In the present study, the 851del4 was accounting for 80% of the mutant alleles identified. We demonstrate 12.8% (5/39) prevalence of NICCD among Thai infants with idiopathic cholestatic jaundice/INH. Ko et al and Fu et al reported 6% (3/47) and 9.5% (38/400) prevalence of NICCD among Korean and Chinese infants with idiopathic cholestasis, respectively [22, 23]. By employing the above prevalence of 12.8%, 40% of Thai infantile cholestatsis being have idiopathic cholestasis  (and our unpublished data), and panethnic incidence of cholestatic jaundice 1/2,500 infants (no specific data for Thai infants) [1, 2], the incidence of NICCD among Thai infants is calculated to be 1 in 48,828. By reverse calculation based on Hardy-Weinberg equation, the frequency of SLC25A13 mutation-carrier among Thai infants calculated from the incidence of NICCD is predicted to be 1 in 110 (q2=1/48,828; 2pq = 2x≅1x1/220). This number is lower than the frequency of SLC25A13 mutations carrier among Chinese (1/63), Japanese (1/65), and Korean (1/108) populations, respectively [15, 20, 25], plausibly due to underestimation of the prevalence in the present study because not all infants with SLC25A13 mutation(s) on both alleles manifest NICCD symptoms and that the carrier rates described in those East Asian populations were obtained through molecular genetic screening of control population which is more reliable method to provide epidemiologic data.
If p.M1? is considered pathologic allele and its carrier rate (3/100) by population analysis is taken into account with the carrier rate of the three mutations (851del4, IVS16ins3kb, and 1638ins23) obtained from the cholestatic infants group, it makes a carrier rate of 1 in 52 (4/210), giving a new estimated incidence of NICCD 1 in 11,025. Given no available data of p.M1? in other population to compare and the absence of functional prove of the p.M1?, it may be too premature to include the carrier rate of p.M1? for estimation of the incidence of NICCD in Thai population.
We tried to compare the genotype of the most severe case (Patient 4) in the present study to those severe NICCD cases published in the literature. Only four NICCD patients have been reported to have progressive liver disease requiring liver transplantation [9, 26–28]. Genotypes of those cases were 851del4/IVS11+1G>A [9, 27], 1638ins23/S225X , and genotype not specified in two cases; therefore, genotype-phenotype association could not be concluded.
Among over 50 mutations of SLC25A13 described , the 851del4 (mutation I) and IVS11+1G>A (mutation II) are the most frequent mutations described among Japanese affected population, accounting for 70% of the mutant alleles, and 851del4 is the most frequent in Chinese [8, 20]. The other common mutations were 1638ins23, S255X, and IVS6+5G>A . There is no doubt about the pathogenicity of the mutations 851del4, IVS16ins3kb, and c.1638-1660dup because these mutations have been identified multiple times in East Asian affected population although, functional data is quite limited.
The pathogenicity of the novel p.M1? variant is uncertain. With the absence of the original initiation start codon, the first AUG codon is located at nt 71-73 leading to translation of a non functional short polypeptide of 21 amino acids and possible degradation of the mutant mRNA through a nonsense-mediated mRNA decay mechanism . Whether or not this particular allele has deleterious effect on AGC2 function, it remains to be elucidated. The p.R605Q mutation occurs at a conserved amino acid position of SLC25A13 (AGC2) and SLC25A12 (aralar) across various species (AGC2: cow, chicken, mouse, rat, chimpanzee, horse; aralar: human, monkey, macaque, dog, mouse); although possibly be deleterious, its pathogenicity remains to be proven.
When comparing with the non-NICCD cases, the NICCD patients had significantly higher ALP and lower ALT levels, higher citrulline concentration and threonine/serine ratio, supporting previous studies [9, 11, 30]. The AST/ALT ratio seemed higher in NICCD group, but was not statistically significant. Failure to thrive was a presenting feature in 37-50% of NICCD patients, but not found in our cohort [11, 12]. NICCD cannot be excluded on the basis of normal PAA profiles, as evidenced in Patients-3. Liver histology was not always diagnostic for NICCD. The characteristic histology of NICCD including cholestatsis and fatty change [10, 11, 31, 32] was found in two cases in this study. Bile duct paucity, an uncommon finding in NICCD which has been demonstrated in one report  was found in one patient (Patient 2). His clinical manifestations were moderately severe with cholestatic jaundice, coagulopathy, galactosuria and jaundice resolved at 6 months of age. Fatty liver was found in a non-NICCD patient (Patient 10) who had jaundice since 1 month and the jaundice resolved by 3 months without identified other metabolic liver diseases. Another non-NICCD case, Patient 11, had markedly high ALP and AFP (1935 and >60,500 respectively, shown in Table 2). Her jaundice started at 1 month and resolved at 6 months of age, and her liver histology was the feature of INH. Of note, Patients 10 and Patient 11 could have NICCD but with unidentifiable mutation(s).
There were limitations of the present study. Firstly, using the panethnic incidence of infantile cholestasis for the estimation of the prevalence of NICCD among Thai infants raises concern on the precision of the predicted prevalence. It would be more reliable to use prevalence of infantile cholestasis among Thai population; however, such data does not exist. Secondly, the prevalence of NICCD obtained by calculation from manifesting infants may be underestimated because AGC2 deficiency is a disease with incomplete penetrance. Thirdly, large deletion and/or insertion could be missed by PCR-direct sequencing. Fourthly, the lack of functional analysis to support the pathologic significance of novel SLC25A13 variants identified in the present study leads to inconclusive data of some patients (Patients 6 and 7). Existing system for functional analysis of AGC2 protein is complicated and difficult to conduct . New method of functional analyses of AGC2 mutation is needed to provide more insight into SLC25A13 variants.