Histological analysis of other organs did not reveal
obvious pathology nor was neurodegeneration
observed in mice heterozygous for this mutation.
We identified the nmf205 mutation as a point
mutation in the consensus splice donor site of
intron 6 of Gtpbp2 that results in misspliced
mRNAs with premature stop codons (fig. S2).
Accordingly, Western blot analysis of mutant
cerebellar extracts failed to detect GTPBP2 (Fig. 2A).
Complementation tests using nmf205 mice and
mice with a targeted deletion of Gtpbp2 confirmed
that loss of Gtpbp2 results in neurodegeneration (fig. S3).
GTPBP2 shares domain homology with a translational guanosine triphosphatase family that is
characterized by the no-go and nonstop decay
pathways ribosome-recycling protein Hbs1 and
the eukaryotic release factor eRF3, which bind
Dom34 and eRF1, respectively (fig. S4A) (7–9).
Although no interaction was observed between
GTPBP2 and eRF1 in coimmunoprecipitation
assays, Pelota (the mammalian Dom34 homolog)
was immunoprecipitated by GTPBP2 (Fig. 2B).
The glutathione S-transferase–GTPBP2 fusion
protein (GST-GTPBP2) also pulled down Pelota
from brain extracts, which demonstrated that
GTPBP2 can interact with endogenous Pelota
(Fig. 2C). Affinity capture of bacterially expressed
GTPBP2 by Pelota demonstrated that these proteins directly interact (Fig. 2D).
Analysis of mice from our mapping cross and
B6J.BALBChr1 congenic mice revealed that homo-
or heterozygosity for a BALB/cJ-derived locus
(Mod205) on distal chromosome 1 suppressed
neurodegeneration in nmf205–/– and Gtpbp2–/–
mice (fig. S5). Mutant mice carrying this BALB locus
routinely survived to 18 months or longer. Further
analysis of multiple other inbred strains including
C57BL/6NJ (B6N) suggested that neurodegenera-
tion in B6J-nmf205–/– mice is likely due to an
epistatic mutation in the B6J strain (table S1).
One single-nucleotide polymorphism (SNP) in
the Mod205 region, rs46447118, was determined
to be a T in the B6J genome but a C in all other
tested strains (fig. S6A). This SNP resides at
nucleotide 50 in the stem of the T loop of n-Tr20,
one of five isodecoders of the nuclear-encoded
tRNAArgUCU family (fig. S6, B and C). Orthologs of
n-Tr20 are widely found in both vertebrates and
invertebrates (fig. S6D). We assayed n-Tr20
aminoacylation and found that the majority of
this tRNA was charged in the B6N brain, but very
low levels were observed in B6J (Fig. 3A). Muta-
tions in the T stem of tRNAs have been shown to
affect pre-tRNA processing and function (10, 11).
In agreement, a 105-nucleotide (nt) band was
detected in the B6J brain, which was confirmed
to be the pre-tRNA retaining the leader and trailer
sequences (Fig. 3B and fig. S7A). In wild-type brains,
the pre-tRNA is 115 nt, which suggests the C-to-T
mutation changes the length of the primary transcript.
Examination of n-Tr20 processing in reciprocal
congenic strains confirmed that this mutation under-
lies the observed maturation defect (fig. S7B).
To confirm that loss of mature n-Tr20 under-
lies neurodegeneration in B6J-nmf205–/– mice
(which are mutant for both Gtpbp2 and n-Tr20),
B6J mice transgenically expressing wild-type
n-Tr20 and harboring the nmf205 mutation
(Tg; nmf205–/–) were examined (fig. S8, A and B).
At 6 months of age, neuron death was greatly
attenuated in the brain and retina (Fig. 3C).
Although Gtpbp2 is widely expressed (fig. S4B)
(12, 13), pathology in mice lacking this gene is
restricted to the CNS. In contrast to other mem-
bers of the tRNAArgUCU family, expression of n-Tr20
and its human ortholog were surprisingly con-
fined to the CNS (Fig. 3D and fig. S8, C and D). In
addition, overall expression of the tRNAArgUCU
isodecoder family was higher in the CNS than in
other tissues (Fig. 3D). Compared with levels of
processed n-Tr20 in age-matched B6N brains,
which show steady postnatal expression, levels in
the B6J brain fell from 50% of B6N levels at
postnatal day 0 (P0) to 19% by P30 (Fig. 3E), and
a concomitant increase in immature n-Tr20 was
also observed. Although B6J brains have a slight
increase in expression of the other members of
the tRNAArgUCU family, a dramatic reduction in
the B6J total tRNAArgUCU pool was observed,
which demonstrated that n-Tr20 normally makes
up ~60% of the expression of this isodecoder
family (Fig. 3F and fig. S9). Spatial differences in
processing of mutant n-Tr20 were also observed
within the B6J brain with significantly lower lev-
els of processed and higher levels of unprocessed
n-Tr20 in the cerebellum compared with the cor-
tex and hippocampus (Fig. 3G). Together, these
data define a CNS-specific tRNA in which levels
of mature transcript correlate with the onset and
severity of cell death in Gtpbp2-deficient mice.
We hypothesized that the n-Tr20 mutation
causes ribosome stalling at AGA codons that is
exacerbated in the absence of Gtpbp2. To test
this, we generated ribosome-profiling libraries
456 25 JULY 2014 • VOL 345 ISSUE 6195
sciencemag.org SCIENCE
Fig. 1. Progressive neurodegeneration in the nmf205–/– mice. (A to F)
Hematoxylin and eosin (H and E)–stained sagittal sections of nmf205–/– and
wild-type (WT; +/+) cerebella. (E and F) Higher-magnification images of cerebellar lobule IX from (C) and (D). (G and H) Cerebellar sections were immuno-stained with antibodies to cleaved caspase-3 (c-Casp3; green) and counterstained
with Hoechst 33342. (I to L) H and E–stained sagittal sections of the dentate
gyrus (I and J) and retina (K and L). Scale bars, 1 mm (D), 50 mm (F) and (J),
and 100 mm (H) and (L). ONL, outer nuclear layer; OPL, outer plexiform layer;
INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer.