key: cord-1019852-h1egervn authors: Hachim, Asmaa; Gu, Haogao; Kavian, Otared; Mori, Masashi; Kwan, Mike Y. W.; Chan, Wai Hung; Yau, Yat Sun; Chiu, Susan S.; Tsang, Owen T. Y.; Hui, David S. C.; Mok, Chris K. P.; Ma, Fionn N. L.; Lau, Eric H. Y.; Amarasinghe, Gaya K.; Qavi, Abraham J.; Cheng, Samuel M. S.; Poon, Leo L. M.; Peiris, J. S. Malik; Valkenburg, Sophie A.; Kavian, Niloufar title: SARS-CoV-2 accessory proteins reveal distinct serological signatures in children date: 2022-05-26 journal: Nat Commun DOI: 10.1038/s41467-022-30699-5 sha: b403cdc6799c9c06e18cde4330c26670172700aa doc_id: 1019852 cord_uid: h1egervn The antibody response magnitude and kinetics may impact clinical severity, serological diagnosis and long-term protection of COVID-19, which may play a role in why children experience lower morbidity. We therefore tested samples from 122 children in Hong Kong with symptomatic (n = 78) and asymptomatic (n = 44) SARS-CoV-2 infections up to 200 days post infection, relative to 71 infected adults (symptomatic n = 61, and asymptomatic n = 10), and negative controls (n = 48). We assessed serum IgG antibodies to a 14-wide antigen panel of structural and accessory proteins by Luciferase Immuno-Precipitation System (LIPS) assay and circulating cytokines. Infected children have lower levels of Spike, Membrane, ORF3a, ORF7a, ORF7b antibodies, comparable ORF8 and elevated E-specific antibodies than adults. Combination of two unique antibody targets, ORF3d and ORF8, can accurately discriminate SARS-CoV-2 infection in children. Principal component analysis reveals distinct pediatric serological signatures, and the highest contribution to variance from adults are antibody responses to non-structural proteins ORF3d, NSP1, ORF3a and ORF8. From a diverse panel of cytokines that can modulate immune priming and relative inflammation, IL-8, MCP-1 and IL-6 correlate with the magnitude of pediatric antibody specificity and severity. Antibodies to SARS-CoV-2 internal proteins may become an important sero surveillance tool of infection with the roll-out of vaccines in the pediatric population. T he spectrum of SARS-CoV-2 infection ranges from asymptomatic to lethal infection, with the immune response playing a major role in the pathogenicity and outcome of COVID-19 1 . Children are generally less affected clinically by SARS-CoV-2 infection and the morbidity and mortality observed in adults increases progressively with age. The viral loads in the upper respiratory tract are reportedly comparable between children of all ages and adults 2 . Various immune functions and physiological differences have also been implicated in differential outcomes with age, such as lower ACE2 expression in children 3 , pre-existing immunity to common cold coronaviruses (CCoV) 4 , elevated baseline IgM 5 , immuno-senescence, inflammatory state 6 , innate immune responses 7 , auto-antibodies 8 , and off-target "trained immunity" 9, 10 . Multisystem inflammatory syndrome (MIS-C) that can develop in children after infection with SARS-CoV-2 is a rare exception (0.002% of pediatric cases) to the generally milder clinical disease observed 11 . Serology is crucial for determining infection attack rates in the population and for assessing the response to current vaccines to curb the global pandemic. Large epidemiological studies reported that children only represent 1-2% of all SARS-CoV-2 cases in 2020 12, 13 . Most serological tests available rely either on neutralizing antibodies or on the detection of binding antibodies targeting the Spike (S) or the Nucleocapsid (N) proteins of the virus 14 . Across three commercial diagnostics S-based assays, children show a lower rate of seroconversion than adults despite having PCR confirmed infection and comparable viral loads 15 , therefore S-based serology in children may not be an accurate marker of recent infection and underestimate seroprevalence in children. Furthermore, N-specific antibody waning is more pronounced than S 16 and children are more likely to test S-antibody positive even in the absence of vaccination 17 . Therefore, N-based serology in children is also limited 18 compared to adults 19 . Salivabased approaches may offer an easier sampling site with longterm duration of IgG and N-specific responses 20,21 over S 22 and IgA, even in children 23 . We have previously demonstrated that antibodies that are directed against non-structural proteins of the virus, namely ORF3d and ORF8, can be used for accurate diagnosis of SARS-CoV-2 infection in adults 24 . Further studies have also contributed data on the SARS-CoV-2 antibody responses to the virus accessory proteins in the adult population [24] [25] [26] but these accessory specific antibody data are lacking for children. For instance, ORF3d, ORF6, and ORF7a, which have been reported to be potent interferon antagonists that may play a role in immune evasion [27] [28] [29] , or ORF8, which seems to participate to the downregulation of MHC I molecules and to viral pathogenesis 30, 31 . In addition, an imbalanced production of cytokines is responsible of severe COVID-19 outcomes in adults 32 which may modulate seropositivity. Finely tuned and balanced antibody response in relation to cytokine responses may impact SARS-CoV-2 infection outcomes, thus the breadth and magnitude of the specificity of antibody responses to non-structural proteins may indicate the extent of virus replication and thus immune control. In the present study, children and adults with SARS-CoV-2 RT-PCR confirmed infection were used to determine the antibody specificity to a comprehensive panel of 14 different structural and accessory proteins by Luciferase Immuno-Precipitation System (LIPS) (l). Antibody responses were then compared relative to circulating levels of a selected panel of cytokines, known to modulate the antibody response and inflammation. The majority of samples were collected between April to November 2020, before the roll-out of COVID-19 vaccines. Furthermore, due to intensive contact tracing and case-finding measures in Hong Kong, asymptomatic pediatric cases with RT-PCR confirmed infections have been included, which represents a rare entity in most countries and are a unique aspect of our study. Different levels of antibodies to structural proteins in children and adults with SARS-CoV-2 infection. We used the unbiased and quantitative LIPS platform to determine the antibody responses to an extensive panel of 14 antigens from structural and non-structural SARS-CoV-2 proteins in plasma samples from a cohort of infected children, in comparison to adults and controls in Hong Kong and the USA (Table 1) . Our first dataset represents the total cohort of SARS-CoV-2 infected cases of mixed time points and symptoms to determine the overall antibody specificity in children (mean ± stdev: 39 ± 47 days, range: 0-206 days), adults (mean ± stdev: 20 ± 23 days, range: 0-123 days) and negative controls (Table 1 : adults with asymptomatic, mild, or severe disease, children with asymptomatic or mild disease, and negative controls). S and N antibodies are the most widely used antibodies in COVID-19 serology testing worldwide. We therefore first determined the levels of antibodies to different S sub-units by using 3 different S constructs in the LIPS assay: S1 which contains the RBD domain, S2, and the S2′ cleaved subunit (Fig. 1 ). The levels of the two Spike antibodies, S1 and S2′ were markedly lower in children compared to the adult cohort (both p < 0.0001, Fig. 1a , c), whereas no difference was observed for S2 antibodies, revealing different antigenicity for the two Spike isoforms S2 and S2′ (Fig. 1b) 33 . Moreover, N antibodies were significantly elevated in the pediatric COVID-19 cohort relative to negative controls (2.45 × 10 5 ± 2.8 × 10 5 LU versus 4.15 × 10 4 ± 1.5 × 10 5 LU (p = 0.0045), but did not differ from levels observed in adults (Fig. 1d) . We also assessed by LIPS antibodies to other structural proteins Matrix (M) and Envelope (E), which are not widely measured in serology. As for S1 and S2′, we found that M antibody levels were lower in infected children compared to infected adults (p < 0.0001, Fig. 1e ), but were still significantly higher than negative controls. E antibodies had an opposite effect, and were significantly elevated in the pediatric COVID-19 cohort (Fig. 1f ) compared to both adult COVID-19 (p = 0.0006) and negative controls (p < 0.0001). This antigen panel revealed that N and E were the best-performing antigens for diagnostics (based on a cut-off of the negative mean + 3x standard deviations) in the pediatrics population with 65% sensitivity and 100% specificity for N and 78% sensitivity and 100% specificity for E (Fig. 1d, f) , which contrasts to our previous analysis in an adult population where E was not immunogenic 24 . Increased antibody response to the accessory protein ORF8 in the pediatric SARS-CoV-2 infected population. We next investigated the levels of antibodies directed against the nonstructural protein 1 (NSP1) and all the ORF proteins of the virus. In line with our previous study 24 , infected adults had elevated levels of NSP1, ORF3a, ORF3d, ORF7a, ORF7b, and ORF8 antibodies compared to negative controls (p < 0.0001, p < 0.0001, p < 0.0001, p = 0.05, p = 0.0009, p < 0.0001, Fig. 2a -c and e-g). No detectable levels of ORF6 and ORF10 antibodies were found in infected adults (p = 0.8691 and p = 0.999, respectively, Fig. 2d, h) . We observed that the COVID-19 children cohort displayed significantly lower levels of ORF3a, ORF7a, ORF7b antibodies than the COVID-19 adult cohort (p = 0.0001, p < 0.0001 and p < 0.0001, respectively, Fig. 2b , e, f). The magnitude of antibody responses to NSP1 and ORF3d (previously referred to as ORF3b 24 , as ORF3d is within frame of ORF3b but ORF3b is not expressed 34 ) were comparable in the pediatric COVID-19 and adult COVID-19 populations (Fig. 2a, c) . ORF8 antibody levels were found significantly elevated in pediatric COVID-19 samples compared to adults (p < 0.0001, Fig. 2g ). In terms of performance as a diagnostic test, ORF8 antibodies by LIPS allows for the detection of nearly all the pediatric population with a sensitivity of 99.2% and specificity of 100% (Fig. 2g) . These results were then confirmed in an in-house ELISA assay for IgG binding using recombinant proteins in both adult and pediatric plasma samples, where ORF8-protein 35 binding antibodies showed 79% sensitivity and 98.4% specificity ( Supplementary Fig. 1c) , whereas N and Spike-protein binding antibodies showed a sensitivity of 88% and 11% and a specificity of 97% and 99%, respectively ( Supplementary Fig. 1d, e) . Furthermore, ORF8 remains a specific diagnostic tool for SARS-CoV-2 infection in vaccinated conditions. ORF8 specific IgG was assessed following vaccination with whole inactivated virion Coronavac and Spike mRNA lipoprotein BNT162b2 ( Supplementary Fig. 1f) , showing ORF8 is likely not incorporated within the virion. We then compared the cumulative SARS-CoV-2 antibody responses from asymptomatic/mild only COVID-19 children and adult in a heatmap (Fig. 2i ) and as percentages of the total SARS-CoV-2 structural and accessory antibody response (Fig. 2j ). Due to the immunodominant effect of the N protein, anti-N antibodies substantially dominate the SARS-CoV-2 humoral response detected by LIPS in both populations (Fig. 2i) , which is consistent with our previous findings in the adult population 24 . There was a significant difference in the distribution of the overall specificity of pediatrics and asymptomatic/mild COVID-19 adults (p < 0.0001 for "observed" pediatric distribution compared with "expected" adult distribution, Fig. 2j , Supplementary Fig. 1a , Supplementary Table 3 ). In the adult population antibody levels of S1, M, ORF3a and ORF7b represented a higher percentage of the response in adults than children ( Fig. 2j and Supplementary Fig. 1a) . While, ORF8 and E antibody responses represented a higher percentage in the pediatric population with 13.4% versus 9.2% and 11% versus 4.8%, respectively ( Fig. 2j and Supplementary Fig. 1a) . The remaining antigens, S2, S2′, N, NSP1, ORF3d, ORF6, ORF7a, ORF10, represent a similar percentage in both populations and hence are not contributing to the differences observed in specificity. Furthermore, there are no differences in total IgG serum concentration with age or infection, despite differences in the magnitude and specificity of SARS-CoV-2 antibodies in adults versus children ( Supplementary Fig. 1b) . SARS-CoV-2 antibody specificity using clusters of points and principal component analysis. A cluster of points depicts each individual sample in a more complete way than a classical single statistical comparison, as it considers a combination of three (or more) different parameters taken together and the relevant relations of these parameters. To decipher the SARS-CoV-2 antibody specificity in children, we used relevant antibody combinations to represent the COVID-19 pediatric samples in clusters of points relative to negative controls and COVID-19 adult populations (Fig. 3a -c and Supplementary Fig. 2) . First, the cluster representing the three antibodies to the S subunit antigens S1, S2′, S2 confirmed that the pediatric population has a S antibody profile that is more closely comparable to negative controls ( Fig. 3a) than an adult COVID-19 response by LIPS (Fig. 3b) . Further cluster analysis of antibodies to other structural proteins N, M, and E reveals that the COVID-19 children population appears to be quite heterogeneous with a large range in response magnitude compared to adults ( Supplementary Fig. 2b, c) . Despite having a different profile than both the adult COVID-19 and the negative populations, the infected pediatric population cannot be clearly discriminated from these two groups using antibodies to structural proteins. We then selected accessory protein antibodies as combinations to investigate the relevance of unique markers. Previously we showed that, ORF3d, ORF8, and N antibodies, can discriminate accurately COVID-19 adults from negative controls 24 . The (N, ORF3d, ORF8) cluster of points can accurately allow the positive discrimination of the pediatric COVID-19 cases from the negatives (Fig. 3c ). In the (N, ORF8; x, y) plane, the negative population is separated from the pediatric positive one by two-segments of straight lines (equations of 830*log (N) + 0.3843*ORF8 = 4801 and −350*log (N) + 1.036*ORF8 = 790), with all pediatric positive samples (red dots) represented above or on these lines, and only one negative sample (gray triangle) being above these lines (specificity of 96.9% and sensitivity of 100% for pediatric cases). Of note, this plane did not allow an accurate discrimination of the adult positive population (blue dots) with the negative one (gray triangles), as some adult samples (n = 8 of 71) were found below these lines, with the negatives. Interestingly, these 8 samples were early time-point samples (mean time-point sampling: 2.6 days of infection). Furthermore, the plane (ORF8, ORF3d; y, z) and a two-segment delineation (equations of 0.035*ORF3d + 0.1334*ORF8 = 409.284 and 0.074*ORF3d + 0.0437*ORF8 = 221.812) separated the negative samples from all the adult and pediatric positive ones (100% sensitivity and 100% specificity), therefore the combined use of ORF3d and ORF8 most accurately discriminates infected samples from uninfected controls. While the combination of ORF3d and N did not allow any clear discrimination of any of the 3 populations: pediatrics infected, adults infected, and negatives. Therefore, using the (N, ORF3d, ORF8) cluster analysis, the pediatric COVID-19 population resembles a COVID-19 adult population when these markers are taken together only, and can be discriminated from negative pre-pandemic controls. Importantly, this is the only combination that allowed us this discrimination of infected samples, as other parameter combinations (e.g., (S1, S2, S2′) in Fig. 3 , (N, E, M) in Supplementary Fig. 2 ) and combinations of antibodies to accessory proteins) were also tested and represented as clusters of points but did not discriminate pediatric samples. Our combined antigen analysis ( Fig. 2i ) and these various data cluster analyses show that the antibody specificity of the COVID-19 children population is distinct from infected adults. To test the hypothesis that the antibody specificity to structural and accessory viral proteins drives the distinct profile of the pediatric population, we undertook a principal-component analysis (PCA) of antibodies to the 14 SARS-CoV-2 antigens for the full dataset (from Figs. 1 and 2). Dimension (principal component) 1 and 2 explained, respectively, 20.6% and 17.7% of Table 2 ), reflecting that antibodies to structural proteins do not solely drive the principal component 1. Particularly, contributions of ORF3d, NSP1, ORF3a, ORF8 were the highest in Dimension 1 (Dim1, Fig. 3d , e and Supplementary Table 3 ). Moreover, PCA showed that ORF3d and ORF7a antibodies highly contributed to the differences seen in both dimensions (Fig. 3e , f) highlighting their discrimination in the serological response. Strikingly, the PCA revealed that pediatric COVID-19 antibody response was also intermediate between COVID-19 adults and negatives (Fig. 3g) . Indeed, the normal-probability representation of the 3 populations showed that only 31.5% of the pediatric patients overlapped with the ellipse of the COVID-19 adults and only 4.72% overlapped with the ellipse of negative controls (Fig. 3g) . Figure 3h and statistical comparison of the 2-dimensional distributions of the pediatric and adult groups "asymptomatics", "mild", and "severe" revealed that the distribution of the population of severe adult patients was significantly different than that of the adult mild or asymptomatic populations (p = 0.027 for severe versus asymptomatic adult cases, and p = 0.011 for severe versus mild adult cases with the Kolmogorov-Smirnov test 36, 37 . Overall, the PCA showed that adults with severe symptoms had a distinct PC value distribution than adults with mild symptoms (Fig. 3h) , which may be driven by these differences in S1, S2′, and ORF8. Further analysis on sex, infection time-point, and neutralization data (PRNT90) values showed they were not significant factors in discriminating the antibody specificity data (Supplementary Fig. 3) . Therefore, the differences in the observed SARS-CoV-2 antibody responses are primarily explained by the age of i Heatmap comparing the mean concentrations (LU) for structural (N, S, S1, S2′, S2, M, E) and accessory proteins (NSP1, ORF3a, ORF3d, ORF6, ORF7a, ORF7b, ORF10) responses in the COVID-19 pediatric, asymptomatic/mild adult populations (excluding severe cases) and negative controls. j Percentages of single antibody levels to SARS-CoV-2 antigens of the cumulative SARS-CoV-2 antibody response in COVID-19 children and asymptomatic/mild adults (excluding severe cases) for the antigen panel including and excluding N. The asymptomatic/mild pediatric and adult distributions were compared using a Chi-square test for expected versus observed distributions. Expected (adult) and observed (pediatric) percentages are detailed in Supplementary Table 3 . P < 0.0001 for pediatric (observed) versus adult (expected) distribution of both antigen panels including N (left) and excluding N (right). Experiments were repeated twice. Two-sided P values were calculated using the Mann-Whitney U test. * shows statistical significance between COVID-19 patients versus negative controls. *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001. Data in (a-h) represents individual time point LIPS responses and the mean ± stdev, data in (i) represents mean values (LU), data in (j) represents percentages. No difference in antibody responses between symptomatic and asymptomatic COVID-19. To assess the potential effect of antibodies to structural and non-structural proteins of SARS-CoV-2, we further stratified data (from Figs. 1 and 2) into symptomatic (including mild (WHO score 1-3) and severe (WHO score 4) and asymptomatic (WHO score 0) for both the adult and pediatric cohorts (Fig. 4) . We found no differences in antibody responses between asymptomatic versus mild COVID-19 children for all 14 antigens. The same trend was observed in adults ( Fig. 4a, b) . More importantly M, NSP1, ORF6, ORF8, and ORF10 antibody levels in asymptomatic children versus asymptomatic adults were not significantly different (p = 0.3676, p = 0.5216, p = 0.1276, p = 0.2775 and p = 0.0521, respectively, Fig. 4 ), while symptomatic adults had an upregulated antibody response to these antigens compared to symptomatic children (p < 0.0001 for all antigens, except p = 0.0001 for NSP1, Fig. 4 ). Antibody specificity at early infection and long-term stability. We previously observed that the SARS-CoV-2 antibody responses can vary in magnitude and specificity in adults between acute and convalescent to memory time-points 24 . To study the effect of time on the pediatric SARS-CoV-2 antibody specificity, we stratified pediatric responses of all 254 samples (Figs. 1 and 2) by early ( 0.05 for all) (Supplementary Fig. 4) . Effect of age within the pediatric population. Further stratification of the pediatric cohort according to the age of the patients was performed but did not reveal a specific pattern for any agegroups (0-2, >2-10, >11 years old) as shown by the representation of the population as a cluster of points for the most relevant (N, ORF8, ORF3d) antibody combination (Supplementary Fig. 5) . Circulating cytokine levels. Severe COVID-19 has been linked with a cytokine storm 38 and some cytokines are now the targets of COVID-19 therapies, i.e., IL-6 for which a monoclonal antibody therapy is being investigated for the treatment of critically ill COVID-19 patients 39 . Because children mainly suffer from mild COVID we sought to investigate the relationship between antibody production and cytokine profile in this population. We selected pro-Th1/Th2, pro/anti-inflammatory cytokines, and chemokines that had also been described as early prognostic markers of severe COVID-19 40, 41 and therefore play an important role in shaping the adaptive immune priming. Cytokines of interest were quantified at acute stages of infection (d14 are ORF3d p = 0.0309. Significant p values for sympto d14 are S2 p = 0.0021, ORF3d p = 0.0044, ORF7a p = 0.0066, and ORF7b p = 0.0123. Two-sided P values were calculated using the Mann-Whitney U test. * shows statistical significance between symptomatic and asymptomatic samples or between early and late samples *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001. All data represents individual LIPS responses and mean ± stdev. strongly associated in Dim1 with disease severity (p < 0.05 and p < 0.0001, respectively, Fig. 7d ). This highlights the close correlation of antibody responses toward the driving antibodies of Dim.1 (ORF3d, NSP1, ORF8, and ORF3a) and cytokine responses (particularly for IL-6 and IL-8) with disease outcome. Young children account for only a small percentage of reported and medically attended COVID-19 infections 9 . This difference is likely contributed to by differences in host responses between children and adults that ultimately drive adaptive immunity. We present herein a comprehensive study of the magnitude, specificity, and duration of SARS-CoV-2 specific antibodies in children. Our data show that children produce antibodies to some accessory proteins at reduced levels compared to adults (namely S1, M, ORF3a, ORF7b), and that only one accessory target induced an increased antibody response: ORF8. Overall, we found a significant diverse distribution of the antigenic targets in children and adults. These diverse levels of antibodies to structural and accessory ORF proteins may reflect different virus pathogenesis in children compared to adults. Viral loads have been shown to be comparable in children and adults, which may reflect similar levels of viral replication 2 . The higher levels of ORF8 antibodies in the pediatric population could be of particular interest as this glycoprotein has been shown to downregulate MHC-I molecules 30 and seems to play an important role in viral pathogenesis 31 . Similarly, we also observed that severe adult cases 33 . c A linear trend on log 10 LIPS values was fitted for longitudinal samples for S1, S2′, N, M E, NSP1, ORF3a, ORF3d, ORF7a, ORF7b, ORF8 (n = 58 pediatric COVID-19 patients). Linear mixed-effects models were fitted to test the trend of the antibody responses. The trend is specific to each SARS-CoV-2 protein and hence adjustment for multiple comparisons is not needed. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30699-5 ARTICLE had lower levels of ORF8 antibodies than mild adult cases, and that infected children had elevated ORF8 antibodies, therefore ORF8 antibodies may function as a positive immune correlate. Whether ORF8 antibodies found in children can block some of the deleterious functions of this protein remains to be determined by functional studies. Higher ORF8 antibodies in children is also in line with reduced cellular immunity 42 , and may also reflect higher expression of ORF8 during pediatric infection that is known to reduce MHC-I presentation 30 . Antibodies to the structural protein E were also present in higher proportions in children than adults, and the E protein is notorious for high turnover due to its pivotal role in viral propagation (reviewed in ref. 43 ). Therefore the extent and speed of virus replication in pediatric cells may be different to adults due to differences in innate pathway activation 44 leading to antibody priming being elevated in children for E, which warrants further investigation. For the antibodies to Spike sub-units, the (S1, S2′, S2) cluster reveals that the children population resembles a negative prepandemic population and not a COVID-19 adult one. A recent study describes a lower anti-S IgG, IgM, IgA in the pediatric population which correlates with our findings 45 . One explanation for the clinical difference between children and adults is that the pre-existing immunity against seasonal human CCoV that crossreacts with SARS-CoV-2 is higher in children, as they have a higher infection rate of seasonal CCoV than adults 46 . Individuals exposed and unexposed to SARS-CoV-2 have cross-reactive antibodies against the proteins of SARS-CoV-2 and seasonal CCoVs 25, 47 . Moreover, because circulating CCoVs have a higher homology to SARS-CoV-2 structural proteins than non-structural proteins (if they exist) 48,49 , we expect a higher cross-reactivity for structural proteins based on pre-existing immunity. SARS-CoV-2 infection back-boosts antibodies against conserved epitopes, including the relatively conserved fusion peptide of the Spike S2 subunit 25, 47 . In our hands, COVID-19 children and adults had comparable levels of S2 antibodies, contrary to S1 and S2′, which shows a possible effect of pre-existing CCoVs immunity for more conserved domains of S such as S2. Shrock et al. used VirScan, a DNA bacteriophage microarray, to investigate cross-reactivity between SARS-CoV-2 and CCoV in COVID-19 patients 25 . They identified cross-reactive epitopes and found that cross-reactivity was weaker in severe patients than in mild patients, but samples from children were not included in this study. Our observations of lower Spike antibodies in COVID-19 children may indicate that there may be lower sensitivity of serological detection for SARS-CoV-2 when using assays based on S alone 50 , leading to an underestimation of SARS-CoV-2 exposed children. S antibodies have been reported in lower magnitude in the majority of mild adult infections, with higher levels being produced in severe cases 51 , which is consistent with our data on Fig. 3 . Severe (triangles) and mild (square) cases are darker than asymptomatic cases and located on the right of the graph meaning that they strongly associated with Dim-1 ((p < 0.05 and p < 0.0001, respectively). Data were analyzed using the two-sample Wilcoxon test for comparison of the responses between asymptomatic and symptomatic cases. low S antibody levels in children which were also asymptomatic or mild clinical scores. Low antibody levels and low affinity have been associated with Antibody-Dependent Enhancement by facilitation of viral uptake by host cells 52 , it has now been demonstrated that binding of antibodies to the N-terminal domain (NTD) of S enhances infectivity 53 , meaning that lower S-NTD antibody prevalence in children could be advantageous. The combinatory use of ORF8 and ORF3d antibodies has been shown to be a highly specific and sensitive tool for COVID-19 serology diagnostic 24 , and appears here as an accurate tool also for the pediatric population. In line with our results, ORF8 has recently been reported as an immunodominant antigenic site with high sensitivity for the serodiagnosis of mild and severe COVID-19 children 54 . The plane (ORF3d/ORF8) in the cluster of points (N, ORF3d, ORF8) reveals that children samples have specific combinatory values of these two antibodies that is consistent with adult populations, and that makes them distinguishable from uninfected controls. The PCA of our dataset confirmed further the importance of antibodies to accessory proteins in characterizing the pediatric samples, notably ORF3d, NSP1, ORF3a, ORF8. Whether these antibodies to accessory proteins play a role in the virus infectivity or in the pathogenesis of the disease and in the milder outcome of SARS-CoV-2 infection in children presents further questions for investigations. We report in children diverse antibody profiles in early versus late samples and the maintenance or increase of all antibodies to structural and accessory proteins, except ORF7b antibodies, for at least 6 months post-infection. Many factors play a role in antibody long-term persistence, such as antigen release, antigen presentation, induction of a germinal center reaction, and a memory B cell pool 55 . Additional studies on viral proteins release, their roles, and their specific B cells are needed to fully understand the pattern of antibody specificity in children. To complement serological data and to understand the link between the inflammatory response and the distinct pattern of antibody specificity observed in children, circulating cytokines were measured in samples from acute time-points (