Prevalence of antibodies to SARS-CoV-2 in healthy blood donors in New York

TL;DR: Results are consistent with seroprevalence studies within the region and with reports that SARS-COV-2 infections can be asymptomatic or cause only mild symptoms, and may represent a source for convalescent plasma donors.

Abstract: Despite the high level of morbidity and mortality worldwide, there is increasing evidence for asymptomatic carriers of the novel coronavirus SARS-CoV-2. We analyzed blood specimens from 1,559 healthy blood donors, collected in the greater New York metropolitan area between the months of March and July 2020 for antibodies to SARS-CoV-2 virus. Using our proprietary technology, SERA (Serum Epitope Repertoire Analysis), we observed a significant increase in SARS-CoV-2 seropositivity rates over the four-month period, from 0% [95% CI: 0 - 1.5%] (March) to 11.6% [6.0 - 21.2%] (July). Follow-up ELISA tests using S1 and nucleocapsid viral proteins confirmed most of these results. Our findings are consistent with seroprevalence studies within the region and with reports that SARS-COV-2 infections can be asymptomatic or cause only mild symptoms. IMPORTANCE The COVID-19 pandemic, caused by the novel coronavirus SARS-CoV-2, has caused vast morbidity and mortality worldwide, yet several studies indicate that there may be a significant number of infected people who are asymptomatic or exhibit mild symptoms. In this study, samples were collected from healthy blood donors in a region of rapidly increasing disease burden (New York metropolitan area) and we hypothesized that a subset would be seropositive to SARS-CoV-2. People who experienced mild or no symptoms during SARS-CoV-2 infection may represent a source for convalescent plasma donors.

Summary

Introduction

• Cloud computing provides on-demand access to IT resources via the Internet.
• The convenience of accessing resources in the cloud is made secure by user-specified access control policies.
• The property to be verified is specified in the policy language itself, eliminating the need for a different specification or formalism for properties.
• (exactly one character) in the string constraints makes the decision problem PSPACE-complete.
• ZELKOVA is the underlying policy analysis engine for a growing number of AWS services.

A. Related work

• Policy languages have been used in a variety of domains, e.g., trust management, distributed authorization, role-based access, access control of resources [1]–[6].
• The AWS policy language is defined with respect to a JSON serialization, and is designed to be used across various cloud services and scenarios of access control.
• ZELKOVA combines all the components of the policy language in a single analysis tool.
• Bounding the analyses, however, makes it unsound.
• The language is robust and flexible, with features that have arisen from practical needs.

II. APPROACH

• When an access request is made to an AWS service, a request context is generated which includes the principal making the request, the resource being requested, and the specific action being requested.
• A policy evaluation engine compares this request context against the policies for the user and the resource to determine if access is granted or denied.
• The fundamental mechanism of ZELKOVA is the ability to say if one policy is less-or-equallypermissive than another.
• Properties can be specified as boundary policies that represent either upper or lower bounds on desired behavior.
• ZELKOVA’s less-or-equally-permissive check then establishes the correctness of these bounds or finds a counterexample.

A. Policy language overview

• The AWS policy language is defined as serialized JSON1, however, in this paper the authors describe the core constructs of the policy language in a simplified abstract syntax.
• Fig. 1 shows the abstract syntax for the policy language.
• In other words, to get access to a resource, there must be some allow statement that grants access and no deny statement that revokes that access.
• Various AWS services publish the set of actions that can be performed by the user for the resources specific to those services.
• The operators are applied to condition keys .

B. Example

• A policy in the simplified abstract syntax for the Amazon Simple Storage Service (S3) is shown in Fig. S3 stores data as objects in these buckets.
• Fig. 3 shows the encoding of the policies from Fig. 2.
• Note that the authors are abusing notation in Y0 to say r = “cs240/*” since this, in fact, will correspond to a form of string matching rather than equality.
• Policy Y additionally grants principals other than “students” and “tas” access to the resources in the bucket “cs240”, since the deny statement only denies “students” access to the “Answer.pdf”.

III. SMT ENCODING

• The authors describe ZELKOVA’s SMT encoding.
• (1) Here Allow and Deny are the set of allow and deny statements in a policy.
• Similarly, A(S) and R(S) return the string values for the actions and resources in the statement.
• The meaning of a condition is encoded by a disjunction over all the listed values.

A. String constraints

• The encoding of policies in ZELKOVA is largely through the use of string constraints.
• The principal, action, and resources constructs in the policy are encoded as string constraints.
• The string operator StringLike is applied to the condition key s3:prefix with a value of “grades/∗”, which limits access so that only objects under the “grades/” directory may be listed.
• The Boolean variables vpcExists and s3PrefixExists encode whether the conditions aws:sourceVpc and s3:prefix are present in the request context.

B. Regular expression constraints

• More complicated string constraints require a more powerful encoding.
• The difficult case is when a condition key is constrained with both case sensitive and case insensitive operators.
• Consider the contrived combinations of conditions in shown in Fig.

C. Bit vector constraints

• The IpAddress condition operator allows users to restrict access based on IP addresses.
• The IpAddress operator is used in combination with the aws:SourceIp condition.
• A bitwise AND operation is used to mask the insignificant bits of the IP address in the constraint.
• Thus, request contexts that are allowed by C0 are also allowed by C1.

D. Other operators

• The conditions on numeric operators only perform integer comparisons.
• There are no arithmetic operations in the policy language and no interactions between numeric values and string values, e.g., you cannot take the length of a string.
• The conditions applicable to the Boolean operators are simply encoded as Boolean constraints.
• Conditions with the IfExists suffix check existence of the condition key in the request context.
• The resulting operator can be applied to the aws:sourceVpc condition key.

IV. Z3AUTOMATA

• Z3AUTOMATA is an in-house extension of Z3 designed to provide a complete decision procedure for the theory of regular expressions.
• The SMT problems that ZELKOVA generates contain a mix of both simple and complex string constraints.
• Z3 and CVC4 can easily and efficiently handle that case, thus Z3AUTOMATA never wins.
• Z3AUTOMATA solves regular expression problems using the standard translation to deterministic finite automata (DFAs) via non-deterministic finite automata (NFAs).

V. ZELKOVA PROPERTIES

• Organizations using cloud services want assurances that policies being authored or modified by users do not violate general security best-practices, adhere to the security guidelines defined by the organization, and do not deny access to the intended users.
• Examples of these properties are as follows: “Ensure that unrestricted public write is not allowed to a particular resource.” (security best-practice), “Ensure access to a resource is only allowed from a certain range of IP addresses.” (organizational security check), and “Ensure a particular user is allowed to perform a specific action on a resource” (availability property).
• These properties can be specified in the policy language and checked by ZELKOVA.
• Verification of properties by ZELKOVA provides assurance that there are no inappropriately configured resources within an organization.

A. Organizational security checks

• The authors use the example in Section II to describe how an organization can specify a property in the policy language such that it can be checked by ZELKOVA.
• As a safeguard measure, suppose, the University administrator wants to ensure that there is no unauthorized access to data in the buckets.
• The first allow statement in Fig. 9 permits getObject only when the request comes from vpc-111bbb222.
• This bound will only be violated if the input policy allows a request which Fig. 9 does not allow.
• On the other hand, if ZELKOVA shows that the input policy implies the policy in Fig. 9 then the upper bound is establish and the proposed property holds true.

B. Security best-practices

• ZELKOVA supports several built-in checks that can be leveraged to check a variety of security best-practices.
• These checks are used internally by AWS to check adherence to security best practices and also available to external customers through services such as Amazon Macie, AWS Config, AWS Trusted Advisor, and the Amazon S3 console.
• Here, the operator ForAllValues:ArnEquals is applied to the condition aws:sourceArn whose value is restricted to mytopic.
• For each of the built-in checks, ZELKOVA takes a policy and returns true or false based on whether the check is satisfied.
• Experiments with other solvers such as Z3Str3 [18] and other automata-based solvers [19] is part of their future work.

B. Usage statistics

• The total number of invocations of ZELKOVA ranges from a few million to tens of millions in a single day.
• The number of invocations varies based on the services invoking ZELKOVA.
• Fig. 13 shows the performance of ZELKOVA on one million randomly selected policy questions.
• The total time includes time to parse the input JSON, encode the policies into SMT, perform the check, and construct the resulting JSON that is returned.
• The y-axis represent the count, i.e., number of policies solved within the time.

VII. CONCLUSION

• The authors have presented a formalization of the AWS policy language that controls access to resources.
• Given the distributed nature of the policy language where different services establish their own list of condition keys, this work provides a single consolidated service to reason about the semantics of policies applicable across different services in AWS.
• Alternatively, given a concrete request context, the policy evaluation engine allows users to test access control.
• In contrast, their formalization into SMT provides the ability to soundly reason about properties of a policy for all valid request contexts.
• One avenue is to improve the existing functionality provided in ZELKOVA.

Figures

Figure 2 148

Figure 1 145

Full text

1
Prevalence of antibodies to SARS-CoV-2 in healthy blood donors in New York 1
2
Kathy Kamath
1
, Elisabeth Baum-Jones
1
, Gregory Jordan
1
, Winston Haynes
1
, Rebecca Waitz
1
, John Shon
1
, Steve 3
Kujawa
1
, Lyn Fitzgibbons
2
, Debra Kessler
3
, Larry Luchsinger
3
, Yale IMPACT Team
4
, Patrick Daugherty
1
4
5
1
Serimmune, Inc., 150 Castilian Dr, Suite 100, Goleta, CA 93117 USA 6
2
Santa Barbara Cottage Hospital, 2400 Bath St., Santa Barbara, CA 93105 USA 7
3
New York Blood Center 310 East 67
th
St., New York, NY 10065 USA 8
4
Yale University School of Medicine 60 College St., New Haven, CT 06510 USA 9
10
Corresponding author: Kathy Kamath 11
Email: kathy@serimmune.com 12
Telephone: +1 (805) 448-6529 13
14
Members of the Yale IMPACT Team
4
15
Kelly Anastasio, Michael H. Askenase, Maria Batsu, Santos Bermejo, Kristina Brower, Molly L. Bucklin, Staci Cahill, 16
Melissa Campbell, Yiyun Cao, Arnau Casanovas-Massana, Rupak Datta, Giuseppe DeIuliis, Charles Dela Cruz, 17
Rebecca Earnest, Shelli Farhadian, John Fournier, Bertie Geng, Nathan Grubaugh, Ryan Handoko, Akiko Iwasaki, 18
William Khoury-Hanold, Lynda Knaggs, Albert Ko, Maxine Kuang, Sarah Lapidus, Anjelica Martin, Irene Matos, 19
David McDonald, Maksym Minasyan, Adam J. Moore, M. Catherine Muenker, Nida Naushad, Allison Nelson, 20
Jessica Nouws, Angela Nunez, Hong-Jai Park, Xiaohua Peng, Tyler Rice, Kadi-Ann Rose, Lorenzo Sewanan, Lokesh 21
Sharma, Denise Shepard, Mikhail Smolgovsky, Nicole Sonnert, Yvette Strong, Codruta Todeasa, Jordan Valdez, 22
Sofia Velazquez, Pavithra Vijayakumar, Elizabeth B. White, Anne L. Wyllie 23
24
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprintthis version posted October 21, 2020. ; https://doi.org/10.1101/2020.10.19.20215368doi: medRxiv preprint
NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.

2
ABSTRACT 25
Despite the high level of morbidity and mortality worldwide, there is increasing evidence for asymptomatic 26
carriers of the novel coronavirus SARS-CoV-2. We analyzed blood specimens from 1,559 healthy blood donors, 27
collected in the greater New York metropolitan area between the months of March and July 2020 for antibodies 28
to SARS-CoV-2 virus. Using our proprietary technology, SERA (Serum Epitope Repertoire Analysis), we observed a 29
significant increase in SARS-CoV-2 seropositivity rates over the four-month period, from 0% [95% CI: 0 - 1.5%] 30
(March) to 11.6% [6.0 - 21.2%] (July). Follow-up ELISA tests using S1 and nucleocapsid viral proteins confirmed 31
most of these results. Our findings are consistent with seroprevalence studies within the region and with reports 32
that SARS-COV-2 infections can be asymptomatic or cause only mild symptoms. 33
34
IMPORTANCE 35
The COVID-19 pandemic, caused by the novel coronavirus SARS-CoV-2, has caused vast morbidity and mortality 36
worldwide, yet several studies indicate that there may be a significant number of infected people who are 37
asymptomatic or exhibit mild symptoms. In this study, samples were collected from healthy blood donors in a 38
region of rapidly increasing disease burden (New York metropolitan area) and we hypothesized that a 39
subset would be seropositive to SARS-CoV-2. People who experienced mild or no symptoms during SARS-CoV-2 40
infection may represent a source for convalescent plasma donors. 41
42
OBSERVATION 43
The COVID-19 pandemic, caused by the novel coronavirus SARS-CoV-2, has exerted high morbidity worldwide, 44
with mortality rates reported greater than 10% for some age groups [1]. Yet several studies have indicated that 45
there may be a significant number of people who are asymptomatic or exhibit mild symptoms [2 - 4], including in 46
the New York metropolitan region [5, 6]. We developed an assay to detect IgG and IgM antibodies to SARS-CoV-47
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprintthis version posted October 21, 2020. ; https://doi.org/10.1101/2020.10.19.20215368doi: medRxiv preprint

3
2 virus and analyzed plasma samples from healthy blood donors collected during the early months of the
48
pandemic in that region. 49
50
Development of SERA assay for antibodies to SARS-CoV-2 51
Our technology, Serum Epitope Repertoire Analysis (SERA) enables discovery and semi-quantitative detection of 52
antibody epitopes with high resolution that can be mapped to eliciting antigens and organisms. The SERA assay 53
platform has been described in detail elsewhere [7], but briefly, it utilizes a random bacterial display 12-mer 54
peptide library of 10
10
diversity in conjunction with NGS to collect and sequence the set of peptides that represent 55
an individual’s serological antibody epitope repertoire. The shared epitopes targeted by antibody repertoires are 56
then identified by custom bioinformatics algorithms [8], using cohort-based subtractive analysis of the mimotope 57
repertoires. For the development of SERA panels to detect IgG and IgM antibodies against SARS-CoV-2, serum 58
samples from confirmed COVID-19 patients (positive for SARS-CoV-2 nucleic assay testing (NAT)) were acquired 59
from the Yale IMPACT biorepository and the Santa Barbara Cottage Hospital (Santa Barbara, CA) in accordance 60
with BRISQ guidelines [9] and the SBCH IRB. The majority of subjects were hospitalized and serum samples were 61
collected between 0 and 57 days post symptom onset. For the discovery of SARS-CoV-2 IgG and IgM antibody 62
epitope motifs, we applied SERA to a Discovery cohort of 164 COVID-19 samples, and 430 (for IgM panel) and 497 63
(for IgG panel) pre-pandemic healthy controls (Table S1). The resulting epitope motifs were compiled into IgG and 64
IgM panels, and the enrichment of each panel motif was normalized and summed to generate a composite 65
diagnostic score [7]. A seropositivity cut-off value of 25 was established to yield a specificity of > 99% for either 66
panel. In an independent verification set of 250 COVID-19 specimens for which we had days since symptom onset 67
data, the sensitivity of SERA IgG and IgM at >10 days post symptom onset was ≥ 91% (Figure 1). In a cohort of 68
1,500 pre-pandemic samples, the specificity of the SERA SARS-CoV-2 IgG panel was 99.3% and the IgM panel was 69
99.1%; the specificity of the IgG and IgM panels together was 98.7%. 70
71
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprintthis version posted October 21, 2020. ; https://doi.org/10.1101/2020.10.19.20215368doi: medRxiv preprint

4
Sample collection from healthy blood donors 72
We initiated a collaboration with Blood Centers of America to collect plasma from healthy donors throughout the 73
US with a primary objective of increasing our healthy subject cohort. Collection and screening began in New York 74
at the New York Blood Center (NYBC) in March 2020. Individuals who were suspected of having COVID-19 were 75
deferred from donation for 14 days after resolution of symptoms. For this study, a total of 1,559 whole blood 76
samples from healthy blood donors were sent to our laboratory within 2 weeks of collection, in seven shipments. 77
The demographics of these samples are shown in Table S2. Upon arrival at our laboratory, blood plasma was 78
isolated, screened by SERA assay and the results were uploaded to the epitope repertoire database. 79
80
Prevalence of antibodies to SARS-CoV-2 in blood donors 81
For each sample analyzed using the SERA assay, the epitope repertoires are compiled into a database that enables 82
assessment of any previously or future developed disease-specific SERA panel, without the need to re-test the 83
sample. We applied the SARS-CoV-2 IgG panel retrospectively to samples collected during the 84
period March 1 through July 7, 2020. Of 1,559 blood samples processed, a total of 68 had IgG antibodies against 85
SARS-CoV-2 on the SERA assay. As collection proceeded, the seropositive rate increased from 0% [95% CI: 0 - 86
1.5%] to 11.6% [6.0 - 21.2%] (Figure 2, Table S3). Eighty-four percent (57/68) of these samples were also positive 87
for IgG antibodies to SARS-CoV-2 Spike S1 and nucleocapsid proteins by follow-up ELISAs (Supplemental Text 1), 88
providing independent verification of the presence of a SARS-CoV-2 antibody response in these subjects. 89
90
Because the age of donors were not equally distributed amongst the sample shipments, we interrogated whether 91
age or shipment date was a stronger predictor of the seropositivity observed. GLM regression analysis of SARS-92
CoV-2 seropositivity relative to donor age or shipment date revealed that sample shipment date contributed more 93
significantly (p=5.5e-8) to the prediction of seropositivity than age (p=0.35) (Supplemental Figure 1). 94
95
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprintthis version posted October 21, 2020. ; https://doi.org/10.1101/2020.10.19.20215368doi: medRxiv preprint

5
Concluding remarks
96
This study provides insight into trends of rate of seropositivity to SARS-CoV-2 among asymptomatic blood donors 97
in the metropolitan New York area during the early months of the COVID-19 pandemic but was not designed to 98
measure population-wide seroprevalence. Nonetheless, our results were consistent with reports that a significant 99
number of SARS-CoV-2 infected people in this region may be asymptomatic or present with mild symptoms [5, 6]. 100
101
Here we demonstrated the utility of SERA for serosurveillance studies of emerging infections through 102
retrospective analysis of archived data in populations. Previously collected data was simply re-analyzed for SARS-103
CoV-2 associated epitopes, thus mitigating the need to store samples or perform additional assays in a surveillance 104
setting. In addition, the new panel was applied retrospectively in silico to large pre-pandemic cohorts in panel 105
development and verification phases, demonstrating high specificity without the need to perform additional 106
immunoassays. Finally, applying the panel to cohorts afforded the opportunity to identify subjects with 107
presumptive asymptomatic or mild disease for research purposes amongst the blood donor population from New 108
York and surrounding areas. 109
110
Our results raise the question of whether a larger number of convalescent plasma donors could be identified 111
among healthy, asymptomatic donors. However, there is concern about providing antibody results to blood 112
donors without an improved understanding of their immunological significance. We also demonstrate the utility 113
of SERA for monitoring the antibody response to emerging infections. 114
115
116
REFERENCES 117
1. Petersen E, Koopmans M, Go U, et al. Comparing SARS-CoV-2 with SARS-CoV and influenza pandemics. 118
Lancet Infect Dis. 2020 ; 20:e238-e244 119
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprintthis version posted October 21, 2020. ; https://doi.org/10.1101/2020.10.19.20215368doi: medRxiv preprint

Frequently asked questions

Q1. What are the contributions in "Prevalence of antibodies to sars-cov-2 in healthy blood donors in new york" ?

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

Q2. What is the significant factor contributing to the increased positivity rates?

These results suggest that shipment date, not the age bias, 174 is the dominant factor contributing to the increased positivity rates. 

Q3. What is the sensitivity of the SERA assay?

Specificity on a cohort of 1,500 pre-159 pandemic controls was 99.3% for IgG, 99.1% for IgM, 98.7% for IgG or IgM, and 100% if a case was positive for IgG 160 and IgM (no pre-pandemic validation controls were positive for both IgG and IgM concomitantly). 

Q4. How many people were able to identify SARS-CoV-2?

applying the panel to cohorts afforded the opportunity to identify subjects with 107 presumptive asymptomatic or mild disease for research purposes amongst the blood donor population from New 108 York and surrounding areas.