PRKCSH contributes to tumorigenesis by selective boosting of IRE1 signaling pathway

Unfolded protein response (UPR) is an adaptive mechanism that aims at restoring ER homeostasis under severe environmental stress. Malignant cells are resistant to environmental stress, which is largely due to an activated UPR. However, the molecular mechanisms by which different UPR branches are selectively controlled in tumor cells are not clearly understood. Here, we provide evidence that PRKCSH, previously known as glucosidase II beta subunit, functions as a regulator for selective activation of the IRE1α branch of UPR. PRKCSH boosts ER stress–mediated autophosphorylation and oligomerization of IRE1α through mutual interaction. PRKCSH contributes to the induction of tumor-promoting factors and to tumor resistance to ER stress. Increased levels of PRKCSH in various tumor tissues are positively correlated with the expression of XBP1-target genes. Taken together, our data provide a molecular rationale for selective activation of the IRE1α branch in tumors and adaptation of tumor cells to severe environmental stress.

reticulum. It is now apparent that tumor survival pathways utilize the UPR as a pro-survival strategy and as such there is interest in drugging PERK and IRE1 pathways. While many stresses activate all 3 branches of the UPR, ATF6, PERK and IRE1, conditions have been reported wherein one pathway is selectively activate, although this is not observed in tumorigenesis. Indeed, protein misfolding reduces binding of BiP to the highly homologous luminal domains of all three UPR transducers permitting oligomerization of PERK and IRE (ER to golgi transport of ATF6), thereby eliciting activation. Whether pathways exist and how the regulate activation of a single branch of the UPR is poorly defined and molecular elucidation of such pathways w ould be of significant interest. The w ork by Shin et al define PRKCSH (a component of ER localized glucosidase II as a direct and specific activator of IRE. While the data does support activation of this pathway, there are gaps. For example, data for IRE specificity over other branches is not strong. It is unclear why PRKCSH would bind selectively to IRE when the luminal domains are highly homologous and in fact interchangeable. In addition, interpretation of results and quality of some key experiments needs to be re-evaluated. Specific points for consideration are listed below. 1. What happens to glucosidase activity during ER stress? I assume that PRKCSH is being redirected and thus w ould limit activity. Alternatively, w hile the authors show direct activation, of this subunit, is the entire enzyme complex recruited and is glucosidase activity needed for activation of IRE? 2. Looking at the data, it is clear that PRKCSH is not required for activation; rather it seems to provide a boost in signaling. Indeed, it appears that the kinetics of Xbp1 splicing are unaltered and differences in spliced Xbp are not apparent until 5 hours suggesting that PRKCSH modifies the response but is not necessary. 3. With respect to the comment above, the dose of tunicamycin used throughout are very high. A dose response and use of a lower dose of TM would be more compelling throughout. 4. In the beginning of the results, the authors never define what PRK cells are either in text or figure legends. 5. Figure 2 and 3 provide evidence for PRKCSH in increasing Xbp1 splicing and later that it increases IRE1 auto-p. However, the two are never tied together. The authors should determine whether Ire1 is necessary for PRKCSH-dependent effects on Xbp1 (knockdown of Ire or Ire1 inhibitor). It w ould be of equal interest to determine w hether overexpression of Ire1 (w hich increases its activity through forced oligomerization) w ould overcome the effect of PRKCSH knockout or knockdown. The issues with the dose of TM, the kinetics of the response, this data w ould help support that the impact is directly on Ire1. 6. The specificity for Ire1 is not convincing given high homology of all UPR transducers in their luminal domains. In the supplemental data, it is not sufficient to only show PERK-p. The authors need to perform the same time course as with Ire`1 and XBP1, and examine downstream signaling; p-eIF2, ATF4 and CHOP. These experiments should also be performed in PRKCSH k/o cells for clarity. 7. The title of the section "PRKCSH directly activates Ire1" is misleading and not accurate. The data demonstrate it binds, but that binding correlates with increased auto-P. How ever, no data demonstrate activation and PRKCSH is clearly not required for binding since there is activity in its absence. 8. The experiments interpreted to demonstrate that PRKCSH increases oligomerization could reflect formation of inclusions at the proteasome that reflects overexpression. These experiments do not in fact establish oligomerization. Use of differential tagging would be one possible approach to compliment and control for this issue. 9. Defining the E/P domain as necessary for binding is excellent. How ever, it is disappointing that the authors did not: 1) demonstrate specificity for Ire; 2) determine whether it is sufficient to activate Ire1 and do so selectively. 10. With regard to specificity, if PRKCSH indeed targets Ire1 specifically, interchanging luminal domains should alter or sw itch specificity and provide direct support for the authors hypothesis.

Reviewer #1 (Remarks to the Author):
The manuscript by Shin et al., describes studies suggesting a novel role to the PRKCSH/Hepatocystin protein, a non-catalytic subunit of glucosidase II as a modulator of the function of the major UPR effector IRE1. A series of biochemical analyses demonstrates that PRK can bind and modulate the kinase activity for IRE1alpha. Coupled with database mining analysis demonstrating increased levels of PRK in tumors compared to normal tissues and data showing that increased PRK levels promote a more pro-survival activity of IRE1alpha against cytotoxic ER stress, the authors propose that PRK may have a protumorigenic role in protect tumor cells from ER stress-induced death. The work presented here contains some novel aspects, particularly the finding and characterization of interactions between IRE1alpha and PRK, which include elegant analysis of the domains involved in this interaction. However, there are significant gaps in the overall rationale linking PRK upregulation with a more aggressive tumor cell phenotype, which are exacerbated by the lack of any in vivo tumor data testing the central hypothesis. These issues are compounded by the modest effects of upregulated or downregulated PRK has on IRE1a activity.
Major concerns: 1. The analysis of PRK expression in human tumors shows statistically significant (based on P value) data for mRNA. However, the magnitude of the differences in expression is marginal and unclear if it is biologically meaningful. Moreover, the use of select datasets of mRNA analyses in the specific tumor types (as opposed to complete data sets from TCGA), raises concerns of selection bias. Similar concerns are raised regarding the levels of PRK protein, shown in Fig. 1C, which appear to be very modest in magnitude. Similar issues are observed in Fig. 3g, where the correlation of mRNA levels between PRK and targets of the Response: Thank you for commenting. Considering this comment, we performed the expression analysis of PRKCSH in various human tumors using complete data sets from TCGA. We analyzed these data sets using GEPIA (Gene Expression Profiling Interactive Analysis, http://gepia.cancer-pku.cn/), a web-based tool that delivers fast and customizable functionalities based on TCGA data. By using GEPIA, we performed differential expression analysis, expression profiling plotting, correlation analysis, and patient survival analysis. In the revised manuscript, the data clearly show that PRKCSH expression is increased in In addition, we preformed further analysis of PRKCSH protein expression in HCC tissues and analyzed the IHC data by using the NIH-ImageJ software with the IHC Profiler plugin (http://rsb.info.nih.gov/ij/) as described previously (Autophagy 2016, 12: 2451-2466; PLoS One 2014, 9: e96801). Staining intensity of the PRKCSH protein is now presented as IMPV (inverted median pixel value; Supplementary Fig. 1c). In the revised manuscript, the data clearly show that level of the PRKCSH protein is higher in liver cancer tissues than in normal liver tissues ( Fig. 1d; described in the Results and Methods sections of the revised manuscript, pages 7 and 28).
We also further analyzed the patient survival curves using the data sets from TCGA and demonstrate that PRKCSH expression is associated with both the overall survival and disease-free survival of patients with liver hepatocellular carcinoma and skin cutaneous melanoma ( Fig. 1g; described in the Results of the revised manuscript, page 8).
Finally, we analyzed the correlation between mRNA levels of PRKCSH and the IRE1/XBP1 target genes using complete data sets from TCGA by GEPIA web tool. We demonstrate that the expression level of PRKCSH is clearly correlated with those of XBP1 downstream target genes ( Fig. 3h and Supplementary Fig. 4; described in the Results of the revised manuscript, page 10).

2.
For the most experiments dealing with overexpression of PRK in untransformed cells (e.g.,Figs. 2b,2b,2f,3a,4a etc), or in some cases of PRK knockout (Fig. 2e), the magnitude of the observed effects is marginal at best (i.e., from 1 to 1.5-fold, or from 2-fold to 3-fold increase in splicing of XBP1). What is of particular concern here, is that these changes arise in response to extreme stress conditions, either with tunicamycin treatment or complete absence of glucose, which are unlikely to be physiologically relevant. Thus, we think that our data is sufficient for explaining the significant differences in XBP1 splicing and MAPK activation between control and PRKCSH-modulated groups under these ER stress conditions. Our data shows that PRKCSH is boosts IRE1α signaling, but is not necessary for its triggering.
Unlike TM, glucose starvation (along with hypoxia, oxidative stress, and nutrient starvation) is generally recognized as one of the most representative physiological stress conditions for  Fig. 5 lack important controls. First, an IgG control Ab needs to be included to determine non-specific IP activity (panels 5a-5e). Second, the reverse IP with anti-PRK antibody should demonstrate activity in bringing down IRE1a.

The co-immunoprecipitation experiments in
Moreover, the quality of Fig. 5f is poor and the blot should be repeated.

Response:
We appreciate these important comments. We performed all the control coimmunoprecipitation experiments using control IgG Ab and included the results in the revised manuscript ( Fig. 5a, 5b, 5g, and Supplementary Fig. 5f). We also performed coimmunoprecipitation experiments using anti-PRKCSH antibody and IRE1-Flag-transfected cells (Fig. 5c). The experiment shown in Fig. 5f in the original manuscript was replaced with better quality blots (Fig. 5c). Overall, the newly added data provides clear evidence of the interaction between PRKCSH and IRE1α. Fig. 6a, does expression of the PRK mutant ∆E/P or MRH constructs affect survival under ER stress condition?

In
Response: Following this comment, we performed cell death assay using WT, ∆E/P, and MRH mutant constructs under ER stress. Our data provide additional evidence that the E/P domain is required for PRKCSH-mediated resistance to ER stress-induced cell death ( Fig.   6f; described in the Results of the revised manuscript, page 16).
6. The cytoprotective data demonstrated by immunoblots against cleaved PARP are also very weak. Clonogenicity assays need to be performed to determine if long-term overall survival (including necroptosis, feroptosis, as well as senescence, necrosis etc) is affected. 7. Since all of the ER stresses used almost exclusively rely on large doses of pharmacological agents such as tunicamycin or thapsigargin, it is critical that in vivo tumor growth experiments, where transformed hepatocytes are exposed to physiological stress (e.g., hypoxia, low nutrient availability).

Reviewer #2 (Remarks to the Author):
The UPR responds to intrinsic and extrinsic stresses that disrupt protein folding in the endoplasmic reticulum. It is now apparent that tumor survival pathways utilize the UPR as a pro-survival strategy and as such there is interest in drugging PERK and IRE1 pathways. 1. What happens to glucosidase activity during ER stress? I assume that PRKCSH is being Following the reviewer's comment, we examined the impact of ER stress on GII activity.
Generally, specific activity of GII is measured by in vitro assay (cell-free system) using recombinant proteins (J Biol Chem. 2013, 23:16460-16475). However, a method to measure GII activity under ER stress in vivo or ex vivo has not been reported yet. Thus, we performed co-immunoprecipitation using anti-GIIα antibody to directly determine the effect of ER stress on the interaction between PRKCSH and GIIα. In the revised manuscript, we provide new evidence that ER stress induces the dissociation of the GII complex ( Fig. 5g; described in the Results of the revised manuscript, page 15; and in the Discussion, page 20). PRKCSH binding to GIIα is required for maintaining the GIIα level in the ER and GII enzymatic activity (Protein Sci. 2016, 25: 2095-2101); therefore, it is plausible that ER stress ultimately reduces glucosidase activity.

2.
Looking at the data, it is clear that PRKCSH is not required for activation; rather it seems to provide a boost in signaling. Indeed, it appears that the kinetics of Xbp1 splicing are unaltered and differences in spliced Xbp are not apparent until 5 hours suggesting that PRKCSH modifies the response but is not necessary.
Response: Thank you for this insightful comment, with which we totally agree. This suggestion was very helpful for interpreting clearly the role of PRKCSH in the regulation of IRE1α activation. Thus, we corrected the sentences in the revised manuscript as suggested (modified in the Results of revised manuscript, pages 9 and 10).
3. With respect to the comment above, the dose of tunicamycin used throughout are very high. A dose response and use of a lower dose of TM would be more compelling throughout.
Response: Treatment with 5−10 μg/ml tunicamycin is generally used in ER stress research. Following the reviewer's comment, we investigated changes in XBP1 splicing in L02-PRK cells treated with 1.25−10 µg/ml tunicamycin (Fig. 2c; described in the Results of the revised manuscript, page 8). Furthermore, we determined the time dependence of the response of XBP1 splicing from 2 h to 16 h of treatment with 10 µg/ml tunicamycin (Fig. 2a, 2b). In the revised manuscript, we revised the description of the results for splicing of XBP1 mRNA and protein and replaced the previously presented data with new data. According to revised  Figure 2 and 3 provide evidence for PRKCSH in increasing Xbp1 splicing and later that it increases IRE1 auto-p. However, the two are never tied together. The authors should determine whether Ire1 is necessary for PRKCSH-dependent effects on Xbp1 (knockdown of Ire or Ire1 inhibitor). It would be of equal interest to determine whether overexpression of Ire1 (which increases its activity through forced oligomerization) would overcome the effect of PRKCSH knockout or knockdown. The issues with the dose of TM, the kinetics of the response, this data would help support that the impact is directly on Ire1.

Response:
We have shown the results for the splicing of XBP1 mRNA and protein, and activation of ERK1/2 MAPK in IRE1α-knockdown L02-PRK cells ( Supplementary Fig. 8 in the original manuscript), which suggest that IRE1α is necessary for PRKCSH-mediated XBP1 splicing and MAPK activation. Considering the reviewer's comment, we moved these data from Supplementary Fig. 8 to Fig. 4 to emphasize this aspect in the revised manuscript ( Fig. 4i, 4j, 4k; described in the Results of the revised manuscript, pages 12−13). Our data showed that the boosting effect of PRKCSH on XBP1 splicing and MAPK activation is almost completely abolished by IRE1α silencing.
Following the reviewer's comment, we performed additional experiments to determine whether IRE1α overexpression would overcome the effect of PRKCSH knockout. In the revised manuscript, we demonstrate that IRE1α overexpression (hIRE1α.pcD construct was provided by Dr. Randal Kaufman) increased the splicing of XBP1 mRNA and protein and activation of ERK1/2 MAPK in L02-PRK KO cells (Fig. 4i, 4j, 4k; described in the Results of the revised manuscript, pages 12−13). Taken together, our results provide clear evidence that IRE1α is a downstream target of PRKCSH during ER stress response.

6.
The specificity for Ire1 is not convincing given high homology of all UPR transducers in their luminal domains. In the supplemental data, it is not sufficient to only show PERK-p. The authors need to perform the same time course as with Ire`1 and XBP1, and examine downstream signaling; p-eIF2, ATF4 and CHOP. These experiments should also be performed in PRKCSH k/o cells for clarity.
Response: According to the reviewer's comment, we performed immunoblot analysis to check the levels of PERK phosphorylation and ATF4 expression using a time course similar to that used for determining IRE1 phosphorylation in PRKCSH-overexpressing and knockdown cells. We also investigated those levels using L02-PRK KO and Huh-PRK KO cells, which were newly established during revision. We examined the expression of Ero1B, a downstream target gene of PERK pathway, using PRKCSH-overexpressing (L02), knockdown (Huh-7), and KO (L02 and Huh-7) cells. We provide evidence that PRKCSH is not involved in the regulation of PERK pathway activation ( Supplementary Fig. 5a−e; described in the Results of the revised manuscript, page 11). 7. The title of the section "PRKCSH directly activates Ire1" is misleading and not accurate.
The data demonstrate it binds, but that binding correlates with increased auto-P. However, no data demonstrate activation and PRKCSH is clearly not required for binding since there is activity in its absence.
Response: Thank you for the insightful comment. According to this, we revised the title and the sentence to explain our result clearly in the revised manuscript (described in the Results title of the revised manuscript, pages 11 and 14).

8.
The experiments interpreted to demonstrate that PRKCSH increases oligomerization could reflect formation of inclusions at the proteasome that reflects overexpression. These experiments do not in fact establish oligomerization. Use of differential tagging would be one possible approach to compliment and control for this issue. Nevertheless, to address the reviewer's concern, we examined whether the oligomerized IRE1α is co-localized with the proteasome. In the revised manuscript, we provide clear evidence that a Venus-tagged IRE1α fusion protein is not associated with the proteasome under either resting or ER stress conditions, excluding the possibility of inclusion formation in the proteasome by overexpressed IRE1α (Supplementary Fig. 6; described in the Results of the revised manuscript, page 12).

9.
Defining the E/P domain as necessary for binding is excellent. However, it is disappointing that the authors did not: 1) demonstrate specificity for Ire; 2) determine whether it is sufficient to activate Ire1 and do so selectively.

Response:
To determine whether the interaction between PRKCSH and IRE1α through the E/P domain is sufficient to activate IRE1α as suggested by the reviewer, we constructed additional PRKCSH mutants. Together with the result shown in Fig. 5e, the experiments using these mutants confirmed that the E/P domain is essential for the interaction between PRKCSH and IRE1α (Fig. 5f, Supplementary Fig. 7a).
To identify the compartment in which the interaction between PRKCSH and IRE1α occurs, we constructed a PRKCSH mutant lacking the N-terminal signal peptide (∆S/G2B). The results provide additional evidence that the interaction between PRKCSH and IRE1α occurs in the ER but not in the cytoplasm (Fig. 5f; described in the Results of the revised manuscript, page 14).
Subsequently, we used the ∆G2B, ∆S/G2B, and E/P mutants to determine whether the PRKCSH−IRE1α interaction is sufficient for IRE1α activation. Both the ∆G2B and E/P mutants interacted with IRE1α not only under ER stress but even under resting conditions (Fig. 5f). Expression of these two mutants promoted both IRE1α activation and splicing of XBP1 under ER stress (Fig. 5i, 5j; described in the Results of the revised manuscript, page 15), whereas these mutants failed to activate IRE1α pathway under resting condition (Fig. 5i,   5j; described in the Results of the revised manuscript, page 15, and Discussion, pages 19−20). Thus, in the revised manuscript, we demonstrate that the interaction between the two proteins is necessary for boosting the activation of IRE1α by ER stress, but is not sufficient for triggering its activation.
The specificity issue is addressed in the response to comment 10.

10.
With regard to specificity, if PRKCSH indeed targets Ire1 specifically, interchanging luminal domains should alter or switch specificity and provide direct support for the authors hypothesis. Considering the reviewer's comment, we performed co-immunoprecipitation with anti-PERK antibody and found that PRKCSH does not interact with PERK under either resting or ER stress conditions, supporting the specific interaction of PRKCSH with IRE1α Fig. 5f; described in the Results of the revised manuscript, page 14, and in the Discussion, page 19).

(Supplementary
Since ATF6 was not detected in cell lysates prepared for co-immunoprecipitation, as shown below, we were unable to assess the interaction between ATF6 and PRKCSH. However, we think that PRKCSH specifically interacts with IRE1α because the luminal domain of ATF6 has no structural similarity with that of IRE1α and PRKCSH does not interact with PERK.

Reviewers' comments:
Reviewer #1 (Remarks to the Author): The authors have responded positively to the majority of the critiques and the manuscript has been improved substantially. Specifically, the new data from analyses of TCGA databases and protein levels in of additional cell lines and samples supports the notion that SPRHCC is upregulated in multiple malignancies and correlates w ith expression levels of XBP-1. Moreover, the authors have generated CRISPR KO cell lines which show a clearer dependency of the UPR and cell survival on PRKCSH expression.
The heavy reliance on high doses of tunicamycin and thapsigargin for some of the key findings is still problematic; how ever, this is counterbalanced by the demonstration that hypoxia and low glucose can also elicit PRKCSH expression. One key unanswered point, w hich in my view diminishes overall impact for this w ork is the lack of any in vivo phenotype when PRKCSH is abrogated. Also, as a more minor issue, the new "clonogenic survival data depicted in figure 6e, are not technically "clonogenicity" data, since the cells w ere plated at high density and plate area coverage rather than individual colonies are measured. Though these results do imply that PRKCSH is impacting clonogenic death, the title of this figure panel and corresponding text should be changed to "Cell survival".
The manuscript is w ell-written, and appropriate statistical analyses appear to have been performed. The experimental design is provided w ith sufficient detail so that the work should be able to be reproduced.
Overall, the w ork presented in this manuscript is novel and adds another layer of complexity in the regulation of the IRE1-arm of the UPR in pathological conditions. Reviewer #2 (Remarks to the Author): The authors have in general addressed my concerns.

Reviewers' comments: Reviewer #1 (Remarks to the Author):
The authors have responded positively to the majority of the critiques and the manuscript has been improved substantially. Specifically, the new data from analyses of TCGA databases and protein levels in of additional cell lines and samples supports the notion that PRKCSH is upregulated in multiple malignancies and correlates with expression levels of XBP-1. Moreover, the authors have generated CRISPR KO cell lines which show a clearer dependency of the UPR and cell survival on PRKCSH expression.
The heavy reliance on high doses of tunicamycin and thapsigargin for some of the key findings is still problematic; however, this is counterbalanced by the demonstration that hypoxia and low glucose can also elicit PRKCSH expression. One key unanswered point, which in my view diminishes overall impact for this work is the lack of any in vivo phenotype when PRKCSH is abrogated.
Also, as a more minor issue, the new "clonogenic survival data depicted in figure 6e, are not technically "clonogenicity" data, since the cells were plated at high density and plate area coverage rather than individual colonies are measured. Though these results do imply that PRKCSH is impacting clonogenic death, the title of this figure panel and corresponding text should be changed to "Cell survival".
The manuscript is well-written, and appropriate statistical analyses appear to have been performed. The experimental design is provided with sufficient detail so that the work should be able to be reproduced.
Overall, the work presented in this manuscript is novel and adds another layer of complexity in the regulation of the IRE1-arm of the UPR in pathological conditions. we also showed evidences for tumor cell survival under ER stress condition using in vitro models with PRKCSH knock-out and over-expressed cells. We think that the most important in vivo relevance of some study can be obtained with the patient data. In this regard, we showed that the level of PRKCSH is associated with multiple human tumor tissues, tumor progressions, and patient survivals. Considering that the mouse study does not always match with the human data, our results from patient data along with in vitro tumor survival study may represent the in vivo relevance on the role of PRKCSH in human tumor promotion.

Response
Thus, we think that the lack of in vivo mouse study will not be able to jeopardize our findings on the novel role of PRKCSH in human tumor promotion.
Finally, we have corrected "clonogenic survival" to "cell survival" in the revised manuscript as suggested (modified in the Results, Methods, and figure legends of revised manuscript, pages 16, 32, 48, and 49).

Reviewer #2 (Remarks to the Author):
The authors have in general addressed my concerns.
Response: Thank you for your helpful comments during the revision.

Reviewers' comments:
Reviewer #1 (Remarks to the Author): The authors have responded positively to the majority of the critiques and the manuscript has been improved substantially. Specifically, the new data from analyses of TCGA databases and protein levels in of additional cell lines and samples supports the notion that PRKCSH is upregulated in multiple malignancies and correlates with expression levels of XBP-1. Moreover, the authors have generated CRISPR KO cell lines which show a clearer dependency of the UPR and cell survival on PRKCSH expression.
The heavy reliance on high doses of tunicamycin and thapsigargin for some of the key findings is still problematic; however, this is counterbalanced by the demonstration that hypoxia and low glucose can also elicit PRKCSH expression. One key unanswered point, which in my view diminishes overall impact for this work is the lack of any in vivo phenotype when PRKCSH is abrogated.
Also, as a more minor issue, the new "clonogenic survival data depicted in figure 6e, are not technically "clonogenicity" data, since the cells were plated at high density and plate area coverage rather than individual colonies are measured. Though these results do imply that PRKCSH is impacting clonogenic death, the title of this figure panel and corresponding text should be changed to "Cell survival".
The manuscript is well-written, and appropriate statistical analyses appear to have been performed. The experimental design is provided with sufficient detail so that the work should be able to be reproduced.
Overall, the work presented in this manuscript is novel and adds another layer of complexity in the regulation of the IRE1-arm of the UPR in pathological conditions.

Response:
We appreciate your positive comments. Considering reviewer's comment, we performed the in vivo tumor growth analysis in xenograft nude mouse model using Huh-PRKCSH WT and KO cell lines. In the revised manuscript, we provide clear evidence that PRKCSH has a role in tumor growth under physiological stress conditions of tumor microenvironment ( Fig. 6f and Supplemental Fig. 9; described in the Results of the revised