Heat Shock Protein 90 Chaperone Regulates the E3 Ubiquitin-Ligase Hakai Protein Stability

The E3 ubiquitin-ligase Hakai binds to several tyrosine-phosphorylated Src substrates, including the hallmark of the epithelial-to-mesenchymal transition E-cadherin, and signals for degradation of its specific targets. Hakai is highly expressed in several human cancers, including colon cancer, and is considered as a drug target for cancer therapy. Here, we report a link between Hakai and the heat shock protein 90 (Hsp90) chaperone complex. Hsp90 participates in the correct folding of its client proteins, allowing them to maintain their stability and activity. Hsp90 inhibitors specifically interfere with the association with its Hsp90 client proteins, and exhibit potent anti-cancer properties. By immunoprecipitation, we present evidence that Hakai interacts with Hsp90 chaperone complex in several epithelial cells and demonstrate that is a novel Hsp90 client protein. Interestingly, by overexpressing and knocking-down experiments with Hakai, we identified Annexin A2 as a Hakai-regulated protein. Pharmacological inhibition of Hsp90 with geldanamycin results in the degradation of Hakai in a lysosome-dependent manner. Interestingly, geldanamycin-induced Hakai degradation is accompanied by an increased expression of E-cadherin and Annexin A2. We also show that geldanamycin suppresses cell motility at least in part through its action on Hakai expression. Taken together, our results identify Hakai as a novel Hsp90 client protein and shed light on the regulation of Hakai stability. Our results open the possibility to the potential use of Hsp90 inhibitors for colorectal cancer therapy through its action on Hakai client protein of Hsp90.


Introduction
Hsp90 (90-KDa heat shock protein) is a molecular chaperone involved in the correct folding of a selected group of proteins, named as client proteins, allowing them to maintain their proper conformation and the preservation of their activity [1]. The regulation of Hsp90 client proteins not only plays a crucial role in several cellular processes, such as cell cycle control, apoptosis and cell survival, but also contributes to the development of pathological conditions, such as neurodegenerative and Moreover, the analysis of Hakai interactome in HCT116, by employing endogenous Hakai immunoprecipitation following by nano-flow liquid chromatography (LC) coupled to a triple TOF Mass Spectrometer, also identified Hakai and Hsp90 proteins in Hakai immunoprecipitation samples compared to IgG control sample, further confirming the interaction between these two proteins. However, no effect on Hsp90 protein levels was detected when Hakai was transiently transfected in HEK293T cells in a concentration-dependent manner (Supplementary Figures S1a and S10) neither in a time-dependent manner (Figure S1b, Supplementary Figure S10). Similarly, we neither detected any effect on Hakai expression by transiently transfecting increasing amounts of Hsp90 (Supplementary Figures S1c and S10). Since we previously demonstrated that Hakai is highly expressed in human colon cancer tissues compared to healthy colon tissues, we also analyzed the possible co-localization by using HT29, LoVo and HCT116 colon cancer cell lines. Endogenous Hakai is highly detected in the nucleus and less intense signal is observed in the cytoplasm, where it is described to exert its E3 ubiquitin-ligase activity. Hsp90 is clearly localized throughout the whole cytoplasm where co-localization with Hakai is slightly detected in HT29 and Lovo cells, with an enriched signal detected in perinuclear areas (Supplementary Figure S2).

Interaction between Hsp90, Hakai and Annexin A2
It is generally reported that chaperone-interacting E3 ubiquitin-ligases induce the degradation of the chaperone client proteins that have not been correctly folded [22][23][24]. Therefore, it is reasonable to assume that additional E3 ubiquitin-ligases are involved in the degradation of HSP90 clients. Given the interaction confirmed between Hakai and Hsp90, our findings led us to ask whether Hakai may also participate in the regulation of other Hsp90 client proteins. Annexin A2 was one of the proteins Moreover, the analysis of Hakai interactome in HCT116, by employing endogenous Hakai immunoprecipitation following by nano-flow liquid chromatography (LC) coupled to a triple TOF Mass Spectrometer, also identified Hakai and Hsp90 proteins in Hakai immunoprecipitation samples compared to IgG control sample, further confirming the interaction between these two proteins. However, no effect on Hsp90 protein levels was detected when Hakai was transiently transfected in HEK293T cells in a concentration-dependent manner (Supplementary Figures S1a and S10) neither in a time-dependent manner (Figure S1b, Supplementary Figure S10). Similarly, we neither detected any effect on Hakai expression by transiently transfecting increasing amounts of Hsp90 (Supplementary Figures S1c and S10). Since we previously demonstrated that Hakai is highly expressed in human colon cancer tissues compared to healthy colon tissues, we also analyzed the possible co-localization by using HT29, LoVo and HCT116 colon cancer cell lines. Endogenous Hakai is highly detected in the nucleus and less intense signal is observed in the cytoplasm, where it is described to exert its E3 ubiquitin-ligase activity. Hsp90 is clearly localized throughout the whole cytoplasm where co-localization with Hakai is slightly detected in HT29 and Lovo cells, with an enriched signal detected in perinuclear areas (Supplementary Figure S2).

Interaction between Hsp90, Hakai and Annexin A2
It is generally reported that chaperone-interacting E3 ubiquitin-ligases induce the degradation of the chaperone client proteins that have not been correctly folded [22][23][24]. Therefore, it is reasonable to assume that additional E3 ubiquitin-ligases are involved in the degradation of HSP90 clients. Given the interaction confirmed between Hakai and Hsp90, our findings led us to ask whether Hakai may also participate in the regulation of other Hsp90 client proteins. Annexin A2 was one of the proteins detected in a proteomic study that turned out to be downregulated in Hakai-overexpressing MDCK Cancers 2020, 12, 215 4 of 20 cells compared to non-transformed MDCK epithelial cells [21]. In addition, it has been reported that Annexin A2 co-immunoprecipitates with Hsp90 [25]. Annexin A2 is a calcium-binding protein reported to be implicated in membrane and vesicle trafficking [26]. Considering these premises, we further analyze the possible interaction between these three proteins. Endogenous interaction of Hsp90 with Hakai and Annexin A2 was confirmed in HCT116 cells. Immunoprecipitation of endogenous Hakai co-precipitated with endogenous Hsp90 (Figure 2a, Supplementary Figure S5). Endogenous Annexin A2 also co-precipitated with endogenous Hakai (Figure 2b, Supplementary Figure S5). Conversely, immunoprecipitation of Annexin A2 also co-precipitated with endogenous Hsp90 (Figure 2c, Supplementary Figure S5). These results confirm that Hakai, Hsp90 and Annexin A2 form an interacting protein complex, either mediated by a direct or indirect interaction.
Cancers 2020, 12, x FOR PEER REVIEW 4 of 20 detected in a proteomic study that turned out to be downregulated in Hakai-overexpressing MDCK cells compared to non-transformed MDCK epithelial cells [21]. In addition, it has been reported that Annexin A2 co-immunoprecipitates with Hsp90 [25]. Annexin A2 is a calcium-binding protein reported to be implicated in membrane and vesicle trafficking [26]. Considering these premises, we further analyze the possible interaction between these three proteins. Endogenous interaction of Hsp90 with Hakai and Annexin A2 was confirmed in HCT116 cells. Immunoprecipitation of endogenous Hakai co-precipitated with endogenous Hsp90 (Figure 2a, Supplementary Figure S5). Endogenous Annexin A2 also co-precipitated with endogenous Hakai (Figure 2b, Supplementary Figure S5). Conversely, immunoprecipitation of Annexin A2 also co-precipitated with endogenous Hsp90 (Figure 2c, Supplementary Figure S5). These results confirm that Hakai, Hsp90 and Annexin A2 form an interacting protein complex, either mediated by a direct or indirect interaction. Hakai, Annexin A2, and Hsp90 interact with other: The whole cell extracts of HCT116 cell line cells were prepared and subjected to a co-immunoprecipitation assay. Subsequent westernblotting analysis was performed by using Hakai, Annexin A2, or Hsp90 antibodies (a) Hakai is specifically detected in anti-Hsp90 immunoprecipitates. (b) Annexin A2 is specifically detected in anti-Hakai immunoprecipitates. (c) Hsp90 is specifically detected in anti-Annexin A2 immunoprecipitates. GAPDH signal was used as protein loading control.

Hakai Regulates Annexin A2 Protein Expression
In order to understand the meaning of the interaction between Hakai, Annexin A2 and Hsp90, first, we decided to confirm the previous proteomic study on which Annexin A2 was downregulated in stably Hakai-overexpressing MDCK cells compared to normal cells [21]. On one hand, it is reported that Hakai interacts with Src tyrosine-phosphorylated substrates, an on the other hand, Annexin A2 is tyrosine-phosphorylated by Src kinase [12,27]. We carried out a transient transfection of Flag-Hakai together with Src and HA-ubiquitin in HEK293T cells to favor the activity of the ubiquitin-mediated degradation system. Besides, co-overexpression of Hakai with Src favors the increase of Hakai levels Figure 2. Hakai, Annexin A2, and Hsp90 interact with other: The whole cell extracts of HCT116 cell line cells were prepared and subjected to a co-immunoprecipitation assay. Subsequent western-blotting analysis was performed by using Hakai, Annexin A2, or Hsp90 antibodies (a) Hakai is specifically detected in anti-Hsp90 immunoprecipitates. (b) Annexin A2 is specifically detected in anti-Hakai immunoprecipitates. (c) Hsp90 is specifically detected in anti-Annexin A2 immunoprecipitates. GAPDH signal was used as protein loading control.

Hakai Regulates Annexin A2 Protein Expression
In order to understand the meaning of the interaction between Hakai, Annexin A2 and Hsp90, first, we decided to confirm the previous proteomic study on which Annexin A2 was downregulated in stably Hakai-overexpressing MDCK cells compared to normal cells [21]. On one hand, it is reported that Hakai interacts with Src tyrosine-phosphorylated substrates, an on the other hand, Annexin A2 is tyrosine-phosphorylated by Src kinase [12,27]. We carried out a transient transfection of Flag-Hakai together with Src and HA-ubiquitin in HEK293T cells to favor the activity of the ubiquitin-mediated degradation system. Besides, co-overexpression of Hakai with Src favors the increase of Hakai levels [12,16]. As expected, Annexin A2 protein expression was significantly reduced by Hakai overexpression, while Hsp90 protein expression was not affected (Figure 3a Figure S6), without affecting Hsp90 protein expression. Moreover, this upregulation of Annexin A2 in HEK293T cell line was accompanied with an increase protein levels of the best-described substrate for the E3 ubiquitin-ligase Hakai, E-cadherin [12,16]. These data suggest that Hakai may act as an E3 ubiquitin-ligase for Annexin A2 protein, inducing its degradation.
Cancers 2020, 12, x FOR PEER REVIEW 5 of 20 [12,16]. As expected, Annexin A2 protein expression was significantly reduced by Hakai overexpression, while Hsp90 protein expression was not affected ( Figure S6), without affecting Hsp90 protein expression. Moreover, this upregulation of Annexin A2 in HEK293T cell line was accompanied with an increase protein levels of the best-described substrate for the E3 ubiquitin-ligase Hakai, E-cadherin [12,16]. These data suggest that Hakai may act as an E3 ubiquitin-ligase for Annexin A2 protein, inducing its degradation.  Whole-cell lysates were subjected to western-blotting 72 h after transfection (top) and protein expression was quantified by densitometry (bottom) using GAPDH as loading control for normalization. Relative quantification of Annexin A2 expression levels was graphically represented as Mean ± SEM for two independent experiments for panel a and three for panel b and c (* p < 0.05, ** p < 0.01).

Hsp90 Inbibitor Geldanamycin Induces Downregulation of Hakai Protein via Lysosome
Geldanamycin is probably the best described Hsp90 inhibitor so far and acts by blocking its ATP binding site, preventing the correct folding of the client proteins that, in consequence, often turns into the degradation through proteasome [28,29]. Given the previously detected interaction between Hsp90, Hakai, and Annexin A2, we decided to study whether geldanamycin may affect these protein interactions. As shown, the described interaction between Hakai, Hsp90 and Annexin A2 was completely disrupted in presence of geldanamycin when using Hakai or Annexin A2 antibodies for the immunoprecipitation assays ( Figure 4, Supplementary Figure S7). Given that Annexin A2 is proposed as a potential new substrate for the E3 ubiquitin-ligase Hakai and that geldanamycin disrupts the interaction between Hakai and Hsp90, it is open the possibility that Hakai might be a direct client protein for Hsp90 chaperone. In order to test whether geldanamycin could downregulate Hakai, we used two different concentrations of geldanamycin Hsp90 inhibitor (10 µM and 20 µM) and we treat two different cell lines, HEK293T and HCT116 cells for 16 and 24 h. The cells were collected and subjected to western-blotting analysis. As shown, geldanamycin treatment decreases Hakai protein levels in both cell lines tested, while an increase of Annexin A2 was detected (Figure 5a, Supplementary Figure S8). Furthermore, we found that treatment with geldanamycin did not decrease Hakai mRNA levels supporting that Hsp90 inhibitor downregulates Hakai at the post-transcription level (Figure 5b, Supplementary Figure S8). Moreover, we tested the effect of Hsp90 inhibitor on the downregulation of Hakai by transiently transfecting Flag-Hakai together with v-Src, and HA-ubiquitin in HEK293T cells. Hakai levels were drastically reduced in presence of geldanamycin compared to non-treated control transfected conditions, accompanied by an increase of Annexin A2 protein levels (Figure 5c, Supplementary Figure S8). These data confirm that geldanamycin is able to downregulate both endogenous and ectopically expressed Hakai while it upregulates Annexin A2. Finally, the effect on Hakai and Annexin A2 protein levels was tested by combining geldanamycin treatment together with two different siRNA Hakai oligos. HCT116 cells were transiently transfected with the indicated siRNA Hakai oligos for 72 h and treated with 10 µM geldanamycin for 24 h. Reduction of Hakai expression levels was accentuated when combining geldanamycin treatment together with the previously tested siRNA Hakai oligos, leading to almost Hakai completely disappearance (Figure 5d, Supplementary Figure S8). On the contrary, Annexin A2 levels were significantly increased. Altogether, these data demonstrate that geldanamycin Hsp90 inhibitor induce Hakai protein downregulation via post-transcriptional mechanism and support that Annexin A2 is a new substrate for the E3 ubiquitin-ligase Hakai protein.   Cell lysates were collected and protein expression was evaluated by western-blotting with the indicated antibodies. Relative quantification of Annexin A2 expression levels was graphically represented as Mean ± SEM for three independent experiments (* p < 0.05, ** p < 0.01).
As previously mentioned, Hakai is an E3 ubiquitin-ligase for E-cadherin that plays a role on the epithelial-mesenchymal transition program. Given the effect of geldanamycin on Hakai expression, we also analyzed the effect of geldanamycin on the cell phenotype. HCT116 epithelial cells were Relative quantification of Annexin A2 expression levels was graphically represented as Mean ± SEM for three independent experiments (* p < 0.05, ** p < 0.01).
As previously mentioned, Hakai is an E3 ubiquitin-ligase for E-cadherin that plays a role on the epithelial-mesenchymal transition program. Given the effect of geldanamycin on Hakai expression, we also analyzed the effect of geldanamycin on the cell phenotype. HCT116 epithelial cells were treated for 24 h with geldanamycin with the indicated concentrations showing a more epithelial phenotype under the treatment. Indeed, HCT116 loses the mesenchymal phenotype accompanied by decreasing cellular protrusions (Figure 6a). Moreover, we also analyzed the localization of Annexin A2 and E-cadherin by immunofluorescence. Annexin A2 expression was increased in presence of the geldanamycin showing a statistically enriched pattern in cytoplasm and cell membrane (Figure 6b). treated for 24 h with geldanamycin with the indicated concentrations showing a more epithelial phenotype under the treatment. Indeed, HCT116 loses the mesenchymal phenotype accompanied by decreasing cellular protrusions (Figure 6a). Moreover, we also analyzed the localization of Annexin A2 and E-cadherin by immunofluorescence. Annexin A2 expression was increased in presence of the geldanamycin showing a statistically enriched pattern in cytoplasm and cell membrane (Figure 6b).
On the other hand, we also observed a three-fold increase expression of E-cadherin at cell-cell contacts in presence of geldanamycin ( Figure 6c). All these data support that Hakai is downregulated by geldanamycin inhibitor, which in consequence may influence the upregulation of E-cadherin and Annexin A2 proteins, and further reinforce the hypothesis of Hakai being an Hsp90 client protein.  Images were taken with confocal microscope by employing 40× magnification objective. A zoom image of 80X magnification was included. Annexin A2 and E-cadherin were stained in green, and cell nuclei were counterstained with Hoechst. Quantification of intensity/area was represented as Mean ± SEM (**** p < 0.0001).
On the other hand, we also observed a three-fold increase expression of E-cadherin at cell-cell contacts in presence of geldanamycin (Figure 6c). All these data support that Hakai is downregulated by geldanamycin inhibitor, which in consequence may influence the upregulation of E-cadherin and Annexin A2 proteins, and further reinforce the hypothesis of Hakai being an Hsp90 client protein.
Then, we further investigated the possible mechanism of Hakai degradation under geldanamycin treatment. First, we analyzed the effect on Hakai protein expression of proteasome inhibitor MG132 and the lysosome inhibitor chloroquine (CQ). As shown, Hakai protein expression was increased in presence of chloroquine while no effect was observed in presence of MG132 (Figure 7a, Supplementary Figure S9), further indicating that Hakai may be degraded in a lysosome-dependent manner. Besides, Hakai levels increase was accompanied by a downregulation of Annexin A2, supporting the previously obtained results that suggest that Annexin A2 is as a new possible target protein for Hakai E3 ubiquitin-ligase. Next, we analyzed the mechanism by which Hakai is degraded in absence of Hsp90 function. In order to better detect Hakai downregulation under geldanamycin treatment, HEK293T cells were transiently transfected with Hakai, Src and Ubiquitin for 48 h and treated in presence or absence of geldanamycin and chloroquine for 24 h and protein lysates were analyzed by western blot. As shown, chloroquine lysosome inhibitor efficiently prevented Hakai degradation induced by geldanamycin (Figure 7b, Supplementary Figure S9). Accordingly, Annexin A2 was upregulated under geldanamycin treatment while this effect was reverted in combination with chloroquine inhibitor. Moreover, the well-described substrate for the E3 ubiquitin-ligase Hakai, E-cadherin, was also regulated in a similar manner than Annexin A2. All these data support that Hsp90 inhibitor geldanamycin induces downregulation of its client protein Hakai in a lysosome-dependent manner. Moreover, Hakai substrate E-cadherin was also regulated in a similar manner than Annexin A2. All these data support that Hsp90 inhibitor geldanamycin induces downregulation of its client protein Hakai in a lysosome-dependent manner. Moreover, E-cadherin and Annexin A2, were also affected by geldanamycin treatment, suggesting the involvement of Hsp90 chaperone in the regulation of Hakai specific substrates.
Cancers 2020, 12, x FOR PEER REVIEW 10 of 20 Then, we further investigated the possible mechanism of Hakai degradation under geldanamycin treatment. First, we analyzed the effect on Hakai protein expression of proteasome inhibitor MG132 and the lysosome inhibitor chloroquine (CQ). As shown, Hakai protein expression was increased in presence of chloroquine while no effect was observed in presence of MG132 ( Figure  7a, Supplementary Figure S9), further indicating that Hakai may be degraded in a lysosomedependent manner. Besides, Hakai levels increase was accompanied by a downregulation of Annexin A2, supporting the previously obtained results that suggest that Annexin A2 is as a new possible target protein for Hakai E3 ubiquitin-ligase. Next, we analyzed the mechanism by which Hakai is degraded in absence of Hsp90 function. In order to better detect Hakai downregulation under geldanamycin treatment, HEK293T cells were transiently transfected with Hakai, Src and Ubiquitin for 48 h and treated in presence or absence of geldanamycin and chloroquine for 24 h and protein lysates were analyzed by western blot. As shown, chloroquine lysosome inhibitor efficiently prevented Hakai degradation induced by geldanamycin (Figure 7b, Supplementary Figure S9). Accordingly, Annexin A2 was upregulated under geldanamycin treatment while this effect was reverted in combination with chloroquine inhibitor. Moreover, the well-described substrate for the E3 ubiquitin-ligase Hakai, E-cadherin, was also regulated in a similar manner than Annexin A2. All these data support that Hsp90 inhibitor geldanamycin induces downregulation of its client protein Hakai in a lysosome-dependent manner. Moreover, Hakai substrate E-cadherin was also regulated in a similar manner than Annexin A2. All these data support that Hsp90 inhibitor geldanamycin induces downregulation of its client protein Hakai in a lysosome-dependent manner. Moreover, Ecadherin and Annexin A2, were also affected by geldanamycin treatment, suggesting the involvement of Hsp90 chaperone in the regulation of Hakai specific substrates.

Downregulation of Hakai May Partially Account for the Pharmacological Anti-Migratory Effect of Geldanamycin Hsp90 Inhibitor
HSP90 is required for the stability and function of numerous oncogenic proteins, and its specific inhibitors display multiple anticancer effects [30][31][32]. On the other hand, Hakai is considered an oncogenic protein that was reported to be overexpressed in various cancers such as colon and lung cancer [16,19,20], furthermore, Hakai knockdown inhibits cell migration [33]. Therefore, we decided to test whether Hsp90 geldanamycin inhibitor may indeed influence the migratory effect driven by Hakai. Transwell migration assay was performed in HEK293T by transiently transfected with pcDNA 3.1 or pcDNA-Flag-Hakai. HEK293T cells transfected with pcDNA-Flag-Hakai would strongly increase cell migration compared to cells transfected with an empty vector (Figure 8). This migratory capability induced by Hakai overexpression was drastically reduced in presence of geldanamycin. herefore, our results support that geldanamycin treatment affects Hakai-mediated cell migration by reducing Hsp90 activity and consequently affecting Hakai-induced migration capacity.
Hakai (4 µg), pBSSR-HA-Ubiquitin (3 µg) and pSG-v-Src (3 µg) for 48 h. The day after transfection, cells were treated with chloroquine and geldanamycin at the indicated concentrations for 24 h. Cell lysates were collected and protein expression was evaluated by western blot analyses using the indicated antibodies. Chloroquine treatment-induced Hakai protein levels recovery during geldanamycin treatment. LC3 I/II levels was used as a positive control in chloroquine treatment and β-catenin as positive control in MG132 treatment.

Downregulation of Hakai May Partially Account for the Pharmacological Anti-Migratory Effect of Geldanamycin Hsp90 Inhibitor
HSP90 is required for the stability and function of numerous oncogenic proteins, and its specific inhibitors display multiple anticancer effects [30][31][32]. On the other hand, Hakai is considered an oncogenic protein that was reported to be overexpressed in various cancers such as colon and lung cancer [16,19,20], furthermore, Hakai knockdown inhibits cell migration [33]. Therefore, we decided to test whether Hsp90 geldanamycin inhibitor may indeed influence the migratory effect driven by Hakai. Transwell migration assay was performed in HEK293T by transiently transfected with pcDNA 3.1 or pcDNA-Flag-Hakai. HEK293T cells transfected with pcDNA-Flag-Hakai would strongly increase cell migration compared to cells transfected with an empty vector (Figure 8). This migratory capability induced by Hakai overexpression was drastically reduced in presence of geldanamycin. herefore, our results support that geldanamycin treatment affects Hakai-mediated cell migration by reducing Hsp90 activity and consequently affecting Hakai-induced migration capacity.

Hsp90 Is Highly Expressed in Colorectal Cancer Samples Compared to Adjacent Normal Epithelial Tissues
The results described thus far suggest a link between Hsp90 and Hakai using an in vitro model. We previously demonstrated that Hakai expression levels are correlated to colorectal tumor progression, being proposed as a novel biomarker for colon cancer progression. Indeed, Hakai expression is gradually increased according to clinical TNM Classification System from UICC in adenoma and in different TNM stages (I-IV) from colon adenocarcinomas compared to human healthy colon tissues [20]. In order to ascertain whether Hsp90 might be involved in colorectal cancer in vivo, we investigated Hsp90 expression in human colorectal cancer patient samples. We performed immunohistochemistry experiments in samples of colorectal cancer patients, including healthy tissue, adenoma, and TNM stages I to IV of colorectal cancer. Hsp90 expression, but not Hsp70 (Supplementary Figure S3), is highly increased in carcinoma samples (TNM stage I-IV) compared to healthy epithelial tissue and adenoma (Figure 9). The changes in Hsp90 expression during colorectal cancer progression further support a pathological role for Hsp90 during tumor progression. Taken together, our studies reveal that Hsp90 is a critical regulator of Hakai protein expression and we propose that its influence on Hakai-regulated genes may, at least partially, impact in tumor progression.

Tissues
The results described thus far suggest a link between Hsp90 and Hakai using an in vitro model. We previously demonstrated that Hakai expression levels are correlated to colorectal tumor progression, being proposed as a novel biomarker for colon cancer progression. Indeed, Hakai expression is gradually increased according to clinical TNM Classification System from UICC in adenoma and in different TNM stages (I-IV) from colon adenocarcinomas compared to human healthy colon tissues [20]. In order to ascertain whether Hsp90 might be involved in colorectal cancer in vivo, we investigated Hsp90 expression in human colorectal cancer patient samples. We performed immunohistochemistry experiments in samples of colorectal cancer patients, including healthy tissue, adenoma, and TNM stages I to IV of colorectal cancer. Hsp90 expression, but not Hsp70 (Supplementary Figure S3), is highly increased in carcinoma samples (TNM stage I-IV) compared to healthy epithelial tissue and adenoma (Figure 9). The changes in Hsp90 expression during colorectal cancer progression further support a pathological role for Hsp90 during tumor progression. Taken together, our studies reveal that Hsp90 is a critical regulator of Hakai protein expression and we propose that its influence on Hakai-regulated genes may, at least partially, impact in tumor progression.

Discussion
Hsp90 chaperone has been widely described to be implicated in cancer progression by regulating client proteins described as hallmarks in cancer disease. Most of the Hsp90 client proteins are implicated in cellular processes such as regulators of cellular proliferation, proteins of oxidative stress or proteins implicated in cellular differentiation between others [2]. Interestingly, a recent study show that more than a hundred E3 ubiquitin-ligases interact with Hsp90 and therefore many of them awaits to be elucidated [1,9]. In this study, we have shown that the E3 ubiquitin-ligase Hakai is a novel Hsp90-interacting protein (Figure 1). It is generally established that the interaction between Hsp90 and their client proteins results on the degradation of the clients by the ubiquitin-dependent proteasome pathway [5][6][7]. Indeed, pharmacological inhibition of the folding activity of Hsp90 is coupled to the action of different E3 ubiquitin-ligases, such as Cullin-5 and CHIP, that signal for ubiquitinization and degradation of the specific Hsp90 client proteins that have not been correctly folded [22,34]. Although this is the best-described mechanism, we demonstrate that Hakai is a novel client protein for Hsp90 ( Figure 5). So far UHRF1 was the only reported E3 ubiquitin-ligase described as a client protein for Hsp90 [35]. Our results underscore that pharmacological inhibition of Hsp90 by geldanamycin results in the degradation of Hakai in a lysosome-dependent manner (Figure 7). A major challenge for future study is to identify the exact ubiquitin-ligase that mediates Hakai degradation upon pharmacological inactivation of Hsp90. To our knowledge, this proteasome-independent mechanism of degradation under Hsp90 geldanamycin inhibition was only previously described for IκB kinase. In this work, authors demonstrate that Hsp90 client protein IκB kinase (IkK), an essential activator of NF-kB, is degraded by autophagy when inhibiting Hsp90 with geldamacyn [8].
In previous work from our lab, we demonstrated that Annexin A2 is downregulated in Hakai-overexpressing MDCK epithelial cells compared to non-transformed MDCK cells [21]. Moreover, it has been previously shown an interaction between Annexin A2 and Hsp90 in vitro an in vivo in diabetic rat's aorta [25]. Based on these reported results, and given the interaction confirmed between Hsp90 and Hakai in different cell lines, we were also interested in determining the possible relationship between Annexin A2, Hakai and Hsp90. As shown in Figure 2, by immunoprecipitation we confirmed an endogenous interaction between Hakai, Hsp90 and Annexin A2 in HCT116 cell line ( Figure 2). Annexin A2 belongs to annexin family and is involved in the dynamic organization of membrane microdomains and the formation of membrane-cytoskeleton and membrane-membrane contacts [36]. Annexin A2 is tyrosine-phosphorylated by Src kinase [27] and, as previously mentioned, Hakai interacts with Src tyrosine-phosphorylated substrates inducing their ubiquitination and degradation, turning Annexin A2 into a potential substrate for Hakai. Transient overexpression of Hakai, Src and ubiquitin significantly downregulates Annexin A2 expression levels, and Hakai silencing give rise to an upregulation of endogenous Annexin A2 expression ( Figure 3). These data confirm the direct regulation of Annexin A2 by Hakai, and suggest Annexin A2 could be a novel substrate protein for its E3 ubiquitin-ligase activity. Furthermore, the inhibition of Hsp90 with geldanamycin completely disrupted the interaction between Hakai, Hsp90 and Annexin A2 (Figure 4), while a decrease Hakai expression and an increase of Annexin A2 expression is observed. Moreover, a recovery of an epithelial phenotype in HCT116 cell line after geldanamycin treatment was observed, accompanied by an increase of Annexin A2 in cell membrane and E-cadherin at cell-cell contacts. Finally, geldanamycin reduces Hakai-induced cell migration. All these data support that the effect of the inhibition of Hsp90 by geldanamycin regulates Hakai client protein stability and, in consequence, E-cadherin Hakai-substrate and Annexin A2 are increased, which may partially account for the pharmacological the anti-migratory and the reversion of the EMT. This data are according to previously reported results on which ganetespib treatment or HSP90 knockdown downregulated molecular pathways associated with EMT and motility [37].
Despite the broadly described role of Annexin A2 as a poor-prognosis marker during cancer progression, it has recently been reported to be induced for degradation by other oncogenic E3 ubiquitin-ligases, such as TRIM65 [38,39]. Furthermore, Annexin A2, involved in the membrane-cytoskeleton dynamics, was demonstrated to play a role in the recovery of E-cadherin at adherens junctions due to its effect on actin cytoskeleton [40], further supporting the above-explained phenotype observed during geldanamycin treatment. On the other hand, it was shown that geldanamycin stabilizes E-cadherin through the degradation of Hsp90 client protein ErbB2, responsible of increasing β-catenin-E-cadherin association and thus, contributing to the maintenance of the epithelial phenotype [41]. These results suggest that geldanamycin might revert EMT process at least partially due to its effect on the stability of Hakai client protein of Hsp90, and in consequence supports Hakai effect on E-cadherin substrate and Annexin A2.
In this study, we evaluated the expression of Hsp90 in different colon adenocarcinoma stages in order to find an implication of this chaperone during tumor progression. Previous findings of our group show that Hakai expression is increased during colon cancer progression and it is considered as a novel biomarker for colon cancer development [20]. We show an increased expression of Hsp90 in human colorectal cancer samples compared to adenoma and to adjacent healthy tissue but no differences were detected between TNM stages I to IV, further suggesting the role of Hsp90 in the acquisition of tumor malignancy in colon cancer. This data are according to the correlation between Hsp90 expression and poor outcome in patients with colorectal cancer [42]. Given that Hakai oncogene is also aberrantly highly expressed in colorectal cancer [10,13,20,43], and the demonstrated interaction between Hsp90 and Hakai in vitro, future investigations on the role of Hsp90 and Hakai in vivo await to be elucidated. Promoting the degradation of Hsp90 oncogenic client proteins by inhibiting Hsp90 is considered as a promising new anticancer strategy [44,45]. Since the appearance of the first generation Hsp90 inhibitors, such as geldanamycin and radicicol, many derivative compounds have been developed and studied for cancer treatment and subjected to clinical trials [29,46]. Although Hsp90 inhibition has been widely studied for the treatment of different types of cancer, Hsp90 inhibition-based monotherapy has not yet reached the expected results due to different resistance mechanisms. So far, for colorectal cancer only combined therapies using Hsp90 inhibitors with chemotherapeutic agents have been successful [47]. Given that one of the drug resistance mechanisms is based on an induced EMT, the development of Hakai inhibitors together with Hsp90 inhibitors could be an attractive strategy for therapeutic interventions.

Cell Culture and Transfection
HEK293T, HCT116, and ACHN cell lines were cultured in Dulbecco's Modified Eagle's Medium (DMEM) (Gibco, Thermofisher). HT29 cell line was cultured in McCoy's Modified medium and LoVo cell line was cultured in F-12K medium (Kaighn's Modification of Ham's F-12 Medium) (Gibco, Thermo Fisher Scientific). All media were supplemented with 10% heat-inactivated fetal bovine serum (FBS) and 1% penicillin/streptomycin. All cell lines were grown at 37 • C in a humidified incubator with 5% of CO2. All cell lines were periodically tested for mycoplasma. Transfection experiments were performed by employing Lipofectamine 2000 Transfection Reagent (Thermo Fisher Scientific) and Opti-MEM (Thermo Fisher Scientific) media following manufacturer's protocol. Transfection was performed for 24 or 48 as indicated in figure legends.
indicated. Cells were starved 18 h prior to assay. After 48 h of transfection, cells were collected and seeded at a confluence of 3 × 10 5 cells/transwell in a gradient of 1-30% FBS between upper and lower chamber respectively. Cells were let to migrate for 16 h and non-migratory cells were carefully removed from de upper chamber with a moistened cotton swap. Migratory cells were fixed for 20 min with 4% PFA, rinsed with PBS pH 7.4 and stained with crystal violet for 20 min. Finally, transwells were rinsed with PBS pH 7.4 and membrane was removed and mounted onto slides. Migratory cells were photographed with an Olympus BX50 employing a 10× objective.

Statistical Analysis
Analysis was performed by employing GraphPad Prism 6 Software (GraphPad Software Inc., San Diego, CA, USA). Western blotting statistical analysis was carried out by using unpaired Student's t-test at the indicated significance levels. Graphical representations of results are expressed as Mean ± SEM. Statistical analysis for immunofluorescence was carried out by performing comparative student's t test. Results are represented as Mean ± SEM for 10 areas of two different photographs. Migration assay quantification was performed by quantifying the number of migratory cells per 10× objective field for three different photographs. Statistical analysis was performed by employing Student's t-test and represented as Mean ± SEM for three photographs of one experiment. Human IHQ quantification was performed by using Kruskal-Wallis with Tukey correction test. Significance of both Student's t-test and Kruskal-Wallis with Tukey correction is indicated in the figures as * p < 0.05, ** p < 0.01 and *** p < 0.001.

Conclusions
In this study, we have demonstrated that the E3 ubiquitin-ligase Hakai is a novel client protein for Hsp90. Although the best-described mechanism for degradation of the Hsp90 clients proteins is through a ubiquitin-dependent proteasome pathway, we have shown that pharmacological inhibition of Hsp90 by geldanamycin results in the degradation of Hakai in a lysosome-dependent manner. Together with the disappearance of Hakai by geldanamycin treatment, a more epithelial phenotype was observed, accompanied by an increase expression of Hakai E-cadherin substrate at cell-cell contacts and Annexin A2 at plasma membrane. More importantly, geldanamycin reduces Hakai-induced cell migration, further underscoring the possible impact of Hsp90 inhibitors on EMT and tumor progression by its action on Hakai stability.