Targeting Backdoor Androgen Synthesis Through AKR1C3 Inhibition: A Presurgical Hormonal Ablative Neoadjuvant Trial in High Risk Localized Prostate Cancer

Abstract

Background:

Localized prostate cancers (PCs) may resist neoadjuvant androgen receptor (AR)-targeted therapies as a result of persistent intraprostatic androgens arising through upregulation of steroidogenic enzymes. Therefore, we sought to evaluate clinical effects of neoadjuvant indomethacin (Indo), which inhibits the steroidogenic enzyme AKR1C3, in addition to combinatorial anti-androgen blockade, in men with high-risk PC undergoing radical prostatectomy (RP).

Methods:

This was an open label, single-site, Phase II neoadjuvant trial in men with high to very-high risk PC, as defined by NCCN criteria. Patients received 12 weeks of apalutamide (Apa), abiraterone acetate plus prednisone (AAP), degarelix, and Indo followed by RP. Primary objective was to determine the pathologic complete response rate (pCR). Secondary objectives included minimal residual disease (MRD) rate, defined as residual cancer burden (RCB) ≤0.25cm³ (tumor volume multiplied by tumor cellularity) and elucidation of molecular features of resistance.

Results:

Twenty patients were evaluable for the primary endpoint. Baseline median PSA was 10.1 ng/mL, 4 (20%) patients had Gleason grade group (GG) 4 disease and 16 had GG 5 disease. At RP, 1 (5%) patient had pCR and 6 (30%) had MRD. Therapy was well tolerated. Over a median follow-up of 23.8 months, 1 of 7 (14%) men with pathologic response and 6 of 13 (46%) men without pathologic response had a PSA relapse. There was no association between prostate hormone levels or HSD3B1 genotype with pathologic response.

Conclusions:

In men with high-risk PC, pCR rates remained low even with combinatorial AR-directed therapy, although rates of MRD were higher. Ongoing follow-up is needed to validate clinical outcomes of men who achieve MRD.

Introduction

Men undergoing radical prostatectomy (RP) for high-risk prostate cancer experience a high rate of relapse and subsequent prostate cancer-related mortality^1,2. Therapy to improve outcomes is therefore of significant interest. Neoadjuvant therapy may improve clinical outcomes through various mechanisms, including pathologic downstaging, enhancing resectability and eradicating microscopic disease outside of the surgical field. Although standard for certain tumor types, including breast and bladder, neoadjuvant therapies prior to RP have not been proven effective in men with prostate cancer (PC)³.

Given PC’s reliance on androgen receptor (AR)-signaling and the proven benefit of AR-signaling targeted therapies for advanced PC, multiple neoadjuvant hormonal therapies have been tested. Trials of androgen deprivation therapy (ADT) alone for 3–8 months prior to prostatectomy reported low pathologic response rates and no difference in biochemical recurrence at 5 years^4,5. This led to trials of combination approaches to further reduce serum and tissue hormone levels. Small phase II trials of ADT combined with the second-generation AR antagonist enzalutamide and/or abiraterone acetate (AA), a prodrug of abiraterone, the potent and specific inhibitor of CYP17, plus prednisone (P), have demonstrated pathologic response rates (minimal residual disease or complete response) of 15–60%^6–8 in high-risk PC. Preliminary results for AAP combined with ADT and apalutamide (Apa), a next-generation AR antagonist, were recently reported with a pathologic response rate of 10–30%^9,10. While pathologic complete response rates remain rare, rates of minimal residual disease (MRD) have increased with these newer agents. Pathologic response in PC neoadjuvant trials have not yet been validated as surrogate endpoints for long-term oncologic outcomes.

One mechanism of proposed resistance to neoadjuvant hormonal therapy relates to the persistence of dehydroepiandrosterone sulfate (DHEA-S), that can be converted by the steroidogenic enzyme AKR1C3 to a the potent AR ligands testosterone (T) and 5α-dihydrotestosterone (DHT)¹¹. AKR1C3 is frequently upregulated in PC, particularly in later stages, and contributes to resistance to treatment with both enzalutamide and AAP^12–14. Indomethacin, a nonsteroidal anti-inflammatory with off-target effect of inhibiting AKR1C3, inhibits enzalutamide and abiraterone resistant cell lines in vivo and in vitro^14–18. In two neoadjuvant trials of ADT including androgen biosynthesis-enzyme inhibition with either AAP or the non-selective agent ketoconazole prior to prostatectomy, a significant depot of DHEA-S in the serum (~20 μg/dL) remained^8,11,19. We hypothesized that residual DHEA-S could serve as a substrate for intratumoral conversion to T and DHT and that inhibiting AKR1C3 with indomethacin combined with androgen blockade prior to RP would lead to favorable pathologic outcomes.

We report the results of a trial of indomethacin combined with the gonadotropin-releasing hormone antagonist degarelix, abiraterone acetate plus prednisone and apalutamide for 12 weeks followed by RP for localized high or very-high risk prostate cancer. We evaluated the association of clinical and pathologic outcomes with serum and tissue hormone levels and analyzed next-generation sequencing and HSD3B1 genotypes.

Materials and Methods

Study Design

This was an open label, single-site, Phase II neoadjuvant trial (Supplemental Figure 1), with all procedures approved by the Fred Hutchinson Cancer Research Center Institutional Review Board (Seattle, WA). All subjects signed informed consent. Eligible men had localized, surgically resectable, high-to very-high risk prostate adenocarcinoma, as defined by NCCN criteria. Patients received 12 weeks of neoadjuvant therapy: three times daily of neoadjuvant indomethacin 50mg; Apa, AA, and degarelix at their respective FDA-approved doses plus prednisone 5mg twice daily. This dose of indomethacin was selected given that it has been shown to result in a maximum plasma concentration (Cmax) of 2.369 ug/mL, which is approximately three times higher than the IC50 for indomethacin shown to inhibit AKR1C3^16,20. The primary objective was determining the pathologic complete response (pCR) rate. Secondary objectives included assessing for MRD, measuring serum and intraprostatic androgens, assessing molecular features associated with drug resistance, and determining the PSA progression free survival (PFS) and metastasis free survival (MFS).

Pathology

Pre-treatment biopsies and radical prostatectomy specimens were processed using standard pathologic techniques. Expert genitourinary pathologists reviewed hematoxylin and eosin (H&E) stained slides. A pathologist upon study completion re-reviewed the slides. All prostatectomy tissue was submitted for processing to minimize sampling error. The extent of residual tumor (i.e. tumor volume) and cellularity were recorded with (1) minimal residual disease (MRD) defined as residual cancer burden (RCB) ≤0.25cm³ (tumor volume multiplied by tumor cellularity) as previously reported and (2) complete response defined as the absence of visible tumor by H&E stained slides⁸.

Hormone and Abiraterone Measurements

Serum and tissue androgen levels were determined on frozen samples by liquid chromatography/tandem mass spectrometry using methods described previously²¹. Hormones analyzed included DHT, T, dehydroepiandrosterone (DHEA), and DHEA-S. Similar methods were used for detection of AA and its metabolites, D4 abiraterone and 3-keto-5α-abiraterone²¹.

Genomic Analysis

Next-generation sequencing of tumor tissue was performed on pre- and post-treatment prostate tissue using the clinically validated UW-OncoPlex platform²². We determined germline HSD3B1 genotype with DNA extraction from prostate tissue using methods previously described²³.

Relapse Free Survival

Treating physician’s per standard of care monitored patients following RP. PSA relapse was defined as a value ≥0.2 ng/mL confirmed on two successive measurements. Metastasis was defined as visible lesions on bone scan, computed tomography (CT), or positron emission tomography (PET)/CT.

Statistical Analysis

The planned sample size of 20 patients provided 91% power to detect a difference in pCR rate of 5% (H0) vs. 25% (H1) based on an exact binomial test with one-sided α=7.5%. Serum androgen levels pre- and post-treatment, as well as serum and intraprostatic androgen levels among pathologic responders and non-responders, were compared using the Mann-Whitney test. PFS was estimated using Kaplan-Meier methods and differences between pathologic responders and non-responders were compared using two-sided log-rank tests. Associations between pathogenic mutations and pathologic outcomes were assessed using logistic regression.

Results

Baseline Characteristics

Twenty-two patients enrolled, with 20 evaluable for the primary endpoint (1 patient discontinued treatment to pursue stereotactic radiosurgery; 1 was removed after developing grade 3 transaminitis). Table 1 shows baseline characteristics. Median age at diagnosis was 63 years (range 51–73). Median PSA was 10.2 ng/mL (range 4.4–159.4). Four patients (20%) had Gleason grade (GG) 4 and 16 (80%) had GG 5.

Table 1.

Baseline characteristics

Characteristic
Patients accrued, n	22
Patients evaluable, n	20
Race, n
Black	1
White	18
Hawaiian or other Pacific Islander	1
Age, yrs median (range)	63 (51–73)
Baseline PSA, ng/mL median (range)	10 (5.6–159.4)
Gleason score median (range)	9 (8–10)
Clinical Tumor Stage
T1c	7
T2a	3
T2b	3
T2c	3
T3a	1
T3b	3

Pathologic and Clinical Outcomes

1 of 20 (5%) patients had a pCR, which was identical to the null hypothesized value. In contrast, 6 (30%) patients had MRD, for an overall pathologic response rate of 35%. Despite this, 4 of 6 patients with MRD had ypT3 disease at the time of surgery. Eighteen (90%) had ypT3 stage at RP and 7 (35%) had lymph node (LN) metastases (Supplemental Table 1). Histologic changes were similar to those previously reported for patients treated with late generation ADT²⁴. Notably, of four patients having focal or multifocal areas of intraductal carcinoma, none had a pathologic response. Eight patients had adjuvant radiation following RP. Over median follow-up for PSA relapse of 23.8 months (IQR 20.3–26.9 months), 1 of 7 (14%) men with pathologic response and 6 of 13 (46%) men without pathologic response had a PSA relapse (p=0.3 from log-rank test; Fig. 1A). Over a median follow-up for metastasis of 24.8 months (IQR 21.4–32.6 months), 1 of 7 (14%) men with a pathologic response and 1 of 13 (8%) without a pathologic response developed metastasis (p=0.7; Fig. 1B), both of whom also had PSA relapses. One patient was lost to follow-up at 18 months.

Figure 1A.

PSA progression-free survival by pathologic response status

Figure 1B.

Metastasis-free survival by pathologic response status

Adverse Events

Treatment was well tolerated with adverse events consistent with each drug’s safety profile (Table 2). One patient was removed due to asymptomatic grade 3 transaminitis that resolved after treatment discontinuation. There were no toxicities attributed to indomethacin. No unexpected complications at time of RP appeared after neoadjuvant therapy.

Table 2.

Adverse events while undergoing neoadjuvant therapy

Adverse Event	Grade 1 or 2, N (%)	Grade 3 or 4, N (%)
General Disorders
Fatigue	16 (73)
Edema	1 (4)
Weight loss	1 (4)
Mood Changes	5 (23)
Rash	1 (4)
Insomnia	3 (14)
Dizziness	2 (9)
Musculoskeletal Disorders	6 (27)
Neurologic Disorders
Cognitive Changes (including decreased concentration and attention)	11 (50)
Ocular migraines	1 (4)
Headache	2 (9)
Vivid dreams	1 (4)
Blurred vision	1 (4)
Balance problem	2 (9)
Endocrine Disorders
Hot Flashes	18 (82)
Decreased libido	1 (4)
Elevated TSH	1 (4)
Gastrointestinal Disorders	11 (50)
Hepatic Disorders
Increased transaminases	3 (14)	1 (4)
Injection Site Reaction	3 (14)
Hematologic disorders
Anemia	1 (4)
Epistaxis	3 (14)
Bruising	1 (4)
Respiratory Disorders- Dyspnea	1 (4)
Cardiovascular Disorders
Hypertension	3 (14)	6 (27)
Hypotension	2 (9)

Hormone and Abiraterone Levels

As expected, serum hormone levels significantly decreased after neoadjuvant treatment (Fig. 2). Testosterone decreased 99.8% from a mean of 354 ng/dL pre-treatment to 0.6 ng/dL post-treatment. DHT decreased by 96.5% from a mean of 33.9 ng/dL pre-treatment to 1.2 ng/dL post-treatment. DHEA-S decreased by 94.7% from a mean of 82.8 μg/dL to 4.4 μg/dL. Tissue levels of T and DHT at time of RP were below the limit of detection in nearly all tissues. No significant differences appeared in: (1) serum or tissue hormone levels by pathologic response status (all p≥0.17 from Mann-Whitney tests); (2) the number of patients with tissue T and DHT levels above the limit of detectability (all p≥0.5); or (3) abiraterone or abiraterone metabolite levels in either the serum or tissue stratified by either pathologic response status or HSD3B1 genotype (Supplemental Figures 2 and 3).

Figure 2.

Serum hormone levels pre- and post- neoadjuvant therapy. Box plots (median, interquartile range, and range) are shown. Comparisons between pre- and post-treatment were made using Mann-Whitney test.

HSD3B1 Genotype

Germline DNA testing was performed evaluating for the presence of at least one HSD3B1 (1245C) allele, the gene encoding for steroidogenic enzyme 3βHSD1; patients with one allele were defined as heterozygous, and those without as wild type. Overall, of 19 men having germline DNA testing, 10 (52.6%) were homozygous wild type and 9 (47.4%) were heterozygous. No patients were homozygous variant. No difference in response, as defined by achieving pCR or MRD, was present when stratified by genotype (Fig. 3A). Two of 10 (20%) with wild type genotypes experienced PSA relapse compared to 4 of 9 (44%) with heterozygous genotypes (p=0.4 by log-rank test, Fig. 3B). There was no difference in tissue hormone or abiraterone levels stratified by genotype.

Figure 3.

Pathologic and clinical outcomes by HSD3B1 c.1245C genotype. (A) Pathologic response by HSD3B1c.1245C genotype. Responder = achieving pathologic complete response or minimal residual disease. (B) PSA progression-free survival by HSD3B1 c.1245C genotype.

Somatic Genomic Alterations

A total of 14 patients had prostatic tissue available for sequencing, with 5 patients having both pre and post treatment tissue sequenced (Table 3). Alterations were consistent with those found in high-risk prostate cancer, with a high prevalence of TP53 mutations (43%). Interestingly, one patient had a germline MSH2 mutation (loss of heterozygosity analysis could not be performed due to low tumor content) but was microsatellite stable without hypermutation; two others had microsatellite unstable and hypermutated tumors (≥10 mutations/megabase). No significant differences appeared in oncogenic alterations between pre and post treatment tissue. Two patients had SPOP mutations and both had pathologic response. No significant association appeared between TP53 or microsatellite instability with pathologic response.

Table 3.

Targeted exome sequencing of prostate tissue.

Patient ID	Pre/Post Treatment	Genomic Alterations	Pathologic Responder
2	Pre	APC (p.T1556Nfs*3)	No
3	Post	TP53 (p.C238Y)	No
4	Pre	TP53 (p.Q331*) FOXA1 (p.T12P) TMPRSS2-ERG fusion TP53 (p.Y220C) MSH2 (p.R524P)- Germline	No
	Post	TP53 (p.Q331*) MSH2 (p.R524P)- Germline TMB 2mutations/Mb
5	Post	FOXA1 (p.G241D)	No
6	Post	MSI and Hypermutation MRE11A (p.N65Kfs*15) BMPR1A (p.L138Cfs*6)	No
7	Post	TP53 (p.K132N)	No
8	Pre	CHD1 copy loss SPOP (p.F133v) CCND1 amplification	Yes
12	Post	MSI-H High TMB (20mut/mb) PIK3CA (p.E542G) FOXA1 (p.G257s)	Yes
13	Pre	KDM6A (p.Q611*) DPYD (p.W621*)- VUS TACC3 (p.Q379*)- VUS
18	Pre	DDX41 (p.M316Dfs*31)- VUS TP53 (p.P190L)- pathogenic CARD11 (p.R848H)- VUS U2AF1 (p.F135C)- VUS	Yes
	Post	DDX41 (p.M316Dfs*31)
19	Pre	SPOP(p.F136V) TP53 (p.S6*) C11orf95 (p.A241_S242dup)- VUS DNMT3A (p.R882C) GRIN2A (p.S1077N)	Yes
	Post	DNMT3A (p.R882C) C11orf95 (p.A241_S242dup)- VUS
20	Pre	BRIP1 (p.D563del)- VUS; favor benign germline variant	No
21	Pre	MSH6 (VUS- germline favored)- not MSIH CDH1 (p.L741Yfs*29) TP53 (p.K120Sfs*3) BRCA2 (partial gene deletion)	No
	Post	TP53 (p.K120Sfs*3) MSH6 VUS (p.E807K) CDH1 (p.L741Yfs*29)	No
24	Pre	ARID1A (p.A2020E)- VUS	No
	Post	No actionable mutations

Discussion

This trial examined the safety and activity of 12 weeks of intense neoadjuvant androgen blockade combined with indomethacin to inhibit AKR1C3 activity prior to RP for high-risk prostate cancer. The primary objective was to assess pathologic complete response rate at time of surgery. Results demonstrate that pCR rate was low but MRD rate, as defined by tumor volume corrected for cellularity, was modest. Additionally, a significant treatment effect was seen in all prostatectomy tissue consisting of diffuse atrophy of benign glands, variable atrophy of cancer glands, areas of cancer cell hypocellularity, variably prominent cytological changes (clear tumor cell cytoplasm and small nucleoli), and foci of lacunae. Lacunae, hypothetically, are regions in the prostate where pre-existing cancer cells were lysed and resulting necrotic debris was metabolized.

A central hypothesis of this trial was that indomethacin, when added to conventional anti-androgen therapies, would inhibit steroidogenic enzyme AKR1C3. Because AKR1C3 catalyzes the conversion of several upstream androgen precursors to potent androgens, we hypothesized that AKR1C3 inhibition would lower intraprostatic levels of DHT and testosterone. If AKR1C3 is effectively inhibited, increased upstream hormone DHEA would be expected, with decreased downstream androgens T and DHT. Although direct cross-trial comparison is not possible, in a previous neoadjuvant trial (NeoAbi), after 12 weeks of neoadjuvant AAP and leuprolide, the intraprostatic DHEA level was 0.455 pg/mg and in our trial DHEA level was 7.72 pg/mg. Intraprostatic T and DHT levels in both studies were too low for meaningful comparison, but our study’s higher DHEA is consistent with the possibility that indomethacin may inhibit intraprostatic AKR1C3; however, randomized trials are needed to confirm this hypothesis.

With combinatorial AR targeted therapy along with AKR1C3 inhibition, pathologic response rates were modest, with 90% of men having ypT3 stage at time of RP, and 35% of men having lymph node involvement. This suggests that despite effective reduction in serum and tissue androgens, as well as effective AR blockade, other drug resistance mechanisms may persist. To explore genomic features that could associate with de novo resistance to AR-signaling inhibition, next-generation sequencing was performed. Pathogenic alterations were generally reflective of the high-risk population included in this study, with the most commonly mutated gene being TP53 in 6 of 15 (43%) of men, and were similar to those reported elsewhere²⁵. Although sample size is too small to make definitive conclusions, and these analyses are intended to be hypothesis-generating, two patients had mutations in SPOP and both had pathologic response. This is consistent with recent reports from a phase II study of AAP, Apa, and leuprolide for 6 months prior to prostatectomy, which also demonstrated that SPOP mutations correlated with response⁹. No major differences appeared in pathogenic alterations between pre- and post-treatment tissue, which is in contrast to a previous report that neoadjuvant therapy may select for genomic alterations that predispose to castration-resistant clones²⁶.

HSD3B1 polymorphisms were explored as a potential resistance mechanism. HSD3B1 encodes the enzyme 3β-hydroxysteroid dehydrogenase-1 (3βHSD1) which has effects on both androgen synthesis and abiraterone metabolism. The HSD3B1(1245C) allelic variant produces a more stable enzyme leading to an increase in extra-gonadal testosterone and DHT synthesis and is associated with worse response to ADT and inferior metastasis free survival in PC^23,27–29. Given the multifaceted potential effects of this genotype, an exploratory aim of our study was the impact of HSD3B1 genotype on pathologic outcome, hormone levels, and progression free survival (PFS). No difference occurred in pathologic response or PSA PFS when stratified by genotype. While recent prospective data suggests that patients with this allele may have worse overall survival and decreased time to castration resistance, our study may not have long enough follow-up to reflect these outcomes²⁹. In our dataset, this genotype lacked any correlation with tissue hormone or abiraterone levels at the time of RP. This might indicate that, (1) at least in the short term, treatment with intense anti-androgen blockade may be enough to overcome the metabolic differences conveyed by this allele; or (2) our population had no patients homozygous for this allele, who are known to have worse outcomes with abiraterone treatment³⁰.

Pathologic response has not been validated in PC as a surrogate endpoint for disease recurrence or overall survival although it is well established in other cancer types. In a pooled analysis of 72 patients enrolled in 3 neoadjuvant androgen-deprivation therapy trials prior to RP, pathologic response was associated with decreased rate of PSA relapse at a median follow up of 3.4 years³¹. This included 4 patients with a pathologic complete response and 7 with MRD, none of whom experienced PSA relapse. In the current study, fewer patients who achieved pCR or MRD developed PSA relapse, but conclusions are limited by the short follow up time. Longer follow up and a larger sample size may be needed to determine whether this is significant.

Our study reports the first published clinical trial using indomethacin to reduce prostate cancer growth. Pre-clinical rationale for using indomethacin in PC is compelling and other groups are currently testing indomethacin as a repurposed PC investigational drug with results not yet reported (NCT02935205). The results from this trial do not support further use of indomethacin in the neoadjuvant setting for high-risk patients as there was not a strong signal that it provided additional benefit over intense androgen blockade alone. Several potent and specific AKR1C3 inhibitors have been reported or are in development, but none have yet demonstrated clinical effectiveness. To our knowledge, only one of these agents has been reported in clinical trials for CRPC. In a small Phase I-II clinical trial, the oral compound ASP9521 was shown to be safe and well tolerated, but without evidence of clinical activity³².

Limitations of this trial include: (1) its small sample size and single center, making broad comparisons difficult; (2) the definition of MRD, while described in the literature, is not yet a validated endpoint in PC⁸; (3) its 12 weeks of neoadjuvant therapy, when some studies demonstrate an improved response with longer therapy, which may impact pathologic outcomes⁸; (4) possible variability in the interval of post-RP surveillance, although we followed our standard practice; and (5) this trial was not randomized, so definitive conclusions concerning the effects of indomethacin on hormone metabolism cannot be determined. While we inferred indomethacin’s effect on AKR1C3 based on upstream and downstream hormone measurements, these are indirect measurements so we cannot draw conclusions about indomethacin’s penetration and pharmacodynamic effect on AKR1C3.

Conclusions

Our data contribute to existing literature that neoadjuvant therapy may have favorable pathologic and clinical outcomes in a subset of high-risk prostate cancer. Phase III clinical trials are ongoing to further investigate this concept (e.g. NCT00430183). Longer follow up is needed to validate pathologic response as a surrogate endpoint for survival. Indomethacin did not appear to add significant benefit as a neoadjuvant strategy. While indirect hormone comparisons suggested it may have attenuated activity of AKR1C3, further study is warranted to determine if more potent inhibition of AKR1C3 would be safe and efficacious.

Supplementary Material

FIG S1

Supplementary Figure 1. CONSORT diagram. Newly diagnosed patients with high or very high risk prostate cancer were given 12 weeks of abiraterone acetate (AA) plus prednisone, apalutamide (Apa), luteinizing hormone-releasing hormone antagonist, and indomethacin. Patients who completed therapy underwent prostatectomy and were included in the final efficacy analysis.

FIG S2

Supplementary Figure 2. Serum Abiraterone and Metabolite Levels Post-Treatment. Box plots (median, interquartile range, and range) are shown.

TABLE S1

Supplementary Table 1. Pathologic characteristics at time of prostatectomy

FIG S3

Supplementary Figure 3. Prostate Tissue Abiraterone and Metabolite Levels at Time of Prostatectomy. Box plots (median, interquartile range, and range) are shown.