Kruppel-like Pluripotency Factors as Modulators of Cancer Cell Therapeutic Responses

Kruppel-like factors (KLF) compose a family of 17 distinct zinc finger transcription factors that function in diverse cell types. A subset of these, including KLF2, KLF4, and KLF5, have been linked to pluripotency (1–3). Pluripotent stem cells (PSC) can differentiate into any of three germ layers including ectoderm, mesoderm, and endoderm. In tumor cells, KLF4 and KLF5, the focus of this review, can exert profound protumorigenic or antitumorigenic effects. Recent results implicate these two factors as mediators of adaptive responses following targeted therapy, and as prosurvival factors that can be preferentially expressed within cancer stem–like cells (4, 5).

Identifying a cooperative relationship within tumor cells, KLF4 and KLF5 were shown to function in concert to promote tumorigenesis and drug resistance in HER2-positive breast cancer models, with potential effects on distant metastasis-free survival in patients (4). These results may mirror a cooperative function of KLF4 and 5 in embryonic stem cells (ESC), where pluripotency is maintained by their conserved (i.e., convergent) and distinctive (i.e., divergent) signaling (3). In this review, we contrast recent insights obtained in breast cancer with observations made in other contexts including embryonic stem cells, normal adult tissues, and select pathologic conditions.

ESCs are derived from the inner cell mass of the blastocyst, and are propagated in vitro as self-renewing, undifferentiated, and rapidly proliferating cells, typically in the presence of the cytokine leukemia inhibitory factor (LIF), which signals through Stat3 to suppress cell differentiation (2). The pluripotency of ESCs is maintained by a core transcription factor network in which extensive cross regulation sustains the various components. Oct4 and Sox2 compose an essential “hub”, with “outer circuit” factors including Nanog, Tbx3, Esrrb, and distinct KLFs (2). Outer circuit members, although individually more or less dispensable for pluripotency, cooperate in ESCs to maintain the naïve, undifferentiated state. In addition, pluripotent cells can be obtained by reprogramming of differentiated adult somatic cells to induced pluripotent stem cells (iPSC), typically mediated by the exogenous quartet of OCT4, SOX2, KLF4, and c-MYC (i.e., OSKM). c-Myc is strictly dispensable for generation of iPSCs. Nevertheless, c-Myc promotes reprogramming, acting early in conjunction with KLF4 (6). The enhancing effect of c-MYC is attributed to its various effects on cell proliferation, metabolism, and the genome-wide regulation of transcription pause release to enhance RNA Polymerase II–mediated transcription (7, 8). Later during reprogramming, KLF4 cooccupies many loci with OCT4 and SOX2, together comprising a “pioneer” activity that efficiently induces chromatin structural changes including decondensation, leading to gene transcription (6, 9, 10). Interestingly, like KLF4, c-Myc has been characterized as a direct target of LIF–Stat3 signaling in ESCs (2, 7). Consequently, enforced expression of the four Yamanaka factors during generation of iPSCs substantially recapitulates the endogenous ESC pluripotency network.

KLF2, KLF4, and KLF5 are coexpressed in ESCs and their combined knockdown in these cells leads to robust differentiation, as does withdrawal of LIF (1, 2, 11). Within ESCs, the core KLFs appear to exert distinct effects (11). Chromatin immunoprecipitation analysis of KLF4 and 5 followed by next-generation sequencing has identified both shared and distinct target genes (3). Although KLF4 and 5 cooperate to suppress differentiation, each appears to preferentially inhibit differentiation of a specific lineage, respectively, the endoderm and mesoderm. While the ability of exogenous KLFs to sustain the ESC phenotype in the absence of LIF is shared, KLF2 and 4 are much more effective compared with KLF5 (2, 11). Similarly, for the reprogramming of somatic cells to iPSCs, KLF2 and 4 are far more efficient compared with KLF5 (11). Despite reports indicating that KLF5 promotes ESC pluripotency (1, 3), others favor the view that KLF5 has no specific or critical role in pluripotency, but rather plays a general role in the early embryo (personal communication, Austin Smith, University of Cambridge, Cambridge, United Kingdom; ref. 2). Such a noncritical function of KLF5 in pluripotency does not preclude a critical role in adult stem cells. Indeed, conditional deletion of Klf5 within the Lgr5+ intestinal stem cells of mice conferred a selective disadvantage relative to residual Klf5+ stem cells, and completely prevented tumorigenesis driven by cotemporal, conditional activation of β-catenin in the same stem cells (12).

When mutually present in epithelial tissues, often times KLF4 and 5 localize to distinct compartments. In normal adult tissues, KLF4 is commonly associated with postmitotic, differentiated cells, whereas KLF5 is more highly expressed in proliferating cells (13, 14). In tumor tissues, this compartmentalization is often lost, and the two appear to be coexpressed in the same cells. In support of the opposing functions of KLF4 and 5 in somatic tissues, each appears to regulate cell-cycle factors such as p21^Waf1/Cip1, p15^Ink4B, p27^Kip1, CCNB1, CCND1, and/or CCNE1, but often with opposite effects (13).

Despite these distinctions, several contexts appear to support convergent signaling. Conditional deletion of Klf4 or Klf5 in the surface ectoderm of the mouse eye reveal similar developmental deficits (15, 16). Indeed, genetic ablation of either Klf in the eye results in dramatically reduced goblet cell number and function. Similarly in the gut mucosa, epithelial knockout of Klf4 or 5 resulted in goblet cell deficits (17, 18).

Early studies implicated KLF4 as a stress responsive gene, with induction by factors such as shear stress or wounding (19). Interactions between KLF4 and 5 have been especially well examined in the cardiovascular system. Although they are highly expressed in the embryonic vasculature, typically levels of these factors are low in normal adult vascular smooth muscle cells (SMC; ref. 20). However, KLF4 and 5 are induced in SMCs in response to various noxious stimuli. Vascular injury models including shear stress, coronary atherosclerosis, vein graft hyperplasia, oxidized phospholipids, and/or ischemia can trigger an adaptive response in SMCs prompting dedifferentiation and proliferation (20,21). In conditional Klf4^−/− mice, the damage-related suppression of SMC differentiation markers is delayed compared with wild-type mice (21). Although the neointimal layers of these mice exhibit elevated cell growth in response to injury, conditional Klf4^−/− mice are nevertheless more vulnerable to vascular insults, suggesting defects in repair due to impaired dedifferentiation. These results appear to mirror KLF4-deficient breast cancer cells, which proliferate somewhat more rapidly than control cells but also show increased cell death following stressors such as matrix deprivation (i.e., anoikis) or exposure to a targeted therapeutic agent (4, 5, 22).

On the other hand, KLF5 may mediate cardiovascular stress responses though enhanced proliferation. Mice heterozygous for Klf5 show reduced response to cardiovascular insults such as arterial polyethylene cuffs, characterized by thinner medial and intimal layers relative to control mice (20). Moreover, KLF5 expression confers increased cell survival in atherosclerotic vascular lesions (23). Interestingly, KLF5 as well as KLF4 is induced by endothelin-1 in neonatal rat myocytes. Endothelins are the most potent known vasoconstrictors, have important roles in pressure regulation of the vasculature, and are implicated in cardiovascular disease (24). These results highlight roles for KLF4 and 5 in the injury response program of vascular SMCs.

In addition to their role in the vasculature, KLF4 and 5 mediate survival to noxious stimuli in other contexts. Early studies reported elevation of KLF4 levels in keratinocytes during wound healing, and the induction of KLF4 by heat stress that is critical to cell viability (19, 25, 26). Subsequent studies demonstrated an important role for KLF4 in cellular survival following γ-radiation (27–29). In gut epithelium or gut-derived malignant cells, KLF4 attenuates radiation-induced damage and cell death (27, 28). Mechanistically, the radioprotective effects of KLF4 are in part attributed to p53 signaling (27, 29). The p53-dependent induction of KLF4 causes growth arrest through p21^Waf1/Cip1, where the cessation of the cell cycle may allow for adequate DNA repair. In addition, KLF4 also inhibits proapoptotic Bax expression in response to γ-radiation (27). More recently, KLF4 was found to be vital for protecting neurons and fibroblasts from oxidative damage (30,31). Taken together, KLF4 is capable of protecting cells from a variety of harmful stimuli and may be associated with a conserved cellular response to toxic insults.

Similar to KLF4, KLF5 participates in the adaptive response to external stressors in several organ systems. In the gut, KLF5 enhances protection and recovery to dextran sodium sulfate–induced injury (32). Likewise, KLF5 is a key mediator in the colonic response to pathogenic bacterial infection. KLF5 promotes survival in the pulmonary blood vessels as well. In hypertensive settings, KLF5 contributes to pulmonary artery SMC survival, and its expression correlates with increased levels of activated survivin in experimental models of congenital diaphragmatic hernia (33, 34). In fact, KLF5 enhances antiapoptotic activity through survivin in a variety of contexts. Alternatively, KLF5 can modulate other aspects of apoptotic signaling. In response to intracerebral hemorrhage, KLF5 protects against neuronal apoptosis by suppressing proapoptotic Bad activation (35).

As for KLF4, KLF5 function is heavily intertwined with that of p53. KLF5 and p53 show coordinate regulation at promoters for genes such as BIRC5 (encoding survivin), hypoxia-inducible factor HIF1A, CDKN1A (encoding p21^Waf1/CIP1), and NOTCH1 (36–39). In keratinocytes, p53 status provides a major contextual determinant for the output of KLF5 signaling. Here, mutation or loss of p53 switches KLF5 signaling so as to suppress cell proliferation, mediated by KLF5 induction of p21^Waf1/Cip1 (39). p53 mutation or deficiency also renders NOTCH1 transcription to be dependent upon KLF5, and consequently tumor cell invasion is associated with loss of KLF5 expression and with loss of KLF5-dependent induction of the NOTCH1, a potent tumor suppressor in this context (38). In summary, the consistent involvement of KLF4 and 5 in the stress response in multiple organ systems argue that these factors are conserved responders to the damage of cells and tissues.

Although elevated expression of KLF4 and 5 in tumor cells relative to normal tissue is quite common, genetic alterations within either gene in cancer is rather uncommon, arguing against roles as a classical oncogene or tumor suppressor. Instead, KLFs appear to be critical to stress-responsive plasticity and prosurvival factors not only in normal cells and tissues but also in their malignant derivatives. In short, based upon current literature, tumor cells appear have acquired these critical regulators of plasticity and survival by inheritance rather than by somatic genetic alteration. Nevertheless, these factors appear formidable as components of the tumor cell stress response that mediates resistance of tumor cells to a wide variety of insults, including therapy.

Consistent with this view are the disparate and context-dependent effects of KLF4 and 5 in cancer models. KLF5 has typically been implicated as protumorigenic, including breast, bladder, colon, lung, gastric, and HPV-associated cancers, and is often associated with the promotion of cell proliferation (29). On the other hand, KLF5 is linked to tumor suppressor activities in prostate cancer and esophageal squamous cell carcinoma. In the squamous cell carcinoma context, it appears to function as a sort of surrogate for p53, inducing tumor suppressors such as p21^Waf1/Cip1 and NOTCH1 whenever p53 is deficient (38, 39).

Given its ability to suppress proliferation, to promote or suppress differentiation, and to impact cell survival, it is not surprising that KLF4 can elicit dual effects on malignant properties. For example, KLF4 has been reported to inhibit tumorigenesis and/or malignant properties in a variety of contexts, including neuroblastoma, leukemia, pancreatic, lung, and colon cancers (29). Conversely, adverse impact on clinical outcome and/or protumorigenic effects in functional assays have been reported for KLF4 in breast, skin, and others, possibly attributed to transcriptional regulation of miR-206 and miR-21 leading to RAS–ERK pathway activation, to prosurvival signaling, and/or to the suppression of TP53 transcription (5, 22, 40). KLF4 appears especially apt to exert protumorigenic effects in cells containing dysregulation of Rb-dependent G₁ cell-cycle progression, as results from expression of DNA tumor virus proteins such as adenovirus E1a, from signaling of RAS to Cyclin D1, or from insufficiency of CDK inhibitors such as p21^Waf1/Cip1 (29, 40).

Without precluding potential roles in the homeostasis of adult stem cells, tissue maintenance, and/or tumor pathogenesis, our studies and those of others appear to support a model in which KLF4 and 5 play a vital role in the response to therapy. Such functions are likely adapted from their role as stress response factors in normal tissues. Consequently, some of the drug resistance of cancer cells and CSCs, in part attributed to their ability to undergo adaptive reprogramming in response to a targeted therapy, may be “inherited” from the tumor precursor cell rather than selected for through clonal evolution during tumor development (41, 42).

As noted above, KLF4 has a well-established role in cellular survival following therapies such as γ-radiation (27–29). Moreover KLF4, in part via miR-206 signaling, promotes breast cancer stem cell features and conveys enhanced cell survival and tumorigenicity in mice, as represented by increased tumor initiation, anoikis resistance, and resistance to cytotoxic drugs (5). Similarly, KLF5 can confer resistance to chemotherapies by upregulation of survivin or HIF1α (36, 43). In addition, depletion of KLF5 conferred colon cancer cell sensitivity to DNA-damaging agents through activated Pim1 (44).

Most cancer studies have examined the function of KLF4 and 5 in relatively permissive contexts, without exposure to any therapeutic agent. Conversely, tumors in human patients are subjected to diverse stresses including antitumor immune responses, conventional chemotherapy, targeted therapy, and radiotherapy. As stress response factors, examining how KLF4/KLF5 influence the cellular adaptation to noxious stimuli may be more informative to their clinical relevance.

In breast cancer, KLF4 and 5 expressions have been repeatedly associated with poor patient outcome (4, 45–47). When KLF4 and KLF5 transcript levels were examined in tandem in breast cancer, higher coexpression was correlated with a poor outcome selectively within the HER2-enriched subtype (4). In this study, endogenous KLF4 and 5 cooperated to suppress the therapeutic response to HER2 inhibition through the induction of antiapoptotic factors MCL1 and BCL-XL, an example of convergent signaling. Consistent results were obtained in human and mouse models of breast cancer. The rapid induction of antiapoptotic factors by KLF4 and 5 appears to represent a type of adaptive reprogramming, a process previously linked to the induction of receptor tyrosine kinases when tumor cells are treated with agents such as MEK inhibitor or lapatinib (41, 42). Interestingly, the individual contribution of KLF4 or 5 paled in comparison with their combined output. Cooperativity in tumor cells was demonstrated for the endogenous factors, which combined to promote not only drug resistance but also the growth of xenografted tumors (4).

In addition to KLF4 and 5, an extensive literature currently indicates that tumor cells can coexpress each of the OSKM factors, indicating the potential for a cross regulating, adaptable pluripotency-like network that could mediate tumor cell plasticity. For example, we recently analyzed expression of OSKM in pancreatic tumor cells and cancer stem–like cells (CSC; ref. 48). Tumor cells expressed all four Yamanaka factors, and three of the four (i.e., OKM) were maintained within CSCs by the transcription factor GLI1, expression of which is driven by activated KRAS in this tumor type. This example provides a link between a common cancer genetic alteration, KRAS mutation, and an endogenous pluripotency factor–like program.

As reviewed elsewhere, there is remarkable overlap in the phenotype of PSCs and tumor cells, including multiple hallmarks of cancer that appear to be present within PSCs (Fig. 1; refs. 2, 6, 9, 49–51). Supporting the idea that tumor cells contain a pluripotency-like network capable of adaptive responses, we have observed upregulation of Nanog and c-MYC transcripts in breast cancer cells deficient in KLF4 and/or KLF5 (unpublished data). Providing some rationale that these results represent a compensatory response, KLF4 and Nanog show cross regulation in ESCs, and Nanog has been used with OCT4 and SOX2 to reprogram human cells (9). Also, c-MYC and KLF4 together compose a first wave of transcription factor binding during iPSC reprogramming (6).

Given the expression of multiple hallmarks of cancer within PSCs (Fig. 1), what factors might distinguish the epigenetically determined PSC phenotype from the invasive and metastatic phenotype of genetically defective tumor cells? One distinguishing trait of PSCs is low RAS–MEK–ERK signaling (2). RAS–ERK activation induces differentiation in PSCs, whereas MEK inhibitor helps to maintain self-renewal. Another distinguishing trait is that PSCs are epithelial, whereas carcinoma cells derive from epithelia but typically show some loss of epithelial features, and may express epithelial and/or mesenchymal markers. Given the considerable phenotypic overlap between PSCs and tumor cells (Fig. 1), it may be revealing of the cancer phenotype to further investigate the potential role of pluripotency OSKM or similarly acting, OSKM-like networks in adaptive processes such as reprogramming induced by drug exposure (6, 52).

Although KLF4 and 5 have multiple effectors to impact cell survival, both KLFs regulate components of the intrinsic pathway of apoptosis, highlighting an attractive node to target therapeutically. Currently under clinical investigation are small-molecule inhibitors that target the KLF-regulated BCL2 family of antiapoptotic factors. These may be useful agents to attenuate the KLF-mediated adaptive responses (4). Obatoclax (GX15-070) and Navitoclax (ABT-263) are BH3-mimetics that disrupt BCL2 family member protein–protein interactions, thereby enabling enhanced proapoptotic signaling. Where Obatoclax broadly inhibits all the antiapoptotic BCL2 family members with a reduced potency, Navitoclax has a high affinity for BCL-2, -XL, and –W (53). Combined, these drugs are involved in 42 active/completed phase I/II clinical trials for various malignancies; however, their efficacy in breast cancer patients has yet to be reported. Although these drugs are generally well tolerated in patients, preliminary data suggest they have limited utility as single-agents, advocating consideration for combination therapy instead.

High-throughput screening identified the small-molecule ML264 as a selective inhibitor of KLF5 expression, with recent documentation of efficacy in a mouse model (54). A promising avenue of research includes the use of iPSC and ESC technology to screen for other inhibitors of core pluripotency factors. Small molecules that impact chromatin modifiers and chromatin-remodeling enzymes have great promise to impact both regenerative medicine and cancer therapeutics. Consequently, the careful dissection of pluripotency factor biology and signaling, including the Kruppel-like factors, their regulators, and their effectors holds substantial promise.

The convergence of developmental biology and cancer has been evident for decades and continues unabated. Kruppel-like factors represent yet another example of this convergence, and it seems likely that there will be conserved roles of KLFs across several fields of research broadly related to epigenetics and cellular reprogramming, including PSC biology, CSC biology, and the cancer cell adaptive reprogramming response that follows treatment with a variety of small molecules. Current research efforts seem likely to generate new candidate therapeutics that are highly relevant across disciplines. To achieve timely progress in the regenerative medicine and cancer therapeutic arenas, it will be critical for researchers to integrate results from disparate areas of research and to adapt and reprogram accordingly.

No potential conflicts of interest were disclosed.

The authors regret that due to space limitations it was not possible to cite all of the relevant literature.

This work was supported by grants NCI RO1 CA127405 (to J.M. Ruppert) and the Jo and Ben Statler Chair in Breast Cancer Research.