key: cord-0907329-3mecdd5m
authors: Benoit, Patrick; Point, Floriane; Gagnon, Simon; Kaufmann, Daniel E.; Tremblay, Cécile; Harrigan, Richard Paul; Hardy, Isabelle; Coutlée, François; Lapierre, Simon Grandjean
title: Impact of cobas PCR Media Freezing on SARS-CoV-2 Viral RNA Integrity and Whole Genome Sequencing Analyses
date: 2021-02-07
journal: bioRxiv
DOI: 10.1101/2021.02.05.430022
sha: 6bae002ee61511cd0dece1509914943fe0cc25cd
doc_id: 907329
cord_uid: 3mecdd5m

SARS-CoV-2 whole genome sequencing is an important molecular biology tool performed to support many aspects of the response to the pandemic. Freezing of primary clinical nasopharyngeal swab samples and shipment to reference laboratories is usually required since RNA sequencing is rarely available in routine clinical microbiology laboratories where initial diagnosis and support to outbreak investigations occur. The cobas PCR Media transport medium developed by Roche facilitates high throughput analyses on cobas multianalyzer PCR platforms. There is no data on the stability of SARS-CoV-2 RNA after freezing and thawing of clinical samples in this transport medium, but potential denaturing of the molecular template could impair test results. Our objective was to compare the quality and results of SARS-CoV-2 genomic sequencing when performed on fresh or frozen samples in cobas PCR Media. Viral whole genome sequencing was performed using Oxford Nanopore Technologies MinION platform. Genomic coverage and sequencing depth did not significantly differ between fresh and frozen samples (n=10). For samples with lower viral inoculum and PCR cycle threshold above 30, sequencing quality scores and detection of single nucleotide polymorphisms did not differ either. Freezing of cobas PCR Media does not negatively affect the quality of SARS-CoV-2 RNA sequencing results and it is therefore a suitable transport medium for outsourcing sequencing analyses to reference laboratories. Those results support secondary use of diagnostic nasopharyngeal swab material for viral sequencing without requirement for additional clinical samples.

(1-4). Viral whole genome sequencing primarily occurs in reference laboratories and is rarely 47 performed where clinical diagnosis or outbreak investigations happen. Therefore, freezing of 48 primary samples is required prior to viral genomic amplification and sequencing. 49 50 cobas PCR Media is a transport medium developed by Roche that simplifies linkage between pre-51 analytical sampling and analytical testing and is adapted for high throughput analyses on cobas 52 multianalyzer PCR platforms. It contains guanidine hydrochloride which is a denaturing agent 53 used to dissociate nucleoproteins and inactivate RNases. The manufacturer does not recommend 54 freezing the cobas PCR Media because of risks of molecular template denaturation (5). Freezing 55 of other transport media was previously shown not to negatively impact the detection of SARS-56

CoV-2 by . However, it is unknown whether, and how, freezing of cobas PCR Media 57 indeed denatures SARS-CoV-2 RNA and if it negatively affects viral genomic sequencing. 58

In this study, we compared the quality and results of SARS-CoV-2 whole genome sequencing 60 between fresh and frozen samples obtained from combined oral and nasopharyngeal swabs 61 (ONPS). We used matched split samples collected in cobas PCR Media and either processed 62 following collection and storage at 4°C or frozen for one week at -80°C and thawed prior to 63 sequencing. Our protocol and analysis address the necessity for most clinical microbiology Viral RNA was extracted from 0.2 mL of cobas PCR Media using Maxwell® 16 instrument 87 (Promega, Madison, WI, USA) for final elution in 30µL. Viral whole genome sequencing was 88 performed using the ARTIC Network V3 protocol on Oxford Nanopore Technologies (ONT) 89 (Oxford, United Kingdom) MinION® long read sequencing platform. Since its initial publication 90 online in January 2020, the ARTIC protocol has become one of the most widely used approach to 91 SARS-CoV-2 genomic sequencing. This protocol has yielded a significant sequence contribution 92 to the GISAID global database and is currently used for surveillance by many public health 93 agencies (8, 9) . Briefly, genome amplification was performed by reverse transcriptase multiplex 94 PCR using nCoV-2019 V3 primer combinations (Integrated DNA Technologies). This set of 95 primers was previously shown to produce high genomic coverage with low variance on the whole 96 viral genome (8). RT-PCR amplicons were assessed by Qubit® fluorometric DNA Quantification 97 (Thermo Fisher Scientific, Waltham, MA, USA). For samples with post RT-PCR DNA quantity 98 below 250 ng, we omitted the dilution step of the sample in 45 µL of molecular grade water before 99 library preparation as recommended in the ARTIC protocol. Such low inoculums were observed 100 in three samples with both ORF1 a/b and E-gene targets Cts over 30 (samples 1, 2, 3). Sequencing 101 libraries were prepared following ONT protocol for genomic DNA with native barcoding and 102 using 9.4.1 flow cells on the MinION® platform. Raw sequencing reads fast5 files were base called 103 with high accuracy using ONT proprietary software Guppy (v3.4.5). Reads were demultiplexed 104 and filtered using the online available ARTIC network bioinformatic pipeline solution (10). This 105 filtering process includes exclusion of sequencing reads respectively below 400 and above 700 106 base pairs which do not correspond to expected amplicons length resulting from the RT-PCR 107 primer set. Reads were mapped to the Wuhan-Hu-1 SARS-CoV-2 reference genome (GeneBank at positions for which a minimal depth of 20 reads had been achieved were used to generate 110 consensus viral genomic sequences. Potential subpopulations or mixed infections were not 111 considered, and hence a unique consensus sequence was generated for each isolate. 112

We compared mean sequencing Q-scores with corresponding error rates and accuracy, single 115 nucleotide polymorphisms (SNPs) identification and diversity of sequenced alleles on identified 116 SNP genomic positions. We used those later metrics as surrogate markers of post-freezing viral 117 RNA integrity. Q-scores represent ONT's sequencing platform and base calling software internal 118 assessment of sequencing read quality. The Q-score of a given base is defined as Q = 10log10 (e) 119 where (e) is the estimated probability of the base call being wrong. We used a two-tailed paired 120 samples t-test with an alpha value of 0.05 to compare pre-and post-freezing variables. All 121 statistical analyses were performed using GraphPad Prism (San Diego, CA USA). 122

To simulate prospective outbreak investigation, we supplemented the pre-and post-freezing 124 sequence datasets with a back catalog of 50 SARS-CoV-2 genomic sequences from our institution 125 (unpublished data) hence generating two mocked nosocomial viral pangenomes. We 126 independently analyzed both augmented data sets as if searching for potential transmission 127 clusters. Consensus sequences were compared, and phylogenetic trees were built using UGENE 128 (v37) with the PHYLIP Neighbor Joining method without bootstrapping. To simulate national 129 surveillance and assessment of circulating viral clades, we independently compared the pre-and 130 post-freezing sequence datasets with published and well described SARS-CoV-2 reference 131 genomes submitted to Nextstrain (https://nextstrain.org/sars-cov-2/) (11).

All laboratory testing including sequencing and data analyses were performed in Centre 134 

Upon initial testing after maintenance of clinical samples at 4°C in cobas PCR Media, RT-PCR 145 the same targets. No statistically significant difference was observed between pre-and post-148 freezing Cts for the ORF1 a/b target (p-value 0.64). One sample only became positive on the E-149 gene target after freezing. Excluding this sample from the analysis, post-freezing Cts for the E-150 gene target were 1.1 Ct higher after freezing (p-value 0.01) ( Table 1) . 151

No statistically significant difference was observed between the sequencing yields before or after 154 freezing. Indeed, freezing did not negatively impact the total number of sequenced bases and mapped reads with pre-/ post-freezing mean deltas of 11 Mb (p-value 0.57) and 938 reads (p-156 value 0.31) for those key metrics. Also importantly, 20X sequencing depth, allowing for wild type 157 or variant allelic identification within our protocol, was achieved for an average of 83.9% and 158 83.7% of the viral genome respectively before and after freezing (p-value 0.90) ( Table 2) . Such 159 similarity was also observed for all other evaluated depth thresholds (1X, 5X, 10X, 50X). As 160 expected, sequencing data yield, depth and coverage were inversely correlated to the Ct value both 161 in pre-(p-value 0.0007) and post-freezing (p-value 0.0003) samples. Less sequencing data was 162 hence generated in the sub-group of low viral inoculum and high Ct samples 1, 2 and 3 but freezing 163 did not negatively impact sequencing yields in this subgroup either (Fig. 1) . 164

No statistically significant decrease was observed in Q-scores (p-value 0.07) and base call 166 accuracy (p-value 0.10) after freezing (Table 3) . Except for samples 1 (ORF1 a/b Ct 34.74) and 2 167 (ORF1 a/b Ct 32.16), freezing did not impact SNP detection and identified mutations were 168 identical in both sequencing analyses. Looking in more depth at each single read for those specific 169 mutation sites, the percentage of alternate bases leading to SNP calling did not significantly change 170 after freezing (p-value 0.31). This ratio of variant versus wild type alleles at each mutation site 171 was the same before and after freezing and suggests that the viral molecular template was not 172 significantly degraded by the freezing process in cobas PCR Media. 173

In the mocked outbreak investigation, samples with higher genomic similarity were identified. previously shown that freezing of cervical samples in cobas PCR Media did not negatively impact 191 the sensitivity of PCR for human papillomavirus detection (12). Our study confirms this holds true 192 for SARS-CoV-2. To our knowledge, this is the first study to formally evaluate the impact of 193 freezing clinical samples in cobas PCR Media for downstream sequencing analyses either for 194 SARS-CoV-2 or for any other target pathogen or molecular template. Our study confirms the 195 ability of cobas PCR Media to maintain SARS-CoV-2 genomic RNA at -80°C for subsequent 196 sequencing analyses. Note that the PCR amplicons generated in this study are relatively small 197 (~400 bp), so this protocol may be more robust to RNA damage than methods which require long, 198 intact starting molecules. Our results should also not be generalized to other transport media 199 without independent confirmation. 200

Our study included three samples with RT-PCR Cts above 30.0 which are considered to have a 202 lower viral load. For those samples, SNP calling showed variability and genomic coverage was 203 insufficient to allow detailed phylogenetic analyses. This phenomenon was observed both before 204 and after freezing and is hence believed to be due to low viral inoculum rather than transport 205 medium related viral RNA denaturation. Our study included only 10 samples but the extensive 206 comparability between pre-and post-freezing sequencing results suggests that a higher 207 denominator would not have led to different conclusions. It is possible that a freezing period longer 208 than 7 days would have led to worse sequencing results after thawing but our protocol did not 209 assess such longer-term effect. Seven days represents a sufficient delay for transportation to 210 reference laboratories performing viral sequencing and our study hence provides meaningful 211 information to clinical laboratories involved in routine diagnostic testing. 

Persistence and Evolution of SARS-233

CoV-2 in an Immunocompromised Host