Technical Comments

T Cell Turnover in SIV Infection

Science  23 Apr 1999:
Vol. 284, Issue 5414, pp. 555
DOI: 10.1126/science.284.5414.555a

In their report of 20 February 1998 (1), H. Mohri et al. analyze the proliferation and death of T cells in normal macaques and in those infected with simian immunodeficiency virus (SIV). With the use of bromodeoxyuridine (BrdU) to label proliferating cells, they observed that the uptake of BrdU by both CD4+ and CD8+ T cells was as much as threefold faster in infected than in uninfected macaques during the 3-week labeling period. After cessation of BrdU administration, the fraction of labeled cells gradually decreased, which indicated that labeled cells were dying more rapidly than they were proliferating. Because Mohri et al. assume that the labeled cells were indicative of the entire pool of T cells that acquired the label during the experiment, and that the T cell population as a whole was essentially at steady state during the study, they conclude that the difference between the rates of death and proliferation must be compensated for by a supply of new T cells (“replacement”) to the pool, possibly from the thymus. Mohri et al. use a mathematical model and adjust its parameters—the various rates of death, proliferation, and replacement—to fit the kinetic data.

One of the assumptions in this model is that all T cells in the pool, labeled and unlabeled alike, are uniformly proliferating at some constant rate and dying at a higher constant rate [see figure 2 in the report (1)]). This assumption is incorrect. It is quite likely that many (probably most) of the labeled T cells in infected macaques had divided in response to immunologic stimulation by their cognate antigen, or by induced cytokines, or by both. It is now well known that such activated cell populations initially expand; the great majority of cells in these populations then die (mainly by apoptosis mediated by fas and fas ligand). Other activated T cells may be converted into “resting” memory cells, or enter a state of anergy in which their proliferation ceases although the cells survive. Thus, it is quite likely that Mohri et al. were mainly tracking the short-term expansion (during labeling) and contraction (during delabeling) of immunologically stimulated cell populations, rather than a stationary turnover of the entire T cell population. When BrdU was no longer administered, peripheral T cells other than those that had earlier been labeled were preferentially activated, in response to new stimuli. The preferential expansion of unlabeled cells is a more likely explanation for the washout dilution of labeled cells than is an influx of new cells from the thymus or some other source. In uninfected macaques, T cell proliferation may be associated to a larger degree with physiologic regeneration and less with immune activation, so that the labeled cells would be more representative of the population as a whole. Indeed, in some of these animals there was almost no delabeling [figure 1 in (1)].

A simple model (Fig. 1) of normal, ongoing immune activation (2) can account (Fig. 2) for the rapid rise in the fraction of labeled cells during the labeling period and for the rapid decline thereafter (3). A crucial aspect of our model is that the processes of T cell regeneration, expansion, and death are largely associated with different sequential compartments (stages of activation and maturation). When labeling of dividing cells occurs for a finite length of time, sufficient only for a fraction of the cells to become labeled, cells in the activated compartments are disproportionally labeled as compared with “resting” cells. The “growing-while-aging” structure of activated cells (Fig. 1) implies that the selected cohort of labeled cells must be essentially transitory. This transitory aspect should not be interpreted to reflect unbalanced overall birth and death rates in the T cell population as a whole or even in the activated cell population (4).

Figure 1

Schematic representation of T cell regeneration and activation in our model. Variables (circled) and parameters are defined in (2).

Figure 2

Changes in the fraction of BrdU-labeled CD4+ T cells during 3 weeks of labeling and 7 more weeks of no labeling [data from Mohri et al. (1)]. Measurements shown are from two representative macaques: (▪), data from an infected macaque; (▵), from uninfected macaque. Curves were generated with the use of our mathematical model; details in (3).

Increased immunologic stimulation in infected individuals most likely accounts for “increased turnover” rates in both CD4 and CD8 cells. Mohri et al. (1) mention some well-documented manifestations of immune activation (p. 1226), but their model (pp. 1223–1224) describes only regenerative proliferation, in which all the cells have the same proliferative capacity, which remains constant over time. This led Mohri et al. to ask questions about the mechanism responsible for the apparent “killing” of both CD4 and CD8 lymphocytes in infected individuals, in excess of proliferation, and about the differential depletion of CD4 cells (1, p. 1226): “Why then is CD4 T cell depletion observed and not CD8 T cell depletion … ?” But if our interpretation is correct, the death of immune activated cells serves to control their earlier accelerated proliferation and need not affect the equilibrium condition of the population as a whole. Accordingly, the death of labeled cells has little bearing on the issue of depletion.


Mohri et al. (1) studied turnover of T cells in SIV-infected macaques by adding BrdU to drinking water of the animals. If their interpretation of BrdU labeling kinetics is correct, then rapid direct or indirect killing of T cells by SIV is compensated, to a large extent, not by proliferation of other T cells, but through supply of new cells from an as-yet-undefined source.

At issue is the explanation of the high rate of loss of BrdU-positive cells after the labeling has stopped. At that stage, unlabeled cells must divide into unlabeled cells, and labeled cells will generate labeled progeny. Given the quasi-steady state that characterizes SIV-infected animals, if proliferation of T cells almost fully compensates for their loss, then rather than decreasing rapidly, as observed, the fraction of BrdU-positive cells should have remained almost constant until the label per cell becomes diluted to a point below the limit of detection, a period much longer than the cell turnover time. To account for this paradox, Mohri et al. postulate that replenishment of T cells is partially provided by a source that has labeled and unlabeled components [equations 1 and 2 in (1), pp. 1223 and 1224], whose relative intensities are constant during the labeling phase [note 11 in (1), p. 1227], and that the labeled source component suddenly becomes very small when external labeling has stopped (S L = 0 to obtain equation 3, p. 1224; note 14, p. 1227). Mohri et al. do not define a biologically consistent model of a source with such hypothetical properties (note 11, p. 1227), although this was a central point in the interpretation of their experiments.

We propose an alternative and simpler explanation, based on dilution of the BrdU label which will occur, according to a model of T cell replenishment (2), as follows. A resting T cell in vivo, on receiving an activation signal, may go through many cycles of division before resuming the resting state. For example, six cycles, each of 1 day or less, would be sufficient to dilute the label per cell below the gate value within 1 week. (Immune activation of T cells in vitro offers a biological example of such proliferation mechanism. The proliferation loop may involve additional cell types, such as effector cells.) In the quasi-steady state, such multiple, rapid proliferation bursts should be compensated by the average rate of death of activated primed T cells. When a labeled cell becomes activated during (or shortly before) the follow-up period after labeling has stopped, this cell and its progeny necessarily “disappear” (become undetectable) within a week, and are replaced, on average, by an unlabeled cell. As a result, the fraction of labeled cells will decay on a time scale of a few weeks, according to the estimated activation rate that can be inferred from the rate of labeling during the labeling phase.

Mohri et al. have argued against this model (loss of labeled cells resulting from dilution of label) as an explanation for the observed decline in the labeled fraction because their “BrdU-intensity data did not reveal any substantially lower values with the passage of time” [note 13 in (1), p. 1227]. (A slow decline in the BrdU amount per labeled cell would be expected in the one-division model used by Mohri et al.) In fact, the multiple rapid division model, although based on dilution, does not predict a gradual decline in the amount of BrdU per cell. Instead, the fraction of labeled cells activated within a time interval is predicted to fall swiftly (within a few days) below the BrdU intensity gate value. Such a fast process is difficult to detect, especially with the wide sampling interval of 1 week used by Mohri et al. The level of BrdU intensity among labeled cells that are not activated yet will always remain high, but the number of such cells will decline according to the activation rate.

We offer these considerations in the hope of encouraging further studies into the nature of the enhanced turnover of different lymphoid cells imposed by SIV in infected animals and by HIV in humans.


Response: We have shown that after BrdU is administered to SIV-infected or healthy macaques for 3 weeks, a population of BrdU-labeled cells is created, which slowly decays (1). Because the progeny of BrdU-labeled cells should remain labeled until a sufficiently large number of divisions have occurred to dilute out the label, we interpreted this loss as suggesting that the subpopulation of cells that were labeled during the 3 weeks of BrdU administration died more rapidly than they were produced. This further implied that in steady state there must be a source of cells that enter this subpopulation.

Grossman et al. and Rouzine and Coffin both state that the subpopulation we studied consists of cells stimulated, possibly by antigen, SIV, or induced cytokines, into rapid division, and which would die rapidly (that is, by Fas/Fas-ligand mediated apoptosis). Although such events undoubtedly occur, the question is whether they are the dominant population or only a subpopulation of the cells under study. Other experiments will be required to answer this definitively, but our impression is that if the subpopulation that we tracked were dominated by cells undergoing short-term antigen-stimulated expansion and contraction, with division rates of a day or less (Rouzine and Coffin), then it is difficult to see why the fraction of labeled cells graduately increases over the 3-week labeling period rather than going through spikes and dips. Further, over the 3-week labeling period, as many as 30 to 40% of CD4+ T cells label in some SIV-infected animals. Because the antigen specificity of these cells was not determined, we cannot address the question of whether these cells derive from a few stimulated clones or are broadly representative of the CD4+ T cells in the animals. But no illness was observed in the animals and no immunizations were performed. Moreover, immune stimulated cells tend to die rapidly, yet even 7 weeks after labeling has stopped, labeling cells still exist and their decay remains on the same straight line in a semi-logarithmic plot. The rate of delabeling, according to our model is d − p. Because p is generally small, our fitting procedure is unlikely to greatly underestimate the death rate. On the basis of the slope of delabeling, the average death rates of both infected and healthy monkeys correspond to cell half-lives of 2 weeks or more, not the few days required by the scenarios of Grossman et al. and Rouzine and Coffin. In some instances, BrdU is toxic to cells. If some toxicity were present, then the natural half-lives would be even longer.

The model proposed by Grossman et al. has several shortcomings. First, for the chosen parameters, the total cell population in infected animals is not in steady state, but in a slow but unbounded exponential growth phase. Second, in uninfected animals, the resting but activatable cell population (denoted by R in their model) would disappear in the long term, since their death rate exceeds their proliferation rate. Grossman et al. appear to misinterpret our model when they say that we assume that all T cells in the pool are uniformly proliferating at a constant rate. Our model describes the mean behavior of a population of cells. Thus, individual cells can proliferate with different rates, or not at all. Our proliferation rate, p, and the death rate, d, are meant to be the average proliferation and death rates, respectively, of the CD4+ cells measured in the periphery. Modeling in this way has the advantage of introducing only a small number of parameters, as opposed to the 15 parameters used in the model by Grossman et al.

Grossman et al. suggest that the turnover that we have measured has little bearing on the issue of CD4+ T cell depletion in HIV or SIV infection. We disagree, and we show (Fig. 1) new data collected at 22 weeks, that is, 19 weeks after the stop of labeling, of our original experiment (1). The percentage of CD4+ cells that were BrdU+ at 10 weeks that are still BrdU+ at 22 weeks versus baseline CD4 count for the monkeys we studied are shown (Fig. 1). Monkeys with low CD4 count (that is, SIV-infected ones) have greatly enhanced loss of labeled CD4 cells as compared with monkeys with high CD4 counts (uninfected ones). If this loss is a result of label dilution, as argued by Rouzine and Coffin, then these results suggest that SIV infection causes enhanced proliferation, whereas if the loss is a result of the rate of cell death being greater than the rate of proliferation, as in our model, then the results show increased death. In either event, the biological conclusion is that SIV infection causes enhanced CD4+ T cell turnover.

Figure 1

Decline of BrdU+ cells during week 10 and 22 correlates with the baseline of CD4 cell count. Data points indicate the percentage of CD4 and CD8 cells that were BrdU+ at week 10 that are still BrdU+ at week 22. Correlation of the decline of BrdU+ cells versus the baseline CD4 cell count are shown. Solid line and dot line are the curve fitting for the data of CD3+CD4+ cells and CD3+CD4 cells only from infected monkeys.

In summary, we agree with both comments that immune stimulation in SIV infection may be elevated and may be the cause of the increased proliferation of all of the various lymphoid populations that we measured. However, because as many as 30 to 40% of T cells are labeled after 3 weeks and no antigen was intentionally given to stimulate a response, we believe that the majority of cells we monitored are representative of the general T cell population, although a minority may be antigen-activated. As explained in reference note 11 of our report (1), our model is consistent with the possibility that the source for cells moving into the subpopulation under study could include a population of resting or slowly dividing T cells, which upon activation would undergo rapid clonal expansion and acquire label (4). Thus, the clonal expansion that Grossman et al. and Rouzine and Coffin are concerned about may be the source. The input of cells that we described from the “source” corresponded to a 1% replacement rate in uninfected animals and 2.5% in highly infected animals, percentages that are consistent with Ki67 labeling and other measures of immune activation that suggest ∼1% lymphocytes are activated at a given time in healthy individuals, increasing three-fold during HIV-1 infection (2–3).


Navigate This Article