Flexible upscaling of laboratory PCR testing capacity at the Robert Koch Institute during the SARS-CoV-2 pandemic

🔗 Access full article via BMC Virology Journal

Comparison of different automated RNA extraction methods

To evaluate the impact of the RNA extraction method on the sensitivity of SARS-CoV-2 detection, we compared five automated extraction methods by using serial dilutions of inactivated cell culture supernatant with SARS-CoV-2 concentrations around the limit of detection. PCR analysis was performed in duplicate by using the E-Gene/KoMa and orf1ab/c-myc duplex PCR assays, which resulted in a total of four PCR replicates for the detection of SARS-CoV-2 RNA in each sample. Table 5 shows the number of positive PCR reactions for each dilution when using the different extraction methods. Detection of SARS-CoV-2 from cell culture supernatant was comparable for all tested extraction automats. Results were confirmed by using two additional, independently generated series of dilutions (data not shown).

Figure 1a, b show the Ct values obtained for each sample in the E-Gene and orf1ab PCR after extraction by using the different instruments. Extraction using the QIAcube Connect was selected as reference method to quantify the differences in Ct values across instruments. For each instrument and sample, the difference of Ct values to the mean Ct value obtained with the QIAcube Connect was calculated (Fig. 1c, d). All automated extraction methods yielded similar Ct values in the SARS-CoV-2 rRT-PCRs. Using the King Fisher Flex and Magna Pure 96 resulted in slightly lower Ct values compared to the QIAcube Connect, which is probably due to the higher concentration factor of RNA extract in relation to the original sample (4 × concentrated when using the King Fisher Flex and Magna Pure 96; 2.3 × and 1.75 × concentrated when using the Qiagen instruments; Table 1). When the Ct values were adjusted for the different ratios of sample volume to RNA volume, the Ct difference of the King Fisher Flex and Magna Pure 96 compared to the QIAcube Connect was largely eliminated, illustrating the equal performance quality of the extraction automats (Fig. 1e, f). Using the QIAsymphony instrument resulted in slightly higher Ct values compared to all other tested methods, a fact which we also observed when applying different non-customized off-board and on-board lysis protocols provided by the manufacturer (data not shown).

The QIAcube Connect, QIAcube HT and QIAsymphony instruments, which at the time could be used for processing infectious samples, were additionally compared by using a panel of 20 SARS-CoV-2-positive patient samples, selected to cover a wide range of viral loads including virus concentrations around the limit of detection. All three compared extraction automats showed a similar performance on the tested patient samples, both with regard to the number of correctly identified samples and the obtained Ct values (Table 6).

Increasing PCR analysis capacity through a higher level of multiplexing

Our SARS-CoV-2 PCR assay was initially developed as a combination of two duplex PCR reactions detecting E-Gene/KoMa and orf1ab/c-myc [7]. In order to be able to increase PCR throughput if needed, we combined the detection of E-Gene, KoMa, orf1ab and c-myc in a quadruplex reaction. As the fluorescence intensity of the orf1ab PCR suffered considerably from the higher degree of multiplexing, we added a second probe to the orf1ab reaction. Notably, this reduces the probability of detection failure due to virus mutations, for which the SARS-CoV-2-specific orf1ab PCR is at a higher risk than the more broadly reactive Sarbeco E-Gene PCR. For further compensation of the expected sensitivity loss in the quadruplex assay compared to the duplex assays, we increased the volume of RNA extract added to the PCR reaction from 5 µl to 10 µl.

The quadruplex assay showed no reactivity for patient samples that had been tested positive for influenza A virus (n = 22), influenza B virus (n = 2), parainfluenza virus 2 (n = 3), 3 (n = 4) and 4 (n = 1), respiratory syncytial virus (n = 24), human metapneumovirus (n = 20), rhinovirus (n = 98), adenovirus (n = 23), bocavirus (n = 8), NL63 (n = 29), 229E (n = 8), OC43 (n = 7) and Bordetella pertussis (n = 3). In addition, we detected no reactivity for MERS-CoV cell culture supernatant (n = 1) and only an E-Gene signal (as expected), but no orf1ab signal for SARS-CoV cell culture supernatant (n = 1). Probit analysis for the quadruplex PCR under the described reaction conditions (AgPath-ID PCR kit, BioRad CFX96) revealed a limit of detection of 28.7 genome copies per reaction for the E-Gene assay and 32.0 genome copies per reaction for the orf1ab assay. The limits of detection of the quadruplex PCR assay are therefore about three and five times higher than the previously determined limits of detection for the duplex PCR assays, which are 9.8 and 6.6 genome copies per reaction for the E-Gene/KoMa PCR and the orf1ab/c-myc PCR, respectively [7].

To evaluate the impact of the slightly reduced sensitivity of the quadruplex PCR on the ability to correctly identify positive patient samples, we compared the duplex and quadruplex PCRs by testing 24 patient samples that were previously identified to be SARS-CoV-2 positive with E-Gene Ct values higher than 30. Fresh RNA extracts were prepared from the original samples by using the QIAcube Connect and tested in parallel with the SARS-CoV-2 duplex PCRs and the quadruplex PCR. Both assays returned similar results for the tested samples (Table 7).

Sample pooling

In order to prepare for peaks of exceptionally high testing demand, we established a two-level pooling protocol with either two or four samples per pool, which can be flexibly utilized depending on the current needs.

To minimize sensitivity loss caused by the dilution effect of the pooling, we developed a modified version of our SARS-CoV-2 quadruplex PCR assay, in which both E-Gene and orf1ab are detected by using the same fluorescence dye (FAM), which results in a higher fluorescence intensity. We refer to this assay as allFAM quadruplex PCR. While in the regular quadruplex assay an additional probe was included only for the orf1ab reaction, the allFAM quadruplex assay also includes an additional probe for the E-Gene reaction to further enhance fluorescence intensity and thus detection sensitivity. The allFAM quadruplex assay does not allow the distinction between E-Gene and orf1ab signals. However, each sample of a positive pool is subsequently analyzed individually by using the standard PCR assays which detect E-Gene and orf1ab separately. While successful RNA extraction and PCR inhibition can still be controlled for in the sample pools by using an exogenous internal control, confirming adequate sampling for each individual sample via an endogenous internal control is not possible when samples are pooled. Nevertheless, we decided to include the c-myc PCR in the assay for pooled samples in order to obtain at least some information on the human nucleic acid content in the pools.

We tested the performance of our sample pooling approach by utilizing the same 24 weakly positive patient samples that were used to compare the SARS-CoV-2 duplex PCRs with the quadruplex PCR (Table 7). In parallel to the individual testing, all 24 positive patient samples were analyzed in a pool with either one or three negative patient samples. RNA extractions were performed by using the QIAcube Connect, and all RNA extracts were subjected to the duplex PCRs, the standard quadruplex PCR and the allFAM quadruplex PCR. Compared to individual testing, both the duplex and quadruplex assays failed to detect three samples in the two-sample pools and six samples in the four-sample pools (Table 8). In contrast, the allFAM quadruplex assay did not miss any samples in the pooled testing compared to individual testing and even detected two additional samples in the two-sample pools.

Figure 2 shows the Ct values for each sample, obtained by individual testing, two-sample pooling and four-sample pooling using either the duplex assay (Fig. 2a, b), quadruplex assay (Fig. 2c, d) or allFAM quadruplex assay (Fig. 2e). Samples that were detected only by individual testing and not in the pools had Ct values of 34 or higher. While the average Ct values obtained from the pools were, as expected, about 1–2 cycles higher than those measured in the individual samples (Table 8), Fig. 2 shows that the difference in Ct values between individual testing and pooled testing varied considerably from sample to sample. In some cases, Ct values were even lower in the pool than in the individual sample.

Continue reading on BioMed Central