Research ArticleInfectious Disease

Immune profiles provide insights into respiratory syncytial virus disease severity in young children

See allHide authors and affiliations

Science Translational Medicine  22 Apr 2020:
Vol. 12, Issue 540, eaaw0268
DOI: 10.1126/scitranslmed.aaw0268

Understanding severity in RSV

Almost all young children become infected with respiratory syncytial virus (RSV), but only a small proportion progress to severe disease, and certain immune responses have been shown to be harmful. To better understand immune signatures associated with mild or severe disease, Heinonen et al. performed transcriptomic and flow cytometry analysis on samples from RSV-infected children under the age of two. Children had either severe (inpatients) or mild disease (outpatients). Severe disease samples showed reduced viral loads and type I interferon signatures, as well as the presence of HLA-DRlow monocytes. Overall, a robust innate immune response seemed indicative of a more favorable outcome.

Abstract

Respiratory syncytial virus (RSV) is associated with major morbidity in infants, although most cases result in mild disease. The pathogenesis of the disease is incompletely understood, especially the determining factors of disease severity. A better characterization of these factors may help with development of RSV vaccines and antivirals. Hence, identification of a “safe and protective” immunoprofile induced by natural RSV infection could be used as a as a surrogate of ideal vaccine-elicited responses in future clinical trials. In this study, we integrated blood transcriptional and cell immune profiling, RSV loads, and clinical data to identify factors associated with a mild disease phenotype in a cohort of 190 children <2 years of age. Children with mild disease (outpatients) showed higher RSV loads, greater induction of interferon (IFN) and plasma cell genes, and decreased expression of inflammation and neutrophil genes versus children with severe disease (inpatients). Additionally, only infants with severe disease had increased numbers of HLA-DRlow monocytes, not present in outpatients. Multivariable analyses confirmed that IFN overexpression was associated with decreased odds of hospitalization, whereas increased numbers of HLA-DRlow monocytes were associated with increased risk of hospitalization. These findings suggest that robust innate immune responses are associated with mild RSV infection in infants.

INTRODUCTION

The burden of respiratory syncytial virus (RSV) infection in infants and young children is substantial, both in terms of morbidity and mortality (14). In high-income countries, RSV is the most common cause of hospitalization in infants, but it also causes considerable morbidity in children who develop mild disease, resulting in an extensive number of outpatient visits (57). The majority of infants and young children hospitalized with RSV infection are previously healthy, without predisposing underlying conditions (8). Although age is a strong determinant of RSV disease severity, with younger children being more at risk, we still lack a comprehensive understanding of the different clinical, immunologic, and virologic factors present in infants with mild RSV disease that protect them from developing severe infection.

Although the contribution of viral factors, such as viral loads or specific RSV genotypes to RSV disease severity, remains to be completely defined (912), there is substantial evidence indicating that a dysregulated host immune response plays an important role in clinical disease severity (1318). An improved understanding of the viral-host interactions may help with the development of RSV vaccines and antivirals, as identification of the immunoprofile associated with mild RSV infection in infants could be used as a marker of ideal vaccine-elicited responses. In addition, understanding the relationship between viral loads, innate immune responses, and age may help in identifying the ideal therapeutic window and specific patient populations that would benefit from antiviral therapy.

We implemented a multidimensional analytical approach to define the distinct variables present in infants and young children with mild versus severe RSV infection stratified according to age by integrating clinical and virologic data concurrently with transcriptomic and cellular immune profiles. Previous studies have shown that innate immunity gene expression and innate cytokine production were inversely correlated with RSV disease severity in hospitalized children (13, 14, 16). We thus hypothesized that the systemic innate immune response plays a major role in determining disease severity and that children with mild RSV infection would demonstrate a more competent innate immune response than those with severe disease.

RESULTS

Characteristics of the study participants

We enrolled 190 children <2 years of age, 125 with RSV infection and 65 healthy controls (Fig. 1 and Table 1); 88% were ≤12 months of age. Clinical data and RSV loads were available for all patients; samples for transcriptome analyses were available from 123 children (tables S1 and S2), for immunophenotypic characterization of white blood cells (WBC) from 104 children (Table 2 and table S3), and for both transcriptome and immunophenotypic analyses from 37 children (table S4).

Fig. 1 Flowchart of the study participants.

Number of healthy controls (HC) is depicted in gray, outpatients with RSV in orange, and inpatients with RSV in blue. Number of children with data available for transcriptome analyses, blood cell immunophenotype, and for the combined transcriptome and immunophenotype analyses are further stratified in <6 and 6 to 24 months of age. From a total of 190 participants, transcriptome data were available in 123 patients, and white blood cell (WBC) count immunophenotype data in 104 patients. The dashed line differentiates the patients included in the transcriptome (left) or immunophenotype (right) analyses. Patients included in the transcriptome analyses were then divided in the discovery (n = 72) and validation (n = 55) cohorts. Of the 104 patients included in the immunophenotype analyses, *37 had also transcriptome data available and **4 HC were hybridized twice and included in both the discovery and validation cohorts. n, number.

Table 1 Study children demographic, clinical, and virologic data.

CDSS, clinical disease severity score; PICU, pediatric intensive care unit; NA, not available. Data are presented medians 25 to 76% IQR in brackets […] and numbers with percentages in parentheses (…). Comparisons were computed using chi-squared, Mann-Whitney, or Kruskal-Wallis tests as appropriate. Values in bold indicate statistical significance.

View this table:
Table 2 Blood cell immunophenotype according to age and severity.

Cell numbers were calculated as the proportion of CD45+ cells × total WBC count and presented as median cells/μl among 93 children with WBC count data available (34 HC, 24 OPs, and 35 IPs). Each age group (<6 and 6 to 24 months) was analyzed using Kruskal-Wallis followed by Benjamini-Hochberg multiple test correction. HLA-DR, human leukocyte antigen–antigen D related; NK, natural killer; mDC1 and mDC2, myeloid dendritic cell type 1 and 2; pDC, plasmacytoid dendritic cell; TFH, T follicular helper. Values in bold indicate statistical significance.

View this table:

Overall, outpatients with RSV [median interquartile range (IQR) age, 7.6 months (5.3 to 11.9 months)] and healthy controls [7.0 months (4.2 to 9.7 months)] were older than hospitalized infants [2.6 months (1.4 to 5.8 months); Table 1]. Regardless of the need for hospitalization, the percentage of lymphocytes was lower, and of neutrophils and monocytes higher in routine WBC counts in RSV-infected children versus healthy controls. For all analyses, we stratified outpatients with RSV, inpatients with RSV, and healthy controls according to age: <6 months versus 6 to 24 months of age. The use of <6 months as a cutoff for age has been a common approach in the RSV literature, likely reflecting that this is the infant population most likely to be hospitalized with severe disease (1, 4). In addition, our previous studies focused exclusively in hospitalized patients with RSV lower respiratory tract infection (LRTI) showed important differences in immune profiles in infants <6 months versus those >6 months of age (13).

Duration of illness at enrollment was not different between groups with a median (IQR) of 3 days(2 to 5 days) for outpatients and 4 days (3 to 5 days) for inpatients (P = 0.15). Results remained unchanged when patients were stratified according to age: <6 months of age [inpatients: 4 (3 to 5) versus outpatients: 3 (2 to 6) days; P = 0.18] or 6 to 24 months of age [inpatients: 4.5 (3 to 7) versus outpatients: 4 (3 to 5) days; P = 0.31].

Median RSV loads were higher in outpatients (7.1 log10 copies/ml) versus inpatients (6.5 log10 copies/ml; P = 0.013). Rates of codetection of RSV with other respiratory viruses were also higher in outpatients (51%) versus inpatients (24%; P < 0.001). Rhinovirus was the most commonly codetected virus in both groups. There were no differences in the proportion of RSV A and B infections between groups.

To determine whether duration of illness may have influenced viral loads, we first explored this association in the whole cohort (n = 125) and found a weak but negative correlation between viral loads and duration of illness [correlation coefficient (r) = −0.21; P = 0.019]. This association was lost when viral loads were analyzed separately in inpatients (r = −0.17, P = 0.104) and outpatients (r = −0.21, P = 0.18). Next, we analyzed viral loads according to days of symptoms in 1 to 2, 3 to 4, 5 to 6, and ≥7 days and found that although RSV loads were higher in outpatients at all times, the differences did not reach statistical significance (fig. S1).

Transcriptional profiles in mild and severe RSV infection

Age appears to influence the immune response to RSV (13). Thus, to account for age, we analyzed the transcriptional profiles of outpatients with RSV, inpatients with RSV, and healthy controls matched for age (discovery cohort, n = 72; Fig. 1 and table S1). Results were then validated in a test set of non–age-matched cohort of outpatients with RSV, inpatients with RSV, and healthy controls (validation cohort, n = 55; Fig. 1 and table S2). The same healthy controls were included in the outpatient and inpatient discovery signatures. Similarly, the same healthy controls were included in the validation outpatient and inpatient signatures. Using principal components analysis (PCA) and principal variance component analysis (PVCA) in R (19), we identified a technical batch effect within the discovery cohort that was associated with globin reduction (fig. S2A). The batch effect was corrected using the ComBat function of the sva package in R (fig. S2B) (20, 21). In addition, to assess the potential impact of duration of illness on transcriptome analyses, we performed PCA and found no clustering according to days of symptoms, suggesting that the effect on transcriptome profiles was not substantial (fig. S3).

Group comparisons [false discovery rate (FDR), P < 0.01, Benjamini-Hochberg multiple test correction, and 1.25-fold change] between outpatients with RSV [n = 24; median age, 6.2 months (4.5 to 8.5 months)] and healthy controls [n = 24; 7.3 months (4.4 to 9.9 months)] in the discovery cohort identified 890 differentially expressed transcripts (RSV outpatient signature; Fig. 2A). This signature was validated by PCA in an independent set of nine outpatients with RSV and 14 healthy controls (validation cohort; Fig. 2B).

Fig. 2 Transcriptional RSV outpatient and inpatient signatures.

(A) The mild (outpatient) RSV transcriptional signature (discovery cohort) was derived by comparing 24 outpatients with RSV and 24 age-matched healthy control using an FDR P < 0.01, Benjamini-Hochberg multiple test correction and 1.25-fold change. Data are displayed in a heatmap format, where each column represents the gene expression of a patient and each row represents a transcript. Red indicates overexpression, blue indicates underexpression, and yellow indicates no difference in expression compared with healthy control. (B) Validation of the outpatient signature in an independent, non–age-matched cohort (n = 23) by principal components analysis (PCA). RSV outpatients are represented in orange and healthy controls in gray. (C) The transcriptional signature for severe RSV disease included 24 inpatients with RSV and 24 age-matched healthy controls (discovery cohort) that was validated (D) in an independent cohort using PCA (n = 46). (E) The Venn diagram shows the number of shared and unique transcripts between the outpatient and inpatient signatures. The size of the circles reflects the overall transcript abundance for the inpatient and outpatient signatures, respectively.

We followed the same approach to define the signature of severe RSV infection (RSV inpatient signature) and identified 1992 differentially expressed transcripts between 24 inpatients with RSV [median age, 6.7 months (3.6 to 9.4 months); discovery cohort] and 24 healthy age-matched controls (Fig. 2C and table S1). This signature was also validated using PCA in an independent cohort of 32 inpatients with RSV and 14 healthy controls (validation cohort; Fig. 2D).

The RSV outpatient and inpatient signatures were combined using a Venn diagram (Fig. 2E) that revealed a total of 2330 transcripts identified in either condition: 24% (n = 552) shared, 15% (n = 338) specific to outpatients with RSV, and 62% (n = 1440) to inpatients with RSV. The top 10 shared and specific over- and underexpressed transcripts in outpatients and inpatients are shown in table S5. Shared transcripts included those related to interferon (IFN) (e.g., IFI27, OTOF, and IFIT3), also identified among the top overexpressed transcripts in outpatients with RSV (i.e., IFIT2 and CXCL10). On the other hand, the top overexpressed transcripts in inpatients with RSV were mostly related to neutrophil function (i.e., MMP9, DEFA1, sDEFA3).

Modular analysis according to disease severity

To further define the biological function and pathways associated with RSV disease severity, we applied a modular analytical strategy as described (22). Modules are groups of coexpressed genes that shared a similar function, where modular over- and underexpression is defined by the percentage of transcripts within each module that are differentially expressed in the condition of interest (in this case, inpatients with RSV or outpatients with RSV) versus healthy controls (13).

Modular maps were first derived from the RSV inpatient and outpatient discovery cohorts and then confirmed in the validation cohorts (Fig. 3, A to C). Whereas modules related to innate immunity were overexpressed in inpatients and outpatients, overexpression of IFN modules was greater in outpatients (P < 0.01), and overexpression of neutrophil and inflammation modules in inpatients (P < 0.01; Fig. 3 and table S6). On the other hand, underexpression of T cell and cytotoxic T cell/natural killer (NK) cell genes was more pronounced in inpatients with RSV than outpatients. Last, outpatients showed greater activation of plasma cell–related genes compared with inpatients (P < 0.01). Modular patterns were confirmed as demonstrated by a strong correlation between the discovery and validation cohorts for outpatients (r = 0.97, P < 0.0001) and inpatients (r = 0.96, P < 0.0001; Fig. 3, B and C). The database for annotation, visualization, and integrated discovery (DAVID) functional annotations and Ingenuity Pathway Analysis confirmed these findings and showed that the top biological pathways associated with outpatients with RSV were related to IFN signaling, whereas those identified in inpatients were associated with inflammation and host defense responses (table S7).

Fig. 3 Modular fingerprints in mild and severe RSV infection.

(A) Modular maps were derived from the outpatient (OP; n = 24) and inpatient (IP; n = 24) discovery cohorts. Each dot represents a transcriptional module with red indicating overexpression and blue indicating underexpression in relation to age-matched healthy controls (n = 24). The number and color intensity on the dot indicate the percentage of differentially expressed transcripts within a module. Findings were corroborated in the outpatient (B) and inpatient (C) validation cohorts using by Spearman correlation analysis.

Influence of age and disease severity on transcriptional immune profiles

To better understand the influence of age on immune gene expression during RSV infection, we analyzed the RSV inpatient versus outpatient signatures in children <6 months (table S8) versus 6 to 24 months of age (table S9). In the younger age group (<6 months), overexpression of inflammation genes was greater in inpatients than in outpatients (P < 0.001), whereas overexpression of neutrophil and IFN modules was comparable irrespective of disease severity (Fig. 4A). On the other hand, inpatients 6 to 24 months old demonstrated greater overexpression of both neutrophil and inflammation modules (P < 0.001) versus outpatients, who showed significantly greater expression of IFN genes (P < 0.001) (Fig. 4B). In summary, overexpression of inflammation transcripts was associated with enhanced disease severity independent of age, whereas only in the older age group, overexpression of neutrophil genes was associated with severe disease (inpatients) and overexpression of IFN genes with mild disease (outpatients).

Fig. 4 Modular expression in children with RSV infection according to age.

Modular analyses were performed separately in (A) 29 children <6 months (outpatients, n = 7; inpatients, n = 12; and healthy controls, n = 10) and (B) 43 children 6 to 24 months of age (outpatients, n = 12; inpatients, n = 14; and healthy controls, n = 17). Significantly different modules identified by chi-square test are presented in spider graph and color coded according to disease severity (blue for inpatients and orange for outpatients).

Blood immune cell populations according to age and disease severity

In parallel, we applied multiparameter flow cytometry for phenotypic characterization of blood immune cells in 104 children: 28 outpatients (mild), 40 inpatients (severe), and 36 healthy controls, who were also stratified in the <6 versus 6 to 24 months of age (table S3). Results are presented in Fig. 5, A to D, Table 2 (cell counts), table S10 (cell percentages), and data file S1.

Fig. 5 Innate and adaptive blood immune phenotype in children with mild or severe RSV infection according to age.

Blood immune cell counts from RSV outpatients (OP; orange), inpatients (IP; blue), and healthy controls (HC; gray) stratified by age. Each dot represents the per patient cell count of a specific innate cell type in infants <6 months (A) and 6 to 24 months of age (B), and adaptive immunity cell type in infants <6 months (C) and 6 to 24 months (D). Horizontal black lines indicate median values with 25 to 75% IQR. Comparisons made by Kruskal-Wallis test, followed by Benjamini-Hochberg post hoc test. Asterisks indicate statistical significance (*P < 0.05 and **P < 0.01). NS, nonsignificant (P ≥ 0.05).

In infants <6 months of age, neutrophil counts were comparable between healthy controls and RSV-infected infants irrespective of disease severity. By contrast, neutrophil counts were significantly higher (P < 0.01) in older (6 to 24 months) inpatients with RSV, followed by outpatients compared with healthy controls. Monocyte counts were higher in inpatients with RSV versus controls irrespective of age and mainly driven by the classical (CD14++ CD16dim) and intermediate (CD14++ CD16+) subsets. In addition, inpatients with RSV irrespective of age had a significant increase in the number of monocytes with low expression of human leukocyte antigen–antigen D related (HLA-DR) compared with healthy controls or outpatients with RSV (P < 0.05). Counts of NK cells and NK cell subtypes (CD56/CD16 bright or dim) were lower in outpatients with RSV and inpatients with RSV versus healthy controls, although differences reached statistical significance only in the older age group (P = 0.01). In infants <6 months of age, dendritic cell (DC) numbers [including plasmacytoid (p)DCs and myeloid (m)DCs], were exclusively decreased in inpatients, whereas, in the 6- to 24-month-old group, DC numbers were decreased in patients with RSV irrespective of disease severity.

CD3+ and CD4+ T cell counts were decreased in RSV-infected children versus healthy controls, but the difference was significant only in outpatients with RSV and irrespective of age (P < 0.05). The decrease in CD4+ T cell numbers was driven by naïve CD4+ cells. There were no differences in CD8+ T cell counts between inpatients and outpatients <6 months, but in children 6 to 24 months old, CD8+ T cells were decreased in patients with RSV versus healthy controls. The decrease in CD8+ T cell numbers was mainly driven by CD8+ central memory and naïve CD8+ T cell counts. Numbers of peripheral blood T follicular helper (TFH) cells (P= 0.02), TFH2 (P = 0.01), and TFH17 (P = 0.03) subsets were significantly decreased in inpatients with RSV versus outpatients <6 months of age but comparable across study groups (inpatients with RSV, outpatients with RSV, and healthy controls) in children 6 to 24 months of age. Absolute B cell (CD19+) numbers, mainly driven by naïve B cells, were decreased in patients with RSV (inpatients <6 months and outpatients 6 to 24 months) versus healthy controls, whereas plasma cell numbers were significantly decreased only in younger inpatients with RSV (P = 0.04). In summary, these data suggest that differences between mild and severe disease in specific innate immune cell populations were more evident in children with a more mature immune system, whereas adaptive immune responses were influenced by severity and age.

Correlations between cellular and transcriptional immune profiles

To better understand the associations between transcriptome and cellular immune profiles, we calculated the correlations between modular gene expression and the different immune cell populations in 37 children who had both datasets available (Fig. 6 and table S4). For this analysis, we used percentages instead of absolute cell counts since WBC data, and thus absolute counts, were missing in 5 of 37 patients.

Fig. 6 Correlations between transcriptional and cellular immunoprofiles.

Correlations between percentages of genes over- or underexpressed per module (y axis), and percentage of immune cell populations (x axis) in 37 children with both datasets available (healthy controls, n = 5; outpatients, n = 20; and inpatients, n = 12). The intensity of the color in the dot indicates the strength of the Spearman’s correlation coefficient, and the dot size indicates the significance of the P value.

We found significant associations (P < 0.0001) between percentages of the major cell subsets and modular expression of transcripts associated with those cell populations, including T cells and T cell modules, NK cells and the NK cell/cytotoxic cell module, B cells and plasma cells and B cell/plasma cell modular expression, and monocytes and monocyte modular expression. We also found significant correlations between the plasma cell module and DC and TFH cell percentages (P < 0.001). In addition, the IFN modules M1.2 and M5.12 correlated with monocytes and TFH cells, respectively, but inversely with T cells, and monocytes expressing low HLA-DR. This integrated analyses, in addition to confirming the findings of the cellular and transcriptome data separately, identified biological meaningful associations, such as TFH cells with expression of plasmablasts and IFN transcripts.

Viral loads and host immune parameters associated with disease severity

Last, we analyzed the associations between viral loads and clinical parameters with the cellular and transcriptome immune profiles in univariate (Fig. 7), followed by multivariable analyses (Table 3). As another parameter of severity, we used a standardized clinical disease severity score (CDSS) (13, 14), which allowed computing correlations with continuous variables. Univariate analyses showed significant associations (P < 0.05) between the CDSS and inflammation genes, but the CDSS inversely correlated with the numbers of TFH cells, plasma cells, and B cells (Fig. 7, A and B). RSV loads correlated with IFN modular expression but negatively with numbers of B cells, monocytes, and HLA-DRlow monocytes.

Fig. 7 Correlations between clinical parameters and transcriptional and cellular immune profiles.

(A) Correlations between transcriptional modular expression (x axis) with clinical parameters and viral loads (y axis) were analyzed in the cohort of RSV+ infants included in the discovery cohort (outpatients, n = 24; inpatients, n = 24). (B) Correlations between immune cell counts (x axis), clinical parameters, and viral loads (y axis) were analyzed in 59 RSV+ children with immunophenotype data available (outpatients n = 24; inpatients n = 35). Significant correlations are represented with color and numeric values, indicating the strength of the correlation by Spearman’s correlation coefficient (blue, negative correlation; red, positive correlation). Analysis adjusted for multiple testing using Benjamini-Hochberg test. Only statistically significant correlations are shown.

Table 3 Adjusted associations between IFN expression and HLA-DRlow monocytes in relation to hospitalization.

Odds ratios (ORs) are presented with 95% lower confidence level (LCL) and upper confidence level (UCL). Values in bold indicate statistical significance.

View this table:

To further assess the impact of IFN expression and HLA-DRlow monocytes on the risk of hospitalization adjusted for age, viral loads, and days of symptoms, we performed logistic regression analyses. Because of the limited number of patients with both transcriptional and cellular profiles datasets available and to avoid overfitting, models were constructed separately so that a maximum of two risk factors were considered within any given model (Table 3). Median IFN expression showed a protective association with odds of hospitalization when controlling for all factors except HLA-DRlow monocytes and inflammation genes. On the other hand, higher numbers of HLA-DRlow monocytes showed a significant association with odds of hospitalization when controlling for most factors, except inflammation. Overall, these data suggest that higher IFN expression exerts a protective effect with regard to hospitalization adjusted for other covariates, whereas increased numbers of HLA-DRlow monocytes are independently associated with a greater risk of hospitalization.

DISCUSSION

In this study, we applied a systems analysis approach and combined two powerful tools, transcriptional profiling and multiparameter flow cytometry, together with measurement of viral loads and comprehensive clinical data to gain a better insight into factors associated with mild RSV disease. By enrolling infants and young children spanning the disease severity spectrum and across different ages, we uncovered a complex interaction between RSV replication, the innate immune response, and age. Counterintuitively, outpatients with mild RSV infection had higher RSV loads than those hospitalized with severe disease. There were also marked differences in immune gene expression patterns, as children with mild disease and of older age showed greater expression of IFN compared with inpatients with severe disease, who demonstrated higher expression of inflammation genes and greater numbers of HLA-DRlow monocytes irrespective of age. The protective role of IFN in terms of hospitalization and the detrimental association between increased numbers of monocytes expressing low HLA-DR and hospitalization was confirmed in multivariable analyses. Overall, these data suggest that RSV infection is associated with distinct patterns of activation of the innate immune response according to disease severity and that these patterns are greatly influenced by age (fig. S4).

The majority of studies evaluating host responses in children with RSV infection have largely focused on children hospitalized with severe disease and suggest that a dysregulated immune response may contribute to disease severity (11, 1318, 23, 24). In the present study, however, we shifted the focus and assessed host immune responses in children who were somewhat “protected” from developing severe RSV disease and were managed as outpatients. We showed that the transcriptional signature of mild RSV infection (outpatients) was characterized by greater overexpression of IFN and plasma cell genes, less activation of neutrophil and monocyte genes, and less suppression of T cell and cytotoxic/NK cell genes, compared with children with severe RSV infection (inpatients). We also found that some responses were greatly influenced by age, as overexpression of inflammation transcripts was consistent in severe RSV infection, irrespective of age, whereas overexpression of neutrophil-related genes was greater in inpatients versus outpatients only in the older age group (6 to 24 months of age). Responses in these older children with less severe disease were also characterized by greater overexpression of IFN genes. Our findings highlight the protective role of robust IFN responses and the detrimental role of an uncontrolled inflammatory response, because these pathways appear to exert opposite effects on the pathogenesis of RSV infection. On the other hand, one can argue that an uncontrolled inflammatory response in children with severe RSV disease may just be a consequence of the innate immune system compensating for an underdeveloped adaptive immune response (13, 25).

In addition to this unique pattern of innate immune activation, outpatients had higher RSV loads. This observation is intriguing, as others have shown a significant association between RSV loads and disease severity in children hospitalized with RSV infection (10, 26). Nevertheless, two recent studies showed that infants (median age, 6 months) with mild RSV infection managed as outpatients had higher RSV viral loads compared with hospitalized children, and suggested a model where higher viral loads in the early stages of infection may promote a robust innate immune response that protects the infant from developing severe disease (11, 27). Another recent study conducted in hospitalized infants with RSV infection found that those with severe disease requiring pediatric intensive care had lower viral loads and weaker mucosal IFN responses than infants with moderate disease hospitalized in the inpatient ward (28). Our findings demonstrating that higher RSV loads and more robust IFN responses were independently associated with lower odds of hospitalization confirm those previous studies and agree with the proposed model.

We also applied multiparameter flow cytometry for phenotypic characterization of blood immune cells and observed that neutrophil numbers were increased only in inpatients of older age who also demonstrated greater overexpression of neutrophil-related genes. We also found that although monocyte counts were increased in severe RSV disease irrespective of age, monocytes expressing low HLA-DR were increased in parallel, suggesting that monocyte activation might be impaired in children with severe disease. Although monocyte function was not directly assessed, reduced monocyte HLA-DR expression has been reported in clinical conditions associated with immune dysfunction such as bacterial sepsis, cardiac surgery, or in infants with severe RSV infection (15, 29). In addition, in our previous studies, we assessed monocyte function after lipopolysaccharide stimulation and found that TNF-α production was significantly impaired in infants with severe RSV disease (14).

In agreement with previous studies (3032), we also found decreased numbers of NK cells and DCs, both in inpatients with RSV and outpatients with RSV, compared with age-matched healthy controls. Although the total CD3+ T cell and CD4+ T cell counts were decreased in outpatients with RSV with mild RSV infection and irrespective of age, CD8+ T cells were decreased in children 6 to 24 months old with severe RSV infection. Cytotoxic T cells have an important role in controlling viral infections (25), and one potential explanation for the reduction in the number of blood T cell or DC counts could be related to their recruitment to the lungs (3335). Moreover, changes in T cell subsets in outpatients and inpatients were characterized by decreased naïve CD4+ T cells and central memory in CD8+ T cells. Last, we found a significant reduction in blood TFH cells and B cells in inpatients <6 months. The latter was also identified in outpatients 6 to 24 months, mainly driven by the reduced numbers of naïve B cells. Overall, these data suggest that there are common and specific responses triggered by RSV according to disease severity in an immune system that is developing. Although older children with mild RSV disease mounted a more robust innate immune response, differences in adaptive immune cell numbers appeared to be driven by differences in severity rather than age.

The combined analyses of transcriptomic and cellular immune profiles showed significant associations between the main cell types and expression of transcripts associated with those cells (modular expression). In addition, we found that the percentage of peripheral TFH cells, which are important in providing help to B cells, significantly correlated with plasma cell and IFN gene expression. IFN modular expression also correlated with monocyte counts but inversely with total T cells, suggesting that global transcriptional profiles not only reflect changes in the WBC composition but also provide further insight into interactions between different immune cell types and underlying mechanisms.

Understanding the complex interactions between the virus and the host in different age groups may have implications on different fronts. On the one hand, defining the different endotypes in response to RSV may help with patient stratification in the clinical setting. It can also help to identify the therapeutic window and ideal patient populations that would benefit from antiviral therapy. In addition, the identification of a “safe and protective” profile induced by natural RSV infection across different age groups could be used as a as a surrogate of ideal vaccine-elicited protective responses in future clinical trials.

This study has limitations. Although a multidimensional analytical approach was undertaken, we focused on systemic immune responses, and information derived from the respiratory mucosa, the primary site of the infection, was not available. Instead, we used systemic responses as a surrogate and were able to derive and validate the transcriptional profiling data in independent cohorts, further strengthening the findings. Our study was also limited by its cross-sectional design, and although analyses were adjusted by duration of illness, the study provides only a snapshot of a dynamic event. Function of specific cell types or integration between immune cell function and expression data was not performed, as single cell data were not available. Nevertheless, we were able to combine transcriptomic, immune cell phenotyping, and clinical and viral load data from a large cohort of children representative of the RSV disease severity spectrum and of different age groups. The large sample size allowed age matching, which is a major challenge when studying infants, especially those with mild RSV disease who do not require hospitalization.

In summary, we defined and validated a transcriptional and cellular signature in children with mild RSV infection. Outpatients with RSV had higher RSV loads, and their immune profiles were characterized by greater overexpression of transcripts related to IFN and less prominent or weaker overexpression of inflammation genes. These findings provide valuable insights into the host immune response to RSV and its relation to disease severity and might be useful for designing clinical studies evaluating novel vaccines and antivirals directed against RSV.

MATERIALS AND METHODS

Study design

We sought to define the clinical, viral loads and immune profile differences between children with mild RSV (outpatients) versus severe RSV infection (hospitalized) across different ages. We hypothesized that RSV viral loads and host immune responses would differ according to disease severity and age at the time of the infection. To this end, we conducted a prospective cohort study involving a convenience sample of previously healthy children <2 years of age with acute RSV infection and a cohort of healthy, asymptomatic age-matched controls. The study was conducted at Nationwide Children’s Hospital (NCH; Columbus, OH) from 2011 to 2015. Children were enrolled at NCH urgent care clinics, the emergency department, or in the inpatient hospital units [ward or pediatric intensive care unit (PICU)] within a median 25 to 75% IQR of 24 hours (17 to 38 hours) of hospitalization and blood and nasopharyngeal (NP) samples obtained. RSV was diagnosed per standard of care using a rapid antigen assay or polymerase chain reaction (PCR) test. Healthy controls were enrolled during well-child visits or minor elective surgical procedures not involving the respiratory tract (13). Exclusion criteria included documented bacterial coinfections, premature birth (≤36 weeks of gestation), congenital or chronic medical conditions, and immunodeficiency. For healthy controls, additional exclusion criteria included presence of fever or symptoms of respiratory tract infection within 1 weeks of enrollment.

Blood samples were analyzed for (i) transcriptome, (ii) cell immunophenotype, and (iii) WBC count with differential. Samples were drawn in this order, but due to the small blood volumes obtained in some children, we were unable to obtain all samples in all participants. NP swabs were collected in inpatients with RSV and outpatients with RSV at enrollment by trained study personnel using an approach that has been validated in multicenter trials (36, 37). To confirm and quantify RSV loads, quantitative reverse transcription PCR analyses targeting the N gene were performed in NP samples of all children included in the study (14). Other respiratory viruses were detected using PCR assays (either Luminex xTAG Respiratory Panel or the FilmArray respiratory viral panel, BioFire, BioMérieux) (27).

Demographic and clinical parameters were collected using a standardized questionnaire and by reviewing electronic health care records. Disease severity was assessed primarily by the need for hospitalization [inpatients (=severe disease) versus outpatients (=mild disease)] and using a standardized CDSS (14, 27). In hospitalized patients, additional parameters of severity were collected including administration of supplemental oxygen, PICU admission, and total duration of hospitalization. In addition, families were contacted 2 and 4 weeks after enrollment to confirm the lack of subsequent readmissions in the inpatient group or hospitalization in the outpatient cohort. Primary data are reported in data file S1.

Transcriptome analyses

Sample collection, processing, and RNA hybridization. Blood was collected in Tempus tubes (Applied Biosystems, Foster City, CA) and stored within 2 to 4 hours at −20°C until further analyses. RNA was extracted, processed, and hybridized into Illumina HumanHT-12 v4 microarray chips (47,323 probes; Illumina, San Diego, CA) as described (13). After hybridization data were scanned on Illumina BeadStation500. Illumina GenomeStudio software was used for background subtraction and to scale average signal intensities as described (13, 38).

Data preprocessing. We first selected transcripts that were “present” (signal precision <0.01) in ≥10% of samples [present in at least (PAL10%); 18,213 transcripts] as described (13, 39). Next, raw expression values <10 were set to 10, and the data were log2 transformed. A technical batch effect, related to globin reduction, was identified using PCA and PVCA (19) in the discovery cohort. This batch effect was corrected using the sva package in R (fig. S2, A and B) (20, 21).

Differential gene expression analysis (transcriptional signatures). Two datasets were used for transcriptome analyses. We first derived the transcriptional signature for mild (outpatients) and severe (inpatients) RSV infection by comparing each RSV cohort to the same healthy controls in the discovery cohort. Patients for this cohort were matched for age and gender. We used the limma package in R (40) and applied stringent statistical filtering (FDR P < 0.01, Benjamini-Hochberg multiple test correction, and 1.25-fold change). The transcriptional signatures identified were validated with PCA in a not age-matched validation cohort that included outpatients with RSV, inpatients with RSV, and healthy controls (Fig. 2, B and D). The two cohorts were independent except for four healthy controls whose samples were hybridized twice and used in the discovery and validation cohorts (Fig. 1). The samples included in the study were processed in two different batches (fig. S5), which largely overlapped in terms of enrollment (batch #1: 2012–2014 and batch #2: 2012–2015). To avoid any potential batch effect, samples included in the discovery cohorts were hybridized in batch #1, whereas those used for validation were hybridized in batch #2.

Functional characterization of differentially expressed transcripts. To define the biological function of the “mild” and “severe” RSV signatures, we applied modular transcriptional analyses. Briefly, this is a systems scale that aims to reduce the abundance of transcriptional data into functional pathways or modules. Transcriptional modules are formed by genes coordinately expressed, thus allowing functional interpretation of the microarray data into biologically useful information. Modular over- and underexpression was defined by the percentage of transcripts within each module that were differentially expressed in outpatients with RSV and inpatients with RSV versus healthy controls. A detailed description of this mining analysis strategy has been reported elsewhere (22).

Modular maps for the outpatient and inpatient discovery cohorts were derived by comparison with the same age-matched healthy controls (FDR P < 0.05, Benjamini-Hochberg multiple test correction). Results were confirmed by (i) computing correlations of modular expression between the discovery and validation cohorts (Spearman’s correlation coefficient; Fig. 3B), (ii) by DAVID bioinformatics tool (version 6.7, available at https://david-d.ncifcrf.gov) (41), and (iii) Ingenuity Pathway Analysis tool (QIAGEN, Redwood City, CA, USA; table S7). Age analyses using modular expression were performed in children <6 months versus 6 to 24 months of age according to disease severity (outpatients versus inpatients) using chi-square test and adjusted for multiple comparisons by Benjamini-Hochberg multiple test (table S5).

Blood immune cell populations

Blood samples (1 to 2 ml) were obtained in acid citrate dextrose (ACD) tubes (BD Vacutainer ACD Solution B; BD, Franklin Lakes, NJ) and processed within 2 to 4 hours of collection. Five blood aliquots of ~200 μl each were stained with different antibody panels for characterization of innate (neutrophils, monocytes—including HLA-DRlow monocytes as a functional marker of activation, NK cells, and DCs) and adaptive immune cell populations [T cells, peripheral TFH cells, and B cells; tables S11 and 12]. Samples were incubated with antibody panels for 15 min, and red blood cells lysed with BD FACS lysing solution (BD Biosciences, San Jose, CA). Stained cells were washed twice with phosphate-buffered saline 1× before storage at 4°C until cell acquisition, which was performed in 1 to 3 days on an LSRII flow cytometry instrument (BD Biosciences, San Jose, CA). Data were analyzed using FlowJo software v9.8.2 (TreeStar, Ashland, OR). Data are presented as absolute numbers (Table 2 and Fig. 5) in children with total WBC data available (89%, 93/104 children) and percentages (table S10). Individual patient-level data are provided in data file S1. We used immune cell counts for all analyses except to compute correlations between immune cell populations and modular gene expression, in which we used percentages because of the lack of WBC count data in 5 of 37 patients with paired data.

Sample size calculations and statistical analyses

For sample size calculation, best practices in the transcriptome field dictate utilization of at least two independent sets of samples of 15 to 20 participants per group for the purpose of validating candidate signatures (or profiles) (13, 21). We analyzed demographic and clinical data using SPSS v22.0 (IBM Corp.) and GraphPad Prism v7.0b software packages, which were presented as medians with 25 to 75% IQRs. We compared continuous variables using either Mann-Whitney or Kruskal-Wallis test with Dunn’s or Benjamini-Hochberg post hoc tests for multiple testing, and categorical variables using either chi-square or Fisher’s exact test.

For transcriptome and cellular immunophenotype data, we used R (R Foundation for Statistical Computing). The statistical tools used for microarray analyses are included in each of the above sections. For correlations between transcriptomic and cellular, clinical, or virology data, we used Spearman’s correlation coefficient.

Last, we analyzed whether the main immune variables found to be relevant in the study (IFN expression and HLA-DRlow monocyte numbers) were associated with hospitalization. To this end, we conducted multivariable analyses using logistic regression with Firth’s correction for small sample size after accounting for other factors. Because of the limited sample size, additional covariates were added in separate models, so that a maximum of two risk factors were included within any given model. Analyses were conducted using SAS 9.4 (SAS Institute), with two-sided P < 0.05 considered statistically significant.

SUPPLEMENTARY MATERIALS

stm.sciencemag.org/cgi/content/full/12/540/eaaw0268/DC1

Fig. S1. Viral loads according to duration of illness in infants with mild (outpatients) or severe (inpatients) RSV infection.

Fig. S2. Batch effect addressed using ComBat batch correction method.

Fig. S3. PCAs of transcriptome data from the inpatient and outpatient discovery cohorts to assess the impact of duration of symptoms on viral loads.

Fig. S4. Schematic figure of the interaction between RSV disease severity, viral loads, and age.

Fig. S5. Flow diagram of patient sample selection for transcriptome analyses.

Table S1. Demographic, clinical, and virologic data of 72 children included in the discovery age-matched cohort for transcriptome analyses.

Table S2. Demographic, clinical, and virologic data of 55 children included in the validation cohort for transcriptome analyses.

Table S3. Demographic and laboratory data of patients with RSV and healthy controls included in the blood cell immunophenotype analysis according to age (<6 months and 6 to 24 months of age).

Table S4. Demographic, clinical, and virologic data of 37 children included in the combined transcriptome and immunophenotype analysis.

Table S5. Specific and shared top over- and underexpressed transcripts included in the outpatient and inpatient biosignatures.

Table S6. Modular analyses of immune response related transcriptional modules.

Table S7. Top biological processes and pathways associated with the specific and shared RSV outpatient and inpatient biosignatures.

Table S8. Demographic, clinical, and virologic data of children <6 months of age included in the age group modular analyses.

Table S9. Demographic, clinical, and virologic data of children 6 to 24 months included in the modular analyses.

Table S10. Percentages of WBC subsets in infants with RSV infection and healthy controls according to age.

Table S11. Antibodies and fluorochromes used to stain WBC for immunophenotype analyses by multicolor flow cytometry.

Table S12. Cell surface markers used to define WBC populations by flow cytometry immunophenotyping.

Data file S1. Primary data.

REFERENCES AND NOTES

Acknowledgments: We are thankful to J. Thompson, V. Olson, S. Sharpe, and G. Wentzel at NCH, Columbus, OH, for the extraordinary efforts with patient enrollment and coordination; C. Smitherman and P. Nguyen at the Baylor Institute for Immunology Research, Dallas, TX, for help with RNA processing; and specially to our patients and their families for agreeing to participate in the study. Funding: This work was supported by an NIH grant (AI112524) to A.M., O.R., and M.E.P.; NCH intramural funds to A.M.; and the European Society for Paediatric Infectious Diseases Fellowship Award and grants from the Finnish Medical Foundation and the Foundation for Pediatric Research to S.H. Author contributions: A.M. and O.R. conceived and designed the study; S.H., C.G.-M., E.B., A.M., D.M.C., and S.O. acquired the clinical data; S.H., B.S., M.E.P., S.A.-B., and M.M.-C. analyzed the data and performed statistical analyses; S.H., V.M.V., F.Y., S.M., C.-G.M., S.A.-B., and E.B. analyzed and interpreted the flow cytometry data; S.H., O.R., and A.M. interpreted the results and wrote the first draft of the paper. All authors have critically reviewed the manuscript. Competing interests: A.M. has received research grants from Janssen, fees for participation in advisory boards from Janssen and Roche, and fees for lectures from AbbVie. O.R. has received research grants from the Bill and Melinda Gates Foundation and Janssen; fees for participation in advisory boards from Sanofi, MedImmune, and Merck; and fees for lectures from Pfizer. M.E.P. has received research grants from the Cystic Fibrosis Foundation and Janssen, and fees for participation in an advisory board or lectures from ReViral and Pfizer. All the above grants and fees were not related to the research described in this manuscript. Data and materials availability: All data associated with this study are present in the paper or Supplemental Materials. The transcriptomic data have been deposited at the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO accession number: GSE105450).

Stay Connected to Science Translational Medicine

Navigate This Article