Cia5dregulates a new fibroblast-like synoviocyte invasion-associated gene expression signature

DA (DA/BklArbNsi, arthritis-susceptible) inbred rats (originally from Bentin & Kingman, CA, USA) were maintained at the Arthritis and Rheumatism Branch (Arb; National Institutes of Health) and then transferred to the Feinstein Institute (previously named North Shore-LIJ Institute; Nsi). The genotype-guided breeding of DA.F344(Cia5d) was previously described in detail [14]. Briefly, a 37.2 megabase interval on rat chromosome 10 was transferred from F344 into the DA background over 10 backcrosses followed by at least five intercrosses (Figure 1). The experiments were conducted with rats homozygous at the congenic interval. All experiments involving animals were reviewed and approved by the Feinstein Institute for Medical Research Institutional Animal Care and Use Committee. Animals were housed in a pathogen free environment, on 12-hour light and dark cycles, with free access to food and water.

Rats aged 8 to 12 weeks received 150 μl of pristane by intradermal injection divided into two sites at the base of the tail [14, 16]. The animals were scored on days 14, 18 and 21 after pristane induction using a previously described arthritis scoring system [17, 18]. On day 21 after injection, the animals were killed and synovial tissue was collected from the ankles for FLS isolation.

FLSs were isolated by enzymatic digestion of the synovial tissue. Briefly, tissues were minced and incubated with a solution containing DNase 0.15 mg/ml, hyaluronidase type I-S 0.15 mg/ml, and collagenase type IA 1 mg/ml (Sigma-Aldrich, St. Louis, MO, USA) in Dulbecco's modified Eagle's medium (DMEM; Gibco, Invitrogen Corporation, Carlsbad, CA, USA) for 1 hour at 37°C. Cells were washed and re-suspended in DMEM supplemented with 10% fetal bovine serum (Gibco), glutamine 30 mg/ml, amphotericin B 250 μg/ml (Sigma), and gentamicin 10 mg/ml (Gibco). After overnight culture, nonadherent cells were removed and adherent cells were cultured. All experiments were performed with cells after passage four (95% FLS purity).

Freshly trypsinized FLSs (105) were re-suspended in phosphate-buffered saline with 0.02% azide (Sigma-Aldrich) and 1% bovine serum albumin (P Biomedicals, Aurora, OH, USA), and incubated with 1 μg anti-CD32 (Pharmingen, San Diego, CA, USA) to block Fcγ II receptors. Cells were stained with saturating concentrations of CD90 (OX-7; PerCP, Pharmingen) or isotype control. Stained cells were fixed with 1% paraformaldehyde in phosphate-buffered saline and analyzed by flow cytometry in a FACSCalibur (Becton Dickinson, Franklin Lakes, NJ, USA), using the BD Cell-Quest™ Pro version 4.0.1 software (Becton Dickinson).

We previously studied the invasive properties of FLSs through a collagen matrix (Matrigel). Cell interactions with the extracellular matrix are known to influence the expression of several genes, including activation of MMP-2 [19], which is a key mediator of the FLS invasive phenotype. Therefore, in order to study the gene expression signature of highly invasive and minimally invasive FLSs, cells were cultured under the same conditions as used in the invasion studies. Specifically, 100% confluent 75 cm² FLS culture flasks were trypsinized (trypsin 0.25% with EDTA 0.1%). The rates of cellular proliferation differed among cell lines, and we previously showed that FLS proliferation does not correlate with the FLS invasive behavior. In order to have similar cell confluence at the time of FLS harvesting for RNA extraction, 10% to 50% of the high-density 75 cm² cell culture flasks (depending on the cell line) were plated in Matrigel-coated 10 cm culture dishes (Becton Dickinson) with DMEM, 10% fetal bovine serum, antibiotics, and glutamine. Cell cultures were maintained at 37°C with 5% carbon dioxide for 24 hours. After 24 hours, FLSs were harvested using a cell scraper (Corning, Acton, MA, USA) followed by digestion of the Matrigel with 10 ml collagenase D 1 mg/ml (Roche Applied Science, Indianapolis, IN, USA) at 37°C for 10 minutes. FLSs were then collected by centrifugation, washed twice with ice-cold phosphate-buffered saline. Cell pellets were re-suspended in RLT lysis buffer (RNeasy Mini Kit; Qiagen, Valencia, CA, USA) with 1% (vol/vol) β-mercaptoethanol (Sigma). Cell-lysis buffer suspension was vortexed, frozen in liquid nitrogen and stored at -80°C until RNA extraction.

Cells in RLT buffer were disrupted using QIAshredder spin columns (Qiagen), and total RNA was extracted using the RNeasy Mini Kit (Qiagen), in accordance with the manufacturer's instructions. Samples were digested with DNase (Qiagen) and eluted with 30 μl RNase-free water. RNAs were quantified and assessed for purity using a NanoDrop spectrophotometer (Rockland, DE, USA). RNA integrity was verified with a BioAnalyzer 2100 (Agilent, Palo Alto, CA, USA).

The t-test was used to compare means of the log-transformed and non-log-transformed data. Genes with a P value under 0.01 between DA and DA.F344(Cia5d) were considered significant and included in additional analysis. The logistic regression model fitting was carried out as previously described [20, 21] using the filtered gene list. The statistical significance of a logistic regression result was obtained by comparing the deviance with the 'null deviance'. This null deviance is the (-2)log-likelihood of a random model in which the probability for a sample to belong to a group (for example, DA) is equal to the proportion of DA samples in the dataset. The difference between the deviance and the null deviance follows the χ² distribution with one degree of freedom by chance alone, and this χ² distribution was used to determine the P value. The R statistical package [22] was used for t-test and logistic regression analyses.

The Ingenuity IPA 5.5.1 program (Ingenuity, Redwood City, CA, USA) and PubMed and GEO (Gene Expression Omnibus) searches were used for pathways detection. CLUSTER [23] and TREEVIEW [24] were used for cluster analysis and generation of a heat map.

The same RNA used for the microarray experiments was also used for the quantitative real-time PCR confirmation experiments. Total RNA 200 ng from each sample was used for cDNA synthesis using the Superscript III kit (Invitrogen). Primers and probe sequences were designed to target the same exon as used in the Illumina RatRef-12 Expression BeadChip. We used Exiqon (Woburn, MA, USA) and Taqman (ABI, Applied Biosystems, Foster City, CA) probes (Table 1). GAPDH was used as endogenous control. Probes were labeled with FAM at the 5' end and TAMRA at 3' end and used at a final concentration of 100 nmol/l. Primers were used at 200 nmol/l concentration with Eurogentec quantitative real-time PCR mastermix (Eurogentec, San Diego, CA, USA). The ABI 7700 quantitative real-time PCR thermocycler was used at 48°C for 30 minutes, 95°C for 10 minutes, and 45 cycles of 95°C for 0.15 minutes and 60°C for 1 minute. Samples were run in duplicates and the means used for analysis. Data were analyzed using Sequence Detection System software version 1.9.1 (ABI). Results were obtained as Ct (threshold cycle) values. Relative expression of all the genes was adjusted for GAPDH in each sample (ΔCt), and ΔCt used for t-test analysis. Quantitative real-time PCR fold differences were calculated with 2^-ΔΔCt [25].

In previous studies we determined that DA FLSs were highly invasive, and that alleles derived from the arthritis-resistant strain F344 at the Cia5d interval, as in DA.F344(Cia5d) congenics (Figure 1), specifically reduced the invasive properties of FLSs. Additionally, FLSs from DA and DA.F344(Cia5d) strains expressed similar mRNA levels of transforming growth factor-β, tumor necrosis factor-α, IL-1β and IL-6, as well as MMP-1, MMP-2, MMP-3, MMP-9, MMP-13, MT1-MMP and MT2-MMP [15]. Both strains had similar collagenase and MMP-3 activity, but levels of soluble MT1-MMP and active MMP-2 were increased in DA. MMP-2 inhibition reduced DA FLS invasion to levels similar to those of DA.F344(Cia5d). Cytoskeleton characteristics were also similar in DA and DA.F344(Cia5d) FLSs [15].

In the present study FLSs were stained with CD90, a marker for FLS [26], and analyzed by flow cytometry. Comparable numbers of CD90⁺ cells were detected both in five different DA and five different DA.F344(Cia5d) rats (percentage of CD90⁺ cells [mean ± standard deviation]: DA 95.46 ± 8.9 and DA.F344 [Cia5d] 96.51 ± 5.9), demonstrating that the cell lines were homogeneously CD90⁺.

A total of 7,665 genes out of 22,228 genes represented in the Illumina RatRef-12 BeadChip were expressed by both DA and DA.F344(Cia5d) FLSs. Log transformation did not significantly affect the list of differentially expressed genes, and therefore results are shown from analyses done with non-log-transformed data.

Sixty-six genes had a P value under 0.01 (Tables 2 and 3) and were used for fold change calculations and pathway detection analyses. Thirty-six genes were expressed in increased levels by DA FLSs, and the presence of F344 alleles at the Cia5d interval, as in DA.F344(Cia5d) congenics FLSs, was enough to reduce their expression significantly (Table 2). Thirty genes were expressed in reduced levels in DA and significantly increased in DA.F344(Cia5d) FLSs (Table 3). These observations demonstrate that alleles within the Cia5d interval, the only genetic difference between DA and DA.F344(Cia5d), are directly or indirectly involved in the regulation of the expression of several genes, and the difference in gene expression correlates with the difference in invasive properties of FLSs. Furthermore, cluster analysis separated DA FLSs from DA.F344(Cia5d) FLSs, demonstrating that the two strains could be reliably differentiated by gene expression (Figure 2).

Cluster analysis identified three main clusters among the genes expressed in increased levels in DA (Figure 2). One of the three clusters contained eight genes, three of which have been implicated in cancer and cancer-related cellular phenotypes such as invasion, and included Cxcl10, Vil2 and Dnmbp (Figure 3). The other genes in this cluster are involved in ion transport (Trpv2), mitosis (Smc1L1), or have incompletely characterized functions (Trim16, Ranbp6 and Hnrpul2). In total, 12 out of the 36 genes (33.3%) expressed in increased levels by DA FLSs and downregulated in DA.F344(Cia5d) are known to regulate cancer-associated processes, including cell cycle progression (Rpa2 and Rpa3), cell invasion (Cxcl10, Vil2, Nras, and Dnmbp), and metastasis (Vil2 and Brms1l), respectively (Table 2). In fact, Cxcl10 was the second best discriminator between DA and DA.F344(Cia5d) cell lines, as per logistic regression (Table 2).

Of additional interest in relation to the MMP-2-dependent difference in FLS invasion that we have observed, three of these genes – namely Cxcl10, Vil2 and Nras – are known to regulate the synthesis or activation of gelatinases. Increased levels of Cxcl10, Vil2, Dnmbp, Trim16, and Trpv2 in DA were confirmed using quantitative real-time PCR, with most of these genes having a nearly fourfold or greater difference in expression (P < 0.05; Figure 4a).

The list of genes with reduced expression in DA, as compared with increased expression in DA.F344(Cia5d) congenics, included seven genes that are involved in tumor suppression-like activity and cell cycle check-points, such as Aph1a, Brwd3, Gadd45b, Gmfg, Lox, and Plekhg2 (Table 3). Gadd45b was chosen for quantitative real-time PCR confirmation (P < 0.05; Figure 4b). These observations, combined with the 11 cancer and invasion associated genes upregulated in DA, suggest an invasion-favoring profile similar to that described in cancer cells, characterized by reduced expression cell cycle check-point and tumor suppressor genes combined with increased expression of invasion genes.

Additionally, Ubxd2, Fzd4, Fkbp7, Olfml2b, Gsdmdc1 and the transcriptional co-repressor Ncor1 were among the genes downregulated in DA and with increased expression in DA.F344(Cia5d). Gtlf3b (predicted), a gene trap fragment with unknown function, was among the most significantly differentially expressed genes (P = 0.000025; 2.2-fold difference; Table 3). The greater than twofold difference in expression of Olfml2b and Gsdmdc1 was confirmed with quantitative real-time PCR (Figure 4b).

Nine genes or 13.6% of the 66 differentially expressed genes were either estrogen-inducible genes, such as Cxcl10, Vil2, Trim16, Gins3 (predicted), and Gadd45b, or genes involved in modulating the estrogen receptor (ER) signaling such as Stub1 and Stip1. Ncor1 negatively regulates ER-mediated transcription and its levels were also reduced in DA, further suggesting unopposed ER-mediated transcription. The differential expression of Cxcl10, Vil2, Trim16, Gins3, and Gadd45b was confirmed with quantitative real-time PCR (Figure 4a, b). The ERs Esr1 and Esr2 were not differentially expressed in the microarray analysis, and those results were confirmed with quantitative real-time PCR (Figure 4b). There was a trend toward increased expression Esr2 in DA.F344(Cia5d), but that difference did not reach statistical significance (P = 0.093; Figure 4b). Taken together, this pattern of gene expression suggests that the invasive DA FLSs have an enhanced ER activity regulated at different levels that could include reduced degradation of the ER, reduced inhibition of the ER-mediated transcription, and increased levels of estrogen-inducible genes.

Five out of the 66 differentially expressed genes were located within the Cia5d interval (Table 4). The number of genes located within the Cia5d interval found to be differentially expressed between DA and DA.F344(Cia5d) FLSs was greater than would be expected by chance (3.3% observed versus 0.8% expected by chance; P = 0.0044 by χ² with Yates correction; Table 5).

Trim16, Trpv2, and Ncor1 are closely located on chromosome 10q23, raising the possibility that a polymorphism in a regulatory region or intron in this region, or even in one of these genes, could account for the difference in expression detected between the two strains.