Skip to main page content

• HOME
• Current Issue
• Archive
• Alerts
• About ASM
• Contact us
• Tech Support
• Journals.ASM.org

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Reprints and Permissions Copyright Information Books from ASM Press MicrobeWorld Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Wick, W. D. Articles by Self, S. G. Search for Related Content PubMed PubMed Citation Articles by Wick, W. D. Articles by Self, S. G.

 Previous Article  |  Next Article 

Journal of Virology, November 2005, p. 13579-13586, Vol. 79, No. 21
0022-538X/05/$08.00+0     doi:10.1128/JVI.79.21.13579-13586.2005
Copyright © 2005, American Society for Microbiology. All Rights Reserved.

How Many Human Immunodeficiency Virus Type 1-Infected Target Cells Can a Cytotoxic T-Lymphocyte Kill?

W. David Wick,^1* Otto O. Yang,² Lawrence Corey,^3,4 and Steven G. Self¹

Statistical Center for HIV and AIDS Research and Prevention, Fred Hutchinson Cancer Research Center, Seattle, Washington 98109,¹ UCLA AIDS Institute, Department of Medicine, Department of Microbiology, Immunology and Molecular Genetics, David Greffen School of Medicine, UCLA, Los Angeles, California 90095,² Department of Medicine, University of Washington, Seattle, Washington 98101,³ Program in Infectious Disease, Fred Hutchinson Cancer Research Center, Seattle, Washington 98109⁴

Received 27 April 2005/ Accepted 26 July 2005

Top Abstract Introduction Materials and Methods Results Discussion References

 
The antiviral role of CD8⁺ cytotoxic T lymphocytes (CTLs) in human immunodeficiency virus type 1 (HIV-1) infection is poorly understood. Specifically, the degree to which CTLs reduce viral replication by killing HIV-1-infected cells in vivo is not known. Here we employ mathematical models of the infection process and CTL action to estimate the rate that CTLs can kill HIV-1-infected cells from in vitro and in vivo data. Our estimates, which are surprisingly consistent considering the disparities between the two experimental systems, demonstrate that on average CTLs can kill from 0.7 to 3 infected target cells per day, with the variability in this figure due to epitope specificity or other factors. These results are compatible with the observed decline in viremia after primary infection being primarily a consequence of CTL activity and have interesting implications for vaccine design.

Top Abstract Introduction Materials and Methods Results Discussion References

 
Evidence for the importance of cytotoxic T lymphocytes (CTLs) in retroviral infection includes the temporal association of the human immunodeficiency virus type 1 (HIV-1)-specific CTL response with reduction in viremia during primary infection (15) and the acute rise in viremia when simian immunodeficiency virus (SIV)-infected macaques (SIVmac) are CD8⁺ cell depleted (12, 19, 27).

These data have prompted many vaccinologists to focus on cellular immunity. Yet, despite clear evidence that CTLs have a crucial antiviral function, their impact in vivo remains obscure. This lack of understanding presents an obstacle to defining the requirements for a successful CTL-based vaccine. Mathematical modeling of the interaction of CTLs and HIV-1 is one approach to elucidating these requirements.

For complicated biological processes, we often rely on mathematical models to explore mechanisms that are beyond direct experimental measurement. Most models require the representation of contributing processes by rate constants, to allow the evaluation of a mechanism of interest. Typically, some of these rate constants are estimated from the data used to build the model. But as model complexity increases, varying a large number of poorly defined parameters risks identifying the best-fitting wrong model. To avoid this pitfall, ideally rate constants should be derived from dedicated experiments, so that the "full model," when finally assembled, is not internally circular.

A key parameter in such a model is the efficiency of CTLs in eliminating infected target cells. A direct estimate is derived from observing the impact of patient-derived HIV-1-specific CTLs on replication in HIV-1-infected cell lines in vitro. From titrations of CTL density against viral growth, we can derive the killing rate by fitting a simplified model. To confirm the relevance of these in vitro measurements using cell lines, we estimated the same parameter by analyzing data from an in vivo "adoptive transfer" experiment conducted by Brodie et al. in 1999 (6, 7). Our results provide similar independent estimates of this key parameter for modeling the impact of CTLs in HIV-1 pathogenesis.

Top Abstract Introduction Materials and Methods Results Discussion References

 
HIV-1 permissive cell lines and HIV-1 stocks. The HIV-1 permissive cell lines T1 (expressing HLA A*02) and H9-B14 (expressing HLA B*14) were maintained in RPMI with 20% heat-inactivated fetal calf serum as previously described (40). These cells were infected during log-phase growth (doubling time, about 2 days). HIV-1IIIB stocks were generated and titers were determined as previously described (13, 40).

HIV-1-specific cytotoxic T-lymphocyte clones. HIV-1-specific CTL clones were obtained by limiting dilution cloning from peripheral blood mononuclear cells of infected individuals, characterized for specificity and HLA restriction, and maintained as previously described (30). Clone 68A62 recognized A*02-restricted reverse transcriptase (RT) epitope ILKEPVHGV (RT amino acids [aa] 476 to 484); 18030D23 recognized the A*02-restricted Gag epitope SLYNTVATL (Gag aa 77 to 85, p17); and clones LWC8 and 115M21 recognized the B*14-restricted Env epitope ERYLKDQQL (Env aa 584 to 592, gp41).

Inhibition assay. Assays for inhibition of HIV-1 replication by CTLs were performed as previously described (41). Briefly, target cells were infected at a multiplicity of 0.01 tissue culture infectious dose per cell and cultured at 5 x 10⁵ target cells per well in 24-well plates with CTL clones at the indicated ratios. Supernatant p24 antigen concentrations were assayed by enzyme-linked immunosorbent assay at 2- to 4-day intervals (Dupont, Boston, MA).

Data from an in vivo study of CTL adoptive transfer. Published data from a 1999 study of infusions of HIV-1-specific CTLs were chosen for analysis because they provided a clear scenario where CTL levels changed with corresponding alterations in the concentrations of infected cells. Brodie et al. (6) derived Gag (p17 or p24)-specific CTL clones from three HIV-infected subjects on antiretroviral therapy with stable CD4 counts (224 to 261 per mm³). The CTLs were genetically modified to express the neomycin phosphotransferase gene (neo), and infusions (1 x 10⁹ to 3 x 10⁹ cells per m² of body surface area) were administered a week apart. After the second dose, neo-modified cells constituted 2 to 3.5% of the patients' CD8 compartment. A few days after each infusion, the concentration of HIV RNA-positive cells decreased dramatically; but, in a few weeks, the neo-CTL disappeared and the infected-cell count rebounded. The data consist of baseline CD4 and the percentages of neo-marked CD8 cells and HIV-producing CD4 cells (productively infected targets [PITs]) at up to 14 time-points. Brodie et al. made several prebaseline measurements of PITs, which, except for patient 1, who seems to have had a spike at baseline, can be averaged to yield a steady-state infection rate. For patient 1, we used the value at day zero. Several CTL measurements were missing for patient 2, and we also omitted the final observation at day 72, which we expect is also an infection spike.

Mathematical models and fitting methods. For the in vivo analysis, we used a rate-equation (ODE) model with five compartments: the two observed, denoted C and P; uninfected, quiescent CD4⁺ T cells, denoted Q; uninfected, activated CD4⁺ T cells (targets) denoted T; and infected CD4⁺ T cells in the "eclipse" phase before transcribing HIV genes, denoted I. Rapid loss of CTLs after infusion allowed autonomous modeling of their effect, independent of the entire CTL activation circuit (involving antigen-presenting cells and CD4⁺ T-cell "help," for which few rate constants are available).

The model's kinetic equations express the rates at which cells enter or leave a compartment. Starting with the dynamics of the uninfected immune system: quiescent CD4⁺ T cells are created at rate , die at rate _Q, and are activated at rate . Activated cells are eliminated at rate _T, with a fraction, , reverting to resting and the others undergoing programmed cell death (apoptosis). (We ignored the naïve versus memory distinction to restrict model complexity, but we might have included it as in reference 32.) Evidence from another experiment (see Discussion) indicated that the CTLs were unable to divide in vivo. Hence we assumed that the initial dynamics reflected cell dispersal and modeled it by a constant growth rate, from zero to the peak. For after-peak dynamics, we assumed a simple exponential decay with rate _C. On the infection side, target cells are infected at rate P and progress to the productive phase at rate . (We ignored the latently infected T-cell compartment, because the drop in viral load occurred too quickly to be explained by latent production.)

PITs are deleted by apoptosis or immune system killing, except by neo⁺ CTLs, at rate _P; killing by neo⁺ CTLs occurred at rate C.

The rate equations are:

(1)

(2)

(3)

(4)

(5)

The Greek letters denote parameters. The infusion term inf(t) is constant for times 0 t 1 and 7 t 8 and is zero otherwise; the function noinf(t) is zero on these intervals and otherwise 1.

We included the unobserved compartments in order to exploit the baseline CD4 and viral load data. We omitted free virus, V, since viral dynamics is fast relative to cellular dynamics (free-virion lifetime in vivo may be less than a half-hour (11); effectively, V is proportional to P.

In this model, some rate constants are known approximately from other studies and are not likely altered by HIV infection or drug therapy, while others are either unknown or will likely be affected. For the former, we adopted standard values (see Table 1). For the latter, we cannot assume the thymopoesis rate, the activation rate, the infection coefficient, or the PIT lifetime is the same in Brodie et al.'s (6) infected, treated subjects as in healthy or untreated ones. The target of this investigation is the killing rate, , while _C, the death rate of the neo-CTLs, can also clearly be estimated. Hence , , , and _P, as well as the initial values of Q, T, and I, remain as unknown parameters. We resolved these issues as follows. We chose to estimate _P, the death rate (inverse lifetime) of PITs, and , the thymopoesis rate, in order to compare with known estimates.

View this table: [in this window] [in a new window]

TABLE 1. Adopted standard values for the rate constants used in this study

The remaining unknowns, and , we derived by solving the steady-state equations of the model.

A steady state exists for equations 1 to 4, omitting C, with (observed) values of P_ss and CD4 = Q_ss + T_ss + I_ss + P_ss, provided – A_P – B > 0, where A = P_ss and B = _Q (CD4 – P_ss) are fixed. Hence we adopted as estimable parameters (_P, , _C, ), yielding a rectangular search region. For each vector , we derived the other parameters and the unobserved steady-state compartment values (used as initial conditions) from the equations

(6)

(7)

(8)

(9)

(10)

(11)

The vector of parameters to be estimated was thereby reduced to (_P, , _C, ).

We assumed Gaussian measurement errors and minimized the sum of squares of differences between model predictions and data points (nonlinear regression). We fit the model by minimizing the sum of squares

(12)

where the pair (P_i, C_i) represents the observations at the i-th time point and X(t) is the five-vector solution of system equations 1 to 4, with initial conditions (Q_ss, T_ss, I_ss, P_ss, 0).

We used the fourth-order Runge-Kutta algorithm or an implicit solver to solve the system (25). For the minimization, we used a derivative-free, hill-climbing routine with random restarts (details available from the authors upon request).

To make confidence intervals, we computed the Fisher information matrix F:

(13)

In equation 13, p and q range over the four parameters and "Cov" stands for a noise-covariance matrix which was taken to be diagonal, with the only non-zero entries on the 4-th and 5-th rows; these covariances were taken from Brodie et al.'s (6) reported standard deviations. The subscripts p and q denote partial derivatives with respect to the parameters X_p(t) = / _p X(t); they were obtained by solving the system

(14)

with initial condition Y_p(0) = / _p X(0). (Recall that the initial conditions depended on a parameter to be estimated.) Vec stands for the right-hand sides of equations 1 to 5, and J (for Jacobian) is the matrix obtained by differentiating Vec with respect to the coordinates. We proceeded by standard likelihood theory to invert F and use the diagonal entries to construct the Wald-type 95% confidence intervals for _P, , _C, and 90% confidence intervals for (which is a linear combination of _P and ). Finally, we assessed convexity of the likelihood by starting the function minimizer at random points within the search region, to ascertain if the same minimum was always reached (it was on 10 restarts); also, we checked that the information matrix evaluated at the minimum was positive definite.

In order to discuss the steady state with a given level of neo-CTL killing, we solved equations 1 to 4 with C fixed to obtain P_new:

(15)

where = _Q + and = _P + C. (Steady states in the model are neutrally stable, an artifact of omitting free virus; nevertheless they attract a perturbed state.)

For the in vitro analysis, we reduced the basic model to three compartments: uninfected target cells, infected cells in eclipse phase, and PITs. That is, we omitted Q and equations 1 and 5, assumed C = rT(0) is constant, set _T = 0, and added a term, T, in equation 2 representing the growth rate of target cells. In equation 4, we replaced C with C/(1 + C); parameter allows for saturation at a high effector/target ratio. Substituting rT(0) for C, we performed a nonlinear regression of observed log inhibition (LI): LI = log[V(0; r)/V(8; r)], where V(t; r) is virus on day t measured in any units, e.g., nanograms/ml of p24 antigen, versus log[P(0; r)/P(8; r)], where P(t; r) is the predicted density of PITs. We estimated and while keeping other rate constants fixed.

Top Abstract Introduction Materials and Methods Results Discussion References

 
HIV-1-specific CTL clones suppress HIV-1 in a dose-dependent fashion. The antiviral activity of CTLs has been studied in an in vitro experimental model evaluating the interaction of HIV-1-specific CTL clones with acutely HIV-1-infected T cells. Using an established assay to measure the ability of CTL clones to suppress the replication of HIV-1 in acutely infected cells in vitro, we assayed the dose relationship (five to seven titrations per clone; see Fig. 1 for the dose-response data for sample clones). As previously reported, each CTL clone demonstrated potent viral suppression in a dose-dependent manner.

View larger version (10K): [in this window] [in a new window]

FIG. 1. Dose-response and model fit to the in vitro data for two CTL clones. Numbers on the x axis are 100 times the CTL/target ratio in the well, reflecting a multiplicity of infection of 0.01; numbers on the y axis are the log10 inhibitions of p24 antigen, relative to controls, after 8 days.

The dose-response curves for inhibition of HIV-1 by CTL reveal killing rates in vitro. The curves show evidence of saturation at high CTL dose. Also, target cell limitation may be significant in the 2-ml wells (the authors of reference 34 presented an argument that it is not important in vivo); the immortalized but uninfected target cells also exhibited log-phase growth.

As simple regression models are inappropriate here, we employed for the statistical estimation a three-compartment ODE model of uninfected target cells, infected cells in the eclipse period, and PITs. In this analysis, sensitivity to unknown parameters mattered more than statistical error. (Indeed, estimating both killing and saturation coefficients, the "best" curves nearly passed through the observations [Fig. 1].)

The point estimates were not sensitive to varying the basic reproductive number (R₀) but were to varying PIT lifetime (parameter _P) and eclipse period (parameter ); neither was known accurately in these cell lines. So we performed a sensitivity analysis, varying PIT lifetime from 2 to 3 days and the eclipse period from 1 to 3 days. We exhibit point estimates and sensitivity ranges in Table 2; Fig. 1 shows several model fits.

View this table: [in this window] [in a new window]

TABLE 2. Point estimates and sensitivity ranges

These estimates attribute the antiviral effects of the CTL clones entirely to their PIT killing activity. Prior work (41) evaluated the role of noncytolytic cytokines in this activity and found that killing is the predominant mechanism and the antiviral effects of cytokines are negligible at the lower concentrations of CTLs used to define killing rate slopes near zero (Fig. 1).

Analysis of adoptive transfer experiments demonstrates similar killing rates in vivo. While analysis of the antiviral activity of CTLs against HIV-1 in vitro affords a controlled setting for defining the killing rate, this model is a highly contrived experimental system. We therefore obtained independent estimates of this parameter from experiments reported by Brodie et al. (6). The estimated parameters and confidence intervals are reported in Table 3; we display one model fit in Fig. 2. (The other two were similar.) The technical definition of the killing parameter in vivo is the following. Let CTLs specific for an HIV antigen, with density denoted "C," kill PITs, with density denoted "P," at a rate proportional to the product P · C. The sought after coefficient is simply the constant of proportionality.

View this table: [in this window] [in a new window]

TABLE 3. In vivo estimated parameters

View larger version (11K): [in this window] [in a new window]

FIG. 2. PIT and neo-marked CTL densities and model fit for subject 1, from the study by Brodie et al (6).

This definition ignores complexities such as target or effector localization in tissues, which cannot be addressed from the available data. Allowing for saturation (as in the in vitro study) did not affect the in vivo estimates, presumably because CTLs are tightly linked to antigen and effector/target ratios are stabler than those used in the in vitro study. From fitting our model to the data of Brodie et al., we found 0.129 µl cell^–1 day^–1 (mean of the point estimates in Table 3; the "meta-analytic" 95% confidence interval of the mean was –0.19, 0.44).If total body cell counts are used instead of densities, 2 x 10^–10 day^–1.

For a more concrete representation, consider the killing rate per CTL per microliter. The subjects of this study had about 250 CD4⁺ T lymphocytes per µl, up to 2% productively infected, providing up to 5 PITs per µl in peripheral blood. From the approximate equation P/C = –P t, it is apparent that each CTL was able to kill an average of about 0.65 PIT per day.

Top Abstract Introduction Materials and Methods Results Discussion References

 
A key goal of current HIV-1 vaccine development is provoking a CTL response that will attenuate disease by reducing viral replication. However, the antiviral capabilities of HIV-1-specific CTLs in vivo are poorly understood. Measurements of HIV-1-specific CTL magnitudes do not correlate with viremia (1), presumably because multiple factors (e.g., viral mutation, varying efficacy of different CTL clones, and varying burst size of infected cells) affect and obscure this relationship. Here we evaluate in vitro and in vivo experiments where individual factors are manipulated to allow evaluation of the interaction of CTLs and PITs. Our results indicate that CTLs can have a significant impact on the number of PITs and therefore HIV-1 replication. We found that each (Gag-specific) CTL was able to kill, on average, about 0.65 PIT per day. This number may seem modest, but consider the effects of the killing rate per CTL per µl of blood in vivo. The peak of infused CTLs for the subjects studied by Brodie et al. reached about 2.5% of the CD8⁺ T-lymphocyte compartment, or about 12.5 CTLs per µl (assuming about 500 total CD8⁺ T lymphocytes per µl). Assuming about 2.5 to 5 PITs per µl in blood (see Results), this implies an effector/target ratio of between 2:1 and 6:1. If the infused neo-CTLs had persisted, the model predicts a reduction of viremia by about 1 log after 3 days (see Fig. 3), in agreement with our in vitro model (where the PIT density was similar and the effector/target ratios spanned this range). These results are also consistent with the observation that the development of CTL responses during acute infection can drop the viral load by 2 to 3 orders of magnitude, as demonstrated in more complete CTL-HIV models (see references 35 to 37)

View larger version (7K): [in this window] [in a new window]

FIG. 3. Theoretical prediction from the model for the impact on PITs in subject 1, if the infused CTLs had persisted at peak level (i.e., setting C = 0).

The study by Brodie et al. demonstrated an average CTL lifetime of about 2 days, which would suggest that the CTLs cleared about 1.3 PITs in their lifetime. However, it is likely that this is an underestimation of the lifetime of naturally occurring CTLs, because the infused CTLs had been manipulated ex vivo and contained a foreign marker gene. A similar study by Greenberg et al. (5), where CTLs were marked with thymidine kinase, demonstrated vigorous thymidine kinase-specific, cellular-immunity-mediated clearance of infused CTLs. Thus, lifetime clearance of 1.3 PITs is a minimal estimate for CTLs in vivo. In addition to the killing rate, our analysis also reveals the lifetime of PITs in vivo. We estimated this lifetime to range from 0.5 to 3 days, slightly longer than estimated by investigations of persons newly placed on combination antiretroviral therapy, (10, 31), probably because Brodie et al.'s subjects were already on treatment. (Treatment decreases the viral load, which decreases the antigen available to stimulate CTLs; as the latter's density drops, PIT lifetime can be expected to increase.) At the height of infused CTLs, the PIT lifetime dropped to less than a day (range, 0.22 to 0.78 day). This figure may be relevant to the design of vaccines with the goal of preventing infection. For natural infection, the time to reach peak viremia indicates a basic reproductive number (number of target cells infected by one PIT in its lifetime, absent CTL killing or therapeutic intervention) of around 3. The drop in PIT lifetime translates to about a threefold decline in the basic reproductive number, which could prevent initial infection. However, an important caveat applies: effector action in CTLs depends on activation and differentiation status. Vaccine-induced memory T cells will likely exhibit a delay before they can function (8, 35).

Despite the observed effects on the lifetime of PITs, the transferred CTLs in the study by Brodie et al. (6) did not perturb the established steady state sufficiently to abolish viremia. According to our model, even if the infused CTLs persisted at their peak level, the impact on PITs would have been about a 10-fold-lowered density (Fig. 3). This is considerably less than the effect required to drive HIV-1 from steady state to extinction (disregarding the latent reservoir). That goal would require about a 30% frequency of these CTLs in the CD8⁺ T-lymphocyte compartment. The observation that viremia was not appreciably reduced in this study is consistent with the modest reduction of PITs, only to 0.5% of the CD4⁺ compartment, still in the range observed in chronic infection. In the context of the above hypotheses regarding the threshold of CTLs needed to prevent acute infection, this likely reflects a greater difficulty in destabilizing an already existing steady state.

Many of these conclusions about steady states and eradication follow simply from contemplating the reproductive number (number of daughter PITs produced by one mother PIT in its lifetime) in the simplest infection model (i.e., ignoring eclipse phase):

(16)

If we substitute the estimated parameters and T = T_ss in this formula, we get 1—not surprising because of the definition of steady state (each PIT produces one replacement in its lifetime). To incorporate neo-CTL killing, _P is replaced by _P + C, which pushes R below 1. The PIT density will subsequently fall. However, that does not imply that the possibility of a steady state is abolished; for as infection lessens T will increase. In fact, in the infected steady state T was small, about 0.13 cell per µl; i.e., most activated CD4⁺ T cells were infected. A new steady state may form at a lower PIT density and a higher target cell density. That is in fact what the model predicts; solving the steady-state equations with C fixed at the peak level gave, e.g., for subject 1, a new solution with P about 0.3% of CD4.

For natural infection in untreated individuals, and T will be larger and _P smaller than in the patients of Brodie et al. The reason that ., the initial PIT death rate in natural infection, will be smaller is that endogenous CTLs will not yet be activated or expanded. All three changes increase R, yielding the basic reproductive number, denoted R₀; as we mentioned, R₀ 3 is reasonable. If we imagine having stimulated Gag-specific CTLs in uninfected individuals to the peak neo-CTL level, with the same killing potential, then R₀ will be multiplied by the factor

(17)

A further interesting point can be made from the in vitro study: the CTL clones appear to vary in their antiviral efficiency. This supports the concept that CTL specificity or other properties can affect the antiviral potential (2, 3, 29, 39, 41). That the estimates of killing rates generally are comparable with those derived from the more difficult and costly in vivo study supports the relevance of this in vitro model and suggests that it may be useful for comparing the antiviral activities of different clones and the factors determining these activities.

Caveats to our analyses include issues of statistical power, data selection, and model choice. Concerning statistical issues, we note that, in the in vivo study, the thymopoiesis rate, , was particularly hard to estimate; the stability of the point estimates is surprising given the wide confidence intervals. Observed T-cell replacement rates range from, depending on health and age of the subject, 0.5 to 5 cells per µl per day, (18, 22, 23), so the estimates are high but reasonable. (Some theories of HIV pathogenesis even support an increased T-cell production [10].) Activation rates in the steady state fit to the baseline data (not included in Table 2) came out to be about 3% of CD4, somewhat elevated over the normal 1% (26), as is usually observed in HIV-infected patients (4, 24). In the in vitro study, the killing rate estimates, which reflect the slopes near zero (e.g., in Fig. 1), were essentially based on 3 data points per clone. Concerning the variation of estimated killing rates, a crude ranking, based on maximum height of the inhibition curves, would give 18030D23 > LWC8 > 68A62 > 115M21, but the ranking based on estimated is 18030D23 > 115M21 > LWC8 > 68A62. The reason that 115M21 rose in this rankings is the single observation at an effector/target ratio of 1.56 (dilution factor, 64:1).

Two other objections to our mathematical techniques are that we assumed only passive measurement errors and that we should have described infection by a stochastic, branching-type model. To the former, we note that, for fitting nonlinear, higher-dimensional, active-noise models to data, adequate methods are lacking at this time (but we have proposed one [38]). To the latter, we remark that, with billions of infected targets and HIV-recognizing CTLs in the study by Brodie et al. and hundreds of thousands in the in vitro experiments, laws of mass action should be adequate. In contrast, for primary HIV, where both infection and response begin at low frequencies, a stochastic process is more appropriate (e.g., see reference 33).

Concerning data selection, we can compare our approach with prior work. The investigators in reference 19 depleted CD8 cells from macaques with monoclonal antibodies; when infected with an SIV-HIV chimera, the monkeys suffered elevated viral loads relative to controls. In reference 27, the investigators similarly perturbed the course of events in primary infection with SIVmac, but also depleted CD8 cells from chronically infected animals. Virus replication peaked again in these animals, but decreased when virus-specific CTLs reappeared. Unfortunately, neither work presented sufficient longitudinal data in the form we required for fitting models; counts of PITs and virus-specific CTLs were the crucial elements for our in vivo analysis. The authors of reference 12 reported a similar experiment and also addressed the theoretical mechanisms in terms of models. However, they concluded that the rapid rise in viremia could not be explained in their model by decreased killing, but rather by increased production. Another possible confounding factor in the CD8-depletion experiments is that the administration of the monoclonal antibody or the destruction of CD8 cells might have produced a general inflammation—resulting in increased CD4 activation and viral production.

In reference 8, the authors studied CTL impact on SHIV-89.6P infection in vaccinated and unvaccinated macaques and concluded that the slopes after peak viremia were statistically indistinguishable between the two groups. Even granting that the vaccine acted by amplifying SHIV-specific CTLs, which moreover were potent killers, their method differs sufficiently from ours (exponential growth or decay equations fitted separately to prior- or post-peak viral load and CTL data, as opposed to nonlinear models fitted jointly to perturbed levels of both populations from an existing steady state), that we cannot assess the reason for the disparity.

Concerning model choice, a key assumption made in order to analyze the data of Brodie et al. was that the CTLs did not divide in vivo. This assumption is supported by an experiment in the humanized mouse system (20). CTLs were labeled with the dye CFSE, which binds to structural proteins and hence is evenly divided between daughter cells at division (17). The investigators observed that transferred CTLs did not divide in vivo. Moreover, the CTLs were rapidly lost in the HIV-infected host, probably due to apoptosis. This experiment should be repeated in the SIVmac system. (In reference 35, it is argued that performing the experiment in monkeys would also help resolve the origin of a putative CTL "memory defect.")

An additional concern in the adoptive transfer experiments is that the engineered CTLs might have had impaired function, either due to the manipulations ex vivo or lack of CD4⁺ T-cell help—so we have probably underestimated the natural killing efficiency of CTLs in vivo. Indeed, CTLs specific for HIV and active in chronic infection are probably not representative of CTLs in primary infection (16) or CTLs active against other viruses, due to a memory or killing defect (modeled in reference 35).

In summary, taken together with the observation that CTLs appear simultaneously with the initial drop in viremia (15), our results provide strong positive evidence that control of viremia in HIV infection is a consequence of CTL activity. (In reference 34, the authors provided a complementary negative argument that control is not due to target cell limitation.) Further studies are needed to elucidate the contributions of CTLs with various specificities and to explore whether enhancing the most potent killers is the key to a successful T-cell vaccine.

 
This work was funded by NIH (NIAID) 1 RO1 AI05428 (W.D.W. and S.G.S.) and PHS AI043203 (O.O.Y.).

We thank S. Brodie and P. Greenberg for valuable discussions about the adoptive transfer experiments and B. Walker and S. Kalams for providing the CTL clones for the in vitro assays.

 
* Corresponding author. Mailing address: SCHARP, Fred Hutchinson CRC, 1100 Fairview Ave. N, LE 400, Seattle, WA 98109. Phone: (206) 667-7908. Fax: (206) 667-4812. E-mail: wick{at}scharp.org.

Top Abstract Introduction Materials and Methods Results Discussion References

 

1
1. Addo, M. M., X. G. Yu, A. Rathod, D. Cohen, R. L. Eldridge, D. Strick, M. N. Johnston, C. Corcoran, A. G. Wurcel, C. A. Fitzpatrick, M. E. Feeney, W. R. Rodriguez, N. Basgoz, R. Draenert, D. R. Stone, C. Brander, P. J. R. Goulder, E. S. Rosenberg, M. Altfeld, and B. D. Walker. 2003. Comprehensive epitope analysis of human immunodeficiency virus type 1 (HIV-1)-specific T-cell responses directed against the entire expressed HIV-1 genome demonstrate broadly directed responses, but no correlation to viral load. J. Virol. 77:2081-2092.[Abstract/Free Full Text]
2
2. Ali, A., R. Lubong, H. Ng, D. G. Brooks, J. A. Zack, and O. O. Yang. 2004. Impacts of epitope expression kinetics and class I downregulation on the antiviral activity of human immunodeficiency virus type 1-specific cytotoxic T lymphocytes. J. Virol. 78:561-567.[Abstract/Free Full Text]
3
3. Allen, T. M., D. H. O'Connor, P. Jing, J. L. Dzuris, B. R. Mothe, T. U. Vogel, E. Dunphy, M. E. Liebl, C. Emerson, N. Wilson, K. J. Kunstman, X. Wang, D. B. Allison, A. L. Hughes, R. C. Desrosiers, J. D. Altman, S. M. Wolinsky, A. Sette, and D. I. Watkins. 2000. Tat-specific cytotoxic T lymphocytes select for SIV escape variants during resolution of primary viremia. Nature 407:386-390.[CrossRef][Medline]
4
4. Ascher, M. S., and H. W. Sheppard. 1988. AIDS as immune system activation: a model for pathogenesis. Clin. Exp. Immunol. 73:165-167.[Medline]
5
5. Berger, C., M.-L. Huang, M. Gough, P. D. Greenberg, S. R. Riddell, and H.-P. Kiem. 2001. Nonmyeloablative immunosuppressive regimen prolongs in vivo persistence of gene-modified autologous T cells in a nonhuman primate model. J. Virol. 75:799-808.[Abstract/Free Full Text]
6
6. Brodie, S. J., D. A. Lewinsohn, B. K. Patterson, D. Jiymmapa, J. Krieger, L. Corey, P. D. Greenberg, and S. R. Riddell. 1999. In vivo migration and function of transferred HIV-1 specific cytotoxic T cells. Nat. Med. 5:34-41.[CrossRef][Medline]
7
7. Brodie, S. J., B. K. Patterson, D. A. Lewinsohn, K. Diem, D. Spach, P. D. Greenberg, S. R. Riddell, and L. Corey. 2000. HIV-specific cytotoxic T lymphocytes traffic to lymph nodes and localize at sites of HIV replication and cell death. J. Clin. Investig. 105:1407-1417.[Medline]
8
8. Davenport, M. P., R. M. Ribeiro, and A. S. Perelson. 2004. Kinetics of virus-specific CD8⁺ T cells and the control of human immunodeficiency virus infection. J. Virol. 78:10096-10103.[Abstract/Free Full Text]
9
9. Dimitrov, D. S., R. L. Willey, H. Sato, L.-J. Chang, R. Blumenthal, and M. A. Martin. 1993. Quantitation of human immunodeficiency virus type 1 infection kinetics. J. Virol. 67:2182-2190.[Abstract/Free Full Text]
10
10. Ho, D. D., A. U. Neumann, A. S. Perelson, W. Chen, J. M. Leonard, and M. Markowitz. 1996. Rapid turnover of plasma virions and CD4 lymphocytes in HIV-1 infection. Nature 373:123-126.
11
11. Igarashi, T., C. Brown, A. Azadegan, N. Haigwood, D. Dimitrov, M. A. Martin, and R. Shibata. 1999. Human immunodeficiency virus type 1 neutralizing antibodies accelerate clearance of cell-free virions from blood plasma. Nat. Med. 5:211-216.[CrossRef][Medline]
12
12. Jin, X., D. E. Bauer, S. E. Tuttleton, S. Lewin, A. Gettie, J. Blanchard, C. E. Irwin, J. T. Safrit, J. Mittler, L. Weinberger, L. G. Kostrikic, L. Zhang, A. S. Perelson, and D. D. Ho. 1999. Dramatic rise in plasma viremia after CD8+ T cell depletion in simian immunodeficiency virus-infected macaques. J. Exp. Med. 189:991-998.[Abstract/Free Full Text]
13
13. Johnson, V. A., and B. D. Walker. 1990.HIV-infected cell fusion assay. In B. D. Walker (ed.), Techniques in HIV research. Stockton Press, New York, N.Y.
14
14. Kaech, S. M., and R. Ahmed. 2001. Memory CD8+ T cell differentiation: initial antigen encounter triggers a developmental program in naive cells. Nat. Immunol. 2:415-422.[Medline]
15
15. Kuroda, M. J., J. E. Schmitz, W. A. Charini, C. E. Nickerson, M. A. Lifton, C. I. Lord, M. A. Forman, and N. L. Letvin. 1999. Emergence of CTL coincides with clearance of virus during primary simian immunodeficiency virus infection in rhesus monkeys. J. Immunol. 162:5127-5133.[Abstract/Free Full Text]
16
16. Lichterfeld, M., D. E. Kaufmann, X. G. Yu, S. K. Mui, M. M. Addo, M. N. Johnston, D. Cohen, G. K. Robbins, E. Pae, G. Alter, A. Wurcel, D. Stone, E. S. Rosenberg, B. D. Walker, and M. Altfeld. 2004. Loss of HIV-1-specific CD8 T cell proliferation after acute HIV-1 infection and restoration by vaccine-induced HIV-1-specific CD4+ T cells. J. Exp. Med. 200:701-712.[Abstract/Free Full Text]
17
17. Lyons, A. B. 2000. Analyzing cell division in vivo and in vitro using flow cytometric measurement of CFSE dye dilution. J. Immunol. Methods 243:147-154.[CrossRef][Medline]
18
18. Mackall, C. L., T. A. Fleisher, M. R. Brown, M. P. Andrich, C. C. Chen, I. M. Feuerstein, M. E. Horowitz, I. T. Magrath, A. T. Shad, S. M. Steinberg et al. 1995. Age, thymopoiesis, and CD4+ T-lymphocyte regeneration after intensive chemotherapy. N. Engl. J. Med. 332:143-149.[Abstract/Free Full Text]
19
19. Mantano, T., R. Shibata, C. Siemon, M. Connors, H. C. Lane, and M. A. Martin. 1998. Administration of an anti-CD8 monoclonal antibody interferes with the clearance of chimeric simian/human immunodeficiency virus during primary infection of rhesus macaques. J. Virol. 72:164-169.[Abstract/Free Full Text]
20
20. McKinney, D. M., D. A. Lewinsohn, S. R. Riddell, P. D. Greenberg, and D. E. Mosier. 1999. The antiviral activity of HIV-specific CD8+ CTL clones is limited by elimination due to encounter with HIV-infected targets. J Immunol. 163:861-867.[Abstract/Free Full Text]
21
21. McLean, A., and C. A. Michie. 1995. In vivo estimates of division and death rates of human T-lymphocytes. Proc. Natl. Acad. Sci. USA 92:3707-3711.[Abstract/Free Full Text]
22
22. Ottinger, H. D., D. W. Beelen, B. Scheulen, U. W. Schaefer, and H. Gross-Wilde. 1996. Improved immune reconstitution after allotransplantation of peripheral blood stem cells instead of bone marrow. Blood 88:2775-2779.[Abstract/Free Full Text]
23
23. Pakker, N. G., D. W. Notermans, R. J. de Boer, M. T. Roos, F. de Wolf, A. Hill, J. M. Leonard, S. A. Danner, F. Miedema, and P. T. Schellekens. 1998. Biphasic kinetics of peripheral blood T cells after triple combination therapy in HIV-1 infection: a composite of redistribution and proliferation. Nat. Med. 4:208-214.[CrossRef][Medline]
24
24. Pantaleo, G., and A. S. Fauci. 1993. HIV-1 infection in the lymphoid organs: a model for disease development. J. NIH Res. 5:68.
25
25. Press, W. H., B. P. Flannery, S. A. Teukolsky, and W. T. Vetterling. 1988. Numerical recipes in C. Cambridge University Press, Cambridge United Kingdom.
26
26. Sachsenberg, N., A. S. Perelson, S. Yerly, G. A. Schockmel, D. Leduc, B. Hirschel, and L. Perrin. 1998. Turnover of CD4+ and CD8 T lymphocytes in HIV-1 infection as measured by Ki-67 antigen. J. Exp. Med. 187:1295-1303.[Abstract/Free Full Text]
27
27. Schmitz, J. E., M. J. Kuroda, S. Santra, V. G. Sasseville, M. A. Simon, M. A. Lifton, P. Racz, K. Tenner-Racz, M. Dalesandro, B. J. Scallon, J. Ghrayeb, M. A. Forman, D. C. Montefiori, E. P. Rieber, N. L. Letvin, and K. A. Reimann. 1999. Control of viremia in simian immunodeficiency virus infection by CD8 lymphocytes. Science 283:857-860.[Abstract/Free Full Text]
28
28. Tough, D. F., and J. Sprent. 1995. Lifespan of lymphocytes. Immunol. Res. 14:1-12.[Medline]
29
29. van Baalen, C. A., C. Guillon, M. van Baalen, E. J. Verschuren, P. H. M. Boers, A. D. M. E. Osterhaus, and R. A. Gruters. 2002. Impact of antigen expression kinetics on the effectiveness of HIV-specific cytotoxic T lymphocytes. Eur. J. Immunol. 32:2644-2652.[CrossRef][Medline]
30
30. Walker, B. D., C. Flexner, K. Birch-Limberger, L. Fisher, T. J. Paradis, A. Aldovini, R. Young, B. Moss, and R. T. Schooley. 1989. Long-term culture and fine specificity of human cytotoxic T-lymphocyte clones reactive with human immunodeficiency virus type 1. Proc. Natl. Acad. Sci. USA 86:9514-9518.[Abstract/Free Full Text]
31
31. Wei, X., S. K. Ghosh, M. E. Taylor, V. A. Johnson, E. A. Emini, P. Deutsch, J. D. Lifson, S. Bonhoeffer, M. A. Nowak, B. H. Hahn et al. 1995. Viral dynamics in human immunodeficiency virus type 1 infection. Nature 373:117-122.[CrossRef][Medline]
32
32. Wick, D. 1999. On T-cell dynamics and the hyperactivation theory of AIDS pathogenesis. Math. Biosci. 158:127-144.[CrossRef][Medline]
33
33. Wick, D., and S. G. Self. 2000. Early HIV infection in vivo: branching-process model for studying timing of immune responses and drug therapy. Math. Biosci. 165:115-134.[CrossRef][Medline]
34
34. Wick, D., S. G. Self, and L. Corey. 2002. Do scarce targets or T killers control primary HIV infection? J. Theor. Biol. 214:209-214.[CrossRef][Medline]
35
35. Wick, D., and S. G. Self. 2002. What's the matter with HIV-directed killer T-cells? J. Theor. Biol 219:19-31.[Medline]
36
36. Wick, D., and S. G. Self. 2004. On simulating strongly-interacting biological populations. II. Multiple compartments. Math. Biosci. 190:127-143.[CrossRef][Medline]
37
37. Wick, D., and S. G. Self. How fast can HIV escape from immune control? In W.-Y. Tan and H. Wu (ed.), Deterministic and stochastic models of AIDS epidemics and HIV infections with intervention. World Scientific, Hackensack, N.J., in press.
38
38. Wick, W. D., and S. G. Self. On fitting deterministic dynamical models to data with small active noise. Submitted for publication.
39
39. Yang, O. O., P. T. Sarkis, A. Trocha, S. K. Kalams, R. P. Johnson, and B. D. Walker. 2003. Impacts of specificity and avidity on the antiviral function of HIV-1-specific CTL. J. Immunol. 171:3718-3724.[Abstract/Free Full Text]
40
40. Yang, O. O., S. A. Kalams, M. Rosenzweig, A. Trocha, N. Jones, M. Koziel, B. D. Walker, and R. P. Johnson. 1996. Efficient lysis of human immunodeficiency virus type 1-infected cells by cytotoxic T lymphocytes. J. Virol. 70:5799-5806.[Abstract]
41
41. Yang, O. O., S. A. Kalams, A. Trocha, H. Cao, A. Luster, R. P. Johnson, and B. D. Walker. 1997. Suppression of human immunodeficiency virus type 1 replication by CD8⁺ cells: evidence for class I-restricted triggering of cytolytic and noncytolytic mechanisms. J. Virol. 71:3120-3128.[Abstract]

---

Journal of Virology, November 2005, p. 13579-13586, Vol. 79, No. 21
0022-538X/05/$08.00+0     doi:10.1128/JVI.79.21.13579-13586.2005
Copyright © 2005, American Society for Microbiology. All Rights Reserved.

This article has been cited by other articles:

• Wick, W. D., Gilbert, P. B., Yang, O. O. (2009). Predicting the Impact of Blocking Human Immunodeficiency Virus Type 1 Nef In Vivo. J. Virol. 83: 2349-2356 [Abstract] [Full Text]  
• Ganusov, V. V., De Boer, R. J. (2008). Estimating In Vivo Death Rates of Targets due to CD8 T-Cell-Mediated Killing. J. Virol. 82: 11749-11757 [Abstract] [Full Text]  
• Mandl, J. N., Regoes, R. R., Garber, D. A., Feinberg, M. B. (2007). Estimating the Effectiveness of Simian Immunodeficiency Virus-Specific CD8+ T Cells from the Dynamics of Viral Immune Escape. J. Virol. 81: 11982-11991 [Abstract] [Full Text]  
• De Boer, R. J. (2007). Understanding the Failure of CD8+ T-Cell Vaccination against Simian/Human Immunodeficiency Virus. J. Virol. 81: 2838-2848 [Abstract] [Full Text]  
• Regoes, R. R., Barber, D. L., Ahmed, R., Antia, R. (2007). Estimation of the rate of killing by cytotoxic T lymphocytes in vivo. Proc. Natl. Acad. Sci. USA 104: 1599-1603 [Abstract] [Full Text]  

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Reprints and Permissions Copyright Information Books from ASM Press MicrobeWorld Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Wick, W. D. Articles by Self, S. G. Search for Related Content PubMed PubMed Citation Articles by Wick, W. D. Articles by Self, S. G.