Simultaneous monitoring of HIV viral load and screening of SARS- CoV-2 employing a low-cost RT-qPCR test workflow

Abstract

The COVID-19 pandemic interrupted routine care for individuals living with HIV, putting them at risk of virologic failure and HIV-associated illness. Often this population is at high risk for exposure to SARS-CoV-2 infection, and once infected, for severe disease. Therefore, close monitoring of HIV plasma viral load (VL) and screening for SARS-COV-2 infection are needed. We developed a non-proprietary method to isolate RNA from plasma, nasal secretions (NS), or both. The extracted RNA is then submitted to RT-qPCR to estimate the VL and classify HIV/SARS-CoV-2 status (i.e., HIV virologic failure or suppressed; SARS-CoV-2 as positive, presumptive positive, negative, or indeterminate). In contrived samples, the in-house RNA extraction workflow achieved a detection limit of 200-copies/mL for HIV RNA in plasma and 100-copies/mL for SARS-CoV-2 RNA in NS. Similar detection limits were observed for HIV and SARS-COV-2 in pooled plasma/NS contrived samples. When comparing in-house with standard extraction methods, we found high agreement (R² > 0.91) between input and measured RNA copies for HIV LTR in contrived plasma; SARS-COV-2 N1/N2 in contrived NS; and LTR, N1, and N2 in pooled plasma/NS samples. We further evaluated this workflow on 133 clinical specimens: 40 plasma specimens (30 HIV-positive), 67 NS specimens (31 SARS-CoV-2-positive), and 26 combined plasma/NS specimens (26 HIV-positive with 10 SARS-CoV-2-positive), and compared the results obtained using the in-house RNA extraction to those using a commercial kit (standard extraction method). The in-house extraction and standard extraction of clinical specimens were positively correlated: plasma HIV VL (R² of 0.81) and NS SARS-COV-2 VL (R² of 0.95 and 0.99 for N1 and N2 genes, respectively); and pooled plasma/NS HIV VL (R² of 0.71) and SARS-CoV-2 VL (R² of 1 both for N1 and N2 genes). Our low-cost molecular test workflow ($1.85/pooled sample extraction) for HIV RNA and SARS-CoV-2 RNA could serve as an alternative to current standard assays ($12/pooled sample extraction) for laboratories in low-resource settings.

Graphical Abstract

This new workflow enables co-extraction of HIV and SARS-CoV2 RNAs from clinical pooled plasma/nasal secretion samples that allows sensitive detection of SARS-CoV-2 and HIV infections in the patients-living with HIV. Combining the detection of these two viruses can help address the challenges associated with the overlay of the COVID-19 pandemic on the existing HIV epidemic.

Introduction

The COVID-19 pandemic has stalled public health responses to address the pre-existing global HIV epidemic.¹ COVID-19 “lockdowns” caused limited transportation, reduced access to in-person doctor visits,² and impacted the global economy, thus interfering with the delivery of effective HIV diagnostics and treatments.³ People living with HIV (PLH) are impacted by the SARS-CoV-2/HIV syndemic in multiple ways.⁴ Reportedly, PLH have fears for SARS-CoV-2 exposure during routine clinical care,^{5, 6} which can impact HIV treatment and result in unsuppressed HIV replication. PLH would have increased morbidity and mortality if infected with SARS-CoV-2;^7–12 especially those with detectable viremia and/or low CD4 counts.¹³ Furthermore, a recent study reports that symptoms of acute HIV infection can be similar to SARS-CoV-2 symptoms;¹⁴ thus clinicians may not test for and miss new HIV infections. Prolonged SARS-CoV-2 infections can occur in immunocompromised individuals, including PLH, in whom multiple variants can evolve.¹⁵ Consequently, early diagnosis of SARS-CoV-2 in PLH, to allow treatment and isolation, could benefit both the individual and public health.

“A call to action,”¹⁶ encouraging HIV researchers to innovate approaches to diminish the impact of COVID-19 on HIV treatment, led us to streamline a workflow for screening SARS-CoV-2 infection while monitoring HIV RNA viral load (HIV VL). Standard laboratory-based molecular methods previously used to detect HIV RNA have been applied to detect SARS-CoV-2 RNA. These molecular tests traditionally involve “extraction” to concentrate and purify the RNA from inhibitors present in biological samples, followed by reverse-transcription quantitative polymerase chain reaction (RT-qPCR) to quantify viral loads. Other than the nasal/oral specimens, SARS-CoV-2 RNA is detectable in plasma for patients hospitalized with COVID-19,¹⁷ and higher viremia is associated with worse clinical outcomes.^{18, 19} Therefore, quantification of SARS-CoV-2 RNA in plasma may be a useful indicator for disease severity. Depending on the sensitivity of the RT-qPCR, plasma SARS-CoV-2 RNA was detected in 19–74% of patients diagnosed via respiratory samples.¹⁷ SARS-CoV-2 viral loads in nasal/oral samples can remain above 10⁴ copies/mL across six weeks after initial infection, whereas SARS-CoV-2 viremia were approximately 500 copies/mL in the first two weeks and completely cleared after the third week of the initial infection.¹⁸ Therefore, this work presents RT-qPCR workflows using an in-house, inexpensive method to extract RNA from plasma, nasal secretion samples, or their combination. Testing SARS-CoV-2 RNA from a plasma sample alone may be beneficial for accessing the severity of the COVID-19 disease in an individual. Here, we investigated the analytical performance of an in-house workflow by testing the contrived samples made up of spiked-in synthetic HIV-RNA in negative plasma, SARS-COV-2-RNA in VTM or HIV and SARS-COV-2 RNAs in pooled NS/plasma matrix to establish the limit of detections (LODs). The major concern of plasma and NS sample pooling is loss of test sensitivity due to specimen dilution. Modified sample input volume for extraction with optimized lysis and purification buffer components of the kit helped maximize purification recovery of the spiked RNA copies. The RT-qPCR assay was optimized for primer/probe concentrations, extracted sample volume/reaction and the number of PCR cycles to avoid PCR reaction inhibition, to avoid competition between the PCR targets for shared reagents in multiplex reaction and to achieve the lowest possible LODs. We tested our in-house workflow with clinical specimens previously characterized in a CLIA certified lab and compared the result with standard Qiagen extraction. The low-cost RNA extraction method for multiple viral RNA targets simultaneously leads to an inexpensive workflow, which could circumvent the supply chain issues in COVID-19 testing, utilizes readily available materials, is easily implemented by test sites, and reduces time and labor required for separate sample testing, thus facilitating HIV testing in COVID-19 patients.

Experimental

Generation and characterization of in vitro transcribed RNA standards

DNA fragments (GenBlock, Integrated DNA Technology, Coralville, IA) containing the amplification regions for HIV LTR assay (base position: 544–621 HXB2; GenBank: K03455.1) or N1/N2 assay (base position: 28,307–28,334 and 29,184–29,212 SARS-CoV-2 respectively; GenBank: NC_045512.2) with upstream T7 RNA promoter sequence that was used as a starting DNA construct for RNA generation. 100ng of the DNA construct was converted to RNA using T7 RNA polymerase (M0251; New England Biolabs (NEB), Ipswich, MA) and after digestion with DNAse I (M0303, NEB) was purified (T2040, NEB) according to manufacturer’s protocols.^{20, 21} Product fragment lengths were confirmed using TapeStation Agilent 2200 (5067–5578), Agilent, Santa Clara, CA). Finally, Crystal Digital PCR (cdPCR) was used to quantify the concentrations of the in vitro transcribed RNA standard (Supplementary Figure 1).

Development of In-house RNA extraction kit

Lysis buffer was freshly prepared by mixing the buffer (4M guanidine thiocyanate, 10mM MES pH 5.5) with 6% (v/v) 1mg/mL tRNA (AM7119, ThermoFisher, Waltham, MA) and 2% (v/v) 2M DTT (P1771, Promega, Madison, WI). VTM (140μL), plasma (140μL), or pooled VTM (140μL)/plasma (140μL) was mixed with 4x volume of lysis buffer and incubated at room temperature for 15 minutes. The ethanol, 4x to sample volume, was added to each sample. The 630μL mixture was then added to the silica column (Epoch Life Sciences, 1920–250) and centrifuged at 6000g for 1 minute. Flow-through fluid was decanted, and the remaining mixture underwent a repeated centrifugation until all the lysed sample was captured by silica. After the last spin of the lysed sample, the silica column was placed in a new collection tube, and 800μL wash buffer 1 (1M guanidinium thiocyanate, 10mM Tris-HCl pH 7.4) was added. The column was centrifuged at 6000g for 1 minute. Flow-through fluid was discarded, and the column was placed in a new collection tube. 800μL wash buffer 2 (80% ethanol; 20% 10mM Tris-HCl pH 7.4) was added to the column, and was centrifuged at 20,000g for 3 minutes. Flow-through fluid was discarded, and the column was centrifuged at 21,000g for additional 1 minutes to completely remove any residual wash buffer. The spin column was then transferred to a clean 1.5mL Eppendorf tube (Catalog No. 022363204, Eppendorf) and 0.04% (w/v) sodium azide in water was added to elute the RNA (60μL for all samples, except 80μL for co-extracted NS/plasma specimens). After 1-minute incubation at room temperature and 1-minute 6000g centrifugation, the resultant RNA was collected and stored at −80°C until tested by RT-qPCR. All samples underwent only a single freeze-thawing cycle to avoid significant RNA degradation.

Commercial RNA extraction method (referred as standard method)

The QIAamp viral RNA kit (52904, Qiagen, Hilden, Germany) was used as a standard method to compare the performance of the in-house RNA extraction method using the same input volume of specimens as described above. All samples were processed according to the manufacturer’s protocol.²²

RT-qPCR assays for HIV viral load and SARS-CoV-2 detection

We adopted clinically-validated primer and probe sets for detection of HIV²³ and SARS-CoV-2.²⁴ The target amplicon of HIV was located in its LTR (long terminal repeat) region. LTR assay was selected as its probe and primers have melting temperatures similar to those used in the US CDC SARS-CoV-2 N1, N2 assays. At the same temperature cycling profile, we reliably amplified the in vitro transcribed HIV and SARS-CoV-2 RNAs down to 10 copies/reaction in a 40-cycle PCR run (Supplementary Figure 2).

Four separate RT-qPCR reactions targeting the HIV long terminal repeat (LTR), SARS-CoV-2 nucleocapsid gene regions N1 and N2, and human ribonuclease P (RP) were set up for each sample. Each 20μL RT-qPCR reaction contained 1x TaqPath master mix (A15299, ThermoFisher, Waltham, MA); HIV, SARS-CoV-2, or human RP primers/probes (Supplementary table 1); and 5μL extracted RNA or RNA standards at 0 – 10⁷ copies/reaction. The reaction mixtures were run on a real-time thermal cycler (CFX96, Biorad) using 2 minutes at 25°C for primer annealing and UNG digestion; 15 minutes at 50°C for reverse transcription; 2 minutes at 95°C for reverse transcriptase enzyme deactivation and initial denaturing; and 50 cycles of 3 seconds at 95°C and 30 seconds at 55°C. At the end of 30 seconds at 55°C, the fluorescence was read using the FAM channel. Sample copies per reaction were calculated using the RNA standard curve included in each run and used to calculate RNA copies/mL of the original specimen. The number of RT-qPCR replicates are provided in the caption of each experiment. RT-qPCR data were analyzed using CFX Maestro (Biorad). Raw fluorescence data were subtracted with the average baseline fluorescence from the cycles 5–10; and samples were determined positive when the signals after baseline subtraction were above 200 RFU. The Cq values of samples were interpolated based on the linear fit to the Cq values and log₁₀ of input concentration of the RNA standards from 5–10⁷ copies/reaction.

Contrived specimens for assay optimization

For HIV testing, contrived samples were prepared by adding in vitro transcribed HIV RNA (for positive samples) or water (for negative samples) to pooled plasma specimens from healthy donors (IPLAWBK2E100ML, Innovative Research, Novi, MI) incubated with lysis buffer for 10–15 minutes. For COVID-19 testing, contrived samples were prepared by using negative VTM (220220, Becton Dickenson, Franklin Lakes, NJ) incubated with lysis buffer for 10–15 minutes, followed by addition of 12ng of human genomic DNA (3041, Promega, Madison, WI, 1ng/μl) and then spiked with in vitro transcribed SARS-CoV-2 RNA (for positive samples) or water (for negative samples).

Panel of clinical specimens for assay validation

We used a total of 113 specimens to evaluate the proposed workflows (defined as OPTION 1 and OPTION 2, Fig. 1a). The remnant nasal secretion (NS) and plasma specimens went through several freeze-thawing cycles, and therefore were suboptimal to directly compare the original results by the certified laboratories to results obtained by the in-house workflow. In this study, we extracted the specimens using the commercial RNA extraction kit (referred to as standard extraction kit) and in-house RNA extraction kit. The extracted RNA was subsequently analyzed using the same RT-qPCR assay and results were compared. Discordant results were ruled out using the original results previously reported by the certified laboratories.

Fig. 1. Workflows for SARS-CoV-2 screening and HIV viral load testing.

(a) Schema of the workflows for analysis of only plasma specimens or the mixtures of contrived plasma and NS samples. Specimens are extracted using our low-cost, in-house extraction method and analyzed using RT-qPCR assays which target HIV LTR gene, SARSCoV-2 N gene (N1, N2), and human RP gene (RP). (b) Assay interpretation of the results based on HIV and SARS-CoV-2 RT-qPCR results (c) Analysis of HIV and SARS-CoV-2 RNA spiked in lysed negative plasma at 0 or 200 copies/mL. (d) Analysis of HIV and SARS-CoV-2 RNA spiked in lysed pooled NS and plasma at 0 or 200 copies/mL. Individual Cq values of three technical replicates (individually extracted and RT-qPCR assayed) are plotted.

We calculated that we would need at least 30 SARS-CoV-2 positive specimens to estimate 95% CI of 88%−100%, if all analysis is accurate (Binomial Exact Proportion Estimation). For OPTION 1 workflow, remnant NS in VTM that were positive for SARS-CoV-2 (n = 31) and negative for SARS-CoV-2 (n = 36) but positive for other respiratory pathogens (e.g., influenza, seasonal coronavirus, adenovirus, parainfluenza virus, metapneumovirus, enterovirus, bocavirus, and pneumoniae) were used.²⁵ These specimens were either nasopharyngeal specimens collected by health personnel or observed self-collected nasal swabs from individuals presenting respiratory symptoms, as part of the Seattle Flu Study in 2020. The swabs were suspended in 3mL VTM and previously tested in a CLIA-certified laboratory.²⁵ Aliquots of these specimens were de-identified prior to testing in the University of Washington’s Lutz laboratory and Seattle Children’s Research Institute’s Frenkel laboratory, as part of this study.

Remnant HIV-seropositive plasma (n = 30) were used to evaluate the performance of the HIV VL tests using the in-house extraction method compared to the standard extraction method. The HIV-seropositive plasma specimens were obtained from the CIENI/INER (WHO HIV certified lab) plasma bank in Mexico City in 2018, as described.²⁶ HIV VL was measured in all the selected samples using the Abbott m2000 system (Chicago, IL), and reported VL of these samples were all above 1000 copies/ml HIV RNA. We randomly selected at least 30 specimens to allow estimation of VL. All samples were de-identified prior to shipment to and testing in the Seattle Children’s Research Institute’s Frenkel laboratory. HIV-seronegative plasma samples (n = 10, screened by the vendor Innovative Research, Inc, MI, USA) were used as negative controls.

To test the OPTION 2 workflow, a subset of specimens (n = 26) remaining from the OPTION 1 workflow experiments were mixed with NS to create 26 HIV-seropositive specimens, of which 10 were SARS-CoV-2-positive and 16 were SARS-CoV-2-negative specimens.

Ethics statement

The remnant respiratory specimens were collected as part of Seattle Flu Study in 2020 (IRB#: STUDY0006181). Informed consent was obtained from adult participants and parents or permanent legal guardians of participant children. Archived specimens were randomly chosen for use in the current study. Participants from CIENI/INER gave written consent for the use of remnant plasma samples, approved by the Institutional Review Board of the National Institute of Respiratory Diseases (Project code: E02–17).

Determination of limit of detection (LOD) and limit of quantification (LOQ) using contrived samples

We spiked in vitro transcribed LTR HIV RNA (100–60,000 copies/mL) or N1 and N2 nucleocapsid SARS-CoV-2 RNAs (100–60,000 copies/ml) to create contrived plasma and NS samples, respectively. Similar copies/mL were spiked in pooled plasma/VTM. We tested 3 samples at each concentration extracted either by in-house extraction or standard extraction to determine the preliminary LOD for LTR, N1, and N2 RNAs, by using the same extracted volume (5μl) in total 20μl RT-qPCR reaction. We confirmed the final LOD with 3 additional replicates as shown positive by RT-qPCR. LOQ was the lowest limit of quantification which was reliably detected with acceptable accuracy (90–150%) and precision (CV% <30%) using replicates of spiked samples.

Interpretation of measured VL results and statistical analysis

In contrived specimens, the measured VL (copies/mL) from RNA extracted by both methods were plotted against the spiked in RNA input concentration (copies/mL) to demonstrate the performance of each extraction method. The measured VL were classified as “correct” if it fell within the upper and lower bounds of precise measurements defined as ±0.3 log₁₀ RNA copies/mL.²⁷ For HIV-seropositive specimens, measured VL values from the RNA extracted by in-house versus standard methods were assessed for their correlation. R² values from a linear regression fit of scatter plots and p-values from Wilcoxon Signed-Rank test (two-tailed) were reported. Next, we used the US NIH virological failure (VF) cutoff²⁸ or the WHO VF cutoff²⁹ to classify specimens as VL suppressed “VLS” (VL<200 or VL<1000 copies/mL, respectively) or virologic failure “VF” (VL≥200 or VL≥1000 copies/mL, respectively). Virologic failure occurs when antiretroviral therapy (ART) fails to achieve and sustain a patient’s viral load to less than 200 copies/mL (US NIH) or 1000 copies/ml (WHO). Percent concordance of the HIV VL classifications from RNA extracted by the in-house versus the standard RNA extraction method was calculated from the number of concordant VL classifications (i.e., VLS vs VF) divided by the number of seropositive samples tested. Similarly, for SARS-CoV-2 specimens, measured VL values from the RNA extracted by in-house vs standard methods were assessed for correlation using a linear regression fit and for statistical difference using a Wilcoxon Signed-Rank test (two-tailed). Next, we classified results as “negative,” “positive,” or “presumptive positive,” based on the results of N1, N2, and RP assays, defined as: “Positive” equates to detection of N1 and N2 regardless of the results from the RP assay; “Presumptive positive” to detection of either N1 or N2 regardless of the results from the RP assay; “Negative” when neither N1 or N2 are detected, but RP is detected; and “Indeterminate” to when N1, N2, and RP are not detected. Percent concordance of the SARS-CoV-2 results by the in-house vs standard RNA extraction methods was calculated from the # of samples with concordant results (i.e., negative, positive, presumptive positive, and indeterminate) divided by the number of samples tested.

Results

Integrated workflows for HIV viral load test and SARS-CoV-2 screening

SARS-CoV-2 RNA has been detected in both plasma and respiratory specimens; and thus, we envision two workflows for screening SARS-CoV-2 infection while performing routine HIV viral load tests (Fig. 1a). OPTION 1 utilizes plasma specimens as the sole sample source for HIV VL test and SARS-CoV-2 screening. Because SARS-CoV-2 viral loads in nasal secretion specimens (NS) are generally higher than in plasma, we propose an alternative workflow (OPTION 2) to improve the sensitivity of SARS-CoV-2 screening, by combining plasma and NS. An in-house RNA extraction method based on the “Boom method”³⁰ is employed in both OPTION 1 and 2. The extracted RNA is subjected to four RT-qPCR reactions: two non-overlapping regions in the SARS-CoV-2 nucleocapsid gene (N1 and N2), HIV long terminal repeat (LTR) gene, and human ribonuclease P (RP) gene. If the result is + for either N1 or N2 gene and +/− for RP gene, the sample would be labeled as presumptive COVID-19 +. If the result is + for both N1 and N2 genes and +/− for RP gene, the sample would be labeled as COVID-19 + (Fig 1b). If the results are negative for all genes, this would suggest procedural errors (Fig. 1b Invalid), and re-testing and/or re-collecting samples is recommended. The assay results also indicate the HIV and SARS-Cov-2 infection status, and that results would be available to clinician to help assess person’s risk of severe disease. Analysis of contrived plasma and pooled plasma/VTM specimens (Fig 1d and 1e) confirmed detection of 200 copies/mL of HIV and SARS-CoV-2 RNAs. Based on the virologic failure (VF) definition from the US National Institutes of Health,²⁸ this limit of detection (LOD) of the HIV assay at 200 copies/mL would allow differentiation of samples collected from PLH who are experiencing VF (>200 copies/mL, as these would be detected) versus viral load suppressed (VLS, undetected). For SARS-CoV-2 RNA, 200 copies/mL is lower than the reported LOD of the US CDC SARS-CoV-2 assay.³¹. In all contrived samples, the human RP gene as extraction control was detected (Fig. 1d and 1e).

Development of RNA extraction protocol

The in-house RNA extraction protocol is based on Boom’s method,³⁰ wherein RNA is bound to a silica column, followed by washing steps to remove PCR inhibitors and then eluted out in a low salt buffer. Herein, a lysis buffer comprised of 4M guanidinium thiocyanate (GuSCN), a potent chaotropic salt, was used to lyse virus and deactivate RNases, as well as serve as a salt bridge between negatively-charged RNA and negatively-charged silica, although 2M GuSCN was previously shown to be sufficient.³² In addition to GuSCN, the lysis buffer incorporated a combination of a reducing agent (dithiothreitol, DTT), to deactivate RNases via disulfide bond reduction, and sacrificial non-target RNA molecules (tRNA) in a relatively low pH buffer (MES, pH 5), which reduces RNA phosphodiester bonds hydrolysis.³³ DTT and tRNA are prone to oxidation in the liquid form at room temperature, so they were stored at −20 °C and mixed with GuSCN in the MES lysis buffer just before the extraction. A more stable reducing agent, β mercaptoethanol, has been successfully used in lysis buffer for RNA extraction,³⁴ but DTT is less toxic³⁵ and less pungent than β mercaptoethanol. Compared to the standard RNA extraction kit, in-house extraction with 40–80mM DTT alone (Supplementary figure 3) recovered 30% of spiked HIV RNA. When combining 4–6% v/v (22.4–33.6μg/560μl lysis buffer) non-target tRNA and DTT (Supplementary figure 4), RNA recovery of the in-house extraction was improved, comparable to the standard RNA extraction kit. The in-house extraction silica column (Supplementary table 2) has sufficient capacity, 40μg, to accommodate absorption of both tRNA and HIV RNA targets. However, tRNA in excess (>33.6μg/60 μl of RNA extraction buffer) was shown to compete with the target RNA during the amplification step (Supplementary figure 5). Also, our prior work observed inefficient target amplification when non-target DNA is present at more than 2 ng/μL depending on the PCR reaction conditions.^{36, 37} To wash loosely-bound charged molecules from the silica column and further purify the RNA, the washing step started from a high salt buffer: 1M GuSCN, 10mM Tris-HCl pH 7.4; and followed by a low-salt buffer: 10mM Tris-HCl pH 7.4 with 80% ethanol. A high concentration of ethanol can solubilize residual salts but not RNA from the silica column when GuSCN is absent to serve as a salt bridge.

Commercial RNA extraction method (referred as standard method)

The analytical sensitivity of both workflows (OPTIONS 1 and 2, Fig 1a) using the in-house extraction method was assessed by comparing directly to the standard extraction method. For the OPTION 1, we analyzed single-target contrived samples (i.e., either HIV or SARS-CoV-2 RNA spiked in lysed human plasma). For the OPTION 2, we analyzed HIV and SARS-CoV-2 RNA spiked in lysed pooled NS/plasma at 0 or 200 copies/mL. Fig. 2a and b are scatter plots of the input VL versus measured VL of contrived plasma and NS samples processed by in-house or standard methods. The extraction efficiencies (slope of fitted linear regression; Fig. 2a and b) were 0.616 – 1.523 and 0.858 – 0.995 for the in-house and the standard methods respectively. HIV VL measurements with both extraction methods are in close agreement as shown by linear regression fits, R² = 0.98 in both plots (Fig. 2a). However, for pooled plasma/VTM samples (OPTION 2), R² of 0.91 and 1 were estimated for measured HIV VL with in-house and standard extraction methods, respectively (Supplementary figure 6a). In Fig. 2a and b, black diamonds are samples with measured VL significantly deviated from the theoretical yield. For both extraction methods, the samples with ≤600 copies/mL were found to be more variable because the theoretical RNA copy number per RT-qPCR reaction was <7 copies/reaction (140μL starting sample volume with <600 copies/mL). Similar variations were also observed in OPTION 2, pooled plasma/VTM (Supplementary figure 6). Overall, the LOQ of HIV VL in contrived plasma (Fig 2a) or plasma/VTM (Supplementary Fig 6a) is 600 copies/mL. However, the LOQ for SARS COV-2 VL in contrived VTM (Fig 2b) or pooled plasma/VTM (Supplementary figure 6b) is 200 copies/ml for N1 and N2 RNAs. The performance of both workflows for single-target and multi-target (pooled plasma/VTM) analyses are summarized in Fig. 2c and 2d, respectively. The analyses classify HIV status as negative (undetectable VL), VLS, or VF and SARS-CoV-2 status as negative or positive. Both analyses show assay results with both extraction methods are in complete agreement except the HIV specimens with 200 copies/mL, where the assay with the standard extraction method only detect 1/3 (Fig. 2c). LODs are 200 copies/mL for HIV in plasma and 100 copies/mL for SARS-CoV-2 in VTM(OPTION 1, Fig 2c), which are identical for HIV and SARS-CoV-2 in pooled plasma/VTM (OPTION 2, Fig 2d and Supplementary figure 6).

Fig. 2. Quantitation and classification of HIV and SARS-CoV-2 RNA obtained from inhouse and standard extraction methods.

Scatter plots of (a) measured HIV VL (n = 3) vs input RNA in contrived plasma samples (lysed plasma and spiked with synthetic HIV RNA at 0, 100, 200, 600, 6K, or 60K copies/mL), and (b) measured SARS-Cov-2 VL (n = 3) vs input RNA in contrived NS (lysed and spiked with synthetic SARS-CoV-2 RNA at 0, 100, 200, 600, 6K, or 60K copies/mL). Diagonal lines represent 100% theoretical RNA recovery from the extraction method based on the spiked-in RNA input. Dashed diagonal lines indicate the bound for precise measurement (i.e., measured VL within ±0.3 log10(VL, copies/mL of input)). Efficiency is the slope of the linear regression fit, reported along with its R2. Summary of HIV VL classification and SARS-CoV-2 detection using both extraction methods using (c) OPTION 1 workflow to analyze either NS or plasma, and (d) OPTION 2 workflow to analyze pooled plasma/NS samples. For HIV samples, we classified results as viral load suppressed (VLS) or virological failure (VF). For SARS-CoV-2, we classified results as positive or negative based on whether they are detected or not detected by both N1 and N2 assays. Supplementary figure 6 is the scatter plots of the OPTION 2 workflow. Data in Fig. 2 and Supplementary figure 2 is collected by a single assay operator.

Clinical evaluation of HIV in plasma specimens

Fig. 3a is the study design for comparing the performance of standard versus in-house extraction methods using plasma specimens from patients, including 30 HIV-seropositive specimens, confirmed by Abbott m2000 VL test, and 10 HIV-seronegative specimens. Assays using the in-house extraction method (Fig. 3b) detected HIV RNA in all (30/30; 95%CI: 88–100) HIV-seropositive plasma (median VL: 30,604 copies/mL; IQR: 6,503 – 96,626 copies/mL) and did not detect HIV RNA in HIV-seronegative plasma (10/10). Human RP genes were detectable in all specimens, confirming a functioning assay. However, human RP levels were significantly higher in HIV-seronegative plasma (Fig. 3b and Supplementary figure 7). Measured HIV VL from assays using in-house and standard methods were positively correlated (Fig 3c, linear regression fit, R² = 0.81) but were statistically different (p<0.05, Wilcoxon Signed-Rank test, two-tailed). Measured HIV VL for 20/30 specimens were within the precise measurement bound (Fig. 3c shaded area). Compared to the standard method, RNA extracted by the in-house method led to 7 overestimated VL values and 3 underestimated VL values (black diamonds, Fig 3c). Overall, classifications (i.e., VLS vs VF) of measured VL results (Fig. 3d) obtained by both extraction methods, had 100% concordance (30/30 specimens; 95% CI: 88–100).

Fig. 3. Quantitation and classification of HIV RNA in clinical plasma specimens.

(a) Schema of experimental design to compare the performance of standard vs in-house extraction methods. Plasma RNA extracted by standard or in-house extraction was evaluated using RT-qPCR. The RT-qPCR results were compared in terms of measured viral load and classifications. All HIV seropositive specimens had VL ≥1000 copies/mL when tested by the reference lab using Abbott m2000. (b) Cq values of the HIV long-terminal repeat (LTR) assay and human ribonuclease P (RP) assay (control for negative specimens) from the plasma RNA extracted by the in-house method. Extracted RNA from HIV-seropositive specimens were analyzed by LTR and RP RT-qPCR assays in duplicate. HIV-seronegative specimens (as tested by the vendor) were each extracted twice, and each extracted RNA aliquot was analyzed by LTR and RP RT-qPCR assays. The average RP Cq values are plotted. (c) Scatter plot of measured HIV VL in specimens extracted by standard vs in-house methods. The diagonal line represents a theoretical 100% correlation of measured HIV VL of specimens extracted by both extraction methods. The shaded box around the diagonal lines indicates a precise measurement bound (±0.3 log10(VL measured in RNA extracted by standard method, copies/mL)). Diamonds are samples with measured VL outside precise measurement bound. (d) Classifications correspond to VL results. Negative results are labeled as “-” in white boxes. Virologic failure results were classified using 200 copies/mL (dark purple boxes) and 1000 copies/mL (light purple boxes) thresholds. RNA extracted by both methods was collected by one assay operator and analyzed by RT-qPCR by another assay operator.

Clinical evaluation of SARS-CoV-2 in NS specimens

Fig. 4a is the study design for comparing the performance of standard versus in-house extraction methods using NS specimens from patients, including 31 SARS-CoV-2-positive, confirmed by a CLIA-certified laboratory, and 36 negative specimens. Due to the limited specimen volume, the study extracted 50μL specimens instead of 140μL (as was done in the aforementioned contrived sample study). Of 31 positive specimens (Fig. 4b and c), assay with the in-house extraction method resulted in 25 (81%) positive (detected both N1 and N2 genes; median VL: 4,180 copies/mL; IQR:3,912–176,168 copies/mL) and 6 (19%) presumptive positive (detected either N1 or N2 gene; median VL: 84 copies/mL; IQR: 39–203 copies/mL). Assays with the standard extraction method for two positive specimens (VL: 162 and 268 copies/mL) resulted in one false negative (Specimen ID 83, Fig 4c) and one presumptive positive (Specimen ID 87, Fig. 4c). Measured SARS-CoV-2 VL obtained by in-house versus standard extraction methods revealed high correlations (R² = 0.95 for N1 in Fig. 4d; R² = 0.99 for N2 in Fig. 4e) and had no significant difference (p>0.05; Wilcoxon Signed-Rank test, two-tailed). The analysis shows that the measured VL of the N1 assay for two low VL specimens falls outside the precise measurement bound (Fig. 4d black diamonds), but the measured VL from the same specimens were within acceptable variation in the N2 assay (Fig. 4e). These specimens exhibited low SARS-CoV-2 RNA that are around the LOD (100 copies/mL), observed from the contrived specimens (Fig. 2b), and the N1 assay has been reported to have lower sensitivity than the N2 assay.²⁴ Of 36 negative specimens, the RP gene was not detected in 2 specimens processed by the in-house method and 5 specimens processed by the standard method (Fig 4c), resulting in indeterminate (IND) results. These specimens contained <250 copies/mL human RP, thus, this discordance is likely due to variations of RT-qPCR near the LOD. Fig. 4f summarizes the SARS-CoV-2 VL detection results in positive and negative specimens processed by standard and in-house extraction methods.

Fig. 4. Quantification and detection of SARS-CoV-2 RNA in clinical NS specimens.

(a) Schema of experimental design to compare the performance of standard vs in-house extraction methods. NS RNA obtained from standard or in-house extraction were evaluated using RT-qPCR. The RT-qPCR results were compared in terms of measured viral load and classifications. Discordant classifications were ruled out based on previous results from the CLIA-certified lab. (b) Cq values of the SARS-CoV-2 N1, N2, and human RP assay (control for negative specimens) from the in-house method. Each specimen was extracted, followed by RT-qPCR, and represented as individual data points. (c) Classifications of specimens based on the results from the two extraction methods. Specimens detected by N1 and N2 (regardless of RP detection) are classified as positive; specimens detected by either N1 or N2 assays are classified as presumptive positive; specimens not detected by N1 and N2 but detected by RP assay are classified as negative; specimens not detected by any assays are classified as indeterminate (IND). Scatter plot of measured SARS-CoV-2 in specimens extracted by standard vs in-house methods in (d) N1 assay (e) N2 assay. The diagonal lines represent a theoretical 100% correlation of SARS-CoV-2 VL measured in RNA extracted from both extraction methods. Dashed lines indicate RT-qPCR precise measurement bound (i.e., ±0.3log10(VL measured in RNA extracted from standard method, copies/mL). Diamonds are samples with measured VL that significantly deviates from those obtained via standard extraction. (f) Summary table for classifications by the two methods. Data were collected by two assay operators.

Clinical evaluation of HIV and SARS-CoV-2 in pooled clinical plasma and NS

We tested the OPTION 2 co-extraction workflow on a subset of available remnant patient specimens, including NS and plasma (Fig. 5a). Here, each HIV-seropositive plasma specimen (n = 26) was randomly assigned to mix with either positive (n = 10) or negative (n = 16) SARS-CoV-2 specimens. Each pooled specimen was then processed using the in-house method and the standard extraction method. Fig. 5b shows the Cq values of SARS-CoV-2 N1 and N2, HIV LTR, and human RP assays using the in-house extraction method. Using the US NIH HIV VF value, 200 copies/mL, assays using both extraction methods resulted in 100% (26/26) concordance (Fig. 5c). Detection of SARS-CoV-2 was also 100% (26/26) concordant. Fig. 5d – f shows high correlations between measured VL of HIV LTR, SARS-CoV-2 N1, and SARS-CoV-2 N2 targets, respectively. Interestingly, the measured HIV VL in these specimens processed by the in-house method were significantly higher than those by the standard method (p<0.05, Wilcoxon Signed-Ranked test, two-tailed); but the measured VL of SARS-CoV-2 are similar between the two groups.

Fig. 5. Comparison of HIV LTR and SARS-CoV-2 Cq values obtained from RNA co-extracted from NS/plasma specimens using in-house vs standard extraction protocols.

(a) Schema of experimental design to compare the performance of standard vs in-house extraction methods. NS/plasma RNA extracted by standard or in-house extraction were evaluated using RT-qPCR. The RT-qPCR results were compared in terms of measured VL and classification of infection status as shown in Fig 1b. (b) Cq values of the SARS-CoV-2 LTR, N1, N2, and human RP assay (control for negative specimens) from the in-house method. Each specimen was extracted, followed by RT-qPCR, and represented as individual data points. (c) Classifications of specimens based on the results obtained from the two extraction methods. Scatter plots showing correlation between the in-house and standard extraction methods for HIV VL by LTR assay (d), SARS-CoV-2 VL by N1 assay (e) and SARS-CoV2 VL by N2 assay (f) in pooled NS/Plasma specimens. The diagonal lines represent a theoretical 100% correlation of VL measured from both methods. Dashed lines indicate precise measurement bound. Diamonds are samples with measured VL that significantly deviates from those obtained via standard extraction.

Discussion

We developed a new workflow with an optimized in-house RNA extraction method to enable simultaneous quantification of HIV VL and detection of SARS-CoV-2 in a single RT-qPCR run. The workflow can be tailored to different clinical use by isolating target RNA from plasma or plasma/NS mixture via a single extraction kit.

In contrived specimens, both OPTION 1 and OPTION 2 workflows can detect low concentrations of the HIV RNA (200 copies/mL) and SARS-CoV-2 RNA (100 copies/mL). We anticipate that this highly sensitive SARS-CoV-2 assay would be able to detect most SARS-CoV-2 positive clinical specimens, as 100 copies/mL is lower than the reported medians VL in blood during the first two weeks of infection and ~50-fold lower than the medians of SARS-CoV-2 viral loads in respiratory samples during the first 6 weeks.¹⁸ This analytical sensitivity was successfully transferred to the clinical samples. Out of 30 HIV seropositive specimens, as confirmed by the Abbott m2000 VL test in CLIA certified laboratory, the VL of 2 specimens was ~ 200 copies/ml of plasma. These specimens were successfully detected by standard and in-house RNA extraction/RT-qPCR workflows in our laboratory. No false-negative results suggested that LOD obtained by contrived samples for HIV RNA is transferable to the detection of clinical specimens with a low HIV VL. When testing clinical NS, we found that the in-house method had positive results for all SARS-CoV-2 respiratory specimens with ≥1000 copies/mL but only around 50% positive and 50% presumptive positive results in ≤1000 copies/mL specimens. For SARS-CoV-2 positive specimens (31 specimens) as confirmed by CLIA certified laboratory, assay with in-house extraction resulted in 6 presumptive positives (detected either N1 or N2 gene) specimens with median VL of 84 copies/ml (IQR: 39–203 copies/ml). The reason for presumptive positive specimens may be partial loss of RNA integrity at multiple freeze-thaw cycles affected the detection efficiency of the N1 gene which has been reported to be less analytical sensitive than the N2 assay. Detection is more affecting for the specimens with VL near to LOD of 100 copies/ml. Due to limited availability of specimens, we only used 50μL specimens from 1mL swab eluate. Processing a larger volume of specimens (i.e., 140μL per extraction such as the case of HIV specimens or contrived SARS-CoV-2 samples), would likely improve detection of SARS-CoV2 specimens with <1000 copies/mL by increasing the RNA input in the final RT-qPCR reaction. In clinical plasma specimens, the in-house method also accurately classified HIV VF compared to the standard method.

The new workflow can potentially streamline HIV management and SARS-CoV-2 detection by determining if a PLH has detectable viremia as well as SARS-CoV-2 infection by extracting RNA from plasma alone or pooled plasma/NS using a single in-house method. Combining the detection of these two viruses can help address the challenges associated with the overlay of the COVID-19 pandemic on the existing HIV epidemic.

Our workflow includes an internal control based on detection of the human RP gene, which is used for many clinical assays.^38–40 In addition to indicating if assay conditions allow amplification, detection of the RP gene confirms collection of adequate specimen, which is especially helpful when NS specimens for SARS-CoV-2 are self-collected. As an alternative to human gene, MS2 bacteriophage can be spiked into the sample as an internal control^{41, 42} to accurately measure the performance of extraction and amplification.

The workflow also helps to circumvent the issues associated with the availability of commercial RNA extraction kits and reduces the cost of RT-qPCR. Because of a high demand for COVID-19 tests, procuring RNA extraction kits has been difficult, which also limits routine HIV testing. RNA extraction kits have contributed significant reagent cost for most RT-qPCR workflows (approximately $6/sample extraction). Our in-house method uses low-cost alternatives that can be obtained as off-the-shelf reagents (Supplementary table 2), which allows laboratories to assemble their own low-cost RNA extraction kits ($1.85/sample extraction). Using the proposed workflow to analyze a specimen of plasma/NS mixture via in-house RNA extraction kit costs $1.85, 5-times lower than the standard extraction, requiring separate plasma and NS specimen extractions ($12 for 2 specimens). Processing individual HIV and SARS-CoV-2 specimens adds indirect costs that can be reduced with a multiplexed assay using pooled specimens. A robust in-house workflow at approximately 20% of the cost of equivalent commercial assays could benefit low-resource environments and make high-volume testing affordable. The goal of this report is not only to assess the performance of our optimized multiplexed in-house workflow but to inform researchers about the easy, reliable, and cost-effective option for screening COVID-19 in HIV patients with high throughput.

Our assay is high-throughput with the advantage of testing SARS-CoV-2 and HIV together in one sample with pooled plasma/NS matrix. Both multiplexing and specimen pooling can substantially increase testing throughput. An optimized RT-qPCR assay for a pooled matrix with a run time ~2h further reduces the time to generate results. With extraction and RT-qPCR controls being included in the test properly, at least two times as many samples with multiplexing can be tested in a single run per plate per real-time PCR instrument compared to samples tested in the separate run of RT-qPCR by FDA or CDC approved standard assay for HIV and COVID-19. This diagnostic assay may allow SARS-COV2 screening in HIV patients in a centralized facility which reduces the risk of undetected covid patients. For institution-based researchers with a BSL-2+ facility, the multiplex assay for quantifying HIV and SARS-COV2 RNAs will be critical to understanding the antiviral treatment effects on COVID-19 in HIV seropositive patients or in preclinical studies where sample volume is limited.

Our workflow detects HIV, SARS-CoV-2, and human endogenous control via separate RT-qPCR reactions in a same run to maximize sensitivity. Bi-plex detection of SARS-CoV-2 and human endogenous control has been performed, but the sensitivity was reduced compared to single-target amplification.⁴³ To improve the detection of the co-amplification, sequence-specific extraction of RNA from HIV or SARS-CoV-2^{36, 44–46} could be useful by increasing the target concentrations. Here we detected each RNA target in a separate well of PCR plate using a single-target RT-qPCR amplification. Instead, using multiplexed amplification could reduce per-test cost since one sample can be analyzed in a single RT-qPCR reaction per sample instead of four RT-qPCR reactions per sample. To replace the need for a real-time thermal cycler, it is possible to couple a standard thermal cycler with an end-point analysis via cell phone images shown to have a high correlation with the end-point signal measured by the qPCR thermal cycler.²⁵

We developed an assay that combines detection of HIV and SARS-CoV-2. The utility of this workflow can be tailored for the detection of other common HIV comorbidities in PLH such as tuberculosis, hepatitis B, hepatitis C, and other AIDS-defining opportunistic infections (e.g., Pneumocystis jirovecii pneumonia, histoplasmosis, toxoplasmosis, cryptococcal meningitis and mycobacteriosis) which are causing high disease burden in low and middle income countries. Our novel and similar assays would allow routine VL tests to also screen for other diseases that if detected early could be treated and improve the health of PLH, reducing both costs of diagnosis costs and morbidity from delays in diagnosis.