The observation of FAH mosaicism in the liver of a HTI patient led us to suspect that reversion of one of the mutations might have occurred, as reported in patients with other mutations [9, 16]. This proved to be the case as confirmed by the mutation detection test in different regions of the liver. The heterozygocity pattern found in the tumoral region was no longer present in the non-tumoral normal looking section of the patient liver. The site-specific reversion of the Q279R mutation in this region was also demonstrated by direct sequencing. This event leads to both the production of normal sized FAH mRNA and expression of functional FAH, as shown by western blot and enzymatic activity measurements. FAH positive cells have been reported to show a selective growth advantage in vivo. Overturf et al.  showed that the introduction of as few as 1000 normal or corrected hepatocytes in a FAH knock-out mice model led to an extensive repopulation of the liver. Therefore, in man, a reversion event might occur in a cell, which expands and forms FAH expressing nodule. The amount of FAH being expressed by those reverted cells appears to be sufficient to reduce the symptoms of HTI. Kelsey et al.  showed that the expression of levels as low as 2% of the FAH level in normal liver was sufficient to rescue the lethal phenotype in the albino mouse model of HTI.
In the present patient, cancer is only present in non-FAH expressing cells. In a mouse model of HTI where the mouse lethal phenotype was rescued by gene therapy with either retroviruses or adenoviruses harboring FAH, tumors and dysplasia were always observed in FAH negative areas [19, 21]. This is consistent with studies showing that the toxic metabolite in HTI, FAA, is indeed mutagenic  and can induce mitotic abnormalities and genome instability . FAH-negative cells may be more prone to cancer development since FAA can accumulate in these cells while restoration of FAH activity by mutation reversion should remove this toxic metabolite in the reverted cells thereby reducing the risk of carcinogenesis.
The present data show that a dysfunctional allele would not be transcriptionally silenced in favor of a functional one since in the reverted region of the liver, the IVS6-1g->t transcripts are still detectable. As shown here, the normal size transcript produced due to reversion of the Q279R mutation seems to be sufficient to allow FAH expression in the non-tumoral liver nodules.
The Q279R mutation has recently been identified in the patient and her father , as a missense mutation. However, the complex pattern of transcripts obtained by RT-PCR in different regions of the liver reported here suggests that the Q279R mutation could act as a splicing mutation in vivo. In fact, we identified at least six alternative transcripts in the tumoral region and in the patient's fibroblasts (Figures 2 and 3). Four of these are produced as a result of the IVS6-1g→t mutation; among these, three are identical to those previously reported in other patients harboring this mutation . We also identified an unreported high molecular weight transcript, with a retention of intron 6, as a minor product. The Q279R mutation seems to be responsible for the other two transcripts. mRNA analysis in the compound heterozygous patient also revealed that this exon 9 mutation partially affects the splicing at the 5' splice site both in vivo and in transfected HeLa cells, whereas the IVS6-1g->t mutation seems to completely abolish the normal splicing. Among the abnormal splicing isoforms produced from either the Q279R or the IVS6-1g->t alleles, three contain premature termination codons (PTCs) and are expressed at low levels when compared to the other major in-frame mRNA transcripts (Figure 3). These PTC-containing transcripts are likely to be targeted for nonsense-mediated mRNA decay (NMD; [23,24,25]).
The nucleotide change in the Q279R mutation (836A->G) is located at position -2 of the 5' (donor) splice site. The wild-type adenine at this position is present in 64 % of all 5' splice sites . The A->G mutation most likely weakens the splice site, but does not completely abolish its utilization as demonstrated in the transfection experiments (Figure 4).
Structural analysis of FAH also suggested that the Gln->Arg amino acid replacement might be structurally tolerated and would not interfere with the structure as predicted by circular dichroism . Indeed functional analysis of a Q279R-containing FAH variant showed that the mutated protein was enzymatically active . Altogether, these data lead to the conclusion that while Q279R introduces a missense mutation at the cDNA level, it likely acts as a splicing mutation in vivo thereby inhibiting the production of normal FAH mRNA and the corresponding protein. This is supported by the analysis of transcripts in the liver and by the minigene assay, but it cannot be excluded that the mutation is leaky as a low amount of normal transcripts is found in both cases. Whether this low amount of FAH mRNA is translated in sufficient functional protein to alleviate the clinical phenotype remains uncertain. While FAH could not be detected in the tumor tissue by western blot analysis, the presence of low amounts of FAH undetectable by the blot assay cannot be excluded.