Selective vertical transmission of HIV-1 antiretroviral resistance mutations

Objective: To examine the patterns of vertical transmission of zidovudine (ZDV) resistance mutations.

Design: HIV-1 reverse transcriptase codons 10-250 were sequenced from 24 pairs of ZDV-exposed women and their HIV-infected infants as part of the Women and Infants Transmission Study.

Methods: Viral RNA was extracted from tissue culture supernatants and sequenced using fluorescent dye-primer chemistry and an automated sequencer.

Results: For 17 of these pairs, maternal and infant sequences were identical to one another and lacking known ZDV resistance mutations. The remaining seven maternal sequences contained known mutations associated with ZDV resistance at reverse transcriptase codons 70, 210, 215 and 219. In each case where the maternal HIV isolate showed a pure mutant species, the infant sequence was identical. When the maternal sequence showed the presence of a sequence mixture at codon 70 or 219, the infant's virus showed only wild-type sequence even when the ZDV-resistant mutant was quantitatively dominant in the mother. The single maternal HIV isolate showing mixed sequence at codon positions 210 and 215 transmitted an unmixed mutant to the infant at both positions. When maternal mixtures were present at sites not associated with ZDV resistance, only the dominant species appeared in the infant.

Conclusions: When maternal HIV isolates contained mixed wild-type and ZDV-resistant subpopulations, only a single component of the mixture could be detected in the infected infants. Resistance mutants without the codon 215 mutation were not transmitted from mixtures, even when the mutants formed the majority of circulating maternal virus. In perinatal HIV transmission, specific ZDV-resistant HIV genotypes circulating in the maternal virus pool may influence whether infection in the infant will be established by a wild-type or ZDV-resistant HIV strain.

Introduction

The spread of an infectious agent within a population involves a balance of selective pressures between optimal growth within an individual host and optimal spread between hosts. For typical pathogens at equilibrium within human populations, these effects are difficult to observe except over very long time periods. Several features of HIV-1 infection, however, make this selection more apparent: its relatively recent introduction into human populations, the chronicity of high-level viral replication within infected individuals [^1-3], the high mutation rate intrinsic to retroviral replication [^4-6], and the complex selection imposed by antiretroviral drugs [^7-10]. In particular, the vertical transmission of HIV-1 from chronically infected mother to newly infected infant reflects the interplay of virological, immunological and pharmacological effects. For vertical transmission, the population of viral sequences are under selection for their ability to grow to high levels within the mother and for their ability to initiate a new infection within the infant. The use of antiretroviral agents during pregnancy adds another level of selection to this process, wherein the ability to grow efficiently in the presence of the drug may select for mutants with altered fitness for spread to the infant.

Most of the currently available antiretroviral agents are associated with the development of particular HIV-1 mutations conferring specific resistance to the individual agent [^11-13]. Zidovudine selects for reverse transcriptase (RT) mutations at codons 41, 67, 70, 210, 215, and 219 both in vivo and in vitro and these mutations display widely varying degrees of zidovudine (ZDV) resistance when expressed in isolation or in combination [^14,15]. Only the codon 70 Lys→Arg and codon 215 Thr→Tyr mutations appear consistently as primary mutations in vivo. Mutations at the other positions appear as compensatory changes, increasing the fitness of the primary mutations under continued selection by prolonged ZDV use. Appearance of antiretroviral resistance has been shown to be associated with worsening clinical course in a number of settings [^16-18].

Zidovudine is the only agent proven to decrease the mother to infant spread of infection [¹⁹]. Despite the modest effects of ZDV monotherapy on viral load and the limited duration of efficacy shown in previous studies, the AIDS Clinical Trials Group (ACTG) 076 trial demonstrated a two-thirds reduction in vertical transmission relative to placebo among mothers treated with ZDV. Multivariate analysis of these data suggest that the prevention of transmission involves effects beyond those accounted for by changes in viral burden alone [²⁰], but the effects of individual ZDV-resistance mutations were not specifically addressed. Current understanding of the kinetics of development of HIV point mutants would suggest that such mutants would be likely to arise within the timeframe of these trials [^5,21,22]. It is possible that the relative fitness of these mutants for transmission and establishment of new infection may account for some of the observed effects of ZDV upon vertical transmission.

We have been following a cohort of ZDV-exposed women as part of the Women and Infants Transmission Study (WITS) in order to determine the effect of antiretroviral resistance mutations upon the vertical transmission of HIV. As part of this study, we have determined detailed HIV RT sequence information for these women and their HIV-infected infants. These data allow an analysis of selective transmission of sequence variants from mother to infant in cases where the maternal sequence shows a mixed population. The transmission of components of a maternal mixture reflects the overall size of the viral inoculum establishing infection in the infant, the relative abundance of subspecies within this inoculum, and the fitness for initiation of new infection of each of these subspecies.

Here we report the analysis of the transmission of individual sequence variants in the women who were shown to harbor ZDV-resistance mutations at the time of delivery and who transmitted HIV to their offspring. The sequencing methodology employed allowed us to assess both the presence or absence of mutations as well as the occurrence of mixed populations at all sites within the region sequenced. Furthermore, the determination of virion RNA sequence from maternal and infant isolates at the time of delivery and amplified by virus coculture, provides a picture of the pool of actively replicating virus at the time of delivery. The direct comparison of the maternal mixtures with the infant-derived HIV-1 virion RNA sequences offers a window into the virological processes operating during the process of vertical transmission.

Methods

Among women enrolled in the National Institutes of Health/National Institute for Allergy and Infectious Diseases/National Institute of Child Health and Human Development/Women and Infants Transmission Study, the subset of women who had received treatment with ZDV during pregnancy were eligible for the current study. Women were enrolled from study sites in New York (Columbia, Brooklyn), Illinois (Chicago), Massachusetts (Boston, Worcester), Texas (Houston), and Puerto Rico (San Juan). For the present study, only the first pregnancy resulting in a live singleton birth was used for analysis. Since the majority of the enrollees predated the ACTG 076 findings supporting ZDV treatment to reduce vertical transmission risk, most of the ZDV treatment was given for therapy of the mothers. However, some of the women enrolled later in the study were also part of the ACTG 076 protocol.

Heparinized maternal and infant blood samples were obtained at the time of delivery and peripheral blood mononuclear cells (PBMC) were isolated via Ficoll-Hypaque centrifugation. Viral cultures were obtained by coculture with fresh, phytohemagglutinin-stimulated donor PBMC as described previously [²³]. Supernatants from these virus cultures were shipped on dry ice to the sequencing site (Boston) for further processing. RNA was isolated from tissue culture supernatants using the Tri-reagent method according to the manufacturer's directions (Molecular Research Center, Inc., Cincinnati, Ohio, USA). Reverse-strand cDNA was generated from this viral template using Moloney murine leukemia virus RT under conditions recommended by the manufacturer (Boehringer, The Roche Group, Basel, Switzerland). PCR was used to amplify a region of the cDNA corresponding to nucleotides 2550-3321 from the HIV-1 nl43 genomic map (Genbank accession number M19921) using AmpliTaq enzyme (Perkin-Elmer/Applied Biosystems, Foster City, California, USA) under recommended conditions. Standard dideoxy-sequencing reactions were carried out using fluorescent dye-labeled primers. Products of the sequencing reaction were then resolved electrophoretically on an Applied Biosystems 7300 Automated Sequencer and analyzed using Sequence Analysis 3.0 software (Applied Biosystems). Sequence chromatograms were inspected visually for presence of ZDV drug-resistance mutations at codons 41, 67, 70, 210, 215 and 219. Each of these sites was categorized as either wild-type, mutant or mixed. Positions where the minor peak was at least 30% of the height of the major peak (25% of total under-peak area) were categorized as mixtures in this analysis. Sequences were also aligned and inspected across the entire sequence for sequence variants and ambiguous positions. In cases where the maternal sequence differed from wild-type, sequencing chromatograms for mother and infant were scanned visually for evidence of possible mixed positions over all regions where double-stranded sequence information was available.

Results

Transmission of pure ZDV-resistance mutants

ZDV-resistance mutations were present in seven (32%) of the 24 maternal isolates. Clinical and virological data for these woman are shown in Table 1. Three mothers showed HIV sequences with chromatographically unmixed ZDV resistance at least one nucleotide position. Each of these transmitted the identical pattern to the infant. One maternally derived HIV-1 RT sequence showed a codon 215 Thr (ACC)→Tyr (TAC) change, the two-nucleotide mutation most highly associated with phenotypic ZDV resistance. The infant sequence was identical to the maternal sequence, including at three silent sites where both diverged from the subtype B consensus.

Table 1

A second pair of maternal-infant HIV isolates displayed an isolated codon 210 mutation, changing the wild-type Leu to Met. This mutation is seen frequently in association with mutation at codon 215 but has not been described as an independent source of ZDV resistance. Several examples of random sequence divergence might have arisen by chance among the four known ancillary ZDV-resistance codons (41, 67, 210 and 215) for the total 143 RT sequences obtained. However, we did not observe such isolated mutations at positions 67, 41 or 219 and did find a second maternal sequence with an isolated 210 mutation (in a mother who did not transmit HIV to her infant).

A third maternal HIV sequence was found to bear an unmixed codon 70 Lys (AAA)→Arg (AGA) mutation, which was transmitted to the infant. This single nucleotide change was the most frequent mutation observed in the overall cohort (56% of all mutations). This particular pair was noteworthy because the maternal HIV isolate was clearly mixed mutant and wildtype at codon 219 as well as unmixed mutant at codon 70, even upon repeat sequencing. Only the wild-type species of this codon 219 mixture appeared in the infant, showing the presence of two cocirculating branches of the maternal HIV quasispecies and the apparent differential transmission of the two mutations. The sequence chomatograms shown in Fig. 1 illustrate this divergence between maternal and infant HIV isolates.

Fig. 1

Transmission of HIV in the presence of maternal mixtures

Isolates from four mothers at the time of delivery demonstrated the presence of mixed sequence of wildt-ype and ZDV-resistant virus. One of the maternal sequences showed a double mixture at both codons 210 and 215 with both wild-type and resistant sequence at each position. Due to the two-nucleotide nature of the codon 215 Thr (ACC)→Tyr (TAC) mutation, it was not possible to distinguish chromatographically a simple mixture of these two species from a more complex mixture containing the intermediate single-nucleotide mutants Ser (TCC) and Asp (AAC). The in vivo prevalence of these intermediate mutations under selection by ZDV is not known with certainty. The infant HIV isolate contained only pure mutant sequence at both positions (Fig. 2). In addition, Fig. 2 indicates where maternal sequence is mixed at two silent third-position sites in codon 206 and 220. At both sites, only the majority species (in each case the HIV B consensus) was transmitted to the infant. Furthermore, the maternal sequence diverges from consensus with an unmixed codon 211 Arg→Lys change transmitted to the infant.

Fig. 2

By contrast, in each of the other maternal mixtures detected among HIV-transmitting women, only the wild-type component was seen in the infant's isolate. One of these was the double-mutant described in the previous section, with an unmixed resistance mutant at codon 70 and a wild-type/mutant mixture at codon 219 (Fig. 1). The other two maternal mixtures were mixed Lys (AAA) and Arg (AGA) at codon 70 and unmixed wild-type at the other five resistance codon positions. In all of these cases, only the wild-type nucleotide sequence appears in the infant. In one case (Fig. 3), the wild-type sequence appeared in the infant even though it was clearly the minor species present in the mother. Unlike the transmission of a pure codon 215 and 210 double-mutant genotype from a maternal mixture (Fig. 2), we observed only transmission of wild-type sequence and failure to transmit the mutant genotype when the maternal sequence mixtures occurred at other resistance codons.

Fig. 3

Transmission of non-resistant genotypes

For the 17 non-resistant maternal-infant pairs, there were approximately 6000 total nucleotide positions of chromatographically readable sequence. Amongst these, there was a total of 75 maternal and 78 infant deviations from the HIV-1 Los Alamos subtype B consensus for an average of 4.5 nucleotide changes per individual or roughly 0.6% per nucleotide pair. Of these sequence variants, 118 were silent third position changes, six were silent first position changes [e.g., Leu (UUA)→Leu (CUA)], 27 were conservative changes, and two were non-conservative changes (codon 30 Lys→Asn in a single maternal-infant pair). Comparing maternal and infant isolates, 74 sequence variants were identical in mother and infant, three (all silent) were present only in the infant isolate, and one conservative change was present only in the mother. There were no positions where polymorphism was observed in more than two pairs of maternal-infant isolates. In both maternal mixtures, the infant HIV sequence showed only the pure B consensus nucleotide at the homologous position. These results are summarized in Table 2.

Table 2

Discussion

This observational study tracked the pattern of vertical HIV-1 transmission of sequence variants and mixtures between ZDV-exposed women and their infants. The data arose via post hoc analysis of a subset of a larger cohort of women and infants and as such were unavoidably limited in both sample size and in the nature of available specimens. These limitations preclude any definitive conclusions regarding the mechanisms of transmission. Nonetheless, the findings are of interest for the insight they provide into the in vivo viral population effects acting during vertical transmission and in the suggestive evidence they provide for the relative fitness for transmission of these resistance mutations.

Another limitation of this study was that the subjects come only from US sites. Since the bulk of vertical transmission occurs in the developing world, effort must be concentrated on understanding and controlling it there. It is uncertain at present how great an effect antiretroviral resistance may eventually play in this setting since limited resources preclude widespread use of antiviral medicines. Even in the developing world, however, there may be cause for concern as black and gray market sources of ZDV find their way to populations desperate for any therapy, potentially leading to the development of resistance in analogous fashion to the history of antibacterials. A troubling possibility would be the advent of significant levels of sporadic ZDV monotherapy in high-risk populations. This would be unlikely to benefit greatly the individuals taking the drug but might potentially compromise the efficacy of perinatal ZDV in blocking vertical transmission. These broader issues lie well beyond the scope of the current study, but the US-based ZDV monother-apy cohort may still hold some relevance in the near future for women from the developing world.

Our sequence data showed a differential transmission of wild-type over mutant virus at positions other than codon 215 that cannot be accounted for by relative abundance alone. In particular, the transmission to the infant of wild-type sequence for each codon 70 mixture even when the Arg-70 mutant was quantitatively dominant may indicate increased fitness of wild-type at this codon either in transmission itself or in the establishment of a new infection in the offspring. Only the codon 215-containing mutant was transmitted to the infant when present as a mutant/wild-type mix. It is impossible to rule out chance effects given the small number of mutant viral isolates, and indeed chance fluctuations may play an important role in the transmission of resistance where the inoculum size is small [²⁴]. Another possibility would be a small decrease in the replicative fitness of the codon 70 mutant whose effects were amplified during the establishment of infection in the fetus. At least in vitro, however, no such difference has been observed in recent measurements [²⁵].

These observations can begin to tease apart the contributions of within-host and between-host selection to the overall population dynamics of HIV evolution. In this setting, the tracking of mixtures offers a window into the virology of transmission by creating a very tightly paired control for multiple viral species under presumably identical in vivo conditions. This sort of analysis should prove useful in understanding the dynamics of viral transmission in a number of settings where the high rate of sequence divergence of HIV can be employed as a probe of the relative fitness of closely related variants.

Another notable feature of the sequence transmission patterns in this population is the absence of transmission to the infant of sequence mixtures even at presumably neutral sites. There might be several trivial explanations for this if the presence of maternal mixtures was so fleeting that they had little chance to be transmitted or if there was some strong selective pressure against one species during the step of culturing the primary isolates. Although we do not have sequential sequences from any maternal transmitters with mixtures present, we have been tracking the viral evolution in the infants as part of a follow-up study and have observed mixtures persisting in the infants at least for several weeks between visits. The presence of sequential infant samples also helps address one potential limitation of this study: the mixing of the infant's birth isolate with maternal HIV. In the cases where the sequence of the infant's initial isolate differs from the maternal isolate, subsequent matching isolates from the infant confirm that it was unlikely to have arisen from contaminating maternal sequences. Furthermore, in each case, mixtures were observed in the maternal sequence but not in the initial infant HIV sequence.

Selection during culture is also unlikely since both maternal and infant virus isolates were cultured under identical conditions and only infant-derived virus showed disappearance of mixtures. More plausibly, the absence of transmission of numerous maternal sequence mixtures to the infant reflects a bottleneck effect wherein the infant infection is derived from a very small number of maternal founder virions and has been observed for envelope sequence variants [²⁶]. This finding may hold implications for the evolution of HIV sequences in the human population as a whole, as it would predict an increased rate of between-host sequence drift by 'founder effect' during transmission in addition to the natural within-host sequence divergence from the intrinsic mutation rate of the virus.

The use of cultured virus rather than direct plasma sequencing further raises the issue of whether HIV sequences derived from these two sources may differ significantly. Only cultured virus was available to us in this study given the very limited supply of plasma samples in the overall WITS population. Proviral DNA may contain HIV sequences from long-lived cells no longer actively transcribed into virion RNA. Cultured virion RNA, however, must come from circulating PBMC actively infected and producing virus. Since these productively infected cells are rapidly turned over in vivo, virus derived from them should track the evolution of the HIV quasispecies closely [^27,28].

An unusual pattern in our sample was the appearance of an isolated resistance mutation at position 210 in two maternal samples and in the infant of one of these mothers. Codon 210 mutants (like mutations at codons 41 and 219) can potentiate the resistance of primary resistance mutations at codon 215. However, expressed in isolation they do not show measurable phenotypic resistance [²⁹] and have not been seen as primary ZDV-resistance mutations in natural populations. This mutation was observed only in a very small number of samples and may reflect a chance clustering of neutral variants. Such isolated mutants were not seen at any other codon position in the sample as a whole (143 total sequenced isolates), however, and thus the codon 210 mutation appearing independently in two maternal isolates may suggest a real, albeit small, in vivo phenotypic effect. Even a small replicative advantage, too small to detect in vitro but propagated in vivo through many cycles of viral replication could nonetheless result in substantial replacement of wild-type by mutant sequences in the intact host.

The transmission of mutations and mixtures at presumably neutral sites not known to be associated with ZDV resistance offers a baseline against which the patterns of resistance transmission can be compared. At these other sites, we did not observe any marked change in frequency or type of sequence variants in the RT region and saw none of the apparent differential transmission observed with the resistance codons. Thus, the skewed transmission of resistance mutant-containing mixtures is more likely a property of the resistance mutants themselves, rather than some general characteristic of HIV genome changes in the presence of ZDV.

This study of heterogeneous viral quasispecies in perinatal transmission can provide insight into the natural selection of viral subpopulations during vertical transmission of HIV. Given the key role ZDV plays in the management of HIV infection during pregnancy, it is critical to understand the transmission of ZDV-resistance mutants to the infant. In particular, the transmission of the codon 215 resistance mutants but not codon 70 mutants when present in a mixture suggests a possible differential fitness for transmission of these mutations. This knowledge may help to guide antiretroviral decisions during pregnancy by clarifying the risks of transmission of resistant virus to the offspring.

Acknowledgments

The authors are grateful to the members of the Women and Infants Transmission Study (see Cited Here...) for their support of this project.