key: cord-0329328-u398lzcl
authors: Menges, D.; Zens, K. D.; Ballouz, T.; Caduff, N.; Llanas-Cornejo, D.; Aschmann, H. E.; Domenghino, A.; Pellaton, C.; Perreau, M.; Fenwick, C.; Pantaleo, G.; Kahlert, C. R.; Muenz, C.; Puhan, M. A.; Fehr, J. S.
title: Heterogenous Cellular and Humoral Immune Trajectories after SARS-CoV-2 Infection: Compensatory Responses in a Population-Based Cohort
date: 2021-12-16
journal: nan
DOI: 10.1101/2021.12.15.21267776
sha: 9649542e5bb077e85d492da925cd8bf6112e6153
doc_id: 329328
cord_uid: u398lzcl

To better understand the development of immunity against SARS-CoV-2 over time, we evaluated humoral and cellular responses a population-based cohort of SARS-CoV-2-infected individuals covering the full spectrum of COVID-19 up to 217 days after diagnosis. We characterized anti-Spike (S)-IgA and -IgG antibody responses in 431 individuals and found that about 85% develop and maintain anti-S-IgG responses over time. In a subsample of 64 participants selected for a detailed characterization of immune responses, we additionally evaluated anti-Nucleocapsid (N)-IgG antibodies and T cell responses specific to viral Membrane (M), N, and S proteins. Most participants had detectable T cell responses to at least one of the four peptide pools analyzed, which were more frequent than antibody seropositivity. We found a moderate correlation between antibody and T cell responses, which declined over time and suggests important variability in response patterns between individuals. The heterogeneity of immune trajectories was further analyzed using cluster analyses taking into account joint antibody and T cell responses over time. We identified five distinct immune trajectory patterns, which were characterized by specific antibody, T cell and T cell subset patterns along with disease severity and demographic factors. Higher age, male sex, higher disease severity and being a non-smoker was significantly associated with stronger immune responses. Overall, the results highlight that there is a consistent and maintained antibody response among most SARS-CoV-2-infected individuals, while T cell responses appear to be more heterogenous but potentially compensatory among those with low antibody responses.

four peptide pools analyzed, which were more frequent than antibody seropositivity. We found a 34 moderate correlation between antibody and T cell responses, which declined over time and 35 suggests important variability in response patterns between individuals. The heterogeneity of 36 immune trajectories was further analyzed using cluster analyses taking into account joint 37 antibody and T cell responses over time. We identified five distinct immune trajectory patterns, 38 which were characterized by specific antibody, T cell and T cell subset patterns along with 39 disease severity and demographic factors. Higher age, male sex, higher disease severity and 40 being a non-smoker was significantly associated with stronger immune responses. Overall, the 41 results highlight that there is a consistent and maintained antibody response among most SARS-42

CoV-2-infected individuals, while T cell responses appear to be more heterogenous but 43 potentially compensatory among those with low antibody responses. 44

Almost two years after its start, the SARS-CoV-2 pandemic remains a threat to public 52 health worldwide and has resulted in hundreds of million cases and millions of deaths globally 53

(1). Control of the pandemic now relies largely on the development of robust immunity in the 54 population after infection, vaccination, or both, which necessitates an in-depth understanding of 55 humoral and cellular responses to the virus. 56

Several studies have characterized B and T cell-mediated immune responses against 57 SARS-CoV-2 and showed that both antibodies and T cells are generated in most people after 58 infection (2-19). Immunoglobulin (Ig) spike specific antibodies have been detected within a few 59 days after infection (4, 14-16, [19] [20] [21] [22] and their neutralizing capacity has been confirmed (5, 12, 60

We identified five distinct joint trajectories of antibody subtypes and peptide pool-specific T cell 233 responses within the subsample ( Fig. 5A and 5B ). Clusters of participants were primarily defined 234 by the presence (clusters 1-4) or absence (cluster 5) of antibody responses, as well as distinct T 235 cell trajectories. When present, antibody trajectories generally followed the decay patterns 236 observed in the overall study population, characterized by waning anti-S-IgA and anti-N-IgG as 237 well as persistent anti-S-IgG. Meanwhile, T cell trajectories between clusters were more 238 heterogenous. We additionally examined T cell subsets using data from flow cytometric 239 analyses, which did not influence clustering, to identify differences in immune phenotypes 240 between clusters. 241

The first cluster (14% of subsample participants) was characterized by high antibody and 242 T cell responses, which remained high across all evaluated timepoints. T cell response 243 trajectories were similar for M, N, S1, S2 peptide pools, peaking at three months with a 244 subsequent slight decline up to six months post-diagnosis. N-and S2-specific T cell responses 245 were the most robust; at six months post-diagnosis, 88% participants had detectable N-specific T 246 cells and 75% had S2-specific T cells (Fig. 5C, Supplementary Fig. S3A ). By flow cytomatric 247 analysis, individuals in cluster 1 had the highest total CD4 + T cell count at one and three months 248 post-diagnosis, but showing a decline at six months ( Supplementary Fig. S3B ). Additionally, 249 total CD8 + T cell numbers were lower compared to clusters 3-5 and showed a decline from three 250 to six months. We further found that individuals in this cluster had higher frequencies of both 251 CD4 + AIM + and CD8 + AIM + virus-specific T cells at all timepoints compared to all other clusters 252 except cluster 2 (Fig. 5C ). We noted that participants belonging to the first cluster were mostly 253 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.15.21267776 doi: medRxiv preprint older than 65 years (89%), male (56%) and 22% were smokers (Fig. 5D ). The majority (78%) 254 had more than six COVID-19 symptoms and 44% were hospitalized within two weeks of 255 diagnosis, indicating more severe disease. Thus, cluster 1 tended to represent older individuals 256 with more severe disease and also robust antibody and cell-mediated immune responses. 257

The second cluster (12% of the subsample) was characterized by persistently high anti-S-258

IgA and -IgG responses in all participants, who also demonstrated a steep increase in virus-259 specific T cells from two weeks to one month post-diagnosis. T cell positivity subsequently 260 declined until six months post-diagnosis, with N-and S2-specific T cells being present in 63% 261 and 13% of participants, respectively (Fig. 5C, Supplementary Fig. S3A ). By flow cytometry, we 262 found that individuals in this cluster had lower total CD8 + T cell numbers than those in all other 263 clusters at two weeks and one months, increasing up to three months post-diagnosis 264 ( Supplementary Fig. S3B ). Meanwhile, they peaked in total CD4 + T cells at one month, which is 265 consistent with the increase in T cell responses observed by ELISpot analysis between two 266 weeks and one month after diagnosis (Fig. 5C, Supplementary Fig. S3B ). As in cluster 1, 267 frequencies of CD4 + AIM + and CD8 + AIM + T cells tended to be higher compared to other clusters 268 ( Fig. 5C, Supplementary Fig. S3A ). Also similar to cluster 1, most (63%) participants had six or 269 more symptoms and 25% were hospitalized (Fig. 5D ). However, participants in this cluster were 270 mostly younger than 65 years (63%), female (63%) and non-smokers or ex-smokers (75%). 271

Thus, cluster 2 was composed mostly of younger females with more severe disease, but robust 272 antibody responses and T cell responses which may tend towards higher CD4 + responses. 273

The third cluster (19% of the subsample) was characterized by antibody responses which 274 were initially high, but declined more sharply than in clusters 1-2. Anti-S-IgG, however, 275 remained detectable in all participants up to the six month follow-up visit. Meanwhile, overall T 276 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.15.21267776 doi: medRxiv preprint cell responses initially waned up to three months post-diagnosis, but then increased again for all 277 peptide pools by month six, at which point 92% of participants had detectable N-specific and S2-278 specific T cells (Fig. 5C, Supplementary Fig. S3A ). By flow cytometry, individuals in this 279 cluster tended to have the highest CD8 + T cell numbers at two weeks compared to other clusters 280 ( Supplementary Fig. S3B ). Meanwhile, they showed a decline in CD4 + and CD8 + T cells and 281 natural killer (NK) cells at one month with a subsequent increase up to six months, which is 282 consistent with the increased T cell responses observed by ELISpot between the three month and 283 six month follow-ups (Fig. 5C ). In addition, participants had higher CD19 + CD27 + memory B 284 cells at three and six months compared to the other clusters ( Supplementary Fig. S3B ). The 285 characteristics of the participants in this cluster were similar to those in the second cluster, but 286 with only 8% requiring hospitalization during acute infection (Fig. 5D) . Five of the participants 287 in this cluster had another PCR-test between three and six months after their infection, none of 288 which was positive. Together, cluster 3 appears to represent individuals with more mild disease 289 and more moderate, less-well-sustained antibody responses. The increase in T cell responses 290 between months three and six in these individuals may possibly represent a re-exposure event. 291

The fourth cluster (16% of the subsample) was characterized by the presence of 292 antibodies and a rapid decline of T cells specific to all peptide pools between the two week and 293 one month follow-up visits. By six months post-diagnosis, only 10% had detectable N-specific T 294 cells and none had S2-specific T cells (Fig. 5C, Supplementary Fig. S3A ). By flow cytometry, 295 individuals in this cluster had higher total CD4 + T cell numbers compared to clusters 2 and 3, but 296 lower compared to clusters 1 and 5 at two weeks and one month, subsequently showing a decline 297 in total CD4 + and CD8 + T cells up to six months post-diagnosis ( Supplementary Fig. S3B ). This 298 cluster was characterized by participants who were predominantly male (80%), younger than 65 299 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.15.21267776 doi: medRxiv preprint years (70%), and non-or ex-smokers (100%; Fig. 5D ). About 40% reported having six 300 symptoms or more and 40% were hospitalized in the acute phase. Together, cluster 4 appears to 301 be comprised of young males with moderate to severe disease, but who also have less-well-302

sustained antibody and T cell responses. 303

The fifth cluster (39% of the subsample) was characterized by persistently low and 304 primarily negative antibody responses and variable T cell responses. About half of the 305 participants had detectable M-and N-specific T cells across all assessments. An increase in S1-306 and S2-specific T cells by three months was noted which was followed by a decline (Fig. 5C , 307

Supplementary Fig. S3A ). Despite T cells responses which tended to be lower than for other 308 clusters by ELISpot, these individuals had higher total CD8 + T cell numbers compared to all 309 other clusters from one to six months post-diagnosis, as determined by flow cytometric analysis 310 ( Supplementary Fig. S3B ). They also had high total CD4 + T cell numbers compared to other 311 clusters (though not as high as cluster 1 at two weeks to three months), and showed a steady 312 increase in CD4 + and CD8 + T cells as well as NK cells up to six months ( Supplementary Fig.  313 S3B). Most individuals within this cluster were younger than 65 years (80%), female (68%), and 314 had mild disease as reflected by the low reported symptom count with none of the individuals 315 requiring hospitalization (Fig. 5D ). That individuals in cluster 5 experience relatively mild 316 disease in the absence of substantial antibody responses suggests compensatory protection by 317 cell-mediated immune responses, possibly including T cells specific to viral proteins not 318 captured by the assays used in this study. 319

Overall, our findings demonstrate the presence of different joint trajectories of antibody 320 and T cell immune responses after SARS-CoV-2 infection, highlighting the large heterogeneity 321 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. We then evaluated whether the demographic and clinical factors described within the 327 clusters were associated with antibody responses in the overall study population using adjusted 328 mixed-effects linear regression analyses. Consistent with other reports (7, 17, 21, 27-30, 48), we 329 found that older age (≥65 years) (p<0.001), male sex (p=0.011), higher symptom severity 330 defined as having one to five (p<0.001) or more than six COVID-19 symptoms (p<0.001), as 331 well as hospitalization (p=0.006) were statistically significantly associated with higher anti-S-332

IgG MFI ratios over time (Fig. 6A , Supplementary Table S5) . Conversely, being a current 333 smoker was associated with lower responses (p=0.010). Results for anti-S-IgA MFI ratios in the 334 overall population were comparable (Fig. 6B , Supplementary Table S5) , and similar trends were 335 identified for anti-N-IgG MFI ratios within the subsample (Fig. 6C ). For T cell responses, we 336 found that older age (≥65 years) (p<0.001) and having more than six symptoms (p<0.001) during 337 the acute infection were associated with higher T cell counts over time within the subsample 338 ( Fig. 6D, Supplementary Table S6 ). Sensitivity analyses for antibody or T cell positivity at two 339 weeks and six months after diagnosis using logistic regression models showed comparable trends 340 (Supplementary Tables S7 and S8 ). Therefore, severity of disease, age and smoking status are 341 associated with the magnitude of immune responses. 342 343 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. respectively. Compared to S1, the S2 domain of the full-length spike protein shares a higher 382 degree of amino acid identity with endemic coronaviruses (53). For example, HCoV-HKU1 383

shares an amino acid identity of 42% with SARS-CoV-2 at the S2 domain of the spike protein 384 compared to 31% at the S1 domain. Therefore, the longer half-life of S2-specific, compared to 385 S1-specific T cells, could potentially be due to intermittent exposure to endemic coronaviruses 386 with similar S2 peptide sequences, resulting in the appearance of a prolonged and more durable 387 T cell response. Overall, our findings suggest that M and S1 responses may initially be more 388 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.15.21267776 doi: medRxiv preprint robust, at least in terms of an IFN-gamma-producing response, but that perhaps responses to N 389 and S2 may be more durable. 390

In evaluating virus-specific CD4 + and CD8 + T cells by AIM assay we found that 391 frequencies of AIM + CD4 + or CD8 + T cells were similar at two weeks post-infection, but that 392 activated, virus-specific CD4 + T cells tended to decline over the six months of follow-up, 393

whereas levels of activated, virus-specific CD8 + T cells remained more stable. Furthermore, 394

CD4 + AIM + T cells were predominantly of TCM phenotype, and this remained consistent 395 throughout the convalescent period, whereas CD8 + AIM + T cells had a predominantly TEMRA 396 phenotype throughout this period consistent with previous studies (8, 54, 55). These findings 397 support the idea that CD4 + T cells may play an important role in maintaining the immune 398 response and immunological memory, whereas CD8 + T cells may play a more direct role in the 399

anti-viral immune response as highly activated and more terminally-differentiated cytotoxic 400 lymphocytes. 401 402

We further evaluated the relationship between antibody and T cell responses within 404 individuals over time, as well as the patterns of these responses within the population. We found 405 a strong, positive correlation between the three antibody subtypes evaluated here. Similarly, T 406 cell responses to M, N, S1, and S2 peptide pools also demonstrated a strong positive correlation, 407

suggesting that, although there may be some variation between distinct subtypes or specificities, 408 both overall antibody and T cell responses tended to behave similarly. 409 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.15.21267776 doi: medRxiv preprint observed a weak to moderate positive correlation early after infection, where increased antibody 411 responses tended to predict increased T cell responses to some degree. Meanwhile, correlations 412 became weaker later during follow-up and were weak or no longer present by six months. In 413 assessing the concordance between antibody and T cell responses, we also found that, at two 414 weeks, more than 70% of participants had concordant results (58% both antibody and T cell 415 positive and 13% both antibody and T cell negative). After six months, this dropped to 55-60%. 416

These findings suggest that individuals had differing immune responses following infection 417 (possibly due to differences in viral load or primary site of infection or previous immune 418 history), and that they perhaps retain differing subsets of immune memory components which 419 could be recalled upon reinfection. 420

Based on this idea, we explored whether there are heterogeneous immune trajectories 421 which individuals tend to follow in response to infection, and which might influence not only 422 their response to infection, but also the immune memory populations which they establish. Using 423 a longitudinal clustering algorithm, we assessed different patterns of immune responses between 424 individuals over time. We identified five distinct joint trajectories of antibody and T cell 425 responses. These trajectories were primarily based on the presence or absence of antibodies and 426 varying T cell responses. Interestingly, we observed distinct clinical characteristics between 427 these clusters. We found that clusters with the most robust immune responses, i.e., clusters 1 and 428 2, included older participants who had more severe COVID-19 as reflected by hospitalization 429 and the number of reported symptoms. This finding is not surprising as older age and disease 430 severity have been consistently reported to be associated with higher immune responses to 431 SARS-CoV-2 (7, 17, 21, 27-30), and other studies primarily conducted in patients with severe 432 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Our cohort is one of few population-based and longitudinal studies assessing various 453 components of the immune system in a sample of patients that is representative of the full 454 spectrum of COVID-19. However, some limitations should be considered when interpreting our 455 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.15.21267776 doi: medRxiv preprint findings. We used single assays to measure antibodies or T cells in our study. The accuracy and 456 detection levels may differ between tests and thus individuals who are negative in one assay may 457 not be so in another. Nevertheless, the Luminex assay that we used for antibody detection has 458 been extensively validated and was shown to be highly sensitive and specific (45). Second, we 459 did not measure the neutralizing capacity of antibody responses. However, other studies have 460

shown that neutralizing capacity correlates strongly with measured levels of binding antibodies 461 Canton of Zurich, understanding the German language, and being cognitively able to follow the 498 study procedures. We obtained written informed consent from all participants upon study 499 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. K2 EDTA blood samples collected from participants at each study timepoint were 519 subjected to initial centrifugation to collect plasma, followed by isolation of PBMCs from the 520 remaining cellular fraction by density-gradient centrifugation using Ficoll-Paque (density 521 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. and samples from PCR-positive donors, cut-off thresholds for MFI ratios used to determine 543 seropositivity were 6.5 for IgA and 6.0 for IgG. The lower limit of measured MFI ratios was 544 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint 

We summarized population characteristics descriptively and report frequencies and 603 percentage or median and interquartile range, as applicable. We report age-stratified summary 604 statistics for the frequency of antibody and T cell responses and calculated 95% Wilson 605 confidence intervals to estimate the associated uncertainty. We excluded any data measured after 606 receipt of COVID-19 vaccination (first dose; n=2 participants at three months and n=78 607 participants at six months) or diagnosed reinfection (based on self-reported positive PCR or rapid 608 antigen test; n=3 participants at six months) from all analyses. All data were transformed using 609 natural logarithms for all comparative and associational analyses due to non-normal distributions 610 and ratio properties of the Ig MFI ratios (zero values being replaced by half of the lowest non-611 zero value). Results are visualized using a log10-transformation in figures. 612 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint To estimate antibody decay times, we first determined the maximum response timepoint 613 for each individual. We excluded data from individuals that were never tested positive for the 614 respective antibody (i.e., anti-S-IgA or -IgG, or anti-N-IgG). We then restricted the data to the 615 maximum and all subsequent timepoints and rescaled the time axis to start with the maximum 616 concentration, in order to restrict and align the data for the descending slope of antibody decay 617 (in line with previous studies (59, 60) ). For T cells, we used the data as measured, as peak T cell 618 expansion typically occurs in the first week after infection (61). We then fitted uni-and 619 multivariable mixed-effects linear decay models on the natural logarithm-transformed data using 620 random intercepts for individuals. We then calculated the half-life in days using the formula = 621 Table S1 . Study population characteristics. 660 Table S2 . Anti-S-IgA and -IgG antibody responses in the overall study population over time. 661 Table S3 . Antibody and T cell responses in the subsample over time. 662 Table S4 . Sensitivity analysis regarding anti-S-IgA and -IgG antibody responses, weighted by 663 age group. 664 Table S5 . Association between demographic and clinical factors and antibody responses over 665 time. 666 Table S6 . Association between demographic and clinical factors and T cell responses over time. 667 Table S7 . Association between demographic and clinical factors and anti-S-IgG antibody 668 positivity at two weeks and six months. 669 Table S8 . Association between demographic and clinical factors and overall T cell positivity at 670 two weeks and six months. 671 Table S9 . Reference table for converting anti-S-IgG MFI ratios to BAU/ml (based on Roche 672

Elecsys Anti-SARS-CoV-2 S immunoassay). 673 674 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.15.21267776 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

Evolution of immune responses to SARS-CoV-2 in mild-681 moderate COVID-19

Dynamics of SARS-CoV-2 neutralising antibody responses and duration of immunity: a 686 longitudinal study

Dynamic 689 changes in anti-SARS-CoV-2 antibodies during SARS-CoV-2 infection and recovery from

COVID-19

Temporal profiles of viral load 699 in posterior oropharyngeal saliva samples and serum antibody responses during infection by

-2: an observational cohort study. The Lancet Infectious Diseases

Magnitude and kinetics of anti-SARS-CoV-2 antibody responses and 705 their relationship to disease severity

Immunological memory to SARS-CoV-2 assessed for up to 8 months after infection

Targets of T Cell Responses to SARS-CoV-2 Coronavirus in Humans with COVID

Disease and Unexposed Individuals

Immune Memory Persists after Mild COVID-19

Neutralizing antibody levels are 723 highly predictive of immune protection from symptomatic SARS-CoV-2 infection

Detection of SARS-CoV-2-Specific Humoral and Cellular Immunity in COVID

Quick COVID-19 Healers Sustain Anti-SARS-CoV-2 Antibody

Antibody responses to SARS-CoV-2 in patients with COVID-19

Antibody Responses to SARS-CoV-2 in Patients With Novel 752

Persistence and decay of human antibody responses to the receptor binding domain of

SARS-CoV-2 spike protein in COVID-19 patients

Cross-Sectional Evaluation of Humoral Responses against SARS-CoV-2

Detection of IgM 782 and IgG antibodies in patients with coronavirus disease 2019

Rapid Generation of Neutralizing Antibody Responses in COVID-19 Patients

COVID-19 patients define multiple targets of vulnerability

SARS-CoV-2 infection persist for months

Protective humoral and cellular immune responses to SARS-CoV-2 persist up to 1 year 817 after recovery

Differences in Antibody Kinetics and 820

Functionality Between Severe and Mild Severe Acute Respiratory Syndrome Coronavirus 2

Antigen-Specific Adaptive Immunity to

SARS-CoV-2 in Acute COVID-19 and Associations with Age and Disease Severity

Naturally enhanced neutralizing breadth against SARS-CoV-2 one year after 860 infection

Persistence 863 of Antibody and Cellular Immune Responses in COVID-19 patients over Nine Months after 864 Infection

CoV-2 infection induces long-lived bone marrow plasma cells in humans

Time series 879 analysis and mechanistic modelling of heterogeneity and sero-reversion in antibody 880 responses to mild SARS-CoV-2 infection

Waning Antibody Responses in Asymptomatic 884 and Symptomatic SARS-CoV-2 Infection. Emerg Infect Dis

Clinical and immunological assessment of asymptomatic SARS-CoV-2 infections

Antibody dynamics to SARS

CoV-2 in asymptomatic COVID-19 infections

Cellular Immunity in COVID-19 Convalescents with PCR-Confirmed

Phenotype and kinetics of SARS-CoV-905 2-specific T cells in COVID-19 patients with acute respiratory distress syndrome

Changes in SARS-CoV-2 Spike versus Nucleoprotein 910

Antibody Responses Impact the Estimates of Infections in Population-Based 911

Arnaud, kml and kml3d: R Packages to Cluster 913 Longitudinal Data

KmL3D: A non-parametric algorithm for clustering joint trajectories

Sex differences in 922 immune responses that underlie COVID-19 disease outcomes

Anti-spike antibody response to natural SARS-CoV-2 infection in the general population

Mapping Neutralizing and Immunodominant Sites on the SARS-CoV-2 Spike 937 52

Decay of Fc-dependent antibody functions after mild 948 to moderate COVID-19

The receptor binding domain of the viral 953 spike protein is an immunodominant and highly specific target of antibodies in SARS-CoV-954 2 patients

SARS-CoV-2-specific T cell memory is sustained in

COVID-19 convalescent patients for 10 months with successful development of stem cell-958 like memory T cells

SARS-CoV-961 2-Specific T Cells Exhibit Phenotypic Features of Helper Function

Differentiation, and High Proliferation Potential

Broad and strong memory CD4+ and CD8+ T cells induced by SARS-CoV-2 in UK 977 convalescent individuals following COVID-19

Zurich Coronavirus Cohort: an observational study to determine long-979 term clinical outcomes and immune responses after coronavirus infection (COVID-19), 980 assess the influence of virus genetics, and examine the spread of the coronavirus in the 981 population of the Canton of

Transcriptional control of effector and memory CD8+ T cell 991 differentiation

R: A language and environment for statistical computing. R Foundation for

Statistical Computing

CRK 1015 Visualization: DM, KDZ, TB, NC 1016 Funding acquisition: MAP, JSF 1017 Project administration: DM, TB, HEA, AD, MAP 1018 Supervision: CM, MAP, JSF 1019 Writing -original draft: DM, KDZ, TB

JSF 1022 each individual in the study. (E) Anti-S-IgG antibody decay estimation based on mixed 1049 linear regression mode. (F) Anti-N-IgG antibody decay estimation within the subsample

cytometry plots depicting AIM + (CD134 + CD137 + ) CD4 + (left) and (CD69 + CD137 + )

CD8 + (right) populations in unstimulated (top) or pooled SARS-CoV-2 peptide-1059 stimulated (bottom) PBMCs. (D) AIM + cells as a fraction of CD4 + or CD8 + T cells at 1060

Percentage of CD4 + or CD8 + AIM + T cells with TCM (CD45RA -CCR7 + ), TEM 1062 (CD45RA -CCR7 -) or TEMRA (CD45RA + CCR7 -) phenotypes by day post-diagnosis

S1-, and S-specific T 1078 cells/1e6 PBMCs for all participants belonging to the identified clusters 1 to 5. Gray 1079 color indicates missing values. (C) Dot plots of anti-S-IgG MFI ratio, N-specific T 1080 cells/1e6 PBMCs, S-specific T cells/1e6 PBMCs and frequency of AIM + of CD4 + T cells 1081 in the five clusters over time. Horizontal lines represent mean at timepoint. (D) Bar plot 1082 showing the proportion of participants according belonging to different age groups (18-1083 39 years, 40-64 years, ≥65 years), sex (male and female), smoking status (non-smoker, 1084 ex-smoker, smoker), number of symptoms reported (asymptomatic, 1-5 symptoms

Association analyses of antibody and T cell responses. (A) Forest plot demonstrating 1090 results from adjusted mixed linear regression analyses evaluating the associations of anti

COVID-19 symptom 1092 count, smoking status, body mass index, presence of at least one comorbidity and 1093 immunosuppression) in the overall study population (n=431)

The authors would like to thank the study administration team and the study 996 participants for their dedicated contribution to this research project. 997