2 to the plasma membrane.

Abstract

The plasma membrane (PM) is a dynamic interface that integrates environmental cues with cellular responses. Insulin is known to remodel the PM primarily by stimulating the translocation of glucose transporter GLUT4, but the full scope of insulin’s PM remodeling remains poorly defined. Here, we performed a meta-analysis of insulin-regulated PM proteins in adipocytes by integrating nine independent proteomic datasets generated using complementary PM enrichment strategies. The meta-analysis identified 37 insulin-regulated candidates detected in at least three datasets, including 30 proteins not previously implicated in insulin action. Among these, we experimentally characterized the insulin-stimulated translocation of two transporters: potassium-chloride cotransporter 1 KCC1 (SLC12A4) and sodium-dependent phosphate transporter PIT2 (SLC20A2), which showed robust and reproducible recruitment to the PM in response to insulin. siRNA-mediated knockdown of KCC1 or PIT2 impaired insulin-stimulated glucose transport, suggesting a role for these transporters in insulin action. Live-cell and fixed-cell imaging revealed that both proteins localize across multiple endosomal compartments, undergo insulin dose-dependent trafficking to the PM, and require PI3K-AKT signaling for their mobilization. Strikingly, insulin-induced translocation of KCC1 and PIT2 to the PM was impaired in adipocytes rendered insulin resistant by chronic hyperinsulinemia, accompanied by increased perinuclear retention under basal conditions. Together, our work provides a valuable resource for understanding insulin-regulated PM remodeling in adipocytes, establishes KCC1 and PIT2 as novel insulin-responsive transporters, and supports the idea that insulin resistance involves defects in cell-surface delivery that extend beyond GLUT4.

Figures

Full text

The plasma membrane (PM) hosts receptors and transporters that mediate interactions between the cell and its environment (1, 2). The protein composition of the PM is essential for these functions and dynamically remodeled to maintain cellular homeostasis (1). Transmembrane protein localization is governed by the exocytic and endocytic recycling pathways, with specific cargoes further modulated by external cues and signaling pathways that fine-tune their surface abundance (3).

The actions of insulin on its target tissues exemplify how extracellular signals can regulate protein localization at the PM to meet physiological demands (4). Following a meal, insulin secreted by pancreatic β-cells acts on metabolic tissues such as skeletal muscle and adipocytes (5). The binding of insulin to its cell surface receptor initiates a signaling cascade that promotes the translocation of glucose transporter GLUT4 from intracellular storage vesicles (GLUT4 storage vesicles, or GSVs) to the PM, thereby enhancing glucose transport into the cell (6). This process is essential for maintaining glucose homeostasis and is disrupted in insulin resistance (IR) and type 2 diabetes (4, 6).

Insulin also increases the surface abundance of several key transmembrane proteins, including transporters (Na^+^/K^+^-ATPase, NHE6, and CD36) (7, 8, 9), receptors (SORT1, TFR, LRP1, and IGF2R) (10, 11, 12, 13), and the peptidase insulin-regulated aminopeptidase (IRAP) (14, 15). These findings suggest that insulin may orchestrate widespread remodeling of the PM proteome. However, the extent to which insulin regulates the PM proteome remains unclear.

Comprehensive analysis of the PM proteome remains technically challenging due to the low absolute abundance, high hydrophobicity, and heterogeneity of PM proteins (16). Various strategies have been developed to enrich PM proteins, including subcellular fractionation via differential ultracentrifugation, chemical and metabolic labeling, and enzymatic surface tagging (17). However, each of these methods has associated challenges (18). For example, subcellular fractionation-based PM cannot separate integral membrane proteins from those peripherally associated with the PM (7, 17). Similarly, cell surface biotinylation is susceptible to cytoplasmic contamination, reducing its specificity for true PM proteins (17). Integrative approaches that leverage complementary techniques remain lacking in PM proteomic analysis.

Here, we profiled the insulin-regulated adipocyte PM proteome by performing a meta-analysis of nine adipocyte PM proteomic datasets generated from complementary PM enrichment methods. This multifaceted approach improved both the coverage and confidence of PM protein identification, enabling robust detection of insulin-responsive targets. We identified two previously unrecognized insulin-regulated solute carriers, potassium-chloride cotransporter [KCC1] (SLC12A4) and sodium-dependent phosphate transporter 2 [PIT2] (SLC20A2), positioning them as novel effectors of the insulin signaling network. This study provides a valuable resource for investigating insulin-driven remodeling of the adipocyte PM proteome, highlighting dynamic changes in cell surface protein abundance and identifying new candidates that may mediate insulin’s diverse physiological actions.

Results

Meta-analysis of insulin-regulated surface proteins in adipocytes

To gain a comprehensive understanding of how insulin remodels the PM proteome in adipocytes, we assembled nine proteomic datasets (DS1–9) that employed distinct methods for isolating PM proteins. These included our previously published cationic colloidal silica isolation (DS1–3) (7) and aminooxy-biotinylation (DS7) (19), as well as previously unpublished datasets obtained using subcellular fractionation (DS4–6) and sulfo-NHS-SS-biotin-based surface biotinylation (DS8, 9) (Table S1 and Fig. 1A). Eight datasets were derived from 3T3-L1 adipocytes, while one from primary mouse adipocytes (DS5). All datasets included samples under both basal and insulin-stimulated conditions, with insulin dose and duration varying across datasets (Table S1).Figure 1**Analysis of multiple adipocyte PM proteomic datasets.**A, schematic of the insulin treatments and PM isolation methods in different PM proteomic datasets. DS, datasets. LC-MS/MS, liquid chromatography-tandem mass spectrometry. GO-CC, Gene Ontology cellular component. B, UpSet plot showing the overlap of insulin-regulated (p < 0.05) integral PM proteins across PM proteomic datasets with ≥3 biological replicates. C, Heatmap of 37 integral PM proteins identified by meta-analysis as significantly insulin-regulated and detected in ≥3 datasets. Values represent log_2_ insulin-over-basal fold change (FOB); ND, not detected. D, insulin responsiveness of IRAP, GLUT4, TFR, KCC1, and PIT2 across nine datasets. E, Spearman correlation between IRAP, GLUT4, TFR, KCC1, and PIT2 mRNA expression in subcutaneous adipose tissue and metabolic clinical features. Figure generated from adiposetissue.org (46) using data from (25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45) (∗ pFDR < 0.05, ∗∗ pFDR < 0.01, ∗∗∗pFDR < 0.001). BMI, body mass index; circ, circulating; CRP, C-reactive protein; HDL, high-density lipoprotein; HOMA-IR, homeostatic model assessment for insulin resistance; iso, isoproterenol; LDL, low-density lipoprotein; LEP, leptin; TG, triglycerides; WAT, white adipose tissue; WHR, waist-to-hip ratio. PM, plasma membrane; IRAP, insulin-regulated aminopeptidase; TFR, transferrin receptor; PIT2, sodium-dependent phosphate transporter 2; GLUT4, glucose transporter 4; pFDR, positive false discovery rate.

Each dataset comprises ∼1000 to 5000 proteins (Table S2 and Fig. S1A), with the DS9 yielding the largest set of proteins (4980 proteins). Across the nine datasets, 26 to 32% of the identified proteins were categorized as PM-associated (Gene Ontology [GO]-cellular component terms containing “plasma membrane” or “cell surface”) regardless of whether they are distributed across multiple compartments (Fig. S1A). Proteins from mitochondria (16–25%), nucleus (13–27%), cytoplasm (9–14%), and endoplasmic reticulum (ER) (8–15%) were also detected, reflecting common contaminants in PM enrichment workflows (Fig. S1A) (17). To focus on functionally relevant membrane proteins, we filtered for proteins containing at least one transmembrane domain, key structural features of transporters and signaling receptors (20). Following filtering, 33 to 45% of the PM-associated proteins were identified as integral membrane proteins according to UniProt (129) or DeepTMHMM 1.0 (21) (Fig. S1, B and C).

Next, we evaluated the insulin responsiveness of the PM proteome (Fig. 1B). As expected, well-characterized insulin responsive proteins GLUT4, IRAP, and TFR exhibited robust insulin-stimulated translocation (Fig. 1, C and D). Fold changes (insulin over basal) (FOB) ranged from 1.51 to 4.74 for GLUT4, 1.38 to 3.83 for IRAP, and 1.17 to 1.77 for transferrin receptor (TFR) across at least six datasets (Fig. 1D).

To identify PM proteins significantly regulated by insulin, we excluded datasets with only one biological replicate (Fig. 1B). Merging the remaining six datasets yielded 237 unique integral PM proteins significantly regulated by insulin in at least one dataset (Fig. 1B). GO enrichment analysis revealed that these proteins were enriched in biological processes such as “sodium ion export across PM”, “amino acid transmembrane transport”, and “organic acid transmembrane transport” (Fig. S1D), highlighting diversity of proteins (and processes) regulated by insulin.

Overlap of insulin-regulated PM proteins across individual datasets was limited (Fig. 1B), likely driven by methodological and biological heterogeneity. Thus, within-dataset analysis alone may be underpowered and prone to missing true insulin responses. To overcome these limitations and integrate insulin responsiveness across heterogeneous PM proteomic datasets, we applied a meta-analysis framework to increase statistical power and identify consistently regulated PM proteins. This approach standardizes study-specific log_2_ FOB values based on the median and median absolute deviation (MAD), then combines them across studies using sample-size-based weights (see Methods for details) (22). Using this method, we identified 216 proteins significantly regulated by insulin (false discovery rate, FDR < 0.05) (Table S2). Restricting the analysis to integral PM proteins detected in at least three datasets yielded 37 insulin-responsive PM proteins (Fig. 1C). Notably, most known insulin-responsive PM proteins were recovered, validating the robustness of our analysis approach (Fig. 1C). In addition, 30 PM proteins not previously reported as insulin-regulated were identified, comprising nine cell adhesion molecules, seven receptors, five transporters/channels, two hydrolases, and seven other regulatory proteins (Table S3). Twenty-two proteins were downregulated by insulin, including the known example INSR, suggesting that insulin exerts bidirectional effects on the PM proteome (Fig. 1C). These findings highlight insulin drives a dynamic remodeling of the PM proteome.

KCC1 and PIT2 undergo insulin dose-dependent PM recruitment

We next focused on two transporters whose cellular localizations were not previously known to be affected by insulin: the potassium-chloride cotransporter KCC1 (SLC12A4) (23) and the sodium-dependent phosphate transporter PIT2 (SLC20A2) (24). These proteins showed some of the largest and most consistent insulin-induced changes across the datasets, ranging from +18% to +176% for KCC1 and +33% to +167% for PIT2 (Fig. 1, C and D). Notably, siRNA-mediated knockdown of either transporter in mature adipocytes impaired insulin-stimulated glucose transport, suggesting a functional role in the cellular insulin response (Fig. S2). Further, both transporters are expressed in human adipose tissue (23, 24) and their expression is correlated with key clinical metabolic traits (Fig. 1E) (25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46).

To characterize insulin sensitivity and trafficking kinetics, we assessed the PM recruitment of KCC1 and PIT2 in 3T3-L1 adipocytes, a widely used model for studying insulin sensitivity (47, 48). Because commercial antibodies recognizing extracellular epitopes of endogenous KCC1 or PIT2 were unavailable, both proteins were tagged with mStayGold (49). PM recruitment was evaluated using total internal reflection fluorescence (TIRF) microscopy (50). As a positive control, we first measured the insulin responsiveness of the canonical regulated protein GLUT4. 3T3-L1 adipocytes expressing HA-GLUT4-mRuby3 were stimulated with insulin concentrations ranging from 0.01 to 10 nM. HA-GLUT4-mRuby3 exhibited robust, insulin dose-dependent translocation, reaching a maximum 3.6-fold increase over basal levels at 10 nM insulin (Fig. 2, A and B and Movie S1). Similarly, KCC1-mStayGold demonstrated a dose-dependent increase in PM localization; however, its maximal response (2.2-FOB) was lower than that observed for GLUT4 (Fig. 2, A and B and Movie S2). In addition, KCC1 displayed modestly lower insulin sensitivity than GLUT4 (EC_50_ = 0.33 ± 0.08 versus 0.22 ± 0.02 nM; mean ± SD; p = 0.0771 from unpaired, two-tailed student’s t test) (Fig. S3A). Of note, the EC_50_ for HA-GLUT4-mRuby3 was lower than reported values for endogenous GLUT4 (EC_50_ = 0.86–1.1 nM), likely due to GLUT4 overexpression (12, 19).Figure 2**KCC1 and PIT2 exhibit insulin dose-dependent PM recruitment.A, 3T3L1 adipocytes electroporated with either HA-GLUT4-mRuby3 or KCC1-mStayGold were stimulated with different insulin doses (0.01–10 nM). PM recruitment was assessed by TIRF microscopy. Representative images for three independent experiments are presented (the scale bar represents 5 μm). Bas, basal condition (0.01 nM insulin). Ins, insulin stimulation. B, quantification of panel A. FOB, insulin-over-basal fold change. C and E, 3T3L1 adipocytes electroporated with either HA-GLUT4-mRuby3 (C) or PIT2-HA (E) were stimulated with different insulin doses (0.01–100 nM). PM recruitment was assessed by immunofluorescence staining and confocal microscopy. Representative images for two to three independent experiments are presented (the scale bar represents 20 μm). D, quantification of panels C and E. F, SGBS adipocytes stimulated with 0 or 10 nM insulin were fractionated to enrich PM proteins. Whole-cell lysates (WCL) and PM fractions were immunoblotted with the indicated antibodies. 14-3-3 and caveolin-1 (CAV1) served as loading controls for WCL and PM fractions, respectively. Representative blots from three independent experiments are shown. G, quantification of *panelF*, statistical analysis was performed by one-way ANOVA with Dunnett’s post hoc; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 (n = 2–3 independent biological replicates). In (B and D), p values were compared to basal (0.01 nM insulin); in (G), p values were compared to GLUT4. PM, plasma membrane; TIRF, total internal reflection fluorescence; SGBS, Simpson–Golabi–Behmel syndrome; PIT2, sodium-dependent phosphate transporter 2; GLUT4, glucose transporter 4.

PIT2-mStayGold expression levels were extremely low following electroporation, with most cells either failing to express the construct or becoming nonviable (Fig. S3B). To address this issue, we utilized a smaller HA epitope tag fused to the extracellular C terminus of PIT2 and performed the translocation assay using an HA-specific antibody. This strategy substantially enhanced PIT2 expression in 3T3-L1 adipocytes (Fig. S3B). Subsequently, we benchmarked PIT2-HA trafficking against cells overexpressing HA-GLUT4-mRuby3 using extracellular HA staining and confocal microscopy (Fig. 2, C and E). HA-GLUT4-mRuby3 showed insulin dose-dependent PM recruitment, reaching a maximal change of 5.2-fold increase at 100 nM insulin (Fig. 2D). PIT2-HA increased at the PM in a dose-dependent manner, with a maximal change of ∼3.1-fold at 100 nM insulin (Fig. 2D). Notably, PIT2 demonstrated greater insulin sensitivity than GLUT4, with EC_50_ values of 0.14 ± 0.02 and 0.31 ± 0.08 nM, respectively (Fig. S3C).

Finally, we validated insulin-stimulated redistribution of KCC1 and PIT2 to the PM in cultured human adipocytes (Simpson–Golabi–Behmel syndrome, SGBS) (51). PM fractions were isolated by subcellular fractionation and immunoblotted for GLUT4, KCC1, and PIT2 (Fig. 2F). Insulin (10 nM) increased PM-localized KCC1 and PIT2 by 2.0 ± 0.6-fold and 2.3 ± 0.3-fold (mean ± SD), respectively, compared with a 3.4 ± 0.7-fold increase for GLUT4 (Fig. 2G). The greater fold-change for GLUT4 likely reflects its lower basal abundance at the cell surface relative to KCC1 and PIT2 (Fig. 2F).

Insulin-stimulated trafficking of KCC1 and PIT2 is regulated by the PI3K-AKT pathway

Since class I phosphoinositide 3-kinase (PI3K) plays a critical role in regulating the cellular effects of insulin (52) (Fig. 3A), we evaluated the dependency of insulin-stimulated KCC1 and PIT2 trafficking on PI3K activity. We pretreated 3T3-L1 adipocytes with 10 μM GDC-0941, a class I PI3K inhibitor, for 10 min prior to insulin stimulation (53). PI3K inhibition was confirmed by suppression of insulin-induced phosphorylation of protein kinase B (AKT) (Fig. 3, B and C). Pretreatment with GDC-0941 completely abolished the PM recruitment of HA-GLUT4-mRuby3, KCC1-mStayGold, and PIT2-HA (Fig. 3, F–I), demonstrating that PI3K activation is essential for the insulin-stimulated trafficking of these proteins.Figure 3Insulin-stimulated translocation of KCC1 and PIT2 to the PM requires PI3K-AKT signaling.A, schematic of the proximal insulin signaling pathway. IRS, insulin receptor substrates. PI3K, class I phosphoinositide 3-kinase. AKT, protein kinase B. B and D, 3T3-L1 adipocytes were pretreated for 10 min with 10 μM DMSO (vehicle; Ctrl), the PI3K inhibitor GDC-0941 (PI3Ki) (B), or the AKT inhibitor MK-2206 (AKTi) (D), then stimulated with 1 nM insulin. Cell lysates were immunoblotted with the indicated antibodies; 14-3-3 served as a loading control. Representative blots from three to four independent experiments are shown. C, quantification of *panelB*. E, quantification of *panelD*. F–H, 3T3L1 adipocytes electroporated with either HA-GLUT4-mRuby3 (F and H) or KCC1-mStayGold (G and H) were treated as in (B and D), and PM recruitment was assessed by TIRF microscopy. I, 3T3L1 adipocytes electroporated with either HA-GLUT4-mRuby3 or PIT2-HA were treated as in (B and D), and PM recruitment was assessed by immunofluorescence staining and confocal microscopy. Statistical analysis was performed by one-way ANOVA with Tukey’s post hoc; ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 (n = 2–8 independent biological replicates). PM, plasma membrane; TIRF, total internal reflection fluorescence; PI3K, phosphoinositide 3-kinase; DMSO, dimethyl sulfoxide; KCC1, potassium-chloride cotransporter; PIT2, sodium-dependent phosphate transporter 2; GLUT4, glucose transporter 4.

We next investigated whether AKT activation is required for insulin-induced trafficking of KCC1 and PIT2 to the PM. 3T3-L1 adipocytes were pretreated with 10 μM of the AKT inhibitor (AKTi) MK-2206 (54) for 10 min prior to insulin stimulation. The effectiveness of the inhibitor was confirmed by suppression of insulin-induced AKT phosphorylation (Fig. 3, D and E). AKT inhibition effectively abolished insulin-induced PM recruitment of GLUT4 (Fig. 3, F, H, and I). Similarly, KCC1 and PIT2 translocation was substantially reduced by 83% and 63%, respectively (Fig. 3, G–I), suggesting that AKT activation is also required for their insulin-stimulated translocation. Collectively, these findings demonstrate that KCC1 and PIT2 undergo insulin-stimulated translocation to the PM in a PI3K-AKT dependent manner.

KCC1 and PIT2 are localized across multiple insulin-responsive compartments

In adipocytes, insulin-responsive trafficking of PM proteins occurs through distinct intracellular pathways (12, 55, 56). For GLUT4, the primary compartment that is mobilized by insulin is the GSV, while a secondary pathway involves the recycling endosome, the latter being the main location of TFR (57). Given the distinct insulin responsiveness observed for KCC1 and PIT2 compared with GLUT4, we hypothesized that KCC1 and PIT2 may be localized to different insulin-sensitive compartments.

To test this, we assessed the colocalization of these proteins using confocal microscopy. To validate our approach, we transfected IRAP-pHluorin, a well-established marker of GSVs (58), and assessed its colocalization with endogenous GLUT4 and TFR in unstimulated 3T3-L1 adipocytes (59). IRAP-pHluorin strongly colocalized with GLUT4 (Fig. S4A), with more than 75% of IRAP-positive compartments overlapping GLUT4-positive compartments (Fig. S4B). In contrast, IRAP and GLUT4 displayed lower colocalization with TFR (51% for IRAP and 39% for GLUT4) (Fig. S4B), and only 41% of TFR-positive compartments overlapped with IRAP/GLUT4-positive compartments (Fig. S4B). These results align with the known subcellular distribution of these proteins in adipocytes (60, 61, 62).

Next, we examined the distribution of KCC1-mStayGold and PIT2-HA relative to GLUT4 and TFR under basal conditions. KCC1-mStayGold colocalized with both GLUT4 and TFR, mainly in perinuclear areas but also within smaller peripheral cytoplasmic puncta (Fig. 4A). Quantitative analysis showed that KCC1-mStayGold overlapped with TFR to a similar extent as with GLUT4 (∼55%) (Fig. 4B). Similarly, PIT2-HA was detected in the perinuclear region, in discrete cytosolic puncta, and at the PM (Fig. 4E). The extent of overlap between PIT2-positive compartments and GLUT4 or TFR was comparable to that observed for KCC1 (approximately 57%, Fig. 4F), suggesting that PIT2 also partitions into multiple intracellular vesicle populations. To confirm these findings, we immunostained endogenous KCC1 and PIT2 in 3T3-L1 adipocytes (Fig. 4, C and G) and SGBS adipocytes (Fig. S4, C−E) and quantified overlap with endogenous GLUT4 and TFR. Endogenous KCC1 and PIT2 displayed distributions similar to the tagged constructs, with signal in the perinuclear region, cytosolic puncta, and at the PM (Figs. 4, C and G and S4, C and D). Both proteins colocalized with GLUT4 and TFR, with overlap values ranging from ∼50 to 62% across cell models (Figs. 4, D and H and S4, B and E). These results suggest that KCC1 and PIT2 localize to multiple insulin-sensitive compartments.Figure 4**Colocalization of KCC1/PIT2 with GLUT4 and TFR in unstimulated 3T3-L1 adipocytes.**A and E, cells electroporated with either KCC1-mStayGold (A), or PIT2-HA (E) (cyan) were fixed, permeabilized, stained for nuclei (gray), GLUT4 (yellow), and TFR (magenta), and imaged using confocal microscopy. Representative images from three independent experiments are shown (the scale bar represents 5 μm). B and F, quantification of panels A and E. C and G, cells were fixed, permeabilized, stained for nuclei (gray), GLUT4 (yellow), TFR (magenta), and KCC1 (C) or PIT2 (G) (cyan), and imaged using confocal microscopy. Representative images from three independent experiments are shown (the scale bar represents 5 μm). D and H, quantification of panels C and G, quantitative analysis includes (i) percentage of each protein volume above threshold colocalized with GLUT4; (ii) percentage of GLUT4-positive volume colocalized with each protein; (iii) percentage of each protein volume colocalized with TFR; and (iv) percentage of TFR-positive volume colocalized with each protein. Statistical analysis was performed by one-way ANOVA with Tukey’s post hoc test (n = 3 independent biological replicates). TFR, transferrin receptor; KCC1, potassium-chloride cotransporter; PIT2, sodium-dependent phosphate transporter 2; GLUT4, glucose transporter 4.

Defective KCC1 and PIT2 trafficking in IR

IR, defined as the impaired ability of cells to respond to insulin, is strongly associated with the development of cardiometabolic diseases (63). To date, most studies examining IR from a trafficking perspective have focused primarily on GLUT4 translocation (4, 6, 64). However, since insulin promotes the translocation of multiple proteins to the PM, we hypothesized that defects in membrane protein trafficking may represent a broader feature of insulin-resistant states than previously recognized. To test this, we examined the PM recruitment of TFR, KCC1, and PIT2 under insulin-resistant conditions.

3T3-L1 adipocytes were exposed to 1 or 10 nM insulin for 24 h to mimic hyperinsulinemia, a well-established model of IR in vitro (12, 65, 66). As expected, chronic insulin treatment impaired GLUT4 PM recruitment in a dose-dependent fashion, reducing surface GLUT4 by approximately 50% and 63% in the 1 and 10 nM chronic insulin groups, respectively (Figs. 5A and S5B) (12, 67, 68, 69, 70). Consistent with previous observations (12), chronic insulin exposure did not alter overall adipocyte morphology (Fig. S5, A and D), and total GLUT4 protein levels remained unchanged (Fig. S5, C and G), indicating that reduced surface GLUT4 is due to defective trafficking (Fig. 5B).Figure 5**Impaired PIT2 and KCC1 translocation to the PM in insulin resistance.**A–D, endogenous PM GLUT4 (A and B) and PM TFR (C and D) in 3T3-L1 adipocytes at basal and after acute 1 nM insulin, following 24-h pretreatment with 0 nM, 1 nM, or 10 nM insulin. Ctrl, control; CI_L, 1 nM chronic insulin; CI_H, 10 nM chronic insulin. Data are shown as raw PM intensity (A and C) or normalized to total GLUT4/TFR abundance (B and D), and expressed as percentage of the acutely insulin-stimulated control. E and F, 3T3L1 adipocytes electroporated with PIT2-HA were exposed to either 0, 1, or 10 nM insulin for 24 h, followed by stimulation with 1 nM insulin. PM recruitment was assessed by immunofluorescence staining and confocal microscopy. Data are normalized to total PIT2-HA and expressed as percentage of the acutely insulin-stimulated control (E) or as the difference (%) between the acutely insulin-stimulated and basal PM intensities within each chronic-insulin condition (F). G and H, 3T3L1 adipocytes electroporated with either HA-GLUT4-mRuby3 (G) or KCC1-mStayGold (H) were stimulated with 1 nM insulin after treatment with 0, one or 10 nM insulin for 24 h. PM recruitment was captured by TIRF microscopy. Data are shown as insulin-over-basal fold change (FOB). Statistical analysis was performed by two-way ANOVA with Tukey’s post hoc test; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001. For (F), one-way ANOVA with Tukey’s post hoc was used (n = 4–6 independent biological replicates). PM, plasma membrane; TIRF, total internal reflection fluorescence; TFR, transferrin receptor; KCC1, potassium-chloride cotransporter; PIT2, sodium-dependent phosphate transporter 2; GLUT4, glucose transporter 4.

In parallel, we quantified endogenous TFR trafficking in IR. Unlike GLUT4, total TFR levels were elevated by ∼29% and ∼44% following 1 and 10 nM chronic insulin exposure, respectively (Fig. S5, F and H). Consistent with this, surface TFR levels were increased under both basal and acute insulin stimulation to a similar extent in insulin-resistant cells (Figs. 5C and S5E), suggesting that this potentiation is due to increased TFR expression and that the insulin-stimulated trafficking of TFR is unaffected in insulin-resistant cells (Fig. 5D).

We next employed a similar approach to examine PIT2-HA trafficking in adipocytes. Chronic insulin reduced surface PIT2-HA in a dose-dependent manner, although to a lesser extent than GLUT4, with 21% and 36% reductions observed following 1 and 10 nM chronic insulin treatments, respectively (Fig. 5E). Total endogenous PIT2 levels remained unchanged in insulin-resistant states (Fig. S5I). Notably, basal surface levels of PIT2-HA were reduced by approximately 17% following chronic treatment with 10 nM insulin (Fig. 5E). This impairment in the basal-state likely contributes to the attenuated insulin-stimulated trafficking response of PIT2-HA observed after chronic 10 nM insulin treatment (Fig. 5F).

KCC1 trafficking was evaluated using TIRF microscopy of KCC1-mStayGold, with HA-GLUT4-mRuby3 included as a positive control. As expected, HA-GLUT4-mRuby3 reproduced the impaired trafficking characteristic of endogenous GLUT4 (Fig. 5G), showing ∼31% and 57% reductions following chronic exposure to 1 nM and 10 nM insulin, respectively. KCC1-mStayGold also displayed a ∼28% reduction in PM localization with 1 nM chronic insulin, and a similar defect was observed at the higher insulin dose (Fig. 5H), indicating that hyperinsulinemia impairs KCC1 trafficking through mechanisms distinct from those regulating other insulin-sensitive proteins. Total endogenous KCC1 abundance was unchanged in IR (Fig. S5J) (71), supporting impaired trafficking rather than altered expression as the primary driver.

To determine whether reduced PM recruitment is accompanied by altered intracellular distribution, we immunostained endogenous GLUT4, PIT2, KCC1, and TFR in unstimulated adipocytes after chronic insulin exposure. Immunostaining revealed increased perinuclear enrichment in insulin resistant models for GLUT4 (Fig. 6A), PIT2 (Fig. 6C), and KCC1 (Fig. 6E), with perinuclear-to-total intensity ratios rising from 13% to 22% (GLUT4) (Fig. 6B), 11% to 17% (PIT2) (Fig. 6D), and 10% to 18% (KCC1) (Fig. 6F). TFR showed a modest, non-significant increase (16% to 19%) (Fig. 6, G and H). These findings suggest that defective trafficking of GLUT4, PIT2, and KCC1 in IR reflects retention in perinuclear compartments rather than being efficiently sorted into insulin-responsive exocytic pools.Figure 6Chronic insulin alters the perinuclear localization of PIT2 and KCC1 under basal conditions. 3T3-L1 adipocytes were treated with 0, 1, or 10 nM insulin for 24 h, serum-starved (basal) for 2 to 3 h, then fixed, permeabilized, and stained for GLUT4 (A and B), PIT2 (C and D), KCC1 (E and F), or TFR (G and H). Cells were imaged by confocal microscopy. Representative images from three independent experiments are shown (the scale bar represents 5 μm). Ctrl, control; CI_L, 1 nM chronic insulin; CI_H, 10 nM chronic insulin. Quantification shows the ratio of summed fluorescence intensity in the perinuclear region (PNR) to total cellular intensity. Statistical analysis was performed by one-way ANOVA with Tukey’s post hoc test; ∗p < 0.05, ∗∗p < 0.01 (n = 3 independent biological replicates). TFR, transferrin receptor; KCC1, potassium-chloride cotransporter; PIT2, sodium-dependent phosphate transporter 2; GLUT4, glucose transporter 4.

Altogether, our data indicate that KCC1 and PIT2, like GLUT4, exhibit reduced PM translocation in hyperinsulinemia-induced IR and are mislocalized to perinuclear regions at the basal state, suggesting a broader disruption of insulin-sensitive trafficking pathways.

Discussion

The meta-analysis of nine PM proteomic datasets uncovered a previously unrecognized collection of insulin-responsive surface proteins in adipocytes (7). Many of these proteins are broadly expressed in many cell types aside from adipocytes (72), suggesting that insulin, and potentially other growth factors, may regulate protein trafficking across multiple cell types. Importantly, most of these insulin-responsive proteins are not known components of the GSVs (10, 73, 74), the canonical insulin-responsive compartment. This suggests that insulin regulates endosomal compartment dynamics more broadly than previously recognized.

Among these candidates, we focused on KCC1 and PIT2 as representative examples for experimental characterization. In cultured adipocytes, both localized to multiple insulin-responsive compartments and underwent dose-dependent translocation to the cell surface in a PI3K-AKT-dependent manner. Strikingly, their trafficking to the PM with acute insulin was disrupted under insulin-resistant conditions.

KCC1 safeguards cell volume by sensing osmotic swelling and exporting K^+^/Cl^-^ in an electroneutral manner, creating the ion gradient that drives water efflux and restores normal cell size (23). Given that insulin promotes cell swelling (75), KCC1 activation may serve as a compensatory mechanism to re-establish cell volume homeostasis. KCC1 also facilitates serine uptake by maintaining chloride gradients for Cl^-^-dependent amino acid transporters (76, 77). Knockdown of KCC1 impaired insulin-stimulated glucose uptake (Fig. S2). As a K^+^/Cl^-^ cotransporter involved in ion homeostasis and cell-volume regulation, KCC1 depletion could influence glucose uptake indirectly. Perturbations in cell volume or ionic gradients can alter intracellular metabolite concentrations, enzyme activities, or membrane trafficking pathways (78). Another possibility is that KCC1 knockdown affects membrane potential in a manner that secondarily influences GLUT4 translocation, as reported for other potassium channels (79). Given the broad physiological roles of KCC1, we cannot exclude the alternative explanation that the reduced glucose uptake reflects pleiotropic effects, including altered cellular energy state or metabolic stress. These possibilities will require dedicated mechanistic investigation.

PIT2 presented a similar phenotype, with knockdown impairing insulin-stimulated glucose uptake (Fig. S2). Because insulin produces a transient drop in cellular energy charge linked to rapid phosphorylation events (80), we speculate that increased phosphate import via PIT2 may help restore phosphate-dependent metabolic intermediates during early insulin action. This is consistent with established roles for insulin in promoting sodium-dependent phosphate uptake (81) and with clinical observations of hyperphosphatemia in insulin-deficient states (82). Genetic associations between SLC20A2 (PIT2) and type 2 diabetes (83) further suggest a possible link between phosphate handling and glucose metabolism. However, these ideas remain hypothetical, and future work will be required to define whether and how PIT2-mediated phosphate transport impacts insulin-regulated glucose uptake.

In insulin-resistant adipocytes, insulin-stimulated trafficking of GLUT4, KCC1, and PIT2 to the PM was attenuated. In contrast, TFR maintained normal insulin responsiveness despite an overall increase in both surface and total protein levels. Consistent with these phenotypes, GLUT4, KCC1, and PIT2 showed significant enrichment in perinuclear regions under basal conditions in IR, whereas TFR did not. The divergent behavior of TFR indicates that IR does not impose a uniform trafficking blockade; instead, it creates cargo-selective vulnerabilities as we previously suggested (67). Together, these data highlight cargo-specific defects in intracellular trafficking, offering new insight into how dysregulated trafficking contributes to distinct aspects of metabolic dysfunction in IR.

This study poses two main questions (1): through what mechanisms does insulin regulate the trafficking of endosomal proteins, such as KCC1 and PIT2, to the PM? (2); why do insulin-responsive proteins exhibit varying degrees of susceptibility to IR? One likely determinant is cargo-specific regulation through post-translational modifications (PTMs) that control sorting and recycling at the PM (84, 85, 86). For GLUT4, PTMs including glycosylation (87), palmitoylation (88, 89), and ubiquitination (90) contribute to trafficking decisions. Insulin-responsive phosphorylation sites have also been reported for GLUT4, KCC1, and PIT2 (91, 92), suggesting phosphorylation may regulate surface abundance by promoting exocytosis and/or inhibiting endocytosis (93), processes that may be disrupted in IR.

A second layer likely involves cargo-selective engagement of trafficking machinery controlling vesicle formation, recycling routes, and fusion with the PM (94, 95, 96). Distinct VAMP isoforms, for example, have been implicated in differential trafficking behaviors of GLUT4 and CD36 (9). Moreover, global analysis of endosomal recycling complexes indicates that insulin-responsive PM proteins can show distinct dependencies on the sorting nexin 27 (SNX27)–retromer and SNX17–retriever pathways (97, 98). How KCC1 and PIT2 are routed through these recycling machineries remains undefined. Together, these findings highlight the complexity of intracellular trafficking regulation in insulin-sensitive cells and suggest that cargo-specific mechanisms may underlie the trafficking defect in IR.

Limitations

This study has several limitations. First, although our colocalization analysis indicates that KCC1 and PIT2 are distributed across multiple insulin-responsive compartments, including GSVs and recycling endosomes, their precise resident compartments remain undefined. Further higher-resolution studies using additional organelle-specific markers will be necessary to delineate their intracellular localization. Second, the mechanistic links between insulin-stimulated KCC1 and PIT2 translocation and metabolic outcomes remain to be established. While knockdown of either transporter impaired insulin-stimulated glucose uptake, the underlying pathways and physiological significance require further investigation. Third, whether the trafficking of these proteins is disrupted in in-vivo settings of IR remains unknown. Here, we employed chronic hyperinsulinemia as an in vitro model of IR. Although prolonged insulin exposure is widely used to induce cellular IR, it may not fully recapitulate the multifactorial pathophysiology of IR in obesity or diabetes, where lipotoxicity, inflammation, and systemic cues act in concert. Future studies using complementary models, such as palmitate- or tumor necrosis factor-alpha-induced IR, or primary adipocytes and tissues from obese or diabetic settings, will be important to establish the broader physiological relevance of these trafficking phenotypes.

Experimental procedures

Cell culture

Mycoplasma-free 3T3-L1 fibroblasts (gift from Howard Green, Harvard Medical School) were cultured in Dulbecco’s modified Eagle medium (DMEM, high glucose) (Gibco, Cat. #11960-044) supplemented with 10% (v/v) fetal bovine serum (FBS) (Sigma Life Science, Cat. #F9423-500 Ml) and 2 mM GlutaMAX (Gibco, Cat. #35050-061). Cells were maintained at 37 °C in a humidified atmosphere of 10% CO_2_. For differentiation into adipocytes, confluent fibroblasts were treated with DMEM/FBS/GlutaMAX containing 0.22 μM dexamethasone, 100 ng/ml biotin, 2 μg/ml insulin, and 500 μM 3-isobutyl-1-methylxanthine (IBMX). After 3 days, the differentiation medium was replaced with fresh DMEM/FBS/GlutaMAX containing 2 μg/ml insulin for an additional 3 days and subsequently switched to fresh DMEM/FBS/GlutaMAX. Adipocytes were refed every 48 h and used for experiments 9 to 12 days after the initiation of differentiation.

Mycoplasma-free SGBS cells (gift from Martin Wabitsch, Ulm University Medical Center, Germany) were cultured and differentiated as previously reported (99, 100), and used for experiments 11 to 14 days after initiation of differentiation. SGBS adipocytes were incubated for 2 days in DMEM/F-12 (Ham)/GlutaMAX (Gibco, Cat. #10565–042) supplemented with 33 μM biotin and 17 μM pantothenate before insulin addition.

Tissue collection and primary mouse adipocyte isolation

Isolated adipose cells were prepared from the whole epididymal fat pads of 6 to 8-week-old male CD1 mice, following the protocol described by Liu et al. (101) with some modifications. The epididymal fat tissue was quickly removed and rinsed in Krebs-Ringer-Hepes buffer (KRH) (140 mM NaCl, 4.7 mM KCl, 2.5 mM CaCl_2_, 1.25 mM MgCl_2_, 2.5 mM NaH_2_PO_4_, 10 mM Hepes, pH 7.4) containing 2.5% (w/v) bovine serum albumin (BSA) and 200 nM adenosine. All buffers were maintained at 37 °C throughout the procedure. The washed tissue was minced in KRH buffer containing 1 mg/ml collagenase Type I (Worthington, Cat. #9001-12-1), 3.5% (w/v) BSA, 5 mM glucose, and 200 nM adenosine, and incubated at 37 °C in a shaking water bath (100 rpm) for ∼20 min until digestion was complete. The resulting cell suspension was filtered through a 250 μm nylon mesh (Lockertex) and returned to fresh KRH buffer containing 2.5% (w/v) BSA and 200 nM adenosine. Cells were allowed to float, and the infranatant was removed using a blunt needle (2 mm × 100 mm) attached to a 20 ml syringe. The washing step was repeated three times with gentle resuspension in fresh buffer. The final cell suspension was adjusted to a cytocrit of ∼20% and cells are then ready to be used. Insulin stimulation was performed with 100 nM insulin for 30 min. Following stimulation, cells were washed once with KRH buffer (without BSA) and once with HES buffer (20 mM Hepes, 1 mM EDTA, 250 mM sucrose, pH 7.4) supplemented with protease inhibitors (Roche Applied Science, Cat. #11873580001) and phosphatase inhibitors (2 mM Na_3_VO_4_, 1 mM Na_4_O_7_P_2_, and 10 mM NaF). Cells were kept at 18 °C and homogenized using 10 strokes of a glass/Teflon Potter-Elvehjem homogenizer (Thomas Scientific) prior to subcellular fractionation.

Isolation of PM proteins

Subcellular fractionation

PM fractions were isolated from primary mouse adipocytes as previously described (102). All centrifugation and handling of samples were carried out at 4 °C. Briefly, the homogenate was centrifuged at 1000g for 1 min to remove debris. The resulting supernatant was then centrifuged at 17,000g for 20 min. The pellet was resuspended in HES buffer and layered over the high-sucrose HES buffer (20 mM Hepes, 1 mM EDTA, 1.12 M sucrose, pH 7.4), followed by centrifugation at 104,000g for 20 min. The PM fraction was collected from the interface between the sucrose layers, resuspended, and subjected to two additional centrifugation steps at 76,000g for 9 min each. The final PM pellet was resuspended in 100 mM Tris buffer containing 2% SDS (pH 8.0).

3T3-L1 adipocytes and SGBS adipocytes were starved for 2 to 3 h in serum-free DMEM/GlutaMAX supplemented with 0.2% BSA before treatment with insulin. Cells were then placed on ice, washed with cold PBS, and harvested in cold HES buffer supplemented with protease inhibitors. All subsequent steps were performed at 4 °C. Cells were homogenized by passing them through a 22-gauge needle 10 times, followed by a 27-gauge needle 6 times. The homogenate was first centrifuged at 500g for 10 min to remove unbroken cells and debris. The resulting supernatant was then centrifuged at 16,000 rpm for 12 min to pellet the PM along with mitochondria and nuclei. The PM/mitochondria/nuclei pellet was resuspended in HES buffer and centrifuged again at 16,000 rpm for 12 min to further purify the fraction. The resulting pellet was resuspended in HES buffer and loaded onto the high-sucrose HES buffer before ultracentrifugation at 36,000 rpm for 60 min to separate the PM from the mitochondria and nuclei. The collected interface (PM fraction) was then mixed with HES buffer and pelleted by ultracentrifugation at 66,000 rpm for 30 min. PM pellets were resuspended in HES buffer containing protease inhibitors.

Amine-reactive cell surface biotinylation

3T3-L1 adipocytes were starved for 2 to 3 h in serum-free DMEM/GlutaMAX supplemented with 0.2% BSA before starting the experiment. Following insulin treatment, cells were subsequently subjected to media removal and three washes with ice-cold PBS followed by a 30-min incubation with EZ-Link Sulfo-NHS-SS-Biotin (0.25 mg/ml in PBS, Thermo Fisher Scientific, Cat. #A39258) at 4 °C. Cells were washed twice with ice-cold tris-buffered saline (TBS, 20 mM Tris, 150 mM sodium chloride, pH 7.4) and the residual biotin reagent was blocked with 100 mM glycine in PBS. Next, the cells were washed twice with ice-cold TBS and harvested by scraping. Cells were lysed in RIPA lysis buffer (50 mM Tris–HCl, 150 mM NaCl, 0.2% SDS (v/w), 0.5% sodium deoxycholate (SDC), 1% Triton X-100, pH 8.0) containing 1× protease inhibitor (Roche Applied Science, Cat. #11873580001) and homogenized by sonication at 4 °C. After centrifugation at 15,000g for 15 min at 4 °C, clear cell lysate was kept at −80 °C.

The clarified cell lysate was added to the NeutrAvidin Agarose slurry gel (Thermo Fisher Scientific, Cat. #29200) and incubated for 3 h at room temperature (RT) with end-over-end mixing using a rotator. After centrifugation for 2 min at 500g, the centrifugate containing the nonbiotinylated membrane was collected. Then NeutrAvidin beads were carefully washed and subjected to the elution buffer (PBS, 6M urea) containing 100 mM dithiothreitol (DTT). After incubating at 65 °C for 30 min to cleave the disulfide bridge in the labeling reagents, the elutes containing biotinylated membrane were collected by centrifugation for 2 min at 1000g. The elution procedure was repeated once for complete elution.

Proteomic sample preparations

Sample preparations and proteomic analysis for published datasets DS1–3 and DS7 are reported in Prior et al. (7) and Conway et al. (19), respectively. Methods for DS4–6 and 8 to 9 are described below.

DS4: sample preparations for proteomics

Filter-aided sample preparation (FASP) and StageTip-based fractionation were performed following the protocol described by Wiśniewski et al. (103), with some modifications. Briefly, proteins were reduced with 1 mM DTT and alkylated with 5.5 mM iodoacetamide for 20 min in the dark. Samples were adjusted to a final volume of 400 μl using 8 M urea in 50 mM Tris–HCl (pH 8.5), transferred to 30-kDa molecular weight cutoff centrifugal filter units, and centrifuged at 14,000g for 10 min. The retentate was washed three times with 400 μl of 8 M urea in 50 mM Tris–HCl (pH 8.0), followed by buffer exchange through two washes with 50 mM ammonium bicarbonate (NH_4_HCO_3_). On-filter digestion was carried out by adding 120 μl of 50 mM NH_4_HCO_3_ containing trypsin at a ratio of 1 μg per 100 μg of protein, and incubating for 16 h at RT. Peptides were recovered by centrifugation at 14,000g for 15 min, followed by a second elution with 100 μl of 50 mM NH_4_HCO_3_ and centrifugation for 10 min. The resulting peptides were dried and reconstituted in a pH 11 buffer containing 20 mM each of acetic acid, phosphoric acid, and boric acid. Peptides were fractionated using strong anion exchange (SAX) StageTips. Peptides were loaded at pH 11 and sequential elution was carried out using buffer solutions (20 mM each of acetic acid, phosphoric acid, and boric acid) at pH 8, 6, 5, 4, and 2.5, respectively. Flow-through and each of the five pH eluted fractions were acidified to pH 2 to 3 with 0.5% trifluoroacetic acid (TFA) and subsequently desalted using C18 StageTips (3M Empore) (104). All procedures were conducted at RT unless otherwise specified.

DS5: sample preparations for proteomics

The sample preparation methods were adapted from Fazakerley et al. (73). Briefly, proteins were reduced with 10 mM DTT, alkylated with 25 mM iodoacetamide, and quenched with 10 mM DTT. Samples were diluted in NuPAGE LDS sample buffer (141 mM Tris base, 2% lithium dodecyl sulfate, 10% glycerol, 0.51 mM EDTA, 0.22 mM SERVA Blue G, and 0.175 mM phenol red) and separated on 8 to 12% bis/acrylamide SDS-PAGE Precast NuPAGE gels. Each lane was fractionated into 10 slices for in-gel digestion with trypsin (10 ng/μl in 50 mM triethylammonium bicarbonate (TEAB)) overnight at 37 °C. Peptides were solubilized by 5% formic acid, dehydrated using 100% acetonitrile, resuspended in 1% TFA, and centrifuged at 21,000g for 10 min at 4 °C. Each sample was desalted using C18 StageTips (3M, Empore) (104).

DS6, 8, and 9: sample preparations for proteomics

The sample preparation methods were adapted from Cutler et al. (105). Briefly, thawed eluates were mixed with 4% SDC and 100 mM Tris–HCl (pH 8.5), boiled at 95 °C for 10 min, and subjected to chloroform/methanol precipitation. Pellets were washed, centrifuged at 2000 g for 5 min, and resuspended in 4% SDC and 100 mM Tris–HCl (pH 8.5). Protein concentration was determined using a BCA assay (Thermo Fisher Scientific, Cat. #A558600). Proteins were reduced/alkylated with 10 mM tris(2-carboxyethyl)phosphine (TCEP) and 40 mM 2-chloroacetamide at 65 °C, 1000 rpm for 10 min. Water was added to dilute SDC to 1% and samples were cooled to RT. Samples were digested overnight (16–18 h) at 37 °C with trypsin and LysC. Digestion was quenched with 1% TFA in ethyl acetate. Peptides were cleaned for MS analysis by StageTip clean-up using styrene divinyl benzene reverse phase sulfonate (SDB-RPS) solid-phase extraction material (104).

Proteomic analysis

DS4–6: proteomic analysis

The proteomic analysis was based on Fazakerley et al. (73). The EASY-nLC II nanoHPLC (Proxeon) coupled to an LTQ-Orbitrap Velos Pro (Thermo Fisher Scientific) was used for proteomic analysis. An in-house packed 75-μm × 17-cm column (1.9-μm particle size, ReproSil Pur C18-AQ) was used for chromatographic separation. The mobile phase comprised Buffer A (0.5% acetic acid in water) and Buffer B (0.5% acetic acid in 90% acetonitrile). The HPLC gradient was 0 to 40% Buffer B over 90 min at a flow of 250 nl/min. A full-scan MS scan [300–1750 m/z; 3 × 10^6^ automated gain control] was recorded in the Orbitrap set at a resolution of 60,000 followed by data-dependent collision-induced dissociation MS/MS of the 20 most intense precursor ions. Parameters for collision-induced dissociation were as follows: normalized energy 35, dynamic duration 60 s, maximum injection time 150 ms, and tandem MS automatic gain control 4 × 10^5^.

Raw proteomic data files were searched using MaxQuant (106) using the Andromeda search engine (107) against the reviewed UniProt mouse proteome. Parameters included ≤2 missed cleavages, 20 ppm MS/MS tolerance, and ≥7 peptide length. The “match between runs” feature was enabled (2 min window). FDR thresholds for protein, peptide, and site were set to 1%. Label-free quantification was used to estimate relative protein abundances.

DS8–9: proteomic analysis

The proteomic analysis was based on Cutler et al. (105). The Dionex UltiMate 3000 RSLCnano LC coupled to a Q-Exactive HF-X mass spectrometer (Thermo Fisher Scientific) was employed for proteomic analysis. An in-house packed 75-μm × 55-cm column (1.9-μm particle size, ReproSil Pur C18-AQ) was used for chromatographic separation. The mobile phase comprised Buffer A (0.1% formic acid in water) and Buffer B (0.1% formic acid in 80% acetonitrile) at a 0.4 ml/min flow rate. The online LC separation followed a time-programmed elution with 250 ng of peptide loaded. In detail, samples were loaded to the column at 100% Buffer A for 12 min, before ramping to 19% Buffer B over 40 min, and then to 98% Buffer B over 20 min and held for 10 min. Eluting peptides were ionized by electrospray with a spray voltage of 2.4 kV and a transfer capillary temperature of 300 °C. The mass spectrometric analysis was performed in data-independent acquisition (DIA) mode with an MS1 resolution of 120,000, between charge/mass ratio (m/z) 350 and 1650. The isolation windows (widths of m/z 27–589) were described in Cutler et al. (105). Ions were fragmented with a higher-energy C trap dissociation collision energy at 25% and MS2 DIA spectra were collected between m/z 300 and 2000 at a resolution of 30,000, with an automatic gain control target of 3e^6^ and the maximum injection time set to automatic.

Raw proteomic data files were searched using DIA-NN using a library-free FASTA search against the reviewed UniProt mouse proteome with deep learning enabled (108). The protease was set to trypsin/P with one missed cleavage, N-terminal methionine excision, carbamidomethylation, and methionine oxidation options on. Peptide length was set to 7 to 30, precursor range 350–1650, and fragment range 300–2000, and the FDR was set to 1%.

siRNA-mediated gene knockdown in adipocytes

For siRNA delivery, 3T3-L1 adipocytes were reverse transfected using the TransIT-X2 (Mirus, Cat. #MIR6006) as previously described (12, 109). Briefly, siRNA targeting mouse SLC20A2 (M-062508–01; siGENOME SMARTpool; Dharmacon Horizons), SLC12A4 (M-044424-01; siGENOME SMARTpool; Dharmacon Horizons) or nontargeting control siRNA (D-001206-13; siGENOME SMARTpool; Dharmacon Horizons) was added to an Opti-MEM/TransIT-X2 mix (final concentration of siRNA added to cells = 50 nM), and incubated at RT for 30 min. Resuspended adipocytes in DMEM/FBS/GlutaMAX were added to the OptiMEM/TransIT-X2/siRNA mixes and immediately seeded into Matrigel-coated 96- (PerkinElmer Cell Carrier Ultra, Cat. #6055300) or 48-well plates for assessing GLUT4 translocation or 2-deoxyglucose uptake, respectively. Twenty-four hours after reseeding, cells were moved to lower media volumes (50 μl per well in 96-well plates, 125 μl per well in 48-well plate) for 48 h prior to assays to increase oxygen delivery to cells and maximize GLUT4 responses (110). Functional assays were performed 96 h following the transfection.

Real-time polymerase chain reaction analysis

To validate the efficiency of siRNA-mediated knockdown in 3T3-L1 cells, RNA was extracted using the RNeasy Mini Kit (Qiagen, Cat. #74104), complementary DNA (cDNA) was synthesized using M-MLV Reverse Transcriptase and for experiments on mature 3T3-L1 adipocytes, real-time polymerase chain reaction was performed using the TaqMan Gene Expression Assay (SLC12A4: Mm00486179_m1, SLC20A2: Mm00660203_m1) (Thermo Fisher Scientific). Gene expression was normalized to housekeeping genes (ACTB: Mm0434036_g1).

[3H]-2-deoxyglucose uptake

Cells were serum-starved for 2 h with 125 μl DMEM/GlutaMAX containing 0.2% BSA at 37 °C, 10% CO_2_. Following serum-starvation, cells were washed and incubated in 200 μl prewarmed Krebs-Ringer phosphate (KRP) buffer containing 0.2% BSA (KRP buffer; 0.6 mM Na_2_HPO_4_, 0.4 mM NaH_2_PO_4_, 120 mM NaCl, 6 mM KCl, 1 mM CaCl_2_, 1.2 mM MgSO_4_, and 12.5 mM Hepes (pH 7.4)) for 10 min. Adipocytes were then stimulated with 0.5 nM insulin for 20 min. To determine nonspecific glucose uptake, 25 μM cytochalasin B (in ethanol, Sigma-Aldrich, Cat. #C6762) was added to control wells before addition of 2-[^3^H]deoxyglucose (2-DG, Revvity Health Sciences, Cat. #NET549005MC). During the final 5 min, 2-DG (0.25 μCi, 50 μM) was added to cells to measure steady-state rates of 2-DG uptake. Cells were then moved to ice, washed with ice-cold PBS, and solubilized in PBS containing 1% (v/v) Triton X-100. Tracer uptake was quantified by liquid scintillation counting on the TriCarb 2900TR (PerkinElmer), and data normalized for protein content.

Plasmids preparations

Human full-length cDNAs encoding SLC12A4 (KCC1) and SLC20A2 (PIT2) were fused at the C terminus with an ALFA tag and the green fluorescent protein mStayGold via a flexible (GGGGS)3 linker. In addition, human full-length cDNA of SLC20A2 (PIT2) was tagged at the C terminus with an HA epitope. These constructs were cloned into the pcDNA3.1(+) vector via the NheI/EcoRI restriction sites and synthesized by GenScript Biotech (Singapore). Mouse cDNA of GLUT4 containing HA in the first exofacial loop was excised using XhoI and BamHI and ligated into the pMO91 vector containing mRuby3 to create the HA-GLUT4-mRuby3 (12).

Electroporation of 3T3-L1 adipocytes

3T3-L1 adipocytes were detached from the culture plates using 5x Trypsin-EDTA (Gibco, Cat. #15400–054) diluted in PBS 7 days post differentiation. The trypsin was neutralized by adding DMEM/FBS/GlutaMAX. The cells were then centrifuged at 150g for 5 min, the supernatant was removed, and the pellet was washed twice with PBS. The cell pellet was resuspended in 400 μl of electroporation solution (20 mM Hepes, 135 mM KCl, 2 mM MgCl_2_, 0.5% (w/v) Ficoll 400, 1% (v/v) dimethyl sulfoxide (DMSO), 2 mM ATP, and 5 mM glutathione, pH 7.4) and 4 to 6 μg of plasmid DNA. The mixture was transferred to a 0.4 cm electroporation cuvette (Bio-Rad, Cat#165–2091) and electroporated using an ECM 830 Square Wave Electroporation System (BTX Molecular Delivery Systems) at 250 V for 20 ms. Immediately after electroporation, 1.5 to 2 ml of DMEM/FBS/GlutaMAX was added to the cuvette, and the cells were seeded onto Matrigel-coated glass bottom dishes (Cellvis, Cat. #D35-28-1.5-N) or PhenoPlate 96-well microplates (Revvity, Cat. #6055300) for further experiments. Following attachment to the dishes or plates (∼1 h), the media were replaced with fresh DMEM/FBS/GlutaMAX.

Live cell TIRF microscopy

Cells were serum-starved for 2 to 3 h, washed once, and then incubated in Krebs–Ringer–phosphate–Hepes (KRPH) buffer (0.6 mM Na_2_HPO_4_, 0.4 mM NaH_2_PO_4_, 120 mM NaCl, 6 mM KCl, 1 mM CaCl_2_, 1.2 mM MgSO_4_, and 12.5 mM Hepes, pH 7.4). The buffer was supplemented with 10 mM glucose, 1×minimum Eagle's medium (MEM) amino acids (Gibco, Cat. #11130-051), 1×GlutaMAX, and 0.2% (w/v) BSA.

Live-cell imaging was conducted using a Nikon Ti-LAPP H-TIRF module with a CFI Apochromat TIRF 60 × oil objective (NA 1.49) or Apochromat TIRF 100 × oil objective (NA 1.49). Image acquisition was controlled by NIS-Elements Imaging Software (Nikon), with frames captured every ∼ 2 min. The TIRF module was angled to achieve an imaging depth of ∼90 nm into the cells. Temperature (37 °C), humidity (≥95%), and CO_2_ level (∼0%) were maintained via an Okolab cage incubator. Viability and overall cell health were assessed using brightfield and epifluorescence imaging. Standard imaging parameters included 2 × 2 binning and a 100 ms exposure time. Laser power was adjusted to minimize photobleaching and phototoxicity while maintaining adequate signal intensity. Treatments were administered through a custom perfusion system, with each addition mixed thoroughly (at least five times via syringing) to ensure rapid and uniform delivery.

Image analysis to quantify PM recruitment using TIRF microscopy was performed using Python 3.10. Cells were segmented using a human-in-the-loop augmented Cellpose-SAM model (https://doi.org/10.1101/2025.04.28.651001). Intensity and geometric features were measured using pyclesperanto (https://github.com/clEsperanto/pyclesperanto) and cell masks were tracked through time with trackpy 0.7 (https://soft-matter.github.io/trackpy/v0.7/). Code is available on request.

Induction of IR in 3T3-L1 adipocytes

The next day after electroporation, 3T3-L1 adipocytes were incubated with a single dose of one or 10 nM insulin for 24 h. Then cells were serum-starved for 2 to 3 h at 37 °C, 10% CO_2_ before the acute insulin (1 nM) stimulation.

Western blotting

Following stimulus, 3T3-L1 adipocytes were placed on ice, washed with cold PBS, and lysed in 2x Laemmli buffer (4% (w/v) SDS, 125 mM Tris–HCl (pH 6.8), 20% (v/v) glycerol, 0.2% (w/v) bromophenol blue) supplemented with 50 mM TCEP (Thermo Fisher Scientific, Cat. #77720). Lysates were collected by scraping, tip-probe sonicated, and clarified by centrifugation at 15,000g for 15 min at 4 °C. The lipid layer was removed, and supernatants were stored at −30 °C until SDS–PAGE.

Protein samples and Precision Plus Protein Kaleidoscope prestained molecular weight markers (Bio-Rad, Cat. #1610375) were resolved by SDS–PAGE and transferred to polyvinylidene difluoride (PVDF) membranes. Membranes were blocked and immunoblotted as described previously (73). Immunoblotting was performed using primary antibodies pan-AKT (Cell Signaling Technology, Cat. #2920), phospho-AKT (Thr308) (Cell Signaling Technology, Cat. #13038), SLC20A2 (PIT2) (Proteintech, Cat. #12820-1-AP), SLC12A4 (KCC1) (Proteintech, Cat. #15927-1-AP), GLUT4 (rabbit polyclonal antibody generated in-house), Caveolin 1 (CAV1) (Abcam, Cat. #ab17052), 14-3-3 (Santa Cruz Biotechnology, Cat. #sc-629), and either infrared dye 700- or 800-conjugated secondary antibodies (Thermo Fisher Scientific, Cat. #A32735 or A21036). Detection was carried out using an Odyssey CLx Imaging System (LI-COR Biosciences). Densitometry analysis was performed in Image Studio Lite version 5.2.5 (LI-COR).

Immunofluorescence staining for endogenous GLUT4 and TFR, HA-GLUT4-mRuby3, and PIT2-HA

The method was adapted from Diaz-Vegas et al. (12). Briefly, adipocytes were washed three times with ice-cold PBS supplemented with 1 mM calcium and 1 mM magnesium chloride (PBS^+/+^) on ice. Adipocytes were incubated with the exofacial GLUT4 antibody LM048 (1:100, Cell Surface Bio, Cat. #CSB0148) in 2% horse serum on ice for 2 h, and then cells were washed three times with ice-cold PBS^+/+^. Cells were subsequently fixed with 4% paraformaldehyde (PFA, ProSciTech, Cat. #C004) in PBS^+/+^ for 5 min on ice and for 15 min at RT. After fixation, cells were washed twice with PBS^+/+^ and incubated with 50 mM glycine in PBS^+/+^ for 5 min at RT. Secondary antibody (AlexaFluor 647-conjugated goat anti-human, Invitrogen, Cat. #A21445) diluted 1:800 and Hoechst 33342 (1:5,000, Thermo Fisher Scientific, Cat. #H3570) in 2% (w/v) horse serum were incubated for 1 h at RT. Secondary antibody was removed, cells were washed three times with degassed PBS^+/+^ and 50 μl/well of degassed imaging buffer was added (2.5% 1,4-diazabicyclo[2.2.2]octane and 10% glycerol, pH 7.8). Solutions with primary or secondary antibodies were filtered through a 0.22-μm syringe filter (Thermo Fisher Scientific, Cat. #SLGP033NB) before use. For intracellular staining, cells were fixed before the blocking step and permeabilized with 5% (w/v) horse serum containing 0.1% saponin. Anti-GLUT4 antibody (rabbit polyclonal antibody generated in-house) was used at 1 μg/ml and was detected using AlexaFluor 647-conjugated goat anti-rabbit secondary antibody diluted 1:800 (Invitrogen, Cat. #A21246).

The staining for TFR, HA-GLUT4-mRuby3, and PIT2-HA was performed using similar protocols described above. TFR was labeled using a monoclonal antibody against TFR (CD71) diluted 1:500 (Thermo Fisher Scientific, Cat. #14-0711-82) and stained with AlexaFluor 555-conjugated goat anti-rat secondary antibodies (Invitrogen, Cat. #A21434). HA-GLUT4-mRuby3 and PIT2-HA were labeled using HA.11 monoclonal antibody (16B12) diluted 1:100 (ENZO Life Sciences, Cat. #ENZ-ABS120-1000) and stained with AlexaFluor 488-conjugated goat anti-mouse secondary antibodies (Invitrogen, Cat. #A21121).

Confocal imaging and (co)localization analysis

Cells were washed three times with ice-cold PBS, fixed with 4% paraformaldehyde in PBS, and quenched with 50 mM glycine in PBS. For immunofluorescence analysis, cells were permeabilized with 0.1% (w/v) saponin in PBS and blocked with 5% (w/v) horse serum. Antibody staining was performed in PBS containing 2% (w/v) horse serum and 0.1% (w/v) saponin.

GLUT4 was detected using either rabbit polyclonal antibody (1:100; generated in-house) with AlexaFluor 555 goat anti-rabbit secondary (1:800; Invitrogen, Cat. #A21428) or mouse polyclonal antibody (1:100; generated in-house) with AlexaFluor 647 goat anti-mouse secondary (1:800; Invitrogen, Cat. #A21237). TFR (CD71) was labeled with monoclonal antibody (1:500; Thermo Fisher Scientific, Cat. #14-0711-82) and AlexaFluor 647 goat anti-rat (Invitrogen, Cat. #A21247) or AlexaFluor 488 donkey anti-rat (Invitrogen, Cat. #A48269). KCC1 was detected using polyclonal antibody (1:100; Proteintech, Cat. #15927-1-AP) with AlexaFluor 555 goat anti-rabbit (Invitrogen, Cat. #A32732). PIT2 was labeled with polyclonal antibody (1:200; Proteintech, Cat. #12820-1-AP) and AlexaFluor 555 goat anti-rabbit (Invitrogen, Cat. #A32732). PIT2-HA was detected using HA.11 monoclonal antibody (16B12) (1:100; ENZO Life Sciences, Cat. #ENZ-ABS120–1000) and AlexaFluor 488 goat anti-mouse (Invitrogen, Cat. #A32732).

Coverslips were mounted with ProLong Glass Antifade Mountant (Thermo Fisher Scientific, Cat. #P36910). Imaging was performed using a Nikon CrestOptics X-Light V3 microscope equipped with a 100 × 1.45 NA PLAN APO oil immersion objective. Z-stacks were captured at an optimal sampling density determined by the Nyquist frequency. Colocalization analysis of immunofluorescence images was performed with the software Imaris version 10.2 (Coloc Module). The degree of colocalization was quantified by calculating the percentage of channel voxels above the threshold that were colocalized.

Data analysis

Protein cellular localization information was analyzed with the GO database using the R programming environment (https://www.r-project.org/). Proteins annotated with “plasma membrane” or “cell surface” in GO-cellular component were classified as PM-associated regardless of whether they are distributed across multiple compartments. Transmembrane annotations were obtained from either UniProt https://www.uniprot.org/ (Accessed July 20, 2025) or predicted using DeepTMHMM 1.0 (21).

The analysis of insulin-regulated proteins within individual datasets was performed on datasets with n > 2. Statistical significance was defined as FDR-adjusted p < 0.05 based on an unpaired, two-tailed Student’s t test. To integrate all datasets, we calculated insulin-over-basal fold-change (FOB) for all proteins across all datasets and performed a robust meta-analysis on log2-transformed FOB values. Within each dataset, log_2_ fold changes were standardized using a robust Z-scoring procedure: for each protein, we subtracted the dataset median and divided by the MAD. To limit the influence of extreme measurements, values were constrained to fall within ±5 × MAD of the dataset median. Robust Z-scores were then combined across datasets using a weighted Stouffer method, where weights were proportional to the square root of each dataset’s effective sample size (22, 111). Final insulin-responsive PM proteins were defined as integral PM proteins detected in at least three datasets and significant at FDR < 0.05 in the meta-analysis.

Data visualization and statistical analysis were conducted using the R programming environment version 4.2.2 (https://www.r-project.org/), GraphPad Prism version 10.4.1 (GraphPad Software Inc.), or Adobe Illustrator version 29.8.1 (Adobe Inc.). Data are reported as mean ± standard error of the mean (SEM) unless stated otherwise. Relevant statistical tests are indicated in the figure legends. EC_50_ values were calculated by fitting data points from individual experiments to a three-parameter Hill equation using GraphPad Prism 10.4. Microscope images were processed using ImageJ (Fiji), with adjustments made to brightness and contrast as needed.

Data availability

The authors declare that all data supporting the findings of this study are available within the manuscript and its supplementary materials. Raw mass spectrometry proteomics data for datasets 8 and 9 have been deposited to the ProteomeXchange Consortium via the PRIDE (112) partner repository (accession: PXD068305; reviewer access token: vmgus5JDjRPj). Datasets 1 to 3 and 7 have been published previously (7, 19). For datasets 4 to 6, complete processed outputs are provided, including protein quantification tables and database search parameters.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.