Despite a number of emerging theories, the mechanisms that cause the body fat and metabolic changes, and the relationship between the physical and metabolic changes, have not been identified.

Lipodystrophy unrelated to HIV

Rare forms of lipodystrophy are inheritable and lipodystrophy is associated with insulin resistance in some of these syndromes (Garg). Many of the defects proposed as mechanisms in HIV-associated lipodystrophy have already been investigated in cases of inherited and/or genetic lipodystrophy to establish whether particular genetic mutations are associated with the syndromes. In the case of familial partial lipodystrophy (Dunnigans type), the gene associated with the syndrome is located very close to CRABP-2, at chromosome IQ21. CRABP-2 is the lipoprotein receptor-related protein which bears a close resemblance to HIV protease.

A form of lipodystrophy called Madelung's disease has been related to a mitochondrial defect (see The role of nucleoside analogues later in this section).

Acquired forms of lipodystrophy which resemble those reported in HIV have also been reported in uninfected individuals, and may be associated with an auto-immune response. The syndrome is 2.6 times more common in women than in men, conforming to the prevalence of auto-immune diseases in the population. In 90% of cases auto-antibody to C3 nephritic factor has been detected. C3neF breaks down adipocytes in test-tube studies. Anything which interferes with the breakdown of fat cells is likely to interfere with peripheral and central fat storage.

A lipodystrophy syndrome has been reported in individuals who received bone marrow transplants and immune suppressive treatment during childhood. The frequency of insulin resistance increased in a cohort of Finnish individuals with the time since bone-marrow transplantation (Taskinen).

Some people may have a genetic predisposition to elevated lipid levels. Apolipoprotein E2 is strongly associated with high triglycerides, as well as LDL-B phenotype, (which is also associated with obesity or a sedentary lifestyle, and with Syndrome X, in HIV-negative people). LDL-B is predictive of atherosclerosis (hardening of the arteries), and can be reduced by diet and exercise in HIV-negative people.

Metabolic problems prior to protease inhibitors

A number of researchers have demonstrated that the lipid changes now being seen in people on HAART were already present to some extent prior to the introduction of effective treatment (Grunfeld 1989). For example, high levels of interferon alfa and triglycerides were reported in individuals with AIDS before the introduction of antiretroviral therapy, and significantly slower rates of triglyceride clearance were also evident. Levels of VLDL cholesterol were also elevated in AIDS, and declining HDL cholesterol levels were reported independent of therapy. This pattern is a classic response to infection. Reductions in insulin sensitivity were also noted, but this pattern appears to be unique to HIV infection.

A case of progressive lipodystrophy in the absence of antiretroviral therapy, commencing in 1993, has also been reported, in a woman with persistently undetectable viral load and a high CD4 cell count. The syndrome developed after two years of AZT monotherapy, which was stopped in 1994. The woman developed high triglyceride levels and low HDL cholesterol levels during the period of AZT treatment, in the absence of any other disease which might cause lipid alterations, and in the absence of high alcohol consumption (Griblin 2001).

How might antiretroviral therapy cause lipodystrophy?

Antiretrovirals have been shown to affect several key stages regulating the development fat cells.

Fat cells (adipocytes) develop from precursors called adipoblasts, which differentiate, or take on the full characteristics of adipocytes, by exposure to a sequence of agents called transcription factors. If one agent in the cascade is withheld, this can affect the activity of the rest of the chain, leading to faulty fat cells.

In the case of HIV-associated lipodystrophy, defects have been noted in the nuclei of fat cells which indicate that a gene called lamin A/C is being impaired. A similar defect is responsible for the inherited form of lipodystrophy called Dunnigan's lipodystrophy. Researchers believe that in the case of HIV lipodystrophy, the defect could be caused by the inhibition by HIV protease inhibitors of cellular proteases that are involved in maturation of a transcription factor involved in adipocyte differentiation called SREBP-1, which must be incorporated into the maturing adipocyte nucleus by the lamina. Any defect in lamin A/C would affect this process of incorporation.

Nucleoside analogues (NRTIs) have also been shown to affect adipocytes by increasing apoptosis (programmed cell death), especially when combined with protease inhibitors. However, they have also been shown to reduce the impact of protease inhibitors on adipocyte differentiation (although not to suppress this effect entirely). A French group found that d4T and AZT decreased cell lipid content and slightly increased apoptosis (Caron 2003).

Another avenue by which adipocyte activity can be altered is through induction of TNF-alpha and interleukin-6 production by protease inhibitors and by NRTIs. TNF-alpha reduces adipocyte differentiation, most specifically by affecting SREBP-1 activity. One study showed that indinavir, AZT and d4T each increased TNF-alpha gene expression, implying increased production (Jones 2003).

What remains unclear is the extent to which elevated TNF-alpha production is a direct effect of individual agents, compared to an effect of dysregulated immune reconstitution. Whilst in vitro evidence has pointed to the former, at least one in vivo study has shown that the dysregulation of TNF-alpha production is not restricted to adipocytes during immune reconstitution, but is a more generalised phenomenon caused by the suppression of lymphocyte apoptosis by protease inhibitors (Gougeon 2000; Ledru 2000). This dysregulation was strongly correlated with the presence of lipodystrophy and metabolic changes such as elevated triglycerides.

Drug-related defects in adipocyte functioning would provide an adequate explanation for many of the phenomena associated with lipodystrophy. Defects in adipocyte differentiation and increased apoptosis would tend to appear first in peripheral fat, leading to loss of fat cells. Increased insulin resistance would lead to increased release of free fatty acids from adipocytes (lipolysis), leading in turn to higher triglyceride levels and further insulin resistance. Increased efflux of free fatty acids from adipocytes has also been associated with d4T treatment and with nelfinavir (Hadigan 2002; Rudich 2001).

Visceral fat accumulation in the abdomen may be attributable to disruption of growth hormone and cortisol production. Whilst growth hormone limits central fat accumulation, cortisol encourages it. If TNF-alpha production is increased in adipose tissue, it will suppress activate cortisol synthesis if human growth hormone levels are deficient. Visceral fat accumulation would be aggravated by any increase in triglyceride levels or reductions in insulin sensitivity.

However, metabolic regulation of adipose tissue is extremely complex, and as the lack of a cure for obesity shows, difficult to unravel.

For example, studies in transgenic mice which lack subcutaneous fat show that these mice have a similar pattern of metabolic abnormalities to those seen in HAART-associated lipoatrophy (elevated triglycerides and diabetes type II). When a large quantity of subcutaneous fat from a normal mouse was transplanted into the fat-less mouse, the metabolic abnormalities disappeared. This study suggests that subcutaneous adipose tissue is a metabolic regulator, and that its disappearance may be partially responsible for the severe metabolic disturbances seen in people with lipoatrophy (Reitman 2000).

A UK Open University group has theorised that the development of lipodystrophy is influenced by the interaction between lymph nodes and adipose tissue. Animal experiments show that chronic stimulation of lymph nodes from the central fat depots led to proliferation of new lymph nodes, whilst peripheral fat stores were drained (peripheral fat does not contain lymph nodes) (Pond 2003).

Further evidence on the role of protease inhibitors

Different protease inhibitors interfere with different steps in lipid metabolism and glucose metabolism, each of which may contribute a different form of metabolic disturbance to the syndrome.

• In healthy HIV-negative adults indinavir causes a 20% reduction in insulin sensitivity within four weeks of starting treatment but was not associated with lipid elevations during the first four weeks of treatment (Noor 2002).

• Nelfinavir, saquinavir and ritonavir are more potent inhibitors of adipogenesis than indinavir and amprenavir.

• Lopinavir, nelfinavir, ritonavir and saquinavir increase triglyceride production in mice, especially if the drugs are taken at the same time as food (Lenhard 2000).

• Nelfinavir and indinavir have been shown to affect gene expression related to lipid metabolism and adipocyte differentiation in pre-adipocyte studies in mice (Stevens 1999; 2000).

• The protease inhibitors ritonavir and saquinavir have also been shown to have a direct effect on the regulation of apoprotein B in liver cells, possibly leading to increased lipoprotein (cholesterol) secretion, but not triglyceride production. Ritonavir also increased the uptake of triglyceride-rich particles by liver cells (Liang 2000; 2002).

• Amprenavir did not affect glucose metabolism but was associated with significant increases in cholesterol after eight weeks (Dube 2002).

• Indinavir has been shown to interfere with the insertion of SREBP-1 into the nucleus of adipocytes, preventing release of PPAR-gamma, and so leading to impaired adipocyte differentiation and reduced insulin sensitivity (Caron 2001). SREBP-1 regulates fatty acid synthesis. It has been suggested that this mechanism may account for the lack of replacement of adipocytes after apoptosis. Apoptosis is upregulated by thymidine analogues (d4T, AZT). Apoptosis is greatest when ddI, d4T and protease inhibitors are both present in cell cultures (Caron 2002).

• Protease inhibitors directly affect an enzyme called mitochondrial processing protease (MPP), which can lead to mitochondrial dysfunction. Protease inhibitors are highly hydrophobic (water insoluble), and hence may be concentrated in fatty tissues and have a greater impact on mitochondria in those tissues with chronic exposure (Mukhopadhyay 2002).

The role of nucleoside reverse transcriptase inhibitors and mitochondrial toxicity

However, lipodystrophic changes can occur in the absence of protease inhibitors, suggesting that other mechanisms are contributing to the development of the problems. At this stage, no one is sure if protease inhibitors contribute relatively little to the process, or if there are multiple potential pathways by which the syndrome can develop.

Dutch researchers led by Dr Kees Brinkman published a paper in The Lancet which argued that mitochondrial toxicity caused by the NRTIs is crucial in the development of fat and metabolic disorders associated with HAART. They drew parallels between HAART-related lipodystrophy and alcohol-related multiple symmetrical lipomatosis type 1 (MSL 1), both linked to mitochondrial dysfunction as well as abnormal fat accumulation and peripheral wasting. See also Lactic acidosis / acidaemia in Symptoms and illnesses: A to Z of illnesses for more on other mitochondrial toxicities and the effects of elevated lactate levels.

Mitochondria are normal structures in human cells which are responsible for energy production. The mitochondria contain genetic material which may mutate to cause mitochondrial disease. Mutations in mitochondrial DNA cause disorders such as mitochondrial encephalopathy (ME), lactic acidosis, stroke-like episodes, a type of epilepsy and an eye disease called Leber's hereditary optic atrophy. Mitochondrial DNA is easily damaged, and cannot be repaired if polymerase gamma (a protein responsible for replication of mitochondria) is inhibited by nucleoside analogues. As mitochondrial DNA becomes more damaged, oxidative phosphorylation declines, leading to cell damage, toxicity and cell death. According to this theory, mitochondrial damage may lead to wasting of subcutaneous fat leading to fat loss in the legs, arms, buttocks and face. A different effect may occur if mitochondrial DNA in central fat cells is impaired; in this situation, lipids could accumulate within cells.

While at first sight this may appear contradictory, different NRTIs are likely to have different effects in different tissues, depending on the distribution and penetration of NRTIs into particular cells, the distribution of adipocytes and the type of fat in different locations, and the cellular capacity to break down NRTIs (Brinkman 1999). It is already apparent, for example, that the NRTIs have different mitochondrial toxicities in other tissues, causing peripheral neuropathy, pancreatitis and myopathy in a drug-specific manner.

The Dutch team has also proposed that the extent to which various NRTIs inhibit mitochondrial polymerase gamma will affect the severity of mitochondrial toxicity associated with particular drugs (Brinkman 1999). Certainly a hierarchy of effects corresponds with the extent of polymerase gamma inhibition. However, a recent in vitro study found toxicity was not enhanced when d4T was added to ddI, although both drugs alone significantly impact on polymerase gamma. This suggests that inhibition of polymerase gamma may not be the key mechanism which produces mitochondrial toxicity (Walker 2002b).

An Australian group has closely examined fat tissue from people treated with PI and non-PI HAART. Electron microscopes revealed abnormal fat cell structure with prominent mitochondrial abnormalities, lipid accumulation in the cytoplasm and loss of fat cell volume. These changes were associated with fat cell loss and lipogranulomata formation (Mallal 2000). The same group found an association between elevated lactate levels and the more rapid onset of lipoatrophy (lactate is elevated when mitochondrial DNA is damaged).

Researchers in Germany and Hawaii have found significantly reduced levels of mitochondrial DNA in the subcutaneous fat tissue of NRTI-treated individuals when compared with HIV-positive untreated controls, and there was also a significantly lower level in NRTI-treated patients who had experienced fat loss on treatment when compared with those on NRTIs who remained unaffected (Walker 2002; Shikuma 2001). Reduced levels of mitochondrial DNA are also associated with symptomatic high lactate in people taking NRTIs (Cote 2002).

The German group found that the extent of mitochondrial DNA reduction was significantly associated with the duration of NRTI therapy and Dr Ulrich Walker, who conducted the study, told the Second International Workshop on Lipodystrophy that if mitochondrial DNA continued to decline at the rate seen hitherto (if compared to the control group and assessed by average duration of therapy), mitochondrial DNA would be reduced by 50% within seven years. The reduction in mitochondrial DNA was significantly associated with d4T use, but it is important to note that this study only looked at 19 NRTI-treated patients, of whom 14 received d4T. There was also a trend towards reduced mitochondrial DNA in 3TC recipients. The study did not investigate the total duration of NRTI therapy and possible associations with NRTIs used prior to the current regimen, so these data should be treated with caution.

The German team has also studied the effect of the NRTIs on human fat cells in laboratory experiments. The NRTIs ddI, ddC and d4T were found to deplete mitochondrial DNA and affect cell growth, lipid levels, lactate production and the expression of a protein encoded by mitochondrial DNA known as cytochrome C oxidase subunit II (COX II). In contrast, AZT did not affect mitochondrial DNA or COX II but it did impair cell growth, and increase lactate and intracellular lipid levels. 3TC had no significant effects (Walker 2002b).

A study in mice found that while d4T did not cause mitochondrial DNA reduction in the adipose tissue of normal mice, obese mice experienced a 45% reduction in adipose mtDNA at peak plasma d4T concentrations three times lower than those expected in humans (Gaou 2000). When the dose was increased fivefold, normal mice had lost nearly 50% of their mtDNA in liver and muscle tissue after six weeks of treatment.

Dutch researchers have reviewed body composition and mtDNA levels in adipose tissue of 28 participants in a randomised comparison of d4T/3TC and AZT/3TC, in which patients added indinavir at week 12. The original diagnosis of lipoatrophy in this study was made by a physician who was blinded to the original study medication, to avoid bias. They found that despite a similar duration of drug exposure for d4T and AZT, d4T-treated patients had a significantly greater prevalence of lipoatrophy (82% vs 9%), significantly less peripheral fat as measured by DEXA and significantly lower levels of mtDNA in subcutaneous fat cells. Stavudine treated patients were 56 times more likely to have lipoatrophy (p=0.0002), but protease inhibitor exposure did not affect the risk of lipoatrophy. The group also found a significant relationship between the amount of mtDNA in subcutaneous fat and the ratio of peripheral to total fat as measured by DEXA, leading the authors to conclude that mtDNA loss is associated with lipoatrophy, and that d4T is more strongly associated with lipoatrophy than AZT (van der Valk 2003).

A limitation of this study is that it included only 62% of the patients originally randomised; 10 withheld consent and seven were lost to follow-up. However, there was no significant difference between the characteristics of participants in this study and those of the entire trial group. Another possible limitation is that the relationship between mtDNA levels and the ratio of peripheral fat to total fat was only found to be significant in thigh fat samples; the relationship was not found to be significant in the fat sample taken from the lower back.

In the TARHEEL study, in which 118 individuals with fat wasting were switched from d4T to either AZT or abacavir, restoration of fat was accompanied by an increase in levels of mitochondrial DNA in PBMCs and muscle tissue. The study did not report on mtDNA levels in adipocytes however, and did not include a d4T-treated control group (McComsey 2002c).

Australian researchers examined mitochondrial DNA levels in 23 HAART-treated patients with lipodystrophy and 11 treatment-naﶥ HIV-positive controls matched for body mass index and found that mtDNA levels were significantly depleted only in d4T-treated patients (13% of control levels). Markers of adipocyte differentiation were also suppressed in d4T-treated patients, but upregulated in AZT-treated patients (Pace 2003). The same group conducted a cross-sectional analysis of mitochondrial DNA levels in the adipose tissue of 60 patients, and longitudinal analysis of 8 patients to determine the relationship between severity of fat wasting and mtDNA levels. They found that patients treated with d4T were likely to suffer significantly greater mtDNA depletion than AZT-treated patients, but that NRTI-treated patients as a whole had significantly lower mtDNA levels than untreated patients, and that NRTI-treated patients experienced mtDNA reductions in adipocytes of between 14% and 81% within eight months of commencing therapy (Hammond 2003).

Updating these findings in 2004, David Nolan reported on what is now a substantial cohort of patients who have undergone fat biopsies. One hundred and three patients underwent 147 fat biopsies. Forty-one treatment-naﶥ patients have been studied, together with 38 AZT-treated patients and 30 d4T-treated patients. The cohort has also gathered data from 24 patients who have started treatment with nucleoside analogue backbones that do not include AZT or d4T, reflecting a growing trend towards avoidance of thymidine analogues in first line HIV treatment.

The mean mitochondrial DNA level at baseline in treatment-naﶥ patients was 1427 copies per cell, with no significant difference according to race, CD4 percentage or gender.

Amongst patients treated with AZT or d4T, mitochondrial DNA levels in adipose tissue declined by at least 60% within six to twelve months and this change was statistically significant in comparison to treatment-naﶥ patients.

In contrast, 17 patients who commenced treatment with an abacavir-containing regimen that excluded AZT showed no significant reduction in mitochondrial DNA levels compared to treatment-naﶥ patients. Similarly, three patients who commenced tenofovir-containing regimens showed no evidence of mitochondrial DNA reduction.

A second longitudinal study, also carried out in Australia, also found that lipoatrophy was associated with mtDNA depletion, but found differential effects depending on the cells sampled. Fat biopsies were obtained from 122 patients.

Whilst current d4T, ddI or ddC treatment was associated with lowered limb fat mtDNA, only ddI and ddC were associated with lowered mtDNA in PBMCs. AZT had a greater effect on mtDNA in suprailiac limb fat than abacavir, 3TC or tenofovir, but a lesser effect than the dideoxynucleotides. No effect of age, duration of HIV infection, time on treatment, protease inhibitor treatment or viral load was seen on mtDNA levels (Cherry 2003).

Although most research has focussed on the effect of drugs on adipose tissue mitochondria, one study has shown that mitochondrial DNA levels in plasma PBMCs are predictive of later development of lipodystrophy.

The analysis was carried out by French researchers using stored blood samples from the ALBI study, which randomised treatment-naive patients to receive either AZT/3TC, AZT/3TC alternating with d4T/ddI, or d4T/ddI for 48 weeks. Stored samples containing peripheral blood mononuclear cells were available for 37 of 51 patients in the AZT/3TC arm and 40 of 51 patients in the d4T/ddI arm. Blood samples were drawn at baseline and when individuals had completed 48 weeks of treatment.

The researchers found that:

• The mean mitochondrial DNA content fell from 5130 copies/cell at baseline to 2450 copies/cell at week 48.

• Although mitochondrial DNA levels were not significantly different at weeks 0 and 24, by week 48 they were significantly lower in the d4T/ddI group (1950 vs. 3020 copies/cell).

• Thirty-nine per cent of patients had at least one symptom of lipodystrophy 30 months after starting treatment. Forty-four per cent of these patients had mitochondrial DNA levels below 1410 copies/cell, compared to 7% of those without lipodystrophy. The odds ratio (OR) of lipodystrophy at month 30 in patients with mitochondrial DNA below 1410 copies/cell was 9.8. Thirty-five per cent of those originally randomised to AZT were still taking the drug at month 30, compared to 63% of the d4T group.

• Although d4T/ddI treatment was significantly associated with lipodystrophy (OR 2.3), after adjustment for mitochondrial DNA at week 48, there was no significant difference in the risk of lipodystrophy in d4T/ddI-treated patients when compared to AZT and 3TC-treated patients.

Evaluating mitochondrial DNA levels in this population as a predictor of lipodystrophy excludes any complicating effect of a third drug that might also influence the way in which lipodystrophy manifests itself, but does not reflect clinical practice today, nor the effect of protease inhibitors on mitochondrial function (Amellal 2004).

While the theory of mitochondrial toxicity has considerable currency, not all research supports this theory. For example, a comparative study found no evidence that anti-HIV treatment was causing mitochondrial DNA mutations in people with HIV (Negredo 2001). This provided support for earlier findings that linked mitochondrial abnormalities to HIV infection itself (Morgello 1995; Simpson 1993). Other studies have shown that some people with lipoatrophy have no reduction in mitochondrial DNA compared with a control group (McComsey 2002a), whereas some HIV-negative people do have lower than average levels, suggesting wide variability within the population (and/or the possibility that mtDNA may be distributed at random levels in fat tissue). In addition, a recent study has shown that measuring levels of mitochondrial DNA does not predict the development of lipoatrophy or other drug toxicities associated with nucleoside analogues (McComsey 2005). In this study, the extent of mitochondrial DNA depletion was not associated with the presence of lipoatrophy, and one third of the nucleoside analogue-treated patients in the study showed an increase in mitochondrial DNA levels.

Although mitochondrial DNA may be depleted, several groups have shown that mitochondria are actually working harder (as measured by fat oxidation) in patients receiving protease inhibitors plus NRTIs, possibly in response to hypermetabolism (Ware 2000; Sekhar 2000). This, rather than NRTI toxicity, might also explain why mitochondria show altered shape (morphology) in patients with lipodystrophy.

Experience in the field of inherited mitochondrial disorder has shown that the brain is commonly affected (either in the form of encephalitis or dementia), but the only nervous system disorder seen with NRTI-associated mitochondrial toxicity is peripheral neuropathy. Indeed, nucleoside analogues have a protective effect against the development of HIV-associated dementia.

Mitochondrial DNA mutations associated with fat accumulation in Madelung's syndrome are not seen in HIV lipodystrophy.

Elevated lactate levels in patients taking NRTIs, often cited as a marker of NRTI toxicity to mitochondrial DNA, may be a consequence of reduced clearance of lactate in the liver as a consequence of elevated triglyceride levels and hepatic steatosis (fat accumulation in the liver) (Roge 2001). Finally, adipocytes are poor at phosphorylating the thymidine analogues AZT and d4T (Moyle 2001; Munch-Petersen 1991) in comparison to activated T-lymphocytes.

One group of researchers has developed a variation on the mitochondrial toxicity theory. The theory is that low levels of high density lipoprotein (HDL) plus mitochondrial toxicity lead to fat redistribution. HDL plays an important role in clearing lipids to the liver and is often low during HIV infection. The theory is that mitochondrial toxicity promotes the loss of white fat and the expansion of brown fat (which contains more mitochondria and may be less susceptible to mitochondrial toxicity). An inverse relationship was found between HDL and visceral fat confirming the relationship between low HDL and lipodystrophy (Fessel 2000).

Mitochondrial activity is also impaired by HIV infection itself. In a Spanish study samples of peripheral blood mononuclear cells were obtained from 25 HIV-positive patients who had never taken anti-HIV drugs and 25 age and sex matched HIV-negative individuals. The HIV-positive patients had been diagnosed with HIV infection for a mean of 44 months, had a median CD4 cell count of 317 cells/mm³ and an HIV viral load of 100,000 copies/ml.

Investigators assessed mitochondrial DNA content, activity in the mitochondrial respiratory chain, the activity of a key enzyme involved in mitochondrial function, and lipid peroxidation, a marker of oxidative damage.

Mitochondrial DNA content was 23% lower in the HIV-positive patients (p < 0.05) than the HIV-negative controls. Activity in the mitochondrial respiratory chain (MRC) was also significantly lower in the HIV infected individuals, MRC complex III being decreased by 38% (p < 0.001) and MRC complex IV by 19% (p = 0.001) (Miro 2004).

The investigators suggest that the effect of HIV on mitochondria may increase their vulnerability to the effects of nucleoside analogues, although no association was found between CD4 cell count and mitochondrial DNA content.

Is mitochondrial toxicity more severe in people with advanced HIV disease?

Although it is widely accepted that toxicities tend to be more severe in people with advanced HIV disease, nobody has been able to explain why. In a review of the literature on nucleoside analogue toxicity, pharmacologist Dr Peter Anderson of the University of Colorado and colleagues offer a potential explanation for the greater severity of nucleoside analogue toxicities in advanced HIV disease and during treatment of hepatitis C infection.

The authors note that people on therapy with CD4 cell counts below 100 tend to have more problems with NRTI-associated toxicities such as peripheral neuropathy, pancreatitis and fat wasting. They cite evidence for the association of higher intracellular NRTI phosphate levels with severity of HIV disease. A few studies also suggest that women have higher concentrations of phosphorylated NRTIs than men, which may explain epidemiological findings that women also have a greater susceptibility to developing NRTI toxicity. Finally, they make a link between cellular activation and NRTI side-effects, pointing to increased levels of chemical markers of activation such as interferon (IFN) and tumor necrosis factor (TNF) found in people with advanced HIV disease. What is missing from the literature, though, are data showing a direct correlation between increased levels of cellular activation markers and increased intracellular levels of phosphorylated NRTIs or increased NRTI toxicity.

Both nucleoside reverse transcriptase inhibitors and protease inhibitors?

Although several clinical trials have found that lipodystrophy is most common in people who combine the two classes of drugs, there is controversy on the question of whether this combination carries a greater risk of lipodystrophy, and whether the two classes of drugs might be interacting at the cellular level to cause more serious problems.

For example, the Prometheus study showed that despite no nucleoside analogue treatment, 8% of PI recipients did develop lipodystrophy. The study also found no significant difference in patterns of body fat changes in each arm of the study; lipoatrophy was somewhat more common in the d4T arm (32% vs 14%), and central fat accumulation was somewhat more common in the PI-only arm (29% vs 9%).

Another long-term study in which people received dual protease inhibitor therapy, with (44) or without NRTIs (39), has reported similar findings. After four years on treatment, people who received ritonavir/saquinavir alone, without NRTI intensification from week 48 (or earlier if their regimen was insufficiently potent) were significantly less likely to develop lipodystrophy, especially in its most severe form. Intensification with NRTIs (predominantly d4T and 3TC) resulted in an 8.4-fold increase in the risk of body fat changes after adjusting for age, CD4 cell count at baseline, duration of prior NRTI treatment, baseline viral load and presence of hyperlipidemia.

Although fat wasting (lipoatrophy) was noted in the protease inhibitor group, the incidence was 1-2%, and it has been suggested that these changes could equally be seen as manifestations of ageing (Cohen 2002). There was no significant difference in lipid elevations between the two arms in the study.

The weakness of this theory is that body fat and metabolic changes have been observed to occur in the absence of both classes of drugs, in the absence of one class of drug, and in the absence of therapy. A prospective study of people starting treatment in Spain found that 11% of 166 individuals who started treatment with either a triple nucleoside analogue or non-nucleoside reverse transcriptase inhibitor (NNRTI)-based regimen developed body fat changes during a mean follow-up period of 14.5 months. The incidence of lipodystrophy, determined by clinician and patient report, was 7.49 cases per 100 patient years of follow-up. The Spanish group reported that the main risk factors for lipodystrophy by multivariate analysis were age over 45 years (RH13.36) and baseline fasting triglycerides > 200mg/dL (RH 14.88) (Martinez 2002).

The role of TNF-alpha: further evidence

Researchers from the Pasteur Institute in France have suggested that HAART dysregulates TNF alpha, specifically by reducing apoptosis (cell suicide) in a subset of CD8 cells which produce TNF alpha. TNF alpha levels become elevated as a consequence. High levels of T cells primed to produce TNF alpha have been correlated with HAART-associated lipid and fat abnormalities (Ledru 2000).

TNF alpha also inhibits triglyceride deposition in adipocytes, and promotes lipolysis. A cellular inhibitor of TNF-induced apoptosis (CIAP2) is more frequently expressed in central (visceral fat) than subcutaneous fat, suggesting that these cells will be protected in the presence of elevated TNF-alpha levels if levels of interleukin-10 are suppressed at the same time, which is an immunological pattern seen in people who experience immune restoration on HAART (Imami 1999). Interleukin-10 acts to upregulate the pathway through which TNF-alpha-induced adipocyte apoptosis occurs in central fat, so IL-10 suppression will inhibit apoptosis of central fat.

TNF has been shown to be synergistic with protease inhibitors in inhibiting lipid accumulation in adipocytes by one study, and another study has found higher levels of TNF in subcutaneous fat compared to visceral fat (Agrawal 2001; Johnson 2001, 2002). This might be expected to promote fat loss in the limbs as compared to the abdomen.

Genetics may play a role in TNF dysregulation. One study found that people with a particular variation in the TNF alpha gene were more likely to develop lipodystrophy. However, only about 15% of the total sample had this variation, so it cannot account for all cases of lipodystrophy (Maher 2002). Although the TNF-alpha-238G/A polymorphism has been correlated with lipoatrophy in several studies (Nolan 2002b; Maher 2002), it is not associated with a greater degree of wasting in cancer or tuberculosis, two other conditions in which TNF-alpha production plays a significant role in weight loss. In one of these studies, fat wasting was more rapid if individuals were heterozygous for TNF-alpha-238G/A, but was not contingent on its presence (Nolan 2002b).

Researchers from Liverpool University have also suggested that protease inhibitors may modulate TNF-a activity and may, in turn, induce some features of lipodystrophy. This theory is based on the observation that 10 patients with lipodystrophy had higher levels of TNF receptors than controls, and that TNF receptor levels fell significantly when PIs were ceased (Maher 2000b).

Further research by the same group has shown that nelfinavir, saquinavir and ritonavir (but not indinavir) increase levels of TNF alpha in adipocytes. Combining any of these protease inhibitors, or combining d4T or AZT with indinavir reduced growth of new adipocytes, and the extent to which differentiation was suppressed correlated with the magnitude of TNF alpha production in the presence of a drug.

NRTI therapy has been shown to result in increased TNF-alpha production in peripheral leukocytes when compared with an HIV-negative control group, but a weakness of this study is the lack of an untreated HIV-positive control group. However, the study also found that increased TNF-alpha levels were associated with insulin resistance in the absence of elevated glucose levels (Limone 2003).

Other researchers have confirmed a link between levels of programmed cell death and soluble TNF receptors (Domingo 2002).

In contrast, a reduction in cytokine activity has been correlated with lipoatrophy in one study, suggesting a different mechanism (Mynarcik 2000). This mechanism focuses on inhibition of acylation-stimulating protein (ASP). ASP upregulates glucose uptake and fat deposition in adipocytes. ASP production is reduced in the subcutaneous fat of individuals with lipoatrophy when compared with HIV-positive individuals without lipoatrophy, or HIV-negative controls. In a study of adipocytes from 24 individuals, lipoatrophy was associated with lower ASP production and reduced conversion of C3a (a component of complement) into ASP, and with higher serum levels of soluble TNF receptor (Ionescu 2002). The function of C3a is to recruit neutrophils and further elements of the complement cascade to locations invaded by microbes, and its production is normally upregulated by TNF-alpha; an increased level of TNF receptors will blunt the inflammatory action of TNF-alpha.

A slightly different approach to the cytokine disruption theory has come from researchers at the Chelsea and Westminster Hospital in London. Evidence that interferon alfa therapy increases lipid levels has been presented as evidence that elevated cytokine levels caused by immune dysregulation may play a role in lipodystrophy (Morlese 2000).

An immune restoration syndrome?

It has also been proposed that lipodystrophy is an immune restoration syndrome, and that it is most strongly associated with the best responses to therapy.

An analysis of the HOPS database in the US has found that whilst there is a strong relationship between d4T, indinavir and lipoatrophy, people whose only risk factor was more than two years therapy with both d4T and indinavir did not develop lipoatrophy. It was only when other risk factors began to be added to the analysis that the risk of lipoatrophy emerged. These risk factors were:

• A CD4 cell count increase of greater than 200 cells/mm³.

• A CD4 cell count nadir of less than 100 cells/mm³.

• A prior AIDS diagnosis.

• Age over 40.

The risk increased with every risk factor that an individual displayed. In other words, people over 40 who had a prior AIDS diagnosis, a very low CD4 cell count in the past and a very good response to therapy, and who received more than two years treatment with d4T and indinavir, were most likely to have lipoatrophy.

The group also found an association between viral rebound, subsequent CD4 cell loss and a subsequent improvement in fat loss, regardless of which drugs the individuals were exposed to during the period of rebound and improvement (Lichtenstein 2001).

However, the association between a good response to therapy and subsequent body fat changes may be a marker for higher levels of exposure to protease inhibitors and better adherence to drugs which cause the syndrome.

The interaction between the immune system and the endocrine system is not well understood, and several recent studies have suggested that further research needs to be directed towards unravelling this relationship if lipodystrophy is to be understood. For example, a study of 135 HIV-positive women showed that baseline body fat predicted the extent of CD4 cell gain after starting therapy. For every 2.5 kg of additional fat, the CD4 cell count was +3.97 cells higher regardless of viral load at any time point (Jones 2001).

Leptin, adiponectin and lipodystrophy

Leptin and adiponectin are adipokines, hormones or cytokines that are released by adipocytes to regulate fat storage.

Leptin is a hormone that regulates appetite and weight; used therapeutically in lipodystrophic mice it has been shown to improve glucose tolerance and at higher doses to reduce obesity. Leptin reduces the lipid content of peripheral tissues, making them more sensitive to insulin. In HIV-negative individuals with Dunnigan's lipodystrophy, lipoatrophy has been improved with leptin replacement therapy, accompanied by improved insulin sensitivity.

Leptin supplementation has been proposed as a treatment for obesity, and in mice leptin deficiency leads not only to severe obesity but also to metabolic abnormalities similar to those seen with HAART. Leptin injections have been proposed as a treatment for fat accumulation, but injections would be necessary at least four times a day in order to have an effect, and leptin supplementation may also have an effect on the reproductive system and other organs (Shimomura 2000).

The clinical evidence from people with lipodystrophy is confusing. A US study of people with lipoatrophy, lipodystrophy, a mixed syndrome or no lipodystrophy found that low leptin levels were correlated with insulin resistance in people with lipoatrophy only (Nagy 2003). A Spanish study similarly found that low leptin levels were correlated with lipoatrophy (Estrada 2002). In children with HIV on HAART, pre-treatment leptin levels were predictive of lipoatrophy, whereas TNF alpha and soluble TNF alpha receptor levels (see above) were not predictive (McComsey 2002d).

However,a case control study of patients with and without lipodystrophy has shown that leptin levels were higher in those with lipodystrophy (Kosmiski 2003), and a group which compared people with and without lipodystrophy found that whilst leptin levels correlated with total body fat, they did not correlate with limb fat levels (Mynarcik 2002).

Adiponectin is a hormone which improves insulin sensitivity. It has been shown to be deficient in people. A study of 131 men receiving PI therapy found that adiponectin levels were positively correlated with insulin sensitivity, and that the ratio of adiponectin to leptin was positively correlated with expression of soluble TNF alpha receptors, suggesting that disruption of the adiponectin/leptin axis is correlated with TNF alpha levels. Patients with mixed lipodystrophy had the most severe changes in adiponectin/leptin ratios (Vigouroux 2003).

A comparative study of lipodystrophic and non-lipodystrophic HIV patients found that those with lipoatrophy had low levels of adiponectin secretion in limb fat, and adiponectin levels were closely correlated with insulin resistance, especially in the liver (Sutinen 2003).

Another group that examined adiponectin levels found that the relationship between adiponectin and insulin resistance disappeared after controlling for NRTI use, suggesting to them that adiponectin levels may be reduced by NRTIs, thus affecting insulin sensitivity (Addy 2003). The most likely mechanism is through the ability of NRTIs to induce TNF-alpha secretion, which in turn suppresses adiponectin production.

Insulin resistance and fatty acid cycling as drivers of the syndrome

In the first few years after the emergence of the lipodystrophy syndrome, enquiry into the possible causes focused on the role of different classes of antiretrovirals (see above).

More recently, evidence has accumulated that suggests a more complex set of pathways to the development of the various body fat changes and metabolic problems.

Studies in HIV-positive and HIV-negative individuals treated with protease inhibitors have shown that that a number of metabolic processes become dysregulated within weeks of commencing protease inhibitor treatment. These metabolic processes all affect the regulatory role of insulin, a hormone that is secreted by B-cells in the pancreas.

Insulin regulates the uptake of glucose into skeletal muscle tissue and other cells in the body, and if insulin sensitivity is reduced, more insulin must be secreted to maintain the same level of glucose uptake. Eventually the pancreas will be unable to produce enough insulin to control glucose levels in the blood, and diabetes will develop.

Several studies have shown that indinavir causes defects in glucose metabolism very quickly. Inhibition of the glucose transporter Glut4 has been demonstrated in rats (Murata 2001) and indinavir has been shown to lower insulin-stimulated glucose disposal in people (Noor 2002). Amprenavir and ritonavir have also shown the same effect in animal studies. What is happening is that the main mechanism for removal of glucose from the blood is being blocked, so tissues show increasing insulin resistance. This effect may be masked by compensatory production of insulin by the pancreas, leading to apparently normal plasma glucose levels.

Free fatty acid circulation is also increased; this is due in part to reduced storage of triglycerides in adipose tissues due to lower uptake of glucose. Adipose cells are experiencing a `fasted` state, because of the lowered uptake of glucose, and they respond to this by releasing free fatty acids for use by other tissues, a process called lipolysis. These fatty acids return to the liver, where they are rejected for storage, and sent back into the circulation again, leading to elevated triglyceride levels. Hypertriglyceridemia may reduce lactate clearance, leading to hepatic steatosis and hence hyperlactatemia.

As discussed above under 'Theory 4: Cytokine disruption' there are some data to suggest that increased lipolysis may be driven by dysregulation of TNF alpha synthesis. This cytokine may be produced in abnormal quantities despite the apparent improvements in the immune system seen in people taking HAART. This mirrors the breakdown of fat tissue seen in acute illness, as the body burns its stores of fat in response to increased TNF alpha production (Maher 2000, 2002). High levels of TNF inversely correlate with reduced expression of factors such as SREBP1 in the fat tissue of HIV-infected people taking protease inhibitors (Bastard 2002).

If TNF is a driver of the development of the syndrome, TNF and hyperinsulinemia need not have mutually exclusive roles. Increased lipolysis and insulin resistance could mutually reinforce each other, and increased TNF production could be one of the factors stimulating both increased lipolysis and increased insulin resistance. Free fatty acids administered to HIV-negative individuals are known to induce insulin resistance, while a study in HIV-positive individuals with hyperinsulinemia and an average of 3.5 years of protease inhibitor treatment found that reduction of free fatty acid levels using the drug acipomox increased insulin sensitivity by 84% after just two doses (Hadigan 2001b).

Consequences of insulin resistance

If insulin resistance persists, it will tend to lead to an accumulation of visceral or central fat, and this will have a reinforcing effect on insulin resistance itself, increasing the risk that insulin resistance will ultimately lead to diabetes, and that central fat will continue to accumulate.

Hypertriglyceridemia is a consequence of increased free fatty acid production, and also drives elevations in total cholesterol. Triglycerides are carried in very low density lipoprotein particles that are also measured as part of the total cholesterol count, which will become elevated when triglyceride levels rise. A small proportion of this cholesterol will be low density lipoprotein, or bad cholesterol, which is implicated in cardiovascular disease. This pathway explains the connection between insulin resistance and lipid elevations, but it is notable that not all individuals who develop lipid elevations will have impaired glucose tolerance or insulin resistance. However, these individuals may have experienced relative declines in insulin sensitivity and some decline in glucose tolerance. It is also possible that studies which measure insulin resistance and lipid elevations may not test insulin resistance at the time when protease inhibitors are at peak levels; a clear temporal relationship was shown between removal of drug and very rapid decline of insulin resistance in the studies mentioned in the previous sub-section.

Is lipodystrophy caused by the effects of antiretroviral drugs on the brain?

Dutch researchers have suggested that redistribution of body fat could be caused by the effects of antiretroviral drugs on the autonomic nervous system, the subconscious network of neurones that controls the activity of the organs and tissues.

They propose that HAART affects the brain regions controlling the amount of subcutaneous fat in opposite ways to the regions controlling the fat tissue around the internal organs.

Their hypothesis states that the drugs cause the sympathetic component of the autonomic regions controlling the subcutaneous fat tissue to become more active than the parasympathetic component. This leads to a loss of fat from beneath the skin. Conversely, the build-up of fat around the internal organs is a result of a greater increase in the activity of the parasympathetic component of the regions controlling the visceral adipose tissue.

They argue that existing theories on lipodystrophy, such as disruption of fat cell differentiation and mitochondrial toxicity, do not account for the different ways in which the subcutaneous and visceral fat stores are affected. Their hypothesis, which remains to be tested, may offer an explanation.

The sympathetic and parasympathetic systems act in opposite directions to one another, and processes including heart rate, blood pressure and the rate of digestion are regulated by a balance between the two systems. The two systems are connected to different autonomic nuclei in the brainstem: the neurones that form the parasympathetic system are controlled by the dorsal motor nucleus of the vagal nerve, whereas the sympathetic nervous system is under the control of a variety of areas including the A1 region.

The researchers have previously demonstrated that selective destruction of the dorsal motor nucleus of the vagal nerve in rats causes increases in insulin resistance, reductions in glucose and fatty acid uptake and an enhancement of the activity of fat-digesting enzymes. This suggests that activation of the parasympathetic nervous system stimulates fat accumulation. Conversely, activation of the sympathetic nervous system induces the breakdown of fat and the mobilisation of free fatty acids.

By tracing the pathway of the nerves from the adipose tissue to the brainstem, the authors have recently demonstrated that there is a clear anatomical separation between the brain cells that control the subcutaneous and visceral adipose tissues within the autonomic nuclei. This leads them to suggest that antiretroviral drugs may affect the activity of nerve connections to the subcutaneous and visceral fat stores differently and that this may account for the loss of fat from beneath the skin and its accumulation around the organs.

In addition to their animal studies, the group support their proposal with some circumstantial clinical evidence. Patients with lipodystrophy tend to have higher plasma levels of noradrenaline, the major neurotransmitter in the sympathetic nervous system, suggesting that alterations in this system may be involved in the redistribution of fat. Patients with HIV and AIDS also tend to exhibit general alterations in the activity of the autonomic nervous system, which may be due to the effects of antiretroviral treatment.

A problem for this hypothesis is how antiretroviral drugs access the brain regions comprising the autonomic nervous system, since most drugs and hormones are prevented from reaching the brain tissue by the blood-brain barrier. The researchers suggest that these drugs could enter at specific sites where this barrier is permeable to certain compounds. These include two regions (the area postrema and the arcuate nucleus of the hypothalamus) that are connected to the parasympathetic regions of the brainstem. Damage to these areas by HAART may affect the activity of the autonomic nervous system, leading to changes in fat build-up or metabolism.

To support this, the researchers point out that detectable levels of amprenavir, saquinavir and ritonavir have been observed in the cerebrospinal fluid of patients. Furthermore, destruction of the arcuate nucleus of rats shortly after birth leads to obesity in adulthood, suggesting a link between damage to this brain region and abnormal fat accumulation.

To test their hypothesis, the researchers suggest that the penetration of antiretroviral drugs into the brain tissue must be confirmed in animal studies, particularly into the regions that control the sympathetic or parasympathetic nervous systems. Subsequently, they propose testing whether localised injections of antiretroviral drugs into these brain regions brings about body fat changes and increases in insulin resistance similar to those observed lipodystrophy. Discovery of this effect would suggest that antiretroviral drugs are likely to act on these brain regions to cause body-fat redistribution in man.

A major question surrounding the theory is whether the proposed effects on the autonomic nervous system are due to reversible changes in the activity of brain cells, or the irreversible death of these neurones. The researchers suggest that this could be addressed by analysing the brains of rats following antiretroviral treatment. By observing the anatomy of these regions under the microscope or by measuring their electrical and biochemical activity, they hope to assess the degree of loss of brain cells or alterations in their function following drug treatment. These findings could be complemented by nerve-tracing techniques to investigate how the connections between the adipose tissue and the brainstem change during antiretroviral therapy.

The investigators state that their hypothesis could eventually be tested in HIV-infected patients with lipodystrophy. One way in which they plan to do this is to sample the levels of neurotransmitters in the subcutaneous and visceral adipose tissue of patients. They point out, however, that sampling of neurotransmitter levels in the fat tissue surrounding the gut may be restricted on ethical grounds (Fliers 2003).

References

See references in Anti-HIV therapy: Body fat and metabolic changes whilst on treatment.