Validation of a UHPLC-MS/MS Method to Quantify Twelve Antiretroviral Drugs within Peripheral Blood Mononuclear Cells from People Living with HIV

Recently, anti-HIV treatment has achieved high efficacy and tolerability. Nevertheless, few data are available about the intracellular penetration of antiretrovirals, partly due to the technical challenges related to intracellular quantification. This work aimed to validate an ultra-high performance liquid chromatography (UHPLC) tandem mass spectrometry (MS/MS) method for the simultaneous quantification of maraviroc, nevirapine, rilpivirine, dolutegravir, raltegravir, cobicistat, darunavir, ritonavir, atazanavir, efavirenz, elvitegravir, and etravirine within peripheral blood mononuclear cells (PBMCs) and apply it to samples from patients. PBMCs were isolated by density gradient on cell preparation tubes (CPT). Samples were prepared by addition of internal standards (IS), sonication, centrifugation, and drying. Reconstituted extracts underwent chromatographic separation by reversed phase UHPLC and detection was performed by electrospray ionization and multiple reaction monitoring. Method validation followed FDA and EMA guidelines, showing acceptable accuracy, precision, recovery and IS-normalized matrix effect. The application to 56 samples from patients undergoing antiretroviral treatment provided description of intracellular penetration, showing method eligibility for future studies.


Introduction
Nowadays, combination antiretroviral therapy (cART), consisting of the combination treatment with several antiretroviral drugs (ARVs) from at least two different ARV classes, has considerably improved tolerability, efficacy, and genetic barrier to resistance, greatly increasing life expectancy in people living with HIV (PLWH) [1]. The major classes of currently available ARVs include nucleoside reverse transcriptase inhibitors (NRTIs), non-nucleoside reverse transcriptase inhibitors (NNRTIs), protease inhibitors (PIs), entry inhibitors (EI), fusion inhibitors (FI), and integrase strand-transfer inhibitors (INSTI) [2][3][4]. Currently, the standard cART entails the administration of two NRTIs as a "backbone" and at least one drug from another class, while in recent years some "nucleoside-sparing" regimens are also being investigated and adopted, combining at least two highly potent drugs from two different classes [5,6]. In particular, dolutegravir (DTG) has recently been identified as a first-line treatment for HIV infection in adults and adolescents [7]. Despite the current effectiveness of cART in suppressing HIV viral load, HIV infection remains a chronic condition, due to latency and residual replication of the provirus in viral reservoirs [8][9][10][11][12][13][14]. These include lymphoid tissues, lymph nodes, the central nervous system and latent infected lymphocytes and monocytes [9,10,14,15].
It is important to note that current monitoring of plasma concentrations of ARVs does not predict their penetration into the target cells and, consequently, into the reservoirs [16]. In recent years, several studies have been conducted to determine the penetration of ARVs into peripheral blood mononuclear cells (PBMCs), used as a surrogate for the viral reservoirs [9,[17][18][19]. It has been demonstrated that ARV concentrations in mononuclear cells extracted from different compartments (e.g., peripheral blood, lymph node, and rectal lymphoid tissue) can differ according to the molecule, but these are usually considerably lower than in plasma, leading to tissue-specific underdosing [9,20]. These findings greatly increased the interest in studying intracellular pharmacokinetics (PK). Moreover, the recent introduction and development of cell-targeted prodrugs (e.g., tenofovir-alafenamide) [21] or nanoformulations [22,23] further support the importance of studying intracellular ARV PK. Nevertheless, to date, the study of intracellular concentrations has been limited to a small number of studies, mainly due to technical challenges and invasiveness in isolating PBMCs or obtaining tissue biopsies, respectively, as well as the need for very high analytical sensitivity. Therefore, the intracellular quantification of ARVs is not widely applied in research, while application in clinical practice, such as for therapeutic drug monitoring (TDM), remains a remote prospect.
Therefore, thoroughly validated methods for the quantification of ARVs in PBMCs are needed in order to estimate their penetration in viral reservoirs. Several methods have been reported for the quantification of ARVs in PBMCs [25,27,31,32] and in other cell types [33], but these included only a few currently used ARVs; moreover, these methods mainly focused on NRTIs [21]. Furthermore, there are no reported methods for the simultaneous intracellular quantification of dolutegravir (DTG), elvitegravir (ELV), and rilpivirine (RPV), together with other ARVs of non-nucleos(t)ide nature.

Calibration Curve and Dilution Integrity
Mean determination coefficients (R 2 ) of all calibration curves ranged from 0.996 to 0.999 (Table 1), with a linear regression model forced through zero. The low concentrations and the presence of SIL-ISs limited the saturation phenomenon. Dilution of samples over the Upper Limit Of Quantification ULOQ led to inaccuracy and imprecision lower than 15%. Table 1. For each compound are reported, in order: mean retention time (RT), highest spiking solution concentration, final drug amounts in the standard (STD) 9 to STD1/LLOQ, LOD, dwell times and mass transitions (parent ions, first and second daughter ions), with the corresponding cone voltages and collision energies. Unit "ng/sample" refers to the absolute amount of drug added to a blank PBMC aliquot of 16 × 10 6 cells, to obtain the calibration STDs. LLOQ: lowest limit of quantification.

Specificity and Selectivity
The assay did not show any significant interference with tested drugs. Analytes' retention times are summarized in Table 1 and an overlaid chromatogram is reported in Figure 1.

Specificity and Selectivity
The assay did not show any significant interference with tested drugs. Analytes' retention times are summarized in Table 1 and an overlaid chromatogram is reported in Figure 1. The blank PBMC samples did not show any endogenous interference causing a significant "noise", taking into account the retention times of the analytes.

Accuracy and Precision
The validation results in terms of accuracy and precision are listed in Table 2. Both imprecision and inaccuracy were below 15%, in keeping with FDA and EMA guidelines [34,35]. The blank PBMC samples did not show any endogenous interference causing a significant "noise", taking into account the retention times of the analytes.

Accuracy and Precision
The validation results in terms of accuracy and precision are listed in Table 2. Both imprecision and inaccuracy were below 15%, in keeping with FDA and EMA guidelines [34,35].

Lowest Limit of Quantification (LLOQ) and Limit of Detection (LOD)
The determination of the LOD was carried out as reported in Section 2.8, executing serial dilutions of the lowest calibrator, standard 1 (STD1) for each drug. LLOQ and LOD are shown in Table 1.
The LLOQ, investigated in five replicates (as other quality control (QC) samples), was at least equal to the lowest point of the curve (STD1) as required by the FDA guidelines, without significant interfering peaks by sample matrix (Figure 2).

Recovery
Recovery data are summarized in Table 3, demonstrating consistency and reproducibility for each compound, as well as concordance between target analytes and their ISs. The lower recovery values for COBI, ATV, and ETV were supposed to be due to a possible adsorption to the cellular matrix but considering that their relative standard deviation (RSD)% was contained, these were considered acceptable.

Matrix Effect
Matrix effect (ME) data are summarized in Table 3. Despite the fact that the mean values of ME were high in some cases, particularly for COBI and ELV, their RSD% was within 15%, thus being acceptable according to EMA guidelines and previous reports [30,34]. The mean values of IS-nME were generally lower than the ones of ME, indicating that the ISs were able to reduce the mean ME. Nevertheless, their corrective impact on the RSD% values was not always appreciable (e.g., NVP, DRV, and ELV), still remaining within the 15% limit.

Carry-Over
Carry-over experiments led to signals which were lower than at the LOD. The mean IS area obtained from the analysis of blank PBMC samples was lower than 1% of the one observed in PBMC samples containing IS, confirming the absence of significant carry-over.

Stability
All drugs were stable bench-top for 24 h (25 • C with artificial light), showing a degradation below 7% at each concentration. All the analytes and IS, kept for 24 h in the autosampler at 10 • C, showed a variation lower than 6% at each concentration.
Taking into account the analytical variability, the processed samples were stable throughout the UHPLC/MS-MS analysis, always completed within 24 h.

Testing of Patients' Samples
The new method was tested on 56 PBMC samples from 30 HIV positive patients undergoing cART mainly based on boosted PIs. Concentrations were evaluated in samples taken at the pharmacological steady state, at the end of the inter-dose interval (trough concentrations) and with detectable drug concentrations in plasma. These samples corresponded to 19 different ARV combinations, including all the drugs considered in this method, except for EFV and NVP. Among these, 34 contained DTG, 2 RAL, 1 ELV, 1 RPV, 5 ETV, 31 DRV, 21 ATV, 29 RTV, and 17 COBI. The quantification of each drug was successful in all samples, showing concentrations within the calibration ranges (Table 4). Concentrations for all drugs were comparable to previously published data [36]. Concerning intra-PBMC/plasma ratios, DTG exhibited the lowest intra-PBMC penetration, followed by RAL, DRV, ATV, MVC, COBI, RTV, RPV, ETV, and ELV (Table 4). The IntraPBMC/plasma ratios were significantly and moderately correlated (R 2 = 0.391; p-value = 0.025) to the LogP values of each drug.

Stock Solutions, Standards, and Quality Controls
Calibration standards (STDs) and quality controls (QCs) were prepared by spiking "blank" PBMC aliquots with 100 µL of previously prepared "spiking solutions". The highest spiking solution (to prepare STD 9) and the ones for the preparation of three QCs (High, Medium, and Low) were prepared at 100 ng/mL, 80 ng/mL, 10 ng/mL, and 1 ng/mL, except for NVP, DRV, and EFV, which were prepared at a 10-fold higher concentration, by directly diluting stock solutions with water:acetonitrile 70: 30 (v:v); the other spiking solutions were prepared by serial (1:1) dilution of the highest one, to obtain nine different concentrations. Drug concentrations in the spiking solutions and the final absolute amounts in the STD samples, respectively, are listed in Table 1.
Flow rate was 0.4 mL/min, with a gradient of two mobile phases (MP): MP-A (0.05% v:v FA in water) and MP-B (0.05% v:v FA in acetonitrile). Briefly, the chromatographic gradient started with 70:30 MP-A:MP-B for 0.3 min, then the percentage of MP-B linearly increased to 47%, 55%, 60%, 75%, and 95% at 8, 9, 10, 11, and 11.7 min, respectively, achieving the separation and elution of all the analytes. The column was washed with 95% MP-B for 1.5 min and reconditioned for 1.8 min, for a total runtime of 15 min.
Two different autosampler washing solutions were used: a "weak-washing solution" (water:acetonitrile 70:30 v:v) and a "strong-washing solution" (water:acetonitrile 30:70 v:v). Two "strong" washing steps and 3 "weak" washing steps (250 µL each) were applied after each injection. Plasma concentrations were evaluated by the same analytical method, following a sample preparation protocol described by Simiele et al. [37]. Plasma concentration data underwent internal and external quality assessment (INSTAND, EQA, Dusseldorf, Germany), with conforming results.

Mass Spectrometry Conditions
Tandem mass spectrometric detection was carried out with a QSight ® 220 (Perkin Elmer, Milan, Italy) tandem mass spectrometer, with electrospray ionization (ESI) interface. The ESI source was set in positive ionization mode (ESI+) for all drugs except for EFV, which was detected by negative electrospray ionization (ESI−).
"Zero-Air" (dry air) was used as a nebulizing and heating gas, while nitrogen was used as "drying and collision" gas: both these gases were produced at high purity (>99.9%) with a Zefiro ® HQ (Cinel, Vigonza, Italy) gas generator.  Table 1. One transition was used as a quantification trace, while the other one was used as confirmation (secondary ion trace). Mass data were acquired and processed through Simplicity ® software ver. 4.6.3 (Perkin Elmer, Milan, Italy).

PBMC Isolation
PLWH receiving several cART regimens containing the target drugs at standard doses and who gave informed consent, as indicated by the local Ethical Committee approvals, were enrolled in the context of the approved clinical study ENHANCERS (Prot. N • 1479/007A/2017, 24/11/2017); a protocol aiming to evaluate the effect of pharmacoki-netic boosters of plasma and intracellular ARV concentrations (COBI, RTV, or no booster). These patients underwent blood sampling at the end of the dosing interval (C trough ), with cell preparation tubes (CPT ® , Becton, Dickinson and Co., Franklin Lakes, NJ, USA, for PBMC isolation) and with lithium heparin tubes (for plasma). Plasma was obtained after centrifugation at 1400× g (3000 rpm) for 10 min at 4 • C (Jouan Centrifuge, Model BR4i, Saint-Herblain, France). It then underwent heat inactivation, drug extraction, and analysis as previously described [37]. Blood samples (2 CPT of 8 mL each, total volume 16 mL) from patients were processed following a PBMC isolation protocol previously described in several studies [38][39][40][41].
Briefly, CPTs were centrifuged at 1700× g for 15 min at 20 • C, then the PBMC layer was transferred to Falcon tubes and adjusted at the volume of 40 mL with NaCl 0.9% (isotonic) solution. After a centrifugation at 700× g for 6 min at 4 • C, the supernatant was discarded and the resulting pellet was washed with 2 mL of an ammonium salts solution (3.5 g ammonium chloride + 0.036 g ammonium carbonate in 500 mL of water) for 1 min, in order to achieve the lysis of residual erythrocytes. Then the solution was diluted to a final volume of 40 mL with NaCl 0.9%: two aliquots of 500 µL were further diluted 1:40 with NaCl 0.9% (final volume 20 mL) and used to perform cell counts and mean cell volume (MCV) measurement with an automated Z2 Beckman Coulter (Instrumentation Laboratory, Milan, Italy). The remaining 39 mL of solution were immediately centrifuged at 700× g for 6 min at 4 • C and the supernatant was discarded. All the washing solutions were kept at 4 • C, in order to reduce any drug efflux during cell isolation.
Cell pellets from patients were resuspended with 1 mL of water:methanol 30:70 (v:v), divided in two aliquots (500 µL) and stored at −80 • C until analysis. The number of cells and MCV data for each sample from patients were used to determine the total volume of cells in each aliquot, allowing the conversion of results from "ng/aliquot" to "ng/mL", as previously described [41].
Blank PBMCs used for the preparation of standards and quality controls were extracted from buffy coats, following the same protocol used for isolation from blood.

STDs, QCs, and Patient Samples Extraction
Blank PBMC samples (500 µL each) were spiked with 100 µL of the corresponding spiking solutions, in order to obtain the STD and QC samples; similarly, STD 0 and patient samples were spiked with 100 µL of water:acetonitrile 70: 30 (v:v), in order to mimic the STDs volume. Each sample was then spiked with 40 µL of IS working solution, and then, after vortex-mixing for 10 s, it underwent sonication for 10 min at room temperature. Subsequently, samples were centrifuged at 21,000× g for 10 min, without brake, at 4 • C.
Supernatants were then transferred to glass tubes and dried in a vacuum centrifuge at 50 • C (nearly 1.5 h). Dry extracts were resuspended with 100 µL of water:acetonitrile 70:30 (v:v) and 10 µL were injected in the chromatographic system. Statistical treatment of analytical data was performed through Microsoft Office Excel (ver. 2007) and SPSS 26.0 (IBM, Armonk, NY, USA).

Specificity and Selectivity
Interference from endogenous compounds was investigated by analysis of six different blank PBMC samples, at medium and high cell numbers (16 × 10 6 and 8 × 10 6 cells/aliquot).
"Interference" was defined as observable ion suppression/enhancement or cross-talk with any of the target analytes.

Accuracy, Precision, Calibration, and Limit of Quantification
Intra-day and inter-day precision and accuracy were determined by analyzing three different QC concentrations (plus the lowest limit of quantification, LLOQ) in six validation sessions.
Accuracy was calculated as the ratio between the mean analytical result and the nominal concentration. Inter-day and intra-day imprecision were expressed as the relative standard deviation (RSD%) at each QC concentration, at the three different QC samples (and at the LLOQ) between sessions and in 5 replicates in the same analytical session, respectively. A maximum RSD% value of 15% and a minimum mean accuracy value of 85% was considered as acceptable.
Each calibration curve was obtained by processing the nine STD points in double replicate.
Calibration curves were processed by analytes' peak area normalized by areas of IS compounds.
A "linear through zero" with 1/x weighting regression model was used for all curves, and model fitting was evaluated up to a concentration twice as high as STD 9. The forced through 0 model was chosen in order to highlight eventual lack of linearity at low levels and to avoid an "overweighting" of the low standard points for the definition of calibration curves (calibration parameters are shown in Table S1). The back-calculated concentrations of at least 7 out of 9 standard samples had to be within a 15% of bias, as suggested by EMA and FDA guidelines, in order to accept the calibration curve (mean accuracy of STD in Table S2).
Dilution integrity was tested by analyzing five replicates of a sample spiked at a concentration twice as high as STD 9 (which was considered as the ULOQ), after a 3-fold dilution with blank PBMC lysates from healthy donors.
Limit of detection (LOD) was defined as the concentration that yielded a signal-tonoise ratio of at least 3:1. "Noise" was evaluated as the mean MS signal at the analytes' retention times corresponding to the injection of a "blank" PBMC sample. Percentage of deviation from the nominal concentration (measure of accuracy) and relative standard deviation (measure of precision) of the concentration considered as the lower limit of quantification (LLOQ) had to be < 20%, and it was considered as the lowest calibration standard, as suggested by FDA and EMA guidelines [34,35].

Recovery
Recovery from PBMC samples was assessed by comparing the peak areas obtained from multiple analyses of QC samples (High, Medium, and Low) with peak areas obtained from the injection of blank PBMC extracts spiked after the extraction (post-extraction addition method) at the same concentrations.

Stability
Stability data for the considered drugs were already reported in previously published studies [42][43][44][45][46][47], so no further extensive long-term stability evaluation was performed in this study. On the other hand, confirmative reanalysis of 11 patient samples containing RPV, DTG, ATV, RTV, DRV, and COBI over a period of three months indicated a maximum deviation of 8% (for RTV), which was within method variability [24,25,27,33,48,49]. Furthermore, a short-term evaluation of stability in the working conditions (in the autosampler and "bench-top") was performed up to 24 h.

Matrix Effect
The "matrix effect" (ME) was investigated in six lots of blank PBMC at two different cell numbers (16 × 10 6 and 8 × 10 6 cells/aliquot) from different donors, as suggested by the guidelines [34,35]. Peak areas from blank extracts spiked with all analytes at three QC concentrations were compared with peak areas from standard solutions (prepared in water:acetonitrile 70:30 v:v) spiked at the same concentrations. The ME was calculated as the percent deviation between the peak areas obtained from the PBMC extracts and those obtained from the standard solutions. Moreover, in order to estimate the capability of IS compounds for correcting ME variability, the "IS-normalized ME" was also calculated, as previously reported [30] and suggested in EMA guidelines [34].

Carry-Over
Carry-over was investigated in triple replicates by injecting extracted blank PBMC extracts immediately after samples containing target analytes at concentrations twice as high as STD 9. A value ≤ 20% of the lower limit of quantification (LLOQ) and a value ≤ 5% for IS were considered as absence of significant carry-over.

Conclusions
In this work, a robust, practical, and relatively economical approach has been described for the quantification of all the current and frequently used ARVs within PBMC within a single assay. This method is relatively cheap since it does not need solid phase extraction, the SIL-ISs are now widely available on the market more cheaply than previously, and the amount of IS needed for each extraction is quite low (1 mg of SIL-IS covers theoretically more than 3 × 10 5 samples). This approach, from sampling, isolation, counting, and final analysis, has the great advantage of obtaining results expressed as concentration, being comparable with those from plasma.
By the application of this method to a small cohort of PLWH treated with a wide range of different combination treatments, several interesting data were observed, which confirmed the marked differences between plasma and intra-PBMC concentration of ARVs.
Interestingly, different classes showed opposite behaviors in terms of intracellular penetration: with the exception of DRV, PIs as well as pharmacokinetic boosters which share similar chemical properties, showed higher intracellular penetration (in accordance with previous data [18,19]), while intra-PBMC concentrations of INSTIs were found to be nearly one third of those in plasma, with the exception of ELV. These results are in accordance with the recent work from Dyavar et al. [50], evidencing a rather poor penetration of INSTIs within lymphoid cells, with the only exception being ELV, which exhibited a very high penetration, probably due to its high LogP value.
It is worth mentioning that the correlation between LogP and intracellular penetration was moderate (R 2 = 0.391), highlighting that it is not the only leading force at the basis of intracellular disposition of ARVs: other important factors, such as drug transporters (e.g., P-gP), are already known to have an impact on intracellular penetration of ARVs, as well as underlying some of its inter-individual variability [51][52][53].
Nevertheless, despite the interesting information about ARV penetration within PBMCs, this study presents some limitations, which should be taken into account. Firstly, the heterogeneous range of regimens enabled simply a description of intracellular penetration of these drugs in a "real life" setting, whereas comparison between regimens was impossible due to the low numbers of samples. Nevertheless, this type of analysis will surely be the object of future studies.
Secondly, although this method allows a reliable quantification of ARVs within the cells, it cannot verify the real disposition of drugs within the cell: in fact, a portion of these drugs, which are quite lipophilic, could be bound to the cell membranes or organelles [17]. For this reason, the direct use of intracellular concentrations of ARVs for pharmacodynamic evaluations might potentially result in an overestimation of their antiviral effect. From a practical point of view, the use of 2 CPT tubes (16 mL) which was chosen in order to ensure a rugged analysis even in potentially lymphocytopenic patients, could be problematic for a large scale application or for particularly critical patients. Nonetheless, the observed concentrations in patients' samples were always abundantly within the calibration ranges, indicating that, probably, the analysis could be successfully conducted starting from a lower blood volume (1 single CPT, 8 mL).
Notwithstanding these limitations, the intracellular quantification of ARVs in PBMC provides an important marker of drug penetration into lymphocytes and monocytes, as well as a better surrogate of the quantification in lymphoid tissues than plasma. The method presented in this work, validated to FDA and EMA standards, will enable such analysis in a greater number of future clinical studies.
Supplementary Materials: The following are available online at https://www.mdpi.com/1424-824 7/14/1/12/s1, Figure S1: Chemical Structures of the target analytes, Figure S2: Chemical structures and labeling of the stable-isotope linked internal standards. Table S1: calibration curves parameters for each analyte. Actual intercepts were set to 0 due to the adoption of a "linear through zero" with 1/x weighing calibration model. Each final model has to be considered as: "Response = Slope * concentration + 0". The mean observed intercept removing "through zero forcing" is also reported, for comparison, Table S2: Mean accuracy percentages of the back-calculated concentrations of calibration standards during the validation sessions.