key: cord-0992199-3z0eotp8
authors: Li, Y.; Xu, Z.; Lei, Q.; Lai, D.; Hou, H.; Jiang, H.; Zheng, y.; Wang, X.; Wu, J.; Ma, M.; Zhang, B.; Chen, H.; Yu, C.; Xue, J.; Zhang, N.; Qi, H.; Guo, S.; Zhang, Y.; Lin, X.; Yao, Z.; Sheng, H.; Sun, Z.; Wang, F.; Fan, X.; Tao, S.-c.
title: Antibody landscape against SARS-CoV-2 proteome revealed significant differences between non-structural/ accessory proteins and structural proteins
date: 2020-12-11
journal: nan
DOI: 10.1101/2020.12.08.20246314
sha: b232d6e4d4acf368ce0f51575102cab3b708c54a
doc_id: 992199
cord_uid: 3z0eotp8

The immunogenicity of SARS-CoV-2 proteome is largely unknown, especially for non-structural proteins and accessory proteins. Here we collected 2,360 COVID-19 sera and 601 control sera. We analyzed these sera on a protein microarray with 20 proteins of SARS-CoV-2, built an antibody response landscape for IgG and IgM. We found that non-structural proteins and accessory proteins NSP1, NSP7, NSP8, RdRp, ORF3b and ORF9b elicit prevalent IgG responses. The IgG patterns and dynamic of non-structural/ accessory proteins are different from that of S and N protein. The IgG responses against these 6 proteins are associated with disease severity and clinical outcome and declined sharply about 20 days after symptom onset. In non-survivors, sharp decrease of IgG antibodies against S1 and N protein before death was observed. The global antibody responses to non-structural/ accessory proteins revealed here may facilitate deeper understanding of SARS-CoV-2 immunology.

The immunogenicity of SARS-CoV-2 proteome is largely unknown, especially for non-27 structural proteins and accessory proteins. Here we collected 2,360 COVID-19 sera and 601 28 control sera. We analyzed these sera on a protein microarray with 20 proteins of SARS-CoV-29 2, built an antibody response landscape for IgG and IgM. We found that non-structural proteins 

In non-survivors, sharp decrease of IgG antibodies against S1 and N protein before death was is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. ; https://doi.org/10.1101/2020.12.08.20246314 doi: medRxiv preprint and entry into the target cells that express the viral receptor, i. e., angiotensin-converting 62 enzyme 2 (ACE2) [5] [6] [7] [8] [9] . While the function, including immunogenicity of most of the non-63 structural proteins and accessory proteins are still elusive.

One of the major features of COVID-19 patients is the extreme variability of clinical severity 65 from asymptom to death 10 . However, the factors that cause this variability are still largely 66 unknown. Humoral immune responses elicited by SARS-CoV-2 play essential roles, especially 67 in diagnosis, neutralizing antibody production and vaccine development [11] [12] [13] . Among all the 68 SARS-CoV-2 proteins, S protein and N protein exhibit high immunogenicity. Antibodies 69 against S protein and N protein are elicited in most patients, and with higher titers in severe 70 patients, demonstrating the association between severity and humoral immune responses 12,14 .

It was reported that the antibodies against peptides derived from non-structural and accessory 72 proteins were also detectable in patients 10, 15, 16 . However, the prevalence, clinical relevance and 

Here, we adopted an updated SARS-CoV-2 proteome microarray that contains 20 proteins, 81 profiled 2,360 sera from 783 COVID-19 patients and 601 control sera. We identified that NSP1, is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. ; https://doi.org/10.1101/2020.12.08.20246314 doi: medRxiv preprint laboratory examinations were missing, for each clinical parameter only the effective patient 99 numbers were given. Expectedly, gender, age, comorbidities of hypertension and diabetes are 100 associated factors of severity, however, only age are significantly associated with clinical 101 outcome for severe patients. In addition, in consistent with many previous studies 18-20 , we 102 identified a set of clinical and laboratory parameters which are highly related to severity or 103 outcome , such as lymphopenia, increased CRP (C reaction protein) and factors associated with 104 blood coagulation, cardiac injury, liver injury and kidney injury. Most of these factors are 105 associated both with severity and outcome, while some are likely associated either with severity 106 or outcome. For instance, thrombocytopenia and some kidney injury related factors are more 107 common in non-survivors as compared to the other two groups, while most liver injury related 108 factors are only associated with severity but not outcome. 112 screened a small cohort of convalescent patients 14 . Here, we aimed to systematically analyze 113 the immune responses and its dynamic change against SARS-CoV-2 proteins with a much 114 larger cohort of samples. In total, we collected 2,360 sera from 783 laboratory confirmed 115 COVID-19 patients as well as 601 control sera ( Table S1 ). All of these sera were analyzed on 116 the SARS-CoV-2 protein microarray. To acquire high-quality data for the microarray 117 experiments, we prepared a positive control by mixing 50 randomly selected COVID-19 sera.

This control was then probed on each microarray to assess and normalize the data. It turned out 119 that high reproducibility was achieved in our assa (Figure S1c, d) . To simplify the analysis and 120 assure the comparability among different SARS-CoV-2 proteins, we defined "initial serum" as 121 the first serum collected 14 days after symptom onset for each patient. The results of the initial 122 sera were used to construct the antibody response landscape (Figure 1 ). Immune response 123 frequency was calculated for each protein with the cutoff value set by mean + 2 x SD of the 124 control group. Except for S1 and N, which are known of highly antigenic, we found that several 125 non-structural and accessory proteins elicited prevalent antibody responses, especially for IgG, 126 including NSP1, NSP7, NSP8, RdRp, ORF3b and ORF9b, for which the positive rates are 38%, 127 48.4%, 27.9%, 30.3%, 52.1% and 28%, respectively. Although the IgM responses were high in 128 some cases, the overall responses are much lower than that of IgG. We then decided to focus 129 on IgG for in-depth analysis.

The IgG pattern of Non-structural and accessory proteins is distinct from that of S1 and 132 N protein

We next asked whether the IgG responses to these proteins are associated with each other. We 134 chose NSP7 as an example. The samples were divided into two groups depending on positive 135 or negative of NSP7 IgG. Positive rates of the rest proteins were calculated for the two groups.

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. ; https://doi.org/10.1101/2020.12.08.20246314 doi: medRxiv preprint Unexpectedly, for all the non-structural and accessory proteins, except for ORF3a, ORF6 and 137 ORF7a which barely elicit antibodies, the positive rates in NSP7-IgG positive group was 138 significantly higher than that in NSP7-IgG negative group, demonstrating high correlations 139 (Figure 2a) . Interestingly, there is no obvious difference for the IgG responses of S1 and N.

To further confirm our observation, we reversely compared the positive rates of IgG to non-141 structural/accessory proteins between the groups of S1-IgG positive and negative (Figure 2b) .

The positive rates of IgG response of N protein are significantly different between these two 143 groups, while no obvious difference was shown for the non-structural/accessory proteins. These 144 observations demonstrate that the structural proteins elicit antibodies with distinct pattern to 145 that of the non-structural and accessory proteins, suggesting that the underlying mechanisms 146 by which the antibodies are triggered are different for these two groups of proteins. To further 147 study the correlations of IgG signal intensity among the proteins, Pearson correlation 148 coefficients between any two of these proteins were calculated and then clustered. The proteins 149 with less than 10% response frequency were not included duo to statistical limit. (Figure 2c ).

Consistently, the S and N proteins have lower correlations with the non-structural/accessory 151 proteins, while the non-structural/accessory proteins were clustered together. In addition, 152 several sub-clusters were shown among the non-structural/accessory proteins. Interestingly,

NSP8 and RdRp have a high correlation (Figure 2d) . It is known that RdRp, NSP8 and NSP7 154 could form a tight complex 21,22 , which might contribute to the high correlation. However, the 155 correlation between NSP8 and NSP7 is less significant (Figure 2c, e) . The structure of the 156 complex shows NSP7 physically connect to RdRp and NSP8, but with most of the protein 157 surface blocked (Figure 2f) , while NSP8 and RdRp are more accessible. However, NSP7 elicit 158 antibody in a higher frequency (Figure 1a) , suggesting NSP7 might mainly exist in other forms 159 rather than complex with NSP8 and RdRp, thus has other yet to be discovered biological 160 function(s). In addition, the IgG responses of NSP2 and NSP16 also have a high correlation 161 (Figure 2c, d-g) . It was reported that PPI (protein-protein interaction) was detected between 162 NSP2 and NSP16 23 . However, we did not detect any direct binding signals between NSP2 and is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. ; https://doi.org/10.1101/2020.12.08.20246314 doi: medRxiv preprint non-survivors. Two statistical methods were applied to assess the correlations. One is to analyze 174 the positive rate of IgG against each protein, and the other is to compare the signal intensity 175 distribution among groups (Figure 3) . For both S1 and N, the overall signals for severe groups 176 were slightly higher than that of non-severe group, but there was no significant difference 177 between the survivors and non-survivors. In contrast, the non-structural/accessory proteins, that 178 with high positive rates/ signal intensities, are more significantly correlated with severity. It is 179 worth noting that for the 6 non-structural and accessory proteins, both the positive rates and 180 signal intensities are significantly higher as the disease exacerbates to more severe stages 181 (Figure 3) . These results indicate that the IgG responses against non-structural/accessory 182 proteins are of higher correlations with the disease severity, and may could serve as a better 183 predictor of COVID-19 severity than that of S1 and N proteins.

We next analyzed the correlation between antibody response and clinical parameter. The 185 clinical parameters, which have statistical correlations with IgG against S1, N, NSP7, NSP8,

RdRp, ORF3b and ORF9b, were determined (Table S3, Figure S3 ). All of these parameters 187 are related with severity, suggesting severity is a major factor and confounder that contribute 188 to the correlations. Interestingly, thrombocytopenia, is related with clinical outcome but not 189 severity (Table 1) , is significantly correlated with NSP7 and ORF3b but not S1 and N, further is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. ; https://doi.org/10.1101/2020.12.08.20246314 doi: medRxiv preprint collapse in SARS-CoV-2 related humoral immune in a majority of patients before death and 212 further study are needed to confirm this.

IgG against the 6 non-structural/accessory proteins decline rapidly during COVID-19 215 progression 216 We next analyzed the dynamic of IgG responses for the 6 non-structural/accessory proteins.

217 Surprisingly, the IgG responses, i.e., the signal intensities and positive rates, against the 6 218 proteins reached plateau in all the three groups at about 20 days after the symptom onset, and 219 then decreased rapidly for all the three groups (Figure 5a, 5b) , this is largely different from S1 220 and N protein (Figure 4) . We next selected NSP7 IgG as an example for further analysis and 221 depicted the change for each patient (Figure 5c-e) . Continuous and dramatic decline of IgG 222 against NSP7 for most patients were observed (Figure 5d, 5e) . These results imply the B cells 223 that producing IgG antibodies against non-structural/accessory proteins might be short-lived, 224 and/ or the underlying mechanism of generating IgG antibodies against non-225 structural/accessory proteins may differ from that of S1 and N proteins.

In this study, we profiled 2,360 sera from 783 COVID-19 patients and 601 control sera using a 229 SARS-CoV-2 proteome microarray. We found that 6 non-structural/accessory proteins elicit is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. ; https://doi.org/10.1101/2020.12.08.20246314 doi: medRxiv preprint to our knowledge, all of these findings are based on small cohorts of samples. Here, by 249 analyzing a large cohort of samples, we actually constructed an antibody response landscape of 250 SARS-CoV-2 proteome. This landscape extent our knowledge of the interaction between 251 SARS-CoV-2 and the immune system. However, due to the difficulty of protein preparation, 252 there are still some proteins that are missing on the SARS-CoV-2 proteome microarray, such 253 as ORF8, which has recently been reported to be able to elicit strong antibody response 30 . Some 254 of these missing proteins will be added when we update the microarray.

Comparison of the IgG responses among the antigenic proteins revealed the possibility that 256 the generation of antibodies against the non-structural/accessory proteins are not independent 257 with each other, which means for one patient that positive for one protein tend to has a 258 significant chance to be positive for other proteins. This is expected since these proteins may 259 be simultaneously exposed to or not to the immune system. Interestingly, high correlations were 260 shown between some particular proteins, likely revealing the high associations of these proteins 

We also observed high correlations between NSP2 and NSP16, suggesting the two proteins 265 might associate in vivo.

The function of the antibodies to non-structural/accessory proteins are still largely unknown, 267 our results reveal that the antibody levels are more associated with disease severity, particularly 268 with the final outcome. These findings imply that the antibodies against non-269 structural/accessory proteins may play more important roles, thus worth further and in-depth 270 investigation. One concern about the antibody against S protein is the possible ADE (antibody is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. ; https://doi.org/10.1101/2020.12.08.20246314 doi: medRxiv preprint structural/accessory proteins might due to the suppressed production of long-lived B cells or 287 tend to generate short-lived B cells with unknown mechanism 40 . It seems that there are two 288 distinct mechanisms through which the proteins of SARS-CoV-2 elicit host humoral immune 289 responses: 1) Viral particle is involved as antigen resource, specifically S and N proteins, which 290 elicit potent antibody responses and tend to generate long-lived B cells. These antibodies 291 mainly play a protective role.

2) The infected cell is involved as antigen resource with non-292 structural/accessory proteins, which elicit weaker antibody responses and tend to be suppressed 293 to generate long-lived B cells. These antibodies might be stronger to induce cytokines to 294 contribute severer outcomes. However, this hypothesis should be confirmed by further studies.

In contrast to the antibody levels of S and N protein are stable for survivors, we observed the 296 overall antibody levels start to decline in non-survivors at about ten days before death. This is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. ; https://doi.org/10.1101/2020.12.08.20246314 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. ; https://doi.org/10.1101/2020.12.08.20246314 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. ; https://doi.org/10.1101/2020.12.08.20246314 doi: medRxiv preprint parameters set as 95% laser power/ PMT 550 and 95% laser power/ PMT 480 for IgM and IgG, 

Table S1. Serum Samples and patients (related to Figure 1 ). 420 Table S2 . SARS-CoV-2 proteins included in the proteome microarray (related to Figure   421 1 and Figure S1 ). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. ; https://doi.org/10.1101/2020.12.08.20246314 doi: medRxiv preprint TOM70. Cell. Mol. Immunol. 17, 998-1000 (2020). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 11, 2020. For a-b , error bar was given as the 95% confidential interval. P-value was calculated by two-sided χ 2 test. *, P < 0.05, **, P < 0.01, ***, P < 0.001, ****, P < 0.0001, n. s., not significant. , RdRp (f) ORF3b (g) and ORF9b (h). For positive rate analysis, error bar was given as the 95% confidential interval. P-value was calculated by two-sided χ 2 test. *, P < 0.05, **, P < 0.01, ***, P < 0.001, ****, P < 0.0001, n. s., not significant. For signal intensity analysis, the middle line was median value; the upper and lower hinges were the values of 75% and 25% percentile. Pvalue was calculated by two-sided t test. The trends of median signal intensities of IgG at different time points for S1 (a) and N (b), among three sample groups, i. e., non-severe, severe (survivors) and severe (nonsurvivors). Samples were grouped per day and the time points with sample number less than 4 were excluded due to lack of statistical significance. c-d. Relative S1-IgG signal levels were calculated for each patient, by dividing the signal intensity of the samples collected at other time points vs. samples collected at 0-2 days before the death of nonsurvivors (c, n = 35) or the discharge of survivors (d, n = 108). The samples were grouped per three days. For each patient, the signals were averaged if there were more than one sample during each three-day. Pvalue was calculated by two-sided t test between the indicated group and the first group (0 -2 days). *, P < 0.05, **, P < 0.01, ***, P < 0.001, n. s., not significant. Table S1 . Serum Samples and patients (related to Figure 1 ). Table S2 . SARS-CoV-2 proteins included in the proteome microarray (related to Figure 1 and Figure S1 ). Table S3 . IgG responses are associated with clinical parameters (related to Figure 3 and Figure S3 ). Figure S1 . SARS-CoV-2 proteome microarray and the assessment of reproducibility (related to Figure 1 ). a. The layout of the SARS-CoV-2 proteome microarray. The locations of proteins and controls are indicated. b. Representative images of the microarray screened by sera from a healthy control and a COVID-19 patient. c. Correlation analysis between two microarrays probed independently with a positive control serum. d. Statistical analysis of the Pearson correlation coefficients between the microarrays incubated with the positive control serum with the averaged data set (see methods). The data are present as mean SD. Table S3 . IgG responses are associated with clinical parameters (related to Figure 3 and Figure S3 )

Red color marks the correlation coefficients more than 0.2, and the green color marks the correlation coefficients less than -0.2.

A pneumonia outbreak associated with a new coronavirus of probable bat origin. 430

A new coronavirus associated with human respiratory disease in China

An interactive web-based dashboard to track COVID-19 in 434 real time

Genome Composition and Divergence of the Novel Coronavirus

Isolation and characterization of a bat SARS-like coronavirus that uses the 438 ACE2 receptor

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

Structural basis for the recognition of SARS-CoV-2 by full-length human ACE2

Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 444 receptor

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is 446 Blocked by a Clinically Proven Protease Inhibitor

Viral epitope profiling of COVID-19 patients reveals cross-reactivity and 448 correlates of severity

Neutralizing Antibodies against SARS-CoV-2 and Other Human 451

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

Immunology of COVID-19: Current State of the Science

SARS-CoV-2 proteome microarray for global profiling of COVID-19 specific

 Table S2 . SARS-CoV-2 proteins included in the proteome microarray (related to Figure 1 and Figure S1)