Implementation of a Point Mutation Assay for HIV Drug Resistance Testing in Kenya: Performance, Challenges, and Cost

Abstract

Objectives:

An increasing prevalence of HIV pre-treatment drug resistance (PDR) has been observed in Sub-Saharan Africa, which could decrease the effectiveness of antiretroviral therapy (ART) programs. We describe our experiences, the costs and challenges of implementing an oligonucleotide ligation assay (OLA) for management of PDR in Nairobi, Kenya.

Design:

An observational report of the implementation of OLA in a Kenyan laboratory for a randomized clinical trial evaluating whether onsite use of OLA in individuals initiating ART would decrease rates of virologic failure.

Methods:

Compared detection of mutations and proportion of mutants in participants’ viral quasispecies by OLA in Kenya vs. Seattle. Reviewed records of laboratory workflow and performance of OLA. Calculated the costs of laboratory set-up and of performing the OLA based on equipment purchase receipts and supplies and labor utilization, respectively.

Results:

OLA was performed on 492 trial participants. Weekly batch-testing of median of 7 (range: 2–13) specimens provided test results to Kenyan clinicians within 10–14 days of sample collection at a cost of US$42 per person tested. Cost of laboratory setup was US$32,594. Challenges included an unreliable local supply-chain for reagents and the need for an experienced molecular biologist to supervise OLA performance.

Conclusions:

OLA was successfully implemented in a Kenyan research laboratory. Cost was twice that projected due to fewer than predicted specimens per batch because of slow enrollment. OLA is a potential simple, low-cost method for PDR testing in resource-limited settings (RLS). Ongoing work to develop a simplified kit could improve future implementation of OLA in RLS.

Introduction

Antiretroviral (ART) delivery programs have been implemented across Sub-Saharan Africa and other resource-limited settings (RLS) during the past 10–15 years. Over this time, an increasing prevalence of HIV pre-treatment drug resistance (PDR) has been observed, which could decrease the effectiveness of ART programs [1, 2]. In the past, transmitted drug resistance (TDR) and PDR became prevalent in epidemics of men-who-have-sex-with-men and intravenous drug users in New York and Zurich, respectively [3, 4], leading to the recommendation to test for HIV drug resistance (HIVDR) to guide selection of initial and subsequent ART regimens [5, 6]. In RLS, the use of drug resistance tests to guide the selection of ART regimens has been extremely limited due the costs of reagents and the infrastructure needed to perform the assay [7]. An effective, affordable, and feasible strategy is needed to address the challenges posed by PDR [8, 9].

Our group developed a low-cost point-mutation assay to detect HIVDR, known as oligonucleotide ligation assay (OLA), that requires relatively simple and inexpensive equipment. OLA is designed to detect “major mutations” commonly selected by WHO-recommended first-line ART regimens based on non-nucleoside reverse transcriptase inhibitors (NNRTI). OLA has been shown to detect PDR among ARV-naïve individuals and selected or “acquired” drug resistance (ADR) in women pre-treated with single-dose nevirapine (NVP) associated with virologic failure [10, 11]. Analyses of a large number of genotypes from cases of transmitted drug resistance (TDR) in RLS confirmed that a point-mutation assay could detect a substantial portion of TDR/PDR [12, 13].

The OLA uses polymerase chain reaction (PCR) to amplify a region of HIV pol and probes with 3’ bases specific for the positions of interest that encode resistance or wild-type template. Ligation of probes that anneal adjacent to region of interest confers ~100% specificity, with the relative proportions of mutant and wild-type variants subsequently detected by an enzyme-immuno-absorbent assay (EIA) [14, 15]. Previously, the OLA was successfully implemented for research purposes by laboratories in low- and middle-income countries, including Thailand[10, 16] and Zimbabwe [17]. In 2013, OLA technology was transferred to the research laboratory at the Coptic Hope Center for Infectious Diseases (CHCID) in Nairobi, Kenya, to conduct a randomized clinical trial (RCT) to test for PDR to guide clinicians’ choice of the initial ART regimen [18].

Given the paucity of published reports analyzing the implementation of HIVDR testing for clinical management of PDR in RLS, we describe our experiences and the cost of implementing OLA for PDR management in one laboratory in Nairobi, Kenya, including discussions of quality control and problems that occurred while performing the assay.

Methods

Laboratory and Clinical Trial Setting

The OLA was setup in the research laboratory at the CHCID to conduct a RCT evaluating whether onsite use of OLA in individuals initiating ART would improve rates of suppression of viral replication [18]. Prior to establishing the OLA, the laboratory consisted of one 10.5m² room equipped with a laminar flow hood, a water bath, two centrifuges, a refrigerator/−20^oC freezer and a −80^oC freezer. The Seattle Laboratory Manager (SLM) trained two Kenyan laboratory technicians. Training included a review of Good Clinical Laboratory Practices (GCLP) and instruction in “standard operating procedures” (SOP) to perform the OLA.

Enrollment of subjects into the RCT started in May 2013 and continued through November 2014. Prior to initiation of ART, peripheral blood mononuclear cells (PBMC) from enrollment blood samples of participants randomized to the OLA arm were isolated and tested by OLA for NNRTI resistance mutations K103N, Y181C and G190A, and lamivudine mutation M184V. The study protocol called for initiating protease inhibitor-based ART when a mutant was found by OLA at a frequency of ≥10% of a participant’s viral quasispecies.

Quality Control of OLA

The Kenyan laboratory was setup by the SLM who routinely performs our CLIA-certified OLA. Once the equipment was installed in Kenya, cloned wild-type (WT) and mutant HIV DNA standards used in the CLIA-certified OLA were evaluated and shown to perform as in the Seattle laboratory. After training, the technicians’ competency in performing the OLA was assessed using a panel of blinded PBMC specimens previously tested in the Seattle laboratory. Ongoing quality control during the study included shipping of PCR amplicons generated from subjects’ pre-ART specimens to Seattle for repeat OLA testing and consensus sequencing (CS), including all amplicons with HIVDR detected, those testing indeterminate by OLA (low or no signal for both the mutant and wild-type genotypes) and 20% of those testing wild-type. The detection and proportion of each mutant variant detected by OLA in each subject’s HIV amplicon tested by OLA in Kenya vs. Seattle were compared for agreement (R²).

Consensus sequencing was performed on the enrollment, month-12 and at times additional specimens from subjects with virologic failure (VF) at month-12 of ART. Neighbor-joining phylogenetic trees were constructed of all sequences generated to check for potential mix-ups of specimens during sample processing or due to PCR carry-over contamination.

Cost Analysis of Laboratory Set-up and Testing

The cost of setting up the laboratory to perform OLA at the CHCID was estimated using records of equipment purchased in Seattle and sent to the site or bought in Kenya. The average cost of performing the OLA for each participant was estimated using the cost of supplies and labor used to assay specimens batch-tested in each run of the OLA. As most reagents and supplies were purchased in Seattle, supply unit costs were based on Seattle prices and included all steps of the OLA process: blood collection and separation, PBMC DNA extraction, PCR/gel, and OLA test.

Results

Laboratory Preparation for OLA Testing in Kenya

The existing research laboratory at CHCID was used to process and aliquot specimens and for administrative activities. Bench space to accommodate the equipment and to perform the OLA was limited, and there were no separate pre- and post-PCR areas. To reduce the risk of “carry-over” of PCR amplicons, the laboratory with the laminar flow hood, measuring 10.5m², was designated as the pre-PCR specimen processing area, and an adjacent office measuring 5.5m² was remodeled into a post-PCR laboratory. Newly purchased equipment necessary to perform OLA included a thermocycler, a gel box and power supply for electrophoresis, a camera, a spectrophotometer and a plate washer. These were purchased in Seattle and shipped or transported to Kenya by travelers. Plastic-ware and basic supplies were purchased in Kenya. However, the probes, enzymes and other molecular reagents needed to perform DNA extraction, PCR and the OLA, were either not readily available in Kenya or were significantly more expensive compared to buying in Seattle and shipping to Kenya. Thus, these were purchased in Seattle. The total cost of equipment needed to set-up the laboratory was $32,594 (Table 1). The SLM traveled to Nairobi and spent three weeks at the CHCID setting-up the laboratory, programming and testing the equipment and training the Kenyan laboratory technicians to perform the OLA.

Table 1.

Cost of Laboratory Set-Up

Equipment	Cost US$
Thermocycler (Applied Biosystems 2720)	$3,080
Microplate reader and software (iMark)	$6,820
Gel box, combs and power supply	$1,240
PhotoDoc-It Imaging System	$3,676
Centrifuge for blood separation	$3,182
Eppendorf Microcentrifuge (1.5mL tubes)	$2,213
2 sets of micropipettes (start pack)	$1,128
Repeater pipettor	$495
Multichannel pipet	$523
Quick spinner 1.5mL tubes	$315
Quick spinner 0.2mL strip tubes	$315
Microwave*	$65
Plate Washer (Biolabs)*	$3,360
Nanodrop*	$6,182
Total	$32,594

^*Purchased in Nairobi. All other items were purchased in Seattle and sent to Nairobi

Training and Educational Background of Kenyan Staff

In January 2013, our SLM trained one Kenyan laboratory technician to perform PBMC separation, DNA extraction, PCR, agarose gel electrophoresis, and OLA. After training, she successfully performed the OLA SOP under the SLM supervision using specimens previously tested in the Seattle laboratory. In October 2013, the SLM traveled to Nairobi and trained a second technician. Prior to learning to perform OLA, one of the two technicians had received diplomas in medical laboratory sciences, and the other had completed a bachelor of science but had limited education in the theory and practice of molecular techniques. Neither of these two technicians had prior experience in molecular techniques. During 1.5 years of enrollment into the study, the SLM reviewed all OLA results and provided ongoing training, coaching, help with troubleshooting and discussion of results during weekly Skype calls.

Laboratory Workflow and Cost of Performing OLA

During the enrollment phase of the clinical trial (May 2013-November 2014), the OLA was performed once weekly to batch-test an average of seven patient specimens (range 2–13). To perform each batch-test of specimens, the technicians spent 10 hours of hands-on time, in addition to each round of a nested PCR lasting two hours. Thus, the assay took 14 hours from start to finish, performed over the course of three days (Table 2). Using the batch-tested size of seven specimens, 10 hours of labor, and average Kenyan technicians’ wage of US$4/hour, the cost of performing OLA was calculated at US$42/specimen (Table 3).

Table 2.

Typical Workflow of OLA

	Monday	Tuesday	Wednesday	Thursday	Friday
Week 1	Enroll subjects, collect blood, isolate PBMC, and freeze samples.
Week 2		PBMC DNA Extraction	Nested PCR/Gel	OLA	Results emailed to Seattle
Week 3	Final OLA results from prior week emailed to Kenya and shared with clinicians	ART Initiation

Table 3.

Cost of OLA Per Person Tested

	Cost Category (US$)†
Procedure	Supplies	Labor°
PBMC DNA extraction	$5.92	$1.14
Nested PCR/gel	$3.65	$0.86
OLA	$26.52	$3.71
Category Totals	$36.09	$5.71
Total Cost Per Person Tested	$41.80

^†All costs presented are per person costs using seven samples per batch.

^°To calculate the cost of labor per procedure per person, we started with the total number of hours of hands-on labor for each procedure, divided by seven people, and then multiplied by an hourly Kenyan wage of US$4.

Kenyan OLA data were emailed on Fridays to the SLM, who spent an additional 2 hours of labor per run to review the data, send an email with verified results and feedback, and hold a Skype call on Mondays. The purpose of the call was to provide further teaching and technical support. In Kenya, OLA results were used to generate a HIVDR genotype report for each study participant that listed the HIV codons evaluated by OLA and whether or not mutations were detected, along with an interpretation: “no resistance, prescription of 1st-line NNRTI-based ART is indicated” or “resistance detected, prescription of 2nd-line PI-based ART is indicated”. Following quality assurance review, the reports were released to the subjects’ clinicians. Based on this workflow, test results were available 10–14 days after a subject’s blood sample was collected. The median time from enrollment to ART initiation, using the HIVDR report, was 18 days (range 11–159). During the trial, Kenyan laboratory technicians performed OLA and reported results on 492 pre-ART blood samples.

OLA performance in Kenya and quality control

On two separate occasions, suboptimal performance of OLA led to month-long pauses in enrollment and testing of specimens. First, in late November of 2013, the OLA mutant standards signal suddenly diminished in intensity, decreasing the sensitivity for detection of mutants at low frequencies in a subject’s HIV quasi-species. Enrollment was paused from December 2–31, 2013, to troubleshoot steps of the assay. Troubleshooting suggested that the locally purchased, double-distilled water was affecting performance of the wash buffers, so the source of water was changed. Also, new aliquots of all standards and ligation and detection reagents were sent from Seattle to Nairobi. These actions led to expected OLA performance parameters. The second pause occurred in August of 2014 when PCR-amplification failed for several specimens and analysis of PBMC DNA by agarose gel electrophoresis revealed degraded DNA. Contamination of isolated specimens with bacterial DNases was suspected. Study enrollment was paused between September 2-October 5, 2014. During this time, the laboratory’s surfaces and all equipment were thoroughly cleaned, including replacement of filters in the biologic safety cabinet and air conditioning ducts. Also, laboratory practices were extensively reviewed to minimize lapses that could lead to bacterial contamination. New DNA extraction reagents were sent from Seattle. Subsequently, DNA degradation was no longer observed.

The quality assurance of OLA results in Seattle, performed for all amplicons generated in Kenya that produced HIVDR or indeterminate OLA results in Kenya plus a random selection of 20% of amplicons that produced WT results, showed that the proportion of each mutation detected at each codon in Kenya and in Seattle correlated closely (R² for K103N, Y181C, M184V and G190A = 0.88; 0.96; 1.00; and 1.00, respectively) (Figure 1). The mean absolute difference in the proportion of each mutation detected in Kenya versus Seattle was 7.4% (range 0–65%), 4% (0–29%), 2% (0–4%), and 3% (0–6%) for K103N, Y181C, M184V, and G190A, respectively. There was one major outlier in which the absolute difference in the proportion of K103N detected in Kenya versus Seattle was 65% (15% vs. 80%). Consensus sequencing of this amplicon did not detect this mutation suggesting that the proportion of K103N present was likely < 20–25%. The 80% result in Seattle could be due to potential contamination of second-round PCR. Given the RCT cutoff threshold for prescription of PI-based ART (i.e., mutant >10% an individual’s quasispecies), it is important to note that only four subjects were identified with ≥10% mutant by the OLA in Kenya (13–17%) who had <10% mutant in Seattle (5–9%). Discordant results in the opposite direction (<10% mutant in Kenya, but ≥10% in Seattle) were not found. One subject had resistance detected at 11% in Nairobi and was prescribed PI-based ART; however, retesting of this sample in Nairobi and Seattle confirmed the original result was an error.

Figure 1.

Correlation of proportion of each mutant detected in Kenya and Seattle. Performance of the OLA in Kenya (y axis) correlated closely with performance by the Seattle CLIA-approved assay across all codons: K103N: R²= 0.88, y=0.91x + 0.056; Y181C: R²= 0.95; y=0.92x + 0.017; M184V: R²=1.00, y=0.98x + 0.023; G190A: R²=1.00, y=1.04x + 0.005

A neighbor-joining phylogenetic tree was constructed using consensus sequences generated from both the enrollment and month-12 time points from 472 subjects. The phylogenetic analysis revealed 12 instances of potential mix-ups between subjects’ specimens either at the time of specimen collection, labeling or downstream processing. To better understand and resolve these errors, we carefully reviewed blood collection/processing dates and PCR and OLA set-up sheets, and performed sequencing of additional aliquots of these samples and multiple other specimens from these subjects as needed. The errors included five instances of mislabeling blood specimens processed on the same day in Kenya and one instance of PBMC mix-up during DNA extraction in Seattle. In addition, there were five instances where the wrong DNA was used for PCR and one instance of cross-contamination with PCR amplicon, but we were unable to determine which of these six mix-ups occurred in Kenya vs. Seattle. These errors caused one participant to incorrectly receive PI-based ART at treatment initiation.

Discussion

Our principle findings are (1) transfer of OLA technology to a Kenyan research laboratory, previously experienced only in collection and processing of blood, allowed local technicians to perform the OLA with test results comparable to the CLIA-certified OLA in Seattle; (2) the OLA was performed on 492 trial participants at a cost of approximately US$42 per test; and (3) weekly batch-testing of specimens provided HIVDR results to Kenyan clinicians within 10–14 days of sample collection to guide choice of ART regimen at treatment initiation.

While OLA was set up and performed successfully in Kenya, we encountered challenges in the implementation process. First, several important test reagents and supplies were not readily available in Kenya, which required us to transport supplies from Seattle. Development of a local, reliable, and affordable supply chain would facilitate implementation of the OLA across multiple sites. Second, the laboratory space used for pre-PCR specimen processing was also used for administrative activities. This led to a high-traffic environment that may have distracted technicians, potentially leading to errors and/or specimen mix-ups. Other small laboratories could experience similar challenges. Use of barcode labeling and reinforcing GCLP could reduce the risk of errors occurring. Third, the cost per test was twice that projected due to slow enrollment of study subjects and thus smaller specimen batches than anticipated. The version of OLA implemented in this study uses expensive streptavidin-coated 96-well plates, so it is better suited for laboratories with a higher volume of specimen testing than in the current study, in order to fully utilize the plate with each batch-test. The cost per test could be as low as $20 per test when making full use of each plate (36 samples/plate). However, to achieve this, our workflow would have required waiting for specimens to accumulate over several weeks, which would have unacceptably lengthened turn-around times.

Our experiences with problem-solving multiple issues related to OLA performance during the trial, such as degradation of extracted DNA, indicated that concepts central to molecular biology were new to our technicians. To competently manage performance of this laboratory-developed OLA, one needs familiarity with molecular theory and techniques, and in our study, supervision from an expert was essential. While remote supervision of the Kenyan technicians was effectively performed by our Seattle-based, highly educated and experienced laboratory manager, including weekly review of results and guidance for troubleshooting when assay issues were detected, this would not be practical for ongoing work. Others have reported a paucity of technicians with training and experience in molecular concepts and laboratory practices in other Sub-Saharan countries [19, 20]. In the absence of sufficiently trained technicians, laboratory tests could be simplified to better match current expertise.

Simplification of the OLA into the OLA_Simple kit is ongoing, with the current “DNA version” requiring a total of ten minutes of hands-on time, excluding isolation of CD4 cells from whole blood, which takes 40 minutes. Also, efforts are ongoing to simply and inexpensively extract RNA from whole blood for our “RNA version”. The current version reduces the number of steps and needed equipment through use of lyophilized reagents in single-use PCR and ligation tubes and detection in paper strips [21]. Multiple other bioengineering groups are also developing simplified assays to detect HIVDR [22]. In addition to reducing the technician’s effort, opportunities for errors, and turn-around time, the OLA_Simple kit addresses the currently unreliable supply chain of laboratory reagents in many RLS, by including all the reagents and plastic-ware needed to test a single specimen. In contrast to the equipment needed to set-up OLA in Nairobi (Table 1), OLA_Simple currently only requires micropipettes, a thermocycler and an inexpensive microcentrifuge. We anticipate that individuals with basic training as a clinical laboratory technologist can perform and retrain others to perform the OLA_Simple with instructions from a software program and on-line support.

This study has a few limitations. First, the generalizability of our study is limited by our implementation of the assay in only one laboratory. The study laboratory space and the education, experience and level of laboratory technician turnover may differ from that at other HIV treatment centers. Second, since OLA was performed within the context of a clinical trial, the Seattle group was motivated to successfully troubleshoot problems with the assay, which may not represent a real-world setting. However, this work builds on the Thai, Peruvian and Zimbabwean experiences that suggests that OLA is feasible in RLS [10, 16, 17]. Third, the applicability of our cost data is limited by our inability to access Kenyan price quotes for reagents and supplies, due to their very limited local availability. However, as we continue to simplify the test, we expect the OLA_Simple kit will have a reliable supply chain and require significantly less labor time and equipment. This should allow the cost of the test to be relatively low, ideally similar to other POC tests, approaching those of HIV self-tests and POC-CD4 tests [23, 24].

If HIVDR testing were widely implemented in RLS, some experts have advocated for a centralized laboratory approach, using CS, as a practical way to concentrate high levels of technical expertise and laboratory infrastructure and to increase economies-of-scale through high-volume testing [25]. In contrast, OLA can potentially be implemented using a decentralized laboratory approach, in small laboratories at or near the point of clinical care, and it offers a few advantages over CS for HIVDR detection in RLS. First, the equipment, laboratory infrastructure, and expertise required by OLA is relatively simple compared to what is needed to perform CS. While some African research groups have developed laboratory-derived CS genotypic HIVDR assays that are less costly compared to commercial assays [26, 27], these assays still depend on sophisticated sequencing equipment that requires maintenance by trained technicians and is rarely available in laboratories in RLS. Second, the costs of equipment and supplies needed for OLA are substantially less than those associated with CS, as the cost of a modern sequencer ranges from tens of thousands to hundreds of thousands of US dollars. Third, decentralized testing can reduce or eliminate the need for specimen transport and facilitate faster turn-around times. Importantly, although OLA was performed near the site of clinical care in this study, the turn-around time from sample collection to generation of results report (10–14 days) was only slightly shorter compared to CS at a centralized research laboratory in South Africa (median 18 days), due to the need for batch-testing to reduce the cost of OLA per test [28]. However, the OLA_Simple kit that is currently in development is designed for single-use, which eliminates need for batch-testing to reduce costs, and it has the potential to provide short turn-around times, ideally while the patient waits for results.

Importantly, point-of-care tests do not inherently address all challenges in testing for HIVDR. Potential barriers to point-of-care testing include shortages in human resources, inadequate space, and inadequate supply chain quality control of kits to ensure they are not mishandled or improperly performed in the clinic [29, 30]. Once HIVDR kits are developed, further research will be needed to optimize their integration into clinic workflow [29]. It is likely that neither centralized nor decentralized testing are “one-size-fits-all” strategies, and the ideal approach in any given setting will likely depend on multiple factors, including regional and local laboratory infrastructure and workforce capacity.

In the future, widespread adoption of generic dolutegravir (DTG)-based ART as 1st-line in Sub-Saharan Africa should reduce the need for PDR testing. However, OLA and other assays may be useful for PDR testing of pregnant women (for whom rapid suppression of viral replication is indicated), young children (for whom DTG is not available at this time) [31], and at the time of VF to differentiate between non-adherence and ADR. Further adaption of the OLA is needed for ADR testing, as the current version tests PBMC instead of plasma, and this provides suboptimal sensitivity for detection of HIVDR emerging at the time of VF. Finally, while DTG monotherapy can select HIVDR [32], the rate of DTG drug resistance when it is substituted for the NNRTI in patients with VF and drug resistance to NRTI is not known. This latter scenario has implications for the utility of PDR testing within the context of switching to DTG-based regimens in RLS.

In summary, OLA was successfully implemented in a Kenyan research laboratory to perform PDR testing for a randomized clinical trial, providing test results to clinicians within 10–14 days of sample collection. The challenges we encountered, including unreliable local supply chains for laboratory reagents and the need for an experienced molecular biologist to troubleshoot assay problems when they arose, suggest that ongoing work to develop an OLA_Simple kit is essential for reducing barriers for future use in RLS. The kit will need to be available within a price range that is cost-effective, and more work is needed to learn how to integrate OLA smoothly into clinical workflows while providing short result turn-around times. If HIV treatment guidelines for RLS recommend HIVDR testing in the future, OLA offers a simpler, lower-cost alternative to conventional consensus sequencing that can be performed at or near the point of care.