Circulating Lipid Profiles Associated With Resistance to Androgen Deprivation Therapy in Localized Prostate Cancer

[podsart]

Circulating Lipid Profiles Associated With Resistance to Androgen Deprivation Therapy in Localized Prostate Cancer

Authors: Hui-Ming Lin, PhD orcid.org/0000-0003-4892-6008, Xiaoyu Yang, MS orcid.org/0009-0001-0035-9289, Margaret M. Centenera, PhD orcid.org/0000-0002-2206-0632, Kevin Huynh, PhD orcid.org/0000-0001-6170-2207, Corey Giles, PhD orcid.org/0000-0002-6050-1259, Jonas Dehairs, PhD, Johannes V. Swinnen, PhD orcid.org/0000-0002-7720-5077, Andrew J. Hoy, PhD orcid.org/0000-0003-3922-1137, Peter J. Meikle, PhD orcid.org/0000-0002-2593-4665, Lisa M. Butler, PhD orcid.org/0000-0003-2698-3220, Mary-Ellen Taplin, MD, and Lisa G. Horvath, MBBS, PhD, FRACP orcid.org/0000-0001-6842-9223 [e-mail] Info & Affiliations

Publication: JCO Precision Oncology

Volume 8

doi.org/10.1200/PO.24.00260

Intense androgen deprivation therapy (ADT) with androgen receptor pathway inhibitors (ARPIs) before radical prostatectomy (RP) produced favorable pathologic responses in approximately 20% of patients. The molecular reason for the low rate of response remains unclear. Lipid metabolism is known to influence androgen receptor signaling and ARPI efficacy. The aim of the study was to identify circulating lipid profiles associated with ADT/ARPI resistance in localized prostate cancer.

Materials and Methods

Two independent experimental approaches were used. Experiment 1: Post hoc analysis of the association between plasma lipidomic profiles and ADT/ARPI response was performed on patients (n = 104) from two phase II trials of neoadjuvant ADT/ARPI. Response to ADT/ARPI was defined by pathologic response. Experiment 2: Patient-derived tumor explants from RP (n = 105) were cultured in enzalutamide for 48 hours. Explant response to enzalutamide was evaluated against pre-RP plasma lipidomic profiles (n = 105) and prostate tissue lipidomic profiles (n = 36). Response was defined by Ki67 (cell proliferation marker) fold difference between enzalutamide and vehicle-treated explants. In both experiments, associations between lipid profiles and ADT/ARPI response were analyzed by latent class analysis.

Results

Pretreatment plasma lipid profiles classified each experimental cohort into two groups with differences in ADT/ARPI response rates. The response rates of the groups were 9.6% versus 29% in experiment 1 (chi-squared test P = .012) and 49% versus 70% in experiment 2 (chi-squared test P = .037). In both experiments, the group with a higher incidence of ADT/ARPI resistance had higher plasma levels of sphingomyelin, glycosylceramides, free fatty acids, acylcarnitines, cholesterol esters, and alkyl/alkenyl-phosphatidylcholine and lower plasma levels of triacylglycerols, diacylglycerols, and phosphoethanolamine (t-test P < .05).

Conclusion

Pretreatment circulating lipid profiles are associated with ADT/ARPI resistance in localized cancer in both human cohorts and explant models.

Introduction

Radical prostatectomy (RP) remains one of the major treatments for localized prostate cancer.1 However, recurrence occurs in a subset of patients. The 15-year disease-specific mortality rate after RP is approximately 7% for all stages of prostate cancer, but up to 30% for high-grade disease.2 Thus, new treatment strategies are needed to improve the cure rate for high-risk localized prostate cancer.

Context

Key Objective

Are circulating lipids associated with intrinsic resistance to antiandrogen therapy in localized prostate cancer?

Knowledge Generated

Pretreatment plasma lipid profiles of patients with localized prostate cancer were associated with resistance to intensive neoadjuvant antiandrogen therapy. Similar pretreatment lipid profiles were also associated with antiandrogen resistance of cancer cells in ex vivo treatment of patient-derived prostate tumor tissue.

Relevance

The lipid profiles may be used to identify lipid biomarkers of response to neoadjuvant antiandrogen therapy and new therapeutic targets to overcome antiandrogen resistance.

Phase II randomized trials have demonstrated that androgen deprivation therapy combined with androgen receptor pathway inhibitors (ADT/ARPI) before RP can produce exceptional pathologic responses (residual tumor ≤5 mm) in 17%-30% of patients.3-7 Biochemical recurrence was lower among exceptional pathologic responders (8% v 94%).8 Ki-67 proliferation index, tumor PD-L1 expression, and androgen receptor (AR) nuclear expression were not associated with response.3,7 However, the role of dysregulated lipid metabolism in hormone resistance has not been examined in this setting.

The human plasma lipidome contains thousands of lipid species, which may provide insights into pathophysiologic processes.9 Lipid metabolism is linked to AR signaling and, in particular, resistance to ARPIs.10-12 For example, elevated circulating levels of ceramide, a type of sphingolipid, are associated with ARPI resistance in metastatic castration-resistant prostate cancer (mCRPC).13 Pharmacologic inhibition of ceramide conversion into proinflammatory and prosurvival sphingosine-1-phosphate (S1P) overcame enzalutamide (ENZ) resistance in cell lines and patient-derived explants of localized prostate cancer.13

The aim of our study was to determine if circulating lipid profiles are associated with ADT/ARPI resistance in localized prostate cancer. We conducted two independent studies involving patients undergoing RP for treatment of localized prostate cancer. In the first experiment, we performed a post hoc analysis assessing the association between plasma lipidomic profiles and pathologic response in a subset of patients from two phase II trials of neoadjuvant ADT/ARPI treatment. In the second experiment, we examined the association between pre-RP plasma lipidomic profiles and the response of the patients' matching tumor explants to ENZ treatment in an independent cohort.

Materials and Methods

Neoadjuvant ADT/ARPI Cohort

Plasma samples were sourced from two previous phase II randomized trials of neoadjuvant ADT/ARPI (six cycles, total of 24 weeks) followed by RP of prostate adenocarcinoma in the United States.3,7 These samples were combined together in this post hoc analysis for higher statistical power. In the first trial (ClinicalTrials.gov identifier: NCT02268175), patients received leuprolide as ADT (22.5 mg once every 12 weeks) and enzalutamide (160 mg once daily) as the ARPI, with or without another ARPI which is abiraterone (1,000 mg once daily with 5 mg of prednisone once daily) for six cycles (total of 24 weeks) followed by RP. In the second trial (ClinicalTrials.gov identifier: NCT02903368), patients received leuprolide (22.5 mg once every 12 weeks) and abiraterone (1,000 mg once daily with 5 mg prednisone twice daily) with or without another ARPI which is apalutamide (240 mg once daily) for six cycles (total of 24 weeks) followed by RP. Plasma samples were collected without previous fasting, at baseline before commencement of ADT/ARPI (pre-ADT/ARPI) and before cycle-4 (mid-ADT/ARPI). Blood was drawn into an EDTA or Streck tube, which was then centrifuged at 300×g for 20 minutes at 22°C. The plasma layer was transferred to a separate tube and centrifuged at 22°C at 5,000×g for 10 minutes to remove residual cells. The supernatant was aliquoted into cryovials and stored at minus 80°C. As reported previously,3,7 responders to ADT/ARPI were defined as those having pathologic complete response (PCR) or minimal residual disease (MRD), which is the largest cross-sectional dimension of residual tumor measuring ≤5 mm; nonresponders were those with any residual tumors >5 mm.

Explant Cohort

Fasting plasma samples were obtained before surgery from patients undergoing RP for prostate adenocarcinoma at St Andrews Hospital in Adelaide, Australia. Blood was drawn into an EDTA tube and centrifuged at 2,500 g for 10 minutes at room temperature to separate the plasma, which was then aliquoted and stored at minus 80°C. Tissue samples from the resected prostate were cultured with 10 μM ENZ (Selleckchem) or vehicle control (dimethyl sulfoxide) for 48 hours, and Ki67 immunostaining of baseline (uncultured) and explant tissues was performed as described previously.14 Response to ENZ was defined by the Ki67 fold difference between ENZ-treated and matching vehicle-treated explants (ie, % of cancer with Ki67 staining in ENZ-treated divided by % of cancer with Ki67 staining in vehicle-treated), categorized as follows—Resistant, ≥1 Ki67 fold difference; Sensitive, ≤0.6 Ki67 fold difference; and Stable, between 1 and 0.6 Ki67 fold difference. Sensitive and Stable were considered as ENZ responders, whereas Resistant was considered as nonresponders. For patients with duplicate explant treatments, the average of the fold difference of Ki67 positivity was calculated for categorization of response. Patients with necrotic/noncancer vehicle controls were excluded from statistical analyses as the Ki67 fold difference could not be calculated. Ethical approval for biospecimen collection and research was obtained from the Human Research Ethics committees of St Andrews Hospital (Approval No. 80) and the University of Adelaide (H-2012-016). All patients provided written informed consent.

Plasma Lipidomic Analysis

Lipids in plasma samples were profiled by liquid chromatography-mass spectrometry (LCMS)15 on separate occasions for the neoadjuvant and explant cohorts. Complete structural characterization was not possible for triacylglycerols (TG), which are reported as the sum of their fatty acyl composition (TG(SIM), single ion monitoring), or with an identified acyl chain (TG [NL], neutral loss). Thus, TG [NL] and TG(SIM) may represent single or multiple species of the same molecular formula. Total TG levels were calculated as the summed concentration of all TG(SIM) species. Further details on lipid nomenclature are available in the Data Supplement. Each plasma lipidomic data set was normalized by the Probabilistic Quotient Normalization (PBQ) method using the average of the samples within each data set as the reference, as described previously. Lipid concentrations were then log2-transformed (logarithm to the base 2) for statistical analyses. Replicate plasma samples (three patients only—two in duplicates, one in triplicate) were averaged before log-2 transformation.16

Tissue Lipidomic Analysis

Lipids in snap-frozen matching baseline (uncultured) and explant tissues were previously profiled using LCMS methodology that differed from that of plasma lipidomic analysis.14 The tissue samples were profiled as two separate batches. Missing values were replaced with half the minimum of the value for that lipid species in each data set. The lipid levels (nmol/mg DNA) from each batch were normalized using the PBQ method and log2-transformed. After normalization, both batches were aligned into a single data set using the ComBat algorithm (sva v3.46.0).

Statistical Analyses

Statistical analyses were performed using R software (v4.2.3) with specific R packages mentioned accordingly. P values <.05 were considered as statistically significant. Lipid levels are log2-transformed for analysis. Latent class analysis (LCA) was used to identify groups of patients with lipid profiles that are associated with pathologic response in the neoadjuvant cohort or Ki67 fold difference in the explant cohort (poLCA v1.6.0.1). LCA is a probabilistic method of unsupervised clustering on the basis of the assumption that the observed distribution of variables (lipid levels) is produced by the mixture of unobserved (latent) underlying distributions.17 In both cohorts, LCA was performed with lipids associated with response, which were identified by linear modeling with empirical Bayes adjustment for the neoadjuvant ADT/ARPI cohort (pathologic response against pre-ADT lipid levels, Data Supplement, Table S1; limma v3.54.2) and correlation analysis for the explant cohort (pre-RP lipid levels against log2 of the Ki67 fold difference between ENZ- and vehicle-treated explants, Data Supplement, Table S2; Hmisc v5.1-1). A value of 0.1 was added to any Ki67 fold differences of zero (caused by the absence of Ki67) to enable log2 transformation. Few lipids were significantly associated with pathologic response (10 lipids) or correlated with Ki67 fold difference (41 lipids) at the statistically significant threshold of P < .05. Therefore, a lower significance threshold of P < .5 was used to select lipids for LCA—373 lipids for the neoadjuvant cohort and 380 lipids for the explant cohort (Data Supplement, Tables S1 and S2), which were similar numbers to those of a previous study (323 lipids in the study by Lin et al16). Correlation analysis was used in the explant cohort to describe the relationship between lipid levels and Ki67 fold change, but linear modeling would have selected the same lipids for LCA with a difference of only four lipids. The lipid levels were categorized into quartiles for the LCA, and the optimum number of LCA classes was determined using the Bayesian Information Criteria. The proportion of responders among the LCA classes was assessed using the chi-squared test. Differences in lipid levels between LCA classes, according to individual lipid species or the sum of lipid species by their lipid class were determined using the t-test.

Results

Plasma Lipids Associated With Neoadjuvant ADT/ARPI Resistance

Plasma samples were obtained from two phase II trials3,7 of neoadjuvant ADT/ARPI followed by RP of prostate adenocarcinoma (Table 1, Fig 1A). These plasma samples consisted of baseline (pre-ADT/ARPI) samples from 104 patients and matching pre–cycle-4 (mid-ADT/ARPI) samples from 91 patients (Fig 1A). The levels of 811 lipid species from 46 lipid classes were measured in these plasma samples.

Table 1. Characteristics of the Neoadjuvant Androgen Deprivation Therapy/Androgen Receptor Pathway Inhibitor Cohort

Characteristic n = 104, No. (%)

Pathologic response

 None 84 (81)

 MRD (residual tumora >5 mm)

9 (8.7)

 PCR (residual tumora ≤5 mm)

11 (11)

Treatmenta

 Leuprolide, enzalutamide, abiraterone, prednisoneb

16 (15)

 Leuprolide, enzalutamideb

11 (11)

 Leuprolide, abiraterone, prednisone, apalutamidec

36 (35)

 Leuprolide, abiraterone, prednisonec

41 (39)

Gleason score

 7 28 (27)

 8 31 (30)

 9 43 (41)

 10 2 (1.9)

RP staging

 <T3 75 (73)

 ≥T3 28 (27)

 Unknown 1

Abbreviations: MRD, minimal residual disease; PCR, Pathologic complete response; RP, radical prostatectomy.

a

Largest cross-sectional dimension.

b

NCT02268175.

c

NCT02903368.

Fig 1. Flow of (A) neoadjuvant ADT/ARPI experiment and (B) explant experiment. ADT, androgen deprivation therapy; ARPI, androgen receptor pathway inhibitors.

The total pathologic response rate was 19.2% (responders), which consisted of 11 patients with PCR and 9 with MRD. LCA of the pre-ADT/ARPI levels of 373 lipids (linear modeling P < .5) classified the patients into two groups, N1 and N2, which have distinct lipid profiles (Fig 2A) and differences in pathologic response rate. N1 had a lower rate of responders (9.6%, 5 of 52) compared with N2 (29%, 15 of 52; chi-squared test P = .012), that is, the lipid profile of N1 is associated with a higher rate of ADT/ARPI resistance. The plasma levels of 380 lipid species and 22 lipid classes were different between N1 and N2 at both pre-ADT/ARPI and mid-ADT/ARPI (P < .05; Data Supplement, Tables S3 and S4).

Fig 2. Plasma lipid differences between latent class groups in neoadjuvant cohort ((A) pre-ADT/ARPI plasma; (B) mid-ADT/ARPI plasma) and explant cohort ((C) pre-RP plasma). Forest plots depict (left) fold differences in individual lipid species or (right) sum of lipids by lipid class; error bars represent 95% CIs. ADT, androgen deprivation therapy; ARPI, androgen receptor pathway inhibitors; RP, radical prostatectomy.

Plasma Lipids Associated With ENZ Resistance in Patient-Derived Explants

Explants of prostate tissue from an independent cohort of 125 patients who never received ADT/ARPI were treated with ENZ or vehicle control for 48 hours (Fig 1B). The levels of 823 lipid species from the same 46 lipid classes as the neoadjuvant cohort were measured in pre-RP plasma from 107 of these patients, where 808 lipid species were in common. Two patients were excluded from the data set (Table 2) as their control-treated explants were necrotic or lacked cancer cells.

Table 2. Characteristics of the Explant Cohort

Characteristic Plasma Lipidomic (n = 105), No. (%) Tissue Lipidomic (n = 36), No. (%)

Explant response, Ki67 fold differencea

 ≤0.6 (sensitive) 30 (29) 14 (39)

 0.6-1 (stable) 30 (29) 10 (28)

 >1 (resistant) 45 (43) 12 (33)

Gleason score

 6 3 (2.9) 2 (5.6)

 7 83 (81) 30 (83)

 8 6 (5.9) 2 (5.6)

 9 9 (8.8) 1 (2.8)

 10 1 (1.0) 1 (2.8)

 Unknown 3

RP staging

 <T3 31 (31) 13 (36)

 ≥T3 70 (69) 23 (64)

 Unknown 4

a

Ki67 fold difference = (% of cancer expressing Ki67 in enzalutamide-treated) ÷ (% of cancer expressing Ki67 in vehicle-treated).

Abbreviation: RP, radical prostatectomy.

The explants from 57% of the patients (60 of 105) were responsive to ENZ (ie, sensitive or stable) according to Ki67 (proliferation marker) fold difference between ENZ-treated and vehicle-treated explants (Fig 3A). LCA of pre-RP lipid levels of 380 lipids (Pearson correlation between pre-RP lipid levels and Ki67 log2 fold difference P < .5) classified the 105 patients into two groups, E1 and E2, with distinct lipid profiles (Fig 2C). E1 has a lower rate of ENZ-responsive explants (49%, 32 of 65) compared with E2 (70%, 28 of 40; chi-squared test P = .037), that is, E1 has a higher rate of ENZ-resistant explants. The plasma levels of 498 lipid species and 30 lipid classes were different between E1 and E2 (Fig 2C, Data Supplement, Tables S5 and S6).

Fig 3. Lipidomic analysis of explants: Waterfall plot of Ki67 fold difference between (A) ENZ-treated and vehicle-treated explants, (B) forest plots, and (C) heatmap of Pearson correlation coefficients between tissue and plasma lipid levels (calculation of lipid class sum includes unmatched lipid species between tissue and plasma); (D) scatterplots of tissue and plasma levels of sphingomyelin.

Interestingly, 320 of the differential plasma lipid species and 21 of the differential plasma lipid classes between E1 and E2 (Fig 2C) were similar to those that were differential between N1 and N2 of the neoadjuvant cohort at both pre- and mid-ADT/ARPI timepoints (Fig 4). When only considering pretreatment timepoints for both cohorts, 28 differential lipid classes were common. Lipid classes that were significantly higher in plasma levels among the patient groups with lower explant response rates in both experiments (E1, N1) and both timepoints (pre-ADT, mid-ADT) compared with their counterparts (E2, N2) included sphingomyelin (SM), the glycosphingolipids (HexCer, Hex2Cer, Hex3Cer, GM1, GM3), alkyl/alkenyl phosphatidylcholine and their lyso forms [PC(O), PC(P), LPC(O), LPC(P)], cholesterol esters, free fatty acids, and acylcarnitines (AC). The lipid classes that were significantly lower for those with lower explant response rates (E1, N1) compared with their counterparts with higher explant response rates (E2, N2) include deoxyceramides [Cer(m)], diacylglycerol (DG), and triacylglycerol [TG(SIM)].

Fig 4. Relative plasma levels of 320 lipid species that were differential between N1 versus N2 at (A and B) both pre- and mid-ADT/ARPI of the neoadjuvant cohort and also (C) between E1 versus E2 at pre-RP of the explant cohort. The 21 lipid classes that were differential in their total levels (sum of all detected lipid species) between the latent class groups in all three comparisons are marked with +. ADT, androgen deprivation therapy; ARPI, androgen receptor pathway inhibitors; RP, radical prostatectomy.

Tissue lipidomic data of baseline tissue specimens of the resected prostate (explant cohort) were available for 36 patients as part of another study using a different lipidomic methodology14 (Data Supplement, Fig S3). A total of 248 lipid species from nine lipid classes were profiled in tissue, of which only 38 lipid species and seven lipid classes overlapped with the plasma lipidomic data set. The pre-RP plasma levels of total SM and SM(d18:1/24:1) (the second most abundant SM species in plasma and tissue) correlated with their respective untreated baseline tissue levels (P < .05; Figs 4B-4D).

Discussion

To our knowledge, this is the first study to demonstrate that circulating lipid profiles are associated with ADT/ARPI resistance in both in vivo human cohorts and ex vivo models of hormone-sensitive prostate cancer. Our two different experimental approaches of studying ADT/ARPI resistance in localized prostate cancer indicate that pretreatment circulating lipid profiles are associated with intrinsic ADT/ARPI resistance. In each experiment, the pretreatment plasma lipid profiles classified the patients into two groups with different response rates to ADT/ARPI. Importantly, most of the differential circulating lipid classes at pre-ADT/ARPI were retained after three cycles of neoadjuvant ADT/ARPI treatment.

Circulating lipids represent a constant flux between tissue uptake and release. Changes in their circulating concentration may influence tissue metabolism or represent an abnormal metabolic state, causing local tissue metabolism to leak into circulation. Thus, it may be possible that circulating lipids influence ADT/ARPI response through their uptake by cancer cells or other cells (eg, immune cells) that can influence cancer cell biology. It is also possible that some of the circulating lipids are produced by the cancer cells and tumor metabolism.

In the neoadjuvant and explant cohorts of our current study, the patient groups with lower response rates to ADT/ARPI have higher plasma levels of the sphingolipids SM and glycosylceramides than the groups with higher response rates. The significant correlation between the tissue levels of total SM in untreated baseline prostate tumor tissue of the explant cohort with plasma SM levels raises the hypothesis that there is tumor uptake of circulating SM, which may be converted into S1P (sphingolipid metabolism pathway displayed in the Data Supplement, Fig S4). We hypothesize that elevated circulating levels of SM and glycosylceramides contribute to intrinsic ADT/ARPI resistance through their metabolism by the tumor into ceramide, S1P, and glycosphingolipids (Data Supplement, Fig S4). S1P can promote cell survival, proliferation, and inflammation and regulate steroid synthesis.18,19 Glycosphingolipids have been implicated in multidrug resistance20,21 and cancer growth.22,23 Elevated circulating SM and glycosylceramides were associated with disease progression in a cohort of early-stage prostate cancer undergoing active surveillance.22 In mCRPC, elevated circulating sphingolipids were associated with shorter radiologic progression-free survival with ENZ or abiraterone and shorter overall survival.13,16,24 Circulating ceramide levels were higher among patients with mCRPC with tumor AR aberrations (known contributors of androgen resistance) compared with those with normal copy number AR.11 Overall, these observations support the role of sphingolipids in promoting ADT/ARPI resistance and cancer cell survival, and targeting aberrant sphingolipid metabolism may overcome ADT/ARPI resistance.

A potential approach to target aberrant sphingolipid metabolism is to decrease circulating sphingolipids with statins and PCSK9 inhibitors.25-27 The combined use of statins with ADT was associated with improved overall survival and reduction in prostate cancer–specific mortality.28 Interestingly, we observed that a higher proportion of patients in the N2 group (more responders) of the neoadjuvant cohort were already on statin medication at the start of the trials compared with the N1 group (33% v 13%, chi-squared test P = .02). However, we are unable to conclude if statin medication contributed to the lower plasma levels of sphingolipids of the N2 group without controlled conditions (eg, length of treatment time and prestatin plasma levels are unknown).

Another approach to targeting aberrant sphingolipid metabolism is to inhibit glycosphingolipid synthesis in cancer cells with eliglustat, which is used in the treatment of Gaucher disease and shown to suppress the growth of prostate cancer xenografts.22 S1P synthesis can be inhibited with sphingosine kinase inhibitors such as opaganib, which overcame ENZ resistance in patient-derived explants of localized prostate cancer.13 Opaganib was well-tolerated in phase I studies for solid tumors and multiple myeloma.29,30

The development of new treatment strategies that targets lipid metabolism to overcome ADT/ARPI resistance also requires the refinement of the circulating lipid profiles into a robust, sensitive, and specific lipid biomarker that can identify resistant patients who will benefit from the targeted therapy. The lipid biomarker may also be used to identify patients who are more likely to respond to neoadjuvant ADT/ARPI and prevent the unnecessary treatment of nonresponders.

Limitations of the study include the small cohort sizes and the low number of tissue lipids profiled, which did not include glycosphingolipids. Thus, a more comprehensive analysis comparing tissue and plasma lipids could not be performed. Another limitation is that the source of the circulating lipids is unknown. It is not known if the lipids are produced by the cancer cells or by other tissues. There may be a genetic basis for the pretreatment circulating lipid levels such as somatic mutations in the tumor affecting lipid metabolism in the cancer cells or polymorphisms in lipid metabolism genes, which predisposes the patients to ADT/ARPI resistance. Alternatively, environmental factors such as pre-existing medical conditions, medications, or dietary factors may be involved. Further studies are warranted to validate the findings.

In conclusion, circulating lipid profiles are associated with ADT/ARPI resistance in both human cohorts and explant models in localized prostate cancer.

[pd63]

I don't know if one can copy and paste in to this forum if not I salute your typing skills.P.S. I haven't read it