Introduction

Over 70% of newly diagnosed bladder cancer (BC) patients have non-muscle invasive BC (NMIBC) confined to the urothelium and underlying lamina propria [1, 2]. NMIBC patients face high rates of recurrence, with two-thirds experiencing tumor recurrence within five years, and up to 88% within 15 years [3]. One proposed explanation for these high rates of recurrence involves the cancer field effect wherein pre-malignant cells are predisposed to tumor development, potentially contributing to the multi-chronotropic and multifocal nature of recurrent NMIBC [4]. How this occurs remains poorly understood, with some work supporting the stepwise accumulation of genetic alterations that ultimately result in tumor formation, whereas other studies suggest clonal expansion from a single common precursor as the major mechanism of recurrent tumor growth [5,6,7]. Field cancerization may drive tumorigenesis regardless of the exact nature of these genetic transformation events. A comprehensive genetic analysis of bladder cancer samples using datasets from The Cancer Genome Atlas (TCGA) project identified multiple genomic alterations, suggesting that these progressive tumors are heterogeneous and can result from a permissive oncogenic environment in the whole bladder [8]. Those tumors can recur anywhere in the bladder and may or may not share similar histology with the primary tumors [9], further supporting the theory that the entire bladder is permissive to tumorigenesis.

The induction of endoplasmic reticulum (ER) stress responses can be cytoprotective [10], with unfolded protein accumulation within the ER lumen triggering a coordinated unfolded protein response (UPR) that can restore ER homeostasis [11]. However, insufficient or sustained ER stress responses result in pathologic alterations that can lead to oncogenesis [12,13,14]. Indeed, ER stress and UPR induction are common features of human cancers [15]. Highly proliferative tumors are exposed to several intrinsic and extrinsic stressors [16], potentially explaining the enhanced UPR signaling activity in these cells as a survival strategy that enables them to better tolerate exposure to stressful environments. The pronounced reliance of many cancer cells on the UPR has prompted interest in targeting this pathway as a form of anti-cancer therapy [15]. However, the advancement of these strategies beyond the preclinical stage has been hampered by concerns regarding off-target effects [17], prompting a need for more in-depth studies of the molecular machinery governing the UPR in specific cancers.

Members of the protein disulfide isomerase (PDI) family, including the canonical PDIA1 encoded by P4HB, are molecular chaperones and thiol-disulfide oxidoreductases abundantly present within the ER lumen [18, 19]. PDIA1 can be phosphorylated and activated in response to UPR induction [20], and it functions in part by catalyzing disulfide bond formation and isomerization to help alleviate this bottleneck in the oxidative protein folding process [18, 19]. Oxidized PDIA1 can effectively donate its disulfide bond to an unfolded or misfolded protein by accepting electrons from the thiol groups of these polypeptides such that they can fold into an appropriate conformation and form proper disulfide bonds [21]. The reduced form of PDIA1 can then recycle to the catalytically active oxidized state by passing these electrons to ER oxidoreductin 1 (ERO1), which, in turn, generates hydrogen peroxide (H2O2) that can contribute to oxidative stress [21, 22]. P4HB upregulation has been reported in bladder tumors and linked to worse pathological staging, overall survival, and recurrence-free survival in patients, with corresponding overexpression in BC cell lines supporting tumor cell proliferation and invasion [23, 24]. P4HB knockdown can sensitize BC cells to gemcitabine [25], and it is a platinum-resistance-related gene in BC patients such that the knockout of this PDIA1-coding gene sensitizes bladder tumor cells to platinum-based treatment [26].

While these prior data support an important pro-tumorigenic role for PDIA1 in BC, relatively little remains known of the precise mechanisms whereby levels of its expression and activity ultimately shape malignant outcomes. Strikingly, extant data suggest that PDIA1 can function as a dual-edged sword such that it can support cancer growth by activating the PERK branch of the UPR pathway to facilitate tumor proliferation and survival [27, 28], whereas its excessive induction of NADPH oxidase activity can result in deleterious levels of reactive oxygen species (ROS) production and cell death [29]. It may thus be incumbent on tumor cells to maintain levels of PDIA1 activity sufficient to adapt to ER stress and maintain proteostasis while mitigating the potential for lethal oxidative stress associated with unrestrained ROS generation.

Extracellular vesicles (EVs) are small membrane-enclosed structures released from cells that enable the intercellular transmission of macromolecular cargos. We and others have demonstrated that certain cargo proteins present within tumor-derived EVs can promote tumorigenesis [15, 30,31,32]. We have further found that BC cell-derived EVs can drive UPR induction and oncogenic transformation in recipient urothelial cells [33]. This, coupled with the potential need for tumor cells to regulate PDIA1 activity and attendant oxidative stress within a tolerable range conducive to proliferation and drug resistance, raises the possibility that EVs may provide a release valve to manage levels of intracellular stress. Indeed, EV-mediated relief of excessive ER stress has been proposed in developmental settings [34], and may be co-opted by tumor cells, indirectly exposing non-malignant recipient cells to this stress in the process.

The goal of this study was to explore the role of PDIA1 as a regulator of BC malignancy, with a focus on the maintenance of homeostatic PDIA1 activity, the role of EVs in this context, and the ability of PDIA1-containing EVs to induce normal urothelial cell transformation through the use of loss-of-function and rescue approaches. We further explored the potential relevance of PDIA1 expression to NMIBC recurrence through a retrospective tissue microarray-based analysis to gain direct insight into the potential clinical relevance of these mechanisms.

Results

Reduced PDIA1 expression promotes bladder cancer cell survival under elevated ER stress

In an effort to begin exploring the role that PDIA1 plays in cancer cells, we used short hairpin RNA (shRNA) constructs to knock down PDIA1 in TCCSUP BC cells (Fig. 1A). To probe the relationship between PDIA1 expression and the survival of tumor cells in the presence or absence of ER stress, tunicamycin (140 nM), a general ER stress inducer, was used to treat these tumor cells. While tunicamycin significantly increased the frequency of propidium iodide (PI)-positive TCCSUP cells expressing the scramble control shRNA, it had no impact on the frequency of PI-positive cells in which PDIA1 had been knocked down (Fig. 1B). Consistently, PDIA1 knockdown reduced the degree of tunicamycin-induced caspase-3 activation in these BC cells (Fig. 1C) while abrogating the ability of tunicamycin to compromise colony formation in a clonogenic assay (Fig. 1D). Similar effects were also observed in the J82 BC cell line (Supplementary Fig. S1). These results suggest that elevated levels of ER stress compromise the survival of BC cells, while the silencing of PDIA1 can restore the tolerance of these malignant cells to this form of stress.

Fig. 1: Reduced PDIA1 expression promotes bladder cancer cell survival under elevated ER stress.
figure 1

A Western blot analysis of PDIA1 abundance in cell lysates derived from scramble control and PDIA1-targeting lentiviral shRNA transduced TCCSUP cells. The two clones exhibiting the lowest PDIA1 expression were used for future experiments. TCCSUP cell death was measured by propidium iodide (PI) staining and quantified using flow cytometry (B). Apoptosis was assessed by Western blotting analyses of cleaved caspase-3 (C), BiP was used as a marker for ER stress. D TCCSUP cancer cell survival following tunicamycin treatment was tested in a clonogenic assay. Scramble control and shPDIA1 TCCSUP cells were treated with tunicamycin (140 nM Tun; in DMSO) or vehicle control (DMSO) and examined to assess H2O2 levels (D), oxidative stress gene expression (E), and the GSH/GSSG ratio (F). For (B), (D), (E), individual genes in (F), and (G), two-way ANOVAs were performed with Fisher’s LSD multiple comparison test; n = 3; error bars indicate means ± standard deviation. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. P parental TCCSUP, SCR scramble control, EV extracellular vesicle, PDIA1 protein disulfide isomerase A1, shPDIA1-1/2 one of two short hairpin RNAs targeting PDIA1, PI propodium iodide, BiP binding immunoglobulin protein, ANOVA analysis of variance, DMSO dimethyl sulfoxide, Tun tunicamycin, NFE2L2 Nuclear factor erythroid-derived 2-like 2, NQO1 NAD(P)H: quinone oxidoreductase 1, GCLC Glutamate-cysteine ligase.

Full size image

Given that ROS production is a byproduct of PDIA1 activity, it is possible that high levels of ER stress may expose BC cells to increased oxidative stress, thereby contributing to the induction of apoptotic death. To test this possibility, H2O2 production was analyzed, revealing that PDIA1 knockdown was sufficient to significantly reduce basal H2O2 levels while also markedly suppressing tunicamycin-induced production thereof (Fig. 1E). In line with these results, tunicamycin significantly induced the upregulation of oxidative stress-related genes (NFE2L2, NQO1, GCLC) in control TCCSUP cells, whereas it failed to do so in these cells following PDIA1 knockdown (Fig. 1F). Moreover, tunicamycin significantly reduced the GSH/GSSG ratio in scramble control TCCSUP cells, whereas it had no impact on this ratio following PDIA1 silencing (Fig. 1G). Together, these data support a potential model wherein PDIA1-mediated ROS production can, under conditions of elevated ER stress, compromise redox homeostasis within BC cells, thereby inducing their apoptotic death. As higher levels of basal caspase-3 activity were observed following PDIA1 knockdown (Fig. 1C), however, this protein may play an important pro-survival role under conditions of reduced ER stress, suggesting that tumor cells need to carefully calibrate their intracellular PDIA1 supply to maintain viability.

Bladder cancer cells mediate PDIA1 homeostasis through EV release

EV-mediated release has been proposed as an important mechanism through which cells can eliminate any undesirable molecules [35, 36]. We thus speculated that BC cells may leverage secreted EVs as a means of maintaining intracellular PDIA1 levels within a tolerable range that balances the beneficial effects of this protein against its potential to induce excessive oxidative stress. Strikingly, tunicamycin treatment markedly increased the levels of PDIA1 found in EVs isolated from TCCSUP cells, whereas it had no impact on PDIA1 cargo levels within EVs isolated from non-transformed SV-HUC urothelial cells (Fig. 2A). This finding was further supported by the observation that PDIA1 colocalized with the exosome/multivesicular body marker TSG101 in tunicamycin-treated TCCSUP and J82 BC cells, while this colocalization was not apparent in SV-HUC cells (Fig. 2B). A reduced thiol quantification approach was further used to examine the redox composition of PDIA1 within these cells and EVs. The majority of PDIA1 detected within both TCCSUP and SV-HUC cells was present in the oxidized (oxidoreductase-active) form, but the frequency of reduced PDIA1 rose significantly in TCCSUP but not in SV-HUC cells in response to tunicamycin treatment (Fig. 2C). Interestingly, we found that the majority of PDIA1 found in TCCSUP cell-derived EVs was present in the oxidized form (89.2%; Fig. 2D). We thus posited that tumor cells may disfavor oxidative protein folding under excessively high levels of ER stress, packaging oxidized PDIA1 into EVs such that it can be exported from the cell to preserve viability and mitigate oxidative stress. In further support of this model, we found that TCCSUP cells released significantly more EVs at baseline as compared to SV-HUC cells, while tunicamycin treatment significantly increased the release of EVs from these cells (Fig. 2E). As tunicamycin also triggered higher levels of EV release in non-transformed SV-HUC cells, such stress-induced EV secretion may be general strategy by which cells can adapt to ER stress, with PDIA1 packaging into these EVs being a particularly beneficial pro-survival mechanism engaged by BC cells.

Fig. 2: Cancer cells mediate PDIA1 homeostasis through EV release.
figure 2

A Western blot showing PDIA1 levels in whole cell lysates and EVs derived from non-transformed SV-HUC and TCCSUP cancer cells with tunicamycin (140 nM Tun; in DMSO) or vehicle control (DMSO) treatment. CD9 was used as an EV marker and normalization control for EV proteins. B Immunofluorescence staining demonstrating PDIA1 and TSG101 intensity and cellular localization in SV-HUC, TCCSUP, and J82 cancer cells with tunicamycin (140 nM Tun) or vehicle (0 nM Tun) treatment. Scale bars: 10 µm. C The percentage of reduced PDIA1 in TCCSUP cancer cells (left) or SV-HUC non-transformed cells (right) with tunicamycin (140 nM Tun) or vehicle (0 nM Tun) treatment as shown by reduced thiol quantification. D Percentage of reduced and oxidized PDIA1 in TCCSUP EVs as estimated by reduced thiol quantification. E EV release kinetics for SV-HUC and TCCSUP cells following tunicamycin (140 nM Tun) or vehicle (0 nM Tun) treatment for 36 h. For (C) and (E), two-way ANOVAs were performed with Fisher’s LSD multiple comparison test; for (C), n = 3, and for (E), n = 9; error bars indicate means ± standard deviation. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. EV extracellular vesicle, PDIA1 protein disulfide isomerase A1, CD9 cluster of differentiation 9, TSG101 tumor susceptibility gene 101, ANOVA analysis of variance, DMSO dimethyl sulfoxide, Tun tunicamycin.

Full size image

PDIA1-enriched BC-derived EVs induce ROS, DNA damage, and colony formation in recipient urothelial cells

We have previously demonstrated that TCCSUP cell-derived EVs can promote the malignant transformation of SV-HUC cells together with the induction of ROS and DNA damage within these recipient cells [33]. We thus sought to determine whether BC-derived EV-borne PDIA1 plays a role in this context. To that end, we isolated EVs from TCCSUPs transfected with shPDIA1 or scramble control constructs, confirming that PDIA1 levels within EVs from cells in which PDIA1 had been knocked down were markedly reduced (Fig. 3A). A 2’, 7’-dichlorodihydrofluorescein diacetate (DCFDA) flow cytometry approach revealed that treatment with these shPDIA1 EVs for 24 h induced significantly lower levels of ROS within non-transformed SV-HUC cells as compared to scramble control EV treatment (Fig. 3B). In line with this result, immunofluorescent γH2AX staining revealed a significant reduction in DNA damage levels following shPDIA1 EV treatment relative to scramble control EV treatment (Fig. 3C).

Fig. 3: PDIA1-enriched BC-derived EVs induce ROS, DNA damage, and colony formation in recipient urothelial cells.
figure 3

A Western blotting analysis of PDIA1 abundance in EVs derived from scramble control and PDIA1-targeting lentiviral shRNA-transduced TCCSUP cells. GAPDH was selected as a normalization control for cell lysates, while CD63 was used as an EV marker and normalization control for EV proteins. B ROS levels in SV-HUC cells following a 24-hour treatment with scramble control, shPDIA1 TCCSUP EVs, or SV-HUC control EVs, as analyzed by DCFDA and flow cytometry assays. Data represent the fold change in the DCFDA histogram geometric mean relative to the PBS control. C Representative images and quantification of DNA double-strand break in SV-HUC cells following an 18-hour treatment with scramble control or shPDIA1 TCCSUP EVs, as assessed by γH2AX immunofluorescence staining. Scale bars: 10 µm. D Representative images and quantification of the number of colonies in soft agar formed by SV-HUC cells following a 13-week treatment with scramble control or shPDIA1 TCCSUP EVs and 5 weeks of regular culture for stabilization. For (B), (C), and (D), one-way ANOVAs were performed with Tukey’s multiple comparison test; n = 4 or 5; error bars indicate means ± standard deviation. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. EV extracellular vesicle, PDIA1 protein disulfide isomerase A1, shPDIA1 short hairpin RNA targeting PDIA1, PERK Protein Kinase RNA-Like ER Kinase, DCFDA 2’,7’-dichlorofluorescein diacetate, PBS phosphate-buffered saline, ANOVA analysis of variance, γH2AX γ histone 2a family member X, WCL whole cell lysate, CD63 cluster of differentiation 63.

Full size image

To directly establish the impact of BC EV-derived PDIA1 on urothelial cell transformation, SV-HUC cells were continuously treated with shPDIA1 or scramble control EVs for 13 weeks. Following a 5-week recovery period, a soft agar colony formation assay was used to assess the tumorigenic potential of these cells, revealing that while scramble EV treatment significantly increased the colony formation rate consistent with malignant transformation, shPDIA1 EV treatment significantly reduced this colony formation rate to vehicle control levels (Fig. 3D).

To ensure result specificity, we used an extrusion approach described previously to restore recombinant PDIA1 (rPDIA1) to the prepared shPDIA1 cancer EVs [37]. Control EVs were instead prepared by extruding scramble and shPDIA1 EVs with PBS, with Western blotting confirming the successful restoration of PDIA1 to shPDIA1 EVs at levels comparable to those in the parental EVs (Fig. 4A) (Supplementary Fig. S2). These EVs were then used to treat non-transformed SV-HUC cells as above, revealing that rPDIA1 extrusion restored the ability of shPDIA1 EVs to induce ROS production and to promote NFE2L2 expression in recipient SV-HUC cells (Fig. 4B, C). PDIA1 restoration similarly increased the levels of DNA damage induced by shPDIA1 EV treatment (Fig. 4D). Importantly, rPDIA1 extrusion restored the ability of shPDIA1 EVs to promote the malignant transformation of SV-HUC cells over a 13-week treatment period (Fig. 4E). To test whether PDIA1 is required as an oncoprotein to maintain tumorigenicity within the resultant transformed SV-HUC cells, it was knocked down in these cells via shRNA, which had no impact on anchorage-independent growth of these cells (Fig. 4F). Together, these data support the central role for PDIA1 as a direct or indirect driver of the EV-induced malignant transformation of recipient urothelial cells following its release from tumor cells, but that PDIA1 is not required to maintain malignancy once it has been established.

Fig. 4: Restoration of PDIA1 to shPDIA1 EVs rescues their ability to increase cellular stress and anchorage-independent growth.
figure 4

A Western blot showing PDIA1 levels in TCCSUP EVs following extrusion with recombinant PDIA1. CD9 was used as an EV marker and normalization control for EV proteins. B ROS level in SV-HUC cells following treatment with EVs derived from scramble TCCSUP control cells, or from shPDIA1 TCCSUP EVs extruded with rPDIA1 or PBS control, as analyzed via DCFDA and flow cytometry assays. Data represent the fold change in the DCFDA histogram geometric mean relative to PBS control. C NFE2L2 gene expression in SV-HUC cells treated with EVs derived from scramble TCCSUP control cells, or from shPDIA1 TCCSUP EVs extruded with rPDIA1 or PBS control. D Representative images and quantification of DNA damage in SV-HUC cells following an 18-hour treatment with EVs derived from scramble TCCSUP control cells or from shPDIA1 TCCSUP EVs extruded with rPDIA1 or PBS control, as assessed by γH2AX immunofluorescence staining. Scale bars: 10 µm. E Representative images and quantification of the number of colonies (top right panel), percent colony area normalized to that in PBS-treated control (bottom left panel), and the product of percent colony area and corresponding intensitites normalized to that in PBS-treated control (bottom right panel) in soft agar formed by SV-HUC cells following a 13-week treatment with EVs derived from scramble TCCSUP control cells, or from shPDIA1 TCCSUP EVs extruded with rPDIA1 or PBS control. The number of colonies was obtained with the Colony Counter plugin of ImageJ whereas the percent colony area and their intensities were derived using the ColonyArea plugin of ImageJ. F Representative images and quantification of the colonies in soft agar formed by transformed SV-HUC established by long-term cancer EV stimulation (parental), and by transformed SV-HUC transduced with scramble control (scramble) and PDIA1-targeting lentiviral shRNA (shPDIA1). For (B–E), one-way ANOVAs were performed with Fisher’s LSD multiple comparison test; n = 4, 3, 5, and 7, respectively; error bars indicate means ± standard deviation. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. EV extracellular vesicle, PDIA1 protein disulfide isomerase A1, shPDIA1 short hairpin RNA targeting PDIA1, rPDIA1 recombinant PDIA1, NFE2L2 NF-E2-Related Factor 2, DCFDA 2’,7’-dichlorofluorescein diacetate, PBS phosphate-buffered saline, ANOVA analysis of variance, γH2AX γ histone 2a family member X, CD9 cluster of differnetiation 9.

Full size image

PDIA1-enriched EV-transformed cells exhibit patterns of ROS-induced DNA damage

Given our observation that PDIA1 is essential for the cancer EV-induced malignant transformation of SV-HUC cells and for the induction of ROS and DNA damage in these cells, we then examined somatic mutation patterns in these transformed cells. Overall, the transformed cells exhibited a higher mutational burden with a somatic mutation prevalence of 8.2 per megabase, a 10.8% increase over parental cells (Supplementary Fig. S3A). Moreover, the transformed SV-HUC cells harbored more unique nonsynonymous mutations and more genes impacted by these unique nonsynonymous variants as compared to parental non-transformed SV-HUC cells (Supplementary Fig. S3B, Supplementary Table 1). Analyses of single base substitutions (SBS) revealed similar patterns in both cell lines (Supplementary Fig. S3C), suggesting that the unique variants found in the transformed SV-HUC cells are likely the result of cancer EV treatment and all six SBS classes were increased in the transformed cells, with the greatest relative increase being seen in C to A transversions (Supplementary Fig. S3D), the type of substitution most characteristic of ROS-induced DNA alterations [38]. Notably, these changes were stable, as we found that they persisted when the transformed cells were passaged >10 times.

Numerous somatic mutational signatures and variant classes have been associated with different types of cancer and are understood to be the result of distinct mutational processes. We next performed strict signature refitting of the genomic variant data from the parental and transformed SV-HUC cells and identified eight previously defined SBS signatures found in the Catalog of Somatic Mutations in Cancer (COSMIC) [39]. The signatures most prominently enhanced in the transformed cells relative to the parental cells were SBS18, SBS85, and SBS40 (cosine similarity > 0.51) (Supplementary Fig. S3E, Fig. S3F). Notably, COSMIC signature SBS18 has been proposed to result from ROS-induced DNA damage [40]. Taken together, the mutational patterns seen in these transformed cells are consistent with oxidative DNA damage in urothelial cells following cancer EV exposure, suggesting that tumor-derived PDIA1 plays a central role in this process.

Interestingly, COSMIC signature SBS85 (Supplementary Fig. S3G), reported to be caused by AID/APOBEC activity, which is one of the major sources of mutations in BC [41], is characterized by a concentration of variants in the T > A and T > C classes. This finding suggests that, besides elevating ROS levels in recipient cells, cancer EVs may also cause DNA mutations through other mechanisms.

High levels of PDIA1 expression in non-muscle invasive bladder cancer patient tumor tissue predicts a higher risk of recurrence

Our in vitro studies indicated that BC cells release PDIA1 packaged in EVs when subjected to high levels of ER stress, and the uptake of these PDIA1-enriched EVs by non-transformed cells was sufficient to drive their malignant transformation. As any recurrent tumors that arise through this mechanism in vivo may be diagnosed as recurrent tumors, these data suggest that PDIA1 may offer value as a prognostic biomarker to predict BC recurrence. While a majority of MIBC patients will ultimately undergo radical cystectomy, NMIBC is characterized by frequent recurrences requiring surveillance cystoscopies as there is no current reliable marker currently available to monitor for recurrence and/or progression. To explore the utility of PDIA1 in this context, we constructed a tissue microarray (TMA) consisting of 121 NMIBC patients who did (n = 55) or did not (n = 67) develop local recurrence with a minimum follow-up of 5 years (Supplementary Table 2).

We analyzed the expression of BiP, a molecular chaperone with multifaceted roles in the regulation of ER homeostasis and activity [42], in tissue samples comprising this TMA as an approach to estimating ER stress levels in these tumor cells, in addition to assessing PDIA1 expression. Significantly higher levels of BiP and PDIA1 were detected in tumor tissues from patients who experienced BC recurrence (Fig. 5A, B). We also quantified the PDIA1 positive area and intensity within BiP negative regions to assess levels of non-ER residential PDIA1 that are likely to be secreted to the extracellular environment. Interestingly, the tumors of patients with recurrent disease exhibited higher levels of both PDIA1 positive area and PDIA1 accumulation in the non-ER cytosolic regions (Fig. 5C). To further investigate the prognostic significance of these findings, we evaluated patient outcomes and found that higher levels of both BiP and PDIA1 expression were associated with significantly reduced recurrence-free survival in patients (Fig. 5D). Overall survival analyses were excluded from this study as TMA samples were from low-grade bladder cancer cases, which typically have a 5-year overall survival rate of 95%. These data suggest that tissues under higher levels of ER stress and PDIA1 expression are at increased risk of recurrence, indicating that the PDIA1 status in tumors offers as a valuable prognostic marker in patients with NMIBC.

Fig. 5: High levels of PDIA1 expression in non-muscle invasive bladder cancer patient tumor tissue predicts a higher risk of recurrence.
figure 5

A Intensity of BiP (left) and PDIA1 (right) immunofluorescence staining. Each dot represents the average values of a patient. B Representative images of tumor tissue derived from patients with non-recurrent (left) and recurrent (right) NMIBC. PDIA1, BiP, and cell nuclei are respectively stained in green, red, and blue. Scale bars: 40 µm. C Area (left) and fluorescence signal intensity (right) of PDIA1 in BiP negative regions. D Recurrence-free survival rates for patients with corresponding PDIA1 (left) or BiP (right) expression. For (A) and (C), two-tailed Mann-Whitney U tests was performed; n = 42 (non-recurrent) or 63 (recurrent). For (D), patients were categorized into BiP- and PDIA1-high and -low categories based on the corresponding median values. Since there were censored observations, log-rank test was performed. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. BiP binding immunoglobulin protein, PDIA1 protein disulfide isomerase A1, NMIBC non-muscle-invasive bladder cancer, ANOVA analysis of variance.

Full size image

Discussion

The high rates of BC recurrence underscore the need to devise new approaches for identifying patients who are more likely to develop recurrent disease and mitigating this risk. Our data suggest that BC cells utilize EVs as a means of ensuring their ongoing survival under conditions of ER stress through the export of oxidized PDIA1. The incidental uptake of these PDIA1-enriched EVs by normal urothelial cells ultimately leads to their malignant transformation through a potential field cancerization mechanism (Fig. 6). While these results offer insight into the mechanisms that may underlie BC recurrence and suggest that levels of PDIA1 and ER stress in tumor tissues are valuable biomarkers for predicting NMIBC recurrence risk, there are several important topics related to our findings that warrant further discussion.

Fig. 6: A proposed model of the role of PDIA1 and PDIA1-enriched EVs in bladder cancer cell survival and disease recurrence.
figure 6

Within both and non-transformed cells, PDIA1 transitions between oxidized (PDIA1oxi) and reduced (PDIA1red) states as it catalyzes disulfide bond formation and isomerization for misfolded and unfolded proteins in the ER. This activity is engaged at elevated levels under conditions of ER stress that arise in bladder cancer cells, triggering the generation of large quantities of H2O2 as a byproduct of this process. Tumor cells release large quantities of primarily oxidized PDIA1 in their EVs, mitigating this oxidative stress in a manner conducive to cell survival. These bladder tumor cell-derived EVs, in turn, can deliver PDIA1 to recipient non-transformed urothelial cells, wherein it can generate elevated levels of H2O2 that, over time, triggers DNA damage contributing to a higher risk of malignant transformation that may contribute to the development of recurrent lesions.

Full size image

A range of neoplasia-related conditions including reduced genomic stability, the accumulation of mutations, increased protein production and secretion, and exposure to hypoxic or nutrient-deprived microenvironmental conditions can compromise proteostasis within tumor cells [15]. ER stress response and UPR induction can thus preserve tumor cell viability under these challenging conditions [10, 11, 16], with UPR-induced oxidized PDIA1 activation providing support to help restore appropriate protein folding [18,19,20]. In this study, we found that BC cells exhibited elevated ER stress levels under basal conditions. However, subjecting them to further ER stress triggered apoptotic death that could be mitigated by knocking down PDIA1, thereby alleviating oxidative stress when PDIA1 levels were reduced. This suggests that the maintenance of PDIA1 homeostasis is vital for the survival of BC cells owing to the unique characteristics of this enzyme. However, we do acknowledge the limitation of using tunicamycin as a general ER stress inducer rather than as a specific modulator of glycosylation. This approach makes it challenging to fully differentiate between the ER-related functions of tunicamycin and PDIA1.

Most cellular compartments maintain a reducing environment, and oxidized proteins are usually unstable in the cytosol [43, 44]. To prevent hyperoxidation and maintain ER homeostasis under conditions of UPR induction when levels of PDIA1 activity are elevated, BC cells must limit the levels of oxidized PDIA1 within the ER, balancing proteostasis against oxidative injury. We found that under ER stress, BC cells released high levels of oxidized PDIA1 within EVs, thereby mitigating the ROS production and consequent cell death, ultimately promoting BC cell survival. This finding is highly innovative and aligns well with the observation that, while cells continuously shed EVs at steady state, multivesicular body formation and EV release are enhanced under conditions of ER stress or oxidative stress [45, 46], supporting this process of EV-mediated oxidized PDIA1 export as a stress relief mechanism [34].

The mechanisms that underlie the packaging of oxidized PDIA1 into EVs by BC cells warrant further clarification. While the loading of protein cargos within EVs may be partially stochastic, ubiquitination, palmitoylation, and other post-translational modifications can facilitate the sorting of certain proteins into EVs [47, 48]. It remains to be determined whether these modifications are relevant for PDIA1 loading into EVs and whether the oxidized form of PDIA1 is preferentially packaged into EVs under varying levels of ER stress. Assessing these factors will help to better elucidate the directed nature of this cargo loading process.

Strikingly, we found that PDIA1-enriched EVs derived from BC cells promoted the malignant transformation of normal urothelial cells, potentially through ROS-induced DNA damage mediated by the oxidized PDIA1 present within these vesicles or the induction of UPR activity within recipient cells. Indeed, we have previously demonstrated the ability of BC-derived EVs to induce UPR activity and malignant transformation in recipient urothelial cells [33], and both persistent UPR and oxidative stress exposure can readily drive tumorigenesis [12,13,14], suggesting that the delivery of PDIA1 to the urothelium may predispose this field to the future development of recurrent bladder tumors. In addition to their effects on tumorigenesis and BC recurrence, these PDIA1-enriched EVs may also serve as pivotal regulators of various malignant processes directly within the tumor microenvironment. For instance, intracellular PDIA1 plays a vital role in the synthesis of type I collagen, a core component of the extracellular matrix in tumors [49, 50], whereas extracellular PDIA1 directly activates integrins through the promotion of thiol-disulfide exchange [51]. Integrin interactions with type I collagen can also facilitate the development of chemoresistance [52], promoting proliferative growth while protecting against apoptotic death. One report has also suggested a potential link between PDIA1 and immune surveillance, with higher PDIA1 levels potentially supporting immune evasion in breast cancer [53]. Inhibition of PDIA1 activity has shown therapeutic potential in reducing breast cancer cell adhesion and migration through the disruption of focal adhesion complex formation and associated phosphorylation events [54]. However, our results do not offer definitive insight into whether the effects of PDIA1-enriched EVs are directly mediated by PDIA1 or are at least partially related to broader differences in the cargo profiles of these EVs given the important role that PDIA1 plays in protein folding, potentially leading to differences in the misfolded protein content in EVs derived from tumor cells with varying levels of PDIA1 expression. Investigating whether EV-derived PDIA1 can alter the composition of the tumor microenvironment, influence chemoresistance, or modulate immune-mediated detection of BC cells in a direct or indirect manner through these mechanisms may represent promising avenues for future research aimed at clarifying the broader role of PDIA1 in bladder carcinogenesis.

We found that both total PDIA1 expression and non-ER-resident PDIA1 levels were higher in those patients who developed recurrent disease. BiP expression was similarly associated with disease recurrence, underscoring the potential value of evaluating ER stress levels, PDIA1 expression, and PDIA1 localization as potential biomarkers of BC recurrence risk. While tumor and urothelial tissue collection are inherently invasive procedures poorly suited for routine monitoring, urine is an EV-rich biofluid that can be readily obtained from patients, providing an opportunity for noninvasive BC-related biomarker testing [55, 56]. Given our finding that BC cells release PDIA1-enriched EVs, this raises the possibility that urine EVs may offer diagnostic or prognostic utility when evaluating BC patients based on the levels of PDIA1 or ER stress-related biomarkers present therein. However, prospective trials will be essential to evaluate this possibility, and the routine implementation of urine EV-based analyses will necessitate overcoming challenges related to the sensitivity, specificity, and standardization of these biomarkers [56,57,58]. Our results also raise the question of whether therapeutic interventions targeting ER stress and/or PDIA1 may afford benefits to BC patients. PDIA1 inhibitors have been evaluated as promising anti-cancer treatments in patients with relapsed ovarian cancer and in various preclinical cancer models [59,60,61], although whether they can protect against field cancerization effects mediated by PDIA1-enriched EVs remains uncertain. Given that we found that lower PDIA1 levels were beneficial to BC cell survival under high levels of ER stress, while ER stress loading has been advanced as an anti-cancer therapy [62], caution and individualized treatment planning would likely be vital for any interventional efforts targeting this regulatory axis. While our results highlight the potential predictive value of PDIA1 as a biomarker of recurrence risk in NMIBC, further prospective clinical research will be essential to clarify whether it reliably serves as a functional mediator of recurrence or an indicator that can be used to gauge patient risk. Additional analyses aimed at clarifying the potential distance between an initial EV-releasing tumor and the site of recurrence, while challenging to implement, also have the potential to provide valuable insights into whether these EVs contribute to a recurrence risk gradient based on the level of local EV exposure. It would also be interesting to explore whether all urothelial cells are equally susceptible to the oncogenic effects of PDIA1-replete EVs, or whether certain clonal populations within the bladder present with greater sensitivity to this mechanism of action, such as cells harboring mutations associated with dysregulated hedgehog pathway signaling linked to bladder tumor development and progression [63, 64].

Together, our results offer new insights into the integral role that EVs play in the maintenance of BC cell homeostasis under conditions of ER stress, while highlighting PDIA1 as an attractive biomarker and therapeutic that may directly contribute to the risk of BC recurrence.

Materials and methods

Cell culture and EV isolation

Cell lines were obtained from the American Type Culture Collection (Manassas, VA, USA) and maintained in a humidified 37 °C incubator under 5% CO2 in media containing FB Essence (3100, Seradigm). SV-HUC cells are an SV40-immortalized human urothelial cell line, while TCCSUP cells are derived from the urinary bladder of a female with grade IV transitional cell carcinoma, and J82 cells are derived from the urinary bladder of a male with transitional cell carcinoma. For EV collection, cells were cultured in medium containing EV-depleted fetal bovine serum (FBS; Thermo Fisher Scientific, Waltham, MA, USA) and penicillin-streptomycin (catalog no. 15140-122, Thermo Fisher Scientific) as described previously [31]. Cell culture supernatants were processed immediately after collection by serial centrifugation at 400 × g for 10 min and 15,500 × g for 30 min to remove cells and debris and then stored at −80 °C. EVs were isolated from thawed samples by ultracentrifugation performed twice at 200,000 × g for 70 min at 4 °C, and the resulting pellets were resuspended in a small volume of DPBS. Aggregates were removed from the samples by another 15,500 × g centrifugation for 5 min. Final total protein concentrations in the samples were measured with a Micro BCA assay (catalog no. 23235, Thermo Fisher Scientific), and samples were stored at −80 °C.

Extrusion

Initially, 250 µL of PBS containing 37.5 µg of shPDIA1 TCCSUP EVs with or without 25 µg of recombinant PDIA1 (catalog no. enz-262, Prospec) were extruded 10 times using an Avanti extruder set with 0.1 µm polycarbonate membrane filters (catalog no. 610023, Avanti, Alabaster, AL, USA). The mixture was then transferred to a 1.5 mL microcentrifuge tube (catalog no. 357448, Beckman Coulter) filled with 900 µL of PBS and subjected to a 2-hour ultracentrifugation step (Optoma MAX-XP ultracentrifuge, Beckman Coulter, Brea, CA, USA) in a TLA110 rotor at 100 K × g at 4 °C. After removing the supernatant, the EV pellet was resuspended in PBS.

Whole genome sequencing

Genomic DNA concentration was assessed with the Qubit Fluorometer (Thermo Fisher Scientific) and quality was assessed using the Agilent Tapestation (Agilent, Santa Clara, CA). The Illumina Nextera Flex kit (Illumina, San Diego, CA) was used for library construction per the manufacturer’s instructions. Briefly, 500 ng of gDNA was tagmented with Bead-Linked Transposome (BLT) beads while simultaneously adding Illumina sequencing primers. Tagmented genomic DNA was purified, followed by reduced-cycle (5-cycle) PCR amplification to add index and adapter sequences for sequencing. Sequence data was generated using Illumina’s NovaSeq 6000 sequencer. Data were analyzed using MuSiCa (Mutational Signatures in Cancer) [65] and the MutationalPatterns R package [66].

Tissue microarray

We retrieved primary tumor specimens from 121 index cases (initial detection) of non-invasive (pTa) low-grade urothelial carcinoma obtained by transurethral resection performed at the University of Rochester Medical Center. These patients included 80 men and 41 women, with a mean age of 69.7 years (range: 46.5–93.1 years) at the time of surgery. All the sections were reviewed for confirmation of original diagnoses, according to the 2004 World Health Organization/International Society of Urological Pathology classification system for urothelial neoplasms [67]. Appropriate approval from the Institutional Review Board was obtained prior to the construction and use of the TMA. Bladder TMAs were constructed from formalin-fixed paraffin-embedded specimens as previously described [68]. Some cases in the initial TMA patient group were not represented in the first sections taken from the TMA paraffin blocks.

Following de-paraffinization of TMA sections, antigen retrieval was performed in heated citrate buffer (Vector H-3300) for 30 min. Primary antibody incubation was conducted overnight at 4 °C using anti-PDIA1 (Cell Signaling 3501, 1:500) and anti-BiP (Santa Cruz sc-166490, 1:100) antibodies. Stained tissues were photographed at 40× magnification using a Leica DM5000 B microscope. Tumor regions in each photographed field were masked manually and confirmed by a genitourinary pathologist. Within each tumor region, total antibody labeling was determined by measuring the mean pixel intensity values with NIH ImageJ/Fiji [69]. To assess the colocalization of two epitopes, the Manders overlap coefficients were determined using the JACoP plugin in ImageJ [70].

Supplementary methods

For further details regarding other assays conducted in this study, see the Supplementary Materials and Methods.

Statistical analysis

All experiments were performed using at least three biological repeats. Utilized statistical tests are indicated in the figure legends. Survival curves were plotted using the Kaplan-Meier method and compared using the log-rank test. Statistical analyses were performed using GraphPad Prism 9.2.0 and the R statistical computing environment, version 4.0.3.