Defects in tRNA anticodon loop 2'-O-methylation are implicated in non-syndromic X-linked intellectual disability due to mutations in FTSJ1

Abstract

tRNA modifications are crucial for efficient and accurate protein synthesis, and modification defects are frequently associated with disease. Yeast trm7Δ mutants grow poorly due to lack of 2'-O-methylated C₃₂ (Cm₃₂) and Gm₃₄ on tRNA^Phe, catalyzed by Trm7-Trm732 and Trm7-Trm734 respectively, which in turn results in loss of wybutosine at G₃₇. Mutations in human FTSJ1, the likely TRM7 homolog, cause non-syndromic X-linked intellectual disability (NSXLID), but the role of FTSJ1 in tRNA modification is unknown. Here we report that tRNA^Phe from two genetically independent cell lines of NSXLID patients with loss of function FTSJ1 mutations nearly completely lacks Cm₃₂ and Gm₃₄, and has reduced peroxywybutosine (o2yW₃₇). Additionally, tRNA^Phe from an NSXLID patient with a novel FTSJ1-p.A26P missense allele specifically lacks Gm₃₄, but has normal levels of Cm₃₂ and o2yW₃₇. tRNA^Phe from the corresponding Saccharomyces cerevisiae trm7-A26P mutant also specifically lacks Gm₃₄, and the reduced Gm₃₄ is not due to weaker Trm734 binding. These results directly link defective 2'-O-methylation of the tRNA anticodon loop to FTSJ1 mutations, suggest that the modification defects cause NSXLID, and may implicate Gm₃₄ of tRNA^Phe as the critical modification. These results also underscore the widespread conservation of the circuitry for Trm7-dependent anticodon loop modification of eukaryotic tRNA^Phe.

Introduction

The numerous post-transcriptional modifications of tRNA are crucial for accurate and efficient translation of the genetic code. In the yeast Saccharomyces cerevisiae, mutations affecting 16 of the 25 tRNA modifications lead to distinct phenotypes, including lethality for three mutants, and poor growth or temperature sensitivity for six other mutants (Hopper, 2013). Modifications in and around the tRNA anticodon loop (residues 31–39) are particularly important in all organisms (de Crecy-Lagard, et al., 2012), often affecting decoding (Agris, et al., 2007; Murphy, et al., 2004), charging by the cognate tRNA aminoacyl synthetase (Muramatsu, et al., 1988; Putz, et al., 1994), and/or frame maintenance (Bekaert and Rousset, 2005; Urbonavicius, et al., 2001; Waas, et al., 2007), whereas modifications in the body of the tRNA often contribute to folding or stability (Hall, et al., 1989; Helm, et al., 1999; Whipple, et al., 2011; Yue, et al., 1994) and are required to avoid decay by two known degradation pathways (Alexandrov, et al., 2006; Chernyakov, et al., 2008; Kadaba, et al., 2004; Kadaba, et al., 2006; LaCava, et al., 2005; Schneider, et al., 2007; Vanacova, et al., 2005).

Emerging evidence shows that tRNA modifications have important roles in human health. Mutations in nine predicted human homologs of tRNA modification genes have been strongly linked to specific diseases, seven of which are linked to neurological disorders, including six linked to intellectual disability (ID) (Abbasi-Moheb, et al., 2012; Alazami, et al., 2013; Anderson, et al., 2001; Cuajungco, et al., 2003; Fahiminiya, et al., 2013; Freude, et al., 2004; Froyen, et al., 2007; Gillis, et al., 2014; Igoillo-Esteve, et al., 2013; Khan, et al., 2012; Martinez, et al., 2012; Najmabadi, et al., 2011; Ramser, et al., 2004; Slaugenhaupt, et al., 2001; Takano, et al., 2008; Yarham, et al., 2014; Zeharia, et al., 2009). However, the molecular basis of the link between the neurological disorders and mutations in predicted tRNA modification genes is not clear. Indeed, decreased levels of tRNA modifications have only been demonstrated in cells derived from patients with two neurological disorders. First, bulk tRNA from brain tissue and cell lines derived from familial dysautonomia (FD) patients, which is due to mutations in IKBKAP in numerous cases (Anderson, et al., 2001; Cuajungco, et al., 2003; Dong, et al., 2002; Slaugenhaupt, et al., 2001), had reduced levels of the mcm⁵s²U₃₄ modification and the IKAP protein (Karlsborn, et al., 2014). This result is consistent with the homology between IKAP and the Elp1 subunit of the S. cerevisiae elongator complex (Hawkes, et al., 2002), which is required for cm⁵U formation (Huang, et al., 2005). Second, tRNA from cells derived from patients with a Dubowitz-like syndrome (characterized by phenotypes including microcephaly and mental and speech delays) linked to an NSUN2 splice-site mutation lacked m⁵C at residues 34, 48, 49, and 50 (Blanco, et al., 2014; Martinez, et al., 2012), consistent with the activity of the yeast and mammalian homologs (Blanco, et al., 2011; Brzezicha, et al., 2006; Motorin and Grosjean, 1999; Tuorto, et al., 2012). NSUN2 mutations are also linked to other neurological disorders including a Noonan-like syndrome similar to the NSUN2-linked Dubowitz-like syndrome (Fahiminiya, et al., 2013) and autosomal recessive ID (ARID) in four independent families (Abbasi-Moheb, et al., 2012; Khan, et al., 2012).

One of the strongest links between ID and a putative tRNA modification gene is that linking non-syndromic X-linked ID (NSXLID) to mutations in FTSJ1(MIM# 300499), a homolog of yeast Trm7, which catalyzes formation of Cm₃₂ and Nm₃₄ on substrate tRNAs (Pintard, et al., 2002) and is critical for normal function of tRNA^Phe (Guy, et al., 2012). Distinct alleles of FTSJ1 from five independent families are linked to NSXLID and each allele results in reduced levels of mRNA or reduced predicted protein function (Freude, et al., 2004; Ramser, et al., 2004; Takano, et al., 2008). In addition, NSXLID is linked to a microdeletion that removes FTSJ1 and SLC38A5 (Froyen, et al., 2007) (Table 1; Supp. Table S1). Moreover, FTSJ1 is the likely human ortholog of the yeast TRM7 gene, since expression of human FTSJ1 under control of the strong P_GAL promoter (P_GAL-FTSJ1) in a high copy (2µ) plasmid suppresses the severe growth defect of S. cerevisiae trm7Δ mutants (Guy and Phizicky, 2015).

Table 1.

FTSJ1 mutations associated with NSXLID analyzed for tRNA modifications in this study

FTSJ1 allele	Family	Mutation	Effect	Reference
FTSJ1Δ	6	Deletion of FTSJ1 and SLC38A5	Loss of FTSJ1	(Froyen, et al., 2007)
FTSJ1 splice site (FTSJ1-ss)	3	c.121+1delG	Significant reduction of FTSJ1 mRNA levels	(Freude, et al., 2004)
p.A26P	7	c.76G>C; p.A26P	Altered FTSJ1 protein function	This study, (Hu, et al., 2015)

In S. cerevisiae, Trm7 is the central component of a complex modification circuitry required for anticodon loop modification of target tRNAs, wherein Trm7 separately interacts with Trm732 and Trm734 to form Cm₃₂, and Nm₃₄ respectively, both of which are required on tRNA^Phe for efficient formation of wybutosine (yW) at m¹G₃₇ by other proteins (Fig. 1A)(Guy, et al., 2012; Noma, et al., 2006). Moreover, the same circuitry appears to be conserved in the phylogenetically distant yeast Schizosaccharomyces pombe, since the nearly lethal phenotype of trm7Δ mutants is suppressed by overproduction of tRNA^Phe, since tRNA^Phe of trm7Δ mutants likewise lack Cm₃₂, Gm₃₄, and yW₃₇, and since the homologous S. pombe Trm732 and Trm734 proteins are required for the respective Cm₃₂ and Gm₃₄ modifications (Guy and Phizicky, 2015). Indeed, this conserved circuitry might be further extended in eukaryotes since suppression of the growth defect of S. cerevisiae trm7Δ mutants by FTSJ1 expression requires the function of Trm732 or its human homolog THADA to form Cm₃₂ on tRNA^Phe (Guy and Phizicky, 2015).

Figure 1.

FTSJ1 mutations associated with NSXLID are defective for 2'-O-methylation on tRNA^Phe.

(A) Schematic of the circuitry for tRNA^Phe anticodon loop modification in S. cerevisiae and S. pombe. Trm7 acts with Trm732 to form Cm₃₂ and separately with Trm734 to form Gm₃₄, and these modifications in turn drive yW formation at m¹G₃₇ on tRNA^Phe. Wider arrow from Gm₃₄ indicates that yW formation is more dependent on this modification than on Cm₃₂. Predicted human homologs of Trm7, Trm732, and Trm734 are in brackets.

(B) HPLC traces of nucleosides of tRNA^Phe purified from human LCLs derived from FTSJ1-associated NSXLID patients. tRNA^Phe isolated from LCLs derived from NSXLID patients with the indicated FTSJ1 alleles, and from control LCLs, was digested to nucleosides and analyzed by HPLC as described in Materials and Methods. (ss), splice site mutation; (wt) wild type.

(C) Quantification of tRNA^Phe nucleoside analysis from panel B. (*), levels below threshold of detection.

(D) Analysis of o2yW levels on tRNA^Phe purified from LCLs of FTSJ1-associated NSXLID patients. Nucleosides of tRNA^Phe from indicated LCLs were analyzed as in panel B under conditions optimized for evaluating o2yW modification.

Although it is well established that human FTSJ1 mutations cause NSXLID, the connection between FTSJ1 and tRNA modifications is not known in human cells, or in any metazoan. Indeed, it is not necessarily true that FTSJ1 modifies human tRNAs based on complementation of a yeast mutant when FTSJ1 is expressed at high levels.

Here we show that cell lines derived from NSXLID patients bearing genetically distinct disease-causing FTSJ1 mutations have pronounced defects in 2'-O-methylation of the anticodon loop of tRNAs, with reduced levels of peroxywybutosine (o2yW₃₇) on tRNA^Phe. These findings provide strong evidence that FTSJ1 catalyzes Cm₃₂ and Nm₃₄ modification in humans, and further support the conserved circuitry for tRNA^Phe anticodon loop modification. Intriguingly, tRNA^Phe from human cell lines with a novel missense FTSJ1-p.A26P mutation (NM_012280.2, c.76G>C, p.A26P; nucleotide numbering is based on cDNA sequence) or from the corresponding yeast trm7-A26P mutant lacks Gm₃₄ but retains Cm₃₂, apparently due to selective loss of substrate recognition and/or catalysis, rather than to reduced expression or protein interactions. These results strongly suggest that FTSJ1-associated NSXLID is due to lack of modifications of the anticodon loop of substrate tRNAs, and suggest the possibility that if tRNA^Phe is the important human FTSJ1 target (as is true in two phylogenetically distant model yeast organisms), that lack of only Gm₃₄ of tRNA^Phe might be sufficient to trigger NSXLID.

Materials and Methods

Sequencing and identification of the FTSJ1-p.A26P variant

DNA extracted from whole-blood (QIAamp DNA blood maxi kit; Qiagen, Limburg, Netherlands) was part of a large X-chromosome exome sequencing study (Hu, et al., 2015), with one affected individual sequenced from this family (III-2, HiSeq; Illumina, San Diego, USA). Confirmation of the FTSJ1 gene (NM_012280.2, c.76G>C, p.A26P; nucleotide numbering based on cDNA sequence) variant and segregation analysis by Sanger sequencing was carried out using standard methods. FTSJ1 exon 2 DNA was amplified and sequenced using the following primers; F - 5’- GCA GTG GAG CCT GAG AGT TC-3’ and R - 5’- CTA TCT TCC TGC CTG TCT CCC T-3’.

Generation of lymphoblastoid cell lines

LCLs derived from patients bearing a c.121+1delG mutation in FTSJ1(Supp. Table S1; FTSJ1-ss; family 3), and from a patient bearing an FTSJ1 deletion (FTSJ1Δ; family 6) have been described previously (Freude, et al., 2004; Froyen, et al., 2007). Other LCLs were generated using standard methods.

Yeast strains and plasmids

Yeast strains are listed in Supp. Table S2. Plasmids used in this study are listed in Supp. Table S3. See Supp. Methods for detailed information on construction of strains and plasmids.

tRNA purification and northern blot analysis

For a detailed description of cell growth conditions and RNA preparation, see Supp. Methods. Briefly, RNA was extracted using TRIzol (Life Technologies) according to manufacturer's instructions. Extraction of RNA using phenol for northern analysis under acidic conditions to preserve aminoacylation was previously described (Alexandrov, et al., 2006). Low molecular weight RNA was extracted from yeast cells as previously described (Jackman, et al., 2003), and appropriate 5’ biotinylated oligonucleotides were used to purify tRNA from yeast and human RNA preparations as previously described (Jackman, et al., 2003).

HPLC analysis of tRNA

Purified tRNA was digested with P1 nuclease and phosphatase as previously described (Jackman, et al., 2003), and nucleosides were subjected to HPLC analysis essentially as previously described (Jackman, et al., 2003). Nucleosides from tRNA^Phe were separated by HPLC at pH 7.0 to maximize separation of Gm and m¹G, as previously described (Guy, et al., 2012), and o2yW was separated by HPLC as previously described (Noma, et al., 2006).

Immunoblot analysis

Yeast crude extracts were subjected to SDS-PAGE and proteins were transferred to nitrocellulose membrane (Bio-Rad). MORF-tagged and PT-tagged constructs were detected with rabbit polyclonal anti-protein A (Sigma, 1:5,000), followed by incubation with goat anti-rabbit IgG-HRP (Bio-Rad, 1:10,000), and visualization with Amersham ECL Plus (GE Healthcare). The 9myc tag was detected with mouse monoclonal anti-[c-myc] (Roche, 1:10,000), followed by incubation with goat anti-mouse IgG-HRP (Bio-Rad, 1:10,000), and visualization.

Affinity purification of tagged proteins

MORF-tagged and PT-tagged proteins were purified by affinity purification with IgG Sepharose followed by elution with GST-3C protease, and removal of the protease using glutathione Sepharose resin, essentially as previously described (Quartley, et al., 2009).

Clinical description

For a detailed clinical description, see Supp. Methods. This research was approved by the Women's and Children's Health Network Human Research Ethics Committee.

Results

FTSJ1 is required for Cm₃₂ and Gm₃₄ modification of tRNA^Phe

To determine if patients with FTSJ1-associated NSXLID have reduced 2'-O-methylation in their tRNAs, we analyzed tRNA^Phe purified from lymphoblastoid cell lines (LCLs) derived from a patient with an X chromosome microdeletion of FTSJ1 and SLC38A5 (FTSJ1Δ, family 6, Table 1, Supp. Table S1) (Froyen, et al., 2007) and from two brothers with a splice site mutation in FTSJ1(c.121+1delG, FTSJ1-ss, family 3, Table 1, Supp. Table S1) (Freude, et al., 2004), as well as from two control LCLs derived from healthy individuals. We found that tRNA^Phe purified from the FTSJ1Δ LCL had no detectable Cm or Gm modification (<0.03 moles/mole for each vs. 0.72 to 0.74 moles/mole for Cm and 0.69 to 0.74 moles/mole for Gm), and that tRNA^Phe from both of the FTSJ1-ss LCLs had undetectable Gm and a small, but detectable, amount of Cm (0.07 and 0.05 moles/mole, respectively), whereas levels of Ψ, m⁵C, m¹A, m⁷G, and m²G were similar to those from control LCLs (Table 2, Fig. 1B,C). Thus, these data strongly indicate that FTSJ1 is the homolog of yeast TRM7 that is responsible for 2'-O-methylation of tRNA^Phe in humans, and link FTSJ1-associated NSXLID with tRNA modification defects.

Table 2.

HPLC analysis of tRNA^Phe nucleoside content from human LCLs derived from patients with mutations in FTSJ1

cell line FTSJ1		(control 1) wild type	(control 2) wild type	(F6) Δ	(F3:1) ss	(F3:2) ss	(F7) p.A26P
mod.	mol. exp.
Cm	1	0.74 ± 0.14	0.72 ± 0.05	<0.03	0.07 ± 0.002	0.05 ± 0.01	0.63 ± 0.14
Gm	1	0.74 ± 0.04	0.69 ± 0.03	<0.03	<0.03	<0.03	<0.03
m1G	0	0.42 ± 0.03	0.15 ± 0.09	0.74 ± 0.06	0.65 ± 0.03	0.65 ± 0.03	0.29 ± 0.08
Ψ	4	3.12 ± 0.29	3.21 ± 0.16	3.17 ± 0.31	3.16 ± 0.27	3.31 ± 0.33	3.32 ± 0.33
m5C	1	0.68 ± 0.14	0.70 ± 0.08	0.62 ± 0.13	0.72 ± 0.08	0.73 ± 0.05	0.58 ± 0.12
m1A	2	0.84 ± 0.11	0.94 ± 0.13	0.78 ± 0.07	0.83 ± 0.13	0.93 ± 0.05	0.91 ± 0.17
m7G	1	0.48 ± 0.15	0.46 ± 0.11	0.39 ± 0.11	0.48 ± 0.16	0.44 ± 0.08	0.46 ± 0.12
m2G	1	0.77 ± 0.10	0.81 ± 0.10	0.78 ± 0.07	0.78 ± 0.08	0.82 ± 0.07	0.80 ± 0.09
o2yW	1	0.60 ± 0.05	1.00 ± 0.12	0.16 ± 0.08	0.22 ± 0.06	0.23 ± 0.09	0.80 ± 0.011

Mean and standard deviation based on three individual growths and RNA preparations.o2yW values are relative to control 2. mol. exp.; moles expected.

We also found that FTSJ1Δ and FTSJ1-ss LCLs have reduced levels of the o2yW modification of m¹G₃₇ of tRNA^Phe. Thus, we observed an increase in m¹G on tRNA^Phe from LCLs with these FTSJ1 mutations compared to control LCLs (0.65 – 0.74 moles/mole, compared to 0.15 – 0.42 moles/mole, Table 2, Fig. 1B,C), and a corresponding decrease in relative amounts of o2yW (to 0.16 – 0.23 compared to control values of 0.6 and 1.0. Table 2, Fig. 1D). This result implies that, as for yW₃₇ formation in both S. cerevisiae and S. pombe (Guy and Phizicky, 2015; Guy, et al., 2012), o2yW₃₇ formation from m¹G is stimulated by Cm₃₂ and/or Gm₃₄ of tRNA^Phe.

We also analyzed Cm levels on tRNA^Trp, which has Cm₃₂ and Cm₃₄ in five of six characterized tRNA^Trp species from eukaryotes (Guy and Phizicky, 2015; Machnicka, et al., 2013), including the mammal Bos taurus (Fournier, et al., 1978) and is modified at both residues by Trm7 in S. cerevisiae (Pintard, et al., 2002). Since tRNA^Trp purified from the FTSJ1Δ LCL lacked Cm, whereas tRNA^Trp from a control LCL had substantial amounts of Cm (<0.03 vs. 1.06 moles/mole, Supp. Table S4), we infer that tRNA^Trp is modified by FTSJ1 in humans.

As in yeast (Guy, et al., 2012), the lack of modification in LCL tRNAs that is due to FTSJ1 mutations does not appear to affect the amounts or the charging levels of substrate tRNAs. Thus, levels of tRNA^Phe and tRNA^Trp from LCLs with FTSJ1-ss or FTSJ1Δ mutations appear to be normal relative to wild type LCLs, as determined by Northern blot analysis (Supp. Fig. S1). Furthermore, lack of these modifications does not appear to affect tRNA charging in these cells, since there is as much or more charged tRNA^Phe and tRNA^Trp relative to uncharged tRNA in the mutant LCLs as in the wild type LCLs (Supp. Fig. S1).

Identification of a novel, missense FTSJ1 variant c.76G>C; p.A26P in a family with NSXLID

By investigation of the X-chromosome exome of a proband (III-2) from a large multigenerational family with NSXLID (Fig. 2A, Table 1, Supp. Table S1, family 7, see Materials and Methods and Hu et al. 2015), we identified a novel FTSJ1 variant c.76G>C, p.A26P (RefSeq NM_012280.2). Subsequent sequencing analysis of six other available family members showed that the c.76G>C; p.A26P allele segregates with ID (Fig. 2A, Supp. Fig. S2). The c.76G>C; p.A26P allele is not currently present in the Exome Variant Server, ExAC Browser, or SNP database from the UCSC genome browser. This variant was submitted to the Leiden Open Variation Database (http://www.lovd.nl/FTSJ1).

Figure 2.

The FTSJ1-p.A26P variant is associated with non-syndromic X-linked intellectual disability (NSXLID).

(A) Pedigree of family 7, with the FTSJ1-p.A26P allele. Filled square, affected male; circle with dot, known or obligate female carrier; open square, normal male; open circle, normal female. X-chromosome exome sequenced proband with the FTSJ1 NM_012280.2, c.76G>C, p.A26P variant is indicated by a red arrow. FTSJ1 genotype is indicated where DNA analysis was carried out. (+), c.76G>C; (−), wt. Nucleotide sequence is based on cDNA numbering.

(B) Amino acid sequence alignment of the N-terminal region of human FTSJ1 with known or predicted Trm7 proteins from diverse eukaryotes. Residue A26 is indicated by magenta arrow, and other residues analyzed in this study are indicated by black arrows.

(C) Predicted location of A26 residue of FTSJ1, based on the Escherichia coli RrmJ crystal structure. Representation of E. coli RrmJ (PDB 1EIZ), with catalytic triad residues in cyan, corresponding human FTSJ1 residues in round brackets, and corresponding S. cerevisiae residues in square brackets. Trp36 of RrmJ (corresponding to FTSJ1 p.A26) is magenta, and the predicted α-helix containing A26 (residues 33–46 of RrmJ) is dark gray.

Based on the fact that the FTSJ1-p.A26P allele was unique, segregated with ID in the pedigree, and is in a highly conserved region of the FTSJ1/Trm7 protein family (Fig. 2B), we predicted that this allele was deleterious to FTSJ1 protein function, thus causing ID in this family, consistent with previous studies linking loss of function FTSJ1 mutations and NSXLID (Freude, et al., 2004; Froyen, et al., 2007; Ramser, et al., 2004; Takano, et al., 2008). Indeed, the FTSJ1-p.A26P variant might also be expected to have impaired methyltransferase activity because p.A26 of FTSJ1 is only two residues from the predicted catalytic residue p.K28 in the same predicted α-helix, based on the structure of the FtsJ protein family member RrmJ (Fig. 2C) (Bugl, et al., 2000; Feder, et al., 2003; Pintard, et al., 2002).

The FTSJ1-p.A26P variant results in tRNA^Phe lacking only Gm₃₄

Consistent with a decrease in FTSJ1 activity, tRNA^Phe from an LCL derived from the proband III-2 (Fig. 2A) with the FTSJ1-p.A26P allele had undetectable Gm (Table 2, Fig. 1B,C); however, this tRNA^Phe had nearly normal Cm levels (0.63 moles/mole vs. 0.72 to 0.74 moles/mole in control LCLs with wild type FTSJ1), and o2yW₃₇ levels (0.80 relative to 1.0 and 0.6 from two control LCLs, Table 2, Fig. 1D). The FTSJ1-p.A26P LCL was also defective for modification of tRNA^Trp with low, but detectable levels of Cm modification for the sample that was analyzed (0.15 vs 1.06 moles/mole, Supp. Table S4), but comparable levels of the other measured tRNA^Trp modifications.

S. cerevisiae Trm7-A26P forms Cm₃₂, but not Gm₃₄ on tRNA^Phe in yeast cells

To further examine the defect of the FTSJ1-p.A26P variant, we determined if high level expression of FTSJ1-p.A26P could complement the growth defect of an S. cerevisiae trm7Δ mutant, by evaluating growth of a trm7Δ [CEN URA3 P_TRM7-TRM7] strain containing a [2µ LEU2 P_GAL-FTSJ1-p.A26P] plasmid, after plating on medium containing 5-FOA and galactose to select against the [CEN URA3 P_TRM7-TRM7] plasmid. Consistent with our earlier results, expression of wild type FTSJ1 from a strong promoter on a high copy plasmid complemented the slow growth phenotype of the trm7Δ mutant (Guy and Phizicky, 2015); however, no complementation was observed upon expression of the FTSJ1-p.A26P variant, since growth on FOA was similar to that of the trm7Δ mutant bearing the vector or expressing the presumed catalytically dead variants FTSJ1-p.K28A or FTSJ1-p.K158A (Fig. 3A) (Feder, et al., 2003). We note, however, that levels of an affinity tagged FTSJ1-p.A26P-PT variant (for a description of the PT tag, see Supp. Methods) were reduced about 10-fold relative to those of FTSJ1-PT or the FTSJ1-p.K28A-PT variant (Supp. Fig. S3), suggesting that lack of complementation in yeast could be due, in part, to reduced protein levels.

Figure 3.

The S. cerevisiae Trm7-A26P variant is specifically defective for Gm₃₄ modification of tRNA^Phe.

(A) trm7-A26P suppresses the slow growth of S. cerevisiae trm7Δ mutants. Wild type and trm7Δ [CEN URA3 P_TRM7-TRM7] strains with [LEU2] plasmids expressing FTSJ1 or TRM7 variants as indicated were grown overnight in S -Leu medium containing raffinose and galactose, diluted to OD₆₀₀ of ~0.5 in H₂O, and serially diluted 10-fold in H₂O, and then 2 µL was spotted onto S medium containing raffinose, galactose, and 5-FOA, followed by incubation for 3 days at 30°C.

(B) An S. cerevisiae trm7Δ mutant expressing trm7-A26P from a low copy (CEN) plasmid efficiently forms Cm₃₂, but not Gm₃₄ on tRNA^Phe. Quantification of nucleosides from tRNA^Phe isolated from indicated yeast strains. (*), levels below threshold of detection.

(C) A CEN plasmid expressing trm7-A26P does not fully suppress the slow growth of S. cerevisiae trm7Δ trm732Δ mutants. Wild type, trm7Δ, trm7Δ trm732Δ, and trm7Δ trm734Δ strains with [CEN URA3 P_TRM7-TRM7] and [LEU2] plasmids as indicated were grown overnight in SD-Leu, diluted, plated on SD-Leu medium containing 5-FOA, and incubated for 2 days at 30°C to select for loss of the URA3 plasmid (top panel), followed by growth analysis on YPD at the indicated temperatures (bottom panel).

However, we found that expression of S. cerevisiae trm7-A26P complements the growth defect of a trm7Δ mutant, with similar effects on tRNA^Phe modification as in the human FTSJ1-p.A26P LCL. Thus, a [CEN LEU2 P_TRM7-trm7-A26P] plasmid fully complemented a trm7Δ mutant, whereas no complementation was observed with plasmids expressing the predicted catalytic dead variants trm7-K28A or trm7-K164A (Fig. 3A). Furthermore, tRNA^Phe from the trm7Δ [CEN LEU2 P_TRM7-trm7-A26P] mutant had nearly normal levels of Cm₃₂ (0.82 moles/mole vs. 1.00 in trm7Δ [CEN LEU2 P_TRM7-TRM7] cells, Table 3, Fig. 3B) but barely measurable Gm₃₄ (0.05 moles/mole vs. 0.81 in trm7Δ [CEN LEU2 P_TRM7-TRM7] cells), almost exactly as observed for tRNA^Phe from the FTSJ1-p.A26P human cell line. As expected of catalytic dead mutants, tRNA^Phe from trm7Δ mutants expressing either trm7-K28A or trm7-K164A had no detectable Cm or Gm (Table 3, Fig. 3B); indeed, tRNA^Phe from a trm7Δ strain overproducing trm7-K28A on a [2µ P_GAL-trm7-K28A] plasmid still had no detectable Cm or Gm modifications. Consistent with the full modification of Cm₃₂ and the partial modification of Gm₃₄ of tRNA^Phe by Trm7-A26P, introduction of a [CEN LEU2 P_TRM7-trm7-A26P] plasmid fully suppressed the slow growth of a trm7Δ trm734Δ mutant, but only partially suppressed the slow growth of a trm7Δ trm732Δ mutant over a range of temperatures from 18°C to 37°C (Fig. 3C).

Table 3.

HPLC analysis of tRNA^Phe nucleoside content from an S. cerevisiae trm7∆ strain expressing TRM7 variants

modification moles expected	medium	Cm 1	Gm 1	m1G 0	Ψ 2	m5C 2	m2G 1
wild type [CENvec]	SD - Leu	1.09 ± 0.23	0.85 ± 0.07	<0.03	2.08 ± 0.10	1.77 ± 0.06	0.95 ± 0.09
trm7Δ[CEN PTRM7-TRM7]	SD - Leu	1.00 ± 0.12	0.81 ± 0.09	<0.03	2.04 ± 0.04	1.75 ± 0.06	0.97 ± 0.06
trm7Δ[CENvec]	SD - Leu	<0.03	<0.03	1.01 ± 0.04	1.98 ± 0.10	1.78 ± 0.05	0.96 ± 0.06
trm7Δ[CEN PTRM7-trm7-K28A]	SD - Leu	<0.03	<0.03	1.02 ± 0.07	2.1 ± 0.11	1.82 ± 0.03	0.97 ± 0.08
trm7Δ[CEN PTRM7-trm7- K164A]	SD - Leu	<0.03	<0.03	1.04 ± 0.08	2.15 ± 0.08	1.83 ± 0.06	0.99 ± 0.09
trm7Δ[CEN PTRM7-trm7-A26P]	SD - Leu	0.82 ± 0.08	0.05 ± 0.01	0.70 ± 0.12	2.01 ± 0.01	1.77 ± 0.04	0.97 ± 0.10
trm7Δ[2µ PGAL-trm7-K28A]	S-Leu + raff gal	<0.03	<0.03	0.87 ± 0.04	2.13 ± 0.07	1.64 ± 0.07	0.83 ± 0.05
trm7Δ[CEN PTRM7-TRM7]	S-Leu + raff gal	1.13 ± 0.06	0.88 ± 0.03	<0.03	2.1 ± 0.08	1.83 ± 0.04	1.0 ± 0.01
trm7Δ[CEN PTRM7-TRM7-MORF]	S-Leu + raff gal	0.66 ± 0.13	0.50 ± 0.06	0.39 ± 0.06	2.06 ± 0.04	1.79 ± 0.03	0.96 ± 0.03
trm7Δ[CEN PTRM7-trm7-A26P]	S-Leu + raff gal	0.75 ± 0.03	0.14 ± 0.04	0.49± 0.07	2.11 ± 0.02	1.79 ± 0.02	0.99 ± 0.02
trm7Δ[CEN PTRM7-trm7-A26P-MORF]	S-Leu + raff gal	0.53 ± 0.10	<0.03	0.75 ± 0.05	2.19 ± 0.08	1.79 ± 0.03	1.01 ± 0.01
trm7Δ[2µ PGAL-TRM7]	S-Leu + raff gal	0.98 ± 0.10	0.93 ± 0.05	<0.03	2.18 ± 0.02	1.85 ± 0.06	0.98 ± 0.02
trm7Δ[2µ PGAL-TRM7-PT]	S-Leu + raff gal	0.89 ± 0.07	0.85 ± 0.04	<0.03	2.12 ± 0.05	1.86 ± 0.04	0.93 ± 0.03
trm7Δ[2µ PGAL-trm7-A26P]	S-Leu + raff gal	0.93 ± 0.08	0.26 ± 0.01	0.38 ± 0.02	2.17 ± 0.03	1.85 ± 0.02	0.99 ± 0.02
trm7Δ[2µ PGAL-trm7-A26P-PT]	S-Leu + raff gal	0.97 ± 0.12	<0.03	0.48 ± 0.08	2.16 ± 0.04	1.84 ± 0.02	1.01 ± 0.02

Mean and standard deviation based on three individual growths and RNA preparations.

To further characterize in vivo activity of the Trm7-A26P variant, we examined modifications of the other S. cerevisiae Trm7 substrates, tRNA^Trp and tRNA^Leu(UAA) (Pintard, et al., 2002). Both Cm₃₂ and ncm⁵Um₃₄ of tRNA^Leu(UAA) were virtually undetectable in the trm7Δ [CEN LEU2 P_TRM7-trm7-A26P] mutant (Supp. Table S5). Similarly, both Cm₃₂ and Cm₃₄ of tRNA^Trp were severely reduced in the trm7Δ [CEN LEU2 P_TRM7-trm7-A26P] mutant (0.09 vs. 1.40 moles/mole in wild type cells, Supp. Table S6), similar to the reduced Cm found on tRNA^Trp from FTSJ1-p.A26P LCLs (Supp. Table S4).

We note that the residual Cm found on tRNA^Trp in the trm7Δ [CEN LEU2 P_TRM7-trm7-A26P] mutant was likely at C₃₂, since similar Cm levels were found on tRNA^Trp from trm7Δ trm734Δ mutants, but no detectable Cm was found on tRNA^Trp from trm7Δ trm732Δ mutants (Supp. Table S6). This finding may suggest that the Trm7-A26P variant has retained some specificity for C₃₂ modification, while losing specificity for N₃₄ modification, regardless of base identity. As expected, the catalytic dead Trm7-K28A variant and the presumed catalytic dead Trm7-K164A variant lacked the ability to 2'-O-methylate tRNA^Leu(UAA) or tRNA^Trp (Supp. Tables S5 and S6). Our finding that the Trm7-A26P variant fully suppresses the growth defect of trm7Δ mutants, but only substantially modifies C₃₂ of tRNA^Phe (and not other tRNAs) is consistent with our previous findings that only Cm₃₂ or Gm₃₄ modification of tRNA^Phe is required for healthy growth in S. cerevisiae (Guy, et al., 2012).

Overexpression of S. cerevisiae Trm7-A26P does not restore full Gm₃₄ modification levels on tRNA^Phe

Although expression of Trm7-A26P on a [CEN LEU2 P_TRM7-trm7-A26P] plasmid fully complemented the growth defect of a trm7Δ mutant, expression levels of the corresponding C-terminally affinity-tagged Trm7-A26P-MORF variant (for a description of the MORF tag, see Materials and Methods) were reduced to ~1/3 the levels observed from corresponding strains expressing wild type Trm7-MORF or the Trm7-K28A-MORF variant (Supp. Fig. S4). Thus, it was possible that the selective loss of Gm₃₄ on tRNA^Phe in the trm7Δ [CEN LEU2 P_TRM7-trm7-A26P] mutant was due to reduced Trm7-A26P. Alternatively, it might be due to reduced interaction of Trm7-A26P with Trm734, since Trm7 forms a distinct complex with Trm734, and both proteins are required for Nm₃₄ formation in S. cerevisiae (Guy, et al., 2012).

To test if there was a reduced interaction between Trm7-A26P and Trm734, we transformed CEN plasmids expressing TRM7-MORF or trm7-A26P-MORF into TRM732-9myc or TRM734-9myc yeast strains, and then purified the Trm7 using the tag. Both Trm732-9myc and Trm734-9myc co-purified with Trm7-MORF and with Trm7-A26P-MORF with comparable efficiency, after taking into account the reduced expression of Trm7-A26P (Fig. 4A,B).

Figure 4.

The defect in S. cerevisiae Trm7-A26P function is not due to reduced expression or reduced Trm734 binding.

(A) Trm7-A26P-MORF expressed from a CEN plasmid efficiently forms a complex with chromosomally expressed Trm732-9myc. Crude extracts were prepared from the indicated strains grown in SD-Leu medium, and proteins were purified with IgG-Sepharose beads (which binds the ZZ domain of the MORF tag), followed by treatment with 3C protease, as described in Materials and Methods, and then samples were analyzed by SDS-PAGE and immunoblot analysis as indicated.

(B) Trm7-A26P-MORF expressed from a CEN plasmid forms a complex with chromosomally expressed Trm734-9myc. Samples were analyzed as in panel A.

(C) Trm7-A26P-PT expressed from a high copy (2µ) plasmid forms more Trm7-A26P:Trm734-9myc complex than the corresponding Trm7:Trm734-9myc complex in wild type cells. Indicated samples grown in S-Leu medium containing raffinose and galactose were analyzed as in panel A.

(D) Overexpressed TRM7-A26P does not fully suppress the slow growth of S. cerevisiae trm7Δ trm732Δ mutants. Wild type and trm7Δ trm732Δ strains with plasmids as indicated were grown overnight in S -Leu medium containing raffinose and galactose as in Figure 3A, and spotted to the indicated media.

To further probe the selective loss of Gm₃₄ on tRNA^Phe in the trm7Δ strain expressing trm7-A26P, we overexpressed Trm7-A26P and examined the interaction with Trm734 and Gm levels of tRNA^Phe. Immunoblot analysis against the ZZ domain of protein A demonstrated that expression of Trm7-A26P-PT (the PT tag has identical components as the MORF tag, see Supp. Methods) from a 2µ P_GALplasmid ([2µ URA3 P_GAL-trm7-A26P-PT]) resulted in more than a 125-fold increase in Trm7-A26P levels, compared to wild type Trm7-MORF expressed from a [CEN URA3 P_TRM7-TRM7-MORF] plasmid (Fig. 4C). Furthermore, purification of overexpressed Trm7-A26P-PT resulted in pulldown of ~10-fold more Trm734-9myc than was pulled down during purification of Trm7-MORF expressed from the [CEN URA3 P_TRM7-TRM7-MORF] plasmid. However, despite the >125-fold overexpression and the ~10-fold increase in Trm734 pull down, Gm₃₄ formation on tRNA^Phe was not fully restored by overexpression of either tagged Trm7-A26P from the [2µ LEU2 P_GAL-TRM7-A26P-PT] plasmid or untagged Trm7-A26P from the otherwise identical [2µ LEU2 P_GAL-TRM7-A26P] plasmid (Table 3). Indeed, under these conditions, overexpression of untagged Trm7-A26P in the trm7Δ strain resulted in only 0.26 moles/mole Gm on tRNA^Phe. Consistent with this finding, overexpression of trm7-A26P also failed to fully suppress the growth defect of a trm7Δ trm732Δ mutant (Fig. 4D).

We note that the C-terminal tag of Trm7-PT or Trm7-MORF appears to interfere with Gm formation, since Gm modification of tRNA^Phe was reduced from 0.88 to 0.50 moles/mole in a trm7Δ [CEN LEU2 P_TRM7-TRM7-MORF] strain relative to a trm7Δ [CEN P_TRM7-LEU2 TRM7] strain (Table 3), and similar results were observed when TRM7 had a C-terminal c-myc tag (data not shown). This loss of Gm modification is even more extreme in the corresponding strains expressing Trm7-A26P-MORF or Trm7-A26P-PT. Nonetheless, we conclude that the loss of Gm₃₄ modification activity in the Trm7-A26P variant is not due to a Trm734 binding defect, since overexpression of Trm7-A26P-PT results in more pulldown of Trm734 than occurs with wild type Trm7-MORF expressed from a CEN plasmid, whereas overexpression of even untagged Trm7-A26P (presumably resulting in the same or more Trm7 than with tagged Trm7) does not fully rescue Gm₃₄ modification of tRNA^Phe (Table 3).

Overexpression of untagged Trm7-A26P also appeared to result in a modest but limited increase in Cm levels on tRNA^Trp purified from trm7Δ mutants (from 0.13 to 0.31 moles/mole, Supp. Table S7), and modest increases in Cm and ncm⁵Um levels in tRNA^Leu(UAA) (Supp. Table S8). Thus, all modification defects associated with normal expression of Trm7-A26P appear to persist even when Trm7-A26P is massively overproduced.

Other mutations in the predicted α-helix bearing A26 of Trm7 result in tRNA^Phe specifically lacking Gm₃₄

We investigated the roles of other Trm7 residues in the region near residue A26 because of its proximity to the catalytic residue K28 and because of the intriguing observation that Trm7-A26P efficiently forms Cm₃₂, but not Gm₃₄, on tRNA^Phe. We constructed [CEN LEU2 P_TRM7-TRM7] plasmids expressing Trm7 variants bearing other amino acid substitutions at A26, and substitutions in the predicted α-helix bearing A26 (which encompasses residues 23–36, based on E. coli RrmJ (Bugl, et al., 2000)), and then tested their ability to suppress the growth defects of trm7Δ, trm7Δ trm732Δ, and trm7Δ trm734Δ mutants.

We found that expression of the trm7-A26W or trm7-A26G variant resulted in complete complementation of the growth defect in all three strain backgrounds, whereas expression of trm7-A26E resulted in a distinct growth defect only in the trm7Δ trm732Δ strain at 30°C (Fig. 5A), similar to the phenotype resulting from the trm7-A26P variant, although there was also a minor growth defect in the trm7Δ and trm7Δ trm734Δ backgrounds that was more obvious at 18°C. Consistent with this result, we found that tRNA^Phe from the trm7Δ [CEN LEU2 P_TRM7-trm7-A26E] strain had undetectable Gm levels and about 50% of the normal Cm levels (0.54 vs 1.00 moles/mole), whereas Cm and Gm of tRNA^Phe were at near normal levels in the trm7Δ [CEN LEU2 P_TRM7-trm7A26W] strain (1.03 and 0.54 moles/mole respectively, compared to 1.0 and 0.81 moles/mole in wild type; Supp. Table S9).

Figure 5.

Genetic analysis of Trm7 variant function in S. cerevisiae.

(A) Growth assay for Trm7 variant function. τρµ7Δ, trm7Δ trm732Δ, and trm7Δ trm734Δ strains expressing CEN trm7 variants as indicated were grown overnight in SD-Leu, diluted, spotted on YPD medium, and incubated at the indicated temperatures.

(B) Schematic of phenotypes associated with loss of FTSJ1/Trm7 activity in humans and yeast. Relative size of yW/o2yW/m¹G corresponds to levels of these modifications on residue 37 of tRNA^Phe.

Similar to our observations with the trm7-A26P and trm7-A26E variants, we found that expression of a trm7-L32P variant resulted in a growth defect specifically in the trm7Δ trm732Δ strain, and not in the trm7Δ or trm7Δ trm734Δ strains (Fig. 5A), and that tRNA^Phe from the trm7Δ [CEN LEU2 P_TRM7-trm7-L32P] strain had barely detectable Gm (0.03 moles/mole) but nearly normal levels of Cm (0.90 moles/mole) (Supp. Table S9). By contrast, expression of the trm7-L32A variant complemented the growth defect of all three strains and yielded normal tRNA^Phe modifications (Fig. 5A; Supp. Table S9), while expression of trm7-A23P, trm7-R24A, or trm7-R24P variants resulted in no complementation in any strain background, suggesting lack of function for these variants.

Thus, these results demonstrate that Trm7 variants with other mutations at A26 or elsewhere in the helix specifically lose the ability to form Gm₃₄ on tRNA^Phe while retaining Cm₃₂ modification ability.

Discussion

In this report we have provided strong evidence that human FTSJ1 is required for Cm₃₂ and Gm₃₄ modification of tRNA^Phe, since these tRNA^Phe modifications were undetectable, or drastically reduced, in each of three LCLs derived from NSXLID patients with two distinct FTSJ1 loss of function mutations, whereas tRNA^Phe from each of two control cell lines derived from healthy individuals with wild type FTSJ1 had essentially normal Cm and Gm levels. Since we also observed parallel selective loss of Gm in tRNA^Phe from the FTSJ1-p.A26P LCLs and from the S. cerevisiae trm7-A26P mutant (Fig. 5B), and since expression of FTSJ1 complements the growth defect of an S. cerevisiae trm7Δ mutant by forming Cm₃₂ on tRNA^Phe (Guy and Phizicky, 2015), we conclude that FTSJ1 catalyzes formation of Cm and Gm on tRNA^Phe in human cells. Furthermore, since FTSJ1 mutant alleles segregate with NSXLID in each of the previously reported families from which the LCLs were derived (Freude, et al., 2004; Froyen, et al., 2007), in the new family with the p.A26P variant reported here, as well as in four other reported families (Freude, et al., 2004; Ramser, et al., 2004; Takano, et al., 2008) and likely in an additional family (Hu, et al., 2015), we conclude further that defective 2'-O-methylation of N₃₂ and N₃₄ of the anticodon loop of substrate tRNAs is a major contributing factor in FTSJ1-associated NSXLID.

FTSJ1 mutations might contribute to NSXLID through lack of modification of any FTSJ1 tRNA substrate. Indeed, we provided evidence that FTSJ1 modifies tRNA^Trp in LCLs, and Cm modification of this tRNA was severely reduced in FTSJ1-p.A26P LCLs. Moreover, it is possible that FTSJ1 modifies any of the four other documented human tRNA species with Nm₃₂ and/or Nm₃₄ (Machnicka, et al., 2013), or any of the numerous uncharacterized tRNAs, and hypomodification of any of these tRNAs could contribute to ID. However, we speculate that defective 2'-O-methylation of tRNA^Phe may play a major role in ID, for two reasons. First, reduced function of tRNA^Phe is the major cause of the slow growth defect of S. cerevisiae and S. pombe trm7Δ mutants (Guy and Phizicky, 2015; Guy, et al., 2012). Second, the tRNA^Phe anticodon loop 2'-O-methylations are very highly conserved, since each of the 17 eukaryotic tRNA^Phe species that have been characterized has Cm₃₂, and 16 of the 17 also have Gm₃₄ (Machnicka, et al., 2013). If lack of tRNA^Phe modification by FTSJ1 is a major contributor to NSXLID, then it would follow that the defective tRNA^Phe would primarily be due to lack of Gm₃₄, because levels of both Cm₃₂ and o2yW₃₇ are normal in the FTSJ1-p.A26P LCL. This importance of Gm₃₄ of tRNA^Phe would be consistent with the finding in S. pombe that Gm₃₄ of tRNA^Phe is much more important for function than Cm₃₂, as measured by growth phenotypes of mutants (Guy and Phizicky, 2015). We note that NSXLID patients with the FTSJ1-p.A26P mutation appear to have ID comparable to that of other FTSJ1-associated NSXLID patients (Supp. Table S1), although there have not been any direct, objective comparisons of the extent of ID for each of the NSXLID patients.

These results also underscore the conservation of the intricate circuitry for anticodon loop modification of tRNA^Phe, as tRNA^Phe from LCLs of patients with splice site and null alleles of FTSJ1 had significantly reduced levels of o2yW₃₇, compared to control LCLs, accompanied by increased levels of the o2yW precursor m¹G. This result suggests that the o2yW₃₇ modification may be dependent on Cm₃₂ and Gm₃₄ modification, as is the yW₃₇ modification in S. cerevisiae and S. pombe (Guy and Phizicky, 2015; Guy, et al., 2012). However, unlike in these yeast species, lack of FTSJ1 in human LCLs did not completely abrogate o2yW modification of tRNA^Phe, lack of Gm₃₄ did not appear to affect o2yW₃₇ levels, and control LCLs had variable levels of o2yW, accompanied by correspondingly variable m¹G levels. The apparent dependence of o2yW₃₇ on Cm₃₂ and Gm₃₄ is consistent with previous observations that tRNA^Phe from neuroblastoma cells and a fraction of tRNA^Phe from Ehrlich ascites cell lines lack Cm, Gm, and o2yW (Kuchino, et al., 1982), although it is not clear why tRNA^Phe was affected in these cell lines.

It is intriguing that both human FTSJ1-p.A26P and yeast Trm7-A26P are specifically defective in Gm₃₄ modification of tRNA^Phe, while retaining normal levels of Cm₃₂ modification. Since our data indicates that overproduction of Trm7 by more than 125-fold can rescue the minor defect in Trm734 binding but not the Gm modification defect, this result suggests that the Trm7-A26P defect is due to some specific aspect of substrate recognition or catalysis for Gm₃₄ modification. This interpretation is consistent with the location of residue 26 near the catalytic residue K28 (Feder, et al., 2003; Pintard, et al., 2002). Alternatively, G₃₄ of tRNA^Phe may simply be more difficult to modify than C₃₂, and the p.A26P mutation might cause a general loss in methylation activity. This interpretation is consistent with the observation that tRNA^Phe from FTSJ1-ss LCLs have distinct but low levels of Cm and no detectable Gm, perhaps suggesting a low level of correct mRNA splicing (Freude, et al., 2004). Moreover, Cm₃₂ of tRNA^Phe appears to be the most efficient Trm7 modification target in yeast since overproduction of the Trm7-A26P variant only results in partial modification of C₃₂ and N₃₄ of tRNA^Trp and tRNA^Leu(UAA), whereas C₃₂ of tRNA^Phe is normally modified even when Trm7-A26P is expressed at normal levels. The observation that yeast trm7Δ mutants expressing Trm7-A26E or Trm7-L32P at normal levels also selectively modify C₃₂, but not G₃₄, of tRNA^Phe emphasizes the sensitivity of the crucial Gm₃₄ modification to perturbations in this helix, but leaves unresolved the mechanism by which this occurs.

Since FTSJ1-associated NSXLID patients have no consistent dysmorphic, metabolic or neuromuscular manifestations of the condition other than intellectual disability, it would follow that the functional levels of substrate tRNAs are insufficient for brain development and/or cognitive function, but are sufficient in other tissues and other aspects of development. Although little is known about the differential expression of the 450 tRNA genes in the human genome, comprising ~270 isodecoders (which share an anticodon but have a different tRNA body) (Chan and Lowe, 2009; Goodenbour and Pan, 2006), there is evidence that an isodecoder can be specifically expressed in the central nervous system (Ishimura, et al., 2014). A role for FTSJ1 in neurological development is consistent with the observation that expression of FTSJ1 mRNA in the fetus is highest in the brain compared to other tissues tested (Freude, et al., 2004).

The linkage between tRNA modifications and NSXLID reported here is part of an emerging theme linking tRNA modifications and neurological function. Previous studies have directly linked defective m⁵C modification to a Dubowitz-like syndrome (Blanco, et al., 2014; Martinez, et al., 2012), and defective cm⁵U modification to FD (Anderson, et al., 2001; Karlsborn, et al., 2014), and lesions in four other predicted or likely tRNA modification genes have been linked to neurological function. Thus, lesions in the TRMT10A gene, a homolog of yeast TRM10 (Jackman, et al., 2003), are linked to short stature, microcephaly, and defects in glucose homeostasis and diabetes, and the purified protein corresponding to a missense allele lacks tRNA m¹G₉ methyltransferase activity (Gillis, et al., 2014; Igoillo-Esteve, et al., 2013). Additionally, a missense allele of human ADAT3, the likely homolog of yeast TAD3, required for deamination of A₃₄ to I₃₄ (Gerber and Keller, 1999), is linked to strabismus and ARID (Alazami, et al., 2013); a frameshift mutation in human TRMT1, the homolog of yeast TRM1, required for m^2,2G₂₆ activity (Liu and Straby, 2000), is linked to ARID (Najmabadi, et al., 2011); and two distinct missense alleles of human ELP2, a member of the ELP complex (Hawkes, et al., 2002; Huang, et al., 2005), are linked to ARID (Najmabadi, et al., 2011). In addition, allelic variants in the human ELP3 gene have been linked by genome-wide association studies to amyotrophic lateral sclerosis, and Elp3 mutants were identified in an accompanying screen for defective neuronal function in Drosophila (Simpson, et al., 2009). We note also that tRNA mutations leading to defects in taurine modifications are implicated in MELAS (mitochondrial encephalomyopathy, lactic acidosis, and stroke-like episodes) and MERFF (myoclonic epilepsy with ragged-red fibers)(Suzuki, et al., 2002; Yasukawa, et al., 2000a; Yasukawa, et al., 2000b), while a TRIT1 mutation resulting in reduced i⁶A modification is linked to myoclonic epilepsy associated with encephalopathy (Yarham, et al., 2014).

Since lack of modifications is often overcome by increased dosage of one or more of the unmodified tRNAs (Esberg, et al., 2006; Fernandez-Vazquez, et al., 2013; Guy, et al., 2012; Han, et al., 2015; Phizicky and Alfonzo, 2010), the numerous links between tRNA modifications and neurological defects suggest that the available pool of functional tRNAs may somehow be limited during development and function of the central nervous system, presumably leading to defects in translation or its regulation (Begley, et al., 2007; Chan, et al., 2012). The specific mechanisms by which defects in tRNA biology impact neurological function remain to be determined.