Autonomic Performance and Dehydroepiandrosterone Sulfate Levels in HIV-1–Infected Individuals: Relationship to T_H1 and T_H2 Cytokine Profile

Background  Products of immune activation, including cytokines and lipid membrane derivatives, have been implicated in the pathogenesis of the neurologic sequelae, including autonomic dysfunction, associated with human immunodeficiency virus 1 (HIV-1) infection. In animal models, autonomic and endocrine dysfunction are associated with an altered cytokine profile.

Objectives  To investigate the relationship between markers of immune activation (β₂-microglobulin), HIV-1 disease progression (CD4⁺ cell count and viral load), and autonomic nervous system performance and to assess the relationship between autonomic performance, plasma levels of dehydroepiandrosterone sulfate (DHEAS), and T_H1 and T_H2 cytokine profile.

Methods  Thirty-one HIV-1–infected individuals and 22 HIV-1–negative controls were evaluated with a comprehensive neurologic, neuropsychological, and autonomic examination. Interleukin 4 and interferon gamma were measured by enzyme-linked immunosorbent assay in the supernatant of stimulated peripheral blood mononuclear cells.

Results  A composite measure of autonomic performance (AZ score) was significantly lower (worse autonomic function) in patients compared with controls (P=.04). A lower AZ score was associated with higher β₂-microglobulin serum levels and a lower CD4⁺ cell count. Interleukin 4 levels were significantly inversely associated with AZ score (P=.01), whereas interferon gamma levels were significantly positively associated with DHEAS levels (P=.04).

Conclusions  Our data show significant associations between markers of immune activation and disease progression and a composite measure of autonomic function in HIV-1–infected individuals. In addition, they suggest that poor autonomic function and low DHEAS plasma levels tend to be associated with an unbalanced cytokine profile.

HUMAN immunodeficiency virus 1 (HIV-1) infection is associated with persistent immune activation.¹ Products of immune activation, including cytokines, arachidonic acid metabolites, and free radicals, may be neurotoxic and have been implicated in the pathogenesis of central and peripheral nervous system manifestations of HIV-1 infection.^2,3 Immune activation, in turn, is regulated in part by the elicited cytokine response. CD4⁺ T cells, often referred to as helper cells, can secrete distinct patterns of cytokines that are broadly associated with either cellular or humoral mediated immunity.⁴ T_H1 cells are characterized in part by secretion of interleukin 2 (IL-2), interferon gamma (IFN-γ), and lymphotoxin, and are associated with an inflammatory cellular immunity. The T_H2 profile is associated with antibody and allergic response and is characterized by the production of IL-4, IL-5, IL-10, and IL-13. Another phenotype of T_H cells (T_H0) can secrete both T_H1- and T_H2-identifying cytokines.⁵

Some of the HIV-1–associated conditions, such as autonomic and endocrine dysfunction, may play a role in the balance of the T_H1 and T_H2 cytokine profile. For example, stimulated spleen cells derived from sympathectomized animals secrete less IL-2 and IFN-γ,⁶ characteristic of a T_H2 predominant response. In vivo and in vitro data also suggest that the adrenal hormone dehydroepiandrosterone (DHEA) and its sulfate (DHEAS) may counteract the effect of glucocorticoids and favor a T_H1 response.⁷ Low serum levels of DHEA (evidence of endocrine imbalance) have also been associated with progression to the acquired immunodeficiency syndrome (AIDS).^8,9

We hypothesized that autonomic dysfunction (AD), a common neurologic complication of patients with AIDS,^10,11 may be associated with markers of immune activation (β₂-microglobulin) and HIV disease progression (CD4⁺ T-cell count and HIV-1 RNA). We also hypothesized that an altered autonomic nervous system and low plasma DHEAS levels might potentiate the immune dysregulation caused by HIV infection, tilting the cytokine balance toward a T_H2 profile. In this regard, we assessed whether associations existed between a composite measure of autonomic function and IFN-γ (T_H1 cytokine) and IL-4 (T_H2 cytokine). Similar relationships were examined between plasma DHEAS and IFN-γ and IL-4. Finally, we assessed whether severity of distal symmetrical polyneuropathy and performance on neuropsychological tests were associated with severity of AD.

Thirty-one HIV-1–positive individuals and 22 HIV-1–negative controls were enrolled after signing an informed consent approved by the institutional review board at the University of Rochester Medical Center, Rochester, NY. The patients were recruited from the university's infectious disease clinic and neurology clinic. Most subjects were coenrolled in a longitudinal study of patients at risk for HIV-1 dementia.¹² Subjects were excluded if other conditions known to cause AD were present, including diabetes mellitus, adrenal insufficiency, heart disease, dehydration, and alcoholism. The subjects were evaluated between 8 and 12 AM either before or at least 2 hours after a meal. They were instructed to avoid caffeinated drinks and to abstain from taking any medication the night before the evaluation.

We performed a standard autonomic battery as described by Ewing,¹³ which included the Valsalva maneuver (Valsalva ratio), heart rate variation with deep breathing (maximum minus minimum heart rate), heart response to standing (30:15 ratio), postural decrease in systolic blood pressure, and increase in diastolic blood pressure with sustained hand grip. Because of the inability of the first few patients tested to perform the Valsalva maneuver and hand grip as described by Ewing, these tests were modified as follows: in the Valsalva maneuver, the patient blew against a resistance of at least 30 mm Hg instead of 40 mm Hg and the sustained hand grip was performed for 1 minute instead of 5 minutes. Mathias and Bannister¹⁴ have suggested that an intrathoracic pressure between 20 and 40 mm Hg should be sufficient to induce appropriate changes during the Valsalva maneuver. All patients and 22 healthy controls were tested with this modified autonomic battery. The sex-specific means and SDs of the controls were used to derive z scores for each autonomic test in patients. The 5 z scores were then averaged, yielding a composite autonomic z score (AZ score). The more negative the AZ score, the worse the autonomic response.

Cognitive performance was assessed with a standard battery that has been described elsewhere.¹² In brief, it included the Rey Auditory Verbal Learning test (total score, trial 5 score, delayed recall, recall after interference, correct recognition), Rey Complex Figure 1 Copy and Recall, Digit Symbol test, Grooved Pegboard (dominant and nondominant sides), Timed Gait, Verbal Fluency, and the Odd-Man-Out test. A z score was computed for each test (except for Timed Gait) using age- and education level–specific norms from the AIDS Link to Intravenous Experience study and the Multicenter AIDS Cohort Study. The z scores were averaged for all tests to obtain a neuropsychological composite z score (NPZ score). Three domain z scores were also created by averaging the z scores for certain tests: memory (Rey Auditory Verbal Learning Test total score, trial 5 score, delayed recall, recall after interference, correct recognition), psychomotor function (Digit Symbol test, Grooved Pegboard), and executive function (Rey Complex Figure Copy and Recall, Odd-Man-Out test). These domains were verified by a principal components factor analysis with Varivax rotation (data not shown).

The macroneurologic examination used in the AIDS Clinical Trial Group protocols was used to assess the presence of distal symmetrical polyneuropathy (DSP). The DSP was graded as absent, mild (impaired sensory modalities limited to great toes), moderate (decreased sensory modalities involving ankles), or severe (sensory impairment extending to knees and wrists). The current use or use within the last 6 months of zalcitabine, didanosine, and stavudine was recorded.

Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll-Hypaque (Pharmacia, Piscataway, NJ) density centrifugation, washed, and cryopreserved in liquid nitrogen using a controlled-rate freezer. The PBMCs were thawed and adjusted to approximately 10⁶ cells/mL in a 1% fetal calf serum, albumin-containing lymphocyte culture media (AIM-V; Gibco, Grand Island, NY). The cells were stimulated with either media control or 5 µg/mL of phytohemagglutinin (Sigma-Aldrich, St Louis, Mo). Supernatants were harvested at 3 days of culture and frozen in duplicate at −70°C.

The concentrations of IL-4 and IFN-γ were determined with a capture enzyme-linked immunosorbent assay using antibody pairs obtained from PharMingen (San Diego, Calif). In brief, enzyme-linked immunosorbent assay plates (Corning, Corning, NY) were coated with mouse anti–human cytokine monoclonal antibodies overnight, washed, and blocked with phosphate-buffered saline and 2% albumin. The samples were then added undiluted (IL-4) or diluted 1:5 (IFN-γ) in duplicate for a 3-hour incubation at room temperature. After washing, a biotin-labeled second monoclonal anticytokine antibody was added, the plates washed, followed by streptavidin-peroxidase, washing, and addition of the substrate 2-azinodio-3-ethylbenzthiazoline (ABTS) (Sigma-Aldrich). The plates were read after 20 to 60 minutes at 405 nm. All samples were run in duplicate. Specific cytokine values were determined by recombinant standards (PharMingen or R & D System, Minneapolis, Minn), which always yielded straight line fits with an R² of 0.97 or more. The lower limits of detection defined in our assays were 1 pg/mL for IL-4 and 30 pg/mL for IFN-γ. Samples of PBMCs from 8 healthy HIV-1–negative individuals were also obtained to compare cytokine profiles between HIV-1–seronegative and HIV-1–seropositive individuals.

The DHEAS plasma levels were quantified by a radioimmunoassay kit (Coat-A-Count DHEA-SO4; DPC, Los Angeles, Calif) according to the manufacturer's directions.

Serum β₂-microglobulin was measured by radioimmunoassay (SmithKline Beecham), with normal values being less than 2.9 mg/L.

Plasma HIV RNA copies were measured with Roche Amplicor assay. The limit of detection of this assay is 400 copies/mL.

The AZ score and cytokine levels were compared among patients and control subjects using Wilcoxon rank sum tests. Associations between the AZ score and markers of immune activation and disease progression (β₂-microglobulin, CD4⁺ T-cell count, viral load) and cytokine levels were examined using standard regression analysis. Associations between DHEAS levels and cytokine levels, neuropsychological test results and AZ score, and markers of immune activation and disease progression and cytokine levels were similarly examined. Analysis of variance was used to compare means among patients with no, mild, and moderate DSP. The Tukey standardized range procedure was used to perform pairwise comparisons among the 3 groups. The DHEAS level, β₂-microglobulin, CD4⁺ T-cell count, viral load, and cytokine levels were transformed using the natural logarithm for purposes of statistical analysis, with log (0) taken to be zero.

Analyses with AZ score as the dependent variable were repeated after adjusting for age, and analyses with DHEAS level as the dependent variable were repeated after adjusting for both age and sex. Since the results were virtually identical before and after adjustment, only the results of the unadjusted analyses are reported for simplicity. A 2-tailed significance level of .05 was used for all tests, unless otherwise noted.

Table 1 summarizes the results of the 5 autonomic tests in patients and control subjects. Patients demonstrated the most impairment in heart rate variation with deep breathing (mean ± SD of z score, −0.56 ± 1.31) followed by the Valsalva ratio (mean ± SD of z score, −0.44 ± 0.86). The AZ score was significantly lower in patients compared with controls (−0.32 ± 0.75 vs 0.00 ± 0.45; P=.04).

β₂-Microglobulin was assayed in 28 patients, and viral load was available in 14 patients. A highly statistically significant linear relationship was found between the AZ score and log (β₂-microglobulin) (r=−0.50, P=.007) (Figure 1). There was no significant association between viral load and AZ score (r=−0.35, P=.22). The mean log (CD4⁺ T-cell count) was significantly associated with the AZ score (r=0.47, P=.008; Figure 1) after exclusion of an outlier.

The DHEAS levels were assayed in 31 patients, whereas 24 of the 31 patients had IL-4 and IFN-γ levels assessed. The PBMC samples from 17 of the 24 patients secreted a measurable amount of IL-4 compared with 4 of 8 controls. The levels of IL-4 in patients (mean ± SD [range], 13.9 ± 23.9 pg/mL [0-102 pg/mL]) and controls (mean ± SD [range], 4.7 ± 8.0 pg/mL [0-20 pg/mL]) were not significantly different. The levels of IFN-γ produced by patients (mean ± SD [range], 11,110 ± 11,706 pg/mL [147-34,931 pg/mL]) and controls (mean ± SD [range], 11,195 ± 8094 pg/mL [2409-27,094 pg/mL]) were also not significantly different. In the patients, higher log (IL-4) levels were significantly associated with a lower AZ score (r=−0.52, P=.009; Figure 2). The relationship between log (IFN-γ) and AZ score was not statistically significant (r=0.18, P=.39). Log (DHEAS) was significantly, positively associated with log (IFN-γ) (r=0.42, P=.04; Figure 2). The correlation between log (DHEAS) and log (IL-4) did not reach statistical significance (r=−0.23, P=.28).

Twenty-one of the 31 individuals were affected by either mild or moderate DSP, but no one met criteria for severe DSP (see the "Subjects and Methods" section for definitions of criteria). Patients with moderate DSP generally had higher serum levels of β₂-microglobulin than those with no or mild DSP (Table 2). The difference among the mild and moderate DSP groups approached statistical significance (nominal P=.02). The difference among the no and moderate DSP groups was less robust (nominal P=.06). The severity of DSP tended to be associated with mean age and the mean log (CD4⁺ T-cell count) (Table 2); however, differences among the groups were not statistically significant. Patients with moderate DSP had a significantly lower AZ score than patients without neuropathy (nominal P=.006). This difference between the moderate and mild DSP groups approached significance (nominal P=.08).

Nucleoside analogues (didanosine, zalcitabine, stavudine) are known to cause a neuropathy clinically similar to HIV-associated DSP; the effect of these drugs on the autonomic nervous system is not well documented. Eighty percent of patients without neuropathy (n=10), 40% with mild DSP (n=6), and 67% with moderate DSP (n=4) were exposed to 1 or more of these nucleosides. Sixty-three percent of patients with an AZ score of 0 or greater (n=8; normal autonomic function) were taking at least 1 of these antiviral drugs compared with 57% of patients with an AZ score less than 0 (n=23).

Cognitive performance (NPZ score) was significantly associated with AZ score (r=0.48, P=.02). Executive function was the domain that correlated best with AZ score (r=0.52, P=.01). For the 3 cognitive domains, only the relationship between executive function and β₂-microglobulin approached significance (r=−0.4, P=.06).

We found a significant association among blood levels of β₂-microglobulin, CD4⁺ T-cell count, and a composite score of autonomic function in HIV-1–infected individuals. We also observed a trend toward a correlation between viral load and AD. Furthermore, our data suggest that in HIV-1–infected individuals with AD or low plasma levels of DHEAS there is a trend toward a T_H2 cytokine profile. Viral load had the weakest association with AD compared with β₂-microglobulin and CD4⁺ T-cell count. However, this observation is limited by the small sample of patients with available viral load. Further studies are necessary to address this issue. In addition, AD has been reported to be present more frequently in patients with AIDS,^10,11 although it can occur in early HIV-1 infection.¹⁵ Our data show that not only CD4⁺ T-cell count but also other markers of HIV-1 disease progression (β₂-microglobulin) are associated with AD, confirming and expanding these previous reports.

Our finding of an association between β₂-microglobulin (a marker of immune activation)¹⁶ and AD is consistent with the hypothesis that the persistent immune activation associated with HIV-1 infection can lead to the release of neurotoxic substances, including cytokines and lipid membrane derivatives, which may then cause neuronal injury. This cascade of events may represent a common mechanism of HIV-1–induced neurologic diseases.^2,3 In the peripheral and autonomic nervous systems, this hypothesis is, in part, supported by the increased numbers of macrophages and T lymphocytes and expression of major histocompatibility class II molecules found in sensory and sympathetic ganglia of HIV-1–infected individuals.^17-19 However, HIV-1 may also contribute more directly to neuronal damage with some of its viral components. For example, gp120 has been shown to bind to rat sensory ganglia.²⁰

Often, HIV-1–associated AD is subclinical and may not represent a serious concern for most patients or clinicians. However, both AD and low DHEAS levels may potentiate the immune dysregulation associated with HIV-1 infection. This hypothesis is based on 2 decades of mounting evidence that indicates the presence of a cross-regulatory interaction among the nervous, endocrine, and immune systems.²¹ The autonomic nervous system represents a direct link between the nervous and immune systems, establishing synapselike contacts with the immunocompetent cells in primary and secondary lymphoid organs.²² Lymphocytes and monocytes express β-adrenergic receptors.²³ The stimulation of these receptors is linked to the activation of adenylate cyclase and consequent accumulation of cyclic adenosine monophosphate. Accumulation of cyclic adenosine monophosphate has been shown to not only alter cytokine production^24-26 but also to enhance HIV replication.²⁷

Further evidence of the immune-modulating role of the autonomic nervous system comes from studies of sympathectomized animals. In animal models of experimental allergic encephalomyelitis and rheumatoid arthritis, chemical sympathectomy worsens the disease.^28,29 Moreover, splenocytes from sympathectomized mice tend to produce less IFN-γ and IL-2.⁶ Our results in part replicate these animal data. We found that HIV-1–infected individuals had a similar cytokine profile to that of HIV-negative controls, in agreement with some^30,31 but not all³² previous reports. The HIV-1–negative group with available cytokines was very small in our study. However, autonomic function (AZ score) in patients was inversely associated with IL-4 and positively, but not significantly, associated with IFN-γ, a trend toward a T_H2 bias.

In addition, DHEA has an in vitro and in vivo protective role against viral infections,³³ including an inhibitory effect on HIV-1 replication.³⁴ Some of activities of DHEA and DHEAS may be explained by their immunomodulating role.^7,35 Clerici et al³⁶ have proposed that an increase in the ratio of cortisol to DHEA in HIV-1 infection may favor a T_H2 phenotype. Although we did not measure cortisol levels, our results partially support this hypothesis in that IFN-γ was significantly positively associated with serum levels of DHEAS, whereas a nonsignificant inverse correlation was observed between DHEAS levels and IL-4.

Distal symmetrical polyneuropathy^37,38 and cognitive impairment³⁹ are 2 other common neurologic complications associated with HIV-1 infection and may share similar pathogenetic mechanisms with AD. We found a significant association between AD and both cognitive performance and severity of DSP. However, we cannot definitely conclude from our results whether these findings reflect the fact that most neurologic complications occur in more advanced HIV-1 disease or that a common pathogenetic mechanism is present. The significant association between AD and β₂-microglobulin and a trend toward increased serum levels of β₂-microglobulin in individuals with moderate DSP would suggest that immune activation could play a role in both these conditions. The relationship between cognitive performance and serum β₂-microglobulin was weaker. However, it has been previously shown that cerebrospinal fluid levels of β₂-microglobulin are a better marker of HIV dementia than serum levels,⁴⁰ but we did not have cerebrospinal fluid available in our patients.

These results must be interpreted in light of the following limitations. First, there were several (planned) outcome variables examined in this exploratory study, and the reported P values are not adjusted for multiple statistical testing. Second, cause-and-effect relationships cannot be reliably inferred in this cross-sectional investigation.

In summary, our data show a significant association between autonomic nervous system function and peripheral markers of immune activation and disease progression in HIV-1 infection. The data also suggest that AD and, to a lesser extent, low plasma levels of DHEAS are associated with a trend toward a T_H2-type cytokine response in stimulated PBMCs. Cytokine dysregulation may play an important role in the pathogenesis of HIV-1 infection^1,41 and its neurologic complications.^2,3 It appears that HIV-1 preferentially infects T_H0 and T_H2 cells³⁰; therefore, HIV-1 infection persistence and dissemination could be enhanced by conditions that potentially contribute toward a T_H2 bias. We speculate that AD and possibly low DHEAS levels may be among those conditions that can exercise positive feedback on cytokine unbalance and ultimately on HIV disease progression. Interventional studies that combine DHEA supplementation with modulation of the autonomic response are needed to further investigate this hypothesis.

Accepted for publication November 29, 1999.

This study was supported in part by a General Clinical Research Center grant (5 MO1 RR00044) from the National Center for Research Resources, National Institutes of Health, Bethesda, Md.

Presented in abstract form at Research Perspectives in Psychoneuroimmunology VIII; Bristol, England; April 4, 1998.

Reprints: Giovanni Schifitto, MD, Department of Neurology, University of Rochester, 601 Elmwood Ave, Box 673, Rochester, NY 14642 (e-mail: gschifitto@mct.rochester.edu).