Prostatic inflammation enhances basal-to-luminaldifferentiation and accelerates initiation of prostatecancer with a basal cell originOh-Joon Kwona, Li Zhanga, Michael M. Ittmannb,c, and Li Xina,b,d,1

Departments of aMolecular and Cellular Biology and bPathology and Immunology and the dDan L. Duncan Cancer Center, Baylor College of Medicine,Houston, TX 77030; and cMichael E. DeBakey Veterans Affairs Medical Center, US Department of Veterans Affairs, Houston, TX 77030

Edited by Owen N. Witte, Howard Hughes Medical Institute, University of California, Los Angeles, CA, and approved December 4, 2013 (received for reviewOctober 15, 2013)

Chronic inflammation has been shown to promote the initiationand progression of diverse malignancies by inducing genetic andepigenetic alterations. In this study, we investigate an alternativemechanism through which inflammation promotes the initiationof prostate cancer. Adult murine prostate epithelia are composedpredominantly of basal and luminal cells. Previous studies revealedthat the two lineages are largely self-sustained when residing intheir native microenvironment. To interrogate whether tissueinflammation alters the differentiation program of basal cells,we conducted lineage tracing of basal cells using a K14-CreER;mTmG model in concert with a murine model of prostatitisinduced by infection from the uropathogenic bacteria CP9. Weshow that acute prostatitis causes tissue damage and creates atissue microenvironment that induces the differentiation of basalcells into luminal cells, an alteration that rarely occurs under normalphysiological conditions. Previously we showed that a mouse modelwith prostate basal cell-specific deletion of Phosphatase and tensinhomolog (K14-CreER;Ptenfl/fl) develops prostate cancer with a longlatency, because disease initiation in this model requires and is lim-ited by the differentiation of transformation-resistant basal cellsinto transformation-competent luminal cells. Here, we show thatCP9-induced prostatitis significantly accelerates the initiation ofprostatic intraepithelial neoplasia in this model. Our results demon-strate that inflammation results in a tissue microenvironment thatalters the normal prostate epithelial cell differentiation programand that through this cellular process inflammation accelerates theinitiation of prostate cancer with a basal cell origin.

prostate stem cells | cells-of-origin for cancer

Chronic inflammation is a significant contributor to the initi-ation and progression of a wide spectrum of malignancies,

including prostate cancer (1). Loss-of-function mutations and cer-tain single-nucleotide polymorphisms in inflammation-associatedgenes and a history of prostatitis have been positively correlatedwith prostate cancer risk (2–5). In addition,DeMarzo et al. (6) havedescribed regions of prostatic atrophy termed “proliferative in-flammatory atrophy” (PIA) that often are associated with in-flammatory cell infiltrates in human samples and have proposedthat these areas are direct precursors of prostatic intraepithelialneoplasia (PIN) or adenocarcinoma. Finally, genetic studies usingmouse models have demonstrated that inflammatory signals in theprostate can either induce or synergize with oncogenic signaling topromote the initiation and progression of prostate cancer (7, 8).Currently, it is widely accepted that chronic inflammation promotestumor initiation and progression through the induction of variousgenetic and epigenetic alterations. For example, inflammationenhances the production of reactive oxygen species, up-regulatestheexpressionof the enzyme that induces nucleotidealterations (9),and modulates the activity of DNA methyltransferases (10). In-terestingly, inflammation also has been shown to deregulate self-re-newal and differentiation of hematopoietic stem cells (11). However,it remains to be determined whether altering cell lineage differenti-

ation programs per se serves as a mechanism through which in-flammation promotes tumor initiation and progression.Human prostate epithelia are composed of a layer of co-

lumnar luminal epithelial cells that rest on a continuous cuboidallayer of basal epithelial cells. In comparison, rodent prostateshave fewer basal cells that instead usually form a punctuatedlayer (12). A third type of epithelial cell, neuroendocrine cells, israre and hence often is not investigated as intensively as theother two lineages (13). Using a lineage-tracing approach, weand several other groups showed that adult murine prostatebasal and luminal cell lineages are largely self-sustained whenresiding in their native microenvironment, although in two of thestudies basal cells were shown to generate luminal cells at anextremely rare frequency (14–18). In stark contrast, other studieshave shown that basal cells are able to differentiate very effi-ciently into luminal cells and neuroendocrine cells in a dissoci-ated prostate-cell regeneration assay (19–25). In the prostate-cellregeneration assay, prostate basal epithelial cells were removedfrom their native environment, dissociated into single cells, andstimulated with embryonic urogenital sinus mesenchymal (UGSM)cells that have strong reprogramming activity (26–28). These ex-perimental conditions recapitulate some processes that also occurduring tissue inflammation. For example, cellular dissociationresembles tissue damage, whereas UGSM cells may mediatesignaling that also can be induced by reactive stroma (29).

Significance

Inflammation promotes the initiation of various malignanciesby inducing genetic and epigenetic changes. Here we showthat bacterial infection-induced prostatitis results in microen-vironmental changes that enhance the differentiation of pros-tate basal cells into luminal cells, a cellular process that rarelyoccurs under normal physiological conditions. Previously, weshowed in a mouse model that disease initiation for prostatecancer with a basal cell origin requires and is limited by basal-to-luminal differentiation and that prostatic inflammation in-duced by bacterial infection accelerates disease initiation byenhancing basal-to-luminal differentiation. Collectively, ourresults show that inflammation-induced microenvironmentalchanges alter the prostate epithelial lineage differentiationprogram, and we propose this alteration as a distinct and com-plementary process through which inflammation promotes tu-mor initiation.

Author contributions: L.X. designed research; O.-J.K. and L.Z. performed research; O.-J.K.,M.M.I., and L.X. analyzed data; and L.X. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

See Commentary on page 1666.1To whom correspondence should be addressed. E-mail: xin@bcm.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1318157111/-/DCSupplemental.

E592–E600 | PNAS | Published online December 23, 2013 www.pnas.org/cgi/doi/10.1073/pnas.1318157111

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

Therefore, we reasoned that prostatic inflammation may in-duce functional plasticity in prostate basal cells in vivo and stimu-late their capacity to differentiate into luminal cells. Previously, weand others have shown that basal-to-luminal differentiation is veryrare under physiological conditions and serves as a barrier fordisease initiation in mouse models of prostate cancer with a basalcell origin (14–16, 30). Therefore, we also hypothesized that in-flammation can accelerate the initiation of prostate cancer witha basal cell origin by stimulating basal-to-luminal differentiation.To test these hypotheses, we took advantage of a mouse model

of bacterial infection-induced prostatitis. One reason for chronicinflammation in the prostate is bacterial colonization in theprostate via reflux of urine into the prostatic ducts of the pe-ripheral zone (31). A reproducible mouse model of bacterialprostatitis has been established in which transurethral inoculation ofuropathogenic strains of Escherichia coli into the prostate resultsin massive inflammation and reactive hyperplasia in the prostate(7, 32–34). As reported here, we established this mouse modelfor prostatitis using a uropathogenic E. coli strain CP9 that wasisolated from the blood of a patient with pyelonephrititis (35).Using this model, we demonstrate that acute inflammation in-duced by bacterial infection causes tissue damage and results ina tissue microenvironment that stimulates the differentiation ofbasal cells to luminal cells and accelerates disease initiation ina mouse model for prostate cancer with a basal cell origin.

ResultsProstate Inflammation Induced by Bacterial Infection Results inHyperplasia Characterized by High Proliferation and Reactive Stroma.To interrogate how the prostate epithelial lineage hierarchy ismaintained in response to acute inflammation, we first charac-terized a mouse model for prostatitis induced by infection witha uropathogenic E. coli strain, CP9. As described in detail inMaterials and Methods, the CP9 E. coli and PBS control solutionswere inoculated transurethrally into 9-wk-old wild-type C57BL/6male mice. Consistent with previous reports, there were signifi-cantly variable degrees of inflammation among experimental miceand among different prostatic lobes (7, 32). In addition, inflamedand noninflamed ducts often were observed concomitantly in alldifferent lobes. Inflammation occurred in anterior and dorsolat-eral prostate lobes at high penetrance but was observed less fre-quently in ventral prostate lobes (Table S1). Epithelial hyperplasiaand immune cell infiltrations were commonly observed in theinflamed area 5 d after bacteria instillation and persisted 3 molater (Fig. 1A).The PBS-treated control mice generally showed normal pros-

tatic histology (Fig. 1A). As shown in Fig. 1B and Fig. S1, a singlelayer of luminal cells that express cytokeratin 8 (K8) and the an-drogenreceptor(AR)surroundtheprostatic luminal space,whereasthe basal cells that express cytokeratin 5 (K5) and p63 form a punc-tuated layer between luminal cells and the basementmembrane.Transit-amplifying cells (36) that are positive for both K5 and K8were rare. In contrast, there was a dramatic increase in basal cellsand transit-amplifying cells in the inflamed prostatic lobes of theCP9-treated mice 5 d after bacteria inoculation, and this increasepersisted 3 mo later (Fig. 1B and Fig. S1). This observation issimilar to the report that intermediate cells in human prostateepithelium are enriched in PIA (37). Finally, piling up of the K8-expressing luminal epithelial cells also was prominent in theinflamed prostates at different time points.There also were considerable changes in prostatic stroma of

the CP9-treated mice. Prostate epithelial ducts of the controlPBS-treated mice are encapsulated by a thin and continuouslayer of smooth muscle actin-positive stromal cells, which becomesslightly thicker upon aging (Fig. 1C). In comparison, in the CP9-treated group this smooth muscle cell layer often was disruptedby 1 mo after bacteria inoculation, indicating damage in tissuestructures. Meanwhile, at the peri-glandular spaces there was an

expansion of myofibroblast stromal cells that were positive forboth smooth muscle actin and vimentin, which are frequentlyreported during wound healing and in reactive stroma associatedwith prostate cancer (29). Masson’s trichrome staining furtherconfirmed a reduction of peri-glandular smooth muscle cells andincreased collagen deposition in the prostate tissues in the CP9-treated group, as shown by the fainter red staining (black arrows,Fig. 1D and Fig. S2) and darker blue staining (yellow arrow, Fig.1D and Fig. S2). Collectively, these observations suggest that CP9-induced inflammation induces stromal fibrosis and instigates a re-active stromal environment.Immunostaining for the proliferation marker Ki67 showed that

the proliferation indices of the basal and luminal cells in inflamedprostatic lobes of the CP9-treated mice increased by more than31.4- and 10.6-fold, respectively, as compared with those of thecells in PBS-treated control mice (Fig. 1E). The urogenital organsof the CP9-treated mice appeared smaller than those of the PBS-treated mice, but the average weight of the prostate tissues in thetwo groups was similar (Fig. S3). We dissociated prostate tissues toquantify the absolute cell numbers of individual cell lineages byFACS analysis as described previously (21). FACS analysis corrob-orated that the basal cell and the stromal cell percentages increasedby 1.5- and 1.2-fold, respectively, in the prostates of CP9-treatedmice (Fig. 1F). The absolute number of basal cells and stromalcells per prostate also increased (basal cells: 7,745.0 ± 991.7 inthe control group versus 20,088.3 ± 1,968.1 in the CP9 group,P < 0.01; stromal cells: 21,281.7 ± 625.0 in the control groupversus 44,326.6 ± 5,343.9 in the CP9 group, P < 0.05) (Fig. 1G).Luminal cell number also increased but did not reach statisticalsignificance (41,998.3 ± 3,573.5 in the control group versus55,305.0 ± 4,682.4 in the CP9 group). The increase in cell numbersis consistent with the increased proliferation indices in these lin-eages and suggests that the decreased tissue volumes in the CP9-treated group may be caused by the impaired secretory function.FACS analysis also revealed a fivefold increase in the per-

centage of CD45+ leukocytes within CP9-treated mouse pros-tates (5.93% in the control group versus 27.53% in the CP9-treated group, P < 0.001) (Fig. 1H). FACS analysis also showedthat macrophages and T cells were the major leukocytes in theprostate in the control group, whereas neutrophils, macrophages,and T cells became the dominant leukocytes in the CP9-treatedgroup (Fig. 1I). Fig. 1J shows that each of the different types ofleukocytes examined increased in the prostates of CP9-treatedmice (116-fold for neutrophils, 8.0-fold for T cells, 3.9-fold formacrophages, and 10.2-fold for B cells). Collectively, these datacorroborate the notion that transurethral instillation of the CP9bacteria induces prostatic inflammation and results in reactivehyperplasia.

Inflammatory Prostatic Microenvironment Induces Basal-to-LuminalDifferentiation. Previously, several different transgenic mousemodels have been used to mark prostatic basal cells specificallywith fluorescent proteins (14–16, 18). We bred a K14-CreER linethat expresses tamoxifen-responding CreER driven by the basalcell-specific cytokeratin 14 (K14) promoter with an mTmGfluorescent reporter line (38) that expresses GFP after Cre-LoxP–mediated homologous recombination (14). When K14-CreER;mTmG bigenic mice are treated with tamoxifen, GFP isexpressed only in prostate basal cells. With this approach, adultmurine prostate basal cells seldom generate other cell lineagesunder their normal physiological conditions (14–16).We sought to investigate whether bacteria-induced tissue in-

flammation affects the in vivo differentiation program of pros-tate basal cells. As schematically illustrated in Fig. 2A anddescribed in Materials and Methods, 5-wk-old K14-CreER;mTmG bigenic male mice were treated with tamoxifen to labelprostate basal cells with GFP. Three weeks later, a group of micewas inoculated transurethrally with CP9 bacteria. Control mice

Kwon et al. PNAS | Published online December 23, 2013 | E593

MED

ICALSC

IENCE

SPN

ASPL

US

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

either were instilled transurethrally with the PBS solution orunderwent no treatment. Prostate tissues were collected andexamined for the expression of GFP and lineage markers 1 and 3mo later (n = 8 per group per time point).

Consistent with our previous studies (14), 5 d after tamoxifentreatment, a small percentage (∼9%) of basal cells were labeledwith GFP under our experimental conditions (Fig. 2B), and allGFP+ cells expressed the basal cell marker K14 but not the luminal

Fig. 1. The mouse model for CP9-induced prostatitis. (A) H&E staining of anterior (AP), dorsolateral (DLP), and ventral (VP) prostate lobes of mice at 5 d,1 mo, and 3 mo after intraurethral instillation of PBS and CP9 inoculum. (B) Immunostaining of K5 and K8. (C) IHC analysis of smooth muscle actin (SMA) andvimentin. (D) Masson’s trichrome staining of prostate tissues at 3 mo after intraurethral instillation of PBS and CP9. The black arrow points to red staining forsmooth muscle cells; the yellow arrow indicates blue collagenous stroma. (E) Coimmunostaining of Ki67, K14, and K8. Data shown in the bar graphs representmeans and SD from four mice. (F and G) FACS plots of prostatic lineage composition in PBS- and CP9-treated mice 2 wk posttreatment. Data shown in the bargraph represent means and SD from three mice. (H–J) FACS plots of CD45+ leukocytes in prostates of PBS- and CP9-treated mice 2 wk posttreatment. Datashown in the bar (H and J) and pie (I) graphs represent means and SD from three mice. *P < 0.05; **P < 0.01; ***P < 0.001.

E594 | www.pnas.org/cgi/doi/10.1073/pnas.1318157111 Kwon et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

Fig. 2. CP9-induced prostatitis stimulates basal–luminal differentiation. (A) Schematic illustration of the experimental strategy. Tmx, tamoxifen. (B) GFP-labeled basal and luminal cells in tamoxifen-treated K14-mTmG mice. (C) Immunostaining of GFP, K14, and K8 in prostates of K14-mTmG mice. (D) Costainingof GFP, K14, and K8 in prostates of tamoxifen-treated K14-mTmG mice that did not undergo transurethral instillation (Tmx) or that underwent instillation ofPBS (Tmx + PBS) or CP9 (Tmx + CP9), performed 1 or 3 mo after transurethral instillation. Red arrows point to GFP+ luminal cells. (E–G) Bar graphs show thepercentages of K14+K8−, K14+K8+, and K14−K8+ cells in GFP+ cells in the three groups. Data were collected from eight mice. (H) Immuno-costaining of K8 andcleaved caspase 3 in the three groups. Data shown in the bar graphs represent means and SD from four mice. *P < 0.05; **P < 0.01.

Kwon et al. PNAS | Published online December 23, 2013 | E595

MED

ICALSC

IENCE

SPN

ASPL

US

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

cell marker K8 (Fig. 2C; n = 1,390 GFP+ cells). In the group thatreceived no transurethral instillation, GFP-expressing cellsremained almost exclusively K14+K8− (1,720 of 1,721 and 1,245 of1,246 GFP+ cells examined 2 and 4 mo after tamoxifen treatment,respectively) (Fig. 2 D and E). This result corroborates previousfindings that adult murine K14-expressing basal cells primarilygenerate basal cells in vivo (14–16). In the control group that re-ceived PBS, the majority of GFP+ cells remained as K14-expressingbasal cells even 3 mo after instillation (Fig. 2 D and F). However,we were able to detect ∼1.05% K8-expressing cells within theGFP+ cells 1 mo after PBS instillation (Fig. S4), and the percentagerose to 1.70% 3 mo after instillation. In sharp contrast, in theCP9-treated group, many GFP+ cells (red arrows, Fig. 2D)were K8+ (21.05% and 20.45% for K8+K14+ cells at 1 mo and 3mo after instillation, respectively; 17.85% and 33.39% forK8+K14− cells at 1 mo and 3 mo postinstillation, respectively)(Fig. 2G). The increasing percentage of K8+ cells within totalGFP+ cells during aging was not caused by the increased pro-liferation of luminal cells compared with basal cells, as shown byKi67 staining (Fig. 1E), nor did fewer luminal cells than basalcells undergo apoptosis, as shown by immunostaining of cleavedcaspase 3 (Fig. 2H). Most likely, this increased percentagereflects a continuous basal-to-luminal differentiation or the dif-ference in the length of the cell cycle in basal and luminal cells.In conclusion, CP9 infection causes tissue damage and micro-environmental changes that stimulate the differentiation of basalcells into luminal cells.The frequency of basal–luminal differentiation also was in-

creased in the PBS-treated group as compared with the un-treated group (Fig. 2 E and F). Intraurethral instillation of PBSmay cause reflux of urine into the prostate, leading to chemicalor physical trauma (39). Therefore, the increased basal–luminaldifferentiation in the PBS-treated group may reflect the occurrenceof a mild inflammatory response, as reported previously (32).We also investigated whether basal cells are capable of gener-

ating neuroendocrine cells during prostatitis. Neuroendocrinecells remained extremely rare within the prostates of the CP9-treated mice, suggesting that the inflammatory response does notinduce neuroendocrine differentiation (Fig. S5). There was noevidence that basal cells can generate neuroendocrine cells, be-cause none of the 187 neuroendocrine cells that we examinedexpressed GFP. One potential caveat regarding these studies isthat we were able to GFP-label only ∼9% of the basal cells.Therefore, we cannot exclude the possibility that nonlabeled basalcells do possess a differentiation capacity or that GFP+ neuro-endocrine cells might be detected if more neuroendocrine cellswere examined.

CP9-Induced Prostatitis Accelerates Initiation of Prostate Cancer inthe K14- Phosphatase and Tensin Homolog Model. Previously, weexamined the capacity of the prostate basal cells to serve as thecells of origin for prostate cancer using a K14-CreER;Ptenfl/fl

(K14-Pten) model, through which Phosphatase and tensin homo-log (Pten) can be ablated specifically in prostate basal cells upontamoxifen induction (14). Pten deletion is not sufficient to trans-form basal cells. Instead, disease initiation in this model requiresthe differentiation of basal cells into transformation-competentluminal cells. However, basal-to-luminal differentiation rarely takesplace; hence this mouse model develops prostate cancer witha long latency and low penetrance. Therefore, we sought to de-termine experimentally whether prostatitis-induced basal-to-luminal differentiation could accelerate disease initiation in theK14-Pten model.We showed previously that at early times after tamoxifen

treatment the K14-Pten mouse model develops rare and tiny PINlesions that are difficult to detect (14). In this study, we gener-ated K14-CreERTg/Tg;PTENfl/fl;mTmGTg/Tg (hereafter referredto as “K14-Pten-mTmG”) mice, reasoning that expression of

GFP from the mTmG allele might facilitate a more sensitive andsemiquantitative detection of multifocal tiny PIN lesions. Theexperimental design was similar to that shown in Fig. 2A. Briefly,PBS or bacteria solution was instilled transurethrally into K14-Pten-mTmG mice 3 wk after tamoxifen treatment. Prostate tis-sues were collected and examined 1 and 3 mo after instillation ofPBS or CP9 (n = 8 per group per time point).As shown in Fig. 3 A and B, GFP-expressing foci were

detected in all experimental mice in both groups. However, onaverage, twice as many GFP foci per mouse were observed in theCP9-treated group (8.4 and 23.6, respectively, at 1 and 3 mopost-CP9 instillation) as in the PBS group (4.3 and 11.0, re-spectively, at 1 and 3 mo post-PBS instillation). The GFP foci inthe CP9-treated group at 3 mo after transurethral instillationalso were larger than those in the PBS group (Fig. 3A).H&E staining of the GFP foci revealed that five mice in the

PBS-treated group developed PIN1/2 lesions (40) 1 mo afterPBS treatment (Fig. 3 C, row 1, and E). The GFP foci in theother three mice contained only normal tissue with clusters ofGFP-labeled basal cells (Fig. 3C, row 2). The GFP+ basal cellsdid not always express pAKT, suggesting that the efficiency ofthe tamoxifen-induced homologous recombination may begreater at the mTmG locus than at the Pten locus. PIN1/2 lesionswere detected in four mice 3 mo after PBS treatment, and PIN3/4 lesions were detected in the other four mice (Fig. 3 C, rows 3and 4, and E). All PIN lesions comprised mainly K8-expressingluminal cells (Fig. S6), as reported previously (14). Cells withinthese lesions stained positively for both GFP and pAKT, con-firming that they were derived from Pten-null basal cells.In contrast, multifocal lesions were observed in all the CP9-

treated mice 1 mo after transurethral instillation (Fig. 3 D andE). The majority of these lesions did not stain positively for GFPor pAKT (Fig. 3D, Upper). Therefore these lesions likely werederived from cells that did not undergo homologous recombinationand may have resulted directly from the prostatitis induced byCP9 infection. Nevertheless, PIN3/4 lesions consisting of cellsthat expressed both GFP and pAKT were detected in all themice, albeit at a low frequency (three to nine foci per mouse)(Fig. 3 D and E). By 3 mo after CP9 instillation, GFP+pAKT+

PIN3/4 lesions were easily detectable in all experimental mice(Fig. 3 D and E). All these GFP+ PIN lesions were composed ofK8-expressing cells (Fig. S6). As expected, we also observedlesions that contained a mixture of GFP+pAKT+ cells and wild-type cells (Fig. S7). The proliferation index in the GFP−pAKT−

lesions was less than that observed in the GFP+pAKT+ lesions,suggesting that Pten deletion confers a stronger or additive mi-totic signaling (Fig. 3F). Collectively, these results demonstratethat CP9-induced prostatitis accelerates disease initiation andprogression in the K14-Pten model, probably through the com-bined effects of enhanced basal-to-luminal differentiation andpromitotic signaling induced by the inflammatory response.

DiscussionInflammation can contribute to tumor initiation and progressionby inducing genetic and epigenetic changes. Our study suggeststhat the inflammation also may accelerate tumor initiation byaltering the tissue lineage differentiation program. Cells inclinedto acquire oncogenic signaling within a tissue may be intrinsicallyresistant to transformation. However, tissue damage induced byvarious insults, including inflammation, may result in microen-vironmental changes that are capable of stimulating the differ-entiation of these cells into other lineages. The prooncogenicsignaling inherited by the daughter lineages may enhance theirsusceptibility to transformation induced by ensuing genetic changes.Of note, we believe that it is not the type of inflammation, but theinflammation- and tissue damage-induced signals, such as the NF-κB signaling, that drive basal-to-luminal differentiation. Thishypothesis is supported by our observation that basal-to-luminal

E596 | www.pnas.org/cgi/doi/10.1073/pnas.1318157111 Kwon et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

Fig. 3. CP9-inducedprostatitis acceleratesdisease initiationandprogression in theK14-Ptenmodel. (A) Fluorescent imagesofprostates fromtamoxifen-treatedK14-Pten-mTmG mice 1 mo and 3 mo after transurethral instillation of PBS (Tmx + PBS) or CP9 bacteria (Tmx + CP9). (B) Dot plots quantify numbers of GFP foci. Datarepresent means and SD from eight mice per group. (C) H&E staining and IHC analysis of GFP and pAKT in prostates from tamoxifen-treated K14-Pten-mTmGmice1moand3moafter transurethral instillationof PBS. (D) H&E stainingand IHCanalysis ofGFPandpAKT inprostatesof tamoxifen-treatedK14-Pten-mTmGmice1 and3moafter transurethral instillationofCP9. (Upper) Representative imagesofpAKT− lesions. (Lower) Representative imagesofpAKT+ lesions. (E) The table summarizesdiseasegrades in PBS-andCP9-treatedgroups.Diseasegrade is definedby themostadvancedGFP+ lesions detected in tissues. (F) IHC stainingofKi67 inGFP+andGFP−

foci in murine prostates 1 mo after CP9 treatment. Data shown in the bar graph represent means and SD from eight mice. *P < 0.05; **P < 0.01.

Kwon et al. PNAS | Published online December 23, 2013 | E597

MED

ICALSC

IENCE

SPN

ASPL

US

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

differentiation was slightly enhanced even in the PBS-treatedgroup, probably because of the very mild inflammation caused bythe urinal reflex-induced chemical and physical trauma.

Mouse Model for Bacterial Infection-Induced Prostatitis. Two modelshave been used frequently in previous studies to investigate theprostate epithelial lineage hierarchy: the aging model and themodel in which epithelial turnover is induced by alternate an-drogen ablation and replacement (14–18). A limitation of thesemodels is that they do not consider some environmental factors,such as inflammation, that affect human prostate epithelial ho-meostasis. Laboratory mice usually do not develop prostatitiseven during aging because they are maintained in aseptic vivariaand fed with a standard laboratory dietary regimen. On the otherhand, even though the infiltration of inflammatory cells into theprostate was observed when mice were subjected to alternateandrogen deprivation and replacement (41, 42), tissue damagesuch as the disruption or ulceration of acinar outlines and epi-thelial reactive hyperplasia that frequently are observed in hu-man chronic prostatitis rarely occur.The model of prostatitis induced by CP9 bacterial infection

complements these models but is not without limitations. First,the bacterial-infection model is not representative of all types ofhuman prostatitis, because many patients with chronic prostatitisand chronic pelvic pain syndrome have no evidence of urinarytract infection (43). In addition, E. coli is not the only source forbacterial infection (39, 44, 45). Second, transurethral instillationof uropathogenic bacteria usually causes an immediate phase ofacute inflammation followed a few months later by a secondphase of multifocal chronic inflammation (34). Because bacteria-induced prostatitis in this model causes the expansion of transit-amplifying cells that express K14 (Fig. 1), we had to label the basalcells in the K14-mTmG model before bacteria instillation. Conse-quently, we were not able to determine the basal cell differentiationprogram under the chronic inflammatory conditions. Thereforeother models for prostatitis may be used to investigate the differ-entiation potential of basal cells during non-bacteria–inducedprostatitis and during chronic prostatic inflammation (46–51).

Functional Plasticity of Prostate Basal Cells. We and others showedpreviously that adult murine prostate basal cells primarily gen-erate basal cells in vivo when residing in their native microenvi-ronment (14–16). However, basal cells exhibit functional plasticityand possess the stem cell potential to differentiate into the otherlineages in prostate epithelia. Rodent prostate basal epithelialcells are capable of adapting a luminal cell phenotype during invitro culture (52). Several independent groups (19–25) have shownin a prostate regeneration assay that both human and rodent basalcells possess the capacity for multilineage differentiation. In thecurrent study, we further showed that acute prostatic inflammationinduced by bacterial infection significantly promoted basal-to-luminal differentiation in situ. Finally, in normal human prostatetissues, luminal cells have been shown to share the same geneticchanges as basal cells; this observation also indirectly supports theoccurrence of basal-to-luminal differentiation (53).The functional plasticity of basal cells may be suppressed by

residence in their natural niches or, alternatively, stimulated bybasal cell-autonomous genetic and epigenetic changes or alter-ations in their native niches. The first mechanism is less likelybecause isolated basal cells are not able to differentiate readilyinto luminal or neuroendocrine cells when cultured in vitro.Hormones or defined growth factors are required for their dif-ferentiation (54–56). Therefore, future studies will focus on theidentification of the basal cell-autonomous or niche-derivedsignaling that actively drives basal-to-luminal differentiation.During inflammation infiltrating immune cells such as macro-phages, T cells, and myeloid-derived suppressor cells can playcritical roles in tumor initiation and progression (1). However,

they may not be directly necessary for basal-to-luminal differ-entiation, because basal cells can generate luminal cells effi-ciently in immune-deficient environments with the help ofUGSM cells (19, 21). Therefore, it is more likely that prostatestromal cells were reprogramed by the inflammatory signalingso that they reactivated signaling shared by the UGSM cells,thereby acquiring the capacity to induce basal-to-luminal dif-ferentiation.

Cells of Origin for Human Prostate Cancer.Recent molecular studieshave demonstrated that multifocal human prostate cancer rep-resents an independent group of diseases (57). These data sug-gest the existence of multiple cells-of-origin for individual prostatecancers. Treatment-naive human prostate cancer consists of cellsthat display a luminal cell phenotype. Therefore, prostate cancerhas been proposed to have a luminal cell origin, so that disease isinitiated directly from luminal cells that have acquired oncogenicgenetic changes de novo. On the other hand, in human pros-tates, basal cells are more likely to acquire and maintain geneticchanges, because they display a higher proliferation index butundergo less apoptosis than luminal cells (58–60). In normalhuman prostate tissues, luminal cells share the same geneticchanges in mitochondria as basal cells, indicating that they havethe same origin or that basal cells can generate luminal cells (53).Therefore, genetic alterations that contribute to the initiation ofsome prostate cancers also may take place in basal cells, so thatthese prostate cancers have a basal cell origin. The differentia-tion of basal cells with preexisting prooncogenic signaling intoluminal cells would enhance the susceptibility of luminal cells totransformation by ensuing genetic changes, as demonstrated byour current study and by several previous studies using mousemodels (14–16, 19, 61). A very recent study using the prostateregeneration model showed that basal cells even may serve as theorigin of prostate cancer stem cells (62). Currently, because oftechnical limitations, no published study has directly demon-strated that basal cells and luminal cells within the same humanPIN lesions share genetic changes. With improved technologies,future studies using laser-capture microscopy in concert witharray comparative genomic hybridization should generate com-pelling results to validate this hypothesis. Finally, our study doesnot imply that all genetic changes associated with prostate cancermust take place in basal cells or that that only basal cells canserve as the cells of origin for human prostate cancer.

Materials and MethodsMice. C57BL/6 mice were from Charles River. The K14-CreER mice (63) werepurchased from the Jackson Laboratory. The mTmG fluorescence reportermice (38) were originally from Liqun Luo (Stanford University, Stanford, CA)and were internally transferred from Jeffery Rosen’s laboratory at the BaylorCollege of Medicine. The Ptenfl/fl mice (64) were made by Hong Wu (Uni-versity of California, Los Angeles). Mice genotypes were verified by PCRusing mouse genomic DNA from tail biopsy specimens. PCR products wereseparated electrophoretically on 1% agarose gels and visualized via ethi-dium bromide under UV light. All animal work was approved by and per-formed under the regulations of the Institutional Animal Care Committee ofthe Baylor College of Medicine.

E. coli Strain. The uropathogenic E. coli strain CP9 was from Allison O’Brien(Department of Microbiology and Immunology, Uniformed Services Uni-versity of the Health Sciences, Bethesda). This strain is a human blood isolatethat expresses cytotoxic necrotizing factor 1 and hemolysin and previouslyhas been reported to induce a reproducible mouse model for prostatitis (7,35). To prepare CP9 inoculum, a single CP9 colony was inoculated and cul-tured in 5 mL of static Luria broth for 24 h and subsequently was passaged in50 mL of static Luria broth for 48 h at 37 °C. Bacteria were adjusted to anA600 of 0.85 (bacterial density equal to 4.5 × 108 cfu/mL), washed, and seriallydiluted by sterile PBS to a final concentration of 1 × 107 cfu/mL.

Bacterial Infection and Tamoxifen Treatment. Intraurethral instillation ofbacterial inoculum was performed based on the procedures described pre-

E598 | www.pnas.org/cgi/doi/10.1073/pnas.1318157111 Kwon et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

viously (7, 34). For characterization of the inflammation model, 9-wk-oldC57BL/6 mice were put under anesthesia (300 mg/kg Avertin; Sigma-Aldrich)and instilled with 200 μL of sterile PBS or CP9 solution (2 × 106 cfu) intra-urethrally using sterile polyethylene tubing (Jelco I.V; Catheter Radiopaque,Smith Medical). For lineage tracing and tumor initiation studies, 5-wk-oldK14-mTmG and K14-Pten-mTmG mice first were treated with tamoxifen asdescribed previously (14). Briefly, tamoxifen (Sigma-Aldrich) was dissolved incorn oil and was injected i.p. at a dosage of 9 mg·40g−1·d−1 for 4 d consec-utively. Three weeks after the last tamoxifen injection, mice were intra-urethrally instilled with either PBS or CP9 inoculum.

Dissociation of Primary Prostates and Flow Cytometry. Dissociation of prostatecells and lineage fractionation were performed as described previously (55).For digestion, tissues were incubated with 10% (vol/vol) FBS DMEM con-taining collagenase/hyaluronidase (StemCell Technologies) for 3 h at 37 °Cwith rotation. After the medium was removed, 0.25% Trypsin-EDTA (Invitrogen)was added for 1-h additional digestion on ice. Subsequently, digested cells wereresuspended in medium containing 5 mg/mL dispase (Invitrogen) and 1 mg/mLDNase I (Roche Applied Science) and were pipetted for 1 min. Finally,dissociated cells were filtered through 100-μm cell strainers (BD Bio-sciences) to collect single cells. For lineage fractionation, single murineprostate cells were stained with FITC-conjugated anti CD31, CD45, andTer119 antibodies (eBioscience), PE-conjugated anti–Sca-1 antibody (eBio-science), and Alexa 647-conjugated anti CD49f antibody (Biolegend). To analyzethe population of immune cells in prostate, digested prostate tissueswere stained with FITC-conjugated anti-CD45, PE-conjugated anti-CD3,APC-conjugated anti-CD19, Pacific Blue-conjugated anti-F4/80, and PerCP/Cy5.5-conjugated anti-Ly-6G (Biolegend). The analyses were performed by usingLSR II or LSR Fortessa (BD Biosciences).

Histology, Masson’s Trichrome Staining, and Immunohistochemical Analysis.Collected prostates were fixed in 10% buffered formalin overnight at 4 °Cand embedded with paraffin. Five-micrometer–sectioned slides were ex-amined histologically by H&E. Masson’s trichrome stain was performedaccording to the manufacturer’s instructions (HT15-1KT; Sigma-Aldrich). Forimmunohistochemistry (IHC), sections were deparaffinized, and antigen re-

trieval was performed by steam heating in 0.01 M sodium citrate buffer (pH6.0) for 10 min in a steamer. Cooled slides were incubated with 5% normalgoat serum (Vector Labs) and with primary antibodies diluted in the 5%normal goat serum overnight at 4 °C. Primary antibodies used were rabbitanti-K5 (PRB-160P; Covance), mouse anti-K8 (MMS-162P; Covance), rat anti-K8 (Troma-1; Developmental Studies Hybridoma Bank), mouse anti-P63(4A4; Santa Cruz Biotechnology), rabbit anti-AR (SC-816; Santa Cruz Bio-technology), mouse anti-smooth muscle actin (Sigma-Aldrich), rabbit anti-vimentin (5741; Cell Signaling Technology), rabbit anti-Ki67 (NCL-Ki67-P;Novocastra), mouse anti-K14 (LL002; Biogenex), mouse anti-GFP (JL8; Clon-tech), chicken anti-GFP (ab13970; Abcam), rabbit anti-synaptophysin (18-0130; Invitrogen), rabbit anti-cleaved caspase 3 (9661; Cell Signaling Tech-nology), and rabbit anti-pAKT (3787; Cell Signaling Technology) Biotinylatedsecondary antibodies and streptavidin-conjugated HRP (Vector Laboratories)were used for chromatic visualization. For fluorescence visualization, slideswere incubated with secondary antibodies (diluted 1:1,000 in PBS containing0.05% Tween 20) labeled with Alexa-Fluor 488, 594, or 633 (Invitrogen).Sections were counterstained with DAPI (Sigma-Aldrich). Immunofluores-cence staining was imaged with a fluorescence microscope (Olympus BX60;Olympus Optical Co Ltd.) or confocal microscope (Leica EL6000; LeicaMicrosystems). Cell counting was performed via Image J Software.

Statistics. All experiments were performed using 3–11 mice in independentexperiments. Data are presented as mean ± SD. A Student t test was used todetermine significance between groups. For all statistical tests, the 0.05 levelof confidence was accepted for statistical significance.

ACKNOWLEDGMENTS. We thank Dr. Allison O’Brien for providing the CP9bacterial strain; Drs. Julienne Carstens and Jonathan Levitt for technicaladvice on transurethral instillation of bacteria; Drs. Jeffrey Rosen and AmyShore for critical comments; Joel M. Sederstrom for expert assistance withflow cytometry; and the Cytometry and Cell Sorting Core at Baylor Collegeof Medicine [National Institutes of Health (NIH) Grants AI036211, CA125123,and RR024574] for technical support. This work was supported by NIH GrantsR00 CA125937 and R01 DK092202 (to L.X.), U01 CA141497 and P20DK097775(to M.M.I.), and NIH Cancer Center Shared Resources Grant P30 CA125123.

1. Grivennikov SI, Greten FR, Karin M (2010) Immunity, inflammation, and cancer. Cell140(6):883–899.

2. Dennis LK, Lynch CF, Torner JC (2002) Epidemiologic association between prostatitisand prostate cancer. Urology 60(1):78–83.

3. MacLennan GT, et al. (2006) The influence of chronic inflammation in prostatic car-cinogenesis: A 5-year followup study. J Urol 176(3):1012–1016.

4. Hochreiter WW, Duncan JL, Schaeffer AJ (2000) Evaluation of the bacterial flora ofthe prostate using a 16S rRNA gene based polymerase chain reaction. J Urol 163(1):127–130.

5. Kazma R, et al. (2012) Association of the innate immunity and inflammation pathwaywith advanced prostate cancer risk. PLoS ONE 7(12):e51680.

6. De Marzo AM, Marchi VL, Epstein JI, Nelson WG (1999) Proliferative inflammatoryatrophy of the prostate: Implications for prostatic carcinogenesis. Am J Pathol 155(6):1985–1992.

7. Khalili M, et al. (2010) Loss of Nkx3.1 expression in bacterial prostatitis: A potentiallink between inflammation and neoplasia. Am J Pathol 176(5):2259–2268.

8. Luo JL, et al. (2007) Nuclear cytokine-activated IKKalpha controls prostate cancermetastasis by repressing Maspin. Nature 446(7136):690–694.

9. Endo Y, et al. (2007) Expression of activation-induced cytidine deaminase in humanhepatocytes via NF-kappaB signaling. Oncogene 26(38):5587–5595.

10. Hmadcha A, Bedoya FJ, Sobrino F, Pintado E (1999) Methylation-dependent genesilencing induced by interleukin 1beta via nitric oxide production. J Exp Med 190(11):1595–1604.

11. Schuettpelz LG, Link DC (2013) Regulation of hematopoietic stem cell activity by in-flammation. Front Immunol 4:204.

12. El-Alfy M, Pelletier G, Hermo LS, Labrie F (2000) Unique features of the basal cells ofhuman prostate epithelium. Microsc Res Tech 51(5):436–446.

13. Vashchenko N, Abrahamsson PA (2005) Neuroendocrine differentiation in prostatecancer: Implications for new treatment modalities. Eur Urol 47(2):147–155.

14. Choi N, Zhang B, Zhang L, Ittmann M, Xin L (2012) Adult murine prostate basal andluminal cells are self-sustained lineages that can both serve as targets for prostatecancer initiation. Cancer Cell 21(2):253–265.

15. Lu TL, et al. (2013) Conditionally ablated Pten in prostate basal cells promotes basal-to-luminal differentiation and causes invasive prostate cancer in mice. Am J Pathol182(3):975–991.

16. Wang ZA, et al. (2013) Lineage analysis of basal epithelial cells reveals their un-expected plasticity and supports a cell-of-origin model for prostate cancer hetero-geneity. Nat Cell Biol 15(3):274–283.

17. Liu J, et al. (2011) Regenerated luminal epithelial cells are derived from preexistingluminal epithelial cells in adult mouse prostate. Mol Endocrinol 25(11):1849–1857.

18. Ousset M, et al. (2012) Multipotent and unipotent progenitors contribute to prostatepostnatal development. Nat Cell Biol 14(11):1131–1138.

19. Goldstein AS, Huang J, Guo C, Garraway IP, Witte ON (2010) Identification of a cell oforigin for human prostate cancer. Science 329(5991):568–571.

20. Goldstein AS, et al. (2008) Trop2 identifies a subpopulation of murine and humanprostate basal cells with stem cell characteristics. Proc Natl Acad Sci USA 105(52):20882–20887.

21. Lawson DA, Xin L, Lukacs RU, Cheng D, Witte ON (2007) Isolation and functionalcharacterization of murine prostate stem cells. Proc Natl Acad Sci USA 104(1):181–186.

22. Xin L, Lawson DA, Witte ON (2005) The Sca-1 cell surface marker enriches fora prostate-regenerating cell subpopulation that can initiate prostate tumorigenesis.Proc Natl Acad Sci USA 102(19):6942–6947.

23. Leong KG, Wang BE, Johnson L, Gao WQ (2008) Generation of a prostate froma single adult stem cell. Nature 456(7223):804–808.

24. Burger PE, et al. (2005) Sca-1 expression identifies stem cells in the proximal region ofprostatic ducts with high capacity to reconstitute prostatic tissue. Proc Natl Acad SciUSA 102(20):7180–7185.

25. Hayward SW, Haughney PC, Lopes ES, Danielpour D, Cunha GR (1999) The rat pros-tatic epithelial cell line NRP-152 can differentiate in vivo in response to its stromalenvironment. Prostate 39(3):205–212.

26. Cunha GR, Lung B (1978) The possible influence of temporal factors in androgenicresponsiveness of urogenital tissue recombinants from wild-type and androgen-insensitive (Tfm) mice. J Exp Zool 205(2):181–193.

27. Xin L, Ide H, Kim Y, Dubey P, Witte ON (2003) In vivo regeneration of murine prostatefrom dissociated cell populations of postnatal epithelia and urogenital sinus mesen-chyme. Proc Natl Acad Sci USA 100(Suppl 1):11896–11903.

28. Taylor RA, et al. (2006) Formation of human prostate tissue from embryonic stemcells. Nat Methods 3(3):179–181.

29. Tuxhorn JA, Ayala GE, Rowley DR (2001) Reactive stroma in prostate cancer pro-gression. J Urol 166(6):2472–2483.

30. Xin L (2012) Cells of origin for cancer: An updated view from prostate cancer.Oncogene 32(32):3655–3663.

31. Krieger JN, Riley DE (2002) Bacteria in the chronic prostatitis-chronic pelvic pain syn-drome: Molecular approaches to critical research questions. J Urol 167(6):2574–2583.

32. Boehm BJ, Colopy SA, Jerde TJ, Loftus CJ, Bushman W (2012) Acute bacterial in-flammation of the mouse prostate. Prostate 72(3):307–317.

33. Elkahwaji JE, Ott CJ, Janda LM, Hopkins WJ (2005) Mouse model for acute bacterialprostatitis in genetically distinct inbred strains. Urology 66(4):883–887.

34. Elkahwaji JE, Zhong W, Hopkins WJ, Bushman W (2007) Chronic bacterial infectionand inflammation incite reactive hyperplasia in a mouse model of chronic prostatitis.Prostate 67(1):14–21.

35. Rippere-Lampe KE, et al. (2001) Cytotoxic necrotizing factor type 1-positive Escher-ichia coli causes increased inflammation and tissue damage to the prostate in a ratprostatitis model. Infect Immun 69(10):6515–6519.

Kwon et al. PNAS | Published online December 23, 2013 | E599

MED

ICALSC

IENCE

SPN

ASPL

US

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

36. Litvinov IV, et al. (2006) Low-calcium serum-free defined medium selects for growthof normal prostatic epithelial stem cells. Cancer Res 66(17):8598–8607.

37. van Leenders GJ, et al. (2003) Intermediate cells in human prostate epithelium areenriched in proliferative inflammatory atrophy. Am J Pathol 162(5):1529–1537.

38. Muzumdar MD, Tasic B, Miyamichi K, Li L, Luo L (2007) A global double-fluorescentCre reporter mouse. Genesis 45(9):593–605.

39. De Marzo AM, et al. (2007) Inflammation in prostate carcinogenesis. Nat Rev Cancer7(4):256–269.

40. Park JH, et al. (2002) Prostatic intraepithelial neoplasia in genetically engineeredmice. Am J Pathol 161(2):727–735.

41. Roden AC, et al. (2004) Augmentation of T cell levels and responses induced by an-drogen deprivation. J Immunol 173(10):6098–6108.

42. Mercader M, et al. (2001) T cell infiltration of the prostate induced by androgenwithdrawal in patients with prostate cancer. Proc Natl Acad Sci USA 98(25):14565–14570.

43. Krieger JN, Riley DE (2002) Prostatitis: What is the role of infection. Int J AntimicrobAgents 19(6):475–479.

44. Campeggi A, et al. (2013) Acute bacterial prostatitis after transrectal ultrasound-guided prostate biopsy: Epidemiological, bacteria and treatment patterns from a4-year prospective study. Int J Urol, in press.

45. Millan-Rodriguez F, et al. (2006) Acute bacterial prostatitis: Two different sub-cate-gories according to a previous manipulation of the lower urinary tract. World J Urol24(1):45–50.

46. Fong L, Ruegg CL, Brockstedt D, Engleman EG, Laus R (1997) Induction of tissue-specific autoimmune prostatitis with prostatic acid phosphatase immunization: Im-plications for immunotherapy of prostate cancer. J Immunol 159(7):3113–3117.

47. Liu KJ, et al. (1997) Identification of rat prostatic steroid-binding protein as a targetantigen of experimental autoimmune prostatitis: Implications for prostate cancertherapy. J Immunol 159(1):472–480.

48. Haverkamp JM, et al. (2011) An inducible model of abacterial prostatitis inducesantigen specific inflammatory and proliferative changes in the murine prostate.Prostate 71(11):1139–1150.

49. Hou Y, et al. (2009) An aberrant prostate antigen-specific immune response causesprostatitis in mice and is associated with chronic prostatitis in humans. J Clin Invest119(7):2031–2041.

50. Birbach A, et al. (2011) Persistent inflammation leads to proliferative neoplasia andloss of smooth muscle cells in a prostate tumor model. Neoplasia 13(8):692–703.

51. Palapattu GS, et al. (2006) Epithelial architectural destruction is necessary for bonemarrow derived cell contribution to regenerating prostate epithelium. J Urol 176(2):813–818.

52. Danielpour D (1999) Transdifferentiation of NRP-152 rat prostatic basal epithelial cellstoward a luminal phenotype: Regulation by glucocorticoid, insulin-like growth factor-I and transforming growth factor-beta. J Cell Sci 112(Pt 2):169–179.

53. Blackwood JK, et al. (2011) In situ lineage tracking of human prostatic epithelial stemcell fate reveals a common clonal origin for basal and luminal cells. J Pathol 225(2):181–188.

54. Lamb LE, Knudsen BS, Miranti CK (2010) E-cadherin-mediated survival of androgen-receptor-expressing secretory prostate epithelial cells derived from a stratified in vitrodifferentiation model. J Cell Sci 123(Pt 2):266–276.

55. Valdez JM, et al. (2012) Notch and TGFbeta form a reciprocal positive regulatory loopthat suppresses murine prostate basal stem/progenitor cell activity. Cell Stem Cell11(5):676–688.

56. Xin L, Lukacs RU, Lawson DA, Cheng D, Witte ON (2007) Self-renewal and multi-lineage differentiation in vitro from murine prostate stem cells. Stem Cells 25(11):2760–2769.

57. Mehra R, et al. (2007) Heterogeneity of TMPRSS2 gene rearrangements in multifocalprostate adenocarcinoma: Molecular evidence for an independent group of diseases.Cancer Res 67(17):7991–7995.

58. Bonkhoff H, Stein U, Remberger K (1994) The proliferative function of basal cells inthe normal and hyperplastic human prostate. Prostate 24(3):114–118.

59. Kyprianou N, Isaacs JT (1988) Activation of programmed cell death in the rat ventralprostate after castration. Endocrinology 122(2):552–562.

60. Hayward SW, Brody JR, Cunha GR (1996) An edgewise look at basal epithelial cells:Three-dimensional views of the rat prostate, mammary gland and salivary gland.Differentiation 60(4):219–227.

61. Lawson DA, et al. (2010) Basal epithelial stem cells are efficient targets for prostatecancer initiation. Proc Natl Acad Sci USA 107(6):2610–2615.

62. Stoyanova T, et al. (2013) Prostate cancer originating in basal cells progresses toadenocarcinoma propagated by luminal-like cells. Proc Natl Acad Sci USA 110(50):20111–20116.

63. Vasioukhin V, Degenstein L, Wise B, Fuchs E (1999) The magical touch: Genome tar-geting in epidermal stem cells induced by tamoxifen application to mouse skin. ProcNatl Acad Sci USA 96(15):8551–8556.

64. Wang S, et al. (2003) Prostate-specific deletion of the murine Pten tumor suppressorgene leads to metastatic prostate cancer. Cancer Cell 4(3):209–221.

E600 | www.pnas.org/cgi/doi/10.1073/pnas.1318157111 Kwon et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0