Characterization of HIV variants from paired Cerebrospinal fluid and Plasma samples in primary microglia and CD4⁺ T-cells

For this cross-sectional study, paired CSF-plasma samples were collected from stored samples in the period 2001–2016 from patients in care at the University Medical Center Utrecht and participating in the Dutch ATHENA observational cohort. Paired samples were obtained for clinical diagnostic purposes [Table S1]. A total of 28 subjects, with and without neurological symptoms, had sufficient material for virological analysis, of which 9 were excluded due to being on different antiretroviral therapies at the time of sampling. CSF in neurosymptomatic subjects was collected for clinical diagnostics purposes. The CSF of neuroasymptomatic patients was primarily collected to exclude neurosyphilis, as part of a standard clinical procedure in patients with TPHA (treponema pallidum hemagglutination assay) or (previous) VDRL- positive plasma syphilis serology and in two instances for a diagnostic work-up not including neurological symptoms [S1]. Paired samples were defined as CSF and plasma-EDTA (or serum) obtained within 7 days from each other. All patients had detectable HIV RNA in plasma at the time of lumbar puncture and were ART-naïve or off-treatment at the time of sampling [Tables 1 and S1].

HIV RNA levels in plasma and CSF were determined by an ultrasensitive viral load assay with a reported cut-off value of 50 copies/ml (Ampliprep/COBAS Taqman HIV-1 assay, Roche).

Viral RNA was isolated according to the method developed by (Boom et al. 1990). The HIV-1 envelope V3 region was amplified by RT-PCR (Titan One Tube RT-PCR kit, Roche) followed by a nested PCR (Expand High Fidelity PCR System, Roche), according to the manufacturer’s protocol. Please refer to supplementary Table S2 for a list of the primers used. PCR products were purified with the Qiaquick PCR purification kit (Qiagen). Library preparation was done using a Nextera-XT DNA Library Preparation and Index kit (Illumina, USA) according to the manufacturer’s instructions. The resulting libraries were normalized and pooled. Sequencing was performed on an Illumina MiSeq platform using the MiSeq Reagent Kit v2 for 500 cycles. After aligning the sequence reads of each subject with the consensus sequence of their respective subtype, reads that overlap the entire V3 region were isolated and trimmed. Unique V3 sequences with a prevalence of > 1% in the population were used for HIV-1 co-receptor tropism. Co-receptor usage was predicted with the Geno2pheno[coreceptor] algorithm version 2.5 for deep sequences with the recommended false-positive rate (FPR) cut-off value for deep V3 sequencing of 3.5% (Swenson et al. 2011). It is predicted that an FPR value below 3.5% indicates an X4-tropic virus, whereas a value above 3.5% indicates an R5-tropic virus.

Genetic compartmentalization between the CSF and plasma-derived viral population, was determined for each subject based on only deep-sequenced V3 sequences using three methods: Wright’s measure of population subdivision (Fst) (Hudson et al. 1992), Nearest-neighbor statistic (Snn) (Hudson 2000), and the tree-based Slatkin-Maddison test (SM) (Slatkin and Maddison 1989). Wright’s measure of population subdivision (Fst) quantifies the genetic variance between populations relative to the total genetic variance. Higher Fst values indicate greater population differentiation. The nearest-neighbor statistic (Snn) assesses how often sequences' nearest neighbors are from the same compartment, with values closer to 1 indicating stronger compartmentalization. The Slatkin-Maddison (SM) test evaluates the number of migrations needed to explain the distribution of lineages between populations, with fewer migrations suggesting compartmentalization. All three methods were conducted with the HyPhy software version 2.2.4 (Kosakovsky Pond et al. 2005). For the two distance-based methods (Fst and Snn), the Tamura-Nei 93 algorithm was applied along with a bootstrap value and permutation test of 10,000 with only deep-sequenced V3 sequences as input. For the tree-based SM test, multiple sequence alignments of the HIV env V3 regions were performed using the online version of the MAFFT software (https://​mafft.​cbrc.​jp/​alignment/​server/​index.​html) (Katoh et al. 2018), accessed 2022/08/09 using standard parameters for nucleic acid alignment. A Nearest Neighbor Joining Tree was constructed using MEGA version 11.0.11(Tamura et al. 2021), and the Kimura 2-parameter substitution model was applied. Compartmentalization was evaluated using the nearest-neighbor statistic (Snn) (Hudson 2000), applying 10 000 permutations, implemented using the HyPhy software package version 2.2.4 (Kosakovsky Pond et al. 2005). CSF viral populations were defined as either compartmentalized (cp), if all three tests were significant (p < 0.01), or equilibrated (eq) if statistical significance was not reached (p > 0.01) in one or more of these tests (Zárate et al. 2007).

The HIV-1 envelope (gp160) region was amplified by RT-PCR (Superscript IV Reverse Transcriptase Kit, Invitrogen) followed by a nested PCR (Platinum Taq Superfi PCR Master Mix, Invitrogen), according to the manufacturer’s protocol. Please refer to supplementary Table S2 for a list of the primers used. Envelope amplicons were introduced into a HxB2 gp160deletion vector with a luciferase reporter gene (HxB2ΔENVluc), previously described in (Gumbs et al. 2022), using the NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs). We used the same vector carrying either the gp160 sequence of JRCSF (R5 T-tropic), YU-2 (R5 M-tropic) or BaL (R5-tropic) as controls. Hek-293 T cells were transfected with the chimeric plasmids using lipofectamine 2000 reagent (Invitrogen). After 48 h, the supernatant containing replication-competent virus was harvested and stored at −80 °C until further use. p24 was determined with an ELISA p24 assay (Aalto Bioreagent, Dublin, Ireland).

Fresh postmortem adult human brain tissue was provided by the Netherlands Brain Bank (NBB). The isolation of primary microglia was conducted according to the protocol described previously with some minor modifications for human brain tissue (Mattei et al. 2020). Following isolation, primary microglia were cultured in poly-L-lysine hydrobromide (PLL)-coated 96-well plates (1 × 10⁵ cells/well) in microglia medium (RPMI 1640 (Gibco Life Technologies) supplemented with 10% FCS, 1% penicillin–streptomycin (Gibco Life Technologies) and 100 ng/mL IL-34 (Miltenyi Biotec)) for 2–3 days before infection. Primary microglia were infected overnight with 10 ng (p24Gag) virus after which the medium was fully replaced and cells were cultured up to 17 days in microglia medium without medium refreshment.

PBMC were isolated from peripheral blood obtained from healthy donors by Ficoll-Paque density gradient. CD4⁺ T-cells were subsequently isolated through negative selection with the CD4⁺ T Cell Isolation Kit (Miltenyi Biotec 130–096-533), according to the manufacturer’s protocol. Before infection, CD4⁺ T-cells were stimulated for 2 days in culture medium (RPMI 1640 (Gibco Life Technologies) with 10% Fetal Bovine Serum, 1% penicillin–streptomycin (Gibco Life Technologies) and IL-2 (20U/mL) (Invitrogen)) supplemented with Phytohaemagglutinin (5 µg/mL). CD4 infection was performed in Duplo for 3 h with 10 ng (p24 Gag) virus per 100.000 cells in Eppendorf’s placed on a tube rotator. For the Maraviroc experiment, CD4⁺ T-cells were treated with 100 nM MVC for 1 h before infection. Following infection, medium was fully replaced with culture medium and CD4⁺ T-cells were cultured for 14 days in 96-well plates without medium refreshment.

Supernatant was collected 2–3 times per week and luminescence was measured according to the manufacturer’s protocol with the Nano-Glo® Luciferase Assay System (Promega). The graphs were created with GraphPad Prism version 8.3.0.

All data were analyzed with GraphPad Prism version 8.3.0. Descriptive statistics were used to compare the characteristics between the paired plasma and CSF samples. The nonparametric Wilcoxon signed rank test is used to compare groups with paired data, including the differences in FPR. Differences within continuous variables (e.g., viral load) compared to categorical data (e.g., neurological symptoms) were performed by using the Mann–Whitney U test. A Pearson correlation is used to determine the association of HIV-RNA levels.

Paired plasma and CSF samples were collected from 19 adult subjects enrolled in the Dutch ATHENA observational cohort subjects were mainly infected with Subtype B virus, except for subject 6 (subtype CRF02_AG) and subject 25 (CRF12_BF), and were ART naïve or off treatment at the time of sampling for at least 1 month. The majority of the study population was male, with a mean age of 47 years and a mean CD4⁺ T-cell count of 312 cells/µL. The clinical and virological characteristics of each subject can be found in Table 1 and Table S1.

Median HIV RNA levels were significantly higher in plasma than in CSF (5.01 vs. 4.12 log10 cp/mL; p = 0.004) and moderately correlated with each other (Pearson r = 0.42; p = 0.04) [Fig. 1a, b]. A similar pattern in virus concentration was also observed in the neurosymptomatic subjects (5.22 vs. 4.25 log10 cp/mL; p = 0.05) [Fig. 1d]. However, further research with a larger sample size is needed to determine whether this finding is also true for neuroasymptomatic subjects [Fig. 1c]. A comparison of the CSF and plasma RNA levels between the neurosymptomatic and neuroasymptomatic subjects, however, showed no significant difference [Fig. 1e, f].

Genetic compartmentalization between CSF and plasma HIV variants was determined based on the V3 region in the viral envelope (env) gene. CSF viral populations were defined as either compartmentalized (cp), if all three compartmentalization analyses (Fst, Snn, SM) were significant (p < 0.01), or equilibrated (eq) if statistical significance was not reached (p > 0.01) (Zárate et al. 2007) [Table 1]. Most of the subjects (89%) had equilibrated viral populations in their CSF and plasma. Significant genetic CNS compartmentalization was detected in two subjects, 13 and 19. Subject 13 had advanced disease (CD4 count 30 cells/μl) and neurological symptoms consisting of balance disturbances and peripheral weakness. In contrast, subject 19 had less advanced HIV infection (CD4 count 387 cells/μl) and had no neurological symptoms, suggesting that CNS compartmentalization does not always present with clinically observable neurological symptoms.

Co-receptor usage was mostly concordant across the paired samples, with 17 out of 19 pairs predicted to consist exclusively of CCR5-using viral strains in both compartments [Table 1]. A comparison of the lowest FPR values in plasma and CSF revealed no correlation or significant difference [Data not shown]. Subject 17 is predicted to have CCR5- and CXCR4-using viral strains in both plasma and CSF, whereas subject 19 is predicted to have CXCR4- and CCR5-using viruses in the plasma but only CCR5-using virus in the CSF. In addition to being the only subject with CXCR4-using virus in the CSF, subject 17 was also the only subject diagnosed with severe symptoms (HIV encephalopathy) and one of three subjects with a CSF HIV RNA load higher than in plasma (difference 1.21 log10 copies/mL).

Within the CNS, HIV DNA is primarily detected in perivascular macrophages and primary microglia (Joseph et al. 2015). We investigated the entry phenotype of the CSF and plasma HIV variants from compartmentalized (cp) neuroasymptomatic subject 19 and two equilibrated (eq) subjects, subjects 8 and 27, with neurological symptoms [Table S1]. These three subjects were selected in a step-wise process. First, we selected subjects with a difference in FPR value between plasma and CSF, then we selected subjects from whom the volume of stored CSF and plasma was sufficient for the experiment and finally the clones needed to be viable from both plasma and CSF. First, we generated CSF- and plasma-derived luciferase reporter viruses using the full-length envelope (Env) gene amplified from the CSF and the plasma of these three subjects. For each subject, we obtained a diverse mixture of viral clones with different FPR values between the CSF and plasma [Fig. 2]. These viral clones were phenotypically characterized for viral entry into CD4⁺ T-cells (high CD4 surface levels) and primary microglia (low CD4 surface levels), the main HIV target cells in the blood and CNS. Considering that CD4 surface expression levels on CD4⁺ T-cells and microglia are likely to differ between donors, we used 3 different donors and three laboratory strains as a control for infection: two R5 M-tropic virus (Bal and YU-2) and one R5 T-tropic virus (JRCSF).

For subjects 19 (cp) and 27 (eq), we observed no significant difference between the ability of the CSF and plasma viruses to infect CD4⁺ T-cells [Fig. 2b, c]. In subject 8 (eq), we observed a ≥ 10-fold higher infectivity with CSF clones 4 and 6, compared to the plasma viruses [Fig. 2a]. Interestingly, this infectivity was also substantially higher than CSF clone 1 which had the same FPR value, suggesting that there are other determinants outside of the V3 loop that can greatly affect cell entry. Furthermore, treatment with the CCR5 inhibitor maraviroc (MVC), supported co-receptor prediction and significantly inhibited T-cell infection by the R5-predicted CSF and plasma viruses, whereas the X4-predicted plasma viruses of subject 19 were resistant to MVC inhibition [Fig. 2d–f]. R5-predicted CSF clone 4 of subject 8, despite having a high FPR value of 28.8, was also greatly resistant to MVC inhibition (≥ 60%), suggesting that this clone is dual tropic [Fig. 2d].

Lastly, phenotyping of the viruses in low CD4-expressing primary microglia revealed that most CSF and plasma clones were unable to efficiently enter these cells, although CSF-derived viral clones were overall better in entering microglial cells than the plasma-derived clones [Fig. 3]. The enhanced ability of CSF clones, to infect microglia was more pronounced in compartmentalized subject 19 [Fig. 3b]. While both BaL and YU-2 are R5 M-tropic viruses, BaL was isolated from infant lung tissue (Gartner et al. 1986), whereas YU-2 was isolated from brain tissue (Li et al. 1991) and hereby potentially more representative of the neurotropic viruses in the CNS. From this perspective, we compared the infection of the viral clones to YU-2 and found that each subject had one CSF clone with an intermediate M-tropic phenotype, defined as ≥ 50% of YU-2 infection, for cell entry in microglia [Fig. 4]. In subject 27, we also found one plasma clone with this intermediate M-tropic entry phenotype. The plasma clone had a higher FPR than the CSF clone with a similar intermediate phenotype (71.1 vs. 58.6), suggesting two separate virus populations possibly originating from different low CD4- expressing cells.