Serum adipocytokines are related to lipodystrophy and metabolic disorders in HIV-infected men under antiretroviral therapy

Objectives: Adipocytokines, secreted by adipose tissue, may regulate fat metabolism, lipid and glucose homeostasis and insulin sensitivity. We analysed the relations between circulating concentrations of adiponectin, leptin, interleukin-6, tumor necrosis factor α and its soluble receptors sTNFR1 and R2, lipodystrophic phenotypes and metabolic alterations in patients under highly active antiretroviral therapy (HAART).

Methods: We studied 131 consecutive HIV-infected males under protease inhibitor (PI)-based HAART, with body mass index < 27 kg/m² and C-reactive protein (CRP) < 10 mg/l. Patients were classified in four groups according to clinical examination: no lipodystrophy (NL), lipohypertrophy (LH), lipoatrophy (LA) and mixed lipodystrophy (ML). In addition to adipocytokines, we measured plasma fasting levels of triglycerides, cholesterol, cardiovascular risk markers (high-sensitivity CRP and apolipoproteins B/A1 ratio), fasted and 2 h post-glucose loading glycemia and insulinemia and calculated the quantitative insulin sensitivity check index.

Results: The patients were HIV-infected and PI-treated for a mean of 8.2 and 1.6 years respectively; 74% presented lipodystrophy, 38% altered glucose tolerance and 42% hypertriglyceridemia. Insulin sensitivity correlated positively with adiponectin and negatively with leptin and interleukin-6. Adiponectin, but not leptin, negatively correlated with all metabolic parameters. Insulin resistance, metabolic defects and cardiovascular risk markers were strongly negatively correlated with the adiponectin/leptin ratio (A/L), and positively with sTNFR1. LA patients had a longer duration of infection but ML patients presented the most severe metabolic alterations, insulin resistance and A/L decrease.

Conclusions: These results suggest that adiponectin and the TNFα system are related to lipodystrophy, insulin resistance and metabolic alterations in patients under PI-based HAART. A/L and sTNFR1 could predict insulin sensitivity and potential cardiovascular risk in these patients.

Introduction

Lipodystrophies represent a heterogeneous group of diseases characterized by generalized or partial alterations in body fat development or distribution, generally associated with metabolic abnormalities: insulin resistance, altered glucose tolerance and dyslipidemia [¹]. Even if their pathophysiology is still unknown, transgenic murine models have provided strong arguments for a primary role of altered fat distribution or development which leads to secondarily insulin resistance and metabolic complications [^2,3]. Adipose tissue, in addition to its well-known role in lipid storage, is a major determinant of insulin sensitivity and modulates glucose and lipid metabolism, in particular through the secretion of adipocytokines [⁴]. Among them, tumour necrosis factor (TNF)α and interleukin (IL)-6 could contribute to insulin resistance [^5,6] while adiponectin and leptin could promote insulin sensitivity [^4,7-9]. Furthermore, adiponectin has been shown to exert protection against atherosclerosis [^7,10,11]. In transgenic lipodystrophic mice with low levels of leptin and adiponectin, exogenous administration of these hormones markedly improved insulin sensitivity, glycemic control, dyslipidemia and hepatic steatosis [^12-14]. In non-HIV-linked human generalized lipodystrophies, levels of serum adiponectin and leptin are reduced and could contribute to the metabolic alterations which are markedly improved by leptin replacement [^15-17]. Thus, the defective endocrine role of the adipose tissue could be an important determinant of metabolic alterations and cardiovascular risk in lipodystrophies.

The most frequent form of lipodystrophy in humans occurs in HIV-infected patients, generally under highly active antiretroviral therapy (HAART) [¹⁸]. Metabolic alterations are probably, at least in part, secondary to fat redistribution, but could also result from the direct effect of drugs on the liver [^19-21]. Altered fat distribution could be linked to the treatment with nucleoside reverse transcriptase inhibitors (NRTI) [²²], but its prevalence and severity are increased in patients treated with both HIV-protease inhibitors (PI) and NRTI [²³]. The deleterious direct action of some PI on adipocytes has been pointed out in vitro [^24-27], with an additive effect of TNFα [²⁸]. It was strongly suggested in analysis of adipose tissue samples from lipodystrophic patients, in which we observed altered expression of leptin and TNFα [²⁹]. Therefore, we hypothesized that the endocrine functions of adipose tissue could be altered in HIV-infected patients under PI-based HAART and could be related to metabolic alterations. To that end, we performed a cross-sectional study in 161 consecutive male patients to study the relationships between serum adipocytokines, the lipodystrophic phenotypes and metabolic alterations.

Patients and methods

Patients

We studied 161 consecutive HIV-1-infected male patients without any active opportunistic infection, followed at the Infectious Diseases Department of Rothschild Hospital, and treated with HAART including PI for longer than 3 months. This work was designed after the results of our previous pilot study of 14 patients [³⁰]. All patients gave their informed consent to take part in this study, which was approved by an institutional review committee. Since both the levels of some cytokines and the metabolic parameters are altered in overweight or obese subjects, only patients with a body mass index (BMI) < 27 kg/m² were selected. In addition, since some cytokines are overexpressed in inflammatory states, we restricted our analysis to patients with C-reactive protein (CRP) < 10 mg/l. Therefore, 30 patients were excluded from the study.

Methods

Clinical examination

The presence of a lipodystrophy was clinically assessed, using a standardized questionnaire. Patients records were validated by a physical examination, performed by the attending physician. Signs of fat loss (empty cheeks, lipoatrophy of the upper or the lower limbs, or of buttocks, phlebomegaly) and of fat accumulation [buffalo neck, increase of the cervical, abdominal or waist girths, or waist-to-hip ratio (WHR) ≥ 0.95] were systematically checked. The presence of at least two signs of fat loss or fat accumulation defined either 'lipoatrophy' (LA) or 'lipohypertrophy' (LH). The patients were classified as having a 'mixed lipodystrophy' (ML) if they fitted both the criteria of LA and LH, and as 'non-lipodystrophic' (NL) if they did not belong to any of the previous groups.

Biological evaluation

Blood samples from the patients were taken after an overnight 12 h fast. Total cholesterol and triglycerides were measured enzymatically (CX7 analyser; Beckman-Coulter, Villepinte, France). Apolipoproteins (apo) A1 and B and high sensitivity (hs) CRP were determined by nephelometry on ARRAY and IMMAGE analysers (Beckman-Coulter). The CD4 lymphocyte count was calculated using standard flow cytometry. HIV-1 RNA viral load was established using the branched DNA technique (Quantiplex HIV-1; Bayer, Tarryton, New York, USA) with a lower limit of detection of 0.5 × 10³ copies/ml. The hepatitis B (HBV) and C (HCV) viral status was assessed from the presence of hepatitis B surface antigen (HbsAg) and anti-HCV antibodies, either from the follow-up medical records or from serum measurements (AGHBS 7A4022 and HCV3.0 3B4420 kits, AXSYM system; Abbott, Abbott Park, Illinois, USA).

A 75 g oral glucose tolerance test (OGTT) was performed with measurements at 0 and 120 min of plasma glucose, and specific insulin without cross-reactivity with proinsulin (AXSYM system; Abbott). Serum leptin was measured using a radioimmunologic assay (HL81K; Linco Research, Saint Charles, Missouri, USA), with a lower limit of detection of 7.8 pg/ml. Serum levels of TNFα, sTNFR1, sTNFR2 and IL-6 were determined by enzyme-linked immunosorbent assay (ELISA) (Quantikine; R&D Systems, Oxford, UK), with a sensitivity of 0.18, 3, 1 and 0.70 pg/ml, respectively. Serum adiponectin was measured using ELISA as previously reported [¹⁵].

Insulin sensitivity was calculated using the quantitative insulin sensitivity check index (QUICKI), defined as: 1/(log G₀ + log I₀), where G₀ and I₀ represent fasting glycemia (expressed as mg/dl) and insulinemia (expressed as μU/ml), respectively [³¹].

Statistical analyses

Results are means ± SEM except for CD4 and HIV RNA which are expressed as median and range. Differences between groups were determined using analysis of variance test followed by a Fisher protected least significant difference test for pair-wise differences. Significance of the correlations was examined using the linear regression analysis. Multiple regression analysis was undertaken for the continuous variable, QUICKI, apo B/A1 and hsCRP. Circulating metabolic parameters and adipocytokines levels were logarithmically transformed for statistical analysis. We used their natural units for presentation in the tables. P values of < 0.05 were considered significant.

Results

Clinical and metabolic features of the patients

Clinical features

The characteristics of the patients are summarized in Table 1. The 131 male patients were 28 to 71 years old (mean 41), and were known to have been HIV-infected for a mean of 8.2 years. They were treated with PI for a mean duration of 19 months. The PI regimen used were: indinavir (86 patients, 65.2%), nelfinavir (29 patients, 22%), ritonavir (10 patients, 7.6%), or ritonavir plus saquinavir (six patients, 4.6%). Most patients (112, 85.5%) were treated with an association of PI and NRTI. Stavudine was the most frequent NRTI used (in 90 patients, 68.7%). Twenty patients were treated with PI ± NRTI ± non-nucleoside reverse transcriptase inhibitor (NNRTI) (one efavirenz plus abacavir, and 19 abacavir and/or nevirapine).

Table 1

Their immunovirological status was generally good, as shown by the median level of CD4 and HIV viral load. However, 11 patients had a CD4 count < 50 × 10⁶ cells/l and a mean loss of weight from baseline of 10%. These 11 patients, at risk for a latent infection with subsequent altered systemic production of cytokines, were considered in a separate group for the interpretation of adipocytokines levels and metabolic features. The remaining patients (n = 120) had a very high frequency of lipodystrophy (89 patients, 74%), that appeared after the introduction of PI therapy in all of them, and could be of the lipoatrophic (LA, n = 29), lipohypertrophic (LH, n = 22), or mixed type (ML, n = 38). Among the 131 patients, 16 and 25 were chronically infected by HBV or HCV, respectively, four being HCV and HBV co-infected. The distribution of HBV- and/or HCV-infected patients among the five phenotypic groups was not statistically different.

Metabolic features

In these lean patients (mean BMI: 21.7 kg/m²), an insulin-resistant state was common since the mean fasting insulinemia reaches 14.1 mU/l and the mean index of insulin sensitivity QUICKI was decreased to 0.338. QUICKI was negatively correlated with BMI and WHR (respectively r = -0.44, P < 0.0001 and r = -0.50, P < 0.0001). Fasting glycemia was normal (< 6.1 mmol/l) in most patients (n = 112, 85.5%). However, OGTT revealed altered glucose tolerance (following the WHO criteria, [³²]) in 42 patients (32%), including 11 with diabetes. In eight of them the diagnosis of diabetes resulted from the values of 2 h post-OGTT glycemia.

Dyslipidemia was frequent, with a mean cholesterolemia reaching 5.7 mmol/l and triglyceridemia 2.9 mmol/l; 81 patients (62%) had a serum cholesterol > 5.2 mmol/l, 55 (42%) serum triglycerides > 2.2 mmol/l, and 42 (32%) a mixed dyslipidemia.

We also evaluated two markers of the cardiovascular risk, the apolipoproteins B/A1 ratio (apoB/A1) and hs-CRP, which did not correlate together in this population, indicating that they independently predict this risk, as previously published [³³]. The mean apoB/A1 level was high in this population (0.99), whereas hs-CRP remained in the normal range (2.42 mg/l).

Adipocytokines levels and their correlations with clinical and metabolic parameters

Leptin

The mean level of leptin in the whole group of patients was of 2.4 ± 0.1 ng/ml, in the low normal range (3.9 ± 0.9 ng/ml for male subjects with a normal BMI) (Table 1). This level positively correlated with BMI (r = 0.48, P < 0.0001) and WHR (r = 0.52, P < 0.0001). Leptin also positively correlated with fasted glycemia, insulinemia and triglycerides, with 2 h post-OGTT glycemia and insulinemia, and negatively with QUICKI (r = -0.35, P < 0.0001). However, leptin no longer correlated with QUICKI when adjusted to WHR.

Adiponectin

The mean level of adiponectin was in the normal range in the group of 131 patients studied (Table 1). It was negatively correlated with BMI (r = -0.31, P = 0.0004), WHR (r = -0.33, P = 0.0002), fasted and 2 h post-OGTT glycemia and insulinemia, triglycerides, apoB/A1 and hs-CRP, and positively correlated with QUICKI (r = 0.41, P < 0.0001). When adjusted to BMI and WHR adiponectin level remained associated with QUICKI (r = 0.25, P = 0.002).

We calculated the ratio of adiponectin to leptin × 10³ (A/L). A/L strongly correlated negatively with BMI, WHR, triglycerides, fasted and 2 h post-OGTT glycemia and insulinemia, and positively with QUICKI (Table 2, and Fig. 1). The A/L ratio was negatively and independently correlated with apoB/A1 and hsCRP, as revealed by multivariable analysis including BMI, WHR and QUICKI as independent variables (r = -0.27, P = 0.013 and r = -0.27, P = 0.019, respectively).

Fig. 1

Table 2

IL-6, TNFα and soluble TNFα receptors

The mean level of IL-6 in the 131 patients was 2.94 pg/ml, in the normal range for the general population, while that of TNFα was increased at 26 pg/ml (normal < 15 pg/ml). Their values correlated together in the whole group of patients (r = 0.17, P < 0.03). The IL-6 concentration was correlated positively with WHR (r = 0.29, P < 0.02) and negatively with QUICKI (r = -0.21, P < 0.01), but not BMI. QUICKI was no longer related to the IL-6 level when adjusting to the WHR. TNFα was negatively correlated with BMI (r = -0.32, P = 0.0002) and WHR (r = -0.20, P = 0.02) but not with QUICKI. Hs-CRP was related to the levels of IL-6 (r = 0.31, P = 0.0003) and TNFα (r = 0.25, P < 0.004).

Serum sTNFR1 was correlated with WHR (r = 0.27, P = 0.002) but not with BMI. sTNFR1 was also positively correlated with most metabolic parameters: fasted insulinemia, 2 h post-OGTT glycemia and insulinemia, triglycerides, apoB/A1 and hs-CRP, and inversely correlated with QUICKI (Table 2). These correlations remained significant when adjusted for WHR. It was also correlated with TNFα (r = 0.33, P = 0.0001), IL-6 (r = 0.38, P < 0.0001), adiponectin (r = -0.21, P = 0.02) and A/L (r = -0.22, P = 0.01) but not with leptin nor sTNFR2.

Viral hepatitis status

QUICKI was not statistically different according to the presence or absence of HCV co-infection. In contrast, it was significantly lower in chronically HBV-infected than non-infected patients (0.318 ± 0.009 versus 0.340 ± 0.004, P < 0.03). It could be noted that HBV-infected patients had slightly higher BMI (22.6 versus 21.5 kg/m², P < 0.05). The serum levels of adiponectin and sTNFR were not different according to the viral hepatitis status.

Treatment with stavudine

Patients treated with stavudine presented similar BMI and WHR, but had significantly lower QUICKI than non-stavudine-treated patients (0.332 ± 0.004 versus 0.351 ± 0.005, P < 0.009). In addition, stavudine- treated patients had significantly lower circulating adiponectin levels (4.3 ± 0.5 versus 6.0 ± 0.8 μg/ml, P < 0.003), whereas leptin was not statistically different.

Analysis of adipocytokines levels and metabolic features according to the body fat distribution and the immunological status of patients

Among the 120 patients with CD4 cell count > 50 × 10⁶ cells/l, the presence of lipodystrophy was associated with insulin resistance whatever its clinical type. Indeed, QUICKI was significantly decreased in the LH, LA and ML groups versus NL. Moreover, in the LA, ML and NL groups, A/L was linked to QUICKI (r = 0.70, P < 0.0001; r = 0.32 P = 0.05; and r = 0.47 P = 0.008, respectively).

The ML group was clearly the most severely affected regarding glucose and lipid alterations and insulin resistance (Table 3). Accordingly, seven of the 11 diabetic patients belonged to this group, and 53% of patients in this group presented hypertriglyceridemia > 2.2 mmol/l (2.2-11.7). Moreover, A/L was significantly decreased in ML patients in comparison with NL and LA groups. The LH group was characterized by an increased WHR and decreased QUICKI, but was almost comparable with the NL group for metabolic parameters. LA patients had significantly increased 2 h post-OGTT glycemia, fasted and 2 h post-OGTT insulinemia and apoB/A1 in comparison with the NL group, although to a lesser extent than ML patients (Table 3). Interestingly, the WHR was significantly increased in the LA in comparison with the NL group, showing that, even if the lipoatrophic signs are prominent in LA patients, they have probably also an increased visceral fat, as previously reported [³⁴]. The duration of PI treatment was higher in ML and LA groups in comparison with the NL patients, suggesting that the lipodystrophic syndrome and the metabolic disturbances are temporally dependent upon treatment with PI.

Table 3

The group of patients with CD4 < 50 × 10⁶ cells/l had a significantly higher HIV viral load, and the mean loss of weight of these patients from baseline was about 10%. This group of patients with deeply altered immunological status showed a marked stimulation of the TNFα system as shown by the significantly higher level of TNFα and its receptors. However, their metabolic parameters including insulin sensitivity were not different from NL patients.

Discussion

In acquired or genetic lipodystrophies, several studies suggest that altered fat distribution per se is a strong determinant of insulin resistance and metabolic alterations [^1,35]. Indeed, in the present study, insulin sensitivity assessed by QUICKI was impaired in HIV-infected, HAART-treated, lipodystrophic patients, whatever the clinical features of fat redistribution, but not in non lipodystrophic patients. This provides further evidence for a strong relation between fat redistribution and insulin resistance. Therefore, the evaluation of the role of adipocytokines in the metabolic alterations of these patients was of particular interest.

The role of leptin in the modulation of insulin sensitivity is complex [³⁶]. Extremely low levels of leptin in congenital generalized lipoatrophies are responsible, at least in part, for severe insulin resistance, as shown by replacement studies in mice and humans [^12,13,16,17]. However, as shown by this study, leptin was not dramatically decreased in lipodystrophic HAART-treated patients. Our results are in accordance with a previous study showing a correlation between leptin and BMI or WHR in HIV-infected, PI-treated patients [³⁴]. They suggest that leptin levels are directly correlated with the amount of adipose tissue but not with insulin sensitivity in this population.

Increasing arguments are in favour of an important physiological role of adiponectin in promoting insulin sensitivity and cardiovascular protection [^7-11,14]. Our study shows that adiponectin levels are negatively correlated with BMI, as previously reported in obese patients [⁷]. However, we found that adiponectin predicts insulin sensitivity independently of the amount and repartition of body fat. Moreover, a decreased adiponectin level correlates with increased glycemia, insulinemia, triglycerides and cardiovascular risk markers. This suggests that altered adiponectin secretion is responsible for defects in insulin sensitivity and glucose and lipid metabolism in these patients.

Adiponectin and leptin levels did not correlate together in this population. Since they are almost exclusively secreted by the adipose tissue and since their levels are both related to the body fat mass, we found interesting to calculate the ratio of adiponectin to leptin × 10³ (A/L). This index offers the advantage to evaluate the relations between these hormones, insulin sensitivity, lipid and glucose homeostasis, and cardiovascular risk markers, independently to the body fat mass. Our results showed that A/L is a pertinent predictor of insulin resistance and correlates with cardiovascular risk markers in patients under PI-based HAART regimens, independently of the effect of body fat mass.

Our study of patients with deeply altered immunological status (low CD4) indicates a systemic stimulation of the TNFα system, probably resulting from HIV infection. However, this stimulation does not result in metabolic alterations in these patients. We also evaluated the level of circulating soluble TNFα receptors, which are thought to be better indicators of the activation of the TNFα system than serum TNFα itself, and are secreted in part by adipose tissue [^37,38]. Interestingly, the TNFα system was also activated, although to a lesser extent, in the LA and ML groups, as shown by the sTNFR1 levels. sTNFR1 correlated with most metabolic parameters, insulin resistance and cardiovascular risk markers, as well as with adiponectin and A/L. A previous study showed that sTNFR2 were increased in HIV-infected lipodystrophic patients, and correlated with insulin resistance [³⁹]. Thus, the local, adipose TNFα system could be, in addition to adiponectin, an important determinant of the metabolic alterations in HIV-infected patients. This is in agreement with our previous results, showing that the TNFα expression is enhanced in the adipose tissue from lipodystrophic patients on PI-based HAART, and correlates with insulin resistance [²⁹]. Concerning IL-6, we confirmed its negative correlation with insulin sensitivity, as reported in other populations [⁶], but found that IL-6 circulating levels were primarily linked with the distribution of adipose tissue in HIV-infected, HAART-treated patients.

In addition, we evidenced that stavudine treatment was associated with insulin resistance and low adiponectin levels, independently to the BMI. This drug could then worsen the metabolic disturbances of PI-treated patients.

The study of the viral hepatitis status of these patients showed that the chronic co-infection by HBV, but not by HCV, could alter the insulin sensitivity. This could be partly explained by the higher BMI of the HBV-infected patients. However, an independent role for HBV in insulin sensitivity could be considered since a previous study also found an association between the HBV co-infection and insulin resistance in HIV- infected lipodystrophic patients [⁴⁰].

In conclusion, this study shows a defective endocrine function of adipose tissue in HIV-infected, HAART-treated patients. This dysfunction, with altered secretion of adiponectin and stimulation of local adipose TNFα system, is associated with metabolic alterations and insulin resistance, and could be an important determinant of them. We propose that the determination of adiponectin/leptin ratio and sTNFR1 serum level could predict insulin sensitivity and potential cardiovascular risk in these patients

Further evaluation is needed concerning the role of these adipocytokines in female patients, in which adiponectin and leptin serum levels were shown to be increased, and sTNFR level decreased, in comparison with males [^4,7,37].

The comparative evaluation of these adipocytokines in patients' serum and adipose tissue samples will allow further study of the role of defective endocrine function in lipodystrophy and metabolic status.

Acknowledgements

The authors thank Dr S. Fellahi and Dr D. Rainteau and the staff of the Biochemistry department of Rothschild Hospital for their co-operation.

Sponsorship: This work was supported by grants from the Institut National de la Santé et de la Recherche Médicale (INSERM) and Agence Nationale de Recherches sur le SIDA (ANRS).