The Forkhead box F1 transcription factor inhibits collagen deposition and accumulation of myofibroblasts during liver fibrosis

ABSTRACT Hepatic fibrosis is the common end stage to a variety of chronic liver injuries and is characterized by an excessive deposition of extracellular matrix (ECM), which disrupts the liver architecture and impairs liver function. The fibrous lesions are produced by myofibroblasts, which differentiate from hepatic stellate cells (HSC). The myofibroblast’s transcriptional networks remain poorly characterized. Previous studies have shown that the Forkhead box F1 (FOXF1) transcription factor is expressed in HSCs and stimulates their activation during acute liver injury; however, the role of FOXF1 in the progression of hepatic fibrosis is unknown. In the present study, we generated αSMACreER;Foxf1fl/fl mice to conditionally inactivate Foxf1 in myofibroblasts during carbon tetrachloride-mediated liver fibrosis. Foxf1 deletion increased collagen depositions and disrupted liver architecture. Timp2 expression was significantly increased in Foxf1-deficient mice while MMP9 activity was reduced. RNA sequencing of purified liver myofibroblasts demonstrated that FOXF1 inhibits expression of pro-fibrotic genes, Col1α2, Col5α2, and Mmp2 in fibrotic livers and binds to active repressors located in promotors and introns of these genes. Overexpression of FOXF1 inhibits Col1a2, Col5a2, and MMP2 in primary murine HSCs in vitro. Altogether, FOXF1 prevents aberrant ECM depositions during hepatic fibrosis by repressing pro-fibrotic gene transcription in myofibroblasts and HSCs.


INTRODUCTION
The liver is the body's filter and insults can result from a variety of infectious, toxic and metabolic agents. Hepatic fibrosis is the common end stage to a multitude of liver diseases (Civan, 2016) and is characterized by an excessive deposition of extracellular matrix (ECM) and collagen (Cheng and Mahato, 2007). Novel animal models of hepatic fibrosis are greatly needed to identify molecular mechanisms responsible for the disease pathogenesis and for the development of therapeutic agents. Hepatic stellate cells (HSC) reside in the space of Disse and are characterized by their storage of lipids when in the quiescent state (Yin et al., 2013;Croci et al., 2013). During fibrogenesis, quiescent HSCs differentiate into myofibroblasts (MF) in response to cytokine signaling from damaged hepatocytes and immune cells after liver insult. MFs secrete ECM and collagen to encapsulate the site of injury and shield the liver from plaguing insults (Cheng and Mahato, 2007). While HSCs and MFs make up only a small number of cells in liver tissue, they are the main contributors of ECM and collagen during liver repair and fibrogenesis (Brenner et al., 2012;Fausther et al., 2013). The TGF-β and PDGF signaling pathways play key roles in hepatic fibrosis and HSC activation (Makarev et al., 2016). TGF-β signaling stimulates cellular transdifferentiation of HSCs to MFs (Hellerbrand et al., 1999;Bachem et al., 1993), whereas PDGF signaling induces cellular proliferation in fibrotic foci (Wong et al., 1994;Kinnman et al., 2002).
The Forkhead Box F1 (FOXF1) transcription factor is expressed in human and murine HSCs and is important in regulating stellate cell activation after acute liver injury . In the advanced disease state of hepatocellular carcinoma (HCC), which is associated with significant fibrotic depositions, FOXF1 expression has been shown to be significantly decreased (Hodo et al., 2013). Foxf1 −/− mice are embryonic lethal due to severe developmental abnormalities in the yolk sac and allantois (Mahlapuu et al., 2001). Murine haploinsufficiency of Foxf1 causes lung hypoplasia, loss of alveolar capillaries in the lung and gall bladder agenesis (Kalinichenko et al., 2002;Bolte et al., 2018), and was associated with delayed lung and liver repair. After acute liver injury by carbon tetrachloride (CCl 4 ), Foxf1 +/− mice exhibited diminished activation of HSCs and delayed liver repair, indicating that FOXF1 is essential for liver repair after acute liver injury . Foxf1 siRNA delivered to mice through nanoparticles prevented activation of HSCs and subsequent collagen deposition after cholestatic liver injury (Abshagen et al., 2015). While these studies have shown that FOXF1 is required for activation of HSCs after acute liver injury, the role of FOXF1 in MFs and in the progression of fibrotic responses remains unknown.
In the present study, we generated a novel genetic mouse model to conditionally delete Foxf1 from MFs (αSMACreER;Foxf1 fl/fl ). During chronic liver injury, deletion of Foxf1 in MFs exacerbated hepatic fibrosis, increased collagen deposition and stimulated expression of profibrotic genes in the liver tissue. Our studies indicate that Foxf1 expression in MFs is critical to prevent MF accumulation and collagen deposition during liver fibrosis.

Deletion of Foxf1 in αSMA-positive cells exacerbates CCl 4 -induced hepatic fibrosis
Previous studies demonstrated that FOXF1 is present in HSCs in murine developing and adult livers Kim et al., 2005). Consistent with these studies, FOXF1 staining was detected in livers of e12.5-e17.5 mouse embryos as well as in mesenchyme of stomach and intestine (Fig. S1). In adult mice, FOXF1 is specifically expressed in the liver parenchyma but not in endothelial or smooth muscle cells surrounding the portal vein or hepatic artery   (Fig. 1A; Fig. S1), and FOXF1 staining co-localized with desmin (DES) (Fig. 1A), a known marker of HSCs (Yokoi et al., 1984). To investigate the role of Foxf1 in liver fibrosis, we utilized a conditional knockout approach. Transgenic mice containing a tamoxifeninducible αSMA-CreER transgene and two Foxf1-floxed alleles (αSMACreER;Foxf1 fl/fl ) were generated by breeding αSMA-CreER and Foxf1 fl/fl mice (Fig. 1B,C). Hepatic fibrosis was induced by chronic liver injury using multiple administrations of CCl 4 , which is known to increase fibrotic depositions and disrupt liver architecture in experimental mice (Martinez et al., 2014). Tamoxifen was given three times per week to achieve a continuous deletion of Foxf1 in αSMA-positive MFs (Fig. 1D) that derive from HSCs after liver injury (Mederacke et al., 2013). Morphological analysis of liver sections revealed increased fibrotic deposition in CCl 4 -treated αSMACreER;Foxf1 −/− livers compared to controls as shown by H&E ( Fig. 1E; Fig. S2) and Masson's Trichrome staining ( Fig. 1F; Fig. S2). Increased fibrosis in αSMACreER;Foxf1 −/− livers was confirmed by significant increases in collagen levels by Sircol (Fig. 1G) and hydroxyproline ( Fig. S2) assays as well as by qRT-PCR for Col1α1 and Col3α1 mRNAs (Fig. 1H). Treatment with tamoxifen alone (without CCl 4 ) did not affect liver architecture or induce liver fibrosis (Fig. S2). Thus, deletion of Foxf1 from MFs accelerates liver fibrosis after chronic liver injury. FOXF1 expression is decreased in hepatic myofibroblasts of αSMACreER;Foxf1 −/− mice Since FOXF1 is expressed in HSCs in the liver , we examined the efficiency of Foxf1 deletion in our experimental model, using immunostaining for FOXF1 and DES. Without CCl 4 treatment, FOXF1 was observed in cell nuclei of DES-positive stellate cells in Foxf1 fl/fl and αSMACreER;Foxf1 fl/fl livers ( Fig. 2A). After CCl 4 and Tam treatment, FOXF1 staining was reduced in DES-positive cells of αSMACreER;Foxf1 −/− livers but not in Foxf1 fl/fl livers ( Fig. 2A). We also immunostained liver sections for FOXF1 and αSMA, a marker of MFs (Rockey et al., 2013). While αSMA was not detected in parenchyma of quiescent livers, αSMA staining was increased after CCl 4 injury. FOXF1 was detected in MFs of control livers but not in αSMACreER;Foxf1 −/− livers (Fig. 2B). Quantitative counts of FOXF1-expressing cells demonstrated that the number and percentage of FOXF1 + MFs (FOXF1 + αSMA + ) were reduced whereas the number and percentage of FOXF1 − MFs (FOXF1 − αSMA + ) were elevated in injured αSMACreER;Foxf1 −/− livers compared to controls (Fig. S3).   (Fig. 2E). The αSMA-CreER transgene allows for the maintained presence of FOXF1 for HSC activation  and only deletes FOXF1 after αSMA is expressed in MFs. Thus, αSMA-CreER transgene effectively deletes Foxf1 from hepatic MFs during CCl 4 -mediated chronic liver injury.

Deletion of Foxf1 reduces MMP9 activity in CCl 4 -injured livers
Histological staining with Sirius Red/Fast Green showed a significant increase in collagen accumulation in αSMACreER; Foxf1 −/− livers after five weeks of CCl 4 treatment ( Fig. 3A; Fig. S4). Increased fibrosis in Foxf1-deficient livers was confirmed by immunostaining for DES and αSMA (Fig. 3B,C). To examine the consequences of extended CCl 4 treatment, we treated mice with CCl 4 for 18 weeks. While hepatic enzymes AST and ALT were increased in blood serum after 18 weeks of CCl 4 , there was no difference between CCl 4 treated αSMACreER;Foxf1 −/− and control mice (Fig. 3D,E). Blood serum protein (albumin, globulin) and bilirubin (direct, indirect) levels were not affected after deletion of Foxf1 (Fig. S5). Collagen accumulation was time-dependent ( Fig. S6), and after 18 weeks of CCl 4 treatment, resulted in widespread liver fibrosis (Fig. S7) and in rare cases, the appearance of visible tumors (Fig. S7).

Deletion of Foxf1 does not influence cellular proliferation in fibrotic livers
We evaluated proliferation markers to investigate if the increased fibrosis in αSMACreER;Foxf1 −/− livers was due to an expansion of the stromal cells. While cellular proliferation was increased after CCl 4 treatment, there were no significant differences in the number of proliferating hepatocytes or non-hepatocytes between Foxf1 fl/fl and αSMACreER;Foxf1 −/− livers ( Fig. 4A-C; Fig. S9). Hepatocytes and non-hepatocytes were identified through distinct morphological appearances (Malarkey et al., 2005) from high magnification images. mRNAs of proliferation-specific genes Foxm1, Ccnb1, Ccnd1, and AurKB (Wang et al., 2009;Kalin et al., 2011;Ren et al., 2013) were unchanged between Foxf1 fl/fl and αSMACreER;Foxf1 −/− livers (Fig. 4D). Proliferating HSCs and MFs were detected in  Protein levels of proliferation-specific genes FOXM1 and CCND1 were unaltered in αSMACreER;Foxf1 −/− livers compared to controls (Fig. 4G). Thus, Foxf1 deletion does not affect proliferation of HSCs and MFs after chronic CCl 4 liver injury.

RNA-seq analysis identified direct FOXF1 target genes critical for ECM deposition and hepatic fibrosis
In order to identify FOXF1 target genes, RNA-seq (GEO accession GSE123726) was performed on primary hepatic stromal cells (containing MFs and HSCs) isolated from CCl 4 -treated Foxf1 fl/fl and αSMACreER;Foxf1 −/− livers. Purified cells expressed Des and Acta2, but lacked hepatocyte (Nikoozad et al., 2014) and Kupffer cell (Yang et al., 2013) markers (Fig. S10). Foxf1 mRNA was lost in isolated αSMACreER;Foxf1 −/− stromal cells (Fig. 5A), a finding consistent with efficient deletion of Foxf1 from CCl 4 -treated livers. The RNA-seq was used to compare differential gene expression patterns between the Foxf1 fl/fl and αSMACreER;Foxf1 −/− stromal cells. The differential gene expression in the two groups are represented in a heat map (Fig. 5B). Gene ontology demonstrated that increased functional pathways for the αSMACreER;Foxf1 −/− mice were related to ECM regulation, while decreased functional pathways included normal liver functions and metabolism (Fig. 5C). RNA-seq analysis was cross referenced with FOXF1 ChIP-seq analysis (GEO accession GSE100149). 905 genes were common between RNA-seq and ChIP-seq (Fig. 5D), which include 74 genes related to ECM deposition and fibrosis. ChIP-seq proximity analysis revealed that 20 of these ECM genes had FOXF1 binding sites within 2KB of the transcription start site (Fig. 5E).
Interestingly, Col1α2, Col5α2 and Mmp2 were among the 20 ECM-related genes that had FOXF1 binding sites within the gene loci (Fig. S11, Table S1). COL1α2 and COL5α2 are common ECM components in fibrotic livers (Mak et al., 2016), whereas MMP2 is a collagenase that is increased during liver fibrosis and associated with disease progression (Benyon et al., 1996). Expression of Col1α2, Col5α2 and Mmp2 mRNAs were increased in CCl 4 -treated αSMACreER;Foxf1 −/− livers as shown by RNA-seq and qRT-PCR (Fig. 5E, Fig. 6D), suggesting a negative regulation by FOXF1. The presence of gene silencing histone methylation marks H3K9me3 and H3K27me3 (Dong and Weng, 2013;Bernstein et al., 2006) in FOXF1-binding promoter regions (Fig. 6A-C) is consistent with negative regulation of these genes by FOXF1. In order to confirm the regulation of Col1α2, Col5α2 and Mmp2 by FOXF1, we overexpressed FOXF1 in isolated murine HSCs (Fig. 6E). Lentiviral-mediated overexpression of FOXF1 decreased Col1α2 and Mmp2 in vitro (Fig. 6F). Thus, FOXF1 negatively regulates expression of pro-fibrotic genes in MFs. Altogether, FOXF1 expression in myofibroblasts is essential to inhibit liver fibrosis after chronic liver injury (Fig. 6G).

DISCUSSION
Myofibroblast activation is a key mechanism in the development of hepatic fibrosis. However, transcriptional regulation of myofibroblasts during liver fibrogenesis remains poorly characterized. In the present study, we found that the deletion of Foxf1 in MFs during chronic CCl 4 -mediated injury exacerbated liver fibrosis, increased collagen deposition and stimulated expression of pro-fibrotic genes. ECM-related proteins were identified as novel FOXF1 transcriptional targets, suggesting that FOXF1 plays an important role in the regulation of ECM and collagen deposition during the progression of hepatic fibrosis.
Previous studies have focused on the role of FOXF1 in acute liver injury using a single CCl 4 administration to Foxf1 +/− mice. These studies demonstrated that FOXF1 is necessary for HSC activation to promote liver repair . CCl 4 -treated Foxf1 +/− mice exhibited diminished collagen depositions and increased mortality after the liver injury . A recently published model of Foxf1-silencing using a lipidbased nanoparticle system to deliver Foxf1 siRNA to the liver demonstrated attenuated collagen deposition when Foxf1 siRNA was delivered 48 h prior to bile duct ligation (Abshagen et al., 2015). It is likely that Foxf1 siRNA inhibited FOXF1 signaling in hepatic stellate cells, decreasing their activation and subsequent collagen depositions into the liver tissue, a finding consistent with previous studies using Foxf1 +/− mice . Recently, a model of chronic hepatic injury using CCl 4 -injections, similar to the present study, was unsuccessful in silencing Foxf1 expression using the same lipid based system to deliver Foxf1 siRNA (Abshagen et al., 2015(Abshagen et al., , 2017. This method involved four weeks of IP CCl 4 -injections before two weeks of treatment with Foxf1 siRNA (Abshagen et al., 2017). It is possible that the lack of Foxf1 silencing was due to inability of nanoparticles to target hepatic MFs. In the current study, we utilized a conditional genetic mouse model to delete Foxf1 in MFs during CCl 4 -induced hepatic fibrosis which shares multiple histological similarities with human disease (Masugi et al., 2018;Bataller and Brenner, 2005). Interestingly, the loss of Foxf1 in MFs resulted in increased collagen deposition, causing severe fibrotic lesions between hepatic portal triads in αSMACreER;Foxf1 −/− livers. Our studies suggest that FOXF1 inhibits production of collagen and ECM during the progression of liver fibrosis. Increased fibrosis in Foxf1-deficient mice was associated with the appearance of liver tumors, a finding consistent with increased tumor formation in patients with liver cirrhosis (EASL-EORTC et al., 2018). Our studies suggest that maintaining Foxf1 expression can be beneficial in patients with advanced liver fibrosis to inhibit fibrotic responses and decrease the risk of liver tumorigenesis.
In the present study, collagens were significantly increased in αSMACreER;Foxf1 −/− livers after chronic CCl 4 -treatment. Desmin and αSMA were both increased in αSMACreER;Foxf1 −/− livers; however, there were no differences in the number of proliferating cells between Foxf1 fl/fl and αSMACreER;Foxf1 −/− livers. Previously, FOXF1 has been shown to stimulate cell proliferation in lung endothelial cells (Ren et al., 2014;Bolte et al., 2017) and in rhabdomyosarcoma tumor cells (Milewski et al., 2017). Surprisingly, we found that deletion of Foxf1 from MFs does not affect their proliferation during liver fibrogenesis. It is possible that FOXF1 requires additional co-activator or co-repressor proteins (that are not present in MFs) to regulate cellular proliferation. Additionally, we found an increase in Timp2 expression with a decrease in MMP9 activity in αSMACreER;Foxf1 −/− livers. Since it is well-known that TIMPs and MMPs regulate ECM depositions to balance the scaring and healing processes during fibrosis (Duarte et al., 2015), it is possible that the loss of Foxf1 alters the TIMP/ MMP balance to allow accumulation of collagens without the degradation mechanisms necessary for proper wound healing. Interestingly, MMP9 has been implicated in HSC to MF transdifferentiation (Han et al., 2007) in addition to its roles in ECM degradation (Duarte et al., 2015;Kurzepa et al., 2014). Therefore, decreased MMP9 activity can contribute to increased liver fibrosis in αSMACreER;Foxf1 −/− mice. Surprisingly, FOXF1 was increased in activated HSCs compared to quiescent HSCs. It is possible that FOXF1 is differentially regulated in HSCs compared to hepatic MFs, and that after liver injury, FOXF1 protects HSCs from differentiating into MFs through transcriptional repression of profibrotic genes.
Consistent with increased fibrosis in Foxf1-deficient livers, RNA-seq analysis revealed increased ECM-related functional pathways in a purified stromal cell population. Comparison with FOXF1 ChIP-seq data revealed 20 novel transcriptional targets of FOXF1, which include Col1α2, Col5α2 and Mmp2, expression of which was increased in Foxf1-deficient cells.
COL1α2 is one of the most abundant ECM components in the liver along with COL1α1 and COL3α1 (Lai et al., 2011). COL5α2 is highly expressed with Collagens 1 and 3 and is important in regulating the assembly and structure of these collagens in the fibrotic matrix (Moriya et al., 2011). MMP2 acts as a collagenase, known to be activated during hepatic fibrosis (Benyon et al., 1996).
In addition to increased mRNA levels of the genes in FOXF1deficient cells, we found multiple FOXF1 binding sites within their gene promoter region and introns, suggesting direct transcriptional repression. This hypothesis is supported by the presence of H3K4me3 and H3K9ac, histone modifications associated with transcriptional repression (Dong and Weng, 2013;Rea et al., 2000), at FOXF1 binding sites. In summary, we have developed a novel genetic mouse model to study the role of FOXF1 in MFs during chronic liver injury. Using this model, we demonstrated that Foxf1 expression in MFs is necessary to inhibit hepatic fibrosis and maintain the balance of collagen depositions, through transcriptional repression of pro-fibrotic genes.

Mice
The Foxf1 fl/fl mouse line was previously generated and bred into the C57Bl/ 6 mouse background (Ren et al., 2014;Cai et al., 2016). Foxf1 fl/fl mice were bred with αSMA-CreER mice (Jackson Laboratory, 029925; Wendling et al., 2009) to generate αSMACreER;Foxf1 fl/fl mice (Black et al., 2018). αSMACreER;Foxf1 fl/fl mice were bred with Foxf1 fl/fl mice and male pups were genotyped and used for all experiments at the age of 6-8 weeks. The following primers were used for genotyping: αSMA-CreER sense: 5′ TGCAACGAGTGATGAGGTTCGC 3′ and anti-sense: 5′ GATCCT-GGCAATTTCGGCTATACG 3′; αSMA-WT sense 5′ GGTTTCTATTGC-TACCAAGAGACAT 3′ and anti-sense: 5′ TGCACCAAACCCTGGA-CTAAGCAT 3′; Foxf1 fl/fl sense: 5′ GCTTTGTCTCCAAGCGCTGC 3′ and anti-sense: 5′ TTCAGATCTGAGAGTGGCAGCTTC 3′. Foxf1 fl/fl littermates were used as controls. To activate the conditional Foxf1 knockout, tamoxifen (Tam) was given via intraperitoneal injection (40 mg/kg of body weight; Sigma-Aldrich) three days in a row at the beginning of each week starting at week 2 over the course of the chronic liver injury period. Hepatic injury was induced by intraperitoneal injections of carbon tetrachloride (CCl 4 ; 1 μl/g of body weight 20% v/v; Sigma-Aldrich; diluted in sunflower seed oil) three times a week every other day over the course of the chronic liver injury period. The levels of aminotransferases AST and ALT, proteins albumin and globulin, and direct and indirect bilirubin were determined by serological analysis of blood serum as previously described (Sun et al., 2017;Ren et al., 2010). All animal studies were approved by the Institutional Animal Care and Use Committee (IACUC) of Cincinnati Children's Research Foundation and the NIH IACUC Guidebook. All experiments were covered under our animal protocol (IACUC2016-0038). The Cincinnati Children's Research Foundation Institutional Animal Care and Use Committee is an AAALAC and NIH accredited institution (NIH Insurance #8310801).

RNA sequencing
RNA was isolated from HSC/MF population purified from CCl 4 -treated Foxf1 fl/fl and αSMACreER;Foxf1 −/− livers using a differential plating method (Giassetti et al., 2016) that we modified. Briefly, liver cell suspension was plated on tissue culture dishes (Corning) and incubated for 2 h at 37°C. Supernatant and non-adherent cells were washed off and the adherent cell population was collected for experiments. Samples were pooled to generate the libraries using the TruSeq RNA library preparation kit and were sequenced on an Illumina HiSeq 2000 sequencer, generating approximately 10 M high quality single end reads (75 base-long reads). Alignment was performed using the Tophat/Cufflink pipeline (Trapnell et al., 2009(Trapnell et al., , 2010. Finally, cuffmerge tool was used to generate Binary Alignment/Map files (BAM files) (Roberts et al., 2011). BAM files of RNA-seq data were analyzed using Avadis ® NGS Version 1.3.0 software. Reads were filtered to remove: (1) duplicate reads, (2) non-primary-matched reads, and (3) reads with alignment scores <95. Quantification was performed on the filtered reads against the RefSeq annotation. Data normalization was performed with the DESeq package. The sequencing depth was estimated by the read count of the gene with the median read count ratio across all genes. The method was based on the negative binomial distribution, which allows for less restrictive variance parameter assumptions than does the Poisson distribution. The false discovery rate was calculated according to the Benjamini and Hochberg algorithm (Benjamini and Hochberg, 1995). Genes with expression altered by a factor of 1.5 and a false discovery rate of 0.05 in Foxf1 fl/fl cells compared with αSMACreER; Foxf1 −/− cells were selected for gene set enrichment analysis using ToppGene Suite. Hierarchical clustering was performed by Ward's method using Euclidean distance metric. RNA-seq data are available at GEO accession GSE123726. RNA-seq data were compared to previously published ChIP-seq data (GEO accession GSE100149) using a two-way Venn diagram.

Statistical analysis
Statistical significance differences in measured variables between control and experimental groups were assessed with a Student's t-test (two-tailed) or one-way analysis of variance (ANOVA) with Bonferroni post hoc test as appropriate. P<0.05 was considered to be significant, with P<0.05 indicated with *, P<0.01 indicated with **, P<0.001 indicated with ***, and P<0.0001 indicated with ****. Values for all measurements were expressed as mean±s.e. of mean.

Competing interests
The authors declare no competing or financial interests.

Data availability
The RNA sequencing microarray data presented in this article has been uploaded to Gene Expression Omnibus (GEO) database at the National Center for Biotechnology Information (NCBI) and the accession number is GSE123726.