Kruppel-like factor 4 signals through microRNA-206 to promote tumor initiation and cell survival

Tumor cell heterogeneity poses a major hurdle in the treatment of cancer. Mammary cancer stem-like cells (MaCSCs), or tumor-initiating cells, are highly tumorigenic sub-populations that have the potential to self-renew and to differentiate. These cells are clinically important, as they display therapeutic resistance and may contribute to treatment failure and recurrence, but the signaling axes relevant to the tumorigenic phenotype are poorly defined. The zinc-finger transcription factor Kruppel-like factor 4 (KLF4) is a pluripotency mediator that is enriched in MaCSCs. KLF4 promotes RAS-extracellular signal-regulated kinase pathway activity and tumor cell survival in triple-negative breast cancer (TNBC) cells. In this study, we found that both KLF4 and a downstream effector, microRNA-206 (miR-206), are selectively enriched in the MaCSC fractions of cultured human TNBC cell lines, as well as in the aldehyde dehydrogenase-high MaCSC sub-population of cells derived from xenografted human mammary carcinomas. The suppression of endogenous KLF4 or miR-206 activities abrogated cell survival and in vivo tumor initiation, despite having only subtle effects on MaCSC abundance. Using a combinatorial approach that included in silico as well as loss- and gain-of-function in vitro assays, we identified miR-206-mediated repression of the pro-apoptotic molecules programmed cell death 4 (PDCD4) and connexin 43 (CX43/GJA1). Depletion of either of these two miR-206-regulated transcripts promoted resistance to anoikis, a prominent feature of CSCs, but did not consistently alter MaCSC abundance. Consistent with increased levels of miR-206 in MaCSCs, the expression of both PDCD4 and CX43 was suppressed in these cells relative to control cells. These results identify miR-206 as an effector of KLF4-mediated prosurvival signaling in MaCSCs through repression of PDCD4 and CX43. Consequently, our study suggests that a pluripotency factor exerts prosurvival signaling in MaCSCs, and that antagonism of KLF4-miR-206 signaling may selectively target the MaCSC niche in TNBC.


INTRODUCTION
Compelling experimental evidence supports the hierarchical organization of certain human tumor types, including breast cancer. [1][2][3][4][5][6] These tumors comprise heterogenous mixtures of tumor cell populations that include cancer stem-like cells (CSCs), typically defined by their ability to initiate tumors in limiting dilution assays (that is, tumor-initiating cells, TICs). Furthermore, CSCs can survive and form tumorspheres in suspension culture, self-renew and differentiate. 7,8 These cells display resistance to chemotherapy, radiation therapy and other triggers of cell death, and are thought to contribute to cancer recurrence. Therefore, CSCs represent an important sub-population for therapeutic targeting. 3,5 In mammary carcinoma, functionally validated CSCs (termed MaCSCs (mammary CSCs)) have been identified by profiling the expression of cell surface markers such as PROCR (P) and ESA (E) and/or by assaying aldehyde dehydrogenase (ALDH) activity. [9][10][11][12] Despite this insight, the underlying mechanisms that mediate the MaCSC phenotype are unclear. For regulation of their abundance and/or their intrinsic properties such as resistance to cell death, several cellular signaling axes have been implicated including the WNT, NOTCH, TGFβ and SHH pathways. [13][14][15] A potential mediator of the MaCSC phenotype is the pluripotency factor Kruppel-like factor 4 (KLF4). This zinc-finger transcription factor promotes the formation of induced pluripotent stem cells from adult somatic cells and can have both antitumorigenic and protumorigenic roles in a context-dependent manner. [16][17][18][19][20] The capability of KLF4 to exert protumorigenic influences may reflect its role as a prosurvival stress response factor. [21][22][23][24][25][26][27][28] In support of a protumorigenic role, KLF4 promotes epithelial transformation in vitro, escape from RAS-induced senescence and skin tumor initiation in transgenic mice. 16,29,30 Furthermore, loss-of-function studies reveal that KLF4 promotes cell survival following radiation-induced DNA damage, and promotes the tumorigenicity of colorectal CSCs-enriched spheroid cells. 26,31 In human breast cancer, KLF4 promoter demethylation and KLF4 protein expression indicate an unfavorable prognosis. [32][33][34] KLF4 expression is positively correlated with tumor size, advanced grade and stage. 35 We previously identified microRNAs, including microRNA-206 (miR-206) and miR-21, as direct transcriptional targets of KLF4 that promote RAS-extracellular signal-regulated kinase (ERK) signaling in triple-negative breast cancer (TNBC) cells. 36,37 Although on its own each miR exerts only subtle influences on RAS-ERK pathway activity, the coexpression of miR-206 and miR-21 potently represses the expression of pathway inhibitors including RASA1 and SPRED1. Furthermore, miR-206 directly represses KLF4 translation, constituting a feedback loop. 36 In this study, we observed elevation of KLF4 and miR-206 in the P + /E + and ALDH High MaCSC fractions. In TNBC cells, both KLF4 and miR-206 were critical for cell survival and in vivo tumor initiation. We identified the tumor-suppressor programmed cell death 4 (PDCD4) as a potential mediator of cell survival by miR-206. Furthermore, in TNBC cells we demonstrated the miR-206 regulation of a previously validated transcript, the gap junction protein connexin 43 (CX43/GJA1). 38 Consistent with the elevated levels of miR-206 in MaCSCs, PDCD4 and CX43 levels were decreased. Supporting functional roles downstream of KLF4 and miR-206, suppression of either PDCD4 or CX43 led to anoikis resistance, an intrinsic property of CSCs. 7,[39][40][41][42][43] Finally, further documenting a prosurvival role, miR-206 promoted chemoresistance of TNBC cells against paclitaxel or doxorubicin. Our studies identify KLF4 and miR-206 as functional MaCSC markers that mediate cell survival. Consequently, KLF4 and/or miR-206 may be therapeutically targeted to selectively cripple MaCSCs in TNBCs.

RESULTS
miR-206 is highly expressed in basal-like breast cancers and MaCSCs KLF4 protein levels correlate with an aggressive phenotype in breast tumors. 32,33,35 Similar to KLF4, miR-206 was increased in human tumors of advanced histological grade (Figure 1a, left panel). Consistent with studies that identified upregulation of miR-206 in ERbreast tumors, miR-206 levels were elevated in TNBCs compared with both ER + and HER2 + human subgroups (Figure 1a, right panel). 44,45 Enrichment of miR-206 was similarly observed in murine basal-like mammary tumors (Figure 1b). Compared with normal mammary tissues or tumors arising in the luminal MMTV_Neu model, 46 we observed upregulation of both KLF4 and miR-206 in basal-like tumors derived from the C3(1)/SV40 large T antigen (C3(1)/TAg) genetically engineered mouse model (GEMM) (Figure 1c). These results are consistent with the direct regulation of miR-206 by KLF4 as previously reported. 37 MaCSCs are enriched in the triple-negative subgroup of breast cancer and are thought to contribute to the aggressive behavior of these cancers. [47][48][49] Similar to human and murine mammary carcinoma cells displaying high ALDH activity, 11,50-52 MDA-MB-231 TNBC cells displaying the P + /E + surface marker profile represent TICs. 12 For SUM159PT cells, CD44 + /CD24 -/ESA + subset was previously identified as TICs. 53 As the P + phenotype is a surrogate for the CD44 + /CD24profile, the P + /E + SUM159PT cells are likely to represent MaCSCs. 10 We analyzed KLF4 and miR-206 levels in flow-sorted subpopulations of MDA-MB-231 cells (Figure 1d, left panel). Compared with non-MaCSCs (that is, P -/E -), miR-206 and KLF4 were increased in the P + /E + sub-population (Figure 1d, middle panels). Using P + /E + cells, we profiled the expression of other genes associated with stem-like cell phenotypes. 9,14,18,19 Compared with P -/Ecells, the expression of CD44, MYC, SOX2, NANOG, ZEB1 and SNAI2 was upregulated in P + /E + cells, whereas CD24 and POU5F1 (OCT3/4) expression were decreased (Figure 1d, right panel). Similarly, the P + /E + fraction of SUM159PT cells displayed elevated levels of KLF4 and miR-206, and showed a similar stem cell marker profile as the MDA-MB-231 cells (Figure 1e). These results associate KLF4 and miR-206 with the MaCSC phenotype in human breast cancer models.
KLF4 and miR-206 are enriched in MaCSCs derived from human patient-derived xenografts (PDXs) and the C3(1)/TAg GEMM KLF4 was similarly consistently elevated in lineage-negative (Lin -)/ALDH High MaCSCs isolated from human mammary tumor tissues that were passaged as PDXs (Figure 2a). miR-206 was upregulated in three of these four cases. Notably, none of these tumors displayed an appreciable CD44 + /CD24 -MaCSC population (data not shown), consistent with the variable expression of these markers in patient samples. 10,54,55 Tumorspheres are enriched for MaCSCs. 7,8 Compared with cells grown in adherent (two-dimensional (2D)) monolayers, tumorspheres formed from the Lincells of C3(1)/TAg mammary tumors showed elevated levels of Klf4 and miR-206 ( Figure 2b). ALDH High cells from other mammary cancer GEMMs were previously shown to have properties of MaCSCs. 50,52 Similar to the human tumors, Lin -/ALDH High cells of C3(1)/TAg mammary tumors also had increased Klf4 and miR-206 relative to ALDH Low cells (Figure 2c). These results identify KLF4 and miR-206 as MaCSC markers and potential mediators of MaCSC malignant properties.  36 Despite the reduced levels of KLF4, miR-206-transfected cells displayed higher P + /E + cell abundance relative to the control cells ( Figure 3c, right panel). Similar regulation of P + /E + cell abundance by miR-206 was observed for SUM159PT cells (Figure 3d). These results establish miR-206 as a potential effector of KLF4 for regulation of MaCSC abundance.
To determine whether miR-206 can promote the MaCSC phenotype, we assayed by limiting dilution the capability of miR-206-transfected MDA-MB-231 cells to initiate tumors in vivo. Consistent with an increased number of P + /E + cells, miR-206transfected cells formed tumors more efficiently in NOD/SCIDgamma (NSG) mice compared with control cells (Figure 3e The critical role of endogenous miR-206 for tumor initiation following orthotopic injection, despite its minimal effects on cell proliferation or MaCSC abundance, pointed to a potential role in regulating cell survival. We therefore assayed for resistance to cell death following matrix deprivation (anoikis), an intrinsic property of CSCs. 7,39-43 Indeed, consistent with our previous report that analyzed two human TNBC cell lines, 37 anti-miR-206 transfection sensitized several human TNBC models and a murine basal-like mammary cancer model (that is, M6 cells) to anoikis (Figure 4d 90 The GEO accession number is indicated. (c) Klf4 and miR-206 expression was evaluated in normal mammary tissues from FVB/N mice and in primary tumors arising in the MMTV-Neu and C3(1)/TAg GEMMs. RNA levels were determined by qRT-PCR. (d) MaCSCs were isolated from MDA-MB-231 cells by sorting using PROCR (P) and ESA (E) as described. 12 Transcript levels were analyzed in P + /E + and P -/Ecells. (e) MaCSCs were isolated from SUM159PT cells and analyzed similarly as described above for MDA-MB-231 cells. 53 For these cells, the P + profile was used as a surrogate for CD44 + /CD24 -. 10 (*P o0.05; **P o0.01; ***P o0.001).
miR-206 suppresses the translation of the tumor-suppressor PDCD4 and promotes tumor cell survival We previously reported that RAS-ERK signaling, a prosurvival pathway, is maintained in TNBC cells by KLF4, in part through its regulation of miR-206. 37 In contrast to the prominent effect of miR-206 on tumor initiation and cell survival, on its own this miR has only limited effects on ERK activity. 37 We therefore sought to better understand how endogenous miR-206 can promote anoikis resistance.
Enhanced KLF4-miR-206 signaling in MaCSCs C-C Lin et al The tumor-suppressor PDCD4 was identified as a potential miR-206 targeted transcript by multiple miR-target prediction tools. 37,56 PDCD4 is a negative regulator of RAS-ERK-AP1 signaling and protein translation, and promotes breast cancer cell apoptosis. [57][58][59] We therefore analyzed PDCD4 as a miR-206regulated transcript.
Consistent with regulation of PDCD4 by miR-206, KLF4 depletion in MDA-MB-231 cells increased PDCD4 expression (Figure 5a, left panel). Similarly, although anti-miR-206 treatment elevated PDCD4, transfection of miR-206 mimic was suppressive (Figure 5a, middle and right panels). Direct regulation of PDCD4 by miR-206 was determined using translational reporter assays.  (Figure 5b and c). In agreement with previous studies, PDCD4 depletion in TNBC cells promoted resistance to anoikis, with little or no effect on 2D proliferation ( Figure 5d).
Consistent with miR-206 regulation of PDCD4 in MaCSCs, the P + / E + sub-population of MDA-MB-231 cells exhibited decreased levels of PDCD4 mRNA and protein compared with non-MaCSCs ( Figure 5e). In TNBC cells, the depletion of PDCD4 was not sufficient to alter the abundance of the P + /E + fraction (Figure 5f). These results appear to support a selective role of PDCD4 for suppression of tumor cell survival.
miR-206 promotes cell survival by suppressing CX43 in MaCSCs Our identification of miR-206 regulation of PDCD4 led us to seek additional targets of this miR that may be important for promoting cell survival. DIANA-miRPath analysis identifies gap junction signaling as the top-ranked miR-206-regulated pathway (P = 2.58 × 10 − 6 ). 60 Among the targeted gap junction proteins, CX43 is a validated miR-206-regulated transcript, as previously shown in muscle cells. 38,61 CX43 is deficient in human breast tumor cells and MaCSCs, and may exert a tumor-suppressor role. [62][63][64][65][66][67][68] Consistent with its regulation by miR-206 in breast cancer cells, CX43 was increased in KLF4-depleted MDA-MB-231 cells (Figure 6a, left panel). Similarly, inhibition of miR-206 led to elevated CX43 levels, and transfection of miR-206 mimic was suppressive (Figure 6a, middle and right panels). In TNBC cells, the activity of a translational reporter containing the CX43 3' UTR was induced by 1.5-fold following anti-miR-206 treatment, and suppressed by 53% following transfection of miR-206 mimic (Figures 6b and c). Supporting the direct regulation of CX43 by miR-206 in breast tumor cells, mutation of site A (mut206-A) abolished regulation by miR-206 (Figure 6c). Similar to PDCD4 depletion, suppression of CX43 in TNBC cells promoted resistance to anoikis, with only subtle effects on cell proliferation (Figure 6d).
Compared with the non-MaCSC fraction, P + /E + MDA-MB-231 cells displayed lower CX43 mRNA and protein (Figure 6e). These results support a previous study that reported low CX43 expression in mammary TICs. 64 Similarly to PDCD4, knockdown of CX43 did not consistently alter the P + /E + cell abundance in TNBC cells, suggesting a selective role in tumor cell survival (data not shown).
miR-206 confers chemoresistance in TNBC cells Consistent with the promotion of cell survival by miR-206 as determined by anoikis assays, TNBC cells transfected with miR-206 mimic were more resistant to paclitaxel or doxorubicin (Figure 7a). Furthermore, inhibition of the endogenous miR-206 moderately sensitized TNBC cells to either agent (Figure 7b). Collectively, these results link pluripotency factor signaling and the enhanced cell survival of MaCSCs, supporting roles of KLF4-miR-206 signaling for breast tumor cell survival, chemoresistance, and tumor initiation through the repression of PDCD4 and CX43 (Figure 7c).

DISCUSSION
CSCs were first identified in hematopoietic malignancies and subsequently in solid tumors such as breast cancer. 1,3,5,69 Despite substantial progress, questions remain regarding the relationship of CSCs to the adult stem cells of normal tissue, and the nature of the signaling pathways that regulate CSC properties. 6 Despite this uncertainty, it is clear that CSCs represent a highly malignant subpopulation of tumor cells with the capability to resist therapy. 3,5 In TNBC cells, KLF4 directly regulates miR-206 transcription, and depletion of KLF4 consistently results in loss of the vast majority of miR-206. 36,37 In this study, we identified KLF4 and miR-206 as critical promoters of breast tumor cell survival. Both factors were preferentially expressed in the MaCSCs purified from 2D cell culture models of TNBC, from tumorspheres cultured in 3D, from human PDXs and from primary mouse mammary cancers. As shown by anti-miR treatment of TNBC cells, endogenous miR-206 directly repressed the translation of the tumor suppressors PDCD4 and CX43 and promoted tumor cell survival, chemoresistance and in vivo tumor initiation. Immunoblot analysis of MaCSCs revealed suppressed levels of both PDCD4 and CX43. Mirroring the role of endogenous miR-206, depletion of each tumor suppressor did not alter the abundance of CSCs, but instead enhanced tumor cell survival consistent with previous reports. 57,70 miRs can act as critical factors for regulating the abundance and/or survival of MaCSCs. [71][72][73][74] In skeletal muscle, miR-206 is important for repression of PAX7 during stem cell differentiation, and for muscle regeneration following injury. [75][76][77][78][79] In a mammary cancer context, miR-206 expression is elevated in ERtumors, which are enriched for MaCSCs. 44,45,[47][48][49] In agreement with previous studies, we observed that miR-206 is upregulated in human breast cancers that display a higher grade, in human TNBCs and in basal-like mammary tumors derived from the C3(1)/ TAg GEMM (Figure 1a and c).
Multiple previous studies have reported that enforced expression of miR-206 can suppress tumor cell proliferation, invasion or   metastasis. 45,[80][81][82][83] These tumor-suppressor-like effects of miR-206 may result from higher level enforced expression of the exogenous miR. In this study, suppression of endogenous miR-206 blocked tumor initiation, and moderate (fivefold) overexpression of exogenous miR-206 promoted initiation in a limiting dilution assay. In addition, we observed that  These results suggest that endogenous KLF4 can signal through miR-206 to promote tumor initiation, probably by impacting cell survival rather than MaCSC abundance. In contrast, exogenous KLF4 or miR-206 promoted MaCSC abundance, mirroring the role of exogenous KLF4 for generation of induced pluripotent stem cells. 18,19 It will be interesting to determine whether miR-206 similarly influences the generation of induced pluripotent stem cells.
In this study, we have identified endogenous KLF4 and a downstream effector, miR-206, as functional markers and prosurvival factors that are enriched in MaCSCs. Prosurvival signaling by miR-206 was attributed to direct regulation of PDCD4 and CX43, and miR-206 enhanced the chemoresistance of TNBC cells.
To construct a WT PDCD4 translational reporter, a 1.7-kb fragment representing the 3′ UTR was excised using MluI and inserted into MluIdigested pMIR-REPORT. To construct a WT CX43/GJA1 translational reporter, a 1.7-kb fragment representing the CX43 3′ UTR was generated by sequential treatment with EcoRI, Klenow fragment and MluI. This fragment was inserted into pMIR-REPORT vector that was prepared by sequential treatment with SacI, Klenow fragment and MluI.
PDCD4 and CX43 reporters with mutation in the miR-seed complementary regions were generated by PCR mutagenesis. Oligonucleotides are listed in Supplementary Table S1. WT reporters were mutated so as to conserve the predicted secondary structure of the 3′ UTR. 89 Cloned PCR products were confirmed by sequence analysis.

Transient transfection and translation reporter assays
The following anti-miR inhibitors (AM) and miR-mimics (PM) were obtained from Ambion and diluted to 20 μM in nuclease-free water: hsa-miR-206 (AM10409, PM10409), AM-negative control (AM17010), and PM-negative control (AM17110). Cells were subjected to reverse transfection and, 24 h later, forward transfection was performed as described. 36 At 24 h after the start of the forward transfection, cell extracts were prepared for expression studies, or cells were used for phenotypic studies. Translational reporter assays were performed following just one transfection, at 24 h after the start of the reverse transfection. Inhibitors/mimics were co-transfected with reporter plasmids, and Dual-Luciferase Reporter Assays (Promega) were performed as described. 36 Immunoblot analysis and antibodies Cell extracts for immunoblot analysis were prepared as previously described. 36 PARP cleavage assays were performed as recommended (Roche, Indianapolis, IN, USA). Following electrophoresis, proteins were
ALDH activity was evaluated by flow cytometry using the ALDEFLUOR assay (STEMCELL Technologies). Cell sorting or flow cytometry was performed on a BD FACSAria using BDFACSDiva software version 6.1, or on a BD Fortessa using BDFACSDiva software version 7.0 (Becton Dickinson, San Jose, CA, USA). For analysis, a minimum of 10 000 events were collected for each sample. The data were analyzed by using FCS Express 4 Research Edition software (De Novo software, Glendale, CA, USA).

Tumorsphere formation and anoikis assays
To grow tumorspheres, 2 × 10 4 Lincells were placed in suspension cultures in low attachment plates (Costar, Corning, NY, USA) using DMEM/F12 supplemented with B27, 4 μg/ml heparin, 20 ng/ml epidermal growth factor, 20 ng/ml fibroblast growth factor and 1% (wt/vol) methylcellulose. For analysis of anoikis, cells were suspended in culture as previously described. 37 Cell death was analyzed by propidium iodide staining and flow cytometry (Invitrogen, Carlsbad, CA, USA), by Trypan blue exclusion, or by analysis of cleaved PARP.

Expression analyses
Microarray data were extracted from GEO accessions GSE45666 and GSE23978 and then normalized to the geometric median. 90,91 For qRT-PCR, total RNA was extracted and mRNA and miR levels were analyzed as previously described. 36 Reactions were normalized to B2M or RPLP0 for mRNA analysis, or to U6 snRNA for miR analysis. Primer sequences are listed in Supplementary Table S2. PCR reactions were performed on a Mx3005P Real-Time PCR System (Stratagene, La Jolla, CA, USA). mRNA and miR levels were determined by the ΔΔC T method. 92 For all RNA measurements, three independent experiments were performed in duplicate manner.

Statistical analysis
Data were analyzed using either the unpaired t-test (two-tailed), or else one-way analysis of variance followed by Tukey's multiple comparison ad hoc post-test. Tumor volumes were analyzed using two-way analysis of variance with a Bonferroni post-test. Tumor initiation was analyzed using a 2 × 2 contingency table with a Fisher's exact test. Statistical analyses were performed in GraphPad Prism 5 (GraphPad Software, La Jolla, CA, USA. Differences were considered significant when the analysis yielded Po0.05.