Research Article

Activation of PKA leads to mesenchymal-to-epithelial transition and loss of tumor-initiating ability

See allHide authors and affiliations

Science  04 Mar 2016:
Vol. 351, Issue 6277, aad3680
DOI: 10.1126/science.aad3680

Have cancer stem cells MET their match?

Solid tumors have been hypothesized to contain a subset of highly aggressive cells that fuel tumor growth and metastasis. The search is on for drugs that selectively kill or diminish the malignant properties of these tumor-initiating cells (TICs; previously called “cancer stem cells”). Pattabiraman et al. hypothesized that compounds that induce TICs to undergo a phenotypic change called the mesenchymal-to-epithelial transition (MET) would therefore cause TICs to lose their tumor-initiating ability. Indeed, drugs activating the protein kinase A signaling pathway triggered an epigenetic reprogramming of TICs that resulted in the cells acquiring a more benign epithelial-like phenotype.

Science, this issue p. 10.1126/science.aad3680

Structured Abstract


Tumor-initiating cells (TICs) have emerged in recent years as important targets for cancer therapy owing to their elevated resistance to conventional chemotherapy and their tumor-initiating ability. Although their mode of generation and biological properties have been explored in a diverse array of cancer types, our understanding of the biology of TICs remains superficial. The epithelial-to-mesenchymal transition (EMT) is a cell-biological program that confers mesenchymal traits on both normal and neoplastic epithelial cells, which enables both to acquire stemlike properties. In the case of carcinoma cells, entrance into a more mesenchymal state is associated with elevated resistance to a variety of conventional chemotherapeutics. This association between the EMT program and the TIC state has presented an attractive opportunity for drug development using agents that preferentially target more mesenchymal carcinoma cells, rather than their epithelial counterparts, in an effort to eliminate TICs. Adenosine 3′,5′-monophosphate (cAMP) is a second messenger that transmits intracellular signals through multiple downstream effectors; the most well studied of these is protein kinase A (PKA). In this study, we explore the role of PKA in determining the epithelial versus mesenchymal properties of mammary epithelial cells and how this signaling pathway affects the tumor-initiating ability of transformed cells.


At least two approaches might be taken to target mesenchymal TICs. One strategy that has been used previously is the development of agents that show specific or preferential cytotoxicity toward TICs. In the current study, we have embraced an alternative strategy that is designed to induce TICs to undergo a mesenchymal-to-epithelial transition (MET). This “induced differentiation” approach would trigger cells to exit the more mesenchymal tumor-initiating state and enter into an epithelial non-stemlike state. In principle, this transition would make the cells more vulnerable to conventional cytotoxic treatments and thereby reduce the likelihood of metastasis and clinical relapse.


To identify agents that might induce an MET in mesenchymal mammary epithelial cells, we performed a screen for compounds that stimulate transcription of CDH1, which encodes E-cadherin, a key epithelial protein. Through this screen, compounds that activate adenylate cyclase (cholera toxin, CTx; and forskolin, Fsk) were identified as key inducers of the epithelial state. We found that mesenchymal cells treated with either CTx or Fsk differentiated into benign epithelial derivatives that had lost their ability to effectively initiate tumors and that were more susceptible to conventional chemotherapeutic agents in vitro. Further interrogation revealed that these agents elevated the intracellular levels of cAMP, which in turn activates PKA. PHF2, a histone H3 with acetylated lysine 9 (H3K9) histone demethylase and PKA substrate, was found to be essential for the cAMP-induced MET. By studying the genome occupancy of PHF2 and the epigenomic state of the cells before and after PKA activation, we determined that PHF2 promotes the demethylation and derepression of epithelial genes that ultimately contribute to acquisition of an epithelial state.


We conclude that PKA participates in the differentiation of TICs by enforcing residence in the epithelial state and preventing or reversing the EMT program. Our study reveals a new direction for targeting the TIC population. We propose that pharmacological induction of epigenetic reprogramming of these cells could promote their differentiation to a more epithelial state and increase their susceptibility to conventional chemotherapeutic drugs.

Induction of the MET as a potential cancer therapy.

TICs have mesenchymal attributes that contribute to their ability to seed new tumors. Treatment of TICs with compounds that increase cAMP levels (e.g., CTx and Fsk) activates PKA. This leads to epigenetic reprogramming through subsequent activation of the histone demethylase PHF2, a PKA substrate, which in turn promotes differentiation of the cells into a more epithelial state, accompanied by a loss of their tumor-initiating ability. Drugs targeting various steps of this signaling pathway might be developed into a differentiation-based cancer therapy for certain breast cancers.

Cite this article as D. R. Pattabiraman et al., Science 351, aad3680 (2016). DOI: 10.1126/science.aad3680


The epithelial-to-mesenchymal transition enables carcinoma cells to acquire malignancy-associated traits and the properties of tumor-initiating cells (TICs). TICs have emerged in recent years as important targets for cancer therapy, owing to their ability to drive clinical relapse and enable metastasis. Here, we propose a strategy to eliminate mesenchymal TICs by inducing their conversion to more epithelial counterparts that have lost tumor-initiating ability. We report that increases in intracellular levels of the second messenger, adenosine 3′,5′-monophosphate, and the subsequent activation of protein kinase A (PKA) induce a mesenchymal-to-epithelial transition (MET) in mesenchymal human mammary epithelial cells. PKA activation triggers epigenetic reprogramming of TICs by the histone demethylase PHF2, which promotes their differentiation and loss of tumor-initiating ability. This study provides proof-of-principle for inducing an MET as differentiation therapy for TICs and uncovers a role for PKA in enforcing and maintaining the epithelial state.

Tumor-initiating cells (TICs), also known as cancer stem cells, are defined operationally by their ability to seed new tumors when implanted in appropriate hosts. They have emerged in recent years as important targets for cancer therapy, owing to their elevated resistance to conventional chemotherapy and their tumor-initiating ability—the latter allows them to metastasize and to drive clinical relapse (1, 2). Although their mode of generation and biological properties have been explored in a diverse array of cancer types (3), our understanding of the biology of TICs remains superficial. Cytotoxic therapies designed specifically to eliminate TICs might be targeted, for example, to interdict the signaling pathways that are used preferentially or uniquely by these cells (4). At present, however, the nature of such TIC-specific signaling pathways remains to be fully elucidated.

The epithelial-to-mesenchymal transition (EMT) is a cell-biological program that confers mesenchymal traits on both normal and neoplastic epithelial cells (5). In addition, activation of an EMT program enables both classes of cells to acquire stemlike properties (6, 7). Indeed, TICs from several carcinoma types have distinct mesenchymal attributes, which suggests that they have passed, at least partially, through an EMT (79). This association between the EMT program and the TIC state has presented an attractive opportunity for drug development using agents that preferentially target more mesenchymal carcinoma cells, rather than their epithelial counterparts, in an effort to eliminate TICs.

At least two approaches might be taken to target mesenchymal TICs. One strategy would be to develop agents that show specific or preferential cytotoxicity toward TICs (1). In this study, we have embraced an alternative strategy that is designed to induce TICs to exit the more mesenchymal tumor-initiating state and enter into an epithelial non-stemlike state. Such induced differentiation should, we reasoned, place cells in a state where they would become more vulnerable to conventional cytotoxic treatments. Accordingly, we screened for agents that could induce a mesenchymal-to-epithelial transition (MET) and, thereby, uncovered the central role of adenosine 3′,5′-monophosphate (cyclic AMP or cAMP) and its downstream target, protein kinase A (PKA), in governing the transition of cells from the mesenchymal to the epithelial state.

cAMP is a second messenger that transmits intracellular signals when certain hormones and neurotransmitters interact with receptors on the plasma membrane (10). cAMP regulates multiple downstream effectors; the first of these to be identified and the most well studied is PKA (11), which plays numerous roles in various cell types and operates in several subcellular locations (11). Because PKA is initially assembled as a heterotetrameric holoenzyme, its activity depends on cAMP binding to its two regulatory subunits, which leads to the release of active catalytic subunits and the phosphorylation of a diverse array of substrates (12).

In previous work, PKA has been shown, under some conditions, to promote an EMT; PKA was shown to regulate the transcription factor Snail in one study; and another study demonstrated that hypoxia-inducible factor 1α (HIF-1α) could regulate transcription of PRKACA under hypoxic conditions (13, 14). However, PKA signaling has been shown to favor the epithelial state, but the mechanistic understanding of this phenomenon is limited. One report showed that schwannomas in mice without Prkar1a (encoding the PKA regulatory subunit) exhibited loss of vimentin and gain of cytokeratins and E-cadherin (15), whereas another study revealed inhibition of the formation of mesoderm-derived structures in Prkar1a null mice (16). A recent study reported that deletion of the Gαs subunit repressed the activity of PKA, which limited the proliferative potential of epithelial hair follicle stem cells (17). Nevertheless, the connection of PKA signaling to TICs and the stemlike state is poorly understood, and the exploitation of this pathway as a differentiation-based cancer therapy has not been explored.

Identification of agents that induce an MET in mammary epithelial cells

Human breast cancers are characterized by cells that show various degrees of epithelial and mesenchymal properties, as revealed by the expression pattern of markers, such as cytokeratins and vimentin (fig. S1). Almost 85% of the carcinomas we examined showed varied expression patterns of cytokeratins, which indicated that the loss of epithelial properties is a commonly occurring event. Notably, ~10% of the carcinomas we examined exhibited high degrees of intratumoral heterogeneity, created in part by the presence of subpopulations of neoplastic cells that have both epithelial and mesenchymal properties. These are reminiscent of cells that have undergone an EMT, which resemble TICs that have a higher tumor-initiating propensity and an increased resistance to chemotherapy (18). To model the behavior of these subpopulations of carcinoma cells from human basal-like breast cancers, we used immortalized human mammary epithelial cells (HMLE cells) (19), which display an epithelial morphology; express E-cadherin at adherens junctions; and have low levels of mesenchymal markers, such as vimentin and fibronectin. They also exhibit a CD44lo/CD24hi cell surface marker phenotype that is characteristic of previously reported cells that lack stemlike properties (non-CSCs) (20). We also used their spontaneously arising mesenchymal derivatives, termed NAMEC8 (N8) cells (21). Relative to their HMLE counterparts, N8 cells express mesenchymal markers, such as vimentin and fibronectin, as well as the EMT-inducing transcription factors Snail and Zeb1 at higher levels; lack expression of E-cadherin at prominent cell junctions; and display a CSC-like CD44hi/CD24lo cell surface marker profile (Fig. 1, A to C). They also have a greater propensity to form mammospheres (Fig. 1, D and E), which is often used as an in vitro surrogate assay for the stemness of mammary epithelial cells. They are more efficient at migration through a trans-well membrane and invasion through a Matrigel-coated Boyden chamber membrane (Fig. 1, F and G); both in vitro assays represent models of cancer cell invasiveness in vivo. N8 cells are also more resistant to treatment with chemotherapeutic drugs, such as doxorubicin and paclitaxel (Fig. 1, H and I), as shown previously (21). Hence, two cell types represent epithelial and mesenchymal derivatives of mammary epithelial cells of common origin that were used to model the two cell states and how they affect tumor initiation and progression.

Fig. 1 Induction of an MET on treatment of N8 cells with CTx or Fsk.

Mesenchymal N8 cells acquire an epithelial morphology as adjudged by their morphology (A), loss of a stemlike CD44hi/CD24lo profile to assume a predominantly CD44lo/CD24hi profile (B), and expression of E-cadherin at cell junctions and loss of vimentin (C). Reverted N2-CTx and N3-Fsk cells lose their ability to form (D and E) mammospheres (P < 0.05, n = 4), (F) to migrate (P < 0.05, n = 4) and (G) to invade in transwell assays (P < 0.05, n = 4) and to acquire increased sensitivity to treatment with (H) doxorubicin and (I) paclitaxel (P < 0.05, n = 4). (J) Heat map of mRNA-Seq data, which demonstrates similarity in gene expression between HMLE, N8, and N8-CTx cells. Data (means ± SD) in (E), (F), and (G) were analyzed by Student’s t test; (H) and (I) were analyzed by Bonferroni correction. All scale bars, 25 μm.

To search for agents that can induce an MET, we performed a screen to identify compounds that could induce transcription of CDH1, which encodes E-cadherin, a key epithelial protein, in N8 cells. As a reporter for the activity of the CDH1 gene, we constructed a lentiviral vector that expresses a portion of the CDH1 promoter fused to luciferase (fig. S2A). We performed a screen using a 400-compound library for agents that were able to induce the CDH1-driven luciferase expression in N8 cells (fig. S2B). Most striking was the behavior of forskolin (Fsk), an adenylate cyclase activator that induced a 40-fold increase in luciferase activity (fig. S2C). Another adenylate cyclase activator, cholera toxin (CTx), was also able to induce an increase in luciferase activity (fig. S2D), which suggested that activation of adenylate cyclase could induce the acquisition of epithelial properties.

Fsk or CTx and the induction of an MET in mammary epithelial cells

We found that treatment of N8 cells in monolayer culture with either CTx or Fsk for a period of 14 days induced the formation of islands of cells with the characteristic cobblestone morphology of epithelial cells; such cells acquired the expression of E-cadherin at adherens junctions along with a loss of mesenchymal markers, such as vimentin (Fig. 1, A and C). Also, the cell surface marker expression profile of the N8 cells switched from a stemlike CD44hi/CD24lo to a nonstem CD44lo/CD24hi phenotype after this treatment (20) (Fig. 1B). These shifts were accompanied by a 100-fold increase in CDH1 mRNA levels, as well as decreases in the mRNA levels of Snail, Twist1, and Zeb1 EMT-inducing transcription factors (EMT-TFs) to 25%, 20%, and 14% , respectively, of the N8 cells before the transition (fig. S3, A and B). Treatment of N8 cells with either CTx or Fsk resulted in a near-complete loss of mammosphere-forming ability (Fig. 1, D and E), as well as their ability to migrate and invade (Fig. 1, F and G). There were no significant differences in the rates of proliferation between the N8 cells and their CTx- and Fsk-treated derivatives (fig. S3C). Of additional interest, withdrawal of CTx after 14 days of treatment led to cell populations that continued to reside in an epithelial state for >2 months in culture.

Reversion to an epithelial state, ostensibly similar to that of HMLE cells, rendered the N8 cells 8 times as sensitive to killing by doxorubicin [lowered the median inhibitory concentration (IC50) from 1.39 μM to 0.159 μM] and 13 times as sensitive to paclitaxel (lowered the IC50 from 4.79 μM to 0.35 μM) (Fig. 1, H and I). Also, the induced MET resulted in increased sensitivity to a range of chemotherapeutic drugs and inhibitors including methotrexate, 90-kD heat shock protein (HSP90) inhibitors, proteasome inhibitors, and epidermal growth factor receptor–mitogen-activated protein kinase (EGFR-MAPK) pathway inhibitors, as observed when we screened against two small-molecule libraries (Selleck Anticancer Compound Library and Enzo Kinase Inhibitor Library) (fig. S4). Hence, the induction of an MET rendered the N8 cells more sensitive to a range of drugs and inhibitors, which points to its utility as a means of overcoming therapeutic resistance. It also reinforces the notion that mesenchymal cells are more resistant to a range of cytotoxic agents.

We then performed mRNA sequencing (mRNA-Seq) to compare the global gene expression profiles of the mesenchymal N8 and the reverted N8-CTx cells in order to view the transcriptional changes that occur after the induction of MET. As determined by differential gene expression (Fig. 1J and tables S1 and S2) and principal component analyses (fig. S3D), the N8-CTx cells assume a gene expression profile that is almost completely converted to that of the epithelial HMLE cells and is significantly different from the mesenchymal N8 cells (Fig. 1J). Gene set enrichment analyses showed that the changes in gene expression from N8 to the N8-CTx cells are highly similar to several previously published EMT and MET gene signatures (2224) (fig. S3E). Taken together, these observations demonstrated a transition of the N8 cells from a mesenchymal-like state to a bona fide epithelial state, which rendered these cells more sensitive to a variety of drugs with potentially important therapeutic implications.

Effects of Fsk and CTx on intracellular cAMP levels and PKA

To confirm that both Fsk and CTx were working through alteration of cAMP levels, we measured the levels of this second messenger in both HMLE and N8 cells using liquid chromatography–mass spectrometry (LC-MS). Treatment with CTx resulted in a six- to eightfold increase in the intracellular levels of cAMP, which could be dampened by exposure to SQ22536, an inhibitor of adenylate cyclase, the enzyme responsible for the formation of cAMP (Fig. 2A).

Fig. 2 cAMP increases activate PKA, which is both necessary and sufficient for the induction of an MET in N8 cells.

(A) Mass-spectrometry measurement of cAMP levels in N8 cells that have been treated with CTx or Fsk alone and in combination with adenylate cyclase inhibitor SQ22536 (means ± SD, P < 0.05, n = 3). (B) Treatment of N8 cells with either 8-CPT-2me-cAMP or 8-Br-cAMP to identify downstream pathways that are responsible for induction of an MET. Knockdown of either PRKACA or PRKACB prevents CTx from inducing an MET in N8 cells as observed by changes in (C) morphology, (D) immunofluorescence for E-cadherin and vimentin, and (E) CD44/CD24 status. (F) Morphological changes of N8 cells undergoing an MET upon ectopic expression of an active mutant of PKA (caPKA). Data in (A) were analyzed using the Student’s t test. All scale bars, 25 μm.

The major downstream targets of cAMP are exchange proteins activated by cAMP (EPAC1/2) (25); cyclic nucleotide–gated ion channels that are primarily found in cells of the kidney, heart, testis, and central nervous system (26); and the most commonly studied downstream effector, PKA (11). To delineate the downstream pathways that are activated in response to increase in cAMP levels, we treated N8 cells with two cAMP analogs—8-bromoadenosine 3′,5′-cyclic monophosphate (8-Br-cAMP), which is known to preferentially activate PKA (27), or 8-(4-chlorophenylthio)-2′-O-methyladenosine-3′,5′-cyclic monophosphate (8-CPT-2Me-cAMP), which is a selective activator of the exchange proteins activated by cAMP (EPACs) (28). As was seen with Fsk or CTx, treatment with 8-Br-cAMP was also able to induce an MET in N8 cells, whereas 8-CPT-2Me-cAMP treatment had no effect on their mesenchymal properties (Fig. 2B). This allowed us to conclude that PKA, rather than the cAMP-activated exchange proteins, was more likely to play a central role in the MET process.

Knockdown of the catalytic subunit of PKA using two different small hairpin RNAs (shRNAs) (fig. S5A) abrogated the CTx-induced MET process in N8 cells, as demonstrated by their inability to develop a clear epithelial morphology; to acquire junctional E-cadherin; and to shed mesenchymal markers, such as fibronectin (Fig. 2, C and D). Moreover, treatment of these PKA-knockdown cells with CTx failed to induce an effective transition from the CD44hi/CD24lo stemlike state to the CD44lo/CD24hi non-stemlike state, which was otherwise observable in the absence of PKA knockdown (Fig. 2E). These results further reinforced the important role of PKA in the MET process.

Then, we tested whether PKA activity, independent of cAMP, was sufficient to induce an MET. Thus, we ectopically expressed a doxycycline-inducible constitutively active, cAMP-independent, constitutively active mutant form of PKA (caPKA) (29) in N8 cells and found that it was capable of inducing a reversion to the epithelial state in 7 to 10 days (Fig. 2F). Hence, it appears as though PKA is both necessary and sufficient to induce the acquisition of epithelial properties in the N8 cells.

We tested the role of CTx or Fsk in inducing an epithelial state in other cell systems to assess the generality of our observations. Removing CTx from the standard culture medium of MCF10A immortalized human mammary epithelial cells (30), caused them to acquire mesenchymal properties, to lose cell-cell adherens junctions, to lose their characteristic cobblestone morphology, and to gain a CD44hi/CD24lo cell surface marker profile. They also lost E-cadherin expression and exhibited an increase in expression of Zeb1, Vim, and FN1 (fig. S6, A to E). Readdition of CTx or forced expression of caPKA led to the reacquisition of epithelial features (fig. S6, A to E). Moreover, the MCF10A cells that lost epithelial properties upon CTx withdrawal were 4 times as resistant to treatment with doxorubicin and so extended our observations made in N8 cells that the mesenchymal variants were more resistant to conventional chemotherapeutic agents (fig. S6F).

We then proceeded to test the role of CTx and/or Fsk in a series of other cell lines. MCF7-Ras human breast cancer cells (31) can be induced to undergo an EMT through the ectopic expression of EMT-inducing transcription factors, such as Slug. Cotreatment of the cells undergoing an EMT with CTx led to a 48-hour delay in the acquisition of mesenchymal morphology and CD44hi cell surface marker expression (fig. S7A). Similarly, the ability of HMLE-Ras cells to undergo an EMT upon ectopic expression of Zeb1 was also hampered upon cotreatment with CTx (fig. S7B). PANC1 pancreatic adenocarcinoma cells undergo an EMT upon treatment with transforming growth factor–β1 (TGF-β1) for 48 hours (32). Cotreatment of PANC1 cells undergoing an EMT with either CTx or Fsk delayed the ability of TGF-β1 to induce an EMT by 48 to 72 hours, which enabled the temporary retention of epithelial properties (fig. S7, C and D).

Treatment with CTx or Fsk induced the acquisition of epithelial properties in a range of cell lines that have mesenchymal traits, including the Hs578T triple-negative breast cancer cell line (fig. S8A), the NCI-H596 lung adenosquamous carcinoma cell line (fig. S8B), and the mesenchymal EpCAMlo CD24lo fraction of the EF021 ovarian carcinoma cell line (fig. S8C). Induction of epithelial properties was also observed in PB3 cells (fig. S8D), which constitute an aggressive cell line isolated from mammary tumors of the genetically engineered MMTV-PyMT transgenic mouse model of breast cancer, in which the expression of the oncogene is driven by the mouse mammary tumor virus promoter (33). Finally, we note that others have recently reported that Fsk promotes the maintenance of an epithelial morphology in primary human mammary epithelial cells, the absence of which led spontaneously to acquisition of mesenchymal attributes, such as down-regulation of E-cadherin expression and up-regulation of mesenchymal markers (34).

Taken together, these data signify the importance of PKA signaling in maintaining epithelial characteristics in a variety of normal and neoplastic epithelial cells. These data give an indication that these responses might be a general property of cAMP-induced activation of PKA in the reversal of phenotypes created by activation of an EMT program.

Although CTx was able to induce entrance of the N8 cells and a range of other cell systems into a stably maintained epithelial state, there were some models in which neither CTx nor Fsk was able to do so, namely, the MDA-MB-231 and SUM159 human breast cancer cell lines, amongst others. These cell lines are maintained in the mesenchymal state through the deletion or stable silencing of several key epithelial genomic loci, which includes the repression of E-cadherin through strong DNA promoter hypermethylation (35). Hence, although the observed effects of PKA activation are applicable to some breast cancer lines and other carcinomas, they are not universal and depend instead on the specific genetic or epigenetic state of the cells.

Essential role of PKA-induced activation of PHF2 in MET

PKA is known to act on many substrates in both the cytoplasm and nucleus (36). Treatment of HMLE and N8 cells with CTx resulted in an immediate increase in the presence of both isoforms of the catalytic subunit in the nucleus (fig. S9, A and B), which suggested that PKA might be regulating nuclear substrates after activation by cAMP. The most well-studied substrate of PKA, CREB1, translocates to the nucleus upon phosphorylation by PKA at Ser133, thereafter altering the transcription of hundreds of target genes (37). In fact, about 300 distinct physiologic stimuli have been described in the literature that can induce CREB Ser-133 phosphorylation (38). It was, therefore, not surprising that CREB was already phosphorylated and present in the nucleus of the N8 cells even before their treatment with either CTx or Fsk (Fig. 3A). Note that knockdown of CREB1 by the use of at least two shRNAs (fig. S5C) did not affect the ability of CTx to induce an MET in the N8 cells (Fig. 3B). Moreover, loss of CREB1 alone induced a partial MET in N8 cells (Fig. 3B), consistent with previous reports of its role in the induction of an EMT (39, 40). On the basis of these observations, we conclude that CREB1 did not play an important role in the PKA-induced MET. We then assessed the localization of Gli1, Gli2, and Gli3, which have been previously reported to be PKA substrates that are retained in the cytoplasm after phosphorylation (41) and found no retention of any of the Gli proteins in the cytoplasm after treatment with CTx or Fsk (fig. S9C). These observations suggest that the Gli proteins may not play a role in the observed PKA-induced MET.

Fig. 3 The PKA substrate PHF2, but not CREB1, is necessary for the MET-inducing properties of CTx.

Activation state of CREB1 as measured by levels of p-CREB1 across HMLE and N8 cells that have been treated with CTx or Fsk (A). Loss of CREB1 through shRNA-mediated knockdown induces a partial MET and permits CTx-mediated complete MET as shown by changes in morphology and immunofluorescence (B). shRNA-mediated knockdown of PHF2 prevents CTx from inducing an MET, which prevents changes in (C) morphology and immunofluorescence-based detection of E-cadherin and fibronectin expression as well as (D) blocking a shift from the CD44hi/CD24lo state to the CD44lo/CD24hi state. Expression of a PHF2 phosphomimetic where the C-terminal serines were modified to aspartate (E) accelerated the MET transition by 5 days, as observed by changes in immunofluorescence (F) and quantitative EMT marker analysis by qPCR (G). Effects of shRNA-mediated knockdown of PHF2 on the ability of HMLE cells to undergo an EMT upon ectopic expression of Zeb1 (H and I) (qPCR data, means ± SD, P < 0.05, n = 3). Immunoprecipitation of PHF2 followed by immunoblotting with a phospho-PKA substrate antibody showed phosphorylation of PHF2 by PKA 24 hours after treatment of N8 cells with CTx (J). (I) was analyzed using the Student’s t test. All scale bars, 25 μm.

Having explored the two most commonly reported nuclear substrates of PKA, we then focused on PHF2, a histone H3 with acetylated lysine 9 (H3K9) histone demethylase, which is known to become activated upon phosphorylation by PKA (42). We found that knockdown of PHF2 expression in N8 cells using either of two shRNAs (fig. S5B) phenocopied the effects of PKA knockdown in that it prevented CTx-induced MET (Fig. 3, C and D). In contrast, knockdown of PHF2 did not alter the ability of HMLE cells to undergo an EMT (Fig. 3, F and G), which indicated that this enzyme, although necessary for induction of an MET, apparently plays no role in the reverse process—the EMT— which suggested that it is specifically important for the derepression of silenced epithelial genes through its function as a H3K9 histone demethylase.

PHF2 can be phosphorylated by PKA at four serine residues in its C terminus (42) (Fig. 3E). Accordingly, we engineered a phospho-mimetic form of PHF2 in which all four of these serines were replaced by aspartate residues. Although expression of this mutant in N8 cells was not sufficient on its own to induce an MET, the phospho-mimetic PHF2 was able to accelerate the rate of CTx-induced transition from the mesenchymal to the epithelial state from 14 days to 7 days (Fig. 3, H and I). Hence, although PHF2 is required for the acquisition of epithelial traits, it appears to be only one of the effectors of the PKA operating during induction of epithelial transition.

To test whether PHF2 can be directly phosphorylated by PKA in our system, we performed an immunoprecipitation of PHF2 followed by immunoblotting using an antibody that recognizes the phospho-PKA substrate motif. As shown in Fig. 3J, 24 hours after treatment of N8 cells with CTx, phosphorylation of PHF2 by PKA can be observed, which provides evidence that PKA phosphorylates PHF2 in the N8 cells. Together, these results suggest an important role for PHF2 as a PKA substrate in the induction of an MET.

PKA-induced activation of PHF2 and the epigenetic reprogramming of mesenchymal cells

The H3K9me2 and H3K9me3 marks have been associated with repression of gene transcription (43). Given the previously reported role of PHF2 as an H3K9me2/3 demethylase, we performed chromatin immunoprecipitation followed by deep sequencing (ChIP-Seq) using antibodies against histone 3 with trimethylated lysine 9 (H3K9me3) and H3K9me2 marks to observe the presence of these marks in untreated N8 cells as well as CTx-treated counterparts in which PHF2 is active. In addition we also performed ChIP-Seq for PHF2, comparing genome-wide occupancy in N8 cells to the N8-CTx cells. We did so in order to monitor PHF2-associated alterations that might enable phenotypic shifts from the mesenchymal to epithelial states, including shifts that might relieve the H3K9-mediated silencing of epithelial genes.

As seen in Fig. 4A, there was a striking inverse correlation at specific loci of the presence of PHF2 with the repressive H3K9me2 or H3K9me3 marks. This suggests that presence of this demethylase may, on it own, suffice to relieve histone-mediated transcriptional silencing. As previously reported, PHF2 appears to occupy the promoter region of genes where it recognizes the H3K4me3 histone mark (Fig. 4B) (9). Interestingly, the total H3K9me3 counts (>4-fold enrichment above control) in N8-CTx cells was almost half of the total counts of the same mark in N8 cells (35,455 versus 18,675). Similarly, the total H3K9me2 counts in N8-CTx cells were also less than a half of that in the N8 cells (1295 vs 473). As shown in the representative circos plots, these data indicate a widespread loss of H3K9-mediated repression of genomic regions upon treatment of N8 cells with CTx and subsequent activation of PHF2 (Fig. 4C).

Fig. 4 Activation of PHF2 leads the epigenetic reprogramming of mesenchymal cells.

Genome-wide occupancy of H3K9me2, H3K9me3, and PHF2 marks shows the inverse correlation between the presence of the histone marks and the demethylase (A), which interacts mainly with the promoter and the first intronic region of genes (B). Circos plots of representative chromosomes 5 and 8 show widespread changes in the H3K9me2 and H3K9me3 profiles (C).

We then sorted for genomic regions present in the N8-CTx but not N8 cells that contained PHF2 binding and lacked repressive H3K9me2/3 marks (table S3). This provided us with a list of genomic regions that were relieved of H3K9me2/3-mediated silencing in the N8-CTx cells, as compared to the N8 cells, owing to PHF2 occupancy. To ensure that these changes were specific for the loss of PHF2, we performed ChIP-Seq for H3K9me2/3 and PHF2 in CTx-treated N8 cells that had an shRNA against PHF2 preventing the MET (table S4). These cells that remained morphologically mesenchymal also demonstrated an epigenetic profile more like that of N8 cells with an overlap of 11,807 peaks compared with the reverted N8-CTx cells, which had an overlap of 6864 peaks. Hence, the list of altered genomic regions outlined in table S3 represents genes that were relieved of H3K9-mediated repression upon CTx-induced activation of PHF2. This suggests that PHF2 activity could be directly responsible for the derepression of these genes that are characteristic of the epithelial cell state. In addition, the expression values of genes that correspond to these genomic loci were also measured in reverted N8-CTx (epithelial) and parental N8 (mesenchymal) cells by RNA-seq, which verified that gain of PHF2 occupancy and loss of H3K9 marks did indeed lead to increased expression (table S3).

Several genes that play a major role in the phenotype and profile of cells in the epithelial state were activated by CTx treatment.The list of genes that were relieved of silencing when treated with CTx includes CDH1 and CDH3 (among other cadherin genes) that code for E-cadherin and P-cadherin (fig. S10A), respectively, which are essential components of adherens junctions and hallmark proteins of basal epithelial cells; KRT8 and KRT18 (fig. S10B), whose gene products are characteristic components of the cytoskeleton of epithelial cells; and AJAP1 and CLDN4 (fig. S10, C and D), which specify genes coding for constituents of adherens and tight junctions that are formed by epithelial but not mesenchymal cells. Other regions include the TP63 gene (fig. S10E), whose product is a hallmark transcription factor of basal mammary epithelial cells, and ITGB2, ITGB6 (fig. S10F), and ITGB8, which code for integrins that are typically expressed on epithelial cells. These observations reveal a mechanism by which activation of this demethylase enables the transcription of genes that induce the acquisition of epithelial properties, ultimately defining the state of the cells.

Activation of PKA and the differentiation of TICs in vivo

We tested the tumor-initiating ability of cells that have been induced to undergo an MET by activation of PKA in vitro. We transplanted at limiting dilutions the neoplastic, RAS-transformed derivatives of HMLE, N8, and the reverted N8-CTx cells, termed HMLE-Ras, N8-Ras, and N8-CTx-Ras, into the mammary fat pads of nonobese diabetic–severe combined immunodeficient (NOD/SCID) mice. As anticipated, the frequency of TICs in the N8-Ras cells was far greater than in the HMLE-Ras cell population, in this case 100-fold as high. Note that the N8-CTx-Ras cells were as inefficient at tumor-initiation as the HMLE-Ras cells (Fig. 5A). The primary tumors that arose upon orthotopic mammary stromal fat pad implantation of N8-Ras tumors spawned 20 to 30 micrometastases in the lungs by 12 weeks after implantation. This property was lost upon induction of an MET by CTx treatment before transplantation (Fig. 5C and fig. S11A), which nevertheless formed primary tumors of comparable size (Fig. 5B). Moreover, this confirmed previous observations that the phenotypic state of these cells before neoplastic transformation strongly influenced their behavior after transformation.

Fig. 5 PKA-induced MET is sufficient to deplete the tumor-initiating ability of N8-Ras cells in vivo.

(A) Differences in tumor-initiating ability of HMLE-Ras, N8-Ras, and N8-CTx-Ras cells upon transplantation with limiting dilution into NOD/SCID mice. Tumors that arose from transplantation of 2 × 106 cells were of similar size (B) with only the N8-Ras cells able to form micrometastases (C). (Each dot represents one mouse; data analyzed using Student’s t test; P < 0.05, n = 10.) (D) Experimental outline to test the tumor-initiating ability of N8-Ras cells upon transient in vivo expression of PKA showing a (E) 20-fold decrease in tumor-initiating ability after secondary transplantation with (F) no significant differences in the tumor volume but a decrease in tumor mass. (Each dot represents one mouse; data analyzed using Student’s t test; P < 0.05, n = 10.)

To better mimic a clinical scenario, we next asked how the induction of an MET would affect preestablished tumors derived from mesenchymal N8 cells. Although we wished to pharmacologically treat mice that already had established N8-Ras tumors, CTx is too toxic to be administered systemically, and the rapid clearance and poor pharmacodynamics of Fsk made it difficult to study its effects upon systemic administration. Such difficulties in treating mice with PKA agonists have also been reported previously (44, 45). For this reason, we focused our efforts on studying the proof-of-principle effects of PKA activation in vivo using the doxycycline-inducible version of constitutively active PKA (caPKA). Thus, we induced expression of the caPKA in already formed N8-Ras tumors of 5-mm diameter (Fig. 5D). On visual inspection, the tumors that had been exposed for 14 days to doxycycline contained pasty, fluid-filled necrotic cores when compared with the tumors that had never been exposed to doxycycline: The latter were solid with a hard center of viable cells. Tumors from mice that received doxycycline weighed less than those that did not receive any (Fig. 5F). Moreover, those tumors in which expression of caPKA had been induced developed a more differentiated histomorphology as revealed by hematoxylin and eosin (H&E) staining of tumor sections (fig. S11, B and C). When tumors were harvested and subjected to fluorescence-activated cell sorting (FACS) analysis, the doxycycline-treated tumors showed a decrease in expression of the CD44 cell surface marker associated with the stemlike population (20), in contrast to the untreated tumors (fig. S11D).

Secondary transplantation of cells isolated from the doxycycline-exposed tumors at limiting dilutions revealed a 20-fold loss of tumor-initiating ability (Fig. 5E), which showed that activation of PKA induces differentiation of TICs and diminishes their ability to subsequently seed new tumors. This result demonstrates that constitutive expression of PKA for a 14-day period in a growing tumor suffices on its own to reduce the tumor-initiating properties of its cells, as indicated by their subsequent inability to propagate upon secondary transplantation.


Cyclic AMP and its main effector, PKA, have been studied for four decades in a variety of cell-biological and physiologic settings, where its effects in activating a number of distinct, tissue-specific traits have been repeatedly documented (11). A role that it might play in governing the epithelial cell state and thus suppressing entrance into the alternative mesenchymal state in breast cancers has not been described. The present work makes it clear that this second messenger and its main effector, PKA, play a key role in determining this epithelial versus mesenchymal balance of mammary epithelial cells, as well as epithelial cells of other tissues. Indeed, in light of these results, it becomes plausible that maintenance of the residence of cells in an epithelial state depends on tonic elevated levels of intracellular cAMP. In retrospect, it now seems likely that the use of cholera toxin as an ingredient in the tissue culture medium of various epithelial cell types [including cells of the epidermis, mammary gland, and bronchus (46, 47)] was motivated by the observation that loss of such cells in culture was accompanied by an overgrowth of fibroblast-like cells (46).

These results collectively indicate a role for PKA in the differentiation of TICs by enforcing residence in the epithelial state and preventing or reversing the EMT program. Although PKA can act via a large number of substrates, we identified PHF2 as an important downstream effector of PKA that mediates the induction of epithelial characteristics through epigenetic reprogramming to a chromatin state that is more favorable for residence in the epithelial state. We find that activating this histone demethylase enables PKA to induce the transcription of genes that play a role in the entrance into and maintenance of residence in the epithelial state.

The EMT program is known to represent one defined route for the generation of both normal and neoplastic epithelial stem cells (6, 7, 48). The observations that PKA-induced activation of PHF2 can either reverse or curtail this program present an opportunity to exploit such a mechanism for therapeutic gain. Indeed, the differentiation of TICs through the induction of an MET is an attractive proposition—one that could be pursued through the induced increase of intracellular cAMP levels, activation of PKA, or activation of PHF2. Nonetheless, it is likely that many such approaches will result in widespread side-effect toxicities, owing to the multitude of signaling pathways that are activated downstream of cAMP increase (11). Specific activation of the PHF2 histone-modifier enzyme may serve as a means of derepressing genes that are essential for the differentiated epithelial state without eliciting many of the toxicities of induced cAMP increases. Along the same lines, identification of a histone methyltransferase that counteracts PHF2 function may also provide an attractive target for therapeutic inhibition, a strategy that has proven successful in the case of DOT1L inhibition against mixed lineage leukemia (MLL)–driven leukemias (49). The role of the G9a histone methyltransferase in establishing the H3K9me2 mark for repression of the CDH1 promoter in breast cancer cells has been reported previously (50).

This study provides mechanistic insight into the benefits of targeting such an enzyme in epithelial tumors, which prevents the constituent cells from undergoing an EMT and thereby acquiring aggressive characteristics, including increases in the numbers of TICs. The pathways explored in this study provide insight into the functions of PKA in the induction of an MET and the differentiation of the more aggressive TICs within a tumor. This study reveals a new direction for targeting the TIC population through epigenetic rewiring that ultimately results in their differentiation and increased susceptibility to conventional chemotherapeutic drugs.

Materials and methods

Cell culture and treatments

HMLE, NAMEC8, and all derived cell lines were grown in Mammary Epithelial Cell Growth Medium medium (Lonza, USA); MCF10A cells were grown in Dulbecco’s minimum essential medium; nutrient mixture F-12 (DMEM/F12) containing 5% horse serum (Sigma, USA; H0146); epidermal growth factor, 20 ng/ml; hydrocortisone, 0.5 mg/ml; CTx, 100 ng/ml; and insulin, 10 μg/ml. EF021 and H596 cells were grown in Roswell Park Memorial Institute (RPMI) medium containing 10% fetal bovine serum. MCF7Ras cells were grown in DMEM containing 10% fetal bovine serum. Hs578T were grown in DMEM containing 10% fetal bovine serum and insulin at 10 μg/ml. Cells were treated with either 100 ng/ml of CTx (Calbiochem, USA; 227036), which was replenished every 2 days, or 1 μM Fsk (Tocris Biosciences, USA; 1099), which was replenished daily over a period of 14 to 16 days. Cells were split to a ratio of 1:6 every 2 to 3 days during the treatments. MCF10A and EF021 cells were a gift from N. Kalaany and R. Drapkin, respectively.


For the CDH1 reporter screen, 500 N8 cells bearing wild-type (wt) CDH1 promoter luciferase were seeded into 384-well plates in a volume of 40 μl. Twenty-four hours later, 100 nl of each compound (200 μM stock) were added using a CyBio liquid handler, which resulted in a final screen concentration of 0.5 μM. Four days later, the plates were read for either firefly luciferase activity (Pierce, 16177) or CellTiter-Glo (Promega, USA; G7572). The Enzo compound library (plate A and plate B; 451 compounds, including repeats) was obtained from the Koch Institute Screening Facility at MIT. Firefly luciferase and CellTiter-Glo assays were performed in triplicates.

The vulnerabilities of the reverted cells were assessed by screening against the Selleck anticancer compound library (400 compounds) and the Enzo kinase library (80 compounds) at the Koch Institute Screening Facility at MIT. N8 or N8-CTx cells (1000 each) were seeded in 384-well plates in a volume of 50 μl. Twenty-four hours later, 50 nl of each compound were added to assay a 5-point dose response. Three days later, the plates were read for CellTiter-Glo, and assays were performed in duplicate.

Flow cytometry

Cells were prepared according to standard protocols and suspended in 2% inactivated fetal bovine serum with phosphate-buffered saline (IFS/PBS). The fluorescent stain 4′,6′-diamidino-2-phenylindole (DAPI) (Life Technologies, USA; D1306) was used to exclude dead cells. Cells were sorted on BD FACSAria SORP and analyzed on BD LSRII, using BD FACSDiva Software (BD Biosciences, USA). Antibodies used were against CD44-PE-Cy7 (Biolegend, USA; 103029); against CD24-PE (BD Biosciences, USA; 555428); against CD45-Pacific Blue (Biolegend, USA; 103125); and against CD31-Pacific Blue (Biolegend, USA; 102421).

Mammosphere and tumorsphere culture

Mammosphere culture was performed as previously described (51). Cells (1000) were seeded per well of a 96-well Corning Ultra-Low attachment plate (Corning, USA; CLS3474) in replicates of 10; sphere numbers were counted between days 8 to 12.

Migration and invasion assays

Cells (25,000) were seeded into 24-well cell culture inserts with 8 μm pores (BD Falcon, USA). After 12 to 24 hours, the cells on the upper surface of the filters were removed with a cotton swab. For visualization, cells on lower filter surfaces were fixed and stained with a Diff-Quick staining kit (Dade Behring/Siemens, Germany). Three to five fields per filter were counted. Data are presented as migrated cells per field.

RNA preparation and polymerase chain reaction analysis

Total RNA was isolated using the RNeasy Plus Mini kit (Qiagen, USA; 74136) and reverse transcription was performed with the High Capacity RNA-to-cDNA kit (Life Technologies, USA; 4387406), both according to the manufacturer’s protocols. A cDNA sample prepared from 1 μg total RNA was used for quantitative reverse transcription polymerase chain reaction (RT-PCR). The PCR reactions were performed with the Fast SYBR Green Master Mix (Life Technologies; 4385612), data collection and data analysis were performed on the ABI7900 machine (Applied Biosystems, USA) by using the SDS2.0 and RQ manager software. The thermal-cycling parameters for the PCR were as follows: 95°C for 5 min, followed by 45 cycles each of 95°C for 10 s, 49°C for 7 s, and 72°C for 25 s. The relative mRNA quantity was normalized against the relative quantity of HPRT1 mRNA in the same sample. The primer sequences in a 5′ to 3′ orientation are shown in table S1.

Immunofluorescence (cultured cells)

Cells were cultured on dishes containing coverslips for 2 to 3 days, after which coverslips were washed in cold PBS, fixed in 4% paraformaldehyde for 10 min at 4°C and permeabilized in 0.2% TritonX in PBS for 2 min. Cells were then washed in PBS, blocked for 1 hour at room temperature in PBS containing 3% normal horse serum (Vector Labs, USA; S-2000). Fixed cells were then incubated with the primary antibody in PBS containing 1% bovine serum albumen (BSA) solution overnight at 4°C. Cells were washed in PBS three times, and secondary antibody was added in PBS containing 1%BSA solution for 1 to 2 hours at room temperature in the dark. Cells were washed three times in PBS and were incubated for 2 min in DAPI solution, after which they were washed in PBS and mounted with a drop of Prolong Gold antifade reagent (Life Technologies, USA; P36961) and placed on coverslips. Slides were viewed on a PerkinElmer Ultraview Spinning Disk Confocal imager and analyzed using Volocity software.

Immunofluorescence (tissue microarrays)

Slides were rehydrated by incubating in Histoclear solution twice for 5 min each, followed by incubation in 100% ethanol twice for 5 min each, in 95% ethanol twice for 5 min each, 70% ethanol twice for 5 min each, once in 35% ethanol for 5 min, and in water for 5 min. Pressure cooker–mediated heat-induced epitope retrieval was carried out in 250 ml of unmasking buffer containing sodium citrate at pH 6. After retrieval, slides were blocked for 30 min in PBS containing 3% normal horse serum after which they were incubated with primary antibody in blocking solution overnight at 4°C. Slides were washed twice with PBS and incubated with secondary antibody at room temperature for 1 hour in the dark. After two PBS washes, 20 μl of mounting medium was added, then slide contents were topped with coverslips, and stored in the dark for 24 hours before imaging. A table of the antigens with source, host, and dilution is shown in table S2.

Proliferation assays

To measure rate of proliferation, 1000 cells were seeded onto a 96-well plate in quadruplicate. Proliferation was measured using CyQuant (Life Technologies, USA; C7026), according to the manufacturer’s protocols.

Protein extraction and Western blotting

To obtain protein extracts, cells were washed with chilled PBS and scraped from culture dishes in aqueous lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 10 mM EDTA pH 8.0, 0.2% sodium azide, 50 mM NaF, and 0.5% NP40) containing complete miniprotease inhibitor cocktail (Roche, USA; 04693159001) and stored at –80°C. The proteins names, sources, and dilutions are shown in table S3. After thawing, they were centrifuged at top speed on a benchtop centrifuge at 4°C for 20 min, and the supernatant was assayed for protein concentration with Bradford Reagent (Bio-Rad; 500-0006). Of the total protein, 30 μg were separated by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) on NuPage gels (Invitrogen, USA) and transferred to Hybond-P polyvinylidene difluoride membrane (GE Healthcare, USA). Membranes were probed with specific primary antibodies and antibody-protein complex detected by horseradish peroxidase (HRP)–conjugated secondary antibodies and SuperSignal West Dura Extended Duration Substrate (Life Technologies, USA; 34075).


N8 cells in 15-cm dishes were scraped and harvested in 0.5 ml of ice-cold immunoprecipitation (IP) buffer (Cell Signaling Technology, USA) containing protease and phosphatase inhibitors (Roche, Germany) after which they were sonicated with three pulses on ice. Lysates were spun down at 14,000g for 10 min, and supernatants were collected for IP. PHF2-specific antibody (Cell Signaling Technology, USA; 3497) was added at 1:25 and incubated overnight at 4°C rotating. The following morning, 40 μl of magnetic protein A/G beads (Thermo Scientific, USA; 26162) were added and incubated for 30 min on a rotator, after which beads were washed five times by using a magnetic separator with cold IP buffer. Beads were then boiled in lithium dodecyl sulfate buffer and run on an SDS-PAGE gel followed by immunoblotting.

RNAi-mediated knockdown

To generate shRNA-expressing plasmids, double-stranded oligonucleotides (oligos) encoding the desired shRNA were cloned into a Tet-pLKO-puro lentiviral vector (Addgene, plasmid 21915). In the absence of doxycycline, shRNA expression is repressed by constitutively expressed TetR protein. With the addition of doxycycline to the growth medium, shRNA expression is triggered, which results in target gene knockdown. The cloning vector has a 1.9-kb stuffer that is released by digestion with Age I and Eco RI. shRNA oligos are cloned into the Age I and Eco RI sites in place of the stuffer. PKA hairpins are shown in tables S4 to S6.

Animal studies

Research involving animals complied with protocols approved by the MIT Committee on Animal Care. For tumor studies, cells suspended in 15 μl 30% Matrigel (GFR)/PBS mix (BD Biosciences; 356230) were injected into the inguinal mammary gland fat pads of age-matched female NOD/SCID mice (Jackson Laboratory). Mice were killed after 10 weeks or when tumors reached a diameter of >1 cm. Lung surface metastases were counted with a fluorescent microscope.

Chromatin IP followed by sequencing

ChIP for PHF2, H3K9me2, and H3K9me3 was carried out using the SimpleChIP Plus Enzymatic Chromatin IP Kit (Cell Signaling Technology, USA; 9005) and the protocols within. The PHF2 rabbit monoclonal antibody (Cell Signaling Technology, USA; 3497) was used at 1:25 per IP; the H3K9me2 mouse monoclonal (Abcam, USA; ab1220); and the H3K9me3 rabbit polyclonal antibodies (Abcam, ab8898) were used at 1:50 (10 μg) per IP. The ChIP DNA was used to prepare libraries for sequencing, which was carried out in the Genome Technology Core at the Whitehead Institute.

Library preparation for sequencing

To prepare libraries for RNA-Seq, the TruSeq stranded mRNA protocol was followed to prep the libraries as described in the kit (Illumina, USA; RS-122-2101) manual. To prepare libraries for the ChIP-Seq, the TruSeq ChIP protocol was followed as described in the kit (Illumina, USA; IP-202-1012) manual.

Deep sequencing and data analysis

Libraries were pooled together and sequenced on the HiSEq 2500 sequencer using the standard sequencing protocols. Images analysis and base calling was done using the Standard Illumina pipeline, and then demultiplexed into FASTQ files. RNASeq paired-end reads from Illumina 1.5 encoding were aligned using TopHat (version 2.0.13) (52) to the human genome (GRCh37) with Ensembl annotation (GRCh37.75) in gtf format. Differential expression was assayed using HTSeq count (53) and DESeq (54). ChIPSeq data were aligned to the human genome (GRCh37) using Bowtie2 (version 2.2.5) (55), base encoding as above, and peaks were called using MACS2 (version (56) with nomodel option, and fragment length was determined by strand cross-correlation (using phantompeakqualtools; Differential binding was determined by using MACS’ bdgdiff tool. Peaks were annotated using Cis-regulatory Element Annotation System (CEAS) (57), and ChIPSeq data profiles were viewed in ngsplot (58). Overlap between peaks, and with expression data, was determined using bedtools (59). ChIPSeq data profiles were viewed in ngsplot (58) and the Integrative Genomics Viewer (60). RNA-seq and ChIP-seq data have been submitted to GEO under the generic stream encapsulation (GSE) ID GSE74883.

LC/MS-based metabolite profiling

LC/MS analyses were conducted on a QExactive benchtop Orbitrap mass spectrometer equipped with an Ion Max source and a HESI II probe, which was coupled to a Dionex UltiMate 3000 high-performance liquid chromatography system (Thermo Fisher Scientific, San Jose, CA). External mass calibration was performed using the standard calibration mixture every 7 days. Polar metabolites were extracted using 1 ml of ice-cold 80% methanol with 10 ng/ml phenylalanine-d8 or phenylalanine-13C9-15N as an internal standard. After 10 min of using a vortex and centrifugation for 10 min at 10,000g, both at 4°C, samples were dried in a centrifugal evaporator. Dried samples were stored at –80°C and then resuspended in 100 μl water; 2.5 μl of each sample was injected onto a ZIC-pHILIC 2.1 × 150 mm (5-μm particle size) column (EMD Millipore). Buffer A was 20 mM ammonium carbonate, 0.1% ammonium hydroxide; buffer B was acetonitrile. The chromatographic gradient was run at a flow rate of 0.150 ml/min as follows: 0 to 20 min, linear gradient from 80% to 20% B; 20 to 20.5 min, linear gradient from 20% to 80% B; 20.5 to 28 min, hold at 80% B. The column oven was held at 25°C. The mass spectrometer was operated with the spray voltage set to 3.0 kV, the heated capillary held at 275°C, and the HESI probe held at 350°C; the sheath gas flow was set to 40 units, the auxiliary gas flow was set to 15 units, and the sweep gas flow was set to 1 unit. To measure cAMP, a positive targeted SIM (tSIM) scan was performed at a resolution of 70,000, an automatic gain control (AGC) target of 1 × 105, and the maximum injection time at 250 ms. The tSIM window was set to a width of 1.0 m/z and centered at 330.05980 m/z, corresponding to the [M+H]+ ion of cAMP. To monitor other endogenous polar metabolites and the internal standard, the tSIM scans were interspersed with positive and negative mode scans in the range of 70 to 1000 m/z, with the resolution set to 70,000, the AGC target at 106, and the maximum injection time at 80 ms. Relative quantitation of polar metabolites was performed with XCalibur QuanBrowser 2.2 (Thermo Fisher Scientific) by using a 5 parts per million mass tolerance and referencing an in-house library of chemical standards.

Statistical analysis

Data are presented as means ± SD. A Student’s t test (two-tailed) was used to compare two groups (P < 0.05 was considered significant) unless otherwise indicated.

Supplementary Materials

Figs. S1 to S11

Tables S1 to S10



Acknowledgments: We thank the Keck Microscopy Facility, Metabolite Profiling Core Facility, Genome Technology Core, Bioinformatics and Research Computing Core, and Proteomics Core Facility at the Whitehead Institute; the Koch Institute Swanson Biotechnology Center (SBC), specifically the Histology Facility and the High-Throughput Screening Facility; J. Benson (SBC) for providing the compound library; D. Bachovchin (Broad Institute) for assistance with automation; and T. DiCesare for assistance with scientific illustration. We thank A. Dongre, S. Iyer, and J. Fröse for technical assistance; all members of the Weinberg laboratory for helpful discussions; and A. Lambert, K. Krishnan, and S. Rajavasireddy for critical reading of the manuscript. D.R.P is supported by a C. J. Martin Overseas Biomedical Fellowship from the National Health and Medical Research Council of Australia (NHMRC APP1071853). W.L.T. is supported by the National Research Foundation, Singapore (NRF-NRFF2015-04); and National Medical Research Council, Singapore (NMRC/TCR/007-NCC/2013). This research was supported by the Ludwig Center for Molecular Oncology at MIT (R.A.W.), Breast Cancer Research Foundation (R.A.W.), Samuel Waxman Cancer Research Foundation (R.A.W.) and NIH R01-CA078461 (R.A.W.). R.A.W. is an American Cancer Society and D. K. Ludwig Foundation Cancer Research Professor. R.A.W. is a shareholder in and the chairman of the Scientific Advisory Board of Verastem, Inc., a company focused on developing drugs to treat cancer by the targeted killing of cancer stem cells. The Whitehead Institute and the authors (R.A.W., D.R.P., B.B., and W.L.T.) have filed a patent application (PCT/US2015/028239) that covers methods of differentiating cancer stem cells through the induction of a MET and the use of these methods for treating cancer. RNA-Seq and ChIP-Seq data from this study have been deposited at GEO under accession number GSE74883.

Stay Connected to Science

Navigate This Article