Mutational Spectrum in the Δ7-Sterol Reductase Gene and Genotype-Phenotype Correlation in 84 Patients with Smith-Lemli-Opitz Syndrome

Patients

The study population included 84 unrelated patients with SLOS who are from the United States, Germany, and the United Kingdom (table 1). All but one patient (D61, who had mostly American Cherokee heritage) were white. Of these 84 patients, 13 had been included in a previous analysis (Fitzky et al. 1998). In 62 patients, the diagnosis was established by quantification of 7DHC, by use of gas chromatography/mass spectrometry (Kelley 1995). When adequate clinical information was available, patients (n=55) were further characterized by a scoring system, wherein malformations in each of 10 embryologically distinct areas were scored as either “1” or “2” and the sum normalized to 100 (Kratz and Kelley 1999). When patients whose SLOS was scored in this manner were adjudged qualitatively, by clinical geneticists, as having mild, moderate, or severe forms, the mean severity scores were 10, 39, and 72, respectively.

Mutation Analysis

Amplification.—High-molecular-weight DNA samples from the probands were isolated from either peripheral blood leukocytes or fibroblasts, by use of established procedures. Exons 1–9 and their flanking sequences were amplified by use of PCR (Saiki et al. 1985). The amplification primers for exons 3–9 were described elsewhere (Fitzky et al. 1998). In the present analysis, the 3′ and 5′ parts of the coding sequence of exon 9 were amplified separately. The primers used for exons 1 and 2 were as follows: 1 forward, gtccggccgctgccgaatggc; 1 reverse, cactgcgcacaccttcccctg; 2 forward, ggccttgagcgtctgccctctcc; and 2 reverse, catagcctgcgcccacgatccagg. The amplification was done over 35 cycles, with each cycle consisting of denaturation for 45 s at 95°C ; annealing for 45 s at 55°C (for exon 7), 57°C (for exons 1–3, 5, 6, and 8), and 62°C (for exons 4 and 9); and extension for 45 s at 72°C. The reaction mixture contained 50–100 ng DNA; 200 ng each primer; 200 μM each of dGTP, dCTP, and dATP; 800 μM dUTP; 3.5 mM MgCl₂; 10 mM Tris-HCl pH 8.8; and 1 U Taq polymerase (Dynazyme II; Finnzymes).

SSCP.—Exons 1–8 and the 5′ and 3′ ends of exon 9 were screened for mutations, by use of SSCP. A total of 7 μl PCR product was subjected to electrophoresis done on a mutation-detection–enhancement (MDE; FMC Bioproducts) gel and was then silver-stained as described elsewhere (Budowle et al. 1991).

DGGE.—For DGGE analysis of exons 4 and 9 (3′ end), 30-bp-long CG clamps were included at the 5′ end of the forward primer (exon 4) and at the 3′ end of the reverse primer (exon 9), respectively (Sheffield et al. 1989). The melting profiles of the fragments were analyzed by use of the computer algorithm Melt 87 (Lerman and Silverstein 1987). Gels were stained with ethidium bromide.

Sequencing.—The amplified DNA was subjected to electrophoresis on agarose gels from which the fragments were excised, were purified with QIA Quick PCR Purification Kit (Quiagen), and were sequenced, on the ABI Genetic Analyzer 310, with the Big Dye Terminator Cycle Sequencing Ready Reaction Kit (PE Biosystems).

In vitro mutagenesis, heterologous expression, and quantitative analysis of overexpressed protein.—Most of the identified point mutations were introduced, by in vitro mutagenesis, into the cDNA of human DHCR7 fused with an N-terminal c-myc epitope, as described elsewhere (Fitzky et al. 1998). Wild-type and mutated cDNAs were subcloned into pCIneo (Promega) with XhoI-NotI restriction sites and were heterologously expressed in the HEK293-derived cell line tsA-201. For quantitative analysis of DHCR7 protein expression, 40 μg microsomal protein were separated on 12% (wt/vol) SDS-polyacrylamide gels, were transferred to polyvinyldifluoride membranes, and were immunostained, against the c-myc epitope, with 50 ng antibody 9E10 per ml (Oncogene Science). 9E10 c-myc immunoreactivity was visualized with the use of 35 μCi purified [¹²⁵J]-labeled antimouse immunoglobulin G (NEN) per ml and was quantified by membrane slicing and γ-counting (Fitzky et al. 1998). The data shown in table 2 are the results from three independent transfections.

Statistics

All patients were included in the genotype-phenotype analysis. To include patients whose exact severity scores were unknown but whose phenotype descriptions were available (n=14), we assigned the scores “10, ” for the mild phenotype; “39, ” for the moderate (classic) phenotype; and “72,” for the severe phenotype, as noted earlier. The Kruskal-Wallis, Mann-Whitney, and χ² goodness-of-fit tests were applied as indicated. Spearman correlation coefficients (R_S) and linear regression were calculated by use of the software package SPSS for Windows (release 8.0.0. 1997; SPSS). Analysis of variance (ANOVA) was used to compare the variance of quantitative biochemical variables both within and between genotypes.

Mutational Spectrum

All of the exons encoding the open reading frame, as well as the untranslated exons 1 and 2, were screened by means of either SSCP (exons 1–9 [5′ end] and 9 [3′ end]) or DGGE (exons 4 and 9 [3′ end]) performed in the 84 patients. Exons showing shifted or ambiguous banding patterns were sequenced. However, in those patients for whom either no mutation or only one mutation was found by use of this strategy, all exons were sequenced. This resulted in a mutational detection rate of 96%. In all patients with biochemically confirmed SLOS, we found ⩾1 mutation. For only six patients were we unable to identify a 2d mutation. Three mutations were found in two of the patients (D80 and D111). This was confirmed by analysis of the parents. The genotypes of these patients were IVS 8–1G→C; R 352Q/R352Q (D80) and IVS 8-1G→C; T93 M/IVS 8-1G→C (D111).

We identified a total of 40 different mutations, several of which were recurrent, in the patients. In addition, several polymorphic silent mutations, some of which have previously been described elsewhere (Fitzky et al. 1998), were found. Newly identified sequence variants causing no amino acid change were 285A→G, 969G→T, 1350C→G, and −223T→C. Mutation −223T→C occurred in 2/40 healthy control individuals.

Table 2 lists all identified SLOS mutations. The majority (36/40 [90%]) were missense mutations. The four nonmissense mutations included one out-of-frame 15-bp deletion (720–735 del), a 33-bp deletion comprising 5 bp of a splice-donor site (385–412del and IVS 5+1–5del), one splice-site mutation (IVS8-1G→C) resulting in a 134-bp insertion because of alternative splicing, and a nonsense mutation (W151X); all of these nonmissense mutations are predicted to result in truncated proteins and may also cause mRNA instability (Maquat 1996). Mutations occurred in the translated exons 4–9 but were clustered in exon 9, in which ∼50% (21/40) of the mutations were found. Several codons had two or three different mutations (e.g., C380S, C380R, and C380Y) (fig. 1).

The highest allele frequency was observed for the splice-site mutation IVS8-1G→C, which was found 48 times among the 168 SLOS alleles (P=.29 of all SLOS alleles). Four homozygotes for this mutation were found, which is close to the expected number (table 3). The missense mutation R404C also occurred with a frequency of SLOS alleles that was >.1, and mutations T93M, W151X, and V326L had frequencies >.05. It is notable that two of the truncating mutations (IVS8–1G→C and W151X) were among the five most frequent SLOS mutations identified, accounting for one-third of all SLOS mutations.

Homozygotes were also identified for R404C, W151X, R352Q, and C380Y. None was from a known consanguineous marriage. However, for one of the mutations (R404C), significantly more homozygotes (n=5) were identified than were expected (n=1). Three of the five R404C homozygotes were from the same Louisiana parish and have in common an ancestor of French origin dating back to the early 18th century. Hence, this mutation most likely represents a French founder mutation.

Of the 36 missense mutations, 31 affected conserved residues (i.e., residues identical in the human, mouse, and plant DHCR7 genes [Fitzky et al. 1998]). Only five mutations changed a nonconserved residue. Three of these changed the same Cys residue (C380 to Y, R, and S, respectively), and one introduced a proline into a presumably α-helical hydrophobic transmembrane domain (L157P). All of these mutations, by virtue of the type of amino acid change introduced, are likely to be SLOS-causing mutations.

The parents of patients D75, SLO14, and 25720, who, according to the results of SSCP/sequencing, appeared to be homozygotes, were analyzed, to establish that they were true homozygotes and that they were not, for example, carrying a deletion for one allele. In all cases, both parents were heterozygous for the respective mutation. Together with the high mutation-detection rate (96%), this result suggests that no frequent large deletions or other SLOS-causing mutations exist and that patients identified as homozygotes are true homozygotes.

Localization of Missense Mutations in the Protein and Classification of Mutations

The DHCR7 protein is localized in the membrane of the ER and has nine putative transmembrane domains and one large cytoplasmic loop (the 4th cytoplasmic loop) (Fitzky et al. 1998; Moebius et al. 1998). The C-terminus is predicted to be located in the lumen of the ER. The 36 missense mutations have a characteristic distribution along this topography (fig. 1). Twenty-one mutations are located in the transmembrane domains, three are at the border of transmembrane domains 8 and 9, seven are in the 4th cytoplasmic loop, and five are located in the C-terminus of the protein. Hence, there is clustering of the missense mutations in the transmembrane domains. These mutations are likely to affect protein stability, which has already been demonstrated for some of the mutations (Fitzky et al. 1998). Notably, all but one (R450L) of the missense mutations analyzed in the present study also reduced protein expression (table 2). At least some of the mutations in the C-terminal ER domain of DHCR7, including R450L, also reduce the enzymatic activity (B. U. Fitzky, unpublished data). On the basis of the localization of the amino-acid sequence changes in the protein and their presumed functional consequences, we categorized SLOS mutations into four classes (table 2): truncating mutations predicted to result in functional null alleles (0 mutations); mutations located in or close to the transmembrane domain (TM mutations); mutations in the 4th cytoplasmic loop (4L mutations); and mutations affecting the C-terminal domain (CT mutations).

Genotype-Phenotype Correlation

The classification of mutations into four distinct types (O, TM, 4L, and CT mutations) was used to analyze whether a relation exists between DHCR7 genotypes and SLOS phenotypes. Of the 10 possible genotype combinations, 7 (0/0 = 7, TM/TM = 6, 4L/4L = 6, 0/TM = 40, 0/4L = 10, TM/4L = 2, and TM/CT = 7) were represented in the sample.

Genotypes were related to the biochemical phenotypes, by comparison of various sterol metabolite concentrations and their ratios with the clinical phenotypes and by application of the described scoring system. When total cholesterol, 7DHC, 8DHC, the 7DHC/cholesterol ratio, and the sum of 7DHC plus 8DHC expressed as a fraction of total sterols (DHC fraction) were compared both within and between genotypic classes, they were more similar within genotypes than between genotypes (ANOVA, P=.001). We first analyzed the relation between the patients' sterol parameters and their severity scores, and we found a significant correlation between the DHC fraction and the severity score (fig. 2A). There was also a clear correlation between the DHC fraction and the genotypes (R_s=0.548, P<.001) (fig. 2B). Patients with two TM mutations or with one TM and one CT mutation had the lowest DHC fraction, on average, whereas those with two 4L mutations, those who were homozygotes and compounds for 0 mutations, and those with one 0 and one 4L mutation had the highest DHC fractions. Patients with 0/TM and 4/TM mutations had intermediate values. Despite the substantial differences among the various means, there was broad overlap between the groups. The correlation of genotypes with the sterol levels suggested that the genotypes might also correlate with the severity scores. Indeed, average disease severity increased in almost the same order with genotypes as did genotypes with the DHC fraction. Patients in the TM/TM and TM/CT groups were, on average, less severely affected than were patients with two 0 mutations or those with two mutations in the 4th cytoplasmic domain (fig. 2C). For this correlation, the coefficient was R_s=.370 (P<.006). Thus, in summary, we found, among our patients with SLOS, a strong correlation of both the genotypes and the cholesterol metabolites with their clinical severity. We further evaluated the genotype distribution among patients in the three classes of severity: severe (score >50), moderate (score >25 and <50), and mild (score <25). Genotypes 0/0 and 4L/4L were significantly more frequent in patients with a severe phenotype, whereas genotypes TM/CT and TM/TM occurred predominantly in patients with a mild/moderate phenotype. This difference was significant (P=.001).

Correlation of Genotypes with Age at Diagnosis

We compared the age at diagnosis of term-delivered patients for the three groups of severity. The group with the lowest severity score (<25) had a significantly higher age at diagnosis (mean age = 8.6 years ± 12.4; n=20), whereas the group with the highest severity score (>50) included the youngest patients (mean .29 years ± .50; n=10). The difference in age at diagnosis between groups was significant (Kruskal-Wallis test P=.050). Because age at diagnosis correlates with clinical severity, it should also correlate with genotype. The genotype distributions were therefore compared between patients in whom SLOS was diagnosed either at age <1 year (n₁=37) or later in life (n₂=29), by means of the Mann-Whitney test, and they were found to be significantly different (table 4).

SLOS Phenotypes in Homozygotes or Compounds for Presumed Functional Null Alleles of Δ7-Sterol Reductase and Mutations in the Most Severe Cases

As shown in the genotype-phenotype correlation studies, patients with two functional null DHCR7 alleles have the most severe phenotypes. Of the four patients homozygous for the IVS8-1G→C mutation, three were severely affected and were detected by prenatal ultrasound. These pregnancies were terminated. The other IVS8-1G→C homozygote was moderately affected. The only patient homozygous for the W151X stop mutation also had a severe phenotype detected prenatally. The compound heterozygote for W151X and IVS8-1G→C had a severe phenotype (severity score 56) and died at age 2 d.

In all, we studied 16 cases with a severity score >50 (table 5). The frequency of 0 alleles (.59) and of null and 4L alleles combined (.78) was significantly higher than that in the mild/moderate group (P=.05), and a significant number of patients (10/16) were 0 and/or 4L homozygotes or compounds. Only one had neither a 0 mutation nor a class 4L mutation.