Functional SARS-CoV-2-sperific immune memory persists after mild COVID-19

Summary: The recently emerged SARS-CoV-2 virus is currently causing a global pandemic and cases continue to rise. The majority of infected individuals experience mildly symptomatic coronavirus disease 2019 (COVID-19), but it is unknown whether this can induce persistent immune memory that might contribute to herd immunity. Thus, we performed a longitudinal assessment of individuals recovered from mildly symptomatic COVID-19 to determine if they develop and sustain immunological memory against the virus. We found that recovered individuals developed SARS-CoV-2-specific IgG antibody and neutralizing plasma, as well as virus-specific memory B and T cells that not only persisted, but in some cases increased numerically over three months following symptom onset. Furthermore, the SARS-CoV-2-specific memory lymphocytes exhibited characteristics associated with potent antiviral immunity: memory T cells secreted IFN-γ and expanded upon antigen re-encounter, while memory B cells expressed receptors capable of neutralizing virus when expressed as antibodies. These findings demonstrate that mild COVID-19 elicits memory lymphocytes that persist and display functional hallmarks associated with antiviral protective immunity.


150
As spike protein, and specifically the RBD, is key for viral entry into the cell, antibodies that target 151 the RBD can be potent inhibitors of infection 8,9 . To determine whether CoV2 + individuals form 152 and maintain neutralizing antibodies, we tested for SARS-CoV-2 neutralization indirectly using a 153 cell-free competition assay (surrogate virus neutralization test, sVNT) and directly in a plaque 154 reduction neutralization test (PRNT) 16 . CoV2 + plasma inhibited RBD binding to ACE2 155 significantly more than HC plasma by sVNT and RBD inhibition correlated strongly with anti-156 RBD IgG levels at both time points (Fig. 1d,e). Further, RBD inhibition capacity was maintained 86% of CoV2 + individuals still had better RBD-inhibiting plasma than HCs and 71% had better 161 neutralizing plasma (measured as above HC mean + 3 SDs). These data are consistent with the 162 emergence of predominantly IgG + RBD and spike-specific LLPCs that maintain detectable 163 neutralizing anti-SARS-CoV-2 antibody to at least 3 months post-symptom onset.

167
The presence of SARS-CoV-2-neutralizing antibodies three months post-symptom onset in CoV2 + 168 individuals suggests GC-derived memory LLPCs have formed. GC-derived MBCs play a critical 169 role in the formation of antibody secreting cells upon antigen re-exposure. Therefore, we tested 170 whether SARS-CoV-2-specific MBCs were also formed and maintained in CoV2 + individuals 171 throughout the study time course. We generated RBD tetramer reagents and used enrichment 172 strategies to identify rare RBD-specific cells that are otherwise undetectable in bulk assessments 17 . 173 We confirmed specificity in RBD immunized mice and then used the RBD-tetramer to identify, 174 enumerate and phenotype rare, RBD-specific B cells in our HN, HC and CoV2 + individuals (E.D. 175 Fig. 5a,b; Fig. 2a). Gates used to phenotype RBD-specific B cells were defined on total B cell 176 populations (E.D. Fig. 5c). At Visit 1, RBD-specific B cells were significantly expanded in CoV2 + 177 individuals compared to HCs and their numbers were increased further at Visit 2 (Fig. 2a, b). The 178 proportion and number of RBD-specific MBCs (defined by CD21 and CD27 expression) in CoV2 + 179 samples was significantly greater than in HCs and increased from Visit 1 to Visit 2 ( Fig. 2c,d,   180 E.D. Fig. 5d). While RBD-specific B cells in HN samples had a similar proportion of MBCs as in 181 CoV2+ samples, they contained substantially fewer cells. In addition, RBD-specific MBCs were 182 largely quiescent with very few expressing Ki67 (Fig. 2e, E.D. Fig. 5e). MBCs expression of class-183 switched B cell receptors (BCRs) is another marker of GC-derivation. We therefore assayed BCR 184 isotype expression on RBD-specific MBCs and found enriched populations of  expressing MBCs in CoV2 + individuals at both time points (Fig. 2f-h, E.D. Fig. 5f). Of note, while 186 small numbers of RBD-specific MBCs were detected in controls, these cells were predominantly 187 unswitched (IgM + and IgD + ), suggesting they may represent cross-reactive MBCs possibly 188 generated in response to one of the human coronaviruses that cause 15% of common colds 18-20 . 189 190 An additional measure of antiviral MBC function is the graded expression of T-bet 14 . MBCs that 191 express low-levels of T-bet are associated with rapid differentiation into secondary PBs that 192 produce high affinity, viral-specific antibodies during a secondary infection 21 . We found a higher 193 proportion and number of T-bet + , and specifically T-bet lo , RBD-specific MBCs in CoV2 + 194 individuals compared with HCs at Visit 1 and the higher numbers were maintained at Visit 2 (Fig. 195 2i-k, E.D. Fig. 5g,h). T-bet hi MBCs are considered to be recently activated and often found  204 205 The presence of T-bet + RBD-specific MBCs suggested that antigen-specific memory T cell 206 responses were also likely to be elicited in CoV2 + individuals. To enumerate SARS-CoV-2-207 specific memory T cells, total PBMCs from control or CoV2 + individuals were incubated with 208 spike protein and expression of activation markers was assessed (Fig. 3a) 23,24 . PBMCs from CoV2 + 209 individuals at Visit 1 and 2 displayed robust re-activation of spike-specific CD4 + memory T 210 responses, as measured by increased expression of ICOS and CD40L (two molecules associated 211 with B cell help upon re-activation), while PBMCs from HC and HN individuals did not ( Fig.   212 3a,b). There were no significant differences in the numbers of responding cells in CoV2 + 213 individuals between the two visits, suggesting spike-specific memory CD4 + T cells were 214 maintained throughout the study (Fig. 3b). Furthermore, greater numbers of CXCR5-expressing 215 circulating T follicular helper (cTfh) cells 25 , which provide B cell help, were found within the 216 population of S-specific ICOS + CD40L + CD4 + cells in CoV2 + individuals than in healthy controls 217 at both visits (Fig. 3c). Together these data suggest that SARS-CoV-2-specific memory CD4 + T 218 cells maintain the capacity to provide B cell help even at three months post-symptom onset.

220
Memory CD4 + T cells produce cytokines within hours of activation, whereas naive T cells take 221 days 26 . We first examined cytokine production from activated CD4 + memory CXCR5non-Tfh 222 cells and CXCR5 + cTfh cells identified in the assay above (Fig. 3b). S-specific CCR6 + CXCR5 + 223 cTfh cells, associated with IL-17 production, and a smaller population of CXCR3 + CXCR5 + cTfh 224 cells, associated with IFNγ production, were recently described in a predominantly mild to 225 moderate cohort 30 days post symptom onset 27 . We therefore analyzed activated ICOS + CD69 + S-226 specific cells for expression of CCR6 and CXCR5 and then cytokine expression was examined in 227 each population based on gating on a PMA positive control (Fig 3d, E.F. 6a). Although multiple  cytokines associated with Tfh function were assessed, only IFNγ, IL-17 and IL-2 cytokine   229 producing cells were significantly expressed in activated S-specific memory CD4 + cells in CoV2 + 230 individuals compared to HCs (Fig. 3d-f). Small numbers of S-specific cells were measured in HCs 231 after stimulation compared to vehicle alone that reflect previously described S-specific cross-232 reactivity 20,28 , but far greater responses were seen in the CoV2 + individuals (Fig. 3e). Three months 233 post symptom onset we found a higher frequency of CCR6 -cTfh cells that produced Th1 cytokines,

236
To further define the types of antigen-specific CD4 + memory T cells in CoV2 + individuals without 237 relying on secretion of specific cytokines, we assessed memory CD4 + T cell proliferation in 238 response to spike restimulation. For this, we sorted CD45RA + naive, CD45RA -CCR7 + central 239 memory (Tcm) and CD45RA -CCR7effector memory (Tem) T cells from HC or CoV2 + 240 individuals (E.D. Fig. 6b), then measured the proliferative capacity of each sorted population 241 following culture with autologous CD14 + monocytes and recombinant spike protein (Fig. 3g, Fig 6c). Only Tcm cells from CoV2 + individuals taken at both Visit 1 and Visit 2 displayed 243 significant proliferation frequencies compared to HC samples, although substantial proliferative 244 responses by Tem cells were observed in some CoV2 + individuals (Fig. 3h). We also examined 245 the expression of CXCR3 and CCR6 on S-specific, proliferated memory cells and found that the 246 majority of cells that had proliferated, as measured by the dilution of cell proliferation dye (CPD lo ) 247 expressed CXCR3, in keeping with Type 1 cytokine production in the previous assay. Spike-248 specific Tcm, and potentially Tem, are therefore maintained throughout our study and have the 249 ability to proliferate and re-populate the memory pool upon antigen re-encounter. While much recent work has focused on antibodies and B cells, memory CD8 + T cells are uniquely 252 positioned to kill virus infected cells through their directed expression of cytokines and cytolytic 253 molecules. S-specific memory CD8 + T cells that persisted for three months after mild COVID-19 254 disease could be identified by expression of the activation marker CD69 and the cytokine IFNγ 255 after overnight stimulation with spike (Fig 3i). Unlike CD4 + memory T cells, activated cytokine-256 expressing CD8 + T cells were significantly increased over vehicle controls in both control and 257 CoV2 + groups (Fig. 3j). Together, these data demonstrate that both CD4 + and CD8 + SARS-CoV-258 2-specific memory T cells are maintained and are able to produce effector cytokines after 259 restimulation three months post-symptom onset in mildly symptomatic COVID-19 individuals.

263
Since SARS-CoV-2 RBD-specific MBC and S-specific CD4 + cTfh were enriched in CoV2 + 264 individuals after 3 months, we assessed whether these MBCs could produce neutralizing antibodies 265 if they were reactivated by a secondary infection. To this end, we index sorted single RBD-specific 266 B cells and sequenced the BCRs from 3 CoV2 + individuals at Visit 1 (E.D. Fig. 7a). Of the class-267 switched (IgG + ) RBD-specific classical MBCs (CD21 + CD27 + ) we sorted, we randomly selected 7 268 to be cloned and expressed as IgG1 monoclonal antibodies (Fig 4a). This set of antibodies utilized 269 a wide variety of heavy and light chains, had all undergone somatic hypermutation and were all 270 unique clones (Fig. 4b, E.D. Table 2). These antibodies were first expressed in small scale 271 cultures. Transfection supernatants were assessed for antibody expression by IgG ELISA (E.D. Fig 7b) and specificity by RBD ELISA where all 7 showed strong binding to RBD (Fig. 4c). The 273 first 4 antibodies cloned were expressed on a larger scale and purified. The specificity of these 274 purified antibodies for RBD was again confirmed by ELISA (E.D. Fig. 7c) and their ability to 275 prevent SARS-CoV-2 infection was tested via PRNT assay. Two of the four tested (#202 and 203) 276 showed strong virus neutralization (Fig. 4d), with IC50 values of 31 ng/ml for both (Fig. 4e). This 277 was comparable to a previously published strongly neutralizing mouse antibody (B04) which was 278 included as a positive control (IC50=7ng/ml) 29 . Two of the RBD-specific antibodies were unable 279 to inhibit virus infection, similar to a non-neutralizing mouse antibody (C02) and an irrelevant 280 Plasmodium-specific human antibody. Three more monoclonal antibodies in addition to the 4 281 above were assessed for their capacity to inhibit RBD binding to the ACE2 receptor by sVNT 282 assay (Fig. 4f). Three of the seven were able to inhibit RBD binding to ACE2, similarly to a 283 strongly neutralizing alpaca nanobody 30 . Interestingly, #203, which neutralized live virus, did not 284 inhibit binding in this assay, while #202 both inhibited binding and neutralized the virus. Overall 285 50% of the antibodies tested showed inhibitory activity by one or both of these methods. Thus, In the absence of a vaccine, natural infection-induced herd immunity could play a key role in 293 reducing infections and deaths. For this to be possible, individuals that experience mild COVID-294 19 would need to develop and sustain protective immune memory. Here, we found that individuals 295 that recovered from mildly symptomatic COVID-19 had an expanded arsenal of SARS-CoV-2-296 specific immune mediators: neutralizing antibodies, IgG + T-bet lo classical MBCs, circulating 297 cytokine-producing CXCR5 + Tfh1 cells, proliferating CXCR3 + CD4 + memory cells and IFNγ 298 producing CD8 + T cells that were maintained to at least three months post-symptom onset. This 299 study predicts that these recovered individuals will be protected from a second SARS-CoV-2 300 infection and, if so, suggests that Th1 memory should be the target of vaccine elicited memory.  SARS-CoV-2-specific MBCs at one and three months from symptom onset. Our study revealed a 313 prominent population of RBD-specific IgG + CD27 + CD21 + T-bet lo MBCs, which has been 314 associated in other infections with rapid differentiation into antibody-secreting PBs upon re-315 exposure 5 , effective antiviral responses 39 and long-lived protection 3 . Furthermore, we found some 316 of the RBD-specific MBCs at Visit 1 expressed BCRs capable of neutralizing the virus when 317 expressed as antibodies. Since the numbers of these IgG + RBD-specific MBCs were not only 318 sustained, but continued to increase between one and three months, we predict they are GC-319 derived. Thus, MBCs at three months would have undergone increased affinity maturation and we 320 would expect an even higher percentage will be capable of producing neutralizing RBD-specific 321 antibodies upon re-infection. CoV2 + individuals rapidly displayed increased levels of ICOS and CD40L on CXCR5 + and 329 CXCR5cells after stimulation as well as expression of Th1-and Th17-associated cytokines.

330
These results are consistent with another recent report of SARS-CoV-2-specific cTfh cells 27 , 331 although they detected a high frequency of Th17-like cTfh cells, which could be due to the earlier 332 time point they were examining as Th17 cells can develop into Th1 cells late in an immune 333 response 40 . The expression of IFNγ and IL-17 by these cells is notable as these cytokines are 334 associated with class-switching to IgG and IgA isotypes, respectively 41,42 . We also found cross-335 reactive memory B and T cells in healthy controls, as has been previously noted 43 . It is difficult to 336 measure their contribution to the expanded populations of SARS-CoV-2-specific cells we found 337 in our CoV2 + cohorts, and therefore impossible to evaluate their protective capacity. However, we 338 can conclude that mild COVID-19 induces an expanded population of functionally diverse 339 memory lymphocytes compared to the cross-reactive pool present in our controls. Studies of reinfection have yet to be done in humans, but macaques infected with SARS-CoV-2 342 were protected from rechallenge 44 . This further suggests that the immune memory induced by mild 343 COVID-19 that we observed will be protective. While additional studies are needed to understand 344 variability of responses in a larger cohort and to determine how long memory to SARS-CoV-2 345 infection is truly maintained, our work suggests that mild COVID-19 induces persistent immune 346 memory poised for a coordinated, protective response to re-exposure that could contribute to herd 347 immunity and curtailing this pandemic.       The study was conceptualized utilizing a prospective case-control design. Cases and controls were 533 identified from a cross-sectional cohort study that recruited via print and online advertising from 534 the Seattle metropolitan area (E.D. Table 1 information was collected from all participants regarding recent illness symptoms and severity.

540
All CoV2 + cases reported at least one symptom but all were classified as mild disease, as none 541 required hospitalization. Historical negative control PBMCs (n=14) were sourced from the BRI 542 PBMC repository. Samples were drawn prior to 2020 and age and sex matched to the CoV2 + cases.

840
Error bars represent mean and SD.     Table 2. RBD-specific MBC-derived antibody amino acid sequences.

877
Extended Data