A pandemic-enabled comparison of discovery platforms demonstrates a naïve antibody library can match the best immune-sourced antibodies

As a result of the SARS-CoV-2 pandemic numerous scientific groups have generated antibodies against a single target: the CoV-2 spike antigen. This has provided an unprecedented opportunity to compare the efficacy of different methods and the specificities and qualities of the antibodies generated by those methods. Generally, the most potent neutralizing antibodies have been generated from convalescent patients and immunized animals, with non-immune phage libraries usually yielding significantly less potent antibodies. Here, we show that it is possible to generate ultra-potent (IC50 < 2 ng/ml) human neutralizing antibodies directly from a unique semisynthetic naïve antibody library format with affinities, developability properties and neutralization activities comparable to the best from hyperimmune sources. This demonstrates that appropriately designed and constructed naïve antibody libraries can effectively compete with immunization to directly provide therapeutic antibodies against a viral pathogen, without the need for immune sources or downstream optimization.


Manuscript details
Tracking number  Submission date 14 April 2021 Decision date 15 June 2021

Title
The unique antibody discovery competition triggered by COVID-19: a naïve library directly delivering antibodies as potent as immune sources

Editorial assessment and review synthesis Editor's summary and assessment
The authors have described antibodies against SARS-CoV-2 spike protein generated via a novel naïve semisynthetic library platform (method unpublished). The method involves embedding CDRs (complement determining regions that mediate antigen recognition/specificity) of the Abs from natural BCRs into clinical Ab scaffolds. The HCDR3 (heavy chain) are amplified from B cell mRNA and the remaining CDRs (LCDR1-3 and HCDR1-2) are identified by NGS of the naive repertoire, filtered using a yeast display and are synthesised on arrays. This HCDR3 is then connected to the rest using a linker and assembled into a scFv (single chain variable fragment).
In the current study, they have compared a panel of anti-S1 Abs generated via this method with Abs generated via other more traditional methods and show that the selected Abs via the semi-synthetic library have a higher neutralization potential than the controls.
Then, they compared the affinities and IC50s for the ten most potent antibodies selected here, with the single most potent antibody from every published report in the same categories (library selected/immune sourced (animals or patient derived). In each case (SPR/psuedovirus neutralization/live virus neutralization), the selected antibodies were comparable with a trend towards enhanced potential.
The strength of the paper is that it shows the potential of antibodies generated via this semi-synthetic library outperforms antibodies generated by other methods.
A weakness of this paper is that their analysis only compares to 5 other control Abs and this is not consistent across experiments. Moreover, the rationale behind choosing these controls is not clear. Finally, the method could be better validated by testing further targets.

Editorial synthesis of reviews
While both reviewers are in general positive about the impact and significance of the findings they raise several serious technical concerns regarding the choice of controls and lack of evidence supporting the claim of improved developability. In addition, there are also concerns about major inconsistencies in the data between the text and the figures.
For further consideration of this study at Nature Communications, the reviewer concerns mentioned above will have to be addressed in full and we would expect the authors to comply with the reviewers requests to perform aggregation/stability studies and add controls for experiments with published antibodies from naïve libraries.
For it to be considered at Communications Biology, we ask that you address all the points too, except for performing the aggregation/ stability studies to strengthen the claims on developability properties or adding additional antibodies (previously published) for your assay.

Editorial recommendation
While we find your study timely and of interest to our readership, we find this manuscript to be an application rather than an introduction of a new method, which precludes publication in Article (or Brief Communication) format. In addition, we find that the performance comparison of antibodies against a single target is not sufficient to justify publication in Nature Methods as an Analysis.
We feel that the study is very timely and the findings are potentially of broad interest, provided that the conclusions can be strengthened as suggested by the reviewers. We agree with the reviewers that the benefits of the newly developed antibodies need to be further demonstrated by including additional published antibodies as controls and providing stability/aggregation data that support the developability properties of the new Abs. In addition, it will be important to address all other reviewer concerns, revise the manuscript and display items accordingly.
We find your study of high interest and useful to the community. While we do ask that you address all the comments of the reviewers, specifically to improve the reproducibility, add controls where prompted, explain the rationale of choosing the control antibodies and other minor comments, we are fine to waive certain referee requests. These include the requests to perform the aggregation/ stability studies (we are happy if you simply tone down the claim) and we also don't ask you to include additional antibodies isolated by other groups working with synthetic naive libraries for your assay.

Minor Revisions
Revision not invited Next Steps

Submission elsewhere
To a journal outside of Nature Portfolio We can share the reviews with another journal outside of the Nature Portfolio if requested. You will need to request that the receiving journal office contacts us at guidedOA@nature.com. We have included editorial guidance below in the reviewer reports and open research evaluation to aid in revising the manuscript for publication elsewhere.

Annotated Reviewer Reports
The editors have included some additional comments on specific points raised by the reviewers below, to clarify requirements for publication in the recommended journal(s). However, please note that all points should be addressed in a revision, even if an editor has not specifically commented on them.

Reviewer #1
This reviewer has not chosen to waive anonymity. The reviewer's identity can only be shared with representatives of an established journal editorial office.

Reviewer #1 expertise
Summarised by the editor Antibody repertoire profiling, systems immunology, protein engineering, vaccine designs, antibody development

Editor's comments about this review
The reviewer agrees on the impact and advance presented in this study which describes potent antibodies generated via a semi-synthetic library. They suggest that the paper would be further strengthened by the addition of more details on the construction of the phage library, as well as more data on factors such as antibody stability and aggregation.
A major concern highlighted by this reviewer are the inconsistencies between the data in the figures and the texts and the lack of details in the tables.

Reviewer #1 comments
Overview Overall significance -The SARS-CoV-2 global pandemic has resulted in many research groups attempting to isolate antibodies against SARS-CoV-2 using a wide variety of different methods. In this manuscript, the authors describe the use of their semi-synthetic naïve scFv library to screen for scFvs that bind to the S1 domain of SARS-CoV-2 spike protein. The library used in the study is constructed using their patented platform that combines antibody scaffolds from clinical mAbs with CDRs (except CDRH3) from BCRs sampled from multiple donors that have sequence liabilities removed to improve the developability properties. The CDRH3 diversity is generated via amplification directly from BCR mRNA. Starting with this semisynthetic library, the authors completed 2 rounds of phage panning before transferring the screened scFv variants into a yeast display system to do 3 additional rounds of FACS sorting. Subsequently, they sequenced 96 clones, which resulted in 31 unique sequences, and they selected 23 to characterize their functional features as full-length IgGs.
Their principal finding is that some of the antibodies they isolated from their naïve scFv library demonstrate extremely high binding affinities (as low as 34 pM to RBD and 38 pM to S trimer) with potent neutralization activities in their lentivirus pseudovirus assay (IC50 of 1.7 ng/mL) as well as live virus assay using Vero cells (IC50 of 1.3 ng/mL). The authors used previously published 5 mAb isolated from B cell profiling of Covid-19 patients as a benchmark, and several of their antibodies demonstrated greater potency than those reference mAbs.
These antibodies represent the best antibodies that have been isolated from naïve libraries, in terms of binding affinities and neuralization activities. Their findings are of great significance that demonstrates the impressive potential of the semisynthetic library they have constructed, especially given their focus on the developability features of the library.

Specific comments # Reviewer comment
Editorial comment 1 Impact -One potential drawback of a synthetic librarybased antibody discovery approach is the lower developability of selected antibodies. The authors address this through their generation 3 antibody library designed to eliminate CDRs with sequence liabilities, and they also demonstrate the antibodies selected from the library having comparable potency as antibodies isolated from B cells of convalescent patients.

2
Strength of the claims -The technical approach for mAb characterization, including detailed binding and neutralization assays, is robust and convincing. On the other hand, there are some key concerns/questions, which are further discussed below (as well as in the "Reproducibility" section).

Construction of the naïve library:
It would strengthen the manuscript to include additional details regarding the construction of the initial phage library used for screening (some of the information included in the accompanying manuscript).
For consideration at Nature Communications, it will be important to provide these details.
The details for construction of phage library as requested here is important for Communications Biology as well.

Developability properties:
While the authors have checked for poly-specificity and sequence liabilities of the selected antibodies, but additional measurements into stability/aggregation (e.g., Tm measurement, SEC, etc.) would further strengthen their claim that naïve antibody libraries can effectively compete with B cell profiling to provide therapeutic antibodies without further downstream optimization.
Nature Communications would expect you to perform the requested stability/aggregation measurements and address the minor points.
For Communications Biology, kindly clarify only the minor points asked here. We do not require the authors to perform the additional measurements into stability, aggregation etc. as requested.

5
Sequencing the 96 clones selected at the end of the screening process resulted in 31 unique sequences. Did the analysis included in Supplementary Figure 2 consider all 31 sequences or the 23 sequences that were selected for further characterization? Figure 2 should include all 31 sequences.

6
Can the authors describe how the 23 antibodies were selected for subsequent biochemical and functional assays? Were all 31 antibodies cloned and tested as IgG but 8 were removed because they didn't convert to the IgG format? The authors should discuss.

7
Control antibodies: CC6.29, CC6.30 (targeting RBD-A), CC12.17, CC12.18 (targeting RBD-B), and CR3022 (targeting RBD-C) were selected as 5 control antibodies to benchmark the antibodies selected from their library. Based on the data presented in the Rogers et al study where these antibodies were characterized, it appears that they were chosen for having high neutralization activities (except CR3022), although they do not have the highest binding affinities. Authors should discuss their rationales for choosing CC6.29/30 for RBD-A and CC12.17/18 for RBD-B. Also, the authors should consider including a couple of previously published antibodies isolated from naïve libraries in their assays.
For Nature Communications, please include published antibodies from naive libraries as additional controls and explain how the control antibodies were selected.
For Communications Biology, kindly explain the rationale behind choosing the five antibodies. We do not ask you to repeat your assay with more previously published antibodies, as requested here. 8 Figure 5 should not include any binding data from antibodies evaluated in non-IgG format (i.e. Fab and scFv) for consistent and unbiased comparisons.
This point must be addressed for both Nature Communications and Communications Biology.

9
Reproducibility -Data description (a major concern): There are numerous inconsistencies regarding how the binding and neutralization data are described in the main texts and presented in Figure 2A/ Table 1, which are detailed below. For clarity, the authors should correct any inconsistencies.
For clarity, the authors should correct any inconsistencies.
Line 116-7: The avidities of the selected antibodies to trimeric spike range from 38 pM to 32.7 nM (not 37 pM to 32 nM) according to Figure 2A. The median values also appear to be inconsistent with the data in Figure 2A.
Line 117-8: 399 pM and 242 pM need to be switched.
For further consideration at either Nature Communications or Communications Biology, it will be essential to address these concerns in full.

10
Some of the figures are not legible (e.g., Supplementary  Figures 3 and 4). Please edit those figures.

11
Supplementary Figure 6 is not referenced in the manuscript.

12
While Figure 4D nicely shows the ratios of neutralization activities of the selected mAbs against the UK and SA variants, please include the neutralization IC50 values for each.

Reviewer #2
This reviewer has not chosen to waive anonymity. The reviewer's identity can only be shared with representatives of an established journal editorial office.

Reviewer #2 expertise
Summarised by the editor Antibody engineering, protein therapeutics, adaptive immune responses, antibody repertoire analysis

Editor's comments about this review
While this reviewer is also positive about the potential impact of this study, they are concerned by the tone of the paper which suggests a competition of antibody discovery methods.
Like reviewer #1, they also recommend addition of more data to evaluate antibody stability and aggregation as well as further discussion of the CDR3 features.

Reviewer #2 comments
Overview Overall significance -Ferrara et al describe the isolation of panel of high affinity and potently neutralizing SARS-CoV2antibodies from a large library of naive scFvs using a combination of phage and yeast display. The authors report the successful isolation of a relatively large panel of high affinity (KD<100 pM) antibodies some of which were shown to be neutralizing and importantly a few also neutralized VOCs including B.1.351. These are interesting and timely findings and point to the utility of the author's naive library and screening and IgG characterization processes.

Specific comments # Reviewer comment Editorial comment 1
Having said that what I find objectionable in the paper is the casting of the story as a "competition" for SARS-CoV2 antibodies compared to other antibody discovery efforts. Frankly, this is pointless and a thinly veiled corporate propaganda. It makes little sense to compare different methods for isolation antibodies that were performed by different labs using very different approaches and with different objectives. For many of the studies referenced here the emphasis was not isolating the highest affinity clones to begin with.

2
Furthermore, why is the isolation of high affinity antibodies from a naive library deemed to be more significant (ie "winning" the "competition") than using screening followed by affinity maturation. For example, using naive libraries + affinity maturation Kossiakoff and coworkers (ref 41) and others have found low pM affinity, neutralizing antibodies.
Before publication it is critical that the paper is revised to recast the storyline and eliminate any mention of competition and simply present the findings for what they are.
Additionally the authors should address the following technical points: While we realize that this comment is rather harsh in tone, we that the claims regarding an "antibody competition" may distract readers from the value of this work. For Nature Communications, we would therefore ask you to substantially tone down or remove these claims.

Major 3
Antibody sequences must be shown. Also some analysis and brief discussion of the CDR3s (length, aa composition etc) should be presented.
Both of these points will have to be addressed for Nature Communications. It will be important to provide additional aggregation/stability data that support the developability claims.
For Communications Biology, please address the first point, regarding adding further analysis on CDR3s. As mentioned previously with R3 (Reviewer #1's comment), we are fine if you tone down the developability claims, without requiring antibody aggregation and stability evaluation.

4
The authors argue that the antibodies isolated are not likely to have developability liabilities. However, this is based solely on sequence analysis and sidesteps perhaps the most critical issues, propensity to aggregation and stability. The authors should at the very least present SEC data to evaluate propensity for aggregation for at least a few of the IgGs (say n=5-6) as well as some stability data.

7
Reproducibility -Need to provide the sequences of the antibodies for reproducibility.

Open Research Evaluation
Data availability

Data Availability statement
Thank you for including a Data Availability statement. However, we noted that you have only indicated that data are available upon request. The data availability statement must make the terpret, verify and extend the research in the article, transparent to readers.
In addition, Nature Portfolio policies include a strong preference for research data to be archived in public repositories. For data types without specific repositories, we recommend that data are deposited in a generalist repository such as figshare or Dryad. More information about our data availability policy can be found here: https://www.nature.com/nature-portfolio/editorialpolicies/reporting-standards#availability-of-data See here for more information about formatting your Data Availability Statement: http://www.springernature.com/gp/authors/research-data-policy/data-availabilitystatements/12330880 In addition, please make antibody sequence data available.
Please cite the link to your publicly available dataset in the reference list

Mandatory data deposition
For your DNA sequencing data, submission to a community-endorsed, public repository is mandatory for publication in a Nature Portfolio journal and is best practice for publication in any venue. Accession numbers must be provided in the paper. Examples of appropriate public repositories are listed below: GenBank Sequence Read Archive (WGS or WES data) The European Nucleotide Archive (ENA) For more information on mandatory data deposition policies at the Nature Portfolio, please visit http://www.nature.com/authors/policies/availability.html#data For an up-to-date list of approved repositories for each mandatory data type, please visit https://www.springernature.com/gp/authors/research-data-policy/repositories/12327124

Other data requests
All source data underlying the graphs and charts presented in the main figures must be made available as Supplementary Data (in Excel or text format) or via a generalist repository (eg, Figshare or Dryad). This is mandatory for publication in a Nature Portfolio journal, but is also best practice for publication in any venue.

Code availability and citation
Please include a statement under the heading "Code Availability", indicating whether and how the custom code/software reported in your study can be accessed, including any restrictions to access. This section should also include information on the versions of any software used, if relevant, and any specific variables or parameters used to generate, test, or process the current dataset. Code availability statements should be provided as a separate section after the Data Availability section.
Upon publication, Nature Portfolio journals consider it best practice to release custom computer code in a way that allows readers to repeat the published results. Code should be deposited in a DOI-minting repository such as Zenodo, Gigantum or Code Ocean and cited in the reference list following the guidelines described in our policy pages (see link below). Authors are encouraged to manage subsequent code versions and to use a license approved by the open source initiative.
Full details about how the code can be accessed and any restrictions must be described in the Code Availability statement.
See here for more information about our code availability policies: https://www.nature.com/natureportfolio/editorial-policies/reporting-standards#availability-of-computer-code We also provide a Code and Software submission checklist that you may find useful: https://www.nature.com/documents/nr-software-policy.pdf Please note: because of advanced features used in this form, you must use Adobe Reader to open the documents and fill it out.

Materials availability
We encourage you to include within the Data Availability statement whether these antibodies can be made available to readers. If so, please also include the email address for requests.

Reporting and reproducibility
Reproducibility Nature Portfolio journals allow unlimited space for Methods. The Methods must contain sufficient detail such that the work could be repeated. It is preferable that all key methods be included in the main manuscript, rather than in the Supplementary Information. Please avoid use of "as described

Statistics
Please see the 'extended comments' file.

Data presentation
The quality of some of the figures appears to be quite low. If possible, we suggest replacing these with higher-resolution images (eg Supp figs 3 & 4).

Other notes
Please include a gating strategy for all flow cytometry experiments.
We have included as an attachment to the decision letter a version of your Reporting Summary with a few notes. This is mainly for your information, but we hope it is helpful when preparing your revised manuscript. If you decide to resubmit the manuscript for further consideration, please be sure to include an updated Reporting Summary.