key: cord-0254954-ufzjvdho
authors: Swadling, L.; Diniz, M. O.; Schmidt, N. M.; Amin, O. E.; Chandran, A.; Shaw, E.; Pade, C.; Gibbons, J. M.; Le Bert, N.; Tan, A. T.; Tham, C. Y. L.; Tan, C.; Kucyowicz, S.; Aidoo-Micah, G.; Rosenheim, J.; Davies, J.; Jensen, M. P.; Joy, G.; McCoy, L. E.; Valdes, A. M.; van Dorp, L.; Altmann, D. M.; Boyton, R. J.; Manisty, C.; Treibel, T. A.; Moon, J. C.; COVIDsortium Investigators,; Balloux, F.; McKnight, A.; Noursadeghi, M.; Bertoletti, A.; Maini, M. K.
title: Pre-existing polymerase-specific T cells expand in abortive seronegative SARS-CoV-2 infection
date: 2021-07-01
journal: nan
DOI: 10.1101/2021.06.26.21259239
sha: 15ee8b11b4f0f24d29194dc6883418675ab752af
doc_id: 254954
cord_uid: ufzjvdho

Individuals with likely exposure to the highly infectious SARS-CoV-2 do not necessarily develop PCR or antibody positivity, suggesting some may clear sub-clinical infection before seroconversion. T cells can contribute to the rapid clearance of SARS-CoV-2 and other coronavirus infections1-5. We hypothesised that pre-existing memory T cell responses, with cross-protective potential against SARS-CoV-26-12, would expand in vivo to mediate rapid viral control, potentially aborting infection. We studied T cells against the replication transcription complex (RTC) of SARS-CoV-2 since this is transcribed first in the viral life cycle13-15 and should be highly conserved. We measured SARS-CoV-2-reactive T cells in a cohort of intensively monitored healthcare workers (HCW) who remained repeatedly negative by PCR, antibody binding, and neutralisation for SARS-CoV-2 (exposed seronegative, ES). 16-weeks post-recruitment, ES had memory T cells that were stronger and more multispecific than an unexposed pre-pandemic cohort, and more frequently directed against the RTC than the structural protein-dominated responses seen post-detectable infection (matched concurrent cohort). The postulate that HCW with the strongest RTC-specific T cells had an abortive infection was supported by a low-level increase in IFI27 transcript, a robust early innate signature of SARS-CoV-2 infection16. We showed that the RNA-polymerase within RTC was the largest region of high sequence conservation across human seasonal coronaviruses (HCoV) and was preferentially targeted by T cells from UK and Singapore pre-pandemic cohorts and from ES. RTC epitope-specific T cells capable of cross-recognising HCoV variants were identified in ES. Longitudinal samples from ES and an additional validation cohort, showed pre-existing RNA-polymerase-specific T cells expanded in vivo following SARS-CoV-2 exposure, becoming enriched in the memory response of those with abortive compared to overt infection. In summary, we provide evidence of abortive seronegative SARS-CoV-2 infection with expansion of cross-reactive RTC-specific T cells, highlighting these highly conserved proteins as targets for future vaccines against endemic and emerging Coronaviridae.

monitored healthcare workers (HCW) who remained repeatedly negative by PCR, antibody binding, and neutralisation for SARS-CoV-2 (exposed seronegative, ES). 16-weeks postrecruitment, ES had memory T cells that were stronger and more multispecific than an unexposed pre-pandemic cohort, and more frequently directed against the RTC than the structural protein-dominated responses seen post-detectable infection (matched concurrent cohort). The postulate that HCW with the strongest RTC-specific T cells had an abortive infection was supported by a low-level increase in IFI27 transcript, a robust early innate signature of SARS-CoV-2 infection 16 . We showed that the RNA-polymerase within RTC was the largest region of high sequence conservation across human seasonal coronaviruses (HCoV) and was preferentially targeted by T cells from UK and Singapore pre-pandemic cohorts and from ES. RTC epitope-specific T cells capable of cross-recognising HCoV variants were identified in ES. Longitudinal samples from ES and an additional validation cohort, showed pre-existing RNA-polymerase-specific T cells expanded in vivo following SARS-CoV-2 exposure, becoming enriched in the memory response of those with abortive compared to overt infection. In summary, we provide evidence of abortive seronegative SARS-CoV-2 infection with expansion of cross-reactive RTC-specific T cells, highlighting these highly conserved proteins as targets for future vaccines against endemic and emerging Coronaviridae.

There is wide variability in the outcome of exposure to highly infectious SARS-CoV-2, ranging from severe illness to asymptomatic infection, to those remaining negative with standard diagnostic tests. Identification of exposure without infection has largely been based on isolated cases with single time-point screening 6, [17] [18] [19] [20] . We undertook a systematic study of an intensively monitored cohort of healthcare workers (HCW) exposed during the first pandemic wave, comparing those with or without PCR and/or antibody evidence of SARS-CoV-2 infection. We postulated that in HCW where PCR, and the most sensitive binding and neutralising antibody (nAb) tests, remained repeatedly negative (exposed seronegative, ES), T cell assays might distinguish a subset with a subclinical, rapidly terminated (abortive) infection. We hypothesised that these individuals would exhibit pre-existing memory T cells with cross-reactive potential, obviating the time required for de novo T cell priming and clonal expansion. 1a). The exposed seronegative (ES) HCW group were defined by negativity on state-of-the-art diagnostic tests carried out weekly for 16weeks (SARS-CoV-2 PCR from nasopharyngeal swabs; anti-Spike-1 IgG and anti-nucleoprotein (NP) IgG/IgM seroassays 36 (Fig. 1b-d) . Having previously reported a range of nAb titres persisting at wk16 in the laboratory-confirmed infection group 21 , we examined nAb in the ES group. Two HCW with nAb titres just above the threshold were excluded from further analyses; the remaining ES were negative by pseudotype assay (Fig. 1e) , with a subset also confirmed negative at 3 time points by authentic virus neutralisation assay (Extended Data Fig. 1a) . Some may have been infected before recruitment but non-seroconverters after PCR positivity were rare (only 2.6% of PCR+ HCW were negative by all 3 serological tests 21 ), and antibody responses were unlikely to have waned before study recruitment 36 . Furthermore staining ES with dual-colour tetramers showed they lacked detectable SARS-CoV-2 spike-specific memory B cells, which we have shown persist after waning of nAb 37 (example plots Extended Data Fig. 1b , comparable frequency to pre-pandemics [below the threshold of detection 37 ], Fig. 1f) , Thus, ES represented a cohort of intensely monitored HCW who resisted classical laboratoryconfirmed infection during the first pandemic wave in the UK.

We quantified SARS-CoV-2-specific memory T cell responses by ELISpot using the unbiased approach of stimulating PBMC with overlapping peptides covering both structural proteins and the less well-studied key non-structural proteins of the RTC in ORF1ab (Fig. 1g) . As previously described, when using sensitive assays [7] [8] [9] 11, [22] [23] [24] [25] (such as IFN -ELISpot with 400,000 PBMC/well 10,21 used here), some SARS-CoV-2-reactive T cells were detectable in the pre-pandemic samples; however, their multispecificity was significantly lower than in the wk16 laboratory-confirmed infected samples (Fig. 1h-i ; structural responses at wk16 previously reported 21 ). By contrast, ES had SARS-CoV-2-specific T cell responses that were comparable in breadth to the infected HCW at wk16 and significantly more multispecific than in individuals sampled prior to the pandemic (Fig. 1h-i) . Not only did ES target more protein pools than pre-pandemics, but they also had an ~5-fold higher cumulative magnitude of responses than pre-pandemics with an overall strength equivalent to the infected cohort at wk16 (Fig. 1j-k) .

. 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.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint Fig. 1 SARS-CoV-2-specific immunity in exposed seronegative HCW. a, Design of COVIDsortium prospective HCW study and pre-pandemic cohort. b, Longitudinal cycle threshold values for E gene PCR in ES (n=58) and laboratory-confirmed infection (n=76) groups . 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.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint (undetectable at 40 cycles assigned value 41). c, Longitudinal anti-Spike S1 and d, anti-NP antibody titres in ES (baseline to wk16; n=58; dotted lines at assay positivity cut-off and at average peak [AvPos] response in laboratory-confirmed infected group). e, Pseudovirus neutralisation at wk16 (n=58). Crossed circles excluded from ES group (IC50 >50). f, Frequency of SARS-CoV-2 spike-specific memory B cells in pre-pandemic or exposed seronegative cohort (wk16; as a percentage of total memory B cells). g, SARS-CoV-2 proteome with RTC and structural regions (and peptide pool numbers) assayed for T cell responses highlighted and number of overlapping 15mer peptides (or mapped epitopes peptides for spike) used in brackets below. h, Viral proteins recognised by individuals, coloured by specificity, and i, number of viral proteins targeted by group. j, Magnitude of T cell response coloured by viral protein and k, cumulative magnitude of T cell response by group. Bars, geomean. l, Proportion of cohorts with T cell responses to NP1/NP2 pools. h-l, IFN -ELISpot. e,f,i Bars, median. i,k, Kruskal-Wallis with Dunn's correction. ES, exposed seronegative; HCW, health care worker; M, membrane; MEP; mapped epitope pool; NP, nucleoprotein; RTC, replication-transcription complex; SFC, spot forming cells.

We noted that T cells from pre-pandemic samples tended not to target both halves of the NP protein (stimulation pools NP1 & NP2), whereas around 50% of ES and laboratory-confirmed donors targeted both NP pools, confirming our early suggestion 10 that this serves as a simple proxy-measure of a more multispecific response (Fig. 1l, Extended Data Fig. 1c-d) . Taken together, we found a higher magnitude and breadth of SARS-CoV-2-specific T cells in repeatedly PCR and antibody negative HCW than in a pre-pandemic cohort.

Having established that T cell reactivity in the ES group differed from pre-pandemic samples, we next sought to further differentiate them from the group with infection confirmed by PCR and/or seroconversion. Anti-viral T cells recognising immunodominant MHC class I restricted peptides from Flu, Epstein-Barr virus (EBV) and cytomegalovirus (CMV) (FEC) were equivalent between the three cohorts (Extended Data Fig. 2a) . However, looking at specificity for structural versus non-structural regions (RTC proteins from ORF1ab) amongst wk16 memory T cells, we found that the relative immunodominance of these regions differed between groups. The infected group had memory T cells dominated by more responses to structural proteins (Spike, membrane, NP, and ORF3a) than to RTC (NSP7, NSP12, NSP13) ( Fig. 2a-b) .

Memory T cells against structural proteins tended to be stronger in those who had a higher viral load, whereas RTC responses did not show this association (Extended Data Fig. 2b) . By contrast, pre-existing T cell responses predominantly targeted RTC proteins, whilst ES recognised both regions (Fig. 2b, Extended Data Fig. 2c-d) , but with a significantly higher 7 magnitude response to RTC-specific than the infected group (Fig. 2a, Extended Data Fig. 2d) .

A further small group (11%) of HCW had PCR-confirmed infection but lacked detectable nAb at wk16, some of whom also lacked binding antibodies; interestingly this sub-group was similarly enriched for RTC-reactive T cells (Extended Data Fig. 2e-f ). Percentage of cohort with a ratio above 1 (RTC>Structural) shown below. c, example CTV and IFN staining (gated on CD4+ [black] or CD8+ [blue] T cells) and d, dual cytokine or activation marker staining of SARS-CoV-2-specific T cells in an ES after 10-day expansion (proliferating T cells become CTV lo as they divide and dilute out marker) with peptide pools. SARS-CoV-2specific highlighted in red (CTV lo IFN + ). Percentage of CD4+ or CD8+ CTV lo IFN + shown. e, Peak and f, longitudinal (first 5 weeks of follow-up) IFI27 transcript signal by RT-PCR in ES with low (in bottom 20 responders) or strong (in top 20 responders) RTC-specific T cells, compared with other baseline samples from HCW who remained PCR negative throughout, and with HCW at the time of PCR positivity. Range of baseline values highlighted in grey. a,b, IFN -ELISpot. a,b, bars at geomean. e, bars at median. a,b, Kruskal-Wallis ANOVA with Dunn's correction.

Taken together, this suggests that the structural proteins, produced in abundance via subgenomic RNA during active infection, are the main targets for T cell responses after mild infection, but that T cells generated by early, transient viral exposure are preferentially focused on the key components of the RTC.

. 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)

The copyright holder for this preprint this version posted July 1, 2021. To confirm the T cell identity of ELISpot responses detected in the ES group we expanded them with cognate peptides from RTC regions and used proliferation and cytokine production for flow cytometric identification of antigen-specific CD4+ and CD8+ T cells. T cell responses that divided (CTV dilution) and produced peptide-specific IFN could be readily expanded from samples taken from ES at wk16 (Fig. 2c ; Gating Extended Data Fig. 3a ; Extended data Table 2 ). Their post-expansion frequencies tended to be lower than control flu/EBV/CMVspecific responses in the same donors but were in proportion to their differing ex vivo frequencies, indicating comparable proliferative potential (Extended Data Fig. 3b ). In vitro expanded T cell responses in ES were also highly functional, producing multiple cytokines in tandem (Fig. 2d) . Most of the SARS-CoV-2-specific T cells expanded from ES were CD4+, however, CD8+ T cell responses were also detectable in most individuals (Extended data Fig.   3c ).

Our T cell data raised the possibility that SARS-CoV-2 infection in HCW represents a spectrum, with some ES expanding T cell responses having had a sub-clinical abortive infection not detectable by PCR or antibody seroconversion. To test this postulate, we applied blood transcript measurements of the interferon-inducible gene IFI27 as a biomarker, which we recently showed discriminates early SARS-CoV-2 infection at, or one week before, PCR positivity (specificity 0.95 and sensitivity 0.84 16 ). ES with the highest post-exposure RTCspecific responses had significantly raised peak IFI27 levels when compared to baseline controls (Fig. 2e) , although levels tended to be lower than in the laboratory-confirmed infected group (Fig. 2e) . A time-course of IFI27 over the first five weeks after recruitment showed a stepwise increase in IFI27 in ES, reaching a plateau by week 3-4, by which time almost all first wave laboratory-confirmed infections had occurred (Fig. 2f) . By contrast IFI27 was unchanged over 5 weeks sampling in ES with low or undetectable RTC-specific T cell responses (Fig. 2f) . Therefore, a low-level systemic interferon response indicative of virus exposure was detectable in individuals who had the strongest SARS-CoV-2-specific T cell response post-exposure, despite them lacking PCR or antibody confirmation of SARS-CoV-2 infection. Extrapolating from our previous data showing that IFI27 is induced at the time of incident infection and correlates with viral load 16 , this supports the ES with stronger RTCspecific T cell responses representing HCW who have experienced a low-level/transient . 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)

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint infection. Thus, ES with SARS-CoV-2 T cell reactivity could be distinguished by both their innate IFI27 signature and the propensity of their T cells to target RTC.

A transient/abortive infection not detectable by PCR or seroconversion could conceivably result from a lower viral inoculum and/or from a more efficient innate and/or adaptive immune response. The latter would be favoured by pre-existing memory T cell responses with the potential to expand rapidly upon cross-recognition of early viral products of SARS-CoV-2

replication. Early T cell proliferation and TCR clonal expansion, even prior to detectable virus, has been observed during mild SARS-CoV-2 22,38 and expansion of virus-specific T cells predates antibody induction after mRNA vaccination 39 . Having found the ES group to be enriched for SARS-CoV-2-specific T cells, particularly against RTC, we therefore investigated the possibility that some of these represented expansions of pre-existing cross-reactive responses.

Likely candidates for the induction of T cells that cross-recognise SARS-CoV-2 are closely related human endemic common cold coronaviruses (HCoV: α-HCoV-229E and NL63, and β-HCoV HKU1 and OC43). We bioinformatically determined the sequence homology of all possible SARS-CoV-2-derived 15mer peptides to a curated set of HCoV sequences (Supplementary Table 1 ). We found that the RTC proteins, expressed at the first stage of the SARS-CoV-2 life cycle 15 , have 15mer sequences that are of high homology to the HCoVs (Fig.   3a ) 24, 40 . In particular, NSP7, NSP12, and NSP13-derived 15mers had 6.3, 29.9 and 31.0% higher average sequence homology to the four HCoVs compared to structural protein-derived 15mers (all p<0.001, Fig. 3b ). NSP12 being the largest of these 3 proteins, represents the region with the most homology overall. Interestingly, the highly conserved RNA polymerase (NSP12) was also the region that was found to be most commonly targeted when screening our cohort of 52 pre-pandemic samples. T cell responses to NSP12 showed the highest average magnitude and frequency of donors responding (Fig. 3c) . Of note, the same preferential targeting of NSP12 was observed in a geographically distinct cohort of prepandemic samples from Singapore (Fig. 3c) .

. 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)

The copyright holder for this preprint this version posted July 1, 2021. Table 1 ). Viral proteins not assayed for T cell responses are shown in grey. c, Magnitude of T cell responses to individual SARS-CoV-2 proteins in pre-pandemic samples taken in London, UK and in Singapore and d, ES at wk16. Frequency of responders shown in doughnuts above. e, Upper panels: alignment of Coronaviridae sequences across two CD8 epitopes; conserved amino acids in yellow. Lower panels: Magnitude of CD8+ T cell response (CTV-IFN +) after 10-day expansion with HCoV variant sequence peptides as a percentage of response with SARS-CoV-. 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)

The copyright holder for this preprint this version posted July 1, 2021.

2 9-mer peptide. f, Example plot of CTV vs. IFN -APC after 10-day expansion with SARS-CoV-2 or HCoV sequence 9-mer peptides (ES wk16 samples; gated on single, live, CD3+, CD8+). c,d Bar, geomean. e Bar, median. nd, not done. c,d Kruskal-Wallis with Dunn's correction.

Pre-existing T cells had the potential to recognise all viral antigens tested, including those with less conservation across HCoV, as previously described 22,41 and responses against all these regions were further enriched in ES, suggesting many sources of pre-existing responses and de novo generated responses can contribute to T cell memory in exposed seronegative individuals. However, as with pre-pandemics samples, ES preferentially targeted NSP12 (Fig.   3d) . Therefore, the viral protein most commonly targeted by pre-existing T cells is also the largest conserved region, suggesting that exposure to HCoV is likely one source of crossreactive T cells.

To further explore the potential for cross-reactivity due to prior infection with seasonal HCoV in this group, we first carried out epitope mapping of RTC-specific T cells from ES. Twodimensional mapping matrices were used to determine individual immunogenic 15mers Table 3 ). Next, we aligned viral sequences for HCoV at the CD4+ and CD8+ epitopes we Fig. 3f ). This suggested prior HKU1 infection primed these NSP7 responses, which were then able to cross-recognise the SARS-CoV-2 sequence, albeit with reduced efficiency. All four HLA*B35 ES also showed some cross-recognition of seasonal HCoV variant epitopes in NSP12, with the extent varying as would be expected in light of heterogeneity in previous HCoV exposure and T cell repertoire composition (Fig. 3e) .

. 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)

The copyright holder for this preprint this version posted July 1, 2021. In summary, key RTC regions like the RNA polymerase, that are expressed in the first stage of the viral life cycle, are highly conserved among HCoV and are preferentially targeted by T cells in pre-pandemic and ES samples. T cells from donors able to abort infection can crossrecognise SARS-CoV-2 and HCoV sequences at individual epitopes within the RTC, pointing to prior infection with HCoV as a source of some pre-existing cross-protective T cells.

To examine whether pre-existing cross-reactive and/or rapidly generated de novo RTCspecific T cells can expand in vivo, we took advantage of unusual access to paired PBMC samples taken pre-and post-SARS-CoV-2 exposure. Firstly, we recruited a cohort of medical students and laboratory staff from whom stored PBMC were available from winter 2018-2019 (n = 23), prior to the COVID-19 pandemic, and sampled them again after known close contact with infected cases, with or without IgG seroconversion +/-PCR positivity (Extended Data Table 4 ). Pre-and post-exposure/infection PBMC were analysed in parallel for ELISpot responses to RTC and structural pools.

There was clear evidence for in vivo expansion of pre-existing NSP12 responses in 4/5 individuals who had exposure to SARS-CoV-2 through a close contact but who remained seronegative, with three cases showing a more than two-fold expansion ( Fig. 4a-b) . The other five seronegative close contacts had no pre-existing NSP12 responses detectable, but four of these had detectable, presumed de novo, low-level responses after exposure (Fig. 4b) .

Overall, the close-contact seronegative group showed preferential expansion of RTC over structural protein responses comparing their pre-and post-exposure samples ( Fig. 4a-b) . By contrast, the group with serological confirmation of infection also showed the expected in vivo expansion of pre-existing SARS-CoV-2-reactive T cells but this was predominantly of responses directed against structural proteins. Only four out of thirteen in the seroconverted group expanded NSP12 responses and overall, they had no significant increase in RTC-specific T cells (Fig. 4a; Extended Data Fig. 4a ).

We then reverted to the ES HCW cohort, where small volume PBMC collections were available from the time of recruitment, allowing targeted analysis of baseline responses in those with the strongest RTC responses at wk16.

. 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.

The copyright holder for this preprint this version posted July 1, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint up, Fig. 1b ) 16, 43 , with seroconversion within the first 3 weeks of follow-up for most 36 . Focusing on the most common specificity of pre-existing responses, NSP12-specific T cells were already detectable at baseline in 74% of those ES with the strongest NSP12 responses post-exposure ( Fig. 4c) . NSP12 responses expanded in vivo on average 8.4-fold between recruitment and wk16 of follow-up, with no corresponding change in Flu/EBV/CMV responses (Fig. 4d) . We could identify three ES with de novo responses, ten with >2 fold-expansion of NSP12-specific T cells and one with a >2 fold-contraction, in line with their reported likely exposure before recruitment (Fig. 4c) .

Interestingly, all HCW with de novo or expanded/contracting NSP12-specific T cell responses also had NP1 and NP2 reactive T cells after exposure (Extended Data Fig. 4b) ; however, of the five individuals who had no change in NSP12 response only 2/5 had these specificities, suggesting they may not have had the same level of SARS-CoV-2 exposure. The fold-change in NSP12 between recruitment and wk16 follow-up correlates with the total SARS-CoV-2 response, suggesting it may be used as a proxy to identify exposed seronegative individuals who have had expanded T cell immunity after exposure (Extended Data Fig. 4c ).

Finally, we compared the magnitude of memory T cell responses of different specificities at wk16 to see if there was a preferential enrichment of RTC-specific responses in ES HCW compared to the laboratory-confirmed infected HCW. Strikingly, NSP12 was the only viral protein to induce T cells to a higher magnitude in the cohort of seronegative individuals in which a successful infection was not established compared to those with classical laboratoryconfirmed infection (Fig. 4e) . Examining the RTC specificities in more detail (breaking down according to peptide pools of equivalent size, ~40 overlapping 15mers) revealed that T cell responses targeting several regions of the RNA polymerase NSP12 were significantly enriched in the exposed seronegative HCW cohort compared to post-infection, whilst other RTC pools also showed non-significant trends for enrichment (Fig. 4e , lower panel) suggesting that they may have played a role in protecting these individuals from PCR-detectable infection and seroconversion.

In summary, we provide T cell and innate transcript evidence for abortive, seronegative SARS-CoV-2 infection. Pre-existing cross-reactive T cells, in particular against the RNA-polymerase, . 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.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint expanded in vivo following SARS-CoV-2 exposure, and were preferentially enriched in individuals in whom SARS-CoV-2 failed to establish a successful infection, compared to those with classical infection. Taken together, these data highlight a role for the in vivo expansion of pre-existing and de novo RTC-specific T cells in aborting early viral infection before the induction of antibodies.

The first proteins to be transcribed during SARS-CoV-2 replication are those of the RTC within ORF1ab 13 , which may make them effective targets for early viral control. Non-structural proteins of ORF1ab are released into the cytoplasm as part of the viral life cycle 13 , giving direct access to the major histocompatibility complex class (MHC) I presentation pathway to activate CD8+ T cell responses to these regions, as has been shown for other single strand RNA viruses including dengue 30 . The formation of the RTC is essential for subsequent transcription of the viral genome, raising the possibility that some infected cells could be recognised and removed by CD8+ T cells before widespread production of structural proteins and mature virion formation 44 . Whereas live virus is likely to be more effectively presented in MHC class I, exogenous viral antigen, for instance from non-replicative particles, can lead to the priming or activation of CD4+ T cells.

The differential biasing of T cells towards early expressed viral proteins at the expense of humoral responses and T cells targeting structural proteins in HCW not seroconverting may reflect repetitive occupational exposure to very low viral inocula, as has been reported in HIV and SIV 27,32,45 . Consistent with this hypothesis, we did note some de novo induction of T cells not detectable prior to exposure in the ES. However, we also documented expansion of preexisting T cells against the highly conserved polymerase, with responses capable of crossrecognising epitope variants between seasonal HCoV and SARS-CoV-2. Such pre-existing T cells, at higher frequency than naïve T cells and poised for immediate re-activation on encountering their antigen, would be expected to favour abortive infection. HCW are particularly prone to exposure to respiratory pathogens 46-48 and have higher frequencies of HCoV-reactive T cells than the general public 19 . Recent HCoV infection is associated with reduced risk of severe COVID-19 infection 49 , likely partly attributable to cross-reactive . 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.

The copyright holder for this preprint this version posted July 1, 2021. Essential viral proteins, such as the RTC, that have less scope to mutate whilst retaining functionality, are more conserved across the Coronaviridae family than structural regions 40 and therefore retain T cell epitopes. Preferential involvement of pre-existing responses, dominated by RTC-specific T cells, in early control would explain their enrichment after abortive infection compared to classical infection.

Although we have shown an association between the presence of certain T cell specificities, in particular to areas of high conservation within HCoV such as NSP12, and resistance to overt infection in exposed HCW, larger cohorts or human challenge studies will be needed to determine their relative contribution to protection. The antiviral potential of CD8+ T cells in SARS-CoV-2 is supported by depletion experiments in macaques 54 and by the resolution of infection in patients lacking humoral immunity because of agammaglobulinemia or B cell depletion therapy 55,56 . It remains possible that innate control alone can mediate abortive infection, with low level antigen production being enough to generate RTC-biased T cell responses simply as a biomarker of low-grade infection. The fact that the ES cohort had much lower levels of IFI27, reflective of less interferon induction, than the infected cohort does not support preferential innate control in the former. However, interferon-independent induction of RIG-I has been proposed as another potential mechanism of aborting SARS-CoV-2 infection 15 . A further caveat in the interpretation of our findings is that we only analysed . 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.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint peripheral immunity; it is plausible that mucosal-sequestered antibodies, recently reported in seronegative HCW 57 , could have played a role in our seronegative cohort.

We have described an under-investigated host-pathogen interaction leading to the induction of innate and cellular immunity without seroconversion. These data will . 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.

The copyright holder for this preprint this version posted July 1, 2021. 

The COVIDsortium bioresource was approved by the ethical committee of UK National

Research Ethics Service (20/SC/0149) and registered on ClinicalTrials.gov (NCT04318314). Full study details of the bioresource (participant screening, study design, sample collection, and sample processing) have been previously described 21,58 .

In this cohort and London as a whole infections peaked during the first week of lockdown (March 23 rd 2020) 43 , and we observed approximately synchronous exposure coincident with recruitment, we therefore used this as the benchmark for assessing exposure generated immunity. Across the main study cohort, 48 participants had positive RT-PCR results with 157 (21.5%) seropositive participants. Infections were asymptomatic or mild with only two hospital admissions (none requiring intensive care admission). The cross-sectional case controlled sub-study (n=129) collected samples at 16-18 weeks after the first UK lockdown (Fig. 1a) . Power calculations were performed prior to week 16 sub-study sampling to determine the sample size needed to test the hypothesis that HCW with pre-existing T cell responses are enriched in exposed uninfected group at a range of incidence of infection, assuming 50% of cohort had pre-existing T cell responses. Sample sizes of 18-64 per group were estimated. An age, sex and ethnicity matched nested sub study was designed within the . 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.

The copyright holder for this preprint this version posted July 1, 2021. and anti-nucleocapsid total antibody assay (ROCHE) described in detail below. The exposed seronegative health care worker cohort were matched for demographics and exposure to the lab confirmed infected cohort and was defined by negativity by these three tests at all 16 time points as well as negative for neutralising antibodies at week 16 and at selected prior time points as indicated.

The cohort of medical students and laboratory staff was approved by UCL Ethics (Project ID Number: 13545/001) and pre-pandemic healthy donor samples were collected and cryopreserved before August 2019 under ethics numbers 11/LO/0421. All subjects gave written informed consent and the study conformed to the principles of the Helsinki Declaration.

Peripheral blood mononuclear cells (PBMC) were isolated from heparinized blood samples using Pancoll (Pan Biotech) or Histopaque®-1077 Hybri-MaxTM (Sigma-Aldrich) density gradient centrifugation in SepMate tubes (StemCell) according to the manufacturer's specifications. Isolated PBMCs were cryopreserved in fetal calf serum containing 10% DMSO and stored in liquid nitrogen.

Whole blood samples were collected in SST vacutainers (VACUETTE) with inert polymer gel for serum separation and clot activator coating. After centrifugation at 1000 X g for 10 min at room temperature (RT), serum layer was aliquoted and stored at -80 °C. All T cell assays reported here were performed on cryopreserved PBMC.

. 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.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint Weekly Euroimmun anti-SARS-CoV-2 enzyme-linked immunosorbent assay (ELISA; anti-SARS-CoV-2 S1 antigen IgG and the Roche Elecsys anti-SARS-CoV-2 electrochemiluminescence immunoassay (ECLIA; anti-SARS-CoV-2 nucleoprotein IgG/IgM) commercial assays were performed by Public Health England as previously described 21 . S1 ELISA: A ratio of ≥ 1.1 was deemed positive. A ratio of 11 was taken to be the upper threshold as the assay saturates beyond this point. NP ECLIA: Anti-NP results are expressed as a cut-off index (COI) value based on the electrochemiluminescence signal of a two-point calibration, with results COI ≥1.0 classified as positive.

SARS-CoV-2 pseudotype neutralisation assays were conducted using pseudotyped lentiviral particles as previously described 21 . Briefly, serum was heat-inactivated at 56 o C for 30 mins. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint (BMG Labtech). Absorbance readings for each well were standardised against technical positive (virus control) and negative (cells only) controls on each plate to determine a percentage neutralisation value. A non-linear regression (curve fit) method was used to determine the dilution fold that neutralised 50% (IC50) using Prism 9 (GraphPad). SARS-CoV-2 is classified as a hazard group 3 pathogen and therefore all authentic SARS-CoV-2 propagation and microneutralization assays were performed in a containment level 3 facility.

Seropositivity against SARS-CoV-2 spike was determined for medical student and laboratory staff cohort between July 2020 and Jan 2021 (Extended Data Table 4 ) by enzyme-linked immunosorbent assay, as validated and described previously 50,60,61 . Briefly, 9 columns of 96half-well MaxiSorp plates (Thermo Fisher Scientific) were coated overnight at 4 °C with purified S1 protein in PBS (3 μg/ml per well in 25 μl), the remaining 3 columns were coated with goat anti-human F(ab)'2 (1:1,000) to generate in internal standard curve. The next day, plates were washed with PBS-T (0.05% Tween in PBS) and blocked for 1 hr at RT with assay buffer (5% milk powder PBS-T). Sera were diluted in blocking buffer (1:50). 25 ul of serum was then added to S1 coated wells in duplicate and incubated for 2 hr at RT. Serial dilutions of known concentrations of IgG were added to the F(ab)'2 IgG-coated wells in triplicate (Sigma Aldrich). Following incubation for 2 hr at RT, plates were washed with PBS-T and 25 µl alkaline phosphatase-conjugated goat anti-human IgG (Jackson ImmunoResearch) at a 1:1000 dilution in assay buffer added to each well and incubated for 1 hr RT. Plates were then washed with PBS-T, and 25 µl of alkaline phosphatase substrate (Sigma Aldrich) added. ODs were measured using a MultiskanFC (Thermofisher Scientific) plate reader at 405 nm and S1specific IgG titers interpolated from the IgG standard curve using 4PL regression curve-fitting on GraphPad Prism 8. Table 3) .

To limit competition for in vitro for peptide presentation we limit stimulations to a maximum of 55 peptides and have, therefore, divided large proteins such as NP into sub-pools: NP (NP1, NP2, 41 peptides each), M (43 peptides), ORF3a (53 peptides), NSP7 (15), NSP12 (36-37 per pool NSP12-1 to NSP12-5) and NSP13 (39-40 peptides per pool NSP13-1 to NSP13-3). In addition 15-mer peptides covering the predicted SARS-CoV-2 spike epitopes 10 to give a total of 55 peptides in this pool (Spike). Optimal 9mer peptides for CD8+ epitopes were custom synthesised by ThinkPeptides (UK) >70% purity.

. 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.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint

IFNγ-ELISpot Assay was performed as previously described on cryopreserved PBMC 10,21,62 .

Unless otherwise stated, culture medium for human PBMC was sterile 0.22 μM filtered RPMI medium (Thermo Fisher Scientific) supplemented with 10% by volume heat inactivated (1 hr, 64 °C) fetal calf serum (FCS; Hyclone, and 1% by volume 100 x penicillin and streptomycin solution (GibcoBRL).

ELISpot plates (Merck-Millipore, MSIP4510) were coated with human anti-IFNγ Ab (1-D1K, Mabtech; 10 μg/ml) in PBS overnight at 4 °C. Plates were washed 6x with sterile PBS and were blocked with R10 for 2 hr at 37 °C with 5% CO2. PBMC were thawed and rested in R10 for 3 hr at 37 °C with 5% CO2 before being counted to ensure only viable cells were included. 400,000 PBMC were seeded in R10/well and were stimulated for 16-20 hr with SARS-CoV-2 peptide pools (2 μg/ml/peptide) at 37 °C in a humidified atmosphere with 5% CO2. Where insufficient cells were available NSP12 pools 1,2 and 3 and NSP13 pools 1,2,3 were combined into a single well. HCW who did not have a full complement of stimulations were excluded from analysis of total magnitude of breadth of response, hence slightly lower n numbers.

Internal plate controls were R10 alone (without cells) and two DMSO wells (negative controls), concanavalin A (ConA, positive control; Sigma-Aldrich) and FEC (HLA I-restricted peptides from influenza, Epstein-Barr virus, and CMV; 1 μg/ml/peptide). ELISpot plates were developed with human biotinylated IFN-γ detection antibody (7-B6-1, Mabtech; 1μg/ml) for 3 hr at RT, followed by incubation with goat anti-biotin alkaline phosphatase (Vector Laboratories; 1:1000) for 2 hr RT, both diluted in PBS with 0.5% BSA by volume (Sigma-Aldrich), and finally with 50 μl/well of sterile filtered BCIP/NBT Phosphatase Substrate (ThermoFisher) for 7 min RT. Plates were washed in ddH20 and left to dry overnight before being read on an AID classic ELISpot plate reader (Autoimmun Diagnostika GMBH, Germany).

The average of two DMSO wells was subtracted from all peptide-stimulated wells for a given PBMC sample and any response that was lower in magnitude than 2 standard deviations of these sample specific DMSO control wells was not considered a peptide specific response (given value 0). Results were expressed as IFNγ spot forming cells (SFC) per 10 6 PBMC after background subtraction. The geometric mean of all DMSO wells was 9.571 SFC per 10 6 PBMC . 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.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint (3.8 spots). We excluded the results if negative control wells had >95 SFC/10 6 PBMC or positive control wells (ConA) were negative. T cell responses to SARS-CoV-2 did not correlate with background spots in DMSO wells (e.g. ES cohort spearman r = -0.068 p = 0.6141).

Frozen PBMC were thawed and washed twice with sterile PBS. PBMC were resuspended in 1 mL R10 culture media (2-10 x 10 6 PBMC) and 0.5 µL of 5 mM stock CellTrace violet (CTV; Thermo Fisher Scientific) was added per sample with mixing. PBMC were stained in the dark for 10 mins at 37 °C in a humidified atmosphere with 5% CO2. Ten-times volume of cold R10 was added to stop the staining reaction, and cells were incubated for 5 mins on ice. Cells were . 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.

The copyright holder for this preprint this version posted July 1, 2021. Optimisation experiments showed use of rhIL2 increases non-peptide specific proliferation of T cells but is essential for optimal expansion of proliferating cytokine producing peptidespecific T cells. CTV dilution and staining with anti-human-IFN antibodies was used to identify antigen-specific T cells. An unstimulated control well (equivalent DMSO to peptide wells added) was included for each PBMC sample and the percentage of CTV lo IFNγ+ CD4+ or CD8+ proliferating was subtracted from all peptide stimulated wells. T cell responses <0.1% of CD4 or CD8 T cells after 10-day peptide expansion or of less than 10 cells were excluded from analysis.

The T cell proliferation assay above was used to expand SARS-CoV-2-specific T cells and a 2dimension matrix (Supplementary Table 2 ) was employed so that each 15mer peptide was represented in two pools aiding the identification of individuals immunogenic 15mer peptides. T cell responses were then confirmed by repeated expansion with individual 15mers.

The sequence homology of SARS-CoV-2-derived peptides to HCoV sequences (Fig. 3a) was computed as previously described 40 . Briefly, the SARS-CoV-2 proteome (NC_045512.2) was decomposed into 15mer peptide sequences overlapping by 14 amino acids. A protein BLAST search of each 15mer peptide was then performed against a custom sequence database comprising 2531 Coronaviridae sequences 40 . Homology values of each SARS-CoV-2-derived peptide to viral accessions with '229E', 'OC43', 'NL63', or 'HKU1' included in the species name and that were isolated from human hosts were retained (Supplementary Table 1) .

Additionally, to determine if the conservation of 15mer peptides differed between the SARS-CoV-2 proteins, the average homology of peptides within each protein was computed. A permutation test was conducted to test if the difference in average homology between the two proteins, Δh, was statistically significant. Briefly, the protein membership of each 15mer peptide was permuted (1000 iterations). The Δh of two proteins were then calculated at each iteration, resulting in a final null distribution of Δh values. P-values were computed as the number of permutations that yielded a Δh at least as extreme as the observed Δh of the two proteins. Custom scripts used to perform the homology searches, heatmap visualisation and permutation testing are hosted on GitHub . 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.

The copyright holder for this preprint this version posted July 1, 2021. 

Power calculations were used to estimate the sample size needed for week 16 sub study (see above). No statistical methods were used to predetermine sample size. For all assays samples from each cohort were run in parallel to reduce the impact of inter-batch technical variation.

. 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.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint IFN -ELISpot assays were performed on HCW cohorts prior to unblinding of group (Laboratory-confirmed-infection or exposed seronegative). Other experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment.

All data analysed during this study are included in this published article (and its supplementary information files). Custom scripts used to perform the homology searches, heatmap visualisation and permutation testing are hosted on GitHub 

We are extremely grateful to all patients and control volunteers who participated in this study and to all clinical staff who helped with recruitment and sample collection. We are grateful to Jamie Evans at the Rayne Building FACS facility for assistance with Flow cytometry assays. 

Extended Data Fig. 1 SARS-CoV-2 immunity in exposed seronegative healthcare workersauthentic virus (Wuhan Hu-1) neutralisation and T cell response in those with NP1+NP2 responses. a, authentic virus neutralisation at 3 time-points, n=6. b, Example plots of SARS-COV-2 spike memory B cell staining (gated on: lymphocytes, singlets, Live, CD3-CD14-CD19+, CD20+, excluding CD38hi, IgD+ and CD21+CD27-fractions) in an exposed seronegative HCW at wk16 and a seropositive individual. c, Magnitude of T cell response coloured by viral protein in ES with T cells reactive against both NP1 and NP (left) and against one of or neither NP1 or NP2 pools (right) at wk16. d, Summed response to RTC and structural regions of SARS-CoV-2 in pre-pandemic samples and ES with and without NP1+NP2-reactive T cell responses at wk16. Kruskal-Wallis with Dunn's correction. Bars, geomean. NP, nucleoprotein.

. 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.

The copyright holder for this preprint this version posted July 1, 2021. . 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.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint Extended Data Fig. 3 Functional and proliferative SARS-CoV-2 specific T cells in exposed seronegative HCW. a, example gating of CTV stained PBMC after 10-day peptide stimulation: Lymphocytes (SSC-A vs. FSC-A), single cells (FSC-H vs. FSC-A), Live cells (fixable live/dead-), CD3+, CD4+ or CD8+. Second row: Gated on CD8+ showing cytokine combinations. Response to immunodominant MHC class I-restricted peptide pool against Flu, EBV, CMV (FEC) in exposed seronegative HCW. b, Correlation between the magnitude of T cells responses to SARS-CoV-2 pools or FEC after 10-day in vitro expansion (% dual staining for two anti-human IFN mAb clones, unexpanded responses <0.1% of CD3 post-expansion excluded) and ex vivo IFN -ELISpot in exposed seronegative HCW. Spearman r. c, Proportion of SARS-CoV-2-specific T cells (CTV-IFN +) that are CD4+ or CD8+ after 10-day expansion (where sub pools were used, they are indicated below bar). d, Example 2D-mapping matrix after 10-day expansion with NSP12-3 peptide pool in an exposed seronegative HCW (Antigen-specific identified as CTV lo IFN -APC+). e, Alignment of Coronaviridae consensus sequences at immunogenic 15mers peptides. Conserved amino acids in yellow. a-d ES at wk16.

. 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.

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint Extended Data Fig. 4 In vivo expansion of pre-existing SARS-CoV-2-reactive T cells postinfection or post-exposure. a, Change in magnitude of T cell response between pre-pandemic and post-infection samples (upper panel: all proteins, lower panel: NSP12) from seropositive medical students and laboratory staff. b, Proportion of ES with NP1 + NP2-reactive T cells grouped by those with and without de novo or expanded NSP12 responses at wk16. c, Correlation between the fold-change in NSP12 between recruitment and wk16 and total response to RTC or structural proteins at wk16 in ES. Spearman r. Table 2 : CTV T cell proliferation in exposed seronegative HCW. . 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)

The copyright holder for this preprint this version posted July 1, 2021. . 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)

The copyright holder for this preprint this version posted July 1, 2021. ; https://doi.org/10.1101/2021.06.26.21259239 doi: medRxiv preprint

Early induction of functional SARS-CoV-2-specific T cells associates with rapid viral clearance and mild disease in COVID-19 patients

Airway Memory CD4+ T Cells Mediate Protective Immunity against Emerging Respiratory Coronaviruses

Correlates of protection against SARS-CoV-2 in rhesus macaques

Generation of a Broadly Useful Model for COVID-19 Pathogenesis, Vaccination, and Treatment

Antigen-Specific Adaptive Immunity to SARS-CoV-2 in Acute COVID-19 and Associations with Age and Disease Severity

Robust T Cell Immunity in Convalescent Individuals with Asymptomatic or Mild COVID-19

Selective and cross-reactive SARS-CoV-2 T cell epitopes in unexposed humans. Science (80-. )

SARS-CoV-2-derived peptides define heterologous and COVID-19-induced T cell recognition

SARS-CoV-2-reactive T cells in healthy donors and patients with COVID-19

SARS-CoV-2-specific T cell immunity in cases of COVID-19 and SARS, and uninfected controls

Targets of T Cell Responses to SARS-CoV-2 Coronavirus in Humans with COVID-19 Disease and Unexposed Individuals

Characterization of pre-existing and induced SARS-CoV-2-specific CD8+ T cells

Coronaviruses post-SARS: Update on replication and pathogenesis