Liver myofibroblasts of murine origins express mesothelin: Identification of novel rat mesothelin splice variants*

Liver myofibroblasts are specialized effector cells that drive hepatic fibrosis, a hallmark process of chronic liver diseases, leading to progressive scar formation and organ failure. Liver myofibroblasts are increasingly recognized as heterogeneous with regards to their origin, phenotype, and functions. For instance, liver myofibroblasts express cell markers that are universally represented such as, ItgαV and Pdgfrβ, or restricted to a given subpopulation such as, Lrat exclusively expressed in hepatic stellate cells, and Gpm6a in mesothelial cells. To study liver myofibroblasts in vitro, we have previously generated and characterized a SV40-immortalized polyclonal rat activated portal fibroblast cell line called RGF-N2 expressing multiple mesothelin mRNA transcripts. Mesothelin, a cell-surface molecule expressed in normal mesothelial cells and overexpressed in several cancers such as, mesothelioma and cholangiocarcinoma, was recently identified as a key regulator of portal myofibroblast proliferation, and fibrosis progression in the setting of chronic cholestatic liver disease. Here, we identify novel mesothelin splice variants expressed in rat activated portal fibroblasts. RGF-N2 portal fibroblast cDNA was used as template for insertion of hemagglutinin tag consensus sequence into the complete open reading frame of rat mesothelin variant coding sequences by extension PCR. Purified amplicons were subsequently cloned into an expression vector for in vitro translation and transfection in monkey COS7 fibroblasts, before characterization of fusion proteins by immunoblot and immunofluorescence. We show that rat activated portal fibroblasts, hepatic stellate cells, and cholangiocarcinoma cells express wild-type mesothelin and additional splice variants, while mouse activated hepatic stellate cells appear to only express wild-type mesothelin. Notably, rat mesothelin splice variants differ from the wild-type isoform by their protein properties and cellular distribution in transfected COS7 fibroblasts. We conclude that mesothelin is a marker of activated murine liver myofibroblasts. Mesothelin gene expression and regulation may be critical in liver myofibroblasts functions and fibrosis progression.


Materials and reagents
Cell culture reagents and media were obtained from Life Technologies (Carlsbad, CA), Fisher Scientific (Pittsburgh, PA) and Thermo Scientific (Rockland, MA). Molecular biology and SDS-PAGE reagents/kits were obtained from Qiagen (Valencia, CA), Bio-Rad Laboratories (Hercules, CA), New England BioLabs (Ipswich, MA) and Life Technologies.

Animal care
All procedures involving animals were reviewed and approved by University of Arkansas for Medical Sciences Institutional Animal Care and Use Committee, protocol AUP#3703. Adult male Sprague-Dawley rats (4 months, two animals) were purchased from Charles River Laboratories (Redfield, AR) and used for two-step collagenase liver perfusion performed as terminal procedure under combined ketamine (80-100 mg/kg)/xylazine (5-10 mg/kg) anesthesia (intraperitoneal administration), as previously described [9]. The method of sacrifice was exsanguination through the inferior vena cava, and all precautions taken to minimize animal suffering.

Primary cell isolation and culture
Primary PF and HSC were isolated from rat livers, as previously described [9,21]. Briefly, hepatocyte and non-parenchymal cell fractions were obtained by in situ pronase/collagenase perfusion of livers. Primary PF were obtained by serial digestion and mesh filtration of hilar remnants, while primary HSC were obtained by density gradient centrifugation of non-parenchymal cell fractions. The resulting cell suspensions were plated onto tissue culture plastic dishes and grown in DMEM/F-12 containing 10% fetal bovine serum and antibiotics. Primary PF and HSC were used on day 3 or prior (quiescent, passage 0) and on day 4 and beyond (myofibroblastic, passage 0 or beyond) after plating, as previously described [9,21]. All cells were maintained at 37˚C, under 95% air-5% CO2.

RT-PCR
Total RNA was isolated from rat and mouse tissues, and primary and immortalized liver myofibroblasts cells using the RNeasy Plus Kit (Qiagen). Each RNA sample (1 μg) was digested with DNase1 enzyme (Life Technologies) to remove any genomic DNA contamination and reverse-transcribed using the iScript RT Supermix (Bio-Rad). Semi-quantitative PCR amplification was performed using RT reaction products and the TopTaq 1 Master Mix Kit (Qiagen) with the following protocol for the PCR reactions: Initialization at 94˚C for 2 minutes followed by 35 cycles of 30 second denaturation at 94˚C, 30 second annealing at 60˚C, 30-150 second elongation at 72˚C; and 10 minutes final elongation at 72˚C, using an S1000 Thermo Cycler (Bio-Rad). Amplification products were visualized on 3% agarose gels via ethidium bromide staining. The primer sequences used are listed in Table 1.

DNA plasmids and in vitro translation
RGF-N2 cDNA sample was PCR-amplified using primers located in the 5'-and 3'-UTR regions of rat Mesothelin coding sequence (NCBI Nucleotide ID: NM_031658.1). All PCR reactions were performed with Phusion 1 (New England BioLabs) or TopTaq 1 High-Fidelity DNA polymerases for maximal elongation fidelity. Purified PCR reactions products were cloned into the pCR4 1 expression vector (Life Technologies), using a TOPO 1 TA cloning kit (Life Technologies). Chemically-competent OneShot 1 TOP10 bacteria cells (Life Technologies) were used for superior transformation efficiency. After PCR analysis of approximately 200 obtained transformants, six clones (A, H, S, U, W, Y) were selected based upon PCR amplicon size/abundance (ranging between 300 and 2100 base pairs approximately), and analyzed by automated sequencing to confirm insert size, sequence and orientation (UAMS DNA Sequencing Core Facility). Each DNA plasmid was then used as template for 5'-end insertion of Hemagglutinin (HA) tag consensus coding sequence (-TACCCATACGATGTTCCAGATTA CGCT-, 27 base pairs) into the complete open reading frame of rat Mesothelin coding sequence, by overlap PCR extension. All constructs were designed, based upon clone sequence homology, so as to insert: 1) HA peptide coding sequence between the 24 th and 25 th base pairs (bp) of rat Mesothelin consensus coding sequence (NCBI Nucleotide ID: NM_031658.1 [24][25]), resulting in a 9-amino acid insertion between the 8 th and 9 th amino acids of rat Mesothelin protein sequence (NCBI Protein ID: NP_113846 [8][9]); and 2) Xba1 and Age1 restriction sites at both 5'-(before ATG start codon, NM_031658.1 [1]) and 3'-ends (after TGA stop codon, NM_031658.1 [625]) of rat Mesothelin coding sequence. All PCR amplicons were purified and used for: 1) adaptor sequences addition by PCR amplification, followed by in vitro translation reactions using the cell-free PURExpress 1 In Vitro Protein Synthesis Kit (New England BioLabs); and 2) directional cloning into the pcDNA™3.3 TOPO 1 TA vector using a Gibson Assembly kit (New England BioLabs), and Xba1 and Age1 restriction enzymes (New England BioLabs). Obtained transformants were analyzed by PCR and automated sequencing. The primer sequences used are listed in Table 1.

Immunofluorescence
Transfected COS7 fibroblasts (grown on coverslips) were fixed with neutral (pH = 7.2) 4% paraformaldehyde solution (diluted in 1X Phosphate-Buffered Saline, PBS) for 20 minutes, washed in 1X PBS, and further permeabilized with Triton X-100 0.1% solution (diluted in 1X PBS) for 10 minutes, all steps at room temperature. After several washes in 1X PBS, coverslips were incubated with a 7% goat serum (Life Technologies), 0.5% bovine serum albumin (Fisher Scientific) blocking solution (diluted in 1X PBS) at room temperature for 1 hour, and then, with rabbit monoclonal anti-Hemagglutinin tag antibody (clone C29F4, Cell Signaling Technologies, Danvers, MA; diluted 1:50000 in blocking solution) at 4˚C overnight. After several washes in 1X PBS, coverslips were further incubated with goat Alexa647-conjugated anti-rabbit IgG antibody (Life Technologies, diluted 1:1000 in blocking solution). After several washes in 1X PBS, DAPI-supplemented Prolong Diamond anti-fade mountant was added to coverslips. Fluorescence microscopy images were acquired using a Zeiss AxioImager imaging system (Zeiss Laboratories, White Plains, NY).

Nucleotide sequences
Complementary DNA sequences containing nucleic acid residues corresponding to the complete open reading frame of rat Mesothelin splice variants were identified and assembled by single-pass automated DNA sequencing. In silico analysis of predicted amino acid sequences corresponding to rat Mesothelin splice variants was performed using open-access Big-PI Predictor, Compute pI/Mw, and SignalP 4.1 Server tools from the ExPASy Bioformatics Resources portal [26], EMBOSS Stretcher (Protein Alignment) tool from the EMBL-EBI bioinformatics web and programmatic tools framework [27], and Prop 1.0 Server tool from the DTU-CBS Prediction Servers [28]. The analysis results are listed in Table 2. A multiple alignment of cloned rat Mesothelin splice variant nucleotide sequences was performed with wild-type rat Mesothelin consensus coding sequence (NM_031658.1) as reference, and nucleotide sequence identity (as percentage) determined, using open-access MView tool from the EMBL-EBI bioinformatics web and programmatic tools framework [23] (see S1 Fig).

Results
Our laboratory has previously generated and characterized two activated RGF and RGF-N2 rat liver PF cell lines that express Mesothelin (Msln) [9]. Here, we cloned and characterized these previously unknown rat Msln transcript variants expressed in RGF-N2 PF cells. In parallel, we analyze established activated HSC-T6 (rat), JS1 (mouse), and Col-GFP (mouse) hepatic stellate cell lines for the expression of prospective Msln transcript variants.
First, semi-quantitative PCR analysis of cDNA samples from primary rat quiescent and activated portal fibroblasts and hepatic stellate cells ubiquitously shows an amplification product corresponding to wild-type Msln (observed molecular weight 857 base pairs, bp) (Fig 1,  left panel). Interestingly, additional PCR amplification products (observed molecular weights varying between 300-800 bp) were detected only in cDNA samples from primary rat activated portal fibroblasts and hepatic stellate cells. PCR analysis was also performed using cDNA samples from immortalized rat activated RGF and RGF-N2 portal fibroblast and HSC-T6 hepatic stellate cell lines, and several amplification products, including one corresponding to wild-type Msln, were also observed. A similar observation was made for rat cholangiocarcinoma BDEneu cell line, used as PCR positive control. When semi-quantitative PCR analysis was performed on immortalized mouse Col-GFP and JS1 HSC cells, only a single amplification product corresponding to wild-type Msln was detected (Fig 1, middle panel), similar to mouse lungs PCR positive control. For each species, sequence verification of purified PCR bands of interest showed that all samples contained cDNA corresponding to Msln gene products (S1 Fig). Second, to further investigate the prospective rat Msln transcript variants, RGF-N2 cDNA was amplified by PCR, using specific oligonucleotide primers located in 5' and 3' untranslated regions of rat Msln mRNA sequence (NM_031658.1) i.e. capable of amplifying the complete rat Msln coding sequence (see Table 1). The resulting PCR amplification products were column-purified and used as templates for 5'-end insertion of Hemagglutinin (HA) tag consensus coding sequence by overlap extension PCR, before cloning into a CMV-driven expression vector. Based on insert size and abundance, six clones (A, H, S, U, W, and Y) were selected and sequence-verified to ensure all clones could be translated using the same open reading frame. Sequence analysis of obtained cDNAs clearly identified clone W as wild-type Msln, while the remaining clones sequences reveal mRNA exon skipping and alternative splice donor site as splicing mechanisms (Fig 2). In silico analysis predicted that encoded unmodified (i.e. HA tag-less) proteins would exhibit molecular weights ranging from 12 to 70 kiloDaltons (see Table 2). From these results, we concluded that multiple rat Msln mRNA transcripts, in addition to the wild-type isoform are expressed by liver myofibroblasts deriving from activated portal fibroblasts and hepatic stellate cells. In contrast, only wild-type Msln is expressed in mouse HSC-derived liver myofibroblasts.
Subsequently, each Msln isoform clone plasmid DNA was used as template for addition of in vitro translation adaptor sequences by PCR (Fig 2). In vitro translation (IVT) reactions were performed using purified PCR products as template for recombinant protein synthesis, and analyzed by immunoblot using antibodies directed against HA tag peptide and rat Msln protein (Fig 3). As expected, the anti-HA antibody detected synthetized fusion protein products in all wells with their molecular weight varying from under 12 to slightly less than 76 kDa approximately, demonstrating that tag was successfully added and that the selected six clones encode viable recombinant proteins. The anti-rat Msln CT antibody detected synthetized protein products in 3 out of 6 wells, corresponding to clones S, U and W, with molecular weights varying from under 52 to slightly less than 76 kDa approximately. Similarly, the anti-Msln SC antibody detected only synthesized fusion protein products in 3 out of 6 wells, with molecular weights varying from under 24 to slightly less than 76 kDa approximately. Of note, an extra band with a molecular weight of 52 kDa was unexpectedly observed after immunoblot analysis of IVT reaction, using clone S plasmid DNA as template with anti-HA and both anti-rat Msln antibodies (Fig 3, asterisk). Because both anti-HA and anti-Msln SC antibodies detected the Msln splicing isoform with a predicted molecular weight of 24 kDa (in contrast with the antirat (specific) Msln CT antibody), this extra band was surmised to represent a reaction artifact potentially resulting from splicing isoform aggregation. Next, monkey COS7 fibroblast cell line that lacks HA tag expression was used for transient heterologous expression of HA-tagged Msln splicing variants. Immunoblot analysis of transfected cell extracts with multiple anti-HA antibodies confirmed our results obtained after similar analysis of IVT reaction products. The molecular weight of fusion protein products ranged from under 12 to 76 kDa approximately. Remarkably, only a single recombinant protein product with a molecular weight of 24 kDa was observed in the sample corresponding to S-Msln transfected cells. No positive band was detected in the sample corresponding to the mock-transfected cells. Afterwards, protein extracts from rat RGF and RGF-N2 PF, HSC-T6 HSC cell lines, and cell lysates from COS7 transfected with W-Msln plasmid DNA or a commercially-available validated full-length rat Msln cDNA clone were analyzed by immunoblot with anti-rat Msln CT and anti-Msln SC antibodies to assess expression of Msln protein species (Fig 4). In RGF, RGF-N2 and HSC-T6 protein samples, both antibodies only detected the mature form of Msln with a molecular weight of 52 kDa. In contrast, in COS7 transfected with W-Msln plasmid DNA and commercial rat Msln cDNA clone, both antibodies mainly detected both precursor (Mpf + Msln) and mature (Msln) forms of rat Msln, with molecular weights of 52 and 76 kDa respectively. The anti-Msln SC antibody also detected in all rat samples tested a band of low molecular weight close to 20-30 kDa that could represent Mpf peptide cleaved off the Msln precursor during protein maturation. Immunoblot analysis of protein extracts from mouse Col-GFP and JS1 HSC cell lines, and control mouse Msln-transfected 293T cell lysate with anti-mouse (specific) Msln CT and anti-Msln SC antibodies indicated that mouse Msln mature form is expressed in both cell lines. The anti-Msln SC also detected a band of low molecular weight close to 20-30 kDa in JS1 samples, similar to the observation made for rat cell line samples. However, this antibody did not produce a signal in the well corresponding to the positive control. Taken as a whole, these results clearly indicate that wild-type Msln protein is expressed in murine liver myofibroblast lines deriving from portal fibroblasts (rat species) and hepatic stellate cells (rat and mouse species).
Finally, the distribution of HA-tagged Msln splicing variants was monitored by immunofluorescence, upon in transfected COS7 cells expressing recombinant GFP at the level of plasma membrane (Fig 5). Of three A, H and W Msln variants predicted to possess a GPI anchor (see Table 2), only clone A and W were observed exhibiting plasma membrane localization (Fig 5). Surprisingly, the distribution of Msln isoform H appeared to be predominantly cytoplasmic/ perinuclear. A similar distribution was also described for S, U and Y Msln variants lacking GPI anchorage.

Discussion
In the present study, we show that liver myofibroblasts deriving from both rat activated portal fibroblasts (primary, immortalized RGF and RGF-N2) and hepatic stellate cells (primary, immortalized HSC-T6) express mesothelin (Msln). We also report Msln expression in liver myofibroblasts deriving from mouse (Col-GFP, and JS1) activated hepatic stellate cells. We identify novel rat Msln transcripts expressed in RGF-N2 cells distinct from the one corresponding to wild-type isoform, and encoding viable proteins that can be produced in both cell-free and cell-based expression systems. Based on sequence analysis, our study identified exon skipping and alternative donor site as differential splicing mechanisms regulating Msln gene expression in rat RGF-N2 cells. However, we need to be cautious in our assessment since these findings can arise from clonal selection bias, and the fact that relative abundance of each novel variant identified was not quantified at the mRNA level. A key point is that our cloning approach allowed us to clone these transcripts along transcripts encoding wild-type Msln isoform, which strongly argues against the notion that these are defective transcripts simply escaping nonsense mRNA-mediated decay quality control [29].
First, when the five newly identified Msln splicing variants were characterized by immunoblot with two commercially available rat Msln-specific antibodies, only three isoforms including the one corresponding to wild-type Msln could be detected at best. Our immunoblot experiments also showed that the same antibodies tested could consistently detect both recombinant uncleaved and mature forms after heterologous expression in COS7 cells, yet only one serum could recognize the native uncleaved form in rat liver myofibroblast cell lines when assayed under similar conditions. Hence, our study clearly demonstrates that available tools to track Msln gene expression products are limited in their capacity of detection. The data provided here should be useful to improve epitope mapping for generation of Msln antibodies. Second, when the cellular distribution of Msln splicing variants was followed by immunofluorescence, one variant could be detected both at the level of plasma membrane and in the cytoplasm, similar to the wild-type protein. Other variants appear to primarily assume a cytoplasmic localization. Remarkably, cytoplasmic MSLN immunohistochemistry signals in human lung adenocarcinoma, lung squamous cell carcinoma, and extrahepatic cholangiocarcinoma tissue samples have been previously reported and attributed to likely unprocessed protein precursors [30][31][32]. We speculate that a potential explanation for this unexplained labeling might be the existence of yet-to-be-described human MSLN splicing variants, distinct from the other variants described in the literature [33][34][35]. Whether this regulatory mechanism of Msln gene expression is species-specific remains an interesting avenue that certainly needs to be further explored.
Third, our results also indicate that occurrence of Msln splicing variants can apparently be dependent on cell activation state, as rat portal fibroblasts express only wild-type Msln transcript at quiescence, with additional transcripts upon activation and phenotypic transition. A similar observation could be made for rat hepatic stellate cells undergoing myofibroblastic transdifferentiation. What could be the impact of Msln splicing variants with regards to liver myofibroblast functions? Recently, Msln, through its interaction with partner surface Mucin 16/CA125 protein, was shown to regulate proliferation, activation, and migration signals in cholestasis-induced liver myofibroblasts, and to a greater extent, liver fibrosis in vivo [10]. Hence, it is well possible that the novel Msln isoforms identified here possess the ability to also regulate the same mechanisms.
Finally, rat Msln gene splicing could also be observed in BDEneu cholangiocarcinoma, which are malignant biliary epithelial (non-fibroblastic) cells. Although our study did not characterize these cholangiocarcinoma-derived transcripts per se, a logical implication is that the observed mechanism of Msln gene expression regulation seems not to be restricted to a single cell type, here liver myofibroblasts. Importantly, this potentially provides new insights on the regulation of Msln gene expression, such as, the pathophysiological conditions and/or related factors involved the observed alternative splicing mechanisms. As suggested for liver myofibroblasts, Msln variants could influence the behavior of cholangiocarcinoma cells, since both Msln and MPF are biologically potent proteins acting as malignant factors to promote directly or indirectly tumorigenesis [36].

Conclusions
We show that alternative splicing of the rat Msln gene takes place in liver myofibroblasts and malignant biliary epithelial cells. Alternative splicing of rat Msln mRNA precursors allows these cells to create distinct protein isoforms that might be functionally relevant to disease progression in conditions such as, fibrosis and cancer.