Mutation in WDR4 impairs tRNA m7G46 methylation and causes a distinct form of microcephalic primordial dwarfism

1Primordial dwarfism (PD) is a term used to describe a wide range of phenotypes that have in common severe prenatal growth deficiency (>3 SD below the mean) that persists postnatally. Although extremely rare, the monogenic nature of PD lends itself readily to gene mapping approaches thus representing a unique resource for understanding the biological networks that control growth through the discovery of genes that are mutated in this condition.

2Impaired DNA damage repair is among the earliest identified mechanisms in PD as revealed by the discovery that ATR is mutated in patients with Seckel syndrome, a clinical subtype of PD characterized by microcephaly and distinct facial features. The same mechanism is invoked in PD caused by mutations in ATRIP, BRCA2, DNA2, and XRCC4. Impaired mitosis due to centrosomal abnormalities has now emerged as a major mechanism underlying many forms of PD. Less common forms of PD were found to be caused by mutations in genes involved in replication licensing, splicing and serine synthesis. Despite the remarkable acceleration of PD disease gene discovery in recent years, one-third of the cases remain undiagnosed molecularly, which suggests that additional disease genes likely exist and these might further expand the known molecular network that controls growth.

3tRNA is a well-studied class of non-coding RNA that plays an essential role in protein synthesis by transferring amino acids to the growing peptide chain as the corresponding mRNA is being decoded by the ribosomal translational machinery. A remarkable multitude of modification reactions (>100) are known, which are often highly conserved in different organisms, including in prokaryotes and archaea, clearly suggesting their importance. Our knowledge of the biology of tRNA modification comes primarily from work on the yeast Saccharomyces cerevisiae and other model organisms. In general, modifications in the tRNA anticodon loop are critical for translational efficiency, frame maintenance, and fidelity, and lack of these modifications often leads to lethality, slow growth, and/or other phenotypic effects. Modifications to the body of the tRNA are generally involved in tRNA folding and stability, and lack of any of several different body modifications in yeast causes temperature sensitivity due to rapid tRNA decay (RTD) of specific tRNAs.

4The recent identification of several links between tRNA modification and human disease have spurred increased interest in this field and its potential to explain the pathogenesis of clinically relevant disorders. In this study, we describe an apparently novel clinical condition characterized by primordial dwarfism and a unique set of additional features. We show that the two families affected by this disorder map to WDR4, the human ortholog of Trm82, which is required for formation of the highly conserved m7G46 (7-methylguanosine) modification of tRNA. The m7G46 modification occurs widely in prokaryotes and eukaryotes, and in S. cerevisiae requires a holoenzyme comprised of the Trm8 methyltransferase subunit and its WD40 repeat-containing binding partner Trm82, which appears to be involved in maintaining Trm8 levels, and in helping Trm8 maintain an active conformation. Yeast trm82Δ mutants, like trm8Δ mutants, are mildly temperature sensitive due to lack of m7G in their tRNA, and have synthetic genetic interactions with trm4Δ mutants (lacking m5C), and a number of other modification mutants, resulting in a severe temperature sensitive growth defect. Human WDR4 and METTL1 are the likely orthologs of S. cerevisiae TRM82 and TRM8 based on homology, and on their complementation of yeast mutants lacking m7G46. Here we show that the WDR4 mutation affects m7G46 methylation suggesting a potential mechanism for this novel form of PD.

5This female infant was born to a 20-year-old mother and 26-year-old father by normal vaginal delivery at 37-weeks’ gestation. The parents are healthy first cousins (Fig. 1a). The family history is non-contributory. The couple had a subsequent boy who died a few days after birth because of growth retardation and multiple congenital heart anomalies. During the gestation of patient 1, the pregnancy was complicated by threatened abortion in the first trimester. Intrauterine growth retardation and weak fetal movements were also documented. The birth weight was 1,600 g. The birth length and head circumference were not recorded but mentioned to be small. She was referred to the Clinical Genetics Department at the age of 4 months because of poor gain of weight and for genetic counseling. On clinical examination she was noted to have a head circumference of 31.5 cm (−5 SD) weight of 2,800 g (−6 SD), and length of 48 cm (−5 SD). The patient had a high forehead, prominent eyes, depressed nasal bridge, short philtrum, tented upper lip and bulged alveolar ridge, and prominent ear lobule (Fig. 1b).

6Her neurological evaluation revealed hypertonia, brisk deep-tendon reflexes, with flexor plantar responses. At that age, radiological examination showed unossified pubic bones, proportionately short femora, and broad metaphyses of the femora and tibiae. The tibiae and fibulae were short and equal in length.

7At the age of 10 months, she developed seizures with only partial response to valproate and lamotrigine. EEG records revealed low voltage slow waves (3–6 cycles/s) mixed with sleep spindle.

8At the age of 20 months, her weight, length, and head circumference were 3,500 g (−6.5 SD), 55 cm (−9.5 SD), and 32 cm (−10.7 SD), respectively. The girl was spastic with contracture deformity in the elbows and hands. Her neurological evaluation revealed hypertonia, brisk deep-tendon reflexes, with flexor plantar responses. She had not acquired any developmental milestones and never recognized her mother. Abdominal ultrasonography and echocardiography showed normal results. Ophthalmological examination showed bilateral optic atrophy.

9Chromosomal examination from peripheral blood lymphocytes and high resolution banding technique revealed normal female karyotype 46,XX.

10Cranial MRI showed partial agenesis of corpus callosum, and abnormal gyral pattern most pronounced posteriorly.

11Diagnosis of microcephalic primordial dwarfism was made at this time based on significant pre- and post-natal growth retardation. Because of the partial overlap with microcephalic osteodysplastic primordial dwarfism I, analysis of RNU4ATAC was undertaken but revealed negative results.

12Patient 2 was the fifth child born to a 36-year-old mother and 39-year-old father. The parents are maternal and paternal first cousins (double consanguineous) (Fig. 1a). The second pregnancy resulted in a similarly affected boy who died at the age of 9 months because of pneumonia accompanied by uncontrolled seizures. A postmortem examination was not performed. There was no other family history of note. Patient 2 was delivered vaginally weighing 1,500 g (−3.7 SD) at 38 weeks’ gestation. Her birth length and head circumference were not recorded. At age 4 months she experienced focal seizures that later evolved into generalized tonic-clonic seizures controlled by combination sodium valproate, levetiracetam, and lamotrigine therapy. EEG records revealed high voltage delta waves (2–4 cycles/s). Abdominal ultrasonography and echocardiography showed normal results.

13At the age of 7 months, she was referred to our Clinical Genetics Department because of the microcephaly and poor weight gain for genetic counseling. On examination, her weight, length and head circumference were 4,200 g (−4.7 SD), 53 cm (−5.7 SD), and 33.2 cm (−7 SD), respectively. Her facial features were quite similar to patient 1 showing rounded face with high forehead, prominent eyes, depressed nasal bridge, short philtrum, bow shaped mouth, and prominent alveolar ridge (Fig. 1c). Her neurological evaluation revealed hypertonia, brisk deep-tendon reflexes, with flexor plantar responses. Ophthalmological examination showed bilateral optic atrophy.

14She presented at the age of 9 months with high fever and chest infection that diagnosed as pneumonia. This was accompanied by status epilepticus and she went into coma for 30 days.

15Follow-up at the age of 17 months, her weight, length, and head circumference were 6,500 g (−3.8 SD), 60 cm (−6.7 SD), and 34.5 cm (−8.9 SD), respectively. The girl made almost no developmental progress and could not recognize the surroundings. Oro-dental examination showed thick alveolar ridge more in the upper than lower and high arched palate.

16Routine biochemical and metabolic screening parameters were within normal ranges.

17Chromosomal examination from peripheral blood lymphocytes and high resolution banding technique revealed normal female karyotype 46,XX. Radiographic examination of the long bones showed proportionately short long bones with broad metaphyses.

18Cranial MRI showed partial agenesis of corpus callosum, and abnormal gyral pattern most pronounced posteriorly (Fig. 1d–g).

19No developmental progress was observed on the last examination of both patients. They were not able to raise their heads or roll, and never laughed. They were unable to follow visually, recognize their mothers, or make eye-to-eye contact. There were no vocalizations beyond an infrequent moaning when in discomfort. Patient 2 sometimes required a tube feeding at the age of 9 months.

20Although the two families have different surnames, they come from the same geographic location in Egypt, raising the possibility of a founder mutation. Indeed, autozygosity mapping and haplotype analysis revealed a single shared homozygous haplotype between the two available patients (chr21:43,809,418-44,828,031 (GRCh37/hg19)) spanning 14 RefSeq genes (Fig. 2a). Whole exome sequencing (WES) was performed separately on each index and the resulting variants were filtered based on frequency (novel or <0.0001), zyogsity (homozygous), position (within the autozygome of the corresponding samples), and nature of the variant (coding/splicing, excluding synonymous changes and those predicted to be benign by two independent in silico prediction tools). Although each index had a few variants that survived these filters, only one variant was shared by the two (WDR4, NM_033661.4:c.509G > T; p.Arg170Leu) (Fig. 2b, c, Additional file 1: Table S1). Reassuringly, this variant was also within the single shared homozygous haplotype, that is, this is the only novel coding/splicing homozygous variant within the critical locus (Additional file 2: Table S2). Segregation analysis using available family members confirmed that only the two patients were homozygous. This mutation is absent in 615 in-house Saudi exomes, 1000 Genomes, Exome Server, and ExAC Browser. It is predicted to be pathogenic by PolyPhen (0.993), SIFT (0), and CADD (PHRED: 20.3).

21We reasoned that homozygosity of the WDR4-R170L allele would result in defective m7G46 modification of tRNA, since WDR4 is the likely human ortholog of S. cerevisiae Trm82, since the corresponding residue in other eukaryotes is almost always an arginine and occasionally a lysine, since this residue is within the most highly conserved region of the Trm8/WDR4 family (Fig. 2c), and since residue K223 of yeast Trm82 (corresponding to R170 in human WDR4) forms a salt bridge with residue E204 of Trm8 that is speculated to be important for maintaining Trm8 in an active conformation (Fig. 3a). We therefore used S. cerevisiae as a model to analyze the effects of the WDR4-R170L mutation. We generated a low copy (CEN) plasmid expressing the S. cerevisiae trm82-K223L from its native promoter (CEN LEU2 PTRM82-trm82-K223L) to test its ability to complement an S. cerevisiae trm82Δ mutant in a trm4Δ background, to amplify the growth defects of the trm82Δ mutation. We also generated plasmids bearing a trm82-K223R variant to determine if, as expected, arginine and lysine residues are interchangeable at this location, and a trm82-K223E variant to test the effects of completely abrogating the salt bridge.

22We found that expression of trm82-K223L suppressed the growth defect of the trm82Δ trm4Δ strain at temperatures up to 37 °C on rich (YP + dextrose) or minimal medium (S -Leu + dextrose), but not at 39 °C (Fig. 3b), indicating that the Trm82-K223L variant is defective for function at high temperature, but is not a complete loss-of-function mutation. By contrast, expression of trm82-K223E did not suppress the growth defect of a trm82Δ trm4Δ strain at any temperature, consistent with a null phenotype, whereas expression of trm82-K223R suppressed the defect at all temperatures (Fig. 3b), demonstrating that arginine and lysine are interchangeable at this residue.

23To determine the extent to which m7G levels were affected in the trm82-K223L mutant (trm82Δ (CEN LEU2 PTRM82-trm82-K223L)), we analyzed modification levels of two of the 11 yeast tRNA species known to have m7G after growth at 30 °C or 37 °C, by purification of the corresponding tRNA species from bulk RNA, followed by nuclease digestion and analysis of nucleosides by HPLC. Strikingly, we found that tRNAPhe from the trm82-K223L mutant had no detectable m7G after growth at 37 °C, but only mildly reduced m7G levels after growth at 30 °C, compared to levels in the wild type or the trm82-K223R controls (0.21 moles/mole vs. 0.34 or 0.32, respectively; Table 1). Other modifications of tRNAPhe were unaffected in the trm82 mutants. By contrast, we observed a milder reduction in m7G levels of tRNAVal(AAC) from the trm82-K223L strain after growth at 37 °C, compared to that from the wild type or trm82-K223R controls (0.25 moles/mole vs. 0.50 and 0.55, respectively; Table 1), and no obvious reduction in m7G levels of tRNAVal(AAC) when strains were grown at 30 °C (0.45 moles/mole vs. 0.53 and 0.46, respectively; Table 2). The temperature-sensitive reduction in m7G levels found on both tRNAPhe and tRNAVal(AAC) from the trm82-K223L mutant is consistent with its temperature sensitive complementation of a trm82Δ trm4Δ strain (Fig. 3b).

24We note that growth of the trm82Δ and trm82-K223L strains at 37 °C resulted in substantially increased levels of m5C on tRNAVal(AAC) (Table 2), similar to the increase in m5C levels observed previously when wild type cells were grown under stress conditions. Unexpectedly, tRNAVal(AAC) in the trm82Δ mutant also had acquired an m2G modification (0.54 and 0.45 moles/mole at 30 °C and 37 °C, respectively, compared to barely detectable levels in the wild type strain). Since tRNAVal(AAC) in the trm82-K223L strain acquired 0.17 moles/mole m2G at 37 °C relative to wild type (from 0.10 to 0.27 moles/mole) while m7G levels were reduced by 0.25 moles/mole (from 0.50 to 0.25), we speculate that m2G levels inversely correlate with m7G levels on this tRNA. Other tRNAVal(AAC) modifications (pseudouridine, inosine, and m1G) were unaffected by the trm82 mutations (Table 2).

25To further define the defect of the WDR4-R170L allele and its relationship to PD, we analyzed the m7G levels in tRNAVal(AAC) and tRNAPhe purified from lymphoblastoid cell lines (LCLs) derived from the PD patient 1 (14DG1160), as well as from control LCLs derived from the healthy mother (heterozygous) and a healthy brother (homozygous wild type WDR4). We found a substantial decrease in m7G on tRNAPhe from the PD LCL (0.08 moles/mole in the WDR4-R170L homozygote vs. 0.49 and 0.46 for the homozygous wild type WDR4 LCL and the heterozygous LCL, respectively; Fig. 3c; Table 3), similar to m7G levels in tRNAPhe from the corresponding yeast trm82-K223L mutant (Table 1). By contrast, levels of the control tRNAPhe modifications Cm, Gm, m2G, pseudouridine, and m5C modifications were similar in all of the LCLs. We also observed a minor reduction of m7G on tRNAVal(AAC) (0.35 moles/mole in the WDR4-R170L homozygote vs. 0.52 and 0.46 for the wild type WDR4 homozygote and heterozygotes, respectively; Table 3), with similar amounts of each of four control modifications observed in the three different LCLs. Thus, our results strongly suggest that the WDR4-R170L mutation causes defects in m7G46 modification in the PD patients.

26Our results demonstrate a clear temperature sensitivity in yeast caused by the trm82-K223L mutation, as reflected by m7G levels in both tRNAPhe and tRNAVal(AAC), and a clear importance of the identity of residue 223, based on the complete lack of complementation of a trm82-K223E mutant. These results cause us to speculate that the Trm82-K223 variant might itself be temperature sensitive, or have a temperature sensitive interaction with Trm8, through the salt bridge described between K223 of Trm82 and E204 of Trm8. It also seems highly likely that the WDR4-R170L allele encodes a protein with similar biochemical properties to the Trm82-K223L variant, based on the high degree of conservation between these regions of Trm82 and WDR4, as well as the similarly reduced levels of m7G modification in the WDR4-R170L LCLs.

27Several human disorders have been associated with perturbation of tRNA modification, although the level of evidence in support of causal links varies. For example, several tRNA modification genes have been found to be significantly dysregulated in various cancers and some have even been used as specific biomarkers, for example, TRMT2A in breast cancer. Associations between variants in genes involved in tRNA modifications and some phenotypes have been reported, for example, ELP4 and epilepsy, and IKBKAP and bronchial asthma. A more directly causal connection was established for several Mendelian disorders. We previously reported that a point mutation in ADAT3 causes intellectual disability. ADAT3 encodes the likely homolog of yeast TAD3, which is the non-catalytic subunit of the complex required for I34 modification of substrate tRNAs. Thus, the association of WDR4 with PD is the second report of a mutation in the non-catalytic subunit of a tRNA modification enzyme that results in disease. More closely related to the PD phenotype we report in this paper are previously reported mutations in TRM10A, the likely human homolog of yeast TRM10 required for m1G9 modification, which cause microcephaly and short stature. Similarly, mutations in human NSUN2, which is required for m5C modification of body residues 48, 49, and 50 in mammals, as well as C34 of the anticodon, cause variable phenotypes that include microcephaly as a feature.

28The pathogenesis of Mendelian diseases caused by tRNA modification genes remains unclear but the predilection to CNS involvement raises interesting possibilities about the vulnerability of the brain to any perturbation in the tight regulation of tRNA modification and, presumably, the consequences of such perturbation on protein synthesis. The mutation we previously reported in ADAT3, for example, is now the single most common cause of autosomal recessive intellectual disability in Saudi Arabia. In this regard, the neurological phenotype we observed in patients with WDR4 mutation (severe microcephaly, agenesis of corpus callosum, and lissencephaly) appears to follow the same pattern of bias towards brain involvement.

29Severe growth deficiency in the syndrome we describe is not limited to the brain but rather generalized, resulting in microcephalic primordial dwarfism. It is unclear if this phenotype is caused by reduced proliferation, increased apoptosis or a combination of the two. Our finding that LCLs derived from a PD patient homozygous for the WDR4-R170L allele had drastically reduced levels of m7G in their tRNAPhe and modestly reduced m7G in their tRNAVal(AAC) strongly suggests that defective modification is a major contributor to the disease pathogenesis in these patients. However, it is not clear if the PD is caused specifically by the m7G defects in tRNAPhe and/or tRNAVal(AAC), by defects in one or more of the other five human tRNA species known to have m7G46, or by defects in other uncharacterized tRNA species. We note that yeast mutants lacking modifications often have growth defects due to only a subset of the tRNA species lacking those modifications.

30The reduction in m7G levels in human tRNAs might result in degradation of one or more specific tRNA species, resulting in reduced or aberrant translation. Yeast mutants lacking m7G, or m7G and m5C, in their tRNA are temperature sensitive for growth due to 5'-3' exonucleolytic degradation of tRNAVal(AAC) by the RTD pathway, which degrades the tRNA because of its more exposed 5' end. Similarly, HeLa cells treated with siRNA to METTL1 and NSUN2 to reduce m7G and m5C modification undergo loss of tRNAVal(AAC) and an accompanying loss of cell viability upon 5-fluorouracil treatment. Reduced levels of specific tRNA species due to reduced m7G levels would be expected to lead to defects in translation and/or reduced growth. However, it is not known if there are other translation effects specifically due to reduced levels of m7G on tRNA, as has been reported for some modifications in the anticodon loop under certain growth conditions.

31Our demonstration that PD patients with the WDR4-R170L missense mutation have defects in m7G levels on tRNA adds to the list of nine known or predicted human modification genes for which mutations have been strongly linked to human disease or pathologies. The underlying molecular causes of these disease associations, and why specific modification defects appear to result in specific disease manifestations remains to be determined. Interestingly, there are forms of microcephalic PD caused by impaired splicing, presumably resulting in impaired protein synthesis, and this may also be invoked as a potential mechanism in PD patients with abnormal tRNA modification although future research will be required to investigate that possibility.

32PD was defined as growth parameters >3 SD below the mean at birth that persist postnatally. Patients were enrolled only after their parents signed a written informed consent form under an IRB-approved protocol (KFSHRC RAC#2080006). Venous blood was collected in EDTA tubes and, when possible, in sodium heparin tubes for DNA extraction and LCL establishment, respectively. The research was carried out in accordance with the principles of the Declaration of Helsinki.

33Autozygosity mapping was performed as described before. Briefly, genomewide genotyping using Axiom SNP chip array was followed by mapping of runs of homozygosity (ROH) >2 Mb in size as surrogates of autozygosity using AutoSNPa, which also delineates the haplotype structure thus allowing the detection of haplotype sharing across families. Exome capture was performed using TruSeq Exome Enrichment kit (Illumina) following the manufacturer’s protocol. Samples were prepared as an Illumina sequencing library, and in the second step, the sequencing libraries were enriched for the desired target using the Illumina Exome Enrichment protocol. The captured libraries were sequenced using Illumina HiSeq 2000 Sequencer. The reads were mapped against UCSC hg19 (http://genome.ucsc.edu/) by BWA (http://bio-bwa.sourceforge.net/). The SNPs and Indels were detected by SAMTOOLS (http://samtools.sourceforge.net/). Variants from WES were filtered such that only novel (or very low frequency 0.1 %), coding/splicing, homozygous variants that are within the autozygome of the affected fetus and are predicted to be pathogenic were considered as likely causal variants. Frequency of variants was determined using publically available variant databases (1000 Genomes, Exome Variant Server, and ExAC) as well as a database of 630 in-house ethnically-matched exomes. Pathogenicity was likely if the mutation is loss of function (splicing/truncating) or, in the case of missense/in-frame indels, removes a highly conserved amino acid and is predicted to be pathogenic by the three in silico prediction modules PolyPhen, SIFT, and CADD.

34Wild-type BY4741 (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0), its trm82Δ::kanMX derivative, and the homozygous trm82Δ::natMX trm4Δ::kanMX diploid (AA0176) were described previously, as was the (CEN LEU2 PTRM82-TRM82) plasmid (AVA0279). Plasmids expressing the S. cerevisiae TRM82-K223L (pMG569A), TRM82-K223R (pMG570A), and TRM82-K223E (pMG571A) variants were generated by QuickChange PCR according to manufacturer’s instructions (Stratagene), and the variant gene was then ligated into the original parent vector to eliminate mutations in the vector that could be introduced by PCR. All plasmids were confirmed by sequencing before use.

35LCLs were grown at 37 °C in 5 % CO2 in RPMI 1640 medium containing FBS (15 %), penicillin (1 U/mL), streptomycin (1 μg/mL), and amphotericin b (0.5 μg/mL) to a density of approximately 1.0 × 106 cells/mL, and bulk RNA from approximately 3.6 × 108 cells was extracted with TRIzol (Life Technologies) according to manufacturer's instructions. S. cerevisiae strains were grown at indicated temperatures to mid-log phase in synthetic (S) dropout media containing dextrose, and bulk low molecular weight RNA was extracted from 300 OD-mL pellets as previously described. For purification of individual tRNAs, appropriate 5’ biotinylated oligonucleotides were used to first purify tRNAPhe from RNA preparations as previously described, followed by purification of tRNAVal(AAC) from the remaining bulk RNA.

36Purified tRNA was digested with P1 nuclease and phosphatase as previously described, and nucleosides were subjected to HPLC analysis at pH 7.0 as previously described.

37Data used in this paper come from a small and well-defined family. To protect the identity of individuals, these confidential data are not publicly available.

38Written informed consent was obtained from the patient’s guardian/parent/next of kin for the publication of this report and any accompanying images.