Malignant cell transformation, invasion, and metastasis are dependent on the coordinated rewiring of gene expression. A major component in the scaffold of these reprogramming events is one in which epithelial cells lose intercellular connections and polarity to adopt a more motile mesenchymal phenotype, which is largely supported by a robust transcriptional machinery consisting mostly of developmental transcription factors. This study demonstrates that the winged helix transcription factor, FOXQ1, contributes to this rewiring process, in part by directly modulating the transcription of TWIST1, itself a key mediator of metastasis that transcriptionally regulates the expression of important molecules involved in epithelial-to-mesenchymal transition. Forced expression and RNA-mediated silencing of FOXQ1 led to enhanced and suppressed mRNA and protein levels of TWIST1, respectively. Mechanistically, FOXQ1 enhanced the reporter activity of TWIST1 and directly interacted with its promoter. Furthermore, enhanced expression of FOXQ1 resulted in increased migration and invasion in colorectal cancer cell lines, whereas knockdown studies showed the opposite effect. Moreover, using the in vivo chicken chorioallantoic membrane metastasis assay model, FOXQ1 significantly enhanced distant metastasis with minimal effects on tumor growth.
Implications: These findings reveal FOXQ1 as a modulator of TWIST1-mediated metastatic phenotypes and support its potential as a biomarker of metastasis. Mol Cancer Res; 11(9); 1017–28. ©2013 AACR.
This article is featured in Highlights of This Issue, p. 965
Metastasis is the leading cause of cancer-related mortality in almost all cancer types, and occurs as a result of a coordinated progression of crucial steps encompassing local invasion, intravasation, survival in the circulation, extravasation, formation of micrometastasis and colonization (1). Intricately interwoven in this progression cascade is an orchestrated series of events that disrupt cell to cell and cell-extracellular matrix (ECM) interactions, enhancing cell motility and liberating epithelial cells from the surrounding tissue. This program involves in most instances the induction of a new transcriptional program that stimulates and maintains a mesenchymal phenotype and has come to be recognized as a fundamental mechanism that induces invasion and metastasis of tumor cells (2–5). Additional molecular processes are also engaged in the completion of this transition program and include the expression of specific cell-surface proteins, reorganization and expression of cytoskeletal proteins, production of ECM-degrading enzymes, and changes in the expression of specific microRNAs.
FOXQ1 belongs to the large forkhead family of transcription factors that share a conserved C-terminal forkhead/winged helix DNA-binding domain (6). Most forkhead family members have confirmed roles in embryonic development and tissue-specific gene expression (6–8). FOXQ1 itself is an approximately 42 kDa protein sharing the same paralogous cluster as FOXF2 and FOXC1 (9), and has been shown to be involved in hair follicle differentiation, gastrulation and mucin production in mice (10–12). Recent reports have confirmed it to be a downstream target of TGF-β, where it enhances mesenchymal transformation that seems to be cell type dependent (13, 14). Furthermore, it has been found to be overexpressed in colorectal and breast cancers, where it has been shown to enhance tumorigenicity, anti-apoptotic behavior, and tumor growth (14–16). A direct modulation of p21CIP1/WAF1 expression was shown to explain the anti-apoptotic effects in colorectal cancer (15), with independent experiments in breast cancer cell lines showing an acquired resistance to chemotherapy-induced apoptosis upon forced expression, and concurrent correlations to aggressive behavior, especially in triple-negative breast tumors (16). It has also been found to be of prognostic relevance in non–small cell lung cancer (NSCLC ref. 17).
Although substantial evidence supports the tumorigenic effect of FOXQ1, and correlation analyses have linked it to aggressive tumor behavior, the exact mechanisms by which this occurs have not been properly elucidated. Clearly, it has been shown to stimulate the expression of the cyclin kinase inhibitor, p21CIP1/WAF1 by binding its core promoter (15), and to induce a suppression of E-cadherin expression, by supposedly binding to E-box elements in its promoter, with conflicting reports regarding the validation of the latter (14, 16). As a transcription factor, it is clear that unraveling its functionality depends to a large extent on identifying its direct targets, and the dissection of its metastatic influence hinges on ascertaining which one(s) of these targets bear consequences on already known/novel metastatic molecules or, on metastatic pathways.
In the search of FOXQ1 targets which could explain its metastatic predisposition, we decided to explore genes whose expressions' were both downregulated when FOXQ1 was suppressed, and conversely upregulated when FOXQ1 was forcibly expressed. The selection of the investigated genes was based on a selected panel of genes known to be intricately involved in epithelial to mesenchymal transition (EMT) in addition to a gene pool, whose expression we found to be simultaneously deregulated in tumor epithelium, stroma, and whole-tissue compared with the corresponding normal equivalents following a microarray analysis of whole and laser-capture microdissected colorectal cancer tissues (18). This additional gene pool was equally implicated in epithelial to mesenchymal transition. The investigation was carried out in endogenously high and low FOXQ1-expressing colorectal cancer cell lines and 2 genes; E-cadherin and Twist1 showed consistent changes in both experimental set-ups.
In this study, we show that FOXQ1 expression correlates substantially with Twist1 in colorectal cell lines, and in resected patient tissues, that Twist1 is a direct target of FOXQ1, that the overexpression and downregulation of FOXQ1 leads to comparable and consistent changes in Twist expression, and this leads to significant changes in migration, invasion, and in vivo metastasis, but with no effect on tumor proliferation.
Materials and Methods
All cell lines, except GEO, a gift from Douglas Boyd (MD Anderson Cancer Center), and Colo206F from the German Collection of Micro-organisms and Cell Cultures (DSMZ), were obtained from the American Type Culture Collection (ATCC). The Colo206f, SW480, and HeLa cell lines were maintained in RPMI media supplemented with 10% fetal calf serum (FCS), GEO was cultivated in Dulbecco's modified Eagle medium (DMEM), whereas WiDR was in maintained in minimum essential medium (MEM). All cell lines were kept in a humidified incubator at 37°C with 5% CO2.
Plasmids and cloning
The full-length FOXQ1 coding region (IRATp970A1179D) was bought from ImaGenes (Source Bioscience Lifesciences) and was subcloned into the pCDNA 3.1 vector between the EcoRI and HindIII sites. The EGFP gene from the pEGFP C1 vector was cut out and cloned into the multiple cloning site of pCDNA 3.1 (NheI/EcoRI) in front of the FOXQ1 gene. This construct was used for all overexpression analysis. The abridged coding sequence was PCR amplified using the following primer sequences FX08 Fwd: 5′TCG GAT CCA TGA AGT TGG AGG TGT TC 3′; FX08 Rev: 5′ TTC TCG AGA TGA AGG GAA GGA GGA GC 3′ and cloned into a HA-tagged pcDNA 3.1 vector in frame with the tag between the Xho and BamHI sites. This construct was used in the chromatin immunoprecipitation (ChIP) assays.
All transient transfections (with Metafectane) were evaluated at 24-hour time-points. Stable Colo206f and GEO cell lines were selected on G418 treatment at a concentration of 400 and 600 μg/mL, respectively.
Two siRNA's targeting the 5′AGG GAA CCT TTC CAC ACT ATA 3′ (Hs_FOXQ1_3) and 5′CGC GCG GAC TTT GCA CTT TGA 3′ (Hs-FOXQ1_4) segments of FOXQ1 were purchased from Qiagen and were applied at a 50 nmol/L end concentration using the HiperFect transfection reagent and protocol (Qiagen). A greater knockdown effect was obtained with Hs-FOXQ1_4, which was then used for all subsequent experiments. The AllStars negative control siRNA (Qiagen) was used, as the name infers. A validated silencer select siRNA for Twist1 (s14523) was purchased from Life Technologies. Knockdown effects were evaluated at 48- and 96-hour time-points for mRNA and protein expression, respectively.
Quantitative real-time PCR (qRT-PCR) was done on the LightCycler480 system (Roche) using the SYBR green detection system (Thermo Scientific). All primers: FOXQ1, Twist1, CDH1, CDH2, CDH3, Snail1, ETV4, TGFB, CLDN1, LPAR1, CTNNB1, Vimentin, KIAA1199, and TBP were purchased from Qiagen from the Quantitect Primer Assay collection (Qiagen). Relative expression was calculated using the δδCT method with normalization to the tata-box binding protein (TBP). Absolute quantification was effected using standard curves generated from a plasmid containing the gene of interest.
Antibodies and Western blots
FOXQ1 (C-16), E-cadherin (G10), N-cadherin (8C11), and Twist (Twist2C1a) antibodies were all purchased from Santa Cruz Biotechnology. Actin (A2066) was purchased from Sigma, whereas β-catenin (L87A12) was from Cell Signaling Technology. Anti-HA tag antibody (ab9110) and rabbit polyclonal IgG (ab27478) were purchased from Abcam.
For Western blotting, cells were harvested and lysed with RIPA-buffer and protease inhibitors (Complete Mini, Roche Diagnostics), protein concentrations were determined with the BCA protein assay kit (Thermo Scientific), and samples were separated by SDS-PAGE as previously described (19). Tissue samples were homogenized with the Mixer Mills tissue lyser before processing, as described for cell lines.
Reporter gene assays
A Twist1 promoter plasmid (product ID S717559) encompassing -742 to +203 nt relative to the TSS was purchased from Switchgear Genomics. Within this construct were 4 putative “GTTT” binding sequences located at +39 to +43, −184 to −188, −195 to −199, and −604 to −608 relative to the TSS. The mutant forms were generated using the QuikChange II site directed mutagenesis Kit (Agilent Technologies). Colo206f cells were plated at a density of 3 × 104 cells/well in 96-well plates, and were transfected after a 24-hour incubation period using Metafectane (Biontex Laboratories). Assays were conducted in quadruplicates and repeated at least 3 times and reporter signals were assessed with the LightSwitch luciferase assay Kit (SwitchGear Genomics).
The ChIP experiments were conducted using the ChIP assay Kit (Upstate) from Millipore, according to the manufacturer's instructions. Colo206f cells were transfected with an empty vector (pcDNA-HA) or HA-tagged FOXQ1 vector. The putative region of the Twist1 promoter was amplified with the following primers ChpX Fwd: 5′ GTC TCC TCC GAC CGC TTC 3′and ChpX Rev: 5′ ATT CGT CCT CCC AAA CCA TT 3′ primers and assayed with both quantitative and semiqRT-PCR.
MTT (3-(4,5-Dimethylthiazol-2)-2,5-diphenyltetrazolium bromide) assay
Cells (1 × 104) were plated in 96-well plates following which 20 μL of MTT was added at 0-, 24-, 48-, 72-, 96-, and 120-hour time-points. Hundred microliters of sodium dodecyl sulphate (SDS) diluted in hydrochloric acid was added 3 hours after the MTT solution (Sigma-Aldrich), followed by measurement of the absorbance at 550 and 640 nm. FOXQ1 and vector-transfected stable cell lines (GEO, colo206f), as well as FOXQ1 siRNA and scrambled transfected cell lines (SW480, WiDr) were evaluated alongside each other.
Migration and invasion assays
Analysis of cell invasion and migration was done using transwell chambers (BD Biosciences) with and without a (10 μg/mL) matrigel coating, respectively. The protocol is as was previously published (20). Migration was additionally evaluated with the wound healing assay. Briefly, cells were seeded in 24-well plates at a density that enabled a confluency of 80% to be attained 24 hours after plating. A 200 μL filter tip was used to gently scratch the cell monolayer across the center of the well 24 hours after seeding. The cells were then gently washed with phosphate-buffered saline (PBS) to remove the dislodged cells, and then replenished with fresh medium, after which the first images were acquired. The cells were incubated for a further 24 hours after which a second set of images were acquired.
In vivo chicken chorioallantoic membrane metastatic assay
This is a well-established model for assessing in vivo metastasis (21–24). Fertilized special pathogen-free (SPF) eggs were purchased from Charles River and incubated for 10 days in a rotary incubator, after which 2 × 106 FOXQ1 and vector expressing GEO cells were inoculated onto the top chorioallantoic membrane (CAM) following a miniaturized dissection of the shell. After a further 7 days of incubation, the developing chicken embryos were sacrificed and lungs, liver, and the developing tumor on the top CAM were removed. Genomic DNA was isolated from the liver and lungs and the number of metastasized cells was evaluated using human-specific Alu-PCR primers and probes, as previously described (25). The tumors removed from the top CAM were weighed to assess in vivo tumor growth.
Significance was set at P ≤ 0.05. Two-tailed unpaired and paired t tests were employed when dealing with independent (cell lines, before/after treatment) and dependent (tumor/normal patient) variables, respectively. Correlation analysis was done with the Pearson's correlation coefficient and lambda coefficient of association algorithms. Calculations were made using Sigmaplot, Microsoft Excel, and Graph pad prism 5.0 tools.
FOXQ1 is overexpressed in epithelial and stromal tumor compartments alongside other EMT genes
Following a microarray analysis of whole-tissue and laser-capture microdissected tissues in a series of 35 patients with colorectal cancer in which epithelial, stromal, and whole-tissue compartments were compared, we found FOXQ1 to be simultaneously overexpressed at the mRNA level in all 3 tumor compartments as compared with the equivalent normal samples. In addition to FOXQ1, the genes identified contained many molecules already implicated in EMT and cellular differentiation, inclusive of TGFB1, TIMP1, ADAMTS1, LPAR1, and CLDN1 (Supplementary Table S1). Details describing the identification and validation of this gene pool are elaborated in our recent publication (18). FOXQ1 was chosen for further evaluation because of its significant expression in all 3 compartments, its relevance to the EMT hypothesis as an embryonic transcription factor, and on its relative obscurity in reports in literature.
Forced expression and knockdown of FOXQ1 modulates Twist1 expression
In the search for possible molecules that were directly affected by FOXQ1 expression, we conducted, using qRT-PCR, an analytical expression screen of genes identified from the common microarray gene pool (Supplementary Table S1), concurrently incorporating other well-described, EMT-related genes. The expression profile of these genes was evaluated following enhanced expression and knockdown of FOXQ1 in endogenously low and high mRNA expressing cell lines, respectively. As the mRNA and protein expression of the cell lines did not always correlate (Fig. 1A), overexpression studies were done in Colo206F and GEO cell lines (Fig. 1B), whereas knockdown experiments were conducted in both high mRNA expressing (SW480 and HCT15) and high protein expressing (WiDR) cell lines (Fig. 1C and D). We observed that of all the genes tested for CDH1 (E-cadherin) and Twist1 expression profiles correlated concurrently and significantly with FOXQ1, with expression levels increasing and decreasing (Twist1 increases, E-cadherin decreases and vice-versa with FOXQ1 expression (Fig. 1B–D). In all situations, there was always a concomitant cadherin switch, between E-cadherin and either N- (CDH2) or P- (CDH3) cadherin (Fig. 1B–D).
As we observed a systematic correlation of expression between Twist1 and E-cadherin following FOXQ1 deregulation, and as E-cadherin is a known target of Twist1, we proceeded to identify if the deregulation of E-cadherin occurred following that of Twist1. Following the knockdown of FOXQ1 in the SW480 cell line, over a 72-hour timeline, we observed that indeed the expression of Twist1 was affected before that of E-cadherin, suggesting that the effect of FOXQ1 on E-cadherin was likely mediated through Twist1 (Fig. 1E).
The Twist promoter harbors putative FOXQ1 binding motifs
Upon the realization of 2 potential candidate genes, whose expressions were affected by FOXQ1, we proceeded to identify the consensus binding sequence of FOXQ1 and find out if this sequence occurred in the promoter on any of these genes, the presence of which would signify a potential direct regulatory effect. Overdier and colleagues (26) identified a core FOXQ1 binding sequence of “TGTTTA,” whereas bioinformatic searches within Transfac suite (www.biobase-international.com) and JASPER CORE database (www.jasper.cgb.ki.se) had matrix motifs from rat studies only (Fig. 2A), with a core “GTTT” motif. As FOXQ1 is evolutionarily conserved (Fig. 2b), and the forkhead DNA binding domain of the human, mouse and rat FOXQ1 proteins share 100% identity (6), it implied that the core GTTT sequence may indeed be valid across many species. A detailed analysis of the respective gene promoters revealed that the Twist1 promoter harbored the cis-elements required for FOXQ1 binding (Fig. 2C).
FOXQ1 transcriptionally activates Twist1
Sequel to our findings above, we went further to investigate if FOXQ1 transcriptionally activated Twist1. Reporter gene assays with the −742 to +203 segment of Twist1 promoter showed a 1.5- to 2-fold increase in Twist1 expression as compared with the control vector (Fig. 3A). In addition, titrated increments in the concentration of FOXQ1 led to increments in reporter activity (Fig. 3b). Within this construct were 4 putative “GTTT” binding sequences located at +39 to +43, −184 to −188, −195 to −199, and −604 to −608 relative to the transcription start site (TSS; ref. 27). Two of these sites were within 7 nt of each other signifying an enhanced binding probability to this location. To verify if these sites were important in the observed induction of Twist1 expression, we mutated these positions, with discernibly decreased effects on the reporter activity at the transcription start site 184 to transcription start site 199 position (Fig. 3C). To further corroborate our findings, we conducted a ChIP assay specifically looking for enrichment of a short fragment containing the 2 putative sites that lay in close proximity to each other, where we observed, after normalizing to the input, an approximately 3-fold increase in the FOXQ1 consensus-binding sites immunoprecipitated from overexpressing cell lines as compared with the vector control (Fig. 3D and E). No significant induction was observed at the +39 to +43 and −604 to −608 sites (data not shown).
FOXQ1 expression correlates with Twist1 expression in resected patient tissues
To assess the in situ validity of our hypothesis, we compared, using RT-PCR, the relative mRNA expression between FOXQ1, Twist1, and other EMT-related genes taken from our microarray experiment in an independent cohort of 45 patients. We obtained a coefficient of association (λ) of 0.66 for FOXQ1 and Twist1, suggesting an in vivo proof of principle. We tested this association between other genes where a direct regulatory effect of one on the other has already been established, where we obtained for instance, in the comparison between Snail and E-cadherin, and between Twist1 and E-cadherin, coefficient of associations of 0.428 and 0.483, respectively (Fig. 4A). We went further to check for a correlation at the protein level, with similar results (Fig. 4B). In a bid to further validate our data, we searched open databases, and in the R2 database (http://r2.amc.nl), where several expression profiling data are available, we found that especially in the MAQC studies, the correlation between FOXQ1 and Twist1 expression had a coefficient of r = 0.9 (Fig. 4c). The MAQC project generated a data set which compared the consistency of microarray data between laboratories and across platforms (28), and which supports our findings in patients with colorectal cancer.
Forced expression of FOXQ1 did not result in mesenchymal transformation in Colo206F, GEO, and MDCK II cell lines
In line with published reports that FOXQ1 induced mesenchymal transformations in certain cell lines and not in others, we analyzed the phenotypic characteristics of 2 colorectal cancer cell lines (Colo206f; GEO) and the canine MDCK II cell line normally used in EMT studies. In all of the 3 cell lines, we found no discernible mesenchymal morphologic features with light microscopy that differentiated FOXQ1 and vector expressing cell lines (data not shown).
FOXQ1 has no significant effect on cell proliferation and in vivo tumor growth
In the bid to unravel the functional significance of FOXQ1 on tumor progression, we analyzed the effects of FOXQ1 overexpression and knockdown on cell proliferation using the MTT assay. We found no significant effect on cell proliferation in both sets of experiments (Fig. 5A and B). To address the question if this inconsequential effect on cell proliferation also occurred in vivo, we analyzed the tumor masses that developed on the top CAM of eggs inoculated with either FOXQ1 or vector expressing cells where no significant difference was observed between the groups further supporting the in vitro cell proliferation assays (Fig. 5C).
FOXQ1 enhances tumor migration and invasion
To dissect the impact that FOXQ1 had on tumor progression, we appraised its influence on migration and invasion using the Boyden chamber assay, without and with a matrigel coating, respectively. This was evaluated in 4 cell lines, 2 overexpressing FOXQ1 (COLO206F and GEO), and 2 in which FOXQ1 expression was suppressed (SW480 and WiDR). We observed a significant increase in cell migration upon forced expression, and a concurrent decrease in migration, when FOXQ1 was suppressed (Fig. 6A). The same results were mirrored in the invasion assays, but with a slightly less significance than was obtained for the migration assay (Fig. 6B). Cell migration was additionally evaluated using wound-healing assay (Supplementary Figure S1).
FOXQ1 enhances distant metastasis
Using the in vivo chicken CAM metastatic assay, we evaluated the effect that FOXQ1 expression had on distant tumor metastasis. The liver and lungs of chicken embryos were dissected following 1 week of incubation after the inoculation of FOXQ1- or vector-GEO overexpressing cells on the top CAM. Using RT-PCR to detect human-specific Alu DNA sequences, we were able to show a significant increase in both lung and liver metastasis, as compared with the vector control (Fig. 7A and B). However, as stated above, we did not observe any significant change in the primary tumor size itself, indicating that the effect of FOXQ1 on metastasis is proliferation independent.
This is the first study to show that FOXQ1 regulates invasion and metastasis in colorectal cancer. Metastasis is a highly complex process, consisting of at least 5 different steps (invasion, intravasation, dissemination in the circulation, extravasation, and colonization of distant sites), each of which depends on the action of diverse molecular events. Over the years, different key molecules have been identified that contribute significantly to at least one phase of the metastatic cascade, some playing multiple roles in more than one part of the process.
The EMT represents an integral part of metastatic progression and interestingly and several of the molecules and pathways involved converge on critical and common end points, one of which is E-cadherin (4, 29). E-cadherin is invariably almost always downregulated during EMT with a concomitant switch to a mesenchymal cadherin that may, or may not be associated with an overt phenotypic mesenchymal transformation (30, 31). This cadherin switch is modulated by a number of transcriptional regulators, one of which is Twist1 (32).
In this study, we evaluated the forkhead box q1 transcription factor, a relatively new addition to embryonic transcription factors that are also involved in metastasis, and show that it modulates its effects in part by acting on Twist1, whose role in EMT is well entrenched (33–35). We postulated that FOXQ1 transactivates Twist1, which in turn suppresses E-cadherin expression, as supported by the concomitant and consistent deregulation of these genes upon FOXQ1 overexpression/knockdown, by the fact that Twist1 is an established regulator of E-cadherin expression and finally by the observation that Twist1 deregulation precedes that of E-cadherin (Fig. 1E). It seems that the expression of FOXQ1 itself is subject to cell-specific posttranscriptional and/or posttranslational regulation that merits further investigation, as supported by the inconsistencies in FOXQ1 mRNA and protein levels we observed in the colorectal cancer cell lines we analyzed, and as corroborated by similar observations in breast cancer cell lines, where for instance, 2 relatively high FOXQ1 mRNA expressing cells; MDA-MB-468 and MB-436 had virtually none, and very high protein expression levels, respectively (16). The cooperation of multiple transcription factors in the modulation of EMT is not new (36–39). For example, following TGF-beta treatment in NMuMG cells, Snail1 was found to rapidly enhance Twist1 mRNA and protein expression, as well as induce the nuclear translocation of Ets1, and both Twist and Ets1 proceeded to induce the expression of ZEB1 by interacting with different cis-elements in its promoter (36). Snail1, Twist1, and ZEB1 all repress E-cadherin expression by binding to E-box elements in its promoter. It is likely that the interaction of FOXQ1 with Twist1 further reinforces the suppression of E-cadherin transcription, as FOXQ1 itself does not have any binding elements in the E-cadherin promoter. In addition, the ensuing cadherin switch could also be likely due to a direct consequence of Twist1 modulation, as it has been proven to be a direct regulator of both E- and N-cadherin expression (40, 41). Interestingly, 2 recent reports have shown that FOXQ1 was able to directly repress E-cadherin expression by supposedly binding to E-box elements in its promoter, however, the in vivo validation was inconsistent, where one group could, and the other could not show a convincing FOXQ1 enrichment at the proximal E-cadherin promoter containing the E-box (14, 16). Despite the changes in expression of EMT markers, we observed at both mRNA and protein levels; we failed to observe an overt phenotypic mesenchymal transformation. This may have been influenced by EMT inhibitors like Grainyhead-like-2 (GRHL2). GRHL2 gene amplification has been noted in several tumor types and directly represses the ZEB1 promoter (42). As ZEB1 is known to be required for EMT in response to Twist1, it is probable that GRHL2 may have contributed to this outcome, but further studies are needed to address this postulation.
The forkhead family of transcription factors are known to be involved in tissue-specific transcription (43, 44), and even though we observed an influence of FOXQ1 on Twist1 in 4 different colorectal cancer cell lines, this effect was not present in the canine epithelial MDCK-II and 293T embryonic kidney cell lines (data not shown), further reiterating this point.
From our experiments, the effects of FOXQ1 on tumor progression were most marked on cell migration, but nonetheless significantly impacted invasion and metastasis. As opposed to the findings of Kaneda and colleagues (15), and in concordance with that of Zhang and colleagues, and Qiao and colleagues (14, 16), respectively, we did not observe any effect on cell proliferation, and in vivo tumor growth, even though the findings of the latter 2 groups were limited to breast cancer cell lines. Nonetheless, it is important to note that the experiments on which Kaneda and colleagues made their observations were carried out in the lung cancer p53-null H1299 cell line, which on the account of this deletion already has an increased proliferative and tumorigenic propensity (45, 46), and could account for the lack of concordance to our observations. Given that EMT represents the critical event in the transition from early to invasive carcinomas (5) and that both overexpression of Twist1 and the downregulation of E-cadherin are associated with poor prognosis in colorectal cancer (47, 48), our finding that FOXQ1 is involved in the regulation of colorectal cancer progression and metastasis is pertinent. However, it is also important to state that Twist1 itself is induced by other strong transcriptional regulators (27, 39, 49, 50), that the effects of FOXQ1 on Twist1 while significant, were modest, and the observed effects of FOXQ1 on invasion, migration, and metastasis are very likely multifactorial. To ascertain if Twist1 was entirely responsible for the modulation of these effects, we knocked down Twist1 in FOXQ1 stable overexpressing Colo206f cells and vector control cells. Interestingly, we observed a significant reduction in both migration and invasion in the Twist1 knockdown cell lines, as compared with the FOXQ1 stably expressing cells. Still, this knockdown did not completely eliminate the increase in migration and invasion associated with FOXQ1 overexpression (as compared with the vector group), suggesting that whereas Twist1 is significant for the increase in aggressive cell behavior attributable to FOXQ1, other hitherto unidentified molecules also contribute to these processes (Supplementary Fig. S2).
Taken together, our results show that FOXQ1 is a novel modulator of Twist1 expression and a regulator of cancer invasion and metastasis. Certainly, further studies need to be conducted to characterize more of its targets, both direct and indirect, that bear consequences on metastasis. The fact that it has been found to be significantly overexpressed in colorectal cancer in 2 independent studies encourages us to support its further evaluation as a potential tumor marker, particularly one that is indicative of metastasis.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: M. Abba, S.A.K. Rasheed, H. Allgayer
Development of methodology: M. Abba, N. Patil, J.H. Leupold, H. Allgayer
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Abba, N. Patil, L.D. Nelson, G. Mudduluru, H. Allgayer
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Abba, L.D. Nelson, J.H. Leupold, H. Allgayer
Writing, review, and/or revision of the manuscript: M. Abba, L.D. Nelson, H. Allgayer
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Mudduluru
Study supervision: H. Allgayer
The authors thank Jürgen Engel and Jacqueline Frohhert for technical assistance. This manuscript contains part of the dissertation (Dr. sc. hum) of MA.
H. Allgayer was supported by Alfried Krupp von Bohlen and Halbach Foundation (Award for Young Full Professors), Essen, Hella-Bühler-Foundation, Heidelberg; Dr. Ingrid zu Solms Foundation, Frankfurt/Main, the Hector Foundation, Weinheim, Germany; the FRONTIER Excellence Initiative of the University of Heidelberg; the BMBF, Bonn, Germany; the Walter Schulz Foundation, Munich, Germany; the Deutsche Krebshilfe, Bonn, Germany; and the DKFZ-MOST German-Israeli program, Heidelberg, Germany.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.