key: cord-1050574-m4dc7x1u
authors: Padmanabhan, P.; Desikan, R.; Dixit, N. M.
title: The quantitative landscape of the neutralizing antibody response to SARS-CoV-2
date: 2020-09-28
journal: nan
DOI: 10.1101/2020.09.25.20201996
sha: 8c3ef3ea7d38ff2916f9134a933461c6254e85c6
doc_id: 1050574
cord_uid: m4dc7x1u

Neutralizing antibodies (NAbs) appear promising interventions against SARS-CoV-2 infection. Over 100 NAbs have been identified so far and several are in clinical trials. Yet, which NAbs would be the most potent remains unclear. Here, we analysed reported in vitro dose-response curves (DRCs) of >70 NAbs and estimated corresponding 50% inhibitory concentrations, slope parameters, and instantaneous inhibitory potentials (IIPs), presenting a comprehensive quantitative landscape of NAb responses to SARS-CoV-2. NAbs with high IIPs are likely to be potent. To assess the applicability of the landscape in vivo, we analysed available DRCs of NAbs from individual patients and found that the responses closely resembled the landscape. Further, we created virtual patient plasma samples by randomly sampling NAbs from the landscape and found that they recapitulated plasma dilution assays from convalescent patients. The landscape thus offers a facile tool for benchmarking NAbs and would aid the development of NAb-based therapies for SARS-CoV-2 infection.

SARS-CoV-2 NAb repertoire, a question that arises is which of these NAbs should be taken 48 up for clinical development. A comparative evaluation of the NAbs has not been performed. 49 Studies identifying NAbs typically report the 50% inhibitory concentration, IC 50 , of the 50 NAbs, the concentration at which viral infectivity is reduced by 50% of that in the absence of 51 the NAbs (Fig. 1A) . The inference drawn is that the lower is the IC 50 , the more potent is the 52 NAb (Fig. 1A) . A limitation of this approach arises from the non-linear dependence of the 53 neutralization efficacy of NAbs on their concentrations because of which a NAb with a lower 54 . 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 (which was not certified by peer review) preprint

The copyright holder for this this version posted September 28, 2020. . https://doi.org/10.1101/2020.09.25.20201996 doi: medRxiv preprint Here, we decided to examine whether the IIP could be applied to comparatively 86 evaluate SARS-CoV-2 NAbs. Unlike IC 50 values, which are routinely reported, the values of 87 . 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) preprint

The copyright holder for this this version posted September 28, 2020. . https://doi.org/10.1101/2020.09.25.20201996 doi: medRxiv preprint m have rarely been reported for SARS-CoV-2 NAbs, precluding the estimation of IIP for 88 most NAbs. We therefore collated all the available in vitro DRCs of SARS-CoV-2 NAbs and 89 analysed them to estimate both IC 50 and m, and then IIP (Fig. 2) 96 We collated and analysed the reported DRCs of over 70 NAbs obtained using SARS-CoV-2 97 pseudotyped virus infection assays (Figs. 2 and 3; Methods; Table S1 ) 21-39 . These NAbs have 98 been proposed as the most promising from among many examined in the respective studies. (Table S1 ). The resulting estimates of 102 IC 50 were in close agreement with the reported estimates, giving us confidence in the fits 103 ( Fig. S3A ; Table S1 ). The IC 50 displayed a wide variation across NAbs, ranging from ~10 -3 104 µg/ml to ~140 µg/ml (Fig. 3C) . m too displayed wide variability, spanning the range ~0.2 to 105 2.3 (Fig. 3D ). As mentioned above, values of m have not been reported in previous studies. 106 We examined whether the variability in IC 50 and m was restricted to a particular pseudotyped 107 virus construct or backbone used (Fig. 3F, 3G) , the cell line used (Fig. 3H, 3I ), or assay 108 conditions, which could vary across studies (Fig. S3B, S3C) , and found that not to be true.

. 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) preprint

The copyright holder for this this version posted September 28, 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) preprint

The copyright holder for this this version posted September 28, 2020. . https://doi.org/10.1101/2020.09.25.20201996 doi: medRxiv preprint (Fig. 4A ). We found that 5 NAbs had IIP > 5. These, from our calculations, would be the . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted September 28, 2020. . https://doi.org/10.1101/2020.09.25.20201996 doi: medRxiv preprint 1 (Table S1 ). (Fig. 4) . We assumed Loewe additivity 41-43 between the different NAbs to describe their 207 overall efficacy. We found that with these in silico samples, we were able to closely 208 recapitulate experimental serial dilution assays (Figs. 6D, S8 ), giving us confidence in the 209 NAb landscape. The values of NT 50 , the dilution at which neutralization efficiency decreases 210 . 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) preprint

The copyright holder for this this version posted September 28, 2020. . https://doi.org/10.1101/2020.09.25.20201996 doi: medRxiv preprint by 50% of the undiluted plasma, we estimated (Fig. 5D inset) were also comparable to the 211 values estimated from patient samples (~10 1 to 10 4 ; see Ref. 24). 212 These comparisons also re-emphasize the need to choose NAbs or convalescent patient 213 plasma for treatment based not only on the IC 50 but also m. To elucidate this further, we 214 repeated our in silico analysis by comparing simulated samples containing NAbs with similar 215 IC 50 values but low (group A) or high (group B) m (Fig. 6E, 6F) . Although simulated plasma 216 samples from groups A and B had similar NT 50 (Fig. 6G) , samples with high m on average 217 had much higher values of IIP 100 than those with low m (Fig. 6H) . Thus, in interpreting 218 plasma dilution assays and in designing plasma and NAb therapies, accounting for m, which 219 has been ignored thus far, would be as important as IC 50 . is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 28, 2020. . https://doi.org/10.1101/2020.09.25.20201996 doi: medRxiv preprint CoV-2, these effects are yet to be elucidated. In many cases, the targets and/or the mechanism 233 of action of the NAbs are not known. For instance, the potent NAb 47D11 targeting the spike 234 protein S of SARS-CoV-2 blocks virus entry without preventing the binding between S and 235 the host receptor ACE-2 required for entry 32 . As future studies establish molecular details of 236 the SARS-CoV-2 entry process 37,47 , identifying unifying characteristics of the NAbs with 237 high IIP values would become feasible, allowing rational design of even more potent NAbs. 

. 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 this version posted September 28, 2020. . https://doi.org/10.1101/2020.09.25.20201996 doi: medRxiv preprint Data 255 We considered data from studies that reported dose-response curves of NAbs using 256 SARS-CoV-2 pseudotyped virions 21-39 . The assays estimate the fraction of infection events 257 unaffected by the NAbs as a function of the NAb concentration (Fig. 2) . Data from such and m for each NAb as well as associated 95% confidence intervals. We then computed 273 . 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) preprint

The copyright holder for this this version posted September 28, 2020. values estimated using the two methods deviated by 20% or more in our analysis, for the 276 deviation indicated that such NAbs either did not conform to the trends expected by Eqs. [1] 277 and [2] or had large uncertainties in the data precluding robust parameter estimation. We also 278 repeated our analysis with a more liberal threshold of 30% deviation for acceptance (Fig. S9) . 279 The details of the NAbs and parameter estimates are presented in Table S1 .

In silico simulation of plasma dilution assays 282 We simulated plasma dilution experiments as follows. We assumed that the plasma 283 contained N NAbs with equimolar concentrations. For each NAb, IC 50 was sampled from the 284 range 0.001 µg/ml to 100 µg/ml (Fig. 3C) and m was sampled from the range 0.2 to 2 ( Fig.   285   3D ). This range was consistent with the range seen from the spectrum of NAbs within 286 individual patients (Fig. 6A) . The reciprocal plasma dilution curve was predicted assuming 287 Loewe additivity between the different NAbs 41,42 using

Here, γ is the plasma dilution factor. ε is the fraction of infection events affected by 290 the plasma in a single round of infection. D i is the concentration of the i th NAb in the plasma 291 before dilution, IC 50 i is its half-maximal inhibitory concentration and m i its slope, with 292 . 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) preprint

The copyright holder for this this version posted September 28, 2020. . https://doi.org/10.1101/2020.09.25.20201996 doi: medRxiv preprint i ∈ 1,2,.., N { } . We assumed N = 10 in our simulations, based on the number of NAbs with 293 significant neutralization efficacy seen in patients 25 . We estimated the value of γ at which 294 ε = 0.5 as the corresponding NT 50 . We chose D i as D 0 /N, and set D 0 = 100 µg/ml. 295 We repeated these simulations 100 times, with each simulation representative of an 296 individual patient. We compared the resulting predictions with observations from 3 297 patients 51 , which also we digitized (Fig. S8) . 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) preprint

The copyright holder for this this version posted September 28, 2020. . https://doi.org/10.1101/2020.09.25.20201996 doi: medRxiv preprint

Developing therapeutic monoclonal antibodies at pandemic pace

Pipeline Brings Hope To Defeat SARS-Cov-2

COVID-19 antibody therapeutics tracker: a global online database of 317 antibody therapeutics for the prevention and treatment of COVID-19

Convalescent Plasma Therapy 320 for COVID-19: State of the Art

Convalescent plasma as a potential therapy for 322 COVID-19

Effectiveness of convalescent plasma therapy in severe COVID-19 324 patients

Passive antibody therapy in COVID-19

Passive immunotherapy of viral infections: 'super-328 antibodies' enter the fray

Broadly neutralizing anti-HIV-1 330 monoclonal antibodies in the clinic

Strategies and challenges for the next 332 generation of therapeutic antibodies

Therapeutic antibodies for autoimmunity and inflammation

Antibody therapy of cancer

Dose-response curve slope sets class-specific limits on inhibitory 340 potential of anti-HIV drugs

This work was supported by the DBT/Wellcome Trust India Alliance Senior 306