Evidence of reactive oxygen species-mediated damage to mitochondrial DNA in children with typical autism

The human mitochondrial genome is a 16.5-kb circular, double stranded DNA that encodes thirteen polypeptides of the mitochondrial respiratory chain, twenty-two transfer RNAs and two ribosomal RNAs required for protein synthesis. The mitochondrial DNA (mtDNA) consists of a heavy (H) and a light (L) strand, in accord with its G and T base composition. mtDNA is particularly susceptible to mutations because of the high level of reactive oxygen species (ROS) (including superoxide anion, hydrogen peroxide, hydroxyl radical and peroxynitrite) generation in this organelle [1, 2], and lack of introns or histones, coupled with a low level of DNA repair [3].

Damage to mtDNA, elicited by ROS continually generated in mitochondria, may result from defective replication and/or repair of mtDNA of primary (genetic) [4] or secondary (for example, oxidative damage to single base pairs inflicted by ROS) [5, 6] origins. ROS-mediated damage is characterized by a variety of lesions to DNA in general [7, 8] and to mtDNA in particular [9], including single- and double-stranded DNA breaks, abasic sites, and oxidized bases (reviewed in [10‐13]). Considering that mtDNA replication occurs as a coupled leading and lagging strand replication pathway, the H strand DNA exists for extended periods in the single-stranded form in this asymmetric mode of mtDNA replication, favoring damage to the exposed bases not protected by the complementary DNA strand [14]. In this regard, mtDNA damage is asymmetric resulting in the majority of single and multiple mtDNA deletions occurring between the O_H and O_L[15]. In addition, cytosine deamination to uracil would be expected to be an asymmetric process because it occurs > 100-fold more rapidly in single-stranded DNA, as it would be when mtDNA is temporarily exposed during ongoing replication and transcription [14]. Increased oxidative stress has been claimed as one of the features present in autism [15‐19], although its precise role in the etiology of autism is still undefined. In support of this hypothesis, higher rates of mitochondrial hydrogen peroxide production (accompanied by lower activities in the pyruvate dehydrogenase complex, Complex I alone, or in combination with other Complexes) from lymphocytic mitochondria [15] and increased markers of oxidative stress (for example, increased oxidized glutathione) [19‐22] had been reported in samples from individuals with autism.

There are limited studies characterizing mtDNA from children with full autism, and none evaluating the putative ROS-mediated damage to mtDNA in the index child. We evaluated the occurrence of ROS-mediated damage to mtDNA in children who met the criteria for presenting full-syndrome autism and age-matched and genetically unrelated typically developing children (TD) without a clinical diagnosis of autism or developmental delays, recruited by the CHildhood Autism Risk from Genes and Environment (CHARGE) Study at the University of California (UC) Davis [23]. To minimize invasive procedures, we evaluated the quality of mtDNA (deletions and sequence variants) in peripheral blood monocytic cells (PBMC) from children with typical autism to ascertain if (i) the diagnosis of autism segregated with a pattern consistent of increased oxidative damage to mtDNA and (ii) if the putative mtDNA damage was de novo (when compared to their parents living in the same household) or inherited (if outcomes followed paternal and/or maternal patterns), when compared to parents of TD or AU children. To our knowledge, there are no systematic and comprehensive studies aimed at investigating the role of ROS-mediated damage to mtDNA in children with autism.

Deletions in mtDNA of TD and AU children were evaluated by qPCR using the mitochondrial gene ratios of CYTB/ND1 and ND4/ND1. The percentage of TD children (n = 46) with deletions (deletion = Z-score < −2SD) encoding for CYTB and ND4 was 8.7% and 6.5%, respectively (Table 1). In samples from AU children (n = 67), these outcomes were significantly higher by 2.4- and 2.3-fold, respectively (Table 1). In both groups, TD and AU children, the frequency of deletions at genes located closer to O_H (CYTB) relative to those located closer to the O_L (ND4) was 1.3- and 1.4-fold, respectively, with no difference between the groups. The higher incidence of individuals with CYTB deletions vs. ND4 ones was also observed in all parents, regardless of sex or diagnosis of child (Table 1, last row). This strand asymmetry of mtDNA deletions was suggestive of ROS-mediated damage to the single-stranded state of the H strand during the asynchronous mtDNA replication.

The extent of the mtDNA deletions at CYTB in AU children was 14 ± 1%, 1.5-fold greater than the corresponding TD values (9 ± 1%, P < 0.005) and was similar to that of older individuals, regardless of sex or the diagnosis of the child (16 ± 2%, P < 0.01).

To discern between de novo (acquired) vs. inherited deletions (from either maternal mtDNA or parental gDNA-inherited mechanisms that favor accumulation of deletions in the mtDNA), deletions in both segments of the mtDNA were evaluated in the parents of TD and AU children. The percentage of fathers of AU children with mtDNA deletions at the segments encoding for CYTB and ND4 was higher than for those of TD children (1.4-fold and 1.9-fold respectively), following the pattern of AU children when compared to TD children for both genes. In contrast, mothers of AU children presented a low incidence of deletions at both segments when compared to mothers of TD children (50%) (Table 1) suggesting lower mtDNA replication.

The ratio of parents of TD children to TD children with mtDNA deletions was 1.2-fold in the segment encoding for CYTB (Table 1). This increase suggested an age-dependent accumulation of mtDNA deletions. Age-dependent mtDNA deletions are thought to accumulate at lower levels in human mitotic tissues than post mitotic ones because they could be lost by negative selection in rapidly dividing cells, by a higher tendency of the cells to undergo apoptosis, mitochondria being able to replicate independently of the cell cycle (avoiding being diluted out by cell division) and/or by preferential replication of deleted mtDNA [35].

To determine sequence variants that were ascribed to ROS-mediated damage in mtDNA from AU and TD children, 31.2% of the entire human mtDNA was sequenced encompassing the following protein-encoding genes: ATP6/8, ND1 and ND2, and parts of CYTB, COX2 and COX3. A relatively higher frequency of transition mutations has been reported in aerobic organisms (from 60% to 68%) [36, 37] consistent with the transitions observed in this study for both children regardless of diagnosis (Table 2). Transitions (90% of base variants) affecting GC and AT pairs in TD children were 44.4% and 55.6% (Table 2). This pattern of base substitutions (90.0% transitions with 55.6% of them converting an AT to a GC pair) resembles that of the human mtDNA sequence polymorphisms reported in the MITOMAP database, which shows an enormous preponderance of transitions (88%) over transversions, with a nucleotide bias approximating that of the genome (that is, the genome is 56% A + T, and 58% of the transitions A-T→G-C). In AU children, transitions affecting GC and AT pairs were 64.7% and 35.3% indicating that CG pairs were replaced 2.5 times more often than AT pairs, which was an opposite trend to TD children (Table 2). The percentage of (G,C,T)→A transitions to the total number of transitions in AU children was 1.6 times that in TD children (Table 2).

In this study, and following the model of strand asymmetry replication of mtDNA, evidence for mtDNA damage has been observed in all individuals, regardless of age and/or diagnosis. The higher frequency of deletions at the segment encoding for CYTB compared to ND4 (Table 1, last row) attests for an asymmetric damage of the mtDNA when the H-strand is exposed in the single-stranded state during replication.

If the hypothesis that higher oxidative damage to mtDNA was occurring in a more exacerbated form in PBMC from AU than from TD children, then the following results would be expected: (i) higher percentage of GC transitions over AT ones because when the well-known marker of oxidative stress, 8-oxo-7,8-dihydro-2′-deoxyguanosine (8oxodG)[1, 38, 39], is effectively repaired and/or removed from the template [40], inducing predominantly G→T transitions by mispairing with A during DNA replication [41‐43]; (ii) higher frequency of mtDNA deletions, especially at the segment encoding for CYTB compared to ND4; and (iii) higher frequency of G,C,T→A transitions because when Polγ bypasses apurinic/apyrimidinic (AP) sites, formed as a result of spontaneous deamination or oxidative lesions, incorporating mainly adenine at these positions [44].

In favor of the model of higher oxidative stress in autism, a higher frequency of GC transitions over AT ones (2 vs. 0.8) (Table 2), higher frequency of deletions (by 2-fold) (Table 1), and higher number of G,C,T→A transitions (1.3-fold those in TD children) (Table 2) were observed in AU children. The lack of G→T transitions could be explained, considering that these types of transitions are rarely observed in both in vivo and in vitro somatic sets of mtDNA point mutations [45, 46]. The extent of the deletions in AU children was 1.6-fold of that in TD children, and was similar to that of all parents, suggesting more damage to their mtDNA. However, given that the percentages observed in AU children were similar to those observed in older individuals in general, and that these percentages are usually much lower than are seen in patients with mitochondrial disorders, in which deletion ≥ 60% is required to demonstrate a mitochondrial defect [47, 48], it is suggested that the extent of the deletions do not seem to be pathogenic per se.

The sequence variants observed in all children were explained by the maternal inheritance of mtDNA, regardless of diagnosis (Table 2). However, given that these outcomes were different between TD child-mother vs. AU child-mother, it suggests that mothers of AU children share a DNA template consistent with a model of higher oxidative stress-mediated damage. It is interesting to note that mothers of AU children, although having more damaged mtDNA, also presented the lowest incidence of deletions when compared to age-matched groups. This might indicate a compensatory mechanism, by which a lower replicative rate might prevent additional accumulation of deletions.

Considering that a higher percentage of AU children exhibited mtDNA deletions compared to TD children, and that this pattern was also present in fathers of AU children, it is likely that a genetic predisposition to accumulate mtDNA deletions was transmitted paternally. It should be noted that paternal mtDNA deletions are not inherited but, accumulation of deletions (or the predisposition to accumulate deletions) resulting from increased ROS production, defective antioxidant/repair system, or defective clearance of damaged mitochondria, could be transmitted from either parent. Alternatively, exposure to epigenetic factors different from those to which families of TD children are exposed, or identical to those of TD families but perceived with a different genetic susceptibility [49], may have resulted in the increased mtDNA deletions observed in AU children and their fathers.

This study has several limitations that need to be considered for a proper interpretation of the results and consequences for the field of autism. First, the number of individuals on which mtDNA sequencing was performed was relatively small, although significant differences were observed between TD child-mother and AU child-mother. Second, the comparisons made in this study reached a significance at the α = 0.001 level minimizing type I errors, whereas type II errors were reduced by increasing the number of observations per group, limited only by the availability of samples. Third, children in this study had not been previously diagnosed with a genetic syndrome, nor had any indications of genetic syndromes been identified by M.I.N.D. developmental pediatricians. Nevertheless, defects (other than deletions) in genes other than those tested could have been present in these samples as recently reported in other studies of ASD [50]. Fourth, although the outcomes reported here for PBMC may represent those present in other cells more relevant to autism (for example, neurons), it is important to consider that the neuroimmune response is characterized by cross-talk between peripheral immune cells and the central nervous system, and that disruption of this process during early life may condition inflammatory responses as well as behavioral changes that persist during adulthood [51‐54]. Finally, inferences about a cause-effect association between ROS-mediated mtDNA damage and typical autism are intricate because this is cross-sectional, and not a longitudinal study. In addition, several factors influence expression of mtDNA damage, for example, nuclear genetic backgrounds [55], mtDNA heteroplasmy in tissues [56], energy thresholds for a given tissue/organ [57], and epigenetic factors [58], in both affected and general healthy populations. Multiple mtDNA deletions for example, may accumulate with age in post-mitotic tissues of apparently healthy individuals [12, 48, 59‐61], or in patients with other disorders not necessarily linked to autism, such as, inherited mutations in nuclear genes [62, 63], neurodegenerative disorders [62, 64], cancer [65], and diabetes [66]. Nevertheless, our study showed that mtDNA in children with autism is more damaged than in age-, sex-, and race-matched TD children, and is more similar to that of older individuals, with a mtDNA template (maternally inherited) consistent with ROS-mediated damage (based on sequence variants), and presenting a predisposition to accumulate damage (deletions) similar to that of their fathers.

Several scenarios may result in increased mtDNA damage, among them: (i) higher oxidative stress not accompanied by antioxidant defenses and/or repair/maintenance of mtDNA; (ii) lower ability to clear mitochondria with damaged mtDNA; (iii) replicative advantage of deleted mtDNA over wild-type mtDNA, or (iv) a combination of any of the aforementioned possibilities.

The fact that mothers of AU children have a template with an array consistent with increased mtDNA damage, and that fathers of AU children accumulate more deletions than fathers of TD children, seems to point to a combination of these factors in addition to the age of onset for these events. In this regard, changes in mtDNA copy number and/or deletions seem to be age-dependent, for the clinical onset of mtDNA depletion is typically in infancy or early childhood whereas multiple deletions seldom present before adolescence. A genetic background, in combination with other genetic (originated from parental genomic DNA or maternal mtDNA) and/or epigenetic factors, for example, dysfunctional electron transport chain [15], low levels of antioxidant enzymes bound to mtDNA [67], environmental factors [49] and dietary deficiencies [68], acting additively or synergistically may lead to more damaged mtDNA during vulnerable windows such as the perinatal periods. This altered process may be causative per se, or may set up the stage for a heightened susceptibility for further insults, which may ultimately alter an appropriate development of energy status and increase autism risk. Structural instability of mtDNA, consisting either of large-scale rearrangements, tissue-specific depletion or deletions, is a major cause of mitochondrial dysfunction and disease in humans [4], and possibly in children with autism.