Evolution of Neuroadaptation in the Periphery and Purifying Selection in the Brain Contribute to Compartmentalization of Simian Immunodeficiency Virus (SIV) in the Brains of Rhesus Macaques with SIV-Associated Encephalitis

ABSTRACT

The emergence of a distinct subpopulation of human or simian immunodeficiency virus (HIV/SIV) sequences within the brain (compartmentalization) during infection is hypothesized to be linked to AIDS-related central nervous system (CNS) neuropathology. However, the exact evolutionary mechanism responsible for HIV/SIV brain compartmentalization has not been thoroughly investigated. Using extensive viral sampling from several different peripheral tissues and cell types and from three distinct regions within the brain from two well-characterized rhesus macaque models of the neurological complications of HIV infection (neuroAIDS), we have been able to perform in-depth evolutionary analyses that have been unattainable in HIV-infected subjects. The results indicate that, despite multiple introductions of virus into the brain over the course of infection, brain sequence compartmentalization in macaques with SIV-associated CNS neuropathology likely results from late viral entry of virus that has acquired through evolution in the periphery sufficient adaptation for the distinct microenvironment of the CNS.

IMPORTANCE HIV-associated neurocognitive disorders remain prevalent among HIV type 1-infected individuals, whereas our understanding of the critical components of disease pathogenesis, such as virus evolution and adaptation, remains limited. Building upon earlier findings of specific viral subpopulations in the brain, we present novel yet fundamental results concerning the evolutionary patterns driving this phenomenon in two well-characterized animal models of neuroAIDS and provide insight into the timing of entry of virus into the brain and selective pressure associated with viral adaptation to this particular microenvironment. Such knowledge is invaluable for therapeutic strategies designed to slow or even prevent neurocognitive impairment associated with AIDS.

INTRODUCTION

Human immunodeficiency virus type 1 (HIV-1)-associated neurocognitive disorders (HAND) remain a persistent complication despite the success of highly active antiretroviral therapy (HAART) in prolonging the progression to AIDS in HIV-1-infected patients (1). Although the post-HAART era has witnessed a decrease in the prevalence of the most severe form of HAND, HIV-associated dementia (HAD), an increase in prevalence has been observed for the milder forms, referred to as asymptomatic neurocognitive impairment (ANI) and mild neurocognitive disorder (MND) (2, 3). Moreover, these assessments have been reported to be an underestimation of the true prevalence, based on a study in which approximately 64% of HIV-1-infected patients classified as “noncomplainers,” self-reporting no symptoms of cognitive impairment, tested positive for HAND using comprehensive neurocognitive testing criteria (4). Furthermore, with the life expectancy for patients receiving HAART now projected to be greater than 70 years (5, 6), the reported prevalence estimations can only be expected to increase, whereas our understanding of the critical components of disease pathogenesis, such as the ability of the virus to adapt to the microenvironment of the brain, remains limited.

Several studies have implicated viral evolution as a major proponent of HAND progression. Increased viral diversity and rapid progression to AIDS have been linked with the development of the neurological complications of HIV infection (neuroAIDS) in HIV-1-infected patients (7). Furthermore, phylogenetically distinct viral subpopulations, or sequence compartmentalization, within the brain has been reported to be associated with neuroAIDS in HIV-1-infected patients and rhesus macaques infected with neurovirulent virus, although this association has been debated (reviewed in reference 8). Characterized by the significant clustering of sequences within a phylogenetic tree according to anatomical location, viral compartmentalization is indicative of either a single introduction of virus into that particular compartment followed by long-term, isolated viral replication, or multiple introductions of distinct viral subpopulations whose ultimate survival demands specific viral genotypic/phenotypic features required for both entry into and replication in the unique cell types found within that compartment. The former scenario offers validity for modeling brain sequence compartmentalization through the limited transport of molecules across the blood-brain barrier (BBB), which separates the brain from circulating virus (reviewed in reference 9). However, recent phylogenetic and immunological studies have suggested continuous infiltration of the brain by “Trojan Horse” cells (10–12). In this alternative scenario, compartmentalization of virus in the brain is likely the result of successful entry and replication of a viral variant that has passed the selective qualifications associated with the unique microenvironment of the central nervous system (CNS) via evolution in similar peripheral cell types. Extensive evolutionary analysis is needed to determine this distinction; however, the ethical limitation of HIV-1-associated neuropathological studies in humans renders this type of analysis difficult.

As in several other diseases, sampling limitation has created a need for animal models that accurately reproduce pathogenesis. Infection of rhesus macaques with pathogenic simian immunodeficiency virus (SIV) clones and/or viral swarms is a widely used model for HIV/AIDS (13) as well as neuroAIDS (14). SIV-infected macaques exhibit clinical manifestations similar to those of HIV-1-infected humans, albeit on a shorter time scale of approximately 1 to 3 years (13), and the incidence (<25%) of SIV-associated encephalitis (SIVE), the pathological hallmark of neuroAIDS, is similar to that of untreated seropositive patients (<30%) (15). However, low incidence, extreme maintenance cost, and disease timeline associated with the macaque model limits its usefulness in terms of producing rapid and statistically robust results. Such limitations have led to the development of rapid disease models (16–18) that utilize antibody-mediated depletion of the CD8⁺ lymphocyte arm of the antiviral response (17, 19, 20), or a combinatorial approach to infection using neurovirulent and immunosuppressive virus (21), resulting in an increased incidence (>75%) of SIVE (17, 20, 21). In this study, we have used phylogenetic analysis of viral sequences derived from a variety of tissues/cell types, including the meninges and distinct CNS cortical regions, for both the CD8⁺ lymphocyte-depleted and non-CD8⁺ lymphocyte-depleted, or naturally progressing, macaques in order to characterize the evolutionary history of compartmentalized virus in the brains of animals that have developed SIVE despite different immunological backgrounds.

MATERIALS AND METHODS

Study population.

Six CD8⁺ lymphocyte-depleted (D01 to D06) and 12 naturally progressing (N01 to N12) rhesus macaques of Indian origin between 4 and 11 years of age were intravenously inoculated with the well-characterized SIVmac251 (1 ng SIV p27) viral swarm (22, 23) and are hereafter referred to as the Mac251-DEP and Mac251-NP cohorts, respectively. Procedures involving the Mac251-DEP animals were performed with the approval of Tulane University's Institutional Animal Care and Use Committee. For additional information on the treatment and handling of macaques in this cohort as well as gross pathology, see the study of Strickland et al. (24). Procedures on macaques with an intact immune system (Mac251-NP), which were allowed to progress naturally to simian AIDS (SAIDS), were conducted according to the standards of the American Association for Accreditation of Laboratory Animal Care and IACUC protocol 04802, and treatment of these animals was in accordance with the Guide for the Care and Use of Laboratory Animals (25). Further detailed information on the handling and supervisional guidelines for these animals can be found in Lamers et al. (26). Plasma viral loads were monitored as previously described by quantitative PCR (qPCR) methods targeting a conserved sequence in gag (27, 28). All possible measures were taken to minimize discomfort of the animals, and the guidelines for humane euthanasia of rhesus macaques were followed. Animals D01, D02, N06, and N07 (two macaques from each cohort) were euthanized at ∼21 days postinfection (dpi) in order to evaluate early evolutionary events leading to brain infection. The remaining animals were euthanized at the onset of SAIDS between 75 and 118 dpi (Mac251-DEP) and 204 and 373 dpi (Mac251-NP), which was confirmed at autopsy (for a brief description of gross pathology reports for the Mac251-NP macaque cohort, refer to Table 1). All animals were perfused with heparinized saline prior to euthanasia to clear the CNS vasculature, and SIV-associated encephalitis (SIVE) was determined in an autopsy by a veterinary pathologist who was blind to rhesus macaque conditions based on the presence of microglial nodules and multinucleated giant cells and confirmed by immunohistochemistry staining for SIV p27, as previously described (15, 29–31). As SIVE is associated with rapid disease progression (15), data are reported in this study for 6 of the 10 longitudinally monitored macaques based on a relatively rapid temporal progression to SAIDS (204 to 373 dpi).

TABLE 1.

Gross pathology of SIVmac251-infected naturally progressing rhesus macaques

Macaque	Diagnosis or outcome	Symptom(s) or gross pathologya	Age (yr)
			At time of infection	At necropsy
N02	SAIDS	Lymphopenia	5.4	6
		Opportunistic infection (primarily lungs)
N04	SAIDS	Opportunistic infection (pancreas and lungs)	5.4	6.1
N05	SAIDS	Multiple opportunistic infections in lungs	6.4	7.4
N06	Euthanized early	No opportunistic infections	4.4	4.4
		Mild inflammation and satellite cell proliferation in DRG consistent with SIV neuropathy
N07	Euthanized early	No opportunistic infections	4.4	4.4
		Variable DRG inflammation
N09	SAIDS	Lymphoma	4.6	5.4
		Arteriopathy
		Myocardial degeneration and fibrosis
		Gliosis
N10	SAIDS/SIVE	Severe meningoencephalitis	7.7	8.3
		Opportunistic infections, including CMV
N12	SAIDS	Moderate endocardial fibrosis	4.2	4.9
		Severe enteritis and colitis
		Gliosis
		Opportunistic infection

^aDRG, dorsal root ganglia.

RNA and DNA extraction.

Viral RNA (and integrated viral DNA when available) was isolated from the SIVmac251 inoculum and cell-free plasma samples (140 to 280 μl) using the QIAamp viral RNA minikit (catalog no. 52904; Qiagen) following the manufacturer's protocol except that two 50-μl final elutions using buffer AVE were performed. Total RNA and integrated DNA, or genomic DNA (gDNA), were isolated from brain tissue (30 to 50 ng), bone marrow aspirates, and sorted bronchoalveolar lavage (BAL) fluid macrophages and peripheral CD3⁺ T lymphocytes and CD14⁺ monocytes (sorted as described in reference 32) using the AllPrep DNA/RNA minikit (catalog no. 80204; Qiagen) according to the manufacturer's guidelines except for two 50-μl final elutions using RNase-free water. Sequences from plasma, bone marrow, and sorted cell samples were obtained for two or three time points, ∼21 days postinfection, ∼60 dpi, and necropsy, from four Mac251-DEP macaques and four time points, ∼21 dpi, ∼90 dpi, ∼180 dpi, and necropsy, from six Mac251-NP macaques. Samples from the meninges and three distinct cortical regions (parietal, frontal, and temporal cortex) were also obtained at necropsy for sequencing when available. The 100-μl final volume of RNA generated from each method was concentrated using RNeasy MinElute cleanup kit (catalog no. 74204; Qiagen).

cDNA synthesis.

cDNA was generated from the RNA of each sample using the SuperScript III first-strand synthesis system (catalog no. 18080-051; Invitrogen Life Technologies). Viral RNA isolated from both the inoculating viral swarm and plasma samples were reverse transcribed using the provided oligo(dT)₂₀ primer. Total RNA from peripheral CD3⁺ and CD14⁺ cells, BAL fluid, and bone marrow aspirate samples was reverse transcribed using the KOUTR primer (5′-TGTAATAAATCCCTTCCAGTCCCCCC-3′) (9479 to 9454 bp of SIVmac239) (33), whereas total RNA from brain tissue and meninges was reverse transcribed using the cULTR primer (5′-ATGGCAGCTTTATTGAAGAGG-3′) (10125 to 10145 bp of SIVmac239) (26), a primer that binds in the SIV 3′ long terminal repeat (3′ LTR) and 5′ LTR regions. A modified cDNA synthesis protocol was followed for the inoculating viral swarm, plasma, and brain tissue RNA to maximize the length of the cDNA (26). Total RNA from the remaining tissues/cell populations was reverse transcribed according to the manufacturer's recommendations as follows: RNA was incubated at 65°C for 5 min with deoxynucleoside triphosphates (dNTPs) (0.5 mM [each]) and 5 μM KOUTR and then cooled quickly to 4°C. First-strand cDNA synthesis was performed in a 40-μl reaction volume containing 1× reverse transcription buffer (10 mM Tris-HCl [pH 8.4], 25 mM KCl), 5 mM MgCl₂, 10 mM dithiothreitol, 2 U/μl of RNase-OUT (RNase inhibitor), and 10 U/μl SuperScript III reverse transcriptase (RT). The reaction mixture was heated to 50°C for 60 min and then 85°C for 5 min. The reaction mixture was cooled to 37°C, and 0.1 U/μl of Escherichia coli RNase H was added, followed by a 20-min incubation. All cDNA was stored at −20°C until needed.

Single-genome sequencing.

Viral sequences derived from the Mac251-DEP cohort were obtained using bulk PCR and cloning methods (10, 24). Due to the potential limitations of clonal analysis, a modified single-genome sequencing protocol based on previously published methods (34) was used for all samples obtained from the Mac251-NP cohort as well as the frontal lobes from the brains of macaques in the Mac251-DEP cohort. The sequencing protocol used for the remaining samples obtained from the Mac251-DEP cohort has been previously described (10). cDNA was serially diluted until an average of 30% or less of the nested PCRs were positive. During the first round of PCR, diluted cDNA was amplified in 25-μl reaction mixtures containing 1× Platinum blue PCR supermix (Invitrogen Life Technologies) and 0.2 μM each primer: SOUTF (5′-GGCTAAGGCTAATACATCTTCTGCATC-3′) and NOUTR (5′-TTTAAGCAAGCAAGCGTGGAG-3′) (coordinates 6565 to 6591 and 10102 to 10122 of SIVmac239, respectively). In the first round of PCR, the following cycling parameters were used: (i) 95°C for 5 min; (ii) 35 cycles, with 1 cycle consisting of 94°C for 1 min, 58°C for 1 min, and 72°C for 4 min; (iii) 72°C for 10 min. If template cDNA was present, a 3.5-kb product was expected containing the complete env and nef gene products. Second-round gp120 PCR amplification consisted of 2 μl of the first-round PCR added to a 23-μl second-round reaction mixture consisting of 1× Platinum blue PCR supermix (Invitrogen Life Technologies) and 0.2 μM each primer: SINF (5′-GTAAGTATGGGATGTCTTGGGAATCAG-3′) and SINR (5′-GACCCCTCTTTTATTTCTTGAGGTGCC-3′) (coordinates 6598 to 6624 and 8158 to 8184 of SIVmac239, respectively). In the second round of PCR, the following cycling parameters were as follows: (i) 95°C for 5 min; (ii) 35 cycles, with 1 cycle consisting of 94°C for 1 min, 58°C for 1 min, and 72°C for 2 min; (iii) 72°C for 10 min. This second-round PCR generates a 1.5-kb product when positive, containing the entire envelope gp120 sequence with flanking sequence on each end. Second-round gp120 PCRs were visualized on 1% agarose gels stained with ethidium bromide, and reaction mixtures containing single, 1.5-kb products were considered positive and selected for sequencing. The SOUTF, SINF, and SINR primers are based on published oligonucleotide sequences (35), while the remaining primers were designed using Primer3 (36), observing regions of conservation in alignments of published SIVmac251 sequences downloaded from the Los Alamos HIV Sequence Database (http://www.hiv.lanl.gov).

RNA and DNA extractions, cDNA synthesis, and first-round PCR set-up were always conducted in a restricted-access amplicon-free room with separate air-handling and laboratory equipment where no amplified PCR products or recombinant cloned plasmids are allowed and work surfaces and equipment are thoroughly cleaned before and after use with Eliminase (Decon Labs, Inc.). Sequencing was performed on an Applied Biosystems 3730xl DNA analyzer (Life Technologies) at the University of Florida Interdisciplinary Center for Biotechnology Research (UF ICBR).

Sequence alignment.

Sequence alignment was performed as previously described (26). Briefly, following sequence assembly in Geneious R7 (available from BioMatters, Auckland, New Zealand), sequences (SIVmac239 coordinates 6706 to 8049) were aligned using the Clustal X algorithm (37) implemented in BioEdit (38; available from http://www.mbio.ncsu.edu/bioedit/bioedit.html); the alignment was further modified by a manual optimization protocol taking into account conserved glycosylation motifs (39). The highly variable region within the V1 domain was removed so as not to confound the genetic analysis. Intrahost recombinants were determined as discussed below and removed prior to sequence analysis. Approximately 20 sequences per tissue per time point were obtained after removal of potential recombinants (Table 2).

TABLE 2.

Spectrum of tissues and cell populations sampled from the rhesus macaques

^a Dashed outlines indicate animals sacrificed early, for which postmortem represents the only sampling time point (21 days postinfection).

Viral diversity and divergence estimates.

Estimates of mean pairwise viral diversity within individual tissues/cell populations, as well as mean divergence of longitudinally sampled sequences from the inoculating viral swarm were calculated in MEGA v5.2.2 (40; available from http://www.megasoftware.net) using the maximum composite likelihood model of nucleotide substitution (41) and 1,000 bootstrap replicates. Due to the influence of the bulk PCR/cloning method on sequence heterogeneity (34), mutated sites representing <1% of observed point mutations (estimated PCR error rate) were removed from the final Mac251-DEP alignment, as described by Strickland et al. (22). Furthermore, after removal of these sites, the overall viral diversity and divergence within tissue/cell populations during early and late infection were compared for the two cohorts (data available upon request) to determine whether differences between the two animal models could be explained by the sequencing methodology.

Compartmentalization analysis.

Compartmentalization analyses results were obtained from two separate compartmentalization tests in HyPhy (42; available from http://hyphy.org/) in order to evaluate the extent of distinct viral subpopulations within individual tissues. Analyses included both tree- and distance-based methods. Tree correlation coefficients were calculated based on the number of branches (r_b) or branch length (r) separating sequences within separate defined compartments (43) in maximum likelihood trees generated using the generalized time reversible (GTR) nucleotide substitution model with gamma-distributed rate variation across sites (GTR + G). Tree reconstruction was performed using RAxML v8.1.23 (44) using 1,000 bootstrap replicates. Statistical significance was determined using a null distribution of permutated sequences (1,000 permutations), whereby a P value of ≤0.5 was considered significant. The Simmonds association index (SAI or AI) was determined for sequence alignments using SIVmac239 as a reference sequence. The SAI represents the mean ratio of 100 bootstrap replicates of the association value, calculated from the test sequences, to those of 10 sample-reassigned controls. The association value (d) is defined as d = (1 − f)/(2n − 1), where n is the number of sequences below the node and f is the frequency of most common sample type (45). Bootstrapping (1,000 replicates) was used as a test of significant compartmentalization, for which support of >80% was considered significant. A more thorough description of such tests can be found in reference 46.

Selection analysis.

Because statistical measures of metapopulation structure can be affected by selection as well as migration dynamics, an unrestricted branch site random-effects model, referred to as BUSTED (Branch-Site Unrestricted Statistical Test for Episodic Diversification; implemented in the datamonkey webserver http://datamonkey.org), was used to test for gene-wide episodic diversifying selection (47). The analysis was restricted to only internal branches, which are assumed to capture at least one round of virus replication, to mitigate the biasing effects of transient deleterious mutations on the ratio of nonsynonymous to synonymous substitution rates estimates along terminal branches, where selection has not had time to fully filter such population level variation (48, 49).

In addition, nucleotide sequence alignments for all tissues/cell populations at individual time points for each macaque were used to determine site-specific selection over the course of infection for both macaque cohorts. The fast, unconstrained Bayesian approximation for inferring selection (FUBAR) model (50; implemented in http://datamonkey.org) was used to identify potential individual amino acid sites under selection within viral gp120 sequences for individual macaques as well as for individual tissues/cell types at each time point. Sites with a posterior probability of >0.9 of an increased (diversifying) or decreased (purifying) rate of nonsynonymous relative to synonymous substitutions were considered to have experienced a significant level of selective pressure. Macaques were then classified according to SIVE diagnosis or early sacrifice in order to determine similarities and differences among classifications across macaque cohorts.

Recombination analysis.

Recombination analysis was carried out for sequence populations in each tissue and in each macaque using three methods, as previously described (51): (i) GARD (52), (ii) an algorithm that uses split decomposition networks in combination with the “phi-test” (53, 54) implemented in Splitstree4 software (55) calibrated specifically to identify likely recombinants within intrahost viral populations (53, 56, 57), and (iii) Recombine (58). Recombinant sequences were also identified using bootscanning, implemented in Simplot (59) and RDP (60), with observed trends similar to those obtained with the algorithms listed above (data not shown). Putative recombinant sequences were screened against viral swarm sequences in order to determine whether they had originated from the inoculum.

Viral RNA and DNA quantification.

Viral RNA and gDNA levels were estimated for longitudinally sampled and postmortem tissues/cell populations of Mac251-NP macaques and postmortem frontal lobes for Mac251-DEP animals. Viral gp120 RNA was quantified using QUALITY (61) based on limiting dilution PCR-positive amplifications (data available upon request). For time points for which only one dilution was used, which cannot be used for copy number estimation in QUALITY, the following calculation was employed: −ln(1 − proportion of successful PCR wells) × (dilution/PCR template volume [2 μl]), assuming a Poisson probability distribution of RNA templates within PCR samples. Integrated viral DNA was quantified using qPCR amplification of Gag p27 using the recommended concentrations of TaqMan universal master mix II (Invitrogen Life Technologies) and probe and the 0.9 μM concentration of primers used by Hofmann-Lehmann et al. (62). Briefly described here, the TaqMan probe (SIVmac239 coordinates 1487 to 1506) consisted of 5′-TGTCCACCTGCCATTAAGCCCGA-3′, whereas the primer sequences were as follows: 5′-GCAGAGGAGGAAATTACCCAGTAC-3′ (sense; SIVmac239 coordinates 1441 to 1464) and 5′-CAATTTTACCCAGGCATTTAATGTT-3′ (antisense; SIVmac239 coordinates 1532 to 1508). Using the Applied Biosystems 7500 FAST PCR system, the cycling parameters were as follows: (i) initial denaturation (95°C for 10 min); (ii) 35 to 40 cycles of 95°C for 15 s; (iii) 60°C for 1 min. Cell-associated viral sequence copy numbers were represented in terms of copies/10,000 cells based on the stock concentration of gDNA and the presumed 6.2 pg gDNA/macaque cell: (viral copy number/ng gDNA) × (62 ng gDNA/10,000 cells).

Nucleotide sequence accession numbers.

Sequences were deposited in GenBank (accession numbers JF764947 to JF766081 [Mac251-DEP], KR998525 to KR999900 [Mac251-NP N02, N10, and inoculating viral swarm], KX068225 to KX068624 [Mac251-NP N04-N07, N09, and N12], and KX081185 to KX082629 [remainder of the Mac251-NP cohort] ).

RESULTS

SIV longitudinal sampling and disease progression in two macaque cohorts.

Viral sequences from the two SIVmac251-infected macaque cohorts (Mac251-DEP and Mac251-NP) were successfully obtained from various tissues/cell populations sampled over time, consisting of plasma, bone marrow, elicited bronchoalveolar lavage (BAL) fluid macrophages, and fluorescence-activated cell-sorted (FACsorted) CD3⁺ T lymphocytes and CD14⁺ monocytes from blood (Table 2). Samples from meninges and from the frontal, temporal, and parietal cortexes were also obtained at necropsy in addition to peripheral tissues/cell populations.

As expected based on previously reported incidence, three (D03 to D05) of the four longitudinally monitored Mac251-DEP animals and one (N10) of the 12 animals in the Mac251-NP cohort were diagnosed with SIV encephalitis (SIVE) during autopsy, as assessed by a veterinary pathologist who was blind to rhesus macaque conditions. The remaining Mac251-DEP macaque, D06, was determined not to have SIVE encephalitis (SIVnoE) but still presented with meningitis, which has often been lumped together with SIVE in the spectrum of SIV-associated CNS neuropathology. The Mac251-NP cohort was thus fundamental to this study, as macaque N02 exhibited no degree of CNS neuropathology (SIVE or meningitis) but still presented with sufficient viral RNA in the brain for sequence analysis (∼20 sequences/lobe), allowing for the unique opportunity to assess evolutionary patterns along this neuropathological spectrum.

Prior to evolutionary analysis, patterns in viral load and CD4⁺ T-cell counts were investigated in order to rule out possible confounding factors responsible for differences in neuropathological disease progression (29). Despite inherent immunological differences due to CD8⁺ lymphocyte depletion of the Mac251-DEP animals, differences in the magnitude of plasma virus at baseline (0 dpi), early infection (7 dpi), peak viremia (8 to 14 dpi), and set point (35 to 60 dpi), were not specific to the cohort or SIVE diagnosis (Fig. 1). Similar observations were reported for differences in CD4⁺ T-cell nadir and counts over time (Fig. 1).

FIG 1.

Viral load and CD4⁺ T-cell population characteristics for naturally progressing (left) and CD8⁺ lymphocyte-depleted (right) macaques. Viral load was determined as previously described by quantitative PCR methods targeting a conserved sequence in gag (27, 28). CD4⁺ T-cell counts were determined using flow cytometry, as previously described (32). (a) CD4⁺ T-cell nadir (lowest reported CD4⁺ T-cell count) and viral load at baseline (0 to 7 days postinfection [dpi]), initial peak (8 to 14 dpi), and set point (35 dpi for naturally progressing macaques and 60 dpi for CD8⁺ lymphocyte-depleted macaques) for both macaque cohorts. (b) Viral load over the course of infection. (c) CD4⁺ T-cell counts over the course of infection.

Sequence compartmentalization and subcompartmentalization occur solely within the brains of macaques with SIVE and are independent of CD8⁺ lymphocyte depletion.

Maximum likelihood (ML) trees for the Mac251-DEP animals have been reported elsewhere (10) but have been included with those inferred for the Mac251-NP cohort using a similar color scheme for easy comparison (Fig. 2). Upon preliminary visualization of the ML trees, brain sequences appeared monophyletic for N10 (SIVE) and D04 and D05 (SIVE), whereas brain sequence monophyly was not present for the macaque with meningitis (D06) or the macaque with no CNS neuropathology (N02). Further subcompartmentalization of sequences belonging to particular cranial lobes was also observed within both cohorts. This phenomenon could not be explained entirely by clonal expansion, as the largest proportion of identical sequences observed was 50% (D05 temporal lobe; Table 3). Results of the quantitative analysis using tree correlation coefficients (TCCs) have been published previously by Strickland et al. (10) for the Mac251-DEP animals and have been reported again here for easy comparison (Table 4). Additional analysis using the Simmonds association index (SAI) was performed in this study to incorporate uncertainty in the ML tree topology (Table 4). Significant brain sequence compartmentalization, as indicated by positive results for both quantitative methods (P value of ≤ 0.05 for TCCs and AI of <0.33 for SAI), confirmed the tree visualizations and was determined to be significantly associated with SIVE diagnosis using Fisher's exact test (P value of <0.05). Subcompartmentalization of sequences within individual cranial lobes was assessed using the more sensitive SAI compartmentalization test alone and was observed for all macaques with SIVE (SIVE macaques), with the exception of D03. This exception was likely due to the relatively short life span of this animal (75 dpi), as the extent of brain sequence compartmentalization has been reported to correlate with length of time of infection (63–65).

FIG 2.

Maximum likelihood (ML) phylogenetic tree reconstruction for each macaque with detectable viral RNA in the brain. Mac251-NP (N02, N09, and N10) and Mac251-DEP (D03 to D06) macaques were sampled longitudinally prior to necropsy, whereas D01 and D02 (boxed) were sacrificed at 21 days postinfection (dpi). Trees for D01 to D06 are adapted from Strickland et al. (10), and taxa are colored according to tissue sampling origin with respect to peripheral and cranial lobe locations. For all unrooted ML trees, the GTR + G model of nucleotide substitution was used in RAxML with bootstrapping analysis (1,000 replicates).

TABLE 3.

Percentage of total brain tissue viral RNA sequences within individual CD8⁺ lymphocyte-depleted and naturally progressing macaques that share 100% sequence identity

^a ES, early sacrifice. Macaques D01 and D02 were sacrificed at 21 dpi.

TABLE 4.

Analysis of tissue-dependent compartmentalization of viral RNA sequences

^a The designated brain tissue consisted of sequences from the parietal, frontal, and temporal lobes when applicable, whereas peripheral tissues includes sorted, circulating CD3⁺ T lymphocytes and CD14⁺ monocytes, BAL macrophages, bone marrow, and plasma. Note that sequences could not be obtained from the temporal lobe of macaque N09.

^a Tree correlation coefficients (TCC) were calculated based on the number of branches (r_b) or branch length (r) separating sequences within separate defined compartments. TCC data are taken from reference 10 and reported here for comparison. Statistical significance was determined using a null distribution of permutated sequences (1,000 permutations). A P value of ≤0.05 was considered significant.

^c Simmonds association index (SAI or AI) represents the mean ratio of 100 bootstrap replicates of the association value, calculated from the test sequences, to those of 10 sample-reassigned controls. The association value (d) is defined as d = (1 − f)/(2n −1), where n is the number of sequences below the node and f is the frequency of most common sample type. AI values of of <0.33 are shown in boldface type, indicating significant compartmentalization. The bootstrap support (BS) for AI values is also provided. SIVmac239 was used as a reference sequence for SAI determination. BS of >80 was considered significant.

^d ES, early sacrifice. The animals were sacrificed at 21 dpi.

*P value of <0.05 for Fisher's exact test of association of SIVE with compartmentalization.

The results described above are consistent with previous reports of association between brain sequence compartmentalization and neuroAIDS (reviewed in reference 8). Furthermore, although only one macaque within the Mac251-NP cohort was diagnosed with SIVE, it is important to note that this macaque (N10) exhibited the highest degree of overall brain compartmentalization (Table 4) and was highly subcompartmentalized, according to the SAI (Table 4). Thus, even if both models appear to be appropriate for further investigation of this evolutionary phenomenon, increased resolution is observed in the naturally progressing macaques, which is likely due to their longer life spans, as has been previously suggested (63–65).

Late entry of a uniform viral population is implicated in the contribution to brain sequence compartmentalization in macaques with SIVE.

Because virus can penetrate the macaque brain early on during infection (as early as 21 dpi in the Mac251-DEP animals), we sought to determine whether this early entry event is implicated in evolution of virus in the brain and the significant compartmentalization of viral sequences obtained from necropsy samples. Pairwise genetic distances estimated for tissue-specific sequences relative to other sequences within the given tissue (diversity) or to the viral swarm at the time of inoculation (divergence) often reveal information regarding evolutionary processes such as timing of a migration event and natural selection leading up to clustering patterns such as compartmentalization, given assumptions about the initial virus population size. For example, in the case of early infection of the brain by few viral variants followed by limited genetic drift, lower levels of sequence diversity within the brain and sequence divergence from the inoculating viral swarm would be expected relative to that of peripheral sequences sampled contemporaneously. Alternatively, late entry of a relatively larger viral population size that has acquired specific substitution(s) critical for migration across the BBB would be characterized by similar divergence but reduced diversity compared to that of peripheral sequences. We hypothesized that the distinct viral subpopulation within the brains of the SIVE animals, resulting in significant compartmentalization, was indeed due to significant differences in diversity and/or divergence, as described in each of these scenarios. Reduced brain sequence diversity and/or divergence relative to peripheral sequences would similarly be expected when comparing SIVE macaque brain virus to that of the SIVnoE animals, for which compartmentalization was not observed, and hence, differences between brain and peripheral sequence diversity and/or divergence in these animals would also not be expected. For this study, necropsy-sampled brain sequence diversity and divergence from the inoculating viral swarm were compared to those of necropsy-sampled peripheral tissues/cell populations for both macaques diagnosed with SIVE and macaques with sufficient (N09 excluded) detected virus in the brain but no SIVE (Fig. 3). Because meninges have been shown previously to harbor sequences genetically similar to virus both in the brain and periphery (66), as well as distinct from the brain parenchyma (67), sequences from this location were considered separately.

FIG 3.

Genetic distance calculations for viral sequences sampled at necropsy from longitudinally sampled, naturally progressing macaques with sufficiently detectable virus in the brain. (a to d) Viral diversity (a and c) and divergence (b and d) were estimated in nucleotide substitutions/site for gp120 sequences collected at necropsy, classified as peripheral tissues/cell populations, meninges, or brain, and were averaged for the naturally progressing macaques N10 and N02 (a and b) and CD8⁺ lymphocyte-depleted macaques D03, D04, D05, and D06 (c and d). Estimations were calculated in MEGA using the maximum composite likelihood model of nucleotide substitution and 1,000 bootstrap replicates. Error bars represent standard errors. Dashed lines represent average diversity and divergence for sequences from plasma/peripheral cell populations and brain and are colored according to SIVE (black) or SIVnoE (gray) diagnosis. The SIVE category includes macaques N10 (a and b) and D03, D04, and D05 (c and d), whereas the SIVnoE category includes macaques N02 (a and b) and D06 (c and d). Black asterisks indicate intrahost statistical significance for N10 (intrahost significant differences were not observed for N02), whereas dollar signs indicate intercategory (SIVE versus SIVnoE) statistical significance for plasma/peripheral cell populations, meninges, and brain sequences. Mean values that are statistically significantly different by a two-tailed Student's t test indicated as follows: *, P < 0.05; ***, P < 0.001; $, P < 0.05; $$, P < 0.01; $$$, P < 0.001. The standard deviation was calculated using the standard error and square root of the total number of sequences per population. NA, not available.

Significant differences were observed between peripheral tissues/cell populations, meninges, and brain sequence diversities for macaque N10 (SIVE), whereby brain sequences within the distinct cortical regions exhibited significantly lower diversity than in meninges (P < 0.05), which, in turn, was significantly lower (P < 0.001) than combined peripheral tissues/cell populations (Fig. 3a). However, a significant difference in divergence was not observed for any of these three anatomical classifications within this macaque (Fig. 3b). Importantly, significant differences in both diversity and divergence between these three tissue compartments were not observed for N02 (Fig. 3a and b), the Mac251-NP macaque with detectable virus in the brain but no pathological markers of SIVE. Indeed, viral diversity and divergence were also significantly lower in the brain (P < 0.001) and meninges (P < 0.01) of macaque N10 versus N02 (Fig. 3a). The difference in diversity was not due to relatively low diversity for N10 sequences, as the difference in viral diversity between the peripheral tissues/cell populations of these two macaques was not significant (Fig. 3a); however, the difference in divergence may have been due to relatively lower N10 sequence divergence from the viral swarm, as the difference in viral divergence between the peripheral tissues/cell populations of these two macaques was approaching significance (P = 0.09) (Fig. 3b). Similar results were observed for viral diversity within the Mac251-DEP cohort (Fig. 3c), wherein brain sequence diversity for macaques D03, D04, and D05 (SIVE) was significantly lower (P < 0.01) than that of D06 (SIVnoE). Unlike N10 in the Mac251-NP cohort, however, brain sequence diversity for D03, D04, and D05 was not significantly lower than that of peripheral tissues/cell populations, although the mean pairwise genetic difference (nucleotide substitutions/site) was almost 1.5-fold, suggesting that, in both models, reduced viral genetic diversity in the brain relative to that of peripheral tissues/cells likely contributes to brain sequence compartmentalization in SIVE macaques. These also exhibited similar divergence estimates between peripheral tissues/cells and the brain (Fig. 3d). Importantly, similar to Mac251-NP animal N02 (SIVnoE), neither brain sequence diversity nor divergence estimates for Mac251-DEP animal D06 (SIVnoE) significantly differed from that of peripheral sequences.

In order to rule out episodic diversifying selection just prior to brain entry as a driving factor for the observation of brain sequence compartmentalization, an unrestricted branch site random-effects model (BUSTED) was used, revealing evidence of transient diversifying selection events along each of the Mac251-DEP and Mac251-NP macaque phylogenies (likelihood ratio test [LRT] P values = 0.0), irrespective of SIVE status. The minimum number of amino acid site changes (SIVmac239 gp120 E78N and S128L/I) (belonging to Mac251-DEP macaque D04) identified by this method were shared among all macaques, with the exception of S128L/I for early sacrificed Mac251-NP macaque N07. These sites were mapped onto individual macaque phylogenies and corresponded to either internal branches that did not give rise to brain sequence lineages or did not directly separate brain sequence clades from sequences belonging to other tissues/cell types (Fig. 4), indicating that these diversification events were not responsible for the observed brain sequence compartmentalization.

FIG 4.

Mapped amino acid changes contributing to episodic diversifying selection along internal branches of representative phylogenies. An unrestricted branch site random-effects model (BUSTED) implemented in the datamonkey webserver was used to test for gene-wide episodic diversifying selection (47) restricted to only internal branches. Amino acid changes at the minimum number of sites contributing to statistically elevated diversifying selection, which were shared by all macaques, were mapped onto the maximum likelihood phylogenetic trees. The phylogenies for two macaques have been chosen as representatives of the two distinct macaque cohorts, macaques D04 and N10, both of which were diagnosed with SIVE. Each of the two amino acid positions (78 and 128, SIVmac239 coordinates) has been represented on separate phylogenies for each macaque, wherein branches are colored according to the amino acid at that position.

When taking into consideration differences in brain sequence evolution alone as well as brain sequences relative to peripheral sequence evolution between SIVE and SIVnoE macaques, the results suggest that, despite early brain infection in the Mac251-DEP cohort, compartmentalization of brain sequences in SIVE animals is likely the result of tissue-specific purifying selection of a viral subpopulation that has acquired, through evolution in the periphery, a more readily neuroadaptive genotype/phenotype. Given the similar divergence estimates between brain and peripheral sequences in the SIVE animals, this neuroadaptive phenotype is likely associated with facilitated entry during late infection, although sufficient time may not have passed after this entry event to observe replicative advantages in the evolutionary history, as would be characterized by the eventual reduction in divergence.

Differences in purifying selective pressure may contribute to the evolution of neurovirulent strains and entry of virus into the brains of macaques with SIVE.

We next sought to determine whether differences in selective pressure over time could be found among the macaques diagnosed with SIVE or without SIVE and whether particular regions of the SIV genome were affected more than others (Fig. 5a to c). Macaques were classified as SIVE, SIVnoE, or early sacrificed (ultimate neuropathological outcome unknown), and sites experiencing selective pressure that were shared among at least two macaques within each category, when available, were plotted along the length of the amino acid sequence. Macaques N10 (SIVE) and D06 (SIVnoE) were an exception, as they were the only macaques in their respective categories within their cohort. A high concentration of sites under selection was observed in the first half of the gene for the Mac251-DEP macaques at ∼60 dpi and the last time point (Fig. 5b), which was similar to the selection pattern observed at ∼180 dpi for the Mac251-NP macaques (Fig. 5a). There was also an increase in the number of sites experiencing purifying selection over time for the constant region C2 within both cohorts, particularly for SIVE macaques. Because this is a heavily glycosylated region, these sites were investigated for the presence of asparagine residues prior to the last time point. Although none of the sites corresponded to previously known or potential glycosylation sites, evidence suggests a role for increased C2 site conservation in the ability of the virus to either enter or replicate in the brain, and this should be considered for further investigation with respect to its role in this process. An increase in the number of sites within pV1-V2 regions (pV1 refers to partial V1) under selection was also observed within both cohorts over time for SIVE macaques; however, this trend was observed for SIVnoE macaques as well, suggesting generalized immune-driven evolution. Across the gp120 gene in general, a larger difference was observed in the proportion of sites experiencing either purifying or diversifying selection between the SIVnoE and SIVE macaques within the Mac251-NP cohort compared with the Mac251-DEP animals (Fig. 5c). Considering the facts that the SIVnoE macaque within the Mac251-DEP cohort was diagnosed with meningitis whereas the SIVnoE macaque within the Mac251-NP cohort exhibited no degree of CNS neuropathology, the results indicate that selective pressures are relatively similar for virus subpopulations responsible for meningitis and SIVE, which may explain the similar brain sequence divergence between the two groups (SIVE and SIVnoE) within the Mac251-DEP cohort, but not the Mac251-NP cohort.

FIG 5.

Site-specific selection over time for SIV sequences within naturally progressing and CD8⁺ lymphocyte-depleted macaques diagnosed with SIVE or without SIVE. The fast, unconstrained Bayesian approximation for inferring selection (FUBAR) model, implemented in HyPhy, was used to measure selection at individual sites within viral gp120 sequences for individual macaques at each sampled time point over the course of infection. (a and b) Naturally progressing and CD8⁺ lymphocyte-depleted macaques were classified according to SIVE diagnosis or early sacrifice, and amino acid sites with a posterior probability of >0.9 of diversifying (+) or purifying (−) selection that were shared among macaques within each category are labeled along the length of the study sequence and colored according to their respective macaque classification (SIVE, SIVnoE, or early sacrifice [Early Sac]). Labels are stacked so as to highlight sites under selection that are present in more than one macaque group. Boxed regions correspond to variable loops V1 to V4, with only partial V1 (pV1) available in the alignment. The last time point is an average of sampling times among all macaques prior to euthanization. The data in panels a and b have been quantified in panel c in order to highlight the differences in the number of sites experiencing selection over time for each cohort and classification.

Different levels of recombination do not readily distinguish macaques based on SIVE status.

In order to determine whether differences in recombination patterns could be observed between the macaques diagnosed with SIVE or without SIVE, percent recombination was compared for macaques N10 (SIVE) and N02 (SIVnoE) within the Mac251-NP cohort (Fig. 6a) and macaques D03, D04, and D05 (SIVE) and D06 (SIVnoE) within the Mac251-DEP cohort (Fig. 6b). Although differences were not significant, the Mac251-NP macaque with SIVE, N10, appeared to have relatively low levels of recombinant sequences in each of the distinct cortical regions examined within the brain, CD14⁺ monocytes and CD3⁺ lymphocytes, plasma, and bone marrow compared to the macaque with no CNS neuropathology (N02). In contrast, virus from the Mac251-DEP macaques diagnosed with SIVE (D03, D04, and D05) exhibited similar levels of recombination in the brain and even greater levels in peripheral tissues/cell populations compared to the macaque with meningitis (D06).

FIG 6.

Percent recombination in necropsy-sampled sequences obtained from individual tissues/cell populations in naturally progressing and CD8⁺ lymphocyte-depleted macaques. (a and b) The percentage of total sequences obtained from individual tissues/cell populations at necropsy containing recombination breakpoints were classified as peripheral tissues/cell populations or brain and were averaged for the naturally progressing macaques N10 and N02 (a) and CD8⁺ lymphocyte-depleted macaques D03, D04, D05, and D06 (b). Dashed lines represent average percent recombination for sequences from plasma/peripheral cell populations and brain and are colored according to SIVE and SIVnoE diagnosis. The SIVE category includes macaques N10 (a) and D03, D04, and D05 (b), whereas the SIVnoE category includes macaques N02 (a) and D06 (b). Significance was determined using Student's t test.

Viral genomic RNA and DNA levels in the brain readily distinguish macaques based on SIVE status.

A similar number of brain sequences obtained from macaques N10, N02, and D03 to D06 were used for sequence and phylogenetic analyses (Table 2); however, the question remained as to whether or not differences in degrees of viral production in the brain would be observed in SIVE and SIVnoE macaques with detectable viral genetic material in the brain in both cohorts, including the Mac251-NP macaque N09, for which relatively low levels of viral RNA were amplified (Table 2). The difficulty in obtaining sequences for N09 was well represented in the limiting dilution PCR-derived viral RNA quantifications, as less than one viral RNA copy per 10,000 cells was estimated to be present (Table 5). Although not statistically significant, viral RNA levels appeared highly elevated in the frontal lobes of the Mac251-DEP macaques compared with those of the naturally progressing macaques (Table 5). Not surprisingly, however, sections of the frontal lobes from the macaques with SIVE from both cohorts contained greater levels of viral RNA compared with the macaques with detectable viral RNA but no SIVE, including the Mac251-DEP macaque (D06) with meningitis (Table 5).

TABLE 5.

Quantification of viral genomic RNA and integrated DNA in longitudinally sampled plasma and peripheral cell populations

^a RNA/DNA are presented in number of copies/10,000 cells. Viral RNA was quantified using QUALITY based on limiting dilution PCR-positive amplifications. Integrated viral DNA was quantified using qPCR amplification of Gag within cellular genomic DNA (gDNA).

^b Light gray cells in the table indicate time points for which only one dilution was used, which cannot be used for copy number estimation in QUALITY. For these quantifications (N), the following calculation was employed: −ln (1 − proportion of successful PCR wells) × (dilution/PCR template volume [2 μl]), assuming a Poisson probability distribution of RNA templates within PCR samples.

^c Viral DNA below qPCR detectable levels are indicated by values based on loaded gDNA quantity (50 or 100 ng) by the white cells in the table.

^d Dark gray cells indicate absence of sampling.

^e Dashed outlines indicate animals sacrificed early, for which postmortem represents the only sampling time point (∼3 weeks postinfection).

^f The length of infection period, or time of postmortem sampling, is reported in days postinfection (DPI).

In order to determine whether viral RNA was a true representation of the presence of virus, integrated viral DNA was quantified using qPCR, including DNA from the frontal lobes of the Mac251-DEP animals. The Mac251-NP macaques for which no viral RNA sequences were obtained from brain tissue using single-genome sequencing also did not present detectable integrated viral DNA and vice versa, suggesting that viral RNA was indicative of viral integration into cells of the brain and that the virus in the brain during the late stage of infection was replicating and not likely dormant, although this may not be the case for earlier virus sampled at 21 dpi for the CD8⁺ lymphocyte-depleted macaques. Unlike RNA levels, viral DNA levels were elevated in the frontal lobes of the Mac251-NP macaques compared with those of the Mac251-DEP macaques, which was primarily attributed to levels reported for N10, as SIV DNA levels were below the limit of detection for N02 and N09 (Table 5). Again, not surprisingly, sections of the frontal lobes from the macaques with SIVE from both cohorts contained greater levels of viral DNA compared with the macaques with detectable viral RNA but no SIVE, consistent with previous studies (68) as well as with the hypothesis presented in this study of late viral entry of a larger viral subpopulation characterized by selection on entry.

With the exception of brain tissue, the presence of viral RNA was not representative of DNA integration in all tissues/cell populations (Table 5). Several instances of absent viral RNA for the Mac251-NP macaques were coupled with the presence of detectable viral DNA, consistent with virus in the CNS with no productive infection. There were also several cases of viral RNA for which viral DNA was not readily detected, predominantly in peripheral CD14⁺ monocytes and bone marrow, but never in CD3⁺ lymphocytes. This observation is consistent with inefficient integration, or the presence of viral episomal DNA, which has been previously noted (69). However, the limit of detection for qPCR to 35 copies per reaction cannot be ignored, considering also the low level of viral RNA in the majority of these cell populations.

Surprisingly, this phenomenon was not the case for virus in the brain, despite results from a previous study by Pang et al. (70) revealing 6- to 81-fold greater levels of unintegrated versus integrated viral DNA in the brains of HIV-1-infected patients diagnosed with HAD. The method used for genomic DNA isolation is able to detect smaller fragments of DNA, 15 to 30 kb in length; however, SIV episomal DNA may still go undetected and requires further investigation. With respect to plasma and peripheral cell populations, ranges of infection as indicated by viral RNA and/or DNA quantification also differed drastically not only between macaques within the Mac251-NP macaque cohort but also between different time points within the same macaque, especially in the case of BAL fluid macrophage viral RNA, for which the range was between approximately 250,000 to 570,000 copies/10,000 cells among the macaques at 6 months postinfection.

DISCUSSION

Significant compartmentalization of viral sequences found primarily within the brains of macaques with SIVE in this study, regardless of immune modulation via CD8⁺ lymphocyte depletion during the acute stage of infection, provides further evidence in support of the link between viral evolution and neuroAIDS. Phylogenetic detection of this event following postmortem sampling suggests either a single, early introduction of a small virus subpopulation into the brain followed by limited genetic drift due to the isolated nature of the brain parenchyma or multiple introductions, whereby specific viral genotypic/phenotypic features are advantageous during entry, ultimately resulting in a bottleneck effect and thus reduced genetic diversity. Although molecular biology techniques are required to definitively determine whether the same virus entering the brain during early infection does not, in fact, give rise to the population observed at necropsy, the results of the present study point to the evolutionary achievement of neurotropism in the periphery prior to late viral entry into the brain, whereby neurotropism is characterized by a metastable genotype/phenotype conferring the ability to enter and replicate efficiently in brain-specific cells, such as perivascular macrophages (71). Rational inferences from these results also suggest that such a phenotype may require preservation of specific amino acids within the C2 region of gp120, although this observation has limited support given the small number of macaques in this study diagnosed with SIVE and without SIVE in the naturally and rapidly progressing cohorts, respectively.

Our results have also indicated that selective pressures are relatively similar for virus subpopulations associated with SIV-associated meningitis and encephalitis, which may be explained by macrophage tropism, as the primary cell types infected in the brain and meninges are macrophages (71). Although distinct genotypic signatures of macrophage tropism have not been definitively associated with CNS neuropathology, a viral genotype conferring increased replication in this particular cell type may facilitate entry into and replication in macrophage-like cells within the brain, ultimately paving the way for neuroadaptation and the emergence of a neurovirulent viral strain, assuming that the entering virus is not already neurotoxic.

Numerous studies have highlighted the importance of understanding viral evolution in the context of neuropathological disease progression, yet additional factors, such as age and coinfection, should not be ruled out. For example, both the Mac251-NP macaque with SIVE (N10) and Mac251-DEP macaque diagnosed with severe SIVE were the oldest macaques at the time of infection, approximately 8 and 11 years of age, respectively, whereas the average ages at time of infection were approximately 5 years for the Mac251-NP macaques and 8 years for the Mac251-DEP macaques (information available upon request; 10). Several HIV-1 studies have found an association between age and neuroAIDS status, with an increased incidence of HAND in individuals greater than 50 years of age (72–74). However, the translation to rhesus macaques is unclear, as neuroAIDS status in human subjects was determined based on neurocognitive testing, rather than neuropathological and histological markers. In addition, macaque N10 also presented with detectable cytomegalovirus (CMV) in the brain. Progression to AIDS is often accompanied by infection with opportunistic pathogens, and CMV is among a subset of pathogens able to enter the brain and potentially lead to neuronal dysregulation (75). However, the qualitative and quantitative results of our study suggest that SIV infection, and not CMV, was the driving factor in the development of SIVE. Overall, our data indicate that despite multiple introductions of virus into the brain over the course of infection, brain sequence compartmentalization in macaques with SIV-associated CNS neuropathology is likely the result of late entry of virus that has acquired through evolution in the periphery sufficient adaptation for the distinct microenvironment of the CNS. Further studies investigating the specific peripheral source (e.g., cell population) of the evolutionary phenomenon leading up to this late entry event would not only improve our understanding of the driving force behind neuroAIDS but aid in the targeted prevention of colonization of the CNS by neurovirulent HIV strains.