A T2 cytokine environment may not limit T1 responses in human immunodeficiency virus patients with a favourable response to antiretroviral therapy

Abstract

Low-level production of interferon-γ (IFN-γ) marks human immunodeficiency virus (HIV)-induced immunodeficiency and has been ascribed to a bias towards T2 cytokines. This was investigated in two cross-sectional studies of HIV patients who were immunodeficient when they began antiretroviral therapy (ART) and had stable increases in CD4 T-cell counts. Blood leucocytes were assessed unstimulated or after stimulation with cytomegalovirus (CMV), anti-CD3 or mitogen. IFN-γ and interleukin (IL)-5 responses were initially assessed by enzyme-linked immunosorbent spot-forming cell assay (ELISPOT) and enzyme-linked immunosorbent assay (ELISA). We then adopted a sensitive reverse transcription–polymerase chain reaction (RT–PCR) system to assess IFN-γ, IL-5, IL-4 and IL-4δ2 (an inhibitory splice variant of IL-4) mRNA. The results were correlated with putative serological markers of a T1 [lymphocyte activation gene-3 (LAG-3), CD26] or a T2 [CD30, immunoglobulin E (IgE)] cytokine environment. IL-5 production and IgE levels were elevated in patients. IgE levels did not correlate with IFN-γ, but showed an inverse correlation with IL-5 released in culture (P = 0·05). The levels of IL-4, IFN-γ, IL-5 and IL-4δ2 mRNA were correlated after anti-CD3 stimulation, where IL-5 was the best predictor of IFN-γ mRNA (P = 0·006). Weak positive correlations were evident between CD30 and cytokine mRNA levels, whilst IgE correlated inversely with IL-4, IL-4δ2, IL-5 and IFN-γ mRNA levels. These analyses provide no evidence for an inverse relationship between T1 and T2 cytokine responses in HIV patients, but suggest that the elevation of IgE marks low cytokine responses.

Introduction

Low-level production of interferon-γ (IFN-γ) by CD4 T cells in response to antigens from human immunodeficiency virus (HIV) or opportunistic infections is a characteristic of HIV-induced immunodeficiency,¹ often ascribed to a bias towards T2 cytokine responses.² Responses improve when patients are given antiretroviral therapy (ART), but previously immunodeficient patients may retain defects of immune function, despite control of HIV replication and increased CD4 T-cell counts. These include impaired lymphoproliferative responses, antibody responses to vaccination and cutaneous delayed-type hypersensitivity responses.^3,4 In patients with nadir CD4 T-cell counts of <50/µl, IFN-γ responses to cytomegalovirus (CMV) or Candida antigens were found to be substantially lower than uninfected controls after several years of ART,^5,6 and IFN-γ mRNA levels were reduced in unstimulated T cells. This was not the result of persistent HIV replication.^6,7

Evidence for a T2 bias in untreated HIV disease includes the low-level production of interleukin (IL)-12⁸ and elevated circulating soluble CD30. T lymphocytes that express and release CD30 produce predominantly T2 cytokines.² We reported that HIV patients responding to ART have low levels of IFN-γ mRNA in unstimulated peripheral blood mononuclear cells (PBMC) and reduced IL-23 mRNA levels in adherent cells.⁹ IL-12 is critical in the induction of T cells and natural killer (NK) cells to produce IFN-γ,¹⁰ while IL-23 particularly affects memory T cells.¹¹ Low levels of IL-23 in HIV patients on ART suggests a continued bias towards T2 cytokine responses. In unstimulated PBMC from HIV patients, IL-4 and IL-10 mRNA was more commonly detected before ART, and the IFN-γ mRNA levels increased in patients on ART. This suggested correction of a T2 bias. However, IFN-γ, IL-4 and/or IL-10 levels rose in parallel in some patients.¹²

Soluble lymphocyte activation gene-3 (LAG-3) and CD26 dipeptidyl peptidase IV (DPPIV) enzyme activity have been proposed as serological markers of a T1 environment. A putative T2 cytokine environment in patients with symptomatic tuberculosis is marked by a decreased ratio of circulating LAG-3 to CD30 and a relative decrease in mRNA for the inhibitory splice variant of IL-4 (IL-4δ2) relative to IL-4 mRNA.^13,14 CD26(DPPIV) cleaves several chemokines, reducing the chemotactic activity stimulated through CCR3 (a chemokine receptor expressed by T2 cells).^15,16 However, circulating levels of CD26(DPPIV) are not inversely related to the levels of CD30 in HIV patients.¹⁷ Serum immunoglobulin E (IgE) may also mark a T2 environment,¹⁸ but may also reflect B-cell activation by IL-6¹⁹ or other cytokines. Hypergammaglobulinaemia has been reported in untreated HIV patients, with a rise in immunoglobulin levels during HIV disease progression.²⁰ ART reduces hypergammaglobulinaemia in HIV-1 patients, but the levels may remain higher than in uninfected controls.^21–23

The view that HIV promotes a T2 cytokine environment is at odds with evidence that IL-5 production is low in severely immunodeficient untreated HIV-1 patients with <50 CD4 T cells/µl, but IL-5 responses are higher in HIV patients responding to ART than in uninfected control donors.⁵ IL-5 responses to mitogen were elevated in HIV-1 patients with >500 CD4 T cells/µl and undetectable plasma HIV RNA on ART. The increase was greatest in patients with a nadir of <300 CD4 T cells/µl.²⁴ The numbers of CD4 T cells producing IL-5 after stimulation with staphylococcal enterotoxin B and anti-CD28 were also elevated in untreated patients with >200 CD4 T cells/µl.²⁵ The consequences of IL-5 excess are unclear. IL-5 influences eosinophil growth and maturation. Transgenic mice over-expressing IL-5 display hypergammaglobulinaemia,²⁶ but the effect of IL-5 on human B cells is controversial.²⁷

In this study, we first investigated the production of IL-5 and IFN-γ by enzyme-linked immunosorbent assay (ELISA) and enzyme-linked immunosorbent spot-forming cell assay (ELISPOT). Then, T1 and T2 cytokines were assessed via their levels of mRNA. This assay is sufficiently sensitive to quantify mRNA in unstimulated cells (reflecting levels in vivo) and has a large dynamic range, so values obtained after mitogenic stimulation do not reach the upper limit of the assay. The strategy allowed us to consider whether different conclusions would be reached if one evaluated IL-2 responses (potentially central memory cells, T_cm²⁸). We correlated IFN-γ responses with the production of IL-5, IL-4 and IL-4δ2, and with putative serological markers of the T1/T2 cytokine balance. This included IgE, CD30, CD26(DPPIV) and LAG-3. Our study specifically addresses patients who were immunodeficient prior to ART and achieved a favorable immunological outcome, as assessed by CD4 T-cell counts.

Materials and methods

Sample collection: patients and controls

Samples were obtained from HIV-1-infected patients attending routine outpatient clinics at the Royal Perth Hospital (Western Australia). All had a nadir CD4 T-cell count of <60 cells/µl prior to ART and achieved >200 CD4 T cells/µl, or at least four times their nadir count, during therapy. Controls comprised healthy CMV-seropositive volunteers (mostly laboratory staff). Prior approval was obtained from the hospital ethics committee, and informed consent was obtained from all participants. PBMC were isolated using Ficoll–Paque density gradients and cryopreserved in liquid nitrogen. Plasma was separated from whole blood collected in lithium heparin, and serum was collected from blood allowed to clot at room temperature. Both were stored at −80°.

Lymphocyte subsets, plasma viral loads, eosinophils and serum IgE

T-lymphocyte subsets were quantified in EDTA-treated whole blood with CYTO-STAT^® triCHROME^™ (CD8-FITC/CD4-RD1/CD3-PC5) (Coulter, Gladesville, NSW, Australia). Plasma HIV RNA levels were determined by quantitative reverse transcription–polymerase chain reaction (RT–PCR) (Amplicor, Version 1·5, 50–75 000 copies/ml; Roche Diagnostic Systems, Castlehill, NSW, Australia). Eosinophil counts were determined on an Abbott Cell Dyn 4000 Analyser (Abbott, North Ryde, NSW, Australia). Serum IgE levels were assayed on an Immulite 2000 Analyser (Immulite, Los Angeles, CA, USA). Reference ranges are presented in Table 1.

Table 1.

Human immunodeficiency virus (HIV) patients display elevated levels of interleukin-5 (IL-5) and immunoglobulin E (IgE)

	Controls	n	HIV patients	n
Age in years	31 (21–59)	26	39 (30–65)2	30
Nadir CD4 T cells/µl	–		18 (0–50)	30
Current CD4 T cells/µl	–		341 (84–1092)	30
Time on ART in months	–		44 (14–56)	30
HIV RNA copies/ml	–		< 50 (< 50–253 000)	30
IFN-γ (IU/ml)1	2·8 (0–11)	7	2·7 (0–17)	16
IL-5 (pg/ml)1	27 (0–377)	13	156 (0–576)3	16
IFN-γ (spots/2 × 105 cells)1	230 (39–468)	22	280 (66–624)	27
IL-5 (spots/2 × 105 cells)1	43 (1–166)	22	93 (5–262)	27
IgE (kU/l)	(0–210)4		156 (9–5541)	30
Eosinophils (× 109 cells/l)	(0·04–0·40)4		0·14 (0·02–0·39)	28

Data are presented as median (range).

¹Cells were stimulated with phytohaemagglutinin (PHA) overnight.

²P = 0·001 relative to controls (Wilcoxon Rank Sum Test).

³P = 0·02 relative to controls.

⁴95% reference range defined in HIV-seronegative donors.

ART, antiretroviral therapy; IFN-γ, interferon-γ.

Antigens and mitogens

CMV strain AD 169 was propagated in human skin fibroblasts.¹⁷ Anti-CD3 (Mabtech, Stockholm, Sweden) was used at 10 ng/ml. Phytohaemagglutinin-P (PHA-P) (Difco, Detroit, MI, USA) was used at 250 µg/ml.

ELISPOT and ELISA assays for IFN-γ and IL-5

ELISPOT microtitre plates (Millipore, Bedford, MA) were coated with anti-IFN-γ (15 µg/ml; Mabtech, Mosman, NSW, Australia) or anti-IL-5 (Pharmingen, San Diego, CA, USA) in 0·1 m bicarbonate buffer (pH 9·5) overnight at 4° and washed with phosphate-buffered saline (PBS). PBMC were added in RPMI 1640 containing 10% human AB serum (First Link, Brierly Hill, UK; Table 1) or 10% FCS (Tables 2 and 4) at 2 × 10⁵ PBMC/well for antigen stimulation and at 5 × 10⁴ PBMC/well for stimulation with anti-CD3. Plates were incubated for 18–20 hr (or 48 hr for IL-5), washed and reacted with biotinylated anti-IFN-γ or anti-IL-5 (2 hr), streptavidin horseradish peroxidase-conjugate (Genzyme, Cambridge, MA; 1 hr) and AEC (Table 1) or TMB (Tables 2 and 4) substrate. PBMC from several patients and controls were depleted of CD4 and CD8 T cells using magnetic beads (Dynal Biotech, Carlton South, VIC, Australia) to determine the source of IL-5 production. Supernatants were assayed for IL-5 (Pharmingen) and IFN-γ (CSL Biosciences, Parkville, VIC, Australia) production by ELISA, following the manufacturer's instructions.

Table 2.

Cytokine responses and levels of serological markers in human immunodeficiency virus (HIV) patients (n = 19)

	IFN-γ ELISPOT1	IFN-γ mRNA2	IL-4 mRNA	IL-4δ2 mRNA	IL-2 mRNA	IL-5 mRNA
Unstimulated	7	12·5	0·10	0·001	0·39	0·001
	(1–47)3	(2·3–20)	(0·002–3·8)	(0·001–0·003)	(0·13–0·79)	0·001
CMV stimulated	10	18·5	0·14	0·001	3·2	0·001
	(0–206)	(10–33)	(0·06–0·47)	(0·001–0·008)	(0·13–18)	(0·001–0·09)
Anti-CD3 stimulated	481	48·8	0·47	0·001	8·4	0·01
	(120–989)	(19–110)	(0·08–2·6)	(0·001–0·11)	(0·13–32)	(0·001–0·07)
PHA stimulated	>400	63·1	1·87	0·06	61	0·04
		(15–171)	(0·16–8·7)	(0·008–0·18)	(0·13–274)	(0·002–0·14)

LAG-3 pg/ml	CD30 U/ml	CD26 U	IgE kU/l
<22	19·7	164	78
(<22–360)	(6·0–64·9)	(76–276)	(0·5–1028)

¹Spots per 2 × 10⁵ peripheral blood mononuclear cells (PBMC).

²Ratio to β-actin mRNA.

³Median (range).

CMV, cytomegalovirus; ELISPOT, enzyme-linked immunosorbent spot-forming cell assay; IFN-γ, interferon-γ; IL, interleukin; LAG-3, lymphocyte activation gene-3; PHA, phytohaemagglutinin.

Table 4.

Levels of T1 and T2 cytokine mRNA correlated directly with CD30 and inversely with immunoglobulin E (IgE) levels

Cytokine response	Serological marker	Correlation coefficient	P-value
Unstimulated IFN-γ mRNA	LAG-3	0·44	0·06
IFN-γ ELISPOT to anti-CD3	CD30	0·41	0·10
IL-5 mRNA to anti-CD3	CD30	0·40	0·10
IL-5 mRNA to PHA	CD30	0·47	0·05
IL-4δ2 mRNA to PHA	CD30	0·51	0·03
IFN-γ ELISPOT to anti-CD3	IgE	−0·46	0·05
IFN-γ mRNA to anti-CD3	IgE	−0·51	0·03
IL-5 mRNA to anti-CD3	IgE	−0·52	0·02
IL-4 mRNA to anti-CD3	IgE	−0·45	0·05
IL-4δ2 mRNA to anti-CD3	IgE	−0·62	0·004
IL-5 mRNA to PHA	IgE	−0·38	0·10
IL-4δ2 mRNA to PHA	IgE	−0·41	0·08

Cytokine mRNA levels are expressed as a ratio to β-actin mRNA. Values for unstimulated cells were subtracted from the values for stimulated cells. Only comparisons yielding P ≤ 0·10 are shown.

ELISPOT, enzyme-linked immunosorbent spot-forming cell assay; IFN-γ, interferon-γ; IL, interleukin; IL-4δ2, inhibitory splice variant of IL-4; LAG-3, lymphocyte activation gene-3; PHA, phytohaemagglutinin.

Serological assays

Soluble CD30 was assayed by ELISA. Half-volume (50 µl) 96-well plates were coated with anti-CD30 (BenderMed Systems, Vienna, Austria) in PBS, blocked with PBS containing 1% bovine serum albumin (BSA) (1 hr) and washed with PBS containing 0·05% Tween 20. Samples, standards and anti-CD30 peroxidase conjugate were prediluted in PBS containing 1% BSA and 0·05% Tween 20, and added to the plate together (3 hr, on a plate shaker). The plate was washed, bound CD30 was detected with TMB substrate, stopped with 0·5 m H₂SO₄ and the absorbance was read at 450 nm.

CD26(DPPIV) enzyme activity was quantified using an antigen-capture enzyme assay.¹⁷ Plates were coated with anti-CD26 (Pharmingen) in bicarbonate buffer and blocked with PBS containing 1% BSA (1 hr). Prediluted samples and a standard control pool were added (4 hr). After a wash in PBS containing 0·05% Tween 20, bound enzyme activity was detected with a chromogenic substrate (Gly Pro pNA; 5 hr, 37°; Sigma, St Louis, MI). The absorbance was read at 405 nm.

LAG-3 was detected in plates coated with 1 ng/ml anti-LAG-3 (11E3) in bicarbonate buffer, blocked with PBS containing 10% BSA (1–2 hr) and washed with PBS containing 0·05% Tween 20. Samples and standards (LAG-3.Ig) prediluted in PBS containing 1% BSA and 0·05% Tween 20 were added (2 hr), followed by biotinylated anti-LAG-3 (BIOT-174; 1 hr) and horseradish peroxidase (1 hr). Bound LAG-3 was detected with TMB substrate. The reagents were donated by Dr Frederic Triebel (Universite Paris-Sud, Chatenay Malabry, France) and Serono Biotechnology (Geneva, Switzerland).

Real-time PCR to quantify IFN-γ, IL-5, IL-4 and IL-4δ2 mRNA

RNA was extracted using RNeasy Mini Kits and cDNA was generated using SensiScript RT Kits (Qiagen, Doncaster, VIC, Australia). All real-time PCR was performed on a Rotorgene^™ (Corbett, Sydney, NSW, Australia). Real-time PCR to quantify mRNA for IFN-γ and β-actin utilized a 20-µl reaction mix containing 5 µl of cDNA, 0·8 mm dNTP, 0·5 µm each primer, 0·5 × SYBR Green fluorochrome and 1·5 U Platinum Taq DNA polymerase. The forward and reverse primer sequences for β-actin were 5′-GAT GAC CCA GAT CAT GTT TGA-3′ and 5′-GAC TCC ATG CCC AGG AAG GAA-3′ and for IFN-γ were 5′-CTC GGA AAC GAT GAA ATA TAC A-3′ and 5′-CAT ATG GGT CCT GGC AGT AAC-3′. PCR protocols comprised denaturation at 95° for 300 seconds, followed by 35 cycles of denaturation (96° for 15 seconds), annealing (64° for β-actin and 60° for IFN-γ for 15 seconds) and extension (72° for 25 seconds). Levels of IL-5 mRNA were quantified using QuantiTect Hs_IL5 and QuantiTect Probe PCR kits (Qiagen). PCR protocols comprised denaturation at 95° for 900 seconds, followed by 45 cycles of denaturation (94° for 15 seconds), annealing (56° for 30 seconds) and extension (76° for 30 seconds). PCR conditions, primers and probes used to quantify IL-4 and IL-4δ2 have been published previously.²⁹ The ratio of IFN-γ, IL-5, IL-4 and IL-4δ2 amplicons to β-actin in each sample were calculated and expressed as arbitrary units. The lower limit of detection for the genes analysed was a ratio of 0·00001.

Statistical analysis

Continuous variables describing patients and controls were compared with Wilcoxon's Rank Sum Tests. Categorical data were compared with two-sided Fisher's Exact tests. IgE and eosinophil levels were compared with the normal range using Exact Binomial Tests. Correlations between continuous variables were evaluated using Spearman's Tests. Multivariable regression analyses were used to identify dominant predictors of IFN-γ mRNA levels in unstimulated cultures and (after logarithmic transformation) with each stimulant. In all cases, P < 0·05 was accepted as statistical significance and 0·05 < P < 0·10 was noted as indicating a trend.

Results

Elevation of IL-5 production in HIV patients on ART is independent of viral load, and of CD4 T-cell and eosinophil counts

Production of IL-5 and IFN-γ by PBMC stimulated by PHA was assessed in HIV patients with a nadir of < 50 CD4 T cells/µl prior to ART (n = 30) and in uninfected control subjects (n = 26). All patients achieved > 200 CD4 T cells/µl, or at least four times their nadir count, during therapy (Table 1). Seventeen patients had < 50 copies/ml of HIV RNA. Patients with a history of drug exposure prior to ART (n = 17), and those without (n = 13), had similar CD4 T-cell counts (median: 446 vs. 324 cells/µl) and plasma HIV RNA (640 vs. < 50 copies/ml; P > 0·05). At the time of testing, 13/23 patients receiving a protease inhibitor (PI), two of four patients on a non-nucleotide reverse transcriptase inhibitor (NNRTI) regimen and two of three patients on a PI + NNRTI combination had < 50 copies/ml HIV RNA. Hence, the results for patients with and without prior therapy and receiving PI or NNRTI regimens were pooled.

IFN-γ levels in cell culture supernatants, and IFN-γ ELISPOT counts, were similar in patients and controls (Table 1). In contrast, IL-5 levels in culture supernatants were higher in HIV patients than in controls (P = 0·02) and IL-5 ELISPOT counts were marginally higher. IL-5 levels did not correlate with CD4 or CD8 T-cell counts (data not shown). HIV patients were significantly older than controls (P = 0·001). However, there was no correlation between age and IFN-γ or IL-5 ELISPOT counts or supernatant concentrations in patients or controls (0·03 < r < 0·32, 0·2 < P < 0·9).

To identify the cells producing IL-5, CD4 and CD8 T cells were depleted from PBMC from five patients and two controls. In controls, depletion of CD4 T cells decreased IL-5 ELISPOT counts (from 29 to 19, and from 94 to 25 spots per 2 × 10⁵ cells), whilst CD8 T-cell depletion increased the IL-5 ELISPOT counts (from 29 to 81, and from 94 to 158 spots per 2 × 10⁵ cells), suggesting that CD4 T cells produce most IL-5. In contrast, depletion of CD4 T cells decreased IL-5 ELISPOT counts in only one patient (from 266 to 69 spots per 2 × 10⁵ cells) and depletion of CD8 T cells increased IL-5 ELISPOT counts in two patients (from 266 to 371, and from 314 to 395 spots per 2 × 10⁵ cells). Moreover, IL-5 ELISPOT counts of two patients were decreased by depletion of CD8 T cells (from 134 to 65, and from 363 to 194 spots per 2 × 10⁵ cells). These findings suggest that IL-5 is produced by mitogen-stimulated CD4 and CD8 T cells in HIV patients.

Plasma HIV RNA levels did not correlate with PHA-stimulated IL-5 ELISPOT counts or with the levels of IL-5 in culture supernatants. However, patients with undetectable plasma HIV RNA had higher IL-5 responses than uninfected controls (median: 156 vs. 27 pg/ml, P = 0·01; 97 vs. 43 spots per 2 × 10⁵ cells, P = 0·02).

Most patients displayed IgE levels above the 95th percentile of the normal range defined in HIV-negative donors (P < 0·001; Exact Binomial Test). IgE levels above and below the median did not split according to detectable plasma HIV RNA (P = 0·7, Fisher's Exact test) and IgE levels did not correlate with IFN-γ production (r = −0·35, P = 0·2) or CD4 T-cell counts (r = −0·14, P = 0·5). However, IgE levels correlated inversely with IL-5 levels in culture supernatants (r = −0·49, P = 0·05). Only one patient had an eosinophil count above the normal range (P > 0·9). Counts did not correlate with IL-5 levels in supernatants, IL-5 ELISPOT counts or IgE levels. Hence, increased IL-5 production was not associated with eosinophilia or high IgE levels.

Evaluation of RT–PCR assay of cytokine mRNA for assessment of responses to an opportunistic pathogen and the T1/T2 balance

Samples of serum, plasma and PBMC were then obtained from an equivalent group of 19 adult HIV patients who had also begun ART with low nadir CD4 T-cell counts [median (range) = 18 (0–54) cells/µl]. Most had achieved a stable virological response, but three patients had detectable viral loads (76, 753 and 9450 HIV RNA copies/ml). Fifteen patients were also in the first study, but 35 (8–60) months had elapsed between sample collections. The group as a whole (n = 19) had been treated for 82 (40–107) months and had 598 (209–1225) CD4 T cells/µl and 1275 (352–2976) CD8 T cells/µl. Consistent with our previous studies,^5,6 IFN-γ responses to CMV were depressed in these patients relative to HIV-seronegative controls [10 (0–206) vs. 41 (1–261) spots per 2 × 10⁵ cells; P = 0·03].

In vitro depletion of CD4 or CD8 T cells using magnetic beads showed that the response to CMV was exclusively CD4 T-cell mediated, so CMV was selected as the antigen for mRNA studies. Cells were also stimulated with PHA and anti-CD3 to achieve measurable IL-2, IL-4δ2 and IL-5 mRNA (Table 2). The IFN-γ ELISPOT response to CMV, assessed by ELISPOT, correlated with the IFN-γ ELISPOT responses to anti-CD3 (r = 0·71, P = 0·0006) and with the IFN-γ mRNA response to CMV (r = 0·63, P = 0·004). ELISPOT and mRNA responses to anti-CD3 were not closely correlated (r = 0·38, P = 0·1), probably reflecting the smaller dynamic range of the ELISPOT assay. Correlations between IFN-γ and IL-2 mRNA levels induced by CMV or anti-CD3 were also poor (r = 0·32–0·38, P = 0·1–0·2).

Cytokine responses correlate positively with CD30 and inversely with IgE levels

The levels of mRNA for T1 cytokines (IFN-γ and IL-4δ2) and T2 cytokines (IL-4 and IL-5) were then compared in the 19 patients described above, and multivariable regression analyses were used to identify dominant predictors of IFN-γ mRNA levels in unstimulated and stimulated cells. The levels of IL-4 and IFN-γ mRNA did not correlate in unstimulated or CMV-stimulated cells but did so after stimulation with anti-CD3. Moreover, both correlated with IL-4δ2 and IL-5 mRNA levels after stimulation with anti-CD3. Positive correlations were retained when the levels of bioactive IL-4 were estimated by subtracting IL-4δ2 from IL-4 mRNA levels (Table 3, Fig. 1). The results were similar when cytokine mRNA levels in stimulated cells were analysed without adjustment for levels measured without stimulation (data not shown). IL-5 mRNA levels were the best predictor of IFN-γ mRNA levels after stimulation with anti-CD3 (P = 0·006). These analyses provided no examples of an inverse relationship between T1 and T2 cytokines.

Table 3.

Positive correlations between levels of T1 and T2 cytokine mRNA induced by polyclonal stimulation

		Correlation coefficient	P-value
Unstimulated cells1
IFN-γ	IL-4	0·06	0·80
CMV-stimulated cells1
IFN-γ	IL-4	0·21	0·40
Anti-CD3-stimulated cells
IFN-γ	IL-4	0·57	0·01
IFN-γ	IL-4δ2	0·73	0·0004
IL-4	IL-4δ2	0·79	< 0·0001
IFN-γ	IL-4 minus IL-4δ2	0·55	0·01
IFN-γ	IL-5	0·66	0·002
IL-4	IL-5	0·81	< 0·0001
IL-4 minus IL-4δ2	IL-5	0·81	< 0·0001
PHA-stimulated cells
IFN-γ	IL-4	0·32	0·20
IFN-γ	IL-4δ2	0·27	0·30
IL-4	IL-4δ2	0·70	0·0008
IFN-γ	IL-4 minus IL-4δ2	0·30	0·20
IFN-γ	IL-5	0·48	0·04
IL-4	IL-5	0·86	< 0·0001
IL-4 minus IL-4δ2	IL-5	0·84	< 0·0001

Levels of cytokine mRNA in cultures from human immunodeficiency virus (HIV) patients (n = 19) are expressed as a ratio to β-actin mRNA. Values for unstimulated cells were subtracted from the values for stimulated cells.

¹Interleukin (IL)-4δ2 and IL-5 mRNA were not detectable in these cultures.

CMV, cytomegalovirus; IFN-γ, interferon-γ; PHA, phytohaemagglutinin.

Figure 1.

The levels of interferon-γ (IFN-γ) mRNA in cultures stimulated with anti-CD3 correlated directly with the levels of interleukin (IL)-4, the inhibitory splice variant of IL-4 (IL-4δ2) and IL-5 mRNA and inversely with the levels of serum immunoglobulin E (IgE). mRNA levels are presented as a ratio to β-actin mRNA. Spearman's coefficients are shown. LAG-3, lymphocyte activation gene-3.

As putative serological markers of a T1 or a T2 cytokine environment, circulating levels of CD30, CD26, LAG-3 and IgE were correlated with levels of IFN-γ, IL-5, IL-4 and IL-4δ2 mRNA and with IFN-γ ELISPOT counts (Table 4, Fig. 1). Plasma LAG-3 correlated weakly with IFN-γ mRNA in unstimulated cells (P = 0·06), although many samples had no detectable LAG-3 (Fig. 1). The effect requires confirmation with a more sensitive assay. There were no significant predictors of baseline IFN-γ levels in the multivariable analysis.

After stimulation, weak positive correlations were evident between levels of CD30 and IL-4, IL-4δ2 and IL-5 mRNA and IFN-γ ELISPOTs. In contrast, IgE levels correlated inversely with the production of IFN-γ, IL-5, IL-4 and IL-4δ2. Differences between the stimuli were marginal.

Discussion

In this study, HIV-infected patients were selected on the basis of a nadir of < 50 CD4 T cells/µl prior to treatment and an increase to > 200 CD4 T cells/µl or four times nadir at the time of testing. For the many HIV patients worldwide awaiting treatment, this increase in CD4 T-cell count would be regarded as a good outcome. Hence, we sought to understand the immunological environment that it creates.

IFN-γ production was similar or depressed in HIV patients compared with uninfected controls, but IL-5 production in PHA-stimulated cell cultures was elevated (Table 1), especially in patients with undetectable plasma HIV RNA. CD4 T cells contributed to the majority of IL-5 production in controls, but CD4 and CD8 T cells contributed to the IL-5 production in HIV patients. Many HIV patients had elevated IgE levels. These did not correlate with IFN-γ production, and IgE levels correlated inversely with IL-5 responses. Eosinophil counts were normal in HIV patients and did not correlate with IL-5 levels in supernatants, IL-5 ELISPOT counts or IgE levels.

In the second study, the levels of IFN-γ, IL-5, IL-4 and IL-4δ2 mRNA were developed as markers of the cytokine balance. The IFN-γ ELISPOT responses to CMV and anti-CD3 correlated with each other and with the IFN-γ mRNA response to CMV. Hence, within the dynamic range of the ELISPOT assay, mRNA levels appear to reflect cytokine production. Quantification of mRNA allowed greater sensitivity and discrimination between IL-4 and IL-4δ2.

The levels of IL-4 and IFN-γ mRNA correlated with each other and with IL-5 and IL-4δ2 mRNA levels after anti-CD3 stimulation, where IL-5 mRNA levels were the best predictor of IFN-γ mRNA levels after stimulation with anti-CD3 (P = 0·006). This is consistent with a positive correlation between the levels of T1 cytokines [IL-2 plus IFN-γ and tumour necrosis factor-α (TNF-α)] and T2 cytokines (IL-4, IL-5 and IL-10) described in HIV- and HCV-infected women,³⁰ and contrasts with active tuberculosis and asthma where there is excess IL-4 mRNA relative to IL-4δ2 or IFN-γ.^14,18,31 Excess IL-4δ2 has been linked with inhibition of IL-4-mediated B-cell activation,³² which provides an explanation for the association between a high IL-4/IL-4δ2 ratio and elevated IgE in asthmatics.³¹ Here, IgE levels correlated inversely with mRNA for T2 cytokines (IL-4 and IL-5) and T1 cytokines (IFN-γ and IL-4δ2) (Table 4). Differences between the stimuli were marginal, so our failure to find inverse relationships between T1 and T2 cytokines is unlikely to reflect our choice of stimuli.

In contrast with the results obtained with IgE, weak positive correlations were evident between plasma CD30 levels and the levels of IFN-γ, IL-4, IL-4δ2 and IL-5 mRNA after stimulation. We have previously demonstrated an inverse correlation between CD30 levels and tuberculin skin test reactivity in HIV patients³³ or IFN-γ production by peripheral blood cells from patients or controls. This correlation was lost in HIV patients with severe disease, and CD30 levels were most consistently elevated in HIV patients who were viraemic.¹⁷ Similarly, elevated CD30 effectively marks disease activity in patients with atopic dermatitis.³⁴ Hence, rather than circulating CD30 being a T2 marker, it may simply reflect ‘disease activity’. Notably, IgE was a better marker of PPD skin non-reactivity than CD30 in healthy adults.³⁵ It is now accepted that CD30 has a regulatory role, which includes stimulation of CCR7 expression required for the migration of T_em into target organs. CD30 ligation stimulates cytolytic activity of CD8 T cells and secretion of soluble CD30, which invokes a negative feedback loop.^36–38 Hence, high levels of circulating CD30 would be expected to parallel increased cytokine responses by PBMC, as observed here.

No serological factors were significantly associated with IFN-γ mRNA levels in unstimulated cells. However, there was a marginal correlation with LAG-3 levels (P = 0·06). Our assay detected ≥ 22 pg/ml LAG-3, which is below the normal range previously defined using these reagents.³⁹ However, 11/19 samples read below this cut-off, so the association with IFN-γ mRNA may strengthen with more sensitive assay. This warrants further attention because it would be consistent with evidence that soluble LAG-3 marks a T1 cytokine response.^13,40 T-cell surface LAG-3 (CD223) has been implicated in T-cell homeostasis⁴¹ and may have a negative regulatory role.⁴² Soluble LAG-3 has adjuvant activity that is attributed to competitive inhibition of these activities.⁴⁰

In conclusion, these analyses provide no evidence for an inverse relationship between T1 and T2 cytokine responses in unstimulated cells or after in vitro stimulation and do not link either serum CD30 or CD26(DPPIV) with a T1 or T2 cytokine environment. Elevated serum IgE warrants further analysis as a marker of persistent T-cell response dysregulation. Follicular T cells (T_FH) may be important in this regard, as they produce IL-4 and IL-5 and are associated with the generation of autoantibody.⁴³

Abbreviations

AEC

3-amino-9-ethylcarbazole

ART

antiretroviral therapy

BSA

bovine serum albumin

CD26(DPPIV)

CD26 dipeptidyl peptidase IV

CMV

cytomegalovirus

ELISA

enzyme-linked immunosorbent assay

ELISPOT

enzyme-linked immunosorbent spot-forming cell assay

HIV

human immunodeficiency virus

IgE

immunoglobulin E

IFN-γ

interferon-γ

IL

interleukin

IL-4δ2

inhibitory splice variant of IL-4

LAG-3

lymphocyte activation gene-3

NK

natural killer

NNRTI

non-nucleotide reverse transcriptase inhibitor

PBMC

peripheral blood mononuclear cells

PBS

phosphate-buffered saline

PHA

phytohaemagglutinin

PI

protease inhibitor

PPD

purified protein derivative

RT–PCR

reverse transcription–polymerase chain reaction

TMB

3,3′,5,5′-tetramethylbenzidine