Longitudinal kinetics of RBD+ antibodies in COVID-19 recovered patients over 14 months

We describe the longitudinal kinetics of the serological response in COVID-19 recovered patients over a period of 14 months. The antibody kinetics in a cohort of 192 recovered patients, including 66 patients for whom follow-up serum samples were obtained at two to four clinic visits, revealed that RBD-specific antibodies decayed over the 14 months following the onset of symptoms. The decay rate was associated with the robustness of the response in that antibody levels that were initially highly elevated after the onset of symptoms subsequently decayed more rapidly. An exploration of the differences in the longitudinal kinetics between recovered patients and naïve vaccinees who had received two doses of the BNT162b2 vaccine showed a significantly faster decay in the naïve vaccinees, indicating that serological memory following natural infection is more robust than that following to vaccination. Our data highlighting the differences between serological memory induced by natural infection vs. vaccination contributed to the decision-making process in Israel regarding the necessity for a third vaccination dose.


Friday, January 21, 2022
Dear PLOS Pathogens editors, We would like to begin by thanking you and the reviewers for the helpful, productive, and constructive reviews and comments for our manuscript entitled; "Longitudinal kinetics of RBD + antibodies in COVID-19 recovered patients over 14 months" manuscript number, PPATHOGENS-D-21-02063. As a result of this review process, we believe our manuscript to be much improved.
We have carefully considered and addressed each of the reviewers' comments and revised the manuscript according to their suggestions.
We have also complied with the editorial requests to style the manuscript according to PLOS pathogens formatting guidelines.
Detailed point-by-point responses to each of the reviewers' comments can be found below.
We thank you for the consideration of our work and look forward to hearing from you.
Sincerely, Yariv Wine, PhD to a lesser extent following vaccination. These reports are important as they informed the community on the duration of immunity and the predicted protection. We have highlighted these reports in the introduction section, along with underlining the methodologies used to determine the longitudinal kinetics of the antibody response. However, there are still knowledge gaps with regard to the longitudinal kinetics of SARS-CoV-2 specific antibodies following recovery, including: i) duration following onset of symptoms (previous reports analyzed the kinetics in a relatively short period following the onset of symptoms); ii) lack of stratification by the robustness of the response (previous reports did not take into account the heterogenicity of the antibody response among COVID-19 recovered patients when analyzing the antibody kinetics).
Our study was conducted on a cohort of COVID-19 recovered patients with follow up samples across a maximum of 14 months following the onset of symptoms. Moreover, we applied a new approach to investigate the longitudinal kinetics by stratifying patients to groups according to their antibody levels at the first visit. This stratification approach allowed us to examine whether the antibody decay rate is associated with the robustness of the antibody response. Overall, our study provides new insights into the persistence of antibodies at the level of individuals who exhibit similar antibody response characteristics and their longitudinal kinetics over a relatively long period of time following the onset of symptoms.
Reviewer's comment: [Bottom of page 11] "As recovered patients received a vaccine at various DFS, we extrapolated the antibody levels to the day that the patients received the vaccine using the regression formula. The antibody levels at 8 DFV were extrapolated as well by using the regression formulas obtained from naïve vaccinees ( Figure 4e)" Is this really valid? If yes, I'm not sure I follow. Antibody levels at 8 DFV can only be predicted like that if the regression is the same in recovered vaccinees and naive vaccinees, right? Why is there no regression analysis of recovered vaccinees? Please explain.

Response:
We thank the reviewer for the request for a clearer and understandable explanation to our approach. We have addressed this question in the revised manuscript -see Results section. To support the understanding of the approach, we have provided an additional supplementary Figure  With regard to the reviewer's comment -"Why is there no regression analysis of recovered vaccinees": We did not have more than one time point following vaccine for the recovered/vaccinated individuals, and for this reason we applied the regression model obtained from the available time points prior the vaccination in order to extrapolate the additional data points (i.e. 8 DFV).
Reviewer's comment: " Fig. 3b and c in the IgG column no real difference in regression discernible by eye yet p<0.0001 due to log scale? Please discuss. Also, add data points in Fig  3c. And change red/green to a colorblind safe color scheme."

Response:
We thank the reviewer for this important comment. Indeed, a graphical error was made in Fig. 3c, resulting in a graph where no real difference in regression was discernible to the naked eye. We have now corrected the graphical error, as is shown in new Fig 4B and

Part III
Reviewer's comment: "Low resolution of most figures, please improve figure resolution."

Response:
We apologize for the low quality of the Figures. It seems that while converting the manuscript to pdf format, the figure resolution was reduced substantially. As proposed by PloS Pathogens editorial team, we have used the PACE system to ensure that the high resolution of the manuscript figures is preserved. Reviewer's comment: Fig 1: b: gender → sex (change also in the text) B: explain MI/SE in figure or just spell it out B: the legend inside the age circle excludes the ages 51 and 51. Please correct numbers and sign. Same, in the text on page 7, you write <50 and >50, which excludes 50.

Response:
The reviewer's comments on these figures is much appreciated and the figures have been revised and explained in the new version of the manuscript. b: gender was changed to sex (also in the text) B: MI/SE abbreviation in the figure was explained in the figure legend. Mild -Mild; Sev -Severe B: We have revised the figure so the age cutoff is 60. Thus, one group will include all the ages below or equal to 60 and one group all ages higher than 60 (excluding 60). We have corrected number and signs and the text on page 7: ≤60 and >60.

Reviewer's comment: Fig 1:
C: is a corr coefficient of 0.4 really signaling high reproducibility? Please discuss. And can you also discuss how this variation may influence the variability of results obtained in the later figures?
Response: We agree with the reviewer's comment regarding the correlation coefficient. We re-visited the correlation test that was applied for these data and considered the Pearson correlation test to be a better fit for this analysis. The results for this test were in accordance with the plots and show the correlation coefficients of r=0.8574, 0.6591 and 0.7805 for IgG, IgA, and IgM respectively. Correlation is an effect size and so we can describe the strength of the correlation using the guide that Evans (Evans, J. D. Straightforward Statistics for the Behavioral Sciences. Brooks/Cole Publishing Company, 1996) suggests for the absolute value of r as follows: : 00-.19, "very weak"; .20-.39, "weak"; .40-.59, "moderate"; .60-.79, "strong"; .80-1.0, "very strong". Thus, we have revised the description of the correlation results that can be found in lines 140-143.
Hence, the moderate to strong positive correlation, together with of the P value for the significance test of the correlation coefficient results, suggest that the measured data can be used as a reliable source for the calculation of the longitudinal kinetics.
Reviewer's comment: Fig 2: "Is it possible to add units to y-axes? Did you consider multiple-testing correction for your analyses?" Response: As suggested, y-axis units have been added. With regard to multiple-testing correction for our analysis, the data source that the graph was based on is derived from serum samples that were collected mostly from individuals at one time point; hence, most of the samples are not repeated measurements. To strengthen the statistical power of the analysis, we plotted both repeated measurement data points as well as single measurement data points (which are the majority of the data points). Thus, we decided to apply multiple testing for this analysis as suggested by the reviewer. The data now is analyzed using the generalized additive mixed model (GAMM), which includes the samples that have repeated measures.
Reviewer's comment: Fig 2.c note: C: can you show the data points?
Response: In Fig 3D (previously Figure 2c) we show a bar plot for a simple visual representation of the comparison of the regression coefficients obtained for IgG, IgA, and IgM. In practice, this comparison was carried out by applying a comparison of fits between the resulting regressions (linear or two-phase decay) from the GAMM analysis performed on the cohort and shown in Figure 3A-C. Thus, all the data points (representing antibody levels for each isotype) were plotted in a single graph for the purpose of calculating the multiple "comparison of fits" of the slope obtained for each isotype. Since the graph derived from this comparison is overloaded and visually incomprehensible, we chose to simplify it via the presented bar plot which demonstrates the significant differences between slopes for each isotype obtained by multiple comparison. Nevertheless, the individual data point can be seen in S1 Fig and the "clean" version of the graph remains in the main text ( Figure 3D). We hope the reviewer finds this explanation satisfactory.
Reviewer's comment: "Is 17 individuals in the naive vaccinated and 20 in recovered+vaccinated cohort enough? Please discuss."

Response:
We thank the reviewer for this important observation. Unfortunately, under the ethical restrictions, our only accesses to naïve vaccinated individuals was from the institute's workers who agreed to volunteer to participate in this study. For this reason, our resources were limited, and thus we reached a pool of only 17 individuals. Nonetheless, it is important to note that this cohort was used as a means for a relative comparison and understanding of the recovered cohort and not as the main studied group. We have added an expanded clarification on this point in the revised manuscript, so that the role this cohort in the study will be clear to readers from all disciplines. This revision can be found in the added "Limitations of study", lines 366-372. Nevertheless, the fact that the cohort of naïve vaccinees included many repeat sample points enabled us to enhance the statistical power of the analysis by using a mixed model regression analysis (GAMM).

Reviewer's comment: [Confusingly worded bottom of page 11]
"Based on the calculated temporal kinetics we found that approximately 80% of recovered vaccinees experienced an increase in IgG levels reaching the highest quartile (Q4) as opposed to 50% of them being designated to Q4 following recovery (Figure 4f)" Response: We thank the reviewer for the request for a more understandable explanation of the matter. Now, we have provided a better description of the presented results in the revised manuscript, so that it may be clearer to readers from all disciplines. The revised description can be found in the new version of the submitted manuscript in the "Results" section, lines 267-271.
Reviewer's comment: "Would like to know: What about different vaccination intervals (e.g., 6 weeks instead of 3 in between shots). What about 2*AstraZeneca and AstraZeneca + mRNA booster and single-shot vaccines?"

Response:
We thank the reviewer for this very important question. Indeed, the effects of a different vaccination interval of BNT162b2, vaccination with 2*AstraZeneca vaccines, AstraZeneca + mRNA booster and single-shot vaccines should all be examined over time. However, for ethical reasons, we were not able to request, instruct or encourage donors to receive vaccination in any manner, timing, or frequency, etc. The consistency in the vaccination profile of our cohort derived from the Israel Ministry of Health's vaccination campaign which presented clear and uniform guidelines for Israeli residents. In addition, at the time of this study, vaccination in Israel was only via the Pfizer BNT162b2 mRNA vaccine. The only additional vaccine available was Moderna, which only arrived in Israel in July 2021. Therefore, no alternative vaccine source was available for us to examine at the time of our study.
The different intervals between vaccine doses is a very interesting and important point. For example, the regimen that was used in UK gave better results with regard to the longevity of immunity. The vaccination regimen in Israel was strictly applied in accordance with Pfizer's recommendation, i.e., a three-week interval between doses, thus we could not obtain samples to support the study of this interesting research question. The manuscript has interesting data and would benefit from additional grammatical editing as there are numerous errors throughout. There are also major concerns with the statistical methods, loss-to-follow up and small sample sizes."

Response:
We thank the reviewer for carefully reading, accurately understanding and summarizing the manuscript, and for acknowledging that the report presents interesting data. We have revised the manuscript according to the reviewer's comments while addressing the concerns regarding the errors throughout the manuscript, statistical methods and loss-of-follow up and small sample sizes.

Part II
only novel for this field as a means of serological classification, but it also proved a sufficient and ethical means of incentive for repeat sample donation.
Reviewer's comment: "The vaccine cohort is extremely small, with only 17 individuals, limiting it's usefulness."

Response:
We thank the reviewer for this important observation. Unfortunately, under the ethical restrictions, our only accesses to naïve vaccinated individuals was from the institute's workers who agreed to volunteer to participate in this study. For this reason, our resources were limited and thus we reached a pool of only 17 individuals. Nonetheless, it is important to note that this cohort exists as a means for a relative comparison and understanding of the recovered cohort and not as the main studied group. We have added an expanded clarification on this point in the revised manuscript, so that the role this cohort in the study might be clear to readers from all disciplines. This revision can be found in the added "Limitations of study" sub-section, line 367-373. Nevertheless, as the cohort of naïve vaccinees included many repeat sample points, we were able to enhance the statistical power of the analysis by using a mixed model regression analysis (GAMM).
Reviewer's comment: "It is not clear how the antibody levels were "extrapolated" to day 0 and day 8. Given the small sample size, it would seem to be very error prone."

Response:
We thank the reviewer for the request for a clearer and understandable explanation to our approach. We have addressed this matter in the revised manuscript-see the Results section. To support the understanding of the approach, we have provided an additional supplementary Figure (S5 Fig) and the revised manuscript includes explanation in lines 248-277.
With regards to the small sample size -Indeed, we have added in the revised manuscript a clarification in the "limitations of this study" section, with regard to donor compliance for follow-up sampling, and therefore the limited certainty of the results and conclusions. The clarification of this can be found in the added "Limitations of study" sub-section, line 364-367.

Part III
Reviewer's comment: "Results are presented in the introduction."

Response:
We thank the reviewer for the request for this important clarification. We have provided an expanded clarification on this point in the revised manuscript, so that the reason for this particular comparison might be clear to readers from all disciplines. This expansion can be found in the "Discussion", line 328-338.
Reviewer's comment: "Given the observation that antibody decay rates in recovered patients correlate with quartile titer at V1, such a quartile-dependency in decay rate may be true for naïve vaccinees, too. In this regard, the analysis by Eyran et al. could be enhanced by the inclusion of more naïve vaccinees, especially those whose vaccine-elicited titers reach only Q3, Q2 or Q1. (I am not asking for additional data [unless it happens to exist] but feel that this limitation of the study deserves more airplay in the Discussion)." Response: We thank the reviewer for this important observation. Unfortunately, under our ethical obligations, our only accesses to naïve vaccinated individuals were from institutional employees who agreed to volunteer for this study. For this reason, our resources were limited and thus we reached a pool of only 17 individuals. Nonetheless, as the reviewer suggested, it is important to note this limitation. We have added a note in the discussion section referring to this limitation (line 338-341) as well as in the limitation of study section (lines 370-375).
Reviewer's comment: "The fact that this study lacks analysis of antibody responses directed to the spike ectodomain, and has focused exclusively on the RBD, should be discussed. For instance, the antibody decay rate of S-directed titers between recovered patient's vs naïve vaccinees might be statistically indistinguishable, thus diminishing the key observation manuscript in the current manuscript (namely, a difference in antibody persistence of RBDdirected IgG between these two groups)."

Response:
We thank the reviewer for this insightful note. Indeed, we agree that it is important to clarify the reasoning for the exclusive focus on RBD, rather than the whole spike. We have taken the reviewers comment into consideration and added and additional elaboration on this point, which can be found in the "Introduction" section, lines 66-79.
In addition, we acknowledge the reviewer's concern for the effectiveness of the antibody decay rate of S-directed titers between recovered patient's vs naïve vaccinees. In the revised manuscript we discuss this limitation as well (lines 376-382).

Part III
Reviewer's comment: "There is no callout of Fig. 3c in the caption."

Response:
We thank the reviewer for this note and apologize for the mistake. Indeed, an error was made in Fig 3c), and in general the figures in the revised manuscript have been improved. The required addition can be found in the legend of Fig 5 (formerly Figure 3c).
Reviewer's comment: "It is unclear how exactly the statistical tests were applied in Fig. 3c" Response: We thank the reviewer for this request for clarification on the statistical comparison made in this instance. In Fig 5 (formerly Figure 3c), we present to summary of the results of multiple comparison of the regression coefficients as determined for each isotype (by applying GAMM analysis. We have revised an expanded clarification on this point in the manuscript, so that this comparison might be clear to readers from all disciplines. This clarification can be found in the legend of Fig 5. Reviewer's comment: "The first paragraph of the Discussion is dispensable and could be deleted."

Response:
We agree and have deleted that text.