Comparison of Dorsocervical With Abdominal Subcutaneous Adipose Tissue in Patients With and Without Antiretroviral Therapy–Associated Lipodystrophy

After approval of the study protocol by the appropriate Ethics Committee, 32 study subjects were recruited from the HIV outpatient clinic of the Helsinki University Central Hospital. All were required to have been on cART for at least 18 months with unchanged regimen for at least 12 weeks before the study. The purpose, nature, and potential risks of the study were explained to the patients, and their written informed consent was obtained. The participants were subgrouped according to the presence or absence of lipodystrophy. Lipodystrophy was determined as self-reported and investigator-confirmed (J.S.) loss of subcutaneous adipose tissue with or without accumulation of adipose tissue intra-abdominally or in the dorsocervical upper trunk. Subjects fulfilling these criteria were included in the lipodystrophic study group (cART+LD+; n = 21). Subjects without lipodystrophy (cART+LD−; n = 11) had received cART without developing the aforementioned changes in body fat composition. Pregnancy and signs, symptoms, or biochemical evidence of active diseases other than HIV-1 were exclusion criteria. All subjects were studied after an overnight fast. The study subjects had participated previously in a study comparing expression of selected genes in abdominal subcutaneous adipose tissue between lipoatrophic and nonlipoatrophic patients (20). The lipoatrophic patients have also participated in a clinical treatment trial with uridine (21). All present data on dorsocervical adipose tissue as well as microarray and histology data from both dorsocervical and abdominal subcutaneous adipose tissue are novel.

After local anesthesia with lidocaine, surgical subcutaneous adipose tissue biopsies were taken from the midway between the iliac crest and the umbilicus as well as from the midline at the base of the neck. Part of the biopsy was immediately snap-frozen in liquid nitrogen, whereas another part was formalin-fixed and paraffin-embedded for subsequent histological analyses.

Genomic DNA was extracted from adipose tissue with the QIAamp DNA isolation kit (Qiagen, Hilden, Germany). Mitochondrial DNA (mtDNA) and nuclear DNA (nDNA) copy numbers were determined by quantitative PCR using the ABI 7700 sequence detection system (Applied Biosystems, Foster City, CA). All samples were run in triplicate. Absolute mtDNA and nDNA copy numbers were calculated using serial dilutions of plasmids with known copy numbers. Details on primers and exact PCR methodology have been published previously (20).

Formalin-fixed, paraffin-embedded subcutaneous adipose tissue samples were sectioned using a standard protocol. The slides were stained with hematoxylin and eosin and examined under the light microscope (Nikon Eclipse 80i; Nikon Corporation, Tokyo, Japan) for the amount of connective tissue; vascularization; lipogranulomata, i.e., lipid-laden macrophages encircling adipocytes; degree of adipocyte size variation; and the presence of cell membrane rupture and brown adipocytes. Images for publication were taken using the computer software compatible with the microscope (NIS-Elements 3.0; Nikon Corporation).

Adipose tissue per patient (100 mg) was homogenized using the isolation kit Geneclean (Bio101 systems; Obiogene, Carlsbad, CA). RNA was extracted with the RNeasy Lipid Tissue Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. Quantity and integrity of RNA were verified using RNA 6000 nanochips (Agilent 2100 Bioanalyser; Palo Alto, CA). Part of the isolated RNA was stored at −80°C until quantification of the target mRNAs by microarray. In addition, of each sample 1.5 μg of RNA were used for reverse transcription using 400U of Superscript II (Invitrogen, Carlsbad, CA) and 100 µM of oligo(dT)_12–18 (Invitrogen) as primer. Another 0.5 μg of RNA was reverse transcribed using the SuperScript VILO cDNA synthesis kit (Invitrogen).

Total RNA (1 μg/sample) was indirectly labeled using the T7 amplification method (Amino Allyl MessageAmpTM II aRNA Amplification Kit; Ambion, Austin, TX) according to the manufacturer’s instructions. The amplified RNA (aRNA; 5 μg/sample) was labeled using monoreactive Cy3 or Cy5 dyes (GE Healthcare, GE Life Sciences, Uppsala, Sweden) followed by purification according to the manufacturer’s instructions. Labeled aRNAs were hybridized onto Whole Human Genome 4 × 44 K microarrays (Agilent Technologies, Santa Clara, CA) according to the manufacturer’s instructions. The slides were then washed according to the instructions with buffers from Agilent and scanned with GenePix 4200 AL (Axon Instruments, Molecular Devices, Silicon Valley, CA) at the resolution of 5 μm/pixel and at 16 bit depth. Microarrays were technically unsuccessful in four samples and these patients (all cART+LD+), consequently, were excluded from further analyses. Microarray results were analyzed using Benjamini and Hochberg post hoc correction for multiple comparisons with statistical significance set to P ≤ 0.01. The analysis was carried out as comparison of dorsocervical to the abdominal subcutaneous adipose tissue in the cART+LD+ and the cART+LD− group separately. DAVID Bioinformatics Resources 6.7 (http://david.abcc.ncifcrf.gov) was used to select over-represented pathways and gene categories (22).

PCR of CD68 was carried out on a LightCycler 480 (Roche, Penzberg, Germany) using SYBR-Green technology by Roche as described previously (20). Expression of CD68 was normalized to the mean of three housekeeping genes (actin β, ACTB; acidic ribosomal protein P0, 36B4; β₂-microglobulin, B2M) commonly used (23). The PCR for UCP1 was carried out in a subset of samples (n = 6) from individuals with the clinically largest buffalo humps using ABI PRISM 7900HT instrument and software (PE Applied Biosystems, Foster City, CA) using Power SYBR Green PCR Master Mix (PE Applied Biosystems). Expression of UCP1 was normalized to ACTB. Expression of the homeobox genes (HOXA10, HOXC9, HOXC8, and SHOX2) chosen to verify the microarray results was measured using SYBR-Green kit according to the manufacturer’s protocol on ABI PRISM 7000 Sequence Detection System instrument and software (PE Applied Biosystems). Expression of homeobox genes was normalized to 36B4. All samples were run in duplicates, and the results were calculated from the real-time PCR efficiency (24). Primers used for PCR analyses are given in Table 1.

Body circumferences were measured in triplicates with a nonstretchable band for the waist midway between the lower rib margin and the iliac crest and, for the hip circumference, over the greater trochanters and recorded to the nearest 0.5 cm. Skinfold thickness (mean values of triplicate measurements) was determined at five sites (triceps, biceps, iliac crest, thigh, and scapula) (25). Limb, trunk, and total body fat were measured using dual-energy X-ray absorptiometry (DEXA; Lunar Prodigy, Madison, WI). Intra-abdominal fat as well as abdominal and dorsocervical subcutaneous fat were quantified by analyzing T1-weighted two-dimensional fast low angle shot trans-axial MRI scans. Details for image acquisition for intra-abdominal and abdominal subcutaneous fat have been given previously (26). Dorsocervical subcutaneous fat volume was measured from scans acquired by the methodology described above covering the area between second and third cervical vertebra to the lower edge of scapula (slice thickness, 5 mm; field of view, 315 × 500 mm²; time of repetition, 159 ms; echo time, 4.1 ms). Liver fat content was measured using proton magnetic resonance spectroscopy, as previously described (27).

Plasma glucose concentrations were measured using a hexokinase method, serum total, and HDL cholesterol and triglyceride concentrations with respective enzymatic kits from Roche Diagnostics using an autoanalyzer (Roche Diagnostics, Hitachi 917; Hitachi, Tokyo, Japan). The concentration of LDL cholesterol was calculated using the formula of Friedewald et al. (28). Serum insulin and C-peptide concentrations were determinedby time-resolved fluoroimmunoassay using Insulin and C-peptide Kits (AUTOdelfia, Wallac, Turku, Finland), respectively. HOMA-IR was calculated from the formula: fasting glucose (mmol/L) × fasting insulin (mU/L)/22.5 (29). Serum ALT activity was determined as recommended by the European Committee for Clinical Laboratory Standards. Serum C-reactive protein (CRP) was analyzed using a commercial kit (Ultrasensitive CRP Kit; Orion Diagnostica, Espoo, Finland). CD4+ T-cell count was determined using flowcytometry (FACSort/FACSCalibur; Becton Dickinson, San José, CA). HIV-1 viral load was measured using HPS Cobas TaqMan 48n HIV-1 Test (Roche Diagnostics, Branchburg, NJ) with a detection limit of 50 copies/mL. Shygmomanometric measurement of brachial pressure was conducted after a minimum of 15 min of acclimatization and before blood sampling.

Demographic and clinical parameters among the study groups were compared using Fisher exact test or χ² test for categorical variables and unpaired t test or Mann-Whitney test for continuous variables, as appropriate. Distribution was tested for normality using a Kolmogorov-Smornov test. Wilcoxon matched pairs test was used to compare intraindividual data. Correlations were calculated using Spearman rank correlation coefficient. For statistical analyses GraphPad Prism version 4 (GraphPad Software, San Diego, CA) and Lotus 1-2-3 of Lotus SmartSuite Release 9.5 (Lotus Development Corporation, IBM Corporation, New York, NY) were used. Data are given as mean ± SEM or median (25% percentile–75% percentile), as appropriate. Two-tailed P values < 0.05 were considered statistically significant. Statistical analyses of the microarray data are described in detail above.

Characteristics of the patients are given in Table 2. The cART+LD+ and the cART+LD− groups were comparable with respect to age, sex, and BMI. The cART+LD+ group had significantly higher ratios of dorsocervical to abdominal subcutaneous and intra-abdominal to abdominal subcutaneous adipose tissue than the cART+LD− group. Serum insulin, HOMA-IR, and triglyceride concentrations were significantly higher and HDL cholesterol concentrations lower in the cART+LD+ compared with the cART+LD− group (Table 2).

The mtDNA copy number was significantly higher in the dorsocervical than in the abdominal subcutaneous adipose tissue in the cART+LD+ group (356 ± 59 vs. 256 ± 44 copies/cell; P = 0.01). A similar trend was found in the cART+LD− group (930 ± 109 vs. 563 ± 161 copies/cell; P = 0.08; Fig. 1).

The mtDNA copy number was 62% lower in the cART+LD+ than in the cART+LD− group in the dorsocervical (P < 0.0001) and 54% lower in the abdominal subcutaneous adipose tissue (P = 0.03; Fig. 1).

The dorsocervical adipose tissue in both the cART+LD+ and the cART+LD− group had significantly less adipocytes (55 ± 6 vs. 77 ± 5%, P = 0.01; and 61 ± 5 vs. 89 ± 4%, P < 0.00001, respectively) and more connective tissue (45 ± 6 vs. 23 ± 5%, P = 0.01; and 39 ± 5 vs. 11 ± 4%, P < 0.00001, respectively) than the abdominal subcutaneous adipose tissue (Fig. 2). The two depots were similar with respect to variation in adipocyte size, cell membrane rupture, presence of lipogranulomata, and cells morphologically typical to BAT irrespective of the lipodystrophy status (data not shown).

The abdominal subcutaneous adipose tissue from patients in the cART+LD+ group contained significantly more lipogranulomata (P = 0.045) than the corresponding tissue from the cART+LD− group (Fig. 2). Other histological parameters did not differ significantly between the cART+LD+ and the cART+LD− group in either depot.

Microarray analysis comparing the dorsocervical to the abdominal subcutaneous adipose tissue within the cART+LD+ group identified 27,819 probes for genes of which 99 were differentially expressed. These 99 probes corresponded to 75 unique Entrez gene IDs. Of these 75 genes, 61 were dowregulated and 14 upregulated in the dorsocervical compared with the abdominal subcutaneous adipose tissue (Supplementary Table 1; full microarray data are available at Gene Expression Omnibus [GEO], http://www.ncbi.nlm.nih.gov/geo/; accession number GSE28073). The differentially expressed genes were involved in regulation of gene transcription; organ development; and regionalization, immune response, and inflammation as well as in composition and organization of extracellular matrix and cytoskeleton. Three of the top six genes downregulated in the dorsocervical adipose tissue were homeobox genes HOXA10, HOXC9, and HOXC8. These genes are known to be transcription factors providing cells with specific positional identities on the anterior-posterior axis. Among genes upregulated in the dorsocervical compared with the abdominal subcutaneous adipose tissue was short stature homeobox 2 (SHOX2), a gene responsible for cranio-facial lateralization of the embryonic cells.

Microarray analysis comparing the dorsocervical to the abdominal subcutaneous adipose tissue within the cART+LD− group identified 27,900 probes for genes of which 18 were differentially expressed. Of these, 13 corresponded to unique Entrez Gene IDs. Of these 13 genes, 9 were downregulated and 4 upregulated in the dorsocervical compared with the abdominal subcutaneous adipose tissue (Supplementary Table 2; full microarray data are available at GEO, vide supra). The differentially expressed genes were involved in regulation of transcription, organ development, and regionalization. The top three downregulated genes were the same HOXA10, HOXC9, and HOXC8 genes as in the cART+LD+ group. Also alike in the cART+LD+ group, SHOX2 was among the genes most upregulated in the dorsocervical as compared with the abdominal adipose tissue.

Expression of SHOX2 was fourfold higher in the dorsocervical compared with the abdominal subcutaneous adipose tissue in the cART+LD+ and sixfold higher for the respective comparison in the cART+LD− group. Expression of HOXA10, HOXC9, and HOXC8 was 18-, 8-, and 11-fold lower in the dorsocervical compared with the abdominal subcutaneous adipose tissue in the cART+LD+ group. In the cART+LD− group, expression of HOXA10, HOXC9, and HOXC8 were 21-, 9-, and 10-fold lower in the dorsocervical than in the abdominal subcutaneous adipose tissue. Fold change differences for any of the homeobox genes analyzed were not significantly different between the cART+LD+ and the cART+LD− groups (Fig. 3).

UCP1 mRNA was detected in similar minute quantities in some dorsocervical and abdominal subcutaneous adipose tissue samples, but expression was approximately 1,000 times lower than that measured by the same laboratory in human BAT (data not shown) (18).

Expression of CD68 mRNA, used as a marker of macrophages and, thus, inflammation, was significantly lower in the dorsocervical as compared with the abdominal subcutaneous adipose tissue in the cART+LD+ group (0.39 [0.08–6.63] vs. 0.56 [0.19–6.48], P = 0.02). The same was true also in the cART+LD− group (P = 0.04). Expression of CD68 mRNA did not differ between the study groups in either adipose tissue depot (data not shown).

Herein we compared in detail the dorsocervical adipose tissue of cART-treated patients with and without lipodystrophy to the abdominal subcutaneous adipose tissue from the same individuals. Composition and characteristics of these two depots are of interest since in lipodystrophic patients one atrophies (abdominal depot), whereas the other is spared (dorsocervical depot).

We found that mtDNA copy numbers are significantly lower in the dorsocervical and the abdominal subcutaneous adipose tissue of the cART+LD+ group than in the respective depots in the cART+LD− group. Dorsocervical adipose tissue had significantly more mtDNA than the abdominal subcutaneous adipose tissue in the cART+LD+ group and almost significantly (P = 0.08) in the cART+LD− group. These results confirm previous reports on mtDNA depletion in lipoatrophic adipose tissue (14,20,30,31). However, the fact that mtDNA content is highly significantly lower (by 62%) also in the dorsocervical adipose tissue of cART+LD+ compared with the corresponding tissue in cART+LD− while the two depots are phenotypically comparable contradicts the prevailing theory that mtDNA depletion causes lipoatrophy. It also implies that once the lipodystrophy syndrome develops, the derangements observed in the adipose tissue are systemic and not restricted to atrophic abdominal subcutaneous adipose tissue.

Dorsocervical adipose tissue in the lipodystrophy syndrome has been suggested to represent BAT based on its anatomical location (14,32). Hallmarks of BAT include ample mitochondria featuring high metabolic activity and high expression of UCP1 (10–12). Previous studies reported significant amounts of UCP1 mRNA in the buffalo hump, but not in healthy or lipoatrophic abdominal subcutaneous adipose tissue (14,15). Protein levels of UCP1 were not quantified. In our dorsocervical adipose tissue samples, we did not find increased copy numbers of mtDNA compared with abdominal subcutaneous adipose tissue and found no evidence of biologically significant UCP1 expression using the same assay as previously used to verify the presence of UCP1 in BAT in human supraclavicular adipose tissue studied by PET (18). Moreover, we did not see typical BAT morphology in our dorsocervical tissue samples, even in those from the clinically most prominent buffalo humps. This is in line with a recent study in subjects with cART-associated lipodystrophy reporting their dorsocervical adipose tissue to lack BAT morphology and not to be activated by cold exposure like true BAT would (19). Thus the present data support the view that dorsocervical adipose tissue in cART-associated lipodystrophy is not BAT.

Lipoatrophic adipose tissue, as compared with nonlipodystrophic adipose tissue, is known to be characterized by smaller adipocytes, greater cell size variation, disruption of cell membranes, and signs of apoptosis as determined by immunohistochemistry staining compared with nonlipodystrophic adipose tissue (33–35). In our histological analyses, irrespective of the lipodystrophy status, we found less adipocytes and more connective tissue in the dorsocervical than in the abdominal subcutaneous adipose tissue. Thus these characteristics seem to be intrinsic to the dorsocervical location. Hitherto, light microscopy and ultrastructural characteristic data of adipose tissue from lipohypertrophic areas of HIV-1–infected subjects on cART have only been examined in one small study (n = 4) (35). Although the study did not specify the anatomical location of the lipohypertrophic areas, adipose tissue from these was reported to show similar, although milder, changes than those seen in lipoatrophic areas (35). These could not be quantitated because of the small sample size. The volume fraction of adipocytes and connective tissue was not reported but the finding of lipohypertrophic fat to be more normal is in line with the present data on lower CD68 expression and more preserved mtDNA copy number in the dorsocervical compared with the abdominal subcutaneous adipose tissue (vide infra).

We found expression of CD68 mRNA to be significantly lower in the dorsocervical than in the abdominal subcutaneous adipose tissue in both the cART+LD+ and the cART+LD− groups. Furthermore, microarray data suggested that genes involved in immune and inflammatory processes are downregulated in the dorsocervical as compared with the abdominal subcutaneous adipose tissue in the cART+LD+ group. These findings are in line with those of Guallar et al. (14), who found that inflammation, as judged from mRNA expression of TNF-α and CD68, is more severe in lipoatrophic abdominal subcutaneous adipose tissue than in the dorsocervical region, although in the latter study samples were not from the same subjects.

The microarray data suggested that dorsocervical and abdominal subcutaneous adipose tissue are largely similar. Using stringent criteria to avoid false-positive data, of all genes in the genome only less than 100 were differentially expressed. The genes differentially expressed in both the cART+LD+ and the cART+LD− groups are most likely to be those different because of the anatomic location of the tissues. The number of genes differentially expressed in the dorsocervical compared with the abdominal subcutaneous adipose tissue was markedly greater in the cART+LD+ than the cART+LD− group. Genes found to be discrepant between the dorsocervical and the abdominal subcutaneous depot only within the cART+LD+ group seem to represent characteristic features of lipoatrophy (e.g., upregulated inflammatory genes).

The foremost difference between dorsocervical and abdominal subcutaneous adipose tissue was in expression of the homeobox genes. The difference was observed in both the lipodystrophic and the nonlipodystrophic adipose tissue using microarray analysis and confirmed by real-time PCR. Homeobox genes are evolutionary highly conserved genes that serve as transcription factors guiding cells of the embryonal neural tube to migrate to their destined location and determining the distinctive traits inherit to the tissue being formed. There are four homeobox gene clusters in mammals: HOXA, HOXB, HOXC, and HOXD (36). Each cluster has several members numbered in ascending order starting from the anterior boundary of the central nervous system. Thus, the lower the order number of a homeobox gene, the more cranially it is located. Our findings are in line with this, since rostral SHOX2 was overexpressed in dorsocervical compared with abdominal subcutaneous adipose tissue. The reverse was true for the caudal homeobox genes HOXA10, HOXC9, and HOXC8. These data support the validity of the microarray data and are in line with previous reports of depot-specific expression of homeobox genes in human adipose tissue in vivo (37–40) and in human adipose tissue–derived preadipocytes (41) and stromal cells (42). These differences were hypothesized to reflect the differentiation stages and adipogenic potential of adipose tissue depots as also suggested by work in murine models and cell cultures (43). The above human studies, however, evaluated expression of homeobox genes in adipose tissue samples from different locations along the ventrodorsal (subcutaneous–visceral–retroperitoneal) rather than the craniocaudal (dorsocervical–abdominal subcutaneous) axis that was examined in the current study. Of interest, HOXC4, which was expressed exclusively in BAT (37), was not differentially expressed between the two adipose tissue depots in the present microarray analysis. This fortifies the notion that dorsocervical adipose tissue in cART-associated lipodystrophy is not BAT.

We conclude that mtDNA is highly significantly lower (by 62%) even in the nonlipoatrophic dorsocervical adipose tissue of the cART+LD+ patients when compared with the corresponding depot of the cART+LD−. Therefore, mtDNA depletion does not seem to lead to lipoatrophy. In lipodystrophic patients, the dorsocervical adipose tissue is less inflamed than the lipoatrophic abdominal subcutaneous adipose tissue. It does not resemble BAT. The most marked difference in gene expression between dorsocervical and abdominal subcutaneous adipose tissue, irrespective of the lipodystrophy status, lay in expression of homeobox genes involved in organogenesis and regionalization. Disparate expression of such fundamental regulators of transcription might ultimately contribute to different patterns of differentiation and affect the susceptibility of the adipose tissue depot to cART-induced toxicity, perhaps making the abdominal subcutaneous depot more vulnerable to atrophy.

This study was supported by grants from EVO and Sigrid Juselius Foundations (to H.Y.-J.), Orion Research Foundation (to K.S.), Lilly Foundation (to K.S.), Biomedicum Foundation (to K.S.), Finnish Medical Foundation (to K.S.), Jalmari and Rauha Ahokas Foundation, and Alfred Kordelin Foundation (to K.S.). No other potential conflicts of interest relevant to this article were reported.

K.S. and J.S. researched data, contributed to discussion, and wrote the manuscript. D.G., M.S., K.S., J.P., V.M.O., D.W., M.E.L., S.E., P.E., U.A.W., P.A., and M.R. researched data, contributed to discussion, and reviewed and edited the final version of the manuscript. H.Y.-J. researched data, contributed to discussion, and wrote the manuscript.

The authors thank Katja Sohlo and Mia Urjansson (Helsinki University Central Hospital, Department of Medicine) as well as Pentti Pölönen (Helsinki University Central Hospital, Department of Radiology) for skilful technical assistance and thank Antti Hakkarainen (Helsinki University Central Hospital, Department of Radiology) for useful discussions.