Persistent mitochondrial dysfunction in HIV-1-exposed but uninfected infants: clinical screening in a large prospective cohort
AIDS August 15, 2003; 17(12):1769-1785
Beatrice Barret a; Marc Tardieu b; Pierre Rustin c; Catherine Lacroix d; Brigitte Chabrol e; Isabelle Desguerre f; Catherine Dollfus g; Marie-Jeanne Mayaux a; Stephane Blanche h; for the French Perinatal Cohort Study Group *
Background: Antiretroviral prevention of mother to child HIV-1 is established but tolerance remains to be assessed.
Aim: To determine the risk for persistent mitochondrial dysfunction in HIV-uninfected children born to seropositive mothers.
Method: An exhaustive study in a large prospective cohort with predetermined algorithm of the unexplained symptoms compatible with mitochondrial dysfunction. A total of 2644 of 4392 children were exposed to antiretrovirals.
Complementary investigations were carried out on a case-by-case basis using classification with a diagnostic probability scale, based on experience with constitutional diseases. A spontaneous notification register for children
not included in the cohort was created.
Results: Good circumstantial evidence of mitochondrial dysfunction was found for twelve children. Seven were from the cohort. All presented neurological symptoms, often associated with abnormal magnetic resonance image (10 of 12) and/or a significant episode of hyperlactatemia (seven of 12). All had either a profound deficit in one of the respiratory chain complexes (11 of 12) and/or a typical histological pattern (two of 12). All were perinatally
exposed to antiretrovirals. None of them had perinatal morbidity that could explain this symptomatology. The 18-month incidence was 0.26% (95% confidence interval, 0.10-0.54) in exposed children, in comparison with the
general figure of 0.01% for paediatric neuro-mitochondrial diseases in the general population. Fourteen other children in the cohort, all exposed to antiretrovirals, had unexplained symptoms, mostly neurological, for which
one of the possible differential diagnoses was mitochondrial dysfunction. Close similarities in clinical, neuroradiological and histological findings strongly suggest a common pathological process in all these 26 children.
Conclusion: Children exposed to nucleoside analogues during the perinatal period are at risk of a neurological syndrome associated with persistent mitochondrial dysfunction.
The effectiveness of prevention of mother-to-child transmission of HIV-1 is now well established and represents one of the major successes of antiretroviral therapy. Tolerance to these agents is generally accepted as being good but remains to be accurately assessed. A long-term analysis of the first placebo-controlled study was reassuring but was based on a small number of mother-child pairs. Antiretroviral nucleoside analogues display mitochondrial toxicity, mostly due to inhibition of the enzyme gamma-polymerase. In a preliminary report we described eight cases of children presenting principally neurological symptoms compatible with persistent mitochondrial dysfunction. The symptoms and biochemical abnormalities were identified several months after
the exposure to antiretroviral drugs. Subsequently, a review of five different US cohorts failed to identify mortality specifically linked to mitochondrial disorder, but it is important to note that the review did not address possible
symptomatology in living HIV-uninfected children. Children born to HIV-infected mothers, even when they themselves are not infected, may show symptoms of various types. These symptoms are often associated with
easily identifiable causes, but in some cases are not readily explained. Investigators of the French paediatric study have for some years been interested in a series of symptoms, mostly neurological, of children born to HIV-infected women. The neurological symptoms include cognitive delay, behavioral disorders, motor abnormalities and convulsions.
Demonstration of drug toxicity during pregnancy is not easy when the suspected event is rare and the symptoms non-specific, especially when the beneficial effects of the drug are large. Only randomized studies can
unambiguously establish a causal relationship but the number of mother-child pairs available to investigators is often too small. Modelling of the denominator required to identify a significant 50-fold increase in mitochondrial dysfunction indicates that a sample size of 4300 mother-child pairs is required. If the effect is a 10-fold increase, 28 500 pairs are needed.
The neurological symptomatology of mitochondrial dysfunction is not specific and similar symptoms could be linked to frequently associated conditions such as prematurity, acute fetal distress, maternal drug addiction, cytomegalovirus infection or even very unfavourable psychological and social environments. Furthermore, sophisticated diagnostic procedures involving genetic, biochemical and histological analysis is required to affirm, strongly suspect or reasonably exclude mitochondrial dysfunction. The diagnosis of mitochondrial disease in children is sometimes considered to be very difficult and even arbitrary. Even blood lactate concentration is seen by some as being so subject to artifacts as to be impossible to interpret. However, these difficulties should not be exaggerated. Using a clinical approach and the diagnostic tools used by specialist teams, despite being imperfect, allow the molecular and genetic identification of an increasing number of constitutional mitochondrial diseases.
We report a systematic analysis of unexplained symptoms compatible with mitochondrial dysfunction in the French cohort of children born to HIV-seropositve mothers. The aim was to determine the incidence and risk factors of this phenomenon. This analysis was restricted to children not infected with HIV, whether or not they were exposed to antiretroviral drugs during the perinatal period. A level of probability for the diagnosis of mitochondrial
dysfunction was established according to more restrictive criteria than those used in our earlier report. During the study period, a national register was set up to collect similar observations in children not included in the cohort.
The French Pediatric Cohort (EPF)
The French prospective cohort (Enqute Perinatale Franaise, EPF) is a national prospective epidemiological study devoted to studying the risk and prevention of mother-to-child HIV transmission. According to unpublished estimates, the cohort covers about 70% of births to HIV-seropositive women in France. To be included, the mother-child pair must be recruited at or before the time of the birth, and the follow-up of the children is strictly prospective after birth. Each investigator has agreed to include all pregnant seropositive women and their newborns diagnosed at the centre except for cases of parental refusal. The children are followed up according to a common timetable and clinical and biological data are collected at regular intervals (birth, 1, 3, 6, 9, 12, 15 and 18 months and every 6 months thereafter). The choice of mode of delivery and of preventive antiretroviral treatment is left to the attending physician. There is no limit to the duration of follow-up of children infected with HIV. Children not infected were initially followed until the age of 36 months but, since 1993, have been followed until the age of 18 months.
Circumstances of the examinations. Expert groups
If mitochondrial dysfunction was a suspected differential diagnosis, specific investigations, adapted on a case-by-case basis to the symptoms, were suggested. The attending clinician was allowed to decide the extent of
these investigations: consequently the investigations performed varied with the severity of the symptoms and the clinician's evaluation of the pertinence of the mitochondrial hypothesis. Hematological and biological markers were
generally assessed at an early stage. One of the next lines of investigation of possible mitochondrial dysfunction was the measurement of lactate concentrations at several pre- and post-prandial time points, in optimal sampling conditions. Magnetic resonance imaging (MRI) of the brain was recommended in all cases of neurological
symptoms. The clinical and biological evaluations were organized for all children in one of the French centres specializing in neurological mitochondrial disease.
We set up three groups of experts (list in the Appendix) for the interpretation and classification of results: (1) a group of pathologists, for the interpretation of histological data; (2) a group of enzymologists, for the interpretation
of enzymological findings concerning the mitochondrial respiratory chain; and (3) a group of neuroradiologists and neurologists for the interpretation of cerebral MRI data.
We analyzed the appearance of mitochondria in muscle tissue obtained by biopsy (under local or general anaesthesia) from the deltoid or quadriceps muscle.
The prospective cohort. Screening overview
A total of 4426 children not infected with HIV-1 (n = 4072) or of undetermined HIV status (n = 354) from the French prospective cohort were included. By definition, the follow-up period for the identification of the first symptoms was restricted to 18 months. Data concerning the use of antiretroviral drugs were missing for 34 children; A total of 1748 of the children (mostly born before 1994) were not exposed to antiretroviral drugs during the perinatal period; 2644 received at least one antiretroviral drug during the pre-, per- or post-partum phase . Around half the children (n = 1311) treated received a standard zidovudine regimen and the other half received a combination of at least two nucleoside analogue drugs (n = 1315). The most frequently prescribed combination was zidovudine- lamivudine (n = 1230, 46.5%).
In the study period, 196 children presented at least one major sign or two minor signs on two different occasions as defined in the screening procedure. The individual re-analysis of each file, in collaboration with the investigator
concerned, clearly identified another cause that could account for the symptoms and/or resulted in no suspicion of mitochondrial dysfunction for 91 children. This group of 91 cases included 11 children who died during the study
period. For 61 children, the symptoms identified in the database, which were confirmed by the investigator, disappeared spontaneously: various investigations during the symptomatic period were not conclusive.
Complementary investigation after clearance of the symptoms was not considered justified. For 15 children, complementary evaluation was not possible (child lost to follow-up, parental refusal). Finally, for 29 children, the
symptomatology was considered compatible with a mitochondrial dysfunction.
For seven of these 29 children, complementary examinations resulted in a diagnosis of 'established' mitochondrial dysfunction, according to the predetermined restrictive criteria. For 14 other children, clinical presentation,
hyperlactataemia and/or histological findings allow mitochondrial dysfunction to be included in the differential diagnosis. Finally, for the remaining eight children, results from complementary investigations are not yet fully available and conclusions not yet possible.
National register for spontaneous notification
During the study period, five cases of children born to HIV-seropositive mothers not included in the EPF cohort, but fulfilling the criteria for 'established' mitochondrial dysfunction, were recorded in the register. Other children
reported to the registry potentially suffering mitochondrial dysfunction, but not fulfilling the restrictive definition of an 'established' mitochondrial disease, are not described here.
All the children in this group had one or several neurological symptoms. One child presented symptoms similar to those of Leigh's disease (motor abnormalities and brainstem symptoms) and nine displayed significant delay in
cognitive development. Five children suffered at least one seizure and six presented diplegia, tetraplegia or other motor abnormality. One child had a myopathy, myocardial dysfunction and presented several episodes of severe
malaise. Seven of the twelve children presented with significant hyperlactataemia on two occasions outside the treatment period. There were only two deaths (previously reported ) in this group. All children were born to mothers infected by the sexual route. Only one was a premature birth (before 32 weeks of gestation) and only one suffered acute fetal distress which rapidly resolved.
All but two of the children had an abnormal cerebral MRI. The most prevalent abnormality was a hypersignal from the supratentorial white matter (n = 9). These lesions were diffuse (n = 8) or localized in the posterior region (n =
1). In six cases, there were abnormal signals from the brainstem (hypersignal of the tegmentum pons) and in four cases abnormalities of basal ganglia. Six of the ten abnormal MRI showed multifocal lesions. Atrophy was observed in five children and necrotic areas in two.
Studies were carried out either on skeletal muscle (n = 11), liver (n = 2) or cardiac muscle (n = 1) biopsies or circulating lymphocytes (n = 6). Enzymological study of the respiratory chain showed a significant deficit in at
least one of the tissues or cell types in eleven children: complexes I and IV deficiency (n = 5), complex IV deficiency (n = 3), complex I deficiency (n = 2) and complex III deficiency (n = 1). As in the initial report, we detected no significant depletion of mitochondrial DNA in any of the children newly identified (data not shown).
Histology and electron microscopy
Muscle biopsies were performed for eleven of these twelve children. Electron microscopy data was missing for one child. For two children (one of them with normal enzymological results) the histological findings were typical
of mitochondrial dysfunction: red ragged fibres in one, massive subsarcolemnal accumulation of abnormal mitochondria in the other. All except one of the others had at least one minor histological finding compatible with mitochondrial dysfunction. In one patient, the ultrastructural picture showed demyelinated small muscle nerve fibres, together with images of remyelination
processes. Five children had enlarged endothelial capillary cells with abnormal mitochondria.
Children from the cohort with 'possible' mitochondrial dysfunction
This group included 14 children with unexplained clinical and/or biological symptoms for which a mitochondrial dysfunction could be included in the differential diagnosis. According to the definition used in this analysis none of them had severe enzymological deficits affecting the respiratory chain or
histological-ultrastructural evidence of mitochondrial disease. However, this group presented close similarities with the established mitochondrial disease group: most of the children (11 of 14) presented neurological symptoms; seven of 12 had abnormal cerebral MRI findings, notably diffuse hypersignal of the white matter (n = 7) or the brainstem (n = 3),. The suggestion of possible mitochondrial dysfunction was based on persistent hyperlactataemia (n = 11 of 14) and/or minor histological or ultrastructural anomalies (seven of eight biopsies). Enlarged endothelial capillary cells with abnormal mitochondria were also observed in four of eight biopsies.
Risk factors for mitochondriopathy in the prospective cohort
The study of risk factors was limited to children included in the prospective cohort, who were followed similarly and for whom the denominator was known. All the children with 'established' or 'possible' mitochondriopathy diagnosed in this study had been exposed to antiretroviral drugs. One of these children was treated with zidovudine only during the prenatal period and received no treatment after birth. For the other children, the treatment was administered in the pre-, per- and post-partum periods. It was zidovudine alone in five cases, a combination of zidovudine-lamivudine in 14 cases and another combination in one. Twenty of the mothers received zidovudine by intravenous perfusion during labor. For six children, other molecules were also administered during the pregnancy. The total (pre and postnatal) mean duration of preventive treatment was 24.6 (SD 14.1) weeks which is not significantly different to the values for the entire cohort: 25.3 (SD 12.0); P < 0.8. The difference in incidence of mitochondrial dysfunction whether established or possible between the exposed and non-exposed groups was significant (21 of 2644 versus 0 of 1748; P < 0.002). The difference remains significant if restricted to the seven cases of established mitochondriopathy (P = 0.047). The 18-months incidence for 'established' mitochondrial disease in the prospective cohort can be estimated to be least 0.26% [95% confidence interval (95% CI), 0.10-0.54] for exposed children. Including both children with possible and with
established mitochondriopathy from the cohort, the risk associated with a combination of nucleoside analogues (n= 15) is significantly higher than that associated with zidovudine (n = 6) monotherapy: RR = 2.5 (95% CI,
1-6.5), P = 0.046. The mode of infection, the geographical origin (African versus European) of the mother and prematurity were not linked to the risk of apparition of these symptoms (data not shown).
We report a systematic detailed re-analysis of the unexplained symptoms presented by uninfected children born to HIV-seropositive mothers prospectively followed in a large cohort. The finding that the use of antiretroviral nucleoside analogues in the perinatal period is associated with persistent mitochondrial disease is confirmed. Of children exposed to nucleoside analogues, 0.3% present, by age 18 months, clinical, radiological,
histopathological and biochemical symptoms similar to those observed in some children with congenital mitochondrial disease. Despite active screening, no similar cases were found in the antiretroviral unexposed group. The diagnostic procedure was performed in three different paediatric neurological centres specializing in mitochondrial diseases. No other usual cause of neurological symptoms were identified, and in particular, none of these children was exposed to illicit drugs during pregnancy, suffered acute fetal distress at birth or had fetal cytomegalovirus or toxoplasmosis infections. The incidence of persistent mitochondrial disease in the group exposed to antiretroviral treatment is substantially higher than that observed in the general population. Two recent studies in northern Europe report attempts to estimate the incidence in the general population of paediatric mitochondrial disease with neurological expression. The results of these two studies are highly
concordant, with estimations of about one case per 10 000 children. The criteria used by these two studies to define mitochondrial disease are very similar to those we used here. Uusimaa et al. over a period of 7 years
identified 17 children with diverse unexplained neurological symptoms associated with mitochondrial respiratory chain enzyme deficit and various ultrastructural anomalies (including those we classify as minor) in a region of
Finland with a population of about 150 000 children under 18 years of age. In a population centre in Sweden with about 360 000 children under 16 years of age (35 000 born annually), Darin et al. between 1984 and 1998 identified 32 children with diverse unexplained neurological symptoms associated with a significant enzyme deficit and ultrastructural and/or enzymatic anomalies, similar to those we describe. The identification of seven such cases in a cohort of 2500 children thus corresponds to a risk about 30 times higher than that in the general population. The true incidence could be higher because additional children in the cohort have similar symptomatology, and several arguments strongly suggest a mitochondrial origin for these cases. The notification
of similar observations in a national register is also interesting, and suggests that such cases are not limited to a particular network or group of investigators. The number of these notified cases from outside the cohort is also consistent with the incidence calculated for the cohort, which includes almost 70% of births to HIV-seropositive women. According to incidence observed in the general population, and considering that approximately 6000
children have been born to seropositive mothers in France since 1994 (two-thirds included in the French cohort), no more than one or two children with constitutional mitochondrial dysfunction would be expected. Comparison
of the incidence of this dysfunction in the treated and untreated groups of the cohort is open to criticism because these two groups essentially correspond to children born in two different periods (before and after 1994). The follow-up of these two groups within the same cohort has nonetheless been highly uniform and the detection of neurological symptoms, whatever their cause, has been one of the concerns of this survey from its inception. Moreover, the difference observed between monotherapy and combination therapy is not influenced by this potential bias and supports the toxicity hypothesis.
As the number of children described increases, a coherent syndrome is appearing with three main features: neurological symptoms (principally developmental retardation, seizures and behavioral disturbances), significant
abnormalities on cerebral MRI (principally lesions of the white matter and brainstem) and often hyperlactataemia either persistent or transient outside the treatment period. These three elements are consistent with those described in constitutional mitochondrial diseases with neurological expression. However, toxic-induced mitochondrial dysfunction may not necessarily match exactly the symptomatology, severity and evolution of constitutional diseases. In some children, the symptoms are very strongly expressed. In others, the symptoms are mild, and only a specific and adapted program of complementary examinations can diagnose or suggest the existence of mitochondrial toxicity. A profound deficit of one of the components of the respiratory chain - principally complexes I and IV - and/or typical mitochondrial histology findings of abnormality were the required characterisitc for diagnosis, in conditions similar to those accepted for constitutional mitochondrial disease. However some children with compatible clinical and radiological symptoms did not fulfill these strict criteria. In these cases, hyperlactataemia persisting after the withdrawal of treatment and 'minor' abnormalities (patchy mitochondrial labelling, excess lipids, or mitochondrial abnormalities at ultrastructural levels) were considered
suggestive of a mitochondrial origin of the symptomatology. Although we intentionally accorded only a moderate weight to these histological anomalies, and considered them 'minor', the frequency with which they were
observed in the cohort is interesting. Unlike in adults and particularly the elderly in who such findings are not specific, they are much more significant in children. The typical red ragged fibre aspect, in contrast, is more rarely
found associated with consitutional mitochondrial disease in children than in adults. Interestingly, images of mitochondrial vessel abnormalities in muscle were identified in half of biopsies of both groups by histoenzymology and/or ultrastructural examination. The extent of this vasculopathy in other organs and notably in the brain remains to be determined. Cerebral MRI showed a similar pattern of white matter and brainstem abnormalities in the majority of children in both groups. Although they are not specific, these images suggest a
common pathophysiological process.
The reason for this persistent mitochondrial dysfunction after the withdrawal of treatment is unknown. First described as a myopathy associated with zidovudine, the issue of mitochondrial toxicity of nucleoside analogues is currently a growing problem. Its clinical expression is highly variable, from peripheral neuropathy to severe lactic acidosis. The central nervous system does not seem to be affected, but central neurological toxicity was recently reported in a child infected with HIV and treated with multidrug therapy. Transplacental passage of nucleoside analogues such zidovudine or lamivudine is high and fetuses and newborns exposed, sometimes for several months, to the drugs must therefore also be exposed to their effects. Morphological and
functional abnormalities of mitochondria have been detected in newborn monkeys after exposure to zidovudine in utero but the clinical significance and the reversibility of these anomalies are unknown. In vitro tissue
studies show that the major physiopathological factor is inhibition of the mitochondrial gamma-polymerase, causing substantial depletion of mitochondrial DNA. This effect could be expected to be reversible following the withdrawal of treatment, and clinical expression of persistent mitochondrial dysfunction after the end of treatment has not previously been described. The diversity of the pattern of biochemical defects should be examined, and investigations are needed to explain why there is no evidence for substantial depletion of mitochondrial DNA. Indeed, the biochemical phenotype is presumably not determined by this parameter (i.e. residual mtDNA). One possibility is that that there is either transitory mtDNA depletion favouring oxidizing stress conditions in the mitochondrial matrix or insertion of analogues into the mtDNA. This may lead to mutations scattered unevenly in the mtDNA which in turn cause the reported mitochondrial anomalies. The extent of clonal expansion of these various mutations would then determine the final and variable biochemical phenotype. Noticeably, even in cases of a sustained mtDNA depletion, a certain degree of heterogeneity in the biochemical phenotype would be predicted by this model. Thus, total depletion of mtDNA, leading to a total lack of the tRNAs required for translation would result in generalized deficiency of all complexes involving mtDNA-encoded subunits. A partial depletion would result in multiple defects variably affecting the different respiratory chain complexes, according to the respective turnover/availability of the mRNAs encoding the various subunits, and the
stringency of the control exerted by these subunit on the electron flow through each of the RC complexes.
It may seem surprising that observations of this type have not yet been reported outside the French cohort. The retrospective analysis of the causes of death in five American cohorts failed to identify observations consistent
with mitochondrial dysfunction. However, the re-analysis of the American data did not address possible mitochondrial morbidity in living children. Independently of the potential effect of the perinatal treatment, few
children with constitutional mitochondrial disease would be identified in this large group of children. Among all the children we describe, there have been only two mitochondria-related deaths and detailed evaluation of morbidity among HIV- uninfected, living children in the various cohorts is warranted.
It is clear that the number of children worldwide who suffer because they have not received antiretroviral treatments is inestimably larger than the number of children who suffer due to the toxicity of these treatments. The use of zidovudine during pregnancy is undoubtedly one of the outstanding successes of antiretroviral treatment. Nonetheless, we feel that it is necessary to define more precisely the tolerance profile of nucleoside analogues on fetal michondria. Although not easy to extrapolate to clinical medicine, several in vitro experiments show that toxicities of the various nucleoside analogues available are different. Additive or synergistic toxic effects have also
been demonstrated. Many antiretroviral molecules of different therapeutic classes are now available and it is very plausible that certain molecules or combinations of molecules will be better tolerated than others by the fetus