Snail upregulates FN and LEF transcription negatively feed backed by PPAR-gamma: a general model established for Snail to transactivate mesenchymal genes

Background Snail (SNA) is responsible for epithelial mesenchymal transition, migration and metastasis of hepatocellular carcinoma. SNA represses the transcription of the essential epithelial marker such as E-cadherin and enhances mesenchymal markers including bronectin and lymphoid enhancer-binding factor. Our previous studies indicated that SNA, in collaboration with EGR1 and SP1, may directly activate transcription of the mesenchymal markers, matrix degradation enzyme matrix metalloproteinases (MMP9) and zinc nger E-box binding homeobox 1 (ZEB1) in HepG2 cell stimulated by the phorbol ester tumor promoter 12-O-tetradecanoyl-phorbol 13-acetate (TPA). Besides, we pinpointed a SNA binding motif (TCACA) upstream of EGR1/SP1 overlapping region on promoters. In this study, we investigated whether LEF and FN are transcriptionally regulated by SNA in a similar fashion. Moreover, a general model for SNA-upregulated mesenchymal markers is proposed. Methods RT/PCR and Western blot were used for analyzing gene expression and shRNA technology for depleting SNA. Dual luciferase assay was used for promoter activation; deletion mapping and mutagenesis were used for conrming the indicated promoter region required for transcription activation. ChIP and EMSA were used for validating the binding of the indicated transcription factor on their putative motifs. Results SNA binding motif and E/S overlapping region are required for TPA-induced of LEF and by TPA-induced binding and on indicated promoter specicity protein 1; FN: bronectin and LEF: lymphoid enhancer-binding factor, PPAR-γ: peroxisome proliferator-activated receptor -γ; COX2: cyclooxygenase 2; COL1A1: collagen type I alpha 1 chain; FOXA2: Forkhead box protein A2; PM: point mutation; ChIP: chromatin immunoprecipitation; EMSA: electrophoresis mobility shift assay; EMT: epithelial mesenchymal transition; HCC: hepatocellular carcinoma.


Background
The poor prognosis of hepatocellular carcinoma (HCC) is due to frequent metastasis occurring via complicated processes, including epithelial mesenchymal transition (EMT), migration and invasion of primary tumor [1]. Among the metastatic transcription factors, Snail (SNA) was not only overexpressed in HCC and associated with poor prognosis of HCC [2,3] but also may accelerate EMT, invasion and metastasis of HCC [1,[3][4][5][6]. Moreover, SNA was regarded as a promising therapeutic target for preventing progression of various tumors including HCC [7]. SNA can be regulated by signalings triggered by a lot of metastatic factors including receptor tyrosine kinase (RTK) and TGF-b. While SNA directly represses epithelial markers like E-cadherin, it also upregulates the mesenchymal markers, including vimentin, bronectin and matrix metalloproteinases (MMP9) [8][9][10]. SNA also upregulates other mesenchymal transcriptional factors such as Twist and zinc nger E-box binding homeobox 1 [11,12] (ZEB1) to complete its EMT program.
The transcriptional mechanisms for SNA to suppress epithelial markers such as E-cadherin are well elucidated. SNA can bind to the E-boxes (5'-CACCTG) on E-cadherin promoter and recruit various epigenetic machineries such as histone deacetylase leading to the repression of E-cadherin expression [13,14]. In contrast, the detailed mechanisms for SNA to upregulate transcription of mesenchymal markers are far less clari ed. Previous studies indicated that SNA activated transcription of MMP9 and ZEB1 indirectly via regulation of other transcriptional factors such as Twist, Ets-1 and Proximal speci city protein 1 (SP1) or microRNAs in different contexts [15,16]. However, our recent report demonstrated that SNA, in collaboration with Early Growth Response 1 (EGR1) and SP1, also directly upregulated transcription of MMP9 and ZEB1 by binding to a speci c putative SNA binding motif "TCACA" for activating their promoter [17]. This is in similar with those observed in the transcription of cyclindependent kinase 4/6 (CDK4/6), p15 INK4b , a cell cycle related gene upregulated by SNA [18]. Intriguingly, after examining sequences of the other known SNA upregulated mesenchymal markers including vimentin, Twist1, vitronectin, a2 smooth muscle actin (ACTA2), bronectin (FN) and lymphoid enhancerbinding factor (LEF), we found the similar sequence architecture, ie, the SNA motif TCACA coupled with neighboring EGR1 and SP1 region, also existing in their promoters around 300-2000 bp upstream of the translational initiation sites. Thus it is tempting to investigate whether SNA can directly activate transcriptional upregulation of other EMT-related genes involved in HCC progression such as FN [19] and and LEF [20] in a similar fashion as observed in MMP9 and ZEB1.
By gene expression and promoter analysis coupled with protein DNA interaction assay, we did nd that both LEF and FN are transcriptionally regulated by SNA via binding to the proposed SNA binding motifs which is close to EGR1 /SP1 overlapping region. Moreover, we found PPAR-g, a well-known antagonist of SNA, plays as a negative feedback regulator in this process. In addition, cyclooxygenase-2 (COX2), collagen type I alpha 1 chain (COL1A1) and other mesenchymal genes are also potentially regulated in the same way.

Materials And Methods
Cell culture, chemical and antibody The cultured conditions for HCC340 and HepG2 cells were the same as described in our previous report [17]. HCC340 derived from a patient-derived hepatocellular carcinoma cell line from Buddhist Tzu Chi Hospital, Taiwan [21]. Tetradecanoyl phorbol acetate (TPA) and GW9662 were purchased from Sigma-Aldrich (Poole, UK). Monoclonal Snail (C15D3) and PPAR-g (81B8) antibodies were from Cell Signaling (Beverly, MA). The antibodies against EGR1, SP1, MMP9, ZEB1, LEF1, Fibronectin, Histone H3 and GAPDH were purchased from Santa Cruz Biotechnology (California, USA). The snail expression plasmid, p-Snail, driven by CMV promoter within the p-cDNA3 vector, is a gift from Dr. Cheng K.K. in Tzu Chi university. The PPAR-g expression plasmid, p-PPAR-g, driven by CMV promoter within the p-cDNA3 vector, was purchased from ORIGENE (Rockville, USA).

Constructions of various promoter plasmids for deletion mapping
The promoter regions in the full-length promoter plasmids of FN, LEF and the other candidate SNA upregulated genes were ampli ed from genomic sequence of each gene. The PCR products were ligated into pGL3 vector (Promega, Madison, WI, USA). The promoter plasmids of deletion constructs were derived from each full-length promoter by double digestion with various restriction enzymes followed by lling in the restriction site overhangs by Klenow enzyme. Subsequently, the digested DNA fragments will be ligated into pGL3 vector.

Site-directed mutagenesis on promoters
The full-length promoter plasmids of LEF, FN and the other Snail upregulated genes were used as templates for site-directed mutagenesis using a GeneEditor in vitro site-directed mutagenesis system (Promega, Madison, WI, USA) to obtain various mutant promoters according to the manufacturer's protocol. The bases changed in the site-directed mutagenesis for proposed Snail binding motif (TCACA), the EGR1/SP1 overlapping and PPAR-g binding region were the same as those described in our previous report [18].
Dual luciferase promoter assay was performed as described in our previous report [18].
Chromatin immunoprecipitation (ChIP) assay was performed as described in our previous report [18]; schematic map as supplemental Fig. 6 showed the PCR fragments (Table 2) ampli ed for the ChIP assay of SNA, EGR1/SP1 overlapping and PPAR-γ.
RT-PCR and Quantitative RT-PCR will be performed as described in our previous report [18] and primers used for RT-PCR and real time PCR in gene expression in Table 3 Electrophoresis mobility shift assay (EMSA) was performed as described in our previous report [18]. Nuclear-cytoplasmic protein fractions were collected using the NE-PER Nuclear and Cytoplasmic Extraction Reagents kit (Thermo Fisher Scienti c) according to the manufacturer's protocol. The sequence of the biotin-labeled probe (25 bp) and un-labeled competitors of FN promoter (20 bp) are 5′CCCCTTCGCT TCACACAAGTCCAGC3′ and 5′CCTTCGCTTCACACAAGTCC3′ (wild type)/ 5′CCTTCGCTTTGTACAAGTCC3′ (mutant), respectively. Also, The sequence of the biotin-labeled probe (25 bp) and un-labeled competitors of LEF promoter (20 bp) are 5′CACACCACACTCACACACCCCAAAA3′ and 5′CACCACACTCACACACCCCA3′ (wild type)/ 5′CACCACACTTGTACACCCCA3′ (mutant).
Western blot was performed as described in our previous reports [17,18].
shRNA technology was performed as described in our previous reports [18]. The sequence of shRNA fragments targeting different regions of Snail will be the same as those used in our previous study [18].

Statistical analysis
Data will be analyzed using Student's t-test in Excel. All the quantitative studies will be performed at least in triplicate, as appropriate. Statistical signi cance between groups will be indicated by *P<0.05 and **P<0.005.

Results
Snail is essential for constitutive and TPA-induced gene expression of FN and LEF in HepG2 and HCC340 cells Initially, by RT-PCR, we found TPA elevated mRNA of FN and LEF as well as SNA by 2.0~2.7-fold within 1-6 h in HepG2, one of the conventional HCC cell lines (Fig. 1a). This was also quantitatively con rmed by Q-RT-PCR (supplemental Fig. 1a). Also, Western blot con rmed FN and LEF proteins were elevated by TPA at 6 h by 1.6-1.8-fold, and declined at 12 h whereas TPA induced dramatic elevation of SNA earlier at 2h and 4h and gradually declined at 6 and 12 h (supplemental Fig. 1b) implicating SNA is an upstream regulator of FN and LEF. In both conventional RT-PCR (supplemental Fig. 1c) and quantitative (Q) RT-PCR ( Fig. 1b) analysis, TPA-induced elevation of FN and LEF mRNA were found to be attenuated by three shRNAs of SNA, shSN18, shSN19 and shSN20 by 50-80% at 4h in HepG2 and HCC340, a patient-derived HCC cell lines used for studying the SNA-upregulated MMP9 and ZEB1 transcription [17]. On the contrary, overexpression of SNA for 6-16 h elevated FN and LEF mRNAs by 3-10-fold in both HepG2 and HCC340 ( Fig. 1c and 1d). Collectively, SNA is essential for gene expression of FN and LEF in HCC. Transfection of SNA shRNAs, including shSN18, shSN19 or shSN20 suppressed the TPA-induced promoter activity of FNpro1938 and LEF pro1897 by 38-60% compared with that of the control (Luciferase, Luc) shRNA group. Similarly, TPA induced promoter activity of FNpro1938 and LEFpro1891 by 25 and 12-fold, respectively, in HCC340, which can be suppressed by aforementioned SNA shRNA by 60-70% (Fig. 2b). Moreover, overexpression of SNA increased constitutive promoter activity of FNpro1938 and LEFpro1897 by 2-2.5-fold in HCC340 (Fig. 2c) and HepG2 (data not shown). Collectively, SNA is essential for constitutive and TPA-induced transcriptional activation of FN and LEF in HCC.
Deletion mapping identi ed SNA motif coupled with the adjacent EGR1/SP1 overlapping regions within TPA-response element on both FN and LEF promoters Further, we sought to identify the TPA-responsive element in promoters of FN and LEF and examined whether it contained the proposed SNA binding region "TCACA" and putative EGR1/SP1 overlapping motif as described previously [17]. According to the Genomatix software, there are three and two proposed SNA binding regions, respectively, locating on promoter of FN (-1187 to -547 bp upstream of the translation start site) and LEF (-1621 to -1211 bp upstream of the translation start site) (supplemental Fig. 3a, b). To examine whether any of these SNA binding motifs were required for promoter activation of both FN and LEF, deletion mapping analysis using 5' end truncated mutants including FNpro1150, FNpro750, FNpro700 and FNpro568 (excluding 0, 1, 2 and 3 SNA region, respectively, from 5' end on fulllength FN promoter) (Fig. 3a, b upper left panel) and LEFpro1595, LEFpro1300 and LEFpro1231 (excluding 0, 1, and 2 SNA region, respectively, from 5' end on full-length LEF promoter) (Fig. 3 a, b lower left panel) were performed. Surprisingly, we found a signi cantly higher TPA-induced promoter activity of Together, these implied that the most proximal (3') SNA motifs are required for TPA-induced promoter activity of both FN and LEF. Interestingly, the TPAresponsive SNA motifs on both promoters are close to a EGR1/SP1 (E/S) overlapping region which was known to be required for TPA-induced, SNA-mediated MMP9 and ZEB1 promoter activation [17]. Thus, it is very probable that the E/S overlapping regions close to the candidate SNA site on both promoters are also involved in SNA-dependent transcriptional upregulation in FN and LEF. Indeed, FNpro330 and LEFpro1100 which lack E/S overlapping and all the SNA regions, showed a further decrease (by 60-80%) of TPA-induced promoter activity compared with those of the FNpro568 and LEFpro1231 that still contained E/S (Fig. 3a, b). Collectively, these strongly suggest that the most proximal (3') SNA region coupled with the adjacent E/S overlapping regions are essential for TPA-induced promoter activations of both FN and LEF.
Mutagenesis on promoter validated SNA binding motif and E/S overlapping region required for transcription of FN and LEF To prove whether the candidate SNA binding motifs are required for transcription of LEF and FN, site directed mutagenesis were performed, producing promoter mutants with altered sequence (TCACA®TTGTA) at indicated SNA motifs (PM 1-3). As shown in Fig. 3 (c, d), FN promoter mutant with altered sequence at the proximal (3') SNA binding motif (PM3) exhibited a 60% decrease of TPA-induced promoter activity compared with that of the wild type FNpro1150 in both HepG2 (Fig. 3c, upper panel) and HCC340 (Fig. 3d, upper panel) at 12 h, whereas mutation of FN promoter at the middle SNA (PM2) and the distal (5') SNA region (PM1) didn't decrease the TPA-induced promoter activity. Similarly, LEF promoter mutant with the altered sequence at the proximal (3') SNA (PM2) but not the distal (5') SNA (PM1) exhibited a signi cantly decrease (by 50-70%) of TPA-induced promoter activity compared with that of the wild type LEFpro1595 in both HepG2 (Fig. 3c, lower panel) and HCC340 (Fig. 3d, lower panel).
Collectively, these validated that the most proximal (3') SNA binding regions on promoters of both FN and LEF are essential for TPA-induced promoter activation. In addition, mutation (CCCCGCCT ®CCCTATCT) on the E/S region (-1195 to -1179 bp) of LEFpro1595 (LEFpro1595 E/S-PM1) close to the active SNA motif, but not the next (proximal) E/S region (-1134 to -1119 bp) (LEFpro1595 E/S-PM2) also attenuated the TPA-enhanced promoter activity of LEFpro1595 by 33% in HepG2 (Fig. 3c, lower panel) and HCC340 (Fig. 3d, lower panel). As a negative control, mutagenesis on the FOXA2 region close to the active SNA region has no effect on TPA-induced promoter activity of LEF (Fig. 3c, d, lower panel). Taken together, the proximal (3') SNA region coupled with the adjacent E/S overlapping region are responsible for TPAinduced promoter activation of FN and LEF.
PPAR-g is the negative feedback regulator in SNA-mediated transcription of FN and LEF In the deletion mapping analysis, we found a negative regulatory region upstream of the distal (5') SNA binding motif on promoters of both genes as described above (Fig. 3a, b). Notably, a putative binding motif of peroxisome proliferator-activated receptor γ (PPAR-g) transcriptional factor according to the JASPAR database, was identi ed within these regions ( Fig. 3 and supplemental Fig. 3). Interestingly, both SNA and PPAR-g may antagonize gene expressions of each other. PPAR-g is well known to be a repressor of SNA for blocking EMT [22]. On the other hand, SNA can inhibit the expression of PPAR-g for adipocyte differentiation [23]. In a Q-RT-PCR analysis, PPAR-g mRNA was increased by treatment of TPA at 6 h by 2.8-fold, further increased to 3.3-fold at 12 h and declined to 2.0-fold at 14 h (supplemental Fig. 2a, left  panel). As expected, it can be suppressed by GW9662, one of the PPAR-g antagonists, by 50% at 12 h time point (supplemental Fig. 2a, left panel). Moreover, TPA-induced PPAR-γ mRNA at 12 h was enhanced by SNA shRNA, shSN18, and on the contrary, it was diminished by overexpression of SNA approximately by 1.5-2-fold at 12 h (supplemental Fig. 2b, right panel). Thus, SNA can be a negative regulator of PPAR-γ in TPA-treated HCC. On the other hand, we examined whether PPAR-γ plays a negative regulatory role in the expression of SNA and the SNA-upregulated mesenchymal genes. In Q-RT-PCR of SNA, FN and LEF (supplemental Fig. 1a), GW9662 reversed the down regulation of mRNA of SNA, FN and LEF by 50~60 % at 12 h of TPA treatment. Similarly, in Western blot of SNA (supplemental Fig. 1b, upper panel), FN and LEF (supplemental Fig. 1b, lower panel), GW9662 dramatically reversed the down regulation of protein of SNA, FN and LEF at 12 h to a level close to those at 6 h of TPA treatment. On the transcriptional level, GW9662 (2.5-5.0mM) dose dependently elevated the TPA-induced promoter activation of both full-length promoters, FNpro1938 and LEFpro1897 (containing PPAR-g region) at 12 h to a level close to FNpro1150 and LEFpro1595 (without PPAR-g region) in both HepG2 (Fig. 4a) and HCC340 (Fig. 4b). To validate the negative regulatory role of PPAR-g, we investigated whether it is also involved in the SNA-mediated MMP9 and ZEB1 upregulation as reported previously [17]. Interestingly, TPA-induced promoter activation of fulllength promoter of MMP9 and ZEB1 (MMP9pro1920 and ZEB1pro2052, containing PPAR-g) were also lower than those of shorter promoter fragments lacking PPAR-g region but contained the TPA-responsive SNA motif (MMP9pro950 and ZEB1pro1079) [17] in both HepG2 (supplemental Fig. 4a) and HCC340 (supplemental Fig. 4b). As expected, cotreatment of GW9662 (2.5-5.0 mM) elevated TPA-induced promoter activation of full-length promoter of MMP9pro1920 and ZEB1pro2052 (containing PPAR-g region) close to that of MMP9pro950 and ZEB1pro1079 (without PPAR-g region) in a dose dependent manner (supplemental Fig. 4a, b). However, cotreatment of GW9662 has no rescue effect on TPA-induced promoter activation of the aforementioned full-length promoters at 4h when PPAR-g was not yet induced (supplemental Fig. 4c). Consistently, the downregulation of MMP9 and ZEB1 protein in TPA-treated HCC340 at 12h can be rescued by co-treatment of GW9926 to a level close to those at 6 h (supplemental Fig. 1b, lower panel). Moreover, overexpression of PPAR-g (using a PPAR-g expression plasmid, p-PPAR-g) suppressed the TPA-induced activations of full-length FN and LEF promoters, FNpro1938 and LEFpro1897, but not 5'-truncated mutants FNpro568 and LEFpro1231 (without PPAR-g region) by 60-70% (Fig. 4c). Prior transfection of p-PPAR-g also prevented TPA-induced activation of full-length promoter of FN, LEF, MMP9 and ZEB1 (FNpro1938, LEFpro1897, MMP9pro1920 and ZEB1pro2052, respectively) at 12 h in HCC340 by 40-50% (Fig. 4d). The same phenomenon was also observed in HepG2 (data not shown).
In summary, PPAR-g is responsible for the negative feedback against the SNA-upregulated transcription for FN, LEF, MMP9 and ZEB1 at late stage of TPA treatment.

ChIP and EMSA assay validated the binding of key transcription factors on putative regions
Thus far, it appears that SNA coupled with EGR1/SP1 activate transcription of the aforementioned mesenchymal genes, and PPAR-g plays as a negative feedback role, whether the indicated putative regions can be bound by the relevant transcriptional factors in these processes was examined, using ChIP assay. As shown in Fig. 5a (upper panel), TPA can induce binding of SNA on a FN promoter fragment (FN280), containing sequences of the critical SNA coupled with EGR1/SP1 region, at 4 h, further increased at 6h and returned to basal level at 12 h in HCC340. Also, TPA can induce binding of EGR1 on the same promoter fragment at 2-4 h, further increased at 6 h and declined at 12 h. In addition, TPA can induce sustained binding of SP1 on the same fragment during 4-12 h (Fig. 5a, upper panel). Similarly, TPA-induced binding of SNA, EGR1 and SP1 on the LEF promoter fragment (LEF280) containing the critical SNA coupled with EGR1/SP1 region begin at 2 h, further increased at 4 h and gradually declined during 6-12 h (Fig. 5a, lower panel). Moreover, GW9662 recovered the binding of SNA on FN280 and LEF280 at 12h (Fig. 5a, last lane), consistent with the rescue effect of this PPAR-g inhibitor on TPAinduced promoter activation (Fig. 4b) and expression (supplemental Fig. 1 ChIP SNA), and SNA with SP1 (by 1 st ChIP SP1, 2 nd ChIP SNA) on FN280 at 4 h, further increased at 6 h and declined at 12 h (Fig. 5b, upper panel). Similarly, TPA induced association of SNA with EGR1, and SNA with SP1 at 2 h on LEF280, further increased at 4-6 h and declined at 12 h on LEF promoter (Fig. 5b,  lower panel). Moreover, the ChIP and double ChIP assay for binding of the aforementioned transcriptional factors on both promoters were quantitatively analyzed by Q-PCR. As shown in Fig. 5c ( Also, double ChIP assay demonstrated 5.5-and 8.5-fold increase of association of (SNA with EGR1) and (SNA with SP1) on FN280, respectively, at 6 h of TPA treatment (Fig. 5c, upper right panel). Similarly, TPA can induce association of (SNA with EGR1) and (SNA with SP1) on LEF280 by 3.5-and 3.0-fold (Fig. 5c, lower right panel), respectively, at 6 h. Moreover, the decrease of SNA binding on both FN280 and LEF280 at 12h can be rescued by cotreatment with GW9662 to those at 6 and 4h, respectively (Fig. 5c, left panel, last bar). On the other hand, TPA can induce binding of PPAR-g on the fragments containing putative region of PPAR-g on FN, LEF, MMP9 and ZEB1 promoter, FN290, LEF230, MMP-9 280, ZEB1 200, respectively, with maximal induction during 6-12h (Fig. 5d). As expected, TPA-induced binding of PPAR-g on the aforementioned promoter fragments can be greatly suppressed by GW9926 at 12 h (Fig. 5d).
We further con rmed the binding of SNA on their putative region in vitro by EMSA. Nuclear extract of HCC340 treated with TPA at different time points were incubated with biotin labelled probe containing the SNA binding motif on FN promoter (FN-Prob). As shown in Fig. 6a, a DNA-protein complex (revealed as a band shift) was observed in the time zero group. This was increased at 4 and 6 h by 1.5-and 2.2-fold, respectively, and declined at 12 h. Addition of 200-fold of the unlabeled wild type but not mutant type of probe reduced the amount of DNA-protein complex by 80-90% at 6 h. Moreover, preincubation of the EMSA mixture with SNA antibody but not IgG control resulted in a supershift of the DNA-protein complex, validating SNA was the protein bound with the DNA probe. The same phenomenon was also observed in EMSA of the DNA probe containing the critical SN1 binding motif on LEF promoter (LEF-Prob). As demonstrated in Fig. 6b, TPA induced band shift of the DNA-protein complex maximally at 4 h, and gradually decreased at 6-12 h. As expected, the TPA-induced band shift of the DNA-protein complex at 4h can be competed by unlabeled wild type but not mutant type probe (Fig. 6b). Moreover, a supper band shift of the protein DNA complex can also be observed if EMSA mixture from 4 h-TPA sample was preincubated with SNA Ab but not IgG control (Fig. 6b).
COX2 and COL1A1 were potentially transcriptionally upregulated by SN as the aforementioned mesenchymal gene.
Thus far, we have demonstrated four of the mesenchymal genes FN, LEF, MMP9 and ZEB1 transcriptionally upregulated by SNA in a same manner, ie, direct binding of SNA coupled with EGR1/SP1 on their putative regions, which can be negatively feedbacked by PPAR-g. To validate whether this can be a general model, we screen the mesenchymal genes that can be upregulated by SNA from PubMed and examine whether they contain the aforementioned SNA binding motif (TCACA) coupled with a downstream EGR1/SP1 overlapping region and an upstream PPAR-g binding motif on the promoter sequences, 2000 bp upstream of translational start site by using gene bank in NCBI. Among them, ten of the well-known mesenchymal genes were proposed ( Table 1). Most of the candidate mesenchymal genes such as COX2, vimentin, vitronectin, COL1A1, α-SMA, N-cadherin, Twist1 have been reported to be upregulated by SNA involved in tumor progression. For example, COX2, which greatly correlates with malignancy, can be upregulated by SNA in head and neck squamous cell carcinomas [24]. To begin with, we validated whether COX2 and COL1A1 can be induced by TPA. As shown in Q RT-PCR (supplemental Fig. 5), TPA increased mRNA of both COX2 and COL1A1 at 2h by 1.5-2.0-fold, further increased at 4h and reach maximum at 6h by 8 and 10-fold, respectively, and nally declined to basal level at 12h. Moreover, GW9662 signi cantly rescued the downregulation of mRNA of both COX2 and COL1A1 at 12h. These suggested that while SNA is essential for gene expression of COX2 and COL1A1, PPAR-g played as a negatively feedback regulator, the same mechanisms as we observed in FN, LEF, MMP9 and ZEB1. We further investigated whether transcription regulation of COX2 and COL1A1 is also the same using deletion mapping for promoter analysis. According to the Genomatix software, there are one proposed SNA binding region and a downstream EGR1/SP1, locate on promoter of COX2 (-1007 to -502 bp upstream of the translation start site) and COL1A1 ( -1122 to -934 bp upstream of the translation start site) (supplemental Fig. 3c, d). Also, there are one putative PPAR-g binding region upstream of SNA region, locating on promoter of COX2 (-1560 to -979 bp upstream of the translation start site) and COL1A1 (-1969 to -1090 bp upstream of the translation start site) (supplemental Fig. 3c, d). Accordingly, we constructed the full-length promoter plasmid of COX2 (COX2pro1560) and COX2pro1228 containing PPAR-g coupled with one SNA motif and EGR1/SP1 overlapping binding regions. Also, COX2pro979, 5' end truncated mutants lacking PPAR-g binding region, and COX2pro870 lacking PPAR-g and one SNA region were obtained. In addition, COX2pro362 and COX2pro175, with deletion of PPAR-g, one SNA region and the downstream EGR1/SP1 overlapping region, were also included. Strikingly, we found a 40-45% higher TPA-induced promoter activity of COX2pro979 (without PPAR-g region) compared with full-length COX2pro1560 and 1228 (with PPAR-g region) whereas the TPA-induced promoter activity of COX2pro870 and COX2pro362 decreased by 50 and 80%, respectively, compared with COX2pro979 in HCC340 (Fig. 7a,  upper panel). Moreover, GW9662 (at 2.5 and 5.0 mM) dose-dependently elevated the TPA-induced promoter activity of COX2pro1560 by 33 to 45%, a level close to that of COX2pro979 (Fig. 7b, upper  panel). The deletion mapping analysis for COL1A1 revealed the similar results. As demonstrated in Fig.   7a (lower panel), the TPA-induced promoter activity of full-length promoter of COL1A1 (COL1A1pro1969) containing PPAR-g and one SNA region was lower than the deleted mutant COL1A1pro1090 (lacking the 5' PPAR-g region) by 46%, whereas the activity of COL1A1pro644 without both SNA and PPAR-g regions was lower than that of COL1A1pro1090 by 60%. Moreover, GW9662 (at 2.5 and 5.0 mM) dosedependently elevated the TPA-induced promoter activity of COL1A1pro1969 by 39 to 46%, a level close to COL1A1 pro1090 (Fig. 7a, lower panel). Thus, we suggested SNA-mediated upregulation of promoter activation COX2 and COL1A1 is also negatively feed backed by PPAR-g, potentially in the same way as that of the aforementioned mesenchymal genes.

Discussion
Multifaceted mechanisms for SNA to upregulate target gene expression In the past decades, it appears that the mechanisms by which SNA upregulates gene expression of mesenchymal genes were more complicated than it downregulated epithelial markers. As demonstrated in this report, SNA upregulate mesenchymal genes such as FN, LEF, MMP9 and ZEB1 by directly binding on a consensus motif to activate target promoters in HCC. However, a lot of previous studies indicated SNA also upregulate mesenchymal markers in an indirect fashion. For examples, SNA mediated the TGFb-induced MMP9 activation by promoting the binding of SP-1/Ets-1 and nuclear factor kappaB (NFkappaB) to the proximal and distal promoter regions, respectively [15], in MDCK cell. Also, SNA upregulated transcription of ZEB1 by elevating gene expression of Twist and triggering the nuclear translocation of Ets1 [16]. Recently, it is emerging that SNA-regulated gene expression involves negative feedback circuit established by the cross talk between SNA and non-coding RNAs, including miRNAs, long non-coding RNAs, and circular RNAs [25,26]. Taken together, there are multifaceted mechanisms for SNA to upregulate target gene expression.

SNA directly upregulated transcription of FN and LEF in tumor progression
Previously, the role of SNA in upregulating transcription of FN and LEF has been intensively studied in different contexts. SNA can mediate TGFb-triggered LEF expression for EMT in MDCKII cells [27].
Overexpression of SNA elevated FN to trigger EMT of retinal pigment epithelial (RPE) cell involved in proliferative vitreoretinopathy [28]. Also, SNA mediated FGF2-dependent expression of FN involved in endothelial-mesenchymal transition leading to retrocorneal membrane formation and blindness [29]. However, the detailed mechanisms by which SNA directly transactivates FN and LEF has not been well established yet. One previous study demonstrated that SNA coupled with the p65 subunit of NF-κB and PARP1 bound to the FN1 promoter at proximal (-236/+72) region for activating FN1 transcription in melanoma cell [30]. Herein, we found SNA coupled with EGR1 and SP1 upregulate both FN and LEF via binding to a distal region (-700 to -547 and -1300 to -1211 bp, respectively) containing a critical TCACA sequence (proposed SNA motif) adjacent to an EGR1-SP1 overlapping region. This mechanism is in similar with what we have observed in the SNA-upregulated MMP9 and ZEB1 transcription [17].
PPAR-γ exerted a negative feedback on SNA upregulated mesenchymal gene PPAR-γ was well known to suppress EMT [31][32][33][34] and inhibit metastases of HCC [35]. PPAR-γ can also affect transcription of mesenchymal genes known to be suppressed by SNA. For example, PPAR-γ prevents the TGF-β-induced vimentin, N-cadherin, bronectin required for metastasis of lung cancer cells [32]. In this study, we found PPAR-γ can suppress the SNA-upregulated transcription of FN, LEF, MMP9 and ZEB1 at later stage of TPA treatment ( Fig. 4 and Fig. 5) suggesting PPAR-g play as a negative feedback regulator.
Mechanisms for SNA to directly upregulate mesenchymal gene transcription Thus far, we have demonstrated four of the mesenchymal genes, MMP9, ZEB1, FN and LEF transcriptionally upregulated by SNA in a same mechanism, ie, direct binding of SNA coupled with EGR1/SP1 on their putative regions, which can be negatively feed backed by PPAR-g. Remarkably, the promoter of most mesenchymal genes known to be upregulated by SNA (Table 1) contain the same sequence architecture: SNA binding motif coupled with downstream E/S overlapping and upstream PPAR-g region. Among them, COX2 and COL1A1 were found to be potentially regulated as the same way ( Fig. 7).

Conclusion
In summary, we proposed a general model for SNA to upregulate transcription of transcription of mesenchymal genes via collaboration with EGR1/SP1 and negatively feed backed by PPAR-g. In the future, this model will be validated after more of the candidate mesenchymal genes are found to be transcriptionally upregulated by SNA in the same way.