Mutation in NSUN2, which Encodes an RNA Methyltransferase, Causes Autosomal-Recessive Intellectual Disability

Abstract

Causes of autosomal-recessive intellectual disability (ID) have, until very recently, been under researched because of the high degree of genetic heterogeneity. However, now that genome-wide approaches can be applied to single multiplex consanguineous families, the identification of genes harboring disease-causing mutations by autozygosity mapping is expanding rapidly. Here, we have mapped a disease locus in a consanguineous Pakistani family affected by ID and distal myopathy. We genotyped family members on genome-wide SNP microarrays and used the data to determine a single 2.5 Mb homozygosity-by-descent (HBD) locus in region 5p15.32–p15.31; we identified the missense change c.2035G>A (p.Gly679Arg) at a conserved residue within NSUN2. This gene encodes a methyltransferase that catalyzes formation of 5-methylcytosine at C34 of tRNA-leu(CAA) and plays a role in spindle assembly during mitosis as well as chromosome segregation. In mouse brains, we show that NSUN2 localizes to the nucleolus of Purkinje cells in the cerebellum. The effects of the mutation were confirmed by the transfection of wild-type and mutant constructs into cells and subsequent immunohistochemistry. We show that mutation to arginine at this residue causes NSUN2 to fail to localize within the nucleolus. The ID combined with a unique profile of comorbid features presented here makes this an important genetic discovery, and the involvement of NSUN2 highlights the role of RNA methyltransferase in human neurocognitive development.

Main Text

Intellectual disability (ID), also called mental retardation (MR), is a neurodevelopmental disorder that can have a devastating impact on the affected individuals and their families. It is believed to occur at a frequency of ∼1%–3% within the population¹ and might often be caused by abnormalities at the genetic level.^2,3 ID might present as the sole clinical feature (nonsyndromic [MIM 249500]) or might be present with additional clinical or dysmorphological features (syndromic).

We ascertained a consanguineous family from a farming community in the Khairpur district within the province of Sindh in Pakistan, where the first-cousin parents had seven children (three of the five female siblings are affected) (Figure 1). A male cousin, offspring of the mother's brother, was also reported to be affected but was not available for the study. Appropriate informed consent was obtained for all study participants, including unrelated healthy Pakistani controls and nonsyndromic ID (NSID)-affected individuals of European descent, and approval was obtained from the institutional research ethics boards at Quaid-i-Azam University and the Centre for Addiction and Mental Health. The affected family members were assessed by a consultant pediatric psychiatrist. The Vinelands II Adaptive Behavior Scale and Diagnostic Interview for Social and Communication Disorders (DISCO⁶) were used as a framework for obtaining information about early development, schooling and academic achievement, current level of social functioning and adjustment, and support and help required for different activities of daily living. On the basis of this information, all three affected individuals, whose intelligence quotients (IQs) ranged from 40 to 50, were diagnosed with moderate ID. No features of autism were noted.

Figure 1.

Pedigree, Mapping, and Mutation Screening for Family MR14

(A) Pedigree of family MR14 from the Khairpur district. Filled circles indicate affected girls.

(B) HomozygosityMapper analysis⁴ of microarray SNP data: Genome-wide significant regions of homozygosity-by-descent (HBD) are seen only on 5p and 14q. The 14q locus was excluded because one of the unaffected siblings was also homozygous at this locus, whereas the unaffected sibling was genotyped as heterozygous at the 5p locus.

(C) Ideogrammatic representation of the critical autozygous or HBD locus in region 5p15.31, as determined from this study and in relation to the MRT5 locus identified by Najmabadi et al.⁵

(D) The c.2035G>A substitution encoding the p.Gly679Arg change in a heterozygous carrier and an affected homozygote.

(E) ClustalW alignment of NSUN2 across multiple species showing conservation of the p.Gly679 residue in vertebrates and also in nonvertebrate animal species.

Neurological assessment was performed by a consultant neurologist. The oldest female (II:3) had significant delay to her development, walking, and talking at 5 years of age. Her speech was limited to just a few words and was dysarthric. Sensory examination and fundoscopy were not possible because she was not cooperative. The individual had a long face and a somewhat long and pointed nose and chin. Height and weight were below the fifth percentile. Her fingers were tapering, but no hyperextensibility was noted. Muscle tone was increased in all limbs (power grade 5), reflexes were brisk, and plantars were equivocal. Other features observed are shown in Table 1. She is toilet trained, can feed and dress and undress herself, and can help with household chores.

Table 1.

Anthropometric Measures and Clinical Details for Affected Family Members

	Individual
	II:3	II:4	II:5
Sex	female	female	female
Age (years)	14	13	6
IQ	40–50	40–50	40–50
OFC (cm)	50	49	46
Height (cm)	152	136	NK
Weight (kg)	32	26.3	NK
LDH (IU/l)	447a	NK	463a
CPK (IU/l)	294b	NK	60b
Epilepsy	–	–	–
Gait	broad	broad	normal
Feet and/or toes	bilateral pes cavus and tight left Achilles	left foot pes cavus, bilateral equinus position and tight Achilles, and hyperseparation between first and second toes	partial bilateral syndactyly of the fourth and fifth toes
Eyes	right-eye strabismus	fundoscopy, eye movement, and pupillary response are normal	intermittent esotropia and fine horizontal nystagmus
Menses	no (Tanner stage 3)	yes (Tanner stage 4)	no
Pelvic ultrasound	small midline uterus, thin endometrium, and ovaries were not visible	NK	NK

The following abbreviations are used: IQ, intelligence quotient; OFC, occipitofrontal circumference; NK, not known or not measured; LDH, lactose dehyrogenase; CPK, creatine phosphokinase; F, female; and NK, not known or not measured. The comparison of clinical features between this family and the Iranian and Kurdish families described in Abbasi-Moheb et al. is shown in Table 1 in Abbasi-Moheb et al.⁷ Photographs of affected individuals are shown in Figure S1.

^aReference level is 122–234 IU/l.

^bReference level is 34–145 IU/l.

The second girl (II:4) also had significant delay; she talked at 4 years of age, but her age of walking was less delayed. Her speech is also limited and dysarthric, and she is unable to recall her name when asked. Like her older sister, she also had a long and pointed face and was below the fifth percentile for height and weight. Muscle tone was increased in all limbs, reflexes were brisk, and plantars were equivocal. Other features observed are shown in Table 1. She is toilet trained and can feed and dress and undress herself, but she is unable to perform household chores. The coordination and sensory exam was grossly normal.

The third girl (II:5) walked at 5 years of age but had no speech at the age of examination (6 years of age). Head and body measurements were below the fifth percentile. Unlike II:3 and II:4, her gait and tone were normal, and her reflexes were brisk with unsustained clonus. A fundus examination could not be performed, but another cranial-nerve examination and a systemic examination were normal. Other features observed are shown in Table 1.

Computed tomography of the brain was performed for two affected individuals, and it showed normal ventricles and cerebral volume and was generally normal. Gray- and white-matter differentiation was preserved, and posterior fossa was unremarkable. Whole-blood count was within the normal range. However, creatine phosphokinase (CPK) and lactose dehydrogenase (LDH) levels were measured in two individuals (II:3 and II:4) and indicated that both had elevated levels of LDH (447 and 463 IU/l, respectively; the reference level is 122–234 IU/L) and that only the older sister (II:3) had elevated levels of CPK.

Photographs of all affected individuals were assessed for dysmorphic features by an experienced clinical geneticist and are shown in Figure S1, available online. The clinical examination of affected individuals is summarized in Table 1. Overall, the general phenotype is of moderate ID, small head and body size, signs of distal neuropathy as indicated by pes cavus, broad gait, and tight Achilles tendons; however, nerve-conductance velocities, wave latencies, and amplitude were apparently normal. Other variable features include a long nose and face, hypoplastic nails, partial syndactyly of the fourth and fifth toes, hyperseparation between the first and second toes, and absent or atrophied ovaries.

Genomic DNA was extracted from peripheral-blood leukocytes by standard methods. We used the Affymetrix GeneChip Mapping 500K array to analyze DNA samples of three affected individuals and one unaffected individual by using just the NspI chip; this allowed us to genotype ∼260,000 SNPs. Microarray analysis was performed at the London Regional Genomics Centre at the University of Western Ontario. Homozygosity mapping, performed with the dChip analyzer,^8–10 identified a ∼5 Mb homozygosity-by-descent (HBD) locus within region 5p15.3. This locus is flanked by SNPs rs2259 (2.635 Mb) and rs2914296 (7.657 Mb) and overlaps another locus (designated MRT5) previously identified in an Iranian family as harboring a nonsyndromic autosomal-recessive ID (NS-ARID)-associated mutation. However, no gene harboring disease-causing mutations has yet been reported for this locus.⁵ The MRT5 locus was defined by SNP markers rs1824938 (5.092 Mb) and rs60701 (10.734 Mb). Thus, the common region shared between our Pakistani family and the Iranian family was from rs1824938 (5.092 Mb) to rs2914296 (7.657 Mb)—a 2.565 Mb critical region.

Additional genotype data from microsatellite markers on 5p also verified this, and linkage analysis gave a maximum LOD score of 2.77 for marker D5S406 (see Table 2). All known coding genes within this 2.565 Mb critical region were screened for mutations by sequencing. We identified a homozygous substitution, c.2035G>A (RefSeq NM_017755.5) (GRCh37/hg19 chr5: 6,600,308C>T), within exon 19 of NSUN2 (MIM 610916), which is one of the eight known genes within this locus. This substitution would result in the missense change p.Gly679Arg. This substitution is not a known SNP in any SNP databases and is not found in the 1000 Genomes Project or the National Heart, Lung, and Blood Institute Exome Sequencing Project (ESP5400 release; 5,379 subjects). We also confirmed this by sequencing over 200 Pakistani control individuals. In addition to c.2035G>A, a 250 bp insertion just after exon 9 was also identified in the members of the family. However, after genotyping was performed for Pakistani controls by PCR amplification across exon 19, it was apparent that this is a relatively common polymorphism. All primer sequences are available upon request. The 679^Gly residue is highly conserved across evolution of the animal kingdom (Figure 1E), and in silico analyses using PolyPhen and SIFT identify the substitution as being “possibly damaging” and “not tolerated,” respectively. A screening set of 45 NSID individuals of European descent (negative for mutations in FMR1 and MECP2) did not show any mutations in NSUN2.

Table 2.

Two-Point Linkage Analysis for Markers across Chromosome 5p

Markers	Position (cM)	Physical Position (Mb)	LOD Score at Recombination Fraction, θ
			0.00	0.05	0.2	0.4
D5S1981	1.7200	1.155	0.4327	0.3972	0.2597	0.0585
D5S406	11.8500	4.994	2.7729	2.4887	1.6231	0.5141
D5S2505	14.3000	5.817	1.6695	1.4721	0.8628	0.1437
D5S580	17.8700	8.141	0.4327	0.3972	0.2597	0.0585
D5S630	19.6710	9.561	0.4327	0.3972	0.2597	0.0585

The EasyLINKAGE program²⁶ was used. The physical positions are according to UCSC Feb 2009 (GRCh37/hg19) assembly. For reference, the common homozygosity-by-descent (HBD) locus for the Pakistani family and the Iranian family⁵ extends from SNPs rs1824938 (5.092 Mb) to rs2914296 (7.657 Mb), and NSUN2 is located between 6.599 and 6.633 Mb.

NSUN2 encodes a methyltransferase that catalyzes the intron-dependent formation of 5-methylcytosine at C34 of tRNA-leu(CAA).¹¹ It also functions in spindle assembly during mitosis as well as chromosome segregation.¹² Previous work on constructs of the mouse homolog Nsun2 carrying the missense change p.Lys190Met have shown ablation of methyltransferase catalytic activity.¹² p.Lys190Met has not been identified in any human subjects.

Using a cDNA clone for NSUN2 in the pcDNA3.1-Myc vector, we performed site-directed mutagenesis to generate the c.2035G>A (p.Gly679Arg) mutation. Wild-type (WT) and mutant constructs were transfected into the human breast cancer cell line HCC1954 and also into COS7 (monkey kidney) cells. 24 hr later, we stained cells with antibodies against the Myc epitope in order to detect transfected proteins. Although WT NSUN2 was detected in the nucleus and nucleolus of transfected HCC1954 cells (Figure 2A), the p.Gly679Arg mutant failed to localize to the nucleoli in most transfected cells (Figures 2B and 2C). We used antibodies against the nucleolar marker protein nucleophosmin (NPM1) to confirm colocalization in the nucleoli (Figures 2A and 2B), and colabeling for endogenous NSUN2 confirmed the nucleolar localization of endogenous NSUN2 (Figure 2C).

Figure 2.

Cellular Localization of WT and Mutant NSUN2 in Human Breast Cancer Cell Line HCC1954

WT (A) and mutant (B) NSUN2 constructs in the vector pcDNA-Myc transfected into the human breast cancer cell line HCC1954. A. WT NSUN2 (A) but not mutant NSUN2 (B) protein colocalizes with nucleophosmin, a nucleolar marker. The scale bar represents 20 μm.

(C) Costaining for endogenous NSUN2 confirms exclusion of mutant NSUN2 (Myc-labeled) from the nucleoli. Arrows indicate nucleoli (A–C). Costaining with DAPI shows nuclear localization. The scale bar represents 20 μm.

Instead of being localized to the nucleoli, the p.Gly679Arg NSUN2 mutant accumulated in the nucleoplasm in 80% of transfected cells (Figures 3A, 3B, and 3D). In fewer than 10% of transfected cells, it was observed that the p.Gly679Arg mutant NSUN2 did localize to the nucleoli and, at the same time, showed intense staining within the cytoplasm (Figures 3A, 3C,and 3D). Also of note was the fact that the mutant protein seemed to be largely excluded from the nucleoplasm in these cells (Figure 3C). Similar results were also seen for the COS7 cells—the mutant NSUN2 localized mainly in the cytoplasm. We conclude that the substitution of glycine to arginine at position 679 impairs the proper cellular localization of NSUN2.

Figure 3.

Cytoplasmic, Nuclear, and Nucleolar Localization of WT and Mutant NSUN2 in HCC1954 and HeLa Cells

(A) Immunostaining shows nucleolar localization of WT NSUN2 in HCC1954 cells. Costaining with DAPI shows nuclear localization.

(B and C) Cellular localization of NSUN2 carrying the 679^Arg variant (p.Gly679Arg) in the nucleoplasm (B) and cytoplasm (C). Costaining with DAPI shows nuclear localization.

(D) Quantification of cellular localizations shown in (A–C). Error bars represent the standard deviation.

(E) Nucleolar localization of GFP-tagged WT NSUN2 in HeLa cells. White arrow heads indicate nucleoli.

(F) Nuclear and cytoplasmic localization in 679^Arg mutant NSUN2 (p.Gly679Arg). White arrow heads indicate nucleoli.

In parallel, WT and mutant cDNA constructs were generated in the pcDNA3.1 vector with GFP tag and were transfected into the human endothelial cell line EA.hy 926 (from the umbilical vein). In EA.hy 926 cells, WT NSUN2-GFP colocalizes with the nucleophosmin 1 antibody (Santa Cruz) in the nucleoli, whereas the p.Gly679Arg mutant NSUN2-GFP remains in the nucleoplasm (data not shown).

In order to gain some insight into NSUN2 function in the brain, we sought to identify specific mouse neural cell types that displayed NSUN2 localization. To this end, we dissected the cortical and cerebellar regions from whole brains of 3-month-old mice. Tissue was fixed for 24 hr in paraformaldehyde, paraffin sections were taken, antigen retrieval was performed, and sections were stained with affinity-purified NSUN2 antibody (Covalab). Although we were able to observe NSUN2 staining sporadically in some cortical and brain-stem neurons (Figure S2), by far the most striking localization was observed in Purkinje cells of the cerebellum (Figures 4A and 4B). Higher-magnification imaging revealed that NSUN2 localizes to the nucleoli of Purkinje cells and that these nucleoli were often located between or adjacent to dense heterochromatic regions (Figures 4C and 4D). Interestingly, in addition to ID, NSUN2-mutation-positive individuals in the current study display features such as poor speech (dysarthria) and broad gait, which have previously been associated with cerebellar defects. It is therefore likely that the disruption of NSUN2 function in Prukinje cells contributes to the phenotype of these individuals. Given the developmental phenotype in the Pakistani family, the fact that we have also demonstrated NSUN2 expression in the cerebellum of 3-day-old mice by using LacZ staining (Figure S2) suggests that NSUN2 is expressed in these tissues during development as well as adulthood. It is also worth noting that, although ID is not usually associated with cerebellar defects (it is associated much more commonly with cortical defects), there are several reports that describe ID in association with cerebellar dysfunction.^13–16 Purkinje cells are a class of GABAergic neuron, and, at the heart of the cerebellar circuitry, they receive more synaptic inputs than any other cell type in the brain. It is interesting that the NSUN2 missense variant identified in the ID-affected family fails to localize to the nucleolus because it hints at a mechanism by which proper Purkinje cell function might be inhibited as a result of aberrant exclusion of NSUN2 from the nucleoli of these cells. In more recent years, the nucleolus has become established as a multifunctional entity rather than simply a ribosomal RNA-processing center as previously envisioned.¹⁷ A better understanding of the nucleolar functioning of NSUN2 in addition to the identification of its methylation substrates will no doubt shed light on the cellular mechanisms disrupted in NSUN2-mutation-positive individuals.

Figure 4.

Localization of Endogenous WT Nsun2 in Mouse Cerebellum

(A) Sections of mouse cerebellum labeled with an antibody against NSUN2. The following abbreviations are used: NF, nerve fibers; GCL, granule cell layer; and ML, molecular layer.

(B) A higher magnification of an insert in (A) shows NSUN2 in Purkinje cells.

(C) Sections of the cerebellum labeled with L7/Pcp-2, a marker for Purkinje cells.

(D) Colocalization of NSUN2 with nucleophosmin (Npm1) in the nucleoli of Purkinje cells. Nuclei are counter stained with DAPI.

In a recent study of Nsun2 in mice, a knockout that leads to the ablation of Nsun2 through the deletion of exon 8 was generated.¹⁸ Heterozygous mice appeared normal and had no visible phenotype. Nsun2^−/− mice were also viable, and the gross phenotype indicated weight loss (∼30% reduction at 3 months old) and partial alopecia at ∼10 months old, suggesting a role for NSUN2 in skin homeostasis. Nsun2^−/− males were sterile. The small size of the Nsun2^−/− mice might be consistent with the reduced growth in the affected individuals from the Khairpur family. Heights of the affected girls (∼tenth percentile for II:3 and first percentile for II:4) are below the mean for Pakistani girls (from the United Kingdom¹⁹). Weight is strikingly low—both II:3 and II:4 are below the first percentile (according to standard growth charts¹⁹). Interestingly, in one of the affected individuals from the Khairpur family, ovaries were either atrophied or absent. However, no information on sexual development or sterility in affected male family members was available. Studies of an Nsun2-knockout mouse generated by the European Conditional Mouse Mutagenesis program (EUCOMM) has been reported by the Wellcome Trust Sanger Institute, and some preliminary data are available online from the Mouse Resources group. This mouse has a deletion of exon 6, resulting in a frameshift. Neurobehavioral and neurocognitive analysis of heterozygous and homozygous mice via hot-plate tests, open-field tests, and a modified SHIRPA test²⁰ showed no overall difference between these mice and control mice. However, male homozygotes are reported as hyperactive in calorimetric studies. Male homozygous mutant mice showed significantly reduced forepaw grip strength, and abnormal humerus morphology is reported for both sexes. Body size was decreased for male and female homozygotes, and abnormal skull and teeth morphology was reported. An eye phenotype, in which the cornea was abnormal and the lens showed increased opacity, was also present in homozygotes. Fertility and fecundity are also reported as abnormal. These data were provided by and can be obtained from the Mouse Resources group at the Wellcome Trust Sanger Institute. More in-depth neurocognitive assessments would be particularly useful on this model. It might also be important to compare future clinical developments in the family with those of the knockout mouse, e.g., whether signs of alopecia develop later on.

NSUN2 is now the third RNA-methyltransferase-encoding gene to be linked to ID. Previously, FTSJ1 was identified in X-linked nonsyndromic ID,^21,22 and very recently, TRMT1 was identified as a cause of autosomal-recessive ID (ARID).²³ Coexpression data from zebrafish, fruit flies, and yeast have linked NSUN2 to both TRMT1 (STRING; coexpression score = 0.877) and FTSJ1. Coimmunoprecipitation assays have also shown protein-protein interactions between NSUN2 and FTSJ1 (IntAct interaction database). Although NSUN2 was initially thought to have a very narrow range of tRNA targets, this range is broadening and is currently also thought to include mRNAs.²⁴ A role for NSUN2 in DNA methylation has not been excluded but seems unlikely given its nucleolar localization.

Thus, our findings strongly support a role for NSUN2 mutations in ARID, for which additional clinical symptoms are apparent. A publication on gene mapping in Iranian families has also identified this region,^5,25 and in this issue of AJHG, an article from this group also reports the identification of truncating mutations in the same gene, NSUN2, in one Kurdish and several Iranian consanguineous ARID-affected families.⁷ Thus, this is one of the few ARID-associated genes for which there is validation in several independent families and from more than one ethnic group or geographic location. Given the similarities in the phenotypes of the families from this study and that of Abbasi-Moheb et al.,⁷ we speculate that the missense mutation reported here has a similar effect to that of the truncating mutations in NSUN2 and that Gly679Arg mainly results in a loss of function due to failure of the protein to localize to the correct cellular organelles; this failure thus results in ID. We speculate further that the myopathy reported in the Pakistani family was not noted in either the Iranian or the Kurdish families affected by NSUN2 truncating mutations and might result from a gain of function due to accumulation of NSUN2 in the nucleoplasm and cytoplasm.

Web Resources

The URLs for data presented herein are as follows:

• 1000 Genome Project, http://www.1000genomes.org

• CLUSTAL 2.1 Multiple Sequence Alignment, http://www.ebi.ac.uk/Tools/msa/clustalw2/

• NHLBI Exome Sequencing Project, http://evs.gs.washington.edu/EVS/

• Online Mendelian Inheritance in Man (OMIM), http://www.omim.org

• PolyPhen, http://genetics.bwh.harvard.edu/pph2

• SIFT, http://sift.jcvi.org/

• STRING, http://string-db.org/

• UCSC Genome Bioinformatics, http://genome.ucsc.edu

• Wellcome Trust Sanger Institute Mouse Resources Portal, http://www.sanger.ac.uk/mouseportal