Club‐like cells in proliferative inflammatory atrophy of the prostate

Club cells are a type of bronchiolar epithelial cell that serve a protective role in the lung and regenerate damaged lung epithelium. Single‐cell RNA sequencing (scRNA‐seq) of young adult human prostate and urethra identified cell populations in the prostatic urethra and collecting ducts similar in morphology and transcriptomic profile to lung club cells. We further identified club cell‐like epithelial cells by scRNA‐seq of prostate peripheral zone tissues. Here, we aimed to identify and spatially localize club cells in situ in the prostate, including in the peripheral zone. We performed chromogenic RNA in situ hybridization for five club cell markers (CP, LTF, MMP7, PIGR, SCGB1A1) in a series of (1) nondiseased organ donor prostate and (2) radical prostatectomy specimens from individuals with prostate cancer. We report that expression of club cell genes in the peripheral zone is associated with inflammation and limited to luminal epithelial cells classified as intermediate cells in proliferative inflammatory atrophy (PIA). Club‐like cells were enriched in radical prostatectomy specimens compared to nondiseased prostates and associated with high‐grade prostate cancer. We previously reported that luminal epithelial cells in PIA can rarely harbor oncogenic TMPRSS2:ERG (ERG+) gene fusions, and we now demonstrate that club cells are present in association with ERG+ PIA that is transitioning to early adenocarcinoma. Finally, prostate epithelial organoids derived from prostatectomy specimens demonstrate that club‐like epithelial cells can be established in organoids and are sensitive to anti‐androgen‐directed treatment in vitro in terms of decreased androgen signaling gene expression signatures compared to basal or hillock cells. Overall, our study identifies a population of club‐like cells in PIA and proposes that these cells play an analogous role to that of club cells in bronchiolar epithelium. Our results further suggest that inflammation drives lineage plasticity in the human prostate and that club cells in PIA may be prone to oncogenic transformation. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.


Introduction
Club cells (formerly called Clara cells) are bronchiolar exocrine cells present in narrow bronchioles in the lower respiratory tract. These cells consist of simple, cuboidal epithelium and function as the major secretory cell in the small airway epithelium [1]. Club cells are considered to serve a protective role to the lung, in part due to their secretion of antimicrobial and immunomodulatory proteins such as the anti-inflammatory component of pulmonary surfactant, Secretoglobin Family 1A Member 1 (SCGB1A1) [2]. Notably, club cells are regarded as the airway 'second stem cell' due to their regenerative potential for damaged lung epithelium [3,4]. In addition to host defense, single-cell transcriptome analysis of human small airway epithelium identified additional roles for club cells in the lung, including xenobiotic metabolism, antiprotease production, and physical barrier function [5].
Recently, single-cell RNA sequencing (scRNA-seq) of the young adult human prostate and urethra identified SCGB1A1-expressing prostate epithelial cells located in the urethra and proximal ductal trunks of the prostate that are similar in morphology and transcriptomic profile to lung club cells [6]. This study likewise identified hillock epithelial cells that are abundant in the prostatic urethra and collecting ducts, as well as the central zone surrounding the ejaculatory ducts [6]. Furthermore, in a study conducted in benign prostatic hyperplasia (BPH) tissues from patients undergoing treatment for prostate enlargement, luminal epithelial cells in 5-alpha reductase inhibitor (5ARI)-induced prostate atrophy were shown to exhibit increased expression of urethral club cellassociated genes, including lactoferrin (LTF), polymeric immunoglobulin receptor (PIGR), Olfactomedin 4 (OLFM4) and Secretoglobin Family 3A Member 1 (SCGB3A1) in addition to SCGB1A1 [7]. The authors conclude that luminal epithelial cells in hormonal therapyinduced prostatic atrophy undergo an adaptation from a prostate secretory luminal cell to a club cell-like state [7]. Finally, it has been suggested that the anatomical location and phenotype of prostate club cells indicate that they play a role analogous to that of pulmonary club cells, e.g. in acting in a protective, antimicrobial function [8]. Aside from their expansion in 5ARI-induced prostate atrophy, club cells are reported to be rare in the prostate apart from the prostatic urethra and collecting ducts [6].
Another recent scRNA-seq analysis conducted on prostate biopsies and radical prostatectomy specimens identified an epithelial cell subset that clustered separately from normal basal and luminal epithelial cells and that was characterized by the expression of PIGR, Matrix Metallopeptidase 7 (MMP7), Ceruloplasmin (CP), and LTF [9]. These cells also expressed SCGB1A1, and their gene expression profile was found to greatly overlap with previously published club cell gene signature scores [9]. These club cell-like epithelial cells were presumably located anatomically distant from the previously reported location of club cells in association with urothelium in normal, nondiseased prostate [6] and in the transition zone in BPH [7]. Rather, these cells were located in the peripheral zone of the prostate and in association with prostatic adenocarcinoma [9]. In the present study, we performed chromogenic RNA in situ hybridization (RISH) for club cell markers (CP, LTF, MMP7, PIGR, SCGB1A1) in a series of nondiseased organ donor prostate specimens and radical prostatectomy specimens from patients with prostate cancer to identify and spatially localize club cells in the prostate, inclusive of the peripheral zone. We further characterize club-like prostate epithelial cells using prostate epithelial organoids derived from prostate cancer specimens in relation to sensitivity to anti-androgen directed treatment in vitro.

Patient population and clinical samples
All specimens were acquired using protocols approved by the Johns Hopkins University and University of California, San Francisco Institutional Review Boards. Formalin-fixed paraffin-embedded (FFPE) whole tissue slides (e.g. standard tissue slides) from tissue blocks were obtained from organ donor prostates with no indication of cancer (n = 10) and from radical prostatectomy specimens (n = 40). One block containing the highest grade index tumor and adjacent benign was chosen from each prostatectomy case. Each slide contained tissue from the peripheral zone, with varying transition zone tissue present. The clinical characteristics of the patient samples are given in Table 1. The patient population for scRNA-seq analysis was previously described [9].  [10], the IGROV-1 cell line for MMP7 [11], human intestine for PIGR, and human lung tissues for SCGB1A1. Normal prostate epithelium was used as a negative control (supplementary material, Figure S1). Positive and negative controls for the LTF RISH assay were described previously [12].

Image analysis
Slides were scanned with a 40Â objective using a Ventana DP200 (Roche Diagnostics) digital whole slide scanner. Digital image analysis was carried out using HALO 3.1 software (Indica Labs, Albuquerque, NM, USA). Slides were annotated for cancer versus adjacent benign regions, and each region (cancer versus benign) was analyzed separately. The Indica Labs Area Quantification version 2.1.11 algorithm was adapted to analyze the positive staining in cancer and benign regions on each whole tissue section. Thresholds for the stain were set up with the real-time tune window. The same algorithm was used to calculate the percentage of brown staining-positive tissue per total tissue area across all whole tissue sections. Data were analyzed and graphed using GraphPad Prism Software version 9 (GraphPad Software, Inc., San Diego, CA, USA), and values were considered significant at p < 0.05. Any areas of prostatic urethra present on whole tissue sections were excluded from image analysis.

scRNA-seq analysis of prostate epithelial organoids
Organoid domes were grown from single-cell dissociated tumors from two radical prostatectomy specimens using previously established methods [9]. To begin, the organoids were passaged, kept in the control medium for 1 day, and then cultured in one of three conditions. In the control condition, 10 ng/ml dihydrotestosterone (DHT) was present in the medium. In the two experimental conditions, either DHT was absent from the medium (NODHT) or 10 μM enzalutamide (Enza), an androgen receptor (AR) inhibitor, was added to the medium. The medium was refreshed twice a week, every 3-4 days. Organoids were cultured for a period of 2 weeks and then dissociated into single cells, captured, and sequenced using the Seq-Well platform. Remaining organoids were grown for an additional 3 weeks, with one additional passage in between, and then sequenced on the Seq-Well platform [14]. After sequencing, organoids were aligned to the hg19 reference genome using the Dropseq workflow on Terra (https://app.terra.bio/). Downstream analysis of the gene expression matrices was conducted in Seurat [15]. Genes that were expressed by less than three cells were removed. Cells were filtered for greater than 500 unique molecular identifiers (UMIs), more than 300 genes, and less than 20% mitochondrial gene content. DecontX [16] was used for ambient RNA decontamination. The filtered count matrix for all cells was normalized and scaled based on the top 2,000 most varied genes. Graphical clustering was performed, and a uniform manifold approximation and projection (UMAP) was generated for visualization. Differentially expressed genes and gene set signatures from previously analyzed nontreated, prostate tumorderived organoids were used for cell type annotation. Signature scores of previously established androgen response gene sets (HALLMARK_ANDROGEN_ RESPONSE, NELSON_RESPONSE_TO_ANDROGEN_ UP) were computed using addmodulescore in Seurat [17]. Two-sided Wilcoxon rank-sum tests were used for each pairwise comparison. Gene Set Enrichment Analysis (GSEA) [18] was also performed in GSEA (version 4.0.2; https://www.gsea-msigdb.org/gsea/index.jsp) between each treatment condition (NODHT or Enza) and control (DHT) condition.

Club cell gene expression profile in association with inflammation in atrophic regions of prostate peripheral zone
Based on the results of our single-cell analyses, we began by performing RISH for peripheral zone club cell markers (CP, LTF, MMP7, PIGR, SCGB1A1) as defined previously [9] in a series of whole tissue sections from radical prostatectomy specimens (n = 9), to identify and spatially localize club cells in the peripheral zone of the prostate. We observed expression of CP and PIGR uniformly along prostatic urothelium, with co-expression of LTF, MMP7, and SCGB1A1 in a subset of urothelial cells and periurethral prostatic glands ( Figure 1A). In the peripheral zone, co-localization of the club cell genes was limited to luminal epithelial cells in focal prostatic atrophy ( Figure 1B, supplementary material, Figure S2), in regions closely associated with overt inflammatory infiltrates (proliferative inflammatory atrophy or PIA, Figure 1C). PIGR was uniformly expressed in focal atrophy lesions/PIA, often together with LTF. Furthermore, CP, MMP7, and SCGB1A1 were observed in a subset of luminal epithelial cells in LTF-and PIGR-positive focal atrophy ( Figure 1C). All five club cell markers were also observed in focal atrophy lesions found in regions where they were admixed with prostate cancer ( Figure 1D). Robust expression of all club cell markers except SCGB1A1 was likewise observed in postatrophic hyperplasia (PAH, Figure 2), which is a morphologically distinct subset of focal atrophy that is often closely associated with chronic inflammation in the prostate [19,20]. As with PIA, SCGB1A1 was only expressed in a subset of the cells in PAH that expressed the other club cell markers.
Co-expression of the club cell markers was limited to regions of focal atrophy/PIA; however, we did observe very focal PIGR positivity in a subset of adenocarcinoma (supplementary material, Figure S3A). Interestingly, these focal areas of PIGR-positive adenocarcinoma were often associated with chronic inflammation (supplementary material, Figure S3B).

Club-like cells in proliferative inflammatory atrophy of the prostate 87
SCGB1A1 was also focally present in normal, noninflamed prostate glands with a staining pattern possibly consistent with infiltrating immune cells (supplementary material, Figure S3C). The staining pattern in benign epithelium that we observed for SCGB1A1 is consistent with RISH analysis in a prior study in cribriform prostate cancer [21].
Club-like cells are enriched in radical prostatectomy specimens and associated with high grade prostate cancer We next questioned whether club epithelial cells could be found in the peripheral zone of nondiseased organ donor prostate or if they were specifically associated with prostate cancer in radical prostatectomy specimens. We performed RISH for PIGR in an expanded series of whole tissue sections from radical prostatectomy (n = 40) as well as organ donor specimens (n = 10, Table 1). The median age of the individuals for organ donor and radical prostatectomy specimens was 21 and 59, respectively. Whole tissue sections were annotated for cancer versus benign tissues (e.g. cancer-adjacent benign) with the exclusion of urothelium. Expression of PIGR was rarely observed in organ donor prostate tissues, with the exception of rare instances of inflammation and PIA in organ donor prostate ( Figure 3A,B). Quantification of PIGR expression across whole tissue sections demonstrated increased expression in the cancer-adjacent benign tissues from radical prostatectomy versus organ donor prostate ( p = 0.0007) and cancer tissue ( p = 0.0005, Figure 3C). Increased PIGR expression in benign versus cancer areas was present in both low-grade (Gleason Grade Groups 1 and 2,   Figure 3D). PIGR expression was likewise higher in benign regions from cases with both low-(p = 0.007) and high-grade cancer ( p = 0.0008) and in high-grade cancer ( p = 0.03) versus organ donor prostate ( Figure 3D). Finally, higher expression of PIGR in benign tissue areas versus cancer in radical prostatectomy was maintained in tissues from both self-identified European American (White, p = 0.03) and Black/African American individuals (Black, p = 0.009, Figure 3E). [22][23][24].

Luminal epithelial cells termed intermediate cells in PIA display an intermediate phenotype between basal and luminal epithelial cells and have been proposed as tumor-initiating cells in prostate cancer
Intermediate cells co-express high levels of the typical basal cell markers cytokeratin 5 (KRT5), GSTP1, and BCL2 along with luminal cell keratin KRT18. These cells also express high levels of hepatocyte growth factor receptor MET [22] and low levels of CD38 [23]. We next explored whether the cells that co-express club cell markers would be defined as luminal intermediate cells using previously defined markers. We stained the series of radical prostatectomy tissues that we assessed for club cell markers with CK903 [high-molecular-weight cytokeratin (HMWK), including KRT5] using IHC. CK903 stains basal cells in the prostate as well as luminal intermediate cells in PIA. We observed almost complete overlap between PIGR-positive luminal epithelial cells and CK903-positive luminal intermediate cells ( Figure 4A). Likewise, areas of PIA with co-staining for PIGR and CK903 had lower expression of AR ( Figure 4A), as was previously described for intermediate cells [22,25]. Analysis of our previously published scRNA-seq data [9] confirmed that cells that expressed a club cell gene signature were positive for MET and BCL2 and negative for CD38 ( Figure 4B), further supporting the idea that club cells in the peripheral zone of the prostate would be classified as luminal intermediate cells found in PIA.

Club-like cells are present in association with putative prostate cancer precursor lesions
We previously reported the identification of TMPRSS2: ERG fusion-positive (ERG+) PIA or low-grade

Club-like cells in proliferative inflammatory atrophy of the prostate 89
prostatic intraepithelial neoplasia (LGPIN) in a series of radical prostatectomy cases that harbored active bacterial infections [13]. In a subset of these cases, ERG+ PIA/LGPIN was found to be in association with early invasive adenocarcinoma (microadenocarcinoma), suggesting that TMPRSS2:ERG fusion events could be initiated within PIA or LGPIN lesions that then progress to early invasive cancer. Since intermediate cells in PIA (which we now demonstrate are club cells) have been proposed as tumor-initiating cells in prostate cancer [22][23][24], we tested whether club cells are present in association with ERG+ PIA lesions. Co-expression of luminal p63/HMWK, PIGR, and LTF was observed in PIA regions adjacent to ERG+ PIA ( Figure 5 and supplementary material, Figure S4). Of interest, PIGR expression was maintained in ERG+ PIA, but LTF was absent. PIGR expression was also found to be variably maintained in the adjacent microadenocarcinoma ( Figure 5 and supplementary material, Figure S4).

Club-like epithelial cells are established in organoids and are responsive to anti-androgen-directed treatment in vitro
Based on our identification of club cells in prostate cancer tissues, we previously identified club-like cells

FW Huang et al
in prostate epithelial organoids derived from prostate cancer specimens [9]. To further characterize the behavior of these club-like epithelial cells in vitro, we derived epithelial organoids from two localized prostate cancer patients (one White, one Black individual). Following a 2-week treatment period with enzalutamide (Enza), androgen deprivation (NODHT), or synthetic androgen (DHT), we isolated a total of 12,502 cells from these organoids. Using signature gene sets from previously analyzed prostate epithelial organoids and gene markers from normal prostate epithelial cell types, four cell populations, including basal epithelial cells, hillock cells, club cells, and dividing cells, were characterized among all treatment conditions. Overall, compared to cells from freshly isolated prostate epithelial tissues [9], the cells from these organoid populations exhibited more homogeneous gene expression profiles, resulting in a UMAP without clear separation of hillock from basal or club cells ( Figure 6A). Similarly, we observed no distinctive separation within the three treatment conditions ( Figure 6B) or by patients ( Figure 6C). Furthermore, we computed androgen response signature scores using two previously established signature gene sets, HALLMARK_ANDROGEN_RESPONSE and NELSON_RESPONSE_TO_ANDROGEN_UP, for comparison between the treatment conditions and cell types. Compared to basal epithelial and hillock cells, club cells exhibited the largest decrease in the two treatment conditions (NODHT and Enza) from the control (DHT) condition (two-sided Wilcoxon rank-sum test, *** p < 0.001) ( Figure 6D). This finding was also confirmed by GSEA, in which the HALLMARK_ ANDROGEN_RESPONSE was found to be the most significantly downregulated gene set in the HALLMARK gene set collection (msigdb.com) in the NODHT/Enza condition for club cells compared to the control (DHT) condition.
Finally, we examined club cell and luminal cell markers in these organoid samples and found that previously documented club cell marker expression (PIGR, CP, MMP7 and SCGB1A1) [6,9] was mostly limited to the club cell population ( Figure 6E). Compared to our previous scRNA-seq study in localized prostate cancer tumor tissues [9] as well as our observations in prostatectomy tissues using RISH in the present study, LTF expression was not detected in club cells in organoids in vitro. We also investigated luminal cell marker expression, such as KLK3, KRT8, Club-like cells in proliferative inflammatory atrophy of the prostate 91

92
FW Huang et al NKX3-1, and AR, within these organoid samples ( Figure 6F) and found a higher percentage of KRT8+ club cells and AR+ club cells compared to basal and hillock cells.
Discussion scRNA-seq of the young adult human prostate identified epithelial cells located in the urethra and proximal ductal trunks of the prostate that are similar in morphology and transcriptomic profile to club cells [6]. Likewise, a study of BPH found luminal epithelial cells in 5ARI-induced prostate atrophy that exhibited increased expression of urethral club cell-associated genes [7]. Aside from the prostatic urethra and collecting ducts and their expansion in 5ARI-induced prostate atrophy, club cells are reported to be rare in the prostate [6]. Our recent scRNA-seq analysis conducted in cells isolated from the peripheral zone of the prostate challenged this dogma and identified an epithelial cell subset that expressed characteristic club cell markers that we called club cells [9]. In the present study, we have now spatially localized these peripheral zone club cells in prostate tissues and identified them in the luminal compartment of focal atrophy/PIA. Furthermore, these club cells expressed characteristic 'intermediate cell' markers including MET, BCL2, and lower expression of CD38. Importantly, prostate intermediate cells with similar gene expression profiles have been reported to exhibit regenerative potential and progenitor-like characteristics both in our prior study [9] and in previous reports, e.g. CD38 lo luminal cells [23] and luminal 2 club cells [26]. This finding would be consistent with our observation here that club cells are present in association with ERG+ PIA lesions that are progressing to early invasive adenocarcinoma. Intriguingly, we found that whereas club cells that are adjacent to/transitioning with ERG+ PIA express LTF, both ERG+ PIA and ERG+ microadenocarcinoma are LTF-negative ( Figure 5 supplementary material, Figure S4). We previously reported that LTF was silenced by CpG island hypermethylation in prostate adenocarcinoma [12]. Our finding that LTF is absent in ERG+ PIA suggests that this gene silencing may occur very early in the carcinogenic process. Collectively, our study adds to a growing body of evidence that PIA (and, hence, peripheral zone club cells) may serve as a risk factor lesion for prostate cancer development. We report that prostate club cells are not normally present in the peripheral zone, since they were largely absent in young organ donor prostate tissues in the absence of inflammation. We therefore conclude that the club cell phenotype is induced in the adult prostate in association with cellular injury and/or inflammation, and possibly with aging, in the development of PIA. In addition to gene expression profiles similar to those of lung club cells, prostate club cells are also histologically similar to lung club cells. Whereas the luminal epithelial cells in normal prostate acini are tall columnar cells, the club-luminal epithelial cells in PIA are shortened and cuboidal [22], akin to club cells in the lung [1]. This intriguing similarity, in terms of both gene expression and phenotype, between lung club cells and prostate club cells suggests that prostate club cells may play a functional role similar to that of club cells in the lung, e.g. a protective, antimicrobial role. We speculate that the presence of club cells in the prostatic urethra and budding prostatic ducts may signify a protective and antimicrobial function for these cells, especially related to the regulation of the urinary microbiome and potential urinary pathogens [27]. The induction of this same cell type in the peripheral zone in PIA may likewise indicate that PIA has formed in response to a pathogen insult. Indeed, the classical club cell genes studied here, such as the antimicrobial innate immune protein LTF [12,28], the extracellular pathogen neutralizer and nonspecific microbial scavenger PIGR [29,30], and the antibacterial peptide activator MMP7 [31], are all involved in pathogen defense.
We report that club cells in PIA, as assessed by quantification of PIGR-expressing cells in whole tissue sections, are more abundant in cancer-adjacent benign prostate tissues than in cancer regions. This is not unexpected, as both inflammation and PIA are more prevalent within benign regions of the peripheral zone than admixed with prostate cancer [32]. However, whereas low-grade cancer was quantitatively similar to organ donor prostate, high-grade cancer had significantly higher quantification of PIGR-positive areas. Of interest, of the five club cell markers that we assessed in this study (CP, LTF, MMP7, PIGR, SCGB1A1), PIGR was the only marker focally expressed in a subset of adenocarcinoma (supplementary material, Figure S3A,B). Future studies will assess whether this focal expression of PIGR is specifically associated with higher grade prostate cancer.
In a prior BPH study [7], gene expression profiling showed that 5ARI-induced atrophy of prostate luminal cells correlated with reduced AR signaling and increased expression of club cell genes. Prostate luminal cells within hormonal atrophy acini adapted to decreased DHT conditions by increasing NF-κB signaling and BCL2 expression, and exhibited increased survival. The club cells that we identified in the peripheral zone also expressed increased BCL2 ( Figure 4B). It has been postulated that focal atrophy/PIA is morphologically distinct from diffuse/hormonal atrophy that is induced after androgen withdrawal or AR blockade [33]. The results of the present study imply a potential overlap between atrophy induced by hormonal treatments and focal atrophy/PIA. Additional study is needed to clarify the similarities and differences between these two types of atrophy. Whether peripheral zone club cells associated with PIA exhibit increased survival with antiandrogen treatment remains to be determined.
Finally, it should be noted that both our RISH staining of club cell markers in prostatectomy tissues ( Figure 1) and single-cell profiling of prostate epithelial cell organoids ( Figure 6) indicate a heterogeneity to club cell gene expression profiles, such that not all cells that we Club-like cells in proliferative inflammatory atrophy of the prostate 93 classify as club cells express all club cell markers. While our characterization of these cells in the luminal compartment of PIA as prostate club cells may be an oversimplification, we assess that multiple cell states likely exist in these lesions, which suggests the transdifferentiation potential of prostate luminal cells. Future studies will clarify these cellular states and their functional role in prostate carcinogenesis. Figure S1. Positive and negative controls for CP, MMP7, PIGR, and SCGB1A1 RISH assays Co-expression of PIGR and absence of LTF is observed in ERG+ PIA (black arrowheads). LTF is absent and PIGR is variably expressed (see also Figure 5) in associated ERG+ microadenocarcinoma (black arrow). Scale bars, 200 μm Club-like cells in proliferative inflammatory atrophy of the prostate 95