Clinical phenotypes and NPC1 mutations in the cohort
Genotypes and phenotypes of the 26 patients are summarized in Table 1 and Table S1. Detailed data on clinical phenotypes are available in  for 21 of them (for correspondence see Table S1). The phenotype classification of patients #5 and #24 was revised and reclassified according to Nadjar and colleagues as late infantile and adolescent/adult, respectively. Juvenile NP-C was by far the most common clinical phenotype. Our cohort also comprised 4 patients with the early infantile form, 5 patients with the late infantile form and 4 patients with the adolescent/adult form. The most frequent NPC1 mutations in our cohort were R1186H (n = 12, 2 homozygotes), S954L (n = 10), and A927V (n = 6, 3 homozygotes). All three are common in European populations. The two variants usually reported as the most prevalent in populations from the Western world, I1061T (n = 2) and P1007A (n = 4), were present in compound heterozygosity with other mutations. One patient was a compound heterozygote for two frameshift mutations. The effects of missense mutations S954L and P1007A were deduced from samples of 7 patients, who carried presumably null frameshift mutations on the other allele. Similarly, impacts of less severe mutations were estimated on the background of known and well characterised severe mutations. The positions of the mutations are highlighted in the NPC1 structure in Figure 1c.
Analysis of the NPC1 promoter region
We identified 6 common variants clustered into 5 haplotypes (Figure 1a) in the NPC1 promoter region in the cell lines of patients and controls. Haplotypes 2, 3, and 4 contained 4, 2, and 1 sequence variants, respectively, that were not present in the reference sequence (haplotype 1). Promoter fragments corresponding to haplotypes 2, 3, and 4 had 20%–30% lower luciferase reporter activities than haplotype 1. The differences were statistically significant (p = <0.001). Rare haplotype 5 (one allele in the cohort only) was not tested by the reporter assay.
Expression of the NPC1 transcripts
NPC1 relative expression measured by two TaqMan assays (Hs00264835_m1 and Hs00975249_m1) were highly correlated (correlation coefficient r = 0.92, p<0.0001), therefore, only results of Hs00264835_m1 assay were used for analysis. Relative expression of NPC1 in patients was comparable to controls. Lowest levels of NPC1 expression (0,25–0,30) were found in samples carrying a frameshift mutation (#1, #21, #22, #23).
NPC1 allelic expression ratios in fibroblasts carrying two missense mutation were ~ 50/50 (Table 1) with the exception of 70/30 ratio in patient #15. This discrepancy was likely caused by combination of promoter haplotypes 1 and 4 and/or altered transcript stability. In patients carrying a nonsense and a missense mutation, expression ratios were usually skewed in favour of the missense mutation (90/10—65/35). In patient #1, who carried two frameshift mutations (A605Cfs*1/A1187Rfs*54), the allelic ratio was shifted in favour of the latter (presumably more stable) allele (10/90).
We did not identify any major promoter haplotype-linked differences in allelic NPC1 expression. For example, there was identical expression of both NPC1 alleles in two siblings homozygous for A927V and heterozygous for promoter haplotypes 2 and 4 and SNP rs1140458 in the coding region.
Semi-quantitative measurement of NPC1 protein in skin fibroblast lines
Fibroblasts with the lowest levels of the NPC1 protein were from patients affected by the early or late infantile forms of NP-C. On the contrary, patients with the highest levels of NPC1 protein presented with the adolescent/adult disease (Table 1, Figure 1b, Figure 2b and Figure S1). Patient #9 (P1007A/R1186H) with relatively high amount of NPC1 protein and a late infantile phenotype was an exception. The lowest amount of NPC1 protein was found in samples from patients #1 (A605Cfs*1/A1187Rfs*54), #2 (N916del/P1245Rfs*12), #7 and #8 (R1186H/R1186H homozygotes). These findings correspond well with the severe impact of the latter mutation. Samples of the two siblings #5 and #6 (Y276H/R1186H) repeatedly displayed slightly altered banding patterns on Western blots compared to controls (Figure 1b).
Confocal microscopic co-localization studies
Subcellular localization of the mutant NPC1 was difficult to assess in cells with extremely low expression of the protein (e.g. fibroblasts carrying R1186H or frameshift mutations). In these cells the NPC1 fluorescent signal was indistinguishable from the background (Table 1 and Figure 1d). For this reason we did not evaluate possible co-localization of NPC1 with the ER-marker protein disulfide isomerase (PDI) that was tested by others. Localization of NPC1 was evaluated only in cell lines expressing higher amounts of the mutant protein. We presume that in cell lines carrying heterozygous frameshift NPC1 mutations the detected protein is expressed from the other allele (e.g. S954L and P1007A). The signal from A927V, S954L and P1007A NPC1 mutant proteins colocalized with the late endosomal / lysosomal (LE / LY) marker LAMP2 (Table 1, Figure 1d and Figure S1).
Concentrations of UC and CE, direct quantitative filipin staining, diagnostic filipin test and LDL-induced rates of cholesteryl ester formation in native cultured skin fibroblast lines
We evaluated UC/CE ratio instead of their separate values (Table S1) as we expected increase of UC and decrease of CE due to the impaired metabolic turnover of cholesterol by the NPC protein mutation. UC/CE ratios were elevated (Table 1, Figure 2d) in 8 patients with the most severe phenotypes (early infantile, late infantile), while in patients with milder phenotypes the values overlapped with controls. However, in patient #9 with late infantile NP-C (P1007A/R1186H) UC/CE ratios were comparable to controls.
The strongest filipin signal under steady-state conditions was detected in cells homozygous or compound heterozygous for null and severe NPC1 mutations, such as R1186H or T1205K (Table 1 and Figure 1d). Conversely, a low direct filipin staining signal was observed in cells compound heterozygous for a null allele and S954L, also showing a good level of NPC1 protein (patients #19–23). Of note, in cells from compound heterozygotes for P1007A, signs of perinuclear cholesterol accumulation by filipin staining were only revealed after preincubation of the cell cultures with lipoprotein-deficient serum followed by LDL loading (diagnostic filipin test).
LDL-induced early rates of cholesteryl ester formation were often close to nil and <150 pmol/4.5h/mg protein (classical profile) in all cell lines with very low NPC1 protein levels, and showed higher values with a wide variation in the other cell lines, falling into two previously described categories, intermediate (values 10–15% of normal) and variant (>15% of normal, i.e.>500 pmol/4.6h/mg protein) (Table 1).