key: cord-0992443-hlyftlv9
authors: Qi, H.; Ma, M.-l.; Jiang, J.; Ling, J.-y.; Chen, L.-y.; Zhang, H.-n.; Lai, D.-y.; Li, Y.; Guo, Z.-w.; Hu, C.-s.; Guo, S.-j.; Meng, Q.-f.; Ren, Y.; Wang, W.; Yang, X.; Zhou, J.; Zhao, X.-d.; Li, H.; Tao, S.-c.
title: Systematic profiling of SARS-CoV-2 specific IgG epitopes at single amino acid resolution
date: 2020-09-09
journal: nan
DOI: 10.1101/2020.09.08.20190496
sha: 467b21151c8d0d3b417019d6f9d4e19ac9ebc506
doc_id: 992443
cord_uid: hlyftlv9

SARS-CoV-2 specific IgG responses play critical roles for patients to recover from COVID-19, in-depth dissecting of the IgG responses on systems level is of great interest. Herein, we adopted a newly developed high-throughput epitope mapping technology (AbMap), analyzed 55 COVID-19 convalescent sera and 226 antibody samples enriched by specific proteins or peptides from these sera. We revealed three areas that are rich of IgG epitopes, two are on Spike protein but outside of RBD, and one is on Nucleocapsid protein. We identified 29 significant epitopes on Spike protein, from two of these significant epitopes, two critical epitope residues were found, i. e., D936 and P1263, which are highly related to the infectivity of SARS-CoV-2. In summary, we provided the first global map of IgG binding epitopes for SARS-CoV-2 at single amino acid resolution. This map will facilitate the precise development of therapeutic antibodies and vaccines.

COVID-19 is a worldwide pandemic and caused by SARS-CoV-2 (1, 2). By September 8, 2020 , 27,249,308 cases were diagnosed and 890,971 lives were claimed (https://coronavirus.jhu.edu/map.html) (3) . Vaccines and therapeutic antibodies of high effectivity are desperately needed. It is known that SARS-CoV-2 encodes 27 proteins, of these proteins, Spike protein (S protein) plays a critical role in the specific binding of SARS-CoV-2 to human cells. During the infection, specific proteases cut S protein at S2' and furin site to split S1 and S2 (4) . S protein is rich of N-glycosylation, i. e., 21 sites (5).

On S protein, RBD domain is presently the central target to develop SARS-CoV-2 antibodies and vaccines (6) (7) (8) (9) (10) . Other areas of S protein can also elicit B cell responses and neutralization antibodies (11) (12) (13) (14) (15) . Nucleocapsid protein (N protein) is highly abundant and triggers significant IgG response (16) (17) (18) .

SARS-CoV-2 specific IgG response is critical for the recovery of COVID-19 patient, since specific and effective therapy is still not available. The comprehensive and in-depth revealing of SARS-CoV-2 specific IgG responses will definitely help us to understand the COVID-19 immunity better, and to facilitate the precise development of neutralization antibodies and vaccines. To reveal the SARS-CoV-2 specific IgG response, several studies have been performed through different angles by applying different strategies. By analyzing 55 convalescent sera on a SARS-CoV-2 proteome microarray, the IgG responses against 18 SARS-CoV-2 proteins were surveyed, including S1, S2, N and RBD (16) . By analyzing sera from 1,051 patients on a S protein peptide microarray of full S protein coverage, several areas and a set of peptides of high IgG reactivity were identified (14) . Similar studies were also performed by using another peptide microarray (17) , and a coronavirus specific phage displayed peptide library (19) . The IgG responses on protein and peptide levels, but not single amino acid level, were at least partially revealed by these studies.

Herein, we aimed to go one step further to dissect the SARS-CoV-2 specific IgG responses systematically at single amino acid level. We adopted AbMap, a method that was developed recently (20) for high-throughput epitope mapping. A set of 55 convalescent sera and 226 samples of protein/ peptide enriched antibodies from these sera . 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint were analyzed. A map of SARS-CoV-2 specific IgG epitopes at single amino acid resolution was generated for the first time.

To globally profile the IgG binding epitopes on SARS-CoV-2 proteins from COVID-19 sera, we collected sera from 55 convalescent patients, and 25 controls. The patients were hospitalized in Foshan Fourth hospital in China from 2020-1-25 to 2020-3-8 for various durations. Patient information is summarized in Table S1 . Serum from each patient was collected on the day of hospital discharge when standard criteria were met.

It is known that serum is rich of antibodies (7, 21) , and it is reasonable to argue that SARS-CoV-2 specific IgGs will compose only a small portion of the total IgGs in serum.

In order to increase the signal to noise ratio of the following-up epitope mapping, the sera from the 55 convalescent patients were subjected for antibody enrichment by specific SARS-CoV-2 proteins and peptides. The three layers of samples, i. e., the serum, the protein enriched antibodies, and the peptide enriched antibodies were then subjected for IgG binding epitope mapping by AbMap, a technology that we have established recently (20) . Statistically significant epitopes were successfully identified for all these three layers of samples (Fig. 1) . A variety of epitopes identified from the protein and peptide enriched antibodies could be matched to the sequence of the corresponding proteins and peptides. These results indicate AbMap is suitable for dissecting the SARS-CoV-2 IgG binding epitopes using COVID-19 sera as input.

To reveal the epitopes at systems level, the 55 convalescent sera were subjected for AbMap analysis directly, following the procedure that we established recently (20) and used a high stringent cutoff, we have identified 418 motifs ( Table S2) , 275 of which could be matched to 27 of the 28 known SARS-CoV-2 proteins (Table S3, Fig. 2A) , we recognized these motifs as epitopes of high confidence. Vertically ( Fig. 2A) , to estimate . 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint the variation of SARS-CoV-2 specific IgG responses among the 55 convalescent sera, we added up the epitope frequency for each patient. The results clearly showed that there are significant differences among the patients, with a frequency from 0 to 27. According to the histogram of the frequency (Fig. 2B) , we noticed that the number of patients is inversely proportional to the frequency, this suggest the SARS-CoV-2 specific IgG response is not randomly distributed among the patients, the underlying mechanism may be worth further exploration. Horizontally ( Fig. 2A) , to check the distribution of the epitopes among the proteins, we also counted the epitopes for each protein, i. e.,

sum_epitopes. There are big variations of the sum_epitopes from 0 for ORF6 to 125 for NSP3 (Fig. 2C) . Since the size of the 28 SARS-CoV-2 proteins varied a lot from 22 aa (ORF3b) to 1,945 aa (NSP3), it is possible that the variations of the sum_epitopes may be due to the different sizes of the proteins. To check this possibility, we divided sum_epitopes by the length of the proteins, indeed, the variation of the sum_epitopes among protein was largely normalized. However, there are still significant variations for sum_epitopes/ length among the proteins, from 0 for ORF6 to 0.08 for NSP9, these differences may be due to the accessibility of these proteins to the immune system during infection, further study may be needed to address this.

Except the SARS-CoV-2 specific antibodies, there are many other antibodies in the convalescent sera, it is possible that a big portion of the 143 un-matched motifs may correspond to the un-related antibodies (22). To reveal more SARS-CoV-2 epitopes of high confidence, it is necessary to improve the signal-to-noise ratio of AbMap. For this purpose, we decided to enrich protein specific antibodies for AbMap. According to one of our previous study (16) , we knew N protein, S1, S2 and RBD are of high reactivity among patients. Because we only had very limited amount of sera, we decided to enrich the antibodies from each sample in a consequential manner, i. e., RBD, S1, S2 and N. To ensure the quality of the enrichment, proteins that biotinylated through Avi-tag and streptavidin coated magnetic beads were adopted (Fig. S1) . Clearly, a variety of highly confident epitopes were identified and could be matched to the sequences of the corresponding proteins (Table S3) .

. 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint Thus, we generated a map of SARS-CoV-2 specific IgG binding epitopes by analyzing 55 convalescent sera and 226 samples of antibodies enriched by specific proteins from these sera.

Two areas outside of RBD that rich of significant epitopes were identified on Spike protein S protein is the key protein for infection and is known to be highly immunogenic (15, 16) .

To present the distribution of the epitopes on S protein, we merged the sequence matched epitopes revealed from both S1 and S2 enriched antibodies (Table S3) . When matching the epitopes to S protein sequence, if an epitope sequence obtained from one sample is the same for other samples, we counted this epitope as 1 for this sample and added it up for all the samples as the final frequency of this epitope. We then plotted the epitopes, the epitopes sequences, and the frequencies alongside the linear sequence and domains of S protein (Fig. 3A) . The results clearly showed that the epitopes are distributed cross but not evenly on S protein. Two hot areas that rich of significant epitopes were readily identified, the first area almost covers the entire CTD (C terminal domain), the second area covers the S2' protease cleavage site and the fusion peptide (FP). Significant epitopes were also identified on the cytoplasmic C-terminal end of S protein. These results are highly consistent to a peptide microarray-based study, of which about 4,000 samples were analyzed against a set of 197 peptides that covers the entire S protein (14, 15) .

It is interesting to test the necessity of enriching SARS-CoV-2 specific antibodies from sera. To this end, we took S protein as an example, we plotted the epitopes identified by using serum directly ( Fig. S2 black dots) and the S protein enriched antibodies ( Fig. S2 red dots) alongside the sequence of S protein. The distributions are similar between sera and S protein enriched antibodies, as expected, many more significant epitopes with higher frequency were identified by using the S protein enriched antibodies (Fig. S2) .

This improvement may be due to the removal of unrelated antibodies from sera, thus the signal-to-noise ratio was significantly enhanced.

. 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint

We defined the epitopes with frequency >= 3 as significant epitopes and obtained 28 of them (Fig. S3A) . To further illustrate the location and distribution of the 28 significant epitopes (Fig. S3A) , we mapped them to the 3D structure of Spike protein (23) monomer ( Fig. 3B) and trimer (Fig. S3B) . It is clear that most of these epitopes locate on the surface of Spike protein, which is consistent to the common notion (24). However, Epitope-S1-6 (KCVNFNFN) and Epitope-S2-3 (QEVFAQVK) are not on the surface of the trimeric Spike protein, but on the surface of the monomer. A plausible explanation is that specific areas of the Spike protein monomer could be exposed to the immune system during the infection. To further pinpoint the locations of the critical residues of the significant epitopes, we took 4 critical epitope residues, i. e., K558, P561, Q563 and Q564 from a significant epitope as examples (Fig. 3C) . As expected, all these residues locate on the surface of S protein (Fig. 3D) . To check the similarity of the epitopes/ critical epitope residues among human coronaviruses, we performed homology analysis for Epitope-S1-8 (KFLPFQQF) (Fig. S3A) . High homologies were observed among SARS-Cov-2, SARS-CoV and BtCoV-RaTG13 (Fig. 3E) .

Structure analysis and homology analysis were also performed for all the rest 27 significant epitopes wherever applicable (Table S4 ). Similar to that of Epitope-S1-8 (KFLPFQQF), most of the critical epitopes of these 27 significant epitopes also locate on the surface of S protein. High homologies were also observed among SARS-Cov-2, SARS-CoV and BtCoV-RaTG13 for most of the epitopes. In addition, an epitope of high homology among BtCoV-RaTG13, the three deadly human coronaviruses (SARS-CoV-2, SARS-CoV and MERS-CoV) and four common human coronaviruses (HCoV-OC43, HKU1, NL63 and 229E) was also identified, i. e., FxxELxxY (Table S4) .

These results indicate that the critical epitope residues of the significant epitopes are unevenly distributed on S protein both linearly and conformationally. And the critical epitope residues are highly homologous among the 7 human coronaviruses, especially between SARS-CoV-2 and SARS-CoV.

Critical epitope residues related to the binding of Spike protein and ACE2/ neutralization antibodies, and the infectivity of SARS-CoV-2 were revealed . 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint

Because the central role that RBD plays in binding of SARS-CoV-2 to the host cell, biotinylated RBD was also applied to enrich specific antibodies from COVID-19 convalescent sera (Fig. S1 ). One relative significant epitope was identified (Fig. 4A) .

This epitope is also consistent to our peptide microarray-based study (14) . To visualize the conformational location of this epitope, we mapped the critical epitope residues (yellow), i. e., R466, D467, I468, S469, E471 and E473 to the 3D structure of RBD (25) (Fig. 4B) . It is clear that the critical epitope residues are not on the exact binding interface of RBD and ACE2 but adjacent (Fig. 4C) . To further investigate whether these critical residues interference with the binding of neutralization antibodies to RBD, we studied the co-crystal structures of RBD bound by several well-studied neutralization antibodies, i. e., CR3022-Fab (8) (Fig. 4D) , S309-Fv (26) (Fig. 4E) , P2B-2F6-Fab (27) ( Fig. S4A) , BD23-Fab (7) (Fig. S4B) , and CB6-Fab (28) (Fig. S4C) . Interestingly, the critical residues are also not on the exact binding interface of RBD and these neutralization antibodies but adjacent. The antibodies targeting this epitope may cause conformational change or providing spatial hinderance, and thus interference the binding of ACE2 and these neutralization antibodies to RBD.

Some of the epitopes and the critical epitope residues may be related to the proper function of S protein and the infectivity of SARS-CoV-2. A recent study surveyed the functional roles of a variety of SARS-CoV-2 S protein mutants, and identified a set of point mutations that cause significant reduction of the infectivity (29). To explore the roles of the critical epitope residues, we matched the mutation sites with the epitope map of S protein (Fig. 3A) , we found three mutation sites are also critical epitope residues, i. e., H49Y, D936Y and P1263L. Except H49Y, D936Y and P1263L could cause 60-70% and 98-99% infectivity loss, respectively. This strongly suggest that D936 and P1263 are critical for maintaining the proper activity of S protein. According to the 2019 Novel Coronavirus Resource (2019nCoVR) database of China National Center for Bioinformation (CNCB) (https://bigd.big.ac.cn/ncov/), those three mutation sites are also of high frequency (Fig. S4D) . Taken together, it is reasonable to argue that D936 and P1263 may could serve as very promising targets for neutralization antibody and vaccine development.

. 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint One area rich of significant epitopes was determined on Nucleocapsid protein N protein is highly abundant (30, 31) and triggers profound IgG response in COVID-19 patients (18, 32) . To present the distribution of the epitopes on N protein, we matched the epitopes to N protein sequence. Following the same procedure as that of S protein (Fig.   3A) , we calculated the frequency of each epitope, and plotted the epitopes, the epitopes sequences, and the frequencies alongside the linear sequence and domains of N protein.

The results clearly showed that the epitopes are distributed cross but not evenly on N protein. One hot area rich of significant epitopes were readily identified (Fig. 5A) , and 6 significant epitopes were identified when set the cutoff as frequency >= 3 (Fig. 5B) .

Because the intact 3D structure of N protein is still not available, we thus decided only to perform homology analysis for N protein. High homologies for the epitopes and the critical epitope residues were observed among SARS-Cov-2, SARS-CoV and BtCoV-RaTG13 for all the 6 significant epitopes (Fig. 5C ).

To validate the key residues of the significant epitopes, according to the frequency and the critical role of RBD, we selected four epitopes, i. e., Epitope-S1-8 (KFLPFQQF) and

Epitope-S1-12 (LTPTWRVY), which are of the highest frequency, Epitope-RBD-1 (RDISTEIY) from RBD, and an epitope of low frequency, i. e., Epitope-S2-15

(KGVKLHYT) was included as control. We then mutated the amino acids of these epitopes one by one to alanine, synthesized the mutated peptides, conjugated to BSA, and fabricated a peptide microarray (Fig. S6) . To validate the epitopes on the peptide microarray, we selected 5 sera, from which at least one of the significant epitopes among Epitope-S1-8 (KFLPFQQF), Epitope-S1-12 (LTPTWRVY) and Epitope-RBD-1 (RDISTEIY) was identified by AbMap (Fig. 6A) . The selected sera were then probed on the peptide microarray. For all the three epitopes, when the corresponding samples were tested, significant binding signal loss was observed when any of the critical residue was . 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint mutated to alanine. For example, for P44 and Epitope-S1-8 (KFLPFQQF), when any of the four critical residues was mutated, the signal was almost completely lost (Fig. 6B) .

As expected, this is not the case for the unrelated samples (Fig. 6B , black caption).

However, the binding intensities varied significantly among samples and epitopes.

It is interesting to see whether the enrichment of SARS-CoV-2 specific antibodies by proteins will enhance the specificity and signal intensity, when probed on the peptide microarray. To test this, we chose Epitope-S1-8 (KFLPFQQF) as an example, probed sera and the corresponding S1 enriched antibodies side by side. As demonstrated in Fig.   S5 , specific bindings were obtained for all the samples. And as expected, the antibody enrichment procedure could significantly enhance both the specificity and the signal intensity. These results further confirmed the necessity for the enrichment of antibodies from sera by specific proteins for epitope mapping.

Taken together, by combining alanine scanning and peptide microarray, the significant epitopes and critical epitope residues could be effectively validated in a fast way.

The aim of this study is to promote our understanding of SARS-CoV-2 specific IgG response in a more precise way. We adopted AbMap technology (20), analyzed 55 sera and 226 antibody samples enriched from these sera by RBD, S1, S2 and N proteins.

Based on these results, we constructed the first map of SARS-CoV-2 specific IgG binding epitopes at single amino acid resolution.

In this study, by applying AbMap, we developed a two-step strategy for profiling pathogen specific epitopes at single amino acid resolution from sera. IgG response is a central part of B cell immunity. Serum or plasma are rich of IgGs for numerous targets that the immune system has encountered, especially when our body is infected by pathogens (22). There is great interest to profile pathogen specific IgG responses by using serum or plasma because of the convenience of sample collection. Protein microarray could be applied to reveal the pathogen specific IgG responses at protein level (16, 33) , . 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint while peptide microarray (15) or pathogen specific peptide library (19, 22) were able to provide information at peptide/epitope level. However, these platforms could not provide epitope information at amino acid level, and pathogen specific protein or peptide library need to be constructed individually, which is costly and time-consuming. Alternatively, random peptide libraries were attempted for dissecting pathogen specific IgG responses from sera (34, 35). For SARS-CoV-2, following the protocol of testing serum directly that we established (34), and the K-TOPE algorithm (35), we tried to profile the virus specific IgG responses at proteome level. However, most of the identified epitopes that match the sequence of SARS-CoV-2 proteins are of low frequency. This indicates a more powerful algorithm is needed to dig meaningful epitopes for the random peptide-based strategy. We then simplified the data analysis part by combining MEME and FIMO (36, 37), and identified a set of epitopes of marginly statistical significance (Fig. 1) . To experimentally improve the performance of AbMap, we tried to reduce the noise by enriching SARS-CoV-2 specific IgGs using specific proteins or peptides, a larger number of high confident epitopes were identified as compared to that of testing serum directly.

Our results clearly point out the directions to improve phage displayed random peptide library-based strategies, such as AbMap, for pathogen specific epitope mapping: 1.

Enrich IgGs by protein/ peptide of interest, and 2. Develop algorithm of better performance.

The epitopes that we identified in this study are highly consistent with studies from our group and others. As expected, S protein and N protein are among the top list according to IgG responses (16) . High reactivities were also observed for other proteins, for example, NSP3 and NSP12. The different reactivities among the SARS-CoV-2 proteins are correlated to the protein size, but only partially. For example, after normalized by protein size, the highest reactivity was observed for NSP9 (Fig. 2C) . It is expected that there are other parameter that cause the difference of IgG reactivity, such as protein level (30, 31), further study is needed to investigate this.

At peptide/epitope level, our results are also consistent to other studies. Though using a totally different strategy, we found RBD is lack of, while CTD is rich of linear epitope, the other "hot area" of linear epitope is around the S2' protease cutting site and the fusion peptide. These results are highly consistent to that of the S protein peptide microarray . 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint that we constructed (15) . According to a SARS-CoV-2 specific peptide library based study (19) , the highly immunogenic regions of Spike protein were also enriched in CTD, region around S2' and the fusion peptide, and HR2, which are highly consistent to our study. While aa362-399 of N protein which identified as high immunogenic region was also determined in our study, but other highly immunogenic regions were not. The inconsistence may be due to the different sample size and the differences of the detailed protocol. According to a peptide microarray (17), 3 of the 6 Spike protein epitopes identified by microarray for serum IgG were also determined as significant epitopes in our study, and aa366-400 of N protein which identified as a long epitope for serum IgG was also enriched in our study. Compared to these epitope related studies, our study can serve as an independent cross validation because of the high consistence. Furthermore, this study is the first of such that providing global IgG epitope information at single amino acid level.

There are a variety of possible applications for the significant epitopes/ critical residues identified in this study. According to S protein, most of the significant epitopes (Fig. S3A) are on the surface of the 3D structure, it is reasonable to argue that the significant epitopes (including that from S protein) are naturally of high immunogenicity. Thus, for a given application, one can select one or a few epitopes to generate site specific antibody on demand for S protein as well as other proteins. High frequencies were observed for some of the epitopes, e. g., Epitope-S1-8 (KFKPFQQF), it is possible this epitope could be used alone or in combination with other epitope for diagnostics. It is known that areas other than RBD on S protein (13, 38) , and furthermore, other than RBM on RBD (receptor binding motif) (8, 39) may also elicit potent neutralization antibodies. Once determined, the corresponding epitopes could be used directly as vaccine or as immunogen for generating site-specific neutralization antibody. Since the COVID-19 pandemic is unfolding and the sequenced genomes of SARS-CoV-2 is piling up (https://bigd.big.ac.cn/ncov/), undoubtedly, more functionally important amino acid sites will be identified, we can match those sites to the identified critical epitope residues, thus link mutation or SARS-CoV-2 evolution with immune responses. Except for the applications that we discuss here, it is anticipated that more applications of the identified significant epitopes/ critical epitope residues could be explored.

. 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint Taken together, we constructed the first map of SARS-CoV-2 specific IgG binding epitopes at single amino acid resolution. We identified several critical epitope residues on S protein that are highly correlate to the infectivity of SARS-CoV-2, and the binding of 

The authors declare no conflict of interest.

. 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint All data supporting the findings of this study could be found within the paper and its Supplementary information, and are available from the related databases. The raw data of NGS were deposited in the NCBI SRA Archive (https://www.ncbi.nlm.nih.gov/sra) with BioProject accession number: PRJNA660911. The raw data of peptide microarray were deposited in the Protein Microarray Database (PMD) (http://www.proteinmicroarray.cn/) with experiment accession number: PMDE246.

Hospital, Foshan, China. The written informed consent was obtained from each patient.

hospital during the period from 2020-1-25 to 2020-3-8 with variable stay time (Table S1) .

When the patients met the standard criteria of hospital discharge, the serum were collected, processed, and stored by standard protocol. As for the hospital discharge criteria, the vital points were normal body temperature, improved respiratory symptoms, lack of inflammation in the pulmonary imaging, and twice consecutively negative for nuclei acid testing. Sera of the control group from Lung cancer patients and healthy controls were collected from Ruijin Hospital, Shanghai, China. All sera were stored at -80 °C until use.

S2 and N were biotinylated according to the manufacturer's protocol. All the biotinylated proteins were incubated with Dynabeads TM Myone TM streptavidin T1 at room temperature for 1 hr, the excess proteins were removed by washing the magnetic beads with PBST (PBS containing 0.1% tween 20). The four biotinylated proteins, i. e., RBD, S1, S2 and N, were immobilized on the surface of magnetic beads, and labelled as RBD::magnetic beads, S1::magnetic beads, S2::magnetic beads and N::magnetic beads, respectively. To deactivate the potential live viruses in the sample, the serum from COVID-19 patients were incubated at 56 for 1 hr. As shown in Fig. S1 , these sera 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint incubated with Dynabeads TM Myone TM streptavidin T1 to remove the antibodies that bind the free magnetic beads. The sera were incubated consecutively by the RBD::magnetic beads, S1::magnetic beads, S2::magnetic beads and N::magnetic beads at 4 , each for 4 hrs. All the magnetic beads, containing the antibodies for each protein, were washed with PBST to remove the non-specific bindings. The antibodies on the beads were eluted with 50 mM glycine (pH2.8), and neutralized with Tris buffer (pH8.0). The quality and concentration of the protein enriched antibodies were checked and measured by SDS-PAGE. All the enriched antibodies were stored in -80 until in use. The same procedure was followed when enrich antibodies by biotinylated peptides.

We followed the AbMap procedure as described previously (20) Dynabesads TM Protein G was added into each well to capture the antibody and phage complex. After incubated for 4 hrs at 4 , the magnetic beads in each well were collected and washed. The beads in each well were suspended to 15 µL ultrapure water. After boiled at 98 for 10 min, the supernatant was collected.

To introduce the adapter sequence and unique barcode or index for each sample, two rounds of PCR were carried out on the phage lysate by Q5 hot start polymerase. The first round of PCR was performed by using primers XX-S5XX-23R and XX-N7XX-18 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint primer for amplifying the corresponding nucleotide acid of the displayed peptide from the genome of phage, and the remaining sequence was the Illumina index). After checked by electrophoresis, all the PCR products were mixed. The PCR products were purified and used as the template for the second round PCR. In the second round PCR, the unique indexes of Illumina next generation sequencing (NGS) were introduced for each mixture.

After purified by gel extraction, the final sample was ready for NGS.

According to the index and barcode combination, the results of NGS were split and assigned to each sample. For each sample, the data of NGS were trimmed further, and only sequence of 36 bp was left, which corresponded to the 12-mer displayed peptide. All the sequences were translated into peptides, and the frequency of each peptide was counted. If stop codon appeared in the translation process, the corresponding sequence was removed. The enrichment factor of each peptide from the sample were calculated.

The reverse enrichment factor of each peptide was also calculated, of which the maximum value was set as the cutoff of the corresponding sample. The peptides, with the enrichment factors lager than the cutoff was determined as the hits. These hits were subjected for MEME analysis, significant motifs were called by setting E value less than 0.05 as the cutoff (40). The significant motifs were matched to SARS-CoV-2 protein/s by FIMO used default settings (36). RBD/CB6 complex (PDB ID: 7C01) were used to analyze the location of the epitopes on these protein complexes. All the structural analyses were processed by using Pymol.

Peptides were synthesized and conjugated to BSA by GL Biochem Ltd. (Shanghai, China). The concentration of each peptide was diluted to 0.5 mg/mL, 0.25 mg/mL and . 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint 0.125 mg/mL with PBS containing 40% glycerin, respectively. According to the pattern showed in Fig. S6 , all the peptides, with extra controls, such as the blank control, human IgG, human IgM and land marker Cy5-labelled donkey anti-human IgM, were printed on the surface of PATH slide with 2 x 7 identical subarrays by using Super Marathon printer (Arrayjet, UK). The peptide microarray was stored in -80℃ until use.

The assay on the peptide microarray was similar to our previous study (15) 

The fluorescent intensity was extracted by GenePix Pro 6.0 software (Molecular Devices, CA, USA). The signal intensity was defined as the mean of the foreground of each spot, and then averaged the three spots of each peptide. All the Signal intensities were analyzed by the software GraphPad Prism version 7.0.

. 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted September 9, 2020. . 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 September 9, 2020. . 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint used as baits to enrich specific antibodies. SA conjugated magnetic beads were adopted to isolate the antibodies. Figure S2 . Protein based antibody enrichment is necessary for AbMap. S protein specific epitopes identified from sera directly (black dot), and S1+S2 protein enriched antibodies (red dot). Pep 1) , which corresponds to Epitope-S1-8 (KLFPFQQF) was selected as an example. Sample with red label indicates the corresponding epitope was identified from this sample by AbMap. The ratio of signal_intensity (S1 enriched)/ signal_intensity (serum) for each residue was plotted (right).

. 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 September 9, 2020. . https://doi.org/10.1101/2020.09.08.20190496 doi: medRxiv preprint Table S1. Serum samples tested in this study. Table S4 . Structure and homology analysis of significant epitopes. Two significant epitopes. The epitopes were matched to S protein, critical epitope residues were labeled as yellow. Homology analysis of the epitope among three deadly coronaviruses, four common human coronaviruses, and bat coronavirus BtCoV-RaTG13.

. 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 September 9, 2020. . 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) 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity . 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 Qi et al., Table S4 . CC-BY-NC-ND 4.0 International license It is made available under a

A new coronavirus associated with human respiratory disease in China

A pneumonia outbreak associated with a new coronavirus of probable bat origin

An interactive web-based dashboard to track COVID-19 in real time

The proximal origin of SARS-CoV-2

Site-specific glycan analysis of the SARS-CoV-2 spike

A noncompeting pair of human neutralizing antibodies block COVID-19 virus binding to its receptor ACE2

Potent Neutralizing Antibodies against SARS-CoV-2 Identified by High-Throughput Single-Cell Sequencing of Convalescent Patients' B Cells

A highly conserved cryptic epitope in the receptor binding domains of SARS-CoV-2 and SARS-CoV

Broad neutralization of SARS-related viruses by human monoclonal antibodies

Antibody cocktail to SARS-CoV-2 spike protein prevents rapid mutational escape seen with individual antibodies

Isolation of potent SARS-CoV-2 neutralizing antibodies and protection from disease in a small animal model

Two linear epitopes on the SARS-CoV-2 spike protein that elicit neutralising antibodies in COVID-19 patients

A neutralizing human antibody binds to the N-terminal domain of the Spike protein of SARS-CoV-2

Linear epitope landscape of SARS-CoV-2 Spike protein constructed from 1,051 COVID-19 patients

Linear epitopes of SARS-CoV-2 spike protein elicit neutralizing antibodies in COVID-19 patients

SARS-CoV-2 proteome microarray for global profiling of COVID-19 specific IgG and IgM responses

SARS-CoV-2 proteome microarray for mapping COVID-19 antibody interactions at amino acid resolution. bioRxiv

Evaluation of Nucleocapsid and Spike Protein-based ELISAs for detecting antibodies against SARS-CoV-2

ReScan, a Multiplex Diagnostic Pipeline, Pans Human Sera for SARS-CoV-2 Antigens. medRxiv

Pep1-WT Pep1-K558A Pep1-F559A Pep1-L560A Pep1-P561A Pep1-F562A Pep1-Q563A Pep1-Q564A Pep1-F565A Buffer Pep2-WT Pep2-L629A Pep2-T630A Pep2-Pep631A Pep2-T632A Pep2-W633A Pep2-R634A Pep2-V635A Pep2-Y636A h-IgM Cy5-BSA

Pep3-WT Pep3-D1153A Pep3-K1154A Pep3-Y1155A Pep3-F1156A Pep3-K1157A Pep3-N1158A Pep3-H1159A Pep3-T1160A Buffer Pep4-WT Pep4-R466A Pep4-D467A Pep4-I468A Pep4-S469A Pep4-T470A Pep4-E471A Pep4-I472A Pep4-Y4T3A h-IgG Cy5-BSA

Qi et al., Figure S3 No.

1 Epitope-S1-1 AYTNSFTR 27-34 12 2Epitope-S1-2 FRSSVLHS 43-50 43 Epitope-S1-3 VTWFHAIH 62-69 3 4Epitope-S1-4 KQGNFKNL 182-189 4 5 Epitope-S1-5 IYQTSNFR 312-319 4 6 Epitope-S1-6 KCVNFNFN 537-544 3 7Epitope-S1-7 GVLTESNK 550-557 7 8Epitope-S1-8 KFLPFQQF 558-565 34 9Epitope-S1-9 VRDPQTLE 576-583 610 Epitope-S1-10 DQLTPTWR 627-634 511 Epitope-S1-11 QLTPTWRV 628-635 1012