Reciprocal negative feedback regulation of ATF6α and PTEN promotes prostate cancer progression

Phosphatase and tensin homolog (PTEN) loss tightly correlates with prostate cancer (PCa) progression and metastasis. Inactivation of PTEN leads to abnormal activation of PI3K/AKT pathway. However, results from clinical trials with AKT inhibitors in PCa have been largely disappointing. Identification of novel regulators of PTEN in PTEN-dysfunctional PCa is urgently needed. Here we demonstrated that the expression level of PTEN is inversely correlated with the signature score of unfolded protein response (UPR) in PCa. Importantly, PTEN suppresses the activity of ATF6α, via interacting to de-phosphorylate ATF6α and consequently inhibiting its nuclear translocation. Conversely, ATF6α promotes the ubiquitination and degradation of PTEN by inducing CHIP expression. Thus, ATF6α and PTEN forms a negative feedback loop during PCa progression. Combination of ATF6α inhibitor with AKT inhibitor suppresses tumor cell proliferation and xenograft growth. Importantly, this study highlighted ATF6α as a therapeutic vulnerability in PTEN dysfunctional PCa.


Introduction
Prostate cancer (PCa) is characterized as being histologically and molecularly heterogeneous, with low-risk localized PCa having an indolent natural history, while others have aggressive and lethal disease [1,2].Phosphatase and tensin homolog (PTEN) is one of the most commonly inactivated tumor suppressor gene in PCa and its loss is associated with poor clinical outcome [3].PTEN deficiency, resulting mainly from PTEN deletion or mutation, could be detected in tissues representing all stages of PCa progression, and especially at much higher frequency in metastatic PCa and castration resistant prostate cancer (CRPC) [4][5][6].Inactivation of PTEN leads to abnormal activation of PI3K/AKT pathway [7].Until recently, however, results from clinical trials with AKT inhibitors in PCa have been largely disappointing [8][9][10].Therefore, an appropriate design of rational combinations of PI3K/AKT inhibitors with other novel agents is urgently needed.Most recently, Sweeney et al. reported that co-targeting AKT and androgen-receptor 292 Page 2 of 17 signaling pathway was a potential treatment strategy for men with PTEN-loss metastatic CRPC (mCRPC) [11].Of note, PTEN function is also regulated at the transcriptional, posttranscriptional and translational levels [12][13][14].Thus far, it remains largely unknown as to what extent dysregulation of these processes could contribute to PCa progression.
The unfolded protein response (UPR) is a homeostatic mechanism to maintain endoplasmic reticulum (ER) function [15] Over the past decade, a large body of evidence has emerged to support a role of UPR in the establishment and progression of several types of malignancies, including PCa [16,17].Recently, an adverse correlation between PTEN and UPR signaling was reported in high-grade serous ovarian carcinoma where knockdown of PTEN leads to increased protein synthesis, upregulation of BiP and elevated levels of p-eIF2α and CHOP [18].Additionally, in another study, Wang et al. showed that UPR pathway is preferential activated in PTEN-deficient triple negative breast cancer [19].However, the link between PTEN and UPR is unclear in PCa progression.
The UPR is mediated by at least three well-conserved stress sensors that are ER-localized transmembrane receptors: pancreatic ER kinase-like ER kinase (PERK), inositolrequiring kinase 1 (IRE1), and activating transcription factor 6 (ATF6) [15].ATF6 consists of two isoform ATF6α and ATF6β, both of which can be activated by ER stress [20].Upon activation, the N-terminal fragment of ATF6α and ATF6β, designated as ATF6α(N) and ATF6β(N), is spliced from their full-length in the Golgi apparatus and transported into the nucleus to induce gene expression [21].ATF6α plays a dominant role in inducing the expression of ATF6 target genes.Clinically, Liu et al. suggested that ATF6α is expressed at a higher level in PCa tissues than in benign prostatic hyperplasia (BPH) [22].In the present study, we found that PTEN expression is adversely related to UPR signature in human PCa tissues.Reintroduction of PTEN into PTENdeficient LNCaP cells leads to ATF6α dephosphorylation, inhibits its nuclear translocation and transcriptional regulatory activity.In turn, ATF6α could promote PCa progression by destabilizing PTEN protein in CHIP-dependent way.Importantly, co-targeting ATF6α and PI3K/AKT pathways may be a promising therapeutic approach for PCa, especially the ones with PTEN dysfunction.

Cell culture and reagents
LNCaP, VCaP, DU145, PC3 and HEK293T cell were purchased from American Type Culture Collection (ATCC).All cell lines were authenticated by short tandem repeat analysis and were tested for mycoplasma contamination.The information of reagents is shown in Supplementary Table 1.

Animals
All experimental protocols for animal studies were approved by the Institutional Animal Care and Use Committee of School of Basic Medical Sciences, Shandong University (Document No. ECSBMSSDU2021-2-126) and conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.Different groups were allocated in a randomized manner and investigators were blinded to the allocation of different groups when doing drug treatment.More detailed information on animals is included in Supplementary Materials and Methods.

Immunohistochemistry (IHC) and immunofluorescence staining
IHC was carried out as described previous [23].Detailed information of the IHC protocol as well as evaluation procedure of tissues microarray were provided in Supplementary Materials and Methods.For immunofluorescence, proteins were detected using goat anti-mouse Alexa Fluor 594.For quantitation of ATF6α subcellular distribution, at least 150 cells were scored for each experiment.Cells were rated as showing nuclear staining (N), cytoplasmic staining (C), or staining within both the nucleus and the cytoplasm (N + C).The ATF6α nuclear location (%) = N + (N + C) / total cells.Nuclei were identified by DAPI staining.The information of antibodies was shown in Supplementary Table 2.

Western blot and co-immunoprecipitation
Preparation of cell lysates and performance of western blot and immunoprecipitation were carried out according to our published procedure [24].Cytoplasmic and nuclear proteins were extracted using the ProteoExtract subcellular proteome extraction kit (#539,791, Millipore).The information of antibodies is summarized in Supplementary Table 2.

RNA isolation and real-time qPCR analysis
Total RNA was extracted using TRIzol reagent according to the manufacturer's protocol (Vazyme Biotech, R401-01 regent kit) and 1 μg of RNA was reverse-transcribed with ReverTra Ace qPCR RT kit (PCR-311, Toyobo).Real-time qPCR was performed using the SYBR Green mix (QPK-201, Toyobo).The primer sequences used are listed in Supplementary Table 3.

Transient transfection and viral transduction
Premade siRNA constructs were purchased from Ribo-Bio (Guangzhou, China) and all the siRNA constructs sequences were list in Supplementary Table 4.For DNA transfections, the plasmid constructs were purchased from Biosun Inc (Jinan, China) and WZ Biosciences Inc (Jinan, China).The transfection procedure was performed as previously described.Human Lenti-ATF6α-GFP and its control Lenti-GFP were obtained from Genecopoeia.To acquire a stable cell line, single-cell clonal isolates were selected by puromycin.

Luciferase reporter gene assay and Chromatin immunoprecipitation (ChIP)
Luciferase activity was determined using the Promega luciferase assay kit (E1910, Promega) following the manufacturer's protocol.Reporter activity was detected using the Molecular devices spectraMzx iD3.ChIP experiments were performed with ChIP assay kit (#17-295, Millipore) according to the manufacturer's protocol.Briefly, Cells were sonicated, pre-cleaned, mixed with magnetic beads and incubated with indicated antibodies.After DNA purification, the enrichment of the DNA template was analyzed by PCR and real-time qPCR using primers.The primers were list in Supplementary Table 5.

Gene set enrichment analysis
We used the java GSEA v4.0 program.GSEA were performed according to the instructions.Gene sets of Hallmark Collection, Canonical Pathway were used.All gene sets are from the Molecular Signatures Database v7.5.1.

Quantification and statistical analysis
All of the statistical details including the statistical tests used, exact number of animals, definition of center, and dispersion and precision measures for the experiments can be found in the figures, figure legends or in the results.All statistical analyses were carried out using GraphPad Prism 8 software (GraphPad software, CA, US).The two-tailed unpaired t-test was used to calculate statistical significance between the two groups.Survival information was verified by Kaplan-Meier analysis and compared using the logrank test.All results were presented as the mean and the standard error of the mean.P values considered to be significant as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

PTEN expression is adversely related to UPR signature and PTEN deficiency synergizes with ATF6α to activate UPR signaling in PCa cells
To determine the correlation between PTEN expression and UPR signaling pathway in PCa, we have analyzed various PCa public datasets.As shown in Fig. 1a, b, gene set enrichment analysis (GSEA) was performed using microarrays from Gene Expression Omnibus (GEO) datasets (GSE3325 and GSE6919).The UPR gene signature was significantly enriched in PTEN-low tumors compared with PTEN-high tumors.Of note, ATF6α target genes (HSPA5 and PDIA6) were up-regulated in PTEN-low samples (Supplementary Fig. 1a, 1b).
To further confirm the relation between PTEN function and UPR activation, prostate-specific Pten knockout (Ptenflox/flox : Pb-Cre + , referred to herein as Pten PC-/-mice) mice were used.As shown in Fig. 1c, PTEN protein levels were significantly decreased in wild type (WT) mice under ER stress inducer tunicamycin (Tm) treatment.In addition, spliced ATF6α (ATF6α-N), and its target gene BiP were increased significantly after Tm treatment in Pten PC-/-prostate tissues than those in WT mice.To further verify the effects of PTEN on UPR activity, we next silenced PTEN with siRNA in DU145 and VCaP (PTEN-WT) cells and overexpressed PTEN in LNCaP and PC3 (PTEN-deficient) cells.Western blot showed that the protein levels of ATF6α and its target gene were increased after silencing PTEN expression, especially in the presence of Tm treatment in DU145 and VCaP cells (Fig. 1d and Supplementary Fig. 1d).In contrast, PTEN overexpression inhibited protein expression of ATF6α and its target gene induced by Tm in LNCaP and PC3 cells (Fig. 1e and Supplementary Fig. 1e).Of note, compared to the changes of phosphorylated levels of IRE1 (p-IRE1) and PERK (p-PERK), the ATF6α protein levels in spliced form increased more significantly in Pten PC-/-prostate tissues (Fig. 1f-h).Real-time qPCR revealed that mRNA levels of the ATF6α target genes were suppressed by overexpression PTEN in LNCaP cells and activated by silence PTEN in DU145 cells (Fig. 1i, j and Supplementary Fig. 1 g).Additionally, as shown in Fig. 1k, l, the BiP luciferase activity was highly induced by Tm.
However, these activity effects of Tm were further enhanced by PTEN silencing in DU145 cells but decreased by PTEN overexpression in LNCaP cells.Collectively, these results demonstrated that ATF6α signaling was adversely related to PTEN expression.
Further analyses on The Cancer Genome Atlas (TCGA) and GSE29010 showed that UPR gene signature was enriched in PTEN-loss prostate tumors when compared with PTEN-WT prostate tumors (Supplementary Fig. 1 h,  1i).We next analyzed the relationship between PTEN loss and ATF6α, PERK and IRE1 pathway signature.As shown in Supplementary Fig. 1j, ATF6α signature was significantly enriched in Pten-low tumors (P = 0.027), whereas the enrichment of IRE1 (P = 0.062) and PERK (P = 0.31) pathways is not statistically significant [25].Additionally, compared with that of PERK or IRE1 knockdown, the proliferation induced by PTEN knockdown could be attenuated more significantly after ATF6α knockdown (Supplementary Fig. 1 k).Collectively, these results supported that PTEN dysfunction can activate the ATF6α signaling.
Since ATF6α could be activated by PTEN knockout or knockdown, we aimed to confirm whether PTEN dysfunction participated in the activity of ATF6α signaling.As shown in Fig. 1m, the combination of PTEN silencing and ATF6α overexpression significantly increased mRNA levels of HERPUD1, SEL1L, and BiP in DU145 cells.However, the increase induced by PTEN silencing could be attenuated when ATF6α was silenced (Fig. 1n).Likewise, we identified that the increase of the luciferase activity of BiP promoter caused by PTEN knockdown could be further strengthened by introducing ATF6α (Fig. 1o) but abolished by silencing ATF6α in DU145 cells (Fig. 1p).

ATF6α protects against ER stress-induced cytotoxicity and promotes PCa progression in vivo
To investigate the correlation between PTEN and ATF6α in clinical PCa samples, we performed IHC assay analysis using 145 Chinese PCa cases from Qilu hospital with three tissue microarrays (TMAs).We found that the ATF6α protein expression was mainly localized in the cytoplasm in BPH samples.Of note, nuclear staining of ATF6α was frequently seen in PCa cases with high Gleason score accompanied with PTEN loss.Representative image of IHC staining for nuclear ATF6α expression of PCa cases were shown in Fig. 2a.A positive correlation between nuclear ATF6α expression with Gleason score was shown in Fig. 2b.Furthermore, high expression of ATF6α was significantly associated with poor clinical outcome (Fig. 2c).We also observed that the expression levels of ATF6α target genes were significantly higher in localized PCa than those in BPH tissues as shown in GSE46602 and GSE70768 (Supplementary Fig. 2a).And the Kaplan-Meier curve showed that patients with high levels of ATF6α target genes had a shorter biochemical relapse free survival rate than those with low expression (Supplementary Fig. 2b).In all, these results suggested that upregulation of ATF6α target gene was associated with PCa progression to advanced clinical states.To elucidate the biological functions of ATF6α in PCa cells, we designed siRNA against ATF6α (siATF6α).As shown in Supplementary Fig. 2c, 2d, siATF6α can significantly deplete mRNA and protein expression levels of ATF6α in DU145 and LNCaP cells, accordingly.Notably, apoptosis could be induced when PCa cells were transiently transfected with ATF6α expression plasmid (5 μg).However, the growth rates of PCa cells that were transiently transfected with ATF6α expression plasmid (2.5 μg) or the ones with stable ATF6α overexpression was higher than those of their parental controls (Supplementary Fig. 2e,  2f).It could be inferred that the role of ATF6α in cellular destiny showed a dose-or context dependent manner.Thus, the dose of ATF6α expression plasmid (2.5 μg) was selected for subsequent experiments.We showed that Tm treatment could significantly reduce cell colony formation and cell proliferation whereas increase apoptosis of DU145 cells compared with control (Fig. 2d and Supplementary Fig. 2j-2 k).Additionally, the inhibitory effects of Tm on cell colony formation were further enhanced by treatment with ATF6α siRNA or its inhibitor but compromised by ATF6α overexpression (Fig. 2d).Next, in vivo experiments demonstrated that, introduction of ATF6α into DU145 cells showed an increased tumor growth Fig. 1 PTEN expression is adversely related with UPR gene signature and PTEN deficiency synergizes with ATF6α to activate UPR signaling in PCa cells.a, b GSEA of the UPR gene signature in GSE3325 and GSE6919 datasets, where genes were ranked by fold changes between PTEN low versus PTEN high in descending order.According to the PTEN expression levels, we arbitrarily defined the top 50% as the PTEN high group and the bottom 50% as the PTEN low group.GSEA, gene set enrichment analysis; NES, normalized enrichment score; FDR, false discovery rate.c Western blot analysis of the UPR related gene in the prostate tissues from WT (n = 6) or Pten PC-/- (n = 6) mice.P-AKT as a positive control to evaluate the efficiency of PTEN overexpression or silencing.WT, wild type; NC, negative control; group mice: every group mouse consists of four mice in total: WT mouse without Tm treatment, WT mouse with Tm treatment, Pten PC-/-mouse without Tm treatment and Pten PC-/-mouse with Tm treatment.d, e DU145 and LNCaP cells were transfected with indicated plasmid or siRNA followed by adding Tm (2.5 μM) for additional 12 h as indicated.Whole cell lysates were subjected to immunoblot with indicated antibodies.f-h Band intensity of ATF6α(N), P-PERK, P-IRE1α and BiP were measured using Image J software, presented as fold change compared with untreated WT, siNC or Vector respectively.i, j Cells as in (d-e) were harvested and real-time qPCR analyses were performed with primers of ATF6α target genes.k, l Cells as in (d-e) were harvested and the BiP promoter activity was identified by luciferase reporter assay.m, n DU145 cells were transfected with siRNA against PTEN.At 24 h post transfection, cells were transfected with ATF6α expression plasmid (m) or siRNA against ATF6α (n) and incubated for additional 48 h followed by adding Tm for another 6 h.Real-time qPCR analyses were then performed with primers of ATF6α target genes.CTL: Control.o, p Cells as in (m-n) were harvested and the BiP promoter activity was identified by luciferase reporter assay.Each data point represents one independent observation.Data represent mean ± SEM.P values were determined by two-tailed unpaired Student's t test; * P < 0.05, ** P < 0.01, *** P < 0.001 ◂ in the xenograft models when compared to its parental controls (998 ± 145.6mm 3 vs.456 ± 93 mm 3 ) (Fig. 2f,  h).Moreover, the protein levels of ATF6α and its target genes in DU145-ATF6α group were relatively increased compared with that in DU145-vector group (Fig. 2i, j).Collectively, our data suggested that ATF6α contributes to PCa progression.

PTEN interacts with ATF6α and inhibits ATF6α activity
To elaborate on the effect of PTEN on ATF6α, we sought to determine whether ATF6α physically associates with PTEN.As shown in Fig. 3a, HA-PTEN co-precipitated with the full-length Flag-ATF6α (short for ATF6α(P)), but not the N-terminal of ATF6α (short for ATF6α(N)) in LNCaP cells.Interestingly, Tm diminished the interaction between HA-PTEN and ATF6α(P).Additionally, the endogenous interaction between ATF6α(P) and PTEN in DU145 cells dramatically decreased with the presence of Tm (Fig. 3b).Finally, to determine whether PTEN directly interacts with ATF6α, GST, and recombinant GST-ATF6α were used to pull down the lysates of HEK293T cells with HA-PTEN overexpression.As shown in Fig. 3c, GST-ATF6α, but not GST protein, interacted with PTEN.Scale bar, 100 μm.e Quantitative analysis of ATF6α nuclear localization as shown in 3d.f LNCaP cells were transfected with PTEN expression plasmid.After 24 h, cells were transfected with ATF6α expression plasmid and then followed by adding Tm for another 6 h.BiP promoter activity was evaluated by luciferase reporter assay.g LNCaP cells were transfected with ATF6α expression plasmid.At 24 h post transfection, cells were transfected with PTEN expression plasmid and incubated for additional 48 h followed by adding Tm for another 6 h.ATF6α target genes were analyzed by real-time qPCR.h, i Following indicated treatment as in (g), the proliferation rates and colony formation number were measured by MTT and colony formation assays.Each data point represents one independent observation.Data represent mean ± SEM.P values were determined by two-tailed unpaired Student's t test; * P < 0.05, ** P < 0.01 Together, these results demonstrated the direct interaction between ATF6α and PTEN.
Immunofluorescent staining analysis showed that the nuclear staining of ATF6α markedly increased when LNCaP cells were treated with Tm, whereas the staining of ATF6α in nucleus significantly decreased when PTEN was overexpressed (Fig. 3d, e).We then investigated whether PTEN could inhibit the transcriptional activity of ATF6α using a BiP promoter-luciferase system.As shown in Fig. 3f, introducing PTEN profoundly diminished the transcriptional activity of BiP promoter in LNCaP cells with overexpression ATF6α.Likewise, PTEN can significantly suppress the mRNA levels of HERPUD1, SEL1L, and BiP induced by ATF6α overexpression in LNCaP cells (Fig. 3g).
We then performed rescue experiments to determine whether ATF6α could reverse the proliferation-inhibiting effect induced by PTEN overexpression.As shown in Fig. 3h and Supplementary Fig. 3a, 3b, ectopic PTEN could decrease the cellular proliferation of LNCaP, PC3, VCaP and DU145.However, such decrease could be partially restored by ectopic overexpression of ATF6α, suggesting that ATF6α could promote cell growth, at least in part, by deregulating PTEN.Accordingly, colony formation assay also confirmed that ATF6α reverses the inhibition of PTEN-induced cell colony formation (Fig. 3i and Supplementary Fig. 3c).Furthermore, overexpression of ATF6α could increase the cell proliferation of PCa cells with PTEN deficiency, which indicates that ATF6α may enhance cell growth independently of PTEN.Since BiP has been shown to control ATF6α export from ER through a dissociation mechanism, we next determined whether PTEN plays a role in their dissociation.The interaction between ATF6α and BiP was determined by immunoprecipitation.In consistent with the previous study, ATF6α dissociated from BiP when PCa cells were challenged with Tm.However, there was no visible change in dissociation when PTEN was overexpressed in LNCaP cells (Supplementary Fig. 3d) or knocked down in DU145 cells (Supplementary Fig. 3e).The micro-scale thermophoresis assay also showed that PTEN failed to affect the binding between BiP and ATF6α (Supplementary Fig. 3f).This suggested that the dissociation between BiP and ATF6α is not affected by PTEN.

ATF6α activity was repressed by PTEN mediated-phosphatase activity
We subsequently explored the molecular basis by which PTEN inactivates ATF6α.Since PTEN is a protein phosphatase that inactivates protein function by dephosphorylation [26,27], we first determined whether ATF6α could be phosphorylated or not in PCa cells.Our results shown that the tyrosine phosphorylation levels of ATF6α were dramatically increased when DU145 cells were stimulated by Tm and Tg (Fig. 4a, b and Supplementary Fig. 4a).To define the exact phosphorylation site on ATF6α, we constructed several plasmids targeting the tyrosine sites of ATF6α.The results showed that compared with ATF6α WT , the expression levels of phosphorylated and spliced ATF6α were decreased more significantly in ATF6α Y111A in the presence of Tm (Fig. 4c and Supplementary Fig. 4b).Concordantly, realtime qPCR analysis demonstrated that the expression level of HERPUD1 was significantly decreased in ATF6α Y111A group than ATF6α WT (Fig. 4d).In addition, PTEN only pulled down ATF6α WT but not ATF6α Y111A , which suggests that Y111 is required for its interaction with PTEN (Fig. 4e).To examine whether PTEN dephosphorylates ATF6α or not, PTEN was silenced in DU145 (Fig. 4f) or overexpressed in LNCaP cells (Fig. 4g).The levels of phosphorylated ATF6α(P) were significantly increased in DU145 cells but decreased in LNCaP cells compared with their control.However, no observable change was detected when PTEN as silenced or overexpressed in DU145 or LNCaP cells with mutant ATF6α Y111A (Fig. 4f, g).In addition, the in vitro phosphatase assay also showed that the phosphorylated levels could not be disturbed in ATF6α Y111A compared to its wild type control (Supplementary Fig. 4c), indicating tyrosine residue Y111 is important for the phosphorylation of ATF6α.
To further confirm the specific phosphatase sites of PTEN for ATF6α inactivating dephosphorylation, we transfected LNCaP cells with a variety of PTEN expression constructs with mutations diagramed in Fig. 4h.These mutations include the protein phosphatase dead PTEN Y138L , the lipid phosphatase-dead PTEN G129E , dual phosphatase-deficient PTEN C124S , and PTEN G129R constructs.After transfection with PTEN G129E , the levels of phosphorylated ATF6α were reduced in LNCaP cells (Fig. 4i).Next, we also observed that PTEN WT and PTEN G129E could dephosphorylate ATF6α directly in vitro (Supplementary Fig. 4d).In addition, none of the PTEN constructs affected ATF6α Y111A phosphorylation levels (Fig. 4j).Collectively, these data indicated that PTEN inhibits ATF6α activation by dephosphorylation.To Fig. 4 ATF6α activity was repressed by PTEN mediated-phosphatase activity.a, b DU145 cells were treated with or without Tm (2.5 μM) (a) or Thapisigargin (Tg) (0.5 μM) (b) for 6 h.Whole lysates were then subjected to immunoprecipitation using anti-phospho-serine (ser), tyrosine (tyr), and threonine (thr) specific antibodies.c DU145 cells were transfected with flag-ATF6α WT or flag-ATF6α mutants (Y111A, Y100A, Y392A) with or without Tm for 6 h.Whole lysates were then subjected to immunoprecipitation using anti-phosphotyrosine antibody.d Cell as in (c) and then the expression of ATF6α target gene were measured by real time-qPCR.e The binding between PTEN and flag-ATF6α WT or flag-ATF6α Y111A were performed using co-immunoprecipitation assays in DU145 cells transfected with flag-ATF6α WT or flag-ATF6α Y111A .f, g DU145 cells were transfected with siRNA against PTEN (f) and LNCaP cells were transfected with PTEN expression plasmid (g) for 24 h and then introduced with flag-ATF6α WT or flag-ATF6α Y111A plasmids for 48 h, followed by treatment of Tm for 6 h.Tyrosine phosphorylation levels of ATF6α were then measured by immunoprecipitation and western blot.h A graphical representation of the PTEN domain structure.Mutations in the phosphatase and tail domains, including C124S (phosphatase-dead), G129R (phosphatase-deficient), Y138L (protein phosphatase dead), G129E (lipid-phosphatase dead) were shown.i LNCaP cells were transfected with empty vector, PTEN WT , or PTEN mutants (Y138L, G129R, C124S, or G129E) for 48 h.Cells were then lysed and subjected to immunoprecipitation using anti-phosphotyrosine antibody.j DU145 cells were transfected with PTEN WT , PTEN mutation overexpression plasmid (C124S or C129E) for 24 h and then introduced with flag-ATF6α WT or flag-ATF6α Y111A for 48 h.Tyrosine phosphorylation levels of ATF6α were then measured by immunoprecipitation. k Cytoplasmic and nuclear ATF6α protein levels in DU145 cells transfected with flag-ATF6α WT or flag-ATF6α Y111A in the presence or the absence of Tm for 12 h were analyzed by western blot.l, m Cells as in (k) and then BiP promoter activity was measured by luciferase reporter assay (l) and the mRNA levels of HERPUD1, SEL1L, and BiP (m) were analyzed by real-time qPCR.Each data point represents one independent observation.Data represent mean ± SEM.P values were determined by two-tailed unpaired Student's t test; * P < 0.05, ** P < 0.01, *** P < 0.001 ◂ examine the functional consequence of PTEN-mediated dephosphorylation of ATF6α, we analyzed its nuclear translocation in DU145 cells.Our data showed that ATF6α WT , but not ATF6α Y111A was transported into the nucleus in the presence of Tm (Fig. 4k).Furthermore, as shown in Fig. 4l, ATF6α Y111A exhibited low transcriptional activity in both unstressed and stressed cells when compared with ATF6α WT , which was consistent with the results obtained Page 11 of 17 292 from real-time qPCR of HERPUD1, SEL1L, and BiP (Fig. 4m).These results highlighted that PTEN-inactivitymediated phosphorylation is required for the protein cleavage, nuclear translocation, and the subsequent transcription modulation of ATF6α.

ER stress induces PTEN protein instability through ATF6α and ATF6α destabilizes PTEN protein via CHIP-mediated polyubiquitin
Since a significant decrease in PTEN protein levels were observed in mice prostate tissues in the presence of Tm (Fig. 1c), we then treated DU145 and VCaP cells with ER stress inducers in vitro.We found that there was a progressive reduction in PTEN expression in a time-and dosedependent manner at the protein levels (Fig. 5a and Supplementary Fig. 5b) but not at the mRNA levels (Supplementary Fig. 5c), suggesting the potential of posttranslational modification.We then aimed to determine of which the UPR signaling pathways (ATF6α, PERK, and IRE1) was required for the decrease of PTEN expression by using MEFs from Atf6α, Perk, and Ire1 knockout mice.Our data showed that the reduction of PTEN protein levels caused by Tm challenge could be partly reversed in Atf6α −/− MEFs compared with those in Perk −/− and Ire1 −/− MEFs (Fig. 5b).Similar effects were also observed when ATF6α, PERK or IRE1 was silenced in DU145 cells (Fig. 5c).However, PTEN protein levels could be rescued after introducing Atf6α into Atf6α −/− MEFs (Fig. 5d).Collectively, these results suggest that Tm-induced decrease of PTEN protein expression was mediated by ATF6α.
To characterize the inhibitory effect of Tm on PTEN expression, we treated DU145 cells with MG132 (a proteasome inhibitor) or leupeptin (a protease inhibitor).Our data showed that MG132, but not leupeptin, could rescue Tm-induced PTEN protein levels downregulation (Fig. 5e and Supplementary Fig. 5d).This suggested that PTEN protein undergoes degradation in a proteasome-dependent manner in the presence of Tm.Our results were consistent with previous report, which demonstrated that PTEN protein levels were regulated by both polyubiquitylation and monoubiquitylation [28].Further analysis showed that more polyubiquitination in PTEN protein occurred in the presence of Tm stimulation or ATF6α overexpression in DU145 cells (Fig. 5f, g).Moreover, polyubiquitination of PTEN proteins could hardly be detected in Atf6α −/− MEFs when compared to WT (Fig. 5h).However, such decrease in polyubiquitination could be restored by introducing ATF6α into Atf6α −/− MEFs (Fig. 5h).These results confirmed that ATF6α plays an important role in PTEN degradation induced by Tm.
We further investigated the molecular mechanism by which ATF6α mediated PTEN suppression.It was previously reported that CHIP, NEDD4-1, XIAP, and WWP-2 [29][30][31][32] regulated PTEN by promoting protein degradation.To search for genes that participate in PTEN degradation induced by ATF6α, we performed real-time qPCR assay in DU145 cells treated with Tm.The data showed that the increase of CHIP mRNA levels in DU145 cells induced by Tm could be reduced by silencing ATF6α (Supplementary Fig. 5e).This result was also confirmed by western blot (Fig. 5i, j and Supplementary Fig. 5f).Furthermore, CHIP protein levels were decreased more significantly in Atf6α −/− MEFs, compared with those of Perk −/− and Ire1 −/− MEFs (Supplementary Fig. 5 g).However, such inability in Atf6α −/− MEFs could be rescued by introducing ATF6α (Fig. 5k).These data supported that ATF6α was important for protecting CHIP protein stability.We next aimed to assess whether ATF6α destabilizes PTEN protein via CHIP-mediated polyubiquitination.As shown in Fig. 5l, the knockdown of CHIP expression inhibited PTEN Fig. 5 ER stress induces PTEN protein instability through ATF6α and ATF6α destabilizes PTEN protein via CHIP-mediated polyubiquitin.a DU145 cells was treated with Tm (2.5 μM) in indicated time or concentration, indicated antibodies were tested by western blot.b, c PTEN protein levels were detected by western blot in WT, Atf6α −/− , Perk −/− , and Ire1 −/− MEFs (b) and in DU145 cells with ATF6α, PERK or IRE1 knockdown.d Atf6α −/− MEFs were transfected with ATF6α expression plasmid and then treated with Tm (2.5 μM) for additional 6 h, indicated antibodies were tested by western blot.e DU145 cells were treated with Tm (2.5 μM) for 6 h and then treated with vehicle, 20 µM MG132 or 20 µM Leupeptin for additional 6 h.PTEN protein levels were detected by western blot.f DU145 cells were transfected with siRNA against ATF6α for 48 h and then were treated with Tm for another 6 h.Cell lysates were subjected to immunoprecipitation using anti-PTEN antibody.g DU145 cells were transfected with siRNA against ATF6α for 24 h and then transfected with ATF6α expression plasmid for another 48 h followed by adding Tm for another 6 h.Cells were then lysed and subjected to immunoblotting using anti-PTEN antibody.h Cell as in (d) and then cell lysates were performed immunoprecipitation with PTEN antibody.i Cell as in (f) and then cells were lysed and blotted for CHIP and tubulin.j DU145 cells were pretreated with Ceapin-A7 for 12 h and then treated with Tm for another 6 h.Cell lysates were subjected to immunoprecipitation using anti-CHIP antibody.k WT and Atf6α −/− MEFs were transfected with ATF6α expression plasmid for 48 h, cells were lysed and blotted for CHIP and actin.l DU145 cells were transfected with ATF6α expression plasmid for 24 h and then transfected with siRNA against CHIP for another 48 h.Cells were then lysed and subjected to immunoprecipitation using anti-PTEN antibody.m DU145 cells were transfected with siRNA against ATF6α for 24 h and then transfected with CHIP expression plasmid for another 48 h.Cells were then lysed and subjected to immunoprecipitation using anti-PTEN antibody.n ChIP analysis of CHIP was performed in DU145 cells transfected with overexpression ATF6α plasmid for 48 h followed by adding Tm for another 6 h.o DU145 cells were transfected with a reporter plasmid containing the luciferase gene fused with intact or mutant CHIP promoters for 48 h, CHIP promoter activity was then quantified by dual-luciferase reporter assays.Each data point represents one independent observation.Data represent mean ± SEM.P values were determined by two-tailed unpaired Student's t test; *** P < 0.001 ◂ 292 Page 12 of 17 polyubiquitination induced by ATF6α in DU145 cells.Similarly, the decreased PTEN polyubiquitination caused by ATF6α knockdown could be reversed by introducing CHIP expression in DU145 cells (Fig. 5m).
To determine whether ATF6α directly regulates CHIP expression, we conducted a luciferase activity assay with CHIP promoter containing reporter plasmid.Our data showed that Tm treatment could significantly increase the transcriptional activity of CHIP promoter, which also could be enhanced by overexpressing ATF6α (Fig. 5n).It was previously reported that TGAC is the core sequence critical for ATF6α binding [33].We aligned this sequence with CHIP promoter from − 2000 ~ + 300 bp and found two perfectly matched sequences.We constructed two TGAC-mutant plasmids and performed transient transfection assays with Tm treatment.As shown in Fig. 5o, mutant construct failed to affect CHIP activity in the presence of Tm.
To determine which one plays more important role in PTEN-ATF6α loop in vitro, we transfected DU145 cells with PTEN or ATF6α overexpression plasmid for 48 h followed by ATF6α or PTEN overexpression.Cell colony formation ability was then evaluated.As shown in Supplementary Fig. 5 h, PTEN reduced cell colony number by 50-60% in ATF6α overexpression cells, while ATF6α rescued colony number by 30% in PTEN overexpression cells irrespectively of ER stress inducers.These results implied that PTEN overweighs ATF6α in deregulating PCa progression.

Combined inhibition of ATF6α and AKT synergistically delays PCa progression
In the context of PTEN inactivity, the most pronounced signaling event is the constitutive activation of AKT which associates with the UPR pathway [34].Indeed, we identified that silencing ATF6α could attenuate the activation of PI3K/AKT pathway in DU145 cells (Fig. 6a).Similar effects were also observed in Atf6α −/− MEFs when compared with WT MEFs (Fig. 6b).Furthermore, the decreased activity of AKT signaling elicited by ATF6α deficiency was rescued by re-introducing ATF6α in Atf6α −/− knockout MEFs (Fig. 6c).Either PTEN silencing or ATF6α overexpression enhanced AKT activity in DU145 cells, and the combination of them was more effective in increasing AKT activity (Fig. 6d).
We then evaluated the potential benefit of co-targeting ATF6α and AKT in PCa cells.As shown in Fig. 6e, combination of AKT inhibitor (AZD5363) with ATF6α inhibitor (Ceapin-A7) was more effective in inhibiting cell proliferation in DU145 and VCaP cells.Similar to the pharmacologic inhibition, siATF6α combined with AKT inhibitor treatment significantly reduced cell proliferation of PCa cells (Fig. 6f).
To determine the impacts of AZD5363 and Ceapin-A7 on tumor growth, we have established DU145 xenograft model.As shown in Fig. 6g-i, either AZD5363 or Ceapin-A7 treatment reduced the tumor volume, and a combination of AZD5363 and Ceapin-A7 induced further inhibition in tumor growth.The expression level of activated ATF6α was detected by western blot and real-time qPCR assay in these tumors (Fig. 6j, k).These results suggested that ATF6α inhibitor combined with AKT inhibitor would be useful for the treatment of PCa.

Discussion
PTEN is one of the most frequently mutated, deleted, and functionally inactivated tumor suppressor gene in PCa.Loss of PTEN is an important event in prostate carcinogenesis and represents one hallmark of PCa [35].Clinically, deletion or mutation of at least one PTEN allele was reported to occur in 20-40% of localized cancers and up to 60% of metastases [5].PTEN inactivation is strongly linked to advanced PCa progression and poor clinical outcome.
Given the high frequency of PTEN loss in PCa, therapeutic strategies that exploit the functional loss of PTEN may prove to be efficacious against PTEN-deficient PCa.Inactivation of PTEN leads to abnormal activation of PI3K/ AKT pathway.Multiple studies have demonstrated a key role of the PI3K/AKT signaling axis in the development of PCa and maintenance of CRPC [36,37].Until recently, however, results from clinical trials with AKT inhibitors in PCa have been largely disappointing.Crabb et al. showed that pan-AKT inhibitor capivasertib with docetaxel and prednisolone failed to improve progression-free survival in patients with mCRPC irrespective of PI3K/AKT/PTEN pathway activation status [8].Therefore, an appropriate design of rational combinations of PI3K/AKT inhibitors with other novel agents is urgently needed.
Of note, many studies have shown that PTEN can synergize with other oncogenic factors or related genes in mouse models, including NKX3.1 [38], p27 [39], and p53 [40] to promote PCa development and androgen independence.Previously, we have demonstrated association of PTEN deletion with ERG rearrangement during PCa progression [5].To the best of our knowledge, these observations suggested that a continuum rather than a stepwise mode of PTEN activity best described the process triggering tumor initiation and progression.In the current study, we discovered a novel reciprocal negative feedback regulation of ATF6α and PTEN that promotes PCa progression (Fig. 7).We demonstrated that 1) ATF6α is a novel negative regulator of PTEN by promoting PTEN protein degradation.2) PTEN can shift the UPR pathway by selective inactivation of ATF6α.3) Cotargeting ATF6α and AKT suppresses tumor cell proliferation and xenograft growth in PCa.These results highlighted co-targeting the AKT pathway and ATF6α oncogenic signaling as potential future therapies for PCa cases with PTEN Page 13 of 17 292 Fig. 6 Combined inhibition of ATF6α and AKT synergistically delays PCa progression with PTEN deficiency.a DU145 cells were transfected with siRNA against ATF6α for 48 h and then treated with 2.5 μM Tm for 12 h.The P-AKT and AKT protein levels were analyzed by western blot.b WT and Atf6α −/− MEFs were treated with Tm for 12 h.The P-AKT and AKT protein levels were analyzed by western blot.c WT and Atf6α −/− MEFs were transfected with ATF6α expression plasmid for 48 h and then treated with 2.5 μM Tm for 12 h.Whole cell lysates were then subjected to immunoblot analyses with indicated antibodies.d DU145 cells were transfected with siRNA against PTEN for 24 h and then transfected with ATF6α expression plasmid for another 48 h.Cells were lysed and blotted for P-AKT, AKT and tubulin.e DU145 cells were treated with increasing doses of AZD5363 (0, 5 μM, 10 μM, 20 μM) or Ceapin-A7 (0, 10 nM, 25 nM, 50 nM) alone or in combination with Ceapin-A7 (25 nM) or AZD5363 (10 μM), cell proliferation was detected by MTT at 72 h.f DU145 cells were transfected with siRNA against ATF6α for 24 h followed by AZD5363 treatment.Cell proliferation rates were measured by MTT.g-i Mice implanted with DU145 cells were treated with vehicle control, Ceapin-A7 (10 mg/kg), AZD5363 (100 mg/kg), or Ceapin-A7 + AZD5363 (n = 5).Tumor size was measured twice a week.At the endpoint, tumors isolated from euthanized mice were weighed and photographed.j, k The Xenograft tumors were collected 4 weeks after injection.The levels of endogenous ATF6α and its target genes in the excised tumors were determined by western blot and real time-qPCR analysis.Data represent mean ± SEM.P values were determined by two-tailed unpaired Student's t test; ** P < 0.01, *** P < 0.001, **** P < 0.0001 dysfunction.Indeed, while this study was in preparation, Yu et al. reported that PTEN phosphatase activity inhibits metastasis by negatively regulating the Entpd5/IGF1R pathway through ATF6, thereby identifying novel candidate therapeutic targets for the treatment of PTEN mutant melanoma [41].
ATF6 is a transmembrane glycoprotein and transcription activator, which functions to initiate the UPR signaling during ER stress.The UPR is divided into three arms, including the PERK, ATF6 and IRE1.These three major effector proteins frequently cooperate in the transcriptional regulation of downstream targets.UPR signaling can affect various aspects of tumor cell biology, including angiogenesis, invasion, and therapy resistance [15].Recent works have established that steroid hormone pathway, in particular androgens and estrogens related signaling, can regulate the UPR and enhance survival benefits for breast cancer and PCa cells [42].In the current study, we demonstrated that UPR pathway is preferential activated in PTEN-deficient PCa.More importantly, ATF6α branch is closely related to PTEN dysfunction compared with that of PERK and IRE1.Functionally, silencing ATF6α, but not PERK or IRE1, can significantly attenuated the increase of cell proliferation induced by PTEN knockdown in vitro.
Although it has been documented that ATF6α is a major mammalian UPR sensor, there is limited data on the potential role of ATF6α in cancer progression.It was reported that activation of ATF6α can have both pro-survival and prodeath outcomes in cancers.There is evidence that ATF6α could drive cell proliferation in hepatoma cells [43], whereas knockdown of ATF6α in colon reduced cell proliferation [44].However, in colorectal cancer cells, enforced expression of ATF6α reduced cell viability after exposure to low levels of Tg by reduced global protein synthesis [45].ATF6 was also suggested to be involved in the process of UPR signaling-induced apoptotic pathways by the ER stress master regulator BiP/Grp78 [46].Therefore, the exact role of ATF6α in tumors may be cell type-and context-specific.Our results showed overexpression of ATF6α increases the growth and colony formation of DU145 especially under the condition of Tm treatment.Wu et al. reported that exposure to either Tg or Tm significantly reduces the proliferative capacity of ATF6α-/-cells relative to wild-type cells [47].Most recently, Pachikov et al. reported that GCC185 played an important role in translocation of S1P and S2P proteases to the ER, which subsequent promoted ATF6 activity [48].However, the exact role of ATF6α in PCa merit further investigation.
PTEN is a dual phosphatase with both protein and lipid phosphatase activities.PTEN has been reported to directly dephosphorylate residues on itself and others, to exert its tumour-suppressive function.PTEN deficiency is resulting mainly from PTEN deletion or mutation.However, PTEN can also be regulated at transcriptional, post-transcriptional and translation levels.Post-translational modifications, such as phosphorylation and ubiquitylation, have been shown to decrease PTEN protein levels [49].Polyubiquitylation and monoubiquitylation play important roles in the regulation of PTEN degradation, in which NEDD4-1 [32], WWP2 [30], XIAP4 [31], and CHIP [29] are involved.Of them, CHIP is a highly conserved ubiquitin E3 ligase that can control the ubiquitination of multiple substrates, including p53, c-Myc, ErbB2, EGFR and PTEN [50].In this study, our results demonstrated that PTEN inhibits the activity of ATF6α mainly by disturpting its phosphorylation at the Y111 site, while its lipid phosphatase is responsible for this function.Conversely, ATF6α promotes the ubiquitination and degradation of PTEN, thereby promoting cell proliferation.
Co-targeting strategies strive to improve cancer outcomes by combining therapies that selectively target exploitable alterations in tumor cells.In this study, we suggested that ATF6α and PTEN forms a negative feedback loop during PCa progression.It is, therefore, reasonable to believe that combinational therapies by co-targeting PI3K/AKT and other oncogenic pathways would provide optimal clinical benefits.We have shown evidence that co-treatment of ATF6α inhibitor (Ceapin-A7) and AKT inhibitor (AZD5363) resulted in stronger tumor suppression than AKT inhibitor alone.Our data demonstrated that co-targeting ATF6α and PI3K/AKT pathways may be a promising therapeutic approach for PCa, especially the ones with PTEN dysfunction.

Conclusions
In summary, ATF6α and PTEN forms a negative feedback loop during PCa progression.Combination of ATF6α inhibitor with AKT inhibitor suppresses tumor cell proliferation and xenograft growth.Importantly, this study highlighted ATF6α as a therapeutic vulnerability in PTEN dysfunctional PCa.Knowledge of the PTEN-ATF6α loop may prove to be valuable in the design of novel preventive and interventive therapeutic strategies in PCa.

A
total of three tissue microarrays were constructed representing 15 patients with BPH and 130 clinically localized PCa patients who underwent radical prostatectomy between 2012 and 2015 at Qilu Hospital of Shandong University (Jinan, China).Three cores (0.6 mm in diameter) were taken from each representative tumor focus and the morphology was confirmed by two pathologists (J.H. and M.Q).The localized tissues consisted of Gleason score 3 + 3 = 6 (n = 17), Gleason score 3 + 4 = 7 (n = 20), Gleason score 4 + 3 = 7 (n = 24), Gleason score 4 + 4 = 8 (n = 30), and Gleason score ≥ 9 (n = 39) tumors.This study was conducted in accordance with the International Ethical Guidelines for Biomedical Research Involving Human Subjects.This study protocol was approved by Shandong University Medical Research Ethics Committee according to the Declaration of Helsinki (Document No. ECSBMSSDU2021-1-61).

Fig. 2
Fig. 2 ATF6α protects against ER stress-induced cytotoxicity and promotes PCa progression in vivo.a, b Protein expression of PTEN and ATF6α in BPH and PCa tumor specimens.a Upper: Representative H&E images of BPH and PCa cases with different GS (original magnification, 400 ×).Middle & Bottom: Representative IHC staining of PTEN and ATF6α in BPH and PCa cases with different GS (original magnification, 400 ×).b H-scores for ATF6α nuclear expression in BPH and PCa cases with different GS.Scale bars, 50 μm.BPH, Benign prostatic hyperplasia; GS, Gleason score.c Kaplan-Meier analysis comparing overall survival of PCa patients with low and high ATF6α expression in TCGA cohort.d DU145 cells were transfected with ATF6α overexpression plasmid or siRNA

Fig. 3
Fig.3PTEN interacts with ATF6α and inhibits ATF6α activity.a LNCaP cells were transfected with flag-ATF6α full length, flag-ATF6α N terminal or empty vector.After 24 h, cells were transfected with HA-PTEN for another 48 h followed by adding Tm (2.5 µM) for another 6 h.Cells were then lysed and subjected to immunoprecipitation using anti-HA antibody.ATF6α(P): ATF6α full length plasmid; ATF6α(N): ATF6α N terminal plasmid.b Co-precipitation between endogenous ATF6α and PTEN in DU145 cells treated with Tm for 6 h.c HEK293T cells were transfected with HA-PTEN and cell lysates were incubated with GST or GST-ATF6α fusion protein and the direct association of ATF6α with PTEN was detected.d LNCaP cells were transfected with PTEN or empty vector.At 48 h post transfection, the cells were treated with Tm for 6 h and then staining with ATF6α antibody (red).Nuclei were counterstained with DAPI (blue).

Fig. 7
Fig. 7 Schematic representation of the ATF6α-PTEN axis in promote PCa progression.PTEN expression level was adversely correlated to ATF6α signaling in PCa.PTEN suppresses ATF6α activity by