Identification of CD166 as a Surface Marker for Enriching Prostate Stem/Progenitor and Cancer Initiating Cells

New therapies for late stage and castration resistant prostate cancer (CRPC) depend on defining unique properties and pathways of cell sub-populations capable of sustaining the net growth of the cancer. One of the best enrichment schemes for isolating the putative stem/progenitor cell from the murine prostate gland is Lin-;Sca1+;CD49fhi (LSChi), which results in a more than 10-fold enrichment for in vitro sphere-forming activity. We have shown previously that the LSChi subpopulation is both necessary and sufficient for cancer initiation in the Pten-null prostate cancer model. To further improve this enrichment scheme, we searched for cell surface molecules upregulated upon castration of murine prostate and identified CD166 as a candidate gene. CD166 encodes a cell surface molecule that can further enrich sphere-forming activity of WT LSChi and Pten null LSChi. Importantly, CD166 could enrich sphere-forming ability of benign primary human prostate cells in vitro and induce the formation of tubule-like structures in vivo. CD166 expression is upregulated in human prostate cancers, especially CRPC samples. Although genetic deletion of murine CD166 in the Pten null prostate cancer model does not interfere with sphere formation or block prostate cancer progression and CRPC development, the presence of CD166 on prostate stem/progenitors and castration resistant sub-populations suggest that it is a cell surface molecule with the potential for targeted delivery of human prostate cancer therapeutics.


Introduction
Despite advances in the early detection and management of prostate cancer, castration resistant prostate cancer (CRPC) remains the second most common cause of male mortality in the United States [1]. Mounting evidence suggests that a subpopulation of prostate cells can initiate prostate cancer and may be responsible for the castration resistance [2,3,4,5]. Therefore, these cancer initiating cells [6] may serve as promising cellular targets for prostate cancer and identification of this subpopulation has become the necessary step toward future effective therapy.
The origins of prostate cancer initiating cells are controversial [7,8]. Normal prostate from human or mouse contains three different types of cells, namely luminal secretory, basal and neuroendocrine cells. Since human prostate cancer is characterized by loss of basal cells and expansion of luminal cells, several animal models posit that luminal-specific progenitors are the sources of prostate cancer initiation [9,10,11]. However, using the tissue regeneration approach, basal cells have proved to be more efficient oncogenic targets for both human and mouse prostate cancer initiation [12,13]. Interestingly, Xin's group demonstrated that adult murine prostate basal and luminal cells are selfsustained lineages that can both serve as oncogenic targets for prostate cancer initiation [14].
PTEN plays an important role in human prostate cancer and CRPC development [15] and is inactivated in 20% of primary and 60% of metastatic lesions [16]. The murine Pten prostate cancer model (Pb-Cre + ;Pten L/L ) recapitulates the disease progression seen in humans, including CRPC [17,18,19,20], and shares many signature genetic changes with human disease [17]. Importantly, the Pb-Cre + ;Pten L/L model provides a unique tool for studying tumor initiating cells as the majority of luminal cells and subpopulations of basal cells have Pten deletion [17,18]. Using this model, we demonstrated that Pten deletion causes an expansion of basal and transient amplifying subpopulations and subsequent tumor initiation in vivo [18]. We further showed Lin -Sca-1 + CD49f hi (LSC hi ) prostate stem/progenitor cells from the Pten null prostate are capable of initiating a cancerous phenotype that mimics the primary cancer in the Pten null prostate model [19].
Here, we report the identification of a cell surface marker, CD166 or Activated Leukocyte Cell Adhesion Molecule (CD166/ ALCAM) that is highly upregulated in human and murine CRPC samples. CD166 can be used to enrich for stem/progenitor sphere-forming cells from both WT and Pten null mutant mouse prostates. In addition, CD166 can separate LSC hi mouse stem/ progenitor cells into CD166 hi and CD166 lo subpopulations, with the LSC hi ;CD166 hi subpopulation having much higher sphereforming activity. We further demonstrate that CD166 can be used as an enrichment maker for isolating human prostate sphereforming cells and tubule-forming cells.

CD166 Expression is Upregulated in Murine Castrated
Prostatic Epithelium and can be used for Enriching Stem/ progenitor Cells Rodent prostate contains stem-like cells that are enriched in the castrated prostate gland and can undergo more than 15 cycles of involution-regeneration in response to androgen withdrawal and replacement [21]. We reasoned that castration may also lead to upregulation or enrichment of those stem cell surface molecules that can potentially serve as marker for isolating stem/progenitor cells and for targeted drug delivery. We therefore mined publically available databases describing gene expression profiles of murine prostates at day 0 and day 3 post-castration [22,23]. We focused on those genes that fell in the gene ontology category of 'plasma membrane' and identified CD166/ALCAM as one of only two common castration-enriched cell surface molecules (Table S1). CD166 was significantly increased (1-tail t-test ,0.015) 3 days after castration as compared to intact mice. While Cxcl12 is also upregulated, we chose not to focus on this gene as it is a chemokine and not amenable for FACS-mediated stem/progenitor cell enrichment.
CD166 is a type I transmembrane protein of the Ig superfamily that mediates cell-cell interactions via heterophilic (CD166-CD6) and/or homophilic (CD166-CD166) mechanisms [24,25]. We found that in the intact mice, CD166 is preferentially expressed in the stem/progenitor-enriched proximal region [21] but low in the stem/progenitor-poor distal region of the WT prostate ( Figure 1A upper panels). CD166 protein levels are also up-regulated immediately following castration (Figure 1A lower panels; comparing day 0 and day 3 post-castration).
Prostate stem/progenitor cells are characterized by their ability to form spheres in vitro [26]. We performed the sphere-forming assay using sorted CD166 hi and CD166 lo cells and found that CD166 hi cells have significantly higher sphere-forming activity compared to CD166 lo cells ( Figure 1B, left). Since we had previously developed the LSC hi enrichment scheme [26], which yields 10-fold enrichment of WT sphere-forming cells, we tested whether CD166 can be used for further enriching sphere-forming activity. We gated LSC hi cells according to their CD166 expression and found that LSC hi ;CD166 hi cells have 5-fold higher sphere-forming activity as compared to their LSC hi ;CD166 lo counterpart ( Figure 1B, right). Therefore, CD166 can be used as a marker to further enrich sphere forming cells within the WT prostate. Serial passaging of the spheres generated from LSC hi ;CD166 hi cells demonstrated that this enhanced sphere-forming activity could be maintained in vitro through at least three passages ( Figure S1A). In contrast, less spheres were generated from LSC hi ;CD166 lo cells (P0-P2) and cannot undergo continuous passage due to the limited cell number. We observed no significant difference in the sphere size distribution between LSC hi ;CD166 hi generated spheres and LSC hi ;CD166 lo generated spheres ( Figure  S1B and S1C). Similar to the LSC hi subpopulation [26], castration also leads to significant enhancement of the LSC hi ;CD166 hi subpopulation ( Figure 1C).

CD166 hi Human Prostate Cells Have Higher Sphere Forming and Regeneration Potential
Certain cell surface markers, such as Sca-1, are only expressed in the mouse and therefore cannot be used for isolation of human stem/progenitor cells. CD166, on the other hand, is expressed in various human organs and upregulated in human cancers, including prostate cancer [27]. To determine whether CD166 can be used for enriching human prostate stem/progenitors, we first examined its expression and found that CD166 is highly expressed in the developing human fetal prostate epithelium ( Figure 2A, left panel) and focally expressed in the benign adult prostate, which overlaps with a subset of TROP2 and CD49fpositive cells (Figure 2, middle and right panels).
We then evaluated whether CD166 could be used as a marker for enriching human stem/progenitor cells. Benign regions of prostate tissue were collected from multiple patients who underwent radical prostatectomy and dissociated to single cells. Consistent with our previous studies [13,28], the percentages of CD166 + cells vary from patient to patient (data not shown). However, the majority of sphere forming activity was identified in the CD166 hi population ( Figure 2B), similar to our findings with murine prostate cells. Data are shown from 6 representative patients.
To evaluate whether CD166 can enrich human prostate tissue regeneration capacity in vivo, benign human prostate cells were dissociated and sorted according to cell surface CD166 expression levels. Equal number of viable CD166 hi and CD166 lo cells (2610 5 ) was implanted subcutaneously into NOD-SCID/IL2rc null mice, in combination with 2610 5 rUGSM inductive mesenchymal cells. After 8-16 weeks, grafts were harvested, fixed and embedded in paraffin for quantification and analyses. CD166 hi cells have more tissue regeneration capacity as evidenced by increased number of tubule-like epithelial structures found in the grafts, which is rarely seen in the CD166 lo grafts ( Figure 2C). Further analyses showed that the tubule-like structures initiated by CD166 hi cells contain CK5 and p63 expressing basal cells, CK8 luminal cells and AR positive cells ( Figure S2).
Combination of markers TROP2 and CD49f can separate lineage-negative human prostate epithelial cells into various subpopulations, with TROP2 hi ;CD49f hi (Lin -T hi C hi or LTC) cells possessing the highest sphere forming capability in vitro [29]. Additionally, LTC cells can develop cancer-like phenotype in vivo following oncogenic transformation [13]. We tested whether CD166 can further segregate this LTC population. FACS analysis of benign human prostate cells indicated that more than 50% of LTC stem/progenitor cells also express the CD166 surface marker ( Figure 2D, left and middle panel). Furthermore, we examined if differences in regeneration potential exist between these two subpopulations. Sorted LTC;CD166 hi and LTC;CD166 lo cells were injected subcutaneously into NOD-SCID/IL2rcnull mice with 2610 5 rUGSM cells and analyzed 8-16 weeks later. Our in vivo data suggest that LTC;CD166 hi cells can induce more tubule-like structures, whereas LTC;CD166 lo cells have less regeneration capacity ( Figure 2D, right panel).  To examine whether CD166 can enrich tumor initiating cells after castration, we compared the percentage of CD166 hi subpopulation between intact and castrated Pten mutant mice and observed the expansion of CD166 hi subpopulation after castration ( Figure 3A). Next, we compared the sphere formation capabilities of LSC hi ;CD166 hi , LSC hi ;CD166 lo , LSC lo ;CD166 hi , and LSC lo ;CD166 lo subpopulations at the pre-cancer PIN (6 weeks) and cancer stages (11 weeks). We found that the LSC hi ;CD166 hi subpopulation has much higher sphere-forming ability, and nearly all sphere-forming activity in the cancer stage resides in the LSC hi ;CD166 hi subpopulation ( Figure 3B). Consistent with our previous observation that Pten mutant spheres are larger than WT control spheres [19], both LSC hi ;CD166 hi and LSC hi ;CD166 lo subpopulations form large prostate spheres ( Figure  S3). Our previous study suggested that Pten deletion promotes the expansion of LSC hi prostate stem/progenitor cells [18,19]. Within the LSC hi population, we observed selective expansion of LSC hi ;CD166 hi cells. Pten mutant mice have more than a 3-fold increase in the percentage of LSC hi ;CD166 hi subpopulation, compared to WT littermates ( Figure 3C).
To further study the LSC hi ;CD166 hi subpopulation, we isolated RNA from LSC hi ;CD166 hi , LSC hi ;CD166 lo subpopulations and the cell fraction depleted of LSC cells (non-LSC hi ) and compared their gene expressions by RT-PCR analysis. LSC hi ;CD166 hi subpopulation expresses similar levels of basal cell markers Ck5 and p63 as the LSC hi ;CD166 lo subpopulation ( Figure 3D, left panel). However, LSC hi ;CD166 hi subpopulation expresses much higher level of luminal marker Ck8 and Trop2, a new epithelial surface marker we recently identified for enriching stem cell activities in both murine and human prostates [13,29] ( Figure 3D, right panel). Further examination of several other epithelial cell stem cells markers [10,30,31,32,33] showed that LSC hi ;CD166 hi cells have significantly higher CD44 and Nkx3.1 expression compared to LSC hi ;CD166 lo cells. Although compared to non-LSC population, LSC hi ;CD166 hi cells express less Nkx3.1. No significant differences were found in CD117, and CD133 expressions between these two populations ( Figure 3D, right panel).

CD166 Expression is Upregulated in Human Castration Resistant Prostate Cancer
Having found that CD166 can be used to enrich for human LTC cells and mouse tumor in itiating cells, we then examined the relationship between CD166 expression and human prostate cancer progression. In clinically annotated data of 218 prostate tumors [34], CD166 gene expression significantly correlates with increased prostate cancer aggressiveness, as indicated by Gleason score, with highest expression in metastasis samples ( Figure 4A). We further surveyed CD166 expression on human prostate cancer tissue microarrays, which consist of 14 castration resistant (CRPC) metastasis samples and 98 hormone naïve primary cancer samples from patients receiving either neoadjuvant hormone treatment (NHT) for various periods or receiving no treatment. CD166 is significantly enhanced in CRPC samples ( Figure 4B for representative images). Compared to the predominant membrane localization of CD166 in hormone naïve primary cancer samples, we observed intense cytoplasmic localization of CD166 in CRPC bone metastasis samples ( Figure 4B, high magnification). CD166 expression levels were scored and p values are computed by Mann-Whitney test. CD166 protein expression level is significantly higher in CRPC samples as compared with primary cancers with (p,0.0001) or without (p,0.02) NHT ( Figure 4C). These data suggest that CD166 is a castration-enriched marker for both murine and human prostate cancer.

Loss of CD166 does not Interfere with WT Prostate Development and Prostate Sphere Formation
While expressed in a wide variety of tissues, CD166 is usually restricted to subsets of cells involved in dynamic growth and/or migration, including neural development, branching organ development, hematopoiesis and immune response [27]. To test whether CD166 plays an intrinsic role in regulating prostate stem/progenitor cells, we analyzed CD166 knockout mice (CD166 2/2 ). Genetic deletion of CD166 gene was achieved by replacing its first exon with a cDNA encoding EGFP [35]. CD166 null mice are phenotypically normal and fertile [35]. We examined the prostate at 8 and 20 weeks of age and found no difference in gross anatomy and histology among WT (data not shown), CD166 +/2 and CD166 2/2 mouse prostates ( Figure 5A).
To further examine whether loss of CD166 has any effect on prostate stem/progenitor cells, we compared sphere formation activities of CD166 +/2 and CD166 2/2 prostate epithelium and found there is no significant difference ( Figure 5B). In addition, spheres generated from CD166 2/2 prostate have similar size distribution compare to those from CD166 +/2 prostate epithelium (data not shown). Similarly, FACS analysis demonstrated that loss of CD166 does not affect LSC hi content of prostates isolated from the CD166 2/2 mice ( Figure 5C), suggesting that CD166 does not play an essential role in normal prostate gland development or prostate stem/progenitor number and function.

Genetic Deletion of CD166 does not Block Prostate Cancer Progression
It has been postulated that CD166 functions as a cell surface sensor for cell density and controls the transition between local cell proliferation and tissue invasion during melanoma progression [36]. To examine whether CD166 plays an essential role in prostate cancer development, especially in the tumor initiating cells, we crossed CD166 2/2 mice with the Pten conditional knockout mice [17]. Histopathologic analysis indicated that loss of CD166 did not significantly change the kinetics of prostate cancer development in Pten null model and all Pb-Cre + ;Pten L/ L ;CD166 2/2 mice developed adenocarcinoma around 9 weeks of age ( Figure 6A and data not shown). We observed similar levels of Ki67 + cells between Pb-Cre + ,Pten L/L ,CD166 +/+ and Pb-Cre + ;Pten L/L ; CD166 2/2 prostates ( Figure 6A). SMA staining also demonstrated that loss of CD166 does not block prostate cancer cells from local invasion ( Figure 6A, right panels).
We then compared the sphere formation between Pb-Cre + ;Pten L/ L ; CD166 +/2 and Pb-Cre + ;Pten L/L ;CD166 2/2 prostates and found that loss of CD166 does not interfere with sphere-forming activity of Pten null epithelium ( Figure 6B). Moreover, CD166 2/2 prostates have similar LSC hi content as compared to CD166 +/2 Pten null prostates ( Figure 6C). Since PI3K/AKT pathway activation is a driving force for cell proliferation and prostate cancer progression in Pb-Cre + ;Pten L/L prostate cancer [17,20], we then examined whether there is any alteration of AKT activation after genetic deletion of CD166. Western blot analysis demonstrated that Pb-Cre + ;Pten L/L ;CD166 2/2 prostate has no CD166 expression, but has similar P-AKT levels compared to Pb-Cre + ;Pten L/L ;CD166 +/+ and Pb-Cre + ;Pten L/L ;CD166 +/2 prostate ( Figure 6D). We further confirmed that there is no negative selection against Pten 2/2 ;CD166 2/ 2 cells since equal intensity of knockin-GFP protein can be detected in all cohorts except CD166 +/+ mice. Since we see significant overexpression of CD166 in human CRPC samples, we next investigated whether CD166 would influence the development of CRPC in the Pten null prostate cancer model. Pb-Cre + ;Pten L/L ;CD166 +/2 and Pb-Cre + ;Pten L/ L ;CD166 2/2 males were castrated at 12 weeks and prostates were isolated 8 weeks later. As shown in Figure 7, deletion of CD166 does not significantly influence the formation of CRPC, as evidenced by similar pathohistology (Figure 7A), CK5/CK8 marker distribution, BrdU pulse labeling and SMA staining in both cohorts ( Figure 7B). Taken together, our genetic studies indicate that CD166 has limited intrinsic function in the prostate, even in the tumor initiating cells.

Discussion
Few surface markers are currently available for enriching both murine and human prostate tissue stem/progenitor cells and for identifying prostate cancer initiating cells. By searching for those cell surface molecules that are upregulated in castrated murine prostate and castration resistant prostate cancers (CRPC) of murine and human origins, we identified CD166 as a surface marker for enriching both murine and human prostate tissue stem/progenitor cells based on in vitro sphere forming and in vivo tissue regeneration analyses. Importantly, upregulated CD166 expression and expansion of CD166 hi cells correlate with Pten null CRPC progression as well as human CRPC development, although genetic deletion of CD166 does not interfere with normal murine prostate development or Pten null prostate cancer progression. Together, our study suggests CD166 can be used as a potential surface marker for identifying castration resistant tumor cells for targeted drug delivery.
CD166 expression has been proposed as a prognostic marker for several cancers, including breast [37], prostate [38], ovarian [39], pancreatic [40], colon [41], oral cancers [42], melanoma [36] and gastric cancers [43]. Importantly, our microarray and TMA studies demonstrate the association of increased CD166 expression with human prostate cancer metastasis and CRPC development. Moreover, within both murine and human prostates, we show that the CD166-high expressing subpopulation encompasses prostate stem/progenitor and cancer initiating cells.
To investigate human prostate tissue stem/progenitor cell properties, we evaluated adult human prostate epithelium dissociated from benign prostate, rather than cell lines and xenografts. The advantage of this approach is to maintain the original heterogeneity in human prostate samples by avoiding the effect of long-term in vitro selection. However, there appears to be greater variability among patient samples in the tissue regeneration assays. This may be due to the difference in sample variability (i.e., ischemia time prior to tissue processing and cell retrieval), individual variability in CD166 expression, and technical challenges related to the tissue regeneration assays using human prostate cells. Therefore, analysis of sufficient patient samples is essential in order to draw a valid conclusion. In the current study, 6 human samples were utilized for the in vitro sphere forming and another 6 samples were used for in vivo regeneration assays. Using this system, we have previously defined TROP2 hi ;CD49f hi as a cancer initiating cell (cell of origin) for human prostate cancer [13]. In the current study, CD166 hi population demonstrated signifi-cantly increased sphere-forming capacity compared to the patientmatched CD166 lo population. In addition, our study demonstrates that CD166 can not only enrich human sphere-forming cells, but also segregate TROP2 hi ;CD49f hi into two functionally different populations, with TROP2 hi ;CD49f hi; CD166 hi having higher regeneration capacity in vivo, compared to TROP2 hi CD49f hi C-D166 lo . CD166 is also highly upregulated in CRPC based on our gene expression analysis and tissue microarray study. Therefore, CD166 may enrich both human prostate tissue stem/progenitor cells and castration resistant prostate cancer cells.
LSC hi subpopulation has been defined as the murine prostate tissue stem/progenitor cells and expands significantly following castration [12,19,26]. LSC hi cells express basal markers and demonstrated robust sphere-forming activity in vitro and prostate regeneration capability in vivo [26]. In contrast to luminal cells, LSC hi cells respond efficiently to multiple oncogenic insults for prostate cancer initiation using a transplantation-based prostate regeneration assay [12]. We and others have demonstrated that the LSC hi population, isolated from Pten null prostate tissue, is sufficient to regenerate cancerous morphology upon transplantation that closely mimics that of primary cancers [19,44]. In this study, we further separated LSC hi subpopulation into CD166 hi and CD166 lo subsets and found that most of sphere-forming activities are associated with the LSC hi ;CD166 hi cells. Importantly, this LSC hi ;CD166 hi population was demonstrated to have selfrenewal activity as spheres from this population could be passaged at least 3 generations with a high rate of sphere formation. Moreover, LSC hi ;CD166 hi cells are expanded upon castration as well as Pten deletion in comparison to LSC hi ;CD166 lo cells. Therefore, CD166 can further enrich murine prostate cancer initiating cells and castration resistant cells.
The relationship of LSC hi ;CD166 hi cancer initiating cells described here to other cell populations is of obvious interest [45]. Using lineage tracing and cell type-specific Cre lines, a recent report demonstrates that both luminal cells and basal cells can initiate prostate cancer upon Pten deletion [14]. This new observation is not in conflict with our previous studies: we showed that Pten deletion mediated by Pb-Cre happens in both basal and luminal cells [18]. In addition, we observed significant expansion of a subset of prostate cancer cells positive for basal cell markers CK5 and p63 and luminal cell marker CK8, suggestive of transient amplifying/intermediate cells [18,46]. Compared to LSC hi ;CD166 lo cells, one of the distinguishing features of LSC hi ;CD166 hi cells is the higher Trop2 expression, a cell surface marker we have used for enriching both murine and human tissue stem cells [13,29]. TROP2 can functionally segregate mouse LSC population but there is no cytokeratin phenotypic difference between LSC hi ;Trop2 hi and LSC hi ;Trop2 lo population [29]. CD166, on the other hand, can enrich Pten null LSC hi population with CK5 + /p63 + /CK8 + /AR 2/ TROP2 hi characteristics, suggesting that CD166 may preferentially enrich for CK5 + /CK8 + transient amplifying/intermediate cells, which currently cannot be prospectively purified. Increased CK5 + ;CK8 + cells have been observed in the Pten conditional knockout model [18,47] as well as Pten 2/2 ;TP53 2/2 prostates cancer model [48]. A recent study also identified a subset of tumor-initiating stem-like cells in human prostate cancer cell lines and xenografts based on co-expression of the human pluripotent stem cell marker TRA-1-60, CD151 and change of Pten mutant LSC hi CD166 hi content compared to WT; right, FACS blots show the expansion of LSC hi CD166 hi cells within LSC population on Pten mutant compared to WT. (D) RNA was isolated from non-LSC, LSC hi CD166 hi , and LSC hi CD166 lo fractions in duplicate experiments. RNA was synthesized into cDNA and subjected to qRT-PCR. Graph shows fold-enrichment over the non-LSC cells for each gene. Gadph was used as the reference gene (*, p,0.05; **, p,0.01; n.s., not significant). doi:10.1371/journal.pone.0042564.g003 CD166 [49]. Interestingly, this subtype of human prostate tumor initiating cells also have the AR 2 ;CK5 + ;CK8 + phenotype [49]. Another characteristic of LSC hi ;CD166 hi cells is relatively higher CD44 expression. Since knockdown of CD44 was very effective to suppress cancer stem cell regeneration and metastasis [30], it will be interesting to examine whether there is any functional role for CD44 in LSC hi ;CD166 hi tumor initiating cells.
As an adhesion molecule, CD166 can initiate homophilic (CD166-CD166) or heterophilic interaction (CD166-CD6), and play important roles in neural guidance and the immune system [27]. CD166 has also been suggested to play a critical role in various human cancers and as a potential therapeutic target for cancer initiating cells, similar to CD44 [30] and CD47 [50]. A truncated CD166 variant has been shown to block melanoma metastasis by interfering with the CD166-CD166 homophilic interaction [51]. Similarly, novel human recombinant single-chain anti-CD166 antibodies have been shown to inhibit colorectal carcinoma growth as well as breast cancer cell invasion [52]. Unlike subcutaneous allograft or xenograft models used in above studies, we defined the functions of CD166 in prostate cancer initiating cells and prostate cancer development in immune competent mice within the natural prostate environment. By generating the Pb-Cre + ;Pten L/L ;CD166 2/2 line, our study demonstrates that loss of CD166 within LSC hi population does not change their ability to form spheres in vitro and block prostate cancer initiation and progression in vivo. As it is possible that other members of the Cell Adhesion Molecule (CAM) family can compensate for the role of CD166 in murine prostate cancer development, we cannot conclude that CD166 has no in vivo function on prostate cancer initiation. Nevertheless, since cancer initiating cell surface markers can be used for molecular imaging [53] and/or for internalizing a death-inducing compound for targeted therapies [54], our work suggests that CD166 may be for a suitable surface marker for future targeted drug delivery [55]. Recently, a promising study showed substantial cytotoxic effects of the CD166 scFv-condugated drugs on three human prostate cancer cell lines (Du-145, PC3, and LNCaP) [55]. Since CD166 is highly expressed on both human and mouse tissue stem/ progenitor cells, it will be interesting to examine the effect of this targeted drug delivery on their prostate sphere forming activity and prostate regeneration potential. The Pb-Cre + ;Pten L/L ;CD166 +/ 2 and Pb-Cre + ;Pten L/L ;CD166 2/2 mouse models generated in this study, therefore, can be used to investigate the efficiency of CD166 -mediated drug delivery to prostate cancer initiating cells in vivo, especially during CRPC development.

Mouse Strains
Mutant mice with prostate-specific deletion of Pten were generated as described previously under a mixed background [17]. The 129/C57 background CD166 knockout (CD166 2/2 ) was generously provided by the laboratory of Dr. Weiner of University of Iowa [35]. Pten L/L mice on a 129/Balb/c background were first crossed to the CD166 2/2 mice [35] to get F2 female Pten L/L ;CD166 2/2 . Pb-Cre + ; Pten L/L ;CD166 2/2 mice were then generated by crossing female Cre -;Pten L/L ;CD166 2/2 with male Pb-Cre + ;Pten L/L ;CD166 +/2 . All animal experiments were performed following Institutional Approval for Appropriate Care and use of Laboratory animals by the UCLA Institutional Animal Care and Use Committee (Chancellor's Animal Research Committee (ARC)), Animal Welfare assurance number A3196-01.

Tissue Collection and FACS
The preparation of prostate epithelial cell suspensions from male mice were described previously [20]. Dissociated prostate cells were suspended in DMEM/10% FBS and stained with antibody for 15 min at 4uC. Antibodies are listed in Table S2. FACS analysis was performed by using BD FACS Canto (BD Biosciences, San Jose, CA). Cell sorting was done by using BD FACS Vantage and the BD FACS Aria II.

In vitro Prostate Sphere-forming Assays
Prostate spheres were cultured and passaged as described previously [56,57]. FACS-isolated prostate cells or unsorted prostate cells were counted and suspended into a 100 mL mixture of 1:1 Matrigel (BD Biosciences, San Jose, CA) and PrEGM (Lonza, Walkersville, MD). Samples were plated around the rims of wells in a 12-well plate and allowed to solidify at 37uC for 45 minutes, before 1 ml of PrEGM was added. Sphere media was changed every three days. Spheres were counted after 8 days. To passage spheres, medium was aspirated off and matrigel was digested with 1 mL Dispase solution (Invitrogen, Carlsbad, CA, 1 mg/ml, dissolved in PrEGM medium) for 30 minutes at 37uC. Spheres were collected, incubated in 1 ml warm Trypsin/0.05% EDTA at 37uC for 5 minutes, passed through a 27-gauge syringe 5 times, and filtered through a 40 mm filter. Cells were counted by hemocytometer and replated.

Immunofluorescence and Immunohistochemistry Analyses
Tissue analysis was carried out using standard techniques as described previously [17]. Sections (4 mm) were stained with hematoxylin and eosin (H&E) or with specified antibodies (Table  S2).

Human Prostate Cancer Tissue Microarray (TMA)
TMA used to survey CD166 expression is composed of 112 patient samples. Written consent was obtained from the patient as well as ethics approval from University of British Columbia-British Columbia Cancer Agency Research Ethics Board (UBC BCCA REB), Vancouver, Canada. The 112 patient specimens were spotted in triplicate to create a tissue microarray with 336 cores as described previously in [58]. Scoring method was based on the intensity of the staining in each core on a 4-point scale from none (0) to high (3). Mann-Whitney test was used to compare CD166 protein expression difference between different groups. p values ,0.05 were considered significant.

Human Prostate Tissue Acquisition and Dissociation
Human prostate tissue was obtained via a research protocol that was approved by the Office for the Protection of Research Subjects at UCLA and the Greater Los Angeles VA Medical Center. Informed written consent was obtained on all participants where identifying information was included. A frozen section was prepared from an adjacent slice of prostate tissue in order to determine the location of tumor nodules. Tumor areas were encircled and dissected away from benign regions within the fresh tissue slice. Benign tissue specimens were placed on ice and brought immediately to the laboratory for mechanical and enzymatic digestion [28]. Prostate tissue was minced into small fragments (1 mm 3 ) in RPMI-1640 medium supplemented with 10% FBS and went through through enzymatic digestion (12 h in 0.25% type I collagenase followed by TripLE (Invitrogen) for 5 min at 37uC). Cell suspensions were passed through a 23-gauge needle and were filtered through 40mm filters. Cells were plated overnight in PrEGM as described above for sphere formation assay or tissue regeneration assay.

Tissue Regeneration Assay
In vivo tissue experiments were performed in male NOD-SCID/ IL2rc null mice in accordance with protocol number 2007-189-11A, approved by the Animal Research Committee within the Office for the Protection of Research Subjects at UCLA. Cells of interest were collected from FACS sorting. 2610 5 viable cells were then mixed with 2610 5 rat urogenital sinus mesenchyme (rUGSM) and suspended in 100 mL with 50:50 matrigel:PREGM [4,28,59]. Cell/Matrigel mixtures were then injected subcutane-ously into NOD-SCID/IL2rc null mice. Animals were supplemented with a 12.5 mg 90-day release testosterone pellet under the skin (Innovative Research of America, Sarasota, FL). Grafts were harvested 8-16 weeks later and subjected to further analysis. Figure S1 WT LSC hi ; CD166 hi prostate cells demonstrate higher self-renewal activity. (A) LSC hi ;CD166 hi and LSC hi ; CD166 lo cells were isolated by FACS from 8-to 10-weekold mice and plated for sphere formation assay. Spheres from the each subpopulation (P0) were dissociated and replated for 3 successive generations (P1-P3). Graph shows the percentage of sphere-forming cells, based on the spheres from each population per 5000 cells plated after 8 days of growth. Error bars represent means and STD from triplicates of one of the two independent experiments (**, P,0.001). (B) Comparison of sphere size distribution between LSC hi ; CD166 hi and LSC hi ; CD166 lo formed spheres. n.s., not significant. (C) Representative sphere images of LSC hi ;CD166 hi and LSC hi ; CD166 lo cells generated spheres. Scale bar: 100 mm.