Ribosomal s6 kinase is a mediator of aquaporin-2 S256 phosphorylation and membrane accumulation after EGFR inhibition with erlotinib

Abstract

Vasopressin (VP) activates protein kinase A (PKA), resulting in phosphorylation events and membrane accumulation of aquaporin-2 (AQP2). Epidermal growth factor receptor (EGFR) inhibition with erlotinib also induces AQP2 membrane trafficking with a phosphorylation pattern similar to VP, but without increasing PKA activity. Here, we identify the ribosomal s6 kinase (RSK) as a major mediator phosphorylating AQP2 in this novel, erlotinib-induced pathway. We found that RSK was expressed in collecting duct principal cells in rat kidneys. RSK inhibition with BI-D1870 blocked erlotinib-induced AQP2 serine 256 (S256) phosphorylation and membrane accumulation. CRISPR-generated RSK knockout (KO) cells failed to show increased S256 phosphorylation in response to erlotinib. Like PKA, RSK was able to phosphorylate AQP2 S256 in vitro. Inhibition of phosphoinositide-dependent kinase-1 (PDK1), a known activator of RSK, blocked erlotinib-induced AQP2 S256 phosphorylation and membrane accumulation. We conclude that RSK is a crucial terminal kinase phosphorylating AQP2 at S256 upon EGFR inhibition by erlotinib.NEW & NOTEWORTHY Epidermal growth factor receptor (EGFR) inhibition with erlotinib induces aquaporin-2 (AQP2) membrane accumulation with a phosphorylation pattern similar to vasopressin (VP). Here, we identify the ribosomal s6 kinase (RSK) as a major mediator phosphorylating AQP2 in this novel, erlotinib-induced pathway. In addition, we show that phosphoinositide-dependent kinase-1 (PDK1), a known activator of RSK, is implicated in this pathway: PDK1 inhibition blocks erlotinib-induced AQP2 S256 phosphorylation and membrane accumulation.

Keywords: EGFR; aquaporin-2; epithelial transport; vasopressin; vesicle trafficking.

Figure 1.. Inhibition of RSK decreases erlotinib-induced S256 phosphorylation of AQP2 in LLC-PK1 cells.

Western blots were performed on lysates of LLC-PK1 cells stably expressing AQP2 (LLC-AQP2) and stained with specific phospho-AQP2 antibodies. Phosphoserine band intensities were quantified and normalized to their respective total AQP2 loading controls. Erlotinib increased S256 phosphorylation similar to VP, and this effect was significantly decreased in cells pre-treated with the RSK inhibitor BI-D1870 (Fig.1, top). The experiment was repeated five times in duplicate (n=5) and quantified (Fig.1, bottom).

Figure 2.. RSK inhibition prevents erlotinib-induced AQP2 membrane accumulation in LLC-AQP2 cells.

Erlotinib-induced AQP2 membrane accumulation in LLC-AQP2 cells was blocked by RSK inhibition. As expected, both VP and erlotinib (Erl) induced an increase in AQP2 membrane accumulation (Fig.2A, top row). When pre-treated with RSK inhibitor BI-D1870, erlotinib-induced AQP2 membrane accumulation was inhibited, and AQP2 was mostly in the cytoplasm (Fig.2A, bottom right panel), similar to untreated cells (Fig.2A, top left panel). BI-D1870 did not inhibit VP-induced AQP2 membrane accumulation (Fig.2A, bottom middle) and BI-D1870 alone did not cause visible changes in AQP2 localization (Fig.2A, bottom left). Quantification showed that pre-treatment with BI-D1870 is indeed able to block an increase in AQP2 in the membrane fraction in response to erlotinib (Fig.2B). Cells pre-treated with BI-D1870 still showed increased membrane accumulation of AQP2 in response to VP, though not to the extent of cells which received VP alone (Fig.2B). These results are representative of three independent experiments (n=3). Scale bar, 10 μM.

Figure 2.. RSK inhibition prevents erlotinib-induced AQP2 membrane accumulation in LLC-AQP2 cells.

Erlotinib-induced AQP2 membrane accumulation in LLC-AQP2 cells was blocked by RSK inhibition. As expected, both VP and erlotinib (Erl) induced an increase in AQP2 membrane accumulation (Fig.2A, top row). When pre-treated with RSK inhibitor BI-D1870, erlotinib-induced AQP2 membrane accumulation was inhibited, and AQP2 was mostly in the cytoplasm (Fig.2A, bottom right panel), similar to untreated cells (Fig.2A, top left panel). BI-D1870 did not inhibit VP-induced AQP2 membrane accumulation (Fig.2A, bottom middle) and BI-D1870 alone did not cause visible changes in AQP2 localization (Fig.2A, bottom left). Quantification showed that pre-treatment with BI-D1870 is indeed able to block an increase in AQP2 in the membrane fraction in response to erlotinib (Fig.2B). Cells pre-treated with BI-D1870 still showed increased membrane accumulation of AQP2 in response to VP, though not to the extent of cells which received VP alone (Fig.2B). These results are representative of three independent experiments (n=3). Scale bar, 10 μM.

Figure 3.. Erlotinib fails to induce AQP2 S256 phosphorylation in RSK knockout cells.

An RSK knockout in the LLC-AQP2 cell line (LLC-AQP2^{RSK KO}) was created using CRISPR/Cas9 and compared to regular LLC-AQP2 cells with intact RSK. Knockout was verified by western blot with clone 3 showing successful knockout (Fig.3 A). Both LLC-AQP2 cells and LLC-AQP2^{RSK KO} cells were treated with erlotinib or VP for 30 or 15 min respectively. Western blots were performed on lysates of cells and stained with specific phospho-AQP2 antibodies. Both cell lines showed increased AQP2 phosphorylation in response to the positive control vasopressin, while only the cells with intact RSK demonstrated a significant increase in AQP2 S256 in response to erlotinib (Fig.3B). This result was repeated six times (n=6) in duplicate and quantified (Fig.3C).

Figure 3.. Erlotinib fails to induce AQP2 S256 phosphorylation in RSK knockout cells.

An RSK knockout in the LLC-AQP2 cell line (LLC-AQP2^{RSK KO}) was created using CRISPR/Cas9 and compared to regular LLC-AQP2 cells with intact RSK. Knockout was verified by western blot with clone 3 showing successful knockout (Fig.3 A). Both LLC-AQP2 cells and LLC-AQP2^{RSK KO} cells were treated with erlotinib or VP for 30 or 15 min respectively. Western blots were performed on lysates of cells and stained with specific phospho-AQP2 antibodies. Both cell lines showed increased AQP2 phosphorylation in response to the positive control vasopressin, while only the cells with intact RSK demonstrated a significant increase in AQP2 S256 in response to erlotinib (Fig.3B). This result was repeated six times (n=6) in duplicate and quantified (Fig.3C).

Figure 3.. Erlotinib fails to induce AQP2 S256 phosphorylation in RSK knockout cells.

An RSK knockout in the LLC-AQP2 cell line (LLC-AQP2^{RSK KO}) was created using CRISPR/Cas9 and compared to regular LLC-AQP2 cells with intact RSK. Knockout was verified by western blot with clone 3 showing successful knockout (Fig.3 A). Both LLC-AQP2 cells and LLC-AQP2^{RSK KO} cells were treated with erlotinib or VP for 30 or 15 min respectively. Western blots were performed on lysates of cells and stained with specific phospho-AQP2 antibodies. Both cell lines showed increased AQP2 phosphorylation in response to the positive control vasopressin, while only the cells with intact RSK demonstrated a significant increase in AQP2 S256 in response to erlotinib (Fig.3B). This result was repeated six times (n=6) in duplicate and quantified (Fig.3C).

Figure 4.. Knockout of RSK inhibits erlotinib-induced AQP2 membrane accumulation.

RSK KO cells showed no increased AQP2 membrane accumulation in response to erlotinib. AQP2 membrane trafficking was assessed in both LLC-AQP2 cells and LLC-AQP2^{RSK KO} cells using immunocytochemistry. Both cell lines were treated with erlotinib, or VP for 30 or 15 min respectively. While vasopressin elicited AQP2 membrane accumulation in both LLC-AQP2 cells and LLC-AQP2^{RSK KO} cells (Fig.4A middle column), erlotinib induced an increase in plasma membrane AQP2 only in cells with intact RSK (Fig.4A right column). Scale bar, 10 μM. The membrane fraction of AQP2 was quantified and showed no significant increase in erlotinib-induced membrane accumulation in the LLC-AQP2^{RSK KO} cells whereas all other conditions showed an intact membrane response (Fig.4B). Immunocytochemistry and quantification were repeated three times (n=3).

Figure 4.. Knockout of RSK inhibits erlotinib-induced AQP2 membrane accumulation.

RSK KO cells showed no increased AQP2 membrane accumulation in response to erlotinib. AQP2 membrane trafficking was assessed in both LLC-AQP2 cells and LLC-AQP2^{RSK KO} cells using immunocytochemistry. Both cell lines were treated with erlotinib, or VP for 30 or 15 min respectively. While vasopressin elicited AQP2 membrane accumulation in both LLC-AQP2 cells and LLC-AQP2^{RSK KO} cells (Fig.4A middle column), erlotinib induced an increase in plasma membrane AQP2 only in cells with intact RSK (Fig.4A right column). Scale bar, 10 μM. The membrane fraction of AQP2 was quantified and showed no significant increase in erlotinib-induced membrane accumulation in the LLC-AQP2^{RSK KO} cells whereas all other conditions showed an intact membrane response (Fig.4B). Immunocytochemistry and quantification were repeated three times (n=3).

Figure 5.. There is predominance of RSK in the collecting duct epithelium, and RSK co-localizes with AQP2 in principal cells.

Immunohistochemistry using specific antibodies to detect RSK (red, left column) and AQP2 (green, middle column). Merged images are shown in the right column. The middle row shows a section of cortex while the bottom row shows a section of inner medulla. The top row shows a higher magnification of the cortical sections for detail. RSK is expressed in collecting ducts in both principal cells (PC, arrow, upper right panel) and intercalated cells (IC, arrow, upper right panel), identified by the presence and absence of AQP2 staining, respectively. Expression in intercalated cells is somewhat higher than in principal cells, and RSK is also expressed at lower levels in other as yet unidentified renal cells in the medulla (lower panels). These images are a representative of three independent experiments on three different animals (n=3). Scale bar, 10 μM.

Figure 6.. Inhibition of RSK decreases erlotinib-induced membrane accumulation of AQP2 in rat kidneys.

Kidney slices were pre-treated with the RSK inhibitor, BI-D1870, followed by erlotinib treatment for 30 min. Apical membrane accumulation of AQP2 was seen in kidney slices treated with vasopressin (VP, B) and erlotinib (Erl, C), but pre-treatment with BI-D1870 abolished Erl-induced apical membrane accumulation and AQP2 instead remained diffusely distributed in the cytoplasm of principal cells (D). Arrows indicate apical accumulation of AQP2 in erlotinib treated kidney slice (C). These images are representative of four independent experiments from 4 different animals (n=4).

Figure 7.. RSK can directly phosphorylate AQP2 at S256.

A C-terminal tail (~7kDa) construct of AQP2 was incubated with RSK and radioisotope ³²P in vitro. As a positive control, the AQP2 c-tail was incubated with recombinant protein kinase A (PKA, left three lanes), and as expected our assay demonstrated phosphorylation of the AQP2 c-tail. A serine to alanine mutation behaves functionally as a permanently dephosphorylated residue, and recombinant AQP2 c-tail with this S256A mutation failed to show phosphorylation by PKA. H89, a non-specific PKA inhibitor, prevented the PKA-mediated phosphorylation of the AQP2 c-tail. RSK (right three lanes) can also directly phosphorylate the AQP2 c-tail, similar to PKA, and cannot phosphorylate the c-tail S256A mutation. BI-D1870 inhibited the ability of RSK to phosphorylate the AQP2 c-tail. The nature of the higher MW bands in all the PKA lanes is unknown. This image is a representative of three independent experiments (n=3).

Figure 8.. Inhibition of Phosphoinositide-dependent kinase 1 (PDK1) abolishes erlotinib-induced AQP2 membrane accumulation.

Membrane trafficking of AQP2 was assessed with immunocytochemistry. Figure A: LLC-AQP2 cells were treated with or without GSK-470, a PDK1 inhibitor, for 120 min and treated with erlotinib for 30 min. Cells showed increased membrane accumulation of AQP2 with erlotinib, consistent with prior results (Fig.8A, top row). PDK1 inhibition with GSK-470 blocked erlotinib-induced AQP2 membrane trafficking (Fig.8A, bottom right). In cells treated with GSK-470 alone, AQP2 was mostly in the cytoplasm (Fig.8A, bottom left), similar to untreated cells. The experiment was repeated three times (n=3). Scale bar, 10μM. Figure B: LLC-AQP2 cells were treated with another PDK1 inhibitor, PS-222. Cells were incubated with or without PS-222 for 120 min and treated with erlotinib for 30 min. Similar to GSK-470, PS-222 inhibited erlotinib-induced membrane accumulation (Fig.8B, left). The experiment was repeated three times (n=3). Scale bar, 10μM. The membrane fraction of AQP2 was quantified and showed no significant increase in erlotinib-induced membrane accumulation in the cells pretreated with PS-222, while cells treated with erlotinib alone showed increased membrane fraction of AQP2 (Fig.8B, right). Immunocytochemistry and quantification were repeated three times (n=3). Figure C: HEK-293 cells were incubated with or without PS-222 for 120 minutes and stimulated with IGF-1 20 minutes prior to lysis. PS-222 inhibited phosphorylation of PIF pocket dependent substrate S6 ribosomal protein, but did not inhibit phosphorylation of PDK1 substrate Akt, which does not depend on PIF pocket binding (Fig. 8C). The experiment was repeated three times (n=3) and quantified.

Figure 8.. Inhibition of Phosphoinositide-dependent kinase 1 (PDK1) abolishes erlotinib-induced AQP2 membrane accumulation.

Membrane trafficking of AQP2 was assessed with immunocytochemistry. Figure A: LLC-AQP2 cells were treated with or without GSK-470, a PDK1 inhibitor, for 120 min and treated with erlotinib for 30 min. Cells showed increased membrane accumulation of AQP2 with erlotinib, consistent with prior results (Fig.8A, top row). PDK1 inhibition with GSK-470 blocked erlotinib-induced AQP2 membrane trafficking (Fig.8A, bottom right). In cells treated with GSK-470 alone, AQP2 was mostly in the cytoplasm (Fig.8A, bottom left), similar to untreated cells. The experiment was repeated three times (n=3). Scale bar, 10μM. Figure B: LLC-AQP2 cells were treated with another PDK1 inhibitor, PS-222. Cells were incubated with or without PS-222 for 120 min and treated with erlotinib for 30 min. Similar to GSK-470, PS-222 inhibited erlotinib-induced membrane accumulation (Fig.8B, left). The experiment was repeated three times (n=3). Scale bar, 10μM. The membrane fraction of AQP2 was quantified and showed no significant increase in erlotinib-induced membrane accumulation in the cells pretreated with PS-222, while cells treated with erlotinib alone showed increased membrane fraction of AQP2 (Fig.8B, right). Immunocytochemistry and quantification were repeated three times (n=3). Figure C: HEK-293 cells were incubated with or without PS-222 for 120 minutes and stimulated with IGF-1 20 minutes prior to lysis. PS-222 inhibited phosphorylation of PIF pocket dependent substrate S6 ribosomal protein, but did not inhibit phosphorylation of PDK1 substrate Akt, which does not depend on PIF pocket binding (Fig. 8C). The experiment was repeated three times (n=3) and quantified.

Figure 8.. Inhibition of Phosphoinositide-dependent kinase 1 (PDK1) abolishes erlotinib-induced AQP2 membrane accumulation.

Membrane trafficking of AQP2 was assessed with immunocytochemistry. Figure A: LLC-AQP2 cells were treated with or without GSK-470, a PDK1 inhibitor, for 120 min and treated with erlotinib for 30 min. Cells showed increased membrane accumulation of AQP2 with erlotinib, consistent with prior results (Fig.8A, top row). PDK1 inhibition with GSK-470 blocked erlotinib-induced AQP2 membrane trafficking (Fig.8A, bottom right). In cells treated with GSK-470 alone, AQP2 was mostly in the cytoplasm (Fig.8A, bottom left), similar to untreated cells. The experiment was repeated three times (n=3). Scale bar, 10μM. Figure B: LLC-AQP2 cells were treated with another PDK1 inhibitor, PS-222. Cells were incubated with or without PS-222 for 120 min and treated with erlotinib for 30 min. Similar to GSK-470, PS-222 inhibited erlotinib-induced membrane accumulation (Fig.8B, left). The experiment was repeated three times (n=3). Scale bar, 10μM. The membrane fraction of AQP2 was quantified and showed no significant increase in erlotinib-induced membrane accumulation in the cells pretreated with PS-222, while cells treated with erlotinib alone showed increased membrane fraction of AQP2 (Fig.8B, right). Immunocytochemistry and quantification were repeated three times (n=3). Figure C: HEK-293 cells were incubated with or without PS-222 for 120 minutes and stimulated with IGF-1 20 minutes prior to lysis. PS-222 inhibited phosphorylation of PIF pocket dependent substrate S6 ribosomal protein, but did not inhibit phosphorylation of PDK1 substrate Akt, which does not depend on PIF pocket binding (Fig. 8C). The experiment was repeated three times (n=3) and quantified.

Figure 9.. Inhibition of Phosphoinositide-dependent kinase 1 (PDK1) abolishes erlotinib-induced AQP2 S256 phosphorylation.

Cells were treated with or without PDK1 inhibitor PS-222 for 120 min and treated with erlotinib or VP for 30 or 15 min respectively. Western blots were performed on lysates of cells and stained with specific pS256-AQP2 antibodies. Erlotinib was unable to significantly increase S256-AQP2 phosphorylation in the presence of PDK1 inhibition, while vasopressin-induced S256 phosphorylation remined intact (Fig.9 top). The experiment was repeated four times in duplicate (n=4) and quantified (Fig.9 bottom).

Figure 9.. Inhibition of Phosphoinositide-dependent kinase 1 (PDK1) abolishes erlotinib-induced AQP2 S256 phosphorylation.

Cells were treated with or without PDK1 inhibitor PS-222 for 120 min and treated with erlotinib or VP for 30 or 15 min respectively. Western blots were performed on lysates of cells and stained with specific pS256-AQP2 antibodies. Erlotinib was unable to significantly increase S256-AQP2 phosphorylation in the presence of PDK1 inhibition, while vasopressin-induced S256 phosphorylation remined intact (Fig.9 top). The experiment was repeated four times in duplicate (n=4) and quantified (Fig.9 bottom).

Figure 10.. Erlotinib increases phosphorylation of RSK T359, a phosphorylation state associated with RSK activation.

LLC-AQP2 cells were treated with or without BI-D1870 for 30 min and then erlotinib for an additional 30 min. Western blots were performed on lysates of cells and stained with specific pT359-RSK antibodies. Erlotinib increased pT359, while cells pre-treated with BI-D1870 failed to show increased T359 in response to erlotinib (Fig.10 A). The experiment was repeated six times in duplicate (n=6) for erlotinib with BI-D1870, and eighteen times (n=18) in duplicate for erlotinib alone and was quantified (Fig. 10 B).

Figure 10.. Erlotinib increases phosphorylation of RSK T359, a phosphorylation state associated with RSK activation.

LLC-AQP2 cells were treated with or without BI-D1870 for 30 min and then erlotinib for an additional 30 min. Western blots were performed on lysates of cells and stained with specific pT359-RSK antibodies. Erlotinib increased pT359, while cells pre-treated with BI-D1870 failed to show increased T359 in response to erlotinib (Fig.10 A). The experiment was repeated six times in duplicate (n=6) for erlotinib with BI-D1870, and eighteen times (n=18) in duplicate for erlotinib alone and was quantified (Fig. 10 B).