Differential gene expression indicates that 'buffalo hump' is a distinct adipose tissue disturbance in HIV-1-associated lipodystrophy

Objective: To elucidate the molecular basis of the progressive enlargement of dorso-cervical adipose tissue, the so-called 'buffalo hump', that appears in a sub-set of patients with HIV-1/HAART-associated lipodystrophy.

Design: Analysis of the expression of marker genes of mitochondrial function, adipogenesis, inflammation and cell proliferation in ten 'buffalo hump' samples and ten subcutaneous fat samples from HIV-1-infected/HAART-treated patients, and in ten healthy controls.

Methods: Quantitative real-time polymerase chain reaction analysis of mitochondrial DNA and gene transcripts, and immunoblot for specific proteins.

Results: 'Buffalo hump' patients had lower levels of mitochondrial DNA and mitochondrial DNA-encoded transcripts with respect to healthy controls. The uncoupling protein (UCP)-1 gene was expressed only in 'buffalo hump' fat. There were no significant changes in the expression of UCP2, UCP3 or of marker genes of adipogenesis in 'buffalo hump' patients relative to healthy controls. 'Buffalo hump' fat did not show the high expression of tumor necrosis factor-α and β2-microglobulin identified in lipoatrophic subcutaneous fat from patients. The expression of the macrophage marker CD68 was also lower in 'buffalo hump' than in subcutaneous fat from patients. In contrast, 'buffalo hump' showed a higher expression of the cell proliferation marker PCNA.

Conclusions: 'Buffalo hump' adipose tissue shows specific disturbances in gene expression with respect to subcutaneous fat from HIV-1-infected/HAART-treated patients. Mitochondrial alterations cannot explain the differential behavior of 'buffalo hump' with respect to adipose depots prone to lipoatrophy. The absence of a local inflammatory status in 'buffalo hump' may explain in part the differential behavior of this adipose tissue.

Introduction

Disturbances in adipose tissue in HIV-1-infected patients under HAART involve a complex set of alterations. A large proportion of patients exhibiting the HAART-associated lipodystrophy syndrome show lipoatrophy in subcutaneous adipose tissue. Lipoatrophy occurs often in the face, arms, and legs. An enlargement in visceral adipose tissue, reminiscent of visceral obesity, also occurs with remarkable frequency, often in combination with peripheral lipoatrophy ^[1,2]. Lipomatosis is also commonly found in HIV-1-infected patients under HAART. It appears usually as an enlargement in the dorso-cervical area ('buffalo humps') ^[3,4], although other anatomical distributions (for example pubis, neck or trunk) have been also noted ^[5,6]. The frequency of 'buffalo hump' in HIV-1-infected patients has been reported to range from 2 to 13%, with a higher prevalence (6 to 13%) in those showing any other feature of the lipodystrophy syndrome ^[7]. The presence of lipomatosis often requires surgical removal due to discomfort, for aesthetic reasons or even because of localized pain. The prognosis after surgery is variable and recurrences are not unusual ^[8].

HAART-associated 'buffalo hump' is unlikely to be simply an additional variation of other enlargements of adipose tissue in lipodystrophy, as those occurring in the visceral area. Enlargement of adipose tissue in the dorso-cervical region or other sites showing lipomatosis involves subcutaneous fat, a type of adipose tissue that, in most of the patients showing lipodystrophy, is characterized by lipoatrophy. In some cases, patients display enlargement in localized subcutanous sites ('buffalo hump') but lipoatrophy in other, even surrounding, regions. Visceral fat accumulation is likely to be a distinct phenomenon, as it involves the intra-abdominal depot of adipose tissue. A recent report has shown that whereas increased visceral adipose tissue gives rise to a phenotype reminiscent of the metabolic syndrome, dorso-cervical fat accumulation is associated only with hyperinsulinemia ^[7]. Moreover, fat enlargements leading to 'buffalo hump' may result from alterations of the adipocyte cell at those sites. Adipocytes in 'buffalo hump' express the brown fat uncoupling protein-1 (UCP1) gene, thus suggesting disturbances in the brown-versus-white adipocyte differentiation pattern ^[9].

The etiopathogenesis of lipomatosis in the context of adipose tissue disturbances in HAART-treated, HIV-1-infected patients, is unknown. Some patients with inherited diseases involving mutations in the tRNA-Lys gene of mitochondrial DNA also show an enlarged, UCP1-expressing, region of adipose tissue in the dorso-cervical area. This led to the hypothesis of the involvement of mitochondrial disturbances in the origin of 'buffalo hump' in HIV-1-infected patients ^[10-12]. It may be hypothesized that 'buffalo hump' develops as a site-specific compensatory response of adipose tissue to metabolic dysregulation, although an alternative hypothesis could be that 'buffalo hump' is caused by enhanced local toxicity elicited by HIV-1 infection and/or HAART in this particular area. Evidence to support this latter model might include evidence of enhanced tissue pathology such as inflammation or mitochondrial toxicity. Understanding the molecular mechanisms of this dramatic modification in the behavior of the adipose depots is of interest in order to prevent and/or treat lipomatosis and to understand the overall etiopathogenesis of adipose tissue modifications associated with HIV-1/HAART. Moreover, it may provide information on the basic aspects of what determines the proliferative capacity of adipose cells in humans. For this purpose, we analyzed the profile of gene expression in 'buffalo hump' from HIV-1-infected patients on HAART, in comparison with that in subcutaneous fat from HIV-1-infected, HAART-treated patients showing lipoatrophy but not lipomatosis, and that from healthy non-HIV-1 infected, nontreated individuals.

Materials and methods

Gene expression in enlarged dorso-cervical adipose tissue ('buffalo hump') from ten HIV-1-positive patients on HAART was studied upon surgical removal. These patients displayed, along with 'buffalo hump', the clinical characteristics consistent with HAART-associated peripheral lipoatrophy (fat wasting from the face, buttocks and limbs). They were compared with ten HIV-1-positive patients also showing HAART-associated peripheral lipoatrophy, but without lipomatosis. Patients in both groups were similar with respect to mean age, male/female ratio, serum CD4+ cells/μl, serum cholesterol, triglycerides and virological control of HIV-1 infection (Table 1). Between the two patient groups, differences in the number of cumulative months of treatment with nucleoside analog reverse transcriptase inhibitors (NRTI), with nonnucleoside analog reverse transcriptase inhibitors (NNRTI) or with protease inhibitors (PI) were not statistically significant. Differences for cumulative treatment with the predominant NRTI (zidovudine, stavudine, lamivudine, didanosine, abacavir), NNRTI (efavirenz, nevirapine) and PI (indinavir, ritonavir/lopinavir, nelfinavir) drugs used throughout the antiretroviral treatment history of the patient groups were also not statistically significant (Table 1). For comparison, a third group was established with sex and age-matched non-HIV-1-infected, nontreated individuals. Biopsy samples of subcutaneous fat from patients showing lipodystrophy, but without 'buffalo hump', and from controls were taken from the abdominal area, whereas samples of 'buffalo hump' were taken from biopsies obtained at the time of surgical removal. All procedures were performed with informed consent from participants and the study was approved by the Hospital de la Santa Creu i Sant Pau ethics committee. Tissue samples were frozen in liquid nitrogen. After homogenization in RLT (Qiagen, Hilden, Germany) buffer, DNA was isolated using a phenol/chloroform extraction methodology and RNA was isolated using a column-affinity based methodology, including on-column DNA digestion (RNeasy; Qiagen). One microgram of RNA was transcribed into cDNA using MultiScribe reverse transcriptase and random-hexamer primers (TaqMan Reverse Transcription Reagents; Applied Biosystems, Foster City, California, USA). For quantitative mRNA expression analysis, TaqMan reverse transcriptase (RT)-polymerase chain reaction (PCR) was performed on the ABI PRISM 7700HT sequence detection system (Applied Biosystems). The TaqMan RT-PCR reaction was performed in a final volume of 25 μl using TaqMan Universal PCR Master Mix, No AmpErase UNG reagent and the following specific primer pair probes: 18S rRNA, Hs99999901; UCP1, Hs00222453; UCP2, Hs00163349; UCP3, Hs00243297; COX4I1, Hs00266371; PPARGC1, Hs00173304; CEBPA, Hs00269972; LPL, Hs00173425; TNF, Hs00174128; PPARG, Hs00234592; APM1(adiponectin), Hs00605917; Pref-1, Hs00171584; b2microglobulin, Hs99999907; CD68, Hs00154355.; COL1A2, Hs00164099 (TaqManGene Expression Assays; Applied Biosystems). Primers and probe for the detection of cytochrome c oxidase subunit II (COII) mRNA and mtDNA abundance assessment were designed (Custom TaqMan Gene Expression Assays; Applied Biosystems) and the sequences were: CAAACCACTTTCACCGCTACAC (forward) and GGACGATGGGCATGAAACTGT (reverse) and the FAM-labeled probe was AAATCTGTGGAGCAAACC. Quantification of mtDNA was referred to nuclear DNA as determined by the amplification of the intronless gene CEBPα. Controls with no RNA, primers, or RT were included in each set of experiments. Each sample was run in duplicate, and the mean value of the duplicate was used to calculate the mRNA expression of the genes of interest which were normalized to that of the reference control (18S ribosomal RNA) ^[13] using the comparative (2^-ΔCT) method, following the manufacturer's instructions. Parallel calculations using the PPIA reference gene (Hs99999904) were performed and results were essentially the same.

Table 1

For immunoblot quantification of protein expression, adipose tissue samples were homogenized in cold buffer (10 mmol/l HEPES pH 7.5, 5 mmol/l ethylenediamine tetra-acetic acid, 5 mmol/l dithiothreitol, 5 mmol/l MgCl₂₎, and protease inhibitors (Complete-mini; Roche, Sant Cugat del Vallès, Spain). 40 μg of homogenate protein were mixed with equal volumes of 2 × sodium dodecyl sulphate (SDS) loading buffer, incubated at 90°C for 5 min, and electrophoresed on SDS/polyacrylamide gels. Proteins were transferred to Immobilon-P membranes (Millipore, Billerica, Massachusetts, USA) and immunological detection was performed using antibodies against PPARγ, SREBP1 and PCNA (H-100, K-10 and sc-25280, respectively, Santa Cruz Biotechnology, Santa Cruz, California, USA), β2-microglobulin (P0163; Dako Cytomation, Glostrup, Denmark), VDAC (31HL-529536; Calbiochem, San Diego, California, USA), COII (12C4; Molecular Probes, Eugene, Oregon, USA) or β-actin (Sigma A5441; Sigma, St Louis, Missouri, USA). Immunoreactive complexes were detected using an antirabbit secondary antibody (sc-2004; Santa Cruz Biotechnology) and the enhanced chemiluminiscence detection system (Amersham Biosciences, Little Chalfornt, Buckinghamshire, UK). The intensity of the signals was quantified by densitometry (Phoretics 1D Softward; Phoretic International Ltd, Newcastle, UK).

Results were expressed as means ± SEM and statistical comparison for differences were performed using nonparametric analysis of variance. Statistical significance of differences was established when P values were < 0.05.

Results

Adipose tissue from 'buffalo hump' showed a significant depletion in mitochondrial DNA (mtDNA) with respect to control adipose tissue (Fig. 1), that was even more pronounced than in nonlipomatous subcutaneous fat from patients. The levels of COII mRNA, a mtDNA-encoded transcript, paralleled these findings, thus indicating an impairment in the synthesis of mtDNA-encoded transcripts as a consequence of mtDNA depletion. The expression of the mRNA for COIV, a nuclear-encoded subunit of the respiratory chain complexes, was similarly lower in 'buffalo hump' adipose tissue and in subcutaneous adipose tissue from patients.

Fig. 1

In accordance with a previous report ^[9], 'buffalo hump' adipose tissue showed substantial levels of expression of UCP1 mRNA, which was not detected in subcutaneous fat from healthy controls or from patients showing only lipoatrophy. UCP2 mRNA and UCP3 mRNA levels were not significantly altered in 'buffalo hump' fat relative to controls (Fig. 1), despite the fact that these transcripts levels were significantly lower and higher, respectively, in subcutaneous fat from patients, as already reported ^[14].

PPARγ mRNA levels in subcutaneous adipose tissue, but not in 'buffalo hump', from patients were substantially lower than those in healthy controls. A similar tendency was observed for C/EBPα mRNA expression although the changes were not statistically significant (Fig. 2).

Fig. 2

The expression levels of both PGC-1α mRNA and Pref-1 mRNA, which were significantly higher in subcutaneous fat from patients showing lipoatrophy, were also not significantly different in 'buffalo hump' adipose tissue relative to those in controls. Similarly, lipoprotein lipase mRNA and adiponectin mRNA levels, which were lower in subcutaneous adipose tissue from lipoatrophic patients, were not significantly different in 'buffalo hump' fat.

TNFα mRNA levels were not significantly higher in 'buffalo hump' relative to control adipose tissue (Fig. 3). This is in contrast with those in subcutaneous fat from patients, which did show significantly higher levels, as previously reported ^[14-16]. The mRNA expression of β2-microglobulin, a target gene of inflammatory cytokines ^[17,18], was also not significantly different in 'buffalo hump' with respect to control adipose tissue (Fig. 3). As a relevant component of the local inflammation status of subcutaneous fat in patients showing lipodystrophy has been attributed to enhanced macrophage infiltration and activation in adipose tissue, the transcript levels of CD68, a marker of macrophages, were determined. Subcutaneous adipose tissue, but not 'buffalo hump' fat, in patients showed moderately higher levels of CD68 mRNA with respect to control adipose tissue. In fact, CD68 mRNA levels were significantly lower in 'buffalo hump' than in subcutaneous fat from patients. Finally, considering reports of enhanced fibrosis in 'buffalo hump' ^[8,19], we analyzed the mRNA levels of the pro-alpha2(I) collagen (COL1A2), a marker of fibrosis ^[20]. Differences between groups were not statistically significant.

Fig. 3

To further analyze the changes in gene expression in 'buffalo hump', we studied the protein levels of representative marker genes of mitochondrial function, either mtDNA-encoded (COII) or nuclear-encoded and unrelated to the respiratory chain [voltage-dependent anion carrier (VDAC)], of adipogenesis (PPARγ) and of inflammation status (β2-microglobulin). Moreover, we also analyzed the protein levels of SREBP1, an adipogenic transcription factor which is post-transcriptionally regulated and is profoundly altered in subcutaneous adipose tissue from HIV-1-infected, HAART-treated, patients showing lipoatrophy ^[15]. Protein levels of 'proliferating cell nuclear antigen' (PCNA), a marker gene of cell proliferation, were also determined.

The results indicated that COII protein abundance, relative to overall mitochondrial protein, was lower in subcutaneous adipose tissue from patients. Even lower levels of COII protein were found in 'buffalo hump' fat, in parallel with data above on mtDNA and COII mRNA levels (Fig. 4). In contrast, the analysis of overall mitochondrial protein levels in patients, as estimated through VDAC protein abundance, indicated higher levels in subcutaneous fat and also, and more intensely, in 'buffalo hump'.

Fig. 4

Changes in PPARγ protein levels paralleled those observed for the corresponding transcripts: lower levels in subcutaneous fat from patients with respect to healthy controls and no significant changes in 'buffalo hump' fat (Fig. 4). For β2-microglobulin, in 'buffalo hump' not only were there not the high levels of expression observed in subcutaneous fat from patients, but levels were even lower than in healthy controls. The expression levels of SREBP1 were dramatically higher in subcutaneous adipose tissue from patients relative to those in controls; however, no such alteration was observed in 'buffalo hump'. Concerning PCNA, 'buffalo hump' adipose tissue showed higher levels of expression with respect to subcutaneous fat from patients or from healthy controls, although the abundance of PCNA protein was remarkably variable in 'buffalo hump' samples from different patients.

Discussion

Whereas several studies have addressed the molecular basis of subcutaneous adipose tissue lipoatrophy through the analysis of gene expression in adipose tissue ^{[14,15,21,22]} no such study was available regarding 'buffalo hump' fat and only a single previous report indicated the expression of UCP1, the marker gene of brown fat ^[9]. The present findings confirm UCP1 gene expression as a disturbance that is highly specific to 'buffalo hump' adipose tissue, not shared by lipoatrophic subcutaneous adipose tissue from patients equally infected by HIV-1 and being under HAART. This establishes that alterations in the white-versus-brown differentiation pattern of adipocytes is a characteristic feature only of 'buffalo hump' and not a common feature of adipose tissue disturbances associated with HAART. Stavudine and nevirapine have been reported to induce UCP1 gene expression and brown adipocyte differentiation in cell culture ^[23]. The limited number of samples and the cross-sectional nature of the present study prevent us from establishing unequivocally the potential role of drug treatment in the differences between UCP1 gene expression in 'buffalo hump' and in subcutaneous fat. A previous report, however, indicated that the appearance of 'buffalo hump' could not be associated with any specific component of HAART regimes ^[7]. Another feature of 'buffalo hump' fat, enhanced mitochondrial biogenesis, is also consistent with this acquisition of a brown fat-like phenotype in this adipose depot. The expression of both PGC-1α and UCP3 is not, however, significantly different with respect to controls, in contrast to the high expression observed in lipoatrophic areas. These two genes are preferentially expressed in brown rather than white, adipose tissue ^[24,25], thus indicating that 'buffalo hump' acquires some features of brown adipose tissue but not a full-blown brown fat phenotype. This is in agreement with reports of a morphology intermediate between typical brown and white in adipocytes from 'buffalo hump' ^[26]. On the other hand, due to the high sensitivity of the UCP3 gene to nonesterified fatty acids ^[27], the high UCP3 mRNA expression in subcutaneous fat of patients with lipoatrophy has been attributed to enhanced lipolysis ^[14], a phenomenon unlikely to occur in 'buffalo hump', a fat-accumulating tissue.

The similar low levels of mtDNA and mtDNA-encoded components of the respiratory chain in 'buffalo hump' and subcutaneous fat from patients is consistent with mitochondrial toxicity being common to both tissues. This indicates that mtDNA depletion is unlikely to be a key event determining whether a given adipose depot will undergo atrophy or hypertrophy in patients under HAART. This is also consistent with the lack of significant differences in the patterns of antiretroviral treatment in patients showing 'buffalo hump' or only lipoatrophy. Recent studies in noninfected patients indicated that it is antiretroviral treatment that causes impaired mitochondrial gene expression ^[28]. Moreover, 'buffalo hump' and subcutaneous fat from patients both show high levels of overall mitochondrial protein. This phenomenon has already been observed in lipoatrophic adipose tissue from patients under HAART regimes containing nucleoside analog transcriptase inhibitors ^[29] and has been interpreted as an attempted adaptive reaction to the altered mitochondrial function due to the low mtDNA levels in adipose tissue.

Concerning adipogenic regulators, PPARγ expression is not significantly lower in 'buffalo hump', in contrast with subcutaneous fat from patients. Conversely, the high expression of the antiadipogenic factor Pref-1 in subcutaneous fat does not occur in 'buffalo hump'. This is consistent with lack of lipoatrophy and progressive increase in adipose mass in the 'buffalo hump' depot. The expression of genes associated with fat replenishment, such as lipoprotein lipase or adiponectin, which are targets of PPARγ ^[30,31], is consistently higher relative to expression levels in subcutaneous fat. On the other hand, a high expression of SREBP1 protein in adipose tissue has been reported in subcutaneous fat from HIV-1-infected, HAART-treated patients, showing peripheral lipoatrophy ^[15] and this has been attributed to the accumulation of abnormally processed SREBP1 protein. This phenomenon is also evident in subcutaneous fat from patients in the present study but, similarly to the other adipogenic regulators, it does not occur in 'buffalo hump'.

A local inflammatory environment appears to be a prominent feature of subcutaneous fat in HIV-1-infected, HAART-treated patients, as evidenced by the high expression of TNFα and other pro-inflammatory cytokines ^[14-16]. Many of the gene expression disturbances in subcutaneous fat from lipoatrophic areas, such as low levels of PPARγ expression and the subsequent low expression of markers of adipogenesis can be explained by the repression exerted by the high levels of TNFα. The unaltered expression of TNFα and of β2-microglobulin, a direct target of pro-inflammatory cytokines, in 'buffalo hump' is consistent with unaltered levels of adipogenic transcription factors and their targets. Thus, although the mechanisms by which HIV-1 infection plus HAART does not led to an increased inflammatory local environment in the anatomical site corresponding to 'buffalo hump' are unknown, the present observation provides an explanation for the lack of impairment in adipogenesis and why there is no lipoatrophic behavior in the 'buffalo hump' fat pad. Microscopic analysis of adipose tissue from lipoatrophic sites in patients indicate increased infiltration of macrophages which can contribute to the local inflammation ^[32-34]. Such induction could be observed to some extent in the subcutaneous fat from patients through the analysis of gene expression of the macrophage marker CD68 ^[35]. Remarkably, CD68 mRNA expression was significantly lower in 'buffalo hump' thus indicating that macrophage infiltration is not enhanced.

The high expression of PCNA supports the notion of a transformed phenotype in adipose cells in 'buffalo hump' indicating an intrinsic enhancement in cell proliferation. Auto-transplantation of adipose tissue from 'buffalo humps' to facial lipoatrophic areas in reconstructive surgery may result in an enlargement of adipose tissue cheeks, a former lipoatrophic area ^[36]. This indicates that cells in the 'buffalo hump' may have acquired a high proliferative capacity that remains even when they are placed in a lipoatrophic environment. PCNA expression was remarkably variable among different 'buffalo hump' samples, however, thus suggesting a distinct extent of acquisition of the proliferative status. This may explain why adipose tissue cheek enlargement after auto-transplantation occurs only in a sub-set of patients. Finally, we did not find statistically significant differences in the expression of the COL1A2 gene, a marker of enhanced fibrosis. This indicates that such phenomenon, if present in the 'buffalo hump' samples analyzed here, does not result in a massive change in gene expression.

Several limitations affect the extent of the conclusions that can be drawn from the present study. The low number of samples limits the capacity to unequivocally clarify issues such as the role of gender, treatment drug patterns or the actual significance of gene expression data when differences between groups lack statistical significance. Moreover, the lack of availability of biopsies from 'buffalo hump' and subcutaneous adipose tissue from the same individuals also precludes a proper assessment of the impact of individual variability on gene expression patterns in adipose tissue.

Despite these limitations, the present results constitute the first systematic analysis of gene expression in 'buffalo hump' adipose tissue from HIV-1-infected patients on HAART and reveal a markedly differential pattern of alterations in gene expression with respect to subcutaneous fat of patients showing lipoatrophy. Whereas low mtDNA abundance and expression is common to the two adipose tissue types, 'buffalo hump' adipose tissue shows unaltered expression of marker genes of adipogenesis, and altered gene expression indicating enhanced cell proliferation and features of a brown-versus-white fat phenotype. The lack of signs of local inflammation in 'buffalo hump' may suggest that enhanced expression of pro-inflammatory cytokines is an important determinant of the specific lipoatrophic behavior of several subcutaneous adipose tissue areas and is consistent with the hypothesis that 'buffalo hump' may represent a dysregulation of adipose tissue in this area, rather than being a reflection of enhanced toxicity. It may also be the case, however, that enhanced inflammation is a response to adipocyte death in lipoatrophic areas and not a cause of the pathology. As with all cross-sectional analyses, this study was not designed to determine cause-and-effect relationships, and unmeasured bias inherent to this type of study may influence the observed changes. Therefore, further studies would be required to establish unequivocally whether a lack of inflammatory signals is a cause or consequence of the lipoatrophic versus proliferative status of adipose depots in patients. It has, however, been reported that adipocytes from 'buffalo hump' release much less TNFα than those from abdominal subcutaneous adipose tissue from HIV-1-infected, HAART-treated, patients when incubated 'in vitro' ^[37].

In summary, 'buffalo hump' is a highly specific alteration in the context of adipose disturbances in HIV-1-infected, HAART-treated patients, and distinct molecular events are associated with their induction: acquisition of brown-fat-like features, induction of proliferation but unaltered inflammatory status and adipogenesis. Furthermore, present results suggest that treatment of HAART-associated lipodystrophy in HIV-1-infected patients that is focused on promoting adipose cell differentiation or proliferation may carry a risk of enhancing lipomatosis.

Acknowledgements

Sponsorship: This study was supported by the Ministerio de Educación y Ciencia (SAF2005-01722), Fundación para la Investigación de la SIDA en España (FIPSE) (36752/06, 36610/06) and Instituto de Salud Carlos III, Ministerio de Sanidad y Consumo (FIS-PI052336) and Red de Investigación en SIDA (RD06/006/0022), Spain.