Subpopulations of long-lived and short-lived T cells in advanced HIV-1 infection (original) (raw)

The immune system may be divided into populations of cells that are functionally and kinetically distinct: long-lived progenitor cells with the capacity to produce many progeny and end-stage effector cells destined to die quickly. As is the case in epithelial tissues or the remainder of the hematopoietic system, progenitor cells are necessary for the maintenance of tissue mass in the face of continuous loss of differentiated progeny. T cell progenitors are found at varying levels of the differentiation tree and include multilineage hematopoietic stem cells in the bone marrow, T lineage-restricted progenitors in the thymus, and naive and “true memory” T cells in peripheral lymphoid organs, such as spleen and lymph nodes. In vivo studies in rodents have documented the presence of kinetically distinct subpopulations of T cells, including long-lived populations of cells that retain tritiated thymidine (26, 27), but analogous T cell subpopulations had not previously been demonstrated in humans. Because specific markers are not currently available to identify different types of memory cells, we instead used differences in dynamic behavior of cells within the surface phenotype-defined m/e-T cell population to establish the existence and quantify the relative abundances of these postulated T cell subpopulations. Although we could not physically isolate effector- or memory-type T cells by this approach, our results clearly demonstrate their presence, quantify their relative contributions in healthy people and in advanced HIV-1 infection, and provide a biomarker for future efforts to identify biochemical correlates and physically isolate these subpopulations.

The combined use of two novel, stable isotope-labeling techniques provided clear evidence for T cell subpopulations with distinctly different life spans and for differential effects of advanced HIV-1 infection on these subpopulations. The two stable isotope-labeling approaches provided complementary and internally consistent results. The long-term 2H2O incorporation curves demonstrated nonlinear or biphasic accrual of total and m/e-phenotype but not naive-phenotype T cells, consistent with the presence of short-lived and longer-lived subpopulations within the m/e-phenotype T cell pool. These results from 2H2O labeling also excluded alternative explanations, such as a late-input source (15), for biphasic die-away curves (see below). The biphasic die-away kinetics of total and m/e-phenotype T cells complemented the long-term label incorporation data. The persistent label retention in a subset of T cells after a prolonged die-away period (Figure 3) confirmed that short-lived and much longer-lived T cells are present in humans. Long-term labeling also allowed identification of quiescent (nondividing or rarely dividing) T cells, functionally distinguished by their failure to divide over 9 weeks of label administration (Figure 2d). Because these cells do not contribute to either immune restoration or immune depletion, the ability to account for their presence is important and is not possible by use of brief labeling protocols. These results provide the first quantitative evidence that long-lived and quiescent T cells do indeed predominate in the T cell pool in humans and determine T cell pool size, as in rodents (1, 2, 8), unlike in other tissues in which short-lived cells predominate and determine pool size. Finally, both labeling approaches indicated that the greatest impact of advanced HIV-1 infection is to reduce the generation of long-lived, potential progenitor T cells, whereas effective LT ARV therapy restores the capacity to generate these cells.

The studies described here are methodologically novel in two ways. Most importantly, these are the first long-term labeling studies (i.e., more than a few days or a week of label administration) of T cells in humans. To observe multiphasic kinetics and thereby tease out cells with different turnover rates in any system by use of a continuous labeling strategy, it is necessary to administer label for a considerably longer period than the turnover time of the short-lived pool (i.e., 2–3 weeks for T cells) (23, 24). This had not been possible before the development of the 2H2O-labeling technique (10, 21–23). Second, these are the first studies of long-term label retention in the m/e-surface phenotype compartment of human T cells. Previous studies of T cell delabeling in humans (15, 16) had only analyzed total CD4+ or CD8+ T cells. By combining the 2H2O technique (10) with 2H-glucose incorporation/die-away measurements and by using flow cytometry to isolate T cells of defined surface phenotype (9, 14), the existence and dynamics of cells with short life span and long life span in the m/e-phenotype compartment of the human T cell system could be characterized for the first time.

An important point to emphasize about the techniques used here is that they reveal heterogeneity only in cell death rates, not in proliferation rates. If proliferation rates varied but death rates were constant in T cell subpopulations, the consequence at steady state would be a difference in pool size but no differences in either fractional replacement or die-away rates (13, 24, 28). Stated differently, the fraction of each subpopulation replaced per day and their subsequent life span would be the same for rapidly proliferating and slowly proliferating subpopulations, and both label incorporation and decay curves would be monoexponential (13, 24). Heterogeneity in T cell proliferation rates would hardly be a novel finding (e.g., related to differences in antigenic stimulation). In contrast, the clear finding of heterogeneity in the fate of T cells (their survival or death) within the phenotypically defined m/e T cell populations is novel in humans and has potential functional importance.

Using BrdU labeling and decay curves, Kovacs et al. (16) also recently concluded that rapidly dividing and slowly dividing subpopulations are present in the total T cell pool in humans. Although this was a well-designed study, interpretation of BrdU die-away curves has long been recognized to be highly confounded (reviewed in ref. 13). In particular, retention of BrdU positivity of cells can persist during several generations of cell division. BrdU-label retention cannot therefore be taken to represent a long life span or absence of division in a population of cells (13). Moreover, BrdU die-away curves do not exhibit monoexponential decay kinetics because persistent BrdU positivity is followed by reduction below a threshold detection level. Biphasic or multiphasic BrdU die-away curves therefore cannot be interpreted in a simple manner (13, 25). These problems do not apply to die-away after deuterium labeling of DNA, which accurately quantifies dilution by unlabeled DNA during semiconservative replication (13) and which exhibits monoexponential die-away curves in single pool systems (13, 15). Also, Kovacs et al. (16) did not analyze m/e- and naive-phenotype T cells separately, and our results demonstrate that kinetically distinct subpopulations are apparent within the m/e-phenotype population.

Mohri et al. (15) recently reported that the fractional death rate was 3 to10-fold greater than the fractional proliferation rate of T cells in both HIV-1–uninfected and –infected humans. When label in a metabolic pool is lost at a higher fractional rate than label is incorporated into the total pool, the traditional kinetic interpretation is that the labeled molecules or cells do not have an identical fate as the total pool (24, 25, 28), i.e., that kinetically distinct subpopulations are present. Mohri et al. (15) instead proposed a model wherein a “source” contributes cells that are initially unlabeled. Over longer-term labeling, these “source” cells would, of course, have to divide and would therefore slowly contribute labeled cells to the peripheral pool in a manner that should increase over time. The long-term 2H2O-labeling results shown here (Figure 2) provide direct evidence against the existence of this postulated “source,” i.e., we did not observe accelerated late incorporation of label, as would be predicted if there were delayed input of labeled cells from a large tissue source. Rather, we observed the opposite (reduced late incorporation). More recently, the data of Mohri et al. (15) have been reanalyzed by Ribeiro et al. (29) using a revised model that includes populations of activated (dividing and dying) and resting cells, in place of a “source.” This revised model (29) is more consistent with our analysis and results. Three weeks after discontinuing 2H-glucose infusions, the fractions of CD4+ and CD8+ T cells remaining were 57% ± 16% and 78% ± 28%, respectively, in healthy controls compared with 30% ± 12% and 40% ± 15% in untreated HIV-1/AIDS patients; 5 to 7 weeks after discontinuing 2H-glucose, the values for CD4+ and CD8+ T cells were 46% ± 6% and 56% ± 21% in controls and 20% ± 8% and 28% ± 15% in HIV-1/AIDS (15). Moreover, the die-away curves generally reached a stable, non-zero value (15). Comparison of proliferation with death rate constants of Mohri et al. (15) suggests that 6% to 8% of the T cell pool is actively dividing in healthy controls and 16% to 33% in untreated HIV-1/AIDS patients. All of these values are similar to the results reported here (compare with Figure 2d).

Recognizing the distinction between long-lived (producing) cells and short-lived (presumably effector) cells is essential for understanding CD4+ T lymphopenia, the defining feature of HIV-1/AIDS. Our results show that the generation of short-lived subpopulations of T cells was not impaired in advanced HIV-1/AIDS. Indeed, the absolute rate of proliferation for short-lived cells is somewhat elevated in advanced HIV-1/AIDS compared with healthy controls (approximately 6 vs. 3 cells/μl of blood per day, Figure 4a). The presence of an increased proportion or total number of cycling cells in HIV-1 infection is not a new observation, having been reported using indirect methods such as Ki67 (30) in addition to labeling approaches (9, 14–16). Our quantitative results regarding production rates of long-lived cells, however, are novel and of fundamental relevance to HIV-1 immunopathogenesis. Advanced HIV-1 infection greatly reduced the percent and total number of CD4+ T cells that are long-lived (Figure 4). Because these cells represent the regenerative source of newly formed CD4+ effector T cells, their loss may underlie the immunodeficiency of HIV-1 disease. It should be emphasized that the HIV-1/AIDS subjects studied here had relatively advanced HIV-1 disease, with CD4+ T cell lymphopenia. These kinetic abnormalities may not be present in early HIV-1 infection and may represent a marker of disease stage or activity (ref. 31; M.K. Hellerstein and J.M. McCune, unpublished observations).

Indeed, death of cells that might otherwise have been long-lived is sufficient to account for CD4+ T lymphopenia of HIV-1/AIDS. By reducing 2 cells/μl/d, the production of long-lived m/e-phenotype cells that have an average life span of 3 to 6 months would reduce the T cell pool by 200 to 400 cells/μl. In contrast, the production of short-lived m/e-phenotype T cells, by 5 cells/μl/d (life span 2–3 weeks), can increase the T cell pool by only 70 to 100 cells/μl. Thus, not all forms of T cell turnover are equal: some T cell proliferative pathways can effectively contribute to immune repletion, whereas others cannot (7, 8, 17). Combadere et al. (30) recently reported that more than 90% of Ki67-positive T cells in HIV-1–infected patients represent cytokine-secreting effector-like cells that are not progressing through the cell cycle, but are arrested in G1-phase. Cells of this type could not contribute to CD4+ T cell repletion.

The kinetics of naive-phenotype T cells also deserves comment. In contrast to m/e-phenotype T cells, naive-phenotype T cells did not show obvious biphasic or nonlinear kinetics either during long-term label incorporation or decay (Figures 2a and 3a). Significantly higher replacement rates were observed for naive T cells in untreated HIV-1 infection, compared with healthy controls or LT ARV-treated patients (Figure 2d) and there was considerable interindividual variability for naive T cell kinetics in the untreated HIV-1–infected group (note SDs in Figure 2a), consistent with our previous observations (9). The interpretation of accelerated naive T cell turnover in advanced HIV-1 infection is not certain, but is consistent with a higher rate of recruitment of naive T cells into the m/e population (18). In future studies, it will be of interest to monitor the die-away curves of naive-phenotype T cells that are labeled during long-term 2H2O administration, to establish whether transitional proliferation of naive-phenotype T cells en route to the m/e-pool explains higher flow-through rates in the naive T cell pool in HIV-1 infection.

The results described here are potentially relevant to the pathogenic locus of action of HIV-1. Two key findings argue against the simple model that the rapid die-away of a subpopulation of CD4+ T cells is explained by direct HIV-1 killing of dividing m/e-phenotype T cells. First, CD8+ and CD4+ m/e-phenotype T cells exhibited similar biphasic kinetics, with a greater die-away fraction, and a reduced long-lived fraction of new cells produced for both. It is unlikely that HIV-1–induced cytopathicity could account for premature CD8+ T cell death of this magnitude (32–35). Second, healthy HIV-1–seronegative control subjects also exhibited biphasic kinetics and a subset of short-lived cells in their m/e-phenotype T cell pool. It is therefore possible to offer an alternative hypothesis to direct HIV-1 cytopathicity; namely, that some feature of HIV-1 infection alters T cell developmental pathways, reducing the fraction of long-lived (progenitor) cells relative to short-lived (presumably effector) cells generated. Whether the causal agent is HIV-1 itself or the effect is secondary to immune activation (7, 17, 18, 36) or other factors cannot be determined from the current data. Experimental testing of these questions is now possible, however, using the kinetic techniques described here.

Alternate explanations of these results deserve consideration. Grossman et al. (6, 17) have proposed that whole-body T cell dynamics are characterized by cumulative, unsynchronized local bursts of T cell division and death. If this model is correct and the fate of most amplificatory bursts is activation-induced cell death, long-term labeling with 2H2O might result in a relatively small cumulative fraction of labeled cells remaining in the T cell pool. Our observation that more than 50% of m/e-phenotype T cells are labeled after 9 weeks of 2H2O administration and more than 40% are labeled after only 5 weeks (Figure 2d) in advanced HIV-1 infection signifies that unsynchronized bursts, if they occur, must directly involve a very high fraction of the total m/e-phenotype T cell pool in the body. These bursts must also result in substantial replacement of the cells initially present by surviving postburst progeny. Our results therefore demonstrate that the subset of T cells that is highly activated and rapidly proliferating in patients with advanced HIV-1 infection cannot represent a minority of the total-body m/e-phenotype T cell pool, whether the model is of asynchronous bursts or uniform and constant proliferation.

Finally, the techniques described here for identifying T cell subpopulations with different life spans may have applications for characterizing the normal antigen-driven T cell proliferative response. Testing the kinetic fate of cells carrying certain surface antigens (e.g., CCR-7 or other markers) (4, 37) may help to determine whether these phenotypic markers represent true memory versus effector lineages within dividing T cells. Establishing the consequences of various pathologic or therapeutic factors (e.g., cytokines, progenitor cell availability, etc.) on each phase of the proliferative responses of antigen-specific T cells is also of potential interest for vaccine research (1–5).