RNA-Seq analysis of chikungunya virus infection and identification of granzyme A as a major promoter of arthritic inflammation

Abstract

Chikungunya virus (CHIKV) is an arthritogenic alphavirus causing epidemics of acute and chronic arthritic disease. Herein we describe a comprehensive RNA-Seq analysis of feet and lymph nodes at peak viraemia (day 2 post infection), acute arthritis (day 7) and chronic disease (day 30) in the CHIKV adult wild-type mouse model. Genes previously shown to be up-regulated in CHIKV patients were also up-regulated in the mouse model. CHIKV sequence information was also obtained with up to ≈8% of the reads mapping to the viral genome; however, no adaptive viral genome changes were apparent. Although day 2, 7 and 30 represent distinct stages of infection and disease, there was a pronounced overlap in up-regulated host genes and pathways. Type I interferon response genes (IRGs) represented up to ≈50% of up-regulated genes, even after loss of type I interferon induction on days 7 and 30. Bioinformatic analyses suggested a number of interferon response factors were primarily responsible for maintaining type I IRG induction. A group of genes prominent in the RNA-Seq analysis and hitherto unexplored in viral arthropathies were granzymes A, B and K. Granzyme A-/- and to a lesser extent granzyme K-/-, but not granzyme B-/-, mice showed a pronounced reduction in foot swelling and arthritis, with analysis of granzyme A-/- mice showing no reductions in viral loads but reduced NK and T cell infiltrates post CHIKV infection. Treatment with Serpinb6b, a granzyme A inhibitor, also reduced arthritic inflammation in wild-type mice. In non-human primates circulating granzyme A levels were elevated after CHIKV infection, with the increase correlating with viral load. Elevated granzyme A levels were also seen in a small cohort of human CHIKV patients. Taken together these results suggest granzyme A is an important driver of arthritic inflammation and a potential target for therapy.

Trial registration: ClinicalTrials.gov NCT00281294.

Conflict of interest statement

The authors have declared that no competing interests exist.

Figures

Fig 1. Global analysis of gene expression.

(A) Venn diagram [70] of up-regulated genes in feet. Only genes with q2 and FPKM >1 were included. Numbers in parenthesis represent the mean fold change of all the genes in the segment; note genes which are shared between all time points show the highest mean fold changes. (B) Heat map of−log10 z scores from Ingenuity upstream regulator analysis of the genes shown in A. Scores are ranked highest to lowest (mean of the 3 time points). The top and bottom 10 upstream regulators are shown on the figure; the full data set is shown in S3 Table. (C) Venn diagram of down-regulated genes in feet. Numbers in parenthesis represent the mean fold change. (D) Heat map of p values from Ingenuity canonical pathway analysis of the genes shown in C, with day 7 pathways ranked most to least significant. A DAVID analysis of the same genes (bar chart) showing similarity scores. The full data sets for these figures are provided in S3 Table. (E) Venn diagram of up-regulated genes in lymph nodes on days 2 and 7. (F) Venn diagram of up-regulated genes on day 2 in lymph nodes (LN) and feet (Ft); and heat map of−log10 z scores from Ingenuity upstream regulator analysis of these genes. The first column of the heat map shows z scores from day 2 lymph nodes ranked highest to lowest (the full data set is provided in S3 Table). (G) Venn diagram of up-regulated genes in lymph nodes on days 7 and feet on day 7. (H) Venn diagram of down-regulated genes in lymph nodes on days 2 and 7.

Fig 2. CHIKV genome.

(A) The percentage of all the reads that aligned to the mouse genome for each tissue and time point. Read alignment numbers are provided in S1D Fig. (B) Examples of alignments of RNA-Seq reads from 3 foot samples mapped to the CHIKV genome (map quality threshold 20) viewed using Integrated Genomics Viewer (IGV version 2.3.34). The CHIKV genome is shown at the top for reference (arrow represents position of the sub-genomic promoter). Upper graphs (dark grey) show sequence coverage (log scale) for each nucleotide position in the CHIKV genome with y axis scale ranges (e.g. 0–300,000) shown in the left hand corners. The bottom graphs are “squished” views of the 100 bp reads aligned to the CHIKV genome (each grey horizontal bar represents one read). Black spots represent deletions/insertions within each read.

Fig 3. The interferon signature.

(A) Expression of IFNα6 and IFNβ as determined by qRT-PCR in different tissues on day 2 post CHIKV infection (n = 3 mice), the time of the peak IFNα/β response [28]. Values are normalized to RPL13a mRNA levels and expressed as fold induction relative to mock infected controls (n = 3). CHIKV titres in the tissues were determined as described [28]. Spearman’s correlation tests showed a significant relationship between IFN mRNA levels and viral titres (IFNα6 Spearman’s rho = 0.829 p = 0.042, IFNβ Spearman’s rho = 0.943 p = 0.005). (B) For all samples the up- and down-regulated DEGs (for which FC>2, q1, see S1 Table) were analyzed using Interferome and the percentage of these DEGs that are interferon regulated genes (IRGs) is shown. (Interferome does not distinguish between genes directly or indirectly stimulated by IFNs, and some type I and/or II IRGs may not be identified by Interferome). (C) Heat map of FPKM values for all IFN genes identified by the RNA-Seq analysis (the same data is plotted as bar chart in S4 Fig). (D) Transcription factors associated with IFN responses in feet. Fold change of indicated transcription factors with vertical numbers representing the mean FKPM values (for the 3 biological replicates). Horizontal bold numbers represent the activation Z scores for the indicated transcription factor as determined by the direct function of the upstream regulator analysis of IPA; corresponding p values are provided in S5 Fig. (E) CiiiDER analysis of putative transcription factor site enrichment in the up-regulated type II IRGs in feet (as identified by Interferome). Color and size of circles reflect p values of the enrichment. Calculations for x and y values and the input/output data for the labeled transcription factors are provided in S7 Fig.

Fig 4. Granzyme gene expression and CHIKV infection in granzyme deficient mice.

(A) Volcano plots of host gene expression from the RNA-Seq analysis of feet and lymph nodes (LN) at the indicated day post infection. Only genes where FPKM ≥1 in at least one sample in the pair wise comparisons were included in the plots. Positions of the granzyme (Gzm) genes are indicated by arrows, with values in parenthesis representing FPKM values. (B) Foot swelling in granzyme deficient mice. Mice were infected as above and foot swelling monitored. GzmA-/- mice; * p<0.03, ** p<0.001 (Kolmogorov-Smirnov tests, n = 8–10 mice per group). GzmK-/- mice; *p<0.031 (Mann Whitney U tests, mean of two independent experiments is shown, n = 12–14 per group). (C) Viraemia in granzyme deficient mice. Granzyme deficient mice were infected with CHIKV and the viraema measure on the indicated days. No significant differences in viraema were apparent: GzmA-/- vs C57BL/6 controls (n = 6 mice per group); GzmB-/- vs C57BL/6 controls (n = 8–10 mice per group); GzmK-/- vs C57BL/6 controls (n = 12–14 mice per group). (D) Tissue CHIKV titers in GzmA-/- and C57BL/6 mice. Feet; n = 6–12 GzmA-/- and n = 12–20 C57BL/6 feet per time point; data obtained from 2 independent experiments. Muscle; n = 3–6 mice per time point.

Fig 5. Histology and immunohistochemistry of GzmA-/- mice feet.

(A) H&E staining of feet on day 6 post infection in C57BL/6 and GzmA-/- mice. Cellular infiltrates are characterized by high densities of blue staining nuclei; areas with pronounced cellular infiltrates are indicated by white oval outlines. (B) Aperio Positive Pixel Count determination of the ratio of blue (nuclear) to red (cytoplasmic) staining areas in whole foot sections day 6 post infection. Leukocytes tend to have a higher nuclear/cytoplasmic area ratio, so elevated ratios reflect increased leukocyte infiltrates [13]; (n = 6 feet from 6 mice per group, 3 sections per foot; statistics by 2 way ANOVA including a term for section). (C) Immunohistochemical staining for NK cells (anti-CD335/NKp46) clearly visible (brown staining) in muscle tissue of feet from CHIKV-infected C57BL/6 mice 6 days post infection (left). NK cell staining was less pronounced in GzmA-/- mice (right). Blue counter staining with haematoxylin. (D) Aperio Positive Pixel Count determination of NK cell staining; strong brown pixels per μm2 in whole feet sections (3 sections per foot; n = 11–12 feet from 11–12 mice per group from 2 independent experiments. Statistics by Mann Whitney U test). (E) As for C for T cell (anti-CD3) staining. (F) As for D for T cell staining. Statistics by Kolmogrov Smirnov test. (G) As for C for monocyte/macrophage (F4/80) staining, which was prominent in subcutaneous tissues (* indicates epidermis). (H) As for D for monocyte/macrophage staining.

Fig 6. Granzyme A inhibitor and granzyme levels in CHIKV infected primates.

(A) On days 2 to 6 post infection with CHIKV, C57BL/6 mice were treated i.v. with 10 μg of the granzyme inhibitor, Serpinb6b, or Trypsinized (inactivated) GzmA inhibitor (green arrows) or were left untreated. The data represents results from 2 independent experiments: GzmA inhibitor (n = 14 feet, 7 mice); Trypsinized GzmA inhibitor (n = 6 feet, 3 mice); untreated mice (n = 30 feet, 15 mice). Statistics by t test; p values provided for comparisons between granzyme A inhibitor and Trypsinized granzyme A inhibitor and are only provided where comparisons between granzyme A inhibitor and untreated mice were also significant. (B) Viraemia for the same mice as shown in A. (C) Aperio Positive Pixel Count determination of the ratio of blue (nuclear) to red (cytoplasmic) pixels in H&E stained whole foot sections day 6 post infection (as in Fig 5B). (n = 6 feet from 3 mice per group, 3 sections per foot; statistics by Kolmogorov Smirnov test). (D) Granzyme A levels in plasma samples from CHIKV-infected non-human primates (NHPs) measured using an ELISA kit. Data for 9 NHPs is shown; all NHPs had samples collected day -1, one day prior to CHIKV infection. (E) Granzyme A levels on day -1 and peak granzyme A levels plotted for the 9 NHPs. Statistics by paired t test. (F) Correlation between peak viral load (S9D Fig) and the increase in granzyme A levels from day -1 to peak (i.e. peak levels minus day -1 levels). Statistics by Spearman correlation. (G) Serum granzyme A levels in healthy controls and IgM positive symptomatic CHIKV patients. Granzyme A levels were determined using cytokine bead array and FACs. The limit of detection is deemed to be 3.7 pg/ml. Statistics by Kruskal-Wallis test.