Longitudinal Study of Cellular and Systemic Cytokine Signatures to Define the Dynamics of a Balanced Immune Environment During Disease Manifestation in Zika Virus–Infected Patients

Understanding the host immune response during Zika virus (ZIKV) infection will provide valuable insights into ZIKV immunopathogenesis. In this study, characterization of cellular changes and immune mediators of ZIKV-infected patients was performed, allowing for the identification of key infection-associated markers.

This is the first comprehensive study that combines cellular changes with specific immune signatures during ZIKV infection, and it provides a better understanding on the pathobiology of the virus during infection. In addition, specific immune signatures allow for the prospect of identifying key markers associated with different disease manifestations.

Standard Protocol Approvals, Registrations, and Patient Consent
Written informed consent was obtained from all participants in accordance with the Declaration of Helsinki for Human Research. Study protocols were approved by the SingHealth Centralized Institutional Review Board (reference 2016/2219) and the National Healthcare Group Domain Specific Review Board (reference 2015/00528). Collection of blood samples from healthy donors was done with written consent in accordance with guidelines from the Health Sciences Authority of Singapore (study approval number National University Singapore Institutional Review Board approval 10-250).

Patients and Sample Collection
The first case of ZIKV infection was reported on 27 August 2016 [15], and detection was confirmed by ZIKV-specific quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis [16]. Thereafter, suspected ZIKV cases were referred/ admitted to the Communicable Disease Centre at Tan Tock Seng Hospital for ZIKV testing. Hematological and biochemistry laboratory tests were performed in parallel with ZIKV-specific RT-PCR analysis [16] upon admission. ZIKV infection was confirmed by a positive result of ZIKV-specific RT-PCR analysis of whole-blood specimens. Subsequently, whole-blood specimens were obtained at 5 collection time points: (1) the acute phase (2-7 days after illness onset), (2) the early convalescent phase (10-14 days), (3) the late convalescent phase (25-35 days), (4) the early recovery phase (2-4 months), and (5) the late recovery phase (5-6 months). Whole-blood specimens were collected in ethylenediaminetetraacetic acid-coated Vacutainer tubes (Becton Dickinson) after peripheral venipuncture. Two milliliters of whole blood was first aliquoted for blood count analysis, wholeblood staining, and viral load quantification, and the remaining sample was centrifuged at 240g for 10 minutes to collect plasma for storage at −80°C. Samples from healthy donors were included and prescreened for the presence of ZIKV viral RNA [16] and ZIKV-specific antibodies [17]. All healthy donors were nonfebrile and had no signs of acute illness during recruitment.

Blood Count
Complete blood count was performed using the Ac·T diff hematology analyzer (Beckman Coulter) according to the manufacturer's instructions. Beckman Coulter 4C Plus Tri-Pack Cell Controls (Beckman Coulter) were conducted to confirm instrument accuracy and precision performance.

Whole-Blood Staining and Flow Cytometry
Staining was performed on 100-μL whole-blood specimens from 44 ZIKV-infected patients and 14 healthy donors. Antibodies were used to identify CD45 + leukocytes (mouse anti-human CD45; Biolegend), CD16 + neutrophils (mouse anti-human CD16; Biolegend), CD14 + monocytes (mouse anti-human CD14; BD Biosciences), CD3 + T cells (mouse anti-human CD3; BD Biosciences), CD4 + helper T cells (mouse anti-human CD4; eBioscience), CD8 + cytotoxic T cells (mouse anti-human CD8; BD Biosciences), CD56 + natural killer (NK) cells (mouse anti-human CD56; Miltenyi Biotec), and CD19 + B cells (mouse anti-human CD19; eBioscience). Subsequently, cell fixation and red blood cells lysis were performed with 1X FACS lysing solution (BD Biosciences). Permeabilization was achieved with 1X FACS permeabilization solution 2 (BD Biosciences), after which cells were stained with ZIKV NS3 protein-specific rabbit polyclonal antibody [14]. Stained cells were counterstained with a fluorophore-tagged secondary goat anti-rabbit IgG (H+L) antibody (Invitrogen), before acquisition with LSR Fortessa analyzers, LSRII 4 lasers, and LSRII 5 lasers (BD Biosciences). Since samples were transported on ice within 3 hours to the laboratory at the Singapore Immunology Network, live/dead staining was not implemented because freshly collected whole-blood specimens typically contain very few dead cells [14]. Numbers of peripheral blood cell immune subsets were subsequently obtained with the following formula: [percentage of a specific immune subset, obtained from immunophenotyping] × [total leukocyte count, obtained from complete blood count] = the number of cells in a specific immune subset.

ZIKV Load Quantification
RNA samples were extracted from 140-μL whole-blood samples, using the QIAamp Viral RNA Mini Kit (Qiagen) according to the manufacturer's protocols. ZIKV quantification was performed by 1-step TaqMan real-time RT-PCR analysis (QuantiTect Probe RT-PCR Kit; Qiagen) as described previously [18,19].

Multiplex Microbead Immunoassay for Cytokine Quantification
Cytokine and chemokine levels in ZIKV-infected patients' plasma were measured simultaneously, using a multiplex microbead-based immunoassay (ProcartaPlex Human Cytokine/Chemokine/Growth Factor Panel 1; Thermo Scientific) as described previously [18]. Preparation of plasma samples and reagents, as well as immunoassay procedures, were performed according to manufacturers' instructions.

Data Processing and Statistical Analysis
Luminex assay-determined concentrations obtained via Bioplex Manager software (using 5-parameter logistic curve fitting) were normalized using median centering to remove potential plate effects. Each analyte was normalized separately, such that the median concentration for each plate was first determined and then the global median concentration from all plates was computed. A scaling factor was then computed for each plate, which adjusted the median concentration for the plate to that of the global median concentration. The final adjusted concentrations were then logarithmically transformed to assume a normal distribution before further data analysis and visualization.
Two-way analysis of variance (ANOVA) with the post hoc Tukey test was used to detect differences between the various sample groups and collections. ANOVA results were corrected for multiple testing, using the method of Benjamini and Hochberg. Nonparametric testing was done for cell percentages and viral load data, using the Mann-Whitney test or the Kruskal-Wallis test with Dunn multiple comparison tests. Data processing was done in the R statistical language (version 3.3.1). The relationship between analytes at different collections was also determined using Pearson correlation analysis. All statistical analyses were performed using R version 3.3.1 or Prism 7.0 (GraphPad software).
Hierarchical clustering and heat map visualization were done using TM4-MeV [20]. In the heat map presentation, the average concentration was computed for each measured analyte in its respective group, and the average values were then scaled between 0 and 1 for visualization [18].
Receiver operating characteristic (ROC) curve analysis for analytes that were differentially expressed between ZIKVinfected patients and healthy controls was performed, and the areas under the curve (AUCs) were calculated. Analytes with AUCs of >0.65 and a P value of <.05 are considered to be potential markers of infection.
A multivariate model was built using patient demographic characteristics, immune mediator profiles, and immunophenotyping data. The analysis was done using a logistic regression fit optimized for analysis by the Akaike information criterion (AIC) in a step-forward fashion. The number of steps was limited to a floor value (calculated as the number of samples/10), to ensure that there were sufficient data for fitting. Leave-one-out cross-validation was then used to assess the predictive quality of the model. Biological processes were predicted from differentially expressed mediators with Ingenuity Pathway Analysis (IPA; Qiagen). Interaction networks of selected immune mediators were predicted with STRING (version 10.5; available at: https://string-db.org/).

Demographic and Clinical Characteristics of ZIKV-Infected Patients
A total of 55 patients were recruited for the study, based on confirmatory results of ZIKV-specific PCR analysis upon hospitalization, and were subsequently recruited into the study. The majority of the patients returned for their follow-up consultations during the acute phase (55 patients), the early convalescent phase (34), the late convalescent phase (36), the early recovery phase (35), and the late recovery phase (32

ZIKV Infection Triggers High Levels of Proinflammatory Mediators
To define the acute host immune markers during ZIKV infection, levels of immune mediators present in patients' plasma obtained during the acute phase were quantified by a 45-plex microbead assay ( Figure 1A).  Figure 1B). Further categorization of ZIKV-infected patients on the basis of viremic status (Table 1) revealed differences between viremic and nonviremic patients ( Figure 1A and Supplementary  Table 1). Collectively, these data highlight the presence of unique cytokine profiles among ZIKV-infected patients.

ZIKV Infection Reduces the Peripheral Blood Cell Count
Blood cell count and subset profiles are important criteria for determining the diagnosis and prognosis of viral infections [21].
Healthy controls  (CD14 + ), B cells (CD19 + ), T cells (CD3 + CD4 + , CD3 + CD8 + ), double-negative (CD3 + CD4 -CD8 -) T (DNT) cells, NK cells (CD56 + and CD56 hi ) NK T cells (CD56 + CD3 + ), and CD14 + CD56 + cells during acute ZIKV infection were profiled (Figure 2A). It was observed that, during acute ZIKV infection, the total number of blood leukocytes in the patients was significantly lower than in healthy controls, owing to the overall reduction of all immune populations except monocytes ( Figure 2B). It was observed that ZIKV-infected patients generally had decreased numbers of B cells, DNT cells, and neutrophils. The presence of detectable viremia affected the numbers of other immune subsets ( Figure 2B), while in nonviremic patients, an intermediate phenotype (ie, a nonsignificant drop in cell number) was observed. Notably, although ZIKV-infected patients had leukopenia, the number of cells in each immune subset was still within the normal range [22].

Disease Symptoms
Because the ZIKV epidemic was anticipated, a study was planned with a standardized care path to collect demographic data, comorbidity data, hospitalization details, travel history and sick contact, pregnancy and vaccination status, signs and symptoms, and laboratory data on complete blood count; renal and liver function test results; diagnostic test results for ZIKV, dengue virus, and chikungunya virus; and clinical outcomes (ie, full recovery, sequelae, and death). In this study, symptoms were defined as moderate if a patient presented with >4 of 8 symptoms presented in Table 1. Patients who did not fulfill this criterion were classified as having mild symptoms ( Figure 3A). It was observed that patients with moderate ZIKV disease experienced higher incidences of myalgia, arthralgia, conjunctivitis, and cough ( Figure 3B). Rash was present in >90% of patients from both groups ( Figure 3B).
To further identify specific unique immune profiles, ZIKVinfected patients were categorized on the basis of both their viremic status and symptom manifestation, as follows: (1)    IP-10, and PIGF-1, compared with other patients. These patients also had increased levels of IL-8 ( Figure 3C and Supplementary Table 2). Levels of these immune mediators were relatively similar among the other 3 groups. In terms of total blood profiling, viremic patients experiencing moderate symptoms of ZIKV infection had reduced numbers of DNT cells, CD4 + T cells, and CD8 + T cells ( Figure 3D). To further understand the link between the host immune response and ZIKV disease progression, multivariate prediction on ZIKV viremic status and disease outcome was conducted using data from both the microbead assay and immunophenotyping. The analysis predicted that HGF could influence disease outcome (Supplementary Table 3). On the other hand, the percentages of monocytes and CD56 hi NK cells were predicted to influence symptom severity, whereas neutrophils were predicted to influence viremia (Supplementary Table 3).

After the Acute Phase of ZIKV Infection
Blood immunophenotyping findings and immune markers were further profiled longitudinally during the acute, early and late convalescent, and early and late recovery phases among ZIKV-infected patients ( Figure 4A). Levels of 14 immune mediators were found to be significantly higher than those in healthy controls during acute ZIKV infection. Of these, 13 could be potential markers for acute ZIKV infection, as calculated by ROC analyses that evaluated their diagnostic values (Supplementary Table 4). Uniquely, there exists a group of immune mediators that remain differentially expressed from the acute phase through the postacute phases of ZIKV infection ( Table 2). Although levels in the majority (ie, in IP-10, IL-10, IL-1RA, IL-12p70, IL-1β, RANTES, and IFN-γ) peaked during the acute phase, levels peaked in later phases for some (ie, IL-18 [convalescent phase] and SDF-1α, IL-21, and MIP-1β [recovery phase]). These immune mediators could potentially be useful markers of these phases, based on ROC analyses (Table 2). Interestingly, levels of IL-21 were significantly lower during the acute phase and peaked only during the recovery phase, making it a potential marker for the recovery phase (Table 2).
To further understand the consequences of these differentially expressed immune mediators in ZIKV-infected patients, cellular developmental processes were predicted using IPA. It was observed that processes involving proliferation and expansion of T cells were important ( Figure 4A). However, blood cell counts at both the convalescent and recovery phases did not indicate any significant differences between ZIKV-infected patients and healthy controls for all immune subsets.

DISCUSSION
In this study, ZIKV-infected patients from the Singapore cohort had high levels of proinflammatory mediators, antiinflammatory mediators, CCL/CXC chemokines, and growth factors. While this robust immune response was expected, it is worthwhile to note that high levels of IFN-γ, IL-18, IL-10, IP-10, and TNF-α are consistent with our earlier report detailing the immune mediator profile from a Brazilian patient cohort [18]. Uniquely, high levels of GM-CSF, IL-2, IL-4, IL-5, IL-6 IL-9, IL-17A, IL-22, MCP-1, HGF, and TNF-α and low levels of PIGF-1 and PGDF-ββ were identified only in the acute phase, making them useful markers for acute ZIKV infection. As such, the prognostic and predictive values on the combinatorial assessment of these markers should be further understood to improve patient management. The duration of detectable acute ZIKV viremia is relatively short [23]. With this in mind, ZIKV-infected patients were further grouped on the basis of both viremic status and symptom manifestations. Viremic patients with moderate disease exhibited high levels of IL-1RA, MCP-1, IP-10, PIGF-1, and IL-8. IL-8 could be a symptom severity marker only in such patients, indicating that subtle unique differences could be missed if detailed clinical and laboratory parameters are incomplete. However, in this study, it was observed that the window for detecting viremic patients with moderate symptoms is short, because levels of IL-1RA, IL-8, PIGF-1, MCP-1, and IP-10 were no longer significant after the acute phase (Supplementary Table 2). While the majority of the aforementioned immune mediators have been previously associated with dengue virus infection [24,25], PIGF-1 could have an important role in ZIKV-induced congenital abnormalities, because placental PIGF-1 has roles in vasculogenesis and angiogenesis in placenta [26]. Coincidentally, levels of MCP-1, IP-10, IL-1RA, and IL-8 were higher in patients with severe dengue [24,27], similar to what was observed in this cohort of ZIKV-infected patients.
a By the Student t test, with correction for multiple testing, using the Benjamini-Hochberg method.
b Immune mediators with a P value < .05 are considered to exhibit a reasonable AUC, signifying their potential as a marker during the different phases of ZIKV infection. chikungunya virus, but not for ZIKV infection. In this cohort, it was observed that the patients experienced transient leukopenia and neutropenia, possibly due to transmigration of immune cells into the tissues during acute infection. The role of blood immune cells in ZIKV infection is understudied, but monocytes are a reported cellular target for ZIKV [11][12][13][14]19]. Moreover, monocytes have been shown ex vivo to modulate the host immune response during ZIKV infection [13]. High levels of MCP-1 and GM-CSF detected in patients during the acute phase further confirm the role of monocytes during infection. Lower numbers of DNT cells, CD4 + T cells, and CD8 + T cells were observed in viremic patients with moderate symptoms during the acute phase. This could be due to high levels of IP-10 in these patients, because IP-10 is responsible for recruitment of T cells to infection sites [28]. CD8 + T cells have been demonstrated to have a protective role during ZIKV infection in immunocompetent animals [29], and their cytotoxicity can be influenced by their responsiveness to IL-8 [30]. Likewise, higher IL-1RA levels in these ZIKV-infected patients could also signal regulated IL-1 signaling in the Th17 subset of the CD4 + T cells [31,32]. On the other hand, little is known about the function of DNT cells in viral infections. DNT cells are commonly associated with autoimmune lymphoproliferative syndrome [33] and are known to share some characteristics with regulatory T cells and CD8 + T cells [33]. Their functional role in ZIKV infection should therefore not be neglected. The overall importance of T cells during acute ZIKV infection is further highlighted by the general presence of T-helper type 1 (Th1)-related (IL-2 and IFN-γ), Th2-related (IL-4), Th17related (IL-17A), and Th9-related (IL-9) cytokines detected in this study, an observation that has been previously reported [34].
Functional prediction by IPA suggested active roles for T cells during the convalescent and recovery phases of ZIKV disease, while STRING analysis (performed with a high confidence threshold of 0.85) revealed key interactions between 11 of these immune mediators ( Figure 4B). As expected, the predicted interactions are linked to T cells. IFN-γ, a known T-cell cytokine [35], was predicted to positively trigger the activity of IP-10 and IL-1β, which are mediators involved in T-cell recruitment and activation [28,31,36]. It was also predicted to interact with IL-10, an antiinflammatory cytokine [37], resulting in reciprocal inhibition. IFN-γ activity could in turn be triggered by IL-12p70 [38], IFN-α [39], and IL-18 [40]. IL-10, a regulator of immunity [41], was predicted to inhibit the activity of IP-10, RANTES [42], and IL-1β. Interestingly, the balance between IFN-γ and IL-10 has been reported to be critical in the pathogenesis of autoimmune diseases [43,44]. Other interactions predicted by STRING analysis suggest binding and catalytic relationships between other T-cell chemokines, RANTES, SDF-1α and IP-10 [45]. Taken together, these findings emphasize the importance of a balanced cytokine environment involved in the T-cell response [46] during the postacute phases of ZIKV infection.
In this cohort, fever, conjunctivitis, arthralgia, and myalgia were the main symptoms in moderately symptomatic patients, in contrast to those reported for dengue virus infection [47]. The finding that viremic and moderately symptomatic patients had higher levels of IL-8, MCP-1, and IP-10 could possibly explain the higher incidence of conjunctivitis, through the recruitment of immune cells to the conjunctiva [48]. Moreover, these cytokines were reported to be associated with acute arthralgia and myalgia in chikungunya virus-infected patients [49,50].
The marked differences in disease presentation between mildly and moderately symptomatic patients reported in this study are useful in developing a more robust definition of ZIKV infection. Importantly, the presence of viremia does not necessarily equate to more-severe disease. Likewise, symptomatic patients can be nonviremic, which suggests the interplay between symptoms, viremia, and host immune response in defining disease severity. Collectively, ZIKV infection stands out on its own apart from other flaviviruses, and the findings reported here will open up new research angles that can only increase knowledge about ZIKV immunopathogenesis.

Supplementary Data
Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.