Systems Analysis of Subjects Acutely Infected with Chikungunya Virus

The largest ever recorded epidemic of the chikungunya virus (CHIKV) began in 2004 and affected four continents. Acute symptomatic infections are typically associated with the onset of fever and often debilitating polyarthralgia/polyarthritis. In this study, a systems biology approach was used to analyze the blood transcriptomes of adults acutely infected with CHIKV. Gene signatures that were associated with viral RNA amounts and to the onset of symptoms were identified. Among those genes, the putative role of Eukaryotic Initiation Factor (eIF) family genes and apolipoprotein B mRNA editing catalytic polypeptide-like (APOBEC3A) in the CHIKV replication process were displayed. We further compared these signatures with those induced by dengue virus infection and rheumatoid arthritis. Finally, we demonstrated that CHIKV infection in mice induced IL-1 beta production in a mechanism highly dependent on the inflammasome NLRP3 activation. The findings provided valuable insights into the virus–host interactions during the acute phase and could be useful in the investigation of new and effective therapeutic interventions.


Introduction
Chikungunya virus (CHIKV) is a mosquito-borne reemerging arbovirus responsible for intermittent and devastating outbreaks (1). The largest epidemic of CHIKV ever recorded started in Africa in 2004 and has spread globally, reaching the Americas in 2014. The disease has reached four continents, affected more than 100 countries and infected over 10 million people (2,3). CHIKV has spread rapidly through several Brazilian states and resulted in a total of 20,598 infected individuals in 2015 (4) and more than 200,000 suspected cases reported in 2017-2018 (5).
CHIKV acute infection typically results in a 5-7 days viremia and the symptoms are characterized by fever, rash and severe polyarthralgia/polyarthritis that can become chronic and persist from months to years (6). Mortality rates are estimated to be approximately 0.1% (7); however, the high attack rates (up to 30-75%) can result in a considerable economic burden (8,9). The current mainstay of treatment involves the use of NSIADs and/or acetaminophen, although relief is often inadequate and more effective treatments are actively being sought (10). There has been a considerable quest to better understand CHIKV disease pathogenesis and virus-host interactions (11) using in vitro approaches (12) or animal models (13,14).
Systems biology approaches have been successfully applied to identify molecular signatures associated with infections (15) and vaccination (16). In this study, a systems approach was performed to investigate subjects naturally infected with CHIKV during the 2016 epidemics in the state of Sergipe, Brazil. The blood transcriptome analyses performed in this study revealed key genes and pathways involved in acute CHIKV infection, providing important insights into how CHIKV interacts with the host's immune system. These analyses helped uncover potential drug targets for improving CHIKV therapy.

Clinical information and diagnosis of subjects infected with CHIKV
In an endemic area for CHIKV, Zika and dengue virus (17) in the northeast of Brazil, whole blood samples were collected from 39 adults with symptoms consistent with arboviral infection including fever, arthralgia, headache and/or muscle pain. Most individuals (n = 22) reported that the onset of the symptoms occurred on the same day or one day prior to the blood collection, indicating a very early stage of the disease ( Fig.   1a and Table S1). Another eight had symptoms for two days and seven of them showed symptoms between three and four days. Two of the subjects reported pain for 20 days or more (Fig. 1a). Blood was also collected from 20 healthy subjects with no symptoms of viral infection and who tested negative for CHIKV RNA from endemic and nonendemic regions in Brazil. Serodiagnostic ELISAs were also performed to detect the presence of IgG antibodies specific to Zika and/or dengue viruses (Fig. 1b). We detected IgG and IgM antibodies specific to CHIKV in only one subject. The presence of CHIKV RNA by real-time RT-PCR or IgM for CHIKV positivity data, and the absence of Zika and dengue RNA were confirmed in the serum of all the patients (Table   S1). The most frequent symptoms of CHIKV-infected patients were fever, arthralgia, headache and myalgia (Fig. 1c). As expected, individuals who reported being in the very early stage of the disease (Fig. 1d) had the highest levels of serum CHIKV RNA.

The impact of CHIKV acute infection on peripheral blood transcriptome
To understand the molecular changes that occurred during acute CHIKV infection, we compared the blood transcriptomes of infected individuals with those from healthy controls. Although the CHIKV RNA amounts had a small contribution to the variance, individuals were not grouped by the period of the onset of symptoms nor their infection status (Fig. S1b). Correlation analysis between gene expression and the amounts of CHIKV RNA in the peripheral blood of infected individuals revealed approximately 3,500 genes associated with CHIKV infection. The expression of most of these genes (> 85%) was negatively correlated with CHIKV RNA (Fig. 2a). Using the correlation values as rank and Blood Transcription Modules (BTM) as gene sets, we run a gene set enrichment analysis (GSEA) to reveal the BTMs related to the present amount of CHIKV virus.
BTMs associated with innate immune cells (dendritic cells and neutrophils), antiviral response, inflammation and toll-like receptor signaling ( Fig. 2b and S4) were positively associated with the amount of CHIKV RNA. BTMs negatively associated with the amount of CHIKV RNA, i.e. with a negative normalized enrichment score (NES) were related to adaptive immune response, such as cell proliferation, activation and differentiation of B and T lymphocytes, in addition to chemokines and cytokines that could be due to the CHIKV infection related lymphopenia (Fig. 2b) (18). A gene network was subsequently constructed by connecting the genes related to BTMs with negative NES score (Fig. 2b) utilizing protein-protein interaction (PPI) data from InnateDB. Genes related to adaptive immune response and cells recruitment, such as CCL5, CD8A and CD8B, CD3D, CD19, CCR7 and TCF7 among others appear in the respective network (Fig. 2c). Several of these genes are related to the homing/recruiting of the effector/memory T lymphocytes as well as to activation and induction of memory cells. The same process was performed for BTMs with positive NES score (Fig. S2).
Genes such as IRF7, OAS3, IFIT2 and IFIT3, CXCL11, SERPING1 are all related to the innate immune response to viral infections (19) or can even be considered as markers of viral infections and some are ISGs (interferon-stimulated genes).  The expression of several eukaryotic translation initiation factors (eIFs) was negatively correlated with the amount of CHIKV RNA (Fig. 3), indicating that these genes could play a major role in viral replication. The eIFs are important proteins that control translation initiation process, which is a major step for viral protein synthesis (20). A network with eIF genes was constructed (Fig. 3b) among which EIF4B, EIF3L, EIF3E and EIF2AK2 presented the highest negative correlation with the amounts of CHIKV RNA (Fig. 3a). Only EIF2AK2 gene also known as PKR had a positive correlation to CHIKV RNA in the blood of the infected individuals. It is well known that this protein could be activated by binding virus-derived double stranded RNA (dsRNA). This IFN-induced dsRNA-dependent serine/threonine-protein kinase inhibits viral replication by phosphorylating EIF2S1 (21) and shuts down cellular and viral protein synthesis.

Consistent transcriptome changes upon CHIKV infection
Due to the natural heterogeneity in human cohorts, differential expression analysis was performed between each infected patient and the group of healthy controls.
The patients whose samples were collected in the initial days post onset of symptoms presented higher numbers of DEGs (differentially expressed genes) detected by RNAseq analysis and a higher mean of CHIKV RNA in their blood samples (Fig.4a). The DEG list of each patient was subsequently combined into a meta-volcano plot (Fig. 4b).
This plot displays the number of genes whose expression was consistently altered in most patients. Approximately 70% of the CHIKV-infected patients presented downregulation of 123 genes and up-regulation of 382 genes (Fig. 4b). APOBEC3A, IFI44 (14), OAS3 are the most represented up-regulated genes and NT5E (also known as CD73) PTGS2, SNORD3C, EEF1A1P13 are the most represented down-regulated ones among the patients. APOBEC3 family members are important cytidine deaminases that control HIV replication (22) and have not been described till now in the CHIKV context. Since APOBEC3 genes are also considered ISGs, CHIKV infection are expected to induce this gene. Interestingly, the analysis also identified the NT5E/CD73 that is an ecto-5′-nucleotidase that has been described as an important molecule to recover the endothelial barrier during the dengue 2 infection leakage (23). IκBs related genes, which are important to NFκB activation and proinflammatory cytokines production when phosphorylated. It is integral to mention here that we also detected the up-regulation of REL, which is a subunit of NFκB.
We also detected DEGs related to NLRP3 inflammasome. Nlrp3 is transcriptionally regulated to guarantee high protein levels for activation of the NLRP3 inflammasome. Its activation could lead to autophagy (SQSTM1), apoptosis of infected cells and to other cytokines production (CARD17) in response to the infection (Fig. 4c).
It has been shown recently that the activation of NLRP3 leads to IL-1β and IL-18 production and was linked to the alphaviral disease severity in an animal model (24).
The random forest method, a machine learning approach, was also employed to rank the importance of the 505 DEGs mentioned in the meta-volcano analysis ( Fig.   4b) in predicting the CHIKV infection status. Fig. S3 displays the top 31 that better distinguish between infected patients and healthy control samples. The gene SERPING1 presented an increased expression in HIV patients' samples and could also be considered a CHIKV marker (25). Significantly, SERPING1 is a C1 complement protein inhibitor (26) and could be involved in signaling at the sites of inflammation contributing to tissue damage and disease severity as was suggested in the case of Ross River Virus infection. (27). Figure S3. Random forest predictive analysis. The genes were sorted by the decreasing order of the Predictor Importance of Status used by random forest in order to prioritize the genes that most distinguish between infected and healthy samples.
We next performed single sample GSEA analysis using the log2 fold-change values of each patient compared to the group of healthy controls as ranks and the BTMs as gene sets (Fig. S4). Similar to the results presented in the Fig. 2, we observed that the up-regulated BTMs were related to innate immunity and anti-viral responses involving dendritic cells and monocytes activation.

Potential Signatures of CHIKV-induced Arthralgia Chronicity
We split the patients (n = 13) who agreed to return for clinical follow-up examinations into those with chronic and non-chronic arthralgia (see methods). These two groups possess no difference in the amounts of CHIKV RNA in the serum (Fig.   S5a). Compared to the group of healthy controls, we found that a total of 1,262 and 1,862 genes were consistently differentially expressed in most of the chronic and nonchronic patients respectively (Fig. S5b). Of those, 514 were commonly up-regulated in both groups whereas 337 genes were down-regulated in both groups (Fig. S5c).
Additionally, there is a high positive correlation between the mean log2 fold change of chronic and non-chronic patients when compared to healthy controls (Fig. S5d).
However, few genes presented an expression with inverted behavior between chronic and non-chronic patients (Fig. S5d). HLA-DRB5 that could be related to antigen peptides presentation is one example of genes that was up-regulated in most nonchronic samples and down-regulated in most chronic samples (Fig. S5d). Genes such as DDX3Y, EIF1AY, and LINC00278 were up-regulated in most chronic samples and down-regulated in most non-chronic samples. The EIF1AY gene belongs to the family of eukaryotic translation initiation factors (and showed in Fig. 3). Interestingly, the DDX3Y is an RNA helicase that was described as an important effector of the herpes virus replication and propagation (28)

Modular expression analysis of CHIKV infection
We performed a gene co-expression network analysis using the expression profiles of all the patients and healthy controls. CEMiTool (29) identified eight coexpression modules containing 74 to over 2,000 genes (Fig. 5a). The expression activity of some of these modules were altered in healthy controls or in patients with different days after the symptoms onset (Fig. 5b). The module M8, which shows higher expression in patients with two to four days of symptoms was enriched for monocytes and neutrophils (Fig. 5c). These immune cell subsets are known for being related to the disease (30). Moreover, CEMiTool can integrate the gene modules with protein-protein interaction data and identify important hubs within each module. In module M8, we found as one of the hubs the gene DDX58 (also known as RIG-1) that has already been linked to CHIKV disease (30) and as a virus replication inhibitor in an animal model for CHIKV infection (31). represented as nodes and its proteinprotein and co-expression interactions as edges. The size of the node represents its degree of connectivity.

Comparing CHIKV signature with Rheumatoid Arthritis and Dengue infection signatures
To check the extent to which the gene signature was specific to CHIKV infection, the CHIKV signature was compared with those from another viral infection (dengue virus) or from an auto-immune related inflammatory disease (rheumatoid arthritis or RA). We re-analyzed two publicly available blood transcriptome datasets and identified the genes whose expression was altered in the dengue infected patients and RA patients compared to healthy controls. A total of 949, 632 and 302 genes were identified as being up-regulated only in RA, dengue infection and CHIKV infection respectively (Fig. 6a). Among those, seven up-regulated genes were shared by all the three signatures, including the following: OAS1, C1QB, ANKRD22, IRF7, CXCL10, IFI6 and IFIT3 (Fig. 6a). More than 300 genes were exclusively up-regulated upon CHIKV infection, including the inflammasome-related NLRP3 genes related to NFκB and genes that belong to the Th2 response (ILI31RA, IL4I1) (32) (Fig.6a). Although no gene was down-regulated in all the three conditions, genes such as IL7R, ESYT1, PCYOX1 and LRRN3 were commonly down-regulated in CHIKV infection and RA samples (Fig. 6b). The cytokine IL-7 is crucial for the survival of naïve and memory T cells, which are important effectors against several pathogens including viral infections. The receptor IL-7R is expressed on the surface of these cells and was shown in an animal model to prevent the chronicity induced by lymphocytic choriomeningitis virus due to the enhancement of CD8 T cells' response and prevention of its exhaustion (33). We identified a specific CHIKV signature composed of 107 down-regulated genes that includes CX3CR1, BTLA, BCL2, GZMK and three genes that encode Fc receptor-like glycoproteins (Fig. 6b) that could be related to the lymphopenia induced by CHIKV infection (34). The down-regulation of CX3C chemokine receptor 1 (CX3CR1) seems to agree with the results of this study that showed that several T cell effector/differentiation genes were down-regulated in CHIKV infection. CX3CR1 has been considered as an important marker of CD8 memory cells (35). Moreover, BCL-2 down-regulation could be a mechanism of CHIKV host response to induce apoptosis of infected cells.
As the inflammasome-related genes are exclusively up-regulated in CHIKV infection and was described before in this context (24), we decided to provide a deeper proof-of-concept related to these findings. The murine bone marrow-derived macrophages were infected with CHIKV virus and the readouts of the inflammasome activation were measured. We used a specific dye that binds to active caspase-1 (FAM-YVAD) to assess caspase-1 activation and it was observed that CHIKV induce caspase-1 activation as measured by the percentage of FAM-YVAD+ cells and the integrated mean of fluorescence (iMFI) (Fig. 7a). Caspase-1 activation was further assessed by western blot and detected caspase-1 cleavage as shown by the presence of Casp1 p20 in cells infected with MOI of 5 (Fig. 6c). We found that CHIKV infection at MOI of 1 and 5 induces IL-1β production and LDH release (Fig. 6d). IL-1β production was dependent on the inflammasome because Casp1/11 deficient macrophages fail to trigger IL-1β production in response to the infection (Fig. 6e). These data in conjunction indicate that the inflammasome is activated in response to the infection, and thereby support the robustness of our transcriptional analyses.

Discussion
CHIKV re-emergence in the past 15 years has been producing major epidemic outbreaks in Asia, Africa, the Indian Ocean and more recently in the Americas after decades of intermittent outbreaks (36). Despite considerable progress in understanding the infection, much of the host-pathogen interplay remains obscure.
Systems biological approaches can provide comprehensive and unbiased dissections of the complex interactions between genes and proteins during an infection (15). Relevant studies related to the natural infection by CHIKV were recently published. However, these were limited to the analysis of a limited number of proteins (37) or on children infected with the virus (38). Since it is extremely difficult to acquire sufficient data from naturally infected individuals for use in systems biological models of analysis, this kind of study is relatively new. To our knowledge, we are the first to investigate the early host response to acute CHIKV infection in adults using a systems biological approach.
The most clinically relevant symptom related to CHIKV infection is peripheral symmetrical joint pain (primarily inflammatory polyarthralgia). Manifested in several CHIKV cases, it can produce a high economic impact and severely affect the patients' quality of life (11). The pain starts in the acute phase and can persist for years in up to 50% of the infected adult individuals (39). Notably, it was proposed that CHIKV infection produced autoimmune sequelae (40). Although polyarthralgia is rarely Consistent with other reports, it was observed the up-regulation of several genes that play roles in the antiviral immune response, and many of them are ISGs (14,19,42). Similarly, other reports that assessed the changes in immune cell subsets during CHIKV infection (38,43) support the findings of this study regarding the most abundant up-regulated genes being related to neutrophils and myeloid populations and especially dendritic cells and monocytes. The monocytes have been reported recently as important inflammatory and regulatory mediators of the innate immune response to different arboviruses (15,44). In corroboration with these findings, we also showed the exclusive up-regulation of inflammasome-related genes. Additionally, the induction of inflammasome activation in macrophages infected with CHIKV in vitro was observed.
Importantly, these data account to explain our previous demonstration that NLRP3 inhibitors can interfere with the pathogenesis of CHIKV virus (24).
Moreover, the negative and strong correlation between CHIKV RNA amounts and most eIF family member genes observed here have not been described previously in CHIKV infections. Furthermore, it is speculated whether the down-regulation of other eIF family members observed in these results in response to the infection could be an important host defense mechanism against the virus replication. We consider that the eukaryotic translation initiation factors are attractive markers of the CHIKV arthralgia chronicity and could be better exploited as novel broad-spectrum antiviral targets.
The CHIKV related arthralgia symptom is very similar to the rheumatoid arthritis (RA) but there are some important clinical and immunological characteristics that differ between these diseases and the different molecular signatures between both CHIKV and RA can be shown here. Previous data showed significant concordance between rheumatoid arthritis gene signatures and a mouse model of CHIKV infection/arthritis (13). Here, we are presenting a more detailed comparison between both diseases in humans.
On the other hand, when compared to the differential gene expression of DENV infected patients, the data in this study shows specific molecular gene signatures in response to CHIKV. It is interesting to note in this context that although previous infections with ZIKV and DENV were detected in the group of non-CHIKV-infected volunteers, specific responses for CHIKV infection were still possible to identify.
Moreover, the comprehensive molecular data in this study showed novel molecules that could play important roles in the acute infection of CHIKV infection thereby providing new candidates as targets for therapies against this incapacitating disease.

Sample collection and clinical information
Blood

Molecular Diagnostics
Real-time RT-PCR was performed to test for CHIKV, ZIKV and DENV as previously described (17,45

Serological Diagnostics
Serum samples were evaluated with a commercial semi-quantitative ELISA kit

RNA-Seq data analyses
Raw paired-end reads were preprocessed for quality control. The Trimmomatic software, version 0.36 (46) has been used to remove adapters, to trim the 5′ and 3′ ends with mean quality score below 25 (Phred+33), and to discard reads shorter than 35 bp after trimming. Paired-end reads mapping to PhiX Illumina spike-in were removed Regarding the pathways that may be related to the progression of the disease, a Gene Set Enrichment Analysis (GSEA) was performed using as ranks the correlation between the genes and inverted Ct. A set of Blood Transcriptional Modules (BTM), previously identified by our group (51) through large-scale network integration of publicly available human blood transcriptome, was used as gene sets.

Gene Co-expression and Network Analysis
We performed the gene co-expression analysis using the R package CEMiTool (52). For this analysis, we normalized the expression data using TMM (Trimmed Mean of M-values) and transformed it to log2 scale. We followed the default parameters with a variance filter of 0.2.
To gain a systems-level understanding of the patterns of a certain disease, one of the steps required is the construction and analysis of the network involving the most interesting genes. For this, we used NetworkAnalyst (https://www.networkanalyst.ca/) with the protein-protein interaction (PPI) database based on InnateDB. To improve the visualization, the software Cytoscape (https://cytoscape.org/) and Gephi (https://gephi.org/) were also used.

Meta-analysis of Dengue and Rheumatoid Arthritis Transcriptome studies
The transcriptome datasets of patients with either rheumatoid arthritis (RA) or Dengue infection were downloaded from the Gene Expression Omnibus (GEO) under the accession GSE51808 and GSE51808. DEGs between RA patients and healthy controls and between Dengue-infected patients and non-infected subjects were identified using the limma package (Adjusted P-value < 0.05 and fold-change > 1.25).

Bone marrow-derived macrophage preparation and infections
Bone marrow-derived macrophages (BMDMs) were prepared using tibia and femur from 6-to 12-week-old mice as previously described (53). Wild type (WT)

Caspase-1 evaluation by western blot analysis and endogenous caspase-1 staining using FAM-YVAD-FMK
In order to measure active caspase-1 we used FLICA assay. Briefly, 10 6 BMDMs were seeded in 12-well plates overnight, and then infected with CHIKV at a MOI of 5 for 24 hours. As a positive control, we used 20 μM of nigericin (Sigma-Aldrich) for 40-60 minutes. After that, cells were harvested and stained for 1 h with a green fluorescent dye that binds specifically to active caspase-1, FAM-YVAD-FMK (Immunochemistry Technologies), following the manufacturer's instructions. The data were acquired on a FACS ACCURI C6 flow cytometer (BD Biosciences) and analyzed with the FlowJo software (Tree Star).
Data were plotted and analyzed with GraphPad Prism 6.0 software (GraphPad, San Diego, California). Multiple groups were compared by two-way analysis of variance (ANOVA) followed by the Bonferroni's post-test were used. The differences in values obtained for two different groups were determined using an unpaired, twotailed Student's t test with a 95% confidence interval. Differences were statistically significant when the p value was less than 0.05.