key: cord-0834594-62gfqb3q
authors: Zhang, Zeli; Mateus, Jose; Coelho, Camila H.; Dan, Jennifer M.; Moderbacher, Carolyn Rydyznski; Gálvez, Rosa Isela; Cortes, Fernanda H.; Grifoni, Alba; Tarke, Alison; Chang, James; Escarrega, E. Alexandar; Kim, Christina; Goodwin, Benjamin; Bloom, Nathaniel I.; Frazier, April; Weiskopf, Daniela; Sette, Alessandro; Crotty, Shane
title: Humoral and cellular immune memory to four COVID-19 vaccines
date: 2022-03-21
journal: bioRxiv
DOI: 10.1101/2022.03.18.484953
sha: adf28a582ba1de88babdd06bd0cb7857d2ef177f
doc_id: 834594
cord_uid: 62gfqb3q

Multiple COVID-19 vaccines, representing diverse vaccine platforms, successfully protect against symptomatic COVID-19 cases and deaths. Head-to-head comparisons of T cell, B cell, and antibody responses to diverse vaccines in humans are likely to be informative for understanding protective immunity against COVID-19, with particular interest in immune memory. Here, SARS-CoV-2-spike—specific immune responses to Moderna mRNA-1273, Pfizer/BioNTech BNT162b2, Janssen Ad26.COV2.S and Novavax NVX-CoV2373 were examined longitudinally for 6 months. 100% of individuals made memory CD4+ T cells, with cTfh and CD4-CTL highly represented after mRNA or NVX-CoV2373 vaccination. mRNA vaccines and Ad26.COV2.S induced comparable CD8+ T cell frequencies, though memory CD8+ T cells were only detectable in 60-67% of subjects at 6 months. Ad26.COV2.S was not the strongest immunogen by any measurement, though the Ad26.COV2.S T cell, B cell, and antibody responses were relatively stable over 6 months. A differentiating feature of Ad26.COV2.S immunization was a high frequency of CXCR3+ memory B cells. mRNA vaccinees had substantial declines in neutralizing antibodies, while memory T cells and B cells were comparatively stable over 6 months. These results of these detailed immunological evaluations may also be relevant for vaccine design insights against other pathogens.

Ad26.COV2.S and Novavax NVX-CoV2373 were examined longitudinally for 6 months. 100% of 28 individuals made memory CD4 + T cells, with cTfh and CD4-CTL highly represented after mRNA or NVX-29

CoV2373 vaccination. mRNA vaccines and Ad26.COV2.S induced comparable CD8 + T cell frequencies, 30 though memory CD8 + T cells were only detectable in 60-67% of subjects at 6 months. Ad26.COV2.S was 31 not the strongest immunogen by any measurement, though the Ad26.COV2.S T cell, B cell, and antibody 32 responses were relatively stable over 6 months. A differentiating feature of Ad26.COV2.S immunization 33 was a high frequency of CXCR3 + memory B cells. mRNA vaccinees had substantial declines in 34 neutralizing antibodies, while memory T cells and B cells were comparatively stable over 6 months. 35 These results of these detailed immunological evaluations may also be relevant for vaccine design 36

insights against other pathogens. 37 38 39 40 41 INTRODUCTION 4 (PSV) neutralization titers ( Figure 2C) were determined, for a total of 1,408 measurements from 352 136 samples. Binding antibody titers and PSV neutralization titers were quantified based on a WHO standard. 137

For mRNA-1273, after 1 st dose immunization, 100% of vaccinees had detectable spike IgG and 138 RBD IgG titers (Figures 2A-B) . 86% of vaccinees had detectable neutralization antibody titers after the 139 1 st dose ( Figure 2C ). These early findings are consistent with a large mRNA-1273 clinical trial cohort that 140 measured serology at early time points (100% positive for RBD IgG and spike IgG, 82% positive for 141 neutralization antibody (Gilbert et al., 2022) . After the 2 nd immunization, antibody levels both spike and 142 RBD IgG were boosted 9-fold (Figures 2A-B) and neutralization antibody titers were boosted 25-fold 143

(GMT 1,399) ( Figure 2C ). 100% of mRNA-1273 recipients remained positive for spike IgG, RBD IgG, and 144 neutralization antibodies at 6-months post-vaccination (T5). From peak (T3) to 6-months (T5), GMTs of 145 spike IgG decreased 6-fold, RBD IgG decreased 9-fold, and neutralizing antibodies decreased 7-fold. 146

For BNT162b2, after 1 st dose immunization, 100% of vaccinees had detectable Spike IgG and 147 RBD IgG titers (Figures 2A-B) . 76% of vaccinees had detectable neutralization antibodies after the 1 st 148 dose, which was slightly lower than the 86% with mRNA-1273 ( Figure 2C ). After the 2 nd immunization, 149 spike and RBD IgG were boosted 9-to 16-fold (Figures 2A-B) , and neutralization antibodies titers were 150 boosted 20-fold (GMT 903) ( Figure 2C ). 100% of BNT162b2 recipients remained positive for spike IgG, 151 RBD IgG, and neutralization antibodies at 6-month post-immunization (Figures 2A-C) . From peak (T3) 152

to 6-month (T5), GMT of spike IgG, RBD IgG, and neutralization antibody titers decreased by 6-fold, fold, and 6-fold, respectively. These antibody declines after BNT162b2 immunization were comparable 154 with declines after mRNA-1273 immunization (Figures 2A-C) . Neutralization antibody titers in 155

BNT162b2 recipients were lower than mRNA-1273 recipients by 1.6-fold (p=0.059), 2.2-fold (p=0.0014), 156 and 1.5-fold (p=0.13), at the T3, T4, and T5 time points, respectively. Neutralization antibody titers 157 trended lower in BNT162b2 than mRNA-1273 recipients when assessed in aggregate across the entire 158 6-month time period (area under curve (AUC), p=0.051, Figures S1B-D) . 159 For Ad26.COV2.S 1-dose immunization, 86% of vaccinees had detectable Spike IgG and 79% 160 RBD IgG at T2 (Figures 2A-B) . 64% of vaccinees had detectable neutralization antibodies at T2, which 161 was somewhat lower than the 86% with mRNA-1273 and 76% with BNT162b2 ( Figure 2C ). 162

Ad26.COV2.S antibody binding and neutralization titers gradually increased over time, with 100% of 163 recipients having detectable Spike IgG, RBD IgG, and neutralization antibodies at 6-month post-164 immunization. Ad26.COV2.S neutralization antibody titers peaked at T5 (GMT 58), but that peak was still 165 24-fold lower than the mRNA-1273 peak (GMT 1,399) and 16-fold lower than the BNT162b2 peak (GMT 166 903). At 6-month post-immunization, Ad26.COV2.S neutralization antibody titers were 3.6-fold lower 167 than mRNA-1273 and 2.4-fold lower than BNT162b2 ( Figure 2C ). Over the entire 6-month time period, 168

Ad26.COV2.S spike IgG, RBD IgG, and neutralization antibody titers were significantly lower than mRNA 169 vaccine recipients (p<0.0001 mRNA-1273, p<0.0001 BNT162b2. Figures S1B-D) . 170 For NVX-CoV2373, antibody titers were available for 3.5 and 6 months. Spike and RBD IgG titers 171 were substantial at 3.5 months post-vaccination and were marginally (not significantly) decreased at T5 172 (Figures 2A-B) . Neutralization antibody titers were comparable at both timepoints ( Figure 2C ). At 6-173 month post-immunization, NVX-CoV2373 neutralization antibody titers (GMT 152) were 2.6 fold higher 174 than Ad26.COV2.S (GMT 58), and were comparable to mRNA-1273 (GMT 209) and BNT162b2 (GMT 175 140). Considering the 3.5-month to 6-month period in aggregate, RBD IgG and neutralization antibody 176 titers in NVX-CoV2373 recipients were comparable to both mRNA vaccines (Figures S1F-G) . 177 Lastly, antibody titers at 6 months were compared to SARS-CoV-2 infected subjects ( Figures 2D-178 F) who were enrolled for a previously reported study . The previously infected 179 individuals were selected randomly. Recipients of the mRNA vaccines (mRNA-1273 and BNT162b2) had 180 4.5-fold higher spike IgG (Figure 2D ), 6.4-fold higher RBD IgG (Figure 2E) , and 3.4-fold higher 181 neutralization antibody titers ( Figure 2F ) compared to previously-infected subjects. Antibody titers from 182 5 NVX-CoV2373 recipients also trended higher than SARS-CoV-2 infected subjects (Figures 2D-F) . 183 Antibody titers from Ad26.COV2.S were similar to titers from SARS-CoV-2 infected subjects ( Figures  184  2D-F) . 185 Overall, antibody titers were significantly higher for mRNA recipients than Ad26.COV2.S 186 recipients. Recipients of NVX-Co2373 immunization also had higher peak antibody titers than recipients 187 of Ad26.COV2.S. Antibody titers to mRNA-1273, BNT162b2, and Ad26.COV2.S changed substantially 188 over the 6+ months of observation, with different patterns seen for the mRNA versus adenoviral vector 189 platforms. 190 191 Spike-specific CD4 + T cell memory elicited by four different vaccines 192 SARS-CoV-2 spike-specific CD4 + T cell responses were measured for all donors at all available 193 timepoints utilizing two previously described flow cytometry activation-induced marker (AIM) assays 194 (OX40 + CD137 + and OX40 + surface CD40L + (sCD40L)) ( Figures 3A, 3D In response to a single dose of the mRNA-1273 vaccine (T2), a majority of subjects developed a 201 spike-specific CD4 + T cell response as measured by both AIM + (Figures 3A and S3 ) and iCD40L + 202 secreted-effector + (ICS + ) CD4 + T cells ( Figure 3C ). Spike-specific CD4 + T cell responses peaked after the 203 2 nd mRNA-1273 vaccination (100% responders, T3) and were well maintained out to 6 months post-204 vaccination, with only a 1.0-to 1.9-fold reduction in AIM + or ICS + CD4 + T cells, respectively (Figures 205 3A,C and S3). mRNA-1273 vaccination induced spike-specific cTfh cells in most donors after the 1 st 206 dose, which peaked after the 2 nd dose (97%, T3), and memory cTfh cells were maintained out to 6 207 months post-vaccination with only a 1.4-fold change from peak (T3 to T5, Figure 3B ). Memory cTfh cells 208

represented 27% of the spike-specific memory CD4 + T cells, on average. 209

Vaccination with BNT162b2 induced spike-specific AIM + and ICS + CD4 + T cells after the first 210 vaccination (T2), with peak responses after the 2 nd immunization (T3) (Figures 3A, 3C and S3) . However, 211 peak responses to BNT162b2 vaccination were significantly lower than mRNA-1273 peak vaccine 212 responses both by AIM and ICS (1.6-fold lower, P=0.019; and 2.5-fold lower, P=0.011. Figures 3A and  213 C). Memory CD4 + T cells were detectable in 85-100% of BNT162b2 vaccinees at 6 months after 214

immunization, but the memory CD4 + T cell frequencies were significantly lower than for mRNA-1273 215

(1.9-fold lower by AIM, P=0.011 and 2.4-fold lower by ICS, P=0.038, Figures 3A and 3C) . Spike-specific 216 memory cTfh cell frequencies were comparable between BNT162b2 and mRNA-1273 vaccination 217 ( Figure 3B ).

Both mRNA-1273 and BNT162b2 vaccination induced ICS + spike-specific memory CD4 + T cells, 219

including iCD40L + IFNg + , iCD40L + TNFa + , and iCD40L + IL-2 + cells, detectable out to 6 months post-220 vaccination. mRNA-1273 vaccinees had significantly higher frequencies of TNFa + and IL-2 + CD4 + T cells 221 at all timepoints and higher levels of IFNg + memory CD4 + T cells at 6 months relative to BNT162b2 222 vaccinees (Figure 4) . GzB + CD4 + T cells (iCD40L + GzB + ) were assessed as indicators of CD4 + cytotoxic T 223 lymphocytes (CD4-CTL). Interestingly, both mRNA vaccines generated CD4-CTLs as a significant fraction 224 of the overall spike-specific CD4 + T cell response (Figures 4E and S4B) . Multifunctional spike-specific 225

CD4 + T cells were observed after the 1 st dose of either mRNA-1273 or BNT162b2, and multifunctionality 226 was stably maintained out to 6 months ( Figures 3C and S4A) . 227

For the Ad26.COV2.S vaccine, spike-specific CD4 + T cell responses were detectable in a majority 228 of individuals and were largely stable out to 6 months post-vaccination (69-100% of individuals with 229 6 spike-specific CD4 + T cells by AIM assays; 46% with spike-specific CD4 + T cells by ICS. Figures 3A, 3C,  230 and S3). cTfh cells were detectable in the majority of individuals ( Figure 3B ). Peak CD4 + T cell responses 231

were lower to Ad26.COV2.S than either of the mRNA vaccines. Peak AIM + CD4 + T cells to Ad26.COV2.S 232 were 2.2-to 3.3-fold lower than BNT162b2 and 3.5-to 4.2-fold lower than mRNA-1273 peak responses 233 (Figures 3A and S3) . Peak spike-specific ICS + CD4 + T cell responses to Ad26.COV2.S were 5.8-fold 234 lower than BNT162b2 and 14-fold lower than mRNA-1273 ( Figure 3C ). Both mRNA vaccines generated 235 significantly higher peak frequencies of IFNg + CD4 + T cells than Ad26.COV2.S vaccination 236 (iCD40L + IFNg + , mRNA1273 P<0.0001, BNT162b2 P=0.001), and mRNA-1273 vaccinees had significantly 237

higher IFNg + spike-specific memory CD4 + T cells than Ad26.COV2.S at 6 months post-vaccination 238 (P=0.007, Figure 4) . The mRNA vaccines also induced significantly more CD4-CTLs at peak than 239

Ad26.COV2.S (mRNA1273 P<0.0001, BNT162b2 P=0.0012, Figure 4E ), and the CD4-CTLs induced by 240 the mRNA vaccines were more sustained as memory cells at the 6-month memory timepoint relative to 241

Ad26.COV2.S ( Figure 4E ). Spike-specific CD4 + T cells induced by Ad26.COV2.S had less 242 multifunctionality at all time points relative to both mRNA vaccines (Figures 3E and S4A) immunized individuals had spike-specific memory cTfh cells ( Figure 3B ). Memory CD4 + T cell responses 249

to NVX-CoV2373 were comparable in magnitude to the mRNA vaccines by AIM (Figures 3A, 3C and 250 S3). By ICS, NVX-CoV2373 responses 6 months post-vaccination were comparable to BNT162b2 (NVX-251

CoV2373 geomean 0.074%, BNT162b2 0.059%), and significantly higher than the Ad26.COV2.S vaccine 252 (Ad26.COV2.S geomean 0.015%, P=0.0057. Figure 3C ). NVX-CoV2373 induced multifunctional 253 memory spike-specific CD4 + T cells comparably to both mRNA vaccines (T4 and T5, Figures 3C and  254 S4A), with a shift in the relative abundance of IL-2 + cells over IFNg + memory CD4 + T cells observed for 255 NVX-CoV2373 (Figures 4B and 4D) . 256 Spike-specific CD4 + T cell responses in COVID-19 recovered individuals were assessed to  257 compare infection-induced versus vaccine-elicited T cell memory (Figures 3D-F) . Spike-specific CD4 + T 258 cell memory at 6 months post-vaccination in mRNA-1273 and NVX-CoV2373 vaccinees was significantly 259

higher than for COVID-19 recovered individuals, both by AIM and ICS (Figures 3D and 3F) . BNT162b2 260 and Ad26.COV2.S generated memory CD4 + T cells frequencies not significantly different than SARS-261

CoV-2 infection (Figures 3D and 3F ). Memory cTfh cell frequencies were similar between all four 262 vaccines and infection ( Figure 3E) . Overall, all four of the COVID-19 vaccines generated memory CD4 + 263 T cells in the majority of vaccinated individuals, with representation of both Th1 (IFNg + ) and Tfh memory, 264

with memory CD4-CTL also generated by mRNA and NVX-CoV2373 vaccines. Additionally, the 265 magnitude of spike-specific CD4 + T cell memory was generally higher for mRNA vaccines and NVX-266

CoV2373 than seen in COVID-19 recovered individuals. 267 268

Spike-specific CD8 + T cells elicited by four different vaccines 269 SARS-CoV-2 spike-specific CD8 + T cells were measured by ICS at all time points to identify IFNg, TNFa, 270 or IL-2 producing cells (CD69 + cytokine + gating = "ICS + ". Figures 5A-C, S5-7) for all vaccine modalities. 271

Spike-specific CD8 + T cells were also measured by AIM (CD69 + CD137 + , Figure S8 ).

For the mRNA-1273 vaccine, 83% of vaccinees had detectable spike-specific CD8 + T cell 273 responses after the 1 st immunization ( Figure 5C ). ICS + CD8 + T cell response rates peaked after the 2 nd 274 immunization (87% T3 responders Figure 5C ). Spike-specific memory CD8 + T cells were largely 275 maintained out to 6 months after mRNA-1273 vaccination (67% responders, Figure 5C ), with only a 2.3-276 7 fold decline in geomean frequency from the peak (0.077% to 0.033%, Figure 5C ). Both acute and 277 memory CD8 + T cell responses were dominated by IFNg-producing cells (Figures 5B-C and S9) , the 278 majority of which co-expressed GzB ( Figure S9 ). The majority of the memory spike-specific CD8 + T cells 279 exhibited an effector memory (TEM) surface phenotype ( Figure S10) . 280 For the BNT162b2 vaccine, IFNg + and total ICS + CD8 + T cell responses also peaked after the 2 nd 281 immunization (T3 73% and 85% responders, respectively Figures 5B-C) . Memory CD8 + T cells were 282 maintained out to 6 months after BNT162b2 vaccination (60% responders, Figure 5C ), with only a 1.8-283

fold decline in geomean frequency ( Figure 5C ). Multifunctional spike-specific memory CD8 + T cells were 284 more common in mRNA-1273 compared to BNT162b2 vaccinees (Figures 5C and S9A) , with the 285 responses dominated by IFNg + cells ( Figures 5C, and S9) . Overall, spike-specific CD8 + T cell acute and 286 memory responses to BNT162b2 were similar to mRNA-1273 but slightly lower in frequency and 287 multifunctionality. 288

The fraction of CD8 + T cell responders to Ad26.COV2.S was lower than both mRNA vaccines 289

(71% compared to 87% and 85%, Figure 5C ). Nevertheless, Ad26.COV2.S spike-specific CD8 + T cell 290 frequencies were relatively stable through 6 months post-vaccination (Figures 5B-C and S5) and 291 geomean frequencies of memory CD8 + T cells after Ad26.COV2.S vaccination were comparable to both 292 mRNA vaccines at 6 months (Figures 5B-C and S5).

For the NVX-CoV2373 vaccine, spike-specific ICS + memory CD8 + T cells were observed in 10% 294

to 40% of donors (T4 and T5, Figure 5C ). There were minimal multifunctional CD8 + T cells ( Figure 5C  295 and S9). 296

Overall, memory CD8 + T cell frequencies and response rates were similar between mRNA-1273, 297

BNT162b2, and Ad26.COV2.S immunizations. Low but detectable memory CD8 + T cells were observed 298 in some individuals after NVX-CoV2373 immunization. CD8 + T cell responses to all COVID-19 vaccines 299

were dominated by IFNg-producing cells. No differences in IFNg MFI were observed between memory 300 CD8 + T cells generated to each of the vaccines ( Figure S7 ). All vaccines elicited AIM + (CD69 + CD137 + ) 301 CD8 + cell responses at levels comparable to, or slightly higher than, frequencies observed in SARS-CoV-302 2 recovered individuals. (Figure S6 ).

Spike-and RBD-specific B cell memory to four COVID-19 vaccines 305

Next, we sought to characterize and compare the development of B cell memory across the 4 306 different COVID-19 vaccines. For that, we utilized spike and RDB probes to identify, quantify and 307 phenotypically characterize memory B cells from vaccinated subjects at 3.5 (T4) and 6 months (T5) after 308 immunization ( Figures 6A-B and S11). Spike-specific and RBD-specific memory B cells were detected 309 in all vaccinated subjects at 6 months ( Figures 6C-D) . RBD-specific memory B cells comprised 15 to 310 20% of the spike-specific memory B cell population, on average ( Figure S12A ). Immunization with 311 mRNA-1273 or BNT162b2 led to higher frequencies of spike-specific and RBD-specific memory B cells 312 compared to Ad26.COV2.S and NVX-CoV2373 at 3.5 and 6 months (each p<0.01. Figures 6C-D) .

Memory B cell responses to the 4 vaccines did not exhibit the same kinetics as the antibody 314

responses. The frequency of spike-specific memory B cells increased over time, (mRNA-1273, p=0.017; 315

BNT162b2, p=0.0018, Ad26.COV2.S, p=0.021. Figure 6C ). RBD-specific memory B cell frequencies 316 increased at 6 months after mRNA-1273 (1.7-fold, p=0.024), BNT162b2 (2.2-fold, p=0.06), Ad26.COV2.S 317

(2.1-fold, p=0.06), and NVX-CoV2373 (3.05-fold, p=0.033) ( Figure 6D ).

RBD-specific memory B cell isotypes were mostly comparable among the different vaccines, with 319 an average distribution of 83.0% IgG, 2.5% IgM, and 2.2% IgA at 6 months ( Figure 6D and Figure S12B ); 320 however, IgA + RBD-specific memory B cells were significantly higher at 3.5 months in mRNA vaccinees 321 compared to Ad26.COV2.S (mRNA-1273 p=0.003. BNT162b2 p=0.04. Figure 6D) . Phenotypically, 322 activated memory B cells (CD21 -CD27 + ) comprised 77-85% of spike-specific memory B cells after mRNA 323 8 vaccination ( Figure 6E and S12C), which was significantly higher than observed for Ad26.COV2.S or 324 NVX-CoV2373 (66%, 61%, mRNA vs. Ad26.COV2.S, p<0.0001. mRNA-1273 p=0.0027; BNT162b2 325 p=0.0038. Figure 6E) , and the differences persisted at 6 months ( Figure 6F) . Reciprocally, the 326 representation of classical memory B cells (CD21 + CD27 + ) was lower in response to mRNA vaccines 327 ( Figure S12D ). To further qualitatively compare memory B cells across vaccine platforms, we assessed 328 CD71, CXCR3, CD95, and CD11c expression by spike-specific memory B cells. CD71 + memory B cells 329

were more common at 3.5 months in response to mRNA vaccines than Ad26.COV2.S or NVX-CoV2373 330 (T4 Figure 6G) , with higher expression on activated memory B cells ( Figure S12E ). Considering that 331 CD71 is a proliferation marker of B cells, this may reflect greater continuing production of memory B 332 cells in response to mRNA vaccines at 3.5 months compared to Ad26.COV2.S and NVX-CoV2373 333

vaccines. At 6 months, the frequency of CD71 + spike-specific memory B cells remained elevated for 334 mRNA-1273 ( Figure S12F ). CXCR3 + spike-specific memory B cell frequencies were substantially higher 335 in response to Ad26.COV2.S compared to the other vaccine platforms (mRNA-1273 p<0.001, 336

BNT162b2 p<0.001, NVX-CoV2373 p=0.008. Figure 6H ) and remained elevated at 6 months ( Figure  337 S12G). 338

Lastly, the frequencies of spike-specific and RBD-specific memory B cells at 6 months post-339 vaccination were comparable to the frequencies found in previously-infected subjects at 6 months 340

( Figures 6I-J) , indicating robust memory B cell development to each of the four COVID-19 vaccines. 341 342

Multiparametric comparisons across vaccine platforms 343 We performed multiparametric analyses, utilizing both correlation matrixes and principal component 344 analysis (PCA) to assess the relative immunogenicity of the four vaccines. Considering all parameters of 345 vaccine antigen-specific immune responses at 6 months after mRNA (mRNA-1273 and BNT162b2) or 346

Ad26.COV2.S vaccination (Figures S13A-B) , we observed strong correlations between spike IgG, RBD 347

IgG, and neutralization antibody titers (Figures 7A-B and 7F) . Neutralization antibody titers correlated 348 with spike-specific and RBD-specific memory B cells for mRNA vaccinees at 6 months ( Figures 7A, C-D) .

Antibody levels and memory CD4 + T cells were significantly associated in mRNA vaccinees by multiple 350 metrics (Figures 7A and 7E) . In contrast, no relationship was observed between antibodies and memory 351 CD8 + T cells ( Figure 7A ). Memory CD4 + T cells and CD8 + T cells were significantly cross-correlated in 352 mRNA vaccinees ( Figure S13A ). For Ad26.COV2.S vaccination, no significant correlations were 353 detected at 6 months between antibodies, memory B cells, memory CD4 + T cells, or memory CD8 + T 354 cells, which may be related to the smaller cohort size (Figures 7A and 7G-I) . 355 Next, we tested for relationships between early immune responses and immune memory 356 (Figures 7J-N, S13B and S14). Peak post-2 nd mRNA immunization cTfh CD4 + T cells were strongly 357 associated with 6-month antibody levels (Figures 7J-L, and S13C-D), providing an early indicator of 358 long term humoral immunity. Early RBD IgG titers after the 1 st mRNA immunization were positively 359 associated with 6-month RBD-specific memory B cell frequencies ( Figure S14E -F). For both mRNA and 360

Ad26.COV2.S, peak ICS + CD4 + and CD8 + T cell responses significantly cross-correlated (T3, Figure 7J ). 361

Overall, these observations suggest that early peak CD4 + T cells responses had a lasting effect on the 362 humoral response. 363

PCA mapping was performed using 3.5-month ( Figure S14G ) and 6-month ( Figure 7O ) post-364 vaccination data. PCA discriminated mRNA-1273 and Ad26.COV2.S, indicating these two vaccines 365 generated distinct immunological profiles ( Figure 7O ). BNT162b2 largely developed the same profile 366

as mRNA-1273 but with more heterogeneity. NVX-CoV2373 generated an immune memory profile 367 overlapping with that of mRNA and adenoviral vectors ( Figure 7O ). Prominent immunological features 368 distinguishing between mRNA and Ad26.COV2.S were CXCR3 + spike-specific memory B cells, ICS + 369 memory CD4 + T cells, CD71 + memory B cells, and spike IgG (Figure 7O and S14G). Notably, 370 9 neutralizing antibody titers and CXCR3 + spike-specific memory B cells were correlated for Ad26.COV2.S 371 vaccinees (r=0.44, p=0.04) but not mRNA vaccinees (mRNA-1273, p=0.25. BNT162b2, p=0.79. Figure  372 7P), corroborating the immunologically distinct outcomes. Overall, substantial relationships were 373 observed between multiple components of immune memory for these COVID-19 vaccines, with distinct 374 immune memory profiles for different vaccine platforms. 375 376 DISCUSSION 377

COVID-19 vaccines have achieved extraordinary success in protection from infection and 378

disease; yet some limitations exist, including differences in VE between vaccines and waning of 379 protection against infection over a period of several months. Here, diverse metrics of adaptive responses 380

were measured to mRNA-1273, BNT162b2, Ad26.COV2.S and NVX-CoV2373, with implications for 381 understanding the protection against COVID-19 associated with each of the vaccines. A strength of this 382 study is that the samples from different vaccine platforms were obtained from the same blood 383 processing facility, from the same geographical location, and were analyzed concomitantly, utilizing the 384 same experimental platform. 385

In the present study, antibody responses were detected in 100% of individuals. At 6 months post-386

immunization, the neutralizing antibody titer hierarchy between the vaccines was mRNA- compared to other vaccines after 6 months have been very limited. Here we observed that NVX-CoV2373 394

neutralizing antibody titers were comparable to that of BNT162b2 and only moderately lower than mRNA-395

1273.

In this side-by-side comparative study, spike-specific CD4 + T cell responses were detected in 397

100% of individuals to all four vaccines. While neutralizing antibody kinetics were different between 398 mRNA and viral vector vaccines, the CD4 + T cell response kinetics were similar. The hierarchy of the 399 magnitude of the memory CD4 + T cells was mRNA-1273>BNT162b2~NVX-CoV2373>Ad26.COV2.S. 400

These with longitudinal data and single-cell cytokine expression resolution providing insights regarding CD4 + 405 T cell subpopulations between the vaccines. Interestingly, multifunctional CD4 + T cells were observed 406 most frequently after mRNA-1273 immunization, and CD4-CTLs represented a substantial fraction of the 407 memory CD4 + T cells after mRNA-1273, BNT162b2, or NVX-CoV2373 vaccination. cTfh memory cells 408

were represented as a substantial fraction of CD4 + T cell memory for each of the 4 vaccines, consistent 409 with these vaccine platforms being selected for their ability to induce antibody responses. Memory CD4 + 410

T cell responses were also comparted to infected individuals, demonstrated that each vaccine was 411 successful in generating circulating spike-specific CD4 + T cell memory frequencies similar to or higher 412 than SARS-CoV-2 infection, though of course infection also generates responses to other viral antigens. 413

The two mRNA vaccines and Ad26.COV2.S induced comparable acute and memory CD8 + T cell 414

frequencies to the overnight stimulation used here. As expected for a protein-based vaccine, CD8 + T cell cytokine 419 responses to NVX-CoV2373 were lower than all other vaccine platforms assessed, but it was notable that 420 NVX-CoV2373 generated spike-specific CD8 + T cell memory in a fraction of individuals. 421

Spike-and RBD-specific memory B cell responses were detected in all individuals to each of the 422 four vaccines. While neutralizing antibody titers declined over time in mRNA vaccinees, the frequency 423 of spike-specific memory B cells increased over time. These divergent antibody and memory B cell 424 kinetics were also observed in SARS-CoV-2 infection ( For baseline determinations, for a subset of donors, blood samples were collected before 678 vaccination (T1), and subsequently 2 weeks (15 ± 3 days) after the first immunization (T2), 2 weeks (45 ± 679 35 days after first immunization) after the second immunization, 3.5 months (105 ± 7 days) and 6 months 680

thereafter (185 ± 6 days). Both cohorts of mRNA vaccinees (mRNA-1273, Pfizer/BioNTech BNT162b2) 681 received two doses of the vaccine (28 and 21 days apart, respectively). In the case of Ad26.COV2.S 682 blood donations were collected after one dose at the same timepoints. Finally, in the case of the Novavax 683 NVX-CoV2373, we advertised locally to recruit subjects who had participated in an investigational NVX-684

CoV2373 trial conducted in the San Diego region, where two intramuscular 5-μg doses of NVX-CoV2373 685 or placebo were administered 21 days apart (Clinicaltrials.gov). The study was structured in such way 686 that donors received either first a placebo injection followed 21 later from the vaccine, or a vaccine 687 injection followed by a placebo injection 21 days later; participants were blinded to their immunization 688 regimen, and LJI had no information on which group the participants were in. An overview of samples 689 analyzed in this study is provided in Figure 1 Exclusion criteria 702

Before analyzing the entire data set for our cohort, we generated exclusion criteria as follows: subjects 703 who tested positive for RBD and neutralization antibodies at baseline were excluded (one subject, 704 mRNA-1273); subjects with no baseline sample available and whose RBD and neutralization antibody 705 reached the peak after first-dose immunization (indicative of memory from previous infection) and were 706 nucleocapsid (NC) antibody-positive were also excluded as previously infected subjects (one subject, 707 CD4 + T cells) were defined by the AIM assay. Briefly, prior to the addition of the Spike MP, PBMCs were 802 blocked at 37°C for 15 min with 0.5 µg/ml anti-CD40 mAb (Miltenyi Biotec S1 and a representative gating strategy of Spike-specific CD4 + and CD8 + T cells using the AIM assay is 814 shown in Figure S2 , respectively. 815

Spike-specific CD4 + and CD8 + T cells were measured as background (DMSO) subtracted data, 816

with a minimal DMSO level set to 0.005%. Response > 0.02% and a stimulation index (SI) > 2 for CD4 + 817 and > 0.03% and SI > 3 for CD8 + T cells were considered positive. The limit of quantification (LOQ) for 818

antigen-specific CD4 + T cell responses (0.03%) and antigen-specific CD8 + T cell responses (0.05%) was 819 calculated using the median two-fold standard deviation of all negative controls. were permeated and stained with intracellular antibodies for 30 min at room temperature in the dark. 834

All samples were acquired on a Cytek Aurora. Antibodies used in the ICS assay are listed in table S2  835 and a representative gating strategy of cytokine-producing spike-specific CD4 + and CD8 + T cells using 836 the ICS assay is shown in Figures 4A and 5A . 837

To define the spike-specific T cells by the ICS assay, we gated the cytokine-or GzB-producing 838 cells together with the expression of iCD40L or CD69 on CD4 + or CD8 + T cells, respectively ( Figures 4A  839  and 5A) . Then, a Boolean analysis was performed to define the multifunctional profiles on FlowJo 10.8.1. 840

The overall response to spike, denoted as Secreted-effector + (IFNg, TNFa, IL-2, and/or GzB) or Cytokine + 841 (IFNg, TNFa, and/or IL-2), was defined as the sum of the background-subtracted responses to each 842 combination of individual cytokines or GzB. The total spike-specific CD4 + and CD8 + T cells producing 843

IFNg, TNFa, IL-2, and/or GzB are shown in Figures 4-5 and Figure S5 , respectively. To define the 844 multifunctional profiles of spike-specific T cells, all positive background-subtracted data (> 0.005% and 845 a SI > 2 for CD4 + T cells and CD8 + T cells) was aggregated into a combined sum of antigen-specific CD4 + 846 or CD8 + T cells based on the number of functions. Values higher than the LOQ (0.01%) were considered 847 for the analysis of the multifunctional spike-specific T cell responses. The average of the relative CD4 + 848 23 and CD8 + T cell responses was calculated per donor and visit to define the proportion of multifunctional 849

spike-specific T cell responses with one, two, three, and four functions (Figures 3E, 5C, S4 and S9) . corr.mtest and graphed based on * p < 0.05, **p < 0.01, *** p < 0.001. The codes used are: 880

M=cor(DataFrame, method="spearman", use = "pairwise.complete.obs") 881

MP=cor.mtest(DataFrame, method="spearman", use = " pairwise.complete.obs", conf.level=0.95, 882 exact=FALSE) 883 corrplot(M, p.mat = MP$p, method = 'square', tl.col="black", tl.cex = 0.7, tl.srt = 45, cl.align="l", type = 884 'lower', sig.level = c(0.001, 0.01, 0.05), pch.cex = 0.7, insig = 'label_sig', pch.col = 'white') 885

The "DataFrame" is the data from each correlation matrix shown in figure 6 , collected and organized in 886 spreadsheet. 887

The codes for PCA analysis are as follows: 888 res.pca=PCA(na.omit(MP), scale = TRUE) 889 fviz_eig(res.pca, addlabels = TRUE) 890 fviz_pca_biplot(res.pca, label ="var", labelsize = 3, repel= TRUE, geom.ind = "point", pointsize= 4, 891

col.ind = Group$Vaccine, palette = c("darkgreen", "blue", "red", "purple"), col.var = "black", alpha.var = 892 0.5, addEllipses = TRUE, ellipse.alpha=0, select.var=list(name=c("RBD IgG", "Spike IgG", "nAbs", "MBC", 893 "aMBC", "cMBC", "CXCR3+ MBC", "AIM2+ CD4", "ICS+ CD4", "ICS+ CD8")), ellipse.level=0.8, 894

legend.title = "Groups", invisible = "quali", title="") 895 24 896

Statistical analysis 897

Cytometry data was analyzed using T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5 T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5   10 T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5 T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5 0.01 )  14  30  30  28  24  14  29  27  28  22  6  15  16  25  14  10  10  Responders (%)  0  70  87  89  88  0  45  78  54  36  0  33  38  32  36  20  50   B   T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 )  14  30  30  28  24  14  29  27  28  22  6  15  16  25  14  10  10  Responders (%)  0  83  100  96  96  0  48  70  64  73  0  33  25  20  36  70  90   T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 )  14  30  30  28  24  14  29  27  28  22  6  15  16  25  14  10  10  Responders (%)  0  87  93  96  92  0  59  81  68  73  0  33  43  46  31  90  88   T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 (A) Representative gating of spike-specific CD8 + T cells. Cytokine-producing ("cytokine + ") CD8 + T cells were quantified as CD69 + along with IFNg, TNFa, or IL-2 expression after stimulation with spike MP. (B) Longitudinal quantitation of CD69 + IFNg + spike-specific CD8 + T cells. See Figure S5 for TNFa and IL-2, and Figure S9 for additional analysis.

(C) Longitudinal quantitation of cytokine + spike-specific CD8 + T cells. CD8 + T cells were quantified as CD69 + along with IFNg, TNFa, or IL-2 expression after stimulation with spike MP. Bottom bars show fold-changes between T3 and T5. The donut charts depict the proportions of multifunctional cytokine + profiles of the spike-specific CD8 + T cells, including IFNg, TNFa, or IL-2 and GzB: 1 (light gray), 2 (dark gray), 3 (black), and 4 (turquoise) functions (See also Figure S9 ). The dotted black line indicates the limit of quantification (LOQ). Graphs are color-coded as per Figure 2 . Background-subtracted and log data analyzed. Data were analyzed for statistical significance using the Mann-Whitney test [(B), (C)]. T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 A   T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 T5  T1 T2 T3 T4 

In addition, any time points following a confirmed COVID-19 booster immunization were 708 excluded for any subject (five subjects, BNT162b2

METHOD DETAILS 711 Peripheral blood mononuclear cells (PBMCs) and plasma isolation 712 Whole blood samples from subjects vaccinated with the mRNA-1273, BNT162b2, Ad26.COV2.S, or 713 NVX-CoV2373 COVID-19 vaccine and convalescent samples after COVID-19 infection were

803 g to separate the 715 cellular fraction and plasma. Blood samples were collected at the times described above. The plasma 716 was then carefully removed from the cell pellet and stored at minus 20°C. PBMCs were isolated by 717 density-gradient sedimentation using Ficoll-Paque (Lymphoprep, Nycomed Pharma, Oslo, Norway) as 718 previously described

supplemented with 10% heat-inactivated fetal 721 bovine serum (FBS; Hyclone Laboratories, Logan UT), and stored in liquid nitrogen until used in the 722 assays. Plasma samples were used for antibody measurements by ELISA and PSV neutralization assay 723 and PBMC samples were used for flow cytometry in the T cell and B cell assays. 724 725 SARS-CoV-2 ELISAs 726 The SARS-CoV-2 ELISAs have been described previously

Plasma was heat-inactivated at 56°C for 30 to 60 min. Plasma was diluted in 1% milk containing 0

PBS starting at a 1:3 dilution followed by serial dilutions by three and incubated for 1.5 734 hours at room temperature. Plates were washed five times with 0.05% PBS-Tween-20. Secondary 735 antibodies were diluted in 1% milk containing 0.05% Tween-20 in PBS. Anti-human IgG peroxidase 736 antibody produced in goat

Spectramax Plate Reader at 450 nm, and data analysis was performed using SoftMax Pro

Negative and 739 positive controls were used to standardize each assay and normalize across experiments. A positive 740 control standard was created by pooling plasma from 6 convalescent COVID-19 donors to normalize 741 between experiments. The limit of detection (LOD) was defined as 1:3 of IgG. The limit of quantification 742 (LOQ) for COVID-19 vaccinated individuals were established based on pre-vaccinated individuals 743 (timepoint 1) and set as the titer at which 95% of pre-vaccinated samples (T1) fell below the dotted line 744 (Figures 2A-B). Titers, LOD, and LOQ were calibrated to the WHO International Reference Panel for 745 anti-SARS-CoV-2 spike

For RBD IgG, the LOD was 0.83 747 with a LOQ of 7.12 (Figures 2B and 2E). For NC IgG, the LOD was 0.68 with a LOQ of 30

S over the entire 6+ month 750 time period, log10 transformed end-point titers (WHO BAU/mL) were used to generate area under the 751 curve (AUC) for each donor (Figures S1B-D). Donors with only 1 timepoint excluded. If there was no 752 (T1), T1 was set as the LOD ET (BAU/mL). Correction factors for AUCs were determined by the number 753 of time points and normalized to compare donor to donor

over the 3.5 months to 6 months period, log10 transformed 755 end-point titers (WHO BAU/mL) were used to generate area under the curve (AUC) for each donor 756 (Figures S1E-G). Donors with only 1 timepoint excluded. Kruskal-Wallis tests for AUC were <0.0001 for 757 Figures S1B-G. Comparison between different vaccines were made by Mann-Whitney

Pseudovirus (PSV) Neutralization Assay

The PSV neutralization assays in samples from vaccinated subjects were performed as previously 762 described

As internal quality 774 control to define the variation inter-assay, a pooled plasma (secondary standard) from 10 donors who 775 received the mRNA-1273 vaccine was included across the PSV neutralization assays. Samples that did 776 not reach 50% inhibition at the lowest serum dilution of 1:20 were considered as non-neutralizing and 777 the values were set to 19. PSV neutralization titers were done with two replicates per experiment. We 778 included the WHO International Reference Panel for anti-SARS-CoV-2 immunoglobulin (20/268) to 779 calibrate our PSV neutralization titers

Spike megapool (Spike MP)

According to this approach, large numbers of different epitopes are solubilized, pooled, and re-786 lyophilized to avoid cell toxicity problems associated with high concentrations of DMSO typically 787 encountered when single pre-solubilized epitopes are pooled

We used a Spike MP of 790 253 overlapping peptides spanning the entire sequence of the Spike protein. As this peptide pool 791 consists of peptides with a length of 15 amino acids, both CD4 + and CD8 + T cells have the capacity to 792 recognize this MP, as described previously

Activation-induced markers (AIM) assay

The AIM assays in samples from subjects vaccinated with mRNA-1273, BNT162b2, Ad26.COV2.S, or 796 NVX-CoV2373 COVID-19 vaccine were performed as previously described

T cells were measured as a percentage of AIM + (OX40 + CD137 + ) CD4 + and 799 (CD69 + CD137 + ) CD8 + T cells after stimulation of PBMCs from subjects vaccinated with the Spike MP

Spike-specific circulating T follicular helper (cTfh) cells (CXCR5 + OX40 + CD40L + , as a percentage of

asterisks as * p <0.05, **p <0.01, *** p <0.001. MBCs indicates memory B cell, AIM1 indicates OX40 + CD137 + , AIM2 indicates OX40 + sCD40L + , nAb indicates neutralization antibody

B-I) The association of indicated parameters shown by scatter plot. Red indicated mRNA, green indicated Ad26.COV2.S. Spearman rank-order correlation values (r) and p values were shown

The blue rectangle indicates the association between CD4 + T cell and antibody. Spearman rank-order correlation values (r) are shown from red (-1.0) to blue (1.0); r values are indicated by color and square size

The association of indicated parameters shown by scatter plot. Red indicated mRNA, green indicated Ad26.COV2.S. Spearman rank-order correlation values (r) and p values were shown

S (n=14), and NVX-Cov-2373 (n=10) on the basis of all parameters obtained 6-month post-vaccination. Only paired subjects were used for the PCA analysis. Arrows indicated the prominent immunological distinguishing features. Ellipse represented the clustering of each vaccine. Red indicated mRANA-1273, blue indicated BNT162b2, and green indicated Ad26.COV2.S. MBCs indicates spike-specific memory B cell, cMBCs indicates spike-specific classical memory B cell, aMBCs indicates spike-specific activated memory B cell

Spearman rank-order correlation between PSV neutralization titers and frequency of spike MBCs expressing CXCR3 at 3.5 months after vaccination. Background-subtracted and log data analyzed

 Figure 3 . Acute and memory CD4 + T cell responses after mRNA-1273, BNT162b2, Ad26.COV2.S, or NVX-CoV2373 immunization. (A) Longitudinal spike-specific CD4 + T cell responses induced by four different COVID-19 vaccines measured by OX40 + CD137 + AIM after spike megapool (MP) stimulation. See Figures S2B-C for the representative gating strategy of AIM + cells. (B) Longitudinal spike-specific circulating T follicular helper cells (cTfh) induced by COVID-19 vaccines. Spike-specific cTfh cells (CXCR5 + OX40 + sCD40L + , as % of CD4 + T cells) after stimulation with spike MP. See Figure S2D for representative gating strategy. (C) Spike-specific CD4 + T cells measured by ICS. Expressing iCD40L and producing IFNg, TNFa, IL-2, or GzB (Secreted-effector + = ICS + ). See Figure 4 for analysis of individual IFNg, TNFa, IL-2, or GzB on spike-specific CD4 + cytokine + T cells expressing iCD40L). Donut charts depict the proportions of multifunctional secreted effector profiles among the spike-specific ICS + CD4 + T cells: 1 (light gray), 2 (dark gray), 3 (black), and 4 (turquoise) functions (See also Figure S4