Highly Uneven Distribution of Tenofovir-Selected Simian Immunodeficiency Virus in Different Anatomical Sites of Rhesus Macaques

Abstract

Antiviral tenofovir monotherapy was used to determine whether drug-selected simian immunodeficiency virus (SIV) variants replaced their wild-type progenitors at the same rate in different tissues of six rhesus macaques. The relative frequencies of drug-resistant and wild-type genotypes were measured longitudinally in blood and in 23 lymphoid and nonlymphoid tissues collected at necropsy. The mutant/wild-type genotype ratio was measured using a heteroduplex tracking assay targeting tenofovir-selected SIV reverse transcriptase codons. After the initiation of tenofovir treatment in animals with high steady-state viremia levels, resistant genotypes emerged in the plasma within 1 to 8 weeks and in five of six cases reached frequencies of nearly 100% within 4 to 25 weeks. The appearance of tenofovir-resistant genotypes in peripheral blood mononuclear cell (PBMC) DNA was generally delayed by 1 to 2 weeks and in one case was completely absent. Necropsies performed 8 to 55 weeks after the initiation of tenofovir treatment showed the frequency of resistant SIV genotypes to be generally higher in tissue RNA than DNA fractions. The frequency of drug-resistant genotypes varied widely between anatomical sites, including different lymph nodes of the same animal. Except for the epidydimis, the tissues with the lowest rates of proviral replacement by tenofovir-resistant genotypes differed between animals. The highly uneven distribution of tenofovir-resistant genotypes in different tissues seen shortly after the initiation of tenofovir monotherapy may reflect differences in local antiviral drug selection pressures and/or the stochastic effect of small effective populations of drug-resistant variants randomly seeding different anatomical sites early in therapy.

Human immunodeficiency virus (HIV) and simian immunodeficiency virus (SIV) RNA and DNA can be detected in many lymphoid and nonlymphoid anatomical compartments of their hosts (25, 28, 30, 31, 36, 40, 44, 47, 60). HIV and SIV proviruses in tissues may reflect the presence of circulating or resident infected T cells (46, 47), infected tissue macrophages (15, 28), and/or other infected cell types (2, 20, 36, 62). Lentiviral quasispecies at different anatomical sites frequently differ at the envelope or pol locus (3, 10, 12, 21, 26, 29, 36, 43, 56, 63). Whether such uneven distribution of sequence variants reflects viral adaptation to different local conditions and/or viral colonization founder effects remains unclear.

The effectiveness of antiviral drugs may vary widely in different tissues, depending on tissue and cell penetration and drug activation (50). Low-level viral replication in a drug sanctuary site where antiviral activity is reduced may go undetected, if it does not result in measurable plasma viremia. Following years of suppression of detectable plasma viremia by use of antiviral drug combinations, plasma viremia will rapidly rebound following drug discontinuation (42). The source of such virus includes latently infected quiescent memory CD4⁺ cells that are reactivated (16, 17). Genetic differences between the rebounding plasma virus after drug discontinuation and viruses grown out from in vitro-activated quiescent memory cells indicate that another source(s) of virus may also seed the rebounding viremia (6).

The antiretroviral drug tenofovir (9-[(R)-2-(phosphonomethoxy)propyl]adenine, or PMPA) selects for SIV variants with fivefold-reduced in vitro drug susceptibility carrying a reverse transcriptase (RT) K65R mutation. In tenofovir-treated SIV-infected macaques, the K65R mutation is sometimes followed by additional RT mutations (K64R, N69S/T, I118V, and S211N) which may serve as compensatory mutations to increase replication fitness (51, 52, 55, 59). Tenofovir disoproxil fumarate (TDF), the oral prodrug of tenofovir approved for the treatment of HIV type 1 (HIV-1) in humans, also selects for the K65R mutation in HIV-1. While mutations selected by other RT inhibitors have been shown to provide some level of cross-resistance to TDF (23), the K65R mutation is the only HIV-1 RT mutation reported to be selected by TDF in vivo (35).

For this study, we used the emergence of K65R resistance mutations as a genetic marker to measure the rate of replacement of wild-type SIV variants with newly selected resistant genotypes. We compared the mutant/wild-type genotype ratios in different tissues with the aim of identifying anatomical sites with the slowest rates of viral turnover to drug-resistant genotypes.

MATERIALS AND METHODS

Animals.

Six juvenile male rhesus macaques (2 to 3 years old) from the type D retrovirus-free colony at the University of California—Davis California National Primate Research Center were inoculated intravenously under ketamine anesthesia with 100 50% tissue culture infective doses (TCID₅₀) of a highly virulent SIV_mac251-5/98 isolate (38). This virus stock, which was propagated on rhesus peripheral blood mononuclear cells (PBMC), has a titer of 10⁵ TCID₅₀ and 1.4 × 10⁹ RNA copies per ml and is highly pathogenic for infant and adult macaques (38, 54). Tenofovir from Gilead Sciences (Foster City, Calif.) was prepared as described previously (53). Daily tenofovir treatment (30 mg/kg of body weight, administered subcutaneously) was started at week 9 postinfection. Blood samples were collected before virus inoculation and at weekly intervals during the first 2 months postinfection and then were collected biweekly and monthly to monitor plasma viral loads (VLs) and the emergence of tenofovir drug resistance mutations. Animals were sacrificed with an intravenous overdose of pentobarbital (60 mg/kg). For reduced contamination with blood of the organs taken at necropsy, several liters of 0.9% NaCl (Irrigation USP; B. Braun Medical, Inc.) was infused into the left ventricle and ascending aorta until only clear fluid exited the severed right atrium. A complete necropsy examination was performed on all animals, and a routine histopathologic examination was done on collected tissues. The animals were treated in accordance with American Association for Accreditation of Laboratory Animal Care standards. The study received University of California—Davis and University of California—San Francisco Institutional Animal Care and Use Committee approvals.

Cell preparation.

PBMC and plasma were obtained from EDTA-preserved blood samples. After the collection of plasma, the blood was centrifuged with standard lymphocyte separation medium (Ficoll-Hypaque; Amersham Pharmacia Biotech, AB, Uppsala, Sweden) at 2,000 rpm in a Beckman Acuspin centrifuge at 20°C for 40 min. Cells were then rinsed twice (2,000 rpm, 20°C, 10 min) in phosphate-buffered saline (Ca²⁺ and Mg²⁺ free), resuspended in freezing medium (90% fetal calf serum, 10% dimethyl sulfoxide), and cryopreserved in liquid nitrogen or frozen as a dry cell pellet. Spleen and bone marrow suspensions were homogenized by passage through a 70-μm-mesh cell strainer before centrifugation with Ficoll-Hypaque by the same procedure as that for whole blood. Cells extracted from lymph nodes (LNs) as well as from bronchoalveolar lavage were rinsed in phosphate-buffered saline. For DNA extraction, dissected tissues obtained at necropsy were snap-frozen in dry ice and stored at −80°C. For RNA extraction, dissected tissues were stored in RNAlater solution at −20°C (Ambion Inc., Austin, Tex.).

Nucleic acid isolation, cDNA synthesis, and PCR.

SIV RNA was isolated from 1.5-ml plasma aliquots spun for 1 min at 13,000 rpm (Eppendorf centrifuge; Brinkmann Instruments Inc., Westburg, N.Y.) to remove cell debris. Supernatants were then centrifuged at 17,000 rpm at 4°C for 1 h. The plasma (140 μl) remaining at the bottom of the tube was used to resuspend the pelleted virus, and SIV RNA was purified by using a QIAamp viral RNA kit (Qiagen, Valencia, Calif.). Total cellular DNA and RNA were purified by using QIAamp DNA and RNeasy kits, respectively (Qiagen). Tissues were first homogenized by grinding in disposable tissue grinders (The Kendall Company, Mansfield, Mass.) before DNA and RNA extractions. Forty units of RNase inhibitor (Roche Diagnostic Corporation, Indianapolis, Ind.) was immediately added to each tube of purified RNA. Ten microliters of plasma RNA (1/4 the total volume) from the viral RNA purification column was used for first-strand cDNA synthesis with 1.5 μg of random primers and murine leukemia virus (RNase H⁻) RT (Gibco Life Technologies, Rockville, Md.) in a final volume of 20 μl. Each PCR was initiated with 0.5 μg of genomic DNA or 5 μl of viral cDNA. PCRs were performed with a 12.5 μM concentration of each primer, 2.5 mM MgCl₂, a 200 μM concentration of each deoxynucleoside triphosphate, and 1 U of Taq DNA polymerase in a final volume of 40 μl (1st round) or 50 μl (2nd round). Reactions were incubated in a PCRExpress Thermocycler (Hybaid, Franklin, Calif.). Cellular RNA was negative for SIV by nested PCR (nPCR) unless it was first reverse transcribed, indicating the absence of contaminating genomic DNA. By using plasmid dilutions, we showed that the nPCR protocol could amplify between 1 and 10 copies of SIV proviral DNA. The first round of PCR consisted of 40 cycles, with a 50°C annealing temperature, and used primers RT-F1A (5′-GAAGCAGTGGCCATTATCAAAAG-3′) and RT-F1B (5′-GAGGTATGGAGAAATATGCATCACC-3′), which generated a fragment of 296 bp. Five microliters of first-round PCR products was used for the second-round PCR, which consisted of 35 cycles, with a 57°C annealing temperature, and used primers SM1A (5′-GTCAGTTGGAGGAAGCTCCC-3′) and SM2A (5′-GGTGTGGTATTCCTAATTGG-3′), generating a 156-bp fragment (SIV SMM239 positions 2985 to 3140) including RT codons 52 to 90.

Plasma VL quantification.

SIV_mac251 RNA was quantified in duplicate by using a real-time RT-PCR assay as previously described (1). The RNA copy number was determined by comparison with an external standard curve consisting of in vitro transcripts representing bases 211 to 2101 of the SIV_mac239 genome. The assay has a sensitivity of ≤160 RNA copies per ml of plasma and a linear dynamic range from 10² to 10⁸ copies/ml (R² = 0.99).

Mutation detection assay.

RT mutations located at and near codon 65 were quantified by using a modification of the universal heteroduplex generator (UHG) concept (45, 64). A 5′ Cy5 fluorescently labeled probe with a 6-bp deletion of codons 64 and 65 was designed to drive a heteroduplex tracking assay (HTA) (8, 9, 11). The probe was obtained by a standard site-directed mutagenesis procedure and was cloned into a TOPO-TA vector (Invitrogen, Carlsbad, Calif.) to produce plasmid pΔ6. To obtain a Cy5-labeled UHG-HTA probe, we amplified 50 ng of pΔ6 by using 2nd round primers SM1A and SM2A-Cy5. The Cy5-labeled Δ6 PCR product was then annealed to unlabeled nPCR products derived from biological samples to generate Cy5-labeled DNA heteroduplexes. Two microliters of the Cy-5-labeled PCR product probe and 10 μl of target PCR samples were therefore mixed, denatured for 2 min at 95°C, and reannealed on ice before being separated in a 12% nondenaturing polyacrylamide gel (29:1 acrylamide-bisacrylamide; Bio-Rad, Hercules, Calif.) in 1× Tris-borate-EDTA for 5.5 h at a constant 235 V. Known sequence variants and the viral inoculum were tested in every HTA gel as electrophoretic mobility standards. Cy5 fluorescence was scanned with the red laser of a Storm 860 phosphorimager (Molecular Dynamics, Sunnyvale, Calif.) and analyzed with ImageQuant 5.1 software (Molecular Dynamics). The gels were then stained with a 0.5-μg/ml solution of ethidium bromide, and UV illumination photographs were taken with a charge-coupled device camera. The mutant/wild-type ratio (reported as percent mutant) was determined for each sample by scanning the intensities of the Cy5-labeled heteroduplex bands exhibiting different mobilities and calculating their contributions to the total DNA heteroduplex Cy5 signal. N69S and K64R mutations were also observed, but in every case they were seen in the context of a K65R mutation. Therefore, the percentage of drug-resistant genotype values reported includes K65R, K65R+N69S, and K64R+K65R+N69S mutants.

DNA heteroduplex sequencing.

A small gel piece (with an approximately 10-μl volume) containing a DNA heteroduplex of interest was obtained under UV light by using a 1,000-μl pipette tip. The acrylamide gel plug was then ejected into PCR tubes and crushed with a pipette tip, and its DNA was reamplified by using second-round PCR primers. The resulting PCR fragments were then purified and directly sequenced, using primer SM2A and an ABI 3700 automatic sequencer. The Δ6 UHG-HTA probe sequence was then subtracted from the mixed sequence electropherogram to derive the sequence of the annealed SIV variant.

RESULTS

UHG-HTA.

We used a modification of HTA based on the ability of universal heteroduplex generators (UHGs) to enhance the detection of single-base-pair substitutions located within a targeted region of interest (45, 64). A Cy5 fluorescently labeled UHG-HTA probe was designed with codons 64 and 65 deleted (see Materials and Methods), which allowed us to distinguish, through their different electrophoretic mobilities, SIV sequence variants in RT codons 64 and 65 and up to 12 bp away in codon 69 (Fig. 1). After the formation of DNA heteroduplexes between the 150-bp Cy5-labeled UHG-HTA probe and 156-bp PCR products from biological samples, the percentages of different codon 64, 65, and 69 sequence variants in PCR products could be determined by the relative intensities of their labeled DNA heteroduplexes with different mobilities. In Fig. 1B, heteroduplex bands of cloned SIV variants that differ by one or more substitutions can be seen to migrate differently through polyacrylamide gels. Such differential electrophoretic mobilities are due to conformational differences in the looped-out region of the DNA double helices caused by the 6-bp deletion in the UHG-HTA probe (Fig. 1A).

FIG. 1.

UHG-HTA. (A) Schematic description of UHG-HTA. (B) UHG-HTA showing the different mobilities of codon 64, 65, and 69 sequence variant heteroduplexes. The three variants in the inoculum are shown together with the sequences and mobilities of subcloned sequence variants detected in the inoculum and after tenofovir selection.

Three RT variants (A, B, and C) could be detected in the SIV_mac251-5/98 inoculum by using this method (Fig. 1B). These variants were isolated by subcloning and were sequenced (Fig. 1B). The three variants differed at two positions, none of which resulted in an amino acid change (i.e., nonsynonymous mutations at codons 62 and 66) (Fig. 1B). The dominant A variant had the same sequence in that region of RT as SIV_mac1A11 cloned from a SIV_mac251 isolate (37).

DNA heteroduplex sequencing.

The identities of sequence variants in distinct DNA heteroduplex bands were also confirmed by directly extracting DNA heteroduplexes from ethidium bromide-stained and UV-illuminated polyacrylamide gels, reamplifying the small quantities of DNA in gel plugs by PCR, and directly sequencing the resulting amplicons (see Materials and Methods). Mixed-sequence electropherograms consisting of two sequences were obtained (i.e., the UHG-HTA probe sequence plus that of the annealed RT variant together making up the heteroduplex) (data not shown). The sequence of the reannealed variant in the heteroduplex was then derived by subtracting the known sequence of the UHG-HTA probe from the mixed sequence.

Measurement of mutant frequencies.

For each sample analyzed, the mutant/wild-type ratio was determined by scanning the intensities of the Cy5-labeled heteroduplex bands as shown in Fig. 2A (also see Materials and Methods). Band fluorescence percentages derived from two independently generated nPCRs (initiated with the same cDNA or genomic DNA) were averaged to derive mutant frequencies. Reconstitution products from mixtures of mutant and wild-type PCR products were annealed to the UHG-HTA probe, and the fluorescence intensities of the resulting bands were measured. Figure 2B shows a plot of the observed versus known frequencies of input variants, showing that the values obtained were close to the known input percentages. Prior HTA mixing experiments have shown a similar linearity between variant frequency and the HTA gel signal (8, 45).

FIG. 2.

Example of UHG-HTA, with DNA band quantitation and pattern reproducibility. (A) UHG-HTA patterns for the SIV plasma quasispecies of macaque 246 showing the transition from a wild-type inoculum-like population at 10 wTFR to a mixed population at 14 wTFR and then a fully drug-resistant population at 19 wTFR. The mobility of the inoculum and the 251-A-K65R variants are used for reference. The percentage of the total heteroduplex fluorescent signal for each labeled heteroduplex band and their averages are shown below. The reproducibility of independent sampling can be seen in the similarity of duplicate nPCRs (lanes 1 and 2) and their quantitation. (B) The percentages of 251-A-K65R in reconstituted PCR mixtures with 251-A were measured by UHG-HTA and plotted against known percentages. The averages of percentage values from the same dilutions analyzed in two HTA gels were plotted.

Validation of reproducible quasispecies sampling.

Population genetic analysis of complex quasispecies requires that a sufficient number of sequence variants be amplified to accurately reflect the actual diversity of the quasispecies. Amplifying a representative number of genomes is especially difficult when analyzing low-copy-number and genetically complex quasispecies (9, 13, 19). To ensure that artifactual mutant frequencies were not generated as a result of insufficient population sampling, we set a stringent quality control criterion. Each sample was analyzed by use of two independently generated PCR products (i.e., products from reactions initiated with different aliquots of the same cDNA or genomic DNA). When the UHG-HTA band pattern was similar for both duplicate PCRs by both visual inspection and fluorescence scanning, the population sampling was considered appropriate. Examples of the reproducibility of the UHG-HTA patterns and of their quantitation are shown in Fig. 2A and 4. When the UHG-HTA patterns were different for the duplicate PCRs, the sampling was considered nonreproducible and therefore insufficient to accurately reflect the actual quasispecies diversity. Sampling was then increased by doubling the cDNA or DNA input, and the nPCR was repeated. Only mutant frequencies derived from reproducible samplings are reported.

FIG. 4.

Longitudinal UHG-HTA analysis of SIV genotypes in plasma and PBMC DNA of macaque 565. The rapid emergence of drug-resistant genotypes in plasma and their delayed and only partial appearance in PBMC DNA are shown. Weeks are labeled starting at the initiation of tenofovir at 9 weeks postinfection.

Longitudinal analysis of tenofovir-resistant genotypes in blood.

Six juvenile macaques were intravenously inoculated with 100 TCID₅₀ of a highly virulent stock of SIV_mac251. VLs were measured at weekly and then monthly intervals (Fig. 3). An initial peak of viremia was detected at 2 weeks postinfection for five of six animals, followed by the establishment of different levels of viremia by 9 weeks postinfection. Prior to the initiation of tenofovir treatment, the SIV quasispecies diversity in vivo in the region of RT being analyzed was highly stable and remained identical to that seen in the inoculum (i.e., with all three RT variants [A, B, and C] coreplicating), except in animal 817, in which the C variant became dominant in the plasma (data not shown). After the initiation of tenofovir treatment at 9 weeks postinfection, the subsequent reductions in plasma viremia were highly variable (Fig. 3). Animals with the highest VLs (animals 205 and 565) showed only a transient or no decrease in plasma viremia. Animals with intermediate VLs (651 and 246) showed transient and ultimately minor (approximately 0.5 to 1 log) decreases in viremia. The animals with the lowest VLs (817 and 803) showed reductions in viremia that at times were below the limits of detection in animal 803 (Fig. 3). Whether all reductions in viremia seen after the initiation of tenofovir treatment were results of antiviral drug activity or were also partially due to developing immune responses is unknown. Within 2 months, the VLs had rebounded to various degrees in all animals.

FIG. 3.

SIV plasma viremia from time of inoculation to necropsy. Tenofovir monotherapy was initiated at 9 weeks postinfection. Numbers correspond to the six animals used in the study.

UHG-HTA was performed with all longitudinally collected plasma and PBMC samples. In five of six animals, the K65R descendant of the major A variant in the plasma quasispecies became the dominant drug-resistant form (Fig. 4). A K65R variant of the C variant was seen only in animal 817. No K65R variants of the minor B variant were observed.

The plasma quasispecies of animal 565, with a steady state VL of approximately 10⁶ RNA copies/ml, showed the most rapid turnover to the K65R genotype (Fig. 4 and 5). The VL fell approximately 1 log for only 1 week. Within 1 week of initiation of tenofovir treatment (1 wTFR), 30% of the plasma RNA consisted of the K65R variant. By 3 wTFR, 95% of the plasma viruses contained the K65R mutation. In contrast, the PBMC DNA showed a 1-week delay in the appearance of the K65R mutation, which by 8 wTFR reached a maximum frequency of only 75% of the proviral population (Fig. 4 and 5). During this period, the K65R frequency in PBMC RNA was even lower than that in the proviral population (Fig. 5).

FIG. 5.

Longitudinal percentages of drug-resistant genotypes in plasma RNA, PBMC DNA, and PBMC RNA. Squares, percentages of K65R mutant in plasma; triangles, percentages of K65R mutant in PBMC DNA; bars, percentages of K65R mutant in PBMC RNA; circles, VLs. Left y axis, plasma VLs in logarithmic scale; right y axis, frequency of K65R-containing SIV variants; x axis, weeks postinfection. Tenofovir treatment was initiated at week 9 postinfection, as shown by dashed vertical lines.

Animal 205, with the highest steady-state VL (approximately 10⁷ RNA copies/ml), showed no decrease in viremia after the initiation of tenofovir treatment. K65R was not detected in the plasma until 4 wTFR (45%), and complete conversion of the viral RNA quasispecies occurred by 14 wTFR. The emergence of the K65R mutation in the proviral quasispecies also showed a delay of 1 week relative to the plasma and reached apparent fixation (100%) by 14 wTFR. For PBMC RNA, the frequency of the K65R mutation went from 0 to 90% within only 1 week at 6 wTFR (Fig. 5).

Animal 651, with a VL of 5 × 10⁵ RNA copies/ml when tenofovir treatment was initiated, showed a transient drop in VL of <1 log followed by a return to baseline by 7 wTFR. The K65R mutation in plasma was first detectable at a frequency of 25% by 2 wTFR and reached 80% by 7 wTFR. Reproducible samplings of the proviral populations were not achieved until 7 wTFR, when it was measured at 38%. The K65R frequency in PBMC RNA closely followed that in PBMC DNA (Fig. 5).

Animal 246, with a starting VL of 1.4 × 10⁵ SIV RNA copies/ml, showed a 1.5-log drop in viremia followed by a modest rebound to 3 × 10⁴ RNA copies/ml within 1 month of initiation of therapy. The emergence of K65R occurred first in plasma RNA at 2 wTFR, and the complete conversion of the viral RNA quasispecies occurred by 10 wTFR. K65R was first detected at 5 wTFR in PBMC DNA. The K65R frequency in PBMC RNA followed that in PBMC DNA (Fig. 5).

Animal 817, with a baseline VL of 1.8 × 10⁴ SIV RNA copies/ml, showed a 2-log drop in viremia to below the detection limit within 1 week of initiation of therapy. By 5 wTFR, viremia was again detectable and had returned to 7 × 10³ by 7 wTFR. K65R was first detectable in plasma at 7 wTFR (8%) and was at 100% frequency only 1 week later. Surprisingly, no K65R mutation was seen in both PBMC DNA and RNA throughout the course of treatment, even while the plasma quasispecies was genetically 100% drug resistant (Fig. 5).

Animal 803 had a VL of 3.7 × 10⁴ RNA copies/ml upon initiation of tenofovir treatment. Viremia was brought to below the detection limit within 2 wTFR but transiently rebounded at 6 wTFR. At 16 wTFR, the VL was again below the detection limit. The VL then showed two low-level viremic episodes until 55 wTFR, when the animal was sacrificed. Due to the low VL, reproducible quasispecies sampling could not be achieved in any blood fraction. Nevertheless, after repetition of the nPCR multiple times, K65R variants were occasionally amplified starting at 45 wTFR for PBMC RNA, 50 wTFR for plasma RNA, and 54 wTFR for PBMC DNA (Fig. 5).

Overall, tenofovir resistance mutations were detected in all five persistently viremic macaques by 8 wTFR (by only 2 wTFR for three of these five animals). In plasma, the K65R mutation reached nearly 100% frequency by 4 to 25 wTFR in all five of the persistently viremic animals. The emergence of K65R in the PBMC DNA fraction was delayed in all five animals, including one (macaque 817) in which it was completely absent. The K65R mutation frequency in PBMC RNA was generally similar to that in PBMC DNA, except for one animal (macaque 205), in which its initial emergence was first seen in PBMC RNA (Fig. 5).

Tissue-associated tenofovir-resistant SIV genotype frequencies.

Animals were sacrificed after 8 to 55 weeks of tenofovir therapy (Fig. 6). The presence of mutant genotypes was then measured in SIV DNA and RNA populations from 23 different tissues collected at necropsy (Fig. 6). Regardless of the length of infection, we were able to amplify SIV RNA and DNA from nearly all tissues tested from most animals, with the frequent exception of some brain tissues (Fig. 6). The K65R frequencies are only shown if they were derived from reproducible population samplings and include all variants containing K65R. The K65R-containing variants that were detected are shown in Fig. 1B (A-K65R+N69S was also detected). Tissues from which reproducible samplings were not possible were particularly numerous in animal 803, which had maintained the lowest plasma VL of all animals. When samples from different anatomical sites of the same animal yielded discordant tenofovir-resistant mutant frequencies, the identities of the variants were further confirmed by purifying and sequencing the DNA heteroduplexes, or when a single UHG-HTA band was detected, by direct PCR sequencing.

FIG. 6.

Percentage of SIV containing K65R in RNA and DNA fractions from different anatomical sites collected at necropsy. Percentage of K65R is in RNA/DNA fractions. NRS, nonreproducible sampling of the quasispecies; NA, nonapplicable; Neg, nPCR negative; and XX, insufficient cells available.

A common observation for the two animals sacrificed the earliest (565 and 246) was a blood-like pattern of K65R emergence, in which a higher frequency of mutants was seen in tissue RNA than in DNA fractions. For animal 565, the lowest percentage of K65R mutation in tissue RNA was measured in the spinal cord (0%), while other tissues contained from 42 to 100% mutant RNA genotypes. The lowest percentage of mutant proviral genotypes was seen in the epididymis (a convoluted tubule in each testis that carries sperm to the vas deferens) (8%), followed by the spleen (9%), thymus (11%), and bone marrow (24%). LNs and gut tissues showed the highest rates of proviral turnover to the mutant genotype, with an average frequency of 62%.

In contrast, in animal 246, which was sacrificed at 11 wTFR, the spinal cord RNA fraction had completely turned over to the mutant. Animal 246 had mild subacute multifocal meningitis at necropsy. Increased cell activation and influx of uninfected target cells may have accelerated productive infection with K65R virus. In animal 246, as in macaque 565, the lowest percentage of mutant proviral genotypes was in the epididymis. Different LNs showed highly variable frequencies of mutant genotypes in both the RNA and DNA fractions (Fig. 6).

Animal 817 (sacrificed at 20 wTFR) had the lowest VL when tenofovir treatment was initiated. After the initial control of viremia, the VL rebounded to low but detectable levels. As was the case for PBMC, many tissues remained entirely wild type in both their RNA and DNA SIV fractions even while the plasma viral population was 100% mutant. The only proviral populations showing the presence of the K65R genotype were two gut tissues (duodenum at 50% and jejunum at 10%) and axillary LNs (27%). The distribution of K65R mutants was also highly variable in the tissue RNA fraction. The RT-PCR-positive central nervous system tissues, the liver, and the lung tissues showed 100% RNA SIV mutant frequencies, while different lymphatic tissues and most LNs showed no detectable tenofovir-resistant RNA genotypes. One notable exception was the axillary LNs, in which the RNA fraction was 100% mutant. It was also noted that, while liver SIV RNA was 100% mutant, liver SIV DNA was 100% wild type (Fig. 6).

Animal 205 maintained the highest VL after the initiation of tenofovir and was sacrificed at 29 wTFR with symptoms of simian AIDS. The histopathological examination at necropsy revealed the inflammation of many organs and the onset of several opportunistic infections. The mesenteric, ileocolic, and submandibulary LNs, the thymus, and the tonsils were severely atrophied, preventing the isolation of sufficient numbers of T cells for SIV nucleic acid analyses. Fewer than 2 × 10⁶ cells were obtained from colonic distal, internal iliac, obturator, inguinal, and axillary LNs. Every tissue tested was populated by 100% mutant SIV variants in both the RNA and DNA fractions (Fig. 6).

Animal 651 maintained a plasma VL of approximately 10⁵ RNA copies/ml throughout 48 weeks of tenofovir therapy. The plasma SIV RNA quasispecies completely turned over to resistant genotypes by 25 wTFR. At necropsy, most analyzed compartments contained 100% tenofovir-resistant genotypes in both the RNA and DNA fractions. The thymus and bone marrow contained the lowest percentages of tenofovir-resistant proviral genotypes (64 and 66%, respectively). Reproducible sampling was not possible for the proviral quasispecies for central nervous system tissues (Fig. 6).

Animal 803 remained on tenofovir therapy for 55 weeks, with only occasionally detectable low-level viremia. Reproducible quasispecies sampling was not possible for the blood compartments as well as for many tissues. The only tissue compartment in which tenofovir-resistant proviral genotypes could be reproducibly measured was the cecum (29% K65R mutants). All other tissues in which reproducible sampling was possible showed 0% tenofovir-resistant proviral genotypes. When measurable, the frequencies of K65R in RNA quasispecies were also 0%, except in the cerebrospinal fluid (CSF) (57% mutant) and a subset of LNs (75 and 69% mutant in the axillary and inguinal LNs, respectively). We therefore observed strongly discordant mutant frequencies between the RNA and DNA fractions as well as between the RNA fractions of different LNs (Fig. 6).

DISCUSSION

Measurements of the frequency of drug-resistant variants by using allele-specific PCR or oligonucleotide probes can be strongly influenced by flanking sequence polymorphisms and insufficient population sampling. The detection of single base substitutions by using a DNA heteroduplex generator HTA probe was previously reported for the HIV protease locus (45). For this study, we used a related approach to detect and quantitate the frequency of sequence variants in a short region of the SIV pol locus. The identities of variants exhibiting different electrophoretic mobilities were also confirmed by sequencing gel-isolated DNA heteroduplexes. Throughout this study, care was taken to ensure that the obtained mutant frequencies were not distorted by insufficient population sampling. We noticed that some samples, particularly those with low SIV copy numbers, resulted in widely different variant frequencies when independently generated PCR products were analyzed. The mutant frequencies reported here were therefore all validated by the generation of identical results for independent sampling sets (8, 9, 13, 19, 34).

As has been observed with anti-HIV-1 monotherapies with lamivudine and protease inhibitors (14, 48), drug-resistant genotypes emerged very quickly in plasma after the start of tenofovir monotherapy in the SIV system. Tenofovir-selected mutations were detected first in the plasma virus population, and after a short delay, in PBMC DNA and RNA. A similar discordance in plasma RNA and PBMC DNA quasispecies has been reported for RT inhibitor-selected HIV resistance mutations (4, 7, 24, 29, 50a, 61). For the five persistently viremic animals, the first time at which K65R plasma viral RNA mutants were detected ranged from 1 to 8 weeks. In animal 205, resistant genotypes had apparently completely replaced wild-type PBMC proviruses by 14 wTFR. It is well established that latently infected human memory CD4⁺ cells can harbor quiescent wild-type HIV-1 proviruses at very low frequencies for extended times (5, 16, 17). While such long-lived latent proviral reservoirs are also anticipated in the PBMC of SIV-infected macaques, their low frequency may have prevented their detection. The rapid turnover of the large majority of wild-type proviruses in PBMC observed for macaque 565 (76% mutant by 8 wTFR) and macaques 205 and 651 (100% mutant by 14 and 25 wTFR, respectively) does indicate that the PBMC proviral compartment can rapidly and almost completely turn over.

The blood proviral quasispecies in animal 817 PBMC behaved very differently from those in the other animals. Even after the plasma virus genotypes had become completely resistant for over 10 weeks, the K65R genotype remained undetectable in PBMC DNA and RNA. The wide difference in mutant frequencies between blood compartments likely reflects the minor contribution that PBMC proviruses made to the synthesis of plasma viruses in this animal. The K65R plasma virus was therefore likely derived from proviral templates in other tissues. Possible sources of mutant plasma viruses in animal 817 include mutant proviruses in axillary LNs, the duodenum, and the jejunum. The CD4⁺ cells in the gut are a major site of viral replication (22, 39, 49, 57, 58). The more rapid genetic turnover of proviruses in the gut tissues of animal 817 may be associated with the gut lymphoid hyperplasia observed in this animal. The 100% mutant virus population in the liver (with 0% mutant proviruses) may reflect the central role of this organ in the clearance of viral particles from the circulation (65).

In some animals tenofovir-selected mutant genotype frequencies measured in 23 different tissues and body fluids showed wide differences with those measured in blood. In two animals (205 and 651), the distribution of tenofovir-resistant genotypes was nearly homogeneous throughout their bodies with almost complete genotypic resistance in both the RNA and DNA fractions from all anatomical sites. These animals had been treated with tenofovir while remaining persistently viremic for the longest times (29 and 48 weeks, respectively). Extended tenofovir selection pressure in the presence of replicating K65R mutants (detected in plasma as early as 2 to 4 weeks after initiation of treatment) likely led to the eventual replacement of nearly all detectable wild-type genomes with tenofovir-resistant genotypes. In persistently viremic animals treated for shorter times (8 to 20 weeks), mixed SIV populations consisting of both wild-type and resistant genotypes were detected in different anatomical sites. Mutant frequencies varied widely within the same animals. For example, in animal 565, the resistant mutant frequencies in the DNA fractions of the liver and epididymis were 84% and 8%, respectively. In animal 817, the axillary LN RNA was 100% mutant while viruses in all other tested LNs were entirely wild type. Unexpectedly, the anatomical sites whose viral variants most rapidly turned over to mutant genotypes differed among animals. For example, at necropsy, the spinal cord viral RNA population of animal 565 remained fully wild type, while it completely converted to a resistant genotype in animal 246. The duodenum SIV DNA population became 92% resistant in animal 565 but only 13% resistant in animal 246.

In conclusion, the present study used a drug resistance mutation marker to study the turnover of viral and proviral populations. Our results indicate that the emergence and dissemination of viral mutants are a highly complex process, with considerable variability among animals and among tissues. The uneven mutant distribution within animals could conceivably reflect different levels of tenofovir activity at different anatomical sites, such that the replication of resistant genotypes was not equally favored in every site. The different mutant frequencies at what are expected to be physiologically very similar anatomical sites in terms of drug activity (i.e., different LNs) indicate that unequal drug pharmacokinetics and activities may not fully account for the uneven mutant distribution that we observed. Furthermore, the disparity seen between animals regarding which sites retained the highest frequencies of wild-type genomes also argues against these results being entirely due to tissue-specific differences in tenofovir selective pressure, since such differences would be expected to be constant between animals. A possible explanation requires that the number of drug-resistant viruses selected by tenofovir very early after the initiation of therapy is low enough that the infection of different anatomical sites with such genotypes is highly variable. Supporting evidence for this model is seen by the changes in envelope sequences in plasma HIV-1 associated with the rapid selection of protease inhibitor-resistant variants (9, 27, 41). An early stochastic distribution of resistant genotypes at different anatomical sites would therefore result in some sites being seeded early with resistant genotypes while other sites would only show the presence of these genotypes at later times. A small effective population size, as hypothesized for HIV-1 in vivo (18, 32, 33), further reduced by selection for drug-resistant mutants could therefore contribute to the highly variable distribution of mutant genotypes seen early after the initiation of therapy.