Fetal bovine serum (FBS) and trypsin (0.25% w/v) were purchased from Hyclone (Thermo Fisher Scientific Inc. Rockford, IL). Topotecan (TOPO) was purchased from 21st Century Global E-Commerce Network (East Sussex, UK). Dimethyl sulfoxide (DMSO), sulforhodamine B (SRB), TRIS buffer, acetic acid, ECL western blotting substrate for chemiluminescence were obtained from Bio-Rad (Hercules CA, United States ). 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) and RNase A were purchased from Sigma-Aldrich Inc (St. Louis, MO). Mouse anti-human antibodies (MMP-9, MMP-1, Ang-2, VEGF, Maspin and c-Fos were purchased from Santa Cruz Biotechnology (uPA; H77A10, MMP-1; SB12e, Ang-2; F-1, VEGF; JH121, Maspin; E10 Santa Cruz, CA) and Cell Signaling Technology (PAI-1; D9C4, uPAR; D7X2N, MMP-9; D6O3H, c-FOs; 9F6, Danvers, MA). Goat anti-Rabbit IgG (H = L) Secondary antibody, HRP, was purchased from Cell Signaling Technology; 345,897 (Danvers, MA). β-actin was purchased from Sigma-Aldrich; A5316. All glass and plasticware were purchased from VWR (Radnor, PA).

Human PCa cell lines (PC-3, PC-3M, and DU145 representing metastatic, castration-resistant with nonendocrine differentiation (NEPC) (Yuan et al., 2007; Sun et al., 2009) and (LNCaP, C4-2B, VCaP, LaPC4, MDA-PCa-2b and 22Rv1 representing androgen-dependent or castration-sensitive (Hooker et al., 2019) were obtained from American Type Culture Collection (ATCC; Rockville, MD). DUTXR (DU145 taxanes resistance) was gifted by our collaborator Dr. Amit Kumar Mitra, Auburn University (Takeda et al., 2007). The cell lines were authenticated at source and tested for mycoplasma contamination. All cell lines are mycoplasma negative, and the other lines show consistent patterns. PC-3, PC-3M and MDA-PCa-2b cells were maintained in 10% (v/v) fetal bovine serum (FBS) supplemented in F-12K, DU145 in Eagle’s Minimum Essential Medium (EMEM), C4-2B, VCaP, and LaPC4 in Dulbecco’s Modified Eagle’s Medium (DMEM), LNCaP, 22RV1 and DUTXR in RPMI-1640 media with 10% FBS. All cells were maintained at 37°C, 21% O₂ and 5% CO₂ in a humidified cell culture chamber (Heracell™ VIOS 160i CO₂; Thermo Scientific™). The RC77, RC165, RC43 cell lines were obtained from Dr. Yates; cell lines were developed from patients that self-identified as African American (Theodore et al., 2014). Cells were cultured in keratinocyte serum-free medium (KGM, Life Technologies, Carlsbad, CA).

RNA-seq data from prostate cancer patient samples was obtained from Dr. Clayton Yates. Details on patient samples have been analyzed and published previously (Theodore et al., 2014). Briefly, AA and CA patient tissue biopsies were obtained along with patient-matched normal adjacent tissues and used for gene expression analysis. No significant differences were detected between the two groups with respect to tumor content. As mentioned in our earlier publication, Institutional Review Board approvals were obtained from the Institutional Review Boards of Tuskegee University (IRB 00001137), St. Frances Hospital and Medical Center, Hartford CT, and UAB (Theodore et al., 2014).

PCa cell lines were seeded at 4 × 10³ cells in 96 well plates with appropriate media with 10% (v/v) FBS supplemented (Fontana et al., 2009). Plates were incubated for 24 h before media change and replacement with serum-supplemented media containing TOPO (0.04–10,000 nM). For CONV treatments, plates were then incubated at the same conditions for an additional 48 or 72 h. Whereas, for METRO treatment media was removed at 24 and 48 h after initial dosing, and cells were exposed to freshly prepared serum-supplemented media containing TOPO (0.04–10,000 nM). The media for control cells was changed daily and did not contain drug (Fontana et al., 2009). All experiments were performed in triplicate with five replicate wells for each concentration. Cell growth and cytotoxicity was assessed at each time point.

The effect of TOPO treatment on prostate cell growth and cytotoxicity was conducted by measuring total protein (SRB), mitochondrial enzyme activity (MTT), and examination of cellular morphology. SRB and MTT staining were performed at 48 and 72 h post initial TOPO exposure as described previously (Fontana et al., 2009). Absorbance was measured at 490 nm using a Synergy 3 Multi-Mode Microplate Reader (BioTek, Winooski, VT). The effect of drug exposure was determined by constructing cytotoxicity (growth) curves (n = 5 wells/group), percent change relative to untreated controls was calculated at each drug concentration. Half-maximal inhibitory drug concentration (IC₅₀) values were estimated by nonlinear regression using a sigmoidal dose-response equation (variable slope - three parameters).

Caspase-3/7 activation was measured using the Caspase-Glo 3/7 assay (Caspase-Glo^® 3/7 Assay - Promega Corporation) according to the manufacturer’s protocol. Briefly, a total of 2×10⁴ cells/well were seeded into 96-well plates (triplicates) and treated at the estimated 72 h MTT IC₅₀ (METRO and CONV). Caspase-Glo 3/7 reagent was added and incubated for 2 h, and luminescence was measured using a Synergy 3 plate reader. We normalized the level of apoptosis in each treatment group with the control group (no drug treatment or baseline caspase 3/7 assay luminescence) for each cell line.

Cell morphology was assessed using phase-contrast microscopy. PC-3 and LNCaP cells were seeded (4 × 10³) in 6-well plates and exposed to TOPO at the estimated IC₅₀ for 72 h (MTT) for each treatment protocol (CONV and METRO). Three areas with approximately equal cell densities were identified in each well, and images were captured with a Nikon AZ100 stereo-fluorescent microscope mounted with a Nikon Digital Sight DS-QiMc camera utilizing NIS-Elements image analysis software (Nikon, Melville, NY). Images were recorded in the modes of bright field and phase contrast at ×20 and ×40 magnifications. Further, images were analyzed using ImageJ (NIH, v. 1.46j, imagej.nih.gov).

Scratch assays were performed by creating a “scratch” in a cell monolayer followed by capturing the images at the beginning and regular interval (0, 48 and 72 h for METRO and CONV treatments) during cell migration to close the scratch and comparing the images to quantify the migration rate of the cells. Briefly, PC-3 cells were plated in 6-well plates at 1 × 10⁵ cells/well and incubated for 48 h to 90% confluency. The monolayer was scratched with a p200 pipette tip at the center of the well. F-12K culture medium, supplemented with 10% FBS, containing the vehicle (0.5% DMSO) was added to the cells in the control wells, and micrographs obtained at 0, 48, and 72 h. The effect of CONV and METRO dosing of TOPO, at their respective IC₅₀, were applied to the cells in the respective wells, and micrographs of the wound areas were obtained at 0, 48, and 72 h using an EVOS FL digital cell imaging system (Thermo Fisher Scientific, Inc.). Images were recorded in the modes of bright field and phase contrast at ×20 and ×40 magnifications. The initial wound area (at 0 h) and the “gap area” were measured at 48 and 72 h with ImageJ software.

Total RNA was isolated from cultured cells and patient samples using standard RNA extraction kits (RNeasy Kits–QIAGEN). RNA concentration and integrity were estimated by a NanoDrop 2000 UV-Vis spectrophotometer (Thermo Scientific, United States ), Qubit^® 2.0 Fluorometer (Invitrogen, Carlsbad, CA, United States ), and Agilent 2,100 Bioanalyzer (Applied Biosystems, Carlsbad, CA, United States ).

Gene expression in PC-3 and LNCaP cells were assessed at the calculated IC₅₀ of 48 and 72 h TOPO treatment (CONV vs METRO) using the RT² Profiler™ PCR Array Human Cancer Pathway Finder (PAHS-033A). The qRT-PCR analysis was performed using a LightCycler^® 480 II Real-Time PCR System (Roche Applied Science, Indianapolis, IN). Three housekeeping genes (GAPDH, ACTB, HGDC), reverse transcription control (RTC - Control RNA + primer) and positive PCR control (PPC) were used for assessing PCR efficacy. Gene expression was calculated by ΔΔCt method. Gene expression in treatment groups were normalized to corresponding controls (no drug treatment) and differentially expressed genes (DEGs) were identified.

Details on NGS-based RNA sequencing and data availability have been published earlier (Theodore et al., 2014). Briefly, Total RNA from patients and cell lines were taken and ribosomal RNA was removed with Ribo-Zero™ Gold kits (Epicenter, Madison, WI, United States ) by the manufacturer’s recommended protocol. Then, the RNA was fragmented and primed for the first-strand synthesis using the NEBnext First-Strand synthesis module (New England BioLabs Inc. Ipswich, MA, United States). The second-strand synthesis was performed with the NEBnext Second Strand synthesis module. Following this, the samples were added to a standard library preparation protocol using a NEBNext^® DNA Library Prep MasterMix Set for Illumina (Theodore et al., 2014). Library quantity and quality were assessed, and paired-end RNA sequencing was performed with an Illumina HiSeq2500 sequencer (Illumina, Inc. San Diego, CA, United States).

IPA (QIAGEN) was used to identify the most significantly affected 1) canonical pathways predicted to be activated or inhibited; 2) upstream regulator molecules like transcription factors, may be causing the observed gene expression changes; 3) downstream effects and biological processes that are increased or decreased; 4) predicted causal networks from our differential mRNA expression analysis (Krämer et al., 2014).

Processed, high-quality RNAseq data PCa patients were obtained from Dr. Clayton Yates’ lab (Tuskegee University) (Theodore et al., 2014). Details on RNA sequencing data availability have also been published earlier (Theodore et al., 2014). These reads were aligned to the reference genome (GRCh37 assembly) using HISAT2 (version 2.0.4), assembled into potential transcripts, and gene expression levels were quantified using Stringtie (version 1.3.3).

The mRNA expression of PCa patients was extracted from The Cancer Genome Atlas (TCGA) Data Portal (cancergenome.nih.gov). 717 PCa patients (500 prostate adenocarcinoma) and (217 cases from the Foundation Medicine Adult Cancer Clinical Dataset) had mRNA expression data available. R-based pipeline was created, and UALCAN, an interactive web-portal was used to perform data download, processing, and in-depth analysis of gene expression data files from TCGA’s Genomic Data Commons (GDCs) server and retrieve transcriptome data on target candidate pathway genes from the prostate expression data matrix containing >60,000 transcripts (Chandrashekar et al., 2017).

PCa cells were seeded and exposed to TOPO at the estimated 72 h MTT IC₅₀ for each treatment protocol (CONV and METRO). Post-treatment cells were lysed in cell lysis buffer (Thermo Scientific™ RIPA Buffer). Equal amount of protein loaded onto 4–15% Criterion™ TGX Stain-Free™ Precast Gels. Proteins will be separated under reducing conditions and then transferred to a PVDF membrane using a Trans-Blot Turbo Mini transfer pack from Bio-Rad (Hercules CA, United States ). Nonspecific binding was limited by incubating the membrane in blocking buffer (2.5% (w/v) casein, pH 7.6, 150 mM NaCl, 10 mM TRIS-HCl, and 0.02% sodium azide). Membranes were incubated with primary antibodies for target protein (1:1000) for overnight and then with the appropriate secondary antibody (1:10,000) at room temperature. Immunoreactivity was detected by Pierce ECL Western Blotting substrate (Bio-Rad, CA). Images were captured and quantify by Gel Doc™ EZ Gel Documentation System and ImageLab™ Software (Hercules CA, United States ). Densitometry analysis was performed using standard image analysis software ImageJ.

All statistical analyses were performed using R statistical programming package (version 3.6.2) and GraphPad Prism (version 7.0). Gene expression profiling (GEP) data were analyzed using Partek Genomics Suite v7.0 software to identify DEGs. Owing to the small sample size, we use limma, an empirical Bayesian method, to detect the DEGs and obtain p values and further provided False discovery rate (FDR) based on the p-value using Benjamini–Hochberg procedure. All tests were a two-sided and p < 0.05 was considered statistically significant. For analysis of RNAseq data obtained from EA/AA patient cohort, FDR adjusted p-values were used to determine significant associations.

The effect of TOPO administration as CONV and METRO on prostate cancer cells were assessed by measurement of MTT (Figures 1A,B) and SRB (Supplementary Figure S1A) absorbance after 48 h and 72 h treatments. Briefly, for CONV treatment, cells were treated with increasing concentrations of (0.04–10,000 nM) at 24 h following incubation, while for METRO, cells were exposed to freshly prepared serum-supplemented media containing TOPO (0.04–10,000 nM) at 24 and 48 h after initial dosing, and cells were exposed to freshly prepared serum-supplemented media containing TOPO (0.04–10,000 nM (Aljuffali et al., 2011). METRO exposure to TOPO in metastatic, androgen-insensitive, NEPC (EA: PC-3M, DU145 and DUTXR) and metastatic, androgen-sensitive (EA: C4-2B, 22RV1and AA: MDA-Pca-2b) cells resulted in significant (p ≤ 0.05) time-dependent decreases in the IC₅₀ value, which reflects increased sensitivity after 48 and 72 h compared to CONV treatment (Table 1). Comparison to CONV, METRO dosing for 72 h decreased the IC₅₀ for MTT by 9.6, 4.0, 9.7, 6.2, 6.4, and 10.7- fold in PC-3M, DU145, DUTXR, 22RV1, C4-2B and MDA-PCa-2b, respectively, as shown in Table 1. Further, TOPO METRO was more potent in AA cell lines (MDA-Pca-2b) by decreasing the IC₅₀ by 11-fold (165.93 vs 15.47 nM by MTT) after 72 h. Similar results were observed after 48 and 72 h of treatment and by the SRB assay as shown by a Supplementary Figure S1A).

Schedule-dependent (METRO vs CONV) effects on apoptosis were determined by measuring caspase-3/7 activity following treatment. METRO dosing of TOPO induced apoptosis in every cell line (PC-3–2.5, PC-3M-5.9, DU145–1.4 and 22RV1 1.6-fold) tested as compared to cells with CONV treatment (Figure 1C). Consistent with results from growth assays, apoptosis assay showed METRO TOPO to be a more effective schedule as compared with CONV TOPO.

Cell morphology studies, as shown in Supplementary Figure 1B, confirmed the differences in the IC₅₀ values (MTT, and SRB staining) as well as apoptosis assays. PC-3 and LNCaP cells exposed to METRO dosing regimens showed greater decreases in cellular density (3.6 and 6.2-fold) compared with CONV dosing.

PC-3 cells were exposed to TOPO CONV and METRO IC₅₀ at 48 and 72 h to examine the effect of different dose regimen on cell migration as measured using the scratch/wound healing assay. Results revealed that METRO-TOPO treatment had a greater effect (∼2-fold) in reducing cell migration in prostate cancer cell lines (p < 0.05) (Figure 1D).

Next, we performed targeted gene expression profiling (GEP) to identify and compare changes in gene expression following TOPO METRO vs CONV treatments in androgen-independent vs androgen-sensitive cell lines. GEP data from each treatment group and treatment time point was normalized to baseline gene expression for the corresponding cell line and heatmaps were generated (Figure 2A). Top differentially expressed genes/DEGs were identified for METRO vs CONV TOPO treatments in PC-3 and LNCaP at 48 (Supplementary Table S1) and 72 h (Table 2).

As shown in Table 2, the top 4 DEGs between pre- (untreated; 0 h) vs post-treatment (72 h), irrespective of androgen-sensitivity or treatment type (CONV vs METRO) were CDKN1A, FOS, TNF and SERPINB5 (Supplementary Figure S2). On the other hand, for METRO treatment only, a total of 34 genes were differentially expressed at 72 h (p < 0.05, fold-change>2). Among these, 17 were unique to PC-3, while 10 were unique to LNCaP. Seven 7) of these DEGs were common to both cell lines signifying a METRO-TOPO-specific GEP signature for PCa (Figure 2B I). Ovrall, the GEPs between treatments (METRO and CONV) identified eleven (n = 11) candidate biomarkers (Figure 2B II). The top five genes were SERPINB5, CDKN1A, TNF, FOS, and ANGPT1 (Table-2).

IPA analysis performed based on the DEGs associated with METRO-TOPO compared to CONV treatment predicted inhibition of tumor progression and angiogenesis as the key pathways for METRO-TOPO. This involved downregulation of SERPINB5, S100A4, MMP9, FOS, CDKN1A, TNF, and upregulation of, PLAU, CASP8, TNFRSF25 and BCL2 (LNCaP) (Figure 3A). Additionally, IPA predicted greater anti-angiogenic activity in PC-3 compared to LNCaP cell line for TOPO METRO treatment (Figure 3B). MMP1 was predicted by IPA as one of the top METRO-TOPO treatment ‘upstream regulators’ for LNCaP and PC-3 cell lines based on differential regulation of many important genes (BAX, TNF, SERPINEB5, S10000A4, PLAU, MMP9, FOS, CXCL8, CDKN1A) which in turn were predicted to result in dysregulation of cell proliferation, migration, metastasis, invasion and angiogenesis. Together, these may explain the potency of METRO-TOPO treatment in PCa (Supplementary Figure 3D). Interestingly, although TOPO is not an FDA approved drug for PCa, IPA analysis predicted PCa signaling (p = 1.40E-14) and PTEN (p = 1.53E-13) signaling pathways as targets for METRO TOPO treatment (Supplementary Table S2). Further, IPA also predicted small cell lung cancer as a top disease target for TOPO treatment, for which it is already an FDA-approved drug (p = 7.55E-17). IL-8 and p53 signaling pathways were also predicted as targets for METRO-TOPO uniquely in LNCaP cells (Supplementary Figure S3A–C).

Next, we assessed the top DEGs for METRO-TOPO using GEP data on PCa patient tumors. FOS, B2M, CDKNA1, and MMP1 were differentially expressed (Figure 4A). The expression of CDKNA1 and MMP1 were lower, while B2M was expressed at a higher level in AA compared to EA patients. Further, ITGA1, and ITGB3, which were unique for METRO-TOPO in metastatic, androgen-insensitive, NEPC had a significantly lower expression in AA, while PLAUR expression was higher in AA compared to EA PCa patients (p ≤ 0.05) (Figure 4B). THBS1—another DEG unique to metastatic, androgen-insensitive, NEPCs for METRO-TOPO, was also significantly downregulated in AA vs EA PCa patient cohorts (p ≤ 0.01) (Figure 4C).

We performed in silico analysis of the top DEGs using patient GEP data available in the TCGA database. We explored the following groups: 1) All men with PCa; 2) AA vs EA men with PCa and 3) high vs low survivors. Differential gene expression between high vs low survival PCa subgroups is depicted by the heatmap in Figure 5A. Results showed that the top DEGs for METRO-TOPO in PCa cell lines models, including SERPINB5, CDKN2A, and MMP9, were also correlated significantly (p < 0.05) with TCGA PCa patient’s survival. Additionally, we used TCGA databases to derive K-means clustering to show the top genes expression profile that are beneficially associated with prostate cancer patient survival. Kaplan-Meier plots were derived following K-means clustering (class = 2), showing stratification in survival among TCGA PCa patient clusters based on the top DEGs for METRO-TOPO treatment (Figure 5B II). Among these, the downregulation of MMP9, FOS, and SERPINB5 was correlated significantly (p < 0.05) with high patient survival (Supplementary Table S4). Further, expression of these top genes was also associated with patient survival specific to AA (Supplementary Table S3). Finally, TOPO METRO specific genes - MMP1, B2M, CXLC8, VEGF, PDGFA, SERPINE1, ERBB2, ITGA1, ITGA3, and JUN were expressed differentially between EA vs AA PCa patients (Figure 5C).

Our TCGA analysis demonstrated that several in vitro METRO-TOPO gene signatures including SERPINB5, MMP9, and FOS were correlated with patient survival and varied significantly (p < 0.05) between AA vs EA PCa patients. However, this analysis was based on one AA PCa cell line and limited number of AA patients within the TGCA. To overcome this limitation, we further compared our signature with gene expression profiles from patient derived PCa cell lines from AA (RC77, RC165, RC43) and EA (LNCaP, VCaP, LaPC4). In untreated cells, we observed increased expression of SERPINB5, CDKN1A, MMP9, ITGA3, and S100A4 in AA cell lines compared to EA cell lines (Figure 6A). Further, the METRO-TOPO DEGs unique to PC-3 were PLAUR and S100A4, and LNCaP were JUN, BCL2L1, and THBS1, were also found to be expressed at enhanced levels in AA compared to EA-derived cell lines (p ≤ 0.05) (Figures 6B,C).

Based on our expression data and in silico analyses, we confirmed protein expression of the top METRO-METRO DEGs via immunoblotting in PCa cell lines. We observed 1) SERPINEB5 and SERPINE1 were downregulated (RNA-seq) in METRO-TOPO; 2) FOS. iii) MMP1 and MMP9—identified as the most important genes by IPA; and 4) ANGPT2 and VEGF. Western blotting results show that protein expression of these top DEGs were concurrently higher in castration-resistant PC-3, and PC-3M compared to hormone-sensitive LNCaP, and 22RV1 (Figure 7). Furthermore, consistently greater downregulation of these top DEGs was observed following METRO-TOPO treatment compared to CONV-TOPO treatments. Figures 7A–E shows that densitometry plots for the proteins encoded by SERPINB5, SERPINE1, FOS, ANG-2, VEGF, MMP1, and MMP9 showed significant differential expression between METRO-TOPO vs CONV-TOPO dosing in PCa cell lines (p-values * = p ≤ 0.05, ** = p ≤ 0.01, and *** = p ≤ 0.001). Further, SERPINE1 (-29.6), FOS (-2.25), ANG-2 (-3.18), VEGF (-3.8), and MMP1 (-13.3) were downregulated the most in PC-3 while SERPINE5 (-3.43) and MMP9 (-21.8) were downregulated most in the LNCaP line. In the PC-3M cell line, which is considered an aggressive metastatic mCRPC/NEPC model (Ohnuki et al., 1980), SERPINEB5, ANG-2, VEGF, MMP1, and MMP9 were among the downregulated genes post-METRO-TOPO treatment (Figures 7A–E). Downregulation of proteins representing the top DEGs was also observed in 22Rv1 and DU145 cell lines; however, the fold changes were lower in comparison (Figures 7B–E and Supplementary Table S4).