Although mitochondrial mosaics after COX staining have been known in skeletal muscle and myocardium since 1983, only a single case has been reported in liver .
COX histochemistry can be carried out within 24 hrs and is easy to perform. In our patients it yielded evidence, already by light microscopy, of the underlying pathological mechanism, i.e. partial defects in COX activity, whereas tests in blood and muscle were normal or not helpful. The COX deficiency directed subsequent analysis towards blue native electrophoresis of the OXPHOS proteins and to mtDNA.
It is striking that COX activity in skeletal muscle cells, visualized also with DAB, had a normal pattern. Non-parenchymal cells in the liver appeared not to participate in the mosaics, which were limited to the hepatic parenchyma, an observation also made by Müller-Höcker et al . But considering the clinical symptoms it is likely that brain cells, exocrine pancreas and bone marrow had COX deficiency as well. Mild brain atrophy on MRI and psychomotor retardation were also present in the patient of Müller-Höcker et al.
When interpreting the absence of, or very weak stain in single mitochondria at ultrastructure, it should be recalled that the tissue was prefixed in aldehyde which inhibits part of COX activity. One can speculate therefore that liver cells devoid of DAB stain, were actually not entirely without COX activity. Indeed it would be difficult for a cell without any COX to survive for long.
How can the cell- and tissue heterogeneity of the COX activity be explained?
Many disorders have been linked to heteroplasmic alterations in mtDNA: in muscle , liver , cardiomyocytes and other tissues. Clinical syndromes associated with heteroplasmy are several types of myopathy [5, 6]; myoclonus epilepsy and ragged red fibers (MERRF) ; progressive external ophtalmoplegia, childhood optic atrophy, followed by deafness and ataxia later ; dilated cardiomyopathy ; multiple lipomas ; Kearns-Sayre syndrome (MIM 530000)[34, 35]; focal segmental glomerulosclerosis ; early onset diabetes mellitus, optic atrophy and deafness (Wolfram syndrome); Alpers-Huttenlocher-like disease ; and Pearson syndrome with Kearns-Sayre encephalomyopathy .
The histochemical COX staining in individual muscle fibers appears to be linked to the expression of a mutation in mtDNA  and heterogeneous defects in COX activity may be due to differing populations of wildtype and mutant mtDNA (. Hudson et al  showed that COX-positive vs. negative fibers in muscle are correlated to their level of mtDNA deletion.
In patient 1, the observed deletion in the mtDNA that encodes for the complex IV protein, might be limited to part of the liver parenchyma if this mutation was generated early during liver histogenesis in a single cell hitting only its descendants. But also blood cells bear the mutation (it was found there). There is no simple model for a common origin of both celltypes, blood cells and (part) of the hepatocytes; less so for the neurons and most glia in the brain some of which must have low COX activity.
A different mechanism proposes that the mutation intervened early in development, possibly even in the oocyte, that both mutated as well as wild-type mitochondria proliferated but were distributed at random in the daughter cells, and thus not equally in every cell. This results in varying proportions of normal and defective mitochondria per cell, so-called "shift", or "drift" of the genotype . Larsson et al  concluded that "the phenotype can change with time and is governed by fractional concentration in different tissues of the mtDNA with the deletion". Chinnery et al  think it is likely that the progression in time of the COX defect in muscle "is due to clonal expansion of mutant mtDNA". The frequently observed sequence in Pearson syndrome of early hematopoietic dysfunction followed later by lactic acidosis, pancreas dysfunction and symptoms of Kearns-Sayre syndrome, has been tentatively explained by "chance distribution of the deleted molecules among the tissues through redistribution" . COX-negative, but also ragged-red muscle fibers (absent in our case) were found in a 8-year old boy who combined the symptoms of Pearson and Kearns-Sayre syndromes . Moreover, we speculate that cell death, ensuing each time COX activity falls below the survival threshold, will trigger compensatory hyperplasia in those tissues that are capable of regeneration, and enable an enrichment with COX-positive cells, i.e. among liver cells of all types, lymphocytes and bone marrow cells, and, in young individuals, even brain cells: glia no doubt, and perhaps even neurons. In neurons proliferative capacity quickly regresses with age; this might explain why the neonate was symptom-free but subsequently developed progressive lesions leading to death.
That cultured fibroblasts of our patients gave normal tests is not unexpected: the phenotypic expression of respiratory chain disorders is unstable in cultured cells and tends to return to normal values [38, 40]. In their recent review on disorders of mtDNA synthesis, Freisinger et al. state that fibroblasts are not suitable for assays of depletion . A possible explanation is that cultured cells originate from mothercells that have wild-type mtDNA while the mutated cells die or do not proliferate.
A mosaic of COX activity in muscle has also been observed in a case of a homoplasmic mutation of mtDNA ; this is more difficult to explain.
On the other hand patients 2 and 3 had a large depletion of the amount of mtDNA.
Depletion syndromes are associated with nuclear mutations of DGUOK , SUCLA2, SUCLG1, TK2, MPV17 , PEO1 , RRM2B, and POLG; the last is the case in two of our patients. Mutations in the POLG gene were previously reported in AHS by three groups [17–19, 45–47] and most recently by Stewart et al.  who describe 9 patients with a COX mosaic in muscle, but none in liver.
How can one explain a mitochondrial mosaic linked to a nuclear mutation?
Comparing muscle of two patients Moraes et al  found a correlation between the degree of mtDNA depletion and COX activity. Since the reduction of mtDNA in cell cultures results in gradual loss of respiratory rate and COX activity , the question can be rephrased as follows: why is the level of depletion varying between individual mitochondria, individual cells, and tissues?
Models for tissue specificity of the expression of nuclear mutations impairing the enzymes required for mtDNA synthesis are proposed by several groups; they are based on interactions of distinct gene products required for this synthesis, enzymes as well as metabolites.
In muscle the dNTP transporter securing import of cytosolic deoxyribonucleotides for mtDNA synthesis, is normally low. Partial loss by a mutation of the mitochondrial enzyme thymidine kinase-2 will result in critically low levels of DNA. In contrast other celltypes where more dNTP transporter is available will be spared although the mutation is present [50–53].
In brain a mutation of succinyl-CoA synthetase-A (required for ADP synthesis) combined to a normally very low mitochondrial NDPKinase (that phosphorylates ADP) will again result in shortage of precursors and thus depletion of DNA . Other tissues have more dNDPK.
The cytosolic deoxycytosinekinase has an overlapping substrate specificity with deoxyguanosinekinase. But in liver and brain it is normally very low; a mutation of deoxyguanosinekinase cannot be compensated and will result in DNA depletion limited to those tissues  or more severe than in muscle . Vice versa, Taanman et al. have shown that supplementation of dGMP and dAMP to fibroblasts can prevent mtDNA depletion . Hepatic and brain-specific phenotypes are known in mutations of deoxyguanosine kinase , and reversal of the phenotype was also described .
Sarzi et al. discussing two sibs with a hepatocerebral syndrome and a TWINKLE mutation, suggest that tissue specific differences of nucleotides as well as the normally higher activity of helicase in muscle compared to liver explain the organ specificity of the disease .
Galassi et al., discussing a patient with one mutation in POLG-A and a second in ANT1, mention the possibility that the complex clinical features "may be due to cooperative effects consequent to the interrelation of two mitochondrial functions both converging on the homeostatic control of mtDNA maintenance and stability" .
Another mechanism for variability in phenotypic expression and possibly for tissue- and cell-specific deficiency is reversal of the mutation in the TK2 gene that caused mtDNA depletion .
When liver is compared to muscle, in 4 unrelated children with AHS, there was respiratory chain enzyme deficiency in liver, but not in muscle . In a recent series of 10 children with mutations in POLG1, 5 had normal enzymes in muscle . The authors conclude that the assay of mitochondrial enzymes in muscle alone is not effective. Mutations in DGUOK causing mtDNA depletion nearly always result in normal muscle enzymes, and the study of liver is recommended . The patients with the mutation in PEO1 had a severe reduction of OXPHOS complexes in the liver and not in muscle . In a large series of patients with mtDNA depletion and mutations in DGUOK, POLG, MPV17 or TK2, 17 cases had lowered respiratory enzymes in liver but normal activities in muscle .
Given the frequency of tissue heterogeneity, mitochondrial mosaics in liver may be more frequent than reported until now.
Whether mosaics after COX staining of the liver were present in these cases from the literature has not been studied.
Finally it is tempting to speculate on the relationship between clinical severity and proportions of COX deficient mitochondria in the mosaics. In patient 1 by far the smallest ratio of stained/unstained mitochondria in the liver coexisted with the severest brain lesions on MRI; on the other hand liver transaminases were unremarkable. Hepatic disease was more pronounced in the AH patient, as estimated by the liver enzymes AST, ALT and gGT. The 3rd patient's liver, with 10% of the hepatocytes with reacting mitochondria, had already developed cirrhosis at 3 months of age, but cerebral symptoms appeared only 3 years later. When taking into account the extraordinary regenerative capacity of liver parenchyma, and the near absence of such capacity in the brain, we see as yet no simple correlation between disease severity and the observed distribution of COX deficiency in the liver.