Differential endocytosis and signaling dynamics of insulin receptor variants IR-A and IR-B

Insulin signaling comprises a complex cascade of events, playing a key role in the regulation of glucose metabolism and in cellular growth. Impaired response to insulin is the hallmark of diabetes, whereas upregulated insulin activity occurs in many cancers (Papa et al., 1993; Bailyes et al., 1997; Pandini et al., 1999; Saltiel and Kahn, 2001; Pandini et al., 2002). Insulin binds to specific receptors at the cell surface, triggering their phosphorylation and thereby activating downstream signaling pathways with concomitant receptor internalization and recycling. The long-term consequences of insulin binding result in increased cell growth and gene expression. However, how receptor dynamics and redistribution regulate the transport and metabolism of glucose, lipids and proteins is not understood in detail.

The insulin receptor (IR) belongs to a subfamily of receptor tyrosine kinases (RTK) that includes the insulin-like growth factor (IGF) I receptor (IGF-1R) and the insulin-receptor-related receptor (Ebina et al., 1985; Seino and Bell, 1989; Patti and Kahn, 1998). The receptors are tetrameric proteins consisting of two extracellular α-subunits and two transmembrane β-subunits linked by disulfide bonds.

The human IR is encoded by a single INSR gene comprising 22 exons. Alternative splicing of exon 11 (Seino and Bell, 1989) gives rise to two protein isoforms differing by a 12-amino-acid insert in the hormone binding domain at the C-terminus of the α-subunit: IR-A (lacking exon 11) and IR-B (with exon 11) (Ullrich et al., 1985; Ebina et al., 1985). The relative abundance of the two transcripts is regulated in a highly tissue-specific manner conserved among mammals (Goldstein and Dudley, 1990), suggesting that functional differences between the two isoforms might underlie the tissue-specific insulin action in vivo.

Although all cell types express both IR isoforms to various degrees, IR-A predominates in fetal tissues and cancer cells, and IR-B in adult differentiated cells. In embryos, IR-A promotes growth owing to its ability to bind insulin and IGF-II; in adults, IR-B is expressed predominantly in insulin-sensitive tissues regulating glucose homeostasis. Misregulation of IR alternative splicing is involved in some diseases and has important consequences for insulin and IGF-II sensitivity and responsiveness (Mosthaf et al., 1990; Mosthaf et al., 1991; Kellerer et al., 1993; Savkur et al., 2001; Sen et al., 2009).

IR-A has a higher affinity for insulin than IR-B. Furthermore, IGF-II binds to IR-A but not to IR-B, with an affinity close to that of insulin, whereas the affinity for IGF-I for both receptors is similar (Mosthaf et al., 1991; Yamaguchi et al., 1991; Yamaguchi et al., 1993; Frasca et al., 1999; Pandini et al., 2002; Benyoucef et al., 2007). Contradictory results were reported in mouse 3T3 cells using a targeted ablation of the IGF-1R gene (Denley et al., 2004). The fact that these studies were performed in different cell types, under different conditions and with diverse approaches could account in part for the discordant findings.

When IR and IGF-1R are coexpressed, the proreceptors can generate IR–IGF-1R hybrids. These hybrid receptors promote insulin resistance in type II diabetes by decreasing the number of insulin binding sites. A larger fraction of hybrid receptors also increases the number of binding sites for IGF-I and IGF-II and signaling by these growth factors in cancer (Pandini et al., 1999; Benyoucef et al., 2007).

Although several studies point to important biochemical features of the insulin signaling mechanism, few have explored these phenomena at the level of individual cells. We have generated tagged insulin and insulin receptors and used advanced imaging tools to study insulin binding and receptor internalization with a focus on the comparative properties of the two IR isoforms. High-speed optical sectioning confocal and structured illumination microscopy with a programmable array microscope (PAM) (Heintzmann et al., 2001; Hagen et al., 2007) revealed the early steps in ligand-dependent receptor activation and endocytosis. We demonstrated a higher initial rate of internalization of IR-A than IR-B and explored the potential influences on the signaling cascades linked to the IR. It is currently accepted that IR-A is more ‘mitogenic’ and IR-B more ‘metabolic’, yet the underlying molecular mechanisms are unclear. We examined certain key points in the insulin pathway: the initial IR activation (autophosphorylation) and ERK activation, a key downstream effector of gene expression. Increased signaling from internalized receptors in endosomes is correlated with increased mitogenicity. In the endosomes, the IR with bound ligand is still activated and can phosphorylate downstream effectors (Di Guglielmo et al., 1994; Pol et al., 1998; Rizzo et al., 2000; Sorkin et al., 2000; Jensen and De Meyts, 2009). Thus, the mitogenic effect of insulin might result from IR signaling from endosomes, whereas the metabolic response of insulin might proceed from IR on the plasma membrane. In view of our microscopy-based evidence for differential internalization, we performed experiments that showed faster endocytosis as well as more persistent and higher activation levels for IR-A than IR-B, reflected in the triggering of the MAPK pathway and ERK1/2 phosphorylation profiles. In addition, IR-A stimulated a higher rate of AP-1 gene transcription whereas IR-B showed higher AKT phosphorylation after insulin stimulation than cells expressing IR-A.

Mammalian expression vectors were generated by fusing IR-A or IR-B to different visible fluorescence proteins (VFPs): eGFP or new variants (Kremers et al., 2006) of CFP (SCFP-3AC) and YFP (SYFP-2F) at the C-terminus of the receptor (IR-GFP, IR-SCFP, IR-SYFP). Specific antibodies against the N-terminus of the IR α-subunit were used in order to evaluate the correct expression and localization of IR-VFPs by immunofluorescence (data not shown). We confirmed correct expression, membrane localization and ligand binding as well as activation, autophosphorylation and internalization after insulin stimulation of the IR chimeras in transfected HeLa cells by microscopy and/or western blots (WBs). After establishing that the transgenes were properly expressed and inserted into the plasma membrane, we determined whether isoforms A and B were colocalized by cotransfecting cells with pcDNA3-IR-A-SCFP and pcDNA3-IR-B-SYFP or with pcDNA3-IR-A-SCFP and pcDNA3-IR-A-SYFP (Fig. 1D,E). Colocalization was quantified by calculation of the Manders coefficient (M) and the Pearson coefficient (Rr). We plotted the product of the differences from the mean (PDM), applying an intensity correlation analysis plug-in (Fig. 1D,E). In addition, frequency scatterplots were computed (Fig. 1D,E). No significant differences in the coefficients for the two pairs were observed. HeLa cells coexpressing IR-A-SCFP and IR-B-SYFP led to Rr=0.90±0.10, M=0.95±0.04, M_CFP=0.86±0.04 and M_YFP=0.82±0.04 (n=26). In cells coexpressing IR-A-SCFP and IR-A-SYFP, the coefficients were Rr=0.91±0.05, M=0.95±0.04, M_CFP=0.87±0.06 and M_YFP=0.85±0.04 (n=16).

Biotinylated ligands in combination with streptavidin-conjugated quantum dots (svQDs) provide valuable tools for studying the dynamics of RTKs in living cells, including the characterization of the internalization process and retrograde transport (Lidke et al., 2004; Lidke et al., 2005; Lidke et al., 2007; Cambi et al., 2007; Echarte et al., 2007; Andrews et al., 2009). The combination of the expression of IR-VFPs, binding to the IR of BAC-Ins and its detection with QDs constituted a sensitive monitor of insulin–IR complexes in membranes and on filopodia of HeLa cells. A PAM provided the sensitivity and speed required to visualize individual QDs bound to BAC-Ins and to track them on the plasma membrane and filopodia of living cells expressing IR-A-GFP (Fig. 1A; Fig. 2C). Confocal microscopy was used on fixed cells expressing either IR-B-GFP (Fig. 1B,C) or IR-A-GFP. Binding of QDs was undetectable in the absence of BAC-Ins. For tracking experiments, 50 pM QDs was applied to cells that were exposed to BAC-Ins shortly before QD addition. Fig. 2C shows QD binding during the first 36 seconds after QD addition (supplementary material Movie 1).

HeLa cells expressing IR-GFP (A and B) or IR-B were stimulated over 10 minutes with 100 nM of recombinant human insulin (rhIns) or BAC-Ins after starvation. The activity of the receptors was assayed by immunofluorescence using a specific antibody against phosphorylated IR (Fig. 3A,B; Fig. 4A). Phosphorylated receptor was detected in plasma membranes of IR-A-GFP, IR-B-GFP and IR-B transfected cells, whereas untransfected and non-stimulated cells were negative (Fig. 3A,B; Fig. 4A, respectively). The β-chain was activated in the Golgi of quiescent cells, reflecting the non-assembled receptor. Untransfected cells in the same field served as the controls for non-specific immunostaining (Fig. 3A,B).

The membrane signal reporting activation levels of IR was quantified for IR-B (Fig. 4B). The activation induced by BAC-Ins did not differ within experimental error from the phosphorylation induced by rhIns (P>0.05, n=5–7; Fig. 4B). In order to evaluate if expression levels had an effect on activation, the ratio between the red (activation level) and green (IR-GFP expression level) signals in the membrane was calculated for each cell. The differences in the activation of the two isoforms were not statistically significant (data not shown).

WB analysis showed that IR and IR-VFPs were equally activated by rhIns (Fig. 3C,D) and by BAC-Ins (Fig. 4C,D). Insulin binding and subsequent IR autophosphorylation trigger mobilization of adaptor proteins (White, 1998; Sebastian et al., 2006). Subsequently, insulin-induced signals propagate through multiple interacting branches, including the MAPK cascade downstream of Ras and the PI3K/AKT pathway (Borisov et al., 2009). Accordingly, we evaluated relevant downstream signaling events in the IR cascade (Fig. 3F). Activation of ERK1/2 was evaluated by WB in order to test the correct functionality of the chimeras. Cells expressing the different chimeras showed that stimulation by 100 nM rhIns or BAC-Ins triggered ERK1/2 phosphorylation (Fig. 3C,E; Fig. 4C,E).

Binding and endocytosis of QD–insulin by IR was assessed by confocal microscopy, before and after short-term acid treatment to remove the acid-labile ligand complex from surface membrane receptors (Haft et al., 1998; Lidke et al., 2004). This procedure demonstrated that the ligand–QD complex was exclusively localized at the plasma membrane after binding and only internalized after subsequent incubation at 37°C.

Cells expressing IR-B, IR-A-GFP and IR-B-GFP were incubated with BAC-Ins followed by QD655, incubated at 37°C and either fixed or treated with acid for 5 minutes before fixation. Confocal microscopy revealed QD binding to the plasma membrane and filopodia at the initial time (0 minutes at 37°C) either in cells expressing IR (supplementary material Fig. S1A,E) or both isoforms of IR-GFP (Fig. 1B,C; Fig. 5A,E), with increasing levels of internalization at later time points of 20, 40 or 150 minutes at 37°C (Fig. 5B,C,E–J; supplementary material Fig. S1B–D,E–H). No significant binding or internalization was detected in untransfected cells, which was as expected because HeLa cells express <5000 endogenous IR per cell (McKeon et al., 1990).

Surface labeling and internalization were recorded by live-cell imaging with a PAM. Before internalization, induction labeling was observed exclusively in the plasma membrane and filopodia. By contrast, QDs were detected inside the cells, colocalizing with GFP after longer incubation periods (Fig. 5K).

The trafficking of BAC-Ins–QD655 was assessed by colocalization with caveolae and early endosomal and lysosomal markers in cells expressing IR-A. At the onset, the BAC-Ins–QD655 complexes colocalized exclusively with caveolin domains in the membrane (supplementary material Fig. S2A). After 20 minutes, the level of colocalization between QD655 and early endosomes increased (supplementary material Fig. S2B,C), whereas at 40 minutes, the QD655 signal was located in perinuclear lysosomes (supplementary material Fig. S2F).

Cells expressing high levels of IR-GFP showed less QD–insulin-IR uptake than cells with lower expression levels. Cells expressing IR-B and exhibiting high levels of QD–ligand binding demonstrated the same tendency. To clarify these observations, cells were classified according to their expression levels: high levels of IR-GFP expression (GFP/cell size >25 cts/pixel) or low levels (GFP/cell size <20 cts/pixel). After 20 and 40 minutes at 37°C, cells with high levels of IR-GFP expression had QDs predominantly on the surface with little internalization (supplementary material Fig. S3A–C). To quantify the effect of expression on insulin-induced internalization, we calculated a ratio between the signal from the BAC-Ins–QD655 complex for the cytosol (QD_i) and that of the plasma membrane (QD_m) plus the cytosol (QD_t=QD_i+QD_m) using image-processing tools on a cell-by-cell basis for both types of cells at different time-points. The differences between the two groups were observed mainly at the initial stage of the endocytosis process. After 20 minutes, cells with low IR-GFP expression levels displayed higher internalization (60±13%, n=13) than cells with high expression levels (35±7%, n=19); the differences were statistically significant (P<0.001; supplementary material Fig. S3D). The effects of receptor abundance decreased with incubation time, leading to smaller differences between the two groups at 40 minutes (high expression, 62±13%, n=8; low expression, 71±13%, n=18) and 150 minutes (high expression, 90±6%, n=14; low expression, 92±4%, n=8).

The series of images acquired during internalization allowed a quantification of QD endocytosis on a cell-by-cell basis, comparing the behavior of each IR isoform. We performed the same type of analysis as described above on cells expressing IR-B, quantifying QD internalization after 20 minutes (48%±7%, n=17), 40 minutes (64%±18%, n=39) and 150 minutes (81%±10%, n=23). Internalization rates at each time-point were significantly different (P<0.001; Fig. 6B), defining a distinctive temporal pattern. We tested these results using fluorescent bovine insulin (FITC-Ins), performing similar experiments but in a single step. Quantification of FITC-Ins internalization (20 minutes: 46%±5%, n=12; 40 minutes: 60%±14%, n=12; and 150 minutes: 75%±12%, n=15) showed no significant differences to that of the BAC-Ins–QD655 internalization (Fig. 6B). These results demonstrate that QD did not interfere with the endocytosis process and provided the advantage of a stronger, more stable signal.

No significant differences were found in QD internalization in cells expressing IR-B or IR-B-GFP at 20 minutes (44±14%, n=21), 40 minutes (67±13%, n=22) or 150 minutes (87±8%, n=9) (Fig. 6C). We conclude that VFPs did not affect the uptake of insulin.

Next, we compared the internalization of both isoforms. At early times (20 minutes), the analysis demonstrated a higher rate of internalization for IR-A-GFP (65±11%, n=10) than for IR-B-GFP (47±13%, n=16; Fig. 6D). These differences were statistically significant (P≤0.001). However, at longer incubation periods [40 minutes (IR-A-GFP: 77±16%, n=3; IR-B-GFP: 70±12%, n=17); 150 minutes (IR-A-GFP: 92±3%, n=5; IR-B-GFP: 87±10%, n=6)], the differences were no longer significant, although there was a higher tendency for IR-A internalization.

The differences observed in the rate of internalization of the receptor isoforms prompted an investigation of downstream signaling. We measured IR phosphorylation after ligand binding. WB experiments were performed with cells expressing IR-A, IR-B or both together, stimulated with 50 nM rhIns for 2, 5 or 15 minutes (Fig. 7A,B). IR-A showed higher and sustained activation levels than cells expressing IR-B or both isoforms together (fold-induction at 2, 5 and 15 minutes, respectively: for IR-A, 1.8±0.4, 1.6±0.3, 1.8±0.4; for IR-B, 1.5±0.3, 1.3±0.4, 1.2±0.2; for both isoforms, 1.6±0.3, 1.3±0.3, 1.1±0.3). Activation of IR after 15 minutes was statistically different (P≤0.05) for cells expressing IR-A (n≥4).

To evaluate the signaling of the IR cascade further downstream, we assayed ERK1/2 phosphorylation. Cells expressing IR-B showed a clear maximum of ERK1/2 activation after 5 minutes of stimulation (fold-induction: 2.4±0.7 at 5 minutes, 1.8±0.6 at 2 minutes and 1.3±0.4 at 15 minutes), whereas activation in cells expressing IR-A peaked earlier: 2.7±0.8-fold at 2 minutes and 2.7±1.5-fold at 5 minutes. In the case of coexpression of both receptors, the temporal profile of activation was similar to that of IR-B alone, although the maximal activation was somewhat higher (2.9±0.9 at 5 minutes; P<0.05) than with either isoform alone. These results demonstrate that ERK1/2 activation occurs more rapidly and is sustained longer upon stimulation of IR-A (Fig. 7C–E).

MAPK activation state is crucial for AP-1-dependent gene expression (Pastore et al., 2005). To evaluate the downstream relevance of the differences detected between IR isoforms, we measured AP-1 gene transcription using a reporter gene. The AP-1-dependent reporter plasmid (AP-1-luc), containing seven tandem copies of the AP-1 site, was used to transfect cells in combination with pcDNA3-IR-A, pcDNA3-IR-B or the empty vector (EV). Induction of luciferase activity was measured after 16 hours of stimulation with 100 nM rhIns in each case (Fig. 7F). AP-1 activity was higher when stimulating IR-A than IR-B and the difference was statistically significant (P=0.004, n=6).

The next step was to evaluate the activation of the metabolic pathway by IR, which regulates and maintains glucose homeostasis in response to insulin (Fig. 7J). AKT activation was measured by phosphorylation of serine 473 in cells coexpressing AKT fused with the HA epitope (AKT-HA) and either IR-A, IR-B or EV, stimulated with 100 nM rhIns. IR-B induced a higher AKT phosphorylation than IR-A (fold-induction at 2, 5 and 15 minutes: for IR-B, 5.8±2.3, 6.8±0.9, 5.1±2.1; for IR-A, 4.1±1.4, 4.1±0.5, 3.3±1.1). The differences were statistically significant for 5 minutes (P<0.001, n=5; Fig. 7G,H). In addition, AKT activation was monitored by wide-field microscopy in cells coexpressing AKT-HA and either IR-A or IR-B and stimulated or not (control) with 100 nM rhIns over 5 minutes at 37°C. These results were consistent with those of the previous WBs (Fig. 7I).

To evaluate the ligand-receptor complex and its dissociation after internalization, we analyzed the colocalization of BAC-Ins–QDs bound to IR-GFP over time. At the onset of endocytosis, BAC-Ins–QDs localized at the cell surface (Fig. 8A), redistributing later to the perinuclear region (Fig. 8B). We evaluated colocalization across the cell body by superimposing the two signals over linear trajectories (Fig. 8D). After 1 hour, BAC-Ins–QD655–IR-GFP complexes colocalized in endosomes (Fig. 8H).

To quantify colocalization using M coefficients, cells were classified into four groups based on the extent of the endocytosis: S0–S3 (Fig. 8H). The analysis of M revealed differences between S0 (0.90±0.03) and both S2 (0.83±0.05) and S3 (0.79±0.07), with a P<0.001. Significant differences were also observed between S1 (0.86±0.05) and S3 (0.79±0.07; P<0.05). S0 and S1 showed no significant differences in their colocalization coefficients (Fig. 8G).

Manders maps showed the contribution of each pixel to the global M. At the onset of endocytosis, higher contributions were in the membrane but this contribution disappeared in S3 cells, leading to maximum contributions at perinuclear regions (Fig. 8H). At intermediate steps, colocalization persisted in the membrane but there were other loci of association, probably in early endosomes. At S3, colocalization was low, indicative of a dissociation of the complexes in late endosomes or lysosomes (Fig. 8H).

Generation of the mammalian expression vectors with IR-A and IR-B fused to eGFP, SCFP or SYFP in combination with biotinylated insulin and QDs allowed the direct visualization of ligand-bound and activated IR in membranes and filopodia of living cells. Quantitative imaging provided a tool for studying the early steps of endocytosis of receptor-ligand complexes in a novel way. Coexpression of both isoforms fused to different VFPs revealed colocalization of IR-A and IR-B. The M and Rr coefficients were similar to those calculated when the same isoform was coexpressed fused to the VFPs SCFP and SYFP. These results are in agreement with previous reports demonstrating (or assuming) the existence of IR-A and IR-B heterodimers (Pandini et al., 1999; Benyoucef et al., 2007; Blanquart et al., 2008). By contrast, Uhles et al. (Uhles et al., 2003) reported experiments on pancreatic beta cells using fluorescence resonance energy transfer, according to which IR isoforms localized to different membrane domains, abrogating the formation of heterodimers and promoting isoform-specific IR signaling.

Cells incubated with the preformed BAC-Ins–QD655 complex (formed with different BAC-Ins:QD655 ratios) showed no labeling. By contrast, biotin insulin (lacking a tether) bound to cells was recognized by fluorescent streptavidin but not by QDs, probably owing to steric hindrance. These results are consistent with the binding model proposed by Ward et al. (Ward et al., 2007).

Either rhIns or BAC-Ins induced receptor autophosphorylation in the membrane, a hallmark of insulin activity. We also observed receptor autophosphorylation in the ER and the Golgi both in starved transfected cells and after insulin addition, probably owing to the inherent activity of the β-subunit prior to IR assembly and dimerization (Patti and Kahn, 1998). BAC-Ins not only induced IR autophosphorylation but also triggered the insulin signaling cascade as revealed by ERK1/2 and AKT activation.

The endocytosed IR-insulin complex visualized at different time-points appeared at the plasma membrane, in endosomes close to the membrane or in perinuclear regions. In the latter case, the insulin-IR complexes dissociated.

Cell-by-cell analysis revealed that high levels of IR-GFP expression led to a slower rate of internalization compared with cells with low expression, a result not available from prior studies based on biochemical and radiolabeled ligand determinations of bulk cell populations. We quantified the internalization properties of each IR isoform and demonstrated a higher initial rate of internalization of IR-A-GFP than IR-B-GFP, concordant with other reports using different experimental approaches (Vogt et al., 1991; Yamaguchi et al., 1991). It is possible that the differences between the endocytic properties of the two IR variants are tissue-specific and potentially important in the regulation of IR recycling or downregulation in vivo.

Impaired receptor internalization has been described in mutants affecting the autophosphorylation (Reynet et al., 1990; Wilden et al., 1990) or altering the intracellular β-subunit, in particular the juxtamembrane region (Backer et al., 1990; Thies et al., 1990) or ATP binding site (McClain, 1987; Hari and Roth, 1987; Russell et al., 1987; Yamamoto et al., 1990). The differences in the rate of internalization of the two IR splicing variants raise the question as to whether the phenomenon could be influencing other steps of the signaling pathway. In view of the well-accepted view that IR-A is more mitogenic than IR-B, we evaluated three points of this pathway: the initial step (IR autophosphorylation), an intermediate step (ERK1/2 phosphorylation) and a late mitogenic effect (gene transcription via AP-1). We observed that IR-A showed a greater and more persistent activation than IR-B, although cells coexpressing both isoforms also showed high activation, albeit more transient in nature. With respect to ERK1/2 activation, there were significant differences consistent with differential endocytosis. Whereas IR-B-induced ERK1/2 activation peaked at 5 minutes, IR-A led to more activation at 2 minutes than the B isoform. The differences between the signals after 2 minutes and 5 minutes of stimulation were larger in the case of cells coexpressing IR-A and IR-B. In endosomes, ligand-bound IR can still phosphorylate downstream effectors (Sorkin et al., 1993; Di Guglielmo et al., 1994; Bergeron et al., 1995; Grimes et al., 1996; Pol et al., 1998; Rizzo et al., 2000; Sorkin et al., 2000), prompting the hypothesis that the mitogenic effect of insulin might originate from IR signaling from endosomes.

In addition, blocking IR internalization led to inhibition of SHC phosphorylation and activation, whereas no effect was detected on IR autophosphorylation or IRS1 and AKT kinase phosphorylation and activation (Biener et al., 1996; Ceresa et al., 1998; Hamer et al., 2002), suggesting that these signaling elements can be fully activated by IRs at the surface. That is, the metabolic response of insulin appears to be fully activated from IRs on the cell surface. Our results are consistent with these interpretations: the isoform (IR-A) that is activated in a higher and more persistent manner (measured as receptor autophosphorylation) by insulin internalizes more readily, triggering the mitogenic pathway and inducing the ERK1/2 cascade earlier and stimulating gene transcription through the AP-1 pathway to a greater degree. By contrast, IR-B internalization is slower, allowing greater signaling from the membrane via AKT activation, an important step in the regulation of glucose metabolism. In conclusion, we identify the internalization dynamics as a (the) possible mechanism responsible for the differential signaling of the two splice variants of IR.

The rhIns was kindly provided by Laboratorios Beta (Buenos Aires, Argentina). Bovine BAC-Ins and FITC-Ins were from Sigma. Monoclonal mouse anti-PY99, polyclonal rabbit anti-ERK2 and polyclonal rabbit anti-AKT-P (Ser473, for WBs) were from Santa Cruz Biotechnology (Paso Robles, CA). Monoclonal rabbit anti-(phosphorylated IR-β-subunit) (Tyr1361), monoclonal mouse anti-IR-β-subunit, polyclonal rabbit anti-phosphorylated (p42/44) MAPK and monoclonal rabbit anti-AKT-P (Ser473, for immunofluorescences) were from Cell Signaling Technology. Anti-HA was from Covance (Emeryville, CA). The QD655 conjugated with streptavidin was from Invitrogen. Anti-CD63 (rabbit) was from Santa Cruz. Caveolae and EEA1 markers (rabbit) were from Abcam (Cambridge, UK).

The pCMV-AKT-HA vector was provided by Anabella Srebrow, Laboratorio de Fisiología y Biología Molecular (LFBM), FCEN, UBA, Argentina. The pCMV-β-galactosidase and the plasmid containing luciferase reporter gene downstream of seven binding sites for AP-1 were provided by Omar Coso (LFBM).

IR-A and IR-B: cDNA of IR-A or IR-B were provided by Axel Ullrich (MPIbpc, Germany) and amplified by PCR using primers containing the HindIII restriction site (IR forward: 5′-aagcttatggccaccgggggccgg-3′) and NheI and XbaI sites (IR reverse: 5′-tctagagctagcgaaggattggaccgaggcaaggtc-3′). The products were digested with HindIII and XbaI enzymes and cloned into the pcDNA3 vector.

IR-GFP, IR-SCFP and IR-SYFP: peGFP-C3 (Clontech) or pSCFP-3AC or pSYFP-2F (provided by T. W. Gadella Jr, University of Amsterdam, The Netherlands) were digested with NheI and ApaI and cloned into pcDNA3-IR-A and pcDNA3-IR-B.

Cell culture reagents [Dulbecco's modified Eagle's medium (DMEM), Glutamax, Optimem and FBS] were from Gibco (Grand Island, NE). HeLa cells were maintained in DMEM supplemented with penicillin, streptomycin and 10% FBS at 37°C in 5% CO₂.

HeLa cells were plated at 1×10⁵ cells per well in 24-well plates and onto 12 mm glass coverslips (microscopy) or at 2.5×10⁵ cells per well in 12-well plates (WBs) 1 day before transfection in DMEM and 10% FBS without antibiotics. Cells were transfected with Lipofectamine Reagent 2000 (Invitrogen, Carlsbad, CA) over 5 hours. After transfection, cells were cultured for further expression in complete medium. For AKT experiments, we used 0.2 μg AKT-HA and 0.6 μg IR per well.

Following stimulation with 50 nM rhIns (stimulation profiles experiments for 2, 5 or 15 minutes) or 100 nM rhIns or BAC-Ins (other experiments for 10 minutes), cells were lysed in buffer containing 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 1 mM EGTA, 150 mM NaCl, 10% glycerol, 1% NP40, 1 mM MgCl₂, 0.1% SDS, 2 mM Na₃VO₄, 40 mM sodium β-glycerophosphate, 1 mM DTT, 1 mM NaF and a protease inhibitor cocktail (Roche) and centrifuged for 10 minutes at 13,000 g at 4°C. The supernatants were mixed with loading loading buffer (50 mM Tris-HCl, pH 6.8, 2% SDS, 0.1% Bromophenol Blue, 10% glycerol and 100 mM β-mercaptoethanol) and heated for 5 minutes at 100°C. Following 10% SDS-PAGE and transfer, membranes were blocked in 5% w/v non-fat dried milk in 0.1% Tween-TBS buffer (TTBS) for 1 hour, washed and incubated overnight at 4°C with primary antibodies diluted in 5% BSA in TTBS [anti-IR β-subunit, 1/500; anti-ERK, 0.4 μg/ml; anti-phosphorylated (p44/42) MAPK, 1:500; anti-HA, 1:1000] or in 2% BSA in TTBS (anti-PY20, 2 μg/ml; anti-AKT-P, 0.4 μg/ml). The following day, membranes were incubated with secondary antibodies for 1 hour and washed. The blots were developed by chemiluminescence using a Bio-Imaging Analyzer Bas-1800 II and Image Gauge 3.12, Fujifilm (FCEN, UBA, Argentina). Quantification was performed using Image J (NIH), measuring the optical density of the blotted membrane bands. For IR experiments, values for stimulated samples were normalized with the corresponding value without stimulation and this was indicated as fold-induction. For ERK and AKT phosphorylation, the signals were normalized by an average of signals from unstimulated cells. In all the experiments, we controlled loading with the total blotting (IR, ERK and HA.AKT) because all of the samples were from an equal initial number of cells. The results were expressed as the mean of at least three independent experiments (±s.e.m.). P values were estimated using two-tailed Student's t-tests.

Cells seeded onto 12-well plates (2×10⁵ cells/well) the day before were transfected using 0.6 μg pcDNA3-IR or EV, 0.1 μg AP-1-luc and 0.1 μg pCMV-β-galactosidase (transfection control). After 24 hours, cells were starved for one day and then stimulated with 100 nM rhIns over 16 hours. Luciferase activity was determined using luciferase reactive and reporter lysis buffer (Promega, Madison, WI) and normalization to the control (non-stimulated cells) was performed (fold-induction). Protein concentration and β-galactosidase activity were measured as controls. The results were expressed as the mean of at least three independent experiments (± s.e.m.).

Before the experiment, cells expressing IR-B, IR-A-GFP or IR-B-GFP and starved overnight were washed with Tyrode's buffer at room temperature (RT) and incubated with 50 nM BAC-Ins or FITC-Ins for 15 minutes at 15°C or RT, washed with Tyrode's buffer and incubated with 50 pM or 1 nM QD655 (in the case of BAC-Ins) depending on the experiment for 10 minutes at 15°C or RT, and washed and either fixed in 3.7% PFA on ice for 20 minutes or incubated at 37°C for different periods of time before fixation. When acid treatment was applied, cells were incubated for 5 minutes at RT with 0.1 M Na-glycine, pH 3 and 0.5 M NaCl.

After overnight starvation, transfected cells were stimulated with 100 nM rhIns or BAC-Ins for 10 minutes (IR activation) or with 100 nM rhIns for 5 minutes (AKT activation), washed with PBS and immediately fixed in cold methanol for 30 minutes at −20°C, blocked with PBS with 0.3% Triton X-100 and 1% BSA for 1 hour at 37°C and incubated with antibodies against phosphorylated IR-β-subunit (0.3 μg/ml), PY99 (1 μg/ml) or AKT-P (1/1000) overnight at 4°C. The following day, samples were incubated with a secondary antibody conjugated to Cy3 or Alexa Fluor 555 for 1 hour at 37°C and washed. IR activation was monitored by confocal microscopy and AKT activation by wide-field microscopy.

Cells expressing IR-A-GFP were starved overnight and labeled as we described. For imaging of QD binding, the cells were incubated with 50 pM QD655 during image acquisition. For internalization experiments, we used 1 nM QD655 at RT.

Confocal laser scanning microscopy was performed with a Zeiss LSM 510 Meta System using a C-Apochromat 63× 1.2 NA water-immersion objective or a Plan-Apochromat 63× 1.4 NA oil-immersion objective. Excitation and filters were as follows: GFP, argon ion excitation, 488 nm, emission 500/20 nm BP filter; QD655, excitation 488 nm and 458 nm, emission >650 nm; Cy3, DSSP laser excitation 532 nm, emission 563–606 nm BP filter.

For experiments with IR-SCFP and IR-SYFP and assays with endocytosis markers, a confocal Olympus Fluoview FV1000 microscope was used with a PLSAPO 60× 1.35 NA oil-immersion objective and a UPLSAPO 60× 1.2 NA water-immersion objective. Excitation and emission filters were as follows: CFP, excitation 458 nm, emission 475/25 nm filter; YFP, excitation 515 nm, emission 530/10 nm BP filter; Alexa555, excitation 543 nm; emission 560–620 nm BP filter; QD655, excitation 405 nm; emission 585–685 nm BP filter.

Wide-field microscopy for AKT experiments was performed with an Olympus IX71 microscope using a 40× 1.15 NA water-immersion objective, with mercury arc lamp excitation and filters suitable for CFP and Alexa555. The camera used was a Hamamatsu Orca CCD C4742-95.

PAM experiments were performed with an LCoS-based system (Hagen et al., 2009) with the following parameters: light source, argon ion 488 nm; sensor, Andor emCCD camera, Ixon DV897-BV, 512×512 pixels, −87°C; exposure, 16 milliseconds, 4× average (for some internalization experiments) and 50 milliseconds, single exposure (for QD binding experiments and some internalization experiments); objective, 60× 1.2 NA water immersion; images, 158 nm/pixel in xy, 0.5 μm z-optical sections; filters, GFP – 525/50 nm BP, QD – 655/40 nm BP.

Confocal microscope and PAM images were processed with Matlab and Image J. In confocal images, the background of each channel (median or mean of empty region) was subtracted and, in some cases, a median filter was applied (radius: 1 pixel) only for presentation. No filter was applied in quantitative analyses.

In the first PAM image (Fig. 1A), we made a maximum-intensity projection from a stack of images after removing the background and applying thresholds for GFP (GFP>40 cts) and QD (GFP>600 cts) signals only for presentation. QD images were multiplied by a factor of two and a median filter (radius: 1 pixel) was applied to the GFP image. The latter was corrected for photobleaching by normalizing the signal of each frame to its maximum value before taking the maximal intensity projection from the stacks. 3D reconstruction of PAM images was performed with Matlab and the Image J plug-in.

Wide-field images were processed with Image J for presentation; the background of each channel, evaluated in regions lacking cells, was subtracted from each foreground value. The z-stacks (80 frames, 0.2 μm step size) were deconvolved using Huygens Professional Scientific Volume Imaging version 3.6 software, applying a quick maximum likelihood estimation algorithm (QMLE).

Cells were classified according to the stage of endocytosis: S0 (internalization not induced and QD signal at the plasma membrane); S1 (QD signal detected very close to the membrane); S2 (more internal red signals but QDs not already in the perinuclear region); and S3 (QD signal predominantly in perinuclear regions).

Colocalization during endocytosis was analyzed by introducing a new mask defined as the pixels for which the QD signal was >2× the mean of the QD signal in the entire cell mask. After that, the Manders coefficient was calculated as described (only cells with a mean GFP signal >500 cts and <5500 cts were considered).

For quantitative internalization experiments, labeling was done at 15°C to minimize endocytosis. We defined t=0 as the end of labeling washings and t=20, 40 and 150 minutes were periods of incubation at 37°C.

Channel backgrounds (median) were subtracted. In QD images, bleed-through contributions were subtracted. Segmentation was performed for each cell using the GFP signal (IR-GFP experiments) or differential interference contrast (DIC) images (IR wild-type experiments). After cell segmentation, the pre-membrane was defined as the difference between the image of the cell and a binary erosion (iterations: 5–20; alternating connectivity); we evaluated the results visually. The pre-interior was defined as the difference between the cell and the pre-membrane. A QD_mask was marked by red pixels. With this mask, the membrane was defined as the product of the QD_mask and the pre-membrane, and the interior was defined as the product of the QD_mask and pre-interior obtained previously.

Values in the membrane and the interior were summed for the red and green channels and sizes were also measured. To compute the relative amount of internalized red fluorescence, we estimated QD_total (QD_t) as the sum of QD_membrane (QD_m) and QD_interior (QD_i) and calculated for each cell the ratio QD_i/QD_t. This ‘internal calibration’ approach was chosen to remove the influence of the amplifier gain and the zoom factor for each image acquisition condition. The expression levels were estimated as the mean of the GFP signal (sum of GFP divided by cell size). The equivalent procedure was applied to images based on fluorescein signals.

HeLa cells expressing IR-A were labeled with BAC-Ins and QD655, internalization was induced and then cells were fixed in PFA. Immunofluorescence experiments were performed with 6 μg/ml caveolae marker, 4 μg/ml anti-CD63 and 1.5 μg/ml EEA1.

The classification into high and low expression levels was performed taking into account the mean of GFP signal (sum of GFP divided by cell size) of each cell. Cells with this value higher than 25 were considered to have high expression level, whereas cells with values lower than 20 were considered to have low expression level. In isoform discrimination of internalization, only the cells with a mean GFP signal <30 were considered in the quantification analysis in order to have an homogeneous cell sample for both isoforms.

We performed a similar analysis using Cy3 signal. After cell segmentation (membrane and interior) and mask generation, quantification of the Cy3 signal in the membrane of stimulated and non-stimulated cells was carried out. The ratio between this measurement and the membrane size (in pixels) was calculated for each cell. For evaluation of the expression level influence on receptor activation, a similar quantification was performed: the ratio between summed activation level (red) and summed expression level (green) signal in the membrane was calculated for each cell.

J.G. is a recipient of a CONICET fellowship and was awarded a short-term EMBO fellowship. This work was supported by the Max-Planck Society, University of Buenos Aires, CONICET (Argentina) and ANPCyT (Argentina). We thank Axel Ullrich for kindly providing plasmids pNTK2-hIR and pNTK2-IR/B, T. W. Gadella, Jr for pSCFP-3AC and pSYFP-2F vectors and Anabella Srebrow for the pCMV-AKT-HA vector. We thank Laboratorios BETA for providing rhIns. We thank Omar Coso for discussions about IR signaling and for providing the anti-ERK antibody and pCMV-β-Gal and AP-1-luc reporter plasmids and Patricia Cuasnicu for anti-PY20 antibody. We thank Wouter Caarls and Maria Julia Roberti for help and suggestions on Matlab use and figure preparation, respectively. We would like to thank Matías Blaustein for discussions about AKT experiments and Alejandro Colman Lerner for providing the anti-AKT antibody (Santa Cruz Biotechnology).