Immune cell landscape in symptomatic and asymptomatic SARS-CoV-2 infected adults and children in urban Dhaka, Bangladesh

Objectives The study of cellular immunity to SARS-CoV-2 is crucial for evaluating the course of the COVID-19 disease and for improving vaccine development. We aimed to assess the phenotypic landscape of circulating lymphocytes and mononuclear cells in adults and children who were seropositive to SARS-CoV-2 in the past 6 months. Methods Blood samples (n = 350) were collected in a cross-sectional study in Dhaka, Bangladesh (Oct 2020-Feb 2021). Plasma antibody responses to SARS-CoV-2 were determined by an electrochemiluminescence immunoassay while lymphocyte and monocyte responses were assessed using flow cytometry including dimensionality reduction and clustering algorithms. Results SARS-CoV-2 seropositivity was observed in 52% of adults (18–65 years) and 56% of children (10–17 years). Seropositivity was associated with reduced CD3+T cells in both adults (beta(β) = −2.86; 95% Confidence Interval (CI) = −5.98, 0.27) and children (β = −8.78; 95% CI = −13.8, −3.78). The frequencies of T helper effector (CD4+TEFF) and effector memory cells (CD4+TEM) were increased in seropositive compared to seronegative children. In adults, seropositivity was associated with an elevated proportion of cytotoxic T central memory cells (CD8+TCM). Overall, diverse manifestations of immune cell dysregulations were more prominent in seropositive children compared to adults, who previously had COVID-like symptoms. These changes involved reduced frequencies of CD4+TEFF cells and CD163+CD64+ classical monocytes, but increased levels of intermediate or non-classical monocytes, as well as CD8+TEM cells in symptomatic children. Conclusion Seropositive individuals in convalescence showed increased central and effector memory T cell phenotypes and pro-resolving/healing monocyte phenotypes compared to seronegative subjects. However, seropositive children with a previous history of COVID-like symptoms, displayed an ongoing innate inflammatory trait.

Previous work has shown that memory B cell responses tend to be short-lived after infection with SARS-CoV (Channappanavar et al., 2014;Tang et al., 2011). In contrast, memory T cell responses can persist for many years (Le Bert et al., 2020;Tang et al., 2011;Yang et al., 2006) and, protect against lethal challenge with SARS-CoV in mice (Channappanavar et al., 2014). Early in the pandemic, studies have shown the kinetics and extent of the immune response in patients with mild to severe COVID-19 (Carsetti et al., 2020;Parrot et al., 2020). It is now clear that inflammation is important in disease progression and pathogenesis of COVID-19, including excess activation of inflammatory cells as well as production of soluble pro-inflammatory mediators (Kvedaraite et al., 2021;Lourda et al., 2021;Openshaw, 2022). In addition, increased activation of memory CD4 + and CD8 + T cells, T cell cytopenia, elevated frequencies of plasmablasts with oligoclonal B cell expansions and enhanced neutrophil responses have been observed in severely affected patients compared to individuals with mild or moderate infection or healthy controls, although the T cell response can be heterogenous with large inter-individual variations (Kuri-Cervantes et al., 2020;Laing et al., 2020;Mathew et al., 2020). With disease progression, CD8 + T and NK cells decrease in numbers, while their cytotoxic activity involving expression of effector molecules such as perforin and granzymes, increases to compensate for this reduction Maucourant et al., 2020;Sekine et al., 2020;Wang et al., 2020).
In the initial phase of the pandemic, infection with SARS-CoV-2 appeared to be less common in children than in adults (Castagnoli et al., 2020;Irfan et al., 2021). Hospital data from across the United States between March 2020 and August 2021, suggests that <2% of hospitalizations due to COVID-19 occurred in children <18 years (CDC, 2020;Delahoy et al., 2021). Moreover, as of July 2022, only about 18% of all COVID-19 cases have been diagnosed among children (<18 years) according to the American Academy of Pediatrics (AAP, 2022). Case Reports indicate that approximately 16% of pediatric COVID-19 infections are asymptomatic (Assaker et al., 2020;Lu and Zhang, 2020), but this may change with the community spread of new variants including delta and omicron, especially among unvaccinated children. The general notion is that a rapid response to the virus by the highly efficient innate immune system in children is the key player, although the actual mechanism of protection remains elusive (Brodin, 2022;Loske et al., 2021;;Mallapaty, 2021).
During the COVID-19 pandemic, several studies on phenotypic and functional immune cell profiling in patients with mild and severe disease have been reported (Kvedaraite et al., 2020;Lourda et al., 2021;Maucourant et al., 2020;Sandberg et al., 2021). However, information on immunophenotyping is limited among seropositive individuals in the largely untested community who got infected but remained asymptomatic or developed mild symptoms and recovered quickly. It is also important to understand the age-related differences in immune responses generated in adults and children. In a cross-sectional study, we evaluated the immune cell landscape including B and T lymphocyte subsets as well as different monocyte subsets, in SARS-CoV-2 seropositive subjects with mild or no symptoms as well as seronegative adults and children from urban communities in Dhaka, Bangladesh. These data will provide important insights into mild SARS-CoV-2 infection in different age groups, and further advance the current knowledge on the immune cell profile in pediatric groups of SARS-CoV-2 infected individuals.

Study settings and subjects
The present study is part of a larger community based cross-sectional serosurvey, where around 3200 community subjects were enrolled from large slums and surrounding non-slum areas of Dhaka and Chattogram cities in Bangladesh during October 2020 to February 2021 (Raqib et al., 2022). During the study was being conducted, Wuhan type variant of SARS-CoV-2 was assumed to be mostly prevalent; with alpha variant being first detected in December 2020 and beta variant in January 2021 (Hossain et al., 2021;Saha et al., 2021). In this study, we targeted 350 participants residing in slums and surrounding non-slum areas of Korail, Mirpur and Dhalpur in Dhaka city. Slums were included to represent primarily impoverished people living in densely packed housing units in unhealthy conditions in urban districts. The non-slum area was selected to include middle-class households that were adjacent to these slums. For selecting study samples in the main cohort, a cluster random sampling procedure was followed as described earlier (Raqib et al., 2022).
Study inclusion criteria included participants of either sex 10-17 years (children) or 18-60 years (adults), willing to provide informed consent/ assent and donate blood. Written informed consent was obtained from adult participants. The children agreed to participate in the study by giving their assent, while their parents provided signed informed consents allowing the children to be enrolled in the study.
Height and weight were measured using the free-standing stadiometer (Seca 217, Hamburg, Germany) and digital weighing scale (Camry-EB9063, China), to calculate body mass index (BMI). Data was collected through a structured questionnaire that included sociodemographic features, ongoing COVID-like symptoms or symptoms present during the past 6 months, presence of confirmed COVID-19 cases in the household and seeking healthcare in the past 6 months (Raqib et al., 2022).

Sample collection and processing
Venous blood samples (7.5 mL) were collected from study participants in Li-heparin coated tubes (S-Monovette Plasma, Sarstedt AG & Co. KG, Nümbrecht, Germany). Plasma and peripheral blood mononuclear cells (PBMCs) were separated by Ficoll-Paque density-gradient centrifugation (GE Healthcare Bio-Sciences AB SE-751 84 Uppsala, Sweden). Isolated PBMCs were cryopreserved in freezing media (10% dimethylsulfoxide in fetal bovine serum (FBS) and stored in liquid nitrogen until analysis. The estimated median time from suspected symptomatic infection to blood sample collection was approximately 6 weeks for adults and ~4 weeks for children.

Hemogram
A complete blood count (CBC) for all participants was performed by using an automated hematology analyzer (XN-1000, Sysmex Corporation, Kobe, Japan).

Flow cytometry of PBMC
On the day of flow cytometry data acquisition, cryopreserved PBMCs were thawed in a water bath at 37℃ and washed twice with cell culture media containing 10% FBS, 1% sodium pyruvate and 1% L-glutamine in RPMI. After counting, PBMCs from each participant were resuspended in FACS staining buffer (Becton Dickinson, Dorset, UK) and divided into three separate 5 mL round bottom polystyrene tubes (Falcon, Corning Science, Mexico) to be analyzed for (i) T cells, (ii) B cells and (iii) monocytes and NK cells (0.5x10 6 cells/100 µl/tube). Cells were stained with a cocktail of monoclonal antibodies conjugated to different fluorochromes for specific cell surface markers (Supplementary Table S1) for 20 min at 4℃ in the dark. Samples were washed twice with staining buffer and after the final wash, cells were resuspended in 350 µl of the buffer. At least 500,000 events per sample were acquired on a BD FACSCanto™ (BD Biosciences, San Jose, USA). All the fluorochromeconjugated monoclonal antibodies were purchased from BD Biosciences.
Acquired data were analyzed using either FACS DIVA software (Tree Star, Inc., Ashland, OR, USA) or FlowJo v.10.8 (BD Biosciences). The different cell populations were identified based on forward-and sidescatter characteristics and cell-specific surface receptors. At least 50,000 lymphocyte-gated cells were analyzed for T and B cells while 100,000 cells were analyzed for NK and monocyte population assessment.

Dimensionality reduction and clustering
For deep assessment of immune cell clusters with Uniform Manifold Approximation and Projection (UMAP v3.1), sixty (30 adults and 30 children) age-and gender-matched study subjects were randomly selected and divided into six sub-groups with 10 participants per group: symptomatic-seropositive, asymptomatic-seropositive, and seronegative adults; and symptomatic-seropositive, asymptomatic-seropositive, and seronegative children. To perform unsupervised analyses, and to define different subpopulations of mononuclear cells, a minimum number of each cell types are required. For monocytes, 2500 cells (CD3 -HLADR + CD14 +/-CD16 +/-) from each participant were down sampled using Down sample plugin (v3.3.1) in FlowJo. For a total of 60 participants, all the down sampled files were concatenated into a single FCS file. Likewise, 5000 CD3 + CD4 + T cells and 4000 CD3 + CD8 + T cells from each of the 60 participants were down sampled and concatenated into single files to avoid data bias.
Thereafter, UMAP algorithm was applied on these concatenated files for dimensionality reduction (Becht et al., 2018) and FlowSOM tool (v2.4) was used for cluster analysis (Quintelier et al., 2021) of these three immune cell populations. For monocytes, the concatenated file was also analyzed by manual gating for different subpopulations and was overlaid onto the UMAP space.

Statistical analysis
A SARS-CoV-2 serosurvey study carried out in Bangladesh including 3220 study subjects showed a seroprevalence of 67.3% (Raqib et al., 2022). The calculated sample size for the present study was approximately 340 (increased to 350) participants based on a seroprevalence of 67.3% with a precision of ±0.05% at 95% confidence interval. Stored PBMC was available from Dhaka city only, and therefore the 350 samples were randomly collected from the Dhaka study site which included 2614 participants. Independent pseudo-random technique was applied to generate a random sample selection of the estimated samples size of 13.4% (350 out of 2614 subjects), by using SPSS-22 for windows (SPSS Inc. Chicago, IL). To find exactly 350 random samples, the process was repeated 57 times. If the selected sample had inadequate PBMCs, the adjacent sample was selected instead.
We performed a descriptive analysis summarizing each participant's demographics and others features with absolute frequencies if categorical and with mean and standard deviation if continuous. Immune cells were tested for normality using Shapiro-Wilk test and Skewness and Kurtosis test, and when normality assumption failed, data were transformed (log2 transformation) as indicated in the footnotes of the tables. To control for multiple testing, we applied the most common family wise error rate (FWER) by Bonferroni method, when the FWER = 0.05, using stepwise procedure. A stratified analysis of seropositive vs seronegative, and symptomatic vs asymptomatic participants among the seropositive group was performed. Multivariate generalized linear regression model was used to assess changes in immune cells in the seropositive compared to seronegative, and symptomatic seropositive compared to asymptomatic seropositive participants. The regression model was adjusted with sex, BMI, monthly family income and area of residence (slum and non-slum areas). Data from the unsupervised analyses was analyzed using Kruskal-Wallis and Dunn's multiple comparisons test to determine the statistical differences between six unpaired groups. Spearmanś correlation test was used for the correlation analyses. The significance level was established at P < 0.05. The statistical analyses were performed with Stata 15 (StataCorp, LP, College Station, Texas, USA) and GraphPad Prism Software (v9.0).

Demography and hemogram of enrolled study subjects
The demography of enrolled subjects is shown in Supplementary  Table S2. In this study, n = 350 participants were enrolled including 56% females and 64% adults. Overall, 54% of the participants were seropositive for SARS-CoV-2, while in adults and children the rate of seropositivity was 52% and 56%, respectively. Among the seropositive participants, about 34% of adults (n = 44) and 41% of children (n = 38) reported that they experienced COVID-like symptoms in the past 6 months. Among the seronegative participants (n = 162), COVID-like symptoms were present in 46 individuals (28%). A total of 21 participants (6%) were tested for SARS-CoV-2 RT-PCR, resulting in 8 PCRconfirmed COVID-19 cases. Among seronegative participants, seven individuals underwent RT-PCR testing and all tested negative. About 79% of adults and 70% of children were vaccinated with the Bacillus Calmette-Guérin (BCG) vaccine during childhood.
The complete blood count (CBC) data of the participants showed normal ranges of total white blood cell (WBC) counts, lymphocytes, monocytes, eosinophils and platelet count in both adults and children (Supplementary Table S3). The various cell count ranges in Bangladeshi population were compatible with those seen in other Asian populations (Sairam et al., 2014). Seropositive adults showed reduced counts of neutrophils (P = 0.031) compared to seronegative adults (Table 1). No significant difference was noted in seropositive compared to seronegative children. Among seropositive adults, symptomatic individuals exhibited no significant difference in the CBC data compared to asymptomatic individuals. Among seropositive children, eosinophil counts were lower (P = 0.002) in symptomatic compared to asymptomatic individuals.
Next, three different monocyte subsets were assessed, classical, nonclassical and intermediate monocytes (Supplementary Table S4 & Fig. 1). Classical monocytes are mostly phagocytic and less inflammatory in nature with a tissue-repair function as compared to non-classical monocytes that are more inflammatory and the primary producers of the inflammatory cytokines IL-1β and TNF-α (Mukherjee et al., 2015). Intermediate monocytes usually constitute a smaller proportion of monocytes and could display both pro-and anti-inflammatory properties (Mukherjee et al., 2015). Seropositive adults displayed increased frequency of classical monocytes as compared to seronegative subjects, however the difference was not statistically significant (P = 0.144). Moreover, the proportion of non-classical monocytes was significantly decreased in seropositive adults (P < 0.001) as well as children (P = 0.008).

Alterations of immune cell subsets in SARS-CoV-2 seropositive subjects with previous symptomatic compared to asymptomatic infection
Differences in immune cell composition in seropositive subjects with symptomatic or asymptomatic suspected SARS-CoV-2 infection were assessed as summarized in Table 2. Relatively few immune cell alterations were observed in adult seropositive subjects with a previously symptomatic infection compared to asymptomatic infection. These changes represented increased levels of CD19 + B cells (P = 0.036), and reduced levels of memory B cells (CD19 + CD27 + ) (P = 0.018) and CD4 + T CM (P = 0.030) in symptomatic subjects, while no other differences in the proportion of lymphocyte subsets were significant.
In contrast to adults, symptomatic infection in seropositive children was associated with low CD19 + B cells (P = 0.021) including low early B cells frequencies (P = 0.094) but higher Breg cells proportions (P = 0.09) Furthermore, significantly higher proportions of naïve CD4 + and CD8 + T cells but lower CD4 + CD25 + T reg , CD4 + T CM and CD8 + T CM were detected in symptomatic compared to asymptomatic infection in children. The proportions of NK cells and NKT cells were also reduced (P = 0.071 and P = 0.035, respectively) in symptomatic children.
Interestingly, the proportion of classical monocytes was decreased (P = 0.002) in symptomatic compared to asymptomatic children, while intermediate monocytes were elevated (P < 0.001) in symptomatic subjects.

Deep characterization of CD4 + and CD8 + T cell phenotypes in the different study subgroups
Next, we sought to investigate if the differences observed in the proportions of T cells and monocytes were also associated to specific phenotypic changes. For these analyses, 10 subjects per group were randomly selected as described in the Materials and Methods and Fig. 2 depicts the hierarchical clustering of immune cells in dendrograms. CD3 + T cells were manually gated to identify distinct subpopulations of CD4 + and CD8 + T cells (Fig. 3A). Expression of CD45RA, CCR7, CD45RO, CD127 and CD25 on the cell surface was used for UMAP projection on CD4 + T cells ( Fig. 3B-C). Deeper analysis using the FlowSOM clustering algorithm, generated eight meta-clusters of CD4 + T cells that were used to identify naïve, memory and effector T cell phenotypes ( Fig. 3D and Fig. 2A) and their distribution in the adult and child study groups (Fig. 3E). Adult subjects exhibited few alterations in immune cells as noted in bulk analysis. A significant decrease in CD4 + T EFF cells was observed in symptomatic (P = 0.02) compared to asymptomatic seropositive children, while no difference in CD4 + T EFF cells was found in the corresponding adult groups (Fig. 3F). CD4 + CD25 + CD127 -Treg cells identified as cluster-6, were expressed at very low levels in all study groups (Fig. 3D-E and Fig. 2A).
A similar approach was taken for UMAP projection and FlowSOM cluster analyses of CD8 + T cells to visualize the dynamics of different phenotypes and subsets comparing the six study sub-groups (see Methods) ( Fig. 4A-D and Fig. 2B). In contrast to CD4 + T EFF cells, CD8 + T EM cells were significantly higher in symptomatic seropositive compared to seronegative children (P = 0.042) (Fig. 4E). Overall, both memory and effector T cells were generally higher in adults as opposed to in the frequencies observed in children (Supplementary Table S5). However, the main differences in T cell frequencies comparing symptomatic and asymptomatic seropositive individuals were observed in the pediatric group only.

Deep characterization of monocyte phenotypes in the different study subgroups
Next, different subsets of classical (c-mono), intermediate (i-mono) and non-classical (nc-mono) monocytes were defined with manual gating of HLA-DR + monocytes followed by identification of CD14 + or CD16 + subsets (Fig. 5A). Dimensionality reduction using UMAP and overlay of the manually gated monocyte populations confirmed the accuracy of the manual gating (Fig. 5B), based on the surface expression of CD14, CD16, HLA-DR, CD64 and CD163 (Fig. 5C-D). The majority (87.1%) of the total monocytes were identified as classical monocytes, followed by non-classical (7.7%) and intermediate (5.1%) monocytes (Fig. 5B). Classical monocytes expressed higher CD14, CD64 and CD163 and relatively lower HLA-DR compared with the non-classical and intermediate monocytes (Fig. 5C-D). FlowSOM clustering analysis generated ten meta-clusters, half of which displayed a classical monocyte phenotype, while the other clusters comprised three non-classical, and two intermediate monocytes phenotypes (Fig. 5E and Fig. C) that were differentially distributed in the study groups (Fig. 5F).
Consistent with the findings from the manually gated flow cytometry data (Table 2), unsupervised analyses revealed that classical monocytes were expressed at significantly lower levels in symptomatic compared to Multivariate generalized linear regression model was used to estimate the p-value and the regression model was adjusted by sex, family income, locality (slum and nonslum) and BMI. asymptomatic seropositive children (P = 0.009) (Fig. 5G). Moreover, symptomatic children showed a reduced frequency of CD163 + CD64 + classical monocyte clusters (P = 0.004) compared to asymptomatic seropositive children (Fig. 5H). Cluster-10, which expressed the highest levels of both CD163 and CD64, and also displayed significantly decreased MFI of CD64 in symptomatic (P = 0.046) compared to asymptomatic children ( Fig. 5I and Fig. 2C). In contrast to classical monocytes, the frequency of non-classical monocytes was higher in symptomatic seropositive children compared to asymptomatic ones (P = 0.016) (Fig. 5G). However, no difference in HLA-DR expression was observed between these groups (data not shown), which indicated that a high CD64 expression did not contribute to enhanced pro-inflammatory activity of these cells (Mily et al., 2020). These results support the findings of the manual gating suggesting that symptomatic children have lower anti-inflammatory monocyte responses compared to asymptomatic children. In adults, no differences in the monocyte subpopulations were observed in symptomatic compared to asymptomatic individuals (Fig. 5G).

Associations of specific monocyte and T cell subsets in SARS-CoV-2 seropositive children
The unsupervised UMAP cluster analyses demonstrated significant differences in the frequencies of classical and non-classical monocytes as well as differences in the frequencies of CD4 + T EFF and CD8 + T EM cells, primarily comparing the children study groups. Accordingly, we observed a significant positive correlation (r = 0.442, P = 0.05) between classical monocytes and CD4 + T EFF (Fig. 6A) but an inverse correlation (r = − 0.479, P = 0.032) with non-classical monocytes and CD4 + T EFF (Fig. 6B) in SARS-CoV-2 seropositive children. Likewise, there was a significant positive correlation (r = 0.542, P = 0.013) between CD163 + CD64 + classical monocytes and CD4 + T EFF (Fig. 6C), but an inverse correlation (r = − 0.487, P = 0.029) between the CD163 + CD64 + subset and CD8 + T EM cells (Fig. 6D) in seropositive children. A negative association was also found between CD4 + T EFF and CD8 + T EM cells (r = 0.422, P = 0.06) (data not shown). Moreover, classical monocyte cluster 10, demonstrated a positive correlation (r = 0.522, P = 0.018) with CD4 + T EFF cells, and a corresponding inverse correlation (r = − 0.450, P = 0.046) with CD8 + T EM cells (data not shown). No associations between CD4 + T EFF or CD8 + T EM cells and the monocyte subsets were found among the healthy seronegative controls (data not shown). These correlation analyses suggest a potential association between reduced numbers of CD4 + T EFF cells and classical (eg. anti-inflammatory) monocytes with elevated CD8 + T EM cells and non-classical monocytes, in seropositive children who developed COVID-like symptoms.

Discussion
In this study, we investigated phenotypes of circulating immune cell subsets among adults and children who were suspected to be infected with SARS-CoV-2 in the previous 6 months as compared to uninfected seronegative individuals living in the same community during the early phase of the COVID-19 pandemic in Bangladesh. The prevalence of SARS-CoV-2 seropositivity was similar in adults and children; seropositivity was associated with reduced proportion of pan T (CD3 + ) cells across both age groups. Bulk flow cytometry analyses demonstrated increased frequencies of CD8 + T CM cells in seropositive adults compared to seronegative subjects. However, seropositive children showed reduced proportion of Breg, Treg and increased proportion of CD4 + T EFF and CD4 + T EM as compared to seronegative subjects. Deep characterization of flow cytometry data revealed that, children displayed the most prominent differences in immune cell frequencies, represented by reduced CD4 + T EFF and enhanced CD8 + T EM , as well as reduced classical monocytes in symptomatic compared to asymptomatic subjects. Bulk flow cytometry analyses suggested elevated frequencies of intermediate monocytes while UMAP analyses indicated elevated levels of nonclassical monocytes in symptomatic children. Furthermore, correlation analyses demonstrated an association between low frequencies of CD4 + T EFF and increased frequencies of non-classical monocytes, as well as an association between low frequencies of CD163 + CD64 + classical monocytes and high frequencies of CD8 + T EM cells in seropositive children. Overall, seropositive adults displayed elevated pro-resolving/ healing monocyte responses compared to seronegative subjects while seropositive children with a history of COVID-like symptoms demonstrated an ongoing innate inflammatory trait.
T cells play a vital role in the defense against viral infections. CD8 + T cells identify and kill virus-infected cells, while CD4 + T cells assist them in eliminating the viral infection, which is followed by generation of immunological memory. However, excessive response to viral infection can cause immunopathology and is detrimental to the host (Abbas, 2020;Saad and Moussa, 2021). Pan-lymphopenia has been widely described in SARS infections, particularly in severe COVID-19, linked to large number of peripheral T cells being recruited to the site of inflammation, functional exhaustion and apoptosis of T cells and activation-induced cell death (Mishra et al., 2020;Radzikowska et al., 2020;Saad and Moussa, 2021). In this study, an overall decrease of total T cell counts but increased frequencies of CD4 + T EM and CD8 + T CM cells in seropositive participants suggest that immunological T cell memory was generated, even though overall T cell numbers had not yet normalized after the infection. Our finding of elevated CD4 + T EFF and T EM cells in seropositive children reflect ongoing inflammation and chemotaxis to the site of inflammation and a capacity to mount rapid response to SARS-CoV-2 infection, whereas increased CD8 + T CM cells in adults may indicate an enhanced capacity to proliferate and create new rounds of effector T cells for protection against future SARS-CoV-2  Gating strategy for CD3 + CD4 + and CD3 + CD8 + T cell identification. (B) UMAP projection on the concatenated file of total CD4 + T cells in symptomatic-seropositive, asymptomatic-seropositive, and seronegative adults (total n = 30) and children (total n = 30) distributed among six study groups. (C) Median fluorescence intensity (MFI) of selected T cell surface markers in the combined UMAP space described in (B). (D) Distribution of eight defined meta-clusters generated by employing FlowSOM algorithm overlaid onto the combined UMAP space for a total of 60 study subjects. (E) FlowSOM meta-clusters overlaid onto the UMAP space for the six different study groups. (F) Dot plots showing the frequency distribution of CD4 + T cell subpopulations in each of the six study groups (n = 10 subjects/group) comparing symptomatic-seropositive subjects (filled circles), asymptomatic-seropositive subjects (open circles) and seronegative subjects (open triangles). Kruskal-Wallis and Dunn's multiple comparisons test was used to determine the statistical difference comparing the unpaired groups of study subjects within each age group (adults or children). Solid bars denote median values in the scatter plots for the corresponding study groups. *P < 0.05; **P < 0.01; ***P < 0.001.
Clinical studies in patients with severe COVID-19 reported both increased and decreased frequencies of central and effector memory T cells (Laing et al., 2020;Shahbaz et al., 2021;Zenarruzabeitia et al., 2021). In a cohort of clinically recovered patients, the frequencies of CD8 + and CD4 + T CM cells as well as the functional capacity of T cells (expression of IFN-γ, IL-2, IL-4, IL-17) were reduced compared to healthy controls showing functional repression (Yang et al., 2021). In agreement with these findings, we observed that seropositive individuals with prior COVID-like symptoms had significantly lower proportions of T CM cells, T reg cells, as well as NKT cells compared with asymptomatic seropositive participants. One explanation for decreased T cell counts could be that viral antigens might persist longer (up to 5.7 months) in symptomatic SARS-CoV-2-infected individuals, who usually have higher level of viral replication (Gaebler et al., 2021). Another potential explanation could be that persistence of inflammation may slow down the restoration of immunological balance. Remarkably, many of the basic T cell subsets analyzed in this study were reduced in seropositive participants having COVID-like symptoms in the previous 6 months and the wide ranging of T cell dysregulation manifested more in children than in adults. Another explanation could be age-related difference in thymic production of T cells. The thymus is typically able to replace the T cells destroyed by virus-mediated apoptosis (Gunes et al.,Fig. 4. Deep characterization of CD8 + T cell phenotypes in seropositive and seronegative adults and children. (A) UMAP projection on the concatenated file of total CD8 + T cells in symptomatic-seropositive, asymptomatic-seropositive, and seronegative adults (total n = 30) and children (total n = 30) distributed among six study groups. (B) Median fluorescence intensity (MFI) of selected T cell surface markers in the combined UMAP space described in (A). (C) Distribution of eight defined meta-clusters generated by employing FlowSOM algorithm overlaid onto the combined UMAP space for a total of 60 study subjects. (D) FlowSOM meta-clusters overlaid onto the UMAP space for the six different study groups. (E) Dot plots showing the frequency distribution of CD8 + T cell subpopulations in each of the six study groups (n = 10 subjects/group) comparing symptomatic-seropositive subjects (filled circles), asymptomatic-seropositive subjects (open circles) and seronegative subjects (open triangles). Kruskal-Wallis and Dunn's multiple comparisons test was used to determine the statistical difference comparing the unpaired groups of study subjects within each age group (adults or children). Solid bars denote median values in the scatter plots for the corresponding study groups. *P < 0.05; **P < 0.01. cluster-10 in the six study groups. Kruskal-Wallis and Dunn's multiple comparisons test was used to determine the statistical difference comparing the unpaired groups of study subjects within each age group (adults or children). Solid bars denote median values in the scatter plots for the corresponding study. *P < 0.05; **P < 0.01. 2021; Wang et al., 2021). In individuals with previous symptomatic infection, high naïve T cell output in children was likely to be derived from the thymus, however, the recovery of naïve T-cell populations in adults was slow, which could be attributed to thymic involution. Exhaustion of central memory T cells and downregulation of T reg cells among symptomatic children were suggestive of sub-optimal control on inflammation perhaps due to the lingering effects of the infection, and a requirement for further generation of memory T cells (Ferreras et al., 2021;Kratzer et al., 2021).
When exposed to antigens, naïve B cells differentiate into antibodyproducing plasma cells and memory B cells, with memory responses being important for protection from future infections. Patients who recovered from COVID-19 also showed reduced CD19 expression, unswitched memory B cells, impaired BCR signaling, and undetectable memory B cells in 30% of the recovered patients (Achiron et al., 2021;Jing et al., 2021). Here we show that compared to asymptomatic individuals, subjects with a history of COVID-like symptoms showed reduced frequency of CD19 + B cells, and memory B cells, although frequency of antibody secreting plasma cells remained unchanged. CD4 + T cells stimulate B cells towards differentiation into plasma cells and memory B cells. It is possible that inadequate pool of CD4 + T cells may contribute poorly towards the transformation into memory B cells (Kaneko et al., 2020) (without altering plasma cell numbers).
Although different monocytes subsets exhibit substantial heterogeneity and plasticity (Kapellos et al., 2019;Merah-Mourah et al., 2020), classical monocytes are mostly non-inflammatory with a tissue-repair function (Wong et al., 2011), while intermediate monocytes and primarily non-classical monocytes, have been found to exhibit more proinflammatory properties and expand in patients with different inflammatory conditions (Ong et al., 2018). In this study, we found that a subset of CD163 + CD64 + classical monocytes was elevated in asymptomatic compared to symptomatic children. As CD163 is a scavenger receptor that usually defines monocytes and macrophages with anti-inflammatory properties (Kowalska, 2020;Moestrup and Moller, 2004), these results may indicate that this monocyte subset has a protective function in asymptomatic children while being selectively downregulated in symptomatic children. This notion is strengthened by the finding that, CD4 + T EFF cells, but not CD8 + T EM cells, are positively associated with classical monocytes as well as CD163 + CD64 + monocytes, which could imply that CD163 + CD64 + monocytes promote CD4 + T cell responses but reduce overt inflammation in seropositive children.
In patients with either sepsis or systemic lupus erythematosus, a significant increment of both intermediate and non-classical subsets were found, while classical monocytes were down-regulated (Mukherjee et al., 2015). The inflammatory trait of non-classical monocytes has been proposed to be due to a senescence-associated secretory phenotype that may be induced by a high basal NF-κB activity and IL-1α production, while their accumulation is also associated with higher levels of plasma TNF-α and IL-8 (Mukherjee et al., 2015;Ong et al., 2018). Non-classical monocytes have also been shown to promote neutrophil adhesion at the endothelial interface via the secretion of TNF-α (Chimen et al., 2017). However, while both bulk flow cytometry and UMAP analyses suggested a trend for elevated frequencies of intermediate monocytes in seropositive compared to seronegative children, there was a discrepancy regarding the frequency of non-classical monocytes. Bulk analysis showed no changes in the levels of non-classical monocytes in symptomatic children, while UMAP analysis suggested significantly elevated levels of non-classical monocytes in symptomatic compared to asymptomatic seropositive children. A limitation of the analyses may be that a smaller sub-sample of study subjects were used for UMAP clustering and/or the manual gating in the bulk analyses was slightly different  compared to the unsupervised analyses. Various studies have shown that SARS-CoV-2 infection promotes profound changes in the innate immune compartment including neutrophilia, depletion of eosinophils, NK cells and basophils, and these changes are related to disease severity (Lourda et al., 2021;Maucourant et al., 2020). Longitudinal studies suggest that neutrophil counts are high in acute stage that slowly decline during recovery, while the numbers of eosinophils and basophils are typically low (Lourda et al., 2021;Radzikowska et al., 2020). Consistent with this, children who recovered from symptomatic infection exhibited higher percentage of inflammatory phenotypes and lower anti-inflammatory monocytes, which suggests ongoing inflammation (Neeland et al., 2021;Tserel et al., 2021). Seropositive adults showed a contrasting picture with down-regulation of pro-inflammatory monocytes and neutrophils but up-regulation of anti-inflammatory monocyte phenotypes (Tables 1, Supplementary Table S4); the findings were in accordance with the reports by others in convalescent phase of SARS-CoV-2 infection (Balta et al., 2021;Carsetti et al., 2020;Zhou and Ye, 2021). Together these findings indicate that symptomatic infection triggered dynamic changes in the innate immune system that continued for longer period in children (Schultze and Aschenbrenner, 2021).
There are several theories about the differences in severity of SARS-CoV-2 infection in children compared to adults (Mallapaty, 2021;Zimmermann and Curtis, 2020). Children with COVID-19 have been found to have less potent T cells, lower frequencies of monocytes, including inflammatory monocytes (Cohen et al., 2021), low neutralizing antibodies and lower Treg cells (Pierce et al., 2021) than adults. Infected children have higher levels of IFN-γ and IL-17A reflecting strong innate and trained immune response from childhood vaccination (BCG, OPV, MMR vaccines) (Zimmermann and Curtis, 2020). Age-related decline (immunosenescence) and dysregulation of immune function (inflammaging), are considered to be partly responsible for heightened vulnerability to severe COVID-19 outcomes in older adults (Yang et al., 2021). In agreement with these studies, we also found differences in T and B lymphocyte subtypes and innate cells, which were suggestive of insufficient adaptive immunity and augmented innate immune response in children compared to adults.
The major strength of this study is the sample size and the comparison of immune responses between adults and children in a study cohort enrolled in a low-income setting. Limitations of this study include lack of knowledge on exact time lapse between infection and sample collection. There may be recall bias of asymptomatic carriers since RT-PCR based confirmation of the infection dates were not available for most individuals. Another potential issue might have been the misidentification of some participants as seronegative subjects, while they might have been previously infected, but due to rapid decline in COVID IgG antibodies, may no longer remained seropositive. Quantitative determination of antibody concentrations would have shown differences in antibody levels between recently infected and those infected several months back. However, a great majority of studies have shown that even though antibody responses after SARS-CoV-2 infection decline, the antibodies persist from >6 to 11 months at detectable levels (Breuer et al., 2021;L'Huillier et al., 2021;Turner et al., 2021). Another limitation was that additional markers in our flow cytometry panels would have provided a better resolution and determination of the different subpopulations found among the basic immune cell subsets. The memory cells determined in this study were generic, and not specifically targeted to SARS-CoV-2 antigens. In addition, in this study we were not able to assess and compare the functional capacity of T and B cells, and monocytes with the proportion of these cells determined in the different sub-groups studied.
In conclusion, previously SARS-CoV-2 exposure leaves imprints in the cellular immune system that includes central and effector memory T and B cells as well as innate cells after recovery. Symptomatic infection results in persistence of the inflammatory responses with elevated neutrophils, inflammatory monocytes, and reduced NK, T reg and memory T cells and could be indicative of heightened innate immune surveillance in children. The findings are intriguing since it appears that even after mild symptomatic or asymptomatic infection with SARS-CoV-2, the individuals experience persistently activated immune surveillance accompanied by a slow return to immune balance. Further investigations are warranted to unravel the mechanisms underlying loss of B and T lymphocytes in COVID-19 patients and retainment of long-term immunological memory in follow-up studies.

Funding
This work was funded by The Foreign, Commonwealth & Development Office (FCDO) through The United Nations Population Fund (UNFPA), and Global Affairs Canada.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.