Androgen Receptor Signaling Induces Cisplatin Resistance via Down-Regulating GULP1 Expression in Bladder Cancer

The underlying molecular mechanisms of resistance to cisplatin-based systemic chemotherapy in bladder cancer patients remain to be elucidated, while the link between androgen receptor (AR) activity and chemosensitivity in urothelial cancer has been implicated. Our DNA microarray analysis in control vs. AR knockdown bladder cancer lines identified GULP1 as a potential target of AR signaling. We herein determined the relationship between AR activity and GULP1 expression in bladder cancer cells and then assessed the functional role of GULP1 in cisplatin sensitivity. Androgen treatment in AR-positive cells or AR overexpression in AR-negative cells considerably reduced the levels of GULP1 expression. Chromatin immunoprecipitation further showed direct interaction of AR with the promoter region of GULP1. Meanwhile, GULP1 knockdown sublines were significantly more resistant to cisplatin treatment compared with respective controls. GULP1 knockdown also resulted in a significant decrease in apoptosis, as well as a significant increase in G2/M phases, when treated with cisplatin. In addition, GULP1 was immunoreactive in 74% of muscle-invasive bladder cancers from patients who had subsequently undergone neoadjuvant chemotherapy, including 53% of responders showing moderate (2+)/strong (3+) expression vs. 23% of non-responders showing 2+/3+ expression (P = 0.044). These findings indicate that GULP1 represents a key downstream effector of AR signaling in enhancing sensitivity to cisplatin treatment.


Introduction
Urinary bladder cancer, mostly urothelial carcinoma, has been one of the most commonly diagnosed malignancies, especially in men [1,2]. Moreover, the number of newly diagnosed bladder cancer cases and related deaths throughout the world has even risen from 429,800 and 165,100 estimated in 2010 [1] to 549,393 and 199,922 reported in 2018 [2], respectively. Clinically, there are two distinct forms of bladder cancer, non-muscle-invasive and muscle-invasive diseases, and the latter is often associated with metastatic disease where the overall 5-year survival rate remains low (i.e., 6.4% [3]). Urothelial carcinoma also occurs in the upper urinary tract, which is often (i.e., 60% [4]) invasive at the time of initial diagnosis.

Associations between AR and GULP1 Expression
We recently employed DNA microarray analysis in control AR-positive UMUC3 versus a UMUC3 subline stably expressing AR-shRNA [16]. Of those expressed at absolutely high levels, several candidate genes were examined if their expression was not only upregulated in AR-knockdown cells but also down-regulated in CDDP-resistant cells. Indeed, a quantitative PCR confirmed the significant increase or decrease in the levels of GULP1 expression in AR-knockdown or CDDP-resistant subline, respectively, compared with control cells. We thus decided to further investigate the functions of GULP1 whose role in neoplastic diseases was largely unknown.
We first examined the expression of GULP1 in four human bladder cancer lines that were known to be AR-negative (i.e., 5637, 647V) or AR-positive (i.e., UMUC3, TCCSUP) [19]. Western blot detected GULP1 signals in all of these cell lines, and the levels were higher in AR-negative cells than in AR-positive cells (Figure 1a). We then compared the levels of GULP1 expression in AR-negative/knockdown versus AR-positive/overexpression sublines. Consistent with the DNA microarray data [16], GULP1 expression was downregulated in cells highly expressing AR (Figure 1b). We further assessed the effects of androgen (i.e., dihydrotestosterone (DHT)) and anti-androgen (i.e., hydroxyflutamide (HF)) on GULP1 expression. DHT treatment (vs. mock treatment) did not significantly change the levels of GULP1 expression in AR-negative 647V or UMUC3-AR-shRNA cells ( Figure 1c). However, in AR-expressing cells, DHT and HF considerably reduced and induced, respectively, GULP1 expression over mock treatment. Thus, the expression of GULP1 was inversely related to the expression/activity of AR in bladder cancer cells.
A bioinformatics-driven search identified three potential AR binding sites in the GULP1 promoter region. We therefore investigated whether AR could regulate the expression of GULP1, using a chromatin immunoprecipitation (ChIP) assay ( Figure 2). DNA fragments from UMUC3 cells immunoprecipitated with an anti-AR antibody or normal immunoglobulin G (IgG) were amplified by PCR with sets of GULP1 promoter-specific primers. The PCR products for all three putative binding sites could be visualized from those precipitated by the AR antibody, but not control precipitations, indicating interactions of AR with the GULP1 promoter. fragments from UMUC3 cells immunoprecipitated with an anti-AR antibody or normal immunoglobulin G (IgG) were amplified by PCR with sets of GULP1 promoter-specific primers. The PCR products for all three putative binding sites could be visualized from those precipitated by the AR antibody, but not control precipitations, indicating interactions of AR with the GULP1 promoter.  Binding of AR to the GULP1 promoter in bladder cancer cells. There are three putative binding sites in the GULP1 promoter region. A ChIP assay using UMUC3 cell lysates immunoprecipitated with an anti-AR antibody or mouse IgG as a negative control. The DNA fragments were PCR amplified with sets of GULP1 promoter-specific primers, and the PCR products (i.e., 338 bp for binding site 1, 173 bp for binding site 2, 131 bp for binding site 3) were electrophoresed on 1% agarose gel. A fraction of the mixture of protein-DNA complex (i.e., 1% of total cross-linked, reserved chromatin prior to immunoprecipitation) was used as "input" DNA.  fragments from UMUC3 cells immunoprecipitated with an anti-AR antibody or normal immunoglobulin G (IgG) were amplified by PCR with sets of GULP1 promoter-specific primers. The PCR products for all three putative binding sites could be visualized from those precipitated by the AR antibody, but not control precipitations, indicating interactions of AR with the GULP1 promoter.  Binding of AR to the GULP1 promoter in bladder cancer cells. There are three putative binding sites in the GULP1 promoter region. A ChIP assay using UMUC3 cell lysates immunoprecipitated with an anti-AR antibody or mouse IgG as a negative control. The DNA fragments were PCR amplified with sets of GULP1 promoter-specific primers, and the PCR products (i.e., 338 bp for binding site 1, 173 bp for binding site 2, 131 bp for binding site 3) were electrophoresed on 1% agarose gel. A fraction of the mixture of protein-DNA complex (i.e., 1% of total cross-linked, reserved chromatin prior to immunoprecipitation) was used as "input" DNA. Binding of AR to the GULP1 promoter in bladder cancer cells. There are three putative binding sites in the GULP1 promoter region. A ChIP assay using UMUC3 cell lysates immunoprecipitated with an anti-AR antibody or mouse IgG as a negative control. The DNA fragments were PCR amplified with sets of GULP1 promoter-specific primers, and the PCR products (i.e., 338 bp for binding site 1, 173 bp for binding site 2, 131 bp for binding site 3) were electrophoresed on 1% agarose gel. A fraction of the mixture of protein-DNA complex (i.e., 1% of total cross-linked, reserved chromatin prior to immunoprecipitation) was used as "input" DNA.

Role of GULP1 in Cell Growth
Using control-shRNA vs. GULP1-shRNA sublines, we assessed the role of GULP1 in the cell proliferation (via MTT assay (Figure 4a)), apoptosis (via TUNEL assay ( Figure  4b)), cell cycle (Figure 4c), cell migration (via wound-healing assay (Figure 4d)), and cell invasion (via transwell invasion assay (Figure 4e)). In these assays without CDDP treatment, there were no significant differences between the two sublines. However, in the of CDDP. Cell viability is presented relative to that of each subline without CDDP treatment. Each value represents the mean (±SD) from a total of 6 determinants. * p < 0.05 (vs. control-shRNA).

Role of GULP1 in Cell Growth
Using control-shRNA vs. GULP1-shRNA sublines, we assessed the role of GULP1 in the cell proliferation (via MTT assay (Figure 4a  Cell migration determined by the rate of cells filling the wound area is presented relative to that of the control subline (original magnification: 40×). Each value represents the mean (+SD) from a total of 10 determinants. (e) Transwell invasion assay in 647V-control-shRNA vs. 647V-GULP1-shRNA sublines. Cell invasion determined by counting the number of invaded cells in the lower chamber under a microscope is presented relative to that of the control subline (original magnification: 40×). Each value represents the mean (+SD) from a total of 10 determinants. * p < 0.001 (vs. control-shRNA).

Expression of GULP1 in Bladder Cancer Specimens
We stained immunohistochemically for GULP1 in two separate sets of bladder tissue microarrays (TMAs). Positive signals were detected predominantly in the cytoplasm of non-neoplastic and neoplastic epithelial cells (Figure 5a). 647V-GULP1-shRNA sublines cultured for 1-4 days. Cell viability is presented relative to that of the control subline at day 1. Each value represents the mean (±SD) from a total of 6 determinants. (b) TUNEL assay in 647V-control-shRNA vs. 647V-GULP1-shRNA sublines in the absence or presence of 5 µM CDDP cultured for 72 h. Apoptosis counted as a percentage of at least 500 cells is presented relative to that of the control subline. Each value represents the median (±SE) from a total of 16 determinants. (c) Cell cycle phase analysis in 647V-control-shRNA vs. 647V-GULP1-shRNA sublines in the absence or presence of 5 µM CDDP cultured for 72 h. Color changes are associated with cells in G1 (yellow), S (light green), G2 (dark green), and M (intense blue) phases (original magnification: 100×). Proportion of G2/M counted as a percentage represents the mean (+SD). (d) Wound-healing assay in 647V-control-shRNA vs. 647V-GULP1-shRNA sublines gently scratched and cultured for 24 h. Cell migration determined by the rate of cells filling the wound area is presented relative to that of the control subline (original magnification: 40×). Each value represents the mean (+SD) from a total of 10 determinants. (e) Transwell invasion assay in 647V-control-shRNA vs. 647V-GULP1-shRNA sublines. Cell invasion determined by counting the number of invaded cells in the lower chamber under a microscope is presented relative to that of the control subline (original magnification: 40×). Each value represents the mean (+SD) from a total of 10 determinants. * p < 0.001 (vs. control-shRNA).

Expression of GULP1 in Bladder Cancer Specimens
We stained immunohistochemically for GULP1 in two separate sets of bladder tissue microarrays (TMAs). Positive signals were detected predominantly in the cytoplasm of non-neoplastic and neoplastic epithelial cells (Figure 5a In the first set of TMA consisting of 129 cases of urothelial neoplasms and corresponding 89 non-neoplastic normal-appearing urothelial tissues, GULP1 was positive in 80 (90%) of non-neoplastic and 96 (74%) of neoplastic specimens (Table 1). Thus, the rate of GULP1 positivity was significantly lower in tumors than in nonneoplastic tissues. In addition, GULP1 expression was marginally or significantly downregulated in high-grade (68%) or muscle-invasive (61%) tumors compared with lower grade (84%) or non-muscle-invasive (83%) tumors, respectively. There was no statistically significant difference in GULP1 positivity between muscle-invasive cases showing pN0 (64%) versus pN+ (46%). We then performed a Kaplan-Meier analysis coupled with a logrank test to assess possible associations of GULP1 expression with patient outcomes. However, there were no significant differences in progression-free survival in patients with non-muscle-invasive disease between GULP1-low and GULP1-high tumors (0 vs.   In the first set of TMA consisting of 129 cases of urothelial neoplasms and corresponding 89 non-neoplastic normal-appearing urothelial tissues, GULP1 was positive in 80 (90%) of non-neoplastic and 96 (74%) of neoplastic specimens (Table 1). Thus, the rate of GULP1 positivity was significantly lower in tumors than in non-neoplastic tissues. In addition, GULP1 expression was marginally or significantly down-regulated in high-grade (68%) or muscle-invasive (61%) tumors compared with lower grade (84%) or non-muscle-invasive (83%) tumors, respectively. There was no statistically significant difference in GULP1 positivity between muscle-invasive cases showing pN0 (64%) versus pN+ (46%). We then performed a Kaplan-Meier analysis coupled with a log-rank test to assess possible associations of GULP1 expression with patient outcomes. However, there were no significant differences in progression-free survival in patients with non-muscleinvasive disease between GULP1-low and GULP1-high tumors (0 vs. 1+/2+/3+: P = 0.396 ( Figure 5b); 0/1+ vs. 2+/3+: P = 0.285), as well as in those with muscle-invasive disease between GULP1-low and GULP1-high tumors (0 vs. 1+/2+/3+: P = 0.744 ( Figure 5c); 0/1+ vs. 2+/3+: P = 0.193).

Discussion
Although resistance to CDDP-based chemotherapy is not uncommonly seen in patients with urothelial cancer, its underlying mechanisms are not fully understood. Meanwhile, AR activation in bladder cancer cells has been implicated to be associated with chemoresistance [12][13][14][15][21][22][23][24]. In the present study, we further investigated the role of GULP1, as a downstream target of AR, in CDDP resistance, using bladder cancer cell lines as well as surgical specimens.
Caenorhabditis elegans CED-6 and its human homologue GULP1, as adapter proteins, play an important role in phagocytosis [18,25]. Specifically, CED-6/GULP1 was shown to interact with phagocytic receptor CED-1 and other transmembrane receptor proteins, such as stabilin-1, and thereby engulf apoptotic cells. GULP1 was also shown to induce the rearrangement of the actin cytoskeleton via MAPK activation [26]. By contrast, its precise function in cancer progression remains largely unknown. Nonetheless, in ovarian cancer cells, GULP1 showed inhibitory effects on their proliferation while inducing phospho-SMAD3 [27] or reducing phosphorylation of AKT/PDK1 and MAPK [28]. In bladder cancer specimens, loss of immunoreactivity for GULP1 was more frequently seen in muscleinvasive tumors (85.8%) than in non-muscle-invasive tumors (39.0%) [29]. The reduced expression of GULP1 is also associated with poorer patient outcomes in The Cancer Genome Atlas dataset [30]. In bladder cancer lines, overexpression or silencing of GULP1, as a tumor suppressor, resulted in reduction or induction, respectively, of cell proliferation, migration, and invasion [29]. Moreover, GULP1 was suggested to involve CDDP resistance by showing an increase in cell viability and a decrease in apoptosis in a GULP1-silenced bladder cancer T24 line cultured in the presence of 20 µM CDDP, as well as a decrease in GULP1 expression in a CDDP-resistant T24 subline [29]. In addition, the levels of GULP1 mRNA expression were found to be low (or undetectable) in all nine bladder tumors from patients who did not respond to CDDP therapy but were relatively high in two of six responders [29]. It is also worth mentioning that phagocytosis has been implicated in chemoresistance. For example, CDDP treatment in lung cancer cells was shown to induce CD47 expression, leading to a reduction in the phagocytic activity of co-cultured macrophages, while the CD47 blockade enhanced the cytotoxic effect of CDDP [31].
Here, we confirmed some of the previous findings [29] in two other bladder cancer lines and transurethral resection specimens. Specifically, GULP1 knockdown via stable expression of its shRNA resulted in a significant reduction in the cytotoxicity of CDDP, along with a significant decrease in apoptosis and a significant increase in G2/M population, both of which were altered by CDDP. Immunohistochemistry in surgical specimens from patients who had subsequently undergone CDDP-based neoadjuvant chemotherapy further showed significantly higher levels of GULP1 expression in those from responders (n = 17) compared with non-responders (n = 26). We additionally demonstrated that the levels of GULP expression were significantly or marginally lower in urothelial neoplasms (vs. non-neoplastic urothelial tissues) and high-grade or muscle-invasive tumors (vs. lower grade or non-muscle-invasive tumors), while we failed to show a strong association between the expression level of GULP1 and the prognosis of the patients who had never received CDDP (prior to tumor recurrence or disease progression). These results indicate that GULP1 plays a critical role in increasing sensitivity to CDDP. Inconsistent with previous observations, however, we found no considerable impact of GULP1 on the proliferation/migration/invasion or apoptosis in 647V cells cultured in the absence of CDDP.
As mentioned above, it remains unclear how AR signals modulate chemosensitivity. Based on our DNA microarray analysis data [16], we expected that GULP1 was a downstream target of AR. We herein demonstrated an inverse relationship between AR activity and GULP1 expression. Specifically, AR overexpression/androgen treatment and AR knockdown/anti-androgen treatment reduced and induced, respectively, the levels of GULP1 expression in bladder cancer cells. Remarkably, a ChIP assay revealed direct interactions of AR with GULP1 at its promoter region, further indicating direct regulation of GULP1 expression by AR. It is thus likely that GULP1 represents a key downstream effector of AR signaling in modulating CDDP sensitivity in bladder cancer. Meanwhile, several molecules/pathways have been suggested to be downstream targets of GULP1. In particular, it has been documented that GULP1 activates SMAD3 [27] or inactivates AKT/PDK1 and MAPK [28] in ovarian cancer cells, while it inhibits the nuclear translocation of NRF2 and subsequently reduces the expression of HMOX1 in bladder cancer cells [29]. Interestingly, all of these potentially downstream of GULP1 have been linked to CDDP resistance [10,24,[32][33][34][35]. Further studies are required for elucidating the molecular mechanisms responsible for AR/GULP1-mediated chemoresistance.
While AR activity must be considerably higher in men than in women, no studies have demonstrated significant sex-related differences in the prognosis of bladder cancer patients who undergo systemic chemotherapy. This may be due to a similar role of estrogen signaling in modulating chemosensitivity. Indeed, we recently showed an association between estrogen receptor (ER)-β activation and CDDP resistance in bladder cancer [36]. Accordingly, modulation of GULP1 expression by estrogen signaling may also need to be further investigated.
An inverse association between AR and GULP1 expression was seen in bladder cancer cell lines. In surgical specimens, we further demonstrated marginal or significant downregulation of GULP1 expression in high-grade or muscle-invasive tumors. The levels of AR expression should therefore be elevated in high-grade/muscle-invasive bladder cancers compared with low-grade/non-muscle-invasive tumors. However, conflicting data exist regarding the expression of AR mRNA [19,37] and protein (as reviewed in [11,12]) in different grades/stages of bladder tumors. Specifically, in a meta-analysis of immunohistochemical studies, AR expression was shown to be rather significantly down-regulated in high-grade or muscle-invasive tumors [38]. These findings may also raise the possibility of GULP1 modulation by others, such as ER pathways, since significant increases in ERβ expression in high-grade/muscle-invasive bladder cancers have been documented [38][39][40].

ChIP
A ChIP assay was performed, using a Magna ChIP kit (Millipore Sigma) according to the manufacturer's recommended protocol with minor modifications. UMUC3 cells were cross-linked with 1% formaldehyde for 10 min at room temperature. The cell lysates were sonicated in nuclear buffer (four 30-s pulses, output 3.0, duty cycle 30% in ice with 120 s rest between pulses; Branson Sonifier 450). Soluble chromatin was immunoprecipitated with an anti-AR antibody and normal mouse IgG (sc-2025, Santa Cruz Biochemistry) directly conjugated with Magnetic Protein A beads. Immuno-precipitated DNA was eluted and reverse cross-linked, and DNA was extracted and purified using a spin filter column. DNA samples were analyzed by PCR. We performed a bioinformatic search (LASAGNA-Search 2.0. Available online at https://biogrid-lasagna.engr.uconn.edu/lasagna_search/ [43]; accessed on 24 November 2020) for potential AR binding sites in the GULP1 promoter and found three target sites (see Figure 2). The sequences of the primers are as follows: Site 1 forward, CTGCGCCTATCACAACTCTATT; Site 1 reverse, GAAAGGGAGCAA-GAAGGAGTATC; Site 2 forward, GGTCAGTAAGAATGGGCTGTT; Site 2 reverse, GGC-CTTAATCTCTGGACTTTGT; Site 3 forward, CCAGAGATTAAGGCCGAGTTAAA; Site 3 reverse, AAACGTGCCGTCTTCACA. The PCR products electrophoresed on 1% agarose gel and stained with ethidium bromide were visualized using Gel Doc XR+ (Bio-Rad).

Apoptosis and Cell Cycle Analysis
The TUNEL assay was conducted on cell-burdening coverslips, using the DeadEnd Fluorometric TUNEL system (Promega, Madison, WI, USA), followed by counterstaining for DNA with 4 ,6-diamidino-2-phenylindole. The apoptotic index was determined in the cells visualized by the fluorescence microscopy.
For cell cycle phase quantification, the Cell Cycle Assay Cell-Clock™ (Biocolor, Carrickfergus, UK) was used according to the manufacturer's recommended protocol. Live cells exhibit color changes associated with cell cycle phases. Data were analyzed using ImageJ version 1.53 (National Institutes of Health, Bethesda, MD, USA).

Cell Migration
A scratch wound-healing assay was adapted to evaluate the ability of cell migration. Cells at a density of ≥90% confluence in 12-well tissue culture plates were scratched manually with a sterile 200 µL plastic pipette tip. The wounded monolayers of the cells were allowed to heal in serum-free medium for 24 h, and the width of the wound area was monitored with an inverted microscope. The normalized cell-free area in photographed pictures (24 h/0 h) was quantitated using ImageJ.

Cell Invasion
Cell invasiveness was determined using a Matrigel-coated transwell chamber (6.5 mm diameter polycarbonate filter with 8 µm pore size, Corning Inc., Corning, NY, USA). Cells (1 × 10 5 ) in 100 µL serum-free medium were added to the upper chamber, whereas 600 µL medium containing 10% FBS was added to the lower chamber. After incubation for 24 h, invaded cells were fixed, stained with 0.1% crystal violet, and counted.

Immunohistochemistry
Two sets of TMA consisting of retrieved bladder tissue specimens obtained by transurethral surgery performed at the Johns Hopkins Hospital were previously constructed [13,39]. The first set consisted of 129 cases of urothelial neoplasm with various tumor grades/stages, along with normal-appearing urothelial tissues from the same patients. In these patients, CDDP was not used prior to tumor recurrence or disease progression. The second set consisted of 43 cases of high-grade muscle-invasive urothelial carcinomas that had subsequently received CDDP-based neoadjuvant chemotherapy prior to radical cystectomy, including 17 responders and 26 non-responders, as defined previously [13,44]. None of the patients included in these sets of TMA had received therapy with radiation or anti-cancer drugs prior to the collection of the tissues.

Statistical Analysis
Student's t-test was used to compare numerical data. Fisher's exact test or chi-square test was used to evaluate the associations between categorized variables. Survival rates in patients were calculated by the Kaplan-Meier method and comparison was made by log-rank test. All statistical analyses were performed using Prism version 5 (GraphPad, San Diego, CA, USA) and EZR software [45], a graphical user interface for R version 4.0.2 (The R Foundation for Statistical Computing, Vienna, Austria). P values less than 0.05 were considered to be statistically significant.

Conclusions
We identified GULP1 as a key downstream effector of AR in modulating CDDP sensitivity in bladder cancer. While GULP1 activators are not currently available, our data further support the notion that concurrent anti-androgen therapy has the potential of being a means of chemosensitization, especially in male patients with AR-positive bladder tumor. In addition, the expression status of GULP1, along with that of AR, may serve as a predictor of chemosensitivity in patients with bladder cancer.

Informed Consent Statement:
The ethical approval included a waiver of patient consent.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author but are not publicly available due to privacy and/or ethical restrictions.

Conflicts of Interest:
The authors declare no conflict of interest.