key: cord-0767718-omihw8qo
authors: Malbon, A.J.; Meli, M.L.; Barker, E.N.; Davidson, A.D.; Tasker, S.; Kipar, A.
title: Inflammatory Mediators in the Mesenteric Lymph Nodes, Site of a Possible Intermediate Phase in the Immune Response to Feline Coronavirus and the Pathogenesis of Feline Infectious Peritonitis?
date: 2018-12-17
journal: J Comp Pathol
DOI: 10.1016/j.jcpa.2018.11.001
sha: f8e1836096b394c7dc65a4bcf35027eb49a930d9
doc_id: 767718
cord_uid: omihw8qo

Feline infectious peritonitis (FIP) is an almost invariably fatal feline coronavirus (FCoV)-induced disease thought to arise from a combination of viral mutations and an overexuberant immune response. Natural initial enteric FCoV infection may remain subclinical, or result in mild enteric signs or the development of FIP; cats may also carry the virus systemically with no adverse effect. This study screened mesenteric lymph nodes (MLNs), the presumed first site of FCoV spread from the intestine regardless of viraemia, for changes in the transcription of a panel of innate immune response mediators in response to systemic FCoV infection and with FIP, aiming to identify key pathways triggered by FCoV. Cats with and without FIP, the latter with and without FCoV infection in the MLN, were compared. Higher expression levels in FIP were found for toll-like receptors (TLRs) 2, 4 and 8. These are part of the first line of defence and suggest a response to both viral structural proteins and viral nucleic acid. Expression of genes encoding inflammatory cytokines and chemokines, including interleukin (IL)-1β, IL-6, IL-15, tumour necrosis factor (TNF)-α, CXCL10, CCL8, interferon (IFN)-α, IFN-β and IFN-γ, was higher in cats with FIP, consistent with inflammatory pathway activation. Expression of genes encoding transcription factors STAT1 and 2, regulating signalling pathways, particularly of the interferons, was also higher. Among cats without FIP, there were few differences between virus-positive and virus-negative MLNs; however, TLR9 and STAT2 expression were higher with infection, suggesting a direct viral effect. The study provides evidence for TLR involvement in the response to FCoV. This could open up new avenues for therapeutic approaches.

Feline infectious peritonitis (FIP) is a well-known and widely distributed coronavirus-induced disease of felids. With as yet no effective vaccine or viable treatment options, FIP is almost invariably fatal, and understanding the pathogenetic and immunological mechanisms involved in disease development is crucial to aiding chances of combating FIP and identifying novel avenues for possible treatment.

After initial enteric infection, feline coronavirus (FCoV) may spread beyond the intestine, resulting in a monocyte-associated viraemia, with or without the development of FIP. In cases progressing to FIP, which may have a time lag of weeks to years, viral and host factors combine to turn an initial, usually subclinical, enteritis into an overt immunemediated disease (Pedersen et al., 1981; Kipar and Meli, 2014) . Much research has focussed on viral mutations and has partially elucidated the function of various viral proteins in the pathogenesis of FIP. The viral spike (S) protein gene has been of particular interest, and a switch from methionine (M) to leucine (L) at amino acid residue 1,058 (M1058L) has been strongly associated with a gain of virulence (Chang et al., 2012) . A second switch from serine (S) to alanine (A) at amino acid residue 1,060 (S1060A) distinguished tissue-associated FCoV in a further small subset of FIP cases from FCoV shed with the faeces by healthy cats (Chang et al., 2012) . These mutations have since been associated with systemic spread of FCoV, rather than providing proof of virulence (Porter et al., 2014; Barker et al., 2017; Felten et al., 2017a) , so the two forms are subsequently referred to here as 'systemic' and 'enteric' FCoV. Early experiments demonstrated that not all cats are susceptible to FCoV infection, even with known pathogenic strains (Pedersen and Boyle, 1980) , indicating the importance of host genetic factors/immune mechanisms in disease development. More recently it was shown that cultured monocytes from different cats vary in their ability to sustain viral replication, again suggesting that there is a subset of animals who can resist disease (Dewerchin et al., 2005; Tekes et al., 2010) . Monocytes/macrophages are not the only cell type beyond enterocytes that may be infected by FCoV, but they are also key cells in the innate immune defence system. They are able to detect pathogen-associated molecular patterns (PAMPs), triggering a number of intracellular signalling pathways leading to activation of an antiviral state in the host (Abbas et al., 2017) . Chief amongst these pathways are those triggered by engagement of toll-like receptors (TLRs); highly evolutionarily conserved, membrane-bound pathogen recognition receptors (PRRs) (Lester and Li, 2014) . Their presence on both the cell surface membrane and internal membrane-bound vesicles allows detection of external and internal PAMPs; their ligands include those associated with viruses, bacteria and fungi (Arpaia and Barton, 2011) . Downstream mediators include inflammatory cytokines and interferons that have been assessed in cats with FIP, with sometimes conflicting results (Dean et al., 2003; Kipar et al., 2006b) . Interferons and the inflammatory cytokine interleukin (IL)-6 can activate members of the signal transducer and regulator of transcription (STAT) family with downstream effects on replication, differentiation or inflammatory potential (Aaronson and Horvath, 2002) . Cats with a compromised immune system appear to be more susceptible to FIP, while, paradoxically, the lesions are caused by an excessive immune response (Pedersen, 1987 (Pedersen, , 2014 Kipar and Meli, 2014) . This has been attributed, at least in part, to increased viral replication in immunosuppressed animals and, therefore, an increased likelihood of viral mutations occurring and accumulating (Poland et al., 1996) .

TLRs have been associated with susceptibility to many diseases, including chronic inflammatory, viral and more specifically coronaviral diseases (e.g. severe acute respiratory syndrome, SARS) (Dosch et al., 2009) . Intriguingly though, both TLR stimulation and antagonism/knock-outs have contributed to exacerbation of disease in different contexts and there exists considerable crossover between receptors and their potential ligands (Arpaia and Barton, 2011) .

When FCoV is able to leave the intestine, the mesenteric lymph nodes (MLNs) are the presumed first site of viral spread, potentially representing the interface between local and systemic immune response; support for this assumption are FIP cases that present only with MLN lesions (Kipar et al., 1999) . We therefore chose the MLN as our organ of interest, with the aim of comparing key mediators of the innate immune system between uninfected cats and FCoV-infected cats with and without FIP. We hypothesized that in addition to an excessive proinflammatory cytokine response, there would be a deficient interferon response, and aimed to gain an insight into which TLR pathways are involved in triggering this response. We also wished to further evaluate the presence and significance of previously published viral S gene variations and determine whether a connection with the host immune response could be detected.

The study was undertaken on cats that had all been seen initially as patients at the university small animal clinics and local veterinary practices of Bristol, UK, or Zurich, Switzerland, and humanely destroyed with or without FIP for clinical reasons unrelated to this study (Table 1) . A post-mortem examination was performed on each cat with owner consent and samples of MLN were collected into RNAlater Ò (Qiagen, Hombrechtikon, Switzerland) within 2 h of euthanasia and stored at À80 C until use. The Bristol cases form part of the University of Bristol FIP Biobank built up as a resource for multiple studies; many of these cases were utilized previously (Porter et al., 2014; Barker et al., 2017) .

Group 1 (G1) comprised of 40 control cats confirmed to not have FIP and with an alternate confirmed diagnosis (Tables 1A and 1B) , and group 2 (G2) consisted of 30 cats confirmed to have FIP (Table 1C) . A diagnosis of FIP was based on relevant clinical findings and compatible gross and/or histological lesions together with immunohistological demonstration of FCoV antigen-positive macrophages within typical lesions (Kipar et al., 1998) . The immunohistochemistry was carried out as previously described (Kipar et al., 1998) , using a mouse monoclonal primary antibody (clone FIPV3-70 SC 65653, Santa Cruz, Heidelberg, Germany). Based on the results of the reverse transcriptase quantitative polymerase chain reaction (RT-qPCR) for FCoV undertaken on the MLNs, group G1 was then subdivided into G1+ (FCoV positive) and G1À (FCoV negative).

RNA extraction was carried out using the RNeasy Plus Minikit Ò (Qiagen) according to the manufacturer's protocol. Briefly, 30 mg of MLN tissue were disrupted in extraction buffer using a tissue homogenizer (Mixer-Mill 300, Retsch, Haan, Germany) for 40 sec at 30 Hz before on-column extraction and elution of RNA. As pilot tests revealed that significant genomic DNA contamination remained, an optional DNase step was included prior to use of the Superscript IV VILO Ò kit (ThermoFisher Scientific, Waltham, Massachusetts, USA) for cDNA synthesis, following the manufacturer's instructions, in order to avoid possible interference with the RT-qPCR results. Starting RNA levels were equilibrated between samples to 400 ng/ml, using a NanoDrop 2000 Ò (ThermoFisher Scientific). Samples were further diluted 1 in 20 prior to RT-qPCR.

TaqMan RT-qPCR was performed on an Applied Biosystems 7500 Fast PCR System Ò (ThermoFisher Scientific) using newly developed, or previously published, primer and probe protocols for: FCoV; feline TLR 1 to 9; STAT 1 to 3; interferon (IFN)-a, -b and -g; IL-1b, -6, -10, -15, and -17; tumour necrosis factor (TNF)-a; CXC motif chemokine 10 (CXCL10); CC motif chemokine ligand 8 (CCL8); transforming growth factor (TGF)-b1; and glyceraldehyde 3-phosphate dehydrogenase (GAPDH), as the reference gene (Table 2 ) . This gene was chosen based on previous experience in our laboratory and following reference gene comparisons during optimization. All primers and probes were manufactured by Microsynth (Balgach, Switzerland). The hydrolysis probes were labelled with a 5 0 reporter dye FAM (6-carboxyfluorescein) and a 3 0 quencher TAMRA (6-carboxy-tetramethylrhodamine). Those primers and probes that were newly developed were designed using Primer Express Ò software (v3.0.1, Thermo Fisher Scientific) to span an exoneexon junction. These were tested for specificity by conventional PCR of a test sample, gel electropho-resis, sequencing of the resulting extracted band (Microsynth) and evaluation using NCBI BLAST. Conditions were as for RT-qPCR except for omission of the probe. Primer concentrations for this step were 900 nM. Varying primer/probe concentrations were then tested to determine the optimal efficiency and dynamic range as well as replicability using a sample dilution series. All final protocols (Table 2 ) had an efficiency >95%. Those previously published were tested again in our system, omitting the conventional RT-PCR step. Each reaction comprised 12.5 ml Taq-Man Fast Universal Master Mix Ò (ThermoFisher Scientific), with 2.5 ml cDNA, primer and probe volumes as per Table 2 , made up to 25 ml with RNasefree water. The thermal profile for all RT-qPCRs was: 50 C for 2 min, 95 C for 10 min, and 45 cycles of 95 C for 10 sec and 60 C for 1 min. All samples were run in duplicate and any samples with discordant results were repeated. Data collection occurred during the extension phase at 60 C. Appropriate controls were included in each run.

The Applied Biosystems 7500 Software Ò v2.0.6 was used to visualize results and allocate a quantification cycle (C q ) to each sample, and the threshold was equilibrated between runs for each target.

The particular codons of interest within the FCoV S gene were 1,058 and 1,060 (Chang et al., 2012) . With reference to the sequence used in the original paper, the mutations in question appear to be at positions 1,048 and 1,050 rather than 1,058 and 1,060 as previously described, and will be referred to subsequently by the former numbers. Following initial FCoV RT-qPCR, all positive samples not analysed for previous studies by Porter et al. (2014) or Barker et al. (2017) underwent additional conventional RT-PCR and Sanger sequencing targeting the S gene region of interest. PCR was performed using the previously published degenerate primers (Porter et al., 2014) . Each reaction comprised 10 ml Phusion Flash Master Mix Ò (ThermoFisher Scientific), with 2 ml cDNA, 0.5 mM each of forward and reverse primers, made up to 20 ml with RNase-free water. Reactions were run on a T Professional Ò thermocycler (Biometra GmbH, G€ ottingen, Germany) with the following thermal profile: 98 C for 10 sec, 40 cycles of 98 C for 1 sec, 52 C for 5 sec, 72 C for 3 sec, followed by 72 C for 1 min. Appropriate controls were included in each run.

The reaction product then underwent gel electrophoresis. Bands of appropriate size were extracted using the GeneJET Gel Extraction Kit Ò (ThermoFisher Scientific) and submitted for Sanger sequencing at a commercial laboratory (Microsynth). When no band was visible, the reaction was repeated using 50 cycles and the product was subjected to gel electrophoresis. Samples still appearing negative were cut out in the region of the expected band, purified and re-subjected to PCR. The bioinformatics software Geneious 9.1.7 Ò , (Biomatters Limited, Silkeborg, Denmark) was used to map the resulting sequences to the reference gene FCoV C1Je (Accession number DQ848678) (Chang et al., 2012) . 

Relative mRNA transcription levels were calculated using the comparative C q method (Pfaffl, 2001) . The C q of each target was first normalized to GAPDH as the endogenous reference (DC q ) and then expressed relative to the G1 DC q mean as the calibrator (2 ÀDDCq ). For FCoV RT-qPCR results, the mean of G1+ was instead used as the calibrator (to allow for visualization graphically).

The statistical programme SPSS Statistics v.25 Ò (IBM, Armonk, New York, USA) was used for all analyses and graphical data presentation. Data were first assessed for normality using a ShapiroeWilk test. As almost all data failed the test, non-parametric measures were applied. A two-tailed ManneWhitney test with a significance level of P #0.05 was used to compare results between groups for each target molecule. Firstly, cats with and without FIP (G1 versus G2) were compared, followed by comparisons between each of the three groups (G1À, G1+ and G2) in turn. Within G2, comparisons were made between cats with and without cavitary effusions and with and without histologically observed FIP lesions in the MLNs. Correlation between relative FCoV levels and inflammatory mediator gene expression levels, and also between individual inflammatory mediator gene expression levels, was analysed within G2 using a one-tailed Spearman's rank test. Here a cut off of P #0.01 was used, with P #0.05 indicating weak correlation.

Signalments of the cats are shown in Tables 1AeC. All MLN samples from cats with FIP (G2) were positive for FCoV (n ¼ 30). Of the 40 cats without FIP (G1), 10 (25%) also had a positive FCoV RT-qPCR result; these were assigned to a new sub-group (G1+). However, the relative FCoV load was clearly, and significantly, lower in G1+ than in G2 (Fig. 1) .

None of the G1 cats exhibited histological changes suggestive of FIP in any tissue examined, including the MLNs when available for histology (25 of 30 from G1À and seven of 10 from G1+). Inflammation Table 2 Primer and probe sequences used for RT-qPCR and conventional RT-PCR

Reference or accession number Primer and probe sequences ( Taglinger Immune Response to Feline Coronavirus of other aetiologies was observed in the MLNs of two of the 30 G1À cats and none of the G1+ animals. All G1 samples were also negative for FCoV antigen by immunohistochemistry.

The MLNs were available for histological examination in 28 of the 30 cats with FIP. In 21 cases (75%), these exhibited the typical pyogranulomatous lesions, with or without associated serosal lesions on the lymph node capsule (e.g. serofibrinous to granulomatous serositis). All samples with typical pyogranulomatous lesions also showed FCoV antigen in lesional macrophages (Figs. 2a, b) . Seven MLNs had no typical lesions; among these was only one case (G2.19) in which FCoV antigen was detected, in low numbers of macrophages within the marginal sinus, suggesting an early lesion (Figs. 2c, d) . There was no significant difference in FCoV load found between MLNs with and without lesions, although those with lesions had a tendency to higher FCoV levels ( Fig. 1) .

Of the 30 cats with FIP, 22 exhibited effusions (Table 1C) . These were not associated with a higher relative FCoV load in the MLNs in comparison with the cats without effusion (n ¼ 6; data not available for two cats).

In order to evaluate the effect of FCoV infection and FIP on target gene transcription, G1 and G2 were first compared with each other before comparisons between all three groups (G1+, G1À and G2). The assessed target genes are described below according to their positions in immune signalling pathways as first line receptors, inflammatory mediators or signal transducers. Detailed results are provided in Table 3 .

Toll-like Receptors: Relative TLR2, 4 and 8 gene transcription levels were significantly higher in G2 than G1. Within G1 there was no difference between virus-positive and virus-negative MLNs for these TLRs; however, TLR9 gene expression, although not elevated in G2, was significantly higher in G1+ than in G1À (Fig. 3) . a Indicates significance level of P # 0.05. In the first three columns, the second group of the comparison is significantly higher in all cases (e.g. for G1 versus G2, G2 levels are higher). In the FIP columns, the value of the 'present' group is in both cases higher than in the 'absent' group.

In G2 cats, gene transcription levels were compared between MLNs with and without FIP lesions, and in relation to the presence of effusions. A significant difference was found only for TLR2 (higher expression in MLNs with lesions) (Fig. 3 , Table 3 ); in contrast, TLR2 expression appeared slightly lower in cats with effusions (Fig. 3) . A possible trend not reaching significance was for a slightly higher TLR4 expression level in MLNs with lesions, while TLR3 and 9 gene expression levels were slightly lower (Fig. 3) . Investigating this further, we found that TLR3 gene expression levels in G2 MLNs without lesions were also slightly higher than levels in G1 (which were similar to those in G2 MLNs with lesions), suggesting a potential negative regulation of TLR3 by FCoV (Fig. 3) .

Cytokines and Chemokines: Relative IL-1b, IL-6, IL-15, TNF-a, IFN-a, -b, -g, CCL8 and CXCL10 gene transcription levels were all significantly higher in G2 compared with G1 (Fig. 4) . None of these showed any significant difference between G1+ and G1À. For most cytokines, G1+ and G1À clustered together; however, for IL-6, TNF-a and IFN-g, G1+ appeared to cluster slightly between the other two groups (G1À and G2), suggesting a possible intermediate stage (Fig. 4) . Between groups, the fold Fig. 3 . Boxplots of relative levels of TLR gene expression in each group. The amount of target was calculated by 2 ÀDDCq , using fGAPDH as the internal reference gene and expressed as an n fold difference relative to the G1 mean as a calibrator. The boxes depict the median and interquartile (IQ) range with whiskers extending to the highest and lowest values, which are within 1.5 Â the IQ range. Outliers beyond this are individually marked. * marks significant differences between individual groups (P # 0.05) or, where joined by a bar, between G1 as a whole and G2. 2E+, 2Ee, L2+ and L2e represent relative gene expression levels among MLNs of cats with and without effusions/lesions. Fig. 4 . Boxplots of relative levels of cytokine and chemokine gene expression in each group. The amount of target was calculated by 2 ÀDDCq , using fGAPDH as the internal reference gene and expressed as an n fold difference relative to the G1 mean as a calibrator. The boxes depict the median and interquartile (IQ) range with whiskers extending to the highest and lowest values, which are within 1.5 Â the IQ range. Outliers beyond this are individually marked. * marks significant differences between individual groups (P # 0.05) or, where joined by a bar, between G1 as a whole and G2. 2E+, 2Ee, 2L+ and 2Le represent relative gene expression levels among MLNs of cats with and without effusions/lesions. differences in the chemokine gene expression levels (CXCL10 and CCL8) were mainly in the range of 10e100Â, while those for the pyrogenic cytokines (IL-1b, IL-6 and TNF-a) rarely exceeded 10Â. IL-10, IL-17 and TGF-b gene transcription levels showed no intergroup differences (Fig. 4, Table 3 ). For IL-1b, IL-6, IFN-g and CCL8, a possible trend towards increased transcription (not reaching significance) was observed in G2 MLNs with lesions compared with those without (Fig. 4) .

Signal Transducers and Activators of Transcription: STAT1 and 2 gene expression levels were significantly higher in G2 than G1. For both transcription factors, gene expression levels were also higher in G1+ than in G1À, significantly so for STAT2. STAT3 gene expression levels were similar across all groups (Fig. 5) .

In G2, STAT2 gene expression levels were significantly higher in cats with effusions (Fig. 5, Table 3 ). For STAT1, there was an insignificant trend to be higher with effusions (Fig. 5) .

The majority of gene targets elevated in G2 also showed significant positive correlation with relative viral load (Supplementary Table 1 ). These included TLR2 and 4, the cytokines IL-1b, IL-6, together with STAT2, CXCL10, CCL8, IFN-b and IFN-g (P # 0.01). TLR8 and IFN-a gene expression showed weaker correlation (P # 0.05), while STAT1, TGF-b, and TNF-a gene expression showed no correlation, and TLR9 a weak, although significant, negative correlation (Table 4) .

Expression of genes encoding IL-6, IL-17 and STAT3, a 'holy trinity' of autoimmunity (Camporeale and Poli, 2012) , was significantly correlated despite the latter two not showing any correlation with FCoV.

Of the 40 cats shown by RT-qPCR to carry FCoV in their MLNs, 38 had analysable S gene sequences following conventional PCR. From the remaining two cats (G1+ cats 1 and 6) it was not possible to obtain samples of sufficient quality even after repeated attempts (Table 1B) .

Of the 30 cats with FIP, one was infected with FCoV serotype 2 for which the previously described S gene sequence characterization is not applicable (Herrewegh et al., 1998; Barker et al., 2017) . Twenty-six MLN samples contained virus that encoded leucine (M1048L) (cDNA sequence TTG, group. The amount of target was calculated by 2 ÀDDCq , using fGAPDH as the internal reference gene and expressed as an n fold difference relative to the G1 mean as a calibrator. The boxes depict the median and interquartile (IQ) range with whiskers extending to the highest and lowest values, which are within 1.5 Â the IQ range. Outliers beyond this are individually marked. * marks significant differences between individual groups (P # 0.05) or, where joined by a bar, between G1 as a whole and G2. 2E+, 2Ee, 2L+ and 2Le represent relative gene expression levels among MLNs of cats with and without effusions/ lesions. [ [ [[ [ [[ [[ [[ [[ [ [[ [[ TLR2 [[ C [[ [[ [ [[ [[ [[ [[ [[ [[ [[ TLR4 [[ [[ C [[ [ [[ [[ [ [[ [[ [[ [ [[ [[ [[ TLR8 [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ C [ [ CXCL10 [[ [[ [[ [ [[ [[ [[ [ [[ [[ [[ [[ [ C [[ CCL8 [[ [[ [[ [ [ [[ [[ [ [[ [[ [[ [[ [ [ [[ C [[, positive correlation at a significance level of P # 0.01; [, positive correlation at a significance level P # 0.05; Y, negative correlation at a significance level of P # 0.05.

CTG or TTA). The remaining three cats encoded methionine at codon 1,048 (cDNA sequence ATG). The results are shown in Table 1C . Of the eight sequences obtained from the G1+ cats, all encoded leucine (M1048L). Of the three cases encoding methionine at codon 1,048, two encoded alanine at codon 1,050 (S1050A), while the third encoded serine (Table 1B) . The small methionine group size (n ¼ 3), including only one cat that carried the 'enteric' virus (M1048, S1050), was not considered valid for statistical comparison with the leucine group, 'systemic' virus. Instead, individual cases were plotted, revealing the methionine group to fall within the range of the leucine group for every target, including FCoV load (Fig. 1 ).

As predicted from previous studies, the results of the present investigation confirm the complex effect of FCoV on the immune system in association with FIP. The disease is caused by an exaggerated immune response to FCoV, but it is well known that cats can also carry FCoV systemically without developing FIP (Meli et al., 2004) . Here we have assessed some of the key mediators of the innate immune response, focussing on the MLN, the most likely first site of infection beyond the intestine and one of the main sites of viral persistence in experimentally infected healthy animals (Kipar et al., 2010) . By comparing FCoV-positive, lesion-free MLNs from cats affected by diseases other than FIP with both FCoVnegative cats without FIP and cats with FIP, we aimed to separate the direct viral effects from the host effects contributing to FIP in a natural setting.

FIP presents as a spectrum of disease with variable duration rather than as a discrete clinical picture; as such the pathological features also vary. This was reflected by variation in organ involvement and presence of effusions in our case cohort. We therefore also wanted to assess whether the inflammatory mediator production in the MLNs showed any correlation with the form of disease. Although vascular permeability is, to a large extent, cytokine mediated (Takano et al., 2011), and we found upregulation of cytokine genes with a role in vascular permeability in the FIP cases, the inflammatory mediator gene expression profile of the MLNs differed only minimally between cats with and without effusions. This suggests that the MLN is unlikely to make a large systemic contribution to vascular permeability. Similarly, Safi et al. (2017) evaluated inflammatory mediators within peripheral blood mononuclear cells of FIP cats with and without effusions and found little consistent pattern to distinguish those with effusions from those without. Both studies therefore provide further, although indirect, support that vascular endothelial growth factor, which was previously shown to correlate with the degree of effusion in FIP, is key to this phenomenon (Takano et al., 2011) . Interestingly, the presence or absence of histological FIP lesions was not correlated with many significant differences between mediator gene expression. Alongside this, and surprisingly, although FCoV levels appeared higher in association with lesions, they were not significantly so. This may partly explain the lack of significant differences in mediator gene expression. Additionally, it cannot be entirely excluded that the area of the MLN sampled for RNA extraction was not within an FIP lesion and vice versa. An alternative explanation for the lack of significant differences between gene expression for most mediators in MLNs, while many exhibit higher overall levels in FIP, would be that systemic stimulation to upregulate inflammatory mediators is more relevant than local or lesion-specific stimulation. Finally, as trends were occasionally observed when cases with and without effusions and MLN lesions were compared, the lack of significance may also be due to the small group sizes once subgroups were created, which was a limitation of this study.

The MLNs of cats without FIP had significantly lower viral loads than their counterparts from cats with FIP. This confirms previous findings in natural infection, where cats with FIP were reported to carry higher viral loads in haemopoietic and lymphoid tissues, including MLNs, than asymptomatic FCoVinfected cats (Kipar et al., 2006a) . Without the disease, however, the presence or absence of FCoV in the MLN seems not to influence the transcription level of most of our target immune mediators. This would indicate that, in the main, the host response has a greater influence than any direct viral effect. Still, there were exceptions. Even among those mediators not attaining significance, IFN-g, IL-6 and TNF-a showed a trend towards higher gene expression levels in FCoV-positive MLNs. This suggests at least a modest direct viral effect; it may have been masked by low group numbers, requiring a larger sample size to confirm or refute. Another study limitation was the composition of the groups. As all were field cases it was not possible to control for confounding factors (e.g. ensuring control cases were free of any inflammatory processes, that FIP cases were at similar disease stages, and that cats were initially subject to the same FCoV infection pressures).

Inflammatory cytokines have been previously studied in FIP, with conflicting results, possibly dependent on variations in disease form between animals included in the different studies and/or the type of sample/organ evaluated. TNF-a gene transcription, for example, was found to be decreased in the MLNs of cats with FIP compared with FCoV-free specific pathogen-free cats, while IL-1b gene expression was elevated (Kipar et al., 2006b) . In the present study, gene expression for all three pyrogenic cytokines (IL-1b, IL-6 and TNF-a) was upregulated in the MLNs in FIP, as well as that for IL-15 (a stimulator of lymphocyte proliferation). We also found significantly higher transcription levels for the monocyte-recruiting chemokines CXCL10 and CCL8, which have both been found to be upregulated in CrandelleRees feline kidney cells after in vitro FCoV infection (Harun et al., 2013) , indicating a mechanism of monocyte recruitment as a direct viral effect. Our results confirm their relevance in vivo, with recruitment of monocytes as the infected cell type being a potential amplifying step that is worthy of further investigation. The increase in inflammatory cytokine gene transcription supports the observation that an overexuberant inflammatory response is a key factor in the development and progression of FIP. Expression of the gene encoding the antiinflammatory cytokine IL-10 was, in contrast, not upregulated in FIP, implying there was no local brake on the inflammatory process. This is in line with previous findings, where IL-10 expression was higher in the spleen of healthy FCoV-infected cats, but not in the MLN (Kipar et al., 2006b) . Gene expression for the interferons was also higher in FIP, IFN-g being one of the cytokines to show an intermediate level in infected asymptomatic cats in our study. These type I and II interferons have major antiviral roles in the innate immune system. IFN-g in particular has been of interest in FIP, as levels of this potentially protective cytokine tend to be low in the peripheral blood of diseased animals and host gene polymorphisms have been identified that may contribute to resistance against the disease (Gelain et al., 2006; Hsieh and Chueh, 2014) . Similarly to the apparent lack of impact of mediator levels on the presence of lesions or effusions in FIP, this suggests that MLN IFN production has a more local effect.

TLRs have been used for targeted therapy against a number of diseases in human medicine, both with adjuvants and inhibitors; however, veterinary medicine lags behind in this respect (Hennessy et al., 2010; Klingemann, 2018) . Here we identified increased gene expression levels of TLRs 2, 4, 8 and 9 with FIP and FCoV infection, respectively, indicating a possible role for these molecules in FIP and hence identifying them as potential targets for FIP control. Assessment at the protein level would be a useful avenue for further investigations; however, this is particularly challenging in feline studies owing to the lack of availability of appropriate antibodies. In most mammals, TLRs 2 and 4 are located on the cell membranes, while TLRs 8 and 9 are found in intracytoplasmic vesicles, most commonly in professional antigen presenting cells (Lester and Li, 2014) . TLR2, together with TLRs 1, 6 and 10, comprise the TLR1 family (Roach et al., 2005) . These latter three receptors arose through evolutionary gene duplication (Hughes and Piontkivska, 2008; Hennessy et al., 2010) . TLR2 is able to signal as a heterodimer with any of its co-family members in order to allow a wider range of antigen recognition. It is typically responsible for detecting bacterial and fungal components (Beutler, 2009 ). TLR2 has been linked to detection of the SARS-CoV S protein in vitro (Dosch et al., 2009); its upregulation in FIP could indicate that the FCoV S protein is also able to act as a ligand.

TLR4 classically detects lipopolysaccharide; however, one study linked it to protection against murine coronavirus, as TLR4-deficient mice were found to exhibit greater susceptibility to murine hepatitis virus infection. The precise mechanism was not established in that case, but it involved inflammatory cell influx in the TLR4-deficient mice (Khanolkar et al., 2009) . No such protective effect was observed in our study, despite upregulation of TLR4 gene expression in the MLNs in association with FIP, although its individual effect in this case cannot be separated from the mediator milieu.

TLR9 gene expression was not elevated in the MLNs of cats with FIP, but was instead increased in the FCoV-positive MLNs of cats without FIP. Considering that a previous in vitro study found reduced viral replication when TLR9 was stimulated with a synthetic CpG ligand prior to FCoV infection (Robert-Tissot et al., 2012) , the increased gene expression in FCoV-infected cats without FIP could indicate that TLR9 has a protective effect, which may even have helped prevent the development of disease. Stimulation by co-infectious agents could therefore also be hypothesized to be protective against FIP. Along these lines, co-infection must also be considered a possible alternative explanation for the raised TLR2 and 4 gene expression levels, as these TLRs are more typically associated with bacterial infections. Enteric coronavirus infection or the generalized inflammatory state induced by FIP may have increased the permeability of the intestinal barrier to microorganisms. The resulting TLR stimulation would therefore not be virus induced. A third alternative is the upregulation of TLR2 and 4 by endogenous ligand stimulation, reported as a response to alarmin release from damaged cells (van Beijnum et al., 2008) . This alternative also fits with upregulation of TLR2 gene expression in lesional MLNs compared with nonlesional MLNs in FIP.

From their known ligands, TLRs 3, 7 and 8 would be predicted to be triggered in infection by FCoV as it is a ssRNA virus (triggering TLRs 7 and 8), possessing a double-stranded RNA intermediate replicating phase (triggering TLR3) (Arpaia and Barton, 2011) . That no upregulation occurs for TLR3 and TLR7 with FIP suggests either the lack of an appropriate trigger (TLR7 and 8 are known to show differing, if overlapping, specificity, and dsRNA intermediate replicates are a minority of the viral RNA present; Jensen and Thomsen, 2012) , or that the virus is able to inhibit TLR transcription. SARS-CoV is known to inhibit both TLR3 and 7 signalling via papain-like protease activity (PLpro) (Li et al., 2016) . This mechanism may also contribute in FCoV infection, but would be expected to affect the signalling pathways rather than the TLR mRNA levels directly. In cats with FIP, we observed slightly lower TLR3 gene transcription in MLNs with typical FIP lesions, as compared with MLNs without lesions, down to the levels seen in MLNs from cats without FIP. This could indicate a general systemic stimulus to upregulate TLR3 in FIP, which is counteracted locally by viral inhibition of TLR3. Prior stimulation of TLR3 has also been shown in vitro to contribute to defence against murine coronavirus via type I interferon induction (Mazaleuskaya et al., 2012) , so is another potential avenue for future FIP research. A larger sample population, in particular with larger numbers of systemically infected cats without FIP, might have revealed significant intergroup differences for TLR3.

The STAT transcription factors are a key part of the antiviral pathways, mediating many downstream IFN effects (Aaronson and Horvath, 2002) . They have also been linked to other coronavirus infections (e.g. STAT1 knock-out mice show a markedly increased susceptibility to SARS-CoV, while avian infectious bronchitis coronavirus uses STAT1 inhibition of IFN responses) (Frieman et al., 2010; Kint et al., 2015) . STAT1 and 2 gene transcription levels correlated with type II and I interferon transcription levels, respectively, in our study, while in virus-positive MLNs of cats without FIP, STAT1 and 2 levels (the latter significantly so), as well as IFN-g levels, lay between the other two groups. This shows that the levels of IFNs and their downstream transcription factors are closely linked. Interestingly, STAT2 gene expression levels were significantly higher in the MLNs of cats with FIP and with effusions, a finding that cannot be readily explained. STAT2 has been linked to IL-6 upregulation, which itself has been linked to increased vascular permeability (Maruo et al., 1992; Nan et al., 2018) ; however, the IL-6 gene was not upregulated in our cohort, suggesting that responsibility lies with another pathway.

The results of our S gene codon mutation analysis add weight to recent findings that the M1058L mutation (referred to as M1048L in the present study due to re-evaluation of the reference sequence) is likely to contribute to systemic spread, but does not itself confer pathogenicity (Chang et al., 2012; Porter et al., 2014) . This indicates that further host and/or viral factors are required for the development of FIP or, more precisely, the activation of virus-infected monocytes as a prerequisite to set off FIP vasculitis (Kipar and Meli, 2014) . Most likely owing to the low viral RNA levels within the MLNs of cats without FIP, obtaining an adequate sequence from this group proved problematic. Other researchers experienced similar problems, often finding that FCoV RT-PCR-positive samples from cats without FIP were not amenable to sequencing (Felten et al., 2017b) . The lack of FCoV antigen expression in these cats was not unexpected and reflects the rarity of infected cells and/or the low virus load in infected cells; this is in line with the results of a previous study that found only rare positive macrophages in the MLNs of experimentally persistentlyinfected cats (Kipar et al., 2010) .

It was not possible to compare statistically the induced immune response of viruses showing S protein amino acid variations (codons 1,048 and 1,050) as only one cat had the 'enteric' form.

The future outcome of our FCoV infected cats without FIP, had they not succumbed to other diseases, is unknown, as is the contribution of yet to be defined viral factors. These cats may have remained carriers or have been demonstrating a transitional phase to later development of disease. However, based on our observations, activation of genes encoding TLRs 2, 4 and 8 in MLNs is associated with a negative outcome (i.e. FIP), while carrier animals upregulated the gene encoding TLR9. IFN-g, and particularly STAT2 with its myriad opportunities to direct cell fate, displayed intermediate levels of upregulation in the MLNs of the carrier/transitional group, not associated with a widespread increase in mediators of inflammation.

This study is only the start of determining the extent of involvement of PRRs in FIP; the downstream effects of these transcriptional alterations must be further investigated. However, our results reinforce the need for a balanced immune response against the virus, with the hypothesis that the moderate response in cats without FIP is part of the key to controlling the virus; when this balance is lost the animal may be at risk of succumbing to disease. measurement of cytokine mRNA expression in the skin of normal cats and cats with allergic skin disease. Veterinary Immunology and Immunopathology, 122, 216e230. Takano 

A road map for those who don't know JAKeSTAT

Cellular and Molecular Immunology, 9 th Edit

Toll-like receptors: key players in antiviral immunity

Limitations of using feline coronavirus spike protein gene mutations to diagnose feline infectious peritonitis. Veterinary Research, 48, 1e14. Beutler BA (2009) TLRs and innate immunity

IL-6, IL-17 and STAT3: a holy trinity in auto-immunity?

Spike protein fusion peptide and feline coronavirus virulence

In vivo cytokine response to experimental feline infectious peritonitis virus infection

SARS coronavirus spike protein-induced innate immune response occurs via activation of the NF-kappaB pathway in human monocyte macrophages in vitro

Sensitivity and specificity of a real-time reverse transcriptase polymerase chain reaction detecting feline coronavirus mutations in effusion and serum/plasma of cats to diagnose feline infectious peritonitis

SARS-CoV pathogenesis is regulated by a STAT1 dependent but a type I, II and III interferon receptor independent mechanism

Whole blood cytokine profiles in cats infected by feline coronavirus and healthy non-FCoV infected specific pathogen-free cats

Transcriptional profiling of feline infectious peritonitis virus infection in CRFK cells and in PBMCs from FIP diagnosed cats

Feline coronavirus type II strains 79-1683 and 79-1146 originate from a double recombination between feline coronavirus type I and canine coronavirus

Identification and genotyping of feline infectious peritonitis-associated single nucleotide polymorphisms in the feline interferon-g gene

Functional diversification of the toll-like receptor gene family

Tolllike receptor expression in feline lymphoid tissues

Toll-like receptor 4 deficiency increases disease and mortality after mouse hepatitis virus type 1 infection of susceptible C3H mice

Infectious bronchitis coronavirus inhibits STAT1 signaling and requires accessory proteins for resistance to type I interferon activity

Natural FCoV infection: cats with FIP exhibit significantly higher viral loads than healthy infected cats

Cellular composition, coronavirus antigen expression and production of specific antibodies in lesions in feline infectious peritonitis

Feline infectious peritonitis presenting as a tumour in the abdominal cavity. Veterinary Record, 144, 118e122

Feline infectious peritonitis: still an enigma? Veterinary Pathology

Sites of feline coronavirus persistence in healthy cats

Natural feline coronavirus infection: differences in cytokine patterns in association with the outcome of infection

Immunotherapy for dogs: running behind humans

Quantitative realtime PCR for the measurement of feline cytokine mRNA

SARS coronavirus papain-like protease inhibits the TLR7 signaling pathway through removing Lys63-linked polyubiquitination of TRAF3 and TRAF6

IL-6 increases endothelial permeability in vitro

High viral loads despite absence of clinical and pathological findings in cats experimentally infected with feline coronavirus (FCoV) type I and in naturally FCoV-infected cats

IRF9 and unphosphorylated STAT2 cooperate with NF-kB to drive IL6 expression

Virologic and immunologic aspects of feline infectious peritonitis virus infection

An update on feline infectious peritonitis: virology and immunopathogenesis

Immunologic phenomena in the effusive form of feline infectious peritonitis

An enteric coronavirus infection of cats and its relationship to feline infectious peritonitis

A new mathematical model for relative quantification in real-time RT-PCR

Two related strains of feline infectious peritonitis virus isolated from immunocompromised cats infected with a feline enteric coronavirus

Amino acid changes in the spike protein of feline coronavirus correlate with systemic spread of virus from the intestine and not with feline infectious peritonitis

The innate antiviral immune system of the cat: Molecular tools for the measurement of its state of activation

Stimulation with a class A CpG oligonucleotide enhances resistance to infection with feline viruses from five different families

Expression profiles of immune mediators in feline coronavirus-infected cells and clinical samples of feline coronavirus-positive cats

Quantitative real-time RT-PCR Immune Response to Feline Coronavirus

This study was supported by the 2017 Journal of Comparative Pathology Educational Trust and Petplan Charitable Trust Joint Research Award in Veterinary Pathology. Earlier sequencing work included in this study was supported by a project grant from The Petplan Charitable Trust (10e27) and by a project grant from the Morris Animal Foundation (#D16FE-507). The authors thank all the practitioners, owners and colleagues who helped in the acquisition of samples used in this study. We thank the technical staff of the Histology Laboratory, Institute of Veterinary Pathology, Vetsuisse Faculty, University of Zurich, for technical support.

The authors declare no conflicts of interest with respect to the publication of this manuscript.

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jcpa.2018.11.001.