Transcriptomic signatures of feline chronic gingivostomatitis are influenced by upregulated IL6

Feline chronic gingivostomatitis (FCGS) is a relatively common and debilitating disease characterized by bilateral inflammation and ulceration of the caudal oral mucosa, alveolar and buccal mucosa, and varying degrees of periodontal disease. The etiopathogenesis of FCGS remains unresolved. In this study, we performed bulk RNA-seq molecular profiling of affected tissues derived from a cohort of client-owned cats with FCGS compared to tissues from unaffected animals, to identify candidate genes and pathways that can help guide future exploration of novel clinical solutions. We complemented transcriptomic findings with immunohistochemistry and in situ hybridization assays to better understand the biological significance of the results and performed RNA-seq validation of biologically relevant differentially expressed genes using qPCR assays to demonstrate technical reproducibility. Transcriptomic profiles of oral mucosal tissues in cats with FCGS are enriched with immune- and inflammation-related genes and pathways that appear to be largely influenced by IL6, and include NFKB, JAK/STAT, IL-17 and IFN type I and II signaling, offering new opportunities to develop novel clinical applications based on a more rational understanding of the disease.

RNA-seq, differential gene expression, cluster analysis and qPCR.We established transcriptomic profiles using bulk RNA-seq on caudal oral mucosal tissues from cats diagnosed with FCGS and gingival tissues from cats diagnosed with PER but not with FCGS, as well as matching tissues from healthy controls.We investigated the RNA-seq data to discover genes dysregulated in FCGS and not (or to a lesser degree) in PER, to distinguish characteristics of FCGS from general inflammation of oral tissues.Of the 19,588 protein-coding genes annotated in Felis_catus_9.0 (Ensembl release 105), 4,207 genes were differentially expressed (q < 0.05) in FCGS when compared to HOM, 748 genes in PER compared HGIN, and 2,891 genes in FCGS when compared to PER (Tables S1-S3).Principal component analysis showed that samples clustered according to clinical phenotype, indicating that the primary global signal in the gene expression profiles distinguished diseased from healthy samples (Fig. 1).Altered expression of 9 biologically relevant genes in FCGS cases was validated with qPCR (Fig. 2).

Functional enrichment analyses.
To complement differential gene expression findings and gain functional insights, we conducted Gene Set Enrichment Analysis (GSEA) 28,29 using the RNA-seq data.When comparing FCGS to HOM, enriched gene sets were predominantly associated with inflammation and the immune response (e.g., NFKB and JAK/STAT) and with cytokine signaling (e.g., IL-6, IFN type I and II, IL-17) (Fig. 3, Tables S4-S6, Figs.S1, S2).Notably, IL6 was either the top leading-edge gene or was among the top leading-edge genes in most inflammation-and immune-related pathways enriched in FCGS; this observation was also reflected when comparing the expression of cytokines and chemokines in the RNA-seq dataset among the different groups (Fig. 4, Table S7).The expression profiles of FCGS compared to HOM and PER revealed enrichment of immune cells with a predominantly myeloid lineage identity (e.g., macrophages, microglia) led by genes typically expressed by myeloid cells (e.g., CD14, CSFR1, CSFR3, HCK, CYBB; The Human Protein Atlas, https:// www.prote inatl as.org) 30 (Table S8, Fig. S3).
Non-host RNA-seq reads, immunohistochemistry and in situ hybridization.To determine whether viral genomic sequences were present in analyzed tissues, we mapped RNA-seq reads that did not align with Felis_catus_9.0 to reference sequences.Results showed sequences mapping to reference genomes of FCV, puma feline foamy virus (PFFV), feline leukemia virus, and feline herpesvirus, among others (Table S9).Of the candidate viruses observed, only FCV and PFFV were significantly more common in cats with FCGS (p = 0.009 and 0.029, respectively, Mann-Whitney test).To determine whether FCV antigen or genome was present in affected tissues, we performed IHC and ISH.Both assays failed to confirm the presence of FCV in any FCGS case (n = 13) while results for positive and negative controls were appropriate (Fig. 5).

Discussion
Despite its clinical relevance, the etiopathogenesis of FCGS remains unresolved.In this study, we performed bulk RNA-seq molecular profiling of affected tissues derived from a cohort of client-owned cats with FCGS and used tissues from cats with PER and unaffected tissues as comparative sets, to identify candidate genes and pathways that can help guide future exploration of novel clinical solutions for FCGS.We complemented transcriptomic findings with IHC and ISH assays to better understand the biological significance of the data and performed RNA-seq validation of biologically relevant differentially expressed genes using qPCR assays to demonstrate technical reproducibility.
Overall, the transcriptional profiles of FCGS tissues were largely dominated by immune-and inflammationrelated genes and signaling pathways including NFKB, JAK/STAT and IFN type I and II signaling, indicating that FCGS is an inflammatory disease potentially related to antigenic stimulation 6,31,32 .As was expected, cluster analysis showed that all samples segregated according to group assignment, indicating that FCGS and PER have distinct molecular phenotypes.Additionally, there were general similarities in the expression patterns of FCGS when compared to those previously reported in three cats that were refractory to surgical therapy 7 , suggesting Notably, most inflammation-related pathways found to be enriched in FCGS tissues appeared to be heavily influenced by IL6.This is a relevant finding given that dysregulated expression of IL-6 (encoded by IL6) could underlie some of the local and systemic events known to occur in cats with FCGS.In general, IL-6 is a finely regulated pleiotropic cytokine that signals via the JAK/STAT pathway.IL-6 can be produced by multiple cell types including mesenchymal, endothelial, epithelial, and immune cells [33][34][35] .During homeostasis, IL-6 blood levels are hardly detectable but can rapidly increase upon stress, tissue injury or antigenic stimulation.Once in circulation, IL-6 activates hepatocytes to produce acute-phase proteins and modulates innate and adaptive immunity to promote healing and help eliminate infections 36,37 .
Sustained overexpression of IL-6 has been shown to promote chronic inflammation and increase susceptibility to viral infections 33,35,37,38 .It is possible that excessive production of IL-6 in cats with FCGS explains the chronic FCV and PFFV infection and elevated globulin and acute-phase protein levels in blood typically found, and may be directly involved in perpetuation of oral mucosal inflammation and ulceration.Similarly, IL-6 signaling is a known inhibitor of apoptosis 39,40 , which might explain the often proliferative nature of mucosal lesions observed in cats with FCGS.Interestingly, recombinant feline IFN-omega has been shown to reduce IL6 expression and IL-6 levels 41 , which could at least partially explain the clinical response reported in some cats with FCGS 14,15 .
Another interesting finding in FCGS tissues in agreement with a previous report 7 was overexpression of IL17A and enrichment of its corresponding signaling pathway.Importantly, IL-17 (encoded by IL17A) is considered the signature cytokine of Th17 cells 42 .Although Th17 cells play an important protective role against microbial and viral pathogens, they are also implicated in autoinflammatory and autoimmune pathology [42][43][44][45] .IL-6 promotes polarization of naïve CD4 + T cells towards a Th17 phenotype 33,35,36 , thus the increased IL17A activity observed in FCGS tissues may be due to IL-6 signaling.However, the cellular origin and biological impact of IL17A overexpression in FGCS, and whether activation of the IL-17 pathway plays a pathogenic or protective role in affected cats is unknown but warrants further investigation.
Coinciding with previous reports [19][20][21][22][23] , this study showed an association between FCGS and FCV and PFFV infection.However, the very low numbers of RNA-seq reads mapping to corresponding viral genomes in affected tissues, and the fact that IHC and ISH failed to detect FCV antigen and genome signals, respectively, suggest that at least FCV does not replicate in areas of mucosal ulceration, or that it is present in such low numbers that it is undetectable using the IHC and ISH techniques used in this study.Regardless, the consistent overexpression of IFNG, which encodes IFN-gamma, and enrichment of IFN type I and II pathways suggest an immune response to viral stimulation 46,47 .Given that IFNG induces expression of CXCL9, CXCL10, and other chemotactic molecules 48 , IFN-gamma signaling is likely to play an important role in immune cell traffic in affected oral mucosal tissues, including attraction of T and B cells.Taken together, these observations implicate viral infection as a possibly required or aggravating element in the pathogenesis of FCGS.www.nature.com/scientificreports/Unexpectedly, the gene expression signatures observed in FCGS tissues revealed a predominantly myeloid lineage identity, which suggests that despite the heavy presence of lymphoid infiltrates, the innate system is transcriptionally more active and is thus likely to be an important driver of the dysregulated immune response.This finding conflicts with the hypothesis that FCGS is primarily a T cell driven disease 7 .As macrophages and monocytes are the main producers of IL-6 34,35 , these new findings support a hypothesis that IL-6 dysregulation underlies some of the pathological events.Regardless, it would be interesting to determine if the typical plasma cell infiltrates correspond to pathogenic memory plasma cells, and how these might contribute to disease.Similarly, whether immune regulatory mechanisms are involved in pathogenesis should be investigated, especially given that FOXP3 appeared differentially upregulated in FCGS tissues.
The caudal oral mucosa of cats is a relatively thin, non-masticatory mucosal barrier that is exposed to mechanical trauma during mastication, and to pathogenic toxins (e.g., LPS) derived from the oral microbiota 49 .Given this environment, it is possible that the caudal oral mucosa is particularly susceptible to immune stimulation, and that in some cases such stimuli result in chronic inflammation.This could explain why extraction of premolar and molar teeth, which typically come in direct contact with caudal and buccal oral mucosal surfaces, results in remission in some animals 9,10 .Alternatively or in addition, based on the transcriptional signatures observed in this study and the epidemiological patterns previously reported 18 , it is possible that viruses and environmental stressors are involved in the pathogenesis of FCGS; both are likely to contribute to upregulation of IL-6 49 .Such a scenario raises the question of why, when subject to similar conditions (i.e., multi-cat environments), only some individuals develop FCGS.Possible explanations could include unknown exposures and/or genetic or epigenetic predisposition.
From a translational medicine perspective, the results of this study provide rational targets for clinical diagnostics and therapy in cats with FCGS.For example, IL-6 could be investigated as a potential diagnostic and prognostic marker that could be used to stage and grade FCGS, determine the best treatment modalities or predict therapeutic response.Similarly, given the precedent of successful targeted inhibition of IL-6 signaling in people diagnosed with certain chronic inflammatory conditions and different forms of cancer 33,34,36,40 , comparable approaches could be tested in cats with FCGS.It should be noted that the expression signatures reported Figure 2. Technical validation of RNA-seq data using qPCR experiments.Scatterplots of qPCR assays (ddCt-RPL13A, X axis) of 9 genes plotted against the results of RNA-seq (RNAseq-log2FC, Y axis), where RNAseq-log2FC is the difference between the log2 normalized expression for each sample compared to the average log2 normalized expression for healthy control samples.The measurements with both gene expression quantification platforms were in excellent agreement for all phenotypes.
here capture global trends, but further studies are required to determine their biological impact.Importantly, bulk RNA-seq techniques using clinical samples do not allow single-cell inferences or insights of how different cell types interact with each other.Therefore, the observations made here represent hypotheses that will require testing using targeted, single-cell and/or spatially resolved approaches, and ideally in vitro and in vivo validation experiments.
We conclude that the transcriptomic profile of oral mucosal tissues in cats with FCGS is enriched with immune-and inflammation-related genes and pathways that appear to be largely influenced by IL6, and include NFKB, JAK/STAT, IL-17 and IFN type I and II signaling.

Materials and methods
Clinical samples.Study material consisted of cryopreserved and formalin-fixed paraffin-embedded (FFPE) oral mucosal and gingival tissues and cryopreserved serum samples obtained from cats presented to the Dentistry and Oral Surgery Service at the Cornell University Hospital for Animals (FCGS and PER tissues), and archival tissues stored in the Cornell Veterinary Biobank (HOM and HGIN tissues).Clinical sample collection procedures were performed in accordance with a protocol (#2005-0151) approved by Cornell University's Institutional Animal Care and Use Committee while animals were receiving standard-of-care intervention under general anesthesia, which was supervised by a board-certified veterinary anesthesiologist following standard-ofcare clinical practices and protocols, as determined by individual patient needs.Informed consent to authorize the use of tissue samples and clinical data for research purposes was obtained from cat owners prior to sample collection, and undue harm was never inflicted to any animal for the purposes of this study; all methods were performed in accordance with the relevant guidelines and regulations as approved in the protocol previously listed.Clinical diagnoses and sample collection were supervised by a board-certified veterinary dentistry specialist (SP).None of the FCGS cats enrolled had been previously treated by surgical means or were considered refractory to therapy at the time of sampling.Control tissue samples were collected from healthy cats, including replicate oral mucosal samples from two separate locations from 3 cats.Age and weight differences across groups were compared using JMP 15 (SAS Institute Inc., Cary, NC); box plots were generated using BoxPlotR (http:// shiny.chemg rid.org/ boxpl otr/) 50 .

RNA isolation, library preparation and sequencing.
Frozen tissue (~ 1 g) was homogenized in 2 mL of Trizol (Thermo Fisher) using 2.8 mm ceramic beads (Hard Tissue Homogenizing Mix, VWR).RNA was extracted with a modified Trizol method as follows: after the addition of chloroform and phase separation of the Trizol lysate, the aqueous phase was combined with an equal volume of 100% ethanol and loaded onto a Zymo-Spin column and purified using the Quick-RNA Prep Kit (Zymo Research).For all samples, RNA concentration was measured with a Nanodrop (Thermo Fisher), and integrity was determined with a Fragment Analyzer (Agilent).If high molecular weight material was evident in the Fragment Analyzer trace, indicating the presence of genomic DNA, samples were treated with DNAse following the instructions of the Zymo RNA RNA Clean & Concentrator Kit (Zymo Research).Ribosomal RNA was depleted with the NEBNext rRNA Depletion Kit v2 (Human/Mouse/Rat; New England Biolabs) using 500 ng input total RNA.All RNA-seq libraries were generated with the NEBNext Ultra II Directional library prep kit (New England Biolabs) and 2 × 150 nt paired-end reads were generated on a NovaSeq6000 instrument (Illumina).
RNA-seq Analysis.Raw reads were trimmed for low-quality and adaptor sequences and filtered for minimum length with TrimGalore (http:// www.bioin forma tics.babra ham.ac.uk/ proje cts/ trim_ galore/), a wrapper for cutadapt 51 and fastQC (http:// www.bioin forma tics.babra ham.ac.uk/ proje cts/ fastqc/) using Figure 3. Functional enrichment analyses.The scatterplots depict normalized enrichment scores (NES) as calculated by GSEA, for represented gene sets from MSigDB for analyses of FCGS vs HOM (i.e., FDR q-value < 0.05, X axis) and comparison to PER vs HGIN (Y axis).Gene sets differentially enriched or depleted (i.e., FDR q-value < 0.05) in FCGS compared to PER appear in purple.For all graphs, relevant gene sets and pathways are labeled.The size of the dots is based on -1og10 transformed p-values based on enrichment scores comparison of FCGS vs PER.parameters '-nextseq-trim = 20 -O 1 -a AGA TCG GAA GAG C -length 50 -fastqc' .Trimmed reads were mapped to the reference genome/transcriptome (Ensembl felCat9) with STAR 52 using these parameters: '-outSAMstrandField intronMotif, -outFilterIntronMotifs RemoveNoncanonical, -outSAMtype BAM SortedByCoordinate, -outReadsUnmapped Fastx and -quantMode GeneCounts' , which also generated raw count outputs per annotated gene.Samples with low rates of reads mapping to the FelCat9 reference genome or transcriptome were excluded from gene expression analysis, including three FCGS cases for which the reads were still analyzed for evidence of feline viruses (cases 4, 9, 13).
Sample clustering and differential gene expression were analyzed with SARTools 53 and DESeq2 54 using these parameters: 'fitType parametric, cooksCutoff TRUE, independentFiltering TRUE, alpha 0.05, pAdjustMethod BH, typeTrans VST, and locfunc median' .Feline gene symbols were converted to human gene symbols using Biomart (Ensembl) one-to-one orthology assignments to enable analysis with gene sets in MSigDB 28 .The human ortholog gene symbols and log2-fold-change values for expressed genes (at least one group with average normalized counts > 50) were used for GSEA 29 'Preranked' analysis.
Reads that did not map to the feline reference genome were defined as non-host and separately analyzed for evidence of viral infection.Non-host reads were mapped to the RefSeq sequence for candidate feline virus genomes using bowtie2 using local alignment settings (-local).Counts per million (CPM) was calculated as the number of mapped reads per million non-host reads, and p-values were determined comparing FCGS vs HOM samples with the Mann Whitney test (https:// astat sa.com/ Wilco xonTe st, default parameters).qPCR validation.Gene expression levels were validated using real-time reverse transcription polymerase chain reaction (qPCR).Genes were selected for validation based on several criteria, including biological relevance to the disease state (e.g., interleukins and chemokines, immunometabolism), evidence for differential expression of gene families (e.g., metalloproteinases, keratins, claudins), level of detection in the RNA-seq dataset, and altered expression in a majority FCGS cases.cDNA was synthesized as previously described 55,56 .All cDNA reactions were diluted 20-fold with water prior to qPCR reaction setup.Primer pairs (Table 2) were designed with Primer-BLAST (NCBI), separated by an intron to minimize amplification of residual contaminating genomic DNA and allow identification of alternate amplicons with melt curve analysis.RPL13A was selected as the endogenous control gene, as this gene showed minimal variation across samples in the RNA-seq data.Each primer pair was validated using a standard curve of six four-fold serial dilutions of a representative sample of pooled cDNA.A 'No-RT' control containing RNA but lacking M-MuLV enzyme and one 'no template' control lacking any cDNA sample was included for each primer pair standard curve validation.Primer pairs that did not generate signal in < 35 cycles or that exhibited non-quantitative performance (i.e., < > 2-cycle shifts for fourfold dilution series), non-specific signal in negative controls, or variable amplicon identities as determined by melt curve analysis were excluded.All primer pairs passed validation by standard curve testing.Each qPCR reaction was prepared in 8 μL reaction volumes in an optically clear 384-well PCR plate with seal using the Luna Universal qPCR Master Mix (New England Biolabs) with 0.25 μM primers and 4 μL pre-diluted sample cDNA.All reactions were performed in triplicate using a Roche LightCycler 480 instrument.Cycles were as follows: initial incubation 5 min at 95 °C; followed by 45 cycles of 30 s at 95 °C; 30 s at 60 °C; 10 s at 72 °C with data acquisition; and final a melt curve with a ramp from 60 to 95 °C at 2 °C per second.Melt curve analysis was used to identify and exclude reactions with alternative amplicons.For relative quantification estimates for each target gene, the ΔΔCt value [ΔCtSAMPLE − ΔCtREF] was calculated for each sample, where ΔCtSAMPLE = average (target gene Ct) − average (all endogenous control Ct) and ΔCtREF was defined as the average ΔCtSAMPLE for the normal samples.The normalized relative amount of the target gene is 2−ΔΔCt77.
Immunohistochemistry and in situ hybridization assays.For IHC, selected FFPE tissue blocks were processed for antigen retrieval and detection by using an automated IHC processor (Leica Bond-Max, Leica Biosystems, Buffalo Grove, Illinois, USA), as previously described 57 .Briefly, sections were dewaxed (cat# AR9222, Bond Dewax Solution, Leica) and processed for epitope retrieval (cat# AR9961 or AR9640, Bond Epitope Retrieval solution, Leica) followed by incubation with an FCV primary antibody (ABCAM Cat#  www.nature.com/scientificreports/AB33990) at a 1:200 dilution for 60 min.Next, polymeric alkaline phosphatase conjugated anti-mouse IgG (cat# PV6110, Powervision™ Poly-AP Anti-Mouse IgG, Leica) was applied for 30 min, followed by Red Detection™ (cat# DS9390, Bond Refine Red Detection Kit, Leica) for 15 min, and hematoxylin counterstain for 5 min.Archival FCV-infected and non-infected tissues were used as positive and negative controls.The same selected FFPE tissues used for IHC were used for ISH.Probes were designed in collaboration with Advanced Cell Diagnostics (ACD, Cat.No. 472281).Briefly, 5-μm sections were cut and stored at − 80 °C prior to staining.Sections were deparaffinized in xylene, washed with ethanol, and dried.Staining was performed according to the manufacturer's protocol for colorimetric ISH.Slides were treated with H 2 O 2 (ACD) to block endogenous peroxides for 10 min.Slides were antigen-retrieved by boiling for 15 min in antigen retrieval solution (ACD) and then treated with proteinase K (ACD) for 30 min.Slides were then incubated for 2 h with FCV probes, and then 6 amplifications steps were performed with ACD reagents.The bacterial gene DapB probe (ACD) was used as a negative control.Slides were developed with DAB chromogen (ACD) for 10 min and counterstained with Mayer's hematoxylin (Dako).Archival FCV-infected and non-infected tissues were used as positive and negative controls.

Figure 1 .
Figure 1.Cluster analysis and differentially expressed genes.(A) PCA plot depicting clustering of RNA-seq gene expression profiles showing grouping based on assigned phenotypes and tissue type according to principal components 1 and 2; reverse triangles correspond to gingival samples, dots correspond to caudal oral mucosal samples, labels indicate case numbers as listed in Table 1.(B) Scatterplot depicting expressed genes based on log2(FC) of affected vs control samples for each disease, with purple points indicating differential expression in FCGS vs PER; genes used for qPCR validation experiments are labeled.

Figure 4 .
Figure 4. Cytokine and chemokine expression profiles.(Panel A) shows a scatterplot depicting log2(FC) values of FCGS vs HOM (X axis) and PER vs HGIN (Y axis), with genes significantly enriched in FCGS vs PER shown in purple.The size of the dots is based on log2(FC) of FCGS vs PER.(Panel B) shows box plots depicting the normalized counts distribution for IL6, IL17A, IFNG, and CXCL10 according to phenotype.

information Assays performed FCV results Case no Sex Age (years) Breed Body weight (kg) Phenotype RNA-seq FCV IHC and ISH RNA-seq reads mapping to FCV FCV IHC FCV ISH
that the molecular mechanisms underlying FCGS are maintained during the natural course of the disease as long as local tissue inflammation persists and regardless of historical surgical interventions.

Table 2 .
Primer pairs used for qPCR.Genes, primer pairs, and product size (bp = base pairs) used for qPCR analysis.*Endogenous control gene.