**3. Discussion**

Our findings in this study indicate that KLF5 is crucial for androgen/AR signaling to function in PCa cells. The first line of evidence is that the transcriptional activity of AR depended on the expression of KLF5. For example, *KLF5* silencing decreased the expression of *PSA*, a classic transcriptional target gene of AR in the prostate [41], and *TMPRSS2* and *FKBP5*, two other AR target genes [42,43], in LNCaP and C4-2B cells (Figure 2). The necessity of KLF5 for AR's transcriptional activity was also

demonstrated by promoter luciferase reporter assays using two androgen responsive promoters (i.e., *PSA* and *MMTV* [48]) (Figure 2) and by the expression of two genes that mediate AR's pro-proliferative function, i.e., *MYC* and *CCND1* [26–28,44–47], in the same cells with *KLF5* silencing (Figure 5). We have also presented evidence from cellular analyses, in which androgen/AR signaling also required KLF5 to maintain a steady proliferation of PCa cells. Specifically, *KLF5* silencing significantly attenuated the functions of AR in the maintenance of colony and sphere formation in vitro and xenograft tumor growth in nude mice (Figure 6a). Consistently, *KLF5* silencing also reduced the number of Ki67-positive cells and the expression of cyclin D1 and MYC (Figure 6). We noticed that, after blocking AR activity with enzalutamide, which had a profound effect (Figures 5 and 6), *KLF5* silencing still had a detectable effect in both the expression of MYC and cyclin D1 and cell proliferation (Figures 5 and 6), which suggests that, while crucial for AR to function, KLF5 can still function when AR is inhibited. Indeed, in androgen-independent PCa cell lines including PC-3 and DU 145, KLF5 is clearly pro-proliferative, even though when TGF-β is activated, TGF-β and KLF5 slow but do not stop cell proliferation [49,50].

Molecularly, the enhancing effect of KLF5 on AR function in cell proliferation is mediated by at least three distinct mechanisms. For example, via direct promoter binding, AR promotes the transcription of *KLF5* to increase its expression [32]. As expected, the upregulation of *KLF5* by androgen was mediated by AR (Figure 1), since inhibition of AR by RNAi-mediated *AR* silencing or enzalutamide treatment eliminated the induction of *KLF5* transcription (Figure 1).

The second molecular mechanism by which KLF5 facilitates AR function is that KLF5 also activates *AR* transcription in PCa cells. For example, *KLF5* silencing by RNAi reduced AR expression in both LNCaP and C4-2B cells (Figure 3). In addition, the *AR* promoter indeed contained multiple consensus KLF5 binding elements that were necessary not only for the *AR* promoter's activities in the promoter-reporter assay and but also for the binding of KLF5 to the *AR* promoter in the ChIP-PCR analysis (Figure 3), and two adjacent KLF5 binding elements in the *AR* promoter have been confirmed to be essential for KLF5 binding and promoter activity (Figure 3).

The third mechanism is that KLF5 and AR coordinate to regulate gene transcription, which is supported by multiple lines of evidence. Firstly, KLF5 and AR depend on each other in their binding to the promoters of *PSA*, *MYC* and *CCND1*, as *KLF5* silencing reduced the amount of promoter/enhancer DNA of these genes in AR-precipitated DNA (Figure 2) while inhibition of AR reduced this DNA in KLF5-precipitated DNA (Figure 5). Nevertheless, the details of KLF5 and AR binding to gene promoters are unclear (e.g., the chromatin landscape for the binding). Secondly, KLF5 and AR, both of which are transcription factors, physically associate with each other to regulate gene transcription. AR was detected in the KLF5 protein complex and KLF5 in the AR complex (Figure 4). In addition, the KLF5-AR interaction occurred not only for ectopically expressed KLF5 and AR in HEK293T cells but also for endogenous KLF5 and AR in both human cells and mouse prostates (Figure 4). Furthermore, the KLF5-AR interaction was mediated by a sequence within residues 1-200 of KLF5 (Figure 4) and was attenuated by enzalutamide treatment (Figure 4).

Therefore, androgen/AR signaling activates *KLF5* expression via the binding of AR to *KLF5* promoter, KLF5 in turn enhances the transcription of *AR* by promoter binding, and AR and KLF5 then coordinate to transactivate a subset of genes to promote the proliferation of PCa cells.

We noticed that, for AR target genes *PSA*, *TMPRSS2,* and *FKBP5*, while *KLF5* silencing reduced their induction by AR (Figure 2a–d), which supports the necessity of KLF5 for AR function in their transcription, AR-bound promoter DNA was detected in KLF5-precipitated promoter DNA only for *PSA* but not for *TMPRSS2* and *FKBP5* (Figure 2). The reason for this discrepancy is unknown. Neither is it known whether the KLF5-AR association depends on promoter DNA or cofactors of AR.

In mouse prostates, castration-mediated androgen depletion increased Klf5-positive cells [29], which is seemingly inconsistent with the induction of KLF5 by androgen in PCa cells (Figure 1). Compared to luminal cells in the prostate, basal cells preferentially express Klf5, particularly acetylated Klf5 [30], castration causes massive death in luminal cells but much less so in basal cells, and basal

cells express much less AR and are androgen insensitive. LNCaP and C4-2B cells are AR-positive and androgen-dependent/sensitive, and thus have a different lineage from basal cells.

The role of KLF5 in androgen-induced cell proliferation and tumor growth could also involve KLF5- s function in tumor microenvironment (TME) and immune responses. For example, pro-inflammatory TNFα and lipopolysaccharide (LPS) induce KLF5 expression, and TNFα depends on KLF5 to induce MCP-1 [51]. In addition, KLF5 directly interacts with NF-κB [52], a potent inflammatory factor, and interruption of this interaction inhibits LPS-induced macrophage proliferation [53]. Knockdown of KLF5 also reduces the expression of p50 and p65 subunits of NF-κB and its downstream target genes TNFα and IL-6 in response to LPS [54]. This and other potential mechanisms for KLF5 function remain to be examined.

During late stages of tumor progression, AR becomes activated even when androgen levels are low, causing CRPC [9,10]. Further studying how KLF5 and AR coordinate to regulate the expression of genes, particularly those mediating cell proliferation/survival and thus likely affecting PCa progression, will facilitate our understanding of AR activation in CRPC and the development of therapeutic approaches for the treatment of CRPCs. For example, in advanced PCa, the *KLF5* locus often undergoes hemizygous deletion [21,55], which downregulates *KLF5* expression because the gene is haploinsufficient [31]. As KLF5 is necessary for AR function in PCa cells, as discussed above, downregulation of KLF5 could generate a feedback signal that leads to the upregulation of AR or functional compensation of AR activity. This hypothesis remains to be tested.

#### **4. Materials and Methods**

#### *4.1. Cell Lines, Cell Culture, and RNA Interference*

Human PCa cell line LNCaP was purchased from American Type Cell Culture (Manassas, VA) and cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS, Gibco, Waltham, MA). The C4-2B cell line, originally derived from a bone metastasis of a LNCaP clone in mice [56], was kindly provided by Dr. Leland W. K. Chung of Cedars-Sinai Medical Center and cultured in the same medium as LNCaP. Cells were maintained at 37 ◦C with 5% CO2. During experiments, cells recovered from a liquid nitrogen freezer were used within two months (<20 passages) with no noticeable morphological changes. All cell lines were authenticated by STR profiling before experiments were started. For all experiments involving R1881 (Melonepharma, Dalian, China, catalog number: MB5484) treatments, the medium was replaced with phenol red-free medium containing 10% charcoal-stripped FBS.

For RNA interference (RNAi) with shRNA, C4-2B cells were infected with lentiviruses expressing an shRNA specifically targeting human *KLF5*, which was developed and validated in a previous study with various PCa cell lines [57], to establish the cell population in which *KLF5* is stably knocked down. For RNAi with siRNAs, siRNA (si*AR*: 5- -CAAGGGAGGUUACACCAAA-3- ; si*KLF5*: 5- -AAGCUCACCUGAGGACUCA-3- ) oligos against human *KLF5* and *AR* were synthesized by Sagon Biotech (Guangzhou, China) and transfected into cells using the Lipofectamine RNAiMAX reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol.
