| |
Mitochondrial dysfunction following perinatal exposure to nucleoside analogues
|
| |
| |
AIDS: Volume 20(13) 22 August 2006 p 1685-1690
[OPINION PIECE]
Blanche, Stephanea; Tardieu, Marcb; Benhammou, Valeriec; Warszawski, Josianec; Rustin, Pierred
From the aHopital Necker Enfants-Malades, Assistance Subligne-Hopitaux de Paris et EA 3620 Universite Paris 5, France
bPaediatric Neurology Service, Hopital Bicetre and Inserm EMI 01 09, France
cINSERM U569, Hopital Bicetre, France
dINSERM U676, Hopital Robert Debre and Universite of Paris 7, Paris, France.
Antiretroviral drugs given during pregnancy and immediately after birth are extraordinarily effective at preventing mother-to-child transmission of HIV-1: the transmission rate is dramatically decreased from 15-20% to less than 1%, at least for women intensively managed with potent antiretroviral therapy [1,2]. Current recommendations are now based on antiretroviral combinations, mostly including at least two nucleoside analogues (NA). Because NA can cross the placenta and can interact with human DNA, fetal tolerance of these drugs used during pregnancy should be carefully assessed [3,4]. Mitochondrial toxicity in adults has been the subject of substantial attention for some years [5]. There is no obvious reason why the fetus and newborn, often exposed for many months, should escape any such effects.
Tolerance data from the initial randomized study ATCG076-ANRS024
Although zidovudine was already known to have an effect on mitochondria when the ATCG076-ANRS024 protocol was designed [6], it was considered to occur only during prolonged treatment and no specific marker was included [1]. The only toxicity formally demonstrated in this zidovudine versus placebo-controlled trial was a transient moderate anaemia of unknown mechanism [1,7]. A detailed neurocognitive analysis was performed on a small subgroup of children followed until 36 months of age and no difference was observed between the group exposed to zidovudine and the unexposed controls [8].
Primate model of mitochondrial toxicity after in-utero exposure
Several experiments performed by Miriam Poirier's group with pregnant patas monkeys have clearly shown that a human-equivalent dose of zidovudine, alone or in combination with lamivudine, leads to multiorgan mitochondrial injury in newborn animals. Mitochondrial dysfunction was demonstrated in various tissues and with variable severity through morphological damage detected by electron microscopy, OXPHOS enzyme activity alterations, zidovudine incorporation into mitochondrial DNA (mtDNA), depletion in mtDNA quantity (shown by Slot blot), and degradation of mtDNA (shown by Southern blot) [9-12]. The reversibility and clinical significance of these biological findings were not known, but more recently, persistent mitochondrial dysfunction, at least in cardiac tissues, was demonstrated after perinatal exposure in the same animal model. Monkeys perinatally exposed to various NA regimens showed extensive mitochondrial damage at 1 year of age, with clusters of mitochondria in the heart having apparently similar damage as revealed by electron microscopy [13]. Interestingly, OXPHOS activity assays were normal, and presumably these assays are not sensitive enough to detect these clusters. In contrast to what is observed at birth, mtDNA quantity was increased, suggesting a recovery process and mitochondrial proliferation.
Tolerance data from epidemiological cohorts
Observations suggestive of possible mitochondrial toxicity in children perinatally exposed to NA originated from epidemiological cohorts of children born to HIV-infected mothers and from anecdotal case reports. The main limitation of epidemiological studies is that comparison between exposed and unexposed children are uncontrolled and necessarily 'historical'. The vast majority of unexposed children were born before 1994, whereas exposed children were born after this date. In addition, specific control groups for biological studies based on children born to uninfected mothers do not exclude potential counfounding bias linked to the HIV infection status of the mother.
Hyperlactataemia
Following the first description of symptomatic lactic acidosis in a child while receiving prophylactic treatment for HIV [14], several reports suggested that an important number of infants perinatally exposed to NA presented hyperlactataemia during the first few weeks or months of life [15-17]. Potential artefacts in lactate measurement were excluded in a recent controlled study showing that lactate levels were significantly higher during the first month of life in NA-exposed children than in controls. In contrast, a recent study in the Ivory Coast did not show any difference in lactate level between controls and children exposed to short-term zidovudine/lamivudine prophylaxis or to nevirapine [18], suggesting that hyperlactataemia could be linked to the level and/or duration of NA exposure. The long-term consequences of this hyperlactataemia, whether asymptomatic or symptomatic, are not known. Although transient mitochondrial injury, lasting several weeks, may have only minor clinical consequences in the muscle or liver, the potential impact in the brain needs careful evaluation.
Other biological markers of mitochondrial dysfunction at birth in humans
Three studies have described biological evidence of mitochondrial dysfunction in asymptomatic children exposed to zidovudine alone or in combination with lamivudine [19-21]. All three showed that exposed children had significantly lower levels of mtDNA in their peripheral blood lymphocytes at birth than unexposed children. In one of these studies, this difference persisted at 1 year of age [22]. Moreover, mtDNA levels were lower in the control group of untreated children born to HIV-infected mothers than in children born to uninfected mothers [19]. Depletion of mtDNA was also observed in the placenta and in the cells of the umbilical cord [20] Finally, abnormalities in mitochondrial ultrastructure, as assessed by electron microscopy, were found to be more frequent in the endothelial cells of the umbilical cord in exposed groups [21]. Studies on mtDNA quantification raised the hope that a simple non-invasive tool for the measurement of mitochondrial dysfunction could be developed. However, in addition to the potential methodological problems, it has been suggested that peripheral blood lymphocytes are not the cells most sensitive to this toxicity and that this marker is not currently suitable for routine use [23]. Discordant unpublished results have subsequently been obtained with this variable. Moreover, mtDNA depletion does not necessarily reflect a clinically detectable pathological process because the thresholds of depletion associated with clinically relevant cellular distress are unknown. In addition, mtDNA depletion is certainly not the only mechanism of NA-induced toxicity.
Neurological symptoms and persistent mitochondrial dysfunction in uninfected children
Full-blown mitochondrial neurological disease was first observed in a tolerance trial of a zidovudine/lamivudine combination during pregnancy [24,25]. The search for similar neurological syndromes associated with persistent mitochondrial dysfunction was then extended to all the 5000 HIV-uninfected children of the French national cohort [26]. After a complex stepwise screening of 150 files of children presenting neurological symptoms of unknown origin, expert groups identified 21 cases as being 'certainly' or 'probably' caused by mitochondrial dysfunction, according to diagnostic criteria similar to those for constitutional mitochondrial diseases [27,28]. The main clinical symptoms were seizures, cognitive delay, motor dysfunction, nystagmus or severe malaise. The mitochondrial origin of the symptoms was established based on profound OXPHOS deficiency and/or a severely disturbed histological pattern. Similar symptomatology, persistent hyperlactatemia and several histological images of abnormal mitochondria led to the notion of 'probable' mitochondrial dsyfunction despite normal or unknown OXPHOS activities. No mtDNA depletion was observed by Southern blot. All the affected children had been exposed to NA, with an incidence in this NA-exposed group of 0.3%, higher than the expected incidence of 0.01% reported for two Northern European general populations using very similar diagnostic criteria [29,30]. However, the neurological symptoms of several children remained unexplained and there were no signs indicative of mitochondrial dysfunction. Other physiopathological processes were considered, together with the difficulty in demonstrating a brain mitochondrial dysfunction, transient or persistent.
Absence of mitochondrial symptoms in other similar cohorts
The publication of these observations led to controversy, with several articles concluded that no similar cases were present in other cohorts. Some analyses quickly followed publication of the first alert [31,32] but were too rapid to be compatible with the long and complex work of investigating diverse, multifactorial and non-specific symptoms sometimes presented by HIV-1-exposed but uninfected children with neurological symptoms. A more structured analysis of five American cohorts, focusing exclusively on retrospective analysis of the medical files of children who had died, showed that none of the 223 deaths among more than 20 000 children born to HIV-seropositive women could be unambiguously attributed to a mitochondrial disease [33]. This analysis is often wrongly cited as the definitive proof of safety, affirming that tens of thousands children have been correctly screened, but this is not the case. To date, only 2 of the 21 children identified in the French cohort have died; the issue is the identification of mitochondria-related morbidity rather than mortality. More recently, two cohorts listed all the symptoms observed in uninfected children: the WITS cohort presented an analysis of the symptoms of 955 children not infected with HIV but indicated that the analysis did not take into account possible symptoms of toxicity [34]; the European ECS cohort did not report any symptom suggestive of a mitochondrial disease in more than 2000 HIV-uninfected children, half of them having been perinatally exposed to NA [35]. The very low rate of identified events of all types - including those normally expected in the general population - strongly suggested insufficient sensitivity to adverse events. These analyses are, therefore, not comparable with ours and the results obtained cannot be considered to be truly contradictory of what we observed in French cohort.
Outside these cohorts, at least two cases of severe neurological symptoms and mitochondrial dysfunction have been reported recently [36,37]. In both cases, neurological symptoms were early and severe, and mitochondrial dysfunction unambiguous. Interestingly, mtDNA was severely depleted both, in contrast with the observations in France for older children. A 5-year follow-up of one child revealed persistent but stable cognitive delay. Since the first publications, several further cases of persistent mitochondrial dysfunction have also been identified in the French cohort.
Non neurological mitochondrial disorders?
In a specific study, echocardiographic data for zidovudine-exposed children showed no significant cardiac perturbation [22]. However, the same group recently presented preliminary data suggesting that findings could be different in children exposed to antiretroviral drug combinations [38].
Other symptoms or biological disturbances have been observed in newborns and infants exposed to antiretroviral drugs [23,39]. The most convincing is certainly the long-lasting, mild, but significant inhibition of haematopoiesis, similarly observed in two independent cohorts [40-42]. However, there is currently no evidence of this intriguing persistent disturbance being of mitochondrial origin.
Diagnostic difficulties for mitochondrial toxicity in children
Mitochondrial dysfunction remains difficult to establish in humans in clinical practice [27,28,43]. The clinical symptoms are extremely diverse, involve various organs and are far from being specific. In the particular context of maternal HIV infection, the first step is to eliminate other possible causes of neurological problems, such as maternal intravenous drug use, prematurity and other infections. Simple biochemical findings, such as true hyperlactataemia and also abnormal transaminases, lipase, creatinine phosphokinases and lactate dehydrogenases may suggest mitochondrial dysfunction. However all these features are non-specific and, more confusingly, can be normal despite genuine mitochondrial dysfunction. Magnetic resonance imaging of the brain may also provide additional non-specific information regarding diffuse hypersignal in the supratentorial white matter or brainstem [44]. The diagnostic gold standard remains the demonstration of OXPHOS deficiency and/or the observation of major abnormalities of mitochondrial structure (histochemical or electron microscopy studies). In cases of significant OXPHOS perturbation or abnormal mitochondrial images, specific mutations associated with constitutional diseases can be sought. In the specific context of potential NA toxicity, mtDNA quantification is also required for complete analysis. With few exceptions, skeletal muscle tissue is analysed, and this necessitates the aggressive procedure of a muscle biopsy. However the diagnosis, even for muscle, remains difficult because assaying activities of the respiratory chain complex is not straightforward; specific histological signs in children can be subtle, with the classic red ragged fibre appearance much less frequently than observed in adults; and techniques for measuring the depletion of mtDNA (if involved) in tissues have not been standardized. Moreover, normal OXPHOS activities do not exclude the diagnosis of mitochondrial dysfunction because the muscle is not necessarily the target organ. As well-described elsewhere for constitutional diseases and applied here, a diagnostic probability scale can be used (with levels of 'established', 'probable', 'possible').
Physiopathology
The physiopathological mechanisms underlying this transient or persistent mitochondrial dysfunction following a relatively short period of NA exposure (several weeks to several months) are unknown. According to current knowledge, there are various potential targets for NA in mitochondria [45-47]. The most frequently cited is mitochondrial gamma-polymerase, and its inhibition by NA presumably results in mtDNA depletion. This inhibition is in principle reversible as long as some mtDNA is still present in the cell. Any persistent dysfunction must, therefore, be the consequence of one or more other mechanisms. A transitorily impaired respiratory chain activity would be predicted to result from any significant mtDNA depletion. This would, in turn, favour overproduction of superoxides by the respiratory chain. Because of their high mutagenic potential, these reactive species may then irreversibly and randomly alter mtDNA. Irreversible mtDNA alterations might also result from impaired proof-reading activity of the mitochondrial gamma-polymerase triggered by NA.
One of the potential alternative targets suggested for zidovudine is the adenylate carrier, involved in ADP/ATP exchange across the mitochondrial inner membrane [48], as blockade of the phosphorylation step of the respiratory process has been shown to produce more toxic superoxides than any other type of respiratory chain impairment [49].
Finally, inhibition by zidovudine of the deoxynucleotide carrier on the inner mitochondrial membrane may impair mtDNA synthesis by reducing or unbalancing uptake of dNTP [50-52]. Recent descriptions of inherited deficiency of the thymidine phosphorylase report that either mtDNA depletion or a shortage of one species of nucleotide necessary for mtDNA synthesis in the mitochondrial matrix may result in inappropriate insertion of an alternative nucleotide in the place of the predicted nucleotide [53]. Any of these hypotheses that involve irreversible and randomly distributed mtDNA mutations are compatible with the apparently discordant results observed: mtDNA quantification data indicate normalization after the marked reduction following exposure, despite persisting signs of mitochondrial dysfunction such as OXPHOS deficiency. It remains to be determined whether there is any such qualitative damage of mtDNA in NA-exposed children.
Although a significant amount of work as been devoted to the impairment of mitochondrial function and mtDNA replication machinery by NA, we are still far from understanding the many potential interactions between these drugs and the various reactions of the nucleotide and nucleoside pathways. At a pharmacological level, we know little about the fetal distribution of the various NA. A recent comparative study indicated that most nucleoside reverse transcriptase inhibitors cross the placenta by simple diffusion and are concentrated in the amniotic fluid, probably through fetal urinary excretion [54]. The efficacy or toxicity of nucleoside reverse transcriptase inhibitors may vary according to placental transfer. There may be interindividual variability in genetic susceptibility for any toxicity that is related to NA pharmacology and/or mitochondria-related genes. A recent study in adults suffering from NA-induced neuropathy suggested a specific suceptibility for some mitochondria haplotypes [55].
Conclusions
Despite many unresolved physiopathological questions, there are coherent experimental and clinical arguments for the existence of mitochondrial toxicity following perinatal exposure to zidovudine, alone or in combination with lamivudine. This effect may be transitory or persistent. Although, the clinical significance of the various biological and/or histopathological mitochondrial anomalies found at birth in NA-exposed animals or children has not been formally established, the high incidence of mitochondrial neurological diseases observed in the French cohort has not been contradicted to date in other cohorts. All data concerning potential toxicity must be balanced against the well-proven benefit of the drugs, but also with a view to possible alternative therapeutics. Given the increasing number of antiretroviral drugs available, a comparative evaluation is required to identify the least toxic therapeutic strategies.
|
|
| |
| |
|
|
|
