MBL-1 and EEL-1 affect the splicing and protein levels of MEC-3 to control dendrite complexity

Transcription factors (TFs) play critical roles in specifying many aspects of neuronal cell fate including dendritic morphology. How TFs are accurately regulated during neuronal morphogenesis is not fully understood. Here, we show that LIM homeodomain protein MEC-3, the key TF for C. elegans PVD dendrite morphogenesis, is regulated by both alternative splicing and an E3 ubiquitin ligase. The mec-3 gene generates several transcripts by alternative splicing. We find that mbl-1, the orthologue of the muscular dystrophy disease gene muscleblind-like (MBNL), is required for PVD dendrite arbor formation. Our data suggest mbl-1 regulates the alternative splicing of mec-3 to produce its long isoform. Deleting the long isoform of mec-3(deExon2) causes reduction of dendrite complexity. Through a genetic modifier screen, we find that mutation in the E3 ubiquitin ligase EEL-1 suppresses mbl-1 phenotype. eel-1 mutants also suppress mec-3(deExon2) mutant but not the mec-3 null phenotype. Loss of EEL-1 alone leads to excessive dendrite branches. Together, these results indicate that MEC-3 is fine-tuned by alternative splicing and the ubiquitin system to produce the optimal level of dendrite branches.


Introduction
Transcription factors (TFs) control gene expression and are essential for specifying different aspects of neuronal cell fate, including neurotransmitter phenotypes [1], neurite morphogenesis and membrane excitability [2][3][4].Many TFs have been reported to affect dendrite morphogenesis [5][6][7].For example, dendrite branching of vertebrate cortical neurons requires the notch signaling and NFkappaB [8].Activity regulated transcription factor CREB is critical for neuronal activity induced dendritic growth and branching [9].The homeobox TF Cux1 is required for dendritic arborization in specific cortical neuron populations [10].In Drosophila, hamlet represses complex dendritic arbor and promotes single-dendrite morphology [11].The development of the dendrites in the Da classes of neurons also requires other families of TFs including Abrupt, Knot, and spineless [12][13][14].Different expression levels of the TF Cut specify the complexity of dendritic arborization in the four different classes of Da Neurons.
The PVD sensory neurons in C. elegans elaborate highly stereotyped and branched dendrites [15].The PVD cell body sends one anterior and one posterior primary dendrite (1˚).The secondary dendrites (2˚) grow orthogonally from 1˚dendrites in both ventral and dorsal directions.Upon reaching the body wall muscle borders, 2˚branches turn anterior and posterior then form tertiary branches (3˚) along the sub lateral nerve cords.From the 3˚dendrites, numerous quaternary dendrites (4˚) grow orthogonally away from the lateral midline [16].PVD neurons sense harsh touch and cold temperature and act as nociceptors [16].They also function as proprioceptive sensors to regulate body bend amplitude [17].Developmentally, the complex and stereotyped dendrites are guided by the neuronal receptor DMA-1 and the epidermal adhesion molecule complex SAX-7-MNR-1-LECT-2 [18][19][20][21][22][23].The elaborate dendritic arbor of the PVD also requires the LIM homeobox TF MEC-3 and the POU domain transcription factor UNC-86 [15,24,25].In addition, AHR-1/spineless suppresses the dendrite branching program and prevents other neurons to adopt the PVD like dendritic arbors [25].
Alternative splicing regulates activities of transcription factors by inclusion or exclusion of exons.For example, Srp is a transcription factor involved in Drosophila mesoderm development.Two isoforms, SrpC (exclusion of N-finger) and SrpNC (inclusion of N-finger) are generated by alternative splicing, which differentially stimulate the expression of crq and gcm, respectively [26].Muscleblind-like proteins (MBNL) belong to a family of RNA splicing regulators of precursor mRNA [27].They regulate alternative splicing by promoting inclusion or exclusion of specific exons on different pre-mRNAs through antagonizing the activity of CUG-BP and ETR-3-like factors (CELF proteins) [28].Pathological CUG and CCUG expansion sequesters MBNLs, leading to loss of MBNLs activity and causes myotonic dystrophy (DM) [29].While MBNLs clearly regulate many muscle genes, in recent years, MBNLs/MBL-1 are reported to be associated with neuronal development [30][31][32].However, the underlying mechanisms are not understood.
Huwe1 is a highly conserved member of the HECT E3 ubiquitin ligase family and functions in many cellular processes such as cell proliferation/suppression, embryogenesis and apoptosis [33].In recent years, Huwe1 has been shown to degrade the N-Myc oncoprotein during nervous system development [34].EEL-1 (the orthologue of Huwe1 in C. elegans) regulates GABAergic presynaptic transmission in C. elegans [35].Human genetic studies show that Huwe1 is associated with multiple neurodevelopment disorders, such as X-linked mental retardation (XLMR) [36].However, little is known about the downstream targets of Huwe1/ EEL-1 in neuronal development and disorders.
Here we report that alternative splicing and proteasome degradation system contributes to the precise regulation of the key TF MEC-3.MBL-1 regulates the alternative splicing of mec-3 to produce the MEC-3 isoform with a higher branching promoting activity.Additionally, EEL-1 may directly or indirectly downregulate MEC-3's protein level to maintain its proper expression level for PVD dendrite morphogenesis.These results demonstrate that the precise control of MEC-3 at mRNA and protein level are required for proper dendrite morphogenesis.

mbl-1 mutants disrupt PVD dendrite morphology
To understand molecular mechanisms regulating PVD dendrite morphology, we carried out a candidate screen using GFP labeled PVD neuron (Fig 1A and 1B).We identified a background mutation wy888 in the strain RB1712, which showed drastically reduced dendrites and simplified arbors (Fig 1B).Through genetic mapping and sequencing, we identified the causative gene for wy888 to be mbl-1.A single point mutation was found in mbl-1, which leads to amino acid substitution of the conserved amino acid histidine to tyrosine (H131Y) (S1 Fig) .Expression of wild type mbl-1 cDNA under a PVD specific promoter could rescue wy888 phenotype, indicating mbl-1 mutant causes the dendrite reduction phenotype and mbl-1 functions cell autonomously in PVD (Fig 1B -1E).To further examine this idea, we analyzed two other putative null alleles of mbl-1, tm1563 (a 513-bp deletion that eliminates the exon 3 of mbl-1) and wy560 (a 70-kb deletion that eliminates eight genes, one of which is mbl-1).The data showed that both of them led to PVD dendrite reduction phenotype (Fig 1B -1E).Interestingly, dendrite phenotype of wy888 is more severe than the two putative null alleles.The mutation in the wy888 allele (H131Y) is a conserved residue located in the zinc-knuckle-like motif CCCH (C3H) that is required for mRNA binding [37].This point mutation might affect binding to its target mRNA and prevent the access of its target mRNA to other splicing factors.

MBL-1 regulates the alternative splicing of mec-3
To identify the downstream factors of MBL-1, we constructed double mutants of mbl-1 with genes involved in PVD dendrite morphogenesis including hpo-30, dma-1, tiam-1, ire-1, and kpc-1 (S2A Fig) .Unlike the other double mutants, ire-1; mbl-1 double mutant did not show a more severe dendrite phenotype compared to that of the mbl-1 single mutant, suggesting that ire-1 may be a splicing target of MBL-1.Therefore, we examined the cDNAs of ire-1 in WT and mbl-1(wy888) mutants and found no differences between these two genotypes (S2B and S2C Fig).Therefore, we do not have evidence that ire-1 is a splicing target of mbl-1.
MEC-3 functions as the vital TF to control essential downstream targets for PVD dendrite development.In mec-3 null mutant, the highly branched PVD dendritic arbors are nearly completely lost with only the unbranched primary dendrite (Fig 2A and 2B).This phenotype was similar, albeit stronger, as the phenotype of the mbl-1 mutants (wy888, tm1563, wy560), suggesting they might function in the same pathway.To test this idea, we constructed the double mutants between mec-3 and mbl-1 and the data showed that mbl-1 could not enhance mec-3 phenotype (Fig 2A and 2B).This result is consistent with the notion that they function in the same genetic pathway, and raises the hypothesis that mec-3 might be the splicing target of mbl-1.To test if there are changes of the mec-3 transcripts in the mbl-1 mutant, we designed primers to examine the alternative splicing of mec-3.Comparing to the wild type, the mbl-1 mutant showed different band sizes for fragments near the 5' end of the mec-3 transcript.Specifically, in wild type the longer transcript, corresponding to the mec-3a isoform, was brighter, while the shorter transcript, corresponding to the mec-3d isoform, was dimmer.In mbl-1(wy888) mutant, the shorter transcript becomes the dominant species (Fig 2D).Sequencing experiments showed that the difference between the two isoforms was that the second exon was included in the longer mec-3a isoform and spliced out for the shorter mec-3d isoform.The second exon is 51 base pair long and would not create frameshift when it is missing.The corresponding protein sequence locates at the N-terminus of the protein, in the first LIM domain (S3 Fig), which is known to be a protein-protein interaction domain.Consistently, the two null alleles of mbl-1 mutant showed similar RT-PCR gel pattern (Fig 2D).To quantitatively assess the splicing defects, we used qRT-PCR to measure the level of the mec-3a isoform in WT, mbl-1(wy888), mbl-1(tm1563), and mbl-1(wy560) animals.The level of mec-3a isoform in wy888 allele is less than 2% of the wild type control, while the two null alleles have about 6% of the level found in the control (Fig 2E).This is consistent with the finding that the wy888 allele of mbl-1 showed more severe PVD dendrite phenotype than the two null alleles.These data suggest that mbl-1 promotes the mec-3a isoform and inhibits the mec-3d isoform.To test if MBL-1 affects known MEC-3's transcriptional targets, we used qRT-PCR to measure the levels of acp-2, hpo-30, T24F1.4,and egl-46 [25].The level of egl-46 transcripts is significantly lower in mbl-1( wy888 Because mbl-1 likely regulates many mRNAs, we wonder whether the reduced mec-3a is largely responsible for the PVD dendrite phenotype in the mbl-1 mutants.We used two approaches to test this idea.First, we expressed the mec-3a cDNA in the mbl-1(wy888) mutant background using a PVD specific promoter.Remarkably, this transgene largely rescued the mbl-1 mutant dendrite phenotype (Fig 2A and 2B), suggesting the mbl-1(wy888) mutant phenotype is mainly caused by the dramatically lowered level of mec-3a.It is reported that mbl-1/ MBNL1 functions through consensus binding sequence GCUU [38].In addition, we found 1 GCUU in the first intron and 3 GCUU in the second intron of mec-3, suggesting the cis-elements might be conserved.
Second, we used CRISPR-Cas9 to generate a mec-3(deExon2) allele by deleting the second exon of mec-3.By RT-PCR assay, we verified that this allele completely lacked mec-3a (Fig 2D).Indeed, mec-3(deExon2) showed dramatic reduction of dendrite complexity.This phenotype was stronger than the mbl-1 mutants but was not severe as the mec-3 null mutants (Fig 2A and 2B).These results are consistent with the fact that there are still 2-6% of the mec-3a transcripts left in the mbl-1 mutants.Together, these data indicate that the MEC-3a isoform carries the majority of MEC-3's activity and that the lack of the mec-3a isoform in mbl-1 mutants account for their dendrite phenotypes.Compared with mec-3 null mutant, mec-3(deExon2) PVD could still elaborate some dendrite branches, suggesting that MEC-3d is partially active, likely with lower activity than MEC-3a.Indeed, overexpression of PVD driven MEC-3(deExon2) rescued the mec-3(deExon2) and CRISPR constructed mec-3(wy50748) null phenotype, suggesting that MEC-3d carries partial transcription factor activity, albeit lower than MEC-3a (Fig 3A -3D).In contrast to the highly branched dendrites of PVD, the other MEC-3 expressing neurons have simple, non-branched dendrites and therefore are not suitable for studying complex dendrite morphogenesis.To test if mbl-1 functions in other MEC-3 expressed neurons, we crossed mbl-1(wy888) into zdIs5 strain, which labelled the touch receptor neurons (AVM and PVM).The results showed that mbl-1 mutant did not affect the neurite morphology of AVM and PVM neurons (S6 Fig) .It suggests that AVM and PVM might only requires a very low functional MEC-3 or mbl-1 might not function in these neurons at all.
We overexpressed mec-3a isoform (full-length) cDNA in the PVD neurons of WT worms.The overexpression resulted in subtle but statistically significant increase in the number of 2˚, 3˚, and 4˚dendrites (S8A Fig) .mec-3a overexpression in eel-1(wy50554) mutation background did not further increase the dendrite branching (S8B Fig) , suggesting that there is a ceiling effect of the level of MEC-3.
To further test if EEL-1 functions as the E3 ligase to degrade MEC-3, we used CRISPR to disrupt the function of the HECT domain, which is the catalytic ubiquitin ligase domain.We isolated one frameshift mutation, a single base deletion (A11171 deletion), in mec-3(deExon2) mutant background.This mutation is immediately upstream of HECT domain and should cause a complete deletion of the HECT domain (S7 Fig) .The dendrite phenotype of mec-3 (deExon2) mutant was significantly suppressed (S9 Fig) , indicating that the HECT ubiquitin ligase domain may be essential for eel-1's function in suppressing mec-3(deExon2) phenotype, although we are not sure if this truncation mutant might affect EEL-1 protein stability or expression levels.The possibility of non-ubiquitin ligases mechanisms for EEL-1 function in shaping dendrite complexity can still not be ruled out.

eel-1 mutants increase the endogenous MEC-3 protein level in PVD neurons
To further verify the idea that EEL-1 controls PVD dendrite morphology through degrading MEC-3 proteins, we generated an endogenous MEC-3::GFP (GFP fused to MEC-3 before stop codon) knockin strain using the CRISPR-Cas9 technique.As expected, MEC-3::GFP localized in PVD nucleus (Fig 5A).To test if eel-1 mutants modulate endogenous MEC-3 protein level, we created three independent eel-1 knock out alleles (wy50884, wy50885, wy50886) using Comparing the GFP fluorescence intensity with wild type, we found that the MEC-3::GFP signal was increased in all three independent eel-1 knock out strains (Fig 5B and 5C).This result is consistent with our hypothesis that the putative ubiquitin ligase EEL-1 downregulates MEC-3 to control PVD dendrite morphogenesis.

Discussion
While it is well appreciated that TFs specify dendrite morphology, the regulation of the TFs themselves is not well understood.In C. elegans, the key TF MEC-3 directs several downstream proteins such as HPO-30 and EGL-46 to shape PVD dendrite.How MEC-3 is regulated at mRNA and protein level is not clear.
We identified a background mutation named mbl-1(wy888) from RB1712 strain.Further investigation reveals that MBL-1 regulates the alternative splicing of mec-3.First, mbl-1 could not enhance mec-3 mutant phenotype but it does enhance hpo-30, dma-1, tiam-1, and kpc-1 phenotype (S2A Fig) , suggesting mbl-1 functions in the same pathway as mec-3.Interestingly, we also found that mbl-1 could not enhance ire-1 mutant phenotype, however, we did not observe ire-1 splicing defect in mbl-1 mutant (S2B and S2C Fig) .Second, the full length mec-3 cDNA (a isoform) rescued mbl-1 phenotype largely, suggesting mec-3 is the main downstream target of mbl-1 although we could not preclude any other putative targets.Third, RT-PCR results demonstrated that the ratio between the mec-3 a isoform and d isoform (lacking the second exon) was reversed with higher level of d isoform and nearly null a isoform in mbl-1 mutant, suggesting mbl-1 regulates the alternative splicing of the mec-3 second exon.Fourth, the orthologue of MBL-1 functions as alternative splicing factor and mbl-1 is reported to splice sad-1 in worm neuron [38].Although it has been well reported that MBLN1 directly interacts with the pre-mRNA of its splicing targets in other models [41], there is still no evidence that MBL-1 binds mec-3 pre-mRNA directly in C. elegans.Interestingly, MBL-1 is also a downstream target of MEC-3 [38], indicating that a positive feedback exists between the two genes to pattern PVD dendrite.
Interestingly, the wy888 allele shows stronger phenotype than the two putative null alleles.The wy888 point mutation locates at H131, which is the conserved C3H motif binding zinc ion.One possible explanation is that the H131Y mutation likely impairs zinc binding and then affects binding between mec-3 mRNA and MBL-1(H131Y).In the two mbl-1 null alleles, MBL-1 protein is completely absent while MBL-1(H131Y) protein might still bind to mec-3 mRNA and block access of other alternative splicing factors.In fact, a previous paper showed that the C3H-type zinc finger modules containing protein Makorin competes with Bru1 for binding to the substrate osk mRNA [42].
Why is mbl-1 mutant phenotype weaker than mec-3 null mutant?We propose that the second exon lacking isoform mec-3d exhibits partial function of the full length mec-3a isoform.First, the exon 2 specific knock-out allele mec-3(deExon2) was weaker than mec-3 null allele, but stronger than mbl-1(wy888), which is probably caused by the few mec-3a isoform in mbl-1 (wy888).Second, overexpression of mec-3(deExon2) construction could rescue mec-3 (deExon2) and mec-3 null allele to a large extent, indicating that although the second exonlacking mec-3d isoform is not potent as the mec-3a isoform, it does show partial function and can compensate the mec-3a isoform function when overexpressed.
Besides regulation at the mRNA level, our results also show that EEL-1, the E3 ligase, downregulates the activity of MEC-3d possibly through proteostasis, but further biochemical studies are needed to determine if MEC-3 is a definitive EEL-1 substrate.First, eel-1 suppressed mbl-1 mutant phenotype, suggesting it functions in the same pathway as mbl-1.Then it could also suppress mec-3(deExon2) phenotype, suggesting it functions through mec-3.Second, eel-1 could not suppress mec-3 null phenotype.The possible explanation is that in mec-3(null); eel-1 double mutants, no MEC-3 protein is produced and the eel-1 mutation could not increase MEC-3 level when there is no MEC-3.Therefore, eel-1 mutation could not suppress the phenotype of mec-3(null), although we could not completely exclude other possibilities.Third, eel-1 mutant exhibited more branches than the wild type, reminiscent of high level of MEC-3 protein, suggesting EEL-1 degrades MEC-3 to shape PVD dendrite.Further biochemical experiments are still needed to strengthen this hypothesis.
During PVD development, the essential TF MEC-3 should be tightly controlled.The splicing factor MBL-1 splices the second exon of mec-3 to provide enough potent mec-3a isoform to produce full functional TF MEC-3.At the same time, the MEC-3 level should be balanced by the E3 ligase EEL-1 to keep normal MEC-3 protein level, avoiding excessive branching of PVD.These elegant regulation mechanisms at post-transcriptional and post-translational ensure the proper level of MEC-3 to pattern appropriate dendrites of PVD (Fig 6A -6C).
Many neuronal diseases have been reported to be associated with MBL-1 and EEL-1, and our findings, which the TF MEC-3 might function as their target, will enlighten the mechanism of these diseases.

C. elegans strains and genetics
C. elegans strains were cultured at 20˚C using nematode growth medium (NGM) plates.Escherichia coli OP50 was seeded in NGM plates.The reference strain is wild-type N2.All the mutants and transgenic strains were genetically modified based on N2.

Isolation and mapping of mutants
We isolated wy888 from RB1712 strain.Based on mbl-1(wy888); wyIs592, we isolated wy50554 by a modifier screen of 3000 haploid genomes.50mM ethyl methanesulfonate (EMS) was used as mutagen in these two screen processes.Genetic mapping and sequencing showed that the wy888 mutant is a C-to-T point mutation in exon 4 of the mbl-1 gene, which changes a histidine to a tyrosine (c.C391T:p.H131Y).In addition, the wy50554 mutant is a G-to-A point mutation in the exon 8 of the eel-1 gene, which changes a glutamic acid to a lysine (c.G7423A: p.E2475K).SNP mapping was performed using standard methods [43].

Constructs and transgenes
Plasmid constructs were generated in pPD95.77vector.The Clontech In-Fusion PCR Cloning System was used for vector construction and these constructs were verified by sequencing to ensure the correct results.The PCR products were amplified with Phusion DNA polymerase (New England Biolabs) or TransStart FastPfu DNA Polymerase (Transgen Biotech) by standard procedures.Standard microinjection techniques were used to generate the transgenic worms.The DNA plasmids were injected into N2, mbl-1(wy888), wyIs592, mec-3(deExon2), and mec-3(wy50748) hermaphrodites at a concentration from 10 to 50 ng/ml.2 ng/ml Pmyo-2::mCherry plasmids were used as the co-injection marker.Knockout and knockin worms were generated by standard CRISPR/Cas9 technique.We generated mec-3(deExon2) allele by co-injection of the mixture of Peft-3_mec-3_gRNA (CRISPR-Cas9 plasmids) and repair templates, deleting the entire exon2 of mec-3 in the genome.

Fluorescent imaging
Young adult animals were anesthetized with 1 mg/ml levamisole in M9 buffer before mounted on 3% (w/v) agar pads.Images of the mec-3::GFP and the three corresponding independent eel-1 KO worms were captured in live animals at the same exposure time (2000 ms) using a 63× objective by AxioImager M2 microscope (Carl Zeiss).And images of PVD dendritic arbors of the related animals were captured by the spinning-disk confocal imaging system which includes an Axio Observer Z1 microscope (Carl Zeiss MicroImaging) equipped with a 40× objective, an electron-multiplying charge-coupled device camera (Andor), and the 488and 568-nm lines of a Sapphire CW CDRH USB Laser System attached to a spinning-disk confocal scan head (Yokogawa CSU-X1 Spinning Disk Unit).Micro-Manager (https://micromanager.org)software and ImageJ (http://rsbweb.nih.gov/ij/)software were used to process the images.

Quantification and statistics
The number of 2˚, 3˚, and 4˚dendrites of young adults was counted under the microscope using a 40× objective.The MEC-3::GFP mean intensity in the nucleus was measured by ImageJ software.All statistical tests were performed using one-way ANOVA with Tukey's multiple comparisons test in Graphpad Prism.

Fig 6 .
Fig 6.Schematic model of the regulation mechanisms of dendrite morphogenesis in C. elegans PVD neuron.(A) In wild-type animals, MBL-1 regulates the alternative splicing of mec-3, producing the great majority of mec-3a isoform and very few of the mec-3d isoform mRNAs, therefore promotes the normal dendrite matures.(B) In mbl-1 mutant, the ratio of mec-3a isoform and d isoform changes in that the d isoform is predominant, hampering the transcriptional efficiency of mec-3 downstream targets, ultimately decreasing dendrite complexity significantly.(C) In mbl-1 and eel-1 double mutants, the increased MEC-3d isoform quantity due to the E3 ligase EEL-1 mutant, although with lower transcriptional activity, restores the dendritic arbors to a WT-like morphology.Alternatively, EEL-1 functions indirectly to regulate MEC-3.https://doi.org/10.1371/journal.pgen.1010941.g006