Genetic analysis of impaired trimethylamine metabolism using whole exome sequencing

We informed adult subjects with body odor complaints who contacted the study investigators about participating in our ongoing sensory and genetic studies. Approximately 130 subjects made contact during an 8-year period from 1999 to 2007 and we evaluated them for the sensory and metabolic arms of the study. From this pool of subjects, we chose ten at random for WES.

For the choline challenge test, we instructed subjects to fast from 10 pm the night before testing and collect their first morning urine while at home. This collection was done using a large, plastic container provided in advance; as previously described [5]. The container was pre-loaded with 2 ml of 6 N HCl to convert free TMA to its non-volatile hydrochloride salt. After the subjects arrived in the lab and following the recording of first sensory measures as described below, they ingested 5 g choline dissolved in orange juice. Each subject took three containers home and collected urine over the next 24 h, using one container for each 8-h period. Subjects returned the following day with their urine. The volume in each container was recorded and a 100 ml aliquot removed to a smaller plastic container and frozen. These urine samples were shipped on dry ice and assayed for TMA and TMAO at the University of Colorado Health Sciences Center following previously published procedures [5], with steps for standard solution preparation, water/urine test solution preparation for TMA analysis, and choline loading analysis. We quantified the amount of TMA for each 8-h sample; we measured the sample with the highest concentration of TMA for TMAO. We used the ratio of TMAO to TMA as the key outcome measure: normal reference ranges are ≥0.90.

For the sensory evaluation of body odor, subjects were asked to come to the laboratory fasted and without having brushed their teeth after 10 pm (22:00) the night before; in addition, subjects were asked to refrain from having applied any scented products (e.g., cosmetics and deodorant products) for 3 days prior to testing. Upon arrival at the laboratory, two trained sensory judges evaluated their breath, axillary and upper body odor with quality descriptors (e.g., fishy or sulfurous) using previously described procedures [7]. We repeated these ratings the following morning, after the choline challenge test.

When subjects returned their urine to the laboratory, they gave a 20 ml sample of whole blood by venipuncture. DNA was extracted (QIAamp DNA Mini Kit; Qiagen, Valencia, CA, USA), quantified by spectrophotometry (Nanodrop, Wilmington, DE, USA), and used in the exome analysis and for other follow-up genotyping.

Genomic DNA of 10ug each from ten subjects was part of a large scale WES project of rare genetic diseases between the Center for Applied Genomics at the Children’s Hospital of Philadelphia and BGI. Exomes were captured using the Agilent SureSelect Human All Exon Kit V2 (Agilent Technologies, Santa Clara, CA, USA) and sequenced using the HiSeq 2000 machine (Illumina, San Diego, CA, USA) with standard paired-end sequencing protocol. We conducted subsequent bioinformatics as previously described [24]. Briefly put, raw sequencing reads were stored as FASTQ files and then aligned to the human reference genome (UCSC hg19) with the Burrows-Wheeler alignment [25]. Variant calling was performed using the Genome Analysis Tool Kit (version 1.4) [26] followed by functional annotation using Annovar [27] and SnpEff [28]. We identified all variant sites in a larger set of exomes, and genotypes for any location with one or more minor alleles were determined for all subjects. Variant calling was also performed independently at BGI as previously described [29], and high concordance of results was obtained. For pathogenicity, each variant was assigned two scores using algorithms implemented in Phenotyping version 2 (PolyPhen-2) and Sorting Intolerant From Tolerant (SIFT) [30, 31], respectively. Genes whose products have oxidoreductase activity or that are in the same pathway as FMO3 were identified through a Gene Ontology website (N = 729) [32]. We also identified variants where two or more subjects had the same rare minor allele. The final filter was for the minor allele frequency (MAF) <0.05 as reported in the 1000 Genomes data set [33]. We identified variants that met the criteria for potential pathogenicity and rarity, and shared by two or more subjects.

Harnessing gene-gene/protein-protein network relationships, we generated a list of genes that may interact with FMO3, PYROXD2 and DMGDH, and investigated rare deleterious variants within these genes. The interacting partners were collected from the STRING database [34] with default parameters below, setting 1) all possible active interaction sources (including text mining, experiments, databases, co-expression, neighborhood, gene fusion, and co-occurrence); 2) minimum required interaction score of 0.4; 3) max number of interactors to show as 10.

We partially Sanger sequenced DNA from each subjects for the FMO3 gene and conducted Taqman genotyping of the FMO3 and PYROXD2 variants to gauge WES reliability. For Sanger sequencing, Exons 2–9 of the FMO3 gene were amplified by validated primers [3] and submitted for analysis to the DNA Sequencing Facility at the University of Pennsylvania Perelman School of Medicine. Sequencing reactions were performed in an ABI GeneAmp 9700 thermal cycler, resolved with an ABI 3730 DNA sequencer, and analyzed using ABI Sequencing Analysis software, version 5.1 (Applied Biosystems, Foster City, CA, USA). Sequences were aligned using the algorithms implemented in Sequencher (Gene Codes Corporation, Ann Arbor, MI, USA), and the variants were identified [35] and matched against those obtained by WES. We used Taqman methods to type four specific variants, three in the FMO3 gene (rs2266780, rs2266782, and rs2066532) and one in the PYROXD2 gene (rs7072216). The PYROXD2 variant is located in the intronic region thus not captured by WES, but it was reported in a previous paper for significant association with TMA in urine [21]. Appropriate primers and probes were cycled (StepOnePlus, Life Technologies, Grand Island, NY), and variants were called using previously described methods [36].

As summarized in Table 1, of the ten subjects studied here, eight were female and two were male. Six subjects reported their race as Caucasian, and four as African American. Table 1 also summarizes subject demographics and experimental outcomes. All subjects produced abnormally low amounts of TMAO from choline and met the criterion for TMAU, but the TMAO:TMA ratios ranged widely, from 0.13 to 0.87 (normative values >0.90). However, the sensory panel judged most subjects to have unremarkable body odors, with one exception: subject 98 had musty and damp odor. No subjects at the time of evaluation had a fishy odor.

From the larger sample set of the parent project (n = 669), 817,028 SNP variant sites and 75,781 indel sites were identified (Table 2 and Additional file 1: Table S1). We found seventy percent of these variants in the homozygous state. Due to differences in sequencing coverage missing genotypes rate varied among individuals and accounted for 15 to 23% of SNP variant sites and 25 to 33% of total indel variant sites.

We identified one rare allele within the FMO3 gene. This subject (#99) was heterozygous for a nonsense variant rs72549325 (chr1:g.171076936G > T; NM_001002294.2:c.442G > T; NP_001002294.1:p.Gly148Ter). We also identified subjects with common variants. Five subjects were heterozygous for rs2266782 (chr1:g.171076966G > A; NM_001002294.2:c.472G > A; NP_001002294.1:p.Glu158Lys), one subject was heterozygous for rs1736557 (chr1:g.171080080G > A; NM_001002294.2:c.769G > A; NP_001002294.1:p.Val257Met) and one subject was heterozygous for rs2266780 (chr1:g.171083242A > G; NM_001002294.2:c.923A > G; NP_001002294.1:p.Glu308Gly). Although common in the population, these three variants of the FMO3 gene have been associated with impairments in metabolism [37‐39]. For the PYROXD2 gene, the allele of interest is common and we found it in five subjects, three of whom were homozygous. We summarize information about these FMO3 and PYROXD2 gene variants in Table 3. In addition, we list 23 more FMO3 rare gene variants that have role in TMAU in Additional file 1: Table S2. We detected none of these variants in the ten people studied here.

We next looked beyond the known genes and their variants previously implicated in TMAU or TMA metabolism for variants that are rare and predicted to have reduced or abolished function. Specifically we focused on novel variants of which at least one of the ten subjects was homozygous for the alternative alleles, identifying 94,535 SNPs and 5,212 indels. The pathogenicity of these variants was predicted by algorithms implemented in either PolyPhen-2 or SIFT; 1,867 SNPs and 163 indels were predicted to be deleterious. Among these pathogenic variants, 57 SNPs and four indels were in oxidoreductase pathways different from that of the FMO3 gene, and four of these SNPs and one indel were rare, with MAF < 0.05. Four of these five variants were found in three subjects with the lowest TMAO:TMA ratios. In addition, within other genes with no known function in TMA metabolism, two subjects shared these rare variants for nine rare putatively pathogenic SNPs and one rare putatively pathogenic indel. We depict the filtering process for the variants in Fig. 1 and we list these results in Table 4. We also conducted the gene-gene network analysis inspired by STRING results, and similar variant analysis was expanded to genes interacted with each of FMO3 (Additional file 1: Figure S1), PYROXD2 (Additional file 1: Figure S2) and DMGDH (Additional file 1: Figure S3). We found rare putatively pathogenic SNPs in multiple genes interacting with DMGDH, including BHMT2 (73% identical to the homologous BHMT which is directly interacted with DMGDH), SARDH and SHMT1 (Table 4). All of the subjects with these SNPs were heterozygous. We found no rare deleterious variants from genes that are currently known to interact with FMO3 or PYROXD2.

We show a summary of the number and type of gene variants detected for each subject in Table 5. Six subjects had variants in the FMO3 gene that reduce or abolish function of the protein but four did not. However, those subjects did have rare variants predicted to be deleterious in other genes that might account for their inability to metabolize TMA. For instance, one subject (with #52) had a very low ratio of TMAO:TMA (0.13) and while she did not have any detectible variants in the FMO3 gene, she did have two rare variants in other genes annotated as oxidoreductases (CP and TXNDC2). Similarly, another subject (#35) had no detectible FMO3 gene variants but was homozygous for the PYROXD2 variant that is associated with high TMA values in urine among healthy people (rs7072216).

When comparing the results of Sanger sequencing and Taqman genotyping versus WES, 29 of the 29 of the Sanger genotypes matched and 30 out of 30 of the Taqman genotypes matched (Additional file 1: Table S3 and S4). Overall, 60 of 60 genotypes matched for a concordance rate of 100%.