Main

Biotin is an essential B-complex vitamin that is the coenzyme for four carboxylases involved in amino acid catabolism, fatty acid synthesis, and gluconeogenesis(1). These carboxylases are activated when biotin is covalently attached to specific ε-amino groups of lysyl residues of the apocarboxylases by biotin holocarboxylase synthetase(2,3). Proteolytic degradation of the carboxylases yields biocytin (biotin-ε-lysine), which is hydrolytically cleaved by biotinidase (EC 3.5.1.2), releasing free biotin and replenishing the biotin pool(1,4). Mean normal biotinyl-hydrolase activity of human serum biotinidase is 7.1 nmol of p-aminobenzoate liberated from biotinyl-p-aminobenzoate per minute per milliliter of serum (range, 4.4-10)(5). Biotinidase has also been shown to have biotinyl-transferase activity, although the physiologic significance of this function is unknown(6).

Biotinidase deficiency is an autosomal recessively inherited metabolic disorder(1,5). Individuals with profound biotinidase deficiency have less than 10% of mean normal biotinylhydrolase activity, which can result in secondary biotin deficiency and subsequently multiple carboxylase deficiency and organic acidemia, if untreated(7). Symptoms of untreated biotinidase deficiency include seizures, hypotonia, ataxia, developmental delay, hearing loss, alopecia, skin rash, lactic acidosis, and ketosis(8,9). Symptomatic individuals exhibit differences in age of onset and varying clinical severity(8). Because oral biotin supplementation can prevent or resolve most symptoms, newborn screening for this disorder is performed in many states and countries(10). Results from these programs indicate that the disorder has an incidence of approximately 1/61 000 newborns.

The cDNA encoding human serum biotinidase has been cloned and sequenced(11) and the genomic structure elucidated(12). Twenty different mutations in the biotinidase gene that cause deficient biotinyl-hydrolase and transferase activities have been identified in symptomatic children with the disorder(13). Because newborn screening should result in complete ascertainment of children with profound biotinidase deficiency, we compared the type and distribution of mutations in this population with those in a population with profound biotinidase deficiency ascertained by exhibiting symptoms(14,15). We found that the two most common mutations in the newborn screening population are absent in the population of children identified by symptoms. These results suggest that some mutations are likely to be associated with the development of mild or no symptoms in untreated individuals. In addition, clinical ascertainment apparently only identifies as subset of all mutation-bearing alleles.

METHODS

Subjects. We studied two distinct groups of unrelated children diagnosed with profound biotinidase deficiency. One group consisted of children who were identified by newborn screening programs throughout the United States; the other group comprised children identified clinically because they exhibited symptoms. All of these individuals have profound biotinyl-hydrolase deficiency as determined by serum assay. All children were treated with 5-10 mg of biotin immediately after diagnosis and, to our knowledge, have remained asymptomatic while on therapy. Blood and serum samples were obtained from these children and their parents for mutation and biochemical analysis. Samples are designated by the letter P and the number of the individual. This study was approved by the institutional review board of the Medical College of Virginia of Virginia Commonwealth University.

Biotinyl-hydrolase activity, biotinyl-transferase activity, and determination of cross-reacting material to antibody prepared against normal human serum biotinidase. Biotinyl-hydrolase activity in serum was determined using the colorimetric assay, which measures the nanomoles per minute of p-aminobenzoate formed from the cleavage of biotinyl-p-aminobenzoate per milliliter of serum(5). Biotinyl-transferase activity was determined by incubating serum samples with histones in the presence of biocytin(6). The amount of biotinylation of the histones is compared with that of normal individuals and semiquantitatively assessed from 0 (undetectable) to +4 (normal). Cross-reacting material (CRM) in each serum sample to antibody prepared against purified human serum biotinidase is compared with normal individuals and designated 0 (undetectable) to +4 (normal)(16).

Mutation analysis by PCR and sequencing of the biotinidase gene. Genomic DNA was isolated from whole blood samples using the Gentra PureGene DNA isolation kit (Research Triangle Park, NC) according to manufacturer's instructions. Direct PCR amplification and sequencing of regions of the biotinidase gene of the study subjects were performed to identify mutations. An alteration in the sequence was considered a mutation if it resulted in a frameshift or a splice site change, or coded for a premature stop. Single base substitutions resulting in an amino acid change were considered mutations if they were previously identified in the symptomatic population or if no other changes were found after sequencing all exons and adjacent intron sequences of the gene.

The human biotinidase gene comprises 4 exons, numbered 1-4, with sizes of 79, 265, 150, and 1502 bp, respectively(12). From previous results(13–15), it was known that most mutations reside in exons 2 and 4. Thus, these regions were analyzed in every child in this study. In accordance with the assumptions stated above, if one or more mutations in a single individual were not identified in exons 2 and 4, exons 1 and 3 were sequenced.

The PCR reactions used to generate fragments for sequencing exons 2, 3, and 4 differ only in primer pairs and annealing temperature used. Cycling parameters and reaction contents are as described previously(14). Primer pairs and annealing temperature used to amplify exon 2 (previously referred to as exon B) and exon 3 (previously referred to as exon C) have been published elsewhere(13). In this study, exon 4 was analyzed in four overlapping fragments spanning the entire 1502 bp of coding sequence as well as 91 bp of 5′ and 43 bp of 3′ noncoding sequence. Primer pairs and annealing temperatures used to generate the two fragments spanning the most 5′ and most 3′ regions of exon 4 are 291.S and B3.A, and 1150.S and 1790.A, published previously(13). The two central fragments were created using primers 533.S (5′-CTTGTCATAGCAGTGACCC-3′) and 353.A(13) with an annealing temperature of 60°C, and 937.S (5′-ATACACACCCCTCTGGAGTC-3′) and 267.A(13) using an annealing temperature of 58°C. Exon 1 was amplified using primer 704.S (5′-ACTAGCAGGAGATTGCTG-3′) and 1016.A (5′-AGCAGCAACGCACAAACAG-3′) in a stepdown reaction(17). The temperature gradient for the stepdown PCR was 70°C to 55°C at increments of 3°C using six cycles per temperature and 10 cycles at the final temperature.

PCR products were purified using Jetsorb DNA purification kit (Genomed, Research Triangle Park, NC) according to manufacturer's instructions. Automated, fluorescent sequencing was performed as described(18) with the same primers used to generate the PCR products. Analysis was performed using Sequencher software, version 3.0 (Gene Codes Corporation, Ann Arbor, MI).

Statistical analysis. The χ2 analysis was used to compare the frequency of each mutation within the symptomatic and newborn screening populations. Individuals homozygous for a mutation are considered to have only one unique allele, although consanguinity has not been confirmed in all cases. All other are considered to have two unique alleles. The number of alleles in the symptomatic population is 55 and the number of alleles in the newborn screening group is 104.

For each mutation, the number of alleles with that mutation and the number of alleles without that mutation were counted for the two populations, and the two percentages were compared using a χ2 test. A χ2. value greater than 3.84 is considered significant at the 5% confidence level and a value greater than 6.64 is significant at the 1% confidence level.

RESULTS

We studied 59 unrelated children with profound biotinidase deficiency who were ascertained by newborn screening programs in the United States. Fourteen of these children are homozygous for a mutation in the biotinidase gene and are considered to have two identical alleles, even though consanguinity was not confirmed in all cases. The remaining 45 subjects are compound heterozygotes, having two different mutation alleles. Therefore, the total number of alleles studied is 104. Among these, 28 different mutations causing profound biotinidase deficiency were identified (Table 1). Because this is the first report of many of these mutations, the nucleotide-based name as well as the amino acid-based name is given for each mutation according to suggested nomenclature(19). Throughout this text, all mutations except frameshift and splice site mutations are referred to by their amino acid-based name.

Table 1 Summary of all mutations and their frequencies in U.S. children with profound biotinidase deficiency ascertained by newborn screening and clinically
Full size table

Four single amino acid substitutions involving a cysteine residue (C423R, R211C, R538C, and Y540C) occur among newborn screening children (Fig. 1,A and B). One of these, C423R, causes the replacement of the normal cysteine with another amino acid and possibly either alters enzyme folding by eliminating a cysteine residue normally involved in disulfide bonding or could disrupt the free sulfhydryl at the active site of the enzyme(4,20,21). Three mutations, R211C, R538C, and Y540C, insert a cysteine residue in place of the proper amino acid. These substitutions may alter the tertiary structure of the enzyme through the formation of aberrant disulfide bonds.

Figure 1
figure 1

Schematic of the biotinidase protein, based on cDNA(11), showing the location of each mutations. A, Mutations found only among children with profound biotinidase deficiency identified by newborn screening programs in the United States. B, Mutations found in children identified by both newborn screening and clinically. C, Mutations found only among symptomatic children. The arrows represent the two in-frame start codons. The bold vertical line indicates the location of the N-terminus of the mature enzyme. Thin vertical lines delineate regions of residues coded by exons 1-4. Locations of cysteine residues are shown by a bold "C", and the six putative N-linked glycosylation sites are noted with shaded hexagons.

Full size image

Twelve other single amino acid substitutions occur among the newborn screening children. One of these is Q456H, a substitution of His for Gln456 (Fig. 1A). This is the most common mutation found among the newborn screening group, accounting for 27.9% of alleles (Table 1), and has been described in detail previously(15). One other relatively common single amino acid substitution, D252G (Fig. 1A), occurs in 6.7% of newborn screening alleles (Table 1). The remaining ten single amino acid substitutions (Table 1), each accounting for no more than 2.9% of newborn screening alleles, are evenly distributed throughout the protein.

Four double mutation alleles are found among the newborn screening children. Each of these alleles code for two separate single amino acid substitutions within the protein. Children who are compound heterozygotes for a double mutation allele and another mutation allele have one parent who carries both alterations of the double mutation, and the other parent carries the third mutation found in the child. In cases of children who are homozygous for a double mutation allele, both parents have both mutations yet each has biotinyl-hydrolase activity in the heterozygote range. This information confirms that in these cases, two mutations occur on one allele, producing a double mutation allele.

The double mutation [A171T;D444H] (Fig. 1A) is the second most common allele found among newborn screening children, occurring in 17.3% of alleles (Table 1), and has been described in detail previously(14). It is interesting that the substitution D444H occurs in two other double mutation alleles, [R157H;D444H] and [F403V;D444H] (Fig. 1A). D444H occurs in the general population with a frequency of 0.039 and results in an approximately 52% decrease in biotinyl-hydrolase activity of the aberrant enzyme encoded by that allele(14). This mutation has also been shown to be responsible for most cases of partial biotinidase deficiency(22). For the two double mutation alleles [A171T;D444H] and [F403V;D444H], we do not know whether A171T or F403V must act in concert with D444H to produce profound biotinyl-hydrolase and transferase deficiencies. By itself, either mutation A171T or F403V could produce a nonfunctional enzyme, although neither has ever been observed alone. F403V occurs within a putative N-linked glycosylation site (Fig. 1A) but does not disrupt the required recognition amino acid sequence for glycosylation. In conjunction with D444H, however, the double amino acid substitutions [F403V;D444H] and [A171T;D444H] cause loss of hydrolase and transferase activities (P615, P21, Table 2). In the case of the double mutation allele [R157H;D444H], however, the R157H substitution alone can result in profound biotinyl-hydrolase deficiency, because it was detected alone in one symptomatic child (P170, Table 2). The amino acid substitutions in the double mutation allele [G45R;A289P] (Fig. 1A) have never been observed alone; therefore we cannot predict their individual effects on the enzyme, but together they produce a nonfunctional enzyme (P82 Table 2).

Table 2 Summary of mutations in the biotinidase gene and biochemical data of children with profound biotinidase deficiency ascertained by newborn screening in the United States
Full size table

There are four mutations causing frameshifts by deletions or insertions of nucleotides. Mutation G98:d7i3(18) is a deletion of seven and insertion of three bases within the putative signal sequence encoded by exon 2 (Fig. 1B) and leads to a premature stop codon. A similar mutation, 1227d15i11(23), is a deletion of 15 and insertion of 11 bases within exon 4 (Fig. 1A). Mutation 933delT (Fig. 1B) is a deletion of a thymine which causes a frameshift about halfway through the coding region. The fourth frameshift mutation, 1616-1617insT, is an insertion of a single thymine only 13 bases from the stop codon (Fig. 1A). This mutation predicts truncation of the protein by three amino acids and a substitution of the two carboxy-terminal residues.

A single nonsense mutation was detected among newborn screening children. This mutation, W386X (Fig 1A), occurs within exon 4 and results in a truncation of the protein by 157 residues. A single in-frame deletion mutation, Y414-V417del (Fig. 1B), was also identified in exon 4. This mutation results in the loss of four residues near the carboxy-terminus of the protein.

Two different mutations that affect splice sites were detected among newborn screening children. Mutation 100G→A within exon 2 activates a 3′ cryptic splice site causing the first 19 amino acids encoded by exon 2, including the second start codon, to be spliced out of the message(24). Mutation 459G→A causes a change in the last base of exon 3. To investigate whether this change might affect splicing, we consulted a scoring system that was devised to compare potential splice sites(25) on the basis of conserved splice junction sequences. A score of 1 indicates perfect consensus. In the biotinidase gene, the normal guanosine at position 459 gives the splice junction a score of 0.79, whereas adenosine at that position gives a score of 0.68. This may be enough of a change to cause aberrant splicing. The two individuals in which this mutation was found have no other changes in their coding sequence, which supports our assumption that this is a mutation. In addition, this change was not detected in any normal individuals.

Mutation analysis and biochemical characterization of the enzyme in children with profound biotinidase deficiency ascertained by newborn screening is summarized in Table 2. Although all children have deficient biotinyl-hydrolase activity, biotinyl-transferase activity and presence of CRM in serum varies among individuals depending on their mutations. For the 14 children who are homozygous for a mutation, the effect of that mutation on the presence of CRM in serum and on the mature enzyme's biotinyl-hydrolase and biotinyl-transferase activities can be determined. For the remaining 45 children who are compound heterozygotes, the biochemical data obtained is a reflection of both mutations, thereby making it difficult to predict the effect of a single mutation on the function of the mature enzyme. Nevertheless, the vast majority of children in the newborn screening group has very low transferase activities and CRM in serum (Table 2).

Our laboratory has previously studied a group of children with profound biotinidase deficiency who were ascertained clinically after exhibiting symptoms of the disorder(13). Mutations in 31 unrelated probands described in that publication and two others who were ascertained subsequently were compared with the type and distribution of mutations in the newborn screening population (Table 2). Of the 20 different mutations found among the symptomatic children, eight are also found in the newborn screening group (Table 1). These 20 mutations include single amino acid substitutions, in-frame deletions, a nonsense mutation, and frameshift and splice site mutations. They are not clustered within one region of the protein, but occur at various sites along the coding region (Fig. 1,B and C).

Of the 40 total mutations identified among both populations, all but four mutations represent a very small percentage of the total alleles studied. Of these four common alleles, two are observed much more frequently in the symptomatic population whereas the other two are exclusive to the newborn screening group (Table 1). The double mutation [A171T;D444H] accounts for 17.3% of newborn screening alleles and none of the symptomatic alleles. Similarly, Q456H occurs in 27.9% of newborn screening alleles and is absent from the symptomatic group. Statistical analysis showed that these frequency differences are significant at the 1% confidence level. The frequency of the frameshift mutation G98:d7i3 within the symptomatic (34.5% of alleles) and newborn screening (12.5% of alleles) populations also is significantly different at the 1% confidence level. The amino acid substitution R538C also occurs at a significantly higher frequency in the symptomatic population (p < 0.01), representing 21.8% of symptomatic and only 2.9% of newborn screening alleles. The differences in frequencies of all other mutations identified within the two populations are not significant at p < 0.05.

DISCUSSION

We studied two distinct groups of children from the United States with profound biotinidase deficiency. The difference between the two populations is the method in which the subjects were ascertained. Children ascertained by newborn screening made up a study group consisting of 104 unique alleles. The symptomatic population consisted of 55 unique alleles. Between the two groups, 40 different mutations were identified, of which eight occurred in both groups (Table 1). The mutations found in the biotinidase gene of children with profound biotinidase deficiency occur along the entire length of the protein and are not clustered in any particular region (Fig. 1).

All individuals diagnosed with profound biotinidase deficiency have less than 10% of mean normal biotinyl-hydrolase activity in serum, measured quantitatively by a colorimetric assay. Biotinyl-transferase activity in serum was markedly decreased or absent in individuals with profound biotinidase deficiency(6). This observation strengthens the assumption that biotinyl-transferase activity is physiologically important and may contribute to the disease state. The amount of CRM to antibody prepared against normal human serum biotinidase is also decreased or absent in serum of children with biotinidase deficiency when compared with that of normal individuals. Depending on the location and severity of the mutation, both transferase activity and CRM are affected to varying degrees. Therefore, slight differences in these properties are observed between profoundly biotinidase-deficient children who have different mutations.

Although the tertiary structure and active site residues of mature biotinidase are unknown, changes such as frameshift mutations, splice site mutations, nonsense mutations, and inframe deletions of amino acids have obvious severe affects on the enzyme, causing a lack of both hydrolase and transferase activities as well as CRM in serum. All children who are homozygous for such a "null allele" or are compound heterozygotes for two different null alleles have deficient biotinyl-hydrolase and undetectable transferase activities and CRM in serum (Table 2).

The biochemical effects of mutations such as single amino acid substitutions that do not result in complete absence of transferase activities and CRM are best characterized by observing individuals who are homozygous for such alleles. Children P21 and P122, for example, are homozygotes for the double mutation [A171T;D444H] and the substitution Q456H, respectively. Results obtained from these individuals indicates that both of these mutations result in trace transferase activity and trace CRM in serum (Table 2). Conclusions about the effect of a single mutation cannot be made in individuals who are compound heterozygotes for two different "non-null" alleles. In such cases, the biochemical data reflect both mutations.

Because biotinyl-transferase activity is not detected in any symptomatic children(13), the degree of transferase deficiency may be a predictor for the development of symptoms in untreated individuals. Lending credence to this possibility is the documentation of two asymptomatic adults who have profound biotinyl-hydrolase deficiency and have never taken biotin supplements(26). One of these individuals is homozygous for D252G, the other is homozygous for [A171T;D444H], and both have trace transferase activity in their serum. Not surprisingly, these two mutations have never been observed in the symptomatic population (Table 1).

If it is true that some mutations causing profound deficiency of biotinyl-hydrolase activity rarely or never result in the development of symptoms if left untreated, then we should observe such mutations exclusively or at a much higher frequency in the newborn screening population. This is the case for the double mutation [A171T;D444H] and the substitution Q456H. Statistical analysis confirms that these mutations appear in the newborn screening population at statistically significantly higher frequencies (p < 0.01) (Table 1). Conversely, mutations that always result in symptoms in untreated individuals should be found in both populations, but at a higher frequency in the symptomatic group because that group consists of only symptom-causing mutations. We found statistical differences between the frequencies of the frameshift mutation G98:d7i3 and the substitution R538C within the two populations. These two mutations occur in the symptomatic group at a statistically significantly higher rate (p < 0.01) than in the newborn screening population (Table 1). These two mutations, therefore, may be severe enough to result in the development of symptoms in any untreated individual.

If the study populations were large enough, all symptom-causing mutations would also be identified in the newborn screening population, and some mutations, those rarely resulting in symptoms if not treated, would appear only in the newborn screening group. Our current study population is only large enough to demonstrate such patterns for the four most commonly occurring mutations mentioned above.

There are eight mutations that are common to both groups (Table 1, shaded region). The fact that two of these mutations, G98:d7i3 and R538C, occur at a statistically higher frequency in the symptomatic group simply reflects how common these alleles are among all children with biotinidase deficiency. Because the symptomatic group is enriched for symptom-causing alleles, these two common symptom-causing alleles are observed among these children at a higher frequency. The 12 mutations unique to the symptomatic group (Table 1, bottom) are probably unique simply because they are rare alleles. Continued study of children identified by newborn screening will undoubtedly detect these mutations in this population in the future.

From our study of these two populations, it seems that clinical ascertainment may lead to the preferential identification of certain mutations. This has been found to be true for other disorders such as medium-chain acyl CoA dehydrogenase deficiency(27). In our case, the newborn screening group is a population from which complete ascertainment of all biotinyl-hydrolase deficiency-causing alleles is possible. Studying a larger group of individuals ascertained clinically and by newborn screening should clarify the existence of genotype/phenotype relationships. Although some mutations may be associated with the development of mild symptoms, biotin treatment should be initiated and continued in all children with biotinidase deficiency ascertained by newborn screening, even if the child has a mutation(s) not yet observed among symptomatic children.