key: cord-0796743-cha7ndv5
authors: Horspool, A. M.; Kieffer, T.; Russ, B. P.; DeJong, M. A.; Wolf, M. A.; Karakiozis, J. M.; Hickey, B. J.; Fagone, P.; Tacker, D. H.; Bevere, J. R.; Martinez, I.; Barbier, M.; Perrotta, P. L.; Damron, F. H.
title: Interplay of antibody and cytokine production reveals CXCL-13 as a potential novel biomarker of lethal SARS-CoV-2 infection
date: 2020-08-31
journal: nan
DOI: 10.1101/2020.08.24.20180877
sha: a8a3ef612b850e8da3ce3abb74e631e35869b853
doc_id: 796743
cord_uid: cha7ndv5

The SARS-CoV-2 pandemic is continuing to impact the global population. This study was designed to assess the interplay of antibodies with the systemic cytokine response in SARS-CoV-2 patients. We demonstrate that significant anti-SARS-CoV-2 antibody production to Receptor Binding Domain (RBD), Nucleocapsid (N), and Spike S1 subunit (S1) of SARS-CoV-2 develops over the first 10 to 20 days of infection. The majority of patients produced antibodies against all three antigens (219/255 SARS-CoV-2 positive patient specimens, 86%) suggesting a broad response to viral proteins. Patient mortality, sex, blood type, and age were all associated with differences in antibody production to SARS-CoV-2 antigens which may help explain variation in immunity between these populations. To better understand the systemic immune response, we analyzed the production of 20 cytokines by SARS-CoV-2 patients over the course of infection. Cytokine analysis of SARS-CoV-2 positive patients exhibited increases in proinflammatory markers (IL-6, IL-8, IL-18) and chemotactic markers (IP-10, SDF-1, MIP-1{beta}, MCP-1, and eotaxin) relative to healthy individuals. Patients who succumbed to infection produced decreased IL-2, IL-4, IL-12, IL-13, RANTES, TNF-, GRO-, and MIP-1 relative to patients who survived infection. We also observed that the chemokine CXCL13 was particularly elevated in patients that succumbed to infection. CXCL13 is involved in B cell activation, germinal center development, and antibody maturation, and we observed that CXCL13 levels in blood trended with anti-SARS-CoV-2 antibody production. Furthermore, patients that succumbed to infection produced high CXCL13 and also tended to have high ratio of nucleocapsid to RBD antibodies. This study provides insights into SARS-CoV-2 immunity implicating the magnitude and specificity of response in relation to patient outcomes.

loading, plates were incubated for 10 minutes at room temperature shaking at 60rpm. 135

Plates were then washed four times with PBS-T. Secondary antibody buffer (100µL of 1% 136 milk diluted in PBS-T containing 1:500 goat anti-human-IgG-HRP; Invitrogen Part #: 137 31410) was added immediately following the washing procedure. The plates were 138 incubated for 10 minutes at room temperature shaking at 60rpm. Plates were washed five 139 times with PBS-T. SigmaFast OPD substrate (Sigma Part#: P9187) was prepared in 140 milliQ (18.2MΩxcm) water and 100µL was aliquoted into each well. Ten minutes after 141 loading the substrate, 25µL of stop solution (3N HCl) was added to end colorimetric 142 development. The absorbance of the substrate in each well was measured on a Synergy 143 H1 (Biotek) spectrophotomer at 492nm. Antibody concentration was calculated based on 144 area under the curve analyses of A492 vs. dilution factor plots for each sample. prepared for analysis by heating at 56 o C for 1 hour. Samples were then centrifuged at 152 13,000 x g for 2 mins to pellet aggregates. Samples (25µL) were diluted 1:2 with universal 153 assay buffer and incubated at room temperature on an orbital shaker at 500 rpm for 1 154 hour. Select samples (based on sample quantity) were diluted 1:4 or 1:5 with the universal 155 assay buffer, which was taken into account during analysis. A standard curve was 156 . 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint generated using antigen standards provided by the manufacturer. Samples were 157 resuspended in 120 µL wash buffer prior to running on a MAGPIX (Luminex) instrument, 158 and 35 µL was analyzed per samples. Bead cytokine production and antibody production were pooled into Microsoft Excel and 177 imported to ClustVis 17 . Data were transformed by the ln(x) transformation provided in the 178 webtool and grouped with a 95% confidence interval. Groups were based on patient 179 . 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 August 31, 2020. 

In-patient anti-SARS-CoV-2 antibody production: Antibody binding target and the timing 192 of the antibody response are critical factors in mediating immunity. We evaluated anti-193 SARS-CoV-2 antibody production to 3 antigens (RBD, N, and S1) in 82 in-patients 194 Table 1 ) by developing a novel rapid-ELISA technique. Our rapid-ELISA 195 technology evaluates IgG antibody production to the SARS-CoV-2 RBD, N, and S1 196 proteins in approximately 1 hour with greater than 99% accuracy (Supplementary Table  197 2). Our survey of SARS-CoV-2 positive patients demonstrated that antibody (IgG) 198 production to RBD, N, and S1 proteins developed over the first 10 to 20 days post-199 symptom onset (Figure 1a 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint Supplementary Data File). To better understand the kinetics of the antibody response, we 203 plotted IgG production of every patient over time to RBD, N, or S1. Patients produced IgG 204 against RBD rapidly after symptom onset with the peak IgG response occurring 10 days 205 after symptom onset (Figure 1g ). Anti-S1 IgG production escalated over a slightly larger 206 period (13 days, Figure 1i ) and anti-N IgG production was slower than either anti-RBD or 207 anti-S1 antibody production (22 days, Figure 1h ). Taken together, these data describe 208 the breadth and timing of the IgG response to SARS-CoV-2 antigens. populations, we analyzed patient groups based on sex, patient mortality, blood type, and 214 age against anti-RBD, anti-N, or anti-S1 antibody production. As IgG production is more 215 consistently detectable after ten days post-symptom onset 18,19 , we assessed differences 216

in IgG production beyond ten days post symptom onset. Limiting sample analysis to those 217 greater than ten days post symptom onset did not significantly impact the mean antibody 218 production of the patients (Supplementary Figure 1) . Patients who did not survive SARS-219

CoV-2 hospitalization produced significantly more antibodies to SARS-CoV-2 N than 220 patients that survived infection ( Figure 2a) . Furthermore, patients that did not survive 221 SARS-CoV-2 infection did not produce different quantities of anti-N antibodies than 222 surviving patients during early infection (Supplementary Figure 2) . To accurately assess 223 differences in antibody production independently of disease outcome, we quantified anti-224 SARS-CoV-2 IgG production in patients who survived infection grouped by biological sex, 225 . 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint blood type, and age. We determined that, in our cohort, females significantly produced 226 more anti-S1 IgG than males ( Figure 2b ). We also observed that blood type was 227 significantly associated with anti-SARS-CoV-2 IgG production ( Figure 2c ). Blood type B+ 228 patients produced significantly more IgG to RBD and S1 than A+ or O+ patients ( Figure  229 2c) and A+ patients produced the lowest quantities of anti-RBD and anti-S1 IgG. O+ 230 patients produced reduced anti-N IgG relative to A+ or B+ patients. Previous studies have 231 identified that age impacts antibody production to SARS-CoV-2 20,21 . Our study 232 demonstrates that antibody production against RBD or S1 antigens increased with age 233 ( Figure 2d ). In contrast, antibody production to N increased in patients over 50 years old 234 but did not continue to increase with age after 80 years of age. This is particularly evident 235 when examining Pearson correlations between age and anti-SARS-CoV-2 IgG production 236 for each antigen (Supplementary Figure 3) . Overall, these data document a significant 237 impact of patient demographics on anti-SARS-CoV-2 antibody production. 238

Changes in SARS-CoV-2 patient cytokine responses correlate with disease severity: 240

Antibody production represents the antigen-specific response to pathogens but is only 241 one facet of immunity. We examined the broader immunological response to SARS-CoV-242 2 infection by quantifying the production of cytokines involved in a representative subset 243 of SARS-CoV-2 or healthy patients. SARS-CoV-2 patients exhibited significant increased 244 pro-inflammatory cytokine production (IL-6, IL-8, IL-18) and increased chemotactic 245 cytokine production (IP-10, SDF-1, MIP-1 , MCP-1 and eotaxin) relative to non-infected 246 individuals (Figure 3 ). Of the SARS-CoV-2-infected patients, mortality was associated 247

with increased IL-6, IL-8, IL-18, IP-10, and MCP-1 production. Patients who succumbed 248 . 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 August 31, 2020. This response is critical for eradicating many pathogens. As many SARS-CoV-2 patients 263 produced robust antibody responses to multiple antigens, we hypothesized that germinal 264 center formation would be increased in these patients. To quantify this, we measured the 265 serum concentration of CXCL13, a critical mediator of germinal center formation and a 266 biomarker of this immunological response 8,10,22,23 . We observed that CXCL13 production 267 primarily correlated with peak antibody production to RBD and S1 antigens across SARS-268

CoV-2 infected patients (Figure 4a-c) . Additionally, we observed that there was a 269 significant increase in average production of CXCL13 in positive patients relative to 270 negative SARS-CoV-2 patients. In addition, we discovered that CXCL13 production was 271 . 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint significantly increased in patients that did not survive SARS-CoV-2 infection compared to 272 those that did (Figure 4d ). When we compared antibody and CXCL13 production based 273 on patient survival over time, we observed that patients who did not survive SARS-CoV-274 2 infection exhibited a sustained increase in antibody and CXCL13 production relative to 275 surviving patients (Figure 4ef) to N increased over a longer period than antibodies against RBD, or the S1 domain. This 288 could be due to a variety of factors including antigen immunodominance 29,30 , incongruent 289 antigen processing and availability 31,32 , differences in antibody utility and turnover, or prior 290 exposure to similar RBD/S1 antigens of other coronaviruses. Theoretically, as N is not 291 expressed on the viral surface, B cells producing antibodies against this antigen may not 292 be selected for as rapidly as those that are specific to the RBD or S1 antigens and may 293 not possess neutralizing function. As infection worsens, more cells lyse. This may 294 . 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint increase the local concentration of free nucleocapsid available for antigen processing and 295 presentation, particularly in lymphoid tissue 33 . In this respect, a more robust antibody 296 response to nucleocapsid later in infection may be due to increased cellular damage. This 297 may initiate a positive feedback loop where infected cells lyse and release nucleocapsid, 298 which induces a less functional anti-nucleocapsid antibody response that fails to alleviate 299 the cell lysis. More evidence is required to support these hypotheses, but these are 300 interesting paradigms to consider in the context of anti-SARS-CoV-2 immunity. 301

Lethal SARS-CoV-2 infection is significantly correlated with higher antibody 302 production 19,20,26 and is described further in this study. In analyzing antibody production 303 between patient demographics, it was important to eliminate increased antibody 304 production due to lethal infection as a source of bias. As such, our analyses presented 305 here describe IgG production of SARS-CoV-2 survivors grouped by demographic. There 306 are a multitude of studies reporting differences in IgG production between demographics 307 including: trends in anti-SARS-CoV-2 antibody production between sexes 20,21,34-36 , a 308 correlation of genetically encoded blood type with SARS-CoV-2 immunity 37 , and 309 variability in antibody production in the aging population 20,21 . From these prior studies and 310 others 38,39 it is known that biological sex can impact antibody production during infection. 311

We observed this phenomenon when quantifying sex specific anti-S1 IgG production. The 312

anti-viral response is mediated in part by Toll-like receptors which are differentially 313 regulated between the sexes 40,41 . A higher frequency of anti-S1 IgG in females would 314 suggest an increased neutralizing response to the virus which has not been thoroughly 315 evaluated to-date. Our data exhibited a modest difference in antibody production between 316 . 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint sexes. As a result, we do not consider biological sex to be a major contributor to anti-317

It is documented that red blood cell phenotypes can influence microbial 319 pathogenesis as antigens can function as receptors and/or co-receptors for pathogenic 320 organisms 42 . Historically, an association was identified between ABO type and pathogen CoV spike protein binding to ACE2 44 . Although the underlying mechanism relating blood 326 type to SARS-CoV-2 pathogenesis remains unclear, it appears there may be a 327 relationship between ABO blood type and coronavirus infection. Recent data identified 328 the 9q34.2 locus (ABO blood group locus) as potentially involved in susceptibility to 329 COVID-19 respiratory failure with evidence that type A phenotypes are at higher risk while 330 type O phenotypes are partially protected 45 . The data generated in these studies show an 331 interesting pattern that may reinforce blood type related outcomes in severe disease due 332 to a previously unreported association to the level and type of antibody response. As seen 333 in Figure 2c , the relative quantity of anti-RBD and anti-S1 antibodies was highest among 334

anti-N antibodies. This is further accentuated by evaluating the ratio of anti-RBD or anti-336 S1 versus anti-N in our patient cohort which shows that higher N:RBD or N:S1 ratios are 337 associated with poor prognosis (Supplementary Figure 5) . It is plausible that type-A 338 individuals may have a misdirected humoral response due to antigenic homology 339 . 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint between N-acetyl-galactosamine sugar moieties on the A antigen and Spike protein 340 resulting in molecular mimicry. This would result in type-O and -B individuals registering 341 more Spike protein epitopes as foreign and eliciting a more robust humoral response; in 342 turn, this putative mechanism could reduce infectious dose and decrease the risk of 343 mortality. Further studies evaluating physiologic modifications of Spike protein and its 344 antigenic moieties would help support or disprove this theory. As the conclusions from 345 these observations are currently theoretical, a more extensive review of comorbid 346

conditionswith a multivariate analysis and estimations of associated odds ratiosmay 347 reveal other associations outside of blood type. 348

The aging process is associated with decreased T-cell functionality 46 , resulting in 349 hyperactive B-cell proliferation that does not confer immunity 47 . We discovered that older 350 patients typically produced more antibodies to RBD and S1 than younger patients. The 351 lack of increase in antibody production to nucleocapsid in the elderly may be a function 352 of antigen availability. To speculate, if elderly patients have higher viral loads due to 353 decreased remediation of virus this would increase the relative abundance of surface 354 exposed antigens (RBD and S1), but not necessarily hidden antigens (N). Increased 355 antibody production would therefore predominantly occur to RBD and S1, and not N. 356

Other challenges are associated with studying this population including co-presentations 357 of multiple diseases which complicates this analysis. Regardless, our study has identified 358 several patient demographics associated with differences in the anti-SARS-CoV-2 359 antibody response. 360

The anti-viral immune response depends on a variety of signaling pathways 361 mediated by cytokines and chemokines. Many of the pro-inflammatory cytokines 362 . 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 August 31, 2020. of cardiac distress. Separately, we discovered increased eotaxin production in SARS-384

CoV-2 patients. Eotaxin was increased or similar to healthy patients during SARS-CoV-2 385 . 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 August 31, 2020. Antibody maturation signaling has not been investigated in the context of SARS-406

CoV-2. We assessed the activity of the antibody maturation pathway by measuring 407 CXCL13 concentrations in the serum of SARS-CoV-2 patients. Increased CXCL13 in 408 . 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 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. (which was not certified by peer review)

The copyright holder for this preprint this version posted August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint to RBD (a), N (b), or S1 (c). Correlation of antibody production to RBD vs. N (d) or S1 (e). 628

Correlation of antibody production to N vs. S1 (f). Antibody production of anti-RBD (g), 629

anti-N (h), or anti-S1 (i) antibodies by SARS-CoV-2 positive patients vs. days post SARS-630

CoV-2 disease onset. 631 . 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 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. (which was not certified by peer review)

The copyright holder for this preprint this version posted August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 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. (which was not certified by peer review)

The copyright holder for this preprint this version posted August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint

Analysis of serum cytokines in patients with severe acute 588 respiratory syndrome

The immunobiology of SARS

Candidate genes associated with susceptibility for SARS-592

CXCL10/IP-10 in infectious diseases pathogenesis and potential 594 therapeutic implications

Highly Efficacious Lymphocyte Chemoattractant, Stromal Cell-derived Factor 1 597 (SDF-1)

SDF-1α and CXCR4 as therapeutic 599 targets in cardiovascular disease

Stromal cell-derived factor-1α is cardioprotective after 601 myocardial infarction

Stromal derived factor 1α: A 603 chemokine that delivers a two-pronged defence of the myocardium

SARS-CoV-2 and cardiovascular complications: From molecular 606 mechanisms to pharmaceutical management

Outcomes of Cardiovascular Magnetic Resonance Imaging 608 in Patients Recently Recovered From Coronavirus Disease

Clinical features of patients infected with 2019 novel coronavirus 611 in Wuhan

Role of eotaxin-1 (CCL11) and CC chemokine receptor

in bleomycin-induced lung injury and fibrosis

Influenza virus a stimulates expression of eotaxin by nasal 616 epithelial cells

The β-chemokine receptors CCR3 and CCR5 facilitate infection by 618 primary HIV-1 isolates

T cell exhaustion

Networking at the level 621 of host immunity: Immune cell interactions during persistent viral infections

International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity

maturation of anti-SARS-CoV-2 antibodies. The significant increase of CXCL13 in 410 patients with lethal disease suggests this may be an emergency response to uncontrolled 411 infection. It is possible that sustained infection stimulates increased antibody affinity 412 maturation that is unable to keep pace with viral replication and the cytokine storm. In this 413 sense, CXCL13 could be used as a marker of SARS-CoV-2 disease severity. There is a 414 precedent for the utility of CXCL13 as a biomarker that is predictive of immune activation 415 during HIV exposure 8, 9, 24 . This adds credibility and feasibility for this utility, but further 416 studies are required to validate this approach. We have provided a schematic of how the 417 CXCL13 response interplays with our other observations of SARS-CoV-2 immunity in 418 Figure 6 . 419To summarize, this study provides insight into the breadth of the immunological 420 response against SARS-CoV-2. We demonstrated increasing antibody production to 421 multiple SARS-CoV-2 antigens over the first ten days of infection using a rapid-ELISA 422 assay. Our results exhibit that patient mortality, sex, blood type, and age impact antibody 423 production to SARS-CoV-2, adding to what is known about SARS-CoV-2 pathogenesis. 424Furthermore, lethal SARS-CoV-2 infection triggers a pro-inflammatory cytokine response, 425 in combination with the secretion of several chemotactic agents. Interestingly, patients 426 with lethal SARS-CoV-2 disease exhibited divergent cytokine production compared to 427 patients with non-lethal disease. Finally, we discovered that a marker of germinal center 428 activity (CXCL13) is upregulated in SARS-CoV-2 patients, and that this upregulation is 429 amplified in lethal disease. Ultimately, these studies help to elucidate the interplay 430 . 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 August 31, 2020. 20-plex cytokine assays. PF produced RBD used in this study. AMH analyzed and 445 compiled assay data and figures. All authors took part in writing and editing the 446 manuscript. We would like to thank BEI Resources for providing the following reagents 447 (NR-52422). We would finally like to express our gratitude to Drs. Laura Gibson and Clay 448Marsh for enabling this research during the global pandemic. 449 . 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 August 31, 2020. 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 August 31, 2020. 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 August 31, 2020. Biol. 3, 1-12 (2020) . 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint production by SARS-CoV-2 production is compared to anti-RBD (a), anti-N (b), or anti-S1 655 (c) IgG quantity over the course of patient disease. Red arrows represent CXCL13 656 maxima, and green arrows represent local IgG maxima. CXCL13 production was 657 . 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint compared between SARS-CoV-2 negative (-) and positive (+) patients, and SARS-CoV-658 2 positive survivors (S), or non-survivors (D) (d). Examples of a surviving patient 659 producing low CXCL13 and low anti-RBD IgG response (e) or deceased patient producing 660 high CXCL13 and high anti-RBD IgG response (f). Statistical significance was assessed 661 with a Brown Forsyth and Welch's one-way ANOVA followed by Tukey's multiple 662 comparison test. **** = p<0.0001, n.s. = not significant. 663 664 . 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 August 31, 2020. . https://doi.org/10.1101/2020.08.24.20180877 doi: medRxiv preprint