Here we describe a total of 91 HGD
variants (62 missense, 13 splice site, 10 frameshift, 5 nonsense and a no-stop), 52 from our study (22 novels) and 39 previously reported. The frameshift variants, the no-stop variant, and the variants modifying the canonical sites for splicing have a high likelihood to be pathogenic and result in a deleterious HGD protein. However, the pathogenicity of the missense mutations is less clear. To investigate the effect of the 62 HGD
missense variants, we evaluated their possible pathogenicity using different bioinformatic prediction tools: BLOSUM [Henikoff and Henikoff, 1992
], SIFT [Ng and Henikoff, 2003
; Xi et al., 2004
], POLYPHEN [Ramensky et al., 2002
], PANTHER [Thomas et al., 2003
; Thomas and Kejariwal, 2004
; Thomas et al., 2006
], PMUT [Ferrer-Costa et al., 2005
] and SNAP [Bromberg and Rost, 2007
; Bromberg et al., 2008
] (Supp. Methods
). All the results of these different tools are indexed the Supp. Table S1
. Based upon the substitution matrix BLOSUM 62, 31 of 62 missense variations (50%) had a value <-2, classifying them as deleterious. Of the remaining 31, 25 (40%) were ambiguous (score between −1 and +1) and 6 (10%) were classified as benign (score > +1) (Supp. Table S1
Among the 62 candidate missense variations submitted to the SIFT program, 55 (88.5%) were identified to be deleterious with a tolerance index score <0.05, one (1.5%) was ambiguous, and 6 (10%) were benign (Supp. Table S1
). SNAP predicted 55 (90%) of these variants to have a non-neutral effect on the protein and 6 (10%) to have a neutral effect. When the same variants were submitted to POLYPHEN, 50 (80.5%) were called probably damaging, 46 based on the alignment and 4 based on their effect on the 3D structure of the protein. Eight (13%) were considered possibly damaging and only four (6.5%) were called benign (Supp. Table S1
The results obtained using PANTHER and PMUT were less significant. PANTHER predicted 39 variants (63%) to be mutations (subPSEC <-3) and 23 (37%) to be nsSNP. PMUT called only 35 variants (56.5%) pathological, with a neuronal score over 0.5; only 14 (22.5%) had a con dence index over 5. The other 27 (43.5%) were considered neutral (neuronal score under 0.5) and only 10 (16%) had a confidence index over 5 (Supp. Table S1
We examined the possible pathogenicity of all the splice site variations using different analysis programs, including the splice site prediction tool from the Berkeley Drosophila Genome Project (BDGP) web site (http://www.fruitfly.org/seq_tools/splice.html
) [Reese et al., 1997
] and NetGene2, available from the Center of Biological Sequence analysis (CBS), http://www.cbs.dtu.dk/services/NetGene2/
[Hebsgaard et al., 1996
]. The results of these two programs are indexed the Supp. Table S2
. Both programs showed a drastic decrease of the score for 11 (84.5%) of the 13 variants. These variants all affected the canonical site for splicing. Interestingly, for 4 variants, BDGP predicted that the alteration may create a new intronic splice site further downstream in the intron (Supp. Table S2
). The two other variants, c.650-56G>A and c.650-17G>A, for BDGP do not modify the score (0.84) and for NetGene2 the NN score does not vary (0.662) or varies only slightly (0.635). Therefore, this seems to predict these two previously reported variants not to affect the splicing site.
We identified three probands (Patients 36, 93 and 94), each with three HGD candidate disease-causing mutations. Patient 36 carried a nonsense mutation p.E168X and a splice site modification, c.1007-2A>T, strongly predicted to affect the splicing by BDGP and NetGene2, since it modifies the AG of the acceptor site. Patient 36 also carried a missense variant p.N149K which was predicted to be deleterious by POLYPHEN, SIFT and SNAP, ambiguous by PMUT and benign by PANTHER. None of these three variants was previously reported.
Patient 93 had three missense variations. One, p.S305F, is universally predicted as deleterious by all prediction tools and has been previously described [Phornphutkul et al., 2002
]. POLYPHEN, PMUT, PANTHER and SNAP predicted the second variant in Patient 93, p.Q258P, to be deleterious; SIFT considered it benign. This variant was also previously reported [Phornphutkul et al., 2002
]. POLYPHEN, PMUT and PANTHER predicted the third variant in Patient 93, p.E3A, to be benign and SIFT predicted it to be deleterious.
The third patient with three candidate disease-causing mutations, Patient 94, carried a previously reported missense variant p.G161R [Gehrig et al., 1997
]. Functional analysis of this variant showed a loss of 99% of HGD enzymatic activity [Rodriguez et al., 2000
]. The second variant, which was previously reported [Beltran-Valero de Bernabe et al., 1999a
], in Patient 94, c.16-1G>A, represents a strongly predicted splice site modification, since it affects the GT of the donor site. The third variant of Patient 94 was an indel, c.1387_1389delGAGinsTA, which creates a frameshift and is expected to produce a truncated protein, p.M339IfsX368. It is a novel mutation, and was detected in three other unrelated probands in this study (Patients 15, 59, 81 and siblings).
We conclude that for these 3 patients, without further functional studies, it is impossible to indicate the disease-causing variants because all three variants have a potentially deleterious effect on protein function.