Mitochondrial toxicity hypothesis for lipoatrophy: a refutation

Only through understanding the aetiology of the metabolic and morphological changes in individuals receiving highly active antiretroviral therapy (HAART) can the optimal management of these problems be established. Evidence from cross-sectional surveys, such as the HOPS cohort ^[1] have pointed to an interaction between disease or immune recovery and drugs. Although both nucleoside and protease inhibitor (PI) use were associated with the presence of lipodystrophy, additional risk factors were necessary for lipodystrophy to be present ^[1]. Patients with the syndrome who have never received PI or nucleoside analogues have been reported ^[2-6], suggesting that although these drug classes or specific members of these classes may influence the time to the onset of clinical manifestations, they are not sufficient alone to cause problems. It is therefore not surprising that switching therapy away from whichever drug or drug class it is the current fashion to blame has not led to the resolution of the syndrome ^[7]. In considering information from surveys or 'cross-sectional studies' it is reasonable to consider that associations found in such analyses do not necessarily point to the causation of the problem. For example, age over 40 years has been associated with an increased risk of lipodystrophy ^[1] but clearly being over 40 is not the cause of lipodystrophy.

Mitochondrial toxicity of nucleoside analogues through the inhibition of DNA-polymerase gamma as been suggested to be a key factor in the pathogenesis of lipoatrophy and visceral fat accumulation ^[8,9]. However, there is little evidence to support mitochondrial toxicity as the mechanism of lipoatrophy, and indeed considerable data has accumulated to refute this hypothesis.

Two fat biopsy studies recently reported ^[10,11] demonstrated only modest reductions (mean 44%^[10]) in mitochondrial DNA, even in samples from lipoatrophic patients. In inherited mtDNA disorders, reductions in mtDNA to less than 20% are generally required for disease to occur ^[12,13]. Importantly, samples from individuals with lipoatrophy in some cases (13% of samples in one study ^[11]) had normal levels of mtDNA and some control samples (from HIV-negative individuals) showed diminished mtDNA ^[10,11]. This underlines that reductions in mtDNA are not necessary for, characteristic of, or diagnostic of lipoatrophy in individuals with HIV. In addition, the mtDNA mutation associated with inherited Madelung's syndrome was absent in all samples in one of the studies ^[10].

Furthermore, two separate groups ^[14,15] demonstrated that fat oxidation, a mitochondrial function, is normal or potentially increased in persons with metabolic disturbances on PI plus nucleoside reverse transcriptase inhibitor (NRTI)-based regimens. An additional study of muscle biopsies pre- and post-exercise and lactate levels found normal oxidative phosphorylation, with a similar recovery rate in lactate and pyruvate levels after exercise to healthy controls, and no significant abnormalities of muscle mitochondria ultrastructure. The authors of the study ^[16] suggested that the hyperlactatemia seen in some patients, often suggested to be a marker of mitochondrial toxicity, may in fact be related to a decreased clearance of lactate. Many of the patients studied, all of whom clinically had lipodystrophy, had hypertriglyceridaemia. One potential explanation for these findings is that hypertriglyceridaemia (or type 2 diabetes) can lead to hepatic steatosis ^[17], hence diminished hepatic function and impaired elimination of lactate. The association of lipoatrophy and hyperlactatemia observed in one cohort analysis ^[18] may be explained by this mechanism, and enables an understanding of why isolated mild hyperlactatemia may be relatively common but lactic acidosis remains rare. Finding hepatic steatosis on a liver ultrasound or biopsy may simply reflect metabolic disturbances in individuals with HIV infection, not mitochondrial injury by nucleoside analogues.

The results of culturing adipocytes with nucleoside analogues or PI yielded different effects on triglyceride accumulation, fat release (lipolysis) and adenosine triphosphate (cellular energy) levels. Although PI had effects at physiological concentrations on triglyceride accumulation and at higher concentrations on lipolysis and adenosine triphosphate production, effects were not observed at concentrations normally seen in treated persons for either zidovudine or stavudine (< 30 μM) ^[19]. Interestingly, when PI and NRTI were given together to the adipocyte culture, a marked synergy was observed on a range of adipocyte functions ^[19]. This may be an important observation, suggesting that the combination of PI and NRTI may have a greater impact on fat cell function than either alone. This is in keeping with clinical observations ^[20]. In a further study of stavudine in mice given very high doses of 100 mg/kg per day of the drug (human dose ∼1.0 mg/kg per day) over 6 weeks ^[21], no effects on skeletal muscle or adipose tissue mtDNA were observed. NRTI alone at current doses thus do not appear to affect adipocyte mitochondrial function. Changes in liver mtDNA were observed only in week 1 samples but not at week 6, suggesting a compensatory recovery in mtDNA ^[21]. This may also explain why mitochondrial toxicity is reported with nucleoside analogues in short-term in-vitro studies ^[22]. Nucleoside mitochondrial toxicity may, to some extent, be only an acute effect, and could be compensated for by the mechanisms that normally regulate mitochondrial numbers in the cell such as an increased expression of mitochondrial transcription factors ^[23]. The findings of such studies are perhaps not surprising, for the thymidine analogues zidovudine and stavudine as adipocytes, and other resting cells, are unlikely to express thymidine kinase type 1, the enzyme involved in the first step of activation of these drugs, and these drugs are poor substrates for the mitochondrially located thymidine kinase type 2 ^[24,25]. Adipocytes are thus unlikely to have active (i.e. potentially toxic) concentrations of phosphorylated thymidine analogues ^[26].

The end of the era of blaming PI exclusively for lipodystrophy came with the recognition that lipoatrophy and metabolic disturbances could occur in PI-naive individuals. However, NRTI-naive but antiretroviral-treated individuals are relatively rare, making and evaluation of NRTI sparing more difficult. A survey from the DP-006 study ^[2] indicated that nine out of 429 (2.1%) patients in the efavirenz-indinavir arm of the study had been diagnosed with 'probable lipodystrophy' despite having never received nucleoside analogues over a maximal follow-up of 88 weeks. Two further recent reports ^[3,4], evaluating patients on dual PI regimens (both ritonavir-saquinavir) observed lipodystrophy in NRTI-spared individuals but higher rates when NRTI were added to the regimen. In the Prometheus study ^[3], after 96 weeks follow-up, 22 out of 88 (25%) of stavudine/ritonavir/saquinavir patients but also seven out of 87 (8%) ritonavir/saquinavir alone treated patients were diagnosed with lipodystrophy, including two out of 44 (5%) ritonavir/saquinavir treated patients who had never received NRTI. A second study with 144 weeks of follow-up ^[4], mostly in persons who discontinued previoius NRTI therapy and then received ritonavir/saquinavir found that 6% of NRTI-spared individuals had both buttock and facial wasting, with 9% having an increase in waist size. Patients who added NRTI during the study were more likely to have lipoatrophy ^[4]. These studies provide several important observations. Most importantly that lipodystrophy/lipoatrophy can occur in the absence of NRTI. A previous study ^[20] suggested that use of PI can accelerate the onset of or increase the incidence of lipoatrophy in NRTI-treated individuals. These data suggest that this interaction works both ways, that NRTI also accelerate the onset or increase the incidence of lipoatrophy in PI-treated individuals.

Available data suggest that we should focus our efforts on understanding lipoatrophy during HAART away from drugs and onto drug-disease and immune recovery interactions ^[1,27]. The recent observation that the metabolic syndrome may occur in individuals several years after bone marrow transplant ^[28], another group of individuals undergoing immune restoration, suggests that this line of investigation may have potential benefits beyond HIV medicine.

Graeme Moyle