Introduction

Cognitive decline and structural brain changes have been described among HIV-infected persons on and off antiretroviral therapy (ART). In the absence of ART, dementia onset is associated with progression to AIDS. With suppressive ART, more subtle forms of neurocognitive decline are more common than frank dementia (Heaton et al. 2010; Sacktor et al. 2002), and cognitive decline likely results from a combination of traditional (age, diabetes, vascular disease, obesity), HIV- and ART-specific risk factors (Ances et al. 2012; Cysique and Brew 2009). With improved control of HIV, white matter abnormalities (which are associated with cognitive impairment) have also declined in this population.(Navia et al. 1986; Petito et al. 1986)

Magnetic resonance imaging (MRI)-derived volumetric analysis is a robust method for assessing brain integrity (Paul et al. 2008). While structural changes occur prior to measurable cognitive decline (Clark et al. 2012), we and others have demonstrated that brain atrophy correlates with poorer neurocognitive performance in HIV-infected persons (Becker et al. 2012; Cohen et al. 2010). In the general population, visceral adiposity and obesity have been associated with declines in cognitive function and brain volume (Debette et al. 2010; Ho et al. 2010; Kanaya et al. 2009). In persons with well-controlled HIV infection, visceral adiposity and generalized obesity are common problems (Jacobson et al. 2005; Sculier et al. 2014; Sharma et al. 2014); however, the contribution of metabolic disease, including obesity and lipohypertrophy, to structural brain changes in HIV-infected persons is not well defined.

We designed a cross-sectional analysis of HIV-infected and HIV-uninfected men enrolled in the Multicenter AIDS Cohort Study (MACS) who underwent both MRI structural brain imaging and computed tomography (CT)-quantified abdominal adipose tissue measurement within a 12-month period. We predicted that there would be independent effects of HIV serostatus, ART, and abdominal fat deposition on brain structure in this convenience sample of men living with or at risk for HIV infection.

Methods

Study population

The MACS began in 1984 to study the natural history of HIV infection among men who have sex with men and to establish a repository of biologic specimens for future study (Kaslow et al. 1987). Participants were enrolled from four sites (Pittsburgh, PA; Baltimore, MD/Washington, DC; Chicago, IL; Los Angeles, CA) over three time periods (1984/85, 1987/90, 2001/03) and return semi-annually for a standardized interview, clinical evaluations, laboratory tests, and storage of specimens.

The MACS Cardiovascular Disease 2 (CVD2) sub-study enrolled MACS participants who were 40–70 years of age, did not have a history of heart surgery (coronary artery bypass grafting or valve surgery) or coronary angioplasty, weighed ≤300 lb and were able and willing to provide informed consent (Hacioglu et al. 2013). Of note, men enrolled in CVD2 are generally similar to the larger MACS cohort. In 2011 and 2012, CT-quantified visceral adipose tissue (VAT) area measurements were obtained on CVD2 participants. Also in 2011, MRI brain examinations were performed on a subset of CVD2 participants, including high-resolution anatomic sequences. We conducted a nested cohort study of 226 MACS participants who (1) participated in the CVD2 and MRI sub-studies and (2) completed both CT and MRI imaging studies within a 1-year period.

Clinical assessments

Adipose tissue quantification

Non-contrast CT assessed VAT area in cm2 at the L4-L5 level using a single slice. Following a centralized training, scans were performed locally (Los Angeles and Pittsburgh: 64-detector LightSpeed Volume CT scanner, GE Medical Systems, Milwaukee, Wisconsin, USA; Chicago: 64-detector SOMATOM Definition Dual Source scanner, Siemens Medical Solutions, Forchheim, Germany; Baltimore: 320-detector row AquilionONE, Toshiba Medical Systems, Tokyo, Japan) and read centrally by an experienced reader at Los Angeles Biomed (Harbor-University of California, Los Angeles, Torrance, CA), who also assessed scan quality and consistency. VAT was defined as tissue density between −190 and −35 Hounsfield units between the internal-most abdominal and oblique muscle walls and the posterior aspect of the L4-L5 vertebral body. Men were stratified by “normal” (nVAT, <100cm2) or “elevated” (eVAT, ≥100cm2) VAT area, which represented both a natural break in our data and is consistent with population-based studies.(Hiuge-Shimizu et al. 2012; Mangili et al. 2015; Williams et al. 1996)

MRI brain volumetric measurement

The MRI scanning sequences were taken from the protocol developed by the Alzheimer’s Disease Neuroimaging Initiative for use with scanners with three Tesla field strengths (http://www.adni-info.org/images/stories//mritrainingmanualv1.pdf). All scans were acquired on Siemens 3T Trio scanners (maximum gradient slew rate: 200mT/m/s; maximum gradient strength 40mT/m), with the Siemens phase-array head coil (Basel, Switzerland). The sagittal magnetization prepared rapid gradient echo (MP-RAGE) sequence used for this analysis was the following: field of view −256 mm; slices 160; repetition time 2300 ms; echo time 2.91 ms; inversion time 900 ms; flip angle 9 degrees; thickness = 1.2 mm. Following scan acquisition, de-identified MRI data was first transferred to the Image Data Archive at the Laboratory of Neuro Imaging at the University of Southern California (https://ida.loni.usc.edu/login.jsp), followed by storage on local, password-protected hard drives at the University of Pittsburgh with the last author of this report. A copy of all MRI data was also sent to the central data repository in Baltimore.

Biomarker assessments

Circulating adipokine levels were measured under the direction of Dr. Russell Tracy at the University of Vermont Laboratory for Clinical Biochemistry Research (Burlington, Vermont, USA). Stored (−70° Celsius) fasting serum and plasma (obtained at the time of CT scan) were analyzed for adiponectin, leptin, high-sensitivity C-reactive protein (hs-CRP), and interleukin-6 (IL-6) levels. Total adiponectin, leptin, and IL-6 were measured by enzyme-linked immunosorbent assays (R & D Systems, Minneapolis, MN, USA). The lower limit of detection for adiponectin was 390 ng/mL, with an inter-assay coefficient of variation of 5.3–10.8%. For leptin, the lower limit of detection was 1300 pg/mL, with an inter-assay coefficient of variation range of 5.9–6.8%. For IL-6, the lower limit of detection was 0.5 pg/mL, with an inter-assay coefficient of variation range of 6.6–12.5%. hs-CRP was measured by nephelometry (lower limit of detection 0.2 μg/mL; inter-assay coefficient of variation range 3.0–6.2%).

Other measurements

Age, race, level of education, smoking history, medication use, diagnosis history, and physical activity level (low, medium, high) were assessed by self-report unless otherwise defined. AIDS and other clinical events were confirmed via medical record review. Education was assessed continuously. Depression was defined as Center for Epidemiologic Studies Depression (CESD) Scale Score >16. Hypertension was defined as resting BP >130/90, self-report of hypertension or use of anti-hypertensive medications. Diabetes was defined as self-report or use of anti-diabetic medications. Chronic Hepatitis C Virus (HCV) infection was defined as plasma HCV RNA positivity. Chronic Hepatitis B Virus (HBV) infection was defined as positive HBV surface antigen or diagnosis of chronic HBV infection. Height and weight were measured using standardized procedures and used to calculate the body mass index (BMI) in kg/m2. Mid-waist and hip circumferences (cm) were measured using a standardized protocol (Westat 1988) and used to calculate waist-to-hip ratios. All socio-demographic and medical data were obtained using information collected at the MACS visit closest and immediately prior to the CT scan date. CD4+ T lymphocyte subsets were enumerated using standardized flow cytometry (Hultin et al. 2007), and CD4+ T lymphocyte nadir was defined as the lowest count prior to and including the CT scan date. Peak HIV-1 RNA was defined as the highest value prior to and including the CT scan date.

Standard protocol approvals, registrations, and patient consents

All participants provided informed consent, and both the parent and sub-study protocols were approved by the Institutional Review Boards of the participating sites. Written informed consent was obtained from all participants prior to their undergoing research procedures. All procedures were performed in compliance with the standards set forth by the Declaration of Helsinki. This study is registered at clinicaltrials.gov (NCT00046280).

Statistical analysis

Metabolic and vascular variables were reduced to categorical classifications (present/absent) based on standard criteria or the distribution of values within the HIV-uninfected group (Becker et al. 2009). MRI data were first processed through a non-parametric, non-uniform intensity normalization (Boyes et al. 2008; Sled et al. 1998) to reduce between-scan and between-site differences. We used a normal template image created with the Template-O-Matic (TOM8) toolbox using 68 randomly selected, HIV-uninfected men from the Information eXtraction from Images (IXI) Project (whose age distribution mimics the controls in this analysis) and estimated the prior probabilities of each tissue class (i.e., gray matter [GM], white matter [WM], cerebrospinal fluid [CSF]) for use in the segmentation routines (Becker et al. 2011). We then used the voxel-based morphometry-8 (VBM8) toolbox SPM12 (http://dbm.neuro.uni-jena.de/vbm8/vbm8_spm12_r442.zip) for affine-registration, normalization, and segmentation of the data. The resulting GM maps were modulated to render the values in each of the 1 × 1 × 1 mm voxels as a volume, which were then smoothed using an 8 × 8 × 8 mm Gaussian filter to reduce the effects of registration error and render the data more amenable to parametric analysis.

Because participants were scanned at different sites (and on two different model scanners), we also entered three binary dummy variables to adjust for possible between-site differences. We also adjusted for enrollment cohort effect. For all analyses, the default threshold for reporting statistical significance was set at a false discovery rate of p ≤ 0.05, with an extent threshold of 100 voxels (Genovese et al. 2002). For the purposes of visualization, we projected the results onto the Colin27 standard brain (Holmes et al. 1998). The mean whole brain image created from the HIV-uninfected participants was used to project all study findings.

Univariate associations between clinical and demographic characteristics and brain volume were then determined. After determining brain regions of interest for key variables, we extracted eigenvariates from the three largest regions associated with VAT and the two largest associated with BMI (Table 2). These individual volumes were extracted from 4-mm radius spheres centered on the peak voxel within each of the five regions of interest. We then used stepwise forward regression (p < 0.05 for model inclusion) to regress these five values on significant predictor variables from the univariate analysis in the following order: age, race (white vs non-white), diabetes, hypertension, BMI (<30 vs ≥30 kg/m2), VAT (<100 vs ≥100 cm2), log IL-6, and adiponectin levels as continuous variables and HIV serostatus. For all analyses, significance was defined using a nominal, two-sided α = 0.05.

Results

Patient population

Complete demographic and clinical characteristics of the 217 men (133 HIV-infected, 84 HIV-uninfected) included in this analysis are presented in Table 1 as a function of VAT area. Overall, the men had mean age of 55 years, 15 years of education, BMI 26 kg/m2, VAT 164 cm2, and 26% were of non-white race. HIV-infected men were slightly younger (55 vs 57 years) but more likely to be of white race, have hypertension and diabetes, and had a greater mean VAT area (171 vs 151 cm2) at a slightly lower mean BMI (25 vs 26 kg/m2). Similarly, HIV-infected men were more likely to have eVAT (73 vs 61%), despite lower BMI. At the time of MRI scanning, the HIV-infected men had a mean CD4+ T lymphocyte count of 593 cells/μL, and 82% had a plasma HIV-1 RNA level <50 copies/mL.

Table 1 Characteristics of participants by HIV serostatus and abdominal fat volume
Full size table

VAT distribution and associated factors

Overall, 65% of men had elevated VAT area, with the mean in the eVAT and nVAT groups being 211 and 61 cm2, respectively (Table 1). Men with eVAT were slightly older, had slightly higher BMIs, and were more likely to have diabetes and hypertension. HIV-infected men with eVAT had higher current CD4+ T lymphocyte counts than men with nVAT and were less likely to have an AIDS diagnosis. GM (F[1214] = 3.86, p = 0.05, ξ p 2 = 0.02) and CSF (F[1214] = 4.36, p = 0.04, ξ p 2 = 0.02) volumes were all significantly different between the two groups, suggesting a whole brain consequence of having elevated VAT.

VBM assessment

With regard to GM volume, univariate analysis demonstrated significant effects of age, VAT area, and BMI, with a smaller effect observed for HIV serostatus (Table 2). Older age was associated with atrophy in the cingulate gyrus (including the precuneus and the anterior cingulate), insula, medial and superior frontal gyri, and middle temporal gyrus (Table 2, Fig. 1). BMI was significantly associated with volume in the right caudate nucleus and the left inferior frontal gyrus. VAT area, by contrast, was associated with volume in the superior temporal gyrus, insula, and right caudate nucleus. HIV serostatus was associated with smaller total GM and caudate volume. There were no interactions between HIV serostatus and any factors associated with body habitus.

Table 2 Areas of significant* associations with gray matter volume
Full size table
Fig. 1
figure 1

Brain regions whose volumes are significantly associated with body mass index (light blue) and visceral adipose tissue area (purple). Areas of significant overlap in caudate nucleus are marked with arrows. The left hand image is taken in the coronal plane, and the right hand image is in the axial plane

Full size image

In the regions associated with VAT area in univariate analysis (superior temporal gyrus, insula, and caudate nucleus), VAT area remained a significant predictor of these volumes after controlling for age, BMI, waist circumference, diabetes, hypertension, and adiponectin, leptin and log IL-6 levels (Table 3). Caudate nucleus volume was significantly associated with the presence of diabetes. In multivariate analysis, the effect of HIV serostatus on total GM volume was smaller and no longer statistically significant. Only a region of the caudate nucleus remained significantly associated with HIV infection after controlling for age, study site, BMI, and VFA. No other covariates demonstrated significant associations in these brain regions. Further, in an HIV-restricted analysis, no significant correlations between regional brain volumes and current or nadir CD4+ T lymphocyte count, or current or peak plasma HIV-1 RNA were apparent (all r’s <0.17).

Table 3 Predictors of regional brain volumes1
Full size table

Older age and white race were significantly associated with same brain regions as BMI (right caudate nucleus and left inferior frontal gyrus [BA11]). Similarly, obesity, defined as BMI ≥30 kg/m2, was associated with smaller brain volumes in these two regions. Log IL-6 level was also inversely associated with caudate nucleus volume, but no other biomarker associations demonstrated significant associations. No interactions were observed between HIV serostatus, BMI, VAT area, or log IL-6 level, nor did level of physical activity predict VAT area (data not shown).

Discussion

This analysis has two main findings: First, we observed significant associations between greater VAT area and higher BMI and brain volume regardless of HIV serostatus. Second, regional brain atrophy is independently associated with older age, white race, and, in the case of the caudate nucleus, by higher IL-6 levels. These data suggest that, among HIV-infected men on suppressive ART (as the vast majority of HIV-infected men were in this analysis), the relationship between visceral adiposity and brain volume is not additionally altered by HIV infection.

These data also highlight the importance of traditional risk factor management in the prevention of chronic, non-infectious comorbid disease in treated HIV infection. In fact, in this analysis, we observed relationships between BMI and brain volume in participants with a mean age of 55 that were similar but attenuated compared to relationships we previously described in HIV-uninfected populations over the age of 65 (Ho et al. 2010; Ho et al. 2011; Raji et al. 2010), suggesting a cumulative adverse effect of obesity on brain volume over time. Further, complicating this relationship is the observation that HIV-infected men may have greater VAT relative to BMI, particularly when clinical lipodystrophy is present (Brown et al. 2009; Joy et al. 2008; Study of Fat and Metabolic Change in 2006), a finding we also observed in this cohort. As such, it has been hypothesized that a higher prevalence of eVAT may contribute to both increased risk for and earlier onset of neurologic disease in HIV-infected persons on ART compared to HIV-uninfected controls (Mateen et al. 2012). Similarly, because VAT may be elevated in the face of a “normal” BMI among HIV-infected men, BMI may not be a clinically sufficient surrogate for visceral adiposity burden in this population.

There is a potential link between visceral adiposity, chronic inflammation, and brain atrophy: HIV infection is characterized by immune activation and systemic inflammation that persists in the face of suppressive ART (Baker et al. 2011), and chronic systemic inflammation has been associated with neuro-inflammation, brain atrophy, and cognitive decline, particularly in the hippocampus (Daulatzai 2014). In fact, poorer neurocognitive function has been significantly linked to abdominal obesity (as measured by waist circumference), systemic inflammation (high IL-6 levels), and immune activation in plasma (high soluble CD14 levels) and CSF (high soluble CD40L levels) in HIV-infected persons with plasma HIV-1 RNA <1000 copies/mL (Sattler et al. 2014). Additionally, in older adults, higher IL-6 levels have been associated with poorer executive function (Mooijaart et al. 2013; Schram et al. 2007). In this analysis, higher log IL-6 levels were associated with greater caudate nucleus atrophy particularly among HIV-infected men. Among the HIV-infected men, the relationship between IL-6 and brain volume was more complex, and likely affected by both cohort sampling difference and selective survival bias from the early days of the HIV epidemic (data not shown). However, these data support a relationship between chronic systemic inflammation and brain atrophy.

Another possible mediating link between visceral adiposity and brain atrophy is adipocytokine imbalance. In HIV-infected persons, visceral adiposity is associated with decreased growth hormone and adiponectin levels (Diaz-Delfin et al. 2013). Leptin levels are suppressed in HIV-associated lipodystrophy but elevated in generalized obesity (Paruthi et al. 2013; Sinha et al. 2012), and both growth hormone and leptin replacement therapy decrease VAT in HIV-infected persons (Diaz-Delfin et al. 2013; Mulligan et al. 2009). Adiponectin is preferentially produced by VAT, whereas leptin is preferentially produced by subcutaneous adipose tissue. Adiponectin and leptin receptors are both present in the brain (Gustafson 2010), and adipokine homeostasis could be important for maintaining cognitive function. Supporting this hypothesis are the findings that lower adiponectin and leptin levels are associated with cognitive impairment and brain atrophy (Kamogawa et al. 2010; Lieb et al. 2009; Narita et al. 2009). Additionally, adiponectin may variably serve pro- and anti-inflammatory roles (Hattori et al. 2008; Tomizawa et al. 2008). In this analysis, adiponectin and leptin were not significantly associated with regional brain volume, although any effect might have been overshadowed by the effect of VAT area.

This study has several limitations. First, it represents a cross-sectional analysis, and therefore cannot account for longitudinal changes in risk factors overtime. However, the MACS is a well-described cohort of HIV-infected and HIV-uninfected men, and both the HIV-uninfected control group and the sample size are strengths of this analysis. Another limitation is that our findings may not be generalizable to women or ART-naïve HIV-infected persons; however, our findings support those observed in the general population, which enhances generalizability.

In conclusion, increased VAT area, obesity, and higher log IL-6 levels were associated with regional brain atrophy in HIV-infected and HIV-uninfected men in the MACS. These findings are consistent with previously published studies in the general population, but have not previously been reported among HIV-infected persons. Additionally, HIV- and ART-specific risk factors did not predict brain atrophy. However, since HIV infection is a chronic inflammatory state and ART-treated, HIV-infected persons may have higher VAT-to-BMI ratios than HIV-uninfected persons; these findings may be important for clinical risk factor management and prevention of cognitive decline in HIV-infected adults on suppressive ART.