Muscleblind-like 2-mediated alternative splicing in the developing brain and dysregulation in myotonic dystrophy

Abstract

The RNA-mediated disease model for myotonic dystrophy (DM) proposes that microsatellite C(C)TG expansions express toxic RNAs that disrupt splicing regulation by altering MBNL1 and CELF1 activities. While this model explains DM manifestations in muscle, less is known about the effects of C(C)UG expression on the brain. Here, we report that Mbnl2 knockout mice develop several DM-associated central nervous system (CNS) features including abnormal REM sleep propensity and deficits in spatial memory. Mbnl2 is prominently expressed in the hippocampus and Mbnl2 knockouts show a decrease in NMDA receptor (NMDAR) synaptic transmission and impaired hippocampal synaptic plasticity. While Mbnl2 loss did not significantly alter target transcript levels in the hippocampus, misregulated splicing of hundreds of exons was detected using splicing microarrays, RNA-seq, and HITS-CLIP. Importantly, the majority of the Mbnl2-regulated exons examined were similarly misregulated in DM. We propose that major pathological features of the DM brain result from disruption of the MBNL2-mediated developmental splicing program.

Figure 1. Mbnl2 Knockout Mice Fail to Model DM-Associated Muscle Deficits

(A) Mbnl2^ΔE2/ΔE2 knockout mice are small at weaning compared to wild-type sibs. (B) Mbnl2 RNAs containing exon 2 are absent in Mbnl2 knockouts. RT-PCR analysis of hippocampus RNA from Mbnl2^+/+, Mbnl2^+/ΔE2 and Mbnl2^ΔE2/ΔE2 mice using primers in Mbnl2 exons 1 and 3. Peptidylprolyl isomerase A (Ppia) was included as a loading control. (C) Immunoblot showing that Mbnl2 protein, present in Mbnl2^+/+ mice (wt), is absent in Mbnl2 knockouts (ko). Mbnl2 protein was detectable in wt quadriceps using longer film exposures. Mbnl1, which is abundantly expressed in lung and upregulated in Mbnl2 knockout brain, and Gapdh were used as family member and loading controls, respectively. (D) Mbnl1 RNA targets are not mis-spliced in either Mbnl2 knockout skeletal (quadriceps) or heart muscles. Splicing patterns were determined using RT-PCR and Mbnl1^+/+ (Mbnl1 wt), Mbnl1^ΔE3/ΔE3 (Mbnl1 ko), Mbnl2^+/+ (Mbnl2 wt) and Mbnl2^ΔE2/ΔE2 (Mbnl2 ko) RNAs. (E) Quantification of exon inclusion for alternative splicing of Clcn1, Serca1, Tnnt2 and Sorbs1 (Figure 1D) RNAs. Data are SEM (n=3) and only the differences between Mbnl1^+/+ and Mbnl1^ΔE3/ΔE3 were significant (*P<0.05).

Figure 2. Altered REM sleep regulation in Mbnl2 knockouts

(A) Baseline wake (top), NREM (middle) and REM (bottom) sleep amount changes in Mbnl2 wild type (wt, blue) and knockout (ko, red) mice across 24 hr (n=8 each for Mbnl2 wt and ko mice; error bars represent SEM). The dark period is indicated by a black bar. (B) Numbers of episodes (left) and mean episode duration (right four bars) of wake (top), NREM (middle) and REM (bottom) sleep. (C) REM sleep latency for all REM sleep episodes in Mbnl2 wt and ko during the dark period. The mean REM sleep latency in ko mice (115.8 ± 5.4 [SEM] sec, n = 229) was significantly shorter than that of wt (132.3 ± 8.7 sec, n = 106), with p= 0.0287 by Mann-Whitney U test. (D) Time course of NREM (top) and REM (bottom) sleep as percentage of time spent every hour during and after sleep deprivation for one day (P<0.05, by one-tailed (*) and two-tailed (**) Student’s t test).

Figure 3. Learning/Memory Deficits, Abnormal Hippocampal Function and Seizure Susceptibility

(A) Mbnl2 expression in the hippocampus and dentate gyrus. Sections were obtained from wild type (left panel) and Mbnl2^GT4 (right) brains, which carry a gene trap in Mbnl2 intron 4 that expresses β-galactosidase (β-geo) and stained for β-gal activity (blue) and counterstained with eosin (pink). (B) Immunohistochemistry of hippocampal sections from Mbnl2^+/+ and Mbnl2^ΔE2/ΔE2 mice using anti-Mbnl2 antibody. Nuclear DNA is indicated by DAPI (blue); scale bars = 25 μm. (C) Spatial memory deficit by the Morris water maze test. Mbnl2^+/+ (wt) and Mbnl2^ΔE2/ΔE2 (ko) mice were evaluated for the number of times the missing platform area was crossed (average crossover). Data are SEM (n=10 for both experimental groups) and significant (P<0.05). (D) NMDA receptor-mediated synaptic potentials are reduced in CA1 of the hippocampus in Mbnl2 knockouts. Representative traces of the NMDAR synaptic responses and fiber potential (arrowhead) obtained from Mbnl2^+/+ (wt, blue) and Mbnl2^ΔE2/ΔE2 (ko, red) male mice. (E) Input-output curves for the mean NMDAR EPSP amplitude versus stimulus intensity. Mbnl2 male knockouts (red circles, n = 5/15 mice/slices) exhibited reduced NMDAR-mediated responses when compared to wild type (blue circles, n = 3/10 mice/slices). Stars indicate significant difference (*P<0.05). (F) Mean presynaptic fiber volley amplitude versus stimulus intensity for Mbnl2 wild type (blue circles, n = 3/15 mice/slices) and knockout (red circles, n = 5/15 mice/slices) male mice. (G) Time course changes in the field EPSP (left) obtained from hippocampal slices 10 min before, and 60 min after, stimulation to induce LTP for Mbnl2 male wild type (blue circles, n = 3/7 mice/slices) and knockouts (red circles, n = 5/10 mice/slices). The bar diagram (right) shows the average magnitude of LTP during the last 5 min of recording (dotted area in time course) from Mbnl2 wild type (blue) and knockouts (red). The * indicates a significant difference from baseline (100%) and # indicates difference between Mbnl2 wt and ko. (H) Mbnl2 heterozygous (+/−, blue) and homozygous (−/−, green) knockout, as well as DMSXL polyCTG transgenic (mu, purple), mice are susceptible to seizures compared to wild type sibs (Mbnl2^+/+, blue; DMSXL wt, brown) using a modified Racine scale: 0, no motor seizures; 1, freezing, staring, mouth or facial movements; 2, head nodding or isolated twitches and rigid posture; 3, tail extension, unilateral-bilateral forelimb clonus; 4, rearing, immobile state on rear haunches with one or both forelimbs extended; 5, clonic seizures with loss of posture, jumping, falling; 6, tonic seizures with hindlimb extension and death. Indicated data are for males (Mbnl2 n=4 for wt and ko, n=6 het; DMSXL n=10 each for wt and mut) and are significant (*P=0.05, ***P<0.001).

Figure 4. Mbnl2 Regulates Alternative Splicing in the Brain

(A) Sepscore analysis of splicing microarray results for Ndrg4 (39 nt alternative exon 14) in Mbnl2^+/+ versus Mbnl2^ΔE2/ΔE2 mice. The negative sepscore (−1.584) for Ndrg4 exon 14 (39 NT) predicts enhanced skipping in the Mbnl2 knockout. (B) Bioinformatics analysis of splicing microarrays predicts that Mbnl2 recognizes the same binding motif as Mbnl1. Mapping of YGCY motifs upstream (left panel) and downstream (right panel) of exons that are preferentially skipped (n=42, red squares) or included (n=47, blue triangles) in Mbnl2 knockouts. The frequency of YGCY in 50 nucleotide windows shifted by 5 nucleotides are compared to a background set of 2905 exons whose inclusion does not change in the Mbnl2 knockout. Error bars mark the 95% confidence interval around the mean frequency. (C) Wiggle tracks of tag coverage obtained from RNA-seq of the Ndrg4 locus flanking exon 14 shows enhanced skipping (three biological replicates each) in Mbnl2^ΔE2/ΔE2 (bottom, blue cassette) versus Mbnl2^+/+ (top, red cassette) mice. Putative Mbnl2 YGCY binding sites located downstream of the Ndrg4 exon 14 5’ss are also indicated (turquoise). (D) Comparison of Mbnl2-dependent exons monitored by microarrays and RNA-seq. Among the 3,222 cassette exons analyzed by both platforms, 116 and 123 exons have high-confidence Mbnl2 dependent splicing, as determined by RNA-seq and splicing microarrays, respectively. Among these, 42 exons were included in the high-confidence set by both platforms (P<1.4×10⁻³², Fisher’s exact test).

Figure 5. Regulation of Adult Brain Splicing by Mbnl2

(A) RT-PCR validation of Mbnl2 target exons identified by microarrays and RNA-seq using hippocampal RNAs. Quantification of alternative exon inclusion (percent spliced in) is shown for 10 genes and Mbnl1^+/+ (blue), Mbnl1^ΔE3/ΔE3 (purple), Mbnl2^+/+(yellow) and Mbnl2^ΔE2/ΔE2 (turquoise). Data are SEM (n=3) and only the differences between Mbnl2^+/+ and Mbnl2^ΔE2/ΔE2 were significant (*P<0.05). (B) Loss of Mbnl2 leads to fetal exon splicing patterns. RT-PCR splicing analysis of select Mbnl2 target RNAs in Mbnl1^+/+, Mbnl1^ΔE3/ΔE3 (Mbnl1^−/−), Mbnl2^+/+ and Mbnl2^ΔE2/ΔE2 (Mbnl2^−/−) were compared to the splicing patterns of wild-type P6 and P42 forebrain (fb) and hindbrain (hb). For some exons (Tanc2 E23a) loss of Mbnl2 completely reproduces the fetal pattern while for others (Ndrg4, Kcnma1) a partial shift is observed. (C) Splicing screen of epilepsy-associated genes in wild-type (Mbnl2^+/+), Mbnl2^+/ΔE2 and Mbnl2^ΔE2/ΔE2 reveals dysregulation of Cacna1d exon 12a in heterozygous knockout hippocampus.

Figure 6. HITS-CLIP and the Normalized Mbnl2 RNA Splicing Map

(A) Protein gel autoradiograph of Mbnl2-RNA ³²P-labeled complexes following crosslinking, RNase A digestion (high and low concentrations indicated by slanted black bar) and immunopurification with anti-Mbnl2. (B) Pie chart of unique CLIP RNA tag distribution. (C) Crosslinking-induced mutation site (CIMS) analysis of Mbnl2 CLIP data. Shown is the base composition of sequences around CIMS for Mbnl2 (top) and Nova (bottom). (D) De novo motif search using sequences around CIMS (−10 to +10 nt) and MEME (Bailey and Elkan, 1994). (E) The majority of alternatively spliced genes in Mbnl2^ΔE2/ΔE2 knockouts are direct binding targets for Mbnl2. (F) Normalized complexity Mbnl2 RNA splicing map. Mbnl2 activated (n=147, red) and repressed (n=143, blue) exons are shown.

Figure 7. MBNL2 Brain RNA Splicing Targets Are Dysregulated in DM1

(A–D) Analysis of DM1 brain for splicing perturbations predicted from the Mbnl2 mouse knockout model. RT-PCR splicing analysis of TANC2, KCNMA1, CSNK1D and CACNA1D using RNAs isolated from normal control temporal cortex (control, n = 4), disease control temporal cortex (disease, n = 9), DM1 temporal cortex (DM1, n=12), fetal control whole brain (fetal, n=1), disease control cerebellum (disease, n=4) and DM1 cerebellum (DM1, n=5) by RT-PCR. (E) MBNL loss-of-function model for myotonic dystrophy. MBNL1 and MBNL2 function as alternative splicing factors during postnatal development of muscle (red myofibers with blue myonuclei) and brain (green neuron), respectively. Note that MBNL1 and MBNL2 cross-regulate the alternative splicing of key MBNL exons, including a 54 nt exon (MBNL1 exon 7, MBNL2 exon 6) which includes a nuclear localization sequence (Figure S7), so loss of MBNL2 could lead to an increase in nuclear levels MBNL1 and partial restoration of adult splicing patterns since both proteins recognize a YGCY motif.