HIV Pre-Exposure Prophylaxis Adherence Test using Reverse Transcription Isothermal Amplification Inhibition Assay

Abstract

Current HIV antiretroviral therapy (ART) or pre-exposure prophylaxis (PrEP) therapy adherence monitoring relies on either patient self-reported adherence or monitored drug dispensing, which are not reliable. We report a proof-of-concept adherence monitoring assay which directly measures nucleotide reverse transcriptase inhibitor (NRTI) concentration using a reverse transcription isothermal amplification inhibition assay. We measure the concentration of Tenofovir diphosphate (TFV-DP)–an NRTI that functions as deoxyadenosine triphosphate (dATP) analog and long-term adherence marker for PrEP–by measuring the inhibition of the reverse transcription of an RNA template. The completion or inhibition of reverse transcription is evaluated by recombinase polymerase amplification (RPA), an isothermal nucleic acid amplification assay commonly used for point-of-care diagnostics. We present and validate a model that predicts the amplification probability as a function of dATP and TFV-DP concentrations, nucleotide insertion sites on the RNA template, and RNA template concentration. The model can be used to rationally design and optimize the assay to operate at clinically relevant TFV-DP concentrations. We provide statistical analysis that demonstrates how the assay may be used as a qualitative or semi-quantitative tool for measuring adherence to NRTI drugs and used to support patient compliance. Due to its simple instrumentation and short runtime (<1 hour), this assay has the potential for implementation in low-complexity laboratories or point-of-care settings, which may improve access to ART and PrEP adherence monitoring.

Introduction

The number of people living with HIV (PLWH) is approximately 38 million globally,¹ including 1.1 million in the US.² More than 25 million PLWH currently have access to anti-retroviral therapy (ART). To reduce new incidences, the WHO recommends pre-exposure prophylaxis (PrEP) to all people at risk of HIV infection.³ When taken daily, PrEP reduces the risk of contracting HIV by more than 90%.⁴ There are 770,000 people prescribed PrEP in the US, and 1.2 million additional people recommended for PrEP therapy because they are considered high risk for HIV.⁵ Despite increasing number of people receiving ART and PrEP, adherence to the therapies is still challenging for various environmental, social, economic, and behavioral reasons.^2,6–9 Given the strong correlation of adherence to ART and PrEP with treatment success,^8,10 continuous efforts and improvement along the patient care continuum are aimed at timely adherence monitoring and support.^11,12

Current methods for monitoring adherence rely primarily on patient self-reporting.¹¹ Self-reporting adherence is inexpensive to implement; however, it has many known limitations such as memory lapse and social desirability bias which undermine its value.^13–15 Many device-based adherence monitoring approaches such as digital pillbox,^16,17 Micro Electro Mechanical System-functionalized pills,¹⁸ text-message and electronic medication reminders^19–22 can be either too costly or do not provide proof of pill ingestion, thus hindering their adoption. Direct drug level quantification based on patient sample such as in blood or urine can provide objective, accurate, and unbiased measurement directly reflective of adherence.^13,14

Tenofovir disoproxil fumarate (TDF) is the prodrug in all PrEP combination regimens currently recommended by the WHO and approved by the US Food and Drug Administration (FDA), and a principal component of first line ART regimens.^23–27 Tenofovir (TFV) is in a class of HIV therapy called nucleotide reverse transcriptase inhibitors (NRTIs), which act as nucleotide analogs to block reverse transcription (RT) of the HIV virus so that they cannot replicate inside the human cell.^28,29 TFV has a short half-life (~15hrs) in plasma^30,31 and can be measured using liquid chromatography-mass spectroscopy (LC-MS).^31,32 LC-MS measurements require significant infrastructure, technical expertise, and investment, preventing its integration in most clinic settings.²⁸ There are also lateral flow point-of-care (POC) TFV competitive immunoassay in urine and blood.^35–38 The assays can accurately classify recent dosage up to a week³⁹ and have been used to identify poor adherence against LC-MS results.³⁸ Unfortunately, because TFV has a relatively short half-life, the immunoassay results are susceptible to the white coat effect and not an accurate biomarker for long-term PrEP or ART adherence.^37,40

Tenofovir-diphosphate (TFV-DP) is the active form of TFV that accumulates in red blood cells with an intracellular half-life up to 17 days.^31,41 As a result, TFV-DP level is a valuable indicator of long-term adherence and predictive of health outcomes.^32,42–44 The median concentration of TFV-DP in red blood cells, as measured by LC-MS, ranges from 15 to 170 fmol/10⁶ red blood cells (RBCs) from 1 dose per week to 7 doses per week.⁴⁵ With an average hematocrit of 40% and average RBC count of 5×10⁶ RBCs/μL for adults, 15 to 170 fmol/10⁶ red blood cells translates to a clinical range of 75 nM to 850 nM of TFV-DP in red blood cells.³¹ A study that correlated LC-MS TFV-DP concentrations from dry blood spots against HIV contraction outcomes showed that PrEP clients taking ≥4 doses/week were protected from HIV infection.^43,46 Four doses per week as measured by LC-MS was 700 fmol/punch in a dried blood spot or 292 nM in red blood cells and has been suggested as an adherence cutoff threshold for distinguishing adherence versus non-adherence.⁴⁶

Recently, our group reported on an enzymatic assay, termed REverSe TRanscrIptase Chain Termination (RESTRICT), that measures DNA duplex production by reverse transcriptase in the presence of NRTI drugs.^47,48 The assay accurately distinguished TFV-DP drug levels within the clinical range for adherence and has the potential to be a useful test to identify patients with poor adherence to ART and PrEP.⁴⁷ Real-time PCR has also been used to quantify inhibitory concentrations of NRTIs in drug development against HIV-1 RT.^49,50,51 These assays represent important advances in adherence monitoring, yet they require complex protocols and expensive instrumentation that limit their use to well-resourced central laboratories with highly trained technicians. There remains a clinical need for a detection assay for TFV-DP with simple user steps and limited instrumentation requirements to enable adherence monitoring in low-complexity laboratories or POC settings.

In this paper, we report on the early-stage development of a novel approach for measuring TFV-DP concentration using reverse transcription (RT) followed by recombinase polymerase amplification (RPA). RPA is an isothermal nucleic acid amplification method that is powered by a mix of recombinase and other accessory proteins, in contrast to PCR which relies on thermocycling to amplify DNA.⁵² RPA is a leading method for the rapid diagnosis of infectious diseases in POC settings due to its low operating temperature (~39 °C), excellent sensitivity and specificity, and short runtime (~20 minutes).^53–55 In this work, we measure the concentration of TFV-DP based on the inhibition of reverse transcription via chain termination of the reverse transcription of a template RNA by a TFV-DP. If the template is not transcribed by RT because the process has been inhibited by the presence of a NRTI, RPA cannot amplify the template. The assay is probabilistic and may require multiple runs on a single sample to determine the experimental probability. We present experimental results for measuring the TFV-DP concentration in the clinically relevant range for ART and PrEP. The assay can be completed in under an hour and requires a single controlled temperature of 39 °C for both RT and RPA steps. We present an experimentally validated, analytical model that predicts the amplification probability as a function of dATP concentration, RNA template number, and number of inhibitive insertion sites on the RNA target sequence. While the assay is in its preliminary stage of development, our results show that this probabilistic approach to RPA-based TFV-DP measurement has the potential to enable rapid and low-cost adherence testing of HIV medication in expanded clinical settings.

Methods

Our TFV-DP detection assay process requires three steps: reverse transcription of the RNA template into cDNA, deactivation of RT enzyme, and RPA, as shown in Figure 1B. The entire process takes approximately 35 minutes, with only one controlled heating temperature of 39 °C in addition to an enzyme deactivation step at any temperature above 70 °C.

Figure 1. Overview of TFV-DP measurement using Reverse Transcription Isothermal Amplification Inhibition.

(A) Molecular structural similarity between TFV-DP and dATP indicates that TFV-DP is a NRTI, blocking dATP binding to terminate HIV RT activities.^59,60 (B). Modulation of the assay by competitive binding between dATP and TFV-DP during cDNA synthesis. The level of TFV-DP is determined by amplifying the RNA template that is reverse transcribed by RT. The ratio, R, is the concentration of dATP divided by that of TFV-DP. (1) If there is little or no TFV-DP present, the RT reverse-transcribes the template RNA to form cDNA with a complete template frame for amplification, and the cDNA is amplified by RPA. (2) If there are moderate concentrations of TFV-DP compared to dATP, some of the cDNA chain may be terminated and thus only some full-length cDNA is formed. In this case only a fraction of amplification reactions will be positive. (3) When high concentrations of TFV-DP are present, nearly all cDNA is chain-terminated, resulting in RPA inhibition of all the samples.

Preparation of RNA template

The RNA template used in this study was extracted from HIV virus (EQUAPOL, Duke University Medical Center, virus name DE00110CN001.S1 from cell culture supernatant). The RNA was extracted by following standard procedures for solid phase extraction with a spin-column (QIAamp Viral RNA Mini Kit, Cat No./ID: 52904, Valencia, CA). All RNA extracts were stored at −20 °C until use. Quantification of the RNA was performed with RT-qPCR methods as previously published.⁵⁶ To avoid unintended amplification by pro-viral DNA contaminants, we added DNase I (AMPD1–1KT, Sigma-Aldrich) to degrade DNA. The RNA samples were incubated with DNase for 15 minutes at room temperature, followed by 10 minutes at 70 °C to deactivate DNase.

Reverse transcription in the presence of TFV-DP

The first step in the assay is reverse transcription of the RNA template into cDNA. cDNA synthesis was completed with final concentrations of 1X AffinityScript RT buffer (600107, Agilent Technologies, Santa Clara, CA), 5 mM DTT (600107, Agilent Technologies, Santa Clara, CA), 1 U/μL SUPERase RNase Inhibitor (AM2694, Invitrogen, Carlsbad, CA), 0.05 U/μL HIV Reverse Transcriptase (HIV RT P66/51, AbCam), 1.5 μM dNTP mix (R0192, Fisher Scientific, Houston, TX), and 0.42 μM HIV reverse primer (Integrated DNA Technologies, Coralville, IA, USA) that contains the complementary sequence of a target region within the HIV-1 pol gene.⁵³ All reagents were diluted in nuclease free water (P1193, Promega, Madison WI). 7.8 × 10³ copies of RNA template and TFV-DP ranging from 0 to 1250 nM (166403-66-3, BOC Sciences Inc.) were spiked into each reaction mix. The final reaction volume was 20 μL. The cDNA synthesis was performed at 39 °C in a heating block (T16, Axxin, Australia) for 10 minutes. After cDNA synthesis, RT enzyme was deactivated at 70 °C for 5 minutes in a mini dry block heater (VWR, Radnor, PA) to remove reverse transcription activity.

RPA assay for cDNA products

After cDNA synthesis in the presence of TFV-DP, we amplified using RPA (TwistAmp® exo, TAEXO02KIT, Cambridge, United Kingdom) in order to detect the cDNA products that may or may not be present in the sample.⁵⁷ Rehydration buffer from the kit was added to the lyophilized RPA reagent, reverse primer, forward primer, and probe⁵³ were added for final respective concentrations of 0.54 μM, 0.54 μM, and 0.12 μM. 0.1% BSA (AM2616, UltraPure, Fisher Scientific, Houston, TX) was added to enhance amplification yield. Nuclease free water (P1193, Promega, Madison WI) was added for a final volume of 37.5 μL per reaction.

After RT deactivation, cDNA products were stored on ice. 10 μL from each reaction was added to the 37.5 μL RPA mix, and 2.5 μL of magnesium acetate (TwistAmp® exo, TAEXO02KIT, Cambridge, United Kingdom) was added last to activate the RPA reactions with a total reaction volume of 50 μL. Pre-incubation was performed at 39 °C for 5 min, followed by a brief manual mixing. The mixed samples were incubated at 39 °C for an additional 20 minutes. Real-time detection of RPA via 6-carboxyfluorescein (FAM) fluorescence was measured (T16, Axxin, Australia) every 20 seconds.

All fluorescence data were normalized between 0 and 1 using the fluorescence value (A.U.) at 6 minutes (after mixing was performed) and the fluorescence value at 17.5 minutes from the positive control (no TFV-DP added). Any increase in normalized fluorescence (ΔAU) greater than 10 AU were considered positive. To conserve reagents, we used different numbers of replicates, N, based on the predicted probabilities and statistical requirements to obtain appropriate confidence levels. For example, we used N=5 for positive controls with no TFV-DP present in reverse transcription because we expected all RPA replicates to amplify. When testing a TFV-DP concentration of 300 nM, we expected inhibited reverse transcription to result in probabilistic amplification events, so we used a higher number of replicates, N=10. The number of RPA reactions that did amplify versus did not amplify were tallied across the replicates to calculate the amplification probability.

Results and Discussion

Theoretical Model

The assay works by measuring the experimental probability that RNA templates are reverse transcribed by RT into cDNA that is subsequently amplified by RPA as shown in Figure 1. When no TFV-DP are present, we expect the RNA template to be reverse transcribed into complete cDNA. cDNA can then be amplified by RPA to provide a positive result. When TFV-DP molecules are present, they are incorporated as dATP analogs during cDNA synthesis, which terminates polymerase extension and results in partial length cDNA that cannot be exponentially amplified in RPA. If all, or nearly all, cDNA strands are chain terminated, then there is insufficient cDNA template for detectable RPA amplification. For intermediate concentrations, some samples have sufficient cDNA synthesized for amplification, while others do not, resulting in a probabilistic response.

Here we present an analytical model to predict the amplification probability as a function of the TFV-DP concentration, dATP concentration, RNA template copy number, and number of potential sites for dATP incorporation on the RNA target sequence. This model assumes there is sufficient RT enzyme, time, and temperature needed for reverse transcription to occur to completion and that sufficient dATP or TFV-DP molecules are present such that their concentrations are constant over the course of the reaction. The model also assumes all cDNA synthesized from the RT reaction are amplified, therefore, the probability of amplification is equal to the probability of cDNA synthesis. We define the variable, R, to be the ratio of the concentration of dATP, [dATP], and concentration of TFV-DP, [TFV-DP]. The probability of reverse transcription and amplification is a function of R, the number of template molecules in the reaction mix, t, and number of potential TFV-DP insertion sites on the RNA template, s. Insertion sites are equal to the number of uracil bases within the RNA sequence targeted for amplification that fall downstream of the reverse primer, at which RT may incorporate either a dATP or TFV-DP molecule. The relationship between R, s, t, and the corresponding RPA readout in the analytical model is illustrated in Figure 1B.

For each potential insertion site on a template RNA molecule, the probability that dATP is incorporated by the RT is,

$$P_{\mathit{\text{site}}}=\frac{[\mathit{\text{dATP}}]}{[\mathit{\text{dATP}}]+K_{aff}\cdot [TFV-DP]}=\frac{\frac{[\mathit{\text{dATP}}]}{[TFV-DP]}}{\frac{[\mathit{\text{dATP}}]}{[TFV-DP]}+K_{aff}\cdot \frac{[TFV-DP]}{[TFV-DP]}}=\frac{R}{R+K_{aff}},$$ (1)

where K_aff is the relative affinity of RT for TFV-DP compared to its native dATP substrate and is determined empirically by fitting K_aff to minimize the difference between the experimental and theoretical data. Assuming each insertion event at the complementary uracil on the RNA template is independent and that there are a total number of s insertion sites, the probability for at least one TFV-DP molecule to bind to a single template RNA within the target sequence is given as,

$$P_{\mathit{\text{template}}}=1-{\left(P_{\mathit{\text{site}}}\right)}^s=1-{\left(\frac{R}{R+K_{aff}}\right)}^s$$ (2)

The synthesis of cDNA is halted via chain termination when a TFV-DP is inserted by RT during polymerization, resulting in a partial-length complementary strand that will not be amplified in the subsequent RPA reaction. At least several, if not tens of copies, of cDNA must be synthesized during the RT step in order to amplify and obtain an RPA signal. If there are t copies of the RNA template, and all of them need to be inhibited for RPA to be negative, the probability for amplification can be written as,

$$P_{amp}=1-{\left[P_{\mathit{\text{template}}}\right]}^t=1-[{\left(1-{\left(\frac{R}{R+K_{aff}}\right)}^s\right]}^t.$$ (3)

This simple analytical model was adopted for the inhibition of RT activity by TFV-DP in competition with dATP.

Experimental Results

The assay measures the probability that TFV-DP is inhibiting reverse transcription and subsequent isothermal amplification. We calibrated the assay to the appropriate clinical range of TFV-DP in ART and PrEP patients with the goal of using the assay to measure patient adherence to the drug treatment. The transition between no inhibition (low TFV-DP) and complete inhibition (high TFV-DP) occurs within a relatively narrow range of clinically relevant TFV-DP concentrations (between 75 nM and 850 nM).³¹

First, we examined whether the RNA template incubated with RT and varying concentrations of TFV-DP will synthesize full-length cDNA and enable subsequent amplification by RPA. Figure 2 shows normalized RPA fluorescence readout as a function of time for 0, 100, 300, 600, and 1,250 nM TFV-DP spiked in the RT mix during the reverse transcription step. When TFV-DP content was low (0 and 100 nM), 6 out of 6 samples amplified, indicating that the RNA templates were successfully reverse transcribed into full-length cDNA, as shown in Figure 2A. In Figure 2C, we show that for moderate TFV-DP levels (300 nM) only 5 out of 10 of the samples amplified, suggesting half of the reactions were chain-terminated from TFV-DP incorporated during cDNA creation. At higher concentrations of TFV-DP (600 and 1250 nM), only 1 out of 14 and 0 out of 5 amplified, respectively, indicating that nearly all RNA templates were chain-terminated by TFV-DP during the RT step, as shown in Figures 2D and 2E. With this work being in the assay development stage, we differ the number of replicates with TFV-DP concentration to conserve reagents, but future work with clinical samples of unknown TFV-DP concentrations will require a fixed number of replicates for each sample analyzed.

Figure 2.

Real-time RPA fluorescence as a result of A) 0, B) 100 nM, C) 300 nM, D) 600 nM, and E) 1,250 nM TFV-DP spiked into the RT reactions with 1.5 μM of dNTP and 7.8×10³ copies of RNA template (n=5 for TFV-DP at 0, 100, and 1,250 nM, n = 10 for TFV-DP at 300 nM, and n = 14 for TFV-DP at 600 nM). All fluorescence data were normalized between a baseline measurement at 360 seconds (after mixing was performed) as 0, and maximum fluorescence at 1,050 seconds from the positive control (no TFV-DP added) as 1.

The experimental probability of amplification is plotted against the model in Figure 3 as a function of the TFV-DP concentration. Here, both the experiments and model have 7.8×10³ copies of RNA template with 21 uracil sites and 1.5 μM dATP. We fit K_aff = 0.7, which is similar to the affinity values reported previously.^47,58 The experimental data and model show that the probability of amplification is unity when no TFV-DP is present, decreases rapidly with incremental addition of TFV-DP, and remains at zero when the TFV-DP concentration is high. Here the assay has been tuned to demonstrate sensitivity between 300 nM and 500 nM of TFV-DP, within the shaded area between 75 nM and 850 nM, which is the clinically relevant TFV-DP concentration range for patients taking ART and PrEP.^31,41

Figure 3.

Analytical model (solid line) and experiments (circles) of the probability of amplification as a function of [TFV-DP] at t = 7.8×10³ copies of RNA template and s = 21 uracil sites. Shaded area indicates the clinically relevant TFV-DP concentration range between 75 nM (1 dose per week) and 850 nM (7 dose per week).

We leverage our analytical model to show the sensitive range of the assay may be further shifted for different TFV-DP concentrations by varying the input RNA template copy number or the number of potential TFV-DP insertion sites. In Figure 4, analytical model outputs of the probability of amplification are plotted against R. P_amp shifts towards higher NRTI concentrations (left) as RNA template copy number, t, increases or the number of insertion sites, s, decreases, respectively. With increasing amount of RNA template, t, the reaction is more likely to amplify because there are more chances to complete the reverse transcription, thus requiring more TFV-DP to inhibit the RT step. When there are increasing number of uracil sites, s, on the template, it is more likely for a TFV-DP molecule to be incorporated at one of the sites, thus terminating strand extension and inhibiting subsequent amplification. Each of the probability curves has a 50% inhibition concentration (IC₅₀) value where 50% of the total reactions are inhibited. When s increases or t decreases, the probability curves shift right towards IC₅₀ values at lower NRTI concentrations and the slope of the curve reduces, extending over a wider range of concentrations. Having a lower slope may be advantageous when the goal is to quantitatively measure drug levels. If s decreases or t increases, the probability curves shift left towards higher IC₅₀ values (in terms of NRTI concentrations), and the curve becomes steeper. Steep probability curves are advantageous for qualitative tests where one wants to distinguish concentrations that are either above or below a certain threshold.

Figure 4.

Probability of amplification from the analytical model for various s and t, and experimental data for s = 21 and t = 7.8×10³ as a function of R, the ratio between dATP and TFV-DP concentration. The analytical model shows the transition zone of the assay and the detection range can be shifted by varying the components in the RT reaction mix, number of sites (s), and number of templates copies (t). All plotted lines (dashed and solid) are outputs from the analytical model for the corresponding experimental conditions of s and t. The circles are experimental data from the four conditions of [TFV-DP] being 0.1 μM (R = 3.75), 0.3 μM (R = 1.25), 0.6 μM (R = 0.625), and 1.25 μM (R = 0.3), using a fixed [dATP] of 0.375μM.

TFV-DP concentration can be deduced quantitatively from outcomes of repeated trials with the same experimental conditions using binomial probability and respective confidence intervals. Each reaction has a binomial outcome with an expected probability of amplification that can be derived from the analytical model. Multiple reactions can be performed to sample the expected probability. Given total number of trials of the same reaction on one sample, N, and the number of amplified (positive) reactions, n, the positive outcome over total number of repeated trials C is defined as C = n/N. This is a binomial random variable with an expected mean of P_amp and confidence interval (C.I.) as estimated using the Clopper-Pearson method. Therefore, for 95% of distinct measurements of a sample, C falls within the range P_amp ± C.I.. In Figure 5, the estimated [TFV-DP] is plotted against C for a specific N (5 and 20 total reactions) [dATP] = 0.375 μM, s = 21 uracil sites per template, and t = 7.8×10³ copies of RNA template. For N=5, we can see that the C.I. is large, and it is difficult to measure the precise concentration of TFV-DP. This is especially true at C=1 or C=0 where the C.I. is asymmetric. For example, at C=0 the lower bound is approximately 300 nM and the upper limit is unbounded. When we increase the number of trials, shown in Figure 5b for N=20, the greater the number of repeated reactions narrows the C.I., indicating increasing confidence in the predicted TFV-DP concentrations. In addition, the predicted TFV-DP concentration has the tightest C.I. for intermediate values of C, and grows wider when C approaches the minimum or maximum values of 0 (all repeated trials inhibited) or 1 (all repeated trials amplified), representing greater quantitative sensitivity obtained towards the intermediate C.

Figure 5.

TFV-DP concentration as a function of the experimental positive trial proportion, C. Figure 5A and 5B represent the calculated TFV-DP for a pre-determined 5 and 20 repeated reactions, and a selected set of conditions of t = 7.8×10³ copies of RNA template, s = 21 uracil sites, and [dATP] = 0.375 μM. The line with the symbols represents the expected [TFV-DP] given the experimental positive trial proportion. The shaded area represents the 95% confidence interval for each experimental outcome.

Alternatively, the assay could be used for qualitative measurement of TFV-DP concentration based on the outcome of a limited number of repeated trials, a one-tailed hypothesis, and p-values from hypothesis testing. Using a standard binomial distribution, with each n, N, and P_amp, given mean μ = nP_amp, and variance σ = nP_amp(1−P_amp), the p-value can be calculated based on the cumulative distribution function.

Table 1 shows a look up table of p-values for the one-tailed hypothesis test that [TFV-DP] > 292 nM, as a function of the total number of trials, N, and the number of positive amplification reactions, n. This hypothesis test is clinically relevant because direct observation clinical studies show that TFV-DP greater than 292 nM, or four doses per week, is a good measure of adherence and represents protection from contracting HIV.⁴⁶ If a patient’s sample fails this hypothesis test, it suggests that the patient does not have sufficient TFV-DP concentration, they may have significant risk of contracting HIV, and should consider adherence counseling and support to improve adherence.

Table 1.

Look up table of p-values for the hypothesis test that [TFV-DP] > 292 nM, as a function of the total number of trials, N, and the number of positive amplification reactions, n. Direct observation clinical studies show that patient samples with TFV-DP greater than 292 nM, or four doses per week, are adherent and protected from contracting HIV.^43,46 Bolded cells in the table represent p < 0.05 or greater than 95% confidence that the hypothesis is true.

	N = 1	2	3	4	5	6	7	8	9	10
n=0	0.43	0.19	0.08	0.03	0.02	0.01	0.00	0.00	0.00	0.00
1	1.00	0.68	0.40	0.22	0.11	0.06	0.03	0.01	0.01	0.00
2		1.00	0.82	0.58	0.37	0.23	0.13	0.07	0.04	0.02
3			1.00	0.90	0.72	0.52	0.35	0.23	0.14	0.08
4				1.00	0.94	0.81	0.65	0.48	0.34	0.22
5					1.00	0.97	0.88	0.75	0.60	0.45
6						1.00	0.98	0.92	0.82	0.69
7							1.00	0.99	0.95	0.88
8								1.00	0.99	0.97
9									1.00	1.00
10										1.00

Conclusions

In this paper, we describe an assay for directly measuring TFV-DP concentration using a reverse transcription and RPA assay for objective adherence monitoring of ART and PrEP. This is the first time that an isothermal amplification has been applied for adherence testing of HIV medications. RPA and other isothermal amplification techniques are increasingly used for POC diagnostics because they are sensitive, rapid, and have simple instrumentation requirements, making them suitable for deployment in the field or clinic-based settings. Our assay for TFV-DP measurement runs in less than one hour and uses a benchtop fluorometer with integrated heat block that is specifically designed to be a low-cost, field deployable instrument for POC molecular diagnostics (Axxin T16 instrument). Other common methods for NRTI measurement have long protocols and require an LC-MS instrument, real-time PCR machine, or microplate reader, which restricts their use to high-complexity laboratories.^31,47,50 Our findings in this report show that isothermal nucleic acid amplification has the potential to reduce the cost and complexity of HIV medication adherence testing, enabling expanded access in clinical settings.

There are several potential directions for future work to improve the performance and clinical utility of this method. We expect that optimization of the experimental conditions, RT incubation length, and reaction temperature may lead to shorter assay times and lower RT enzyme concentrations, thereby reducing assay costs. The sensitivity of the assay is dependent on the number of uracil sites on the RNA template during reverse transcription, and this number is readily increased by using a different primer to initiate reverse transcription at a region upstream of the designated amplification region of the template. Detection of significantly lower concentrations of TFV-DP may enable our assay to directly process diluted whole blood samples, simplifying sample preparation. Our assay could also be applied to measure other NRTIs (e.g. emtricitabine) to provide additional information on adherence for patients receiving combination therapies. One limitation of our work is that RNA templates are relatively unstable and should be handled with care as repeated freeze-thaw cycling can result in degradation of viable RNA. We also do not work with clinical samples and do not address the sample preparation challenges that may be needed for detecting TFV-DP from whole blood or dried blood spots. We hypothesize that a simple dilution step and direct addition of crude sample into the assay may be possible due to the known resilience of RPA to inhibitors present in blood.⁵⁴ To ultimately implement this assay for point-of-care HIV adherence testing in clinical settings, extensive experimentation will be needed to characterize the inherent variability of the assay and its repeatability across diverse clinical samples.