Pregnancy-induced physiological changes reduce plasma concentrations of antiretrovirals (ARV), especially in the third trimester (1–4). Mother-to-child transmission (MTCT) of HIV is reduced significantly at the standard dose of current ARVs in use (1, 4–7). Cases of perinatal transmission, although not common, and vaginal shedding of HIV RNA among pregnant women with undetectable or low plasma HIV RNA suggest that declining MTCT may not be attributed to low plasma HIV RNA viral load alone (8–10).

The use of ARVs suppresses plasma HIV RNA levels below the limit of detection (11). However, rapid viral rebound in non-adhering patients suggests that replication-competent viruses persist in HIV reservoirs during treatment (12, 13). Suboptimal adherence may therefore cause a viral rebound in pregnant women, which can increase the risk of mother-to-child transmission (MTCT) (14–16). The lymphoid tissues constitute the largest HIV reservoir sites because they are the primary sites for viral replication, and therefore contain a high proportion of viral genetic components and free virions (15, 17, 18). Furthermore, persistent isolates of HIV particles in lymph nodes of patients on active ART also suggest that the virus may be capable of evading lethal ARV concentrations in maternal plasma. This has constituted a major barrier in HIV eradication (12, 19–22). Penetration of ARVs into the lymphatic tissues is crucial for prevention of viral replication, rebound, drug resistance and MTCT (23).

Quantification of drug distribution into the lymphatic system of living persons has not been studied due to the challenges with sample collection. Macaque mass spectrometry imaging, human lymph node mononuclear cells, and human primary lymphoid endothelial cells are methods that have been reported so far in the literature for drug quantification in lymphoid tissues (24–26). Ethical considerations around sample collection and safety concerns limit fetal pharmacokinetics studies before delivery (27). These gaps may be filled through physiologically-based pharmacokinetic (PBPK) modeling and simulation.

Materno-fetal PBPK (m-f-PBPK) modeling strategy has advanced from simple models to using highly representative models that include gestational-age dependent changes in maternal and fetal anatomy and physiology (27–30). M-f-PBPK models have been used to reliably estimate fetal concentrations of emtricitabine, tenofovir, nevirapine, darunavir, efavirenz, and thalidomide (28–30). These predictions sometimes rely on a number of assumptions based on data derived from in vitro or animal models in the absence of relevant clinical pregnancy data. However, a robust mechanistic workflow on PBPK models starting from simple non-pregnant models validated with available clinical data to more complex materno-fetal models, often builds confidence in the data output from such models. Applications of such models to HIV tissue reservoirs could support the development of molecules with optimal characteristics for enhanced distribution in HIV eradication studies. There is currently no published description of ARV distribution into lymphoid tissues during pregnancy.

In the current work, we describe the extension of our previous m-f PBPK model (28) to describe the penetration of efavirenz, dolutegravir and rilpivirine into the lymph and fetal plasma during pregnancy.

Our extended m-f-PBPK model which incorporated lymphatic circulation into an existing whole-body pregnancy model was successfully used to predict maternal lymph and fetal plasma pharmacokinetics of the ARVs efavirenz, dolutegravir, and rilpivirine. Model predictions for maternal plasma pharmacokinetics during the third trimester and infant delivery drug exposures were within 1.08–1.99 average fold difference of clinical data. Predicted maternal lymph-to-plasma AUC ratio was highest for efavirenz at 6.4, followed by rilpivirine at 1.7 and lowest for dolutegravir at 0.016. Model-predicted fetal plasma-to-maternal plasma AUC ratios were 0.59 for efavirenz, 0.78 for dolutegravir, and 0.57 for rilpivirine. The median predicted lymph concentration at 24 h after dose was above the published MEC for efavirenz and rilpivirine only. The predicted low lymphoid tissue penetration of dolutegravir appears to be significantly counterbalanced by its extended dose forgiveness (42 h compared with 12 h for efavirenz and 0 h for rilpivirine) and adequate fetal compartment exposure. Hence, it is unlikely to be a predictor of maternal virological failure or mother-to-child transmission risks.

Although ART suppresses plasma viraemia below the limit of detection, persistence of latent but replication-competent HIV in sanctuary tissues during active treatment constitutes a challenge in HIV cure research (38). Despite lymphoid tissues having the highest proportion of latent HIV (14, 15), no comprehensive assessment of lymph pharmacokinetics of ARVs in humans (pregnant and non-pregnant) is available due to the invasiveness of the conventional lymph node aspiration technique. Although, a number of studies have used in-vitro, ex-vivo, and in vivo animals models to determine lymphatic exposure of efavirenz, dolutegravir, and rilpivirine (16, 24, 67), such models are known to be inadequate representations of what is expected in humans. It is known that suboptimal adherence to ART may lead to subtherapeutic drug levels in the systemic circulation, stimulating latent HIV in lymphoid tissues to resume active replication, thereby causing viral rebound (19). Detectable viral load is a known risk factor for MTCT and optimal adherence during pregnancy remains critical.

Administration of ARVs during pregnancy has several benefits, notably PMTCT which may be partly due to fetal prophylactic pre-exposure to ARVs. Past studies have relied on umbilical cord blood concentration at delivery to measure the extent of fetal exposure of ARVs. This method has limitations such as single time-point measurement and sample-time variation relative to maternal dosing. In this study, the infant plasma concentration prediction was validated with efavirenz and dolutegravir clinical data for infant washout in non-breastfed babies 2–10 h post-delivery. Post-delivery scenarios were simulated by stopping maternal dosing at term, and then estimating median fetal concentration after 2–10 h. The results were within the acceptable 2-fold difference for efavirenz and dolutegravir. The validated model was applied to rilpivirine and its fetal plasma concentration-time profile was also predicted successfully. The predicted fetal-to-maternal plasma ratio of efavirenz, dolutegravir, and rilpivirine were 0.591, 0.781 and 0.573, respectively. The predicted median fetal concentration at 24 h was higher than MEC for efavirenz (1,123 vs. 800 ng/mL) and dolutegravir (927.4 vs. 300 ng/mL), but lower for rilpivirine (41.26 vs. 50 ng/mL). Differential concentrations of efavirenz, dolutegravir and rilpivirine in maternal lymph and fetal plasma is, to a certain extent, as a result of differences in their physicochemical properties such as plasma protein binding, log P, molecular weight, and pKa (68–70). For instance, high pKa, log P and hydrophobicity of efavirenz were identified to be responsible for high penetration of efavirenz into human lymphoid endothelial cells compared to dolutegravir (24).

Our present study predicted C_trough of 992 ng/mL and AUC of 46,114 ng/mL for dolutegravir in non-pregnant women, these are similar to predictions by Freriksen et al. (71), and Liu et al. (72) respectively, and are within 1.5-fold error to clinically observed data of the drug. This further indicated a strong reliability in the non-pregnant model and the confidence to extend it to incorporate the pregnancy model. Furthermore, the maternal dolutegravir PK parameters during pregnancy predicted by the model employed in this present study were similar to those predicted by Liu et al. (72). Additionally, the predicted fetal exposure to dolutegravir in the present study was comparable to that reported by Freriksen et al. (71). Although, fetal-to-maternal AUC plasma exposure ratio was predicted in our study, it was assumed to be similar to and within the range of the cord blood-to-maternal blood concentration ratios predicted in their study and observed clinical data. Likewise, the C_max and the C_min of the maternal plasma concentration predicted by our model was lower and higher, respectively, in comparison to their reported values (71). Further studies are still suggested to establish this assumption of similarity.

Dose forgiveness was used to estimate how long it would take for drug concentration in maternal lymph and fetal plasma to reduce below MEC in non-adhering pregnant mothers. Efavirenz and rilpivirine lymph concentrations remained above MEC for 126 and 32 h, respectively; efavirenz may therefore offer a longer protection in lymph against latent HIV in non-adhering pregnant women. While this may be an advantage, the ability of wild-type HIV to develop resistance to efavirenz monotherapy is of concern (67). Therefore, further investigation is required to know the extent of lymph exposure of tenofovir and emtricitabine which are commonly used in combination with efavirenz. In fetal plasma, efavirenz and dolutegravir concentration remained above MEC for 12 and 42 h, respectively. These results showed that efavirenz and rilpivirine may offer adequate protection against viral rebound from the maternal lymph nodes, but in rare situations where the virus enters the systemic circulation, efavirenz and dolutegravir may offer adequate fetal pre-exposure prophylaxis. Longer dose forgiveness of dolutegravir in fetal plasma offers sustained pre-exposure prophylaxis to fetus in pregnant women with suboptimal adherence. These results do not reflect the enzyme induction or inhibition effect of other drugs used as combination therapy. Therefore, interpretation of these results may be limited clinically.

Although, the current model reliably predicted lymphatic and fetal exposure to efavirenz, dolutegravir, and rilpivirine during the third trimester, a number of limitations are identifiable. Firstly, our model did not account for the possible role of transporter activities in placental drug transfer due to lack of sufficient data for model parameterization. The use of data from cell lines expressing relevant transporters such the BeWo monolayer are possible options to mechanistically describe these processes. Unfortunately, these data are not currently available in the literature and thus the option of relying only on passive processes for our predictions. Additionally, such models do not adequately recapitulate these processes in humans. The inclusion of drug transporters and associated variability in their expression can potentially improve the accuracy of model predictions, particularly for drugs that are substrates for these transporters. Secondly, due to lack of data we relied on key assumptions supported by sensitivity analyses for placental diffusion constants of study drugs. Model predictions were fitted to clinically observed infant plasma concentration at delivery. While this resulted in adequate predictions of infant exposure of the study drugs at delivery, further enhancement is desirable in future studies. Thirdly, there was no previous study to validate rilpivirine infant washout data. The validated model with available clinical data on infant washout for efavirenz and dolutegravir was extended to predict for rilpivirine. Furthermore, there are no clinical data available in humans to validate the predictions of the lymphatic model. Although lymph exposure data are available from ex-vivo, in-vitro, and animal studies, we could not rely on them to assess the predictive performance of the lymphatic model due to well-established inter-species variation.

In conclusion, predictions from our extended m-f-PBPK model showed differences in the distribution of efavirenz, rilpivirine, and dolutegravir into the lymph during pregnancy and the fetal compartment. Importantly, the inclusion of dose forgiveness predictions indicate alignment with recommendations of no dose adjustment despite moderate changes in exposure during pregnancy observed in clinical studies. This is an important new application of PBPK modeling strategy to evaluate the adequacy of drug exposure in an otherwise inaccessible compartment.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

All co-authors contributed equally to the conception of the ideas presented here, the conduct of the research, and the preparation of this manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.