key: cord-0276244-br8fgpkl
authors: Hernandez, M. M.; Banu, R.; Shrestha, P.; Gonzalez-Reiche, A. S.; van de Guchte, A.; Farrugia, K.; Sebra, R.; PSP Study Group, M. S.; Gitman, M. R.; Nowak, M. D.; Cordon-Cardo, C.; Simon, V.; van Bakel, H.; Sordillo, E. M.; Luna, N.; Ramirez, A.; Castaneda, S.; Patino, L. H.; Ballesteros, N.; Munoz, M.; Ramirez, J. D.; Paniz Mondolfi, A. E.
title: A robust, highly multiplexed mass spectrometry assay to identify SARS-CoV-2 variants
date: 2022-05-29
journal: nan
DOI: 10.1101/2022.05.28.22275691
sha: b48ead4638e96667f6d587b864c0ef4418378ae3
doc_id: 276244
cord_uid: br8fgpkl

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants are characterized by differences in transmissibility and response to therapeutics. Therefore, discriminating among them is vital for surveillance, infection prevention, and patient care. While whole viral genome sequencing (WGS) is the "gold standard" for variant identification, molecular variant panels have become increasingly available. Most, however, are based on limited targets and have not undergone comprehensive evaluation. We assessed the diagnostic performance of the highly multiplexed Agena MassARRAY(R) SARS-CoV-2 Variant Panel v3 to identify variants in a diverse set of 391 SARS-CoV-2 clinical RNA specimens collected across our health systems in New York City, USA as well as in Bogota, Colombia (September 2, 2020 - March 2, 2022). We demonstrate almost perfect levels of interrater agreement between this assay and WGS for 9 of 11 variant calls ({kappa} [≥] 0.856) and 25 of 30 targets ({kappa} [≥] 0.820) tested on the panel. The assay had a high diagnostic sensitivity ([≥]93.67%) for contemporary variants (e.g., Iota, Alpha, Delta, Omicron [BA.1 sublineage]) and a high diagnostic specificity for all 11 variants ([≥]96.15%) and all 30 targets ([≥]94.34%) tested. Moreover, we highlight distinct target patterns that can be utilized to identify variants not yet defined on the panel including the Omicron BA.2 and other sublineages. These findings exemplify the power of highly multiplexed diagnostic panels to accurately call variants and the potential for target result signatures to elucidate new ones.

Since the onset of the coronavirus disease 2019 (COVID-19) pandemic, suboptimal surveillance 72 and diagnostic efforts have not been able to prevent the rapid, unchecked spread of severe acute 73 respiratory syndrome coronavirus 2 (SARS-CoV-2) (1-4). In conjunction with various factors (e.g., 74

We recovered residual viral RNA from all 391 specimens from -80°C storage to undergo testing run. This resulted in a total of 8 wells with 1,500 SARS-CoV-2 genome copies/well and 10ng human 179 . CC-BY-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 May 29, 2022. ; https://doi.org/10.1101/2022.05.28.22275691 doi: medRxiv preprint liver total RNA/well for all positive controls. A negative control of nuclease-free water was included in 180 each RT-PCR run. RT-PCR thermocycler conditions are depicted in Table S1 . 181 RT-PCR products underwent reaction with shrimp alkaline phosphatase (SAP). Directly to each 182 RT-PCR well, a mastermix of 1.53μL nuclease-free water, 0.17μL SAP Buffer, and 0.30μL SAP was 183 added for a total volume of 7μL including 2μL of SAP mastermix. SAP reaction thermocycler 184 conditions are described in Table S2 . 185

Extension products were generated with SARS-CoV-2 Variant v3 Extend Primers using the 186 iPLEX ® Pro Reagent Set. Mastermixes were created as follows: 1.06μL nuclease-free water, 0.20μL 187 iPLEX Buffer Plus, GPR; 0.20μL iPLEX Termination Mix, 0.04μL iPLEX Pro Enzyme, and 0.50μL 188 MassARRAY ® SARS-CoV-2 Variant v3 Extend Primers (E01 or E02. Two microliters of E01 189 mastermix were added to each well amplified by P01 primers, and 2μL E02 mastermix was added to 190 each well amplified by P02 primers for the total extension reaction volume of 9μL. Extension 191 thermocycler conditions are detailed in Table S3 . Table S4 . 202 . CC-BY-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 May 29, 2022.

To compare performance of the panel to the WGS "gold standard", we generated 2x2 204 contingency tables for detected and not detected results of each variant call or target call. To measure the 205 level of agreement between WGS and the variant panel, we performed agreement analyses with kappa 206 (κ) results and 95% confidence intervals (95% CI) using the publicly-available GraphPad Prism web 207 calculator (https://graphpad.com/quickcalcs/kappa2/, last accessed April 20, 2022). Level of agreement 208 was interpreted from kappa values as previously described (39). Interpretations included no (κ < 0), 209 Statistical analyses were performed for variant call results of 12 of the possible 16 variants on the 216 panel. We also performed these analyses for 30 of the 36 possible targets on the panel as clinical 217 specimens that encoded 6 specific amino acid polymorphisms (D80G, Y453F, E484Q, Q493K, N501T, 218 I692V) were not recovered for this study. In addition, we did not recover any specimens that harbored 219 the native D614 amino acid (A23403 nucleotide), and, therefore, we were not able to compute level of 220 agreement or diagnostic specificity for the D614G variant or D614G target calls. In addition, for 221 performance analyses of targets H69_V70del, N439K, and E484K, we excluded specimens that resulted 222 in target dropout as we could not infer nucleotide polymorphisms that caused dropout given that 223 primer/probe sequences are proprietary and not known. 224

. CC-BY-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) Table S4 ]. The remaining 10 genomes with mixed patters of mutations were not deposited 241 into GISAID as single variant consensus genomes could not be resolved (data available upon request). 242 . CC-BY-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 May 29, 2022. ; https://doi.org/10.1101/2022.05.28.22275691 doi: medRxiv preprint

We analyzed a diverse set of 391 SARS-CoV-2 viral RNA specimens that were collected from 244 infected patients over 18 months of the COVID-19 pandemic in the NYC metropolitan area and 245

Colombia. These RNA all underwent WGS that resulted in consensus genomes that comprise 56 distinct 246 phylogenetic Pango lineages and corresponded to 12 of the 16 possible variant calls on the panel (Fig.  247   1C We measured the level of agreement between the overall variant calls from WGS and the variant 261 panel ( Table 1) . Of the 12 panel variants in our study set, we performed agreement analyses on 11. We 262 could not measure level of agreement -or diagnostic sensitivity/specificity -for the isolated D614G 263 result because specimens which concurrently encoded the native D614 amino acid and did not yield any 264 other variant result were not recovered for this study. 265 . CC-BY-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) These did not meet the minimum number of detectable targets to yield the Eta result and only resulted in 276 2-3 of the 4 minimum required targets. These results may be the consequences of RNA degradation over 277 long-term storage. For example, the six discordant specimens encode the H69_V70del by WGS, but all 278 yielded dropout of that target on the panel which further supports this scenario. 279

We also measured the diagnostic sensitivity and specificity of the panel ( Fig. 2A-B (Fig. 2B) . Furthermore, excluding the Zeta variant, the panel 287 results showed high PPVs (≥0.933) and high NPVs (≥0.984) for all variant calls (Table S5) . 288 . CC-BY-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. When we performed agreement analyses on each of these 30 targets ( Table 2) , 25 demonstrated 294 almost perfect levels of agreement (κ ≥ 0.820). The targets with suboptimal levels of agreement for our 295 dataset included D215G (κ = 0; no agreement), L242_244del (κ = 0.568; moderate agreement), N501Y 296 (κ = 0.528; moderate agreement), and K1191N (κ = 0.799; substantial agreement). This low level of 297 agreement may be impacted by small sample sizes tested. Indeed, we only recovered 2-4 specimens that 298 encoded each of the D215G, L242_244del, and K1191N targets, Therefore, small frequencies (e.g., 2-4) 299 of inaccurate calls may explain this result. It is important to note that our study set did not include 300 specimens with the native D614 amino acid (A24303 nucleotide), and level of agreement could not be 301 calculated for the D614G target. 302

Interestingly, for N501Y target, we found that of 158 specimens with the polymorphism by 303 WGS, 79 yielded a false-negative result on the variant panel. All 79 belong to the Omicron (BA.1) 304 variant lineage, and when reanalyzed excluding these BA.1 specimens, interrater agreement was almost 305 perfect (κ = 0.975) ( Table 2) . This suggests genomic variation outside original assay design may impact 306 primer/probe binding and yield distinct target results for novel variants. 307

Across the 30 targets tested, the average diagnostic sensitivity measured was 90.2% (Fig. 2C) . 308

The targets with the lowest sensitivities included D80A (75.00%, 95% CI: 30.06-98.72%), D215G (0%, 309 95% CI: 0.00-56.00%), L242_244del (50.00%, 95% CI: 9.00-91.00%), and N501Y (50.00%, 95% CI: 310 . CC-BY-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 May 29, 2022. ; https://doi.org/10.1101/2022.05.28.22275691 doi: medRxiv preprint 42.00-58.00%). Notably, when all BA.1 specimens were excluded from the analyses, the sensitivity of 311 the N501Y target improved to 100.00% (95% CI: 95.00-100.00%). 312

The variant panel assay demonstrated a high diagnostic specificity across nearly all 30 targets 313 tested in this study (Fig. 2D) . On average, the diagnostic specificity was >99.00% across all of the tested 314 diagnostic targets. Specificity was not calculated for the D614G target because no clinical specimens 315 with the native D614 amino acid were recovered in this study. In addition, across the 30 targets, the 316 PPVs and NPVs were 0.900 and 0.959, respectively (Table S6) . nucleotide sequence encodes these amino acid polymorphisms. In addition to these 5 substitutions, there 342 are 3 Mu specimens which yielded a detectable Q677H target. However, these genomes harbor the 343 G23593 nucleotide which encodes the native Q677 amino acid and suggests this is a nonspecific result 344 or detection of a minor intra-host variant. 345

We also found that the 9 Omicron BA.2 specimens generated a distinct target result signature on 346 the variant panel assay. All 9 resulted in detection of S477N, T478K, N501Y, D614G, and P681H 347 targets as well as dropout of the N439K target. The five detected targets each were confirmed by the 348 presence of the amino acid substitutions in WGS data. We cannot delineate the cause of the N439K 349 target dropout because we do not know primer/probe sequences at the site of the targeted nucleotide 350 substitution at position 22879. However, one can speculate that sequence variation around this region 351 may interfere with primer/probe binding. Indeed, all 9 of the BA.2 specimens harbor the T22882G 352 polymorphism which results in the N440K substitution. Furthermore, the K417N substitution is found in 353 all BA.2 consensus genomes but is only detected in 7 specimens. This may be the result of different 354 nucleic acid quantities across specimens and reflect a limit in analytic sensitivity for the target. 355 . CC-BY-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) target substitution among publicly-available genomes (GISAID) (Fig. 3B) CC-BY-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. Here, we report a comprehensive diagnostic evaluation of one of the highest multiplexed variant 380 panel assays on the market. Based on a diverse cohort of clinical specimens across two continents and a 381 wide timeline of the pandemic, we highlight almost perfect levels of interrater agreement between this 382 assay and the "gold standard" WGS for 9 of 11 variants and 25 of 30 distinct targets tested. The assay 383 has a high diagnostic specificity across all variants (≥96.15%) and all targets (≥94.34%) tested. 384 Furthermore, the panel shows a high diagnostic specificity and sensitivity for contemporary variants in 385 global circulation (e.g., Delta, Omicron (BA.1)). 386

Our study does present some limitations particularly with respect to limited sampling. While the 387 panel has defined target signatures for 16 different variants, we were only able to recover clinical 388 specimens that corresponded to 11 of these variants for testing. Indeed, variants with the lowest level of 389 agreement and diagnostic performance metrics were those with some of the fewest specimens recovered 390 and tested (e.g., Zeta (n = 1), Beta (n = 4), Eta (n = 7). We also did not include specimens from the early 391 . CC-BY-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. Accurate identification of currently circulating and emerging SARS-CoV-2 variants is key to 413 effective pathogen surveillance and providing optimal care to patients. Although WGS is the mainstay 414 . CC-BY-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 May 29, 2022. ; https://doi.org/10.1101/2022.05.28.22275691 doi: medRxiv preprint for surveillance, it is important to consider in the global context of a pandemic, this may not be a 415 realistic technology for many LICs and LMCs (29). Therefore, there is a great need for rapid, cost-416 effective, conventional technologies (e.g., RT-PCR) in the clinical laboratory. However, these require 417 increased diagnostic resolution to adequately capture viral evolution and meet the needs of pathogen 418 surveillance. Thus, highly multiplexed molecular assays such as the one presented benefit from high 419 discriminatory power and are a vital tool to shed light on changing viral dynamics. 420 421 . CC-BY-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. . CC-BY-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) . CC-BY-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) . CC-BY-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) CC-BY-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 May 29, 2022. ; https://doi.org/10.1101/2022.05.28.22275691 doi: medRxiv preprint K 1 1 9 1 N S 8 9 2 A T 7 1 6 I A 7 0 1 V I 6 9 2 V P 6 8 1 H L 5 F S 1 3 I L 1 8 F T 1 9 R H 6 9 _ V 7 0 d e l D 8 0 A D 8 0 G T 9 5 I Y 1 4 4 d e l W 1 5 2 C D 2 1 5 G L 2 4 2 _ L 2 4 4 d e l D 2 5 3 G K 4 1 7 N K 4 1 7 T N 4 3 9 K L 4 5 2 R Y 4 5 3 F S 4 7 7 N T 4 7 8 K E 4 8 4 Q E 4 8 4 K Q 4 9 3 K N 5 0 1 Y N 5 0 1 T A 5 7 0 D D 6 1 4 G Q 6 7 7 H Q 6 7 7 P P 6 8 1 H P 6 8 1 R I 6 9 2 V A 7 0 1 V T 7 1 6 I S 8 9 2 A K 1 1 9 1 N 8/10 . CC-BY-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) . CC-BY-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 May 29, 2022. ; https://doi.org/10.1101/2022.05.28.22275691 doi: medRxiv preprint

L 5 F S 1 3 I L 1 8 F T 1 9 R H 6 9 _ V 7 0 d e l D 8 0 A D 8 0 G T 9 5 I Y 1 4 4 d e l W 1 5 2 C D 2 1 5 G L 2 4 2 _ L 2 4 4 d e l D 2 5 3 G K 4 1 7 N K 4 1 7 T N 4 3 9 K L 4 5 2 R Y 4 5 3 F S 4 7 7 N T 4 7 8 K E 4 8 4 Q E 4 8 4 K Q 4 9 3 K N 5 0 1 Y N 5 0 1 T A 5 7 0 D D 6 1 4 G Q 6 7 7 H Q 6 7 7 P P 6 8 1 H P 6 8 1 R I 6 9 2 V A 7 0 1 V T 7 1 6 I S 8 9 2 A K 1 1 9 1 N BA.5 L 5 F S 1 3 I L 1 8 F T 1 9 R H 6 9 _ V 7 0 d e l D 8 0 A D 8 0 G T 9 5 I Y 1 4 4 d e l W 1 5 2 C D 2 1 5 G L 2 4 2 _ L 2 4 4 d e l D 2 5 3 G K 4 1 7 N K 4 1 7 T N 4 3 9 K L 4 5 2 R Y 4 5 3 F S 4 7 7 N T 4 7 8 K E 4 8 4 Q E 4 8 4 K Q 4 9 3 K N 5 0 1 Y N 5 0 1 T A 5 7 0 D D 6 1 4 G Q 6 7 7 H Q 6 7 7 P P 6 8 1 H P 6 8 1 R I 6 9 2 V A 7 0 1 V T 7 1 6 I S 8 9 2 A K 1 1 9 1 N Omicron Sublineage 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 May 29, 2022. ; https://doi.org/10.1101/2022.05.28.22275691 doi: medRxiv preprint

Spike mutation D614G alters Ramírez JD. 2022. Epidemiological Dynamics of SARS-CoV-2 Variants During Social Protests

Reduced sensitivity of SARS

CoV-2 variant Delta to antibody neutralization

Reduced 622 neutralization of SARS-CoV-2 B.1.617 by vaccine and convalescent serum

Omicron-included mutation-induced 635 changes in epitopes of SARS-CoV-2 spike protein and effectiveness assessments of current 636 antibodies

Evidence that D614G Increases Infectivity of the COVID-19 Virus

Evolutionary and 644 structural analyses of SARS-CoV-2 D614G spike protein mutation now documented worldwide

D614G 647 mutation eventuates in all VOI and VOC in SARS-CoV-2: Is it part of the positive selection FIG 1. Detection of viral variants by the Agena MassARRAY ® SARS-CoV-2 Variant Panel

S gene 714 polymorphisms targeted by the variant panel (lollipops) and corresponding amino acids are depicted 715 below. (B) A color map depicts algorithms of target combinations that define 16 distinct SARS-CoV-2 716 variants on the panel

Target results are depicted as colored cells 721 indicating detectable native (e.g., unchanged from Wuhan-Hu-1 reference) amino acids which do not 722 contribute to the variant target algorithm (grey), detectable native amino acids which do contribute to the 723 algorithm (yellow), detectable amino acid polymorphisms (red), and dropout of the given target 724 polymorphism. (C) Phylogenetic composition of 391 clinical specimen viral RNA recovered for 725 diagnostic evaluation of the variant panel

FIG 2. Diagnostic sensitivity and specificity of the Agena MassARRAY ® SARS-CoV-2 Variant

A) Diagnostic sensitivity and (B) diagnostic specificity of eleven variant calls on the panel are 729 depicted. The number of specimens that correspond with each variant according to WGS are annotated 730 in brackets. Depiction of (C) diagnostic sensitivity and (D) diagnostic specificity of each of thirty 731 distinct panel targets. The number of specimens that correspond with each amino acid polymorphism 732 according to WGS are annotated in brackets for each target. Asterisks (*) indicate targets for which 733 dropout results were excluded from analyses (see Methods). For target N501Y, a separate diagnostic 734 analysis was conducted excluding BA.1 specimens

We thank the members of MSHS MML, Simon, and van