key: cord-0690726-iobtq1za
authors: Pearson, Joel D.; Trcka, Daniel; Hyduk, Sharon J.; Aynaud, Marie-Ming; Hernández, J. Javier; Peidis, Filippos; Lu, Suying; Chan, Kin; Woodgett, Jim; Mazzulli, Tony; Attisano, Liliana; Pelletier, Laurence; Cybulsky, Myron I.; Wrana, Jeffrey L.; Bremner, Rod
title: Comparison of SARS-CoV-2 Indirect and Direct Detection Methods
date: 2020-05-13
journal: bioRxiv
DOI: 10.1101/2020.05.12.092387
sha: c65a438ff7af6a37c489d42c3ad6dcf71cc03aab
doc_id: 690726
cord_uid: iobtq1za

The COVID-19 pandemic caused by the SARS-CoV-2 virus has placed extensive strain on RNA isolation and RT-qPCR reagents. Rapid development of new test kits has helped to alleviate these shortages. However, comparisons of these new detection systems are largely lacking. Here, we compare indirect methods that require RNA extraction, and direct RT-qPCR on patient samples. For RNA isolation we compared four different companies (Qiagen, Invitrogen, BGI and Norgen Biotek). For detection we compared two recently developed Taqman-based modules (BGI and Norgen Biotek), a SYBR green-based approach (NEB Luna Universal One-Step Kit) with published and newly-developed primers, and clinical results (Seegene STARMag RNA extraction system and Allplex 2019-nCoV RT-qPCR assay). Most RNA isolation procedures performed similarly, and while all RT-qPCR modules effectively detected purified viral RNA, the BGI system proved most sensitive, generating comparable results to clinical diagnostic data, and identifying samples ranging from 65 copies – 2.1×105 copies of viral Orf1ab/μl. However, the BGI detection system is ∼4x more expensive than other options tested here. With direct RT-qPCR we found that simply adding RNase inhibitor greatly improved sensitivity, without need for any other treatments (e.g. lysis buffers or boiling). The best direct methods were ∼10 fold less sensitive than indirect methods, but reduce sample handling, as well as assay time and cost. These studies will help guide the selection of COVID-19 detection systems and provide a framework for the comparison of additional systems.

The SARS-CoV-2 coronavirus is a positive-strand RNA virus with a large genome of about 30kb, 41 which encodes up to 14 open reading frames, including several structural genes; e.g. Nucleocapsid 42 (N), Spike (S), Membrane (M) and Envelope (E), accessory genes, and a large open reading frame 43 (Orf1a/Orf1ab) that encodes a polypeptide that is cleaved into 16 non-structural proteins (1, 2) . It is 44 related to the SARS-CoV and MERS-CoV coronaviruses, which cause severe respiratory illness in 45 humans, and is the causative agent of the COVID-19 respiratory disease (3). Since the first 46 documented case in Wuhan, China in December 2019, the virus has spread rapidly across the globe. 47

On March 11, 2020, the WHO officially declared COVID-19 a pandemic (4, 5). As of May 12, 48 2020, there have been over 4.2 million reported cases of COVID-19 and over 286,000 deaths 49 worldwide (6). 50

The wide range of disease symptoms, including a large portion of mildly or asymptomatic people, 51 has facilitated rapid dissemination (7, 8) . Efficient diagnosis, allowing rapid and accurate patient 52 testing remains the key to limiting disease spread. Rapid disease spread has strained the capacity of 53 diagnostic facilities, and the availability of standard reagents. The principle means of diagnostics for 54 COVID-19 relies on RNA extraction from nasal swabs followed by reverse transcriptase-55 quantitative PCR (RT-qPCR) detection of viral genes (e.g. N, E and RdRp). Rapid development of 56 SARS-CoV-2 RT-qPCR detection systems from many companies has helped to alleviate some of 57 the strain, and many new systems have been given Emergency Use Authorization (EUA) for clinical 58 use. However, comparison of new systems with clinical diagnostics is largely lacking. A limited 59 number of studies have evaluated some kits and compared efficiency of different RT-qPCR primer 60 sets for COVID-19 detection (9-12). These studies have revealed large differences in sensitivity of 61 95°C for 5 min or treated with MyPOLS Bio VolcanoCell2G lysis buffer, 1% Triton X-100, 1% 138

Tween-20 or 1% Saponin and incubated on ice 15min. Samples were then directed added to the qPCR reaction mixture and compared to UTM samples that had been left untreated. 140

Plate format. Many diagnostic protocols utilize 20 µl reactions in 96-well plates, but reducing 142 volume in a 384-well format increases throughput and reduces costs. Using the Norgen RT-qPCR 143 COVID-19 detection kit (which utilizes CDC-approved N1 and N2 primers), we observed similar 144

Ct values in a comparison of 20 vs 10 µl reactions in 96-or 384-well plates, respectively (Fig. 1A) , 145 thus in subsequent analysis we focused on 384-well plates. 146 RNA Extraction Methods. Qiagen RNA extraction systems are used extensively for viral RNA 147 isolation, but availability has become limited. Thus, we first compared the Qiagen RNeasy RNA 148 extraction kit to a similar kit from Norgen Biotek, both of which utilize silica-based columns. None 149 of the SARS-CoV-2-negative samples generated any signal, and we detected no significant 150 difference in Ct values across four clinically-diagnosed positive patient samples (Fig. 1B) , thus the 151

Norgen extraction system performs similarly to standard Qiagen kits. We next compared efficiency 152 of the Norgen (column based), Invitrogen Purelink (column-based) and BGI (magnetic bead-based) 153 RNA isolation systems. Using two new positive patient samples, we observed similar recovery with 154 both the Norgen and BGI systems, but considerably higher Ct values were observed for viral (N1 155 and N2 primers) and human control (RNase P) genes with the Invitrogen kit (Fig. 1C) . Thus, for 156 isolating SARS-CoV-2 RNA from nasopharyngeal patient samples in UTM, Norgen, Qiagen 157

RNeasy and BGI extraction methods are all comparable, but the Invitrogen Purelink kit is less 158 efficient. 159

TaqMan Primers/RT-qPCR mix. Next we compared the efficiency of two recently developed 160

TaqMan-based SARS-CoV-2 RT-qPCR detection kits from Norgen and BGI. The BGI protocol 161 uses one primer set against Orf1ab, while as noted above, Norgen uses two separate reactions 162 targeting the N gene. We first compared RNA from the two positive patient samples extracted with 163 the Norgen, BGI or Invitrogen Purelink methods, as well as RNA isolated from one negative patient 164 sample using the BGI extraction method. These assays confirmed that Norgen/BGI extraction 165 methods are more efficient than Purelink, but also revealed greater sensitivity (~1-3 Ct values) with 166 BGI vs Norgen primers/RT-qPCR mix, particularly with the lowest level sample (L028-Purelink) 167 ( Fig. 2A) . We observed a similar trend across seven additional samples all isolated using the 168 Norgen RNA extraction kit (Fig. 2B ). Comparison to clinical values for the viral N, E and RdRp 169 genes obtained using the Seegene STARMag RNA extraction kit and Allplex 2019-nCoV RT-qPCR 170 assay analyzed using the Bio-Rad CFX96 IVD real-time qPCR detection system, showed a strong 171 correlation with Ct values obtained using either the BGI or Norgen detection modules (Fig. 2C) . We 172 observed no significant difference between the BGI and clinical values, although there was a trend 173 toward lower Ct values with the clinical lab E gene primers/probe, particularly with higher-level 174 samples (Fig. 2C ). In contrast, the Norgen detection module showed significantly higher Ct values 175 compared to clinical data (median 1.4 to 3.5 Ct values higher depending on the primer sets used) 176 ( Fig. 2C) , similar to what we observed in comparison to the BGI detection system. Using synthetic 177 TWIST Bioscience SARS-CoV-2 standards we found that the BGI detection kit routinely detected 178 2.5 copies/µl (6.25 copies/10µl rxn), and 1 copy/µl (2.5 copies/10µl rxn) 80% of the time, whereas 179 the detection limit of the Norgen system was 10 copies/µl (25 copies/10µl rxn) (Fig. 2D) , with 180 lower concentrations being detected < 50% of the time. 181

To determine if sensitivity using the Norgen RT-qPCR mix could be enhanced, we tested different 182 annealing/elongation temperatures in the qPCR reaction along with two other published SARS-183

CoV-2 primers/probes shown to have high sensitivity (E Sarbeco and HKU Orf1) (9, 12, 21, 22) , 184

and new primers/probes we designed to target the viral N gene. Increasing the annealing/elongation 185 temperature from the manufacturer's recommended 55°C did not affect Ct values for either the N1 186 or N2 primers provided with the Norgen system (Fig. 2E ). Using the Norgen RT-qPCR mix, we 187 observed poor performance of the HKU Orf1 primer set, and the newly designed N gene primers 188 provided higher Ct values compared to the CDC N1 and N2 primers, but the E gene primers/probes 189 demonstrated lower (more sensitive) Ct values compared to the N1/N2 primers, particularly at 59°C 190 annealing/elongation (Fig. 2E ). This improvement, however, did not translate to a lower detection 191 limit (Fig. 2D) . Thus, while both systems easily detect purified SARS-CoV-2 RNA from infected 192 patients, the BGI primers/RT-qPCR system provides a lower detection limit and similar Ct values to 193 clinical data, while Ct values for the Norgen detection module are ~2-3 cycles higher. 194 SYBR green detection. We next compared the more sensitive BGI detection system to a SYBR 195 green-based method. We tested various published primers, some designed for SYBR green and 196 some from TaqMan assays (9, 12, 20, 23) , and designed our own. One published set for the viral S 197 gene (20) and two new N or S gene primer sets gave little/no signal in no-template control (NTC) 198 and generated a linear response across 8 -800,000 viral copies/μl (Fig. S1A ), and were thus selected 199 for future analyses. We then compared SARS-CoV-2 standards using the SYBR green primers and 200 the BGI detection kit and observed comparable Ct values between the two systems across 20 to 201 20,000 genome copies/μl (Fig. 3A) . Identical Ct values were obtained using SARS-CoV-2 RNA 202 from WRCEVA (not shown). The BGI system provided a slightly lower detection limit than the 203 SYBR green systems (compare Figs. 2D and 3D). 204

We next analyzed patient samples comparing the SYBR green primers to previous data obtained 205 with the BGI kit (Fig. 3C) . One of the primers (SH S1) did not perform well on patient samples and 206 was excluded from these experiments. The other SYBR green primers reliably identified all 207 negative and positive patient samples, with SH-N1 primers generating slightly lower Ct values (0.3 208 to 1.1 Ct values, p = 0.02) and S1 primers providing slightly higher Ct values compared to the BGI 209 system (-0.2 to 1.6 Ct values, p < 0.01). Quantification of gene copy numbers generated similar 210 results for SYBR green and BGI, and ranged from 24 copies to >120,000 copies/µl (Fig. 3D ). Non-211 specific melt peaks were occasionally observed in negative and low virus copy positive samples, 212 which could easily be identified and excluded (Fig. S1B ). All patient samples were positive for 213 human RNase P (not shown). 214

One step detection without RNA purification. To reduce the number of steps required for viral 215 detection we tested RT-qPCR direct from patient samples in UTM. For this, we added 2.5 μl of 216 sample directly to the RT-qPCR mix and compared this to an equivalent input of extracted RNA. 217 UTM blocked SYBR-green detection of SARS-CoV-2 RNA standards (data not shown), but both 218 the BGI and Norgen TaqMan detection systems identified positive patient samples (Fig. 4A ). Ct 219 values were lower for BGI vs Norgen, consistent with data with purified RNA (c.f. Figs. 2 & 4A). 220 Furthermore, the Norgen system did not reliably identify some positive samples with lower levels of 221 virus (Fig. 4A ). Relative to extracted RNA, direct RT-qPCR with the BGI detection kit was 2-26 222 fold less sensitive (except sample L021, which was ~600-fold reduced, see below for an 223 explanation), whereas with the Norgen kit it was 20-1000's fold lower (L033 with the N2 primers 224 was an exception at 4.4-fold). Despite the reduced sensitivity, the strong correlation between BGI 225 and clinical Ct values was maintained (Fig. 4B) . 226

Others have reported that reduced sensitivity in direct vs. extracted RNA analyses can be partially 227 overcome by heat or different lysis buffers/detergents (14) (15) (16) 19) . Thus, we assessed the effect of 228 adding an RNase inhibitor (RNaseOUT), heating samples at 95°C for 15 minutes, or five different 229 lysis buffers/detergents (Lucigen QuickExtract DNA extraction solution, MyPOLS Bio VolcanoCell2G lysis buffer, 1% Triton X-100, 1% Tween-20 or 1% Saponin). Simply adding 231

RNase inhibitor was sufficient to dramatically increase detection >100 fold using the Norgen 232 system, bringing Ct values to levels comparable to those obtained with the BGI RT-qPCR system, 233

and, most importantly, allowing for detection of previous "false-negative" samples L021 and L032 234 ( Fig. 4C) . Furthermore, RNase inhibitor brought direct RT-qPCR results with the Norgen detection 235 kit to within 3 Ct values (~10-fold) of those obtained with extracted RNA (compare Figs. 4A and 236 C). Treatment with heat, lysis buffers or detergents did not appreciably increase sensitivity further, 237 and in some cases reduced sensitivity (higher Ct values). For the BGI detection system, none of the 238 treatments dramatically improved detection, with the exception of sample L021 (Fig. 4C ), which 239 previously showed the largest difference between extracted RNA and direct UTM analysis (Fig. 240 4A). We presume, therefore, that L021 had higher RNase levels that were not fully inhibited by the 241 (proprietary) RNase inhibitor already present in the BGI mix. Thus, RNase inhibitor is sufficient to 242 improve direct detection and under these conditions BGI and Norgen kits perform similarly. 243

Here, we comprehensively compared four different RNA isolation methods, two recently released 245 SARS-CoV-2 TaqMan RT-qPCR detection modules and a SYBR green-based RT-qPCR approach 246 for SARS-CoV-2 detection using published and newly-developed primers. Furthermore, we tested 247 and optimized extraction-free SARS-CoV-2 detection using these same detection modules. 248

For RNA extraction, we tested three different column-based systems from Qiagen (RNeasy), 249

Invitrogen (Purelink) and Norgen Biotek, as well as a magnetic silica bead system from BGI. While 250 only the BGI system is specifically marketed for viral RNA isolation, we observed similar results 251 using the Qiagen RNeasy, Norgen and BGI systems, and while it was only tested on two samples, 252 we observed lower recovery of viral RNA using the Invitrogen Purelink system. Cost analysis of the 253 BGI and Norgen Biotek RNA isolation systems revealed that the latter is ~40% more expensive 254 than that of BGI ($6.55 CAD vs. $4.68 CAD/sample, Fig. 4D ), but we found that for small batches 255 of samples the bead-based BGI kit was slower, increasing sample preparation time by about 50% 256 over the Norgen kit (~30 vs. 45 min). This difference was largely due to two incubation steps in the 257 BGI protocol, so the relative difference in sample preparation time may diminish as larger numbers 258 of samples are processed. Furthermore, magnetic beads facilitates large-scale, automated sample 259 extraction. 260

For RNA detection, we tested TaqMan-based detection systems from BGI and Norgen Biotek, as 261 well as a SYBR green method using a commercially available RT-qPCR mix and published primers 262 (some used for SYBR green and others from probe-based methods) along with new primers we 263 developed. All systems could accurately detect SARS-CoV-2 positive patient samples using 264 extracted RNA, and generated Ct values that strongly correlated with clinical diagnostic values. 265

However, the BGI and SYBR green methods routinely produced lower Ct values for patient 266 samples, which closely match clinical results, and had lower detection limits compared to the 267 Norgen system. The BGI system also performed slightly better that the SYBR green methods with 268 low-level standards. One drawback to the SYBR green method was reduced specificity, as we 269 sometimes observed non-specific products in negative or low-level samples, although these could 270 be identified by monitoring melt curves. These non-specific products were not routinely observed in 271 NTC reactions. Thus, melt curve analysis is an essential component of SYBR green qPCR. We also 272 tested 8 other published and newly designed primers and all yielded non-specific PCR products (not 273 shown). Whether non-specific products can be eliminated using alternative RT-qPCR mixes 274 remains to be determined. The BGI detection module is over four-times more expensive than 275 Norgen or SYBR green methods (Fig. 4D) , providing a significant financial drawback. Cost savings 276 with the Norgen kit could be even greater if multiplexing primers/probes were utilized; currently 277 this system follows the CDC guidelines with three separate reactions, one each using FAM-labelled 278 viral N1, viral N2 or human RNase P primers/probes. The Norgen and SYBR green systems also 279 provide more flexibility than that of BGI. Primers/probes come pre-mixed in the BGI system and 280 cannot be altered, whereas they are added separately in the others, allowing alternative primer/probe 281 options and concentrations. We tested three alternative primers/probes with the Norgen system. 282

Those targeting the E gene performed similarly to the provided N1/N2 primers/probes, while 283 alternatives for the viral N or Orf1a gene had reduced sensitivity, although only a single 284

primer/probe concentration was tested. Sequences of the BGI primers/probes are unavailable, and 285 only a single primer/probe set targeting the viral Orf1ab gene is used. Mutation could affect 286 detection and generate false negatives. Thus, while the BGI system provides a lower detection limit 287 with extracted RNA than Norgen or SYBR green detection systems, all accurately identified SARS-288

CoV-2-positive patients, and latter systems detect multiple viral targets and offer greater flexibility 289 and substantially reduced costs. 290

Finally, we tested direct, extraction-free detection of SARS-CoV-2. This approach reduces cost, 291 increases throughput, and circumvents the need for RNA extraction systems that may be scarce 292 during a pandemic. Others have shown that SARS-CoV-2 can be detected from patient samples, 293 although this typically comes with reduced sensitivity, which can at least partially be overcome by 294 heat and/or detergent lysis (14) (15) (16) 19) . We found that SYBR green-based detection was 295 incompatible with direct detection of samples in UTM. The unmodified BGI detection system 296 performed well in the direct detection of unprocessed patient samples, and confirmed all positive 297 samples tested across a wide range of clinical values, but had a reduced median sensitivity of ~12-298 fold compared to extracted RNA. The Norgen system initially performed poorly on direct UTM 299 samples, generating much higher Ct values than extracted RNA (in some cases 1000s of fold 300 higher), and resulted in several false-negatives. Critically, however, adding RNase inhibitor 301 increased sensitivity of direct RT-qPCR with the Norgen system > 100-fold, allowing detection of 302 all previously false-negative samples. This modification did not, in most cases, dramatically 303 increase sensitivity of direct sample analysis with the BGI detection system, suggesting it already 304 contains an RNase inhibitor. Even in that case however, detection of one patient sample was 305 markedly improved, implying higher RNase levels. Thus, addition of RNase inhibitor is a simple 306 and sufficient step to facilitate diagnosis of SARS-CoV-2 direct from patient samples. 307

Our results provide in depth analysis of recently released SARS-CoV-2 detection systems from BGI 308 and Norgen Biotek and compare these to a SYBR green-based approach and to clinical diagnostic 309 values. Each system provides advantages and disadvantages depending on sensitivity, specificity, 310 flexibility and cost. Our findings will help guide selection of SARS-CoV-2 detection systems, and 311 provide an outline for others to compare alternative systems. TaqMan assays. Linear regression was used to determine the R 2 . BGI data is from Fig. 2A RNaseOUT with or without heating at 95°C for 15 min, or treated with the indicated lysis 436 buffers/detergents and then directly analyzed using the BGI or Norgen (N1/N2 primers) RT-qPCR 437 detection systems. Note sample L020 (clinical negative) was also tested under these conditions and 438 was confirmed as SARS-CoV-2 negative. (D) Cost analysis comparing Norgen, BGI, and SYBR 439 green systems. Price is in CAD at the time these studies were initiated (late March/early April 2020) 440 for 10µl RT-qPCR reactions and include relevant processing and shipping fees. * BGI RNA 441 extraction module is based on the 96-sample format, price can be reduced ~15% by purchasing the 442 1728-sample format, and bulk pricing with a ~25% discount of the detection module is available for 443 >10,000 samples. ** Pricing for the Norgen detection module is based on the 50-sample format 444 running three separate wells (N1, N2 and RNaseP) per sample, pricing can be reduced if purchasing 445 the larger 500-sample format. *** Pricing for SYBR green detection is based on the 200 reaction 446 size LUNA Universal One-Step RT-qPCR Kit (NEB) running three separate wells/sample (two 447 viral genes and one human control gene). Pricing can be reduced up to 30% with larger kit sizes. showing both specific and non-specific melt peaks. NTC, no template control (water). Invitrogen Purelink or BGI RNA isolation kits (C) using the Norgen RT-qPCR detection system and N1, N2 or human control (Rnase P) primers. Samples L015, L018 and L019 are the mean +/-range of technical duplicates run independently on two separate plates, other samples were analyzed once, although L028 and L029 are rerun in Fig. 2a. In (B) Comparison of Ct values obtained for each patient sample with the SYBR green and BGI TaqMan assays. Linear regression was used to determine the R 2 . BGI data is from Fig. 2A C.

Primers: S1 SH N1 SH S1

10 copies/μl 10/10 10/10 10/10 5 copies/μl 9/10 9/10 9/10 2.5 copies/μl 7/10 7 (E, RdRp and N genes) to data obtained for direct analysis with the BGI detection system. (C) Patient samples in UTM were left untreated, or treated with the RNase inhibitor RNaseOUT with or without heating at 95°C for 15 min, or treated with the indicated lysis buffers/detergents and then directly analyzed using the BGI or Norgen (N1/N2 primers) RT-qPCR detection systems. Note sample L020 (clinical negative) was also tested under these conditions and was confirmed as SARS-CoV-2 negative. (D) Cost analysis comparing Norgen, BGI, and SYBR green systems. Price is in CAD at the time these studies were initiated (late March/early April 2020) for 10µl RT-qPCR reactions and include relevant processing and shipping fees. * BGI RNA extraction module is based on the 96-sample format, price can be reduced ~15% by purchasing the 1728-sample format, and bulk pricing with a ~25% discount of the detection module is available for >10,000 samples. ** Pricing for the Norgen detection module is based on the 50-sample format running three separate wells (N1, N2 and RNaseP) per sample, pricing can be reduced if purchasing the larger 500-sample format. *** Pricing for SYBR green detection is based on the 200 reaction size LUNA Universal One-Step RT-qPCR Kit (NEB) running three separate wells/sample (two viral genes and one human control gene). Pricing can be reduced up to 30% with larger kit sizes. N/A: not applicable. Ct Copies/μL SH S1

A.

L024 (specific peak) L017 L024 NTC L017 (non-specific) S1 Primers (Lancet) L024 (specific peak) SH N1 Primers LO17 (non-specific) L017

L024 NTC L032 L024 L032 (non-specific peak) S1 Primers (Lancet) L024 (specific peak) L032 (specific peak)

A new coronavirus associated with human respiratory disease in China

Genomic 321 characterization of the 2019 novel human-pathogenic coronavirus isolated from a patient with 322 atypical pneumonia after visiting Wuhan

Genome Composition and Divergence of the 325

Novel Coronavirus (2019-nCoV) Originating in China

WHO Director-General's opening remarks at the media briefing on COVID-19 -11 March 328 2020. 329 6. COVID-19 Map -Johns Hopkins Coronavirus Resource Center

Covid-19: identifying and isolating asymptomatic people helped eliminate virus 331 in Italian village

The Incubation Period of Coronavirus Disease

Publicly Reported Confirmed Cases: Estimation and Application