Subjects

The total sample comprised 72 individuals (40 males), including one pair of siblings, aged one month to 18 years, who were recruited from 2012 to 2015 by contacting Brazilian medical geneticists through the mailing list of the Brazilian Medical Genetics Society. The inclusion criterion was reduced biotinidase activity (a value below the normal reference range for the testing laboratory, i.e., heterozygous level, partial or profound BD) detected by quantitative testing after abnormal neonatal screening (n = 59 patients) or after clinical suspicion (n = 6 patients). In addition to these participants, seven subjects (identified by neonatal screening) who had no quantitative test were also included in the study because the special interest of the families and geneticists in the molecular analysis, once they did not have access to the confirmatory test.

Clinical information and blood samples were collected from patients and their parents (samples from both parents were available for 55 patients, samples from only one parent were available for nine patients, and no parental samples were available for eight patients). The patients came from several regions of Brazil (South = 42, Southeast = 15, Northeast = 15).

Biotinidase activity

Quantitative enzyme activity testing results were available for 65/72 patients. All laboratories (n = 11) that performed the tests used the colorimetric testing in plasma or serum described by Wolf et al. [2], and the normal reference range of 5.0–10 nmol/min/mL was applied. The mean value of 7.5 nmol/min/mL was used to classify the biochemical phenotype among patients who presented activity lower than 5.0 nmol/min/mL: <0.75 (<10%), profound BD; 0.75–2.25 (10–30%), partial BD; and 2.26–4.99 (30.1–66.5%), heterozygous activity. Patients whose biotinidase activity was 5.0–10.0 nmol/min/mL were considered to have 100% activity. Values within ± 0.1 of a cut-off point were classified as borderline. When more than one measurement was available, the highest value was used for classification purposes. For the remaining patients (n = 7/72), only qualitative enzyme activity testing results were available; thus, these patients were not included in the biochemical phenotyping.

DNA analysis

In our previous study, the DNA of 38 patients was extracted from blood samples in EDTA, and exons 2, 3, and 4, as well as the exon/intron junctions of the BTD gene, were sequenced [12]. In the present study, the same analysis was performed for the additional 34 participants. Furthermore, the promoter region, 5'UTR and exon 1 (according to annotation at GenBank, accession no. AF018630.1) were analysed for all 72 samples. A fragment of 715 base pairs (bp) including the BTD promoter, 5’UTR, and exon 1 was amplified by PCR using the primers 5’-CCCCACCATATCGCCACATCC-3’ and 3’-CCGCACGCCCTTACCACA-5’ and an annealing temperature of 67°C, followed by Sanger sequencing. This strategy allowed for analysis of the 599 bp upstream of the first translation start site. Parental samples were assessed to define whether the identified variants were in cis or trans. The sequence NG_008019.1 was used as the reference sequence for the BTD gene.

Since the sample included one pair of siblings and five children of consanguineous parents, the total number of alleles was 137.

The establishment of the expected biochemical phenotype to each genotype found was based on the literature [7]. As a result, we could only include genotypes composed of variants with known clinical significance (reported at ARUP database, http://utahedu.aruplab.com/database/BTD/BTD_welcome.php). Non-coding variants were not considered in this classification.

A portion of exon 4 of the BTD gene (exon 4d) was sequenced in anonymous samples from 100 healthy, adult controls from Southern Brazil as part of our previous study [12]. Those results were reassessed to verify the frequencies of the missense variants described for the first time in the present study, c.1337T>C or p.L446P and c.1466A>G or p.N489S.

The frequency of the p.D444H allele and the frequency of patients classified as partial BD based on enzyme activity and genotype were compared between South, Southeast and Northeast Brazil using the Chi-square test. The significance level of 5% was adopted and IBM SPSS Statistics for Windows version 23.0 software (Armonk, NY: IBM Corp) was used.

Comparison between genotype and biochemical phenotype

For each genotype observed, the biochemical phenotype (profound BD, partial BD, heterozygous activity, normal activity) presented by the patients was verified. Furthermore, the expected biochemical phenotype (based on the genotype) and the observed biochemical phenotype were compared in patients for whom both classifications were available, and these comparative results were classified as being either concordant or discordant. When the biotinidase activity was borderline between two categories (e.g., between profound and partial BD), the expected biochemical phenotype was considered concordant if one of the two categories was consistent with the expected biochemical phenotype.

Bioinformatics analysis

The pathogenicity of the novel variants p.L446P and p.N489S was evaluated using PolyPhen-2 [16], SIFT [17] and Mutation Taster [18] software.

Conservation in the region surrounding the non-coding variants in the BTD promoter (chr3:15637844–15648175 according to GRCh37/hg19) was evaluated using ECR Browser [19] (http://ecrbrowser.dcode.org/). Evolutionary conservation of the nucleotides at the non-coding variant positions was then assessed using the multiple alignment tool MAFFT version 7 [20] (http://mafft.cbrc.jp/alignment/server/). The Alibaba version 2.1 tool was used to predict transcription factor binding sites that could be created or abolished by the non-coding variants (http://www.gene-regulation.com/pub/programs.html). The Gene Cards database [21] (www.genecards.org) was consulted to compare the expression profiles (mRNA and protein) of biotinidase and transcription factors that had binding sites predicted to be created/abolished in the BTD gene. This comparison focused on liver and pancreas expression patterns, where serum and intestinal biotinidase is produced. Finally, the ENCODE ChIP-Seq Significance Tool [22] (http://encodeqt.simple-encode.org/) was used to verify whether the predicted binding sites for transcription factors were corroborated by the ChIP-Seq experiments.

Clinical data and biochemical classification

Sixty-six patients were identified by neonatal screening, 55 of whom were receiving biotin supplementation at inclusion; none had clinical manifestations suggestive of BD. Six patients were preterm (35–36 weeks of pregnancy), and neonatal jaundice was reported for another six patients.

Six patients were diagnosed based on clinical suspicion. In these patients, the most common symptoms were visual disturbances, neurological manifestations, and skin lesions. The age at onset of clinical manifestations ranged from one day to 10 years, and the age at diagnosis ranged from 40 days to 18 years. All symptomatic patients were receiving biotin supplementation (10–20 mg/day) at the time of the study. Of the 71 unrelated patients, 5 (7%) had consanguineous parents; one (1.4%) had a family history suggestive of BD (the patient had an older sister who died at 3 years old with a clinical picture consistent with BD but with no biochemical or genetic diagnosis), and one (1.4%) had a confirmed family history of BD (the pair of siblings included in the study).

The biotinidase activity (quantitative assay) and observed biochemical phenotype were determined for 65/72 patients (Table 1). Four (6.1%) patients were classified as profound BD, 19 (29.2%) as partial BD, 34 (52.3%) as heterozygous, six (9.2%) as borderline activity (three between partial BD and heterozygous and three between heterozygous and normal activity), and two (3%) as abnormal activity at inclusion in the study but normal thereafter (patients with genotypes 08 and 36 in Table 1). The frequency of partial BD tended to be higher in the South Brazilian region (38.9%) than in the Southeast (20%) and Northeast (14.3%), although this was not statistically significant (exact p-value = 0.198).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Biotinidase deficiency: Genotypes and biochemical phenotypes presented by 72 Brazilian patients.

https://doi.org/10.1371/journal.pone.0177503.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Clinical information for patients with discordance between the observed and expected biochemical phenotypes (n = 17/72 patients).

https://doi.org/10.1371/journal.pone.0177503.t002

Among all patients, 22 (14 male, 8 female) had more than one biotinidase activity measurement (Table 3). The median difference in age between the first and second blood collection was 70 days (range: 2 to 572 days), and the median difference in biotinidase activity was 9.3% (range: -41.3 to 82.7%). The classification of 11 (9 male, 2 female) patients changed to a better or worse enzyme activity classification between consecutive tests. The changes occurred between the first and second blood collection, with most showing improvement, i.e., moving from a more to a less severe category (n = 8/11). Prematurity and jaundice were present only in the group of patients who moved to a better classification.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Clinical information of patients with more than one biotinidase activity measurement (n = 22/72 patients).

https://doi.org/10.1371/journal.pone.0177503.t003

BTD variants

Thirty-nine different genotypes were found in the entire group of patients (Table 1). In addition to the variants found in our previous study [12], five known pathogenic variants, one synonymous variant (c.1284C>T or p.Y428Y) and three novel variants were identified in the 34 patients added to this study. The novel variants were c.1337T>C (p.L446P), c.1466A>G (p.N489S) and c.962G>A (p.W321*) in exon 4. The variants p.L446P and p.N489S were not found in the controls and were both considered damaging by PolyPhen-2 (scores of 1 and 0.85, respectively) and Mutation Taster (probability of 0.99 for both) but were tolerated by SIFT (scores of 0.1 and 0.22, respectively).

In the BTD promoter region, 5'UTR and exon 1 analyses, three variants in the promoter region were identified: c.-514C>T (rs41284037), c.-315A>G (rs2019160, not shown in Table 1) and c.-183G>A (rs2279841), with allele frequencies of 3.6% (allele T), 99.3% (allele G) and 7.3% (allele A), respectively. Except for one patient (genotype 5 in Table 1, p7 in Table 3) who was heterozygous for the variant c.-315A>G, all other patients were homozygous for allele G.

Among known pathogenic variants, the most frequent were c.1330G>C (p.D444H), c.755A>G (p.D252G) and c.[1330G>C;511G>A] (p.[D444H;A171T]), with allele frequencies of 46.7%, 5.8% and 3.6%, respectively. The frequency of p.D444H (not considering this variant in complex alleles) tended to decrease in patients from the South (51.2%) to the Southeast (44.8%) and Northeast regions of the country (35.7%), although these differences were not statistically significant (exact p-value = 0.38).

Since 11 patients had variants of unknown phase (genotypes 30–33, Table 1) and/or variants not described previously (genotypes 34–39, Table 1), the expected biochemical phenotype could be determined for only 61/72 patients. Based on genotype, four patients (6.5%) were expected to present profound BD, 19 (31.1%) to present partial BD, 23 (37.7%) to present heterozygous activity (of these, 16 were homozygous for p.D444H), and 15 (24.5%) to present normal activity (of these, nine were heterozygous for p.D444H). Among patients classified as having profound BD, three were from the South and one was from Southeast Brazil. The frequency of patients classified as partial BD according to genotype was not significantly different among the South (42.9%), the Southeast (21.4%) and the Northeast regions (8.3%) (exact p-value = 0.058).

Association between genotype and biochemical phenotype (Table 1)

Patients with no BTD variants or with only polymorphic variants (genotypes 01–03) were found among those patients who had lower than normal biotinidase activity, in the range for heterozygotes. Most patients heterozygous for p.D444H (genotypes 05–07) and homozygous for p.D444H (genotype 08) possessed activity within the heterozygous range. The majority of patients who were compound heterozygous for p.D444H and previously described severe variants (genotypes 16–25) had enzyme activity consistent with partial BD, except for five with heterozygous levels. The recurrent genotypes associated with different biochemical phenotypes are shown in S1 Fig (genotypes 05, 08, 17 and 22).

Among the novel genotypes, the patient who presented heterozygosity of the variant p.W321* (genotype 34) exhibited 56% activity (heterozygous range). An enzyme activity measurement was not available for the patient’s father, who had the same variant. A patient with the p.L446P variant in one allele and the normal sequence in the other (genotype 36, p18) had enzyme activity of 21.3% (partial BD) and in a second sample 100% (normal); his father was also heterozygous for the novel variant and exhibited activity of 100% (normal). The patient who possessed the p.N489S variant also possessed a known pathogenic variant in trans (genotype 39, p11) and had enzyme activity of 9.3% (borderline profound/partial BD) and, in a second assay, 18.7% (partial BD); the patient’s father, who was heterozygous for p.N489S, had enzyme activity of 53.3% (heterozygous range).

Comparisons between the expected and observed biochemical phenotype were possible for 54/72 patients, and concordance was observed in 37 (68.5%) cases. The clinical information for patients who were discordant is shown in Table 2. No discordant patients were premature, and neonatal jaundice was reported in only one patient from this group (dp1). Among the patients with an observed biochemical phenotype worse than expected, the biotinidase activity test was performed in a state other than the state of residency for 9/11, and so we assume the plasma samples for these patients underwent a prolonged period of transport prior to the enzyme test being performed. For 5/6 patients with an observed biochemical phenotype better than expected, the biotinidase activity test was also performed in a state other than the state of residency and plasma samples likely also underwent a prolonged period of transport before the enzyme test was performed. Among patients expected to have biotinidase activity in the normal range, patients dp1-dp4 and dp6-dp11 exhibited enzyme activity suggestive of heterozygosity, and dp5 exhibited enzyme activity suggestive of partial BD. Patients dp3 and dp4 showed the c.-514C>T variant, patient dp7 was the only heterozygous patient for c.-315A>G (allele A was in trans with the other variant presented), and patient dp10 had the non-coding variant c.-183G>A. Conversely, patient dp12 exhibited normal enzyme activity, and patient dp13-17 showed levels suggestive of heterozygosity (considering the highest result), although their genotypes suggested heterozygous levels and partial BD, respectively. Patient dp17 exhibited the non-coding c.-183G>A variant. Among the 17 discordant patients, the c.-183G>A, c.-315A>G and c.-514C>T variants were present in heterozygosis in two (12%), one (6%) and two (12%) patients, respectively. The c.-183G>A and c.-514C>T variants were also present in eight (22%) and two (5%) of 37 concordant patients, respectively.Patients with borderline biotinidase activity had an expected biochemical phenotype equal to the higher classification. Among patients whose classification changed between consecutive tests (Table 3), the expected biochemical phenotype matched the worst result for 3/9 patients and the best result for 6/9 patients.

In silico evaluation of non-coding variants

The c.-514C>T, c.-315A>G and c.-183G>A variants map to an evolutionarily conserved region in chimpanzee (1,002 bp) and rhesus monkey (3,212 bp), and c.-514C>T is also conserved in dog (228 bp), mouse (444 bp) and rat (221 bp). S2 Fig shows that the cytosine at position -514 is not conserved in rat; at position -315, there is an adenine or guanine; and at position -183, guanine is not conserved in mouse and rat.

According to in silico prediction, the c.-514C>T variant is located in an Sp1 transcription factor (Sp1) binding site, and the T allele creates one more site for Sp1 and two for neurofibromin 1 (NF-1) (S3A Fig). In the c.-315A>G variant, no binding sites for human transcription factors are predicted, independent of the allele. Finally, allele A of the c.-183G>A variant is predicted to change the site for CCAAT/enhancer binding protein alpha (C/EBPalpha) to a site for CCAAT/enhancer binding protein delta (C/EBPdelta) and to create a site for Sp1 (S3B Fig). The BTD gene and the aforementioned transcription factor genes exhibit mRNA expression in most of the tissues assessed in the studies in Gene Cards database. Protein expression differed, however, with NF-1 found in liver and pancreas, C/EBPalpha detected in the liver, and C/EBPdelta and Sp1 present in the pancreas. The transcription factors Sp1 and C/EBPdelta were included in the list of transcription factors identified in the 5’ region of the BTD gene by the ENCODE ChIP-Seq Significance Tool.

Genotype-biochemical phenotype association

In our study, the genotype-biochemical phenotype association is not always consistent. For example, some patients who presented with the same genotype had different biochemical phenotypes. Patient dp12 (genotype 08), who presented with normal activity despite homozygosity for D444H, is an interesting case, having only been described once before [23].

Neonatal jaundice, prematurity and transportation conditions of collected sample do not appear to be the main causes of the discordance between the observed and expected biochemical phenotypes because none of the discordant patients was premature, and only one discordant patient experienced jaundice. In addition, for patients with biotinidase activity that was both worse and better than expected, the biotinidase activity test had been performed on samples that underwent a period of transport to another state before analysis.

The variants we identified in the promoter region of BTD do not seem to have a relevant effect on biotinidase activity. However, it cannot be excluded that the variant c.-514C>T contributes to a subtle decrease in gene expression since patients without pathogenic mutations (dp3 and dp4) and heterozygous for this variant have activities consistent with heterozygosity. The c.-315A>G (allele A) might also have a subtle decreasing effect since it was only present in a discordant patient who presented with an activity worse than expected (dp7).

The bioinformatics analysis suggested that the c.-514C>T variant is located in a more evolutionarily conserved region than the other variants and that the c.-514C>T and c.-183G>A variants might be critical for transcription factor binding. The co-expression of biotinidase and transcription factors predicted to have binding sites that are created or abolished by these variants reinforces the likelihood of these variants affecting BTD gene expression, but further studies are needed.

Non-genetic modifier factors

Our data on consecutive tests performed on the same patient suggest that enzyme activity increases with age, in particular during the neonatal period. In addition, prematurity and neonatal jaundice might have contributed to the first lower results. Schulpis et al. [5] suggest that the reduced biotinidase activity in jaundiced babies may be due to impaired liver function and that high concentrations of total bilirubin in plasma may act as an inhibitor of biotinidase.

Our data does not support the concept that male babies have lower biotinidase activity due to a higher degree of hepatic immaturity since the majority of patients whose condition changed to a better or to a worse classification were male. It is important to note that the comparison of consecutive measurements could be biased since the reason for performing a second test was not always clear, and consecutive tests were not always performed by the same laboratory.

Genetic data

The BTD gene variants p.D444H, p.D252G and p.[D444H;A171T] accounted for 56% of the known pathogenic alleles in this study. These variants were also among the most frequent identified in the Brazilian sample evaluated by Neto et al. [10] (n = 21). The frequencies of p.D444H, partial BD and partial BD according to genotype tend to decrease from the South to the Southeast and from the Southeast to the Northeast. Swango et al. [27] observed p.D444H more frequently in patients whose parents were of German descent, suggesting a potential founder effect. According to “The 1000 Genomes Project (Phase 3)”, the p.D444H allele frequency differs among populations, with 4% prevalence in European populations, to 2–3% in South Asian and American populations (excluding Brazilians) and zero in African and East Asian populations. European ancestry is the major contributor to the genetic background of Brazilians (followed by African and Amerindian ancestries), and the South has a greater European contribution than the Southeast and Northeast [28]. These geographical differences provide a likely explanation for the apparent difference in p.D444H frequencies observed in Brazil. The frequency of p.D444H in the Brazilian population from the South is 4% [12], but similar data are not available for the populations of other regions.

The novel variants p.L446P and p.N489S do not appear to be polymorphic in the Brazilian population. The in silico pathogenicity predictions for the variant p.L446P are divergent, but based on the enzyme activity exhibited by the patient and his father, this variant appears to be non-pathogenic. The variant p.N489S is located at a potential N-linked glycosylation site [29] and could be considered mildly pathogenic, similar to p.D444H, according to the patient’s highest activity (partial BD), although the patient’s first biotinidase activity measurement was compatible with profound BD, and the activity shown by the patient’s father was at the level of heterozygosity. Previously, another nucleotide substitution in the same position, c.1466A>C (p.N489T), was reported and considered pathogenic [30]. The novel variant p.W321* creates a stop codon at the end of the nitrilase/amidase homologous domain (which reaches amino acid 367), and thus this variant must be considered pathogenic.

Exon 1 contains the first translation start site and encodes the peptide signal. Some groups have analysed the initial portion of the BTD gene in patients with BD [31–34]. These studies observed a point variant in the promoter [32] and another in the 5’UTR [35]. In addition, a 107-kb contiguous deletion including the promoter, exon 1 and a portion of intron 1 of the BTD gene, the entire HACL1 gene and a small portion of the of the COLQ gene [36] have been reported. Previous studies in Brazil have not identified variants in the 5'UTR and exon 1 [9, 10], and variants in these regions were not observed in the larger and more diverse sample examined in the present study. This is the first study to sequence the BTD promoter in Brazilian patients with reduced biotinidase activity, and the c.-514C>T, c.-315A>G and c.-183G>A variants were identified and are considered polymorphic based on their frequencies of 6% (allele T), 96% (allele G) and 17% (allele A) in “The 1000 Genomes Project (Phase 3)”, respectively.