Combined Env- and Gag-specific T cell responses in relation to programmed death-1 receptor and CD4⁺ T cell loss rates in human immunodeficiency virus-1 infection

Abstract

Additional progression markers for human immunodeficiency virus (HIV) infection are warranted. In this study we related antigen-specific responses in CD4⁺ and CD8⁺ T cells to CD38, reflecting chronic immune activation, and to CD4⁺ T cell loss rates. Clones transiently expressing CD107a (CD8⁺) or CD154 (CD4⁺) in response to Gag, Env and Nef overlapping peptide pools were identified, along with their expression of the inhibitory programmed death-1 receptor (PD-1) in fresh peripheral blood mononuclear cells (PBMC) from 31 patients off antiretroviral treatment (ART). HIV-specific CD8⁺ T cell responses dominated over CD4⁺ T cell responses, and among CD8⁺ responses, Gag and Nef responses were higher than Env-responses (P < 0·01). PD-1 on CD8⁺ HIV-specific subsets was higher than CMV-specific CD8⁺ cells (P < 0·01), whereas PD-1 on HIV-specific CD4⁺ cells was similar to PD-1 on CMV-specific CD4⁺ cells. Gag and Env CD8⁺ responses correlated oppositely to the CD4 loss rate. Env/Gag CD8⁺ response ratios, independently of PD-1 levels, correlated more strongly to CD4 change rates (r = −0·50 to −0·77, P < 0·01) than the total number of Gag-specific CD8⁺ cells (r = 0·44–0·85, P ≤ 0·02). The Env/Gag ratio performed better than CD38 and HIV-RNA in logistic regression analysis predicting CD4 change rate as a measure of progression. In conclusion, HIV-specific CD8⁺CD107a⁺ Env/Gag response ratio was a stronger predictor for progression than CD38 and HIV-RNA. The Env/Gag ratio may reflect the balance between possibly beneficial (Gag) and detrimental (Env) CD8⁺ T cell responses and should be explored further as a progression marker.

Introduction

Anti-retroviral treatment (ART) effectively reverses immune deficiency in human immunodeficiency virus (HIV)-infected individuals who have HIV-related symptoms or opportunistic infections; however, the immune system is better preserved when ART is started early in an asymptomatic phase [1]. For such patients, low current CD4⁺ T cell counts have predominated as an indication for ART, accompanied by secondary criteria such as rapid CD4 decline or high HIV-RNA concentrations [2–5]. The probability of low CD4 cell-associated clinical complications was previously considered acceptable at CD4 counts down to 200 × 10⁶/l, but deferring treatment based on such a threshold has been reported recently to be associated with increased risks for complications [6,7]. The trend has therefore emerged to start ART at higher CD4 counts for all patients. Alternatively, an early start of ART could be recommended primarily to those patients who have a higher risk of complications or more rapid disease progression [8–10]. However, this approach probably requires better clinical predictors than CD4⁺ T cell counts and HIV-RNA concentrations [11,12]. Currently, predictors reflecting HIV-related chronic immune activation have proved promising, particularly the expression of CD38 on CD8⁺ T cells [12–14].

Progression markers should reflect the development of HIV-related pathogenetic events. For example, chronic immune activation is associated with enhanced mucosal translocation of endotoxin into the circulation [15,16], whereas slow disease progression has been related to high frequencies of HIV-specific T cell responses with polyfunctional [17] and proliferative capacity [18]. Unfortunately, assessment of these parameters may require cautious standardization which may complicate clinical evaluation. In this exploratory study of new putative prognostic markers in untreated, asymptomatic patients we used CD4⁺ loss rates and CD38 as measures for actual progression and progression risk. Furthermore, progression was related to T cell response distributions to three major HIV antigenic regions (Gag, Env and Nef) and the expression of inhibitor programmed death receptor-1 (PD-1; CD279) on these specific T cells for the following reasons: first, T cell responses to certain HIV epitope sequence regions, such as Gag and Env, may be more or less important for clinical progression [19–22]. The individual frequencies and their distributions between CD8⁺ and CD4⁺ T cell responses to three different optimized peptide panels [23] representing Gag, Env and Nef were tested on freshly isolated peripheral blood mononuclear cells (PBMC). Antigen specificity was ensured by a robust one-step detection of the activation-specific transient expression of CD107a on CD8⁺[24] and CD154 on CD4⁺[25] T cell subsets, respectively, although mobilization of CD154 (CD40 ligand) on CD4⁺ cells may be hampered in chronic HIV infection [26]. Secondly, PD-1, a reversible inhibitor of T cell-specific activation [27–29], may be elevated particularly on HIV-specific CD8⁺ T cells [28,30–32]. This explorative study showed that both the magnitude and relations between Env and Gag responses and their PD-1 expression were better predictors for CD4⁺ T cell loss rates than the conventional indicators for ART in asymptomatic patients, and probably even better than expression of CD38.

Materials and methods

Patients

Thirty-one asymptomatic, HIV-1 seropositive, adult patients without ART were included from our out-patient clinic (Table 1). With the exception of four patients who had their first CD4⁺ T cell count taken less than 6 months prior to the study, CD4 change (ΔCD4) rates were defined as the time-adjusted change in CD4 counts per year and calculated as follows: the common baseline CD4 count was determined as the median of two consecutive dates with corresponding CD4 counts, both taken approximately 1 year prior to inclusion. The last CD4 count determining ΔCD4 was either at the point of immune response determination (current ΔCD4) or the last available sample post-study (prospective ΔCD4), determined 12·5 (11·7–13·9) and 32·2 (22·5–37·1) months from baseline, respectively. Prospective ΔCD4 rates were available for 14 patients, as the remaining participants were included in a clinical trial testing immunomodulating therapy. CD4⁺ T cell counts were analysed in asymptomatic phases. The patients were anti-retroviral treatment-naive (n = 22) or temporary ART had been terminated at least 18 months prestudy (n = 9). In the latter group, ART had been initiated due to primary HIV infection (n = 8) and pregnancy (n = 1), but stopped 46 months prior to inclusion (range 22–64). All patients gave their informed consent according to the approval by the Regional Committee for Medical Research Ethics.

Table 1.

Study cohort characteristics.

n = 31	Median (IQR)
Males : females (n)	27 : 4
Risk groups (MSM : heterosexual : IDU : unknown) (n)	18 : 9 : 0 : 4
Race (Caucasian : Asian : Black) (n)	27 : 1 : 3
Age (years)	39(31–44)
Time HIV-infected(n = 31)(months)	35·9(14·5–58·3)
Time since HIV seroconversion(n = 9) (months)	16·9(7·8–28·1)
CD4 cell count(×106/l)	400(320–500)
CD8 cell count(×106/l)	1230(800–1660)
HIV-RNA(copies/ml)	36 000(20 000–170 000)
β2-microglobulin(mg/l)	3·1(2·6–4·0)
D-dimer(ng/ml)	298(186–347)
ΔCD4(number/year)(×106/l)	−159(−198to−38)
ΔCD4 % (relative change/year)	−38(−58to−10)
CD38 density on CD8+CD38+PD-1+ (molecules/cell)	6933(4507–9189)
CD4+PD-1+(%)	26·4(20·7–37·8)
CD8+PD-1+(%)	40·2(32·6–51·0)
Env/Gag CD8 response ratio	18·4(3·2–66·9)
Env/Gag PD-1-negative CD8 response ratio	33·7(5·3–175·4)

HIV, human immunodeficiency virus; IDU, intravenous drug use; IQR, interquartile range; MSM, men who have sex with men.

Laboratory parameters and reagents

Routine clinical chemistry profiles were collected, including C-reactive protein, β₂-microglobulin and D-dimer. CD4⁺ and CD8⁺ T lymphocyte counts in peripheral blood and HIV-1 RNA with a detection limit of 50 copies/ml were obtained as described [33]. The antibodies and reagents were obtained from Becton Dickinson (BD, San Diego, CA, USA) [anti-CD3 allophycocyanin, anti-CD4 and anti-CD8 peridinin chlorophyll protein, anti-CD38 Quantibrite phycoerythrin (PE), QuantiBRITE PE Beads, anti-CD107a fluorescein isothiocyanate (FITC), anti-PD-1 (FITC or PE) and isotype control antibodies] and eBioscience (San Diego, CA, USA) [CD154 (PE), co-stimulatory anti-CD28 and monensin].

Flow cytometry and immune activation assay

Two-laser four-colour flow cytometric analyses were performed on a FACSCalibur (fluorescence activated cell sorter) instrument (BD), adjusted and compensated as detailed elsewhere [34]. CD38 density (molecules/cell) in T cell subsets was determined in fresh ethylenediamine tetraacetic acid (EDTA)-containing full blood by means of QuantiBRITE (BD) PE-labelled anti-CD38 in conjunction with PE-labelled standard beads according to the manufacturer's instructions, and calculated as described previously [14]. Concurrently, PBMCs were isolated in the Cell Preparation Tube (CPT™, BD) containing sodium heparin and directly stimulated by antigen (see below) along with co-stimulatory unlabelled anti-CD28 (1 µg/ml), monensin (2 µM) and 10% autologous serum for 6h. CD8⁺ and CD4⁺ T cell specific responses were based on T cell receptor-dependent transient surface expression of CD107a [24] and CD154 [25], respectively, which were detected by soluble anti-CD107a (FITC) and anti-CD154 (PE), added to the cell culture medium together with the antigens. Antigens included HIV-1 group M panels of overlapping 15-mer peptides at 2 mg/l from Gag, Env and Nef, respectively (a gift from the NIH AIDS Research and Reference Reagent Program, MD, USA) and cytomegalovirus (CMV) lysate proteins [33]. After 6 h PBMC were surface-stained with CD3, CD4 or CD8 and PD-1 monoclonal antibodies, before flow cytometry. Data analyses were performed with Winlist analysis software (Verity SH, Topsham, ME, USA). Antigen-specific responses were measured as subset-specific responses above the median background in two control cultures.

Statistical analyses

Statistical analyses were performed with Statistica™ software (StatSoft™ Inc., Tulsa, OK, USA). Data are presented as median values [25–75 interquartile range (IQR)] unless stated otherwise. Non-parametrical two-tailed statistical methods were used throughout; i.e. Spearman's rank correlation analysis, Mann–Whitney U-test for groupwise comparison, and the two-tailed Wilcoxon matched-pairs test for dependent variables. Probability values ≤0·05 were considered significant. Binary logistic regression was used to determine odds ratios.

Results

HIV-1-specific T cell responses dominated by CD8⁺ responses to Gag and Nef

Stimulating PBMC with three panels of overlapping 15-mer peptides gave heterogeneous antigen-specific CD4⁺ and CD8⁺ T cell response patterns (Table 2). This variability between patients was supported by a lack of correlation between the proportions of CD8⁺ and CD4⁺ Gag-, Env- or Nef-specific T cells [r ≤ 0·20, not significant (n.s.)]. A greater than 10-fold dominance was observed in CD8⁺ response frequencies compared to the corresponding specific CD4⁺ cells (P < 0·01, Table 2). In contrast, CMV lysate proteins induced mainly CD4-mediated responses (data not shown), but this difference may be difficult to evaluate, as proteins are more aptly processed and presented by class II major histocompatibility complex (MHC) molecules in vitro (Fig. 1a). CD8⁺ Gag- and Nef-specific responses dominated over Env (P < 0·01), and Gag responses were possibly higher than Nef (Table 2). Among CD4⁺ T cells, this predominance of Gag-specific clones was not observed (Table 2).

Table 2.

HIV-specific T cell responses.

		Responses		Responding cells		PD-1 on responding cells		PD-1-negative responses		PD-1-neg. responding cells
	Antigen	(‰)	(P)†	(106 cells/l)	(P)	(%)‡	(P)	(‰)	(P)	(*10^3 cells/l)	(P)
CD8+CD107a+	Gag	13·3(4·2–28·1)§		16·1(4·4–34·5)		86·3(77·3–90·3)		1·2(0·5–3·3)		178(54–353)
	Env	2·3(0·9–7·3)	(<0·01)	3·0(1·2–7·3)	(<0·01)	67·4(50·0–84·3)	(<0·01)	0·4(0·2–1·8)	(0·09)	60(25–219)	(0·07)
	Nef	8·8(4·9–19·0)	(0·06)	11·6(3·5–28·9)	(0·06)	83·4(69·2–88·5)	(n.s.)	1·8(0·5–3·1)	(n.s.)	198(60–450)	(n.s.)
CD4+CD154+	Gag	1·0(0·0–3·6)		0·3(0·0–1·5)		73·1(45·5–84·2)		0·1(0·0–1·2)		2(0–44)
	Env	0·0(0·0–1·8)	(n.s.)	0·0(0·0–0·8)	(n.s.)	n.a.		n.a.		n.a.
	Nef	0·5(0·0–3·0)	(n.s.)	1·4(0·0–1·1)	(n.s.)	66·7(35·4–89·6)	(n.s.)	0·0(0·0–0·8)	(n.s.)	0(0–29)	(n.s.)

^†Differences from Gag responses (Wilcoxon).

^‡Only calculated for responders >0.

^§Data expressed as median (interquartile range). HIV, human immunodeficiency virus; n.a., not applicable; n.s., not significant (P > 0·10).

Fig. 1.

(a) Box plots showing proportions of programmed death receptor-1 (PD-1) on all CD8⁺ and CD4⁺ T cells and on antigen-specific CD8⁺CD107a⁺ and CD4⁺CD154⁺ subsets as indicated. Median with interquartile ranges indicated. Median Env-specific CD4⁺CD154⁺ T cells responses was zero, i.e. PD-1 fractions listed are from Env-responders only. Differences in PD-1 between antigen-specific subsets (Wilcoxon) indicated, *P < 0·01, **P = 0·01 and ***not significant (n.s.). (b) Fluorescence activated cell sorter analyses scatter plots gated on live gate lymphocytes/CD3⁺/CD8⁺ showing PD-1 fractions and HIV-specific responses of CD8⁺ T cells defined by CD107a expression after stimulation with Env and Gag peptide panels as indicated. Background thresholds for both markers indicated based on relevant controls. Left scatters illustrate Env- and Gag-responses of a slow progressor with a CD4 change rate at −12 × 10⁶ CD4⁺ T cells/l/year and E/G ratio of 12·0. Right scatters depict a rapid progressor with CD4 change rate at −186 × 10⁶ CD4⁺ T cells/l/year and an E/G ratio at 81·3.

When the absolute numbers of antigen-responsive cells were determined by adjusting for the current CD4⁺ and CD8⁺ T cell counts in peripheral blood, the distributions of these effector cells were comparable to the corresponding response frequencies (Table 2). Interestingly, total CD8⁺ T cell counts correlated well with total numbers of Gag- and Nef-specific CD8⁺ T cells (r = 0·58 and r = 0·51, respectively, P < 0·01), but not with Env-specific cells (r = 0·05, n.s.).

PD-1 is up-regulated on HIV-1-specific CD8⁺ and CD4⁺ T cells

PD-1 is up-regulated on HIV-1-specific CD8⁺ T cells, at least on certain clones, which were detected initially in selected patients by means of human leucocyte antigen (HLA) class I HIV epitope-specific tetramers [30,35]. In this study we found that PD-1 was up-regulated uniformly on all Gag- Nef- and Env-specific CD8⁺ T cells (Table 2) (Fig. 1a), irrespective of HLA class I constitution. PD-1 on all HIV-specific CD8⁺ cells (>>67%, Table 2) was higher compared to both CMV-specific cells (48·9%, P < 0·01) as well as the total CD8⁺ T cell population (40·2%, P < 0·01) (Fig. 1a). Moreover, no correlation was found between PD-1 expression on HIV-specific CD8⁺ T cells and the remaining non-activated, non-HIV-specific CD8⁺ cells; this suggested that PD-1 levels on cytotoxic T cells for a given individual were not set at a generalized level, but were rather dependent upon the nature of the antigen and infection activity. Due to technical limitations in the flow cytometry analyses, PD-1 estimates were not available for the naive, memory and effector CD4⁺ and CD8⁺ T cell subsets, thus some of the antigen-specific differences in PD-1 expression might have been attributed partly to different distributions of resting and effector CD8⁺ T cells [35,36].

Day et al. [30] found that PD-1-blocking monoclonal antibodies (mAbs) enhanced CD4⁺ T cell responses to HIV antigens, which suggests indirectly that PD-1 is up-regulated even on HIV-specific CD4⁺ T cells. Here, we confirmed this concept because PD-1 was up-regulated particularly on Gag- and Nef-responsive CD4⁺CD154⁺ T cells compared to the majority of non-activated cells (Fig. 1a). In contrast to PD-1 on CD8⁺ T cell subsets, PD-1 on CMV-specific CD4⁺ cells was both similar to (Fig. 1a) and correlated with PD-1 on both Gag- (r = 0·57, P = 0·02) and Nef-specific (r = 0·72, P < 0·01) CD4⁺ T cells.

HIV-1-specific responses in relation to CD4 loss and prognostic markers

Subsequently, we examined how HIV-specific immune responses to Gag, Nef and Env related to progression and other predictors including CD38, current CD4 count and viral load in asymptomatic untreated patients. In the lack of clinical events, progression was measured as current and prospective CD4⁺ T cell change rates. CD38 density was measured on CD8⁺ T cells and on the CD8⁺PD-1⁺ subset. These measures for CD38 correlated (r = 0·80, P < 0·01), but in accordance with our previous results [14], CD38 on the PD-1⁺ subset was, in general, statistically stronger. CD38 density will henceforth therefore be reported only for the CD8⁺PD-1⁺ T cell subset (Table 1).

Gag-specific CD8⁺ T cell responses relate to the CD4 change rate and markers of chronic immune activation

Only Gag-specific CD8⁺ T cell responses correlated with both the current and the prospective CD4 count change rates, particularly the total concentrations of CD8⁺ Gag-specific T cells in the circulation (Table 3). Moreover, patients who had the highest frequency of Gag-specific CD8⁺ cells (upper tertile) demonstrated substantially slower current CD4 loss rates than those having few (lower tertile) [−62·9 versus−195·1 CD4 cells/µl/year (medians), respectively, P = 0·04] (Fig. 2a). Furthermore, these observations were confirmed in those patients whose prospective CD4 change rate could be calculated (r = 0·85, P < 0·01) (Table 3). In agreement with these results, CD38 correlated only with Gag-specific responses (Table 3), but not with Env- and Nef-responses, current CD4⁺ T cell count, viral load, D-dimer, nor to time infected or age.

Table 3.

Correlations between HIV-specific CD8⁺ T cell subset responses and prognostic parameters.

		CD38 on PD-1+CD8+		Current CD4 change rate		Prospective CD4 change rate
Specificity	CD8+ T cell subset	r	(P)*	r	(P)	r	(P)
Gag	Frequency (%)	−0·34	(0·06)	0·39	(0·04)	0·85	(<0·01)
	Total specific (×106/l)	−0·43	(0·01)	0·44	(0·02)	0·85	(<0·01)
	Frequency of PD-1 (%)	0·03	(0·86)	0·12	(0·53)	−0·07	(0·84)
	Frequency of PD-1-negative (%)	−0·27	(0·14)	0·19	(0·35)	0·79	(<0·01)
	Total PD-1-negative (×106/l)	−0·41	(0·02)	0·25	(0·20)	0·80	(<0·01)
Env	Frequency (%)	0·01	(0·97)	−0·29	(0·13)	−0·26	(0·42)
	Total specific (×106/l)	−0·14	(0·46)	−0·13	(0·49)	0·01	(0·97)
	Frequency of PD-1 (%)	−0·09	(0·65)	0·37	(0·07)	0·29	(0·39)
	Frequency of PD-1-negative (%)	0·06	(0·74)	−0·41	(0·03)	−0·23	(0·47)
	Total PD-1-negative (×106/l)	−0·09	(0·64)	−0·26	(0·19)	−0·16	(0·62)
Nef	Frequency (%)	−0·11	(0·55)	−0·04	(0·85)	0·04	(0·90)
	Total specific (×106/l)	−0·27	(0·15)	−0·14	(0·48)	0·09	(0·78)
	Frequency of PD-1 (%)	0·10	(0·60)	0·18	(0·37)	0·35	(0·27)
	Frequency of PD-1-negative (%)	−0·16	(0·39)	−0·05	(0·8)	−0·36	(0·25)
	Total PD-1-negative (×106/l)	−0·29	(0·11)	0·07	(0·71)	−0·10	(0·75)
Ratios	Env/Gag ratio	0·30	(0·10)	−0·50	(<0·01)	−0·77	(<0·01)
	PD-1-negative Env/Gag ratio	0·31	(0·09)	−0·43	(0·02)	−0·71	(<0·01)

^*P ≤ 0·05 shown in bold type, P between 0·05 and 0·10 in bold italic type. HIV, human immunodeficiency virus; PD-1, programmed death-1.

Fig. 2.

Relations between Env- and Gag-specific CD8⁺ T cell responses and CD4 loss rates. (a) CD8⁺ Gag-specific responses (left panel) and Env/Gag ratios (right panel) (tertiles) in relation to CD4 loss rate. Median and interquartile ranges with differences between tertile indicated. (b) Correlations between Env/Gag CD8⁺ T cell response ratios and current (left panel) and prospective (right panel) CD4 change rates (r: Spearman's rank correlation coefficient).

No correlation was detected between Nef- or Env-specific T cell response frequencies and CD4 change rates or other prognostic markers (Table 3).

PD-1 negative subsets of Env- and Gag- specific CD8⁺ T cells

PD-1-negative HIV-specific T cells may theoretically represent ‘true’ effector T cell capacity against the virus. PD-1-negative CD8⁺ T cell responses were also dominated by Gag and Nef, but the predominance of CD8⁺ Gag compared to Env responses (×5–6) became less pronounced (×3) among CD8⁺ PD-1-negative T cells (P < 0·01) (Table 2). However, when PD-1 expression on specific T cells was related to prospective CD4 loss rates and CD38, Gag-specific CD8⁺ PD-1-negative T cells were again superior to the corresponding Env- and Nef-specificities (Table 3). The impact of PD-1-negative Gag-specific cells was supported by lower CD38 levels in patients with a high number of Gag PD-1-negative CD8⁺ cells [5698 (highest Gag tertile) versus 7634 CD38 molecules/cell (lowest tertile); medians, P = 0·01].

Interestingly, Env-specific cells correlated with current CD4 change rate (r = −0·41), but inversely, so compared with the corresponding Gag subsets (r = 0·79, prospective CD4 change rate) (Table 3). In fact, Env-specific CD8⁺ T cells were the only cells where high PD-1 was favourable in terms of positive correlation with CD4 change (r = 0·37, Table 3). These results correspond with the hypothesis that Env-specific CD8⁺ T cells may be directly or indirectly harmful [20,37].

The ratio between Env- and Gag- specific CD8⁺ T cells

The inverse correlations between CD4⁺ T cell change rates for Gag- and Env-specific CD8⁺ responses (positive and negative correlations, respectively; see above) combined with the lack of correlation between these two antigen responses (r = 0·09, n.s.) prompted us to analyse the Env/Gag CD8⁺ response ratio (E/G). The E/G ratio for PD-1-negative CD8⁺ T cell subsets (E/G neg) were also included in the analyses, as the E/G and E/G neg ratios did not correlate completely (r = 0·79, P < 0·01). It should be noted that the inverted Gag/Env ratios correlated more strongly with CD4 change rates, but were mathematically inapplicable in three of the 31 cases due to undetectable Env-responses (data not shown).

The E/G and E/G neg ratios correlated more favourably than all of the other pseudomarkers tested with the two CD4 change rate parameters (Table 3, Fig. 2b). This was supported by significantly higher current CD4 change rates in patients with low E/G ratio (approx. −50 CD4 cells/µl/year, lower tertile) compared with those having a high ratio (approximately −200 CD4 cells/µl/year, highest tertile, P < 0·01) (Fig. 2a). The same was true for the E/G neg ratios (P < 0·01, data not shown).

E/G ratio best predictor of CD4 loss in logistic regression analysis

All predictive markers were compared in a binary logistic regression analysis where the median current absolute and relative CD4 change rates represented the binary breakpoints (−158 CD4⁺ T cells/µl/year and −38·2%/year, respectively). We also included CD4 loss rate of 100 cells/µl/year (a separate indication for ART [2,4,5]) as a breakpoint. All independent predictor candidates were transformed into variable-dependent tertile numbers, which were arranged in such a manner that a high tertile number was considered unfavourable in terms of CD4 loss. In univariate analysis, the E/G and E/G neg ratios were not only the strongest predictors of current CD4 change rate, but in this limited cohort also the only significant predictors. For example, the odds ratio for rapid CD4 loss was 8·0 between patients within the lowest and the highest tertile of E/G ratios (i.e. 4·0 × 2 tertiles, Table 4). CD38 expression and Gag-specific CD8⁺ responses per se were also predictive for high relative and guideline-restricted CD4 loss rates, in contrast to HIV-RNA and β2-microglobulin (Table 4). No significant results in multivariate binary regression model were found.

Table 4.

Univariate logistic regression estimates to predict CD4 loss rates.

	Current CD4 loss rate		Current CD4 loss rate (relative)		CD4 loss rate (guidelines)
	<>median (−158 CD4 cells/µl/year)		<>median −38.2%		<>−100 cells/µl/year
Parameters (tertiles)	OR (CI)	P*	OR (CI)	P	OR (CI)	P
Env/Gag ratio	4·0(1·3–12·6)	0·01	4·0(1·3–12·6)	0·01	4·2(1·3–13·7)	0·01
PD-1 neg. Env/Gag ratio	3·8(1·2–12)	0·02	3·8(1·2–12)	0·02	4·1(1·2–13·6)	0·02
CD8+CD38+PD-1+(CD38 density/cell)	1·7(0·64–4·6)	0·26	2·8(1·0–8·2)	0·05	2·8(1·0–8·3)	0·05
Gag response frequencies(relative to all CD8+ T cells)	2·2(0·79–6·0)	0·12	1·7(0·6–4·6)	0·26	3·8(1·2–12·3)	0·02
Gag response(absolute number of CD8+ subsets)	1·5(0·59–3·9)	0·37	1·9(0·7–4·9)	0·19	2·3(0·8–6·3)	0·09
HIV-RNA(copies/ml)	0·6(0·2–1·7)	0·34	0·6(0·2–1·7)	0·34	0·9(0·3–2·5)	0·89
β2-microglobulin	1·4(0·53–3·6)	0·49	2·2(0·8–6·0)	0·12	2·2(0·8–6·0)	0·12

^*P-values by Wald's χ². CI, confidence interval; HIV, human immunodeficiency virus; OR, odds ratio; PD-1, programmed death-1.

Discussion

Clinical evaluation of asymptomatic and untreated HIV-infected patients should be based upon prognostic markers with sufficient statistical power for individual counselling. HIV-RNA levels, for example, correlates clearly with clinical progression in large cohorts but predicts progression poorly at the individual level [11–13]. Thus, optimal markers of progression should provide significant information even in small cohorts. This explorative study investigated new parameters for HIV-specific immunity in the search for optimal prognostic markers.

The main goals of this study were to investigate prognostic significance of HIV-specific T cell responses to Gag, Env and Nef and of PD-1 on such HIV-specific cells. Specific clones were detected through transient expression of CD107a and CD154. These data were compared to quantitative measurements of CD38 on CD8⁺ and CD8⁺CD38⁺PD-1⁺ T cells and correlated subsequently to progression, which in asymptomatic patients may be best described by CD4⁺ T cell loss rates. Furthermore, fresh blood samples as opposed to thawed PBMC were analysed due to the decay of CD38 on thawed PBMC [14], possible preferential loss of CD8⁺ T cells [38] and limited robustness of the CD107a assay.

Two mainly affirmative observations were made: a predominance of Gag-restricted CD8⁺ T cell responses and their relation to prognosis [20] and a high expression of PD-1 molecules on such HIV-specific CD8⁺ T cells [30]. In addition, this study provided new data showing up-regulated PD-1 on HIV-specific CD4⁺ T cells, but differently than on the CD8⁺ subset as well as a lower expression of PD-1 on Env-specific CD8⁺ T cells compared with Gag-specific cells (Fig. 1a).

Subsequently, the data on relative and absolute abundance of HIV-specific responses, including the estimates of PD-1, were related to CD4 loss rates. The total number of Gag-specific CD8⁺ cells were correlated even stronger with CD4 loss rates and immune activation than the conventional frequency estimates (Table 3) supporting the relevance of taking the CD8⁺ T cell count into consideration. Moreover, the total CD8⁺ T cell counts correlated well with the total number of favourable Gag-specific cells; a rapid decline in total CD8⁺ T cell counts may involve loss of Gag-specific cells and may thereby explain why CD8 count reduction is a strong independent predictor for acquired immune deficiency syndrome (AIDS) [39]. In terms of PD-1-negative HIV-specific CD8⁺ T cells, two phenomena were particularly interesting (Table 3): the total number of Gag-specific PD-1 negative cells was correlated inversely and favourably with CD38 and immune activation, whereas Env-specific PD-1 negative cells did not correlate to CD38 and correlated unfavourably to CD4 change rates (r = −0·41), in accordance with the fact that more PD-1 on Env-specific cells possibly correlated positively to CD4 change rates (r= 0·37).

The lack of correlation between Env- and Gag-specific CD8⁺ T cell responses in combination with their opposite correlation to CD4 loss rates prompted us to investigate the Env and Gag response ratios. Indeed, the Env/Gag ratios correlated more favourably to CD4 loss rates than the individual antigen-specific responses themselves. Moreover, the poor correlation between the E/G ratios and CD38 suggests that these parameters provide supplementary biological information. In logistic regression analysis the odds ratio for progression was clearly most favourable for the E/G ratio, particularly compared to CD38. As the E/G ratios of the PD-1 negative subsets were comparable to the E/G ratios of the total CD8⁺ subsets, PD-1 assessments may even be unnecessary.

In conclusion, Gag- and Env-specific CD8⁺ T cell responses offer significant prognostic value. Furthermore, opposite relations to CD4 loss rates and CD38 were found between possibly beneficial Gag and detrimental Env CD8⁺ T cell responses in asymptomatic patients who were not on treatment for chronic HIV-infection. Env/Gag response ratios, independently of PD-1 levels, predicted progression better than the currently best prognostic marker CD38. These promising observations should be confirmed and evaluated further in a larger prospective cohort.

Disclosure

None of the authors have conflicts of interest, or any relevant financial interest, in any company or institution that might benefit from this publication.