Use of recombinant viruses to assess the pattern of early human immunodeficiency virus breakthrough infection in the presence of stavudine

The measurement of plasma human immunodeficiency virus (HIV) RNA copy number after the initiation of antiviral therapy has provided several insights into the kinetics and dynamics of HIV infection. Initial studies quantitating HIV RNA after the initiation of a non-nucleoside reverse transcriptase inhibitor (NNRTI), nevirapine, indicated a very rapid turnover of plasma HIV (Wei et al., 1995 ). In those studies there was an initial decline in HIV RNA followed by a rapid increase in plasma viral RNA. These studies with nevirapine demonstrated that the rapid rebound in HIV RNA levels was a consequence of the outgrowth of HIV with phenotypic and genotypic resistance to nevirapine (Wei et al., 1995 ). Subsequent kinetic analysis indicated a prevalence of nevirapine-resistant HIV-1 variants in pre-treatment plasma of approximately 0·001 of the HIV-1 (Havlir et al., 1996 ). An in vitro model of HIV infection in the presence of a different NNRTI (TIBO) has indicated a very similar prevalence of unselected variants capable of infection in the presence of the drug (Strair et al., 1993 )

Similar clinical and laboratory studies analysing early HIV infection in the presence of 3'-azido-3'-deoxythymidine (AZT) have also been undertaken (Loveday et al., 1995 ). In contrast to the clinical studies with nevirapine, early HIV infection in the presence of AZT does not appear to be predominated by the outgrowth of AZT-resistant HIV. While the amount of virus circulating in plasma shortly after the initiation of AZT rapidly declines, the remaining circulating virus early after this decline does not contain mutations known to encode resistance to AZT (Loveday et al., 1995 ). A component of this genetically AZT-sensitive virus may be derived from longer-lived cells with more latent infection (Perelson et al., 1996 ); however, the amount of virus present and the ultimate emergence of genetically resistant virus containing multiple mutations associated with AZT-resistance implies that there may also be virological, cellular or pharmacological features which allow HIV-1 that is initially genotypically and phenotypically sensitive to AZT to replicate in the presence of AZT. These latter features may impact on the kinetics of virus production in the presence of AZT and allow ongoing replication that supports the development of variants with the multiple mutations known to be associated with high level genetic resistance to AZT.

In prior studies, a quantitative in vitro model system has been used to analyse a single cycle of recombinant HIV infection in the presence of antiviral drugs (Strair et al., 1993 ; Medina et al., 1995 , 1998 ). This model system has demonstrated a pattern of virus infection that shares some important characteristics with the pattern of infection seen in the initial clinical studies analysing virus dynamics: (i) early breakthrough infection in the presence of an NNRTI that is dependent on genetic drug resistance; and (ii) early breakthrough infection in the presence of AZT that is, at least partially, independent of the presence of genetic resistance (Strair et al., 1993 ; Medina et al., 1995 ). Furthermore, this recombinant virus system has allowed an estimate of the prevalence of unselected NNRTI-resistant variants in a population of unselected replication-competent HIV-1 that has subsequently been approximated by mathematical modelling of HIV-1 kinetics after the initiation of an NNRTI (Havlir et al., 1996 ; Strair et al., 1993 ). Therefore, some features of early recombinant virus infection in the presence of antiviral drugs in vitro are similar to those described in clinical studies of HIV dynamics after the initiation of an NNRTI (Wei et al., 1995 ) or AZT (Loveday et al., 1995 ) and further analyses of this in vitro recombinant system may allow a characterization of some of the mechanisms that allow HIV replication in the presence of antiviral drugs.

To determine the predominant mechanisms responsible for early HIV infection in the presence of another nucleoside reverse transcriptase inhibitor, stavudine (d4T), we utilized a similar in vitro recombinant system. The results of this analysis indicate that a significant component of early infection in the presence of d4T is a consequence of infection by genetically sensitive virus that has infected a subset of the host cell population which is refractory to the antiviral effects of d4T.

Cells.
The lymphoid cell lines H9 and JE6.1 were cultured in RPMI 1640 medium supplemented with antibiotics, 2 mM l-glutamine and 10% FBS. Peripheral blood mononuclear cells (PBMC) isolated from healthy HIV-1 seronegative donors were activated with PHA (10 µg/ml) for 72 h prior to infection. After PHA stimulation, PBMCs were maintained in RPMI 1640 supplemented with 10% interleukin-2 (Advanced Biotechnologies), 20% FBS, 2 mM l-glutamine and antibiotics.

Virus.
Production of recombinant HIVgpt has been described elsewhere (Strair et al., 1993 ; Page et al., 1990 ). The amphotropic cell line PA317 was transfected with the recombinant murine retrovirus pLXSN (Miller & Rosman, 1989 ) and was used as the source of the recombinant MLVneo virus. Stock preparations of HIV-1 IIIB were harvested from H9 cells by the `shake off method'. Stock virus infectivity was determined by end-point dilution in MT-2 cells. Virus-induced cytopathic effect was scored 7 days post-infection and the TCID₅₀ was calculated with the Reed and Muench equation (Reed & Muench, 1938 ).

Compounds.
Stavudine (d4T) and floxuridine (5-fluoro-2-deoxyuridine, FUdR) were purchased from Sigma and were dissolved in PBS, sterile-filtered and stored at -20 °C.

HIV-1 RT assay.
HIV-1 production in infected cells was determined by a ³²P-based assay as described by Willey et al. (1988) . RT activity was determined by quantification of ³²P bound to DE81 paper by using a Molecular Dynamics phosphorimager. The results are reported as pixel units/µl of the reaction mixture.

Cytotoxicity assay.
A checkerboard analysis of the cytotoxicity of d4T and FUdR alone and in combination was assayed. Triplicate wells of 24-well plates containing 1x10⁵ cells were cultured in the absence or presence of various concentrations of each drug alone or in combination. Samples were taken every 2 days for 810 days. Drug cytotoxicity was quantified by the MTT reduction assay (Mosman, 1983 ). The amount of formazan produced in 4 h was determined by dissolving the product in 0·1 M HCl made with 2-propanol and then measuring the absorbance at 570 nm.

Recombinant HIV and murine leukaemia virus (MLV) infection in the presence of d4T
To determine the predominant mechanisms responsible for early HIV infection in the presence of d4T, three different populations of recombinant viruses were utilized (Fig. 1). Recombinant replication-defective HIVgpt was prepared by transfection of COS cells with complementing plasmids encoding the RNA and proteins necessary for the production of a replication-defective recombinant HIV encoding gpt (Strair et al., 1993 ; Page et al., 1990 ) (Fig. 1A). Such viruses are produced without major genetic heterogeneity as the predominant mechanisms responsible for the generation of heterogeneity (e.g. cycles of reverse transcription) are not involved in the production of these viruses. In contrast, replication-defective HIVgpt made by rescue with replication-competent HIV-1 (Fig. 1B) will contain proteins encoded by the replication-competent virus used for rescue, and are anticipated to have greater heterogeneity. Prior experiments have demonstrated the close relationship between the drug sensitivity phenotype of the recombinant virus and the drug sensitivity phenotype of the virus used to rescue the recombinant virus. For example, the use of an NNRTI-resistant virus to rescue HIVgpt results in HIVgpt that is resistant to the NNRTI (Strair et al., 1993 ; Medina et al., 1998 ), and the use of an AZT-resistant virus to rescue HIVgpt results in HIVgpt that is resistant to AZT (D. J. Medina & R. K. Strair, unpublished). Therefore, recombinant virus produced by rescue with unselected replication-competent virus will be heterogeneous and may reflect the drug sensitivity profile of the virus used for rescue (Strair et al., 1993 ; Medina et al., 1995 ). In a prior experiment, the heterogeneity introduced by the replication-competent virus used for rescue resulted in a calculation of the prevalence of HIV genetically resistant to an NNRTI in an unselected population that was very similar to the prevalence subsequently calculated from in vivo studies of HIV dynamics after the initiation of an NNRTI (Havlir et al., 1996 ; Strair et al., 1993 ).

(26K): Fig. 1. Schematic representation of the production of the recombinant viruses. (A) HIVgpt is produced by transfection of plasmids into COS cells. The two plasmids, HIVgpt and HIVenv (Page et al., 1990 ) encode the complementing proteins and RNA which allows assembly of the replication-defective recombinant HIVgpt virus. (B) `Rescued HIVgpt' is obtained after infection of the H9/HIVgpt cell line with HIV-1 IIIB (Strair et al., 1993 ). The H9/HIVgpt cell line contains stably integrated HIVgpt. (C) MLVneo is obtained after infection of a murine retrovirus packaging cell line, PA317, with a plasmid encoding a replication-defective murine retrovirus containing the neo gene.

A comparison of the two virus populations described above (Fig. 1A, B) demonstrated very similar rates of infection in the presence of 50 µM d4T (Table 1). The similar levels of infection with these two recombinant viruses is in contrast to infections performed in the presence of the NNRTI TIBO (Strair et al., 1993 ;Medina et al., 1995 , 1998 ). In those prior studies with the NNRTI, there was a 20-fold increased rate of infection with the rescued HIVgpt virus (Fig. 1B) compared to the virus produced by transfection of COS cells with plasmids (Fig. 1A) (Strair et al., 1993 ). This disparity of infection rates with the two recombinant virus preparations used for infection in the presence of the NNRTI was interpreted as an indication of the presence of genetic NNRTI-resistant HIV-1 variants. The presence of these variants was subsequently confirmed by genetic analysis of unselected HIV (Najera et al., 1995 ), and kinetic analysis of the emergence of resistance in treated patients (Havlir et al., 1996 ; Havlir & Richman, 1994 ). The absence of a disparity of infection rates with the two virus populations in the presence of either AZT (Medina et al., 1995 ) or d4T (Table 1) suggests that a component of early infection in the presence of these drugs may be a consequence of infection with virus that is not genetically resistant to the drug.

Table 1. Frequency of HIVgpt and MLVneo colony formation in the presence and absence of d4T

The virus produced by COS cell transfection of plasmids (Fig. 1A) infected HeLa-T4 cells in the presence of d4T at a relatively high frequency (1·7x10^-3 of the rate of infection in the absence of d4T). To confirm this high rate of infection by recombinant viruses not anticipated to have a high rate of genetic heterogeneity, we used other recombinant viruses and host cells. MLV-based recombinant viruses encoding the neo gene (Fig. 1C) showed a similar high rate of infection in the presence of d4T (Table 1). To assure that the high level of infection in the presence of d4T with the viruses depicted in Fig. 1(A, C) was not specific for HeLa-T4 cells, similar infections of Jurkat cells were undertaken. Those experiments indicated a similar high frequency of infection of Jurkat cells in the presence of d4T (see Table 3).

Table 3. Infection of Jurkat cell clones sensitive and resistant to d4T

The similar rates of infection with virus prepared by plasmid transfection and virus prepared by rescue with replication-competent HIV suggested that features other than genetic resistance might be contributing to early in vitro HIV breakthrough in the presence of d4T. This interpretation was supported by evidence of high rates of infection with MLV-based recombinant virions (also anticipated to have a low level of genetic heterogeneity). Infections of Jurkat cells indicated that the high rate of infection in the presence of high concentrations of d4T was not cell line specific. Although it is possible that a low frequency of genetically resistant virus present in the replication-defective recombinant HIV obtained by COS cell transfection and/or MLV virus populations was solely responsible for the infections seen in the presence of d4T, the similar rates of infection seen with both viruses and the lack of an increase in infection with virus obtained by rescue with replication-competent virus indicated that a component of early HIV breakthrough infection in the presence of d4T was likely to be unrelated to infection by genetically resistant virus.

Isolation and characterization of cells infected in the presence of d4T
The presence of a selectable marker gene in the recombinant retroviruses depicted in Fig. 1 allowed the direct isolation of cells infected by these viruses in the presence of d4T (Fig. 2) and the demonstration that a component of the infection was a consequence of infection by viruses that were not genetically resistant to d4T. HeLa-T4 cells infected with MLVneo (Fig. 1C) in the presence of d4T were isolated, cloned and expanded into cell lines that were subsequently infected with recombinant HIV, recombinant MLV (MLVLacZ; see Strair et al., 1991 , 1993 ) or replication-competent HIV-1. As demonstrated in Table 2, approximately 37% of the isolated HeLa-T4 cells were repeatedly refractory to the antiviral effects of d4T (they could readily be infected with recombinant MLVLacZ in the presence of 50 µM or 100 µM d4T). The same cells readily infected with recombinant MLV in the presence of d4T were also readily infected with recombinant HIV or replication-competent HIV-1 in the presence of d4T (data not shown). Similar experiments with AZT, ddI and ddC indicate that refractoriness to the antiviral effects of each of these nucleoside analogues contributes to early breakthrough infection in vitro in the presence of the drug (Table 2). A comparison of the prevalence of cells refractory and persistently refractory to the antiviral effects of these drugs is shown in Table 2. The prevalence of cells persistently refractory to the antiviral effects of d4T is higher than the prevalence of cells persistently refractory to the antiviral effects of AZT, ddI or ddC. These studies directly demonstrate a subpopulation of host cells that are infected with recombinant viruses in the presence of d4T and can be infected repeatedly with other recombinant viruses or replication-competent HIV in the presence of d4T. These cells were not pre-selected by exposure to antiviral drugs.

(32K): Fig. 2. Isolation and characterization of cells infected with recombinant HIV in the presence of d4T. Recombinant MLVneo, produced as depicted in Fig. 1(C), was used to infect cells in the presence of d4T. Infected cells were selected in G418 selective media and individual infected cells were expanded into cell lines. These cells were subsequently infected in the absence and presence of d4T with replication-competent HIV-1 IIIB, recombinant HIVLacZ (Strair et al., 1993 ) or recombinant HIVgpt (Fig. 1B). Control cells were obtained in an identical fashion with the exception that they were initially obtained after infection with MLVneo in the absence of d4T.

Table 2. Colony formation and persistent resistance after HeLa-T4 infection with MLVneo virus in the presence and absence of various antiretroviral agents

A high percentage of Jurkat cells (JE6.1) infected in the presence of d4T also were persistently refractory to the antiviral effects of d4T (Table 3). Infections with replication-competent HIV-1 demonstrated that the refractoriness to infection detected in these clones was not a phenomenon solely associated with recombinant viruses (Fig. 3). The d4T IC₅₀ of these infections with replication-competent HIV was 50350-fold higher than the d4T IC₅₀ determined by infection of parental JE6.1 cells (Table 4). This marked increase in d4T IC₅₀ is in contrast to the only moderate increase in d4T IC₅₀ seen when d4T-resistant virus selected in vitro or in vivo is used to infect unselected cells (Lacey & Larder, 1994 ; Lin et al., 1994 ). Therefore, the relative resistance to d4T in the infection of these refractory clones is far greater than the degree of resistance seen with virus that has been selected for prolonged periods either in vitro or in vivo.

(24K): Fig. 3. Infection of JE6.1 cell clones refractory to the antiviral effects of d4T [D4T bulk () is a mixed population of cells initially infected with MLVneo in the presence of d4T; D4T R1 () and D4T R3 (") are isolated clones of cells initially infected with MLVneo in the presence of d4T and subsequently demonstrated to be persistently refractory to the antiviral effects of d4T] and a control clone of JE6.1 cells () with HIV-1 IIIB in the presence of various concentrations of d4T. RT activity was assayed and compared with that for a control infection in the absence of d4T.

Table 4. d4T/FUdR susceptibility in cells sensitive and refractory to the antiretroviral activity of d4T

A subset of the persistently refractory cells (approximately 20%) were also refractory to the antiviral effects of AZT. Cells refractory to the antiviral effects of d4T and AZT were not refractory to ddI, ddC or TIBO (data not shown).

These results provide support to the concept that a component of early infection of these cell lines in vitro occurs as a consequence of mechanisms independent of genetic drug resistance. The phenomenon of early virus breakthrough in the presence of these nucleoside analogues has been demonstrated with replication-defective recombinant HIV, replication-competent HIV and two replication-defective recombinant murine retroviruses. In addition, early infection as a consequence of cellular features has been seen with two different host cell types (HeLa-T4 and Jurkat 6.1).

As has been demonstrated previously for AZT (Medina et al., 1996 ), a component of the refractoriness to the antiviral effects of d4T can be reversed by the addition of FUdR (Table 4). The antiviral efficacy of the combination therapy, as measured by the d4T IC₅₀, is markedly improved with combination therapy. Prior studies of the antiviral efficacy of FUdR have demonstrated limited antiviral efficacy of FUdR alone, but marked antiviral efficacy of combined AZT and FUdR (Medina et al., 1996 ). Table 4 shows the capacity of the FUdRd4T combination to reverse some of the cellular refractoriness to d4T described above. The FUdRd4T combination has marked antiviral activity in cells demonstrated to be persistently refractory to the antiviral effects of d4T. To assure that FUdR did not inhibit the antiviral activity of d4T in normal unmanipulated PBMC, HIV-1 was used to infect PBMC in the presence of various concentrations of d4T and FUdR. As demonstrated in Fig. 4, the d4TFUdR combination resulted in markedly enhanced antiviral activity (compared to d4T alone) in PBMC. FUdR alone resulted in minimal antiviral activity.

(23K): Fig. 4. A d4TFUdR combination inhibits HIV-1 infection of PBMCs. PBMCs were infected with HIV-1 in the absence of drug (+), with d4T alone (0·01 µM; x), with various concentrations of FUdR alone (, 0·005 µM FUdR; , 0·01 µM FUdR; , 0·025 µM FUdR) or with combinations of FUdR and 0·01 µM d4T (", d4T+0·005 µM FUdR; , d4T+0·01 µM FUdR; , d4T+0·025 µM FUdR).

Recent clinical analyses have emphasized the potential of prolonged suppression of HIV viraemia when antiviral drug combinations are used (reviewed in Finzi & Siciliano et al., 1998 ; Gulick et al., 1997 ). However, long term eradication of virus has not yet been consistently demonstrated and virus regrowth with cessation of antiviral drugs can occur. In the face of this uncertainty, more detailed information concerning the mechanisms contributing to HIV breakthrough in the presence of antiviral drugs is needed. Although the clinical relevance of the studies presented in this report are unclear, the heterogeneity of cell types hosting in vivo HIV infection raises the possibility that similar `cellular' mechanisms contribute to HIV breakthrough at some point in the natural progression of the disease.

In this report we utilized an in vitro model of HIV infection to provide two lines of evidence that early HIV breakthrough infection in the presence of d4T is not solely a consequence of infection by HIV with genetic drug resistance. Initial studies demonstrated that the frequency of HIV infection in the presence of d4T was very similar with several stocks of virus predicted to have significant differences in genetic heterogeneity. These studies suggested that any pre-existing unselected d4T-resistant HIV in the population of HIV used to rescue the replication-defective HIV was not detected above the very high level of infection occurring with the other virus populations. The fact that nearly 50% of the Jurkat cells infected with HIV in the presence of d4T are readily re-infected in the presence of high concentrations of d4T provides definitive evidence of a high level of infection in the absence of genetic d4T resistance. These results are not limited to recombinant HIV and were also demonstrated with MLV-based viruses as well as replication-competent HIV.

A subset of the cells infected with HIV in the presence of d4T does not have a persistent phenotype of being refractory to d4T. These cells may have been infected as a consequence of cell cycle related phenomena, intravirion reverse transcription (Zhang et al., 1994 ) or other viral, cellular or pharmacological features that are not characterized by persistent cellular phenotypic change.

Previous studies have demonstrated that the combination of FUdR with AZT or d4T has significant antiretroviral activity (Medina et al., 1996 ; Ahluwalia et al., 1996 ). Furthermore, the combination of FUdR and AZT has potent antiretroviral activity in cells refractory to the antiretroviral activity of AZT (Medina et al., 1996 ). These studies demonstrate the capacity to improve the antiviral efficacy of d4T by the addition of drugs such as FUdR with the capacity of interacting with the biochemical mechanisms responsible for AZT and/or d4T metabolic activation. This increased antiviral efficacy in cells refractory to the antiretroviral effects of AZT or d4T may have therapeutic implications for the treatment of early HIV infection in the presence of d4T (e.g. cells infected with d4T-sensitive HIV in the presence of d4T). As this early replication may contribute to the ultimate emergence of genetic resistance, determination of the mechanisms contributing to the presence of virus after initiation of therapy may impact on efforts to achieve long term suppression of the virus.

In summary, we have demonstrated that early HIV breakthrough infection in vitro in the presence of d4T is not solely a consequence of infection with virus that is genetically resistant to d4T. As with AZT, infection with drug-sensitive virus is a major mechanism of infection early after the introduction of the drug in vitro. Further analyses of the mechanisms responsible for HIV breakthrough in the presence of antiviral drugs are essential to efforts to define drug combinations that provide durable suppression of HIV infection and viraemia.