Postmortem Analysis of Dolutegravir, Tenofovir, Lamivudine, and Efavirenz Penetration in Multiple Central Nervous System Compartments

Abstract

Background

Central nervous system (CNS) compartmentalization provides opportunity for human immunodeficiency virus (HIV) persistence and resistance development. Differences between cerebrospinal fluid (CSF) and cerebral matter regarding HIV persistence are well described. However, CSF is often used as surrogate for CNS drug exposure, and knowledge from solid brain tissue is rare.

Methods

Dolutegravir, tenofovir, lamivudine, and efavirenz concentrations were measured across 13 CNS regions plus plasma in samples collected during autopsy in 49 Ugandan decedents. Median time from death to autopsy was 8 hours (interquartile range, 5–15 hours). To evaluate postmortem redistribution, a time course study was performed in a mouse model.

Results

Regions with the highest penetration ratios were choroid plexus/arachnoid (dolutegravir and tenofovir), CSF (lamivudine), and cervical spinal cord/meninges (efavirenz); the lowest were corpus callosum (dolutegravir and tenofovir), frontal lobe (lamivudine), and parietal lobe (efavirenz). On average, brain concentrations were 84%, 87%, and 76% of CSF for dolutegravir, tenofovir, and lamivudine, respectively. Postmortem redistribution was observed in the mouse model, with tenofovir and lamivudine concentration increased by 350% and efavirenz concentration decreased by 24% at 24 hours postmortem.

Conclusions

Analysis of postmortem tissue provides a unique opportunity to investigate CNS antiretroviral penetration. Regional differences were observed paving the way to identify mechanisms of viral compartmentalization and/or neurotoxicity.

The central nervous system (CNS) is a reservoir of human immunodeficiency virus (HIV), which contributes to compartmentalized replication of the virus [1–4], and, in some cases, subsequent development of resistance to antiretroviral therapy (ART) [5–7], viremic relapse after treatment interruption [3], and HIV-associated neurocognitive disorders (HAND) despite viral suppression in the peripheral blood [8–10]. One hypothesis of compartmentalized viral replication in the CNS is that the CNS exposure of antiretrovirals is insufficient to eliminate the virus and provides a favorable environment to select less susceptible variants. This hypothesis is indirectly supported by studies that showed use of CNS-penetrating antiretrovirals is associated with lower CSF viral load [11–13] or better neurophysiological outcomes [10, 12, 14, 15]. Meanwhile, some antiretrovirals are associated with neurotoxicity [16] and potentially contribute to development of HAND [17]. Therefore, understanding antiretroviral CNS penetration is critical to improving therapy to balance between protecting the CNS from HIV persistence and from drug-induced neurotoxicity. As brain tissue is not readily available from clinical routine investigations, cerebrospinal fluid (CSF) has been widely used as a surrogate of the entire CNS for investigation of viral replication and antiretroviral distribution. However, the adequacy of considering CSF as a surrogate for brain has not been well justified, and some studies found CSF is not representative of brain tissue [18, 19].

The current study used human postmortem brain tissues, including multiple anatomical regions, to determine differences in distribution and penetration of 4 antiretrovirals (dolutegravir, tenofovir, lamivudine, and efavirenz) that are first-line or alternative first-line treatments in multiple guidelines [20–22]. The effect of demographic and clinical factors on CSF or brain penetration was also examined. In addition, because the phenomenon of postmortem redistribution has been reported in multiple forensic toxicological studies [23], but has never been described for antiretrovirals, a mouse model was used to investigate the potential for postmortem redistribution of tenofovir, lamivudine, and efavirenz in brain tissues.

METHODS

Human Postmortem Study

Decedents who were hospitalized due to advanced HIV/AIDS were enrolled with written informed consent from next of kin for autopsy at Mulago National Referral Hospital and Kiruddu National Referral Hospital in Kampala, Uganda from 2017 to 2020, including 4 participants whose data have been previously reported [24]. A confirmed HIV diagnosis was the only inclusion criteria. Individuals who had trauma precluding reliable autopsy or tissue collection were excluded. Medical history, including opportunistic infection diagnoses, use of ART and concomitant medications, and adherence, were collected from a combination of hospital records (when available) or from interviews with next of kin or carers. Autopsies were performed within 32 hours of death, during which approximately 1–2 g brain tissues/fluid were collected from 13 regions of the CNS (CSF, frontal lobe, corpus callosum, parietal lobe, occipital lobe, globus pallidus, hippocampus, cerebellum, midbrain/substantia nigra, pons, medulla oblongata, cervical spinal cord/meninges, and choroid plexus/arachnoid) and immediately snap-frozen in liquid nitrogen. Whole blood from the femoral vein was collected into EDTA vacutainers and centrifuged at 4°C to separate plasma. CSF was collected via cisternal puncture. Specimens were transferred via liquid nitrogen to −80°C freezers where they were stored until analysis. The study protocol was approved by the Research Ethics Review Committee at Mulago National Referral Hospital. Plasma creatinine concentrations and plasma and CSF albumin concentrations were measured from stored postmortem samples by Fairview Diagnostic Laboratories (Minneapolis, Minnesota).

Mouse Postmortem Redistribution Study

Animal experiments were conducted under approved protocol by the Virginia Commonwealth University institutional animal care and use committee. Institute of Cancer Research mice (Envigo, Dublin, VA) were administered a daily intraperitoneal injection at a dose of 60 mg/kg tenofovir disoproxil fumarate, 60 mg/kg lamivudine, and 30 mg/kg efavirenz for 5 consecutive days. On the fifth day of administration, all mice were sacrificed by cervical dislocation 2 hours after the last intraperitoneal injection. Brains from 30 mice were harvested in groups of 6 at multiple time points postsacrifice: 0, 2, 6, 12, and 24 hours after death. Tissue samples were snap-frozen immediately and stored at −80°C until sample analysis.

Quantification of Antiretrovirals With LC-MS/MS

Quantification of dolutegravir, tenofovir, lamivudine, and efavirenz in human and mouse tissues was performed with liquid chromatography with tandem mass spectrometry (LC-MS/MS) at the Clinical Pharmacology Analytical Services laboratory at the University of Minnesota College of Pharmacy. The method for tenofovir, lamivudine, and efavirenz in human plasma, CSF, and brain tissue has been previously described [24, 25]. Efavirenz in CSF was not quantified due to low solubility of efavirenz in CSF solution and expected low concentrations of efavirenz in CSF. Dolutegravir was measured in a separate simultaneous assay with 4 other antimicrobials. Details of both methods can be found in the Supplementary Materials.

Data Analysis

Penetration of antiretrovirals in each anatomical compartment is described by tissue penetration ratio, the ratio of concentration in tissue over concentration in plasma. Tissue to plasma ratio is only reported for human samples when the drug concentration was quantifiable in both the plasma and the tissue compartment. Arithmetic means of concentrations and penetration ratios of 12 solid brain regions were calculated for the human samples and defined as “composite brain.”

Summary statistics were calculated for absolute concentrations and penetration ratios. For each drug, comparison of natural log transformed concentrations and penetration ratios across all brain regions were conducted with the 1-way analysis of variance (ANOVA) test, followed by a post hoc pairwise comparison for all pairs using the paired t test (P value adjusted for multiple comparisons by the Holm method). Tests of normality and homogeneity are in Supplementary Tables 1 and 2. The effect of age, sex, Cryptococcus, tuberculosis, postmortem interval (time interval between death and autopsy), concomitant use of rifampin, estimated glomerular filtration rate (estimated with the Chronic Kidney Disease Epidemiology Collaboration [CKD-EPI] equations), plasma creatinine, time since last dose, and CSF/plasma albumin ratio on the penetration of antiretrovirals were tested by simple linear (univariate) regression. All statistical analyses were conducted at the significance level of α = .05. The data were analyzed using R statistical software (version 4.2.2).

RESULTS

Characteristics of Study Participants

Forty-nine participants were included in the final analysis, with clinical and demographic characteristics shown in Table 1. Age of decedents ranged from 18 to 79 years, with 75% of participants under 47 years of age. The time to autopsy was a median of 8 hours (interquartile range, 5–15 hours) from death. Causes of death included HIV-related opportunistic infections (ie, cryptococcal meningitis, tuberculosis, pneumonia, and toxoplasmosis), malignancies (ie, Kaposi sarcoma and pancreatic tumor), anemia, respiratory/cardiac/hepatic failure, etc. The most commonly prescribed antiretroviral combinations were tenofovir disoproxil fumarate/lamivudine/efavirenz, followed by tenofovir disoproxil fumarate/lamivudine/dolutegravir. Concomitant medications for Cryptococcus and tuberculosis infections included fluconazole, flucytosine, amphotericin B, isoniazid, ethambutol, rifampin, and pyrazinamide.

Concentrations of Dolutegravir, Tenofovir, Lamivudine, and Efavirenz in Plasma and Brain Subcompartments

We initially assayed samples from 53 participants. Samples from 4 participants were excluded as their plasma was below the limit of quantification (BLQ), indicative of no recent antiretroviral dosing, leaving 49 participants in the final analysis. Antiretroviral concentrations in plasma, CSF, and the 12 brain regions are shown in Supplementary Table 1, as well as the number of BLQ samples by each brain region. Figure 1 displays the distribution of the concentrations in each anatomical compartment relative to 50% maximum effective concentration (EC₅₀) or EC_90–95 as reported in Food and Drug Administration package inserts [26–29]. Dolutegravir had the highest proportion of BLQ samples followed by tenofovir, while lamivudine and efavirenz were detectable in all brain samples, regardless of region (Supplementary Table 3). Notably, the BLQ samples in brain specimens were not randomly distributed across participants as 74% of BLQ samples came from the same 2 individuals. The highest median concentration of dolutegravir, tenofovir, lamivudine, and efavirenz of all 13 brain regions (Figure 1 and Supplementary Table 3) were seen in cervical spinal cord/meninges, parietal lobe, medulla oblongata, and pons. The lowest median concentrations of the 13 brain regions (Figure 1 and Supplementary Table 1) were midbrain/substantia nigra for dolutegravir, corpus callosum for tenofovir, frontal lobe for lamivudine, and occipital lobe for efavirenz. Overall, brain concentrations above the highest EC₅₀ for dolutegravir or EC_90–95 for efavirenz were noted for most participants. In contrast, for tenofovir and lamivudine, most participants had brain concentrations that fell inside the range of reported EC₅₀ (Figure 1). Comparisons between brain regions of log-transformed concentrations were significant for lamivudine (P < .001) and efavirenz (P = .021), but not for dolutegravir (P = .117) or tenofovir (P = .073). In post hoc pairwise comparisons for lamivudine, 21 pairs were significantly different, with significantly lower frontal lobe concentrations than found in 9 other brain regions (Figure 1). Other significant pairwise differences are shown in Figure 1. Dolutegravir CSF and plasma concentration were highly correlated (r = 0.89, P < .001), as were tenofovir (r = 0.79, P < .001) and lamivudine (r = 0.89, P < .001) (Supplementary Figure 1).

CSF concentrations tended to be lower than plasma concentrations for dolutegravir, tenofovir, and lamivudine (geometric mean ratio [GMR] of CSF to plasma 0.075, 0.172, and 0.341, respectively). Correlation between CSF and composite brain were stronger for dolutegravir (r = 0.95, P < .001) and lamivudine (r = 0.93, P < .001) than tenofovir (r = 0.54, P = .013) (Supplementary Figure 1).

Penetration of Dolutegravir, Tenofovir, Lamivudine, and Efavirenz in Brain Regions

Antiretroviral penetration was determined by the ratio of concentration between each brain region and plasma. Distribution and summary statistics of these ratios are illustrated in Supplementary Figure 2 and reported in Table 2. All ANOVA tests of penetration ratios across regions were significant (dolutegravir, P = .002; tenofovir, P < .001; lamivudine, P < .001; efavirenz, P = .021). In post hoc pairwise comparisons for dolutegravir only corpus callosum was significantly smaller than choroid plexus/arachnoid; for tenofovir, penetration in corpus callosum was significantly lower than 5 other regions (Supplementary Figure 2); 21 pairs were significantly different for lamivudine, and the penetration ratio in the frontal lobe was significantly lower than in 10 other regions; and 3 pairs were significantly different for efavirenz (data not shown). Regions with the highest penetration ratios were choroid plexus/arachnoid (dolutegravir and tenofovir), CSF (lamivudine), and cervical spinal cord/meninges (efavirenz); the lowest were corpus callosum (dolutegravir and tenofovir), frontal lobe (lamivudine), and parietal lobe (efavirenz) (Table 2).

CSF concentrations tended to be greater than concentrations in the composite brain (0.84, 0.87, and 0.76 tissue to CSF GMR for dolutegravir, tenofovir, and lamivudine, respectively). Geometric mean CSF to tissue ratios for each brain region are shown in Table 3.

Covariates on CSF and Brain Penetration

Results from the univariate regression between multiple demographic or clinical factors and penetration ratio in CSF or composite brain are shown in Supplementary Table 4. Male participants had lower dolutegravir penetration in CSF than female participants (P = .045), but no significant sex differences were observed for drug concentrations with the composite brain (P = .438). Participants with cryptococcal meningitis had higher CSF penetration of lamivudine (P = .027) and a trend towards higher CSF penetration of dolutegravir (P = .050). The ratio of CSF to serum albumin, which reflects the permeability/integrity of the blood-brain barrier (BBB) [30] and blood-CSF barrier (BCSFB) [30, 31], although not statistically significant, was possibly associated with increased CSF penetration of dolutegravir (P = .054), tenofovir (P = .07), and lamivudine (P = .069). Concomitant use of rifampin was associated with a higher penetration ratio in composite brain for dolutegravir (P = .027) and tenofovir (P = .037). Postmortem interval, the time between death and autopsy, was significant for penetration in composite brain for tenofovir (P = .035). Variables that were identified in univariate analysis as potentially important (P < .05) for penetration in CSF or composite brain are shown in Supplementary Figure 3. Age, coinfection with tuberculosis, glomerular filtration rate, serum creatinine, and time since last dose were not significant predictors of drug penetration into CSF nor brain compartments.

Postmortem Redistribution of Tenofovir, Lamivudine, and Efavirenz in Mice

Concentrations of tenofovir, lamivudine, and efavirenz in mice brain changed over a 24-hour period after death (Figure 2). While tenofovir and lamivudine displayed similar trends of increasing concentrations over time, efavirenz had a decreasing trend. At 2, 6, 12, and 24 hours after death, median concentrations of tenofovir were 1.7, 1.4, 2.6, and 3.6-fold of the median concentration at time of death (0 hour), median lamivudine concentrations were 1.6, 1.7, 3.3, and 4.5-fold of the 0 hour, while the median efavirenz concentrations were 0.88, 0.75, 0.76, and 0.76-fold of the 0 hour. The t tests between concentrations postmortem and at time of death (0 hour) showed no significant difference (P adjusted > .05) except for lamivudine at 24 hours postmortem (P adjusted = .045) (Figure 2).

DISCUSSION

Our robust sampling approach of 12 brain regions in addition to plasma and CSF allowed for an unprecedented and comprehensive analysis of antiretroviral distribution throughout the CNS in humans. Regional differences were discovered. Notably, there was lower accumulation of lamivudine and efavirenz in cortical regions (eg, frontal lobe, occipital lobe, and parietal lobe) compared to the rest of the regions examined. Regional differences in antiretroviral exposure, particularly when concentrations fell below inhibitive concentrations, could create a selective pressure that drives the mutation of HIV towards certain regions of the brain and promotes HIV mutation against the antiretroviral. Therefore, this pressure could contribute to the creation of HIV reservoirs. Antiretrovirals with relatively low genetic barriers to resistance, such as tenofovir, lamivudine, and efavirenz [32], showed significant regional differences in our analysis, highlighting the importance of combining these with antiretrovirals that exhibit higher genetic barriers, such as dolutegravir [33, 34] or protease inhibitors [35]. On the other hand, for antiretrovirals with high penetration and/or with well-known neurotoxicity (eg, efavirenz), the regional difference identifies potentially more vulnerable regions (ie, regions with higher penetration ratios). This could warrant in vitro or preclinical studies to identify specific tissue concentration cutoffs.

Most studies investigating antiretroviral penetration have used CSF as a surrogate of the whole CNS; however, how representative CSF is of brain concentrations is unknown. Best et al reported tenofovir CSF concentrations ranging from 0.4% to 84% (median, 5%) of plasma, similar to the 17% we observed [36]. Lamivudine CSF concentrations have previously been reported to be between 15% and 40% of plasma, consistent with the median of 34% observed in our study [37, 38]. Dolutegravir CSF concentrations have been reported to range from 0.2% to 5% of plasma, slightly lower than the median 7.5% we report here [39]. These differences may be partially explained by differences between CSF collected via lumbar puncture and cisternal puncture. While data on drug exposure in human brain parenchyma are limited, our data are in contrast to Ferrara et al who reported that on average tissue concentrations in brain were higher than CSF concentrations; however, they did not measure CSF in their study and were comparing to historical controls [40].

In our study, the ability of CSF concentrations to predict brain tissue concentrations varied by drug. For dolutegravir and lamivudine, strong correlations were identified between CSF and brain tissue. However, for tenofovir, the relationship between CSF and brain tissue was more variable. We found that, overall, penetration varied across regions and the difference varied by drug, and CSF could be lower or higher than the solid brain tissues depending on the specific region. Both raw concentrations as well as tissue to plasma ratios are presented, as both have strengths and limitations in their interpretation. Raw concentrations provide insight into concentrations achieved in brain tissue. Using penetration ratios allows adjustment for factors such as variable adherence and renal function that may influence overall systemic exposures.

Different dolutegravir CSF penetration between sexes has not been previously reported. However, this may explain reports that women are more vulnerable to neurorelated adverse effects than men. For example, a study investigating CNS adverse effects among German patients on dolutegravir-based therapy reported that women were more likely to discontinue therapy due to CNS adverse effects than the men [41].

The ratio between CSF and serum albumin has been widely used as a marker of the integrity of BBB [30] and BCSFB [31] because more albumin leaks into CSF as the integrity of the barriers deteriorates. We saw a positive trend between CSF penetration and CSFto serum albumin ratio for dolutegravir, tenofovir, and lamivudine, although this was not statistically significant. The difference in lamivudine CSF penetration with and without Cryptococcus infection is consistent with our previous analysis [24], and implies that meningeal infection is likely to increase drug distribution into CSF. Although the CSF to serum albumin ratio did not differ by Cryptococcus infection status in our study, the increased drug distribution could be mediated by the increased permeability of BBB and/or BCSFB.

Postmortem redistribution is a well-known phenomenon. The challenges have long been recognized by forensic toxicologists, especially where postmortem analysis is used for drugs with neuropsychiatric impacts or narrow therapeutic windows and those often associated with overdose, such as some antidepressants, benzodiazepines, opioids, etc. [23]. However, postmortem redistribution of antiretrovirals has not been investigated. In the current study, we used a mouse model to investigate the diffusion of 3 antiretroviral drugs in the brain over a 24-hour period after death. Efavirenz concentrations decreased in the brain over time, which could be a result of continued metabolism by cytochromes P450s and UDP-glucuronosyltransferases in the brain, even after death [42]. Redistribution to surrounding tissues appeared to be along the concentration gradient because we observed higher concentrations of efavirenz in brain relative to the periphery in human. In contrast, brain concentrations of tenofovir and lamivudine increased over time. This could be due to redistribution from surrounding tissues because their initial brain concentrations were relatively low. In addition, there were notably larger changes in the first 2 hours postmortem compared to later time points, which may be a transient period during which there was still active enzyme-mediated metabolism and blood movement. This is also why we did not force straight regression lines across the postmortem concentrations to derive a postmortem rate of change. In addition, median tissue concentrations were calculated from results in different mice and the change over time may be an artifact of subject variability. Overall, from the animal study, we confirmed the postmortem redistribution of antiretrovirals in the brain; however, the pattern and extent differed by drug. Therefore, the concentration of efavirenz we measured in human brain tissues may be underestimated, while the concentration of tenofovir and lamivudine may be overestimated. Of note, postmortem change in the tissue penetration ratios was not available in the animal study due to blood coagulation occurring soon after death. Although we have demonstrated drug redistribution in the postmortem brain, we remain uncertain whether or how the postmortem penetration ratios would change. Furthermore, we hesitate to extrapolate the results to humans as the quantitative extrapolation would require a good understanding of the postmortem allometric scaling between preclinical species and humans, while postmortem changes are distinct between humans and animals, such as the rate of drug metabolism, rate of decomposition, and coagulation [43]. Additionally, disease-related factors that confound the drug penetration in humans may not be reflected by the animal model.

To our knowledge, the current study is the most extensive investigation of antiretroviral distribution and penetration in a variety of subanatomical brain tissues. Nonetheless, there are limitations. First, despite our large total sample size relative to other studies, there were fewer participants on dolutegravir and efavirenz than tenofovir and lamivudine, which may have hindered statistical interpretations for those drugs. Second, our participants were critically ill upon death, which might be accompanied by multiple organ dysfunction that impacts pharmacokinetics of drugs in the body, although we attempted to quantify for this by measures of kidney function such as creatinine clearance. We expect that this would contribute more to absolute concentrations rather than relative distribution between compartments, which is why we report both in this study. Also, because these participants were critically ill, medication adherence information came mostly from interview with caretakers. This is because in Uganda, caretakers, not hospital staff, are responsible for administration of oral medications. However, we used detectable plasma concentration as criteria to confirm the recent administration of antiretrovirals and, once again, we used relative concentrations rather than raw concentrations to inform on relative penetration to account for variable adherence. Here we report total, not free, concentrations of drugs in all compartments. Dolutegravir and efavirenz are both highly protein bound and therefore comparison of free drug would likely be more informative. Of note, most of the participants had increased BBB permeability as measured by CSF to serum albumin ratio; therefore, caution must be taken in extrapolating results to the general population. Lastly, our work is limited to the Ugandan population and more research is needed to determine how generalizable our findings are to other populations.

These findings pave the way for future studies to explore potential significance of these regional differences in antiretroviral penetration. In particular, our findings regarding antiretroviral penetration into a putative HIV reservoir (the CNS) has significant relevance to the HIV cure agenda and the hope for eradication of HIV. Future studies could investigate the neurological effect of high penetration of efavirenz by incorporating neurocognitive and neuropsychiatric assessments and brain distribution of efavirenz. Future studies could also explore the correlation between tissue viral load and mutation/resistance and regional exposure of antiretrovirals. In conclusion, with postmortem analysis, we measured the penetration of dolutegravir, tenofovir, lamivudine, and efavirenz in various brain regions that are rarely available in common clinical or research settings, and significant regional differences were discovered that may inform future clinical care.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Author contributions. D. M., M. M., D. B., and M. N. conceived and designed the research. F. W., K. R., O. N., K. A., J. F., and R. L. performed the research. F. W., K. R., and M. N. performed data analysis. F. W. and M. N. wrote the initial draft of the manuscript. All authors critically reviewed the manuscript.

Acknowledgments. We gratefully acknowledge the study participants and their next of kin for making such generous contributions to close the knowledge gap and benefit the successors. We acknowledge the following collaborators: Lilian Tugume, Morris Rutakingirwa, Enock Kagimu, Kenneth Ssebambulidde, and John Kasibante. Also, we are grateful for the active cooperation of the research teams from different institutions.

Financial support. This work was supported by the National Institute of Neurologic Disorders and Stroke (grant numbers R21NS108344 and R01NS086312); National Institute of Allergy and Infectious Diseases (grant numbers K08AI134262 and U01AI125003-04S1); National Institute on Drug Abuse (grant number R01DA05315); and National Institutes of Health Clinical and Translational Science Award (grant number 1UM1TR004405-01A1 to University of Minnesota).

Presented in part: Conference on Retroviruses and Opportunistic Infections, 2022. 12–16 February 2022, Virtual. Nicol MR et al. Postmortem tissue sampling to describe exposure of eight anti-infectives in the brain. Poster No. 450.

References

Author notes