Accuracy of real-time PCR and digital PCR for the monitoring of total HIV DNA under prolonged antiretroviral therapy

Total HIV DNA is a standard marker to monitor the HIV reservoir in people living with HIV. We investigated HIV DNA quantification accuracy by a real-time PCR kit (qPCR) and digital PCR (dPCR) method within the same set of primers and probes. Among 48 aviremic patients followed for up to 7 years with qPCR, the mean coefficient of variation of total HIV DNA between two successive measurements was 77% (± 0.42log10 HIVDNA copies/106 PBMC). The total HIV DNA quantified by the two PCR methods has a high correlation (0.99 and 0.83, for 8E5 and PLHIV samples, respectively), but we observed better repeatability and reproducibility of the dPCR compared to the qPCR (CV of 11.9% vs. 24.7% for qPCR, p-value = 0.024). Furthermore, we highlighted a decay of the number of HIV copies in the 8E5 cell line qPCR standard over time (from 0.73 to 0.43 copies per cell), contributing to variations of HIV DNA results in patients whose HIV reservoir should be theoretically stabilized. Our study highlighted that absolute quantification of total HIV DNA by dPCR allows more accurate monitoring of the HIV reservoir than qPCR in patients under prolonged antiretroviral therapy.

Human immunodeficiency virus (HIV) provirus persists in the genome of latently infected cells despite suppression of viral replication by current antiretroviral therapies 1 . HIV reservoir is therefore considered a major obstacle to HIV cure. The quantification of total HIV DNA in whole blood or peripheral blood mononuclear cells (PBMC) by quantitative polymerase chain reaction (qPCR) is the standard method to estimate the HIV reservoir in people living with HIV (PLHIV) 2,3 . HIV DNA can be quantified in routine practice following treatment initiation and PLHIV under prolonged antiretroviral therapy 3 . The total HIV DNA quantification has been envisaged in French guidelines as a decision support tool for reducing the number of active ARVs 4 . Total HIV DNA includes integrated genomes that can be functional and non-functional forms (1 and 2 LTR-circles that are episomal genomes representing abortive integration events and linear forms), forming the total HIV DNA reservoir 3 . Studies have shown that the total HIV DNA is correlated with cells containing a replicationcompetent virus 3,5 . A rapid decay of HIV DNA is observed during the first year after the start of antiretroviral therapy 6 . Afterward, HIV DNA decay is slower or absent 2,6,7 . In people who achieve prolonged HIV suppression under antiretroviral treatment, the estimate of this marker of the viral reservoir is between less than 100 to 3000 HIV DNA copies per million PBMC 2,8 . Because of this relatively small concentration range, the quantification of HIV DNA requires a highly reliable method.
qPCR is based on the detection of target DNA relative to a standard and on the measurement of a fluorescence signal over time. qPCR is a sensitive method to detect low concentrations of virus nucleic acids 9 . However, instability of the standard curve, changes in amplification efficiencies, and measurement during an exponential amplification process contribute to limit the accuracy of DNA/RNA quantification using the qPCR method 10,11 . The reproducibility of qPCR for HIV DNA quantification may be insufficient for therapeutic purposes, especially for low HIV DNA values that exhibit the highest variations 12 . Digital PCR (dPCR) is a promising alternative to qPCR for HIV DNA quantitation [13][14][15] . dPCR is based on a limiting dilution, splitting the target DNA into multiple partitions. This approach eliminates the need for a standard curve and allows absolute quantification of the target. The amplification in each partition is detected at the end of the PCR, and the DNA concentration www.nature.com/scientificreports/ is calculated from the number of positive partitions. As an endpoint, the PCR method measures the nucleic acid concentrations through the ratio between positive and negative partitions. However, dPCR would be theoretically more reproducible 15 . The greater analytical precision of dPCR compared to qPCR makes theoretically possible the detection smaller variations in HIV DNA concentrations. Based on this, many authors recommend using dPCR rather than qPCR for quantifying the HIV reservoir 14,16 . Therefore, the limits of qPCR for HIV DNA quantification in clinical practice and the potential benefit of dPCR over qPCR for improved monitoring of HIV DNA needs to be evaluated. In this study, (i) we analyzed variations of HIV DNA levels using qPCR during the long-term follow-up of outpatients with durable viral suppression; (ii) we developed a dPCR using a chamber-based system, dedicated to HIV DNA quantification according to the recommendations of the dMIQE Group 17 ; (iii) we compared the two PCR methods to highlight the highest precision of dPCR for HIV DNA quantification; (iv) we showed that variations in HIV DNA content of the 8E5 batches of standard contribute to the low reproducibility of qPCR.

Results
Clinical monitoring of total HIV DNA using qPCR. Total HIV DNA was quantified in 48 subjects with prolonged virological success and 3 subjects initiating antiretroviral therapy using the commercial HIV DNA kit. Sequential measurements of HIV DNA were performed using 215 clinical samples collected between 2014 and 2021, with two to seven samples per subject. The mean coefficient of variation (CV) of HIV DNA concentration between all measurements of HIV DNA was 77%, corresponding to a mean variation of ± 0.42 log HIV-DNA copies/10 6 PBMC between two successive measurements (Fig. 1a). A decrease in HIV DNA was observed for all three patients initiating antiretroviral therapy (Fig. 1b). In these subjects, the mean CV was 82%, corresponding to a decrease of 0.68 HIV DNA copies/10 6 PBMC, one to five years after treatment initiation.

Development of HIV DNA dPCR.
The following criteria were used to select the optimal PCR conditions: the ratio of the mean fluorescence amplitude of HIV positive cell-line (8E5) to the mean fluorescence amplitude of PBMCs from uninfected controls, the visual separation of the positive and negative clouds allowing easy thresholding, and the number of false positives partitions present in PBMCs from uninfected controls. Different primers and probe concentrations, temperatures and time of hybridization and elongation, and numbers of cycles were tested to optimize the dPCR (Supplemental Fig. S1). An appropriate time of hybridization and elongation, and an appropriate number of cycles, were crucial to obtaining optimal discrepancy between positive and negative partitions. Different primer and probe concentrations were tested, ranging from concentration of 200 to 666 nM. For hybridization and elongation, we have chosen for optimal conditions a temperature of 57 °C for one minute, with 35 cycles where a concentration of 400 nM HIV DNA primers and 466 nM B-globin probe were retained in the PCR mix. The specificity was tested on ten human genomic DNA samples and ten clinical samples positive for different pathogens: hepatitis B virus, leptospirosis, and syphilis. No samples tested positive for HIV DNA (100% specificity, data not shown). dPCR analytical sensitivity. We performed serial measurements in slightly varying concentrations to determine the limit of detection (LOD) and linearity of the HIV DNA dPCR. The 95% LOD was determined by testing 14 concentrations ranging from 12 to 1250 copies/10 6 PBMC in multiple replicates ( Table 1). The LOD95% result was estimated at the concentration of 75 copies/10 6 PBMC, i.e. 0.6 copies/µL PCR, or 9 copies/ replicate. The LOD95% determined by probit analysis was 74 copies/10 6 PBMC (Fig. 2a). This concentration of HIV DNA was detected 19 times out of 20 replicates performed. The limit of blank was tested on 14 blank samples and estimated at 0.2 copies/µL PCR (data not shown). The established limit of quantification (LOQ) was 125 copies/10 6 PBMC for this method with an accuracy of 98.4%, compared to the lower concentrations with 94.6% for 112 copies and 92.0% for 100 copies (Fig. 2b).
To evaluate the linearity of the dPCR, we assessed a range of HIV DNA concentrations that included more than 75% of the theoretical values of the HIV reservoir in subjects with prolonged virological success 2 , i.e., between 100 to 1250 copies/10 6 PBMC (Fig. 2c). The dPCR exhibited high linearity with an coefficient of determination (R 2 ) close to 1 (R 2 = 0.937).
We evaluated the analytical sensitivity of the dPCR for the β-globin target to determine if a low number of cells could be accurately detected and quantified by this technique. A range of 15 samples with a volume of 100 µL and concentrations ranging from 100 to 6000 cells per sample were measured in triplicates to test the LOD95% and LOQ of the β-globin dPCR (Fig. 2d). All concentrations tested were detected at 100%, so we did not determine a LOD95%, the lowest detected concentration being 100 cells/sample.The linearity experiment was then conducted on these same concentrations, demonstrating good linearity with an R 2 very close to 1 (R 2 = 0.996) for the β-globin target of digital PCR (Fig. 2d). dPCR technique is more accurate than qPCR. At a concentration of 100 HIV DNA copies/10 6 PBMC, the CV for dPCR was 21.29% for both the repeatability and reproducibility (Table 2a). By comparison, qPCR had a CV of 31.16% for repeatability and 36.79% for reproducibility. At 1000 copies/10 6 PBMC, dPCR had a CV of 7.96% for repeatability and 11.92% for reproducibility (Table 2b), while qPCR showed a CV of 8.96% and 24.74% for repeatability and reproducibility, respectively. A non-parametric test (Siegel-Tukey) showed that the reproducibility at 1000 copies/10 6 PBMC was significantly higher for the dPCR than qPCR (p-value = 0.024). Therefore, the dPCR had a more accurate inter-experimental precision than the qPCR (Fig. 3).
Evaluation of HIV DNA content of the 8E5 cell-line batches used as a standard by the qPCR kit. The 8E5 cell line standard from several batches of the qPCR kit was quantified using the dPCR to assess www.nature.com/scientificreports/ possible variations in HIV-1 DNA concentrations. Theoretically, the standard has a concentration of 1 copy of HIV DNA per cell or 3 × 10 6 copies/mL 18 . Absolute quantification with dPCR showed lower concentrations in all the lots tested, ranging from 0.43 to 0.73 copies per cell (1.3 × 10 6 to 2.2 × 10 6 copies/mL). A progressive decrease in cellular HIV DNA was observed between successive lots over time (Fig. 4). To confirm this difference, we quantified two 8E5 cell-lines stored in our laboratory (8E5A and 8E5B) and the 007 Lot standard on two additional dPCR thermocyclers, Bio-Rad and Stilla, in addition to the ThermoFisher thermocycler (Fig. 4). The laboratory cell line 8E5A was quantified at 1 copy of HIV DNA per cell by qPCR, but at 0.23, 0.32, and 0.33 copies per cell by dPCR performed on all three machines, respectively. The 8E5B cell-line was quantified at 0.13 copies per cell by the two dPCR thermocyclers, Bio-Rad and Stilla. The 007 Lot standard was quantified at 0.40, 0.41, and 0.52 copies per cell by dPCR. As a comparison, the ThermoFisher thermocycler quantified the chronically infected promonocytic cell line U1 at 1.17 copies per cell ( Fig. 4) 19,20 . Therefore, the results obtained with several dPCR thermocyclers allowed us to confirm the deviation of the HIV DNA cell concentration of the standard used in qPCR from the theoretical value used as a reference.
Comparison of total HIV DNA levels using dPCR and qPCR. Bland-Altman analysis was performed using 12 dilutions of the 8E5 cell line in HIV-negative PBMC (Fig. 5a) and 52 individual patient samples (Fig. 5b). The dPCR tended to have lower quantification than qPCR, with a median difference between the two techniques of -0.32 log 10 HIV DNA copies/10 6 PBMC for the 8E5 cell line (p-value = 0.0005) and -0.35 log 10 HIV DNA cop-

Discussion
Total HIV DNA is considered a clinically relevant marker for long-term monitoring of HIV reservoir in patients under antiretroviral therapy 3 . Quantification of total HIV DNA in PBMC or whole blood does not permit analyzing the HIV reservoir in tissue compartments, nor distinguishes between replication-competent and defective HIV genomes. However, this marker only requires a simple venous blood sample and a readily available PCR test 3 . Accuracy and sensitivity are required to analyze HIV DNA levels in the narrow range of concentrations observed in patients under prolonged antiretroviral therapy. During the follow-up of patients under prolonged antiretroviral therapy in whom the HIV reservoir was theoretically stabilized, we observed an important variation of total HIV DNA using qPCR. The accuracy of the qPCR technique must be evaluated when taking into account the mean HIV DNA level, generally below 3 log 10 copies HIV DNA/10 6 PBMC. Therefore, a variation of ± 0.4 log 10 limits the usefulness of the HIV DNA qPCR for therapeutic monitoring. The reproducibility of total HIV DNA quantification using this qPCR assay has been evaluated in previous studies. The reproducibility of HIV DNA by qPCR is better than viral outgrowth assays; however, qPCR is not precise enough to properly quantify HIV DNA [21][22][23] . Based on 69 testing in two patients followed for nine months, the standard deviation of HIV DNA with Biocentric PCR was 0.11 and 0.18 log 10 with 95 CI intervals of ± 0.22 and 0.35 log 10 18 . In this study, the qPCR of both calibrated curves and clinical samples were carried out in duplicate. Reproducibility across laboratories of this qPCR reported a 0.2 log 10 standard deviation in 10 laboratories using three dilutions of HIV cell line (2.0, 2.8, and 3.4 log 10 HIV DNA copies/10 6 PBMC) 24 . This reproducibility of qPCR may be sufficient to monitor the rapid decay of total HIV DNA in patients initiating antiretroviral therapy, but a better accuracy is required to compare HIV DNA in CD4 subpopulations, or in different tissue compartments 25-27 . Our results suggest that the instability of HIV DNA content in 8E5 cell line may significantly contribute to the inaccuracy and intra-individual variations observed in patients. We observed a progressive decay of HIV copy per cells over time as successive lots were used. Previous studies have reported that the 8E5 cell line, which theoretically has 1 copy of HIV DNA per cell, can 'lose' the provirus during multiple cell passages 28,29 . To date, the deletion or mutation mechanisms probably involved in this pheromone remain unknown 28,30 . Based on the result of HIV DNA content measured in the 007 8E5 batch (0.48 HIV DNA copies/cell), the standard would be the source of a 0.3 log 10 difference in quantification between qPCR and dPCR. This gap in HIV DNA corresponded to the difference observed between the two PCR methods. Variations in HIV DNA content of the standard compromised the accuracy of results recorded during the patients over several years. The U1 cell line, containing on average two proviruses per cell, has been proposed as a possible alternative to 8E5 cells 30,31 , but using dPCR, we also observed a lower HIV DNA copy per U1 cell than expected. Furthermore, we observed that depending on the HIV DNA concentration range of a sample, the difference in quantification between qPCR and dPCR changed. In the samples with high HIV DNA concentrations, qPCR tends to overestimate the concentration while dPCR underestimates it. Technical variations due to the equipment, and induced by users, are responsible for uncertainty when using quantitative PCR 32 .
The dPCR developed in this study with the same set of primers and probes that qPCR offers significant advantages when compared to qPCR. First, dPCR enables HIV DNA quantification in absolute concentration; secondly, we observed a higher reproducibility using dPCR than dPCR; finally, the dPCR includes an internal cell www.nature.com/scientificreports/ control targeting the β-globin gene to quantify the number of cells tested. Because of its characteristics, dPCR is a more promising method than qPCR for the analysis of the HIV reservoir. The LOD of the HIV DNA dPCR based on the same HIV genomic target (LTR) and the same set of primers and probe is slightly higher than the qPCR LOD (i.e., 75 vs. 40 copies/10 6 PBMC, respectively). Considering the difference in quantification in favor of qPCR (+ 0.3 log 10 ), the LOD value for dPCR is theoretically around 150 copies of HIV DNA/10 6 PBMC in qPCR. After correction, this value remains below the threshold of 2.39 log 10 (200 copies/10 6 PBMC) established with the qPCR kit and indicates a low HIV reservoir and an optimal therapeutic response 2 . Hence, the sensitivity of dPCR may be sufficient to quantify the HIV reservoir in most patients under prolonged antiretroviral therapy 2,8 .
The dPCR may also be more appropriate to quantify HIV DNA in a limited number of cells because this method uses a small volume of eluate after extraction (4 µL) compared to qPCR (25 µL). This low volume facilitates multiplicate assays. The use of quantification replicates allows the results to be reported as a median, eliminating outliers through statistical analysis and improves the accuracy of the measurement. For example, on a 100 µL eluate after extraction of DNA from cells, dPCR allows us to experiment 25 times for only 4 times in qPCR. It is possible to collect and extract more significant volumes of blood to multiply the qPCR tests to improve sensitivity and accuracy, but this approach has limits in the case of the HIV reservoir explored in CD4 subpopulations and rare cells. Our results are in line with previous studies using dPCR for HIV DNA quantification [33][34][35] . dPCR techniques based on multiple targets could also further increase the accuracy of HIV DNA quantification [36][37][38] . Although dPCR assays allow the study of the HIV reservoir with more accuracy, standardized protocols are  (c), the x-axis shows the theoretical concentration, whereas the y-axis shows the concentration measured by dPCR in copies/10 6 PBMC, which allows the estimation of the LOD95% and LOQ for (b). The x-axis of (a-d) and the y-axis of (b-d) are displayed with a log scale. www.nature.com/scientificreports/ needed, and manufacturers should pursue regulatory approval for monitoring of HIV infected patients using dPCR kits 39,40 .
In conclusion, we observed critical intra-individual variations of HIV DNA when using the qPCR in aviremic subjects. We showed that both the poor reproducibility of qPCR and changes over time in the HIV-1 DNA concentration in the 8E5 standard fuel these variations. HIV DNA dPCR technique allows independence from two pitfalls of qPCR, namely real-time measurement from exponential amplification and quantification relative to a standard. The dPCR offers better performance to measure HIV DNA because of its reproducibility and absolute quantification of HIV DNA. Table 2. Precision experiments to compare the dPCR and qPCR methods with the coefficient of variation (CV%). For the two concentrations, (a) 100 copies/10 6 PBMC and (b) 1000 copies/10 6 PBMC, the tables show the results of the repeatability and reproducibility studies in CV% in normal and log-transformed data calculated with the formula 1 = σ/µ and the CV% of log-transformed data calculated with the formula 2 = e σ 2 − 1 . Formula 1 is commonly used to calculate the CV of untransformed data, and formula 2 the CV of log-transformed data. The CV calculation with formula 1 is often used in the literature. However, with a serious statistical error, Formula 1 is not correct for calculating the coefficient of variation of logtransformed data because the mean μ depends on the unit.  Figure 3. Precision experiments to compare the dPCR and qPCR methods with the coefficient of variation (CV%). The data of the repeatability and reproducibility studies for the two concentrations (100 and 1000 copies/10 6 PBMC) are represented in log-transformed data in a violin plot. The violin plot shows median values and 25% and 75% interquartile ranges of the CT and 5% and 95% ranges. The non-parametric test (Siegel-Tukey) shows that the CV% of the reproducibility at 1000 copies/10 6 PBMC is significatively different between the dPCR and qPCR with a p-value = 0.024.

Methods
Study design. In the first part of the study, we evaluated the variation in HIV DNA concentration quantified with a commercial qPCR kit in subjects followed for HIV infection [41][42][43] . Intra-individual variation in HIV DNA concentration over time was assessed in subjects who had been aviremic for at least 24 months and whose total HIV DNA level was considered stabilized.
In the second part of the study, we evaluated the potential gain of using a dPCR technique compared to qPCR. For this purpose, we: (i) developed a dPCR technique and evaluated its characteristics and analytical performances according to the dMIQE guidelines 17 ; (ii) evaluated the variations in HIV DNA concentrations in different batches of the 8E5 line used as a standard for qPCR to highlight possible variations in the qPCR standard that would contribute to the lack of precision of this technique; (iii) compared HIV DNA values obtained in dPCR and qPCR from 8E5 cells diluted with PBMC and from clinical samples.

Patients.
Outpatients consulting for HIV therapeutic monitoring in the infectious and tropical diseases department of the Montpellier University Hospital were included after having provided written informed consent to the use of their clinical and biological data for the study (DC-2011-1401). This study has received an approval from Comité de Protection des Personnes Sud Méditerranée III institutional ethics committee and followed the guidelines and regulations of the same institute (ID RCB no 2011-A01566-35, and DC-2011-1401). Blood samples from 51 PLHIH were analyzed in the HIV reservoir follow-up study. 48 patients were in a prolonged virological success (HIV reservoir considered stabilized), and 3 patients were considered positive controls (less than 24 months after the treatment initiation phase with an HIV reservoir considered not stabilized). Patient characteristics are presented in Table 3.
Cells samples. The 8E5 cells constituting the standard for the qPCR assay were used for dPCR and qPCR analytical performance analysis. Theoretically, cells from this cell line contain one copy of mutated HIV DNA per cell 44 . The laboratory's 8E5 cell line (8E5A)(ATCC CRL-8993TM) was thawed and cultured in RPMI 1640 medium complete with 9% SVF and stored at 37 °C and 5% CO 2 . Positive cell samples containing 10 6 8E5 cells and 10 6 PBMC and negative samples containing 10 6 PBMC were performed using the Bio-Rad TC10TM automated cell counter. PBMCs were obtained from healthy blood donors and cryopreserved until used. DNA was extracted from these samples on the QIAcube from QIAGEN using the QIAamp® DNA Mini Kit protocol. Quantification with Nanodrop One confirmed their cellular concentration at 61.85 ng/µL of DNA (i.e., an equivalent of 2 × 10 6 cells) for HIV-positive samples and 33.88 ng/µL of DNA (i.e., an equivalent of 10 6 cells) for HIV negative samples. These samples were then aliquoted and stored at − 20 °C before being used for dPCR development and as positive and negative controls. To test the dPCR's internal control, the negative samples were also used to Aliquots of an 8E5 cell line (8E5B) and a U1 cell line 19,20 , donated by the laboratory IRIM of CNRS (Montpellier) at a theoretical concentration of 1 copy per cell, were also quantified by dPCR.
As previously described, the Biocentric, Generic HIV DNA CELL kit standard used in qPCR comprises DNA extracted from this cell line 24 . Dilutions of this standard, provided at a concentration of 3 × 10 6 copies of HIV DNA/mL, were used to evaluate the analytical performance of the two PCR methods. Serial dilutions were performed with the human genomic DNA solution, also provided in the kit, to obtain concentrations from 12 to 1250 copies/10 6 PBMC. For each test performed, dilutions were made on the same day and stored as aliquots at − 20 °C. Sampling procedure. DNA extraction was performed on a 200 µL volume of fresh blood on the QIAcube automated system from QIAGEN following the QIAamp ® DNA Mini Kit protocol. The extracted DNA eluate was then assayed with Nanodrop One and stored at − 20 °C before PCR techniques.
Precision study. Repeatability and reproducibility experiments were performed simultaneously on both methods to test and compare the accuracy of dPCR and the qPCR. 8E5 cells samples at concentrations of 100 and 1000 copies/10 6 PBMC, consistent with the values generally found for the HIV reservoir in patients with prolonged therapeutic success, were used. One of these concentrations was close to the LOD at 100 HIV DNA copies/10 6 PBMC (1.25 times the LOD of dPCR and 2.5 times the LOD of qPCR), and the other at 1000 copies/10 6 PBMC (12.5 times the LOD of dPCR and 25 times the LOD of qPCR). For repeatability, the samples were tested ten times on a single run and by a single operator. In the reproducibility study, the two concentrations were tested twenty times on twenty different runs over ten days and by two operators.  www.nature.com/scientificreports/ The results of these precision experiments are presented by the coefficients of variation (CV%) calculated from the untransformed data (formula 1 = σ∕μ) and on the log 10 transformed data, by applying two different formulas to calculate the CV of the latter (formula 1 = σ∕μ and formula 2 = √(e (σ2) − 1)). Formula 1 is commonly used to calculate the CV of untransformed data, and Formula 2 the CV of log-transformed data. Calculating the CV of log-transformed data using formula 1 is often used in the literature but has severe statistical errors, as this formula is incorrect because the mean μ depends on the unit used. Thus, using formula 1 by transforming data into logs could make the coefficient of variation vary to almost zero. To compare with the results reported in the literature, the coefficients of variation according to these three cases have been indicated.

Probes and primers.
To closely compare between the dPCR and qPCR techniques, we have used the same pair of HIV DNA primers and probe targeting HIV LTR, already used in the literature 18  After distribution, the chips were loaded onto the thermal cycler to undergo the following amplification program: an initial denaturation time of 10 min at 96 °C, followed by 35 cycles of hybridization and elongation of 2 min at 57 °C and 30 s at 98 °C, then a final elongation time at 60 °C for 2 min. The chips were transferred and read on the QuantStudio chip reader at the end of this program. The dMIQE guidelines were followed to validate our digital PCR technique 17 .
Bio-Rad and Stilla digital PCR thermocyclers were used to test the Biocentric qPCR range. The PCR mix and program used on these dPCRs were the same as those developed on the QuantStudio 3D dPCR. qPCR from Biocentric. qPCR was performed following the protocol of the Biocentric Generic HIV DNA CELL kit on the Roche Life Science LightCycler ® 480 18 . According to the manufacturer and previous study the LOD was 40 copies/10 6 PBMC 18 . Data analysis. qPCR results were analyzed on LightCycler ® 480 SW 1.5 software from Roche Life Science. dPCR results were analyzed on QuantStudio 3D Analysis Suite Cloud Software v3.1 from Thermo Fisher Con-nectTM. As the automatic fluorescence threshold setting proposed by this software was not adapted to the analysis of a set of samples, several control points were put in place to perform a manual threshold setting 45,46 . Firstly, the chip had to be of high quality (no bubbles or reading problems) and have a total number of partitions greater than 15,000 out of the 20,000 possible partitions. A minimum threshold fluorescence level was established through simplex dPCR experiments for each target, on FAM fluorescence targeting HIV DNA and VIC fluorescence targeting the internal control β-globin. An assessment of the percentage of false positives was performed on a range of negative samples, allowing us to estimate the possible number of false positives at each threshold value 47 . A threshold of 2300 was chosen, giving a 0% median false positive (data not shown). With this minimum threshold and the study estimating the percentage of false positives at each threshold value, we manually choose the most appropriate threshold for each sample. Finally, the experiments were also analyzed in triplicates with a time-lapse between each analysis. Statistical analyses and figures were made on GraphPad Prism 8 and Inkscape.