Reciprocal and Autonomous Glucocorticoid and Androgen Receptor Activation in Salivary Duct Carcinoma

Abstract

Purpose

To determine the expression of glucocorticoid receptor (GR) and androgen receptor (AR) in salivary duct carcinoma (SDC) and analyze the role of these proteins in the development and management of this disease entity.

Experimental Design

We performed a phenotypic assessment of GR and AR localization and expression and determined in the association with clinicopathologic factors in 67 primary SDCs. In vitro functional and response analysis of SDC cell lines was also performed.

Results

Of the 67 primary tumors, 12 (18%) overexpressed GR protein, 30 (45%) had constitutive expression, and 25 (37%) had complete loss of expression. Reciprocal GR and AR expression was found in 32 (48%) tumors, concurrent constitutive GR and AR expression in 23 (34%), and simultaneous loss of both receptors and high GR with AR expressions were found in 12 (18%). GR overexpression was significantly associated with worse clinical outcomes. In vitro ligand-independent AR response was observed in both male- and female-derived cell lines. GR antagonist treatment resulted in decreased cell proliferation and survival in GR-overexpressing cells, irrespective of AR activation status. Reciprocal GR and AR knockdown experiments revealed a lack of interaction.

Conclusions

Our study, for the first time, demonstrates differential GR and AR expressions, autonomous GR and AR activation, and ligand-independent AR response in SDC cells. These findings provide critical information on the roles of GR and AR steroid receptors in SDC tumorigenesis and may lead to the development of a biomarker-based strategy to guide targeted steroid receptor therapy trials in patients with these tumors.

Introduction

Salivary duct carcinoma (SDC), a rare and highly aggressive malignancy, presents in elderly male and female patients and runs a lethal clinical course (1–4). Exclusive of other salivary carcinomas, SDC selectively expresses androgen receptor (AR) and its isoforms in the majority of untreated tumors in male and female patients (5–9). Successful androgen deprivation therapy (ADT) in prostate cancer has led to empirically use in some patients with AR-positive SDC, however, current results of ADT clinical trial in SDC has shown the variable and inconsistent response (10–14). The mechanism underlying AR nuclear expression and its presumed activation in adenocarcinoma of non-reproductive organs is unknown. Possible mechanisms of AR signaling activation in SDC include a constitutive mechanism or inter-dependence on the activation of closely related steroid receptors. Recent studies in tumors of reproductive organs and breast and prostate adenocarcinomas have demonstrated inter-dependence and interactions between AR and glucocorticoid receptor (GR) (15–23). Therefore, insights into the association between the expression and activation of AR and the closely related GR are critical to understanding the role of these proteins in SDC tumorigenesis and in the response of this disease to targeted steroid-based or combined therapy. We hypothesize that AR activation in SDC could be induced through interactions with GR.

GR, in contrast to AR, is ubiquitously expressed in most normal tissues via ligand-mediated binding and plays critical roles in cellular and biological pathways in normal tissue homeostasis and in cancer (19–21). The GR gene is mapped to the chromosome 5q31–32 region and generates multiple isoforms through alternative splicing and initiation sites (19–21). GR isoforms have been shown to differentially transduce glucocorticoid signals to target genes in a tissue- and organ-specific context (19–21,24). Glucocorticoid binds GR to induce transcriptional activation of critical biological pathways and has been used to alleviate cancer-related symptoms in patients with solid malignancies (25,26). Recently, however, evidence has shown that glucocorticoid treatment promotes cancer cell survival and blocks chemotherapy-induced cell death (16,27–29). Notably, studies of castration-resistant prostate cancer (CRPC) have demonstrated close interaction between AR and GR through binding to similar DNA response elements to induce common downstream targets including SGK1 and FKBP5 genes and their transcriptional programs (16–18,30–32). Similarly, FOXA1 transcriptional factor has been identified as a common effector of both AR and GR in CRPC (33) and to be associated with tumor progression and resistance to chemotherapy in triple-negative breast cancer, which is an entity with remarkable phenotypic and biological resemblance to SDC (6,16,34–39).

In this study, we determined the expression and clinicopathologic correlation of GR and AR in primary SDCs and evaluated the in vitro functional and therapeutic effect to anti-ligand treatment.

Materials and Methods

Primary SDC tumors

We searched the database of the head and neck tissue bank at The University of Texas MD Anderson Cancer Center to identify all patients with primary SDC for whom archival tissue blocks were available; we identified 67 cases, who were treated primarily at MD Anderson between 1983 and 2011. The majority of the tumors had been included in previous studies (7,40). This study was conducted in accordance with Declaration of Helsinki and approved by the MD Anderson Institutional Review Board. The written informed consent for molecular analyses was provided by all patients. Clinicopathologic information was extracted from patients’ medical records.

Immunohistochemical (IHC) analysis

We performed an IHC analysis using 4 μm-thick unstained TMA sections and an Autostainer Link 48 (Dako), according to the manufacturer’s instructions. In brief, sections were incubated with primary anti-GR antibody (mouse, BD Biosciences, 1:100 dilution) and the secondary antibody was applied. GR expression status in tumors was evaluated when GR staining appeared in stroma or normal salivary gland cells on the same tissue section. GR nuclear staining in normal salivary parenchyma and host elements was considered constitutive (normal) expression. Among GR-positive cells, stronger GR staining intensity in tumor cells compared to in normal salivary or stroma cells was scored as high. Equal expression in tumor and host cells was scored as normal. A complete lack of staining in tumor cells was scored as a loss of GR. AR staining and scoring has been previously reported (6,7).

Western blot analysis

Whole-cell lysate protein was extracted from fresh tumor tissues and cell lines using RIPA buffer that contained freshly added protease (no. 0505648900; Roche Applied Science) and phosphatase (no. 04906837001; Roche Applied Science) inhibitor cocktails. Aliquots of 20 μg of protein were loaded on SDS-PAGE gel and transferred to a nitrocellulose membrane. Immunodetection was performed by Western blot analysis using anti-GR (mouse, BD Biosciences), anti-AR (rabbit [EPR1535(2)], Abcam), anti-cleaved PARP (rabbit, Promega) and anti-beta actin (ACTB, AC-74, mouse, Sigma-Aldrich) antibodies.

SDC and prostate cancer cell lines

SDC cell lines female-derived RET981 (7,41), male-derived MDA-SDC-04 (42), and MAC (43). Both RET981 and SDC-04 cells express full-length AR. The MAC cell line is a salivary mucinous adenocarcinoma cell line that lacks AR expression. RET981 and MAC cells were maintained in RPMI 1640 medium (Thermo Fisher Scientific) with 10% FBS (Thermo Fisher Scientific). SDC-04 cells were maintained in DMEM medium (Thermo Fisher Scientific) with 10% FBS using standard cell culture techniques. Prostate cancer cell lines, 22Rv1 and LNCaP, were purchased from ATCC and maintained in RPMI 1640 medium with 10% FBS.

GR knockdown by small interfering RNA (siRNA)

Cells were transfected with siRNAs designed GR exon 7 (7–2, 5′-GGCUUCAGGUAUCUUAUGAAG-3′), ON-TARGET plus siRNA GR pool (Dharmacon), AR exon 1 (1–5, 5′-GAAGAUACUGCUGAGUAUU-3′), and ON-TARGET plus siRNA AR pool (Dharmacon) using jetPRIME reagent (Polyplus transfection), according to the manufacturer’s instructions. The ON-TARGET plus siRNA control pool (Dharmacon) was used as a control. Cells were collected 72 hours after siRNA transfection. Data are representative of three independent experiments.

AR stable transfection by lentiviral vector

Human full-length AR was cloned into pLenti-CMV-GFP-2A-Puro lentiviral vector (Abm), which co-expresses green fluorescent protein (GFP). Lentivirus packaging was described previously (40). In brief, 293FT cell lines (Invitrogen) were transfected with the AR lentiviral vector (pLV-AR), pMD2.G (#12259; Addgene), and pCMVR8.74 (#22036; Addgene) using jetPRIME reagent, and then lentivirus supernatant was collected 48 hours after transfection. MAC cells were infected with lentivirus supernatant in the presence of 8 μg/mL polybrene. GFP sorting was performed after puromycin selection (2 μg/mL) at the MD Anderson Flow Cytometry and Cellular Imaging Core Facility.

Cell proliferation assay

Cultured tumor cells were seeded at a density of 2500 cells per well on 96-well plates under androgen-depleted conditions using phenol red-free medium supplemented with 10% charcoal-stripped FBS (CSS; Invitrogen). For androgen response experiments, cells were treated with 10% CSS by adding an appropriate concentration of methyltrienolone (R1881), dihydrotestosterone (DHT) or DMSO (control) after 48 hours of seeding (0 day) and then monitored at 1, 2, 4, and 6 days using the MTT (3-[4,5-dimethylthiazol-2-yl]-2,5 diphenyl tetrazolium bromide) assay. For the AR or GR inhibition assays, cells were treated with 10% FBS medium by adding various doses of enzalutamide (Selleck Chemicals) or mifepristone (Selleck Chemicals) after 48 hours of seeding (0 day) and then monitored at 3 days using the MTT assay. The data are representative from three independent experiments. Growth inhibition by 50% (GI50) was calculated using GraphPad Prism 7 software.

Clonogenic survival assay

One thousand cells per well were seeded in 6-well plates for 24 hours and then treated with various doses of mifepristone or 10 μmol/L enzalutamide for 7 days. This was followed by reconstitution of regular culture conditions for an additional 7–10 days until cell colonies could be discerned. Cells were then fixed with 4% paraformaldehyde and stained with 0.05% crystal violet. Three independent experiments were performed, and colonies were counted using OpenCFU software (44).

Statistical analysis

Fisher’s exact test was used to evaluate clinicopathologic findings and GR expression status. Overall survival curves were created on the basis of the results of a Kaplan-Meier analysis using GraphPad Prism 7 software. If the P value was less than 0.05, the result was declared statistically significant.

Results

Patients’ clinicopathologic characteristics are presented in Table 1. The cohort was composed of 48 men and 19 women who ranged in age from 26 to 85 years (median, 62 years). Tumor sizes ranged from 1.0 cm to 10.0 cm, with a median of 3.0 cm. Only four patients were classified as having stage I and II; 40 had III and IV, and 23 patients had unknown cancer staging. Forty patients (60%) experienced recurrence and metastasis, and 25 (37%) had no evidence of progression. Of the 63 patients for whom 5-year follow-up information was available, 46 (73%) had died of their disease and 17 (27%) were still alive.

Table 1:

Incidence and clinicopathologic correlation of GR expression in salivary ductal carcinomas

Parameter	GR expressiona, no. (%)
	High	C/L	P value
Age, years
> 50	11 (16)	48 (72)
≤ 50	1 (1)	7 (11)	1.00
Sex
Male	10 (15)	38 (57)
Female	2 (3)	17 (25)	0.49
Tumor size, cm
≥ 4	6 (9)	13 (19)
< 4	5 (7)	27 (40)	0.29
AR (IHC)
Positive	5 (7)	41 (61)
Negative	7 (11)	14 (21)	0.039
Stage
I or II	0 (0)	4 (6)
III or IV	9 (13)	31 (46)	0.57
PNI
Yes	6 (9)	31 (46)
No	1 (1)	11 (16)	0.67
Recurrence
Yes	10 (15)	30 (45)
No	2 (3)	23 (34)	0.11
Follow-up (5 years)
Dead	12 (18)	34 (51)
Alive	0 (0)	17 (25)	0.026

GR, glucocorticoid receptor; C/L, constitutive or loss; AR, androgen receptor; IHC, immunohistochemical analysis; PNI, perineural invasion.

^aGR status (total, 67 tumors) was determined by IHC, where stronger nuclear staining in tumor cells than in normal salivary or stroma cells was scored as high (12 tumors [18%]). C/L includes loss of GR expression (25 tumors [37%]) and constitutive status (30 tumors [45%]), in which GR expression was detected at the same or lower level in tumor cells than in normal salivary or stroma cells.

Complete data were only available for age, sex, and AR expression.

P value was calculated by Fisher’s exact test.

GR and AR protein expression in SDCs

Fig. 1A shows GR and AR expression in SDCs. Nuclear GR expression was observed in normal salivary parenchyma and host stroma cells and was considered to represent constitutive expression of GR. In tumors, high GR expression was found in 12 of 67 tumors (18%), constitutive in 30 (45%), and complete loss in 25 (37%) (Fig. 1A and 1B). Concurrent AR activation and high GR expression were found in five tumors (7%), and loss of both receptors’ expression was found in seven (11%). Reciprocal expression of AR and GR was found in 32 tumors (48%). We found a statistically significant inverse relationship between AR activation and high GR expression (P = 0.039, Table 1, Fisher’s exact test).

Figure 1. GR expression and its association with poor survival in SDCs.

(A) Glucocorticoid receptor (GR) and androgen receptor (AR) protein expression by immunohistochemical (IHC) analysis. Case SDC-09 demonstrated high GR (nuclear staining intensity in tumor cells is higher than in stroma cells) and were AR negative. Case SDC-17 cells demonstrated normal GR expression levels, which were the same as those in stroma cells, and were AR positive. Case SDC-66 showed both loss of GR (stroma is positive GR) and AR in tumor cells. (B) GR and AR expression profile in SDCs. (C) Western blot analysis of GR and AR protein expression in selected SDCs. High GR proteins by IHC analysis corresponded to stronger GR expression in SDC tissue than in normal salivary gland tissues. Beta-actin (ACTB) was used as the loading control. Status of AR and GR by IHC analysis, shown at bottom; positive (P) and negative (N). Status of GR by IHC analysis; high (H), constitutive (C), and lost (L). (D) Kaplan-Meier analysis revealed a significant correlation between high GR expression and a low overall survival rate (upper panel, P = 0.046, log-rank test). Of note, there was a trend towards a higher overall survival rate in patients with constitutive GR expression and loss of GR expression (lower panel).

Figure 1C illustrates GR and AR protein expression, as determined by Western blot analysis, in tumor and normal tissue samples, along with the matched corresponding IHC analysis results. All AR-negative cases showed GR expression except one tumor with GR loss. Most AR-positive tumors had variable GR protein bands that corresponded to constitutive expression or loss of GR expression, according to an IHC analysis.

Differential GR expression and AR activation and clinicopathologic correlation

The association between GR and AR expression and clinicopathological factors is shown in Table 1 and Fig. 1D. High GR expression was significantly associated with low 5-year survival (P = 0.026, Table 1, Fisher’s exact test) and overall survival rates (Fig. 1D, upper, P = 0.046, log-rank test). Although statistically not significant, a trend of association between tumors with loss or constitutive expression of GR and better overall survival was also noted (Fig. 1D, lower). No association was found between GR loss, clinicopathologic parameters, and overall survival (Supplementary Fig. 1A, P = 0.67, log-rank test).

AR ligand-independent cell growth of SDC cells

We previously reported that AR knockdown led to inhibition of cell proliferation and ligand-independent cell growth under androgen-depleted conditions in a female-derived AR-positive SDC cell line (RET981) (7). In this study, we extended this analysis to a male-derived AR-positive SDC cell line (SDC-04) (42) and found similar results with an AR ligand agonist. Figure 2A and 2B represent cell proliferation of SDC and prostate cancer cell lines in response to AR agonist R1881. No effect on cell growth was found at low and high concentrations of R1881 (0.1–10 nmol/L) in SDC-04 cells, in contrast to the effects found in control prostate 22Rv1 cells. Treatment with DHT also resulted in no cell proliferation changes in both RET981 and SDC-04 cells (Supplementary Fig. S2A). In addition, AR ligand-independent nuclear localization was uniformly detected in SDC-04 cells in charcoal-stripped FBS medium (Supplementary Fig. S2B).

Figure 2. No effect of enzalutamide on SDC cell proliferation.

(A) Synthetic androgen agonist R1881 treatment did not affect the proliferation of male-derived cells (SDC-04). MTT assay was used to determine the cell growth curve under AR-depleted (CSS) or CSS with R1881 (1 nmol/L) conditions. MAC and 22Rv1 cell lines were used as AR-negative and -positive controls, respectively. (B) RET981 and SDC-04 cells did not show a dose-dependent response of cell proliferation after exposure to R1881 (0.1–10 nmol/L, 72 hours). (C) Treatment with AR siRNA (72 hours) inhibited the proliferation of SDC-04 cells, similar to its effects in 22Rv1 cells. (D) RET981, SDC-04, and LNCaP cells were treated with variable doses of enzalutamide (Enz) for 72 hours. No noticeable differences were found in SDC cells. (E) Combined androgen agonist (R1881, 1 nmol/L) and antagonist (Enz, 0.1–10 μmol/L) did not affect RET981 or SDC-04 cell proliferation. Error bars in all graphs represent standard deviation (SD) from three independent assays. Asterisks (*) represent a significant difference in proliferation activity compared with the control (P < 0.01).

To further confirm AR ligand independence in SDC-04 cells, we evaluated known AR-targeted genes in prostate tumors, namely serum/glucocorticoid-regulated kinase 1 (SGK1) and FKBP prolyl isomerase 5 (FKBP5). In contrast to prostate cell lines, we found no changes in SGK1 gene expression in SDC cells and elevated FKBP5 expression in only RET981 cells in response to R1881 treatment (Supplementary Fig. S2C). These findings support tumor type differences between prostate tumors and SDC. Figure 2C shows decreased cell proliferation in SDC-04 cells in response to AR siRNA treatment, as in RET981 (7) and 22Rv1 cells, but not in MAC cells (AR negative). In contrast to in prostate tumors (45), SGK1 expression in RET981 cells increased after AR siRNA treatment (Supplementary Fig. S2D), suggesting that neither of the common AR-targeted prostate tumor genes plays a role in SDC tumorigenesis.

To understand the AR biological differences between SDC and prostate tumors, we performed a differential expression analysis of RET981 cells treated with AR siRNA or control siRNA (Supplementary Fig. S3A and Supplementary Table S1). Among AR SDC-targeted genes, both BCL2 and PLHPP1 were consistently downregulated; these results were validated by AR knockdown experimentation (Supplementary Fig. S3A). Interestingly, the expression levels of both genes remained unchanged or slightly decreased after R1881 treatment in both SDC cell lines, in contrast to in prostate tumor cells 22Rv1 and LNCaP (Supplementary Fig. S3B). These data provide further evidence of AR ligand-independent activation of SDC cell growth.

Effect of enzalutamide treatment on SDC cell growth

AR positive SDC cell lines were tested for sensitivity of the enzalutamide treatment and the results showed no effect on cell proliferation or cell death in male- (SDC-04) and female-derived (RET981) cell lines (Fig. 2D). We also noted that neither RET981 nor SDC-04 cells demonstrated a response to the combination of androgen agonist (R1881, 1 nmol/L) and antagonist (enzalutamide, 0.1–10 μmol/L) in charcoal-stripped serum medium (Fig. 2E), in contrast to LNCaP cells. Together, these findings are consistent with ligand-independent activation and AR response in SDCs.

GR and AR interaction

We evaluated GR and AR expression in SDC cells by Western blot analysis. Only RET981 cells had high GR protein expression (Fig. 3A). A GR knockdown analysis using siRNAs (targeted exon 7 and pools) demonstrated a marked reduction in cell proliferation in all cell lines (RET981, SDC-04, and MAC; Fig. 3B) and no effect on the AR expression level (Fig. 3C). Similarly, the GR expression level remained unchanged after AR knockdown (Fig. 3D) and the exogenous induction of AR (Fig. 3E). Our data are at variance with those studies of prostate carcinoma cell lines and support tumor-context specificity (18).

Figure 3. GR inhibition did not affect AR expression.

(A) Protein GR and AR expression in SDC cells. All three cell lines expressed GR. Of note, RET981 cells expressed relatively high GR. Beta-actin (ACTB) was used as a loading control. (B) GR siRNAs inhibited cell proliferation in all SDC cells after 72 hours. Error bars in all graphs represent standard deviation. Asterisks (*) represent significant differences in proliferation activity compared with the control (P < 0.01). (C) Knockdown GR by siRNA (72 hours) did not affect AR expression in RET981, SDC-04, and MAC cells. (D) Inhibition of AR by siRNA (72 hours) had no effect on GR expression. (E) Overexpression of AR did not affect GR expression in stable AR-transfected MAC cells. AR and GR Western blot bands were quantified and referred to ACTB; the ratio under those bands indicates the fold change between the control (100) and the subsequent treatment.

We also identified variabilities in downstream target genes between SDC cells: RET981 cells showed downregulation of PHLPP1, while MAC cells showed downregulation of BCL2. There were no changes in these targets in SDC-04 cells after GR knockdown (Supplementary Fig. S4).

GR blockade suppresses SDC cell growth and survival

We evaluated the effect of the GR inhibitor mifepristone on SDC cells. MTT assay demonstrated (Fig. 4A) robust growth inhibition in RET981 cells, which had high GR expression (GI50: 6.98 μM), but not in SDC-04 and MAC cells, which had low GR expression. A clonogenic assay (Fig. 4B) revealed dose-dependent colony inhibition by mifepristone treatment in RET981 cells and in 22Rv1 cells, which also had high GR expression. In contrast, MAC and SDC-04 cells showed no response to mifepristone treatment (Fig. 4B and Supplementary Fig. S5). In addition, the cleaved PARP level was elevated in RET981 and 22Rv1 cells after mifepristone treatment but not in SDC-04 and MAC cells (Fig. 4C), consistent with cell proliferation and clonogenic findings. Similarly, decreased GR protein expression was found in response to different concentrations of mifepristone in RET981 and 22Rv1 cells, but not in MAC cells. Interestingly, SDC-04 cells demonstrated dose-independent GR suppression by mifepristone, and differences in response to mifepristone were found between RET981 and SDC-04 cells. In contrast, no change in AR protein level in response to mifepristone was found in RET981 and SDC-04 cells. Interestingly, prostate 22Rv1 cells demonstrated increased AR expression after mifepristone treatment. Enzalutamide treatment had no effect on cell growth (Fig. 2D and E), colony formation (Fig. 4B), or apoptosis (Fig. 4C) in AR-positive male (SDC-04) and female (RET981) cells, confirming ligand-independent activation of AR in both sexes.

Figure 4. Mifepristone inhibits SDC cell proliferation and colony formation through induction of apoptosis.

(A) The mifepristone concentration needed for growth inhibition of 50% (GI50) for 72 hours was determined using an MTT assay. RET981 and 22Rv1 cells were treated with 6.98 μM and 13.3 μM of the GI50 dose, respectively. SDC-04 and MAC cells showed no response (GI50 was not specified). (B) A clonogenic assay confirmed the effectiveness of mifepristone treatment in RET981 cells. Of note, SDC-04 cells remained viable but did not form colonies (see Supplementary Figure S5). Error bars in all graphs indicate standard deviation (SD). (C) RET981 and 22Rv1 cells demonstrated increased cleaved PARP, with decreasing GR expression after 3 days of mifepristone treatment as a dose-dependent response. AR and GR Western blot bands were quantified and referred to beta-actin (ACTB); the ratio under these bands indicates the fold change between the control and the subsequent treatment. ACTB was used as the loading control.

Discussion

Our findings demonstrate complex AR and GR expression patterns in primary tumors and a lack of direct in vitro interactions between both receptors in SDCs. In primary SDCs, except for a small subset of concurrent expression and loss, the majority of tumors showed reciprocal nuclear localization of GR or AR, suggesting the dependence on steroid receptor activation in this entity. These results are at variance with prostate cancer studies in which GR activation was reported to be induced in androgen-refractory prostate carcinoma and led to resistance to anti-AR treatment (16,17,30,46,47). Considering the constitutive lack of AR in normal salivary glands and the selective AR expression in SDC exclusive of all other salivary carcinomas, these findings support tumor context differences. In male- and female-derived SDC cell lines, we found ligand-independent AR nuclear expression and putative activation; we also identified distinct downstream effector genes, including BCL2 and PHLPP1, which is concordant with the results of CRPC studies (48–50). Interestingly, PHLPP1 has been shown to play a role in the regulation of the PI3K pathway (50), which is frequently altered in SDC (40). Together, these findings confirm organ and tumor differences in AR and GR relationship between SDC and reproductive organ-derived adenocarcinomas.

Our findings support the biological and therapeutic significance of assessing GR expression status in patients with SDC, considering the common use of glucocorticoids to alleviate tumor- and therapy-related side effects in these patients. Glucocorticoids binding to GR induced multiple critical biological effects, including anti-proliferative and anti-apoptotic effects and immune suppression, which may contribute to tumor progression (19,51–55). Recent studies of several solid tumors reported evidence of disease progression in these patients (16,27–29). Therefore, a GR expression analysis may serve as a biomarker for stratifying patients for combined chemotherapy and glucocorticoid therapy. Patients with GR loss could be eligible for glucocorticoid treatment, with minimal effect on tumor progression. In that context, patients with GR-expressing tumors would be eligible for anti-GR therapy, especially in view of recent data indicating that GR antagonism sensitizes tumor cells to chemotherapy in ovarian (56,57), prostate (16–18,30), and triple-negative breast carcinoma (16,22,36).

According to the results of this study, differential and variable AR and GR expression may allow for subclassifying patients for hormone-based targeted therapy. In contrast to AR, GR is constitutively expressed in salivary glands, and its complete loss is a tumor-related event. Comparative analyses of our findings and those of other malignancies are difficult to conduct because of the widely variable methods used to evaluate GR expression. Using immunohistochemical analyses, we determined the expression and localization of GR in tumor cells and compared it to the expression of host elements in each case. The underlying cause of GR loss is unknown, but possible factors include epigenetic post-translational modification (58–60), GR haploinsufficiency (61), or enhanced histone H3K27me3 (62), resulting in GR downregulation in these tumors. Notably, a trend towards better survival was noted in patients with constitutive GR expression and loss of GR expression; these findings may guide future glucocorticoid therapy in patients with SDC.

Our in vitro findings support a differential effect of anti-GR treatment on cell proliferation and survival that is dependent on the level of GR expression. Similar findings have been reported in prostate and breast carcinoma cell lines (30,36,46,47). GR expression may, therefore, be useful as a marker to guide the selection of patients who will benefit from anti-GR or combined anti-GR and anti-AR therapy. Patients with low GR expression and AR activation can be targeted for anti-AR therapy and spared corticosteroid treatment. GR antagonism has been shown to sensitize tumors to a chemotherapy-induced response in multiple epithelial malignancies and can be effective in SDC. However, we found no evidence of inter-dependence or cooperative interactions between GR and AR in SDC cell lines. Loss or activation of GR or AR is an independent event in both male- and female-derived cell lines. We extended our previous study of ligand-independent AR activation in female-derived cell lines (7) to male-derived cell lines, and our findings indicate constitutive activation in both sexes. The AR ligand-depletion strategy may not be effective in SDC in both sexes. This contention is empirically supported by the variable and inconsistent responses to ligand-based therapy (12–14). In SDC, therefore, drugs that directly target AR receptors, alone or in combination with chemotherapy or anti-GR, are likely to be effective.

Collectively, our study demonstrates variable GR expression and a reciprocal and independent relationship between GR expression and AR activation. The use of IHC analysis is practical and cost effective for determining GR and AR expression in tumor specimens and thus guiding the therapeutic stratification of patients with SDC. Tumors with high GR alone can be targeted by GR blockade and patients with concurrent high GR and activated AR will be eligible for concurrent blockade (30,36,46,47). Tumors with AR alone will be targeted with AR blockade, and the early effect on GR expression can be assessed by iterant biopsy during therapy. It will be important to identify GR and AR downstream target genes and their stochastic effects on tumor and host elements (17,62–64). The development of tissue-selective GR and AR antagonists might address the potential side effects of their systematic inhibition.

Translational Relevance.

Salivary duct carcinoma (SDC), an uncommon aggressive salivary gland malignancy, uniquely expresses androgen receptor (AR) in untreated tumors in both males and females. The underlying mechanism of AR activation in SDC is currently unknown; it could be constitutive, or it could occur through cross-talk with the closely related glucocorticoid receptor (GR). Our study documents the reciprocal expression of AR and GR in untreated SDCs and a significant association between GR overexpression and poor clinical outcomes. In vitro experiments demonstrated no evidence of direct AR and GR interactions and ligand-independent nuclear localization and response of AR. Our study provides valuable tumor context-specific characteristics of AR and GR expression and response that could guide a rational steroid-receptor therapy strategy for patients with SDC.