Comprehensive Mutation Screening in 55 Probands with Type 1 Primary Hyperoxaluria Shows Feasibility of a Gene-Based Diagnosis

Abstract

Mutations in AGXT, a locus mapped to 2q37.3, cause deficiency of liver-specific alanine:glyoxylate aminotransferase (AGT), the metabolic error in type 1 primary hyperoxaluria (PH1). Genetic analysis of 55 unrelated probands with PH1 from the Mayo Clinic Hyperoxaluria Center, to date the largest with availability of complete sequencing across the entire AGXT coding region and documented hepatic AGT deficiency, suggests that a molecular diagnosis (identification of two disease alleles) is feasible in 96% of patients. Unique to this PH1 population was the higher frequency of G170R, the most common AGXT mutation, accounting for 37% of alleles, and detection of a new 3′ end deletion (Ex 11_3′UTR del). A described frameshift mutation (c.33_34insC) occurred with the next highest frequency (11%), followed by F152I and G156R (frequencies of 6.3 and 4.5%, respectively), both surpassing the frequency (2.7%) of I244T, the previously reported third most common pathogenic change. These sequencing data indicate that AGXT is even more variable than formerly believed, with 28 new variants (21 mutations and seven polymorphisms) detected, with highest frequencies on exons 1, 4, and 7. When limited to these three exons, molecular analysis sensitivity was 77%, compared with 98% for whole-gene sequencing. These are the first data in support of comprehensive AGXT analysis for the diagnosis of PH1, obviating a liver biopsy in most well-characterized patients. Also reported here is previously unavailable evidence for the pathogenic basis of all AGXT missense variants, including evolutionary conservation data in a multisequence alignment and use of a normal control population.

Type 1 primary hyperoxaluria (PH1; OMIM 259900) is a rare but potentially life-threatening inborn error of metabolism. Inherited deficiency of a liver-specific enzyme (alanine:glyoxylate aminotransferase [AGT]; E.C.2.6.1.44) causes impaired glyoxylate metabolism in peroxisomes of human hepatocytes (¹). This autosomal recessive trait is invariably characterized by marked hyperoxaluria with or without associated hyperglycolic aciduria, calcium oxalate urolithiasis or nephrocalcinosis, and progressive loss of renal function over time.

In 1990, normal human AGT cDNA was isolated and sequenced (²) (GenBank X53414 and NM_000030). Characterization and mapping of a genomic clone (designated as AGXT) to the telomeric region of chromosome 2 (2q36–37) followed, with ascertainment of a coding composition of 11 exons spread across 10 kb (GenBank M61755 to M61763 and M61833) (³). Southern blotting using an isolated full-length AGT cDNA probe determined that human AGT was encoded singly (³).

Early description of human AGT using protein A-gold immunocytochemistry and isopycnic density gradient centrifugation studies revealed a unique AGT-targeting defect: Mislocalization of 90% of the protein from peroxisomes to mitochondria in some patients with PH1 (^4–6). Cloning followed by sequencing of human AGT cDNA that was isolated from livers of patients with this peroxisome-to-mitochondria mistargeting phenotype showed three sequence variants in the coding region (P11L, G170R, and I340M) (⁷).

Of these, only G170R has been shown to be disease specific, although P11L has been demonstrated to modify disease expression in vitro (⁸). In 2000, Lumb et al. (⁸) showed that presence of P11L alone reduced the activity of AGT by a factor of 3, whereas coexpression with four of the most common mutations (G41R, G170R, F152I, and I244T) caused protein aggregation. These in vitro observations have been corroborated by the fact that in patients with PH1 described so far, three of these four mutations (G170R, F152I, and I244T) seem to segregate only in cis with P11L. It is postulated that inheritance of these common variants opposite P11L would not give rise to the PH1 phenotype (⁸).

Subsequent to detection of the P11L and I340M polymorphisms in patients with PH1 and mitochondrial AGT, Purdue et al. (⁹) also identified a closely linked 74-bp duplication in intron 1 (IVS1 + 74 bp). These three polymorphic variants (P11L, I340M, and IVS1 + 74 bp) are now collectively referred to as the “minor” allele of AGXT. A second normal haplotype of AGXT that lacks these changes is recognized as the “major” allele. Published frequencies for the minor AGXT allele in normal populations range from approximately 2.3% in Chinese to as high as 28% in Saami (¹⁰). In PH1, the frequency of the minor AGXT allele is higher (approximately 50%), attributed to the predilection for more common mutations (G170R, F152I, and I244T) to segregate solely with this allele (¹¹).

As of 2004, there were a total of 55 AGXT sequence variants reported in the Human Gene Mutation Database (www.hgmd.cf.ac.uk), 34 (approximately 62%) of which are missense or nonsense changes. The remaining mutations include six splicing changes, eight small deletions, four small insertions, one small insertion/deletion, and two large deletions. Fifty of these variants, many to date largely unclassified in terms of pathogenicity, were recently summarized by Coulter-Mackie and Rumsby (¹²) in the single available review of published AGXT sequence changes. In a separate report from these same authors, molecular analysis sensitivity was 62% for a large series of 287 probands with liver biopsy–proven PH1, using restriction enzyme-based screening for the now recognized three most common AGXT mutations (G170R, c.33_34insC, and I244T) (¹³). Detection of two mutant alleles was feasible in only 99 (34.5%) patients.

Given the disappointing results of this and earlier reports (^14,15) regarding the application of limited mutation screening using restriction enzyme digestion for the molecular diagnosis of PH1, in this investigation, we assessed the diagnostic relevance of performing whole-gene sequencing. In an effort to establish the pathogenicity of all previously described and newly discovered AGXT missense variants, we also report a classification strategy that is based on the scheme developed by Grantham (¹⁶) while at the same time using evolutionary sequence conservation and normal population data.

Materials and Methods

To determine the feasibility of using comprehensive mutation analysis for establishing a molecular-based diagnosis of PH1 and to expand further on the heterogeneity of AGXT, we sequenced the entire coding region of AGXT in 64 patients with PH1 from the Mayo Clinic Hyperoxaluria Center. A definitive diagnosis of PH1 was based on biochemical evidence along with hepatic enzyme analysis that documented AGT deficiency in the patient (n = 48), an affected sibling (n = 8), a first cousin (n = 1), or supporting molecular data (n = 7). For our normal control population, we screened 50 DNA samples of individuals of predominantly European and North American descent. The study was approved by our institutional review board, and all participants provided informed consent or assent.

Genomic DNA was extracted from peripheral blood leukocytes using standard methods. The primer pairs and PCR reaction conditions that were used to amplify and sequence the 11 exons and exon-intron boundaries of AGXT are listed in Table 1. Primer design was based on the available published genomic sequence of AGXT (GenBank NT_005416). The promoter region of AGXT was not screened. For all PCR reactions, we used 50 to 100 ng of genomic DNA, 5 to 10 pmol of reverse and forward primers, 0.25 U of AmpliTaq Gold (Applied Biosystems, Foster City, CA), and 200 μM dNTP (Invitrogen, Carlsbad, CA) in a total volume of 25 μl, with addition of DMSO for optimization. Amplification (94°C 30 s, Ta 58 to 62 °C 30 s, and 72°C 30 s for denaturation, annealing, and extension steps, respectively) was performed in an MBS Satellite 0.2G Thermal Cycler (Applied Biosystems) for 25 to 30 cycles. PCR products were cleaned using ExoSAP IT, per the manufacturer’s (USB, Cleveland, OH) instructions. Sequencing was performed in both directions using the ABI PRISM 3700 DNA Analyzer (Applied Biosystems), and chromatograms were analyzed with the 4.5 version of Sequencher Software (Gene Codes Corp., Ann Arbor, MI). Positive results were sequenced in duplicate, using a separately amplified PCR product.

To screen for complex alleles (large deletions or insertions) and to increase the sensitivity of molecular analysis, we applied Luminex FlexMAP tag/anti-tag system technology (^17,18) to the multiplex ligation-dependent probe amplification (MLPA) technique initially described by Schouten et al. (¹⁹). In this method, two template-specific probes are designed to screen for gene copy changes in the genetic region of interest: A short probe (approximately 25 bp) that consists of template-specific sequence and universal primer, along with a longer probe (approximately 75 bp) that is made of template-specific sequence, universal primer, and “stuffer” sequence complementary to a selected FlexMap bead. Long probes are phosphorylated to carry out the ligation reaction. The selected intragenic and extragenic probe sequences are listed in Table 2.

For the multiplexing steps, we used the commercially available MLPA Kit from MRC Holland (Amsterdam, The Netherlands). PCR was performed for 24 cycles (30 s at 95°C, 30 s at 60°C, and 60 s at 72°C) using Platinum Taq (Invitrogen). Bead hybridization followed, using 40 μl of FlexMap bead mix plus 10 μl of PCR product, for a 1-h incubation period at 37°C. We then added 25 μl of a streptavidin phycoerythrin/tetramethyl-ammonium chloride solution to each reaction well (2 μl of Streptavidin, R-phycoerythrin conjugate [1 mg/ml] and 250 μl of 1× tetramethyl-ammonium chloride). This mixture was incubated at room temperature for 15 min, and the number of FlexMap beads that successfully hybridized was counted with a Luminex LX100. Fluorescence intensity that was generated from each sample compared with control probes (expressed as a peak ratio) was then used to assess its copy number. In the experience of D.B.D. and P.A.L., this method has proved robust in quantifying gene dosage changes in patients with Niemann-Pick type C (^17,18).

To classify the new AGXT missense variants described here, we developed and applied a scoring system that is based on the matrix of Grantham (¹⁶) and Abkevich et al. (²⁰). Overall scores are provided for all described and newly detected AGXT missense mutations. Finally, we used the 8.1.1 version of the Mac Vector program to generate a multiple sequence alignment of the following AGT orthologs (corresponding GenBank accession numbers): Homo sapiens (BAA02632), Canis familiaris (XP_848328.1), Felis catus (CAA53527.1), Oryctolagus cuniculus (S24155), Rattus rattus (CAA29656.1), Mus musculus (AAH25799.1), Xenopus tropicalis (NP001006705.1), and Danio rerio (AAH76465.1).

Results

Overall, our fully sequenced cohort consisted of 64 patients with PH1 (55 unrelated probands, eight affected siblings, and one first cousin). AGXT genotyping in the 55 unrelated probands is listed in Table 3. The 21 newly discovered mutations (12 missense, three nonsense, three splice site, two microinsertion/deletion, and one large deletion) are shown in boldface type. Direct sequencing of the entire AGXT coding region revealed two disease alleles in 52 (95%) patients and a single pathogenic change in the remaining three. In the unrelated PH1 group as a whole, minor and major allelic frequencies were 58% (64 of 110 alleles) and 42% (46 of 110 alleles), respectively.

Two additional AGXT sequence variants are worthy of special mention. The first is a T → C transition in exon 8 (c.836 T → C, I279T), which was detected in probands 31, 33, 43, and 46, the significance of which was not clear because it was not detected in any of the control subjects screened and was found in tandem with two disease alleles in probands 33, 43, and 46. Previous expression of this variant by Coulter-Mackie et al. (²¹) on the background of the major AGXT allele, however, yielded only a minimal effect on AGT catalytic activity. The second is an intronic change (IVS10–91 G → A) in proband 42, which segregated with disease in her affected pedigree and again was detected in the setting of two other seemingly pathogenic changes.

Of the 48 unrelated probands with PH1 and availability of hepatic enzyme analysis, two disease alleles were satisfactorily detected in 46 (96%), and a single disease allele was detected in two, yielding a sensitivity of 98% (94 of 96 alleles detected) for this whole-gene sequencing approach. When limited to the three exons with the highest mutation frequencies (exons 1, 4, and 7; Table 4), the sensitivity of molecular analysis for the liver biopsy cohort was 77% (72 of 94 alleles detected).

To date, only two large partial AGXT deletions (5′ untranslated region [UTR] to IVS5 and 5′ UTR to IVS7) and a single case of segmental maternal isodisomy of 2q37.3 have been published (^22–24). A likelihood of hemizygosity is also supported by a few other rare (frequency <5%) mutations in PH1 that have been detected only in homozygous form (S205P and G82E) in families of reportedly nonconsanguineous backgrounds (^25,26).

To exclude the possibility that large deletions may have gone undetected in our purely homozygous probands for rare mutations (48, 49, 51, and 52) and in probands with a singly identified mutation (31, 54, and 55), we strategically designed MLPA probes (exons 1, 4, and 11) across the AGXT coding region. Haplotype analysis for probands 48, 49, 51, 52, and 54 is listed in Table 5.

We did not detect changes in gene copy in any of these tested patients, with the exception of proband 54 (Figure 1), in whom we successfully detected a new deletion that involved exon 11, also confirmed separately in an affected sibling. Because the haplotype analysis in this patient suggested a potential partial deletion that affected intron 8 to the 3′ end of AGXT (Table 5), we designed six additional probes (exon 9, 10, 3′ 1 kb, 3′ 2 kb, 3′ 80 kb, and GAL3ST2) to delineate its extent. These added probes verified that this new deletion encompasses solely exon 11 of AGXT and that it extends >2 kb downstream from the 3′ UTR of the gene (Figure 1).

The classification system that was developed here for AGXT (Table 6) represents the first attempt to gauge the pathogenicity of all reported missense variants. Our control population data, including number of alleles tested and allelic frequencies for the seven new (shown in boldface type) and described AGXT polymorphisms in comparison with PH1 is depicted in (Table 7). None of the new AGXT missense variants was detected in the normal control population tested. The multisequence alignment is shown in Figure 2.

Discussion

Using this whole-gene sequencing approach, we show a sensitivity for molecular analysis of 98% in 48 liver biopsy–proven cases of PH1, an improvement of 36% over the previously published restriction enzyme–based screen for the three most common mutations in AGXT (G170R, c.33_34insC, and I244T) (¹³). Comprehensive mutation screening by direct sequencing therefore seems to be a satisfactory method for detection of sequence variation in AGXT and appropriate for molecular diagnosis of PH1. Given the small size of the gene, molecular analysis is relatively inexpensive and easily achieved. Because of the wide range of mutation types detected (missense, nonsense, microinsertion/deletions, and splice variants), direct sequencing of AGXT has the added benefit of contributing to our molecular understanding of PH1, despite the high prevalence of private mutations.

Even if sequencing is limited to the three exons that contain the more common mutations (1, 4, and 7), the sensitivity remains considerably higher (77%) than the reported mutation-specific restriction enzyme approach (62%) (¹³). For diagnostic purposes, a prioritization scheme that consists first of limited sequencing (of exons 1, 4, and 7) and then direct sequencing of the remaining exon and exon-intron boundaries with addition of family studies when indicated seems to be a suitable approach. Mutation analysis may also be targeted to a particular ethnicity for which there are known AGXT associations (e.g., I244T in Spanish populations). A liver biopsy would then be required only when this molecular approach proves nondiagnostic.

Both published reports (^27–31) and data that were obtained here confirm that less common mutations are also found on these same exons (1, 4, and 7), substantiating the strategy to sequence these coding regions directly, in lieu of performing mutation-specific restriction enzyme–based assays. A second exon 4 mutation in particular (F152I) has been reported in sufficient frequency (6.6 to 19%) in two different PH1 populations (Dutch and Canadian) (^23,28) and again here (6.3%), making it a worthwhile addition to the PH1 molecular diagnostic service. G156R, a less common exon 4 mutation, previously reported in patients of Israeli Arab and Italian descent (^30,31), was also found here in a frequency of 4.5%.

We detected G170R, the most common mutation in PH1, in 41 of our 110 unrelated alleles (frequency of 37%). This allelic frequency most closely resembles that of a Dutch PH1 cohort of 33 patients (allelic frequency 43%) (²⁸). The pathogenic basis (peroxisome-to-mitochondria mistargeting) of this change has been well established in vitro (⁸) and is supported by segregation analysis (³²) and by its absence in screened normal controls, both in this cohort and elsewhere (⁷).

The c.33_34insC microinsertion has been documented in people of various ethnicities (^13,27,30), in frequencies (12 to 14%) that qualify for the second most common AGXT mutation. In our cohort, the frequency of c.33_34insC was comparable (11%). In addition to detection of c.33_34insC and other, less frequent changes that are located on exon 1, direct sequencing of this part of the coding region has the advantage of supplying P11L genotyping. Because of the strict association between P11L and IVS1 + 74 bp documented in the past (^9,11), the latter change has been used as a marker for the minor AGXT allele. The recent recognition of a breakage in this linkage, documented in an African population (³³), underscores that IVS1 + 74 bp may no longer be a suitable surrogate for P11L in certain populations.

To date, several mutations have been documented on exon 7 of AGXT (²⁹), the most common of which is I244T, a founder mutation in Spanish patients who originated from a small island of the Canary Islands called La Gomera (³⁴), where its frequency is 92% (³⁵). The reported frequency for I244T has otherwise ranged from 6 to 9% (^11,13). We detected I244T in only three of the 110 unrelated alleles screened (frequency of 2.7%). Both patients (probands 40 and 50) were of Spanish descent (from Spain and southwest United States).

After exons 1, 4, and 7, exon 6 contained the next most frequent number of mutant alleles (five of 110) in our PH1 cohort. Sequence conservation in this part of the coding region is essential to AGT catalytic activity because exon 6 contains the highly conserved Lys209 residue, which is critical for co-factor (pyridoxal phosphate) binding via Schiff base formation. A single sequence change (S205P) has been reported on this exon, in a patient of Japanese descent (²⁵). It is interesting that we detected three new sequence changes (D201E, Y204X, and G216R) in this part of the AGXT coding region. G216R was also just reported in one other patient with PH1 and shown to segregate with disease in the affected family (³⁶).

Additional noteworthy observations that arose from this fully sequenced PH1 cohort included detection of a new 3-bp, in-frame microdeletion (V139del) on exon 3 (proband 43), bringing the total number of reported (¹²) AGXT microinsertion/deletions to 14. V139del is the first reported mutation for this exon, the smallest for AGXT, consisting of only 65 bp (GenBank accession no. M61756). Direct sequencing also facilitated detection of three new splice variants (probands 25, 26, and 30), for an overall frequency of AGXT splice-site variants of approximately 10%.

Excluding our homozygous probands for common mutations (probands 1 to 10, 46, 47, and 50), there were an additional four apparently homozygous patients in our group (probands 48, 49, 51, and 52). Despite the observed absence of heterozygosity across all of the described intragenic AGXT polymorphisms in four of these patients who were purely homozygous for rare mutations, we did not detect any large gene rearrangements, suggesting ancestral founder haplotypes rather than undetected hemizygosity. We did, however, identify a new deletion that involves the 3′ end of AGXT in proband 54 (Ex11_3′UTR del). The haplotype and MLPA data in this proband and her affected sibling suggest that this deletion extends from exon 11 to at least 2 kb downstream from the 3′UTR. The 3′ deletion end point is outside of the coding region; therefore, in contrast to an intragenic breakpoint whose effect may cause a frameshift in the reading frame, the predicted effect of this new deletion is truncating.

Apart from common or well-studied mutations, interpretation and classification of the increasingly detected sequence variation in AGXT are difficult, especially for diagnostics. As such, another goal of our investigation was to provide a classification scheme of pathogenicity. Missense variants in particular, which make up the majority of the described unclassified variants in PH1, are notoriously problematic to characterize in the absence of functional assays, segregation analysis, or normal population frequencies.

Recently, a scheme developed for the BRCA1 gene has become a model for classification of missense variants, taking into account the scoring system of chemical differences between amino acids that initially was developed by Grantham (¹⁶) and sequence conservation data that were taken from multiple sequence alignment (²⁰). Since such an approach had not yet been instituted for PH1, we applied a similar strategy for classifying both our newly discovered and all previously described AGXT missense variants.

Except for T9N, the overall scores that were calculated for our 13 newly discovered missense variants suggest that all are likely pathogenic. Purely based on the Grantham matrix score and sequence alignment data, T9N would not be predicted to be pathogenic, but its absence in any of the tested control subjects argues against its being a polymorphism. Screening in a larger cohort of control subjects may prove otherwise. On the basis of similarly derived data for the 23 already described AGXT missense variants, all are predicted to be pathogenic, whereas (intriguingly) I244T, the Spanish founder mutation, is not. These three methods (Grantham matrix score, multi-sequence alignment score, and normal control population data) therefore can be taken as being complementary rather than exclusive and predictive rather than definitive. As such, confirmatory in vitro functional assays should be performed whenever feasible.

Conclusion

We report our experience with comprehensive AGXT mutation analysis in a cohort of 55 unrelated probands with a definitive diagnosis of PH1. Our data suggest that in a majority of well-characterized patients with PH1 (i.e., those with availability of complete biochemical data and a high clinical index of suspicion), a molecular diagnosis using direct sequencing is feasible, having higher sensitivity (98%) than the current restriction enzyme–based approach (62%) (¹³), even when limited to the three exons with the highest mutation frequencies (77%).

Furthermore, we herewith provide the first pathogenic classification scheme for all new and old AGXT variants of unknown clinical significance via provision of evolutionary conservation and normal control population data. Similar analyses of additional AGXT missense variants that are detected in the future may prove useful in interpreting their functional significance.

Disclosures

None.

Figure 1: Multiplex ligation-dependent probe amplification analysis. The five control (breast cancer anti-estrogen resistance 3 [BCAR3], catenin, β1 [CTNNB], HIR histone cell regulation defective homolog A [HIRA], TNF receptor superfamily 7 [TNFRSF7], and dystrophin [DMD]) and nine experimental probes (AGXT exons 1, 4, 9, 10, 11, 3′ 1 kb, 3′ 2 kb, 3′ 80 kb, and galactose-3-O-sulfotransferase 2 [GAL3ST2]) are listed sequentially above the selected patient panels (female control, male control, proband 54). For AGXT, probes 3′ 1 kb, 3′ 2 kb, and 3′ 80 kb are located 1, 2, and 80 kb from the 3′ untranslated region (UTR), respectively. Large gene rearrangements were not detected in probands 48, 49, 51, and 52 (data identical to controls not shown). In proband 54, a deletion that encompasses exon 11 and extends at least 2 kb from the 3′ UTR is shown. For the control probes, BCAR3, chromosome 1; CTNNB1, chromosome 3; TNFRSF7, chromosome 12; HIRA, chromosome 22; DMD, exon 11, gene dosage control, X-chromosome.

Figure 2: Multiple sequence alignment of alanine:glyoxylate aminotransferase (AGT) orthologs. Conserved amino acids are shown in black, similar amino acids are in gray, and mismatches are in white. A letter indicating the corresponding amino acid change for all newly discovered missense variants is shown directly above the Homo sapiens sequence.

Table 1: Flanking primer pairs and annealing temperatures that were used to amplify the 11 exons and exon-intron boundaries of AGXT

Table 2: MLPA probe sequences for exons 1, 4, 9, 10, and 11 of AGXTa

Table 3: AGXT genotyping in 55 unrelated PH1 probandsa

Table 4: Exon-specific combined mutation frequencies for the 48 unrelated probands with PH1 and availability of hepatic enzyme analysisa

Table 5: Haplotype analysis in the four seemingly homozygous probands for rare mutations and in proband 54 (patient with newly detected partial gene deletion, Ex 11_3′UTR del)a

Table 6: Classification of new and described AGXT missense variants

Table 7: Frequencies of AGXT sequence variants detected in normal control subjects and in patients with PH1a

We kindly thank Dr. Christopher J. Ward for invaluable assistance with bioinformatics, our patients for their gracious participation, and the National Institutes of Health (DK 64865 and DK 73354) and Oxalosis and Hyperoxaluria Foundation for research funding.