Adult murine prostate basal and luminal cells are self-sustained lineages that can both serve as targets for prostate cancer initiation

Summary

The prostate epithelial lineage hierarchy and the cellular origin for prostate cancer remain inadequately defined. Using a lineage tracing approach, we show that adult rodent prostate basal and luminal cells are independently self-sustained in vivo. Disrupting the tumor suppressor Pten in either lineage led to prostate cancer initiation. However, the cellular composition and onset dynamics of the resulting tumors are distinctive. Prostate luminal cells are more responsive to Pten null-induced mitogenic signaling. In contrast, basal cells are resistant to direct transformation. Instead, loss of Pten activity induces the capability of basal cells to differentiate into transformation-competent luminal cells. Our study suggests that deregulation of epithelial differentiation is a critical step for the initiation of prostate cancers of basal cell origin.

Introduction

Defining the cells-of-origin for cancer is of great value for accurate tumor prognosis and efficient prevention and therapeutics (Visvader, 2011). Previously, the identities of the cells-of-origin for cancer were assumed based on histological characterization of cancers. However, recent transcriptome studies have revealed that molecular signatures of cancer cells do not always match their histological appearance (Lim et al., 2009). Therefore, it can be misleading to determine cells-of-origin in the absence of functional lineage tracing studies (Molyneux et al., 2010).

Though prostate cancer is the second leading cause of cancer related death in males in the United States, it is still mainly described by qualitative clinical measurements including the TNM system and the Gleason grading system (Iczkowski and Lucia, 2011). The cellular origin of prostate cancer has not been definitively defined partly because the prostate epithelial lineage hierarchy per se has not been clearly characterized. Prostate epithelia are composed of three types of epithelial cells: luminal cells, basal cells, and neuroendocrine cells that are extremely rare (Abate-Shen and Shen, 2000). When prostate epithelial cells are cultured in vitro, a population of transit-amplifying cells is frequently observed (Litvinov et al., 2006; Peehl, 2005). These transit-amplifying cells express antigenic markers for both basal and luminal cells. Transit-amplifying cells are abundant at the developmental stage but are not detectable in vivo in adults under physiological conditions (Wang et al., 2001).

Using a functional prostate regeneration assay, several independent groups including ours have demonstrated that some basal cells in human and murine prostates can generate all three prostate epithelial cell lineages (Burger et al., 2005; Goldstein et al., 2010; Lawson et al., 2007; Leong et al., 2008; Xin et al., 2005; Xin et al., 2007; Zhang et al., 2011). In addition, Wang et al discovered that a rare castration-resistant luminal prostate cell population also possesses multipotent stem cell activity (Wang et al., 2009). Of note, the conclusions of those studies are based on experimental conditions involving cell transplantation that do not reflect physiological conditions. Therefore, it remains an open question whether the activities measured in those assays reflect the obligate or facultative function of prostate stem cells. Pertaining to this caveat, Liu et al recently showed by lineage tracing that prostate luminal cells are derived from preexisting luminal cells (Liu et al., 2011). In addition, a very recent comprehensive lineage tracing study on the mammary gland epithelial lineage hierarchy showed that in the postnatal mammary gland distinct stem cells contribute to the maintenance of the myoepithelial and luminal cell lineages (Van Keymeulen et al., 2011). A similar comprehensive study on prostate lineage hierarchy is required to address these controversies.

Previously, prostate luminal cells, transit-amplifying cells, and basal cells have all been implicated as the cells of origin for prostate cancer (Lawson and Witte, 2007). Two distinct functional approaches have been used recently to directly investigate the identity of the cellular origin for prostate cancer. One approach is to employ genetically engineered mouse models (Ellwood-Yen et al., 2003; Foster et al., 1997; Iwata et al., 2010; Majumder et al., 2003; Wang et al., 2003; Wang et al., 2009) to introduce oncogenic signaling in different prostate cell lineages. Most of these studies utilized two prostate specific promoters (Probasin and Nkx3.1). However, recently it has been shown that these promoters are active in both luminal cells and some basal cells (Mulholland et al., 2009; Wu et al., 2007; Zhang et al., 2011). Therefore, one can not determine definitively the identities of the cells of origin for cancer using this approach. Wang et al recently demonstrated that some very rare NKX3.1 positive luminal cells in castrated mice can serve as targets for transformation (Wang et al., 2009), but it remains undetermined whether other luminal cells also can serve as the cells of origin for cancer. A prostate-specific antigen-CreER^T2 model has recently been demonstrated to mediate luminal cell specific gene expression, but to date this model has not been extensively utilized (Liu et al., 2011; Ratnacaram et al., 2008).

The other approach is to genetically modify different prospectively isolated prostate epithelial cell lineages and investigate their tumorigenic potential by in vivo transplantation assays. Using this approach, we and others have demonstrated that murine and human prostate basal cells, but not luminal cells, can efficiently initiate prostate carcinogenesis under various oncogenic contexts (Goldstein et al., 2010; Lawson et al., 2010; Xin et al., 2005). However, a caveat for this approach is that luminal cells by nature do not proliferate and regenerate prostate tissues in the in vivo transplantation assay; hence these studies can not exclude that luminal cells also serve as the cells of origin for cancer in vivo. In summary, it has not been comprehensively determined previously whether individual cell lineages can serve as targets for transformation orthotopically in the prostate due to a lack of mouse models that enable strict lineage targeting in the prostate. This study aims to reveal how prostate basal and luminal cell lineages are maintained and the roles of these two lineages in prostate cancer initiation.

Results

Lineage tracing shows that prostate basal cells only generate basal cells in vivo

We employed a lineage-tracing approach to determine in vivo whether adult murine prostate basal epithelial cells would generate all three prostate epithelial cell lineages. A K14-CreER transgenic mouse line was generated previously (Vasioukhin et al., 1999), in which CreER is driven by the promoter of Keratin 14, a prostate basal cell marker. CreER encodes a Cre recombinase fused to a mutant estrogen ligand-binding domain so that its activity is activated only in the presence of tamoxifen. To investigate whether CreER is specifically expressed in prostate basal cells, K14-CreER transgenic mice were bred with mTmG reporter mice to generate K14-CreER^Tg/Tg;mTmG^Tg/Tg mice (hereafter referred to as the K14-mTmG mice). The mTmG mouse line is a double fluorescent reporter line that replaces the expression of a membrane-targeted Tomato-Red (mT) protein with a membrane-targeted enhanced green fluorescence protein (mG) upon Cre-LoxP mediated homologous recombination (Muzumdar et al., 2007) (Fig. 1A).

Fig. 1. Lineage tracing shows that prostate basal cells only generate basal cells in vivo.

(A) Schematic illustration of the lineage tracing strategy.

(B–D) Co-staining of GFP with K5 and K8 (B), K5 and P63 (C), and Synaptophysin (D) in tamoxifen-treated K14-mTmG mice. Arrowheads in B and C indicate GFP-labeled basal cells. Arrowhead in D points to a neuroendocrine cell.

(E) Timeline for androgen deprivation and replacement experiments.

(F–H) Co-staining of GFP with K5 and K8 (F), K5 and P63 (G), and Synaptophysin (H) in tamoxifen-treated K14-mTmG mice after induced epithelial turnover. Arrowheads in F and G indicate GFP-labeled basal cells. Arrowhead in H points to a neuroendocrine cell.

(I) Bar graph shows the percentage of GFP-labeled basal cells in lateral prostate lobes of K14-mTmG mice 5 days after tamoxifen induction (6 weeks) and after 2 cycles of epithelial regression-regeneration. Data represent means with SD. Also see Table S1.

(J) GFP-labeled basal cells (arrowhead) incorporated BrdU.

See also Fig. S1 and Table S1.

Tamoxifen was injected i.p. into male K14-mTmG bigenic mice. In contrast to the highly efficient Cre-mediated recombination observed in the skin (Fig. S1A), the recombination efficiency in the prostate was lower and varied among different prostate lobes. On average, 17% of basal cells in lateral lobes were pulse-labeled with GFP (Table S1). In contrast, rare and heterogeneously distributed GFP positive basal cells were observed in other lobes, which were hard to quantify. The recombination frequencies among lobes did not correlate with the K14 promoter activity (Fig. S1B). All GFP positive cells expressed the basal cell marker Keratin 5 (K5) (Fig. 1B, Table S1, n= 2095 cells from 5 mice). All of the observed cell nuclei of these GFP positive cells were positively stained with another nuclear-localized basal cell marker P63 (Fig. 1C, n= 1952 cells). In contrast, none of the examined GFP positive cells expressed the luminal cell marker Keratin 8 (K8), or the neuroendocrine cell marker Synaptophysin (Fig. 1B, D). These data demonstrate that the CreER expression is restricted to prostate basal cells in the K14-CreER model.

Adult murine prostate epithelia turn over extremely slowly under physiological conditions. To determine the fate of prostate basal cells, we induced extensive epithelial turnover by a classic prostate regression-regeneration model as schematically illustrated in Fig. 1E. In this model, prostate tissues atrophy and regenerate repeatedly in response to fluctuating serum testosterone levels. Substantial epithelial cell turnover was induced after two cycles of prostate regression-regeneration. IHC analyses showed that all GFP positive cells remained basal cells because they expressed K5, but not K8 or Synaptophysin (n= 1295 cells from 4 mice, Fig 1F–H). In addition, the percentage of GFP-labeled basal cells in lateral prostates did not change after induced epithelial turnover (Fig. 1I, Table S1), suggesting that either all basal cells possess equal regenerative capacity or the unipotent basal stem cells and differentiated basal cells are labeled at equal frequency. The GFP-labeled basal cells have proliferated because they incorporated BrdU (Fig. 1J). Collectively, these data demonstrate that during prostate regeneration the GFP labeled basal cells proliferated but only gave rise to prostate basal cells and did not differentiate or undergo lineage conversion to generate other cell lineages.

Prostate luminal cell lineage is self-sustained in vivo

We generated a K8-CreER^T2 mouse model via BAC transgenesis (manuscript submitted), in which CreER^T2 is driven by the promoter of the prostate luminal cell marker Keratin 8. The same lineage-tracing approach was employed using K8-CreERT2^Wt/Tg;mTmG^Wt/Tg (hereafter referred to as K8-mTmG) mice to determine how prostate luminal epithelial cells are sustained. GFP was undetectable in vehicle-treated K8-mTmG mouse prostates (data not shown), but was expressed abundantly in tamoxifen-treated mouse prostates (Fig. S2A). All GFP positive cells expressed K8, but not K5 or synaptophysin (n= 31,319 from 7 mice; Fig. 2A–B), demonstrating that CreER^T2 is only expressed by prostate luminal epithelial cells in this model. The recombination efficiency varied among different prostate lobes, with that in lateral lobes the highest (Figs. 2E and S2A, Tables S2 and S4). The recombination frequencies did not correlate with the K8 promoter activity (CreER expression level) among lobes (Fig. S2B). Therefore, the variation in recombination efficiency may be due to differential local tamoxifen concentrations and CreER activation status as a result of distinct blood vessel densities among these different lobes.

Fig. 2. The prostate luminal cell lineage is self-sustained in vivo.

(A–B) Co-staining of GFP with K5 and K8 (A), and Synaptophysin (B) in tamoxifen-treated K8-mTmG mice. Arrowheads and asterisks in A denote basal cells that are not GFP-labeled and GFP-labeled luminal cells, respectively. Arrow in B points to a neuroendocrine cell.

(C–D) Co-staining of GFP with K5 and K8 (C), and Synaptophysin (D) in tamoxifen-treated K8-mTmG mice after induced epithelial turnover. Arrowheads indicate that all basal cells are GFP negative. Arrow points to a neuroendocrine cell.

(E) Bar graph shows the percentage of GFP-labeled luminal cells 5 days after tamoxifen induction (6 weeks, black bars), after 4-month-ageing (red bars), after the first androgen deprivation (blue bars) and after 2 cycles of androgen deprivation-replacement experiments (green bars) in anterior (AP), ventral (VP), dorsal (DP) and lateral (LP) prostate lobes. Data represent means with SD. See Table S2–4 for more details.

(F) Timeline for investigating the origin of newly formed luminal cells by BrdU labeling.

(G) Bar graph shows that the frequency of newly formed GFP-positive luminal (GFP+K8+BrdU+/ K8+BrdU+) cells reflects that of GFP-labeled luminal (GFP+K8+/K8+) cells in all four prostate lobes. Data represent means with SD. Also see Table S2–4.

(H) A representative image showing BrdU incorporated into luminal cells during prostate regeneration.

See also Fig. S2 and Tables S2–4.

To determine how the prostate luminal cell lineage is maintained, we aged tamoxifen-treated K8-mTmG mice for 4 months. As shown in Fig. 2E and Table S2–4, the percentage of GFP positive luminal cells in individual lobes did not change after ageing. Since prostate tissues gain weight significantly from 6 weeks to 24 weeks (Fig. S2C), it is unlikely that the invariant percentage of GFP positive luminal cells after ageing is due to rare cellular turnover. Instead, this result suggests that the luminal cell lineage may be self-sustained. To further interrogate this possibility, tamoxifen-treated K8-mTmG mice were subjected to 2 cycles of prostate regression-regeneration to induce extensive epithelial turnover as illustrated in Fig. 1E. IHC analyses showed that the percentages of GFP positive luminal cells were not statistically different before and after induction of epithelial turnover (Fig. 2E, Table S2–4). In addition, all GFP positive cells still expressed K8, but not K5 or Synaptophysin (Fig. 2C–D), demonstrating that GFP-labeled luminal cells only generate other luminal cells. Overall these results imply that prostate luminal cells are not replenished by stem/progenitor cells from other cell lineages during prostate regeneration.

In addition, the approach illustrated in Fig. 2F was employed to directly investigate the origin of newly formed prostate luminal cells. Tamoxifen-treated K8-mTmG mice were subjected to one cycle of prostate regression-regeneration to induce extensive epithelial turnover. BrdU was administered to label dividing cells during the period of regeneration. As shown in Fig. 2G–H and Table S2–4, the GFP⁺ new luminal cells (BrdU⁺K8⁺) in individual prostate lobes were generated at a frequency equal to that of the initial pulse-labeling (i.e. the percentage of GFP⁺ luminal cells). These data suggest that if the labeling efficiency of luminal cells is 100%, all new K8⁺ luminal cells will be GFP⁺, which means that they would all come from cells in the luminal lineage existing at the time of castration and androgen replacement. In summary, these data directly demonstrate that the luminal cell lineage is self-sustained.

Prostate cancer initiated from basal cell-specific loss of function of PTEN

To determine the susceptibility of the prostate basal cell lineage to transformation induced by the loss of function of the tumor suppressor Pten, we generated K14-CreER^Tg/Tg;Pten^fl/fl (K14-Pten) bigenic mice. Tamoxifen was administered to 5-week-old K14-Pten mice so that Pten was specifically disrupted in prostate basal cells. An ARR2PB-Cre;Pten^fl/fl mouse model for prostate cancer was used as a control, in which Pten is disrupted in both basal and luminal cells (Wang et al., 2006; Zhang et al., 2011).

The disease progression status observed in all experimental mice is summarized in Table 1. Tamoxifen-treated K14-Pten mice displayed a shaggy fur phenotype (Fig. S3A) with complete penetrance as reported previously (Backman et al., 2004; Yao et al., 2006). Disrupting PTEN activity leads to phosphorylation of AKT that accumulates at the plasma membrane, which serves as a reliable marker for PTEN loss. AKT phosphorylation was uniformly detected in the epidermis (Fig. S3B–C), demonstrating an efficient Pten knockout in the skin. As we showed previously the activation of the CreER activity in K14-Pten mouse prostate basal cells was less efficient (Fig. 1), but Pten was disrupted only in prostate basal cells as demonstrated by co-staining pAKT with P63 (Fig. 3A-A″). Pten deletion in prostate basal cells was further confirmed by PCR genotyping in vitro cultured prostate spheres derived from basal cells in tamoxifen-treated K14-Pten mice (Mulholland et al., 2009) (Fig. S3D). Surprisingly, no abnormal epithelial growth was noted in all the mice examined 1 month post tamoxifen treatment (Table 1). However, by 3 months post tamoxifen treatment, focal hyperplastic growth (Fig. 3B) was observed in 4 out of 11 mice. Those neoplastic foci were graded between PIN1 to PIN4 using the nomenclatures and criteria developed by Park et al (Park et al., 2002). The PIN lesions, but not the adjacent normal tissues, expressed pAKT (Fig. 3C). They contained both K5 positive basal cells and K8 positive luminal cells (Fig. 3D–E). Luminal cells expanded in the PIN lesions as measured by the ratio of basal versus luminal cells in epithelia (Fig. 3F). Fig. 3G–H show clearly that pAKT was also expressed in some basal cells, further corroborating that Pten deletion has taken place in those basal cells.

Table 1.

Summary of disease progression in K14-Pten and K8-Pten mice

K14-PTEN
Month(s) post Tmx treatment	Tumor incidence	Disease stage/Numbers of mice	Mouse ID
1	0/5	Normal/5	321,322,323,324,411
3	4/11	Normal/6	293,504,505,509,544
		PIN1/1	295
		PIN2/1	294
		PIN3/1	55
4		Normal/1	92
		PIN4/1	91
6	29/31	Normal/2	711,713
		PIN2/2	58,59
		PIN3/2	52,712
		PIN4/4	51,53,56,112
		Early cancer/2	57,111
7		PIN3/1	50
		PIN4/9	47,48,49,93,94,95,98,456,460
8		PIN2/1	105
		PIN3/2	100,102
		PIN4/5	99,103,104,106,865
		Early cancer/1	101
K8-PTEN
Month(s) post Tmx treatment	Tumor incidence	Disease stage/Numbers of mice	Mouse ID
1	3/3	PIN1/1	89
		PIN2/1	69
		PIN3/1	146
2	9/9	Early cancer/2	159,924
3		PIN4/3	86,87,773
		Early cancer/3	89,772,774
4		Early cancer/1	307
6	15/15	PIN4/1	374
		Early cancer/12	226,229,237,375,451,454,455,498,499, 503,876,898
		Adenocarcinoma/2	456,903

Note: disease stage is defined by the most advanced foci observed in tissues, most of which are within dorsolateral lobes.

Fig. 3. Prostate cancer initiated from basal cell-specific loss-of-function of PTEN.

(A): Costaining of basal cell marker P63 with pAKT in K14-Pten mice 1 month post tamoxifen induction. Arrowheads point to pAKT⁺P63⁺ cells.

(B) H&E staining of a K14-Pten prostate 3-month post tamoxifen induction reveals the formation of PIN lesions (asterisks).

(C) PIN lesions (asterisks), but not adjacent normal glands, express pAKT.

(D–E) PIN lesions contain both K5-expressing basal cells and K8-expressing luminal cells. Arrowheads indicate K5-expressing basal cells that encapsulate glands.

(F) Quantification of the expansion of luminal cells in PIN lesions (region B) as compared in normal glands (region A). Data represent means with SD.

(G–H) Pten is disrupted in some basal cells. Arrowheads in G and G′ denote a P63-expressing basal cell that expresses pAKT. Arrowhead in H points to an anatomically typical basal cell that expresses pAKT.

See also Fig. S3.

Six to eight months post tamoxifen induction, 29 out of 31 mice developed neoplastic foci at various stages (Table 1). Fig. 4A-A′ shows images of K14-Pten mouse prostate glands 8-months post tamoxifen or vehicle treatment. The prostate lobes appeared less transparent in tamoxifen-treated mice. H&E staining revealed multifocal lesions that developed in all four lobes in this mouse (Fig. 4B–E). A majority of mice developed focal PIN4 and a few mice developed early prostate cancer most frequently in the DLP. In most cases, the lumen of a few DLP lobes in individual mice was filled with epithelial cells with nuclear pleomorphism and hyperchromasia displaying tufting and cribriform patterns. The K14-Pten cancers were mainly composed of cells that were either positive for K5 or K8 (Fig. 4F, H), with a small percentage of cells dual positive for K5 and K8. These double positive cells express pAKT, suggesting they were also derived from Pten null basal cells (Fig. S4A–C). In contrast, there were many K5⁺K8⁺ cells within the ARR2PB-Pten tumors (Fig. 4G, H). Most of the K5 positive cells in K14-Pten mice were also positive for P63 (Fig. 4I, J). Cells within the cancerous foci of K14-Pten mice express the androgen receptor and display secretory function as demonstrated by the immunostaining of the secretory proteins of murine dorsolateral lobes (mDLP) (Donjacour et al., 1990) (Fig. S4D–G). Finally, very few cells in K14-Pten tumors expressed the neuroendocrine cell marker synaptophysin (Fig. 4K).

Fig. 4. Progression of prostate cancer in the K14-Pten model.

(A-A′) Representative images of urogenital organs and dissected prostate lobes from K14-Pten mice 8 months after induction with vehicle (A′) or tamoxifen (A).

(B–E) Representative images of H&E staining of total prostate (B), anterior (AP, C), dorsolateral (DLP, D), and ventral (VP, E) prostate lobes of a K14-Pten mouse 8 months after tamoxifen treatment.

(F–H) Immunostaining of K5 and K8 of a K14-Pten mouse 3 months after tamoxifen treatment (F) and a 4-month old ARR2PB-Pten mouse (G). (H) Quantification of the percentage of K5⁺K8⁺ and K5⁺K8⁻ cells in the two models.

(I–J) Immunostaining of K5 and P63 in K14-Pten mice (I). (J) Quantification of the percentage of K5⁺P63⁺ and K5⁺P63⁻ cells.

(K) Synaptophysin-expressing neuroendocrine cells (arrow) are rare in K14-Pten prostate tumor. Arrowheads point to P63-expressing basal cells.

(L-L″) Prostate basal cells in normal glands in tamoxifen-treated K14-Pten mice express pAKT. Arrowheads point to P63-expressing basal cells that express pAKT.

(M–O) Representative images of H&E staining show that only mild focal PIN lesions (dot-circled or arrow-pointed regions) are developed in many K14-Pten mice 6–8 months post tamoxifen treatment.

(P–S) Dissociated prostate cells from K14-Pten tumors are capable of regenerating abnormal glandular structures. (P) H&E staining of the outgrown tissues. (Q–S) Immunostaining of K5 (Q, arrow) and K8 (Q, arrowhead), P63 (R, arrowhead), and pAKT (S, arrowhead). Insets in S indicate that both basal (arrowhead) and luminal (arrow) cells express pAKT.

See also Fig. S4.

Sporadic basal cells that express pAKT were observed in glandular structures displaying normal histology (Fig. 4L), suggesting that ablating Pten in basal cells is not sufficient to initiate prostate cancer. In contrast, cancer only initiated upon the emergence of pAKT-expressing luminal cells, thus suggesting that transition of basal cells into luminal cells may both be essential and be a limiting step for cancer initiation in this model. We found that disease progression dynamics among experimental mice varied, with some developing into PIN4 to early cancer; while others remained as PIN2/3 even 6–8 months post tamoxifen induction (Fig. 4M–O). This is probably because the transition from basal cells to luminal cells occurred with different kinetics in those mice. Overall, our study suggests that loss of function of PTEN induces differentiation of prostate basal cells into luminal cells, which is an essential step for disease initiation in this model.

To investigate whether K14-Pten tumor cells have functional repopulating activity, we enzymatically dissociated primary tumors into single cells, mixed them with embryonic urogenital sinus mesenchymal cells, and transplanted them subcutaneously into immunodeficient host mice for 2 months, as done previously using the PB-Pten mouse models (Mulholland et al., 2009). Fig. 4Q showed that K14-PTEN tumor cells were capable of regenerating hyperplastic prostate glandular structures. Those glands were composed of basal cells and luminal cells that both expressed activated pAKT (Fig. 4P–S), further confirming that they were Pten null.

Prostate cancer derived from luminal cell specific loss of function of PTEN

K8-CreERT2^Tg/wt;Pten^fl/fl (K8-Pten) mice were generated to determine whether prostate cancer initiates as a result of luminal cell-specific loss-of-function of Pten (Table 1). K8-Pten mice developed low grade PIN lesions in all lobes at full penetrance one month post tamoxifen treatment (Fig. 5A-A″). There were fewer pAKT-expressing cells in the AP (data not shown), suggesting that the homologous recombination is less efficient in the AP. We showed that even when Pten disruption was induced in very few luminal cells using a lower dosage of tamoxifen, hyperplastic foci still formed 1 month post tamoxifen induction (Fig. S5A–G). This suggests that the more rapid disease progression in the K8-Pten model, compared with the K14-Pten model, is due to the intrinsic differential response of these two cell lineages to Pten ablation, but is not likely because that Pten deletion was induced in more cells in the K8-Pten model. The prostate hyperplasia progressed to PIN4 or early cancers 2–4 months post-treatment (Fig. 5B-B″). Mitotic figures and large atypical cells with enlarged nuclei, prominent nucleoli, and hyperchromasia were observed. Early cancer or even frank adenocarcinoma formed unanimously 6 months post tamoxifen induction (Fig. 5C-C″). More than 95% of the glands in DLP and VP prostates displayed a tufting and cribriform growth pattern, while the anterior prostates (AP) were more histologically heterogeneous, consisting of both normal and cancerous glands. This is probably due to the relatively lower frequency of homologous recombination in AP, as mentioned above.

Fig. 5. Prostate cancer derived from luminal cell specific disruption of PTEN.

(A–C) H&E staining of anterior (AP), ventral (VP), and dorsolateral (DLP) prostate lobes of K8-Pten mice at 1, 3, and 6 months post tamoxifen induction. Arrows point to focal hyperplasia in AP.

(D–F) Immunostaining of K5 and K8 reveals a distinct lineage composition among VP (D), AP (E), and DLP (F).

(G–I) P63-expressing basal cells (arrowheads) do not express pAKT in VP (G), AP (H), and DLP (I).

(J–L) Only K5-expressing basal cells residing at the basement membrane express P63 (arrowheads). VP (J), AP (K), and DLP (L).

(M–R) K8-Pten prostate tumor can repopulate. Representative images of H&E staining of outgrown tissues from dissociated K8-Pten tumor tissues (M–O). Immunostaining of K5 and K8 (P and Q), and pAKT (R) of outgrown tissues. Arrows indicate cancerous foci.

See also Fig. S5.

The PIN lesions in ventral prostates were composed predominantly of K8 positive luminal cells, while K5 or P63 positive basal cells were almost completely lost (Fig. 5D). In comparison, in the AP and DLP there was a dramatic expansion of the cells that were dual positive for K5 and K8 (Fig. 5E, F), similar to the double positive cells observed in the ARR2PB-Pten model (Fig. 4G). The K5 staining in these dual positive cells was cytoplasmic, in sharp contrast to the typical K5 staining in basal cells that highlights cellular contour. Since these double positive cells expressed activated AKT, they should have been derived from K8 positive luminal cells, presumably becoming putative “transit amplifying cells” through dedifferentiation (Fig 5H, I) (Litvinov et al., 2006). They did not express P63 (Fig 5G–I). In contrast, P63 was only expressed by the K5 positive basal cells residing at the basement membrane and did not express activated AKT (Fig 5J–L), suggesting that these P63-expressing cells were bona fide basal cells which did not undergo homologous recombination. The differential phenotypes among lobes were observed consistently along the 6 months post tamoxifen treatment. Cancer cells in all lobes express the androgen receptor and mDLP, while synaptophysin-expressing neuroendocrine cells were very rare in all lobes (Fig. S5H–N). The same observations were made when both Pten and P53 were knocked out in luminal cells using K8-CreERT2^Tg/Wt;Pten^fl/fl;P53^fl/fl (K8-Pten;P53) mice (Data not shown).

Transplantation assays again were performed to determine whether K8-Pten tumor cells were capable of functional repopulation. H&E staining show that the outgrowth tissues contained both normal glandular structures and cancerous lesions (Fig. 5M–O). The normal prostate glandular structures were composed of a single layer of epithelial cells encircling lumen filled with eosinophilic secretions. Both prostate basal and luminal cells were detected in those glands (Fig. 5P). They did not express pAKT (data not shown). We previously showed that only prostate basal cells are capable of forming such prostate glandular structures upon transplantation (Lawson et al., 2007; Xin et al., 2005). These results suggest that those normal glands were derived from prostate basal cells and further corroborate that Pten was not disrupted in prostate basal cells in the K8-Pten model. In contrast, the focal cancerous lesions were composed of cells that were double positive for K5 and K8 and expressed pAKT (Fig. 5Q, R). They did not express P63 (data not shown). This phenotype recapitulates that of the original tumors in the AP and DLP, suggesting that those lesions were derived from Pten-null luminal cells. Collectively, these data further support that disruption of Pten in prostate luminal cells causes prostate cancer.

Castration resistant prostate cancer cells exist in K8-Pten prostate tumors

To investigate the response of K8-Pten tumors to androgen deprivation, we castrated K8-Pten mice 4 months after tamoxifen induction. Seminal vesicles and prostate tumors shrunk significantly after androgen ablation (Fig. 6A, A′), which demonstrated successful androgen ablation and corroborated that many cancer cells were dependent on androgen for their survival. Histological analysis confirmed substantial cell death in the prostate 10 days and 2 months post castration as evidenced by significant apoptotic bodies inside tumor masses and cellular debris inside prostate lumen (Fig. 6B, C, D). IHC analysis showed that two months post castration the remaining VP tumors were composed predominantly of luminal cells (Fig. 6E). The K5 and K8 double positive cells decreased significantly, but still persisted in the AP and DLP tumors (Fig. 6F, G). These data demonstrate that these luminal cells survive androgen deprivation and suggest that they may serve as the cellular origin for castration-resistant prostate cancer.

Fig. 6. Castration resistant prostate cancer cells exist in K8-Pten prostate tumors.

(A-A′) Representative images of urogenital organs excluding seminal vesicles from K8-Pten mice 2 month after a mock surgery (A) and castration (A′). BL: bladder. UR: Urethra. AP: anterior prostate. VP: ventral prostate. DP: dorsal prostate. LP: lateral prostate.

(B–G) H&E staining (B–D) and immunostaining of K5 and K8 (E–G) of VP (B, E), AP (C, F), and DLP (D, G) of K8-Pten mice before castration (B–D), and 10 days (B′–D′), and 2 months (B″–D″) after castration. Insets in 6E–6G show staining of K5 and K8 at bracketed regions in respective images.

Prostate basal cells are resistant to direct oncogenic transformation

The above studies showed that prostate basal cells were less sensitive than luminal cells to mitogenic signals mediated by Pten deletion. We wondered whether this was due to the insufficient oncogenic potency conferred by Pten loss or the indolent nature of basal cells to transformation. P53-dependent PTEN deletion-induced cellular senescence has been shown to impede the rapid progression of fully developed prostate adenocarcinoma in the Pten null prostate cancer model (Alimonti et al., 2010; Chen et al., 2005). We generated K14-CreER^Tg/Tg;Pten^fl/fl;P53^fl/fl (K14-Pten;P53) mice and sought to determine whether prostate basal cells can be directly transformed upon simultaneous deletion of Pten and P53. Two months post tamoxifen induction, experimental mice developed severe hyperplastic growth in facial skin, showed signs of morbidity such as loss of weight and hunched postures, and had to be euthanized (Fig. S6A–D). Seminal vesicles and prostates from tamoxifen treated K14-Pten;P53 mice were much smaller as compared to the control mice, probably due to morbidity (Fig. S6E). H&E staining showed that prostate epithelial cells packed tightly due to reduced cytoplasmic volumes. There was no signs of hyper-proliferation (Fig. 7A, B, N=9 mice), which was corroborated by a lack of Ki67 expressing cells (data not shown). The glandular structures were mostly composed of K5 positive basal cells that encapsulated K8 positive luminal cells (Fig. 7C). Occasionally, basal cells were reduced in number in some glands (Fig. 7D). IHC analyses showed clearly that pAKT was activated only in P63-expressing basal cells, demonstrating that Pten was only disrupted in basal cells. (Fig. 7E). PCR analysis confirmed successful ablation of P53 (Fig. S6F). We were able to keep four K14-Pten;P53 mice for 3 months after tamoxifen treatment and observed one PIN4 lesion in one mouse (Fig. 7F, arrow), which demonstrates that cancer can initiate from basal cells with Pten and P53 deletion. The PIN4 lesion is mainly composed of luminal cells expressing K8, with some cells expressing K5 only or both K5 and K8 (Fig. 7G). Some P63-expressing basal cells expressed pAKT, suggesting Pten was disrupted in those cells (Fig. 7H). To exclude that the morbidity of experimental mice interferes with disease progression in the prostate, we collected intact prostate lobes from K14-Pten;P53 mice 2-month post tamoxifen treatment and transplanted them under the kidney capsules of immunodeficient male hosts. Only a few PIN1 lesions were noted in 7-week transplants (Fig. 7I–K). In contrast, massive lesions at the PIN4 or early cancer stages were observed in 15-week transplants (Fig. 7L–N). Within those cancerous regions, some P63-expressing basal cells express pAKT while other do not, suggesting that Pten was disrupted in only a fraction of basal cells (Fig. 7O–P). Collectively, these results imply that basal cells in situ are by nature relatively resistant to direct transformation by oncogenic stimuli, which partly explains why prostate basal cell carcinoma is so rare (Ali and Epstein, 2007).

Fig. 7. Prostate basal cells are resistant to direct oncogenic transformation.

(A–B) H&E staining of K14-Pten;P53 mouse prostates 2 months after vehicle (A) or tamoxifen (B) induction. A′ and B′ show images of higher magnification.

(C–D) Representative images of immunostaining of K5 and K8 of prostates from K14-Pten;P53 mice 2-month post tamoxifen induction.

(E) Pten is disrupted only in basal cells in K14-Pten;P53 mice, as demonstrated by costaining of pAKT (E, E′) and P63 (E, E″). Arrowheads indicate pAKT-expressing cells that also express P63.

(F–H) H&E (F) and immunostaining of K5/K8 (G) and pAKT/P63 (H) of K14-Pten;P53 mouse prostates 3 months after tamoxifen induction. Arrow in F points to a dot-circled PIN4 lesion. Arrowhead in H points to a basal cell expressing pAKT.

(I–P) Two months after tamoxifen treatment, K14-Pten;P53 prostate glands were transplanted under renal capsules of immunodeficient male hosts. H&E staining (I, L, M) and immunostaining of K5/K8 (J, N) and pAKT/P63 (K,O,P) of 7-week (I–K) and 15-week (L–P) transplants. Arrowheads in O and P denote P63-expressing cells that express and do not express pAKT, respectively.

See also Fig. S7.

Discussion

Obligate versus facultative stem cell capacity of the prostate basal cells

Our study demonstrates that adult murine prostate basal and luminal cell lineages are independently sustained. We and others also showed previously using a transplantation-based prostate regeneration assay that some prostate basal cells in both human and rodents are capable of generating all three prostate epithelial cell lineages (Burger et al., 2005; Goldstein et al., 2010; Lawson et al., 2007; Leong et al., 2008; Xin et al., 2005; Zhang et al., 2011). These two observations are not mutually contradictory since they may illustrate the obligate versus facultative activities of stem cells in the prostate basal cell lineage. In the prostate regeneration assay, basal cells are dissociated into single cells, removed from their natural environmental cues, and co-cultured with embryonic urogenital sinus mesenchyme (UGSM) cells (Xin et al., 2003). It has been shown that embryonic stromal cells provide inductive signals that are absent in adult murine stromal cells, which affects epithelial cell differentiation or even changes their lineage status (Neubauer et al., 1983; Taylor et al., 2006). On the other hand, stromal-epithelial and epithelial-epithelial interactions in adult mouse prostate may mediate signaling that prevents basal cells from differentiating into other lineages. Our observations from the K14-mTmG and K14-Pten models demonstrate that oncogenic signals like Pten loss can alter the differentiation program of basal cells. In conclusion, we demonstrated that at least some prostate basal cells possess unipotent stem cell activity to maintain this lineage during prostate homeostasis and regeneration. In contrast, the results obtained using the prostate regeneration assay revealed the plasticity of the basal cell lineage in response to changes in environmental cues.

Of note, since only up to 21% of prostate basal cells were labeled with GFP in the K14-mTmG model, we can not exclude the possibility that some of unlabeled basal cells can generate luminal cells and neuroendocrine cells. However, our complementary lineage tracing experiment using the K8-mTmG model suggests that the luminal cell lineage is mainly self-sustained. Therefore, even if these additional multi-potent basal cells exist, differentiation of basal cells into luminal cells most likely would be rare during prostate regeneration. Recently, similar conclusions have been made in the mammary gland (Van Keymeulen et al., 2011). Though the mammary gland myoepithelial cells are capable of regenerating the mammary gland in vivo in transplantation assays, genetic lineage tracing experiments demonstrated that the myoepithelial and luminal cell lineages are independently maintained in adults.

Maintenance of the luminal cell lineage

Previously, a population of castration-resistant NKX3.1-expressing (CARN) luminal cells was shown to be able to generate all three prostate epithelial lineages (Wang et al., 2009). However, we did not observe in the K8-mTmG mice any descendant of GFP positive luminal cells that expressed K5 or Synaptophysin, suggesting that GFP-labeled luminal cells do not differentiate or convert into the other two epithelial cell lineages at least in this mouse model. Since not all prostate luminal cells were labeled with GFP in the K8-mTmG model, we can not exclude the possibility that CARN cells were preferentially not labeled under our experimental conditions.

Several lines of evidence imply the existence of a lineage hierarchy within the luminal cell lineage. Luminal cells are heterogeneous with regard to their capacity to retain BrdU labeling (Tsujimura et al., 2002). The androgen receptor expression level and activation status are heterogeneous among adult murine luminal cells (L.X. unpublished observation). In addition, some luminal cells can survive androgen deprivation for an extended period (Tsujimura et al., 2002). It has been suspected that those androgen-independent luminal cells represent the committed progenitor cells in the luminal lineage. Our results showed that the percentage of GFP positive luminal cells in castrated mice was approximately the same as that in intact mice (Fig. 2G), which suggests that differentiated androgen-dependent luminal cells and androgen-independent luminal progenitor cells were GFP-labeled at a similar frequency. To date, the identity of the putative “luminal progenitor cells” remains undefined.

Alternatively, the androgen-independent survival of prostate luminal cells may be accounted for by a stochastic model, in which any luminal cells could be conferred with the capacity for castration resistance when they happen to reside in a specific niche, such as a specialized anatomical location or a direct contact with a certain sub-type of basal cells. In this scenario the luminal cell lineage may be sustained simply by cell duplication, like the β cells in the pancreas (Dor et al., 2004). Future effort should be made to distinguish the two models.

Prostate basal cells as the cellular origin for prostate cancer

Our studies demonstrated that although both prostate basal and luminal cells can serve as the cellular origin for prostate cancer, prostate luminal cells are more sensitive to mitogenic signaling while basal cells are relatively resistant to transformation. This is consistent with the fact that luminal cells possess low levels of H2A.X hence are more vulnerable to oncogenic stress (Jaamaa et al., 2010). The distinctive responses of these two cell lineages to oncogenic insults explain why treatment-naïve prostate cancers are mostly composed of luminal cells while prostate basal cell carcinoma is very rare.

Intuitively, prostate basal cells would seem to be the preferred cellular origin for cancer because they are more prone to accumulating genetic alterations than luminal cells. They are less well-differentiated and proliferate more frequently (Bonkhoff et al., 1994). Additionally, they are proposed to act as a natural barrier to protect the luminal cell lineage from various insults. Thus they are exposed to a more “hostile” environment than luminal cells. For examples, the basal cells are in closer contact with various cancer-promoting cytokines generated by surrounding reactive stroma as a result of chronic inflammation (Tuxhorn et al., 2001). However, since most prostate cancers display a luminal cell phenotype, differentiation of basal cells into luminal cells becomes an essential and probably a rate limiting step for cancer initiation and progression, if the cellular origin for cancer is of the basal cell lineage. This is supported by our result from the K14-Pten model. Prostate cancer was initiated in the K14-Pten model with an increased latency. Though Pten was disrupted specifically only in the prostate basal cells, the initiation of cancer did not start until the emergence of pAKT-expressing luminal cells. Since direct differentiation of basal cells to luminal cells is absent under physiological conditions based on our lineage tracing experiments, these results suggest that deregulation of the normal prostate epithelial differentiation program is a critical step for initiation of human prostate cancer with a basal cell origin.

It should be noted that the most ideal way to investigate the cells of origin for cancer is to perform lineage tracing in the cancer models. However, we were not able to perform our study in this way for two reasons. First, activation of CreER^T2 by tamoxifen is transient. Therefore, homologous recombination may not be achieved in all the cells that express CreER^T2, as we have shown in the K14-mTmG and K8-mTmG lineage tracing experiments (Figs. 1 and 2). Second, the recombination efficiencies at different genomic loci are not identical. As shown in Fig. S5O–P, GFP expression does not guarantee Pten deletion in tamoxifen-treated K8-Pten^fl/fl-mTmG triple transgenic mice. Nevertheless, we found that the specificities of the K14 and K8 promoters are not affected by the genetic background of the experimental mice (Fig. S5Q–R). In addition, we confirmed lineage specific Pten deletion by costaining pAKT with lineage markers in the K14-Pten and K8-Pten models. Therefore, we can conclude that neoplasia and tumors in the K14-Pten and K8-Pten models are derived from basal and luminal cells, respectively.

A unique feature of prostate cancer is that the disease is strictly age-dependent. Men under 35 seldom develop prostate cancer. Our data showed that it takes at least 3 months for K14-Pten mouse prostate basal cells to differentiate into luminal cells, which is almost equivalent to 10 years of human life, suggesting that this is still an extremely lengthy and inefficient biological process under certain genetic contexts. This may provide as an additional explanation for the strict age-dependent nature of human prostate cancer.

Prostate luminal cells as the cellular origin for primary and castration resistant prostate cancer

Our genetic model also supports that luminal cells can be the cellular origin for prostate cancer. An intriguing observation from our K8-Pten model is that the same genetic change causes tumors with lobe-specific phenotypes with regard to their cellular composition. The phenotype in VP resembles that of the human disease because markers of basal cells are rare in prostate carcinoma, which has served as a diagnostic criterion for prostate cancer. In comparison, tumors in the AP and DLP are composed of cells that express both K5 and K8, and basal cells remain largely unaffected. A previous study has revealed a compartmentalization of gene expression between prostate lobes and identified dozens of differentially expressed genes between prostate lobes (Abbott et al., 2003). It is possible that the cellular context, i.e. the intrinsic differences in gene expression between prostate lobes lead to their differential responses to the oncogenic insult. It will be interesting to investigate whether human prostate luminal cells share more similarity with murine VP luminal cells in terms of gene expression profiles. The mechanism by which basal cells are depleted in the VP is also unknown. We recently reported that dissociation of basal cells leads to cellular apoptosis induced by activation of the RhoA/ROCK kinases (Zhang et al., 2011). One potential mechanism, therefore, could be that the basal cell layer in ventral prostate is less resilient to cellular perturbations, so that when prostate glands are enlarged and distorted due to excessive proliferation of prostate luminal cells, the cellular contact between ventral prostate basal cells is attenuated, which alters the signaling that regulates basal cell survival, such as the RhoA/ROCK-mediated signaling.

Our study also demonstrates that castration-resistant cells in prostate cancers may originate from luminal cells. This result excludes the possibility that prostate basal cells are the only cellular origin for castration resistant disease. Since there is no evidence that CARN cells (Wang et al., 2009) were genetically manipulated by tamoxifen induction in the K8-CreER^T2 model (Fig. 2), our study also suggests that CARN cells are not the only cells of origin for prostate cancer in the luminal cell lineage, and that castration resistance is not a unique feature of CARN cells. This observation further highlights the need to identify the androgen-independent prostate luminal progenitor cells and to determine the essential signaling that confers the capacity for androgen-independent survival to luminal cells.

Experimental procedures

Mouse procedures

The sources of experimental mice and the genotyping strategies were described in Supplementary Experimental Procedures. Castration, androgen replacement, dissociation of primary prostate tumors, and tumor cell transplantation were performed using standard techniques described in Supplementary Procedures. Tamoxifen (Sigma-Aldrich, St. Louis, MO) was dissolved into corn oil and was administrated i.p. into experimental mice at the age of 5 weeks once a day for four consecutive days. The dosages used for K14-CreER and K8-CreER^T2 mice were 9mg/40g and 2 mg/40g, respectively. All animal work were approved by and performed under the regulation of the Institutional Animal Care Committee of the Baylor College of Medicine.

Histology, immunohistochemical and immunofluorescent analysis

The procedures for histological and IHC analyses and the information about primary antibodies were described in Supplementary Experimental Procedures. Paraffin embedded sections were stained and counterstained with 4,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich, St. Louis, MO). Secondary antibodies were labeled with Alexa Fluro 488, 594, 633 (Invitrogen, Carlsbad, CA). Immuno-fluorescence staining was imaged using an Olympus BX60 fluorescence microscope or a Leica EL6000 confocal microscope. Cell counting was performed either manually or via ImagePro Software.

Statistics

All experiments were performed using 4–10 mice in independent experiments. Data are presented as mean ± SD. Student’s t test was used to determine significance between groups. For all statistical tests, the 0.05 level of confidence was accepted for statistical significance.

Significance.

Understanding the cellular origin for cancer can help improve disease prevention and therapeutics. Our genetic studies directly demonstrate that prostate cancer can initiate from both the basal and luminal cell lineages. However, prostate basal cells are less susceptible than luminal cells to direct transformation. Instead, disease initiation from prostate basal cells requires oncogenic signaling-induced differentiation of adult prostate basal cells into luminal cells, that is absent under physiological conditions. These studies suggest that deregulation of the normal prostate epithelial differentiation program is a critical step for initiation of human prostate cancer with a basal cell origin. Furthermore, suppressing signaling pathways that induce basal-luminal differentiation may provide an efficient approach to prevent prostate cancer initiation.

Highlights.

• Adult prostate basal and luminal cell lineages are independently self-sustained.

• Prostate cancer can initiate from both basal and luminal cells.

• Altering prostate basal cell differentiation is a major step for cancer initiation.

• Prostate basal cells in situ are relatively resistant to direct transformation.