Expression of CD28 and CD38 by CD8⁺ T lymphocytes in HIV-1 infection correlates with markers of disease severity and changes towards normalization under treatment

Abstract

The relationship between blood CD8⁺ T lymphocyte subsets, as defined by CD28 and CD38 expression, and plasma viraemia and CD4⁺ T cells in HIV-1 infection was investigated. In a cross-sectional study of 46 patients with either no or stable anti-retroviral treatment, there was a strong negative correlation between the percentage of CD8⁺CD28⁻ and the percentage of CD4⁺ T cells (r = − 0.75, P < 0.0001), and a positive correlation between absolute numbers of CD8⁺CD28⁺ and CD4⁺ T cells (r = 0.56, P < 0.0001). In contrast, the expression of CD38 by CD8⁺ T lymphocytes correlated primarily with plasma viraemia (e.g. the percentage of CD38⁺ in CD8^bright cells, r = 0.76, P < 0.0001). In the 6 months following triple therapy initiation in 32 subjects, there was a close correlation between changes (Δ) in CD8⁺CD28⁺ or CD8⁺CD28⁻ and in CD4⁺ T cells (e.g. Δ% CD8⁺CD28⁺versusΔ% CD4⁺, r = 0.37, P = 0.0002; Δ% CD8⁺CD28⁻versusΔ% CD4⁺, r = − 0.66, P < 0.0001). A marked decline of the number of CD8⁺ T cells expressing CD38 was also observed. These results suggest the existence of a T cell homeostasis mechanism operating in blood with CD4⁺ and CD8⁺CD28⁺ cells on the one hand, and with CD8⁺CD28⁻ cells on the other. In addition, the percentage of CD38⁺ cells in CD8⁺ cells, generally considered an independent prognostic factor, could merely reflect plasma viral load.

INTRODUCTION

Blood CD8⁺ T cells are composed of several subsets that can be differentiated by the expression of transmembrane proteins, such as CD28 and CD38. CD28 conveys essential costimulatory signals during antigen presentation to T lymphocytes [1]. CD8⁺CD28⁺ T cells include naive and memory cytotoxic T lymphocytes (CTL) precursors [2]. In HIV-1 infection, these cells also produce, in a non-MHC-restricted way, partly identified anti-HIV soluble factors [3,4]. CD8⁺CD28⁻ T cells are terminally differentiated CTL effectors [2,5,6] which display a major expansion in the blood of HIV-infected persons [7–13]. These expanded effector cells, which contain a high proportion of HIV-specific CTL, may play an important role by efficiently containing HIV replication [4,14] or, on the contrary, by contributing to the loss of CD4⁺ T lymphocytes [15]. Alternatively, these CD8⁺CD28⁻ T cells may also display low cytotoxic effectiveness due to their propensity to undergo apoptosis upon stimulation [16], and their expansion may merely reflect the operation of a blind T cell homeostasis mechanism [17].

CD38, an ecto-enzyme [18] and a counter-receptor of CD31 [19], is present on immature (HLA-DR⁻) and activated (HLA-DR⁺), but not quiescent, T cells [20]. A high proportion of CD8⁺ T cells positive for CD38 expression is known to be a strong prognostic marker of disease progression in HIV-infected subjects [21–28]. This increased expression of CD38 is found in both the CD28⁺ and the CD28⁻ subsets [12].

The possible association between CD8⁺ T lymphocyte subsets defined by CD28 and CD38 expression, and immunological and virological parameters in the course of HIV infection has, to date, received little attention [29–31]. Thus, we have here addressed this question by performing a cross-sectional study involving either untreated or stably treated patients, and by investigating changes in these parameters during a 6-month period following initiation of anti-retroviral treatment.

PATIENTS AND METHODS

Patients

Forty-six patients were included in the cross-sectional study. They were either untreated or had had no change in their anti-retroviral treatment in the preceding 6 months. All had uncontrolled plasma viraemia and were candidates for anti-retroviral treatment initiation or change. Thirty-two of these patients were followed up in different therapeutic studies. Eleven of them received didanosine (200 mg bid) and stavudine (40 mg bid), plus hydoxyurea (500 mg bid) or placebo in a blinded protocol [32]; eight had a combination of stavudine (30–40 mg bid), ritonavir (400–600 mg bid) and saquinavir (600 mg bid); and 13 received various combinations of two nucleoside reverse transcriptase inhibitors (NRTI) and one protease inhibitor. Except for two patients in whom treatment was discontinued, the mean time of follow up was 195 days (range 151–259 days).

Flow cytometry

Flow cytometry was performed using standard procedures, the Coulter Q-Prep apparatus and the Coulter EPICS XL flow cytometer [12]. Three-colour analysis was performed using CD8–FITC/CD4–PE/CD3–ECD or CD8–FITC/CD4–PE/CD3–PC5 from Coulter (Beckman Coulter Int. S.A., Nyon, Switzerland). For two-colour analysis, MoAbs (all from Dako (Zug, Switzerland), except CD28–PE from Becton Dickinson (Basel, Switzerland)) were combined as follows: CD8–FITC/CD3–PE, CD8–FITC/CD28–PE, CD8–FITC/CD38–PE. CD8^bright cells were separated into a CD38^high (referred to here as CD38⁺) and a CD38^low/neg (referred to here as CD38⁻) population according to the isotypic control. All results expressed as percentages refer to the proportion of the cells under investigation among all lymphocytes, except where otherwise indicated.

CD8⁺ cells contain both cytotoxic/suppressor T lymphocytes and some natural killer (NK) cells. The presence of CD8⁺ NK cells was circumvented as follows. Since all CD8⁺CD28⁺ are T cells [33,34], the percentage of CD8⁺CD28⁻ T cells was obtained by calculation: ((% CD8⁺CD28⁻ T) = (% CD8⁺CD3⁺) − (% CD8⁺CD28⁺)) [12]. This strategy cannot be used when CD38 expression is considered, as NK cells may be CD38⁺ [35,36]. In general, NK cells are CD8^dim, whereas most cytotoxic/suppressor T cells are CD8^bright. Therefore, the expression of CD38 was measured only among CD8^bright cells [20].

Absolute lymphocyte counts were obtained by combining flow cytometry data with total leucocyte counts and differentials measured in a haematological analyser (Cell-Dyn 3500R; Abbott, Baar, Switzerland).

Plasma HIV RNA

Plasma viraemia was quantified using the Amplicor HIV-1 Monitor assay (Roche, Basel, Switzerland), with a detection level of 400 copies/ml.

RESULTS

Cross-sectional study

The percentages and absolute numbers of cells in the several lymphocyte subsets investigated, as well as plasma viraemia, are shown in Table 1. The correlation coefficients found between these parameters are summarized in Tables 2 and 3. In particular, there was a weak positive correlation between percent T CD4⁺ and percent T CD8⁺CD28⁺, a very strong negative correlation between percent T CD4⁺ and percent CD8⁺CD28⁻, and a marked positive correlation between absolute numbers of CD4⁺ and CD8⁺CD28⁺ T cells (Table 2). There were some significant—but weaker—associations between CD8⁺CD28⁺ or CD8⁺CD28⁻ subsets and plasma viraemia (Table 2). Table 3 shows very strong correlations between CD8⁺ subsets defined by CD38 expression and plasma viraemia. In comparison, the correlation coefficients between CD4⁺ T cells and plasma viral load were much lower (Table 3).

Table 1.

Cross-sectional study of untreated patients or patients with stable treatment (n = 46). Immunologic and virologic data

Table 2.

Cross-sectional study of untreated patients or patients with stable treatment (n = 46). Relationship between CD8⁺ T subsets defined by CD28 expression and plasma viraemia or CD4⁺ T cells

Table 3.

Cross-sectional study of untreated patients or patients with stable treatment (n = 32). Relationship between CD8⁺ T subsets defined by CD38 expression and plasma viraemia or CD4⁺ T cells

Therapeutic study

The following changes (Δ), expressed as means ± s.d. and maximum (max.), were observed in the 30 patients that were followed up for > 150 days (mean follow-up time, 195 days): an increase in percent T CD4⁺ (+4.7 ± 3.9, max. +14.7), in percent T CD8⁺CD28⁺ (+2.0 ± 4.5, max. +16.4), in percent CD28⁺ among CD8⁺ T cells (+10.5 ± 10.1, max. +31.8), in percent CD8^brightCD38⁻ (+8.9 ± 9.1, max. +34.8), in CD4⁺ T (+136 ± 143, max. +580/μl), and in CD8⁺CD28⁺ (+75 ± 154, max + 401/μl); a decrease in percent T CD8⁺CD28⁻ (−10.1 ± 8.2, max. −27.3), in percent CD8^brightCD38⁺ (−17.6 ± 12.2, max. − 45.0), and in percent CD38⁺ among CD8^bright cells (−24.1 ± 16.4, max. −62.7). Changes in absolute numbers of CD8⁺CD28⁻ (−124 ± 332/μl), CD8^brightCD38⁺ (−310 ± 334/μl), and CD8^brightCD38⁻ (+249 ± 318/μl) T cells were much more variable and ranged from −729 to +959, −1251 to +337 and −159 to +1455/μl, respectively. Plasma viraemia became undetectable (< 400 copies/ml) in 22 subjects, whereas a 1.53 ± 0.90 log₁₀ decrease was observed in the eight others. Figure 1 illustrates the changes observed in one representative patient.

Fig. 1.

Changes observed in a representative patient (no. 28) after treatment initiation at day 0. The treatment consisted of stavudine, lamivudine and nelfinavir. In this patient, changes in the percentages of CD4⁺, CD8⁺CD28⁺, CD8⁺C28⁻, CD8^brightCD38⁺ and CD38^brightCD38⁻ T cells among all lymphocytes were quite similar to the corresponding changes in absolute cell numbers and are therefore not depicted.

As shown in Table 4, changes in the CD8⁺ T subsets defined by CD28 expression were strongly related to changes in CD4⁺ T lymphocytes in a manner similar to that observed in the cross-sectional study.

Table 4.

Changes from baseline during anti-retroviral treatment. Relationship between CD8⁺ T subsets defined by CD28 expression and CD4⁺ T cells

Changes in % CD8^brightCD38⁺, % CD8^brightCD38⁻, % CD38⁺ in CD8^bright cells, and absolute numbers of CD8^brightCD38⁻ cells correlated significantly both with changes in the CD4⁺ subset and log changes in plasma HIV RNA level, computed using a value of 2.60 = log₁₀ 399 for undetectable levels (< 400 copies/ml). Thus, for instance, the correlation coefficient between Δ% CD8^brightCD38⁺ and Δ% T CD4⁺ was −0.49, while the correlation coefficient between the same Δ% CD8^brightCD38⁺ and Δlog plasma viraemia was 0.48, both with P < 0.0001 (other data not shown). Figure 2 illustrates such changes in four patients. Patient 12 discontinued successful treatment on his own initiative after 35 days, and the plasma viraemia and percentage of CD38⁺ among CD8^bright cells returned to baseline values within 20 days (Fig. 2).

Fig. 2.

Changes in the percentage of CD38⁺ cells among CD8^bright cells, compared with the evolution of plasma viraemia and the increase in the CD4⁺/CD8⁺ ratio. Patient 12 (□) received stavudine, ritonavir and saquinavir, and discontinued treatment on his own initiative after the third visit. Patients 18 (▪) and 19 (○) also received stavudine, ritonavir and saquinavir, and patient 27 (▵) stavudine, lamivudine and ritonavir.

DISCUSSION

In the cross-sectional part of this study, we found that CD28 expression primarily reflected immune deficiency as assessed by CD4⁺ T cell depletion. Indeed, we noted a strong inverse correlation between the percentage of CD8⁺CD28⁻ and that of CD4⁺ T cells. There was also a positive correlation between the percentage (although of low statistical significance) and absolute counts (P < 0.0001) of CD8⁺CD28⁺ cells and CD4⁺ T cells. These results suggest the existence of a homeostatic regulation inside the T cell compartment [17], operating at least in peripheral blood [37], by which losses of CD4⁺ and CD8⁺CD28⁺ cells are balanced by an increased number of CD8⁺CD28⁻ cells, as recently proposed by Caruso et al. [38]. This interpretation is supported by our observation that the coefficient of variation (CV) of the percentage of total T cells (CV 8.3%) was much smaller than either the CV of the percentage of CD4⁺ plus CD8⁺CD28⁺ cells (CV 33.2%) or that of the percentage of CD8⁺CD28⁻ cells (CV 31.5%). When absolute counts were considered, the high inter-patient variation in total T lymphocyte numbers precluded the demonstration of such T cell homeostasis. Indeed, the corresponding CVs were 28.1% (CD3⁺), 40.3% (CD4⁺ plus CD8⁺CD28⁺), and 45.0% (CD8⁺CD28⁻). This may explain in part the different correlations found when absolute numbers or percentages of the same subpopulations were considered (e.g. CD4⁺versus CD8⁺CD28⁻). An additional possible explanation is that some of the several complex mechanisms that determine T cell homeostasis could be more sensitive to the ratio of T cell subsets in blood than to their absolute number [37].

During anti-retroviral therapy, increases in the relative and absolute numbers of CD4⁺ T cells were clearly correlated with increases in the CD8⁺CD28⁺ T cells, whereas increases in the percentage of CD4⁺ cells were strongly associated with decreases in the percentage of CD8⁺CD28⁻ T lymphocytes. These changes were observed whether or not treatment included protease inhibitors. The mean of all changes in the percentage and absolute number of total T cells (−3.2%, +69/μl) was significantly smaller than the mean of changes in the percentages and absolute numbers of either CD4⁺ plus CD8⁺CD28⁺ cells (+5.2%, +154/μl) or CD8⁺CD28⁻cells (−7.9%, −83/μl) (P < 0.001, Friedman test followed by Dunn's test). These findings again point at a T cell homeostasis phenomenon similar to that described by Caruso et al. [38].

CD8⁺CD28⁺ T cells include naive cells which are decreased in number in HIV infection [39] and memory CTL precursors [2]. An increase in CD8⁺CD28⁺ T cells in the course of treatment may have two beneficial effects. Indeed, this subset contains precursors for efficient CTL effectors [40,41], including HIV-specific cells which are generally considered to play a protective role in HIV infection [4,14]. CD8⁺CD28⁺ T cells also produce partly identified soluble factors that inhibit viral replication in CD4⁺ T lymphocytes [3,4,11]. On the other hand, CD8⁺CD28⁻ T cells consist of terminally differentiated cells [5,6]. Some of them (0.1–2% [42]) have been shown to display HIV-specific CTL effector activity without prestimulation [2,40,43,44]. Nevertheless, the level of the latter cells in blood is not related to the prognosis of HIV infection, in contrast to the level of anti-HIV memory CTL [41,45]. This could be attributed to the fact that, under certain circumstances, in the absence of a CD28 costimulatory signal, CD8⁺CD28⁻ effector cells may die by apoptosis at the same time as their target cell [16,46].

In our cross-sectional study, we uncovered a very strong association between plasma viraemia and the number of cells in the CD8⁺ T subsets defined by CD38 expression. The correlation was much stronger than that observed between plasma viraemia and the number of CD4⁺ cells. These results raise the possibility that the percentage of CD38⁺ cells among CD8⁺ T lymphocytes merely reflects plasma viraemia, hence viral replication in lymphoid tissues, rather than being an independent prognostic marker as generally believed [21–28]. There is no evident explanation for the tight correlation between the expression of CD38 by blood CD8⁺ T cells and plasma viral load. Infection of these cells by HIV, as observed in vitro by De Maria et al. [47], has not been convincingly demonstrated to occur in vivo and therefore cannot account for our observation. A more likely explanation is that the general activation of CD8⁺ T cells known to occur in lymphoid tissue, due at least in part to cytokine dysregulation [48], could be proportional to the level of viral replication in this tissue, which in turn determines the concentration of HIV RNA in plasma [49].

Changes in the CD8⁺CD38⁺ and CD8⁺CD38⁻ subsets upon anti-retroviral treatment were equally correlated with changes in plasma viraemia and in CD4⁺ T cell number. This does not contradict the findings of the cross-sectional study. Indeed, correlation coefficients involving changes in plasma viraemia are only rough estimates, since most post-treatment values are below detectable levels (400 copies/ml). As a matter of fact, we observed that the kinetics of changes in the CD38⁺ and CD38⁻ subsets did not closely match the kinetics of changes in plasma viraemia or in CD4⁺ T cell number.

In conclusion, our results are in agreement with the hypothesis that CD8⁺CD28⁺ and CD8⁺CD28⁻ subsets are differently involved in a T cell homeostasis phenomenon, and that expansion of CD38⁺ T lymphocytes in HIV infection is closely related to viral replication. Abnormalities in the number of cells present in the CD8⁺ T subsets defined by CD28 or CD38 expression regressed remarkably under highly active anti-retroviral treatment.