Stimulating Soluble Guanylyl Cyclase with the Clinical Agonist Riociguat Restrains the Development and Progression of Castration-Resistant Prostate Cancer

Abstract

Castration-resistant prostate cancer (CRPC) is incurable and fatal, making prostate cancer the second-leading cancer-related cause of death for American men. CRPC results from therapeutic resistance to standard-of-care androgen deprivation (AD) treatments, through incompletely understood molecular mechanisms, and lacks durable therapeutic options. Here, we identified enhanced soluble guanylyl cyclase (sGC) signaling as a mechanism that restrains CRPC initiation and growth. Patients with aggressive, fatal CRPC exhibited significantly lower serum levels of the sGC catalytic product cyclic GMP (cGMP) compared to their castration-sensitive stage. In emergent castration-resistant cells isolated from castration-sensitive prostate cancer (CSPC) populations, the obligate sGC heterodimer was repressed via methylation of its beta subunit. Genetically abrogating sGC complex formation in CSPC cells promoted evasion of AD-induced senescence and concomitant castration-resistant tumor growth. In established castration-resistant cells, the sGC complex was present but in a reversibly oxidized and inactive state. Subjecting CRPC cells to AD regenerated the functional complex, and co-treatment with riociguat, an FDA-approved sGC agonist, evoked redox stress-induced apoptosis. Riociguat decreased castration-resistant tumor growth and increased apoptotic markers, with elevated cGMP levels correlating significantly with lower tumor burden. Riociguat treatment reorganized tumor vasculature and eliminated hypoxic tumor niches, decreasing CD44⁺ tumor progenitor cells and increasing the radiosensitivity of castration-resistant tumors. Thus, this study showed that enhancing sGC activity can inhibit CRPC emergence and progression through tumor cell-intrinsic and extrinsic effects. Riociguat can be repurposed to overcome CRPC, with noninvasive monitoring of cGMP levels as a marker for on-target efficacy.

Introduction

Prostate cancer (PC) is the second leading cause of cancer-related death among American men. Most PC patients present with castration-sensitive disease (CSPC) for whom initial treatment with androgen deprivation therapy (ADT) and agents targeting the androgen receptor (enzalutamide, apalutamide, darolutamide) or steroidogenesis inhibitors (abiraterone acetate) is very effective. However, upon progression to castration resistance (CRPC), PC is invariably fatal with a median overall survival of 15–36 months (1). To eradicate CRPC, new therapies are critically needed and an incomplete understanding of its molecular drivers is a barrier to therapeutic innovation.

To uncover novel targets underlying CRPC, we leveraged a prior finding from our lab and others that AD induces senescence in CSPC cells as a CRPC-inhibitory mechanism (2–6), and our unique discovery that evasion of AD-induced senescence (ADIS) promotes outgrowth of castration-resistant variants (4,7). Through pilot hypothesis development studies leveraging ADIS evasion as a key step in the progression to CRPC, we identified several novel targets. We previously reported that one such target, thioredoxin-1 (TRX1), is elevated in CRPC as a redox-protective response to AD-induced reactive oxygen species (ROS), and that its inhibition re-sensitizes CRPC cells to AD (7). Significantly, we also uncovered new targets that were decreased in the CRPC variants. These candidates comprised novel pathways likely to enforce the ADIS barrier to CRPC emergence and thus are unique in the therapeutic space for PC.

These pilot studies in ADIS-resistant variants led us to investigate soluble guanylyl cyclase (sGC) signaling, a well-studied vasodilatory pathway, as a novel CRPC-restraining mechanism. When activated by freely diffused nitric oxide (NO), the sGC complex catalyzes the conversion of 5’-guanosine triphosphate (5’-GTP) to its second messenger molecule, cyclic guanosine monophosphate (cGMP) (8) (Fig. 1A). Through cGMP, sGC exerts anti-inflammatory, anti-proliferative and vasodilatory functions via downstream effectors, cGMP-dependent kinases, phosphodiesterases, and cyclic nucleotide-gated ion channels (9). Significantly, stimulating sGC activity via on-target agonists is an FDA-approved intervention for pulmonary hypertension, with minimal side-effects (9,10). Thus, given the excellent long-term safety profile of sGC agonists, there is imminent therapeutic relevance to preclinically testing sGC agonists for potential repurposing in CRPC, a fatal and therapeutically under-served disease. Moreover, the role of sGC signaling in modulating tumor biology is not well-understood in PC or indeed in other cancers. Therefore, in this study, we investigated the mechanisms by which sGC complex activity is altered with PC progression, and complementarily whether an FDA-approved sGC agonist, riociguat (aka Adempas, BAY 63–2521) (11), could limit CRPC growth. Our results show that sGC activity enforces the ADIS barrier to castration resistance and is dampened through transcriptional repression as well as through reversible protein oxidation during different stages of castration resistance. Importantly, our findings indicate riociguat limits castration-resistant tumor growth through apoptosis, and that its on-target treatment efficacy can be noninvasively monitored through systemic cGMP levels.

Figure 1. Castration resistance is marked by decreased sGC expression and activity.

(A) The α1β1 heterodimer is the predominant form of the soluble guanylyl cyclase (sGC) complex (created with Biorender.com). (B-E) The mRNA levels of sGC subunits GUCY1A1 (sGCα1) and GUCY1B1 (sGCβ1) in LNCaP SB0 vs. SB5 cells (B, C) and LAPC4 SB0 vs. SB1 (D, E). AD induced for 5 days. Loading controls were ActinB (LNCaP), and GAPDH (LAPC4). (F) Comparison of GUCY1A1 and GUCY1B1 expression (normalized to their respective GAPDH values) from TCGA vs. SU2C patient datasets (cBioportal). (G) Comparison of GUCY1A1 and GUCY1B1 expression from the Grasso dataset (GSE35988). (H) Trends of pre- and post-castration-resistant patient sera cGMP levels. (I) Sera cGMP levels at CSPC vs. CRPC stage for each patient in (H). (J) Percentage change in sera cGMP levels (from H) vs. patient outcomes. Mann-Whitney test.

Materials and Methods

Patient-derived serum

De-identified matched sera from CSPC/CRPC patients were received from the NCI biospecimen repository (NCT00034216). We confirm that these samples were obtained through IRB-approved studies from patients who provided written informed consent, and that the studies were conducted in accordance with institutional ethics guidelines.

Cell lines

LNCaP (CRL-1740, denoted as LNCaP SB0 in this study, RRID:CVCL_0395) and 22Rv1 (CRL-2505, RRID:CVCL_1045) were obtained from ATCC. LAPC4 (RRID:CVCL_4744) cells were a kind gift of Dr. John Isaacs, Johns Hopkins University (Baltimore, MD). None of the cells used in this study are listed in the ICLAC database of commonly misidentified lines. All cell line stocks used in this study, including our in-house LNCaP CRPC derivatives, LNCaP SB5 variants and LNAI, are validated yearly by short term-repeat (STR) profiling (Genetica Corp), last validated in 2023. Additionally, all cells used in xenograft animal experiments were certified pathogen-free prior to use. LNCaP cells engineered for disruption of functional sGC activity (sGC knockout pools) were generated by Synthego Corporation. The sgRNA for sGCβ1 (ENST00000264424; sequence: UGCCACACUGAGUGACCACU) targets exon 6. All cell culture reagents described in this section were obtained from Gibco, Life Technologies, except for sera, FBS and charcoal stripped serum (CSS), which were obtained from Hyclone. Cell lines were cultured in RPMI-1640 medium (Gibco, 11875093). All media were supplemented with either 5% (LNCaP, LNCaP SB5, LNAI, LAPC4) or 10% (22Rv1) fetal bovine serum (FBS; Hyclone, SH30396) for cell line growth and maintenance. To produce androgen-deprived conditions, cells were cultured in their appropriate base media supplemented with 5% or 10% charcoal-stripped serum (CSS; Hyclone, SH30068). To deplete androgen, cultures were washed three times with CSS-supplemented media with 5-minute incubations between washes at 37°C in a humidified incubator. All media were supplemented with 100 U ml⁻¹ penicillin/streptomycin (Gibco, 15140122). Cells were cultured at 37 °C in a humidified incubator at 21% oxygen/5% CO₂ or 5% O₂/5% CO₂ (HeraCell Tri-Gas, Thermo Fisher Inc).

Quantitative PCR (qPCR) analyses

The mRNA from cultured cells was extracted using the RNeasy (Qiagen, 74136) or RNAqueous-4PCR kit (Life Technologies, AM1914) as previously described (7). The following gene-specific TaqMan primer/probe sets were used: GUCY1A1 (Hs01015574_m1), GUCY1B1 (Hs00168336_m1), AR (Hs00171172-m1), ACTB (internal normalization control; Hs99999903_m1) and GAPDH (internal normalization control; Hs99999905_m1) (ThermoFisher Scientific).

DNA constructs and viral transduction

The TRX1 overexpression construct was generated as previously described (7). Retroviral supernatant production was carried out in HEK 293 T cells (ATCC), and infection of target cells was performed as described previously (7). Transduced cells were selected in 250 μg ml⁻¹ G418 (Life Technologies, 11811031)-containing media, corresponding at a minimum to the time taken for untransduced cells to die completely in selection media. Protein knockdown or overexpression was verified via western blotting.

Measurement of cGMP levels

Steady state cGMP concentrations in cells, mouse plasma, tissues and human serum were measured by cGMP Enzyme Immunoassay (Arbor Assays, K065-H). Lysates were prepared according to the manufacturer’s instructions with minor modifications. Briefly, lung and tumor samples were mechanically pulverized following flash-freezing in liquid nitrogen. For lung samples, every 50 mg of powdered tissue was diluted in 500 μl of provided sample diluent, For tumor samples, 250 μl of sample diluent was added for every 50 mg of powdered tissue. For measurement of cGMP levels in cell lines, cultures were treated for 30 minutes at 37 °C with 20 μM drug (riociguat, cinaciguat) or an equivalent DMSO dose. For cell and tissue samples, the BCA assay was used to measure protein lysate concentration, and cGMP levels were expressed as picomoles of cGMP per milligram of total protein. Measurements in sera and plasma samples were carried out as follows. Mouse plasma was diluted 1:5 in the provided sample diluent solution. Human sera from PC patients were assayed undiluted using the Cyclic GMP Direct Assay, as cGMP is lower in sera than plasma. An acetylation format was used to enable detection of low concentrations of cGMP. The cGMP signal was read and calculated at 450 nm on a Spectramax iD3 or iDx Microplate Reader. Plasma cGMP levels were obtained by multiplying concentration readings by their dilution factors and expressed in nM concentration.

Western blotting

Western blotting was performed as previously described (7). Protein concentrations were measured using the Pierce BCA Protein Assay kit (ThermoFisher, 23225). Blots were probed with antibodies against the following proteins: sGC (1:2,000, ThermoFisher MA5–17086, RRID:AB_2538557), sGC (1:2,000, Cayman Chemicals,160897, RRID:AB_10080042), AR (1:8,000, Santa Cruz Biotech, sc-816, RRID:AB_1563391), GAPDH (1:10,000, Abcam, ab9485, RRID:AB_307275), cleaved-PARP (1:1,000, Cell Signaling, 9541, RRID:AB_331426), p16^INK4a (1:1,000, Abcam, ab108349, RRID:AB_10858268), TRX1 (1:1,000, BD Biosciences, 559969, RRID:AB_398684), phospho-VASP(Ser239) (1:1,000, Cell Signaling, 3114S, RRID:AB_2213396), VASP (1:1,000, Santa Cruz, sc-46668, RRID:AB_2213431), β-actin (1:5,000, Abcam, ab8226, RRID:AB_306371), actin (1:2,500, Sigma, A2066), Bcl2 (1:2,000, Proteintech, 12789–1-AP, RRID:AB_2227948), survivin (1:1,000, Cell Signaling, 2808, RRID:AB_2063948). Following incubation with the appropriate secondary horseradish peroxidase-conjugated antibodies, blots were developed using SuperSignal West Femto Maximum Sensitivity Substrate (ThermoFisher, 34095). Two molecular markers were used in this study: Cytiva Rainbow Molecular Weight Markers (Sigma, RPN800E) and Spectra Multicolor Broad Range Protein Ladder (ThermoFisher, 26634) depending on the size(s) of the protein(s) being targeted for detection. Densitometry of images was carried out via the ImageJ Analyze Gels (NIH) module (RRID:SCR_003070) and normalized to the loading signal for each band.

Drug treatments and dose response assays

For all cell lines, 750 cells per well were plated in triplicate in 96-well plates in FBS-supplemented culture medium. Twenty-four hours after plating, cultures were treated with either riociguat (Cayman Chemicals, 9000554) or cinaciguat (Cayman Chemicals, 17468) at various concentrations. For androgen blockade, LNAI cultures were pre-treated for 24 hours in 10 μM enzalutamide (Selleckchem, S1250), followed by treatment with riociguat alone or combined with 10 μM enzalutamide. Drug in fresh media was replenished daily for 72 hours, after which viability was measured using the luminescence-based the CellTiter-Glo^® Assay (Promega, G7571) on a SpectraMax iD3 or iDx Microplate Reader (Molecular Devices, LLC). Data were normalized to luminescence values from vehicle-treated controls within each group and plotted as %viability (relative to vehicle). Curve fits were modeled using four-parameter logistic (4PL) regression to derive best fit value, IC₅₀. Individual areas under the dose-response curves (AUCs) were calculated and compared by group. All tests comparing aAUC utilized the Wilcoxon test for two groups or the Kruskal-Wallis test for more than two groups. To determine whether ferrostatin could mitigate the effects of riociguat, LNAI cells were plated in FBS media, and twenty-four hours after plating, cells were changed into media supplemented with 5% CSS. The cells were subsequently treated with indicated concentrations of riociguat and 2 μM ferrostatin-1 (Sigma, SML0583) for 48 hours.

5-Aza-2’-deoxycytidine and RG108 treatment

LNCaP SB0 or SB5 lines were each seeded into 10 cm cell culture plates, in either FBS- or CSS-supplemented media. Twenty-four hours after plating, cultures were treated with either 5 μM or 10 μM 5-aza-2’-deoxycytidine (MedChemExpress, HY-A0004) or 100 μM RG108 (Selleckchem, S2821) or equivalent DMSO as the vehicle control. Cultures were subsequently collected by scraping on ice for Western blotting qPCR or immunoblotting, four days following initiation of 5-aza treatment or 7 days following initiation of RG108 treatment.

Senescence-associated beta-galactosidase assay (SA-beta-gal)

SA-beta-gal staining was carried out as previously described (4). To quantify positive staining, 100 cells were counted for each sample over multiple fields of view, excluding fields at the very edge.

ROS Measurements

LNAI cells were seeded into 96-well black-walled plates at 2,000 cells/well in FBS culture medium and incubated at 37°C. Twenty-four hours after plating, cells were changed to their corresponding culture medium supplemented with 5% FBS or 5% CSS and treated with either 100 μM or 200 μM riociguat (Cayman chemical, 9000554) or DMSO (Sigma, D8418) as the vehicle control. At 20 h following riociguat treatment, cells were washed in ice-cold Ca² and Mg²⁺ free 1X Hank’s balanced salt solution (HBSS, Thermo Fisher, 14175–095), and incubated with freshly prepared 10 μM dihydroethidium (hydroethidine, HEt) (ThermoFisher, D11347), 5- (and −6)-chloromethyl- 2′,7′-dichlorofluorescein diacetate (CM-H₂DCFDA, ThermoFisher, C6827), or hydroxyphenyl fluorescein (HPF, ThermoFisher, H36004) for 30 min at 37°C. The cells were then washed once in 1X HBSS, and 150 μl 1X HBSS was added to each well. The plate was then read on the SpectraMax iD3 reader (Molecular Device) for detection of fluorescent signal. Unstained cells were used as negative controls, and these baseline values were subtracted from signal in the samples with added fluorophores. Data analysis and graphing were performed in GraphPad Prism (v.10).

Mitochondrial bioenergetics profiling

LNAI were seeded in FBS media at a density of 14000 cells per well in 96-well Seahorse cell culture plates (Agilent, 103777–100) and incubated at 37°C. Twenty-four hours after plating, cells were changed to their corresponding culture medium (supplemented with either 5% FBS or 5% CSS for AD) and treated with either 100 μM or 200 μM riociguat (Cayman chemical, 9000554) or equivalent DMSO control (Sigma, D8418). At 20 hours following riociguat treatment, the media were replaced by Seahorse XF assay media, and oxygen consumption rates were measured by using Mito Stress test assay (Agilent, 103015–100) on XF analyzer (Seahorse Bioscience) according to the manufacturer’s instructions. Oxygen consumption values were normalized to the OD 562 nm values for each well using Pierce BCA Protein Assay kit (ThermoFisher, 23225). ATP production was calculated by subtracting the value for proton leak from the value for basal respiration. Spare respiratory capacity was calculated by subtracting the value of basal respiration from the value for maximal respiration.

Xenograft tumor experiments

All animal studies were performed in accordance with the University of Miami Institutional Animal Care and Use Committee (IACUC)-approved protocol. Numbers of animals to be used for the tumor formation experiments were determined through power analysis to provide 90% statistical power to detect a mean difference of 2.6 between two groups, assuming a two-sample Student’s t-test and a standard deviation for both groups of 1.5 at two-sided 5% significance level.

As previously described (7), for all xenograft experiments, cells were resuspended in a 1:1 matrigel (BD Biosciences, 356237): full FBS/RPMI-1640 media mixture and injected subcutaneously using a 26-gauge needle into one flank of immunocompromised 5–6-week-old castrated male mice (Nu/Nu, Envigo). Tumor length, width, and height were measured using electronic precision calipers (VWR, 90028). Tumor measurements and animal weights were monitored three times a week in a non-blinded manner. Tumor volumes were calculated according to the following formula: 0.52 × (height × width × length). Tumor-bearing animals were euthanized through AALAC-approved procedures when tumors in any group exceeded 10% of animal body weight (~1000 mm³). Immediately following euthanasia, blood was collected through cardiac puncture from experimental animals and processed for sera and plasma. Tumors were excised, cut sagitally where possible, photographed and sectioned into samples for formalin fixation or flash-frozen in liquid nitrogen. Lungs were also collected and flash-frozen for systemic measurement of cGMP levels.

To assess whether loss of the sGC heterodimer promoted castration resistance, we subcutaneously injected 2 × 10⁶ WT or KO sGCβ1 LNCaP cells into the flanks of 6-week-old, male, castrated Nu/Nu mice and allowed tumors to reach approximately 1000 mm³.

To assess whether riociguat could mitigate in vivo CRPC tumor growth, we subcutaneously injected 2 × 10⁶ LNAI cells. Animals were randomized into a vehicle or treatment groups once tumors were palpable (100–150 mm³) and treatment was initiated with either DMSO:Tween80:PBS (10%:10%:80%) or 20 mg kg⁻¹ riociguat (Cayman Chemicals, 9000554) or cinaciguat (Cayman Chemicals, 17468) diluted in Tween80 and Mg²⁺ and Ca²⁺-free DPBS. Injections were given intraperitoneally (IP) seven days a week starting at palpable tumor formation.

For combinatorial ionizing radiation (IR)/riociguat studies, we subcutaneously injected 2 × 10⁶ LNAI cells on the flanks of 6-week-old male, castrated Nu/Nu mice and allowed tumors to reach approximately 100 mm³ prior to randomization into vehicle and riociguat treated groups. Mice were then treated with riociguat (20 mg/kg/day, i.p.) or vehicle (DMSO) daily for 72 hours, following which animals from each treatment cohort were randomized into a group receiving either a one-time dose of 6 Gy ionizing radiation (IR) or 0 Gy radiation (mock IR). Non-tumor tissues were shielded with lead covers and tumors were irradiated using the RadSource RS2000 as the radiation source. Mice continued to receive vehicle or riociguat as per initial cohort assignment for the remaining duration of the experiment.

Immunohistological staining

For histopathological analysis, fine sections (4 μm) were cut from formalin-fixed, paraffin-wax-embedded samples, and stained with hematoxylin and eosin. Immunohistochemical analyses were performed utilizing ready-to-use Ki67-specific antibody solution (K2, Leica Biosystems, PA0230). Sections were dewaxed, rehydrated, and pretreated in a high pH (pH 9) solution at 100 °C. Slides were dehydrated using a standard protocol and mounted using Permount. Ki67 images were taken using a Leica microscope and associated LAS V4.9 software at 40X objective magnification.

For evaluation of tumor vasculature, thin-tissue FFPE slides were dewaxed and rehydrated as per standard protocols. Antigen retrieval was performed using low pH (pH 6) citrate solution at 15 PSI for 15 minutes. Following appropriate blocking PBS/10% goat serum, slides were incubated with primary antibody CD31 (Cell Signaling, 77699, RRID:AB_2722705) in PBS/1% BSA, followed by polymer treatment, 3% peroxide blocking, DAB chromogen (Vector, SK-4800), hematoxylin treatment (Leica, 3801), dehydrated and mounted using Permount. CD31-stained slides were imaged using the Olympus Whole Slide Scanner VS200 and processed using OlyVIA 2.9.1 software.

Immunofluorescent staining

For determination of hypoxia, pimonadizole (Hypoxyprobe, HP10-100) was diluted in saline and administered to study animals intraperitoneally at a dose of 60 mg/kg two hours prior to euthanasia. Tissues (tumor, lung, and liver) were then collected and fixed in formalin and embedded in paraffin. Paraffin-embedded fine sections were deparaffinized by heating at 56°C for two hours and then hydrated by treating with Xylene (15 minutes, 2 times), 100% ethanol, 90% ethanol, and 70% ethanol (2 times) at 5 minutes each. The slides were steamed for 35 minutes with a pH 6 Dako Target Retrieval (Agilent Technologies, S169984-2) and permeabilized with 0.1% Triton X-100 (Sigma-Aldrich, T8787) for 5 minutes, followed by a 5-minute wash with PBS. Slides were then blocked in serum-free, protein blocker (Agilent Technology, X090930-2). Slides were incubated with a biotinylated anti-pimonidazole mouse IgG1 monoclonal antibody diluted 1:50 in Background Sniper (Fisher Scientific, 5082385) for one hour at room temperature. This step was followed by washes with PBS. Streptavidin, AlexaFluor 568-conjugated secondary antibody (Life Technologies, S11226) was applied for 30 minutes at a 1:200 dilution in the dark at room temperature followed by PBS washes. Slides were then mounted using ProLong Gold anti-fade with DAPI (Molecular Probes, P36935) for immunofluorescence staining.

For CD44 immunofluorescence, the same procedure was followed except the primary antibody CD44 (Abcam, ab157107) was incubated on the slides at a 1:200 dilution in Background Sniper overnight at 4°C. Following this step, the secondary antibody, goat anti-rabbit IgG Superclonal Alexa Fluor 488 (Life Technologies, A27034, RRID: AB_2847859) was diluted 1:200 in Background Sniper and applied for 30 minutes at room temperature.

Apoptosis was assessed by terminal deoxynucleotidyl transferase-mediated dUTP nick end-labeling (TUNEL), using the In Situ Cell Death Detection assay, Fluorescein (Roche, 11 684 795 910). Slides were dewaxed, rehydrated and blocked as described and incubated with the TUNEL reaction mixture for 1 h at 37 °C in a humidified environment in the dark. After rinsing with PBS, the sections were mounted using Prolong Gold Antifade Mountant with DAPI. Images were acquired using a Leica fluorescence microscope keeping identical exposures across the samples, per channel. Quantitation of images was carried out via the Image J module, by measuring the percentage signal across the area of the field in each channel of interest.

Cytokine profiling

Sera from a vehicle-treated, riociguat-treated (non-responder; NR), and two riociguat-treated (responder; R) male mice bearing LNAI subcutaneous tumors were assayed for systemic (murine) cytokines using the Proteome Profiler Mouse XL Cytokine array (R&D Systems, ARY028). The indicated cytokines in Fig. 8 fit the criteria of being the only ones where the non-responder and vehicle-treated values are similar and where both responder values trend in the same direction. Each cytokine value represents the average of n=2 samples (each measured in duplicate). Results are represented as a fold-change relative to the values established from vehicle-treated sera. Membranes were exposed on film using the provided chemiluminescent reagent. Pixel densities of duplicate signals for the 111 cytokines were measured using the ImageJ Analyze Blots module.

Figure 8. Riociguat induces tumor oxygenation and enhances anti-tumor response to radiation.

(A) Representative immunofluorescent Hypoxyprobe images with endpoint tumor volumes noted (50 μm scale bar). Quantitation of stained areas from a minimum of 6 tumors/treatment group, 3–7 images per tumor (right). (B) Representative CD31 IHC staining from the same tumors as (A). Arrows indicate canalised blood vessels (100 μm scale bar). (C) Representative immunofluorescent images of CD44 (green) and DAPI (blue) co-staining (25 μm scale bar). (D) Cytokine profiling. (E) Tumor formation kinetics per treatment group: vehicle/mock IR (n=3), riociguat/mock IR (n=4), vehicle/6Gy IR (n=10), riociguat/6 Gy IR (n=11). Timing of riociguat dosing and one-time 6 Gy irradiation are noted. (F) Endpoint tumor volumes (mm³) from all treatment groups from (E), each point representing an individual tumor. (G) Intratumoral cGMP levels from all tumor groups in (E). (H) Immunoblotting for cl-PARP in xenograft tumor lysates from vehicle/IR or riociguat/IR-treated mice. (I) H&E and co-localized Hypoxyprobe/DAPI immunofluorescent staining from representative tumors across all treatment groups in (E), 50 μM scale bar shown. Quantitation of staining (below) derives from at least 3 tumors per mock treatment group (3–5 fields/tumor), and at least 10 tumors per IR group (3 fields/tumor). (J) Overview of our findings describing inhibition of sGC signaling in progression to castration resistance and the opportunities for therapeutic targeting (created with Biorender.com).

Statistical analyses

Unless otherwise specified, data are presented as ± standard error of the mean (SEM). In the Figure legends, unless otherwise noted, all two-sample mean comparisons were analyzed for significance via two-tailed unpaired Student’s t-test with Welch’s test for unequal variances, or a Mann-Whitney test. For three or more comparisons, data were analyzed using a one-way analysis of variance (ANOVA) with Tukey’s test to multiple paired comparisons of means with the groups. The extent of correlation between two variables were determined by calculating the Spearman coefficient for monotonicity or Pearson coefficient for linearity. All tests for comparing area under the curve (aAUC) used the Wilcoxon test for two groups, or the Kruskal-Wallis test for more than two groups. Dose response curves were estimated using 4-parameter log-logistic function and the IC₅₀ was obtained from the estimated parameters. Results with p-values < 0.05 were considered statistically significant, with * p< 0.05, ** p≤ 0.01, *** p≤ 0.001, **** p≤ 0.0001. Statistical analyses were performed using statistical software package R (v. 4.2.2) or GraphPad Prism (v. 10).

Data availability statement

Single cell RNA sequencing (sc RNA-seq) data were analyzed through the Broad Institute single cell portal (https://singlecell.broadinstitute.org/single_cell). The expressions of GUCY1A1 and GUCY1B1 were visualized in dot plot format as generated on the website. Raw gene expression data from the TCGA (prostate cancer, Firehose Legacy) and SU2C patient datasets were mined from cBioportal (https://www.cbioportal.org/) (12,13). GUCY1A1 and GUCY1B1 levels from TCGA and SU2C were normalized to their respective GAPDH levels, as these two datasets utilized different methods of transcript quantitation (FPKM vs. RSEM). Datasets (GSE35988, GSE82071) from the Gene Expression Omnibus (GEO: https://www.ncbi.nlm.nih.gov/geo/) were obtained and analyzed through the database’s GEO2R processing pipeline. All the statistical analyses were performed using GraphPad Prism v.10 (RRID:SCR_002798). All other raw data are available upon request from the corresponding author

Results

Progression to castration resistance correlates with decreased sGC expression and activity

To investigate acute molecular changes associated with resistance to ADIS, we previously utilized cyclic AD in parental (SB0) CSPC LNCaP and LAPC cells to rapidly select for naturally-occurring ADIS-resistant variants (denoted respectively as LNCaP SB5 and LAPC4 SB1) (4). The LAPC4 SB0 line is already enriched for AD-refractory subpopulations as they continue to grow slowly under AD (4,14), despite a subset of the culture undergoing ADIS (4).The LNCaP SB0 cells are unable to grow under AD or in castrated athymic male mice. However, their ADIS-resistant SB5 counterparts can proliferate under AD conditions (4) and form castration-resistant tumors (7). We initially identified the sGC-cGMP pathway as a possible CRPC-restraining mechanism from a pilot experiment in CRPC models repurposing FDA-approved drugs, including the sGC agonist, riociguat. We subsequently noted, in an Illumina-based screen we had conducted comparing LNCaP SB0 vs. SB5 cells, that levels of sGC subunit genes, GUCY1A1 (sGCα1 protein) and GUCY1B1 (sGCβ1 protein) were significantly decreased under AD in the emergent CRPC SB5 variants but were increased in the androgen-deprived parental CSPC cells (Supplementary Fig. S1A). NOS3, the gene for endothelial nitric oxide synthetase, was also significantly altered in these settings. However, we chose to focus on the sGC component of the NO-sGC-cGMP pathway for various reasons. The significance of NOS3 in PC is controversial, with studies reporting increased NOS/NO as variously inhibiting and driving PC progression (15–17). Moreover, therapeutic use of NO donors has several known downsides, including unpredictable rate of NO release, lack of selectivity, induced vascular resistance and nitrosative stress (18). These issues do not affect sGC agonists which, unlike NO donors, are not a source of nitrosative stress - indeed, the adverse systemic effects of NO are thought to be largely cGMP-independent (19). Moreover, in PC, NO donors likely act through nitrosative inactivation of androgen receptor (AR) (15), a vulnerability that then adaptively suppresses NOS activity in advanced PC. By contrast, sGC agonists act on the sGC complex independently of NO levels and can also sensitize sGC to low NO concentrations to elevate cGMP levels (11). Moreover, overexpression of sGC is reported to decrease growth of some breast cancer lines (20). We likewise found no dependency by any cancer line (including PC) on the sGC complex for survival (Supplementary Fig. S1B), further supporting that sGC activity has a tumor-inhibitory rather than tumor-promoting role in cancer. Finally, the safe long-term use of the sGC agonist, riociguat, for pulmonary hypertension led us to undertake this study.

We independently verified changes in GUCY1A1 and GUCY1B1 via qPCR in CSPC vs. their emergent CRPC counterpart cells under AD (CSS) as well as androgen-replete (FBS) culture. Consistent with Supplementary Fig. S1A, both subunits were elevated in senescent LNCaP SB0 vs. their proliferating counterparts (Fig. 1B), and both sGC subunits were significantly downregulated under AD, in the ADIS-resistant SB5 cells (Fig. 1C). The qPCR results however showed that the magnitude of GUCY1B1 decrease in SB5 CSS cells was significantly underestimated in the Illumina dataset, a known caveat of such screening methodology. Expression of both subunits was also higher in AD-deprived LAPC4 SB0 and lower in the ADIS-resistant SB1 under AD (Figs 1D–E). The relative magnitude of differences was smaller than in LNCaP, likely due to LAPC4 SB0 cultures already containing a relatively high fraction of emergent CRPC clones (4,14) that resemble the SB1 variants. Regardless, these results supported that lower sGC expression coincided with enhanced ability to evade ADIS.

GUCY1B1 (but not GUCY1A1) decreased significantly in the LNCaP SB5 vs. SB0, even under androgen-replete conditions (Supplementary Fig. S1C). A prior report indicated GUCY1A1 is positively regulated by androgens (21). Analyses of two GEO datasets (GSE128749 and GSE172205) (22,23), describing global changes on gene expression induced by the synthetic androgen R1881, supported that GUCY1A1 (but not GUCY1B1) is positively regulated by androgen signaling in LNCaP and the LNCaP-derived C4–2 line (Supplementary Fig. S1D). GUCY1A1 however was not significantly upregulated by androgens in LAPC4. Analysis of TCGA PC patient datasets indicated a moderate positive correlation between AR and GUCY1A1 expression (correlation coefficient ~ 4) and low correlation with GUCY1B1 (correlation coefficient ~ 2.8) (Supplementary Fig. S1E). Furthermore, despite decreased AR (4), both GUCY1A1 and GUCY1B1 levels were elevated in LNCaP SB0 that had undergone ADIS (Fig. 1B). Thus AD-induced changes in sGC subunits are likely governed by mechanisms beyond simply changes in androgen availability or AR signaling.

Comparing the SU2C (metastatic CRPC) vs. TCGA (CSPC) patient datasets (cbioportal.org) (12,13), we found normalized GUCY1A1 and GUCY1B1 expression was lower in the former (Fig. 1F). GUCY1A1 and GUCY1B1 levels were similarly lower in CRPC vs. CSPC in the Grasso dataset (GSE35988) (24), which transcriptionally profiled patient samples for drivers of progression to lethal metastatic CRPC (Fig. 1G). These patient-derived results mimic those from the ADIS-resistant cells, verifying that lower sGC expression marks the CRPC state. PC patient datasets also showed a trend (p=0.07) of poorer survival correlating with lower levels of GUCY1A1 and GUCY1B1 (Supplementary Fig. S1F). Nevertheless, transcriptional levels of sGC subunits are unlikely to fully reflect sGC enzymatic activity in PC.

Therefore, to assess whether changes in sGC activity marked PC progression and outcome, we measured cGMP levels in matched sera taken at the CSPC stage and then again following progression to CRPC from 10 de-identified patients (cohort details, Supplementary Fig. S1G). As the cGMP phosphodiesterase, PDE5A, is expressed at very low levels in PC (25), sGC activity is expected to be the major contributor to systemic cGMP levels. All cGMP measurements were conducted in a blinded manner, and subsequent unblinding indicated an overall trend of lower cGMP correlating with the castration-resistant state (Fig.1H). Furthermore, we found a >50% decline in cGMP levels correlated with rapid disease progression and mortality at relatively young ages following CRPC diagnosis (patients 1, 3, 4, 10; Figs. 1I–J). By contrast, the sole patient (PC-8) with increasing cGMP levels at CRPC had less aggressive disease and longer post-diagnosis survival (Figs. 1I–J, Supplementary Fig. S1G). With the obvious caveat of small sample size, our results support that significantly attenuated sGC activity correlates with CRPC disease progression and fatality. More generally, our collective results here point to low sGC expression and activity as an early marker of CRPC progression.

The sGCβ1 gene is repressed during CRPC progression, preventing functional heterodimer formation

The heterodimeric structure of the sGC complex is essential for its catalytic activity (26). Reduced or altered expression of either subunit leads to impaired downstream signaling, and is implicated in several pathologies, including cardiovascular (27,28) and genitourinary disorders (29). However, whether stoichiometric dysregulation of the sGC complex plays a causal role in cancer progression is not known. Therefore, we assessed sGC expression and activity across an isogenic panel of progressively more castration-resistant lines derived from LNCaP SB0. The variants are denoted as follows, in order of increasing castration resistance (Fig. 2A): SB0 (CSPC), SB5 (emergent CRPC), SB5X (derived from an SB5 castration-resistant tumor), SB5XX (derived from an SB5X castration-resistant tumor). The SB5 variants retained AR expression under AD (Fig. 2A) and showed progressively lower incidence of ADIS compared to their parental SB0 counterparts, as assessed by SA-beta-gal staining and p16^INK4a induction (Supplementary Figs. S2A–B). The SB5X line formed castration-resistant tumors more readily than its originating counterpart SB5 (Supplementary Figs. S2C–E), just as the SB5 cells formed castration-resistant tumors more readily than their parental SB0 line (7). However, SB5 variants retain mechanisms that restrain castration resistance, as none show 100% tumor incidence in the castrate setting.

Figure 2. Emergent CRPC variants exhibit loss of sGC activity.

(A) Schematic showing development and nomenclature of isogenic LNCaP-derived lines representing emergent, progressively castration-resistant variants. The emergent CRPC variants retain AR expression under AD (right). (B, C) The mRNA levels of GUCY1A1 and GUCY1B1 (ActinB as loading control) under the indicated culture conditions. (D) Immunoblotting of total protein lysates from the lines in (A). Cells cultured under AD for 30 hrs. (E) Cell lines were cultured in FBS or CSS-supplemented media for 5 days and then treated with riociguat, following which cGMP levels were measured. (F) Methylation of GUCY1B1 gene vs. its expression in Gleason 6 vs. Gleason 9 prostatic tumors (TCGA patient dataset). (G) GUCY1B1 mRNA levels (ActinB as loading control) following 5-aza-2’-deoxycytidine (Aza)-treatment in the indicated cell lines. (H) Immunoblotting for sGCα1 and sGCβ1 in Aza-treated LNCaP SB0 and SB5 cells (as in G). (I) Stimulation of cGMP levels by riociguat in control or Aza-treated SB5 cells (as in G). (J) Immunoblotting for sGCβ1 and cl-PARP in riociguat (± Aza)-treated SB5 cells. (K) GUCY1A1 and GUCY1B1 mRNA levels following RG108 treatment (ActinB as normalization control). (L) Immunoblotting of counterpart RG108-treated cells from (K) for sGC subunit protein expression.

Analyses of sGCα1 and sGCβ1 levels across this matched panel revealed the emergent CRPC lines possess significantly dysregulated sGC stoichiometry, manifesting as very low sGCβ1 expression (Figs. 2B–D). By contrast, sGCα1 was not limiting for heterodimer formation in any of the lines. We noted a mismatch between sGCα1 mRNA and protein levels due to its high protein stability (26), which has been previously noted in cancer lines (30).Thus, functional sGC stoichiometry in LNCaP cells appears to be critically regulated by β1 subunit expression. Indeed, the low sGCβ1 levels in emergent CRPC cells were sufficient to dramatically dampen sGC function, preventing riociguat stimulation of its activity (Fig. 2E, Supplementary Fig. S2F).

The sGCβ1 subunit is repressed by gene methylation in some breast cancer lines (20). Notably, methylation at the GUCY1B1 locus correlated more strongly with downregulation of β1 expression in advanced CSPC tumors most likely to progress to CRPC (Gleason 9, TCGA) than in early-stage CSPC (Gleason 6, TCGA) (Fig. 2F). Congruently, treating emergent CRPC SB5 cells with the nucleoside-analog DNA methyltransferase inhibitor, 5-aza-2’-deoxycytidine (Aza), restored sGCβ1 mRNA and protein expression as well as riociguat stimulation of sGC activity (Figs. 2G–I). Consistent with the Gleason 6 methylation data, Aza treatment did not alter β1 expression to the same extent in the CSPC SB0 cells. Aza treatment also enhanced riociguat-induced apoptosis in SB5 cells (Fig. 2J). Whereas GUCY1A1 was negatively regulated by methylation at both the CSPC and emergent CRPC stages (Supplementary Figs. S2G–H), its baseline expression was not limiting for heterodimer formation. Thus, restoration of sGC activity in SB5 correlates with β1 re-expression. We ruled out the possibility that Aza-induced changes in sGCβ1 expression occurred through AR, which was largely unchanged under Aza treatment (Supplementary Figs. S2I–J). We also confirmed that sGCβ1 expression is not directly repressed by AR in SB5 cells (Supplementary Fig. S2K).

Curiously, use of RG108 (31), a non-nucleoside blocker of DNA methyltransferase enzymatic activity, induced the expression of both sGC subunits more strongly in the SB0 CSPC cells compared to their SB5 counterparts (Figs. 2K, L). Differences in activity and cellular outcomes between these two DNA methyltransferase inhibitors have been previously reported and are believed to correlate to differences in baseline gene methylation (31,32). Methylation signatures vary significantly among early and advanced PC stages (33) although epigenetic regulation of PC progression remains poorly understood. Our results suggest epigenetic changes at the sGC loci in CSPC vs. CRPC could constitute novel markers of progression, requiring further investigation. Regardless, our collective results here point to transcriptional repression of sGC heterodimer formation as an early alteration in the progression to CRPC.

Abrogating sGC activity in CSPC promotes ADIS evasion and emergence of castration resistance

We next determined whether inhibiting sGC function could confer castration resistance. Whereas sGCβ1 was notably absent in ADIS-resistant LNCaP SB5 (Fig. 2), its progressive elevation in androgen-deprived CSPC SB0 coincided with p16^INK4a induction which marks ADIS (4) (Fig. 3A). Moreover, riociguat treatment elevated SA-beta-gal activity in androgen-replete SB0 cultures and decreased cell viability under FBS culture comparably to CSS (Supplementary Figs. S3A–C), indicating that sGC stimulation in androgen-replete CSPC cells recapitulates ADIS. Thus, we postulated that inhibiting sGC would lower the ADIS barrier to CRPC progression.

Figure 3. Inhibition of sGC activity facilitates ADIS evasion and castration-resistant tumor growth by LNCaP CSPC cells.

(A) Immunoblotting to show sGCβ1 protein levels under ADIS. (B) GUCY1A1 and GUCY1B1 mRNA levels (GAPDH as normalization control) in LNCaP SB0 KO sGCβ1 vs. WT cells. (C) Depletion of sGCβ1 validated by immunoblotting. (D) Baseline and riociguat-stimulated cGMP levels from WT and KO sGCβ1 LNCaP lines. (E) Cell proliferation curves for KO sGCβ1 or WT cells under indicated conditions. (F) Senescence-associated beta-galactosidase (SA-beta-gal) activity following 13d AD culture. Representative light microscopy images of stained cells (left), 100 μM scale bar shown, and quantitation (right). (G) Immunoblotting of KO sGCβ1 vs. WT LNCaP cells for p16^INK4a protein levels vs. WT counterpart cells, following 13-day AD to induce ADIS. (H) Tumor formation kinetics for LNCaP KO sGCβ1 or matched WT cells. (I) Endpoint tumor volumes (mm³) from (H). Each point represents an individual tumor. Tumor incidence as follows: WT 1/5; KO sGCβ1 4/7. Wilcoxon two-sample test. (J) Endpoint tumor weights (mg) from (H). Each point represents an individual tumor. Wilcoxon two-sample test. Absence of a tumor mass upon necropsy was indicated as zero. (K) Representative images of castration-resistant LNCaP KO sGCβ1 (n=4) tumors (tumor weights noted below the images). The sole wildtype tumor could not be wholly resected in a single piece due to its small size and was thus not photographed. (L) Intratumoral cGMP levels in KO sGCβ1 tumors vs. WT counterpart. (M) Immunoblotting for sGCα1, sGCβ1 and AR levels in tumor lysates corresponding to the tumors evaluated for cGMP in (L). Endpoint tumor weights are indicated above the blot. (N) Meta-analysis of an scRNA-seq dataset from the Single Cell Portal at Broad Institute for GUCY1A1 and GUCY1B1 mRNA level. Four distinct epithelial clusters (two basal, two luminal) derived from human prostate tissue, representing histologically normal regions of hormonally intact and ADT-treated tumors, are shown. (O) GUCY1A1, GUCY1B1 expression in basal-like benign and cancer cells vs. luminal-like cells (GSE82071).

Loss of sGCβ1 eliminates physiologic sGC enzymatic activity in cells and tissues from germline-deleted mice (34). Therefore, we deleted sGCβ1 via CRISPR-Cas9 gene editing in LNCaP SB0 cells (Figs. 3B–C) to see if this loss would abrogate sGC activity and confer ADIS resistance. Riociguat was unable to stimulate cGMP production in sGCβ1 KO LNCaP (Fig. 3D), recapitulating the loss of sGC function seen in LNCaP SB5 (Figs. 2D–E). Under androgen-replete conditions, sGCβ1 KO cells displayed enhanced proliferation compared to their WT counterparts (Fig. 3E, left; Supplementary Figs. S3D–E, top), consistent with cell cycle-inhibitory effects previously reported for the β1 subunit (35). More significantly, the sGCβ1 KO LNCaP population acquired key ADIS evasion traits (4), namely proliferation and lower SA-beta-gal activity under AD (Figs. 3E–F, right; Supplementary Figs. S3D–E) as well as lower p16^INK4a levels (Fig. 3G) relative to WT counterparts.

To determine whether loss of sGC activity promoted emergence of castration resistance in vivo, we subcutaneously injected either sGCβ1 KO or counterpart wildtype (WT) LNCaP cells into the flanks of castrated Nu/Nu male animals and monitored tumor growth. Neither cohort of injected animals exhibited any significant differences in weight or other ill-effects (Supplementary Fig. S3F). Yet the sGCβ1 KO tumors showed markedly greater CR tumor kinetics relative to their wildtype (WT) counterparts (Fig. 3H). Differences in endpoint tumor volumes and weights (Figs. 3I–K) were borderline non-significant due to the wide variability in CR tumor size formed by the sGCβ1 KO cells. Nevertheless, 4 out of 7 sites injected with the LNCaP sGCβ1 KO cells robustly formed CR tumors, compared to a single very small tumor formed by WT LNCaP (Figs. 3I–K). During necropsy, a fifth sGCβ1 KO site was found to have formed a very small tumor. However, none of the other WT injection sites showed any sign of tumor formation upon necropsy.

Intratumoral cGMP levels were also much higher in the one WT sGC tumor that formed compared to the four sGCβ1 KO CR tumors (Fig. 3L). Furthermore, as we started with a pooled CRISPR population rather than a single clone, we noted that the largest sGCβ1 KO tumor had near undetectable sGCβ1 levels compared to the smaller tumors (Fig. 3M), again reinforcing that loss of the functional sGC complex promotes castration-resistant tumor growth by CSPC cells. Unexpectedly, this tumor also had very low AR protein (Fig. 3M). This corresponded to lower AR expression in the sGCβ1 KO LNCaP CSPC bulk population vs. WT under AD (Supplementary Fig. S3G). It is unlikely low sGCβ1 levels or concomitant low sGC activity directly repress AR; under androgen-replete (FBS) conditions, AR expression was unaltered by sGCβ1 depletion (Supplementary Fig. S3G). Thus, this observation suggested that in vivo emergent CRPC subpopulations may benefit from low AR.

While this seems paradoxical given that CRPC is marked by AR stabilization under AD, low AR-expressing populations could reflect a transitional state necessary to evade the ADIS barrier. Normal adult prostatic tissue comprises of AR-low basal cell populations that possess prostate regenerative properties and are naturally AD-refractory (36). Moreover, prior studies have reported that aggressive metastatic CRPC tumors are enriched for cells that share transcriptional programs reminiscent of AR-low normal basal stem cells, whereas organ-confined castration-sensitive disease is enriched for AR-expressing luminal cell signatures (37). Comparison of sGC subunit levels in basal vs. luminal prostatic cells from a single cell RNA-seq profile of histologically normal human prostatic tissue (38), revealed that basal cells possess barely detectable sGC expression vs. luminal cells (Fig. 3N). Similar results were obtained through analysis of published GEO bulk transcriptomics datasets (GSE82071) (37) comparing basal vs. luminal signature-expressing cells derived from benign or cancerous human prostatic tissue (Fig. 3O). Thus, low sGC expression is a feature of naturally AD-resistant cells in prostatic tissue, suggesting that loss of sGC complex may promote CRPC emergence by enabling tolerance to low AR levels under AD or promoting transient dedifferentation to a basal-like cell state through as yet-unknown mechanisms. Nevertheless, our results support that sGCβ1 deletion and concomitant loss of sGC activity lower the ADIS barrier to emergence of castration-resistant sub-populations, suggesting inhibiting sGC is a key adaptation for castration resistance.

Androgen deprivation enhances riociguat-mediated stimulation of sGC in established CRPC cells

We next determined whether treatment with the sGC agonist riociguat, an FDA-approved treatment for pulmonary hypertension (39,40), would inhibit CRPC growth by enhancing activity of the attenuated sGC pathway. The physiologic activities of sGC are critically dependent on the obligate heterodimer, with lack of functional complex conferring resistance to riociguat. Thus we examined GUCY1A1 and GUCY1B1 expression in a spatial transcriptomics dataset and ascertained that the sGC subunits are being co-expressed in human metastatic CRPC (41) (Fig. 4A, Supplementary Fig. S4A). Because riociguat has the advantage of stimulating cGMP production from relatively low levels of the intact complex and the expression of PDE5A (which hydrolyzes cGMP) is low in CRPC (Fig. 4A), steady state increases in cGMP through riociguat action is favored.

Figure 4. AD enhances riociguat-mediated stimulation of sGC activity in CRPC cells.

(A) Publicly available spatial transcriptomics profiling in metastatic human CRPC. GUCY1A1, GUCY1B1 and PDE5A expression in PC and indicated immune cells are shown. (B) GUCY1A1 and GUCY1B1 mRNA levels across PC lines (ActinB as normalization control). (C) Immunoblotting for sGCα1, sGCβ1 and AR under the indicated conditions. (D) Relative cGMP levels under indicated culture conditions in riociguat-treated LNCaP SB0 vs. LNAI cells. (E) Baseline and riociguat-stimulated cGMP levels from CRPC cell lines LNAI and 22Rv1. (F) Cell viability curve. LNAI cells treated with riociguat under the indicated culture conditions (72 hours, with daily drug redosing). IC₅₀s for each curve were modeled using a four-parameter logistic function. The Wilcoxon two-sample test. (G) LNAI cells were plated in FBS-supplemented media as in Fig. 4F, then incubated with vehicle or 10 μM enzalutamide for 24 hours before dosing with riociguat (72 hr). Viabilities, IC₅₀s and p-values were established as in Fig. 4F. (H) 22Rv1 cells were treated with riociguat and analysed as in Fig. 4F.

We then evaluated expression of the sGC heterodimer subunits in fully established CRPC cells, LNAI and 22Rv1. LNAI CRPC cells, originally derived from LNCaP CSPC (4,7), expressed detectable levels of both sGCα1 and sGCβ1 (Figs. 4B–C). Thus, although sGC expression in LNAI was lower than in the parental SB0 cells, we surmised they possess functional sGC heterodimer unlike the emergent CRPC SB5 variants. Consistent with this idea, riociguat was able to stimulate sGC activity in LNAI, albeit to a lesser extent than in parental LNCaP cells (Fig. 4D). The lower level of sGC expression and stimulation in LNAI is consistent with the idea that sGC is inhibited in established CRPC relative to CSPC, as we saw in patient data (Figs. 1I–J).

Riociguat also stimulated cGMP production in the constitutively active AR-V7- expressing 22Rv1 CRPC cells, however at much lower levels than in LNAI (Fig. 4E). The 22Rv1 cells exhibited indications of dysregulated sGC stoichiometry, expressing very low levels of sGCα1 but very high levels of sGCβ1 (Figs. 4B–C), potentially explaining the relatively lower degree of riociguat stimulation vs. LNAI (Fig. 4E). Significantly, under AD, both LNAI and 22Rv1 showed an approximately two-fold enhancement in stimulation of sGC activity by riociguat (Fig. 4E, note the difference in FBS vs. CSS y-axis scales). Therefore, unlike the emergent CRPC variants, established CRPC cells possess sufficient intact sGC complex for effective stimulation by riociguat and its efficacy is enhanced under AD. We next determined whether riociguat-mediated stimulation of sGC activity affected the viability of CRPC cells. Congruent with the enhanced cGMP stimulation by riociguat in CSS culture, LNAI and 22Rv1 CRPC cells were most sensitive to riociguat-induced cytotoxicity under depletion of androgens or AR blockade via enzalutamide (in FBS-cultured LNAI) (Figs. 4F–H, Supplementary Figs. S4B–D). Thus, our results support that riociguat is a promising candidate for combination with the gold-standard ADT agents, enzalutamide or abiraterone, in CRPC.

The sGC complex is reversibly inactivated by oxidation in CRPC

We next investigated why AD stimulated sGC activity. We and others have previously shown that CRPC cells sustain a more pro-oxidant cellular environment than CSPC cells (7,42), and are buffered from oxidative stress by AD-induced redox protective responses (7). In several inflammatory and age-associated pathologies (26), the sGC complex is functionally inactivated through oxidation of its ferrous heme to the ferric state (43) (Fig. 5A). The implications of this redox regulation of sGC function have not been previously studied in cancer. Riociguat requires the reduced ferrous heme in order to stimulate cGMP production from sGC (Fig. 5A) (43). We therefore investigated whether the sGC complex is intrinsically inhibited by oxidation in CRPC (but not CSPC) cells. To do so, we utilized cinaciguat, an sGC activator that stimulates sGC activity only in the absence of a functional (reduced) heme group (44) (Fig. 5A). Although cinaciguat is not approved for clinical use due to hypotensive side-effects (45), it is the optimal tool to assess cellular sGC oxidation state, as its specificity for sGC redox state is orthogonal to that of riociguat.

Figure 5. The sGC complex is reversibly inactivated by oxidation in CRPC cells.

(A) Schematic of sGC redox state specificities for riociguat vs. cinaciguat (created with Biorender.com). (B) Cinaciguat-induced stimulation of cGMP levels. (C) Cinaciguat treatment of LNAI. Treatment and analyses as in Fig. 4F. (D) Immunoblotting for cl-PARP in riociguat vs. cinaciguat-treated cells. (E) Cinaciguat treatment of 22Rv1. Treatment and analyses as in Fig. 4F. (F) Riociguat treatment of control or TRX1-overexpressing LNAI cells (immunoblot in inset) under androgen-replete conditions. Treatment and analyses as in Fig. 4F.

Cinaciguat treatment stimulated cGMP production in LNAI but not in LNCaP SB0 cells (Fig. 5B), indicating that sGC is oxidatively inactivated in the former but not the latter. Cinaciguat was also able to significantly stimulate cGMP production in the most CRPC-like (SB5XX) variant (Fig. 5B). This suggests SB5XX possesses low levels of sGCβ1, albeit undetectable by immunoblotting, that is stabilized by cinaciguat (46). By contrast, the earliest CRPC emergent SB5 line could not be stimulated to produce cGMP by either riociguat or cinaciguat (Fig. 5B, Supplementary Fig. S2F), supporting that full abrogation of sGC activity is required to evade ADIS. These results indicate that as PC cells progress towards full-blown castration resistance, control of sGC signaling shifts from repression of heterodimer formation towards reversible oxidative inactivation of the functional heterodimer. This portends a baseline requirement for sGC activity in fully CRPC cells, which is promising for riociguat efficacy in this setting.

Consistent with the ability of cinaciguat to stimulate cGMP production, CRPC LNAI cells showed a robust apoptotic response to cinaciguat treatment even under FBS culture (Figs. 5C, D, Supplementary Fig. S5A). Cinaciguat also decreased the viability of androgen-replete 22Rv1 cells (Fig. 5E, Supplementary Fig. S5B). To further support that sGC is oxidatively inactivated in CRPC cells, treating LNAI cells with riociguat under 5% oxygen culture, known to reduce intracellular ROS levels (7,47) induced a comparable response to AD, even in FBS culture (Supplementary Fig. S5C). Thus, sGC is functionally regenerated under redox-protective cellular conditions. We previously reported that CRPC cells elevate the redox-protective protein, TRX1, to protect against AD-induced ROS levels and cell death (7). TRX1 is a known mechanism for restoring sGC activity via thiol reduction (48), and its levels progressively increased across the CRPC continuum in LNCaP-derived cells (Supplementary Fig. S5D). Indeed, TRX1-overexpressing LNAI were more sensitive to riociguat cytotoxicity relative to WT counterparts, even under androgen-replete conditions (Fig. 5F). Collectively, these results highlight that redox-protective responses evoked by CRPC to protect against AD-induced death can create a paradoxical therapeutic vulnerability to riociguat in the AD setting.

Riociguat cooperates with AD to induce high potency oxidants and dysregulate mitochondrial function in established CRPC cells

We have previously shown androgen-deprived CRPC cells are uniquely vulnerable to redox stress (7). Thus, we next assessed whether sGC stimulation in the AD setting decreased CRPC cell viability by altering redox balance. Levels of intracellular superoxide, hydrogen peroxide (H₂O₂) or high potency ROS were measured in riociguat-treated LNAI cells using specific cell-permeant fluorophores. Superoxide and H₂O₂ are largely signal-transducing entities in cancer cells and promote oncogenic kinase-driven signaling (49) whereas high potency oxidants are short-lived, cell-damaging species (50). Riociguat treatment did not significantly alter total cellular superoxide (measured via hydroethidium (HEt)) under FBS or CSS culture (Fig. 6A). Total cellular H₂O₂ levels (measured by chloromethyl-dichlorofluorescein diacetate, CM-DCFDA) showed some fluctuations under FBS riociguat treatment but did not change significantly under AD (Fig. 6B). CM-DCF-DA specificity was verified by determining fluorescence signal increases following addition of H₂O₂but not the NO donor, DETA-NO, nor the mitochondrial uncoupler/superoxide generator, CCCP (Supplementary Fig. S6A). However, reactive high potency oxidants (including hydroxyl radical and peroxynitrite anion), measured by hydroxyphenyl fluorescein (HPF), were elevated under AD but not androgen-replete conditions (Fig. 6C). Given the very short lifetimes of the oxidants measured by HPF, it is likely our measurements are underestimates, and that the apparent steady-state increase under AD may in fact reflect continual production of these entities by riociguat. Specificity of the HPF probe was verified by the ability of the iron chelator, deferoxamine (DFO) to reverse increased signal following DETA-NO or H₂O₂ treatment (Supplementary Fig. S6B).

Figure 6. Riociguat cooperates with androgen deprivation to induce redox and mitochondrial dysfunction in CRPC cells.

(A-C) ROS profiling of riociguat-treated LNAI cells (24 hr treatment) under indicated culture conditions. (A) Intracellular superoxide levels (HEt signal, PI channel). (B) Intracellular peroxide levels (DCF signal, FITC channel). (C) Intracellular high-potency ROS (HPF fluorescence, FITC channel). (D) Cell viability at 72 hrs, riociguat ± ferrostatin. Unpaired Student’s t-test with Welch’s correction. (E, F) Oxygen consumption rates (OCR) profiles under indicated culture conditions (24 hr riociguat treatment). (G) Mitochondrial bioenergetics profiling. Basal respiration rates, maximal respiration rate, ATP production and spare respiratory capacity under FBS or CSS culture from the data in (E, F). (H) Immunoblotting for pro- and anti-apoptotic proteins, following riociguat treatment.

Because high potency oxidants can readily react with DNA and produce genotoxic breaks, we next assessed whether riociguat induced DNA breaks in androgen-deprived LNAI cells via the alkaline single-cell gel electrophoresis (“comet”) assay. Induction of DNA breaks is visualized as cells with increased DNA “tail” moments under electrophoretic conditions, whereas non-damaged cells visualize as intact spheres (Supplementary Fig. S6C). However, our results did not support induction of DNA strand breaks as the source of riociguat/AD-mediated cytotoxicity (Supplementary Fig. S6D). High potency ROS can also produce highly damaging lipid peroxides. We found that the cytotoxic effects of riociguat under AD could be partially mitigated by the lipophilic antioxidant, ferrostatin (Fig. 6D). However, riociguat treatment did not induce steady-state changes in overall cellular redox state as measured by the ratio of reduced to oxidized glutathione (GSH/GSSG) (Supplementary Fig. S6E).

We next assessed whether dysfunctional mitochondrial bioenergetics were the source of riociguat-mediated cytotoxicity, via Seahorse cellular respiratory profiling (Figs. 6E–F). Our results indicated that riociguat cooperated with AD to decrease maximal aerobic respiration and, concomitantly, spare respiratory capacity, a measure of cellular stress tolerance (51) (Fig. 6G). We further noted that riociguat treatment increased the apoptotic marker, cleaved (cl) PARP, in androgen-deprived LNAI and decreased levels of mitochondrial pro-survival proteins, Bcl-2 and Survivin (Fig. 6H). Our collective results indicate that the effects of combined AD/riociguat treatment on CRPC cells are associated with elevated high potency ROS, and a tumor-suppressive stress response that converges on mitochondrial stress and apoptosis induction.

Riociguat treatment decreases CRPC growth through on-target stimulation of sGC activity and AD-induced apoptosis

To test whether riociguat could restrain in vivo CRPC growth, we subcutaneously injected LNAI cells into the flanks of athymic castrated male mice and following palpable tumor growth, randomized the animals into groups receiving daily intraperitoneal (IP) injections of either vehicle (DMSO) or 20 mg/kg/day riociguat. This dose was well tolerated as riociguat-treated animals did not exhibit weight loss (Supplementary Fig. S7A) or other physical signs of drug toxicity. Riociguat treatment decreased tumor growth kinetics as well as endpoint tumor sizes and tumor weights (Figs. 7A–C). Hematoxylin & eosin staining (H&E) indicated a perceptible reduction of the pan-proliferative marker, Ki67, in riociguat-treated tumors and morphology consistent with cell death (reduced cellularity, pyknosis; Fig. 7D). Apoptosis is a key determinant of AD efficacy in preclinical PC models (52,53) and TUNEL staining confirmed a striking increase in apoptosis-associated DNA fragmentation following riociguat treatment (Fig. 7E). Riociguat-treated tumors also had higher cl-PARP levels relative to vehicle (Fig. 7F). Riociguat treatment did not impair the ability of prostate-specific antigen (PSA) levels to correlate positively with castration-resistant tumor growth (Supplementary Fig. S7B). Intriguingly, we found AR levels increased following riociguat treatment (Supplementary Figs. S7C–D), with significantly higher expression in riociguat-responsive tumors (tumor volume < 500 mm³) compared to vehicle or less treatment-responsive tumors (tumor volume> 500 mm³) (Fig. 7G). We have previously reported AR levels are increased by oxidative stress in LNAI tumors (7). Therefore, this increase could reflect the dysregulated redox conditions associated with riociguat-induced cytotoxicity and/or a shift towards a less aggressive or more-differentiated AR-high PC subtype.

Figure 7. Riociguat treatment decreases CRPC growth through on-target stimulation of sGC and induction of apoptosis.

(A) Tumor growth kinetics for LNAI tumor-bearing castrated male Nu/Nu mice from riociguat (20 mg/kg/day) or DMSO treatment groups (n=14/group). Inset shows plotted aAUCs. Wilcoxon two-sample test. (B) Endpoint tumor volumes (mm³) and tumor weights (g) per treatment group. (C) Representative images of castration-resistant tumors derived from the groups in (B). (D) Representative H&E and Ki67-stained images (50 μM scale bar). (E) TUNEL staining. Representative images of co-localized dUTP/DAPI staining for two representative tumors in each group (100 μm scale bar) with quantitation (right). (F) Immunoblotting for cleaved (cl)-PARP in tumor lysates (representative blot shown) with quantitation of the actin-normalized immunoblot signal (right). (G) Normalized AR protein levels from immunoblotted vehicle, riociguat-responsive or non-responsive tumors lysates. (H) Immunoblotting for p-VASP(Ser239)/VASP in tumor lysates (representative blot). Quantitation of protein signal, normalized to total VASP (right). (I) Intratumoral cGMP levels. (J) Intratumoral cGMP levels vs. tumor volumes in riociguat-treated animals. Spearman correlation coefficients are shown. (K) Plasma cGMP levels. (L) Correlation between plasma cGMP levels vs. tumor volumes in riociguat-treated animals. (M) Plasma cGMP levels vs. intratumoral cGMP levels in riociguat-treated animals.

Riociguat-treated tumors had significantly higher levels of phosphorylated vasodilator-stimulated protein (p-VASP), a downstream marker for cGMP-dependent signaling (54) (Fig. 7H). Consistent with this finding, cGMP levels were significantly elevated in the lungs (site of maximal bioactivity) and in the tumors of riociguat-treated animals (Supplementary Fig. S7E, Fig. 7I). We further noted cGMP levels were highest in the smallest treated tumors (Fig. 7J), supporting cGMP as a reliable marker for on-target riociguat anti-tumor efficacy. We also measured cGMP levels in plasma, which can be obtained noninvasively compared to tumor tissue, and found these levels similarly were higher in riociguat-treated animals and correlated inversely with tumor growth (Figs. 7K, L). Plasma and intratumoral cGMP levels were strongly correlated (Fig. 7M), indicating that monitoring blood cGMP can accurately reflect intratumoral sGC stimulation by riociguat. Collectively, our data herein show that riociguat effectively inhibits castration-resistant tumor growth by inducing apoptosis, and that the degree of its on-target efficacy can be readily measured by elevated intratumoral or plasma cGMP levels, in addition to declining PSA.

Riociguat treatment in CRPC is associated with improved tumor oxygenation and enhanced sensitivity to radiation

Castration induces hypoxia (55) which promotes metastasis (56), and targeting hypoxic cells can inhibit PC progression (57). Because sGC stimulation is expected to increase tissue oxygenation through its physiologic function, we evaluated whether riociguat treatment mitigated tumor hypoxia. Two hours prior to euthanasia, we randomly assigned experimental animals per group for injection with pimonidazole (Hypoxyprobe). Hypoxyprobe forms protein adducts in hypoxic tissues, detectable via specific fluorescent antibodies to enable visualization of hypoxic tumor regions (58). Riociguat-treated tumors showed a striking decrease in Hypoxyprobe fluorescent signal, independent of tumor size (Fig. 8A). Decreased Hypoxyprobe staining was accompanied by endothelial (CD31-positive) cell reorganization around an open lumen in riociguat-treated tumors, indicative of improved vasculature (Fig. 8B). This was in sharp contrast to the narrow constricted CD31 staining in vehicle-treated counterparts. Additional images from different tumors are provided in lieu of quantitation to support these observations (Supplementary Fig. S8A). Human prostate cancers express sGC in vascular cell types (59) (Supplementary Fig. S8B) and, post-castration, leaky poorly-formed vasculature and vasoconstriction mark hypoxic prostatic tissue (60,61). Thus riociguat treatment likely oxygenates CRPC tumors through canalisation of tumor blood vessels.

Because hypoxic niches house CRPC-promoting tumor stem cells, we also stained tumor sections for CD44, a key PC stem cell marker (62,63). Cell membrane-localized CD44 staining was lower or delocalized in smaller riociguat-treated tumors (responders R1, R2, tumors < 500 mm³; Fig. 8C, Supplementary Fig. S8C). These changes were not observed to the same extent in riociguat non-responder (NR, tumor volume > 500 mm³) or vehicle-treated tumor (Fig. 8C). This finding indicates riociguat anti-tumor response correlates with reduction in CRPC-driving stem-like populations. Riociguat responders were also marked by changes in host (murine) cytokines that correlate functionally with improved vasculature and reduced blood pressure namely, lower levels of angiopoietin-1 (64) and elevated IGFBP1 levels (65) (Fig. 8D, topmost graph). Riociguat responders also exhibited unique changes in paracrine factors that portend decreased tissue stemness (Fig. 8D, middle and bottom) namely, decreased levels of regenerative factors HGF and DKK1 (66,67) and elevated FGF21, which also improves vascular function (68) and promotes apoptosis in PC cells (69). Thus, the paracrine changes in riociguat-responsive tumors functionally mirrored their increased cGMP levels and decreased membrane-localized CD44 expression. By contrast, the cytokine profiles of riociguat nonresponders were essentially identical to vehicle-treated animals (Fig. 8D), suggesting tumor oxygenation is necessary but not sufficient for riociguat anti-tumor activity and that its effects on the paracrine microenvironment (among other factors) could also influence CRPC response.

We noted that despite variability in its CRPC-inhibitory effects (Fig. 7B), riociguat oxygenated even non-responder tumors (Fig. 8A), creating a vulnerability to additional treatment modalities. Castration-resistant tumors also become resistant to radiation therapy (70) in large part due their hypoxic state (71). This radioresistance is partly mediated by self-renewing PC stem cells in hypoxic niches (72). Therefore, we tested whether riociguat treatment, through its mitigation of tumor hypoxia, could increase the efficacy of ionizing radiation (IR) in CRPC tumors. Following palpable tumor growth by injected LNAI cells, animals were randomized into groups that received either DMSO or 20 mg/kg/day riociguat daily for 72 hours prior to receiving either a one-time dose of 6 Gy ionizing radiation (IR) or 0 Gy radiation (mock IR) (Fig. 8E). Mice continued to receive vehicle or riociguat for the remainder of the experiment. Combining riociguat with IR significantly improved on tumor suppression over radiation or riociguat alone, with combined radiation and riociguat showing the greatest efficacy among the groups (Figs. 8E–F, Supplementary Fig. S8D). We further noted that riociguat-mediated CRPC inhibition was comparable to radiation alone (Figs. 8E–F). CRPC inhibition under these conditions was paralleled by the highest induction of cGMP levels in the combined riociguat/IR-treated tumors (Fig. 8G). The mechanism underlying this enhancement is not presently clear. Furthermore, riociguat combinatorial treatment further enhanced cl-PARP levels induced by IR (Fig. 8H). Moreover, the lowest Hypoxyprobe signal was seen in riociguat/IR tumors, coincident with their reduced tumor cellularity (Fig. 8I), suggesting IR efficacy correlates with riociguat-induced tumor oxygenation. Collectively, our results demonstrate riociguat enhances the efficacy of the two treatment modalities to which CRPC acquire resistance, namely AD and radiation therapy. An overview of our findings, describing changes in sGC expression and activity across the spectrum of PC disease and highlighting the novel therapeutic opportunities in CRPC, is provided (Fig. 8J).

Discussion

Our study is the first to report a therapeutically actionable role for sGC agonists, notably the FDA-approved vasodilator riociguat, in limiting CRPC growth through induction of apoptosis. Our studies indicate that riociguat induces tumor oxygenation, thus paving the way for improving other standard-of-care treatments, such as radiation. We find noninvasive cGMP measurements can predict the efficacy of on-target riociguat action in CRPC, without interfering with the predictive capabilities of PSA levels. In patients, a >50% decline in systemic cGMP levels portends worse disease prognosis and mortality following progression to CRPC, pointing to cGMP’s potential as a much-needed prognostic and predictive biomarker.

Riociguat treatment in androgen-deprived CRPC cells converges on redox and mitochondrial stress-mediated pathways, increasing high-potency short-lived oxidants while decreasing mitochondrial bioenergetics and anti-apoptotic proteins. Our results are consistent with other studies reporting that the sGC activator YC-1 or addition of the phosphodiesterase-resistant cGMP homolog 8-bromo-cGMP in breast cancer cells induces apoptosis through caspase activation (73). However, a question that remains unanswered from our studies is the precise nature of stress response that connects sGC stimulation, dysregulated oxidative metabolism and AD-induced apoptosis. Despite partial rescue of riociguat-induced cytotoxicity by ferrostatin, we could not establish (or indeed negate) a role for ferroptosis or other lipid oxidation-centered phenomena in riociguat/AD-mediated stress response. Riociguat treatment in CRPC cells did not alter GSH/GSSG ratios nor did it induce irreparable DNA double strand breaks. Going forward, resolving this question will likely require investigations into how riociguat treatment affects cross-talk among factors such as high-potency ROS (e.g. peroxynitrite), protein nitrosation, cGMP-mediated cellular calcium regulation and other multifactorial determinants of apoptosis that converge on the NO-sGC-cGMP pathway (74,75).

Although sGC stimulation has clear CRPC cell-intrinsic components i.e. tumor apoptosis, our in vivo studies indicate riociguat treatment also affected the CRPC microenvironment, notably remodeling the CD31 endothelial cells into canalised structures and inducing tumor oxygenation independent of endpoint tumor volumes. The decrease in CRPC-driving PC stem cells (CD44-high) in riociguat-responsive tumors suggest a potential mechanism for its anti-CRPC efficacy. Riociguat anti-tumor response was also associated with distinct host paracrine changes, congruent with vascular remodeling and loss of cellular plasticity. It is unclear why the growth of some tumors was not decreased to a greater extent by riociguat, despite efficient oxygenation. One explanation is that riociguat bioavailability through our IP procedure is sufficient for raising cGMP levels to promote oxygenation (a known acute effect of sGC signaling) but not apoptosis, which may be a chronic effect requiring a higher cGMP threshold or additional signaling effectors. Further studies involving measurements of tumor oxygenation and apoptosis at different time points following treatment with riociguat are required to clarify this point. Nevertheless, the consistent tumor oxygenation effect and improved vessel canalisation by riociguat opens the way for increasing the efficacy of radiation and radioligand therapies. CSPC tumors are very sensitive to radiotherapy (76) and riociguat treatment restored this vulnerability in CRPC tumors, in addition to resensitizing them to AD.

In addition to potential translational benefits, our study provides several mechanistic insights into how the sGC pathway regulates progression to CRPC. Here we provide the first evidence that, in CSPC cells, sGC complex activity enforces ADIS, a key therapeutic barrier to CRPC emergence, and that its loss enables ADIS evasion and castration-resistant tumor growth. Under AD, we found acute loss of the sGC complex associates with decreased AR expression but not decreased cell proliferation. Transcriptomic profiles of benign and cancerous prostatic patient tissues revealed that sGC subunits are significantly lower in naturally AD-refractory AR-low basal-like cells, which give rise to metastatic CRPC, compared to their luminal counterparts (37). Thus, suppression of sGC may lower the ADIS barrier to CRPC emergence through transient selection of or de-differentiation into AR-low basal-like cell states, enabling tolerance to AD. Conversely, riociguat-responsive CRPC tumors exhibited decreased markers of stemness as well as elevated AR, further supporting this idea.

Our results further indicate that in the naturally-occurring emergent CRPC LNCaP variants, GUCY1B1 expression is repressed through methylation, consistent with a prior report in breast cancer cells (77). As disruption of sGC activity induced by loss of sGCβ1 promoted castration-resistant tumor growth in CSPC populations, AD efficacy and durability in high-risk CSPC could potentially be enhanced through use of 5-aza-2’-deoxycytidine or other clinically-approved demethylating agents to restore or increase sGCβ1 expression. By contrast, unlike emergent CRPC, we found fully CRPC cells expressed low levels of the sGC complex, in a reversibly oxidized and inactive state not readily stimulated by riociguat. However, this functional inactivation could be overcome by combined riociguat and AD treatment, ostensibly through redox-protective responses mounted by CRPC cells under AD. An important component of this response is elevated TRX1, which protects against AD-induced apoptosis (7) and also maintains sGC in its reduced functional state (48). It should be noted TRX1 cannot efficiently decrease high potency oxidants. Thus, TRX1 represents a unique redox Achilles heel in CRPC cells, able to reductively regenerate the target for riociguat but unable to detoxify the specific oxidant species that accompany riociguat anti-tumor effects.

The fact that sGC activity was fully abrogated in emergent CRPC cells but restored at low levels in established CRPC cells indicates it acts as a rheostat for CRPC progression and growth, rather than an absolute barrier. This allows sGC to be an effective therapeutic target in full-blown CRPC, as riociguat can stimulate cGMP production even from low levels of the sGC complex. Notably, combining riociguat with AD or androgen blockade induced robust cytotoxicity in fully CRPC cells regardless of whether they were driven by full-length AR (LNAI) or the ligand-independent AR-V7 mutant (22Rv1). Our collective findings thus show riociguat exploits unique and intrinsic CRPC vulnerabilities associated with the sGC pathway and is likely to have clinical benefit across the spectrum of advanced PC.

We note that prior studies focusing on the sGC-cGMP pathway in PC (78–80) proposed inhibition rather than stimulation of sGC signaling as a tumor-suppressive strategy. Several points of difference exist between these published studies and ours. In a subset of studies (78,79), in vivo experiments were carried out through peptide inhibition of the sGCα1 subunit in CRPC lines without accompanying changes in cGMP to indicate on-target effects, and in vivo experiments were carried out in non-castrated animals, thus losing the critical clinically relevant effects of systemic AD. An additional study which concluded that inhibition of sGC limits PC growth was carried out in the VCaP CSPC line (80), which does not depend on the sGC subunits for survival (depmap.org, Supplementary Fig. S1B). Furthermore, sGC was inhibited in these studies using compounds based on the oxidizing agent 1-H-[1,2,4]oxadiazolo[4,3-a]quinoxalin-1-one, which is known to induce cytotoxicity via sGC-independent nonselective effects on multiple heme proteins (81,82).

Our study explicitly shows that inhibition of sGC activity promotes acquisition of castration resistance in CSPC cells and that low sGC activity marks aggressive clinical CRPC. The transcriptional repression of sGC complex formation and oxidation-induced sGC inactivation in CRPC settings shown by us further underscore the selective pressure to repress sGC activity as the tumor evolves to castration resistance. Furthermore, our results robustly support that the FDA-approved sGC agonist, riociguat, possesses several advantages in the therapeutic space for CRPC, being clinically well-tolerated and safe, able to reduce CRPC growth through induction of apoptosis, and with treatment efficacy measurable through on-target cGMP increase. There are no reports associating long-term riociguat use with increased incidences of cancer. In 2020, a post-marketing surveillance study on 1031 patients from Japan reported adverse drug reactions (headache, dyspepsia/gastritis, dizziness, nausea, diarrhea, hypotension, vomiting, anemia, gastroesophageal reflux, and constipation) similar to the initial Phase 3 studies conducted internationally (83). The riociguat FDA package insert reports that studies in mice and rats, treated for two years and achieving six times the plasma exposure of unbound riociguat vs. the human plasma concentration, did not reveal any evidence of carcinogenesis. Thus, it is highly unlikely sGC activity is associated with enhanced malignancy (as suggested by the prior studies in PC), or indeed any other cancer.

Finally, the regulatory mechanisms we describe for sGC activity may represent more general tumor-modulating mechanisms beyond CRPC, as differential expression of sGC obligate heterodimeric subunits have been reported in breast cancer progression (20,77) and malignant gliomas (84). Thus, our findings describing the tumor-inhibitory mechanisms of sGC signaling and its stimulation via riociguat have significant potential for broader impact on the field of cancer biology and therapeutics, in addition to providing new treatment options for CRPC.

Statement of significance.

Soluble guanylyl cyclase signaling inhibits castration-resistant prostate cancer emergence and can be stimulated with FDA-approved riociguat to resensitize resistant tumors to androgen deprivation, providing a strategy to prevent and treat castration resistance.