Differential activation of the inositol 5-phosphatase SHIP2 by EGF and insulin signaling pathways

Abstract

The importance of phosphatidylinositol 3,4,5- trisphosphate (PIP₃) in cell signaling has been well established. Despite phosphatidylinositol 3,4-bisphosphate [PI(3,4)P₂] emerging as an actor independent of PIP₃, its exact signaling role remains poorly understood, and the precise dynamics of PI(3,4)P₂ and PIP₃ upon receptor tyrosine kinase (RTK) stimulation are still inadequately investigated. In this study, we employed bioluminescence resonance energy transfer (BRET) sensors to monitor plasma membrane phosphoinositide (PIP) dynamics in HEK293-derived and HeLa cells following stimulation with epidermal growth factor (EGF) and insulin. Our findings reveal significant differences in PIP regulation: The increase in PI(3,4)P₂ compared to PIP₃ was larger with EGF stimulation relative to insulin. Using siRNA-mediated knockdown, we identified SH2-domain containing inositol polyphosphate 5-phosphatase 2 (SHIP2) as the key enzyme responsible for PI(3,4)P₂ production in the EGF pathway, which was further supported by a bioinformatics analysis. Moreover, we detected increased phosphorylation at two tyrosine sites in SHIP2 upon EGF stimulation, which was shown to be dependent on PI3K activation and PLC-induced calcium signal. These findings help refine our understanding of receptor-specific phosphoinositide dynamics and the enzymatic machinery involved, as well as their potential influence on downstream cellular responses.

Phosphoinositides (PIPs) are a class of specialized phospholipids that exist in dynamic pools, undergoing rapid interconversion by various enzymes (1). Despite their low abundance, these molecules are vital regulators of numerous essential cellular processes, primarily through their interactions with effector proteins that contain specific lipid-binding domains (2). There are seven distinct types of PIPs, each characterized by unique phosphorylation patterns on the 3-, 4-, and 5- hydroxyl groups of the inositol ring. PIPs are distributed across different intracellular membranes, where they are key to defining membrane identity (3). Furthermore, they act as crucial second messengers, with those localized to the plasma membrane (PM) playing a particularly central role in signal transduction (1). Moreover, dysregulation of PIP metabolism has been implicated in a range of pathologies, including cancer and metabolic disorders (4, 5).

Phosphatidylinositol 3,4,5-trisphosphate (PIP₃) has long been recognized as a crucial signaling molecule, particularly in receptor tyrosine kinase (RTK) signaling cascades. Activation of RTKs triggers the phosphorylation of phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂] at the 3-position by class I phosphatidylinositol 3-kinases (PI3Ks), leading to generation of PIP₃ at the PM (6). PIP₃ then acts as a critical signaling lipid, recruiting and activating various effector proteins, including protein kinase B (PKB or Akt), a serine/threonine kinase essential for regulating cell survival, growth, and proliferation (7, 8). Additionally, PIP₃ is involved in a wide range of other cellular processes, such as exocytosis, cytoskeletal remodeling, apoptosis, and metabolism (9).

Once formed, PIP₃ can be either dephosphorylated back to PI(4,5)P₂ by the 3-phosphatase PTEN or converted into phosphatidylinositol 3,4-bisphosphate [PI(3,4)P₂] by 5-phosphatases, such as SH2-domain-containing inositol polyphosphate 5-phosphatase (SHIP) enzymes (10, 11). Previously regarded as merely a degradation product of PIP₃, PI(3,4)P₂ has recently gained recognition as an active signaling molecule with functions that are, at least partially, independent of PIP₃ (12). Studies have revealed that PI(3,4)P₂ plays crucial roles in regulating cytoskeletal organization, cell polarization, and migration (13). Additionally, PI(3,4)P₂ can also be synthesized directly from phosphatidylinositol 4-phosphate (PI4P) via the catalytic action of class II PI3Ks (PI3KC2), a pathway that is pivotal for endocytosis (14, 15).

As PI(3,4)P₂ emerges as a distinct and important signaling molecule, understanding the enzymes that regulate its synthesis and turnover becomes increasingly critical. Central to this regulation are the SHIP enzymes, which play a key role in modulating PI(3,4)P₂ levels at the PM. Two major isoforms of SHIP exist: SHIP1, predominantly found in hematopoietic cells, and SHIP2, which is more widely expressed across various tissues (16, 17). SHIP enzymes have been linked to various diseases, such as Alzheimer's disease and metabolic disorders (18, 19). The role of SHIP enzymes in cancer is complex and remains a topic of debate (20). On the one hand, by reducing PIP₃ levels, SHIP enzymes can act as tumor suppressors. However, their product, PI(3,4)P₂, can also activate Akt isoforms, raising the possibility of oncogenic functions as well (17). Consequently, SHIP enzymes have become targets of interest in cancer research, with ongoing studies exploring their suppression as a potential therapeutic strategy (21).

Epidermal growth factor receptor (EGFR) and insulin receptor (IR) both belong to the RTK family and share many similar features and downstream signaling pathways, including activation of PI3Ks, consequent elevation of PIP₃ levels, and recruitment and activation of the Akt kinase. However, there are also substantial differences in their signaling patterns: They use a different set of adaptor proteins for PI3K and Akt activation and activate divergent downstream pathways (22, 23). As an example, EGFR stimulation involves activation of phospholipase C-gamma (PLCγ), which in turn generates a calcium signal and activates protein kinase C isoforms (22, 24), whereas IR stimulation can lead to activation of Rho and Rab family GTPases to promote glucose uptake in skeletal muscle (23, 25). PI(3,4)P₂ has been implicated in the signaling network of both EGFR and IR, but its relative contribution and the details of its regulation are still not entirely understood.

In this study, we investigated the dynamics, effects, and mechanisms of PM PIP levels after stimulation of EGFR and IR in both HEK293-derived and HeLa cells. Leveraging a sensitive bioluminescence resonance energy transfer (BRET)-based method and utilizing various PIP biosensors, we were able to demonstrate a distinctive pattern of changes in PIP₃ and PI(3,4)P₂ levels depending on the activated receptor. With knockdown experiments, we confirmed the central role of the SHIP2 enzyme in orchestrating the phosphoinositide cellular response to EGF stimulation. We also demonstrated that SHIP2 undergoes PI3K- and PLC-dependent phosphorylation in EGFR signaling, while remaining unaffected in the context of insulin stimulation.

Results

Monitoring plasma membrane phosphoinositide levels with BRET-based biosensors

As many of the effects of PIPs are mediated by the recruitment of soluble proteins to the membrane via specific PI- recognizing domains, such as the pleckstrin homology (PH) domain, we leveraged this mechanism to monitor PM phosphoinositide changes by fusing specific PH domains to fluorescent proteins to follow their localization. For this, we used lipid-binding domains that have been recognized and extensively employed in the field. To follow changes in PIP₃, we used two distinct peptide domains: the PH domain from the GRP1 protein and the PH domain of Bruton’s tyrosine kinase (Btk), while for PI(3,4)P₂ we used the PH domain of the TAPP1 protein in tandem (TAPP1-2xPH).

First, we assessed the intracellular localization of our lipid-binding PH domains with confocal microscopy in HEK293-AT1 cells, a HEK293-derived line stably expressing the type-1A rat angiotensin II receptor. These cells were transfected to express either the yellow fluorescent protein Venus directed to the cell membrane using the targeting sequence of Lck (L10), or the lipid-binding domains, also tagged with Venus. Since production of PIP₃ and its derivatives is known to be dependent on serum-derived factors, we captured images of native cells (control) and compared them to images captured following 5 h of serum deprivation. As expected, membrane-targeted Venus was confined to the PM. However, the biosensors exhibited varied intracellular localization. Specifically, in control cells kept in serum-containing medium, Btk-PH was localized in the cytosol, whereas GRP1-PH exhibited a clear association with the PM. TAPP1-2xPH also associated with the membrane, although less distinctly (Fig. 1A, upper images). In contrast, following 5-h serum starvation, the membrane association of GRP1-PH, although still present, was noticeably diminished. Concurrently, both Btk-PH and TAPP1-2xPH were restricted to the cytosol, with no evident membrane association (Fig. 1A, lower images).

Figure 1.

Biosensors for plasma membrane phosphoinositide detection.A, representative confocal images of native (control) or 5 h serum-starved HEK293-AT1 cells expressing plasma membrane targeted Venus or the indicated lipid-binding domains tagged with Venus. Scale bar: 10 μm. B, schematic representations of domain structures of the phosphoinositide BRET-based biosensors used in this study. PM represents the plasma membrane target sequence of Lck, while T2A is a viral protein sequence. The sequence is transcribed as a single mRNA molecule, but translation is interrupted between the two amino acids indicated by the redline, resulting in two separate polypeptide chains. C, the bar graph shows the basal BRET ratios of the indicated sensors measured in native cells or after 5 h of serum deprivation. Data are the mean ± SD of 3 independent experiments. Statistical significance was tested with two-way ANOVA and Holm-Sidak post hoc test, with only the comparisons of interest indicated. NS: p > 0.05, ∗p < 0.05, ∗∗∗p < 0.001 between the Btk-PH group and GRP1-PH group. NES, nuclear export signal; SD, serum deprivation.

Given the higher sensitivity of BRET for observing PIP dynamics, we adapted our confocal microscopy sensors into BRET sensors to monitor PIP levels upon receptor stimulation. To achieve this, we fused the lipid-binding domains to the BRET donor molecule, luciferase. Concurrently, to specifically monitor PIP levels at the PM we anchored the BRET acceptor, Venus, to the cell membrane using the above-mentioned targeting sequence (Fig. 1B). This setup allows for direct monitoring of PIP levels at the PM as shifts in PIP concentrations result in altered recruitment of the lipid-binding domains, thereby modulating the BRET signal. A more precise description of the method was published recently (26, 27).

To compare our BRET-based biosensors, we calculated the basal BRET ratio (in the absence of stimuli) of the three sensors, which correlates with the resting PM localization of the lipid-binding domains, in native (control) and 5-h serum starved cells. Consistent with the confocal images, the basal BRET ratio in the control cells was markedly higher for GRP1-PH compared to the Btk-PH and TAPP1-2xPH sensors. Notably, 5-hour-long serum starvation resulted in decreased basal BRET ratios for GRP1-PH and TAPP1-2xPH when compared to their respective control states (Fig. 1C). Taken together, these findings highlight the distinct intracellular localization and cellular dynamics of the two PIP₃ biosensors, Btk-PH and GRP1-PH. These differences prompted us to use both biosensors during our experiments, which allows for a more comprehensive assessment of fluctuations in PIP₃ levels.

Comparison of EGF and insulin effects on PIP levels in HEK293-AT1 cells

We employed our BRET-based biosensors described earlier to quantify alterations in PM PIP levels in response to different stimuli, using HEK293-AT1 cells. Tyrosine kinase receptors are known to activate PI3K and generate PIP₃ and also PI(3,4)P₂, so we decided to first test the effects of EGF stimulation on these lipids. HEK293-AT1 cells were transfected with the above-mentioned biosensors, and to overcome the variability associated with endogenous EGFR expression in HEK293-derived cell lines, cells were also transfected with EGFR. 28 h after transfection, we applied 100 ng/ml EGF stimulation to serum-starved HEK293-AT1 cells and subsequently calculated changes in the BRET ratio, which represents the lipid-dependent membrane binding of the PH domains of the sensors. An increase in the signal of a particular sensor upon stimulation represents an elevation in the corresponding PIP level at the PM. EGF stimulation resulted in a slight increase in PIP₃ levels measured with both sensors, coupled with a pronounced rise in PI(3,4)P₂ levels (Fig. 2A, blue curves). Since the changes for both PIP₃ sensors were surprisingly low, we chose to confirm their lipid binding affinity by comparing the PIP profile of EGF stimulation to that of another tyrosine kinase receptor, the insulin receptor. We treated the cells with an insulin concentration that elicited a PI(3,4)P₂ response comparable to that of EGF, which was determined to be 70 nM. In contrast to the EGF response, insulin stimulation led to a substantial increase in both PIP₃ and PI(3,4)P₂ levels (Fig. 2A, red curves).

Figure 2.

Comparison of EGF and insulin effects on plasma membrane PIP₃ and PI(3,4)P₂ levels in HEK293-AT1 cells.A, the levels of PIP₃ and PI(3,4)P₂ were measured with BRET in HEK293-AT1 cells. Cells were transiently transfected with CMV EGFR construct, as well as Btk-PH or GRP1-PH sensors to measure PIP₃, or TAPP1-2xPH sensor to measure PI(3,4)P₂. 28 h after transfection, serum-starved cells (5 h) were treated with EGF (100 ng/ml, bluecurves) or insulin (70 nM, red curves) at time point indicated by the arrows, and the BRET ratio was calculated as described in Methods. BRET ratio values were normalized for the baselines (100%) for each curve. Data are the mean ± SD of 7 independent experiments, each performed in duplicate. B, for better comparison of EGF and insulin effects on PIP₃ and PI(3,4)P₂ levels, the PI(3,4)P₂/PIP₃ ratios were calculated, using either the Btk- or GRP1-based PIP₃ signal, and the TAPP1-based sensor response for PI(3,4)P₂ in both cases. C, the bar graph shows the PI(3,4)P₂/PIP₃ ratio with each stimulus. Data are means ± SD of the last 5 measurement points of the curves from panel B. EGF- and insulin-induced changes were compared with an unpaired, two-tailed t test for each sensor separately. ∗∗p < 0.01.

For better visualization of this different pattern, we calculated the PI(3,4)P₂/PIP₃ ratio under each stimulus condition. Ratios were calculated using the TAPP1-based sensor response for PI(3,4)P₂, and either the Btk- or GRP1-based signal for PIP₃. These ratios were significantly higher with EGF compared to insulin stimulation (Fig. 2, B and C), suggesting distinct cellular responses to these two stimuli in terms of phosphatidylinositol signaling.

Comparison of EGF and insulin effects on PIP levels in HEK293A cells

Building on our findings from the specialized HEK293-AT1 cells, we sought to extend the validity of our results by performing parallel experiments on a more ubiquitously utilized cell line, HEK293A. In this new set of experiments, we introduced a novel BRET PIP₃ biosensor, Btk-2xPH, a tandem version of the previously employed Btk-PH sensor, to overcome the lower affinity of the Btk-PH sensor compared to GRP1-PH, reflected by its significantly lower basal BRET ratio (Fig. 1C). Confocal microscopy experiments investigating the biosensor localization revealed a lack of PM association under serum-starved conditions. However, upon 70 nM insulin stimulation, a pronounced association with the PM was observed (Fig. 3A). We also checked the basal BRET ratios for each sensor in HEK293A cells under serum-starved conditions (Fig. 3B) which proved to be slightly different from those measured in the previous cell line (Fig. 1C).

Figure 3.

Comparison of EGF and insulin effects on plasma membrane PIP₃ and PI(3,4)P₂ levels in HEK293A cells.A, representative confocal images of serum starved (5 h) HEK293A cells expressing the Btk-2xPH biosensor tagged with Venus, before and 3 min after stimulation with 70 nM insulin, as indicated. Scale bar: 10 μm. B, the bar graph shows the basal BRET ratios of the indicated sensors measured in serum starved cells (5 h). Data are the mean ± SD of 3 independent experiments. C, the levels of PIP₃ and PI(3,4)P₂ were measured with BRET in HEK293A cells. Cells were transiently transfected with CMV EGFR construct, as well as with the Btk-PH, Btk-2xPH or GRP1-PH sensors to measure PIP₃, or with the TAPP1-2xPH sensor to measure PI(3,4)P₂. 28 h after the transfection, serum starved cells (5 h) were treated with EGF (100 ng/ml, blue curves), a low (L) or high (H) dose of insulin (70 nM and 0.7 μM, red and green curves, respectively), at time point indicated by the arrows. BRET ratios were calculated as described in Methods and were normalized for the baselines (100%) for each curve. Data are the mean ± SD of 3 independent experiments, each performed in duplicate. D, PI(3,4)P₂/PIP₃ ratios were calculated similarly to Figure 2B using all three PIP₃ sensors in the calculation, as indicated. E, the bar graph shows the PI(3,4)P₂/PIP₃ ratios with each stimulus. Data are means ± SD of the last 5 measurement points of the curves from panel D. For statistical analysis, one-way ANOVA with Holm-Sidak post hoc test was used, separately for each sensor. ∗∗∗p < 0.001, NS p > 0.05. F, the PI(3,4)P₂/PIP₃ ratio was measured using the tandem Btk-PH sensor for PIP₃ detection. Experimental conditions were identical to those in Panel D, except data were collected at the highest achievable temporal resolution. Results are shown as mean ± SD of 3 independent experiments.

In addition to EGF and the original insulin concentration, we also explored the effects of a tenfold increase in insulin concentration. Initially, we transfected HEK293A cells with EGFR and our specific biosensors targeting PI(3,4)P₂ and PIP₃, using all three PIP₃ sensors mentioned. After 28 h, serum-starved cells were stimulated with 100 ng/ml EGF or the two insulin concentrations (70 nM and 0.7 μM), and BRET ratio variations were calculated. Here, EGF and the original insulin concentration induced a similar increase in PIP₃, but a more significant elevation in PI(3,4)P₂ levels was observed with EGF (blue and red curves on Fig. 3C, respectively). Notably, the higher insulin concentration caused a more substantial increase in both PI(3,4)P₂ and PIP₃ levels compared to the previously used concentration (green and red curves on Fig. 3C, respectively). Response to all three stimuli were substantially higher for the tandem version of the Btk sensor compared to the original version, indicating a higher affinity for PIP₃. Taken together, these findings demonstrate that the Btk-2xPH biosensor can recognize PIP₃ in the PM with sufficiently high affinity, rendering it an additional tool for measuring its fluctuations.

Analysis of the PI(3,4)P₂/PIP₃ ratios demonstrated that, similar to HEK293-AT1 cells, EGF yielded a higher ratio than insulin regardless of the PIP₃ sensor used in the ratio calculation (Fig. 3D). Notably, the ratios were consistent across both insulin concentrations (green and red curves on Fig. 3D). We also tested these ratio changes statistically and found that EGF stimulation was significantly different from both insulin concentrations, but there was no difference between high and low insulin (Fig. 3E). Next, as our earliest post-stimulation time point was already near the plateau, we examined PI(3,4)P₂/PIP₃ ratio dynamics at the highest temporal resolution achievable with our setup (every 2 s), using the tandem Btk-PH sensor. This revealed a transient decrease in the ratio immediately following EGF stimulation, followed by a rise to its elevated level (Fig. 3F).

Surprisingly, despite the extensive similarities between the two cell lines, the absolute values of the ratios were different in the HEK293-AT1 and HEK293A cells. Specifically, the ratios in the HEK293A cells, calculated using the GRP1-PH sensor measurements (Fig. 3, B and E), were lower than those in HEK293-AT1 cells (Fig. 2, B and C), while maintaining the relative differences between EGF and insulin stimulation.

The effect of siRNA-mediated SHIP2 knockdown on EGF-induced plasma membrane PIP level changes

As EGF elicited a pronounced increase in PM PI(3,4)P₂ in both HEK cell lines, we next aimed to investigate the underlying mechanism responsible for this increase. There are two primary pathways through which PI(3,4)P₂ can be synthesized: either through the 5-phosphatase activity of SHIP2 enzyme which transforms PIP₃, produced by class I PI3Ks, into PI(3,4)P₂, or via the direct phosphorylation of PI4P by class II PI3Ks (Fig. 4A). Nevertheless, the consensus is that the predominant source of PI(3,4)P₂ is PIP₃ (28). To validate this consensus in our model, we employed siRNA-mediated knockdown of SHIP2 and class II PI3Ks. HEK293-AT1 cells were first transfected with the indicated siRNAs, 48 h later with EGFR and the GRP1-PH or TAPP1-2xPH sensors. 28 h later, serum-starved cells were stimulated with EGF and BRET ratios were calculated. With SHIP2 knockdown, we observed a slight increase in PIP₃ levels accompanied by a pronounced reduction in PI(3,4)P₂ levels compared to control (Fig. 4B, orange and black curves, respectively). Contrarily, knockdown of class II PI3Ks considerably amplified the EGF-induced PI(3,4)P₂ levels (Fig. 4B, green curves). Consequently, the PI(3,4)P₂/PIP₃ ratio upon EGF stimulation was decreased with SHIP2 knockdown and increased with class II PI3K knockdown when compared to control (Fig. 4C).

Figure 4.

The effect of siRNA-mediated SHIP2 knockdown on EGF signaling pathway.A, schematic depiction of the two PI(3,4)P₂ synthetic pathways and the inhibitors used. PI(3,4)P₂ is synthesized either through the 5-phosphatase activity of the SHIP2 enzyme on PIP₃ or via the direct phosphorylation of PI4P by class II PI3Ks. B, the levels of PI(3,4)P₂ and PIP₃ were measured with BRET in HEK293-AT1 cells. Cells were transfected with scrambled control siRNA (scr – blackcurves) or siRNA targeting SHIP2 (siSHIP2 – orangecurves) or PI3KC2 (siPI3KC2 mix – green curves). 48 h later, cells were transfected with CMV EGFR construct as well as the GRP1-PH sensor to measure PIP₃, or the TAPP1-2xPH sensor to measure PI(3,4)P₂. 28 h after transfection, serum starved cells (5 h) were treated with 100 ng/ml EGF, at time point indicated by the arrows, and the BRET ratio was calculated. BRET ratio values were normalized for the baselines (100%) for each curve. Data are the mean ± SD of 5 independent experiments, each performed in duplicate. C, the effects of siRNA-mediated knockdown of the PI(3,4)P₂ synthetic pathways on the EGF induced PI(3,4)P₂/PIP₃ ratio in HEK-AT1 cells (leftgraph). The bar graph shows the PI(3,4)P₂/PIP₃ ratio in case of each siRNA, calculated from the last 5 measurement points of the curves from the ratio graph. Data are mean ± SD and for statistical analysis ANOVA on ranks was used with Student-Neuman-Keuls post hoc test. ∗p < 0.05. D, the levels of PI(3,4)P₂ and PIP₃ following transfection with scrambled siRNA (blackcurves) or siSHIP2 (orangecurves) were measured with BRET in HEK293A cells as detailed in panel B. Data are the mean ± SD of 5 independent experiments, each performed in duplicate. E, the effects of siRNA-mediated SHIP2 knockdown on the EGF-induced PI(3,4)P₂/PIP₃ ratio in HEK293A cells (leftgraph). The bar graph shows data as mean ± SD of the last 5 measurement points of the curves from the ratio graph, statistical significance was evaluated by unpaired, two-tailed t test. ∗∗∗p < 0.001. F, the effects of siRNA-mediated SHIP2 knockdown on the insulin-induced PI(3,4)P₂/PIP₃ ratio in HEK293A cells (leftgraph). The bar graph shows data as mean ± SD of the last 5 measurement points of the curves from the ratio graph, statistical significance was evaluated by unpaired, two-tailed t test. NS: p > 0.05. G, lysates from HEK293A cells transfected with scrambled control siRNA (scr) or siRNA targeting SHIP2 (siSHIP2) were analyzed by Western blot using anti-SHIP2 and anti-EGFR antibodies. Molecular weights (in kDa) are indicated with arrows. H, the effects of siRNA-mediated SHIP2 knockdown and rescue on EGF-induced PI(3,4)P₂ level changes in HEK293A cells (leftgraph). Knockdown was carried out as specified in panel D, and for the rescue either wild-type (WT) or phosphatase domain deleted (PD) SHIP2 were co-transfected with CMV EGFR and the sensor for PI(3,4)P₂. The bar graph shows data as mean ± SD of the last 5 measurement points of the curves from the time-course graph, statistical significance was evaluated by one-way ANOVA with Holm-Sidak post hoc test. ∗∗∗p < 0.001, NS: p > 0.05. I, the levels of PIP₃ and PI(3,4)P₂ were measured with BRET in HEK293A cells expressing the CMV EGFR construct. Cells were serum-starved (5 h) before being pre-treated with DMSO or wortmannin (Wm, 300 nM) for 30 min. Cells were then stimulated with EGF (100 ng/ml), as indicated by the arrows. Data are the mean ± SD of 3 independent experiments, each performed in duplicate.

Similarly, the EGF-induced response was also tested in HEK293 A cells, targeting only SHIP2 as it effectively reduced PI(3,4)P₂ levels in the other cell line. Consistent with previous results, siRNA-mediated SHIP2 knockdown mildly elevated PIP₃ levels upon EGF stimulation while simultaneously substantially attenuating the rise in PI(3,4)P₂ (Fig. 4D), thereby significantly reducing the PI(3,4)P₂/PIP₃ ratio when compared to control (Fig. 4E). These findings suggest that mainly SHIP2 activity is responsible for the observed PI(3,4)P₂ surge upon EGF stimulation in both investigated cell lines. In contrast, SHIP2 knockdown had no effect on the PI(3,4)P₂/PIP₃ ratio following insulin stimulation, suggesting a restricted role of this phosphatase in insulin-induced PIP dynamics (Fig. 4F). Since SHIP2 knockdown did not result in complete inhibition of the PI(3,4)P₂ increase with EGF stimulation, we decided to assess the efficacy of SHIP2 depletion in HEK293A cells by immunoblotting, which demonstrated a marked but not complete reduction in SHIP2 protein levels, while EGFR expression remained unaffected (Fig. 4G). We further confirmed the specificity of the knockdown with rescue experiments, which showed that the exogenous expression of wild-type (WT), but not the phosphatase domain deleted (PD), SHIP2 was able to restore EGF-induced PI(3,4)P₂ dynamics after SHIP2 knockdown (Fig. 4H).

Furthermore, to confirm that the PI(3,4)P₂ surge observed is the consequence of its enhanced synthesis rather than altered degradation, we pre-treated HEK293A cells with 300 nM wortmannin (Wm), a non-selective PI3K inhibitor, for 30 min before stimulating with EGF. The presence of wortmannin completely eliminated any increase in the levels of PIP₃ and PI(3,4)P₂, thus confirming that the observed PI(3,4)P₂ surge originates from its increased synthesis (Fig. 4I). Taken together, our findings demonstrate that the alterations in PI(3,4)P₂ following EGF stimulation are primarily due to enhanced synthesis of PI(3,4)P₂ via the class I PI3K-SHIP2 pathway. This is further supported by the transient decrease in PI(3,4)P₂/PIP₃ ratio immediately following EGF stimulation before its elevation, highlighted by the enhanced temporal resolution data shown in Figure 3F.

The effect of EGFR stimulation on SHIP2 phosphorylation

Having established the pivotal role of SHIP2 in the elevation of PI(3,4)P₂ following EGF stimulation, we aimed to investigate its regulation in response to EGF and insulin stimulation. To this end we measured its phosphorylation levels, known to correlate with enzymatic activity (29). In this series of experiments, we overexpressed EGFR in HEK293A cells using two distinct approaches: a substantial overexpression and a more physiologically representative, moderate overexpression achieved through different promoters: cytomegalovirus (CMV) and thymidine kinase (TK) promoters for high and low expression, respectively. 28 h following transfection, serum-starved cells were stimulated with EGF and the two previously selected insulin concentrations for 5 min. Cell lysates were analyzed through Western blotting.

Stimulation with EGF resulted in the phosphorylation of SHIP2 at tyrosine positions 986/987 and 1135. Contrarily, stimulation with insulin did not induce SHIP2 phosphorylation, even after a tenfold increase in the concentration (Fig. 5A). As expected, EGF-induced SHIP2 phosphorylation was more pronounced in cells with higher EGFR overexpression (Fig. 5A, left images) but was still detectable in cells with moderate overexpression (Fig. 5A, right images). To confirm that the used stimuli concentrations are effective we further compared the signaling of EGF and insulin in two additional pathways, a PI3K-dependent pathway, Akt, and a PI3K independent one, ERK (22, 30). As PI3K activation is shared between EGF and insulin, both stimulations resulted in Akt phosphorylation. However, ERK was phosphorylated with the EGF pathway only. We also examined EGFR phosphorylation which occurred only with EGF, as expected. These observations were consistent across both levels of receptor overexpression (Fig. 5A). Immunoblotting images of SHIP2 phosphorylation at tyrosine position 1135 were quantified and statistically tested (Fig. 5B).

Figure 5.

The effect of EGFR stimulation on SHIP2 phosphorylation.A, EGFR was expressed at two different levels in HEK293A cells, using either CMV or TK promoters to achieve substantial and moderate overexpression, respectively. Serum starved (5 h) samples were stimulated with 100 ng/ml EGF, 70 nM insulin (L) or 0.7 μM insulin (H), as indicated, for 5 min. Cell lysates were then analyzed with Western blot using antibodies for the indicated proteins. Phospho-specific antibodies are marked with an initial p, and the sites of phosphorylation are indicated in brackets. The total protein amounts are shown as loading controls. Molecular weights (in kDa) are indicated with arrows. B, immunoblot quantification of phospho-SHIP2 at Tyr1135 as shown on panel A, separately for substantial overexpression (left graph) and moderate overexpression (right graph) of EGFR. Quantification values were normalized for the control (100%) for each experiment. Data are the means ± SD from 4 to 8 independent experiments. For statistical analysis ANOVA on ranks was used with Student-Neuman-Keuls post hoc test, and EGF-treated samples were significantly different from all other samples (∗p < 0.05), while insulin treatment caused no significant change. C, Western blot analysis of SHIP2 phosphorylation at tyrosine 1135 following EGF stimulation over various time points (5 s to 20 min, as labeled above each lane) in HEK293A cells overexpressing EGFR (CMV). The upper panel displays phosphorylated SHIP2 (pSHIP2 Y1135), while the lower panel shows total SHIP2, which served as a loading control. Arrows on the left indicate the molecular weight marker at 140 kDa. D, immunoblot quantification of phospho-SHIP2 on panel C. Quantification values were normalized for the 1 min EGF stimulation (100%) for each experiment. Data are the means ± SD from 3 to 4 independent experiments. E, the levels of PIP₃ and PI(3,4)P₂ were measured with BRET in HEK293A cells transiently transfected with the different EGFR constructs (as indicated) and PIP sensors (as detailed in Fig. 3). Serum starved cells (5 h) were treated with EGF (100 ng/ml) or insulin (70 nM) at time points indicated with arrows, and BRET ratios were processed as detailed in the methods. Data are the mean ± SD of 3 independent experiments, each performed in duplicate.

As our BRET measurements indicated that changes in PIP levels following EGF stimulation were rapid (Fig. 3F), we examined the dynamics of SHIP2 phosphorylation at tyrosine position 1135 with a higher temporal resolution, using eight different durations of EGF exposure rather than only 5 min. Results demonstrated that SHIP2 phosphorylation was discernible as early as 5 s post-EGF stimulation and increased steadily with longer durations of stimulation, reaching its peak intensity at 10 min (Fig. 5, C and D).

To address potential concerns that EGFR overexpression could affect PIP dynamics, we conducted BRET-based lipid measurements in cells exhibiting either endogenous EGFR levels or overexpression driven by the TK or CMV promoters. All conditions displayed a consistent pattern of phosphoinositide response upon EGF stimulation, with higher EGFR expression correlating with more sustained lipid alterations (Fig. 5E, upper graphs). Insulin-induced PIP patterns were also not affected by EGFR overexpression (Fig. 5E, lower graphs).

In conclusion, our findings demonstrate that the observed surge in PI(3,4)P₂ following EGF stimulation can be attributed to the enhanced activity of SHIP2, likely facilitated by its phosphorylation. Notably, this phosphorylation of SHIP2 is absent with insulin stimuli, highlighting the different signaling patterns for these two tyrosine kinase receptors.

PIP signaling dynamics upon EGF and insulin stimulation in HeLa cells with endogenous EGFR expression

Given that EGF receptor expression can be variable in HEK cell lines thus necessitating receptor overexpression, we sought a cell line with stable, endogenous EGF receptor expression. We therefore selected HeLa cells, which respond robustly to both EGF and insulin (31, 32, 33). These cells were transfected with our PI(3,4)P₂ and PIP₃ biosensors. 28 h later, serum-starved cells were stimulated with 100 ng/ml EGF or 70 nM insulin, and BRET ratio changes were measured. The resulting PI(3,4)P₂ and PIP₃ dynamics closely resembled those observed in HEK cells. Although insulin produced a comparatively larger increase in PIP₃, EGF elicited a more pronounced elevation of PI(3,4)P₂ (Fig. 6A, red and blue curves respectively). Analysis of the PI(3,4)P₂/PIP₃ ratio confirmed that, similar to both examined HEK cell lines, EGF consistently yielded a higher ratio than insulin, regardless of which PIP₃ sensor was used (Fig. 6, B and C). We also performed the same immunoblotting analysis as in HEK293A cells, revealing comparable outcomes, including phosphorylation of SHIP2 at two tyrosine sites upon EGF stimulation (Fig. 6D). In conclusion, the observed PIP patterns and SHIP2 dynamics are not exclusive to HEK-derived cell lines and do not require receptor overexpression. Instead, they are also evident in a cell line with endogenous EGF receptor expression, reinforcing the broader applicability of our findings.

Figure 6.

Comparing the effect of EGF and insulin stimulation in HeLa cells.A, the levels of PIP₃ and PI(3,4)P₂ were measured with BRET in HeLa cells. Cells were transiently transfected with the Btk-2xPH or GRP1-PH sensors to measure PIP₃, or with the TAPP1-2xPH sensor to measure PI(3,4)P₂. 28 h after the transfection, serum starved cells (5 h) were treated with EGF (100 ng/ml, blue curves) and insulin (70 nM, red curves), at time point indicated by the arrows. BRET ratios were calculated as described in Methods and were normalized for the baselines (100%) for each curve. Data are the mean ± SD of 4 independent experiments, each performed in duplicate. B, PI(3,4)P₂/PIP₃ ratios were calculated similarly to Figures 2B and 3D using both PIP₃ sensors in the calculation, as indicated. C, the bar graph shows the PI(3,4)P₂/PIP₃ ratios with each stimulus. Data are means ± SD of the last 5 measurement points of the curves from panel B. For statistical analysis, EGF- and insulin-induced changes were compared with an unpaired, two-tailed t test for each sensor separately. ∗∗∗p < 0.001. D, serum starved (5 h) HeLa cells were stimulated with 100 ng/ml EGF, 70 nM insulin (L) or 0.7 μM insulin (H), as indicated, for 5 min. Cell lysates were then analyzed with Western blot using antibodies for the indicated proteins. Phospho-specific antibodies are marked with an initial p, and the sites of phosphorylation are indicated in brackets. The total protein amounts are shown as loading controls. Molecular weights (in kDa) are indicated with arrows.

Bioinformatics analysis of SHIP2 involvement in intracellular signaling pathways

Taken together, our findings suggest a more significant involvement of SHIP2 in the EGFR signaling pathway as compared to the insulin receptor (IR) pathway. To further substantiate this observation, we utilized a bioinformatics approach with the L-1000 database (refer to Methods for details). This analysis involved comparing gene expression signatures from SHIP2 manipulation (both inhibition and activation) against those resulting from the manipulation of other proteins. The results, presented in Figure 7, show the distribution of normalized correlation scores, which quantify the similarity between SHIP2-associated gene expression changes and those induced by other proteins. Specifically, positive scores denote concordant expression patterns, while negative scores indicate opposing effects. The results display a distinct bimodal distribution with sharp peaks at −1 and +1, underscoring strong reciprocal or parallel transcriptional responses linked to SHIP2 activity.

Figure 7.

Bioinformatics analysis of SHIP2 involvement in intracellular signaling pathways. This bar chart illustrates the distribution of Pearson correlation scores that quantify the degree of similarity between SHIP2 perturbation profiles and various other perturbation profiles within the L-1000 database (detailed in the Methods section). Each bar corresponds to the frequency of scores across specified intervals. Above these bars, colored triangles mark specific correlation values: blue triangles for scores derived from SHIP2/EGFR perturbation profile correlations, and red triangles for those from SHIP2/IR perturbation profile correlations. These color-coded markers help illustrate the comparative analysis of how SHIP2 interacts with the EGFR and IR signaling pathways.

In the graph, significant hits are concentrated in the two tails, with their importance increasing the farther they are from the zero point. Triangles mark the correlation scores: blue for EGFR and red for IR manipulations relative to SHIP2 manipulations. The results clearly show a disparity in the correlation values. While the IR profiles exhibit only moderate similarity to SHIP2 perturbations, appearing in the less significant portions of the distribution, the EGFR profiles predominantly cluster in the positive tail of the distribution (in the higher, more significant values). This indicates strong similarities between EGFR and SHIP2 perturbations, affirming the critical role of SHIP2 within the EGFR signaling network, far more than within the IR pathway.

Mechanism of SHIP2 phosphorylation upon EGF stimulation and its effect on PIP dynamics

After noting enhanced SHIP2 phosphorylation following EGFR stimulation, which produces a unique PIP pattern distinct from IR stimulation, we sought to explore the specific molecular mechanism driving SHIP2 phosphorylation during EGFR activation. Both EGFR and IR are receptor tyrosine kinases and share many signaling pathways such as PI3K activation, but a unique feature of EGFR signaling not associated with the IR pathway is PLCγ-induced calcium increase (22, 24). Consequently, we started with exploring the potential role of both PI3K and PLCγ pathways in SHIP2 phosphorylation. To investigate the role of calcium signaling, we employed immunoblotting utilizing the calcium chelator BAPTA-AM in HEK293A cells. Initially, cells were transfected with EGFR and, after 28 h, serum-starved cells were pre-treated for 30 min with either BAPTA-AM or DMSO before stimulation with EGF and the two previously selected insulin concentrations for 5 min. Calcium chelation resulted in a partial but significant attenuation of EGF-induced phosphorylation at both Y1135 and Y986/987 of SHIP2 (Fig. 8A). To confirm effective calcium chelation with BAPTA-AM pre-treatment we conducted cytoplasmic calcium measurements in HEK293A using our previously developed BRET calcium sensors (34), and were able to confirm effective calcium buffering by BAPTA-AM compared to DMSO pre-treatment following EGF stimulation (Fig. 8B).

Figure 8.

Mechanism of SHIP2 phosphorylation upon EGF stimulation.A, Western blot analysis of phospho-SHIP2 levels at tyrosine positions 1135 and 986/987 in HEK293A cells expressing the CMV EGFR construct. Serum-starved (5 h) cells, pre-treated for 30 min with either DMSO or BAPTA-AM (20 μM), were stimulated with 100 ng/ml EGF, 70 nM insulin (L) or 0.7 μM insulin (H), as indicated, for 5 min. Anti-SHIP2 measurement was used as loading control. Molecular weights are indicated with arrows (140 kDa). B, cytoplasmic calcium measurements with a BRET-based sensor in serum-starved HEK293A cells pre-treated for 30 min with either DMSO or BAPTA-AM. Cells were stimulated with EGF as indicated, and BRET ratios were normalized to the control period of DMSO-pretreated samples. Data are the mean ± SD of 3 independent experiments, each performed in duplicate. C, Western blot analysis of phospho-SHIP2 and phospho-Akt levels in HEK293A cells. All conditions were the same as on panel A, except for the 30-min pre-treatment with 100 nM Wortmannin (Wm). Anti-SHIP2 and anti-Akt measurements were used as loading control. D, Western blot analysis of phospho-SHIP2 levels. All conditions were the same as on panel A, except for the 30-min combined pre-treatment with BAPTA-AM and Wm. E, immunoblot quantification of EGF-induced phospho-SHIP2 levels at tyrosine sites Y1135 and Y986/987 with the indicated pre-treatments. Values were normalized for the DMSO pre-treated samples (100%) for each experiment. Data are the means ± SD from 3 to 4 independent experiments. For statistical analysis two-way ANOVA was used with Holm-Sidak post hoc test. All pre-treated samples were significantly different from their respective control. Significance is indicated only for comparisons of particular interest. ∗∗p < 0.01, ∗∗∗p < 0.001. F, Western blot analysis of EGF-induced SHIP2 phosphorylation levels following 30 min pre-treatment with DMSO, U73122 (10 μM) or U73343 (10 μM, negative control). G, immunoblot quantification of EGF-induced phospho-SHIP2 levels at tyrosine sites Y1135 and Y986/987 with the indicated pre-treatments. Values were normalized for the DMSO treated samples (100%) for each experiment. Data are the means ± SD from 3 independent experiments. For statistical analysis, one-way ANOVA with Holm-Sidak post hoc test was used, separately for each phosphorylation site. There was no significant difference between DMSO and U73343 pre-treatment. ∗∗∗p < 0.001. H, the PI(3,4)P₂/PIP₃ ratio was analyzed using the tandem Btk-PH-based PIP₃ sensor and the TAPP1-2xPH sensor to measure PI(3,4)P₂. HEK293A cells expressing CMV EGFR construct were serum-starved (5 h) before being pre-treated with DMSO, U73122 (10 μM), or U73343 (10 μM) for 30 min. Cells were then stimulated with EGF (100 ng/ml, upper graph) or insulin (70 nM, lower graph), as indicated by the arrows. Data are the mean ± SD of 3 independent experiments, each performed in duplicate. The bar graphs illustrate the PI(3,4)P₂/PIP₃ ratios associated with each pre-treatment, presented as mean ± SD of the last 5 measurement points from the ratio graphs. Statistical significance was determined using one-way ANOVA and Holm-Sidak post hoc test was used for EGF-stimulated groups. ∗∗∗p < 0.001, NS p > 0.05.

To determine the role of PI3K in SHIP2 phosphorylation we conducted similar immunoblotting experiments using wortmannin pre-treatment in HEK293A. Similarly to BAPTA-AM, 30 min of wortmannin pre-treatment resulted in a partial but substantial attenuation of EGF-induced SHIP2 phosphorylation at both tyrosine sites. To confirm effective inhibition of PI3Ks by wortmannin, Akt phosphorylation was also measured with immunoblotting, which showed complete suppression (Fig. 8C). Finally, 30-min pre-treatment using both BAPTA-AM and wortmannin resulted in a strong attenuation of SHIP2 phosphorylation (Fig. 8D). This combined pre-treatment proved to cause significantly greater inhibition compared to individual treatments, when immunoblotting images were quantified and statistically tested (Fig. 8E).

Recognizing the PLCγ-induced calcium signaling as a key determinant in SHIP2 phosphorylation, we aimed to investigate this pathway in a more specific manner. Cells transfected with EGFR were serum-starved and then pre-treated for 30 min with U73122, a PLC inhibitor, or with DMSO and U73343 (specific negative control), prior to a 5-min stimulation with EGF. Inhibition of PLC markedly reduced EGF-induced phosphorylation of SHIP2 at both tyrosine residues Y1135 and Y986/987 (Fig. 8F), which proved significant after quantification (Fig. 8G). Given ongoing debates about the impact of SHIP2 phosphorylation on its phosphatase activity, we further explored the effects of PLC inhibition on the PI(3,4)P₂/PIP₃ ratio using BRET. Our results indicated a significant decrease in this ratio following PLC inhibition with EGF stimulation, while, as expected, insulin-induced changes remained unaffected (Fig. 8H).

Discussion

In this study, we employed our BRET-based biosensors to monitor PIP levels at the PM in real-time. This approach allowed us to recognize distinct signaling behaviors in response to EGF and insulin stimulation. Notably, although both EGF and insulin elevated PIP₃ and PI(3,4)P₂ levels, EGF consistently generated a higher PI(3,4)P₂/PIP₃ ratio compared to insulin in both HEK cell lines and in HeLa cells. This observation suggests a preferential production of PI(3,4)P₂ during EGF signaling. Using siRNA-mediated knockdown of both PI(3,4)P₂ synthetic pathways, we demonstrated that SHIP2 is primarily responsible for regulating PI(3,4)P₂ levels at the membrane during EGF stimulation. This was corroborated by siRNA rescue experiments and enhanced temporal resolution BRET measurements of the PI(3,4)P₂/PIP₃ ratio. On the other hand, SHIP2 knockdown had no influence on insulin-induced PIP dynamics. Additionally, phosphorylation of SHIP2 was observed following EGF but not insulin stimulation, indicating a distinct role of SHIP2 in modulating the PI3K signaling pathway in response to EGF. Moreover, we found that while SHIP2 phosphorylation in the EGF pathway depends partially on PI3K activity, it is primarily driven by the PLC-induced calcium signal. Beyond reduction in SHIP2 phosphorylation, inhibition of PLC also significantly reduced the PI(3,4)P₂/PIP₃ ratio, strongly suggesting a role for SHIP2 phosphorylation in regulating its phosphatase activity. These findings provide new insights into receptor-specific phosphoinositide dynamics and emphasize how differential growth factor signaling can shape unique lipid signaling profiles at the PM, potentially influencing downstream cellular responses.

In our experiments, we used several different sensors for PIP₃. Although these sensors contain only the lipid-binding PH domains of their respective original proteins (GRP1 or Btk), they might still differ in their affinity and binding kinetics and thus behave differently under the same conditions, as can be seen on the confocal images of Figure 1B. This discrepancy between GRP1- and Btk-based sensors is reflected in the basal BRET ratios as well (Fig. 1C), although this signal might be additionally influenced by the relative orientation of the BRET donor and acceptor molecules in each sensor (35). To enhance the relatively small signals, we introduced a tandem dimer version of our Btk-based sensor which we expected to have increased avidity. Indeed, our artificially duplicated sensor showed a slightly higher signal already at resting PIP₃ levels (Fig. 3B), and a substantially larger increase after receptor stimulation compared to the single Btk PH domain version (Fig. 3C). This finding is in line with a recent study reporting on the complex dynamics of Btk lipid binding and activation which features a sharp sensitivity for PIP₃ density in the PM (36).

We started our studies with HEK293-AT1 cells, a cell line stably expressing type 1A angiotensin receptor, which has been used by our group in several previous studies (27, 37, 38). However, we wanted to check our findings in a more generally used cell line as well, which led us to use HEK293A cells in subsequent experiments. Although these two cell lines have common origins, we found striking differences in their PIP response to the same stimuli, especially in PIP₃ changes as measured by the GRP1-based sensor (Figs. 2A and 3C). We speculate that this discrepancy might be due to higher resting PI3K activity in HEK293-AT1 cells, supported by elevated GRP1-PH basal BRET ratios in these cells (compare Figs. 1C and 3B). However, we did not investigate this further as our main focus was the divergent responses evoked by EGF and insulin, which were more consistent in both cell lines as determined by the PI(3,4)P₂/PIP₃ ratio (see Figs. 2C and 3E).

Our results indicate that SHIP2, using PIP₃ generated by class I PI3Ks, serves as the primary source of PI(3,4)P₂ in the EGF signaling pathway, rather than class II PI3Ks (Fig. 4B). This is in line with the notion that class II PI3K does not contribute significantly to signaling-related PI(3,4)P₂ dynamics (28) but rather generates localized, small pools of PI(3,4)P₂ at endocytic vesicles (39). Interestingly, however, siRNA-mediated knockdown of class II PI3K resulted in an increase in PI(3,4)P₂ levels following EGF stimulation. Since class II PI3K is one of the main enzymes involved in PI(3,4)P₂ synthesis (14) and is known to be recruited and activated by EGFR (40, 41), this finding was surprising. A plausible explanation for this result could lie in the critical role of Class II PI3K in receptor endocytosis (14). The knockdown of class II PI3K may disrupt EGFR internalization upon EGF stimulation, thereby prolonging EGFR activation at the PM and enhancing PI(3,4)P₂ production.

To assess SHIP2 regulation in response to EGF and insulin, we examined its phosphorylation status, specifically at tyrosine residues 986/987 and 1135. Our observations confirm phosphorylation at these sites following EGF stimulation (Fig. 5A), aligning with previous studies (42). We further explored the temporal dynamics of SHIP2 phosphorylation across different EGF exposure times, noting discernible phosphorylation as early as 5 s post-stimulation, which steadily increased, peaking at 10 min (Fig. 5, C and D). The impact of SHIP2 tyrosine phosphorylation on its phosphatase activity remains a topic of debate. While some studies suggest that it enhances the phosphatase activity of SHIP2 (42), others propose that its activity is regulated more by changes in subcellular localization than by its phosphorylation state (43). Our results demonstrate that SHIP2 tyrosine phosphorylation coincides with significant PIP₃-to-PI(3,4)P₂ conversion in the EGF signaling pathway, as shown by BRET lipid measurements, suggesting that phosphorylation may enhance SHIP2 function, either by modulating its localization or by directly increasing its phosphatase activity.

To confirm our results on a broader spectrum of conditions and cellular contexts, we used a bioinformatics approach where gene expression changes after perturbations (genetic or chemical activation or inactivation) of specific proteins were compared across many cell types and different conditions, based on an existing database named LINCS-L1000. We set out to determine the correlation between SHIP2 perturbation induced gene expression profiles and that of other proteins, and after quantification, we observed much higher correlation scores for EGFR perturbations compared to insulin receptor (IR) manipulations (blue and red triangles on Fig. 7, respectively). This confirmed a stronger connection between the signaling networks of SHIP2 and EGFR overall. However, we also detected negative scores (representing anti-correlation) which highlight that the interaction between SHIP2, and these receptors can be highly dependent on the cellular context. Furthermore, the cell lines in the LINCS-L1000 database are mostly of tumor origin and their gene regulatory machinery is likely to be damaged. In addition, the time course of the response of the gene expression level to a perturbation is different for the different genes in the profile. The actual time point presented in the database may not optimally capture the similarity between two patterns. Despite these limitations, we believe that this statistically robust method supports the involvement of SHIP2 in the EGF response.

We demonstrated that EGF-induced SHIP2 phosphorylation at the investigated tyrosine residues is partially reduced by both calcium chelation and PI3K inhibition, with varying effects at each site (Fig. 8E). This variation may be due to differences in antibody sensitivity or distinct regulatory mechanisms affecting each phosphorylation site. Given the varied impacts of phosphorylation on SHIP2 function, where some sites correlate with increased degradation (44) and others with enhanced enzymatic activity (29, 42), further research is essential to delineate the specific roles and regulatory mechanisms of each site. As the calcium signal appears to be the dominant driver of SHIP2 phosphorylation, we used a PLC inhibitor prior to EGF stimulation, which revealed a nearly complete inhibition of SHIP2 phosphorylation as well as a significantly reduced PI(3,4)P₂/PIP₃ ratio (Fig. 8, F–H). This highlights the potential impact of SHIP2 phosphorylation on its phosphatase activity. Moreover, while some studies have suggested that the kinase responsible for SHIP2 phosphorylation is likely a Src family member (43, 44), additional studies are required to identify the specific kinase involved. Since SHIP2 is being targeted pharmacologically, a thorough understanding of its regulatory mechanisms is crucial for therapeutic development (21).

In this study, we utilized U73122 to inhibit PLC to investigate its influence on SHIP2 phosphorylation within the EGF signaling pathway (Fig. 8, F–H). However, despite its widespread use, U73122 is notorious for its off-target effects (45, 46) and inconsistent efficacy, with some reports even suggesting paradoxical activation rather than inhibition of certain PLC isoforms in vitro (47). These effects are likely contingent on the specific conditions under which U73122 interacts with PLC. It has been proposed that while U73122 may activate cytosolic PLC enzymes, it could inhibit membrane-associated PLCs (47), the latter being the focus of our in vivo studies. Furthermore, there is no valid pharmacological alternative for targeting PLC, with the other options also known for their problematic nonspecific inhibition (48, 49). Therefore, U73122 was used as a complementary approach to calcium chelation, both of which had similar effects on EGF-induced SHIP2 phosphorylation. Nevertheless, due to the known limitations of U73122, these findings must be interpreted with caution. This highlights the critical need for further research to clearly define the role of PLC in inducing SHIP2 phosphorylation within the EGF pathway.

Our study did not detect SHIP2 phosphorylation at the examined sites following insulin stimulation, even with a tenfold increase in concentration (Fig. 5). In contrast, a previous study has observed SHIP2 phosphorylation in response to insulin, in a model with stable overexpression of the insulin receptor (50). Another study found that this insulin-induced phosphorylation varied even between different developmental stages of the same cell line (51). Taken together, these findings suggest that SHIP2 phosphorylation may depend on specific experimental conditions and cellular context. Despite these discrepancies, the role of SHIP2 as a key negative regulator in insulin signaling is well established. This important role is well demonstrated by SHIP2 overexpression being associated with insulin resistance, while targeting it has been shown to improve insulin sensitivity and glucose homeostasis, making it a potential therapeutic target for managing insulin resistance and metabolic disorders (21, 52).

EGFR overexpression in our experiments in HEK293-derived cell lines was necessary due to variability in endogenous receptor levels. As this may alter downstream signaling, we performed BRET-based lipid measurements with cells expressing either endogenous EGFR or the two overexpression levels, driven by either the TK or CMV promoter. All expression levels showed a consistent phosphoinositide pattern upon EGF stimulation, with higher EGFR expression leading to more sustained lipid changes. Notably, PIP dynamics were nearly identical in case of endogenous and TK-driven overexpression, suggesting that the TK promoter closely mirrors physiological conditions (Fig. 5E). To validate these findings beyond HEK cells, we utilized HeLa cells, chosen for their stable, robust response to both EGF and insulin through endogenous receptors. BRET lipid measurements and immunoblotting results in HeLa cells (Fig. 6) were consistent with those observed in the EGFR-overexpressing HEK cells. This confirms that the PIP patterns and SHIP2 dynamics identified are not unique to modified HEK cell lines but are also present in cells with native EGFR expression, thus underscoring the broader applicability of our findings.

Our BRET measurements, together with the PI(3,4)P₂/PIP₃ ratio analysis, reveal a preferential production of PI(3,4)P₂ in the EGF signaling pathway compared to insulin signaling. Given that PI(3,4)P₂ is a well-established signaling molecule involved in various cellular processes, these distinct dynamics in PI(3,4)P₂ production may contribute to downstream functional differences between the two pathways. However, experimentally evaluating these differences poses a significant challenge, as the role of PI(3,4)P₂ is highly context-dependent. PI(3,4)P₂ has been shown to play a crucial role in diverse cellular contexts, such as platelet function (53) and cataract prevention (54), which limits the applicability of testing its effects in our current system. Nevertheless, one pathway we were able to evaluate was the impact of differential PI(3,4)P₂ and PIP₃ production in the EGF and insulin signaling pathways on the activation of Akt isoforms. Since Akt1 is predominantly activated by PIP₃ and Akt2 by PI(3,4)P₂ (55), we hypothesized that EGF signaling would preferentially activate Akt2 compared to insulin. To test this, we analyzed Akt isoform activation using immunoblotting for Akt phosphorylation. However, our results revealed no significant difference in the phosphorylation patterns of Akt isoforms between EGF and insulin (data not shown). The precise role of preferential PI(3,4)P₂ production in modulating signaling pathways remains unclear and warrants further investigation.

Experimental procedures

Materials

Molecular biology reagents, Lipofectamine 2000 (Cat. No. 11668019) and Lipofectamine RNAiMAX (Cat. No. 13778150), were purchased from Thermo Fisher Scientific. Coelenterazine-H was purchased from Regis Technologies (1–361304–200). BAPTA-AM (Cat. No. 2787), U73122 (Cat. No. 1268), and U73343 (Cat. No. 4133) were from Tocris, and insulin was from Eli Lilly. Unless otherwise stated, all other chemicals and reagents were purchased from Sigma-Aldrich.

DNA constructs

To prepare the CMV EGFR and TK EGFR constructs, the coding sequence of the previously used EGFR was cloned to the Clontech pEGFP-N1 vector containing either the cytomegalovirus (CMV) or the thymidine kinase (TK) promoters (56). Using the NheI and Bsp1407I restriction enzymes, the EGFP sequence was replaced with the receptor sequence. In addition to the previously created and characterized PM PIP₃ and PI(3,4)P₂ BRET-based biosensors (L10-mVenus-T2A-Btk-PH-Luciferase, L10-mVenus-T2A-NES-Luciferase-GRP1-PH and L10-mVenus-T2A-Luciferase-TAPP1-2xPH) (27, 57) we created the tandem version of Btk-PH-containing biosensor (L10-mVenus-T2A-Btk-2xPH-Luciferase) by inserting the sequence of a second PH domain between the PH and luciferase sequences using the AgeI restriction site. In the construct, the two PH domains are connected by the DPPVAAGAGGAG linker. Using the same strategy, mVenus-tagged Btk-2xPH constructs was also created. Mammalian expression vectors coding wild-type (WT) and phosphatase domain deleted (PD) human SHIP2 were purchased from Addgene (No. 214907 and 214,901) (58). All constructs were checked by sequencing.

Cell culture

HEK293-AT1 cells (a line that stably expresses the type-1A rat angiotensin II receptor (AT1AR) (37)), HEK293A (Thermo Fisher Scientific, Cat. No. R70507) and HeLa cells (ATCC, CCL-2) were maintained in Dulbecco’s modified Eagle’s medium (DMEM, Gibco 11,995,065) supplemented with 10% fetal bovine serum (Biosera, FB-1200), 50 U/ml penicillin and 50 μg/ml streptomycin (Gibco, 15,140–122). HEK293-AT1 cells were also treated with 5 μg/ml Plasmocin (InvivoGen, ant-mpp) as a mycoplasma prophylactic and were regularly tested for mycoplasma contamination. All cell lines were cultured in a 5% humidified CO₂ incubator at 37 °C in 10 cm tissue culture plastic dishes.

Confocal microscopy

HEK293-AT1 or HEK293A cells were seeded at a density of 3 × 10⁴ cells/well on poly-L-lysine-pretreated (0.001%, 1 h) IBIDI 8-well μ-Slides (Cat. No. 80826) and cultured at 37 °C in DMEM for 1 day. For transfection, the culture medium was changed to 200 μl transfection mixture containing the indicated DNA constructs (0.2 μg total DNA/well) and 0.33 μl/well Lipofectamine 2000. After 6 h, the medium was changed to 300 μl supplemented DMEM medium. Cells were inspected 24 to 26 h after transfection using a Zeiss LSM 710 laser confocal microscope with a 63 × /1.4 oil-immersion objective. Cells were serum-depleted for 5 h where indicated and medium was changed to modified Krebs–Ringer buffer (120 mM NaCl, 4.7 mM KCl, 1.2 mM CaCl₂, 0.7 mM MgSO₄, 10 mM glucose and 10 mM Na-HEPES, pH 7.4) at room temperature before experiments. Post-acquisition picture analysis was performed using Fiji and Photoshop (Adobe) software. Only linear changes were made during picture analysis and processing.

Bioluminescence resonance energy transfer (BRET) measurements

For BRET measurements, HEK293-AT1, HEK293A or HeLa cells were trypsinized (0.25% Trypsin, Lonza BE17–160E) and plated on poly-L-lysine-pretreated (0.001%, 1 h) white 96-well plates (Greiner bio-one, 655,083) at a density of 6 × 10⁴ cells/well together with the indicated DNA constructs (0.03–0.04 μg of total DNA/well) and the cell transfection reagent (0.25 μl/well Lipofectamine 2000) in Opti-MEM reduced serum medium (Gibco, 31,985,047). Cells were serum-starved (5 h) before starting the BRET measurements, which were performed 28 h after transfection. Before the measurements, the medium over the cells was changed to a modified Krebs–Ringer buffer (50 μl) (described above) for 1 h at 37 °C. Measurements were performed at 37 °C using a ThermoFisher Scientific Varioskan LUX Multimode Microplate Reader, and started with the addition of the cell-permeable luciferase substrate coelenterazine-H (40 μl; final concentration of 5 μM), and counts were recorded using 480 nm and 530 nm emission filters. Detection time was 250 ms for each wavelength. The indicated reagents were also dissolved in modified Krebs-Ringer buffer and added (20 μl). All measurements were done in two biological replicates. To calculate changes in the PM lipid levels, BRET ratios were calculated by dividing the 530 nm and 480 nm intensities and normalized to the baseline. Because the absolute ratio values depended on the expression of the sensors in the case of the intermolecular inositol lipid sensors, the resting levels were considered as 100%, whereas 0% was determined from the values of those experiments where cytoplasmic luciferase construct was expressed alone.

Knockdown experiments

HEK293-AT1 or HEK293A cells were plated on 6-well plates at a density of 1.5 × 10⁵ cells/well. One day later, the medium was changed to fresh culture medium and cells were transfected with 200 μl transfection mixture containing 6 μl/well Lipofectamine RNAiMAX and the indicated siRNAs (120 pmol/well). All siRNAs were ON-TARGETplus SMARTpool (Horizon Discovery) mixes of 4 different sequences against the target gene (L-004152–00–0005 for SHIP2, L-006771–00–0005 for PI3KC2A, L-006772–00–0005 for PI3KC2B) and we used ON-TARTGETplus Non-targeting Pool as scrambled control (D-001810–10–05). 48 h later, DNA transfection and BRET measurements or immunoblotting were carried out as detailed in the relevant sections.

Immunoblotting

Cells were cultured on 6-well plates (4 × 10⁵ cells/well) in cDMEM at 37 °C. One day later, HEK293A cells were transfected by adding 200 μl transfection mixture containing the indicated DNA constructs (0.5 μg DNA total/well) and 2 μl/well Lipofectamine 2000. 23 h after plating (HeLa) or transfection (HEK293A), cells were made quiescent by incubation in serum-free DMEM for 5 h. After agonist stimulation, reactions were stopped by placing the 6-well plates on ice and washing each well with ice-cold modified Krebs–Ringer buffer (described above). Cells were then scraped with 2 × concentrated Laemmli buffer supplemented with 2-mercapto-ethanol (5%) and protease and phosphatase inhibitors (cOmplete protease inhibitor cocktail, Roche 11,836,145,001). This cell lysate was briefly sonicated, then boiled at 95 °C for 10 min, and equal amounts of samples were loaded into 4 to 15% SDS-polyacrylamide gels (BioRad, 4,561,083). Proteins were transferred to PVDF membranes (BioRad, 1,704,156). Membranes were blocked with a 5% blocking solution (fat-free milk powder diluted in TBST) for 1 h at room temperature and incubated with the primary antibody (diluted 1:1000 in TBST containing 5% fat-free milk powder) overnight at 4 °C. Membranes were washed 3 times with TBST for 10 min, then incubated with HRP-linked secondary antibody (anti-rabbit, Cell Signaling #7074, or anti-mouse, Cell Signaling #7076, diluted 1:5000 in TBST containing 5% fat-free milk powder) for 1 h at room temperature and washed 3 times. The signals were visualized with enhanced chemiluminescence, using Immobilon Western HRP substrate reagent, and were then detected with an Azure c600 device. All immunoblot images shown are representative of at least 3 independent experiments. Bands were quantified by measuring the total pixel intensity with background subtraction using ImageJ. The primary antibodies used were rabbit anti-phospho-SHIP2 (Tyr-986/987, Cell Signaling, #2008), rabbit anti-phospho-SHIP2 (Tyr-1135, Thermo Fisher Scientific, PA5104651), rabbit anti-SHIP2 (Cell Signaling, #2839), rabbit anti-Akt (Cell Signaling, #4691), rabbit anti-phospho-Akt (Ser-473, Cell Signaling, #4060), rabbit anti-ERK (Cell Signaling, #9102), mouse anti-phospho-ERK (Thr-202/Tyr-204, Cell Signaling, #9106), mouse anti-EGFR (Santa Cruz, sc-373746), and mouse anti-phospho-EGFR (Tyr-1068, Cell Signaling, #2236).

Bioinformatics analysis

The LINCS-L1000 database (https://lincsproject.org/LINCS/) is a collaborative effort for the systematic screening of gene expression pattern changes of selected human cell lines responding to various treatments (perturbations). These treatments include drug induced selective activation or inhibition of various proteins as well as knock-out or knock-down of selected genes via CRISPR or siRNA techniques. The cell lines, the drugs, the experimental protocols, all the related parameters and data processing steps are standardized across the project. The results are collected in the L-1000 database as normalized sets of gene expression change profiles of the given perturbation.

From the database, the INPPL1 (the gene symbol of SHIP2) perturbation profiles were selected. The selection was further narrowed down to cell lines containing data for more than one perturbation experiment for SHIP2. This restriction resulted in 16 cell lines, mostly of tumor origin. Their perturbation profiles were screened for similarity with the ones of SHIP2 using Spearman correlation as the metrics of similarity. Gene set enrichment analysis was applied on the gene similarity sets using the python “decoupler” module with Z-score normalization. In the case of ligand induced perturbation experiments activation and inhibition data were treated separately (59). The procedure resulted in a ranked list of protein perturbation profiles, correlating or anti-correlating with the perturbation profiles of SHIP2 (positive and negative values, respectively).

Mathematical and statistical analysis

SigmaPlot 10.0 was used for statistical analysis, the tests applied (including post hoc tests after ANOVA) for each comparison are indicated in the figure legends. The program automatically tested for normal distribution and equal variance of the samples, and t test or ANOVA were used only when both tests passed, otherwise non-parametric tests (e.g. ANOVA on Ranks) were used.

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Conflict of interest

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

Reviewed by members of the JBC Editorial Board. Edited by Kirill Martemyanov