A high-throughput pipeline for scalable kit-free RNA extraction

The TRIzol method, while widely proven effective in RNA extraction, poses challenges to adaptation for high-throughput processing performed on automated liquid handling instruments due to the multiple steps involved. The need for phase separation, precipitation by centrifugation and potential phenol carryover further complicates the workflow. In this study, the conventional protocol was adapted to a 96-well format with emphasis on miniaturization and reduced human intervention (Fig. 1). To validate that the semi-automated, non-kit extraction method was able to isolate RNA at sufficient yield and quality for sensitive downstream applications, biological samples were spiked with non-infectious, synthetic SARS-CoV-2 virus-like-particles (VLPs) and heat-inactivated 2019 novel coronavirus and the extracted RNA evaluated for SARS-CoV-2 detection via real-time reverse transcription qPCR (RT-qPCR). While prevailing clinical diagnoses of SARS-CoV-2 are mostly conducted on nasopharyngeal or oropharyngeal specimens, several studies have shown that saliva presents a low-cost, non-invasive and reliable alternative for SARS-CoV-2 detection^5,6,7. As such, saliva and pharyngeal swabs were used as diagnostic samples in this study due to the ease of self-collection from volunteers, negating the need for skilled personnel for sample collection. Importantly, saliva can be highly viscous and difficult to pipet and was presented as an assessment on the automated liquid handlers’ capability to exact extraction from specimens with varying viscosity. For comparison, analyses were also carried out on oropharyngeal nasal mid-turbinate (OP-NMT) swabs performed by a trained medical personnel. Two laboratory automation platforms were employed, namely, the Opentrons OT-2 lab robot and the Eppendorf epMotion® 5075, both which were equipped with an 8-channel volumetric dispensing arm and a thermal block for 96-well plates.

Scalable high-throughput kit-free RNA extraction workflow. Biological samples are manually transferred into a 96-well PCR plate and total RNA extraction was carried out by automated liquid handlers using a modified TRIzol-based method. Manual intervention is required for plate centrifugation. Viral diagnostic analyses is performed on extracted RNA via qPCR-based detection. A complete run of 96 samples takes 4 h. The cost (USD) per sample was calculated based on actual local prices and includes laboratory consumables such as tips and reagents but excludes the cost of instruments.

In order to perform RNA isolation in 96-well format, the reaction volume was scaled down proportionally in a 3:1 TRIzol reagent-to-sample ratio, reducing the volume of reagent required by sevenfold to 105-μL, thereby directly lowering the consumable cost per sample extraction. With a concomitant decrease in input sample to a volume of just 45-μL, glycogen was added as a co-precipitant to aid in RNA recovery. In a kit-free extraction method, precipitation and purification of RNA by centrifugation remain ineludible steps, making manual intervention obligatory in the workflow. The addition of glycogen as an inert carrier for nucleic acids was crucial in overcoming the challenge of low centrifugal force when using swing-out rotors for plates. The adapted procedure successfully extracted 1 to 4 ng/μl of RNA from both saliva and throat swabs with the yield slightly higher from saliva (Fig. 2A).

Efficacy of TRIzol RNA isolation from saliva and throat swabs using automated liquid handlers. (A) Concentration of RNA of saliva (blue circle) and throat swabs (red triangle) determined using Quant-IT. (B) Threshold cycles (C_T) of one-step RT-qPCR analysis on total RNA extracted from saliva (blue cirlce) and throat swabs (red triangle) using Opentron-2 (OT2) and Eppendorf epMotion (epM). Analysis targeting human Ribonuclease P (RNase P) was performed on five technical replicates. (C) Detection of RNase P using multiple qPCR master mixes. Analysis was performed with Luna (NEB; blue diamond), iTaq (Bio-Rad; red square) and KAPA (Kapa Biosystems; green inverted triangle) one-step SYBR RT-qPCR kits. Total RNA was extracted from throat swabs using Eppendorf epMotion. Statistical differences were analysed using unpaired, two-tailed t-test where ns indicates non-significance (p > 0.05).

During transfer of the aqueous phase following phase separation, cross-contamination with proteins, lipids, DNA and phenol can occur. Indeed, ensuring high quality of isolated RNA, particularly in a miniaturized format, proved challenging. Phenol carryover was identified in sample extractions using both automated liquid handlers as indicated by a characteristic absorbance maximum at 270 nm. Despite the relatively low purity and quantity, the extracted RNA was tested for its suitability for use in sensitive downstream applications such as real time RT-qPCR. The phenol contamination did not appear to interfere with the efficacy of the RT-qPCR targeted at the ubiquitous human Ribonuclease P (RNase P) from both the saliva and throat (Fig. 2B). To determine the reproducibility in analytical efficiency, the validation was repeated using three commercially available one-step RT-qPCR master mixes (Fig. 2C). The Ct values showed slight variability for the KAPA qPCR mix compared to the Luna and iTaq mix but the results, nonetheless, suggested that the extracted RNA was of sufficient yield and quality for downstream diagnostic application irrespective of the molecular reagent kits.

To assess viral RNA extraction efficiency of the miniaturised, semi-automated TRIzol-based extraction protocol against commercial RNA purification kits, biological samples were spiked with heat-inactivated 2019 novel coronavirus (nCoV) and RNA extracted using the QIAamp Viral RNA Mini Kit specific for viral RNA isolation from body fluids, the RNeasy Mini Kit for generic purification of total RNA from cells as well as the TRIzol-based method performed manually at standard volume. To ensure consistent sample input, 45 ul of biological samples were each spiked with a fixed quantity of 4,500 copies of nCoV, and topped up with DMEM to the varying starting sample volumes recommended for each kit. Kit-extracted RNA was eluted according to manufacturer’s instructions, followed by volume reduction by evaporation to 10 ul to standardize with that used in the miniaturised TRIzol-based method before performing RT-qPCR assay targeted at RNaseP and SARS-CoV-2 N gene. The CDC 2019-nCoV N1 primer set was used as it has been shown to be more sensitive than the N2 or N3 sets, typically generating lower Ct values from positive samples^8,9. The standard TRIzol-based method gave RNA recovery comparable to the QIAamp Viral RNA kit and RNeasy Mini kit with Ct values of 24.3, 24.1, 24.7 for RNaseP and 27.7, 28.1, 28.1 for nCoV-N1 detection, respectively (Fig. 3A, B). The results were similar to that reported by Won et. al., 2020, indicating that TRIzol was equally effective in extracting RNA compared to commercial available kits¹⁰. While the miniaturised, 96-well-adapted TRIzol-based extraction yielded adequate RNA to amplify the target genes, the method showed a lower sensitivity as observed by the higher Ct values for RNaseP (Ct = 26.3) and nCoV-N1 (Ct = 30.6).

Efficiency of TRIzol RNA extraction relative to commercial kits. RNA was isolated from swab samples spiked with heat inactivated 2019 novel coronavirus (nCoV) using the Qiagen QIAamp Viral RNA kit, Qiagen RNeasy Mini Kit, standard TRIzol-based method performed manually, or semi-automated, 96-well-adapted TRIzol-based protocol by Eppendorf epMotion. SYBR-based amplification curves for (A) human Ribonuclease P, and (B) SARS-CoV-2 N1 gene.

The limit of detection of the semi-automated workflow was also examined with tenfold serial dilutions (10¹–10⁵ copies per reaction) of SARS-CoV-2 VLPs and nCoV. The highest quantified concentration in the standard curve for VLPs was limited to 10⁴ copies per reaction as the available stock concentration was 1700 copies per μL. The extracted RNA showed a detection limit of 10 copies of VLP per reaction, while that of nCoV was 100 copies per reaction, suggesting good efficiency of extraction using the high-throughput platform (Fig. 4).

Efficiency of TRIzol RNA extraction from serial-diluted SARS-CoV-2 spiked samples. SYBR one-step RT-qPCR analysis targeting SARS-CoV-2 N gene was performed on DMEM spiked with tenfold serial dilutions of (A) SARS-CoV-2 virus-like-particles (VLPs; blue diamond), and (B) heat inactivated 2019 novel coronavirus (nCoV; red square). Total RNA was extracted from each concentration using Eppendorf epMotion. Results are shown as mean ± SE of three technical replicates.

Upon demonstrating the efficiency of the miniaturised, TRIzol-based RNA extraction using the semi-automated workflow, the platform was further validated on mock patient samples loaded with SARS-CoV-2 RNA. Saliva and throat swab samples were collected from 24 volunteers and SARS-CoV-2 VLPs and nCoV were separately added into 8 samples each at 1 × 10⁵ copies/ml to simulate clinically relevant concentrations¹¹. In the SYBR RT-qPCR assay, up to 3 false positive signals with significant Ct values (Ct < 37) was detected out of the expected 16 positive spiked samples (Fig. 5). Examination of the melt profiles revealed nonconcordant melt peaks, indicating the occurrence of amplification artefacts in both saliva and throat negative control samples. Amplification of nonspecific products in quantitative PCR using SYBR Green-based detection is not uncommon and presents compromises in specificity and reproducibility¹². Negative detection was observed for the internal control primer set targeted at RNaseP, especially in throat swab samples, in which up to 50% of the 24 samples did not yield signals. This could largely be due to the variability in skill of swab sampling since this was self-administered by untrained volunteers. Non-detection of internal control was noticeably reduced in saliva samples, occurring in 1 to 3 out of 24 samples. These results lend support to the use of saliva as a reliable alternative for detection of SARS-CoV-2 as it alleviates the need for trained personnel for specimen collection and reduces the reliance on swabs and viral transport media.

SARS-CoV-2 detection on hypothetical patient samples. Saliva and throat swab samples were collected from 24 volunteers and spiked 8 each with 1 × 10⁵ copies/mL heat inactivated 2019 novel coronavirus (nCoV; blue diamond), VLPs (red diamond), or water (green diamond). Total RNA was extracted using (A) Opentron-2 and (B) Eppendorf epMotion. Biological samples were analysed by SYBR one-step RT-qPCR targeting SARS-CoV-2 N gene. Artefact amplification was observed in water-spiked saliva and throat swab samples (green diamond).

To circumvent the challenge of false positives, which puts the diagnostic reliability into question, the analysis was repeated using a multiplex, probe-based RT-qPCR kit deploying three primer–probe sets targeting the ORF1ab region 1 and 2 of the SARS-CoV-2 genome as well as an internal control (Fig. 6). Expectedly, the probe-based assay displayed higher specificity compared to the SYBR Green-based method, producing no false positives in any of the SARS-CoV-2-negative samples. Indeed, all biological samples spiked with VLPs did not generate positive signals as these recombinant VLPs were engineered to carry the SARS-CoV-2 N gene and lacked the ORF1ab region targeted by the RT-qPCR assay used.

SARS-CoV-2 detection on hypothetical patient samples. Saliva and throat swab samples were collected from 24 volunteers and spiked 8 each with 1 × 10⁵ copies/ml heat inactivated 2019 novel coronavirus (nCoV; blue diamond), VLPs (red diamond), or water (green diamond). Total RNA was extracted using (A) Opentron-2 and (B) Eppendorf epMotion. Biological samples were analysed by multiplex, probe-based RT-qPCR targeting SARS-CoV-2 ORF1a gene.

In a bid to verify if the partial detection of internal control was indeed the result of inconsistencies in self-sampling by untrained volunteers, sampling was repeated by a medically trained personnel familiar with oropharyngeal nasal mid-turbinate swabbing performed in routine Covid-19 testing. This time, detection for the RNaseP internal control in the SYBR RT-qPCR assay was positive for all 24 volunteers (Fig. 7A). The result suggests that the RNA extraction platform is reliable in a realistic, diagnostic setting. Instead of VLPs, 8 biological samples were spiked with 1 × 10⁵ copies/ml nCoV and another 8 with a tenfold lower concentration of 1 × 10⁴ copies/ml nCoV. Occurrence of amplification artefacts recurred as observed in 4 random, false positive signals (Ct ≤ 37) in the negative, water-spiked controls, underlining the shortcoming in specificity of SYBR Green-based detection, particularly for diagnostic analyses (Fig. 7A). Analyses by both SYBR Green- and probe-based RT-qPCR gave positive detection in all 16 SARS-CoV-2-spiked samples, with Ct values higher for those spiked with 1 × 10⁴ copies/ml nCoV (Fig. 7A, B).

SARS-CoV-2 detection on hypothetical patient samples. Oropharyngeal nasal mid-turbinate samples were collected from 24 volunteers and spiked 8 each with 1 × 10⁴ copies/ml (blue square, blue circle), 1 × 10⁵ copies/ml (red square, red circle), heat inactivated 2019 novel coronavirus (nCoV), or water (green square, green circle). Total RNA was extracted using Opentron-2 (OT-2; squares) and Eppendorf epMotion (epMotion; circles). (A) Biological samples were analysed by SYBR one-step RT-qPCR targeting SARS-CoV-2 N gene. Artefact amplification was observed in water-spiked samples (green square, green circle). (B) Biological samples were analysed by multiplex, probe-based RT-qPCR targeting SARS-CoV-2 ORF1a gene.

As exemplified by the recent, unanticipated surge in Covid-19 testing across the world, supply chain issues in commercial RNA extraction kits can present a rate-limiting obstacle to swift and extensive clinical diagnosis. In a bid to ease the reliance on proprietary, high-throughput RNA extraction kits and their dedicated instruments, a semi-automated, high-throughput kit-free workflow was designed with miniaturization and minimal manual operation in mind (Fig. 1). This approach is notably more cost-effective as kit-based RNA isolation is more expensive than TRIzol-based RNA extraction (Table 1). By further lowering the sample volume from hundreds down to less than 50 µl and by processing in the 96-well format, the semi-automated workflow significantly reduces processing time, but, more importantly, lowers the operational cost which can translate to substantial savings for healthcare test facilities. Based on actual prices for all consumables used, including general laboratory plastic wares such as collection tubes, pipette tips and 96-well plates, as well as chemicals and reagents purchased from local vendors, the cost per sample was as low as less than USD 5 to up to USD 15 without taking into consideration the capital cost for instruments. The price disparity is primarily attributable to the broad price range between the affordable universal SYBR-Green dye-based kits and the more specific but costlier probe-based assays. Even in the absence of automated liquid handlers, the 96-well adapted protocol can still be performed manually using an 8-channel pipette at comparable speed, albeit more labour-intensive. It can be argued that emerging diagnostic strategies involving extraction-free direct RT-qPCR detection offers an even more attractive option that will not only drive the cost lower but also greatly reduce time required for sample processing^13,14,15. While holding great promise for SARS-CoV-2 detection, its prospective application for medical surveillance of future novel infectious diseases will require re-validation, thereby necessitating the reliance on the assured reliability of conventional, prior RNA isolation, at least in the early stage of routine clinical testing.