Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Cancer Res. Author manuscript; available in PMC 2010 July 15.
Published in final edited form as:
PMCID: PMC2720524

Computational identification of a p38SAPK regulated transcription factor network required for tumor cell quiescence


The stress activated kinase p38 plays key roles in tumor suppression and induction of tumor cell dormancy. However, the mechanisms behind these functions remain poorly understood. Using computational tools we identified a transcription factor (TF) network regulated by p38α/β and required for human squamous carcinoma cell quiescence in vivo. We found that p38 transcriptionally regulates a core network of 46 genes that includes 16 TFs. Activation of p38 induced the expression of the TFs p53 and BHLHB3, while inhibiting c-Jun and FoxM1 expression. Further, induction of p53 by p38 was dependent on c-Jun downregulation. Accordingly, while RNAi downregulation of BHLHB3 or p53 interrupted tumor cell quiescence; downregulation of c-Jun or FoxM1 or overexpression of BHLHB3 in malignant cells mimicked the onset of quiescence. Our results identify components of the regulatory mechanisms driving p38-induced cancer cell quiescence. These may regulate dormancy of residual disease that usually precedes the onset of metastasis in many cancers.

Keywords: p38, FoxM1, BHLHB3, c-Jun, p53, quiescence


The stress activated protein kinase p38 is an evolutionarily conserved pathway involved in inflammation, apoptosis and stress signaling (1). In mammals there are four p38 isoforms encoded by different genes (α,β, γ and δ) (1). The best functionally characterized isoforms are p38α and p38β (1). Deletion of p38α is embryonic lethal in mice due to problems with placental development.

Activation of p38 is required for suppression of oncogene-induced transformation and tumorigenesis (1). These functions depend in part on p53, the cellular senescence pathway and the regulation of genes important for G1-S and G2-M checkpoints (13). Several tumors overexpress the p38 phosphatase WIP1, suggesting that downregulation of p38 is important to allow tumor progression (2). Further, conditional deletion of p38α in lungs and livers favors unscheduled proliferation of progenitor cells and tumor formation through enhanced JNK and c-Jun signaling in mice (4, 5). Although increased p38 activity may be advantageous for tumor cells in highly advanced cancers (6), in some instances, tumors cells may still be susceptible to the negative regulation of p38 (711). For instance, MKK4 mediates suppression of metastases in ovarian cancer via activation of p38 (10).

Metastatic squamous carcinoma HEp3 cells (T-HEp3) adapted to the in vitro microenvironment for >40 generations, re-program and lose their tumorigenic and metastatic potential in vivo (9, 12). Loss of malignancy is not due to enhanced apoptosis, but due to a G0-G1 arrest that leads to the acquisition of a dormant phenotype upon re-inoculation in vivo (13). This phenotype is also not due to selection of pre-existent “dormancy predisposed” clones as it happens in multiple clones isolated from HEp3 tumors at almost 100% cloning efficiency (12). Furthermore, the dormant phenotype is transient because after ~11 weeks in vivo all individual clones or polyclonal tumor cell populations resumed growth (12). Cells from these tumors when placed in vitro can once more adopt a dormant behavior, suggesting that epigenetic mechanisms might drive this reversible dormancy behavior (12).

We hypothesized that if tumor cell dormancy does not result from the appearance of rare clones, then an epigenetic reprogramming might be responsible for the phenotypic shift. Previous studies revealed that loss of malignancy of HEp3 cells is due to the activation of p38, which inhibits ERK signaling, causing a downregulation of the urokinase receptor (uPAR) and reduced trans-activation of EGFR (13). Genetic or pharmacologic inhibition of p38 was sufficient to restore ERK activation, uPAR expression and proliferation of these cells in vivo (8, 9). In addition to HEp3 cells, activation of p38 was also predictive of proliferation vs. growth arrest in breast, prostate and fibrosarcoma cancer cell lines (7). Further, independent studies showed that in malignant cervical carcinoma HeLa cells, p38 activation suppresses their tumorigenicity (11). However, the global changes in gene expression responsible for such phenotypic outcomes were unknown.

Here we have combined gene expression profiling with a powerful bioinformatics analysis to provide a systems view of p38 and the transcription factor (TF) network signaling it regulates in dormant HEp3 cells. We identified a putative TF network using data from conventional gene array platforms and through in vivo validation and found that it was predictive of gene function in dormant HEp3 cells. Using this approach we discovered that p38-induced tumor cell quiescence is controlled in part by the regulation of TFs including R213Q mutant p53, BHLHB3, Jun and FoxM1. These results provide insight into the mechanisms by which stress signaling might reprogram tumor cells to acquire a quiescence program and may help identify genetic determinants of cancer cell dormancy.

Materials and Methods

Reagents, siRNAs and antibodies

SB203580, SB202190 and PD98059 were from Calbiochem (Beverly, MA). Doxorubicin and DMSO were from Sigma (St. Louis, MO). Rabbit polyclonal anti-p53 and anti-Bax antibodies were from Cell Signaling (Danvers, MA); rabbit polyclonal anti c-Jun (H-79) and FoxM1 (C-20) and mouse monoclonal anti phospho-Erk (E-4) were from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-Erk, anti-HA, anti-V5, anti phospho-p38 and anti p38α monoclonal antibodies were from BD Biosciences (San Jose, CA). Horseradish peroxidase (HRP)–conjugated anti-mouse IgG and anti-rabbit IgG antibodies were from Vector Laboratories (Burlingame, CA) and Chemicon International (Temecula, CA) respectively. siRNAs to p38α were from Ambion (p38α-1, Austin, TX) and New England Biolabs (p38α-2, Ipswich, MA); siRNAs to p38β and DEC2 were from Santa Cruz Biotechnology (Santa Cruz, CA); siRNA to c-Jun and custom FoxM1 siRNA (sense sequence: CCUUUCCCUGCACGACAUGUU) were from Dharmacon (Chicago, IL); control and GAPDH siRNAs were from Ambion (Austin, TX).

Cell lines and Xenograft studies

Tumorigenic (T-HEp3) and "spontaneous" dormant (D-HEp3) human epidermoid carcinoma HEp3 cell (8), D-HEp3-neo, and D-HEp3-p38 DN cell lines were described previously (9). Tumor growth on chick embryo CAMs or Balb/c nude mice has been described previously (8). All animal experiments were approved by IACUC (SUNY-Albany and MSSM).

Microarray Analysis and TF Network Mapping

A total of 20 Affymetrix Hgu133a gene chips were run with the following samples: 4 SB203580 treatments (5 µM, 48h) and 4 DMSO controls; 3 SB202190 treatments (5 µM, 48h) and 3 DMSO controls; 3 DNp38α expressing D-HEp3 cells and 3 empty vector (Neo) expressing D-HEp3 cells. Raw CEL file data from GCOS was imported into a Bioconductor session using the Affymetrix package (S-Fig 1). The batch was background-corrected, normalized, and summarized using both gcrma and mas5 functions in parallel. Linear models of differential expression were fit to each of the transformed datasets and contrasts were estimated between each group of p38-inhibited samples and its corresponding set of controls. Probability values of the F-statistic were used to gauge the likelihood a gene was significantly differentially expressed (S-Fig 1). The distributions of the F- statistic P value as determined by the classify TestsF function of the limma package (8) were examined and a cutoff of significance was established at the inflection point of 0.05 occurring in both transformed sets. Genes significant in both MAS5 and gcRMA sets at this level were marked as a low-stringency group. The response of each gene in each contrast (SB203580, SB202190, and DNp38α) was then calculated using the same p-value of 0.05 for a cutoff and a response matrix with values −1 (repressed by p38), 0 (no change), and +1 (induced by p38) was created. Response matrices for gcrma and mas5 datasets were merged to show only common responses. Genes significantly changing here in all three treatments were marked as a high-stringency group. Tables of differentially expressed genes in each list were generated (S-Fig 1).

TF Network Mapping

A database of known and predicted interactions between transcription factors and target gene promoters was developed using the TRANSFAC and LocusLink as described in Tuck et al. (8) and mapped to Affymetrix identifiers using Bioconductor's hgu133a package. For each combination of a transcription factor T and a gene G present on the array: first, the scores for all known or potential binding sites for T in the promoter of G were summed up as the co-regulation score; next, the non-parametric Spearman correlation between T and G across all samples was taken as the co-expression score; and finally, the co-regulation and co-expression scores were multiplied together to obtain a combined score whose magnitude reflects the strength of both the co-regulation and co-expression scores and whose sign reflects correlation or anti-correlation (and by extension transcriptional induction or repression) (S-Fig 1). This inferred transcriptional regulation network was examined across different thresholds of the combined score for both the high and low stringency responder gene sets and including or excluding transcription factors that were not in the responsive gene sets. Networks that proved unmanageable for a given threshold were further filtered to include only TFs.

RNA interference studies

Retroviral delivery of shRNAs targeting the genes of interest (Open Biosystems, Huntsville, AL) or, firefly luciferase or an empty vector, as control and the development of stable cell lines was done as previously described (14). Transfections of siRNAs targeting the desired sequences or, as controls, a siRNA targeting GAPDH or scrambled siRNA, were performed as previously described (14). Cells were either used for in vivo experiments or lysed 24–72 h later for immunobloting and/or qPCR.

Immunoblotting, RT-PCR and quantitative PCR


Luciferase reporter assays

In vitro and in vivo dual luciferase assays were performed using the Dual Luciferase Reporter Kit (Promega, Madison, WI) following the vendor’s instructions as described previously (7).

Statistical analysis

In vivo assays were evaluated by the non-parametric tests Mann Whitney or Kruskal-Wallis followed by Dunn’s multiple comparison test. Luciferase assays were evaluated by ANOVA followed by Bonferroni multiple comparison tests. A p value of less than 0.05 was considered statistically significant.


Identification of a putative TF network embedded in the p38SAPK-regulated gene expression profile

We showed that dormant D-HEp3 cells have higher P-p38 and P-Hsp27 levels than tumorigenic T-HEp3 cells (S-Fig2). Further, pharmacologic or genetic inhibition of p38α/β signaling with SB203580 or SB202190 (S-Fig2) or dominant negative p38α (DNp38) caused a reversion of quiescence in vivo (79). Similar results were obtained with two different si or shRNAs targeting p38α (S-Fig2). Further, activation of p38 signaling in T-HEp3 cells using an MMK6 active mutant inhibited T-HEp3 cell proliferation in vivo, mimicking the induction of dormancy (S-Fig2). Of note is the fact that ~75% of the genes that p38 induced in D-HEp3 are also induced in D-HEp3 vs. T-HEp3 cells and 97% of the genes down-regulated by p38 in D-HEp3 cells were lower in expression levels in D-HEp3 than in T-HEp3 cells (data not shown). This suggests that majority of the genes induced or repressed by p38 are also similarly regulated between T-HEp3 and D-HEp3 cells. Thus, we used this model of forced inhibition of p38 to reveal the gene program driving D-HEp3 tumor cell quiescence in vivo.

We compared using Affymetrix gene array profiling the gene expression changes of cells treated with or without p38 pharmacological inhibitors SB203580 or SB202190 (5µM) for 48 hrs in the absence of FBS or a genetic approach comparing D-HEp3 cells stably expressing an empty neomycin resistant vector (Neo) or a DNp38 (9) (Fig1A). The genes changing in expression in these treatments were grouped into two sets: a low-stringency set, which changed significantly in any of the three treatments (15); and a high-stringency set in which gene expression changed significantly in all of the three treatments.

Figure 1
A Venn diagrams of genes changing significantly (91 up, 133 down) to each of the three strategies using both the gcRMA and MAS5 methods. Significance was established with a p value of the F-statistic at 0.05. Total number of genes induced or repressed ...

So that we could identify TFs responsible for the p38-regulated growth arrest program, we looked in both the high and low stringency sets for TFs whose activities might explain the observed expression changes (16). For this we used a database in which associations between TFs and their target gene promoters are represented numerically with interaction scores, (Fig1 and S-Fig1). For each combination of a TF, T, and gene, G, present on the array, we asked the following questions. First, were there known or potential binding sites for T in the promoter of G? If so, we summed the scores for these to use as a “co-regulation score” (S-Fig1). Next, we asked if the expression change of T correlated with that of G. We took the non-parametric Spearman correlation between T and G across all samples as our “co-expression score” (S-Fig1). Finally, we multiplied the co-regulation and co-expression scores together to obtain a combined score. The magnitude of the combined score reflects the strength of the scores for both co-regulation (i.e. if there are no binding sites for T upstream of G, the co-regulation score and the product overall score will be zero) and co-expression (i.e. low correlations will be weighted proportionally lower). The sign of the combined score reflects correlation or anti-correlation, which by extension corresponds to potential transcriptional induction or repression (see Methods and S-Fig1). We used a combined score threshold of 0.75, restricting genes to the high-stringency set, and restricting TFs to the low-stringency set to keep the networks manageable.

The low-stringency set (S-Fig3) revealed an association network containing 129 genes. Of these 43 represent known TFs (boxes), the remainder correspond to other non-TF genes (circles) (Fig1B and S-Fig3). In the high-stringency set (Benjamini-Hochberg corrected test p<0.05 in all treatments), 16 TFs were found to be consistently changing in expression (up, red; down, green) and correlation levels across the three strategies. Some of these (e.g., c-Jun, FoxM1, BHLHB3 and NR2F1, see below) appear to be connected to a larger number of target genes (Fig1B and S-Fig3). We found that the majority of the genes regulated by p38 were similarly regulated between T-HEp3 and D-HEp3 cells. Thus, the p38-regulated genes are also changing during the reprogramming of tumorigenic to dormant phenotypes of HEp3 cells. Using the high-stringency set we obtained a TF network amenable for experimental testing (Fig1). Nodes in this network included c-Jun and FoxM1, as genes negatively regulated by p38 signaling. In contrast, p38-induced the tumor suppressor p53, the transcriptional repressors Nr2f1 (17) and the basic helix loop helix TF BHLHB3/Dec2/Sharp-1) (18), two genes of relatively unknown function in cancer and not previously linked to p38. Next we determined whether these findings were predictive of TF function in p38-induced quiescence of D-HEp3 cells. We selected four TFs, BHLHB3, c-Jun, p53 and FOXM1 for our initial analysis (see below), either because they were known targets of p38 (e.g., p53 and c-Jun) or because they were new targets not previously implicated in p38-mediated inhibition of tumor cell proliferation.

BHLHB3 is required for D-HEp3 cell quiescence and regulated by ERK and p38

BHLHB3 is a clock gene (19, 20), a transcriptional repressor (18) and importantly a tumor suppressor in lung cancer (21). In D-HEp3 cells BHLHB3 is induced by p38 and was positively linked to genes involved in extracellular matrix biology (e.g., collagen-I α1, MMP2) (Fig1B). Its induction by p38 suggested a role in tumor cell quiescence. T-HEp3 cells express less BHLHB3 mRNA than D-HEp3 cells and inhibition of p38 by SB203580 reduced BHLHB3 expression in D-HEp3 cells (Fig2A). Inhibition of Mek1/2 with PD98059 upregulated BHLHB3 mRNA expression in D-HEp3 cells but to a lesser degree in T-HEp3 cells (Fig2A). We conclude that ERK and p38 signaling results in opposing regulation of BHLHB3 expression.

Figure 2
BHLHB3 is required for D-HEp3 cell quiescence A, basal expression of BHLHB3 detected by RT-PCR (left panel) or (right panel) qPCR under the indicated treatments for 24hrs. PD = 20 µM PD98059 Mek1/2 inhibitor, SB=10 µM SB203580 p38 inhibitor; ...

In D-HEp3 cells siRNA targeting BHLHB3 resulted in a slightly stronger downregulation of BHLHB3 to that obtained by a p38α siRNA as measured by RT-PCR and qPCR (Fig2B). SiRNA to p53 did not affect BHLHB3 expression (Fig2B), suggesting that its expression is not downstream of p53. Upon inoculation in vivo, we observed that like p38α knock-down, downregulation of BHLHB3 also restored proliferation of D-HEp3 cells in vivo (Fig2B). We also tested D-HEp3 stably transduced with a different shRNA to BHLHB3 or a shRNA to luciferase or an empty vector (Fig2C). Upon in vivo inoculation D-HEp3 cells expressing a shRNA to BHLHB3 contained >6 fold more cells per tumor nodule than the tumors from control cells (Fig2C). The same cells inoculated in nude mice showed a statistically significant shortening in the dormancy period for D-HEp3 cells with BHLHB3 knock down (Fig2C). In agreement, over-expression of BHLHB3 in T-HEp3 was able to significantly inhibit their proliferation (Fig2D). This effect could be further enhanced by co-expression of an MKK6 active mutant that activates p38α (Fig2D) (7). We conclude that BHLHB3 is a novel negative regulator of tumor growth functionally linked to p38 signaling.

Downregulation of FoxM1 or c-Jun inhibits T-HEp3 growth in vivo

We next tested whether FoxM1 and c-Jun down-regulation by p38 was linked to D-HEp3 cell dormancy (Fig 1). QPCR analysis revealed that c-Jun expression was higher in T-HEp3 than D-HEp3 cells (Fig3A). Further, p38 inhibition resulted in c-Jun upregulation only in D-HEp3 cells while Mek1/2 inhibition downregulated c-Jun expression in both cells (Fig3A). SiRNAs inhibition of c-Jun expression (Fig3B) did not affect BHLHB3 and FoxM1 mRNA levels (data not shown) suggesting that these genes are not downstream of c-Jun. However, c-Jun downregulation by RNAi (si c-Jun-I) stimulated p53 expression in T-HEp3 cells and to some extent in D-HEp3 cells (Fig3B). Thus, our data support that p38-mediated upregulation of p53 transcript is at least in part mediated by a p38-dependent inhibition of c-Jun induction. In addition, RNAi to c-Jun was sufficient to cause a strong inhibition of T-HEp3 proliferation in vivo (Fig3B). This correlated with reduced proliferation marker phospho-Histone H3 staining in siRNA treated tumors, but not with increased apoptosis as measured by cleaved-caspase3 staining (Fig3C).

Figure 3
A, Basal c-Jun expression is lower in D-HEp3 cells than in T-HEp3 cells, increased by 10 µM SB203580 (SB) treatment in D-HEp3 cells and reduced by 20 µM PD98059 (PD) in T-HEp3 cells as measured by RT-PCR (upper panel) and qPCR (middle ...

FoxM1 expression was repressed by p38 as SB203580 treatment upregulated FoxM1 mRNA in D-HEp3 cells (Fig3D). RNAi-mediated downregulation of FoxM1 was detected by RT- and qPCR (Fig3D) and was also very efficient in inhibiting T-HEp3 tumor growth in vivo (Fig3D). Reduced tumor growth was also associated with reduced phospho-H3 levels in these tumors. However, inhibition of FoxM1 expression induced apoptosis by ~2 fold in T-HEp3 cells (Fig3C). We conclude that both FoxM1 and c-Jun are important to promote T-HEp3 tumor growth (and survival in the case of FoxM1).

Opposing regulation of p53mutR213Q by ERK and p38 contributes to D-HEp3 tumor cell quiescence

In D-HEp3 cells p38 induced p53 mRNA (Fig 1), which is in agreement with lower p53 mRNA level in tumor vs. normal head and neck cancer tissue (22) (S-Fig4). Furthermore, in ~50% of head and neck squamous cell carcinomas (HNSCC) p53 is mutated (23). Indeed we found in both T- and D-HEp3 cells a mutation comprising an Arg to Gln substitution at codon 213 (R213Q) (Fig4), which is present in oral cancers (24). Sequencing data in S-Fig5 showed that the CGA to CAA G->A substitution appears to be complete suggesting that both alleles are mutated. Although mutations or post-translational modifications can inactivate p53, we tested whether p53mutR213Q was functional in HEp3 cells and whether transcriptional downregulation of p53mutR213Q may be an additional mechanism to overcome p53 inhibitory effects.

Figure 4
A, protein sequence alignment along aa 201–250 between wt (NM_000546) and mutant p53 in HEp3 cells shows an R213Q substitution (upper panel). Basal p53 mRNA (middle panel) and protein (lower panel) expression is higher in D-HEp3 cells than in ...

As shown in Fig4, p53 mRNA and protein levels were upregulated in D-HEp3 vs. T-HEp3 cells. Basal luciferase reporter expression driven by a p53-binding element was 6–10-fold higher in D-HEp3 than T-HEp3 cells (S-Fig5), suggesting a functional p53 protein despite the R213Q mutation. SB203580 (10µM) treatment, which almost completely reduces Hsp27 phosphorylation by p38α/β although p38γ and p38δ are also expressed (S-Fig5), inhibited p53 mRNA (48hrs), protein expression and activity (Fig4 and S-Fig5) in D-HEp3 cells. Also, Mek1/2 inhibition caused a dramatic upregulation in p53 mRNA and activity in both cells (Fig4 and S-Fig5). In addition, phosphorylation of p53 at Ser15 after modulation of ERK or p38 activity and protein turnover following proteasome inhibition were unaffected in T-HEp3 and D-HEp3 cells (data not shown). Endogenous p53 protein level and activity could be induced by doxorubicin in T-HEp3 and D-HEp3 cells albeit to a lesser extent in the latter (S-Fig5). However, the downstream target Bax was induced poorly in T-HEp3 cells and did not change in D-HEp3 cells, which have high basal p53 and Bax expression (S-Fig5). Marginal Bax induction is in agreement with the abrogated capacity of the R123Q p53 mutant to induce apoptotic genes (24). Finally, p53 activity was significantly higher in D-HEp3 cells compared to T-HEp3 cells after three days in vivo. This suggests that higher p53mutR213Q expression and activity in D-HEp3 cells, persists upon in vivo growth arrest (Fig4) and that this mutant p53 may allow HNSCC cells to uncouple growth arrest from apoptosis in response to stress.

We next tested whether p53 is required for p38-induced quiescence via RNAi. The shRNAs to p53 showed a consistent and specific downregulation of p53 protein of ~50–60% (Fig4C) compared to a shRNA to luciferase. SiRNAs to p53 targeting a different region of the mRNA generated an almost complete reduction in mRNA and protein after 48 hrs by qPCR and Western blot, respectively (Fig4D). In all instances downregulation of p53 was sufficient to allow these cells to resume proliferation in vivo (Fig4C&D). These results further support the predictive value of the TF network and suggest that p53mutR213Q is in part required for p38-induced quiescence in D-HEp3 cells.


We have identified a gene expression program responsible for p38-induced tumor cell quiescence and the TFs executing this program by using computational tools that (i) revealed TF networks, (ii) were predictive of gene function and (iii) were of a complexity level addressable experimentally. Our analysis identified p38 co-regulated targets were functionally linked to the quiescent phenotype of D-HEp3 cells. Further, we revealed clues about the activating or repressing functions of several TFs (present in TRANSFAC) on numerous gene promoters.

Our analysis identified TFs potentially responsible for the quiescent phenotype of D-HEp3 cells that could be validated in vivo. For example, inhibiting p53 and BHLHB3 resulted in a similar phenotype to that of p38 inhibition. BHLHB3 and p53 appear to have similar abilities to suppress growth despite the apparent difference in the number of predicted downstream targets in the network. This also suggests that the number of connected genes may not directly assign phenotypic importance to a TF. It was noticeable that for c-Jun and FoxM1 or Nr2F1 and BHLHB3, although clearly co-regulated by p38, there was no single common regulating TF. Further, identification of p38 target genes in vivo where D-HEp3 cells fully growth arrest (25) and/or temporal studies will help identify additional p53 targets and upstream regulators of these TFs.

We found that p38-induced D-HEp3 cell quiescence required downregulation of c-Jun, which is consistent with its role in G1-S transition (26) and the upregulation in tumor vs. normal HNSCC tissue (S-Fig4B). Accordingly, recent data showed that p38α deficient mouse fetal livers strongly upregulated c-Jun and resulted in accelerated liver cancer development (5). Similar results were observed in mice lacking p38α in the lung (4). We found that the upregulation of p53 by p38 required downregulation of c-Jun, as its knockdown was sufficient to induce p53 and inhibit T-HEp3 tumorigenesis (Fig5B). The antagonistic effect of c-Jun on p53 was shown to be linked to p38α signaling in liver regeneration (27) and mouse liver cancer development (28). Further, c-Jun −/− MEFs have a proliferative defect mostly attributed to the transcriptional upregulation of p53, which was dependent on c-Jun mediated transcriptional repression from PF-1 site on p53 promoter (29). Similarly, blocking FoxM1 in T-HEp3 cells inhibited their growth in vivo. Still this was linked to reduced proliferation and to a 2-fold increase in apoptosis, suggesting that in D-HEp3 cells, downregulation of FoxM1 and the potential loss of prosurvival signaling might be compensated by other recently described survival pathways (30). FoxM1 is a known regulator of G1-S and G2-M transition located at the 12p13 locus, usually amplified in HNSCC(31). Further, the Oncomine database shows that FoxM1 transcript is up-regulated in HNSCC vs. normal tissues (32) (S-Fig4C) and if expressed at high levels in primary breast tumors it might be a poor prognosis indicator (33, 34). Our results suggest that reprogramming by p38 of D-HEp3 cells requires decreased FoxM1 and c-Jun expression to enter prolonged G0-G1 arrest in vivo.

Transcriptional regulation of p53 was important for p38-induced D-HEp3 cell quiescence. This was interesting, since T- and D-HEp3 cells carry an R213Q mutation in p53. We show that p53mutR213Q can still induce quiescence and that it is regulated by p38 (1, 35). Our data diverged from other studies on post-translational regulation of p53 (1, 35), in that ERK and p38 had opposing roles in the regulation of p53 at the transcript level. The Oncomine data supports that transcriptional downregulation of p53 occurs in patients (22) (S-Fig4A). This suggests that spontaneous reprogramming and quiescence in HNSCC might occur upon upregulation of p53 even in patients displaying the p53mutR213Q. Thus, this mutation might predict for dormant disease as it may eliminate p53 pro-apoptotic functions (24) but allows tumor cells to enter quiescence. This may be due to the inability of codon 213 mutants to bind the p53 binding proteins ASPP1/2, which stimulate p53 to trans-activate apoptosis but not growth arrest genes (36, 37).

Another required component that appears to operate in parallel to p53 is BHLHB3. This transcriptional repressor is very important for p38-induced D-HEp3 cell quiescence and we are currently dissecting the mechanisms by which it induces tumor cell dormancy. BHLHB3 is expressed at higher levels in tumor vs. normal HNSCC tissues (S-Fig4D), suggesting that in the primary tumor BHLHB3 expression may vary between different cancer types. However, the Oncomine database contains studies showing that high expression of BHLHB3 in primary breast cancer tumors is a good prognosis indicator (33, 34). In agreement, while this manuscript was under review, Adorno et al.,(38) showed that BHLHB3 is a TGFβ target gene downregulated in breast carcinoma cells that bear a p53R175H mutation that blocks p63 function. Consequently they found that BHLHB3 serves as a metastasis suppressor and that its downregulation was associated with poor prognosis in breast cancer (38). However, a mechanism was not described. We speculate that induction of quiescence and consequently metastatic dormancy dependent on p38 and BHLHB3, both targets of TGFβ signaling (3840), may be responsible for the metastasis suppression observed by Adorno et al. (38). We also showed that the ERK/p38 ratio is predictive of dormancy/quiescence of fibrosarcoma, breast, prostate and squamous carcinoma cell lines (79) and similar results were observed with cervical carcinoma HeLa cells (11). Thus, collectively our findings in this model of HNSCC quiescence might reveal important mechanisms driving dormancy and subsequent metastatic growth in different types of cancer

Our studies reveal a previously unrecognized network of TFs regulated by p38 and required for the induction of tumor cell quiescence. Understanding how these TFs contribute to p38 induced growth arrest is of importance because inhibiting and/or reactivating more than one gene may be required to inhibit tumor progression. The strategy presented here may help identify new therapeutic targets or prognostic indicators when applied to large array datasets from patient samples.

Supplementary Material



We acknowledge the help of Drs. Thomas Begley (SUNY-Albany), Ari Melnick (Cornell University) and Yang Zheng (MSSM). This work is supported by grants from the NIH/National Cancer Institute grant CA109182 (J.A.A-G), the U.S. Army Medical Research Acquisition Activity W8IWXH-04-1-0474 (D.S.C) and by the Samuel Waxman Cancer Research Foundation Tumor Dormancy Program (J.A.A-G and D.S.C). A.C.R. is a recipient of a Ruth L. Kirschstein National Research Service Award (NIH/NCI) Fellowship. D. M. S. is a recipient of a Dr. Mildred-Scheel postdoctoral grant by the Deutsche Krebshilfe.


1. Bulavin DV, Fornace AJ., Jr p38 MAP kinase's emerging role as a tumor suppressor. Adv Cancer Res. 2004;92:95–118. [PubMed]
2. Bulavin DV, Demidov ON, Saito S, et al. Amplification of PPM1D in human tumors abrogates p53 tumor-suppressor activity. Nat Genet. 2002;31:210–215. [PubMed]
3. Bulavin DV, Phillips C, Nannenga B, et al. Inactivation of the Wip1 phosphatase inhibits mammary tumorigenesis through p38 MAPK-mediated activation of the p16(Ink4a)-p19(Arf) pathway. Nat Genet. 2004;36:343–350. [PubMed]
4. Ventura JJ, Tenbaum S, Perdiguero E, et al. p38alpha MAP kinase is essential in lung stem and progenitor cell proliferation and differentiation. Nat Genet. 2007;39:750–758. [PubMed]
5. Hui L, Bakiri L, Mairhorfer A, et al. p38alpha suppresses normal and cancer cell proliferation by antagonizing the JNK-c-Jun pathway. Nat Genet. 2007;39:741–749. [PubMed]
6. Junttila MR, Ala-Aho R, Jokilehto T, et al. p38alpha and p38delta mitogen-activated protein kinase isoforms regulate invasion and growth of head and neck squamous carcinoma cells. Oncogene. 2007;26:5267–5279. [PubMed]
7. Aguirre-Ghiso JA, Estrada Y, Liu D, Ossowski L. ERK(MAPK) activity as a determinant of tumor growth and dormancy; regulation by p38(SAPK) Cancer Res. 2003;63:1684–1695. [PubMed]
8. Aguirre-Ghiso JA, Ossowski L, Rosenbaum SK. Green fluorescent protein tagging of extracellular signal-regulated kinase and p38 pathways reveals novel dynamics of pathway activation during primary and metastatic growth. Cancer Res. 2004;64:7336–7345. [PubMed]
9. Aguirre-Ghiso JA, Liu D, Mignatti A, Kovalski K, Ossowski L. Urokinase receptor and fibronectin regulate the ERK(MAPK) to p38(MAPK) activity ratios that determine carcinoma cell proliferation or dormancy in vivo. Mol Biol Cell. 2001;12:863–879. [PMC free article] [PubMed]
10. Hickson JA, Huo D, Vander GriendDJ, et al. The p38 kinases MKK4 and MKK6 suppress metastatic colonization in human ovarian carcinoma. Cancer Res. 2006;66:2264–2270. [PubMed]
11. Timofeev O, Lee TY, Bulavin DV. A subtle change in p38 MAPK activity is sufficient to suppress in vivo tumorigenesis. Cell Cycle. 2005;4:118–120. [PubMed]
12. Ossowski L, Reich E. Changes in malignant phenotype of a human carcinoma conditioned by growth environment. Cell. 1983;33:323–333. [PubMed]
13. Liu D, Aguirre GhisoJ, Estrada Y, Ossowski L. EGFR is a transducer of the urokinase receptor initiated signal that is required for in vivo growth of a human carcinoma. Cancer Cell. 2002;1:445–457. [PubMed]
14. Ranganathan AC, Ojha S, Kourtidis A, Conklin DS, Aguirre-Ghiso JA. Dual function of pancreatic endoplasmic reticulum kinase in tumor cell growth arrest and survival. Cancer Res. 2008;68:3260–3268. [PMC free article] [PubMed]
15. Smyth GK. Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol. 2004;3 Article3. [PubMed]
16. Tuck DP, Kluger HM, Kluger Y. Characterizing disease states from topological properties of transcriptional regulatory networks. BMC Bioinformatics. 2006;7:236. [PMC free article] [PubMed]
17. Ferrer-Martinez A, Marotta M, Baldan A, Haro D, Gomez-Foix AM. Chicken ovalbumin upstream promoter-transcription factor I represses the transcriptional activity of the human muscle glycogen phosphorylase promoter in C2C12 cells. Biochim Biophys Acta. 2004;1678:157–162. [PubMed]
18. Azmi S, Ozog A, Taneja R. Sharp-1/DEC2 inhibits skeletal muscle differentiation through repression of myogenic transcription factors. J Biol Chem. 2004;279:52643–52652. [PubMed]
19. Kawamoto T, Noshiro M, Furukawa M, et al. Effects of fasting and re-feeding on the expression of Dec1, Per1, and other clock-related genes. J Biochem (Tokyo) 2006;140:401–408. [PubMed]
20. Rohleder N, Langer C, Maus C, et al. Influence of photoperiodic history on clock genes and the circadian pacemaker in the rat retina. Eur J Neurosci. 2006;23:105–111. [PubMed]
21. Falvella FS, Colombo F, Spinola M, et al. BHLHB3: a candidate tumor suppressor in lung cancer. Oncogene. 2008;27:3761–3764. [PubMed]
22. Toruner GA, Ulger C, Alkan M, et al. Association between gene expression profile and tumor invasion in oral squamous cell carcinoma. Cancer Genet Cytogenet. 2004;154:27–35. [PubMed]
23. Gath HJ, Brakenhoff RH. Minimal residual disease in head and neck cancer. Cancer Metastasis Rev. 1999;18:109–126. [PubMed]
24. Pan Y, Haines DS. Identification of a tumor-derived p53 mutant with novel transactivating selectivity. Oncogene. 2000;19:3095–3100. [PubMed]
25. Aguirre GhisoJA, Kovalski K, Ossowski L. Tumor dormancy induced by downregulation of urokinase receptor in human carcinoma involves integrin and MAPK signaling. J Cell Biol. 1999;147:89–104. [PMC free article] [PubMed]
26. Nicolaides NC, Correa I, Casadevall C, et al. The Jun family members, c-Jun and JunD, transactivate the human c-myb promoter via an Ap1-like element. J Biol Chem. 1992;267:19665–19672. [PubMed]
27. Stepniak E, Ricci R, Eferl R, et al. c-Jun/AP-1 controls liver regeneration by repressing p53/p21 and p38 MAPK activity. Genes Dev. 2006;20:2306–2314. [PubMed]
28. Eferl R, Ricci R, Kenner L, Liver tumordevelopment, et al. c-Jun antagonizes the proapoptotic activity of p53. Cell. 2003;112:181–192. [PubMed]
29. Schreiber M, Kolbus A, Piu F, et al. Control of cell cycle progression by c-Jun is p53 dependent. Genes Dev. 1999;13:607–619. [PubMed]
30. Schewe DM, Aguirre-Ghiso JA. ATF6alpha-Rheb-mTOR signaling promotes survival of dormant tumor cells in vivo. Proc Natl Acad Sci U S A. 2008;105:10519–10524. [PubMed]
31. Laoukili J, Stahl M, Medema RH. FoxM1: at the crossroads of ageing and cancer. Biochim Biophys Acta. 2007;1775:92–102. [PubMed]
32. Rhodes DR, Yu J, Shanker K, et al. ONCOMINE: a cancer microarray database and integrated data-mining platform. Neoplasia. 2004;6:1–6. [PMC free article] [PubMed]
33. van't Veer LJ, Dai H, van de Vijver MJ, et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature. 2002;415:530–536. [PubMed]
34. van de Vijver MJ, He YD, van't VeerLJ, et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med. 2002;347:1999–2009. [PubMed]
35. Bulavin DV, Saito S, Hollander MC, et al. Phosphorylation of human p53 by p38 kinase coordinates N-terminal phosphorylation and apoptosis in response to UV radiation. Embo J. 1999;18:6845–6854. [PubMed]
36. Samuels-Lev Y, O'Connor DJ, Bergamaschi D, et al. ASPP proteins specifically stimulate the apoptotic function of p53. Mol Cell. 2001;8:781–794. [PubMed]
37. Thukral SK, Blain GC, Chang KK, Fields S. Distinct residues of human p53 implicated in binding to DNA, simian virus 40 large T antigen, 53BP1, and 53BP2. Mol Cell Biol. 1994;14:8315–8321. [PMC free article] [PubMed]
38. Adorno M, Cordenonsi M, Montagner M, et al. A Mutant-p53/Smad complex opposes p63 to empower TGFbeta-induced metastasis. Cell. 2009;137:87–98. [PubMed]
39. Cocolakis E, Lemay S, Ali S, Lebrun JJ. The p38 MAPK pathway is required for cell growth inhibition of human breast cancer cells in response to activin. J Biol Chem. 2001;276:18430–18436. [PubMed]
40. Gal A, Sjoblom T, Fedorova L, et al. Sustained TGF beta exposure suppresses Smad and non-Smad signalling in mammary epithelial cells, leading to EMT and inhibition of growth arrest and apoptosis. Oncogene. 2008;27:1218–1230. [PubMed]