Relationship of Fat Distribution with Adipokines in Human Immunodeficiency Virus Infection

Abstract

Background and Methods: HIV-infected patients receiving antiretroviral therapy often develop changes in body fat distribution; the dominant change is reduction in sc adipose tissue (SAT). Because adipose tissue makes important hormones involved in whole-body energy metabolism, including leptin and adiponectin, we examined plasma concentrations and their relationship to regional adiposity measured by magnetic resonance imaging in 1143 HIV-infected persons (803 men and 340 women) and 286 controls (151 men and 135 women) in a cross-sectional analysis of the FRAM study.

Results: Total and regional adiposity correlated positively with leptin levels in HIV-infected subjects and controls (P < 0.0001). In controls, total and regional adiposity correlated negatively with adiponectin. In HIV-infected subjects, adiponectin was not significantly correlated with total adiposity, but the normal negative correlation with visceral adipose tissue and upper trunk SAT was maintained. However, leg SAT was positively associated with adiponectin in HIV-infected subjects. Within the lower decile of leg SAT for controls, HIV-infected subjects had paradoxically lower adiponectin concentrations compared with controls (men: HIV 4.1 μg/ml vs. control 7.5 μg/ml, P = 0.009; women: HIV 7.8 μg/ml vs. control 11.6 μg/ml, P = 0.037). Even after controlling for leg SAT, exposure to stavudine was associated with lower adiponectin, predominantly in those with lipoatrophy.

Conclusion: The normal relationships between adiponectin levels and total and leg adiposity are lost in HIV-infected subjects, possibly due to changes in adipocyte function associated with HIV lipodystrophy, whereas the inverse association of adiponectin and visceral adipose tissue is maintained. In contrast, the relationship between adiposity and leptin levels appears similar to controls and unaffected by HIV lipodystrophy.

In HIV patients the effects of abnormalities in body fat and of antiretroviral drugs on adipocytokines has been unclear. In this study, normal relationships between adiponectin levels and total and leg adiposity were lost in these patients, while its inverse association with visceral adiposity was maintained. In contrast, the relationship between adiposity and leptin levels was unaffected by HIV lipodystrophy.

Adipose tissue is a storage depot for fuel and an endocrine organ (1). Leptin and adiponectin are two adipokines produced by adipose tissue that act locally and systemically to influence whole-body energy balance and metabolism.

In the general population, plasma leptin concentrations strongly and positively correlate with adipose tissue mass (2). Obese individuals have higher leptin levels than lean individuals. Very low concentrations are found when fat mass is profoundly reduced as in anorexia nervosa (3) and some lipodystrophy syndromes (4). Individuals with congenital generalized lipodystrophy, characterized by almost complete absence of body fat, have leptin levels 5- to 10-fold less than healthy controls.

However, leptin concentrations vary in partial lipodystrophy syndromes characterized by loss of adipose tissue from sc depots (5,6,7). When partial lipodystrophy is associated with an overall reduction in fat mass, leptin levels are low (5,6), but leptin levels may be normal in individuals with partial lipodystrophy who have overall fat mass similar to healthy controls (7,8). Thus, in lipodystrophy syndromes, leptin concentrations remain proportional to total adiposity.

Adiponectin levels are inversely related to adipose tissue mass (9). Concentrations rise when obese persons lose weight (10). Patients with anorexia nervosa have higher adiponectin levels compared with controls (11,12). However, in lipodystrophy syndromes, including HIV lipodystrophy, adiponectin levels have been found to be relatively low despite the lower fat mass (5,6,7,13,14), suggesting that adiponectin does not retain its normal inverse relationship with total adiposity in lipodystrophy.

Although in vitro data suggest that sc adipocytes make more leptin and adiponectin than visceral adipocytes (15,16,17), the contribution of each depot to circulating adipokine levels is unclear. Given that sc adipose tissue (SAT) accounts for approximately 85–90% of adipose tissue in lean individuals (18), SAT may produce the majority of circulating adipokines.

The syndrome of lipoatrophy in HIV infection exists on a continuum rather than being a condition that is fully absent or present. Thus, there are quantitative changes in body fat in HIV-infected persons exposed to antiretroviral (ARV) therapy (19,20). Longitudinal studies in ARV-naive patients initiating therapy demonstrate initial increases in extremity and trunk fat, with subsequent decreases in extremity fat and relative preservation of trunk fat (21). In the study of Fat Redistribution and Metabolic Change in HIV Infection (FRAM), MRI was used to quantify regional adipose tissue to understand the characteristics of fat distribution in HIV infection and how regional adiposity relates to metabolic and hormonal parameters. Hence, FRAM offers a valuable opportunity to explore the relationship of serum leptin and adiponectin concentrations to direct measures of total and regional adiposity as well as other factors unique to HIV infection.

Subjects and Methods

Study design

This was a cross-sectional analysis of 1143 HIV-infected participants and 286 controls in the FRAM study with leptin or adiponectin measurements. For comparisons of HIV and control characteristics, 592 (of the 1143) HIV-infected participants of similar age to controls (33–45 yr old) were analyzed. FRAM was designed to evaluate the prevalence and correlates of changes in fat distribution, insulin resistance, and dyslipidemia in a representative sample of HIV-positive participants and controls in the United States. The methods, including MRI measurement of adipose tissue volume, have been described in detail previously (19,20,24) and appear in brief published as supplemental data on The Endocrine Society’s Journals Online web site at http://jcem.endojournals.org. The protocol and consent procedure was approved by institutional review boards at all sites.

Adipokines

Total adiponectin and leptin were measured by RIA at Linco Research Inc. (St. Louis, MO). The sensitivity of the human adiponectin assay is 1.0 μg/ml (at 3.0 μg/ml, the inter- and intraassay coefficients of variation were 6.2 and 6.9%). The sensitivity of the human leptin assay is 0.5 ng/ml (at 7.2 ng/ml, the inter- and intraassay coefficients of variation were 4.6 and 5.0%).

Magnetic resonance imaging (MRI)

Adipose tissue volumes were normalized in all analyses by dividing by height squared with summaries back-transformed to 1.75 m of height. We did not adjust for body mass index (BMI) because BMI is influenced by the very phenomenon being studied, i.e. quantity of fat. The HIV-specific decrease in sc fat, which does not affect visceral adipose tissue (VAT) (19,20), makes such adjustments improper. Using these methods, we quantified adipose tissue volume in the leg, lower trunk (abdomen and back), upper trunk (chest and back), arm, and VAT. Because control subjects on average were heavier than HIV-infected subjects, we examined the effect of controlling for lean mass, which is the part of weight that is fully distinct from the regional fat amounts that we examined as separate predictors.

Other measurements

Age, gender, ethnicity, medical history, and risk factors for HIV were determined by self-report, and alcohol, tobacco, and illicit drug use were assessed by standardized questionnaire. Height, weight, and blood pressure were measured by standardized protocols.

CD4 lymphocyte count and HIV RNA were measured in HIV-infected participants. Research associates performed medical chart abstraction of medications and medical history at HIV sites.

Statistical analysis

Analyses were restricted to men and women with adipokine measurements. Analyses that compared HIV-infected persons with controls excluded HIV-infected individuals with recent opportunistic infections (OI), and age was restricted to 33–45 yr old (n = 592 for HIV-infected) to match the age range of the controls. Characteristics of HIV-infected participants and controls were compared and tested for statistical significance using the Mann-Whitney U test for continuous variables and Fisher’s exact test for categorical variables. Spearman correlation coefficients were calculated to examine the relationship of adipokines with adipose tissue.

To assess the relationship of adipose tissue depots with adipokines, we performed multivariable linear regression analyses in separate models in HIV⁺ and control, using stepwise regression with P = 0.05 for entry and retention. Because of their skewed distribution, the adipokines were log-transformed, and results were back-transformed to produce estimated percentage effects of each factor. Gender, age, and ethnicity were forced to be included in every model, and HIV RNA level (log10) and CD4 count (log2) at the time of study visit were forced to be included in the HIV model. To test the validity of pooling men and women in this analysis, interactions between gender and other factors in the model were assessed. Additional candidate variables tested in the multivariable models included MRI-measured adipose tissue volume from five anatomic sites (plus total SAT, total adipose tissue, and percent total body adipose tissue), tobacco use, alcohol intake, adequate food intake, physical activity level, and current illicit drug use (marijuana, speed, heroin, crack, cocaine, and combination use of crack and cocaine). The linearity assumption for continuous predictors was also tested. Confidence intervals were determined using the bias-corrected accelerated bootstrap method (22), with P values defined as the one minus the highest confidence level that still excluded zero; this was necessary because the error residuals appeared to be non-Gaussian.

The five adipose tissue sites considered were visceral, lower trunk, upper trunk, arm, and leg. Measurements were normalized by dividing by height squared, analogous to BMI, and summaries were back-transformed to 1.75 m of height. Because of their skewed distribution, the adipose tissue measures were also log-transformed. To properly model the relationship of adipose tissue with the adipokines, smoothing splines were constructed using generalized additive models (23). Examination of these smoothed curves suggested a linear relationship in some depots but in most instances suggested quadratic relationships. To simplify the presentation of results, we also fit quartiled versions of the adipose tissue sites, created using quartile cutoffs from the control group (men and women were done separately) to facilitate comparison of similar quantities of adipose tissue. Similarly, deciles of leg SAT were defined using cutoffs from the control group, with men and women done separately.

In addition to the variables listed above, candidates tested in the HIV model included AIDS diagnosis by CD4 or OI, HIV duration, hepatitis C virus (HCV) infection (by virus detection), recent OI status (last 100 d), days since last OI, and HIV risk factors. In multivariable models controlling for the factors above that were independent predictors, we evaluated total duration of each individual ARV drug and ARV class: nucleoside reverse transcriptase inhibitor, nonnucleoside reverse transcriptase inhibitor, protease inhibitor, and highly active ARV therapy, as previously defined (19).

All analyses were conducted using the SAS system, version 9.1 (SAS Institute, Inc., Cary, NC).

Results

Characteristics of the FRAM participants have been described in detail previously (19,20,24). The demographics of the slightly smaller subset that had adipokines measured are shown in Table 1. Overall, 32 HIV-infected subjects and 11 controls were excluded from analysis due to missing adipokine measurements but were similar in characteristics to those included. As shown previously (19), total adipose tissue was lower in HIV-infected men compared with controls (Table 1), with a similar trend in women. In HIV-infected men, total SAT volume and all SAT depots were smaller compared with controls. In HIV-infected women, total SAT was also smaller compared with controls as were most sc depots except for upper trunk SAT, which tended to be larger in HIV-infected women. VAT was lower in HIV-infected men compared with controls (Table 1), whereas HIV-infected women had higher VAT compared with their controls.

Table 1.

Demographic and clinical characteristics

	Males (age restricted)			Females (age restricted)			HIV+ (full cohort)
	HIV+	Control	P value	HIV+	Control	P value	Males	Females
n	415	151		177	135		803	340
Age (yr)	40.0	40.0	0.90	39.0	42.0	0.0001	43.0	41.0
	37.0–43.0	37.0–43.0		36.0–42.0	37.0–44.0		38.0–49.0	36.0–47.0
Ethnicity
Caucasian	56%	53%	0.021	32%	50%	0.024	55%	33%
African-American	31%	47%		55%	50%		33%	55%
Hispanic	12%	0		11%	0		10%	9%
Other	1%	0		2%	0		2%	3%
Height (cm)	176.0	178.5	0.0002	162.6	164.5	0.004	175.3	162.6
IQR	171.5–181.0	174.5–183.0		158.8–166.4	161.0–170.5		171.0–180.3	158.1–166.9
Weight (kg)	75.6	85.1	<0.0001	71.8	75.2	0.028	74.7	69.6
IQR	68.2–84.0	77.5–98.0		58.7–83.2	63.7–88.8		67.8–82.9	59.2–82.5
BMI (kg/m2)	24.3	27.0	<0.0001	26.6	27.9	0.25	24.4	26.4
IQR	22.3–26.8	24.6–30.3		22.2–32.2	23.0–33.2		22.2–26.7	22.5–31.6
Total AT (liters)	12.0	17.0	<0.0001	27.0	31.3	0.055	11.8	27.3
IQR	8.8–17.3	12.8–21.4		17.3–38.8	20.7–41.5		8.8–17.1	17.6–38.5
Total SAT (liters)	10.2	14.7	<0.0001	25.2	30.1	0.033	10.1	25.5
IQR	7.1–14.7	11.1–18.8		15.9–36.8	19.7–39.5		7.1–14.2	16.6–36.8
VAT (liters)	1.6	2.0	0.010	1.3	1.1	0.067	1.7	1.4
IQR	0.7–2.7	1.1–3.0		0.6–2.3	0.4–2.0		0.8–3.0	0.6–2.4
Leg SAT (liters)	2.9	4.5	<0.0001	7.3	10.0	<0.0001	2.8	7.6
IQR	2.0–4.2	3.7–5.7		5.1–10.8	7.2–13.9		2.0–4.0	5.1–11.0
Upper trunk SAT (liters)	2.5	3.1	0.0001	5.9	5.0	0.16	2.5	5.8
IQR	1.7–3.5	2.3–4.1		3.2–8.8	2.9–8.1		1.7–3.6	3.4–8.3
Lower trunk SAT (liters)	3.4	5.6	<0.0001	9.6	10.9	0.029	3.3	9.5
IQR	2.1–5.3	4.1–7.4		5.6–13.3	7.1–15.2		2.1–5.1	5.7–13.4
Arm SAT (liters)	1.0	1.1	0.036	1.8	2.2	0.007	1.0	1.9
IQR	0.8–1.3	0.9–1.4		1.2–2.4	1.4–2.9		0.8–1.3	1.2–2.6
HCVRNA+	16%	3%	<0.0001	27%	1%	<0.0001	20%	26%
Duration HIV (yr)	8.0			8.5			8.0	8.0
IQR	5.2–12.0			6.0–10.9			5.1–12.0	5.8–10.8
Current CD4 (cells/μl)	358			372			347	364
IQR	227–516			186–565			220–526	198–557
HIV RNA (1000/ml)	0.4			0.8			0.4	0.5
IQR	0.4–14.6			0.4–14.6			0.4–11.5	0.4–14.8

Note that the table contains medians and interquartile ranges (IQR). AT, Adipose tissue. 

Adipokine levels in HIV-infected participants vs. controls

Leptin levels are positively correlated with total SAT. Consistent with the lower fat mass found in HIV-infected men, median leptin levels were substantially lower in HIV-infected men than in controls (2.9 vs. 4.5 ng/ml, P < 0.0001, Table 2). Although HIV-infected women had somewhat less total adipose tissue, leptin levels in HIV-infected women were similar to those in control women (median 14.3 vs. 14.1 ng/ml, P = 0.43).

Table 2.

Adipokine levels in men and women by HIV status

	Leptin (ng/ml)			Adiponectin (μg/ml)
	Median	IQR	P value	Median	IQR	P value
Control men	4.5	3.2–6.6	<0.0001	5.5	4.0–8.3	0.12
HIV+ men	2.9	1.9–5.2		5.2	2.9–8.1
Control women	14.1	7.8–20.0	0.43	8.7	5.6–12.4	0.12
HIV+ women	14.3	7.7–23.2		8.0	4.6–10.8

IQR, Interquartile range. 

Adiponectin levels have been reported to be inversely correlated with total body fat in controls (9,25), but not in HIV infection (13,26). Despite the lower adipose tissue volume in HIV-infected men, adiponectin levels were not higher in HIV-infected subjects compared with controls (men: median 5.2 vs. 5.5 μg/ml, P = 0.12; women: median 8.0 vs. 8.7 μg/ml, P = 0.12).

Leptin and adiponectin levels were significantly greater in women than in men in both HIV-infected and control groups (P < 0.0001), both before and after multivariable adjustment.

Associations of adipose tissue volumes with leptin and adiponectin

In unadjusted models in both HIV-infected and control participants of both genders, leptin was positively and significantly correlated (P < 0.0001) with percent adipose tissue as well as with each regional depot (Table 3). However, associations of VAT with leptin in HIV-infected men and leg SAT in HIV-infected women were weaker than controls after multivariable adjustment.

Table 3.

Associations of body composition with adipokines

	% Fat	Total SAT	VAT	Leg SAT	Upper trunk SAT	Lower trunk SAT	Arm SAT
Leptin
Control men	0.82a	0.82a	0.69a	0.68a	0.79a	0.83a	0.65a
HIV+ men	0.77a	0.74a	0.47a	0.56a	0.68a	0.74a	0.60a
HIV × depoteP values	0.97	0.21	0.007	0.10	0.19	0.74	0.86
Control women	0.85a	0.85a	0.71a	0.82a	0.78a	0.82a	0.79a
HIV+ women	0.83a	0.82a	0.51a	0.66a	0.77a	0.80a	0.76a
HIV × depoteP values	0.43	0.57	0.109	0.002	0.053	0.57	0.46
Adiponectin
Control men	−0.25c	−0.31b	−0.21d	−0.19d	−0.28c	−0.28c	−0.32b
HIV+ men	0.03	0.07	−0.37a	0.29a	−0.16a	0.10c	0.04
HIV × depoteP values	0.022	0.002	0.45	<0.0001	0.17	0.002	0.006
Control women	−0.40a	−0.46a	−0.53a	−0.29c	−0.57a	−0.47a	−0.42a
HIV+ women	−0.06	−0.13d	−0.37a	0.10	−0.32a	−0.15d	−0.12
HIV × depoteP values	0.062	0.086	0.44	<0.0001	0.78	0.14	0.023

Note that table contains unadjusted Spearman correlations (r). Analysis was age restricted, and subjects with recent OI were excluded. 

^aP < 0.0001; 

^bP < 0.001; 

^cP < 0.01; 

^dP < 0.05. 

^eHIV × depot interactions P values are from multivariable models with log-transformed adipokines as outcome; models were adjusted for HIV status, age, ethnicity, lifestyle factors, and the depot of interest. 

In control men and women, adiponectin levels showed the expected negative correlations with percent adipose tissue (Table 3). Adiponectin was also significantly, negatively associated with all regional adipose depots in controls.

In contrast, in HIV-infected men and women, there was little association of adiponectin levels with percent adipose tissue (Table 3). However, in both genders, univariate analysis found the expected negative correlation of VAT and upper trunk SAT with adiponectin in HIV infection, whereas in HIV-infected women, adiponectin was also weakly negatively correlated with lower trunk SAT (Table 3); these associations did not change significantly after multivariable analysis, including adjustment for lean mass. Of note, the association of leg SAT with adiponectin was positive in HIV-infected men and women (Table 3), a reversal of the normal relationship. In HIV-infected men, lower trunk SAT also had a weak positive association with adiponectin. The paradoxical associations of these depots with adiponectin remained significantly different between HIV-infected and control subjects after multivariable analysis (Table 3).

Lipoatrophy and adiponectin

Because leg and lower trunk SAT, the two depots most affected by HIV lipoatrophy, had unexpected associations with adiponectin in HIV-infected subjects, we examined the distribution of leg SAT and adiponectin levels. Figure 1 shows the distribution of leg SAT (height-normalized) in control and HIV-infected subjects. Both HIV-infected men and women demonstrated a dramatically lower distribution of leg SAT than controls. To place the adiponectin findings in perspective, we analyzed how many HIV-infected subjects fell into the lowest decile of controls. Among HIV-infected men, 49% had leg SAT below the cutoff point marking the lower decile (10%) of control men. For HIV-infected women, 32% had leg SAT below the cutoff point of the lower decile (10%) for control women.

Figure 1.

Distribution of leg SAT and adiponectin levels. Distribution of height-normalized leg SAT is shown by histogram with smoothed density curve created using kernel density estimation. Decile reference line was defined using cutoffs from the control group, with men and women done separately, as described in Materials and Methods. A, Men; B, women. Light bars and dashed lines represent control subjects. Dark bars and solid lines represent HIV-infected subjects. Analysis was age restricted, and those with recent OI were excluded.

There was little difference in median adiponectin levels between HIV-infected and control men or women with leg SAT in the upper 90% for controls (men: HIV 5.9 μg/ml, control 5.5 μg/ml, P = 0.30; women: HIV 8.3 μg/ml, control 8.6 μg/ml, P = 0.61). In contrast, for those within the lower decile of leg SAT of controls, HIV-infected subjects had lower median adiponectin levels than controls (men: HIV 4.1 μg/ml, control 7.5 μg/ml, P = 0.009; women: HIV 7.8 μg/ml, control 11.6 μg/ml, P = 0.037). These results may account for the positive association of adiponectin with leg SAT in HIV-infected subjects vs. the negative association in controls.

Factors associated with leptin and adiponectin in HIV-infected subjects after multivariable adjustment

We next used multivariable analysis to examine associations of adipose tissue depots with adiponectin in HIV-infected participants (Table 4). In this analysis, women were found to have higher adiponectin levels than men and African-Americans were found to have lower adiponectin levels than Caucasians, even after adjusting for adipose tissue volume and other factors.

Table 4.

Multivariable analysis of factors associated with adiponectin and leptin in all HIV⁺ men and women

	Adiponectin			Leptin
	% effect	95% CI	P value	% effect	95% CI	P value
VAT Q2 vs. Q1	−34.0	−42.2, −24.5	<0.0001	19.7	8.1, 31.8	<0.0001
VAT Q3 vs. Q1	−43.8	−51.1, −35.5	<0.0001	22.4	10.1, 35	<0.0001
VAT Q4 vs. Q1	−45.6	−53.1, −36.9	<0.0001	46.5	29.8, 63.7	<0.0001
Leg SAT Q2 vs. Q1	30.1	13.1, 49.6	0.0002	23.3	10.4, 37.1	<0.0001
Leg SAT Q3 vs. Q1	30.0	10.3, 53.3	0.002	28.0	13.6, 43.9	<0.0001
Leg SAT Q4 vs. Q1	43.7	18.3, 74.6	0.0003	36.2	16.6, 58.7	<0.0001
Upper trunk SAT Q2 vs. Q1	1.7	−10.8, 15.9	0.80	19.8	8.5, 32.2	<0.0001
Upper trunk SAT Q3 vs. Q1	−12.1	−25.0, 3.0	0.11	28.9	13.8, 46.7	<0.0001
Upper trunk SAT Q4 vs. Q1	−17.7	−32.0, −0.5	0.044	56.5	35.7, 83.5	<0.0001
Lower trunk SAT Q2 vs. Q1	16.4	1.1, 34.0	0.035	29.0	15.6, 45.1	<0.0001
Lower trunk SAT Q3 vs. Q1	12.5	−6.6, 35.6	0.21	55.2	34, 79.2	<0.0001
Lower trunk SAT Q4 vs. Q1	16.4	−7.5, 46.5	0.19	88.2	58.8, 121.6	<0.0001
Arm SAT Q2 vs. Q1				17.5	7.5, 28.2	0.001
Arm SAT Q3 vs. Q1				24.6	11.6, 39	<0.0001
Arm SAT Q4 vs. Q1				12.0	−1.3, 27.5	0.078
Current HIV viral load (log 10)	−3.5	−8.4, 1.8	0.19	0.5	−3.3, 4.6	0.77
Current CD4 (log 2)	−6.5	−11.5, −1.3	0.015	−0.4	−4.3, 3.6	0.84
Total duration of stavudine	−8.0	−10.3, −5.7	<0.0001
Total duration of ritonavir	4.4	0.5, 8.6	0.029	6.6	3.7, 9.7	<0.0001
Total duration of indinavir				3.0	0.6, 5.5	0.018
Total duration of nelfinavir				3.4	0.9, 6	0.006
HCV (RNA > 615)	27.2	14.4, 41.5	<0.0001	10.5	2.1, 19.3	0.013
Adjusted r2			0.307			0.780

Note that outcomes are log(adiponectin) and log(leptin), respectively. Results are back-transformed. Both adiponectin and leptin models also control for age, ethnicity, sex, and alcohol and current cocaine use. Both models control simultaneously for all depots listed above. CI, Confidence interval. 

After multivariable adjustment, VAT remained strongly associated with lower adiponectin levels. In contrast, after multivariable adjustment including VAT, leg SAT was strongly associated with higher adiponectin. The association with other depots was weaker. Lower trunk SAT was associated with higher adiponectin, whereas upper trunk SAT was associated with lower adiponectin. Although the upper trunk effect was stronger in women than in men (P = 0.039 interaction), we report results from the pooled model because upper trunk was negatively associated with adiponectin in both genders. No other statistically significant interactions were identified.

Among HIV-related factors, duration of stavudine exposure and higher current CD4 counts were associated with lower adiponectin levels. Coinfection with HCV and duration of ritonavir exposure were associated with higher adiponectin levels.

Because both lower amounts of leg SAT and stavudine exposure were independently associated with lower adiponectin levels, we examined the impact of these two factors further. Figure 2 shows adiponectin levels among those with 1) no stavudine exposure, 2) less than 2 yr exposure, and 3) more than 2 yr exposure (ranges close to tertiles of exposure) plotted within control quartiles of leg SAT. For HIV-infected subjects in the upper three quartiles of leg SAT for controls, the relationship between leg SAT and adiponectin was similar to controls and showed little effect of duration of stavudine. In contrast, stavudine duration was negatively associated with adiponectin in HIV-infected subjects in the lowest leg SAT quartile; adiponectin levels in these subjects were inappropriately low compared with controls, even in those with no stavudine exposure. Finally, although the number of HIV-infected subjects who never received ARV was small (n = 47, 4%), they did not show paradoxically low adiponectin levels (median 8.3 vs. 5.7 μg/ml in those who had received ARV), including those in the lowest leg SAT quartile (median 9.1 μg/ml in those never on ARV vs. 4.7 μg/ml in those who had received ARV).

Figure 2.

Adiponectin levels in HIV-infected subjects by quartile of leg SAT and stavudine duration. Median adiponectin levels in HIV-infected subjects are plotted as a function of quartiles of leg SAT based on control cutoff points. Regression line based on control subjects (Q1–Q4) illustrates that HIV-infected subjects with normal amounts of leg SAT (Q2–Q4) are similar to controls. White bars represent no stavudine exposure; gray bars represent less than 2 yr of stavudine exposure; black bars represent 2 or more years of stavudine exposure. Dashed line represents trend line for controls. Full cohort of all HIV-infected subjects are included.

In contrast to adiponectin, in a similar multivariable model, leptin levels were positively associated with each adipose tissue depot (Table 4). Higher leptin levels were associated with female gender, being African-American (data not shown), coinfection with HCV, and duration of ritonavir, nelfinavir, and indinavir exposure.

Discussion

In this study of adiposity and adipokine levels, there were several noteworthy findings. Adiponectin and leptin concentrations were higher in HIV-infected women compared with HIV-infected men. Therefore, the normal gender differences in these adipocytokines are preserved in HIV infection (2,9).

Both HIV-infected men and women had less total body fat compared with healthy controls, primarily due to differences in SAT. Median total SAT was 4.6 liters lower in both HIV-infected men and women compared with healthy controls, whereas median VAT was 0.3 liters lower in HIV-infected men and 0.3 liters greater in HIV-infected women compared with respective controls (Table 1). Because of SAT’s contribution to total body weight, weight and BMI showed the same pattern.

As expected, total and regional adiposity were strongly and positively associated with leptin in healthy controls. A similar relationship was found in HIV-infected men and women, suggesting that adiposity remains the major determinant of plasma leptin levels in HIV infection.

Previous studies of leptin in HIV-infected subjects have mostly compared those with a clinical diagnosis of HIV lipodystrophy to HIV-infected subjects thought to be free of body fat abnormalities and healthy controls. In these studies, leptin concentrations in HIV lipodystrophy were lower than or similar to that of controls, depending on subject selection. When patients with HIV lipodystrophy have total body fat mass similar to controls, leptin levels are similar to those of both HIV-infected and healthy controls (7,8). However, when lipodystrophy is accompanied by lower total body fat mass, leptin levels are lower compared with controls (27). Earlier studies examining those with and without HIV cachexia and controls also found leptin levels were proportional to total body fat (28). Together these studies suggest that leptin levels reflect total adiposity in both HIV lipodystrophy and wasting.

The data on adiponectin are more complex. In our controls, both total and regional adiposity were significantly and negatively correlated with adiponectin levels, similar to other reports of adiponectin correlations in the general population (9,25). Recently, abdominal SAT was found to be negatively correlated with adiponectin concentrations in Hispanics and African-Americans in univariate analyses (29). After multivariable analysis, abdominal SAT was a positive independent correlate of adiponectin. These data highlight the complex relationship of adiponectin with regional adiposity.

However, in HIV-infected populations, the normal relationships between adiponectin concentration and sc adipose depots appear to be lost or even reversed (13,26). Our analysis suggests that the loss of the normal relationship between total adiposity and adiponectin may be due to a subset of HIV-infected patients with significant HIV lipoatrophy by direct measurement. Although there is no cutoff that defines HIV-associated lipoatrophy, HIV-infected patients with leg SAT below the 10th percentile of controls had significantly lower adiponectin levels than controls with similar leg SAT volumes. Remarkably, almost half of HIV-infected men and one third of HIV-infected women fell into this category. Furthermore, in both HIV-infected men and women, the relationship between leg SAT and adiponectin was reversed compared with healthy controls, with leg SAT having a positive correlation with adiponectin.

Similar findings have been reported in smaller studies of HIV-infected subjects. For example, in 112 HIV-infected subjects (67 with lipodystrophy), there was no association between adiponectin and total adiposity, whereas adiponectin was positively correlated with both extremity and abdominal SAT (13). Other studies have also found a positive correlation between adiponectin concentrations and extremity fat (26).

What is the significance of the positive correlation between limb SAT and adiponectin levels in HIV-infected patients in this and other studies? On the simplest level, it suggests that the remaining SAT in HIV-infected patients with moderate to severe lipoatrophy is dysfunctional rather than simply reduced in size. If the only finding were reduction in adipose tissue, adiponectin levels should be higher with less body fat, as is seen in the severe fat depletion of anorexia nervosa (11). In one study, SAT had significantly lower levels of adiponectin expression in HIV-infected patients with lipodystrophy compared with HIV-infected controls without body fat changes (30). The independent association with stavudine use found here suggests a possible mechanism for this dysfunction. An alternative possibility is that the difference in HIV represents adipocyte destruction (e.g. apoptosis).

Other forms of partial lipodystrophy are also associated with low adiponectin concentrations. In familial partial lipodystrophy due to lamin a/c mutations, or dominant-negative mutations in the peroxisome proliferator-activated receptor-γ, adiponectin levels are low compared with controls despite significantly lower total body fat mass (5,31). These findings of abnormal function or loss of adipose tissue in partial lipodystrophy syndromes also support the concept that low adiponectin concentrations in HIV-infected subjects with low levels of SAT are due to lipodystrophy per se. In contrast, very low levels of fat without loss of fat cells, as in anorexia nervosa, are accompanied by high levels of adiponectin (11,12). Thus, it would appear that lipodystrophies are characterized not only by decreased fat but also by dysfunction or loss of adipocytes.

In the HIV-infected population, we also show that additional factors, such as certain ARVs, may influence adiponectin levels. Indeed, those with low leg SAT who never received ARV did not appear to have paradoxically low adiponectin levels.

HIV lipodystrophy, familial partial lipodystrophy, and obesity have been associated with high circulating levels of TNF-α (5,32,33) as well as low adiponectin. Increased TNF-α gene expression has also been found in sc fat from patients with HIV lipodystrophy (34). TNF-α inhibits adiponectin gene expression and secretion from adipocytes in vitro (35). Therefore, this cytokine may play a role in the reduced adiponectin concentrations.

The independent association of stavudine use with low adiponectin levels suggests that stavudine may not only reduce fat but also alter adipocyte function. In vitro, both stavudine and protease inhibitors have been shown to reduce adiponectin secretion from adipocytes (36). However, when HIV protease inhibitors are given to healthy volunteers, circulating adiponectin levels increase (37). In this study, ritonavir was associated with higher adiponectin levels.

In HIV-infected men and women, the normal negative relationship between visceral adiposity and adiponectin was maintained. In multivariable analysis, visceral adiposity remained a strong negative predictor of adiponectin levels in this HIV-infected population. VAT and other measures of central adiposity have also been shown to be inversely correlated with adiponectin levels in several studies of HIV-infected patients with and without clinical lipodystrophy (7,13,26). In the general population, there is also an inverse correlation between abdominal adiposity and adiponectin levels; in one study, VAT was the primary determinant of adiponectin concentrations (25). The normal negative relationship between VAT and adiponectin in HIV-infected subjects is consistent with our previous findings that VAT is not affected by the ARV drugs that affect SAT (19,20).

Several limitations of this study should be noted. A cross-sectional design cannot prove causality between ARV use, lipoatrophy, and alterations in adiponectin levels. Although use of stavudine has declined greatly in industrialized nations, stavudine-induced lipoatrophy persists after discontinuation. Furthermore, stavudine is still frequently used in the developing world. It is not possible to determine whether adiponectin abnormalities are due to reduced hormone expression in adipose tissue, decreased adipocyte differentiation, or decreased adipose cell number (14). It is possible that different results would have been found with high molecular weight adiponectin, which has been shown in some studies to have stronger correlations with metabolic factors (38,39,40,41). Finally, control subjects were taller and heavier than HIV-infected subjects. We adjusted for height and examined the effect of controlling for lean mass divided by height squared in multivariable models, thus controlling for the part of BMI not influenced by amount of fat. The associations of regional fat remained similar with adjustment for lean body mass and other potential confounders, suggesting that differences in height and weight did not distort our results.

In summary, leptin levels appear to be primarily determined by total adiposity in HIV-infected individuals with and without lipoatrophy. Although visceral adiposity had the expected inverse association with adiponectin levels in our HIV-infected population, the normal relationship between adiponectin levels and total adiposity was lost. Finally, leg SAT and adiponectin were positively correlated in this HIV-infected population, a reversal of the normal relationship, driven by decreased levels in those with low leg SAT. This decrease may be due to changes in adipocyte function or number associated with the HIV lipoatrophy syndrome.