Upper airway gene expression reveals a more robust innate and adaptive immune response to SARS-CoV-2 in children compared with older adults

Unlike other respiratory viruses, SARS-CoV-2 disproportionately causes severe disease in older adults and only rarely in children. To investigate whether differences in the upper airway immune response could contribute to this disparity, we compared nasopharyngeal gene expression in 83 children (<19-years-old; 38 with SARS-CoV-2, 11 with other respiratory viruses, 34 with no virus) and 154 adults (>40-years-old; 45 with SARS-CoV-2, 28 with other respiratory viruses, 81 with no virus). Expression of interferon-stimulated genes (ISGs) was robustly activated in both children and adults with SARS-CoV-2 compared to the respective non-viral groups, with only relatively subtle distinctions. Children, however, demonstrated markedly greater upregulation of pathways related to B cell and T cell activation and proinflammatory cytokine signaling, including TNF, IFNγ, IL-2 and IL-4 production. Cell type deconvolution confirmed greater recruitment of B cells, and to a lesser degree macrophages, to the upper airway of children. Only children exhibited a decrease in proportions of ciliated cells, the primary target of SARS-CoV-2, upon infection with the virus. These findings demonstrate that children elicit a more robust innate and adaptive immune response to SARS-CoV-2 infection in the upper airway that likely contributes to their protection from severe disease in the lower airway.


Introduction 46
One of the defining features of the COVID-19 pandemic has been the striking relationship 47 between disease severity and age [1][2][3] . While infection with other respiratory viruses, such as 48 influenza or respiratory syncytial virus (RSV), causes significant morbidity and mortality in both 49 young children and older adults 3-9 , severe COVID-19 occurs disproportionately in older adults 50 and only very rarely in children 1,3,10-14 . A recent comprehensive modeling study estimated that the 51 infection fatality rate is lowest for children ages 5-9 (~0.001%) and that even adults in their 40s 52 may already be at 100-fold greater risk of death from COVID-19 1 . The age-dependent effect on 53 disease severity and mortality has been shown even when accounting for age-associated 54 comorbidities 15 . 55 A few studies have examined differences in systemic immunological profiles of 56 hospitalized children and adults with COVID-19, revealing greater breadth and neutralizing activity 57 of SARS-CoV-2 specific antibodies as well as stronger CD4+ T cell responses to viral spike 58 protein in adults 16,17 . These studies suggested that poorer outcomes in adults are not due to failure 59 to engage a systemic adaptive immune response, and that children may exhibit a less pronounced 60 adaptive response because their infection course is milder. One study reported higher serum 61 levels of IL-17A and interferon-γ (IFNγ) in children early in their hospitalization 17 , suggesting a 62 more robust innate immune response may contribute to the milder disease course. 63 Our understanding of age-related differences in the immune response at the site of initial 64 infection, the upper airway, remains limited. Several large-scale studies have found no systematic 65 differences between children and adults in the distribution of SARS-CoV-2 viral load measured in 66 nasopharyngeal (NP) swabs 18,19 , suggesting children are not generally better able to control viral 67 replication in the upper airway. However, differences in the upper airway microenvironment and 68 immune response could still contribute to protection from severe disease in children, for example, 69 by limiting migration of the virus into the lower airway. 70 Small-scale studies have begun to compare the upper airway immune response to  CoV-2 in children and adults, with some contradictory results. One study found that children with 72 COVID-19 expressed higher levels of genes associated with innate immune pathways, including 73 interferon-stimulated genes and genes related to NLRP3 inflammasome signaling 20 . Another 74 study, however, found no age-related differences in interferon-stimulated gene expression and 75 reported globally similar host transcriptional responses between adults and children with diverse 76 types of viral infections 21 , highlighting the need for further investigation. Importantly, neither study 77 directly controlled for SARS-CoV-2 viral load when comparing immune-related gene expression 78 between children and adults. 79 We previously used metagenomic RNA-sequencing of NP swabs to compare upper airway 80 gene expression in adult patients with COVID-19, other viral acute respiratory illnesses or non-81 viral illnesses 22 . Our analysis revealed a pronounced interferon response in COVID-19 patients, 82 proportional to SARS-CoV-2 viral load, but attenuated activation of additional innate immune and 83 pro-inflammatory pathways compared to patients with other viral infections 22 . Here, we report new 84 sequencing data from a relatively large pediatric cohort to enable assessment of age-related 85 differences in the upper airway transcriptional response to SARS-CoV-2 infection and shed further 86 light on these outstanding questions. When controlling for viral load, our results suggest that 87 differences in the overall magnitude of interferon-stimulated gene expression in the upper airway 88 of children and adults with COVID-19 are relatively subtle and seem unlikely to explain their 89 distinct clinical outcomes. However, we also find clear evidence of more robust pro-inflammatory 90 and adaptive immune responses in the upper airway of children, which may contribute to their 91 protection from severe disease. 92 93

Results and Discussion 94
To compare the upper airway gene expression response to SARS-CoV-2 infection in 95 children and adults, we utilized our previously published dataset of NP swab RNA-sequencing 96 from an adult cohort alongside newly sequenced swabs from a pediatric cohort. All samples in 97 both cohorts were obtained in the course of clinical testing for SARS-CoV-2 by reverse 98 transcription polymerase chain reaction (RT-PCR) at the University of California San Francisco 99 or Children's Hospital Colorado. We included patients up to 19 years of age in the pediatric cohort 100 and restricted the adult cohort to those 40 years of age or older to impose clearer age separation. 101 As in our previous analysis, we divided each age cohort into three viral status groups: 102 1) patients with PCR-confirmed SARS-CoV-2 infection ("SARS-CoV-2" group), 2) patients 103 negative for SARS-CoV-2 by PCR with no other pathogenic respiratory virus detected by 104 metagenomic RNA sequencing ("No Virus" group), and 3) patients negative for SARS-CoV-2 who 105 had another respiratory virus detected by sequencing ("Other Virus" group). Finally, we limited 106 the samples in the SARS-CoV-2 group to those with at least 10 viral reads-per-million ( Table 1A). Most of the patients in the SARS-CoV-2 group in both age cohorts 114 were tested as outpatients, indicative of an early/mild stage of disease (Supplemental Table 1A). 115 Samples in the SARS-CoV-2 group in both age cohorts spanned several orders of magnitude of 116 viral load, and while viral load trended higher in the children this did not reach statistical 117 significance ( Figure 1C). Rhinovirus was the most prevalent other respiratory virus detected in 118 both age cohorts and a diversity of other viruses were also identified ( Figure 1D). 119 We began by performing differential expression (DE) analyses between the SARS-CoV-2 120 and No Virus groups within each age cohort separately. This approach minimizes confounding by 121 age-related gene expression differences unrelated to SARS-CoV-2 infection and any potential 122 batch effects, though it could be affected by differences among the patients in each cohort's No 123 Virus group. The analyses yielded 1,961 and 1,216 differentially expressed genes at an adjusted 124 p-value < 0.1 for the pediatric and adult cohorts, respectively (Supplemental Data 1). We then 125 performed gene set enrichment analyses 26 (GSEA) using Gene Ontology (GO) biological process 126 annotations 27 on the DE results in each cohort and compared the enriched pathways. 127 As expected, a range of immune related pathways were upregulated in both adults and 128 children with SARS-CoV-2 compared to those with no virus ( IFNγ production, IL-2 and IL-4 production) and chemokine/cytokine signaling were more highly 145 expressed in children with  While some immune pathways did not reach statistical significance in the direct 147 comparison between children and adults with SARS-CoV-2, they typically trended in the same 148 direction ( Figure 2B). On the other hand, the stark disparity in neutrophil activation observed in 149 the comparison to the No Virus groups was only weakly supported in the direct comparison, likely 150 reflecting differences among the No Virus patients themselves. The direct comparison clearly 151 revealed lower expression of cilia-associated genes in children with SARS-CoV-2 and also 152 suggested a trend toward lower expression of interferon-stimulated genes, though the pathway 153 just missed the statistical significance cutoff. As expected, many developmental processes 154 unrelated to infection also differed in the direct comparison between children and adults 155 (Supplemental Data 4). 156 Many of the pathways identified in the GSEA results as differentially expressed between 157 children and adults with COVID-19 were tightly related to particular cell types. We therefore 158 applied in silico estimation of cell type proportions 30 based on marker genes derived from an 159 airway single-cell study 31 as an additional approach to contextualize our findings (Figure 3, 160 Additionally, pediatric samples were obtained from patients tested for COVID-19 by RT-PCR from 228 NP swabs at Children's Hospital Colorado (CHCO). CHCO specimens and data were obtained 229 under IRB protocols #0865, #20-1617 and #20-0972, which also granted a waiver of consent. Genes were retained for each DE analysis if they had at least 10 counts in at least 20% 297 of the samples included in the analysis. All analyses were performed with the R package limma 41 , 298 using quantile normalization and the voom method. The design formula for the comparisons within 299 each age cohort was ~viral status, where viral status was either "SARS-CoV-2" or "No Virus". The 300 design formula for the direct comparison between children and adults with SARS-CoV-2 was 301 ~log 10 (rpM) + age cohort, where age cohort was either "children" or "adults". 302 proportions between comparator groups were evaluated for statistical significance using a Mann-317 Whitney test with Holm's correction for multiple testing. 318

Regression of gene counts against viral load 319
We performed robust regression of the limma-generated quantile normalized gene counts 320 (log 2 scale) against log 10 (rpM) of SARS-CoV-2 for selected genes. The analysis was performed 321 within each age cohort separately using the R package robustbase 44

465
Unknown values were excluded from analyses.

466
Age was analyzed as one-way ANOVA and all categorial variables using chi-squared tests.

467
Supplemental or African American race were excluded from statistical analysis due to insufficient numbers.

475
All unknown values were excluded from analyses.

Figure 3: Cell-type proportion differences between children and adults. 509
In silico estimation of cell-type proportions in the bulk RNA-sequencing using single-cell 510 signatures. Black lines denote the median. The y-axis in each panel was trimmed at the maximum 511 value among all groups of 1.5*IQR above the third quartile, where IQR is the interquartile range. 512 For each cell type, we formally compared each viral status group between the two age cohorts as 513 well as the No Virus and SARS-CoV-2 groups within each age cohort. Pairwise comparisons were 514 performed with a two-sided Mann-Whitney test followed by Holm's correction for multiple testing.