ATF3 downmodulates its new targets IFI6 and IFI27 to suppress the growth and migration of tongue squamous cell carcinoma cells

Activating transcription factor 3 (ATF3) is a key transcription factor involved in regulating cellular stress responses, with different expression levels and functions in different tissues. ATF3 has also been shown to play crucial roles in regulating tumor development and progression, however its potential role in oral squamous cell carcinomas has not been fully explored. In this study, we examined biopsies of tongue squamous cell carcinomas (TSCCs) and found that the nuclear expression level of ATF3 correlated negatively with the differentiation status of TSCCs, which was validated by analysis of the ATGC database. By using gain- or loss- of function analyses of ATF3 in four different TSCC cell lines, we demonstrated that ATF3 negatively regulates the growth and migration of human TSCC cells in vitro. RNA-seq analysis identified two new downstream targets of ATF3, interferon alpha inducible proteins 6 (IFI6) and 27 (IFI27), which were upregulated in ATF3-deleted cells and were downregulated in ATF3-overexpressing cells. Chromatin immunoprecipitation assays showed that ATF3 binds the promoter regions of the IFI6 and IFI27 genes. Both IFI6 and IFI27 were highly expressed in TSCC biopsies and knockdown of either IFI6 or IFI27 in TSCC cells blocked the cell growth and migration induced by the deletion of ATF3. Conversely, overexpression of either IFI6 or IFI27 counteracted the inhibition of TSCC cell growth and migration induced by the overexpression of ATF3. Finally, an in vivo study in mice confirmed those in vitro findings. Our study suggests that ATF3 plays an anti-tumor function in TSCCs through the negative regulation of its downstream targets, IFI6 and IFI27.


Introduction
Head and neck squamous cell carcinomas (HNSCCs) are one of the top ten most fatal cancers in the world. Twenty-five % of HNSCCs are oral squamous cell carcinomas (OSCCs) [1,2] of which nearly half are tongue squamous cell carcinomas (TSCCs). TSCCs are highly aggressive tumors that are prone to local recurrence and metastasis and are difficult to treat as indicated by a five-year survival rate of only 50% [3]. Many factors, including genetic predisposition, lifestyle (such as alcohol and tobacco consumption) and viral infection (human papillomavirus), are associated with the risk for TSCCs, however the detailed molecular mechanism involved in TSCC development is unclear [3,4]. More knowledge of the molecular mechanisms underlying TSCC development is therefore urgently required as a basis for new therapeutic strategies for the clinical treatment of TSCCs.
ATF3 (activating transcription factor 3), a member of the CREB (ATF/cyclic AMP-responsive element binding protein) family, is a key transcription factor involved in regulating cellular stress responses with different functions in different tissues [5,6]. The roles of ATF3 in tumor cells have been intensively investigated in recent years [6,7]. Interestingly, ATF3 can act as an oncogene or as a tumor suppressor gene, depending on the type of tumor and the cellular context [8][9][10][11][12]. For instance, ATF3 promotes tumor cell growth in skin, breast and lung cancer cells [13][14][15][16]. In contrast, ATF3 was found to be down-regulated in esophageal squamous cell carcinomas, hepatocellular cancers, prostate and colon cancers and thus acts as a tumor suppressor [17][18][19][20][21][22]. Recent studies have identified ATF3 as a downstream target of microRNAs miR-488 and miR-431, which play important roles in the progression of TSCCs [23,24]. However, the downstream effects of ATF3 in TSCCs have not been explored so far.
The present study investigated the role of ATF3 in the growth and migration of TSCC cells and explored its underlying molecular mechanism. The results demonstrate that ATF3 plays a tumor suppressing function in TSCCs through the negative regulation of its novel targets IFI6 and IFI27.

Lower levels of ATF3 are expressed in poorly differentiated TSCCs derived from clinical patients
In order to characterize the potential role(s) of ATF3 in TSCCs, the expression of ATF3 was analyzed by immunohistochemistry (IHC) in highly (n = 11), moderately (n = 8) and poorly differentiated TSCCs (n = 10), as well as in normal tongue tissues (n = 6), graded previously by histological analysis (Fig 1A). Immunohistochemical staining of ATF3 was carried out in those tumors and representative images of ATF3 staining in 3 different tissues (Cases 1, 2 and 3) from each group are shown in S1 Fig. One of them (Case 1 from each group) are shown at high magnification in Fig 1B. From the staining patterns visible in Figs 1B and S1, we can observe that ATF3 is localized in the nuclei of normal oral epithelial(E) cells but nuclear staining of ATF3 is decreased in poorly differentiated TSCCs (Figs 1B (red arrows) and S1). Interestingly, the cytoplasmic staining of ATF3 in TSCC cells, especially in highly differentiated TSCC cells, was also strong, suggesting there is a potential redistribution of ATF3 between the nucleus and cytoplasm during tumorigenesis and malignant progression (Figs 1B and S1). These results indicate that ATF3 has various expression patterns in different OSCCs. Therefore, we assessed the ATF3 mRNA expression levels in 5 oral SCC lines, and found that two of those 5 cell lines expressly relative high levels of ATF3 (CAL 27 and SCC-25) compared to the other three cell lines (SCC-4, SCC-9 and SCC022) (S2 Fig). We then stained ATF3 in CAL 27 and in SCC-9 cells as well as in normal oral derived keratinocytes (OHKCs) and found a significantly higher nuclear staining of ATF3, which colocalized with nuclear staining (DAPI, blue), in normal oral epithelial cells and slightly higher levels of ATF3 in the nuclei of CAL 27 cells compared to SCC-9 cells (red arrows, Fig 1C and Fig 1D and 1E, respectively, indicated that the percentage of cells positive for ATF3 in their nuclei was significantly higher in normal oral epithelial cells compared to tumor cells (Fig 1D and 1E). That result was validated by western blot analysis of ATF3 protein levels in nuclear fractions prepared from cultured epithelial and TSCC cells ( Fig  1F). To further confirm the lower expression of ATF3 in oral SCCs, we analyzed ATF3 expression in HNSCCs in the TCGA dataset using the UALCAN tool (http://ualcan.path.uab.edu) [25]. That analysis revealed a slightly lower expression level of ATF3 in primary tumors compared with normal tissues (S4 Fig), although the difference was not statistically significant. However, when we selected only TSCCs in the TCGA dataset with NIH GDC portal for analysis of ATF3 expression level (https://portal.gdc.cancer.gov), significantly lower mRNA expression levels of ATF3 were found in oral SCCs compared to normal tissues ( Fig 1G).
These data suggest that the expression of ATF3, especially ATF3 localized in nuclei, is downregulated in TSCCs, which suggests that ATF3 may play a tumor suppressing function in TSCC cells.

Loss of ATF3 promotes TSCC cell growth and migration
To further understand the role of ATF3 in TSCCs, we deleted the ATF3 gene in CAL 27 cells, which express relatively high levels of ATF3 (S2 Fig) using CRISPR genome editing. In order to validate the effect of ATF3 deletion, we designed two independent small guide RNAs (sgRNA1 and sgRNA2) to create frame-shift mutations in ATF3. Those sgRNAs resulted in the expression of a truncated form of ATF3 by sgRNA1 which specifically targeting the end region of the ATF3 exon right before the 3'-UTR, or in an efficient loss of ATF3 by sgRNA2, which targeting the end of the 5'-UTR of the ATF3 gene (S5A Fig). Both genome edited populations showed significantly increased growth compared to cells transfected with an empty vector (V) starting at day 3 after infection (Fig 2A). We didn't observe a further increase of cell growth after day 3, which could result from an unknown feedback mechanism due to the deletion of ATF3 in vitro. To verify this result, we inhibited ATF3 expression using the transfection of specific siRNAs, whose knockdown efficiencies were reported in our previous publication [13] and were also validated by qRT-PCR (S6A Fig) in these cells and produced the same phenotype (S6B Fig). Furthermore, cell migration was increased by sgRNA1 and by sgRNA2 as revealed by trans-well migration assays (Fig 2B and 2C) and by wound healing assays (Fig 2D  and 2E). These results were further validated using another TSCC line SCC-25 which has a Representative images of histological analysis (H&E stain) of highly, moderately and poorly differentiated clinical TSCC tumors and normal oral tissues. B. Representative images of immunochemical staining of ATF3 from one patient of each different grade of TSCC tissues from A, red arrows indicate nuclear staining of ATF3, the green dot lines indicate the border between epidermal (E) and mesenchymal (M) compartments in the normal oral or tumor tissues. Lower magnification images together with representative images of ATF3 staining of two other patients from each different grade of TSCCs are shown in S1 Fig. C. Representative merged images of immunofluorescence staining of ATF3 (red) and DAPI (blue) in cultured primary oral human keratinocytes (OHKCs), CAL 27 and SCC-9 cells, DAPI for nuclei staining. Red arrows indicate nuclear staining of ATF3. Separate images of ATF3 and DAPI staining are shown in S3 Fig. A-C: Scale bars = 100 μm. D. Quantification of the percentage of normal oral epithelial cells and tumor cells with nuclear ATF3 staining (red arrows in B) calculated by counting the number of ATF3 nuclear stained cells in a total of 200 epithelial or tumor cells in B. E. Quantification of the percentage of nuclear positive staining of ATF3 in C. F. Nuclear and cytoplasmic fractions prepared from the cells in C were used for western blot analysis of ATF3. Lamin B1 was used as a loading control for nuclear fractions, GAPDH was used as a loading control for cytoplasmic fractions. G. Normalized expression level of ATF3 (relative log expression) in normal tissues and in oral tumor tissues derived from ATGC data. D, E, F: � p<0.05; �� p<0.01; ��� p<0.005 when comparing two groups as indicated.
https://doi.org/10.1371/journal.pgen.1009283.g001 relatively high expression of ATF3 (S2 and S7 Figs). The inhibition of TSCC growth and migration by disruption of ATF3 expression was found to be correlated with the increased activation of AKT and ERK signaling pathways, as shown by significant increases of pAKT and pErk, while total levels of those factors were unaffected (red arrows, Fig 2F and 2G) and the efficient deletion of ATF3 was also confirmed ( Fig 2F). These data indicate that the loss of ATF3 function enhances TSCC cell growth and migration.

High expression levels of ATF3 suppress TSCC cell growth and migration
Next, we tested the effects of increasing ATF3 expression in oral SCC cell lines. We stably overexpressed ATF3 in SCC-9 cells, which showed relatively low expression levels of ATF3 (S2 Fig). The high expression level of ATF3 in SCC-9 cells infected with a retrovirus expressing ATF3 was confirmed (S5B Fig). Ectopic expression of ATF3 significantly suppressed the growth of SCC-9 cells (Fig 3A), and also inhibited their migration in Transwell assays (Fig 3B and 3C) and in wound healing migration assays (Fig 3D and 3E). These results were also validated using another TSCC cell line SCC-4 that overexpresses ATF3 (S8 Fig). The corresponding activation of AKT and ERK was also suppressed in SCC-9 cells overexpressing ATF3 (Fig 3F and 3G).
Taken together, our findings suggest that ATF3 negatively regulates the growth and migration of TSCC cells, possibly by downmodulating the activation of the AKT and ERK pathways.

ATF3 downregulates its downstream targets IFI6 and IFI27, which are highly expressed in TSCC cells
In order to understand the molecular mechanism by which ATF3 negatively regulates the growth and migration of TSCC cells, we analyzed the mRNA gene expression profile of CAL 27 cells with the deletion of ATF3 using RNA-seq. Gene Ontology (GO) enrichment analysis was performed and the Top 15 upregulated pathways were identified in the ATF3 deleted (SG1) group ( Fig 4A). The GO analysis showed that most of those pathways are cytokine and chemokine regulated pathways, including "regulation of immune cell migration", "response to virus/bacterium" and "response to type I interferon or interferon gamma", as well as "epidermal cell differentiation/development" (Fig 4A). Gene cluster analysis revealed the upregulation of cytokine genes including known ATF3 target genes, such as IL-6 and IL-1β, and agreed with the GO analysis ( Fig 4B). Two members of the IFI (interferon (IFN) alpha-inducible) gene family, IFI6 and IFI27, were identified as the Top upregulated genes in the ATF3-deleted cells ( Fig 4B). Since interferons, especially the cytokine interferon alpha (IFNα), have been shown to mediate diverse immune responses to tumors [26], we further investigated the roles of IFI6 and IFI27 in ATF3 deleted and in ATF3 overexpressing TSCC cells.
The upregulation of IFI6 and IFI27 was validated by RT-PCR analysis both in SG1 and in SG2 ATF3 deleted cells ( Fig 4C). Conversely, the lower expression of both IFI6 and IFI27 was observed in SCC-9 cells that overexpress ATF3 compared to control cells (Fig 4D). To investigate whether IFI6 and/or IFI27 are direct targets of ATF3, we analyzed published ATF3 CHIPseq data (http://cistrome.org) and found that both IFI6 and IFI27 are putative targets of ATF3 in several cell types, including colon cancer cells, liver epithelial cells, erythroblasts, embryonic stem cells and skin dermal fibroblasts [27][28][29][30]. To determine whether ATF3 directly targets IFI6 and/or IFI27 in TSCC cells, CHIP assays were performed. Those results confirmed that ATF3 can directly bind predicted ATF3 binding sites in the IFI6 and IFI27 promoter regions (Figs 4E, 4F and S9). That binding could be reduced by ATF3 deletion or enhanced by ATF3 overexpression (Fig 4E and 4F), but was not detected by control non-immune IgGs (S9 Fig). Taken together, these results suggest that ATF3 directly targets IFI6 and IFI27 to negatively control their expression in TSCCs.
IFI27 has already been reported to be highly expressed in oral SCCs [31]. We checked the expression level of IFI6 in TSCCs using IHC, and found high expression levels of IFI6 in clinical tumor tissues as well (Fig 4G and 4H). Importantly, TCGA data analysis further revealed that both IFI6 and IFI27 are highly expressed in TSCCs compared to normal tissues ( Fig 4J). These data suggest that IFI6 and IFI27, which are direct targets of ATF3 and are negatively regulated by ATF3, are upregulated in TSCCs.

ATF3 suppression of TSCC growth and migration in vitro depends on the expression of IFI6 and IFI27
Next, we tested whether ATF3 modulates the growth and/or migration of TSCC cells through the regulation of IFI6 and/or IFI27. First, we confirmed the knockdown efficiency of IFI6 and  . The results showed that the knockdown of IFI6 or IFI27 significantly blocked the increased growth of CAL 27 cells induced by the deletion of ATF3 (Fig 5A), and the knockdown efficiency of IFI6 and IFI27 in ATF3-deleted cells was also validated ( Fig 5B). Next, we used a lentivirus expression vector to overexpress IFI6 and IFI27 in SCC-9 cells, then infected those cells with a retrovirus expressing ATF3. We found that the overexpression of IFI6 or IFI27 together with ATF3 (ATF3+IFI6 or ATF3+IFI27) counteracted the reduced growth and migration of SCC-9 cells with only the overexpression of ATF3 (ATF3+GFP) (Fig 5C), and the high expression levels of IFI6 or IGI27 were also verified by qRT-PCR ( Fig 5D).
To analyze whether ATF3 regulates TSCC migration depending on the expression of IFI6 or IFI27, in vitro wound healing assays were carried out with ATF3 deleted cells with either knockdown IFI6 or IFI27 (SG1+siIFI6 or SG1+siIFI27) (the knockdown efficiencies of IFI6 or IFI27 are shown in S10 Fig) or the overexpression of IFI6 or IFI27 in ATF highly expressing cells (ATF3+IFI6 or ATF3+IFI27). We found that knockdown of either IFI6 or IFI27 inhibited the migration increased by the deletion of ATF3 (Figs 5E and S11A) and that the forced expression of IFI6 or IFI27 could rescue the migration inhibition induced by the high expression of ATF3 (Figs 5F and S11B). The enhancement of TSCC migration by IFI6 and by IFI27 was further confirmed using trans-well migration assays (Figs 5G and S11C). Importantly, the corresponding changes of AKT or ERK activity induced by the deletion or overexpression of ATF3 were counteracted by IFI6 or IFI27 expression (Fig 5H-5K).
These data suggest that ATF3 modulates the growth and migration of TSCC cells through the downregulation of IFI6 and IFI27.

ATF3 negatively modulates TSCC growth through the regulation of IFI6 and IFI27 in vivo
To validate the biological role of ATF3 in TSCCs through IFI6 and IFI27 in vivo, we first tested whether the inhibition of ATF3 promotes tumor growth of TSCCs in mice. CAL 27 cells with or without the CRISPR-Cas9 mediated deletion of ATF3 were xenografted on the dorsal skin of nu/nu mice. One month after grafting, the mice were sacrificed and the grafted tumors were collected (S12A Fig). ATF3-deficient TSCC cells (SG1 and SG2) had formed significantly larger tumors than control cells (V) (Fig 6A and 6B). The proliferation and differentiation status of the tumors formed were then evaluated by IHC. Ki-67 staining showed that there were many more Ki-67 positive cells (red arrows) in tumor tissues with the deletion of ATF3 (SG1 or SG2) (Figs 6C, 6D and S12B). All tumor cells from the different groups were positive for the epidermal basal cell marker keratin 14 (K14) (Figs 6C, 6D and S12C), which is normally expressed in the oral stratum basal layer (undifferentiated cells) [32]. IHC of keratin 4 (K4) and keratin (K13), which have been reported to be expressed in normal oral differentiating suprabasal cells [32], but not in poorly differentiated oral SCCs [33], was positive but neither of those markers was clearly expressed in TSCC cells from any of the groups (S12D-S12F Fig).
These data indicate that the loss of ATF3 could significantly enhance TSCC growth in vivo.
In contrast, ATF3 overexpressing SCC-9 cells (ATF3+GFP) generated smaller tumors than SCC-9 cells transfected with an empty vector (NEO+GFP) (Figs 6E, 6F and S13A). The effects of ATF3 overexpression on tumor growth were diminished by the overexpression of IFI6 or of IFI6 expression in 6 clinical TSCC tumors and in 6 normal tissues. H. Quantification of IFI6 staining (n = 6) in G. I. IFI6 expression in TSCC tissues and in normal tissues from the ATGC database. J. IFI27 expression in TSCC tissues and in normal tissues from the ATGC database. Both I and J are shown with relative log expression for the Y-axis. C, D, H-J: � p<0.05, �� p<0.01 and ��� p<0.005 when compared with the corresponding control group as indicated.
https://doi.org/10.1371/journal.pgen.1009283.g004 RT-PCR analysis of IFI6 or IFI27 from TSCC cells pre-infected with a CRISPR/Cas 9 ATF3 virus, then transfected with siRNAs of IFI6 or IFI27, at 48 h after transfection. C. Growth of SCC-9 cells infected with a retrovirus overexpressing ATF3 (ATF3) plus stable expression of IFI6 (ATF3+IFI6) or IFI27 (ATF3+IFI27) or GFP(ATF3+GFP) and with expressing neomycin and GFP as a control (NEO+GFP) were analyzed using a CCK8 kit at different time points. D. RT-PCR analysis of IFI6 or IFI27 from TSCC cells at 48 h after infection with ATF3 expressing virus together with either IFI6 or IFI27 expressing retrovirus as in C. E, Wound healing migration assays were performed with CAL 27 cells in the same groups as in A, the healing IFI27 (ATF3+IFI6 or ATF3+IFI27), which were further supported by staining of the cell proliferation marker Ki-67 (red arrows), which showed that the overexpression of ATF3 significantly suppressed the proliferation of tumor cells, and that suppression was blocked by the stable expression of IFI6 or IFI27 (Figs 6G, 6H and S13B). All tumor cells from the different groups expressed K14, however, we found a lower expression of K14 in TSCC cells induced by the high expression of ATF3, but not in cells with co-expression of ATF3 and IFI6 or IFI27 (Figs 6G, 6H and S13C). Interestingly, we also observed that higher levels of K4 and K13 were expressed in ATF3 over-expressing cells (red arrows), but again not in cells co-expressing ATF3 and IFI6 or IFI27 (Figs 6K-6N and S13D and S13E). These data suggest that ATF3 can inhibit tumor cell growth and affect TSCC cell differentiation in vivo.
Taken together, these results suggest that ATF3 negatively modulates TSCC tumor growth and differentiation in vivo through the expression level of IFI6 or IFI27.

Discussion
ATF3, a stress-response gene, is expressed at low levels in normal cells but can be induced by a variety of signals including hypoxia, anoxia, DNA damage and chemicals to play important roles in cellular processes that respond to signals disrupting homeostasis such as tissue damage [5]. The role of ATF3 in cancers has been extensively investigated and accumulated evidence has demonstrated that ATF3 can play tumor promoting or tumor suppressing functions depending on the type of tumor cell and the stromal context [10,12,13,[15][16][17][18], and particularly, on the conditions of the cells as well [6]. For instance, the loss of ATF3 was found to promote prostate cancer development, but the overexpression of ATF3 enhanced the metastasis of prostate cancer [34,35]. ATF3 has been reported to play dichotomous roles by enhancing normal mammary epithelial cell apoptosis and by promoting breast cancer cell proliferation [8,36] and a similar situation was observed in TSCCs. In this study, IHC analysis revealed that ATF3 is localized in the nuclei of normal tongue epithelial cells and that nuclear expression levels of ATF3 are negatively correlated with the differentiation status of clinical TSCCs. This finding is inconsistent with two recent studies that showed higher expression levels of ATF3 in clinical TSCCs compared to normal tissues [23,24]. This difference in results is likely due to the conditions of the clinical tumors. Interestingly, the overexpression of ATF3 driven by the bovine keratin 5 (BK5) promoter in mice didn't result in the development of TSCCs, but did elicit the formation of oral mucosa SCCs [37], indicating a complicated role of ATF3 in oral SCCs. Studies by Shi et al. and by Hu et al. identified ATF3 as a downstream target of microRNAs that play important roles in TSCCs [23,24], but the direct effect of ATF3 on TSCCs has not been fully explored. Moreover, neither of those studies investigated the subcellular localization of ATF3 in TSCC cells. Both the nuclear and cytoplasmic localization of ATF3 has been reported in different situations related to the cell type [38], and the nuclear localization of ATF3 has been linked to its activation [39][40][41]. From our observations, we propose that ATF3 loses its nuclear localization with the progression of TSCCs. Importantly, analysis of the ATGC database showed that there was a significantly lower expression of ATF3 in tumor samples compared to normal tissues (P = 0.0037, Fig 1D), which supports our findings. Our gain-or loss-percentage at 24 h after wounding is shown and images of wound healing are shown in S11A Fig. F. Wound healing migration assays were performed with SCC-9 cells in the same groups as in C, the healing percentage at 24 h after wounding is shown, and images of wound healing are shown in S11B Fig. G. Transwell migration assays were performed with cells in the same groups as in C. The numbers of migrated cells in the different groups are shown, and images of migrated cells at 24 h are shown in S11C Fig. H. Western blot analysis of the indicated protein levels in cells in the same groups as in A. I. Quantification of phosphorylated AKT (p-AKT) and ERK (p-ERK) levels normalized with the total levels of the corresponding proteins in H. J. Western blot analysis of the indicated protein levels in cells in the same groups as in C. K. Quantification of phosphorylated AKT (p-AKT) and ERK (p-ERK) levels normalized with the total levels of the corresponding proteins in J. � A-K: � p<0.05, �� p<0.01, ��� p<0.005, ���� p<0.001 compared with the corresponding control groups as indicated.
https://doi.org/10.1371/journal.pgen.1009283.g005 of function analyses of ATF3 in four different TSCC cell lines (Figs 2 and 3) showed that ATF3 negatively modulates the growth and migration of TSCC cells associated with activation of the AKT/ERK pathways. Furthermore, suppression of TSCC growth by ATF3 was verified in an in vivo mouse model, which also showed that ATF3 could promote TSCC cell differentiation ( Fig  6K-6N), agreed to GO analysis result of RNA-seq data (Fig 4A) and the induction of differentiation has been considered to be an anti-tumor function of ATF3 [42]. Therefore, we conclude that ATF3 plays an anti-tumor function in human TSCCs.
Consistent with our study, the tumor suppressing function of ATF3 has been reported in multiple studies with different types of tumors. ATF3 was found to be downregulated in several human cancers, including liver cancer, colon cancer, bladder cancer, prostate cancer and other epithelial cancers [12,17,18,[20][21][22]. It has been shown that ATF3 negatively regulates the progression of different cancer cells through the modulation of different downstream pathways. As an example, in skin keratinocytes, ATF3 targets p53 to suppress cellular senescence to promote keratinocyte tumorigenesis [13]. However, ATF3 suppresses esophageal SCC growth by downregulating ID1 (inhibitor of DNA binding 1), blocks the metastasis of bladder cancer by regulating gelsolin-mediated remodeling of the actin cytoskeleton and inhibits the progression of hepatocellular carcinoma cells by upregulating CYR61 expression. Recently we reported that ATF3 suppresses expression of the cytokine IL-6 in dermal fibroblasts to inhibit melanoma cell growth through paracrine pathways [43]. Taken together, these results suggest that ATF3 negatively regulates tumor progression probably through specific downstream targets. In our study, RNA-seq analysis revealed that "response to type I and type II interferon" pathways were upregulated in ATF3 deleted TSCCs, and two IFN inducible genes, IFI6 and IFI27, were upregulated in ATF3-deleted TSCCs, and were downregulated in ATF3-overexpressing cells. Both IFI6 and IFI27 are type I interferon alpha inducible proteins, which are encoded by genes with high expression levels in microarray analysis of OSCCs compared to normal tissues [44]. However, the detailed functions of IFI6 and IFI27 have not been fully characterized. Both IFI6 and IFI27 were highly expressed in our clinical TSCC tumors, and again, that finding was supported by analysis of the TCGA database. Importantly, our in vitro study showed that the overexpression of IFI6 or IFI27 could counteract the inhibition of OSCC cell growth and migration induced by ATF3 overexpression and vice-versa, the knockdown of IFI6 or IFI27 blocked increased cell growth and migration with corresponding activation of the ERK/AKT pathways induced by the deletion of ATF3. and this finding was further verified using an in vivo mouse model. IFNs, a well-known family of cytokines, play an essential role in antiviral effects, and the function of IFNs modulate cellular proliferation and immune responses mainly through interferon stimulating genes (ISGs). Both IFI6 and IFI27 are type I IFNalpha ISGs, and increasing evidence shows that IFNalpha plays dual opposing roles in cancer development based on ISGs, which determine whether it has anti-or pro-tumorigenic functions. IFI27 has been reported images are shown S13B-S13E Fig. H, J, L, N. Corresponding quantification of G, I, K and M as labeled. All quantification data were from 6 tumors of each group (n = 6): � p<0.05, �� p<0.01, ��� p<0.005 compared with the two corresponding control groups as indicated. All scale bars = 100 μm.
https://doi.org/10.1371/journal.pgen.1009283.g006 to be involved in different biological processes including the regulation of cellular apoptosis; modulation of the transcriptional activity of nuclear receptors NR4A1, NR4A2 and NR4A3; involvement in the ubiquitin-mediated proteasomal degradation pathway and playing antivirus functions [45][46][47][48][49], IFI27 has been shown to inhibit Heptatis C virus replication by through the activation of JAK/STAT signaling pathway [50], which has been reported to be interact with AKT and MAPK/ERK pathways and to be associated with cancer development [51,52]. IFI27 has been also shown to play a tumor promoting function in multiple cancers and the knockdown of IFI27 was reported to inhibit the proliferation and invasion of TSCC cells [31], but the underlying molecular mechanism by which IFI27 regulates tumor cell growth and migration is still not clear. IFI6, has been shown to play anti-apoptosis role in gastric cancer cells, breast cancer cells, human myeloma cells and vascular endothelia cells mainly through inhibiting activation of caspase 3 and suppressing proapoptotic gene BCL-2 expression [53][54][55][56], and anti-apoptosis plays a crucial role for tumor development. Especially, IFI6 was recently reported to promote metastatic potential of breast cancer cells through mitochondrial ROS (mtROS) [57]. Therefore we conclude that ATF3 plays an anti-tumor function in TSCCs likely through the negative regulation of expression of its downstream targets IFI6 and IFI27. Future studies will be aimed at characterizing whether IFI6 and IFI27 control the growth and migration of TSCCs through regulation of JAK/STAT, anti-apoptotic or other potential downstream signaling pathways.
Oral human keratinocytes (OHKCs) were isolated from tongue tissues according to a previously published protocol for isolation of keratinocytes from skin tissues [59]. The OHKCs were cultured in keratinocyte serum free medium (K-SFM, Thermo Fisher Scientific, #10725018) in culture dishes pretreated with a coating matrix containing type-I collagen (Gibco, R-011-K) and the medium was changed every 2 days.

Tumor tissues from patients
Clinical normal tongue and TSCC tissues were obtained from patients who had provided verbal and written informed consent at the Stomatological Hospital, Shandong University. The procedures for obtaining those tissues were approved by the Medical Ethical Committee of the School of Stomatology, Shandong University (Protocol No. 20150401, Date: 12-05-2015).

Virus production and infection
Retrovirus and lentivirus preparations and infections were performed as previously described [13]. Viral vectors with corresponding packaging vectors pCL-Eco (Addgene, Plasmid #12371), for retrovirus, and pMD2.G (Addgene, Plasmid #12259) together with psPAX2 (Addgene, Plasmid #12260) for lentivirus, were transfected into 293T cells using Lipofectamine 2000 (Invitrogen, Carlsbad, USA, #L3000015). Starting one day after transfection, the virus-containing supernatants were collected daily for three days and were stored at -80˚C. For virus infection, 3 x 10 5 TSCC cells were seeded into 10 cm dishes. The following day, 6 ml virus-containing medium with 8 μg/ml polybrene were added to the culture medium. After 6 h incubation, the virus-containing medium was removed and changed to normal growth medium. After 48 h of infection, the infected TSCC cells were selected by adding 1 μg/ml puromycin into the growth medium for another 6 days. Surviving cells were analyzed by western blot analysis to validate the deletion or overexpression efficiency of the corresponding genes.

Cell proliferation assays
Cell proliferation assays were carried out using a Cell Counting Kit-8 (CCK-8; Dojindo, Tokyo, Japan). Three x 10 3 cells per well were plated into 96-well plates and were cultured as indicated in the Figures. At 0, 2, 3, 4, 5 and 6 days, 10 μl CCK-8 working solution were added to each well, and incubated for 1.5 h at 37˚C. Optical densities were then measured at a wavelength of 450 nm using a microplate reader (Spectrostar Nano, BMG Labtech, Offenburg, Germany).

Trans-well migration assays
Cell migration assays were performed using 8.0 μm pore size Transwell TM permeable supports (Corning Costar, Lowell, MA). Three x 10 4 TSCC cells were seeded in the upper chamber of each insert, which was placed into 6-well plates with medium containing 5% FBS. At 24 h, cells remaining on the upper filter were scraped off gently using a cotton swab and the inserts were gently washed with PBS. Cells that had migrated to the lower chamber were fixed with 4% paraformaldehyde for 10 min and then stained with 0.5% crystal violet solution for 20 min. Filters were then imaged in a microscope. Five non-overlapping, representative images were captured for each insert and were used to manually count the number of cells. Results are presented as the mean number of cells per field ± standard deviation.

Wound-healing assays
Wound-healing assays were used to evaluate the migration ability of TSCC cells under conditions indicated in the Figures. Briefly, 8 x 10 5 cells per well were seeded in 6-well plates. When the cells reached nearly 90% confluence, the cell layer was scratched with a sterile 200 μl pipette tip, then washed twice with PBS and cultured with medium containing 1% FBS. At 0,12 and 24 h, photographic images of the plates were acquired using a microscope.

Quantitative RT-PCR (qRT-PCR) assay
Total RNAs were extracted from cells using an RNA Extraction kit (Takara Bio Inc., Kusatsu, Shiga Prefecture, Japan) according to the protocol of the manufacturer. The RNAs were dissolved in nuclease-free water and the concentrations were measured using a NanoDrop spectrophotometer. One μg of each total RNA was reverse transcribed to complementary DNA (cDNA) using a Takara PrimeScriptTM RT kit (Takara Bio Inc.). PCR reactions were performed with Takara SYBRR Premix Ex TaqTM II (Takara Bio Inc.) using a LightCyclerR 480 II (Roche Diagnostics Ltd., Rotkreuz, Switzerland). One hundred ng cDNA and 250 nmol gene specific primers were used for amplification in 20 μl qRT-PCR reactions. The PCR reactions were carried out at 95˚C for 30 s followed by 40 repeated cycles of 95˚C for 5 s and 60˚C for 20 s and then were terminated with an elongation step for 15 s at 72˚C. Data were acquired and analyzed with a LightCycler 480 system. The PCR primers used are listed in S1 Table.

Chromatin immunoprecipitation (ChIP) assays
Chromatin immunoprecipitation (ChIP) assays were performed as described in the EpiQuik™ Chromatin Immunoprecipitation Kit protocol. Briefly, 1 x 10 6 CAL 27 cells were cross-linked with 1% formaldehyde, sonicated, precleared and then incubated with 2-5 mg antibody per reaction (rabbit monoclonal anti-ATF3 antibody (1:100), Abcam, Cambridge, MA, USA, #ab207434). Immune complexes were washed with low and high salt buffers, and the DNA was extracted and precipitated. The enrichment of the DNA template was analyzed by conventional PCR and qRT-PCR, using primers specific for the IL-27 and IL-6 gene promoters. The primers used are listed in S2 Table. The conventional PCR products were analyzed by gel electrophoresis, and the qRT-PCR followed the above procedure.

Western blot analysis
To prepare whole cell protein extracts, the cells were washed twice with ice-cold PBS and then lysed with RIPA buffer containing 1% phenylmethylsulphonyl fluoride (PMSF) for 30 min on ice, followed by centrifugation at 12,000 r.p.m. at 4˚C for 10 min. The supernatants were collected, and protein concentrations were quantified using a BCA protein quantitation kit (Solarbio, Beijing, China). Equal amounts of protein per lane were separated on 10% SDS-PAGE gels (Beyotime, Shanghai, China) and were then electrotransferred to polyvinylidene fluoride (PVDF) membranes (Invitrogen) following the manufacturer's protocol. After blocking with 5% bovine serum albumin, the membranes were incubated with the primary antibody overnight at 4˚C with gentle shaking. The next day, blots were washed 3 times for 10 min each with 1xTBST (20 mM Tris-HCl, 150 mM NaCl), containing 0.05% (v/v) Tween-20 and then incubated with the secondary antibody for 2 h at room temperature. Finally, detection was performed using enhanced chemiluminescence reagents (Millipore, Billerica, MA, USA, #1702303). The following primary and secondary antibodies were used: rabbit monoclonal anti-ATF3 antibody (1:1000, #ab207434)

Histology and Immunofluorescence analysis
Formalin-fixed tumor issues were embedded in paraffin and 10 μm sections were prepared for histological analysis following standard protocols. IHC analyses were performed as follows: each section was deparaffinized in xylene and post-fixed in ice-cold acetone, incubated with blocking buffer (2% bovine serum albumin and 5% donkey serum in PBS with 0.1% Triton 100) at room temperature for 1 h, and then incubated with the primary antibody at 4˚C overnight. On the second day, after removing the primary antibody solution and washing with PBS, each specimen was incubated with the secondary antibody at room temperature. After 1 h incubation, the sections were washed with PBS and stained with DAB (ZSGB-BIO) for 5 min. The following primary and secondary antibodies were used: rabbit monoclonal anti-

siRNA transfection
Transfection of siRNAs into TSCC cells followed the standard protocol. Briefly, TSCC cells at 70% confluence were transfected with 20 nM various siRNAs using Lipofectamine 3000 (Invitrogen). Seventy-two h later, the cells were collected for analysis by RT-PCR to determine knockdown efficiency. siRNA oligos that target IFI6, IFI27 and 36beta4 (used as an internal control) or a negative control scramble siRNA (siCtrl) were purchased from Gene Pharma (Shanghai, China). Three independent siRNAs for each gene were tested and the oligo sequences of all siRNAs are shown in S3 Table. RNA sequencing (RNA-seq) and data analysis RNA sequencing of control (V) and ATF3-knockout CAL 27 cells (SG1) were performed using a BGISEQ-500 sequencing system at BGI (Shenzhen, China). Gene expression levels were normalized and quantified using the RSEM tool [61]. Differentially expressed genes (DEGs) between the V and SG1 groups were screened using the PossionDis method [62].

In vivo tumor formation
To test whether the loss of ATF3 enhances TSCC tumor growth in vivo, 8-week-old female nu/ nu nude mice (Crl:NU-Foxn1 nu Immunodeficient Outbred)(Beijing Vital River Laboratory Animal Technology Co., Ltd., Beijing, China) were randomly assigned to three groups: a control group (n = 6, Cal 27 cells infected with CRISPR empty vector lentivirus, V), a SG1 group (n = 6, Cal 27 cells infected with CRISPR-sgRNA1 lentivirus) and a SG2 group (n = 6, Cal 27 cells infected with CRISPR-sgRNA2 lentivirus). To test whether ATF3 affects TSCC tumor growth through the regulation of IFI6 and/or IFI27, nude/nude mice were randomly divided into the following four groups: a control NEO group (n = 6, SCC-9 cells infected with neomycin expressing retrovirus and an empty vector lentivirus, NEO), an ATF3 group (n = 6, SCC-9 cells infected with ATF3 expressing retrovirus and an empty vector lentivirus), an ATF3+IFI6 group (n = 6, SCC-9 cells infected with ATF3 expressing retrovirus and an IFI6 expressing lentivirus) and an ATF3+IFI27 group (n = 6, SCC-9 cells infected with ATF3 expressing retrovirus and an IFI27 expressing lentivirus). One x 10 6 TSCC cells were injected subcutaneously into the dorsal skin of mice as described previously [13] and each mouse received a total of 4 injections. Tumors were observed and measured every day, and one month later, the mice were euthanized and their tumors were collected and weighed.
All animal procedures were performed in accordance with National and International Animal Welfare Regulations and with approval from the Stomatology School of Shandong University Committee on Animal Care.

Statistical analysis
Statistical analysis was performed using GraphPad Prism 7 and data are presented as means ± standard error of the mean (SEM). All experiments were performed at least three times. Student's t-test was used to compare two groups and one-way or two-way ANOVA analysis were used for comparison of more than 2 groups. A p<0.05 is considered to be statistically significant, and p values are labeled in the Figures.  Fig. Expression of ATF3 is mediated by CRISP-cas9 and a retro-expressing vector. A. CAL 27 cells infected with a lentivirus carrying CRISPR/Cas9 with ATF3 sgRNA1 (SG1) or sgRNA2 (SG2) or with an empty vector as a control (V). After selection with puromycin, the cells were collected for western blot analysis of ATF3 protein to determine the deletion efficiency. GAPDH was used as a housekeeping gene for a loading control. B. SCC-9 cells were infected with a retrovirus expressing ATF3 (ATF3) or neomycin (NEO) as a control; 48 h after infection, the cells were collected for RT-PCR (left panel) and western blot analysis for ATF3 expression. Relative mRNA levels of ATF3 were normalized with the 36beta4 gene, and GAPDH was used as a loading control for the protein level. ��� p<0.005 compared with the control group.  Fig. ATF3 directly binds the promoter regions of the IFI6 and IFI27 genes. A,B. Extracts of CAL 27 cells were processed for CHIP assays with the anti-ATF3 antibody and nonimmune IgG followed by PCR analysis of either the IFI6 (A) or the IFI27 promoter regions containing ATF3 binding sites as shown (maps) in Fig 4E and 4F. Gel electrophoresis (1% agarose gel) shows the PCR amplified 160 bp fragment for IFI6 and the 463 bp fragment for IFI27. Labels for lanes: M: molecular markers, 1: Input, 2,3: ATF3 antibody; 4: IgG. C,D. Extracts of CAL 27 cells with either the deletion of ATF3 (SG1 or SG2) or the overexpression of ATF3 (ATF3) and the corresponding controls (V or NEO) were processed for CHIP assays with the anti-ATF3 antibody or non-immune IgG followed by PCR analysis of either IFI6 (C) or IFI27 (D) promoter regions containing ATF3 binding sites. PCR products were analyzed by gel electrophoresis. Labels for lanes: M: molecular markers, 1: Input, 2: V+ATF3 antibody, 3: SG1 +ATF3 antibody, 4: SG2+ATF3 antibody; 5: IgG, 6: NEO+ATF3 antibody, 7-8: overexpression of ATF3+ATF3 antibody. (JPG) S10 Fig. siRNA-mediated knockdown of IFI6 or IFI27 expression in TSCC cells. A, B. CAL 27 cells were transfected with 3 independent siRNAs of IFI6 (A) or IFI27 (B) or a scrambled siRNA (siCtrl). 72 h after transfection, the cells were collected for RT-PCR analysis of IFI6 and IFI27 expression. The relative mRNA levels of IFI6 or IFI27 were normalized with the 36beta4 gene, �� p<0.01, ��� p<0.005 compared with the control group (siCtrl) as indicated.