Multilayered regulations of alternative splicing, NMD, and protein stability control temporal induction and tissue-specific expression of TRIM46 during axon formation

The gene regulation underlying axon formation and its exclusiveness to neurons remains elusive. TRIM46 is postulated to determine axonal fate. We show Trim46 mRNA is expressed before axonogenesis, but TRIM46 protein level is inhibited by alternative splicing of two cassette exons coupled separately to stability controls of Trim46 mRNA and proteins, effectively inducing functional knockout of TRIM46 proteins. Exon 8 inclusion causes nonsense-mediated mRNA decay of Trim46 transcripts. PTBP2-mediated exon 10 skipping produces transcripts encoding unstable TRIM46 proteins. During axonogenesis, transcriptional activation, decreased exon 8 inclusion, and enhanced exon 10 inclusion converge to increase TRIM46 proteins, leading to its neural-specific expression. Genetic deletion of these exons alters TRIM46 protein levels and shows TRIM46 is instructive though not always required for AnkG localization nor a determinant of AnkG density. Therefore, two concurrently but independently regulated alternative exons orchestrate the temporal induction and tissue-specific expression of TRIM46 proteins to mediate axon formation.

N eurons are highly polarized, containing an axon that conducts action potentials to directionally transmit information between neurons. An axon is specified during neuronal differentiation and is a prerequisite to synapse and neural circuit formation. Newly generated immature neurons typically contain multiple seemingly indistinguishable neurites. One neurite extends its growth and is specified for an axonal fate, leading to the establishment of axonal and somatodendritic compartments. Axon formation appears to involve a diverse set of molecules, including growth factors, signal transducers, kinases, phosphatases, cytoskeletal proteins, and molecular motors [1][2][3][4][5][6][7][8][9][10][11][12] . Substantial progress has been made toward understanding the cellular processes mediated by these molecules, although the gene expression regulation underlying these processes remains elusive.
The search for axon determinants has focused on symmetrybreaking molecules exhibiting differential localization at the future axon vs other neurites. Localization of kinesin motor proteins (e.g., KIF5C, KIF1A, and KIF3A) in the nascent axon plays an essential role in establishing the axonal domain by promoting polarized cargo trafficking 6,13,14 . Other proteins separating neuronal compartments include widely used axonal marker Tau1 and dendritic marker microtubule-associated protein 2 (MAP2) 15 . However, knockout (KO) of Tau1 or Map2 in mice show no obvious effects on brain development or neuronal polarity, probably due to redundant functions of other MAPs such as MAP1B [16][17][18] . Many kinesin knockouts show diverse phenotypes in mice, commonly disrupting cargo delivery and reducing axonal transport rates. However, such mice do not exhibit clear developmental defects of axon formation either [19][20][21] .
Amidst the multi-decade chase for the earliest molecular determinants of axon specification, tripartite-motif containing 46 (TRIM46) has emerged as another promising candidate. TRIM46 was recently shown to localize to future axons earlier than Tau1, suggesting TRIM46 is one of the earliest, if not the earliest, markers for axon specification 4 . TRIM46 localization at the proximal axon partly overlaps with the future axon initial segment (AIS). The AIS is a distinct neuronal compartment at the base of the axon best known for the generation of the action potential and regulation of cargo trafficking 9,18,19 . AIS is marked by the accumulation of AnkG, which is essential to the maintenance of axo-dendritic polarity. It has been proposed that TRIM46 is transported under low MARK2 activity to the future AIS where TRIM46 organizes microtubule fascicles leading to enrichment of AnkG at the AIS site 4,[22][23][24] .
While regulation of TRIM46 subcellular localization is a key component influencing axonal fate, temporal control of TRIM46 expression likely serves another important checkpoint but this remains to be determined. Axon consistently emerges during a specific developmental time window, arguing for the importance of temporal regulation. Furthermore, neural-specific expression of axonal determinants could be a parsimonious explanation for neurons' exclusive ability to generate an axon. However, gene expression controls of molecular candidates, including Trim46, have rarely been examined. Finally, despite accumulating evidence for TRIM46's important role, the effect of a complete depletion of TRIM46 is unknown (e.g., whether TRIM46 is always required for AnkG localization). This knowledge is critical to guide the future direction of research.
Alternative splicing of precursor mRNA produces mRNA isoforms from a single gene and is an essential regulatory mechanism of gene expression. Mammalian brains exhibit prevalent splicing control governed by RNA binding proteins (RBPs) [25][26][27][28][29][30][31] . Early cortical axonogenesis is controlled by a neuralspecific alternative splicing program coordinated by polypyrimidine tract binding protein 2 (PTBP2) 32 . The RBFOX family proteins control alternative splicing of many genes (but not Trim46) encoding the AIS components to regulate AIS formation 33 . Axon guidance of dorsal interneurons in the spinal cord is controlled by NOVA proteins, which regulate isoform expression of attractive DCC and repulsive ROBO receptors of growing commissural axons [34][35][36] . Mutation of these RBPs show detrimental defects in brain development, likely resulting from mis-regulation of a large set of alternative exons. However, the biological significance of most individual alternative splicing events and their possible contribution to axon formation are largely unknown 28 .
The nonsense-mediated mRNA decay pathway (NMD) functions as a mRNA quality control mechanism that selectively degrades transcripts containing a premature termination codon (PTC), preventing translation of abnormal C-terminal truncated proteins [37][38][39] . While NMD is classically recognized as a surveillance pathway, it is increasingly viewed as a regulatory process with an active role in fine-tuning the abundance of normal transcripts [40][41][42][43][44] . In mammalian cells, alternative splicing coupled with nonsense-mediated decay (AS-NMD) is a conserved mechanism of post-transcriptional gene regulation. A common NMD-inducing feature is a PTC at least 50 nt upstream of an exon-exon junction [37][38][39] . Normal stop codons are typically found in the last exon downstream of any exon-exon junction. Alternative splicing can alter the reading frame to produce a PTCcontaining spliced transcript subject to NMD 31,[45][46][47][48] . By shifting the splicing of NMD-associated alternative exons, AS-NMD effectively controls homeostatic expression of multiple splicing factors [49][50][51][52][53][54] . AS-NMD also regulates the expression of developmental genes in the brain (e.g., Psd-95, Bak1, and histone modifiers) 40,46,49,55 . For Psd-95 and Bak1, the NMD-associated exons exhibit tissue-specific splicing, resulting in neural-specific expression of the PSD-95 protein and neural-specific loss of the BAK1 protein. Therefore, AS-NMD can play an important role in regulating the overall abundance of a functional protein during neural development.
In this study, we investigated whether and how Trim46 expression is developmentally regulated during neuronal differentiation to control axon formation. We found Trim46 expression is increased in concordance with axonogenesis. Surprisingly, Trim46 upregulation is orchestrated by alternative splicing of its exon 8 and exon 10 in addition to transcriptional upregulation. Exons 8 and 10 are independently regulated and lead to different downstream control of Trim46 expression, but resulting in convergent upregulation of the functional TRIM46 protein to enable axon formation. Using CRISPR/Cas9-mediated exon knockout, we show alternative splicing controls of exons 8 and 10 are functional for TRIM46 induction and AnkG recruitment to the proximal axon. Therefore, multiple control mechanisms act synergistically to determine the temporal induction of TRIM46 during early axon formation.

Results
Trim46 exon 10 is developmentally regulated. The success in differentiating embryonic and induced pluripotent stem cells into functional neurons in the absence of an in vivo microenvironment presents a unique system for study of the temporal control of axon formation. We differentiated the well-established 46C Sox1-GFP knock-in mouse embryonic stem cells (ESC) into neurons (Figs. 1a and 1b) [56][57][58][59] . The GFP expression under the control of the Sox1 promoter provides a convenient readout of successful differentiation toward the neural lineage and of obtaining neural progenitors. Briefly, ESCs were dissociated in single cells and replated in 96-well u-bottom and agarose-coated low cell adhesion plates. With differentiation medium, these cells reaggregate over time and form embryoid bodies (EB) resembling ectoderm cells 57,60 . On day 4, retinoic acid (RA) is added to activate neurogenic pathways, and the induction of neural lineages is highlighted by increasing GFP expression from day 6 to day 8 (Fig. 1b). On day 8, Sox1-GFP expressing EBs were dissociated and plated as a monolayer (Figs. 1a and 1b). EBderived cells became neural progenitor cells (NPC) on day in vitro 1 (DIV 1) [56][57][58]61 . Further differentiation led to the polarization of post-mitotic neurons around DIV 2/3. Many neurons display a singular long cellular process (the future axon) and other short processes (the future dendrites) on DIV 3. By DIV Supplementary Fig. 1b, 1c). Its expression further increased at DIV 5 and DIV 7 (Figs. 1c, 1d, and Supplementary Fig. 1b). Therefore, TRIM46 protein is not expressed in neural progenitors and is induced concurrent with the emergence of an axon.
We searched databases of NCBI RefSeq and Ensembl for annotated Trim46 transcripts and found exon 8 and exon 10 are the two cassette exons consistently annotated in both databases. Importantly, skipping exon 10 causes a shift in the reading frame at the C-terminus, resulting in a much shorter isoform (TRIM46S) with an alternative stop codon in the last exon, exon 11 (Fig. 1e). We note that the TRIM46 protein reported in previous studies is the long isoform (TRIM46L) including exon 10 and TRIM46S is an uncharacterized isoform 4,22 . The large difference between TRIM46L (83 kDa) and TRIM46S (60 kDa) proteins suggests alternative splicing of exon 10 has functional implications.
To test whether exon 10 is alternatively spliced during neuronal differentiation we designed primers annealing to exon 9 and 11 to simultaneously detect the inclusion and skipping isoforms and performed quantitative alternative splicing analysis (Figs. 1e and 1f). We found EBs expressed Trim46S transcripts almost exclusively. During the EB-to-DIV 1 transition, the dominant Trim46 mRNA isoform switched from Trim46S to Trim46L, and the trend continued until DIV 7 (Fig. 1f). Note that on Western blots, the upregulated TRIM46 protein is the long isoform mirroring the upregulation of the Trim46L mRNA. Therefore, the induction of the TRIM46 protein follows the mRNA isoform switch from Trim46S to Trim46L. Surprisingly, despite clear expression of Trim46S mRNA in EBs and on DIV 1, we could not detect the TRIM46S protein during these stages, indicating additional regulation to restrain TRIM46 protein expression.
The major wave of neurogenesis in the mouse neocortex occurs during embryonic day 13 (E13) and E16. Axons emerge about 1-2 days after neurogenesis and continue to grow and mature during the embryonic period. We found exon 10 splicing increases from 12.7% in E14.5 to 62.7% in E18.5 WT cortices (Figs. 1g and 1h), consistent with the splicing switch during the ESC differentiation that generates the Trim46L isoform (Fig. 1f). This upward trend is also consistent with the reanalysis of cortex RNA-seq data from the ENCODE project ( Supplementary Fig. 2a,  2b). In the Ptbp2-cKO cortices, exon 10 inclusion was consistently and significantly higher than in WT (22.5% vs 12.7% on E14.5, 43.3% vs 24.3% on E16.5, and 76.6% vs 62.7% on E18.5, Figs. 1g and 1h). These results demonstrate that loss of splicing regulator PTBP2 leads to an increase in the exon 10 + Trim46L variant during development.
In further support of PTBP2 regulating exon 10 splicing, we examined the PTBP2 CLIP-Seq dataset of E18.5 mouse brains 63 . PTBP2 typically binds flanking intronic sequences to regulate an alternative exon 62,63 . As shown in Fig. 1i, PTBP2 interacts with multiple polypyrimidine tracts within intron 9 and intron 10 of Trim46. These polypyrimidine tracts run from 11 to 24 nt and on average consist of 89% C and U nucleotides. There was also a strong PTBP2 CLIP-Seq signal in exon 11, represented by two notable CU-rich tracts from 9 to 24 nt ( Supplementary Fig. 3a). Based on the PTBP2 CLIP-Seq data from embryonic mouse brains and analyses of embryonic PTBP2 conditional KO brains, we conclude PTBP2 represses the inclusion of Trim46 exon 10.
Presumptive endogenous TRIM46S protein isoform is undetectable. Interestingly, both Trim46L and Trim46S mRNA transcripts were easily detected, but only the TRIM46L protein isoform was detected. We needed to rule out the absence of TRIM46S protein as an artifact of TRIM46 antibodies. We expressed N-terminal GFP-tagged TRIM46 protein isoforms (GFP-TRIM46L and GFP-TRIM46S) in N2a cells for immunoblot analysis. Two different TRIM46 antibodies (from the Hoogenraad lab and Proteintech Inc.) readily detected the ectopically expressed TRIM46S protein isoform (Supplementary Fig. 4a and  4b). By comparing the GFP signal to the TRIM46 signal, the Hoogenraad lab antibody was modestly less efficient in detecting TRIM46S, but the Proteintech antibody showed the same level of sensitivity to both isoforms. To further validate the antibody detecting TRIM46S, we ectopically expressed N-terminal FLAGtagged TRIM46 proteins (FLAG-TRIM46L and FLAG-TRIM46S) and confirmed detection of both protein isoforms with the same sensitivity by the Proteintech TRIM46 antibody ( Supplementary  Fig. 4c). The Proteintech antibody was used in the remaining experiments. Our results demonstrated that the absence of endogenous TRIM46S protein was not due to detection sensitivity of TRIM46 antibodies. Fig. 1 Trim46 exon 10 is developmentally regulated. a Schematic protocol for embryoid body (EB) based neuronal differentiation using 46C embryonic stem cells (ESC). Differentiation medium included Glasgow's modification of Eagle's Minimal Essential Medium (GMEM) and Knockout Serum Replacement (KSR) for EB day 0 to day 8 while retinoic acid (RA) was used from EB day 4 to day 8. b Representative image of various differentiation stages from ESC to EB to neural progenitor cells (NPC), and into neurons from 45 independent experiments. Scale bar, 100 µm. c Western blot for TRIM46 protein expression from ESC-derived neurons on DIV 3, 5, and 7. d Quantification shows TRIM46 protein is upregulated during neuronal differentiation. Data represent mean ± SEM from 4 independent experiments. e Schematic showing mRNA isoforms produced from alternative splicing of Trim46 exon 10. The inclusion of exon 10 represents the long isoform or (Trim46L) while the exclusion of exon 10 leads to shorter isoform (Trim46S). F1 and R1 show the location of primers for RT-PCR analysis of Trim46 splicing. f Gel image shows isoform expression of Trim46 mRNA in WT EBs on day 8 to ESC-derived neurons on DIV 7 during differentiation from 4 independent experiments. g Gel image highlighting alternative splicing of Trim46 exon 10 between CTRL and Ptbp2-cKO mouse neocortices on E14.5, E16.5, and E18.5 from 3 independent experiments. h Splicing quantification shows upregulation of Trim46 exon 10 PSI (percent spliced-in) in Ptbp2-cKO neocortices. Data is represented as mean ± SEM. Statistics: t-test, two-tailed, unpaired; (E14.5: CTRL, n = 7 and KO, n = 3; *p = 0.02133; E16.5: CTRL, n = 7 and KO, n = 3; **p = 0.003421; E18.5: CTRL, n = 3 and KO, n = 2; *p = 0.02448). i Genome browser track of E18.5 PTBP2 CLIP-Seq shows multiple distinct PTBP2 binding sites located in Trim46 introns 9 and 10. Two prominent binding regions are highlighted. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-29786 -4 We hypothesized that the lack of TRIM46S protein detection was due to the insolubility of TRIM46S in protein lysates. We overexpressed FLAG-TRIM46L and FLAG-TRIM46S proteins in N2a cells and compared TRIM46 expression in the supernatant versus the pellet fractions. We found most TRIM46L and TRIM46S proteins remained in the supernatant, as for either isoform only about 17% of TRIM46 proteins were in the pellet fraction ( Supplementary Fig. 5a). Considering overexpression could enhance misfolding and insolubility of the ectopic proteins, endogenous TRIM46 proteins may have a lower insoluble proportion. Indeed, when we examined the pellet fractions from protein lysates of ESC-derived neurons, we observed no presumptive TRIM46S proteins by Western blot (Fig. 2a). These data suggest other causes for this absence of detection of the endogenous TRIM46S protein.
To enhance detection sensitivity, we attempted to enrich endogenous TRIM46S proteins by immunoprecipitation (IP). We first tested whether the TRIM46 antibody was suitable for IP using ectopically expressed TRIM46 proteins. We transfected N2a cells with plasmids expressing FLAG-Trim46L and FLAG-Trim46S cDNA and collected whole cell lysates 48 h after transfection. We used the TRIM46 antibody along with protein A magnetic beads in complex to pull down FLAG-TRIM46L or FLAG-TRIM46S proteins. FLAG-TRIM46L and FLAG-TRIM46S proteins were easily immunoprecipitated by the TRIM46 antibody but not the control IgG antibodies (Fig. 2b).
Next, we performed TRIM46-IP using P10 mouse cortices and E12.5 mouse brains. Endogenous TRIM46L protein was effectively immunoprecipitated and detected in P10 mouse cortices (Fig. 2c). TRIM46L proteins were also detected by TRIM46-IP in the E12.5 mouse brains but at a much lower level (Fig. 2d). In neither condition did we observe a protein band corresponding to the approximate 60 kDa size of TRIM46S protein. Importantly, E12.5 cortices expressed predominantly the Trim46S mRNA variant but only the TRIM46L protein isoform was detectable in this experiment.
Another possibility for the missing endogenous TRIM46S protein was that its epitope for the TRIM46 antibody was somehow masked (e.g., by post-translational modifications). This is unlikely: two different effective polyclonal TRIM46 antibodies failed to detect TRIM46S proteins under the denaturing conditions of SDS-PAGE. Nevertheless, epitope masking remained a possibility for the IP experiment. We decided to knock in a 3x-FLAG tag sequence to fuse in-frame at the N-terminal of the endogenous Trim46 locus and replace the start codon. This would create an allele endogenously expressing N-terminal FLAG-tagged TRIM46 protein and we would apply a commonly used FLAG antibody with high IP efficiency to detect and identify the uncharacterized short isoform.
Western blot analysis using a FLAG antibody further confirmed the KI allele. However, using total protein lysates we only detected the 3xFLAG-TRIM46L protein from the lysate of neurons derived from 3xFKI:Trim46 ESC (Fig. 2i). We conducted FLAG-IP experiments using protein lysate from 3xFLAG-TRIM46-expressing neurons. Again, we were able to only identify FLAG-TRIM46L proteins consistently. An approximate 60 kDa band that migrated slightly faster than the expected FLAG-TRIM46S proteins occasionally showed up but was inconsistent among biological replicates (Fig. 2j). Nevertheless, this band was dramatically weaker than FLAG-TRIM46L. In summary, we conclude that the endogenous TRIM46S protein is difficult to detect with existing methods.
TRIM46S protein variant is less stable than TRIM46L protein.
We hypothesized protein stability might be different between the two protein isoforms. We tested this using ectopically expressed TRIM46 protein isoforms. The difference in protein stability can be reflected by steady-state abundance when protein translation is modest and at par with the protein decay level; overabundant translation can shadow protein degradation activity. We transfected increasing amounts of FLAG-Trim46L and FLAG-Trim46S cDNA plasmids (in the CAGIG backbone) in N2a cells and quantified protein isoform expression as a function of Trim46 plasmid amount. Backbone plasmid CAGIG was added to make the total transfection amount consistent and allowed EGFP as an internal expression control. At limiting dosages of 0.1, 0.2, and 0.4 µg of transfected plasmid, the FLAG-TRIM46S protein was expressed at a significantly lower level than the FLAG-TRIM46L isoform (Figs. 3a and 3b). There was no difference in expression levels with 1.0 µg DNA transfection, likely due to excess plasmid (Figs. 3a and 3b). Different protein outputs can be due to different mRNA levels of the two FLAG-tagged isoforms. To clarify this potential effect, we measured FLAG-Trim46L and FLAG-Trim46S transcripts at corresponding transfection dosages and found negligible difference of mRNA levels between isoforms (Supplementary Fig. 6a). These results indicate that the lack of TRIM46S protein detection in comparison to TRIM46L protein can be mimicked by expressing low amounts of exogenous TRIM46 proteins.
To further evaluate protein stability differences between the two isoforms, we performed a protein stability assay using translation inhibitor cycloheximide (CHX). N2a cells expressing FLAG-TRIM46L or FLAG-TRIM46S proteins were treated with CHX to inhibit protein synthesis and the remaining FLAG-TRIM46 protein isoforms were measured at subsequent time points. Western blot analysis showed FLAG-TRIM46 protein abundance decreased over a 24-h period after CHX treatment. FLAG-TRIM46 proteins were first normalized to GFP and then to starting levels of the first collection time point (0 h, i.e., right before CHX treatment). There were significant differences in the remaining protein proportions between FLAG-TRIM46L and FLAG-TRIM46S at 6, 12, and 24 h after CHX treatment (Figs. 3c and 3d). For example, at 12 h, 31% FLAG-TRIM46L protein and 6% FLAG-TRIM46S protein remained (Figs. 3c and 3d).
The difference was also obvious when we co-transfected FLAG-Trim46L and FLAG-Trim46S cDNA plasmids together and detected the remaining protein isoforms after CHX treatment ( Supplementary Fig. 6b), suggesting the stability of the two protein variants is differentially regulated even if they would complex with one another. We calculated the protein decay rate for each isoform. FLAG-TRIM46S had a shorter half-life of 1.9 h versus FLAG-TRIM46L of 4.3 h (Fig. 3d). Overall, these results showed the TRIM46S protein isoform is less stable than the TRIM46L protein.
Next, we tested whether TRIM46L and TRIM46S were subject to differential regulation by well-known protein degradation pathways. Over 90% of all proteins in cells are removed by the ubiquitin-proteasomal system (UPS) and the lysosomal pathway [68][69][70][71][72] . To investigate if the UPS pathway regulates protein degradation of TRIM46 proteins, we treated N2a cells expressing FLAG-TRIM46L or FLAG-TRIM46S with proteasome inhibitor MG132. We found FLAG-TRIM46S was stabilized by MG132 in a dose-dependent manner ( Supplementary Fig. 6c). We chose 10 µM MG132 for subsequent experiments because higher concentrations appeared to induce cellular toxicity and because proteasome inhibition was evident with increased ubiquitinated proteins ( Supplementary Fig. 7a). In comparison to DMSO control, 10 µM MG132 treatment upregulated both FLAG-TRIM46L and FLAG-TRIM46S proteins. However, the FLAG-TRIM46S protein was more sensitive to MG132 than the FLAG-TRIM46L protein (Figs. 3e and 3f).
We attempted the MG132 experiment on DIV 1 neurons derived from 3xFKI:Trim46 ESCs. However, in 3xFLAG-KI neurons treated with MG132 from 10 nM to 100 µM there was   Fig. 7a). We also extended the treatment time to 24 h, which did not upregulate the ubiquitin signals but did cause massive cell death at various concentrations ( Supplementary Fig. 7b). The MG132 toxicity (probably proteasome-independent) in ESC-derived neurons precluded further investigation.
To test for differences in TRIM46 protein isoform degradation under the lysosomal pathway, we inhibited the pathway with NH4Cl in N2a cells expressing FLAG-TRIM46 proteins. We observed insignificant effects on TRIM46S expression by up to 10 mM NH4Cl but a substantial reduction after treatment of 50 or 100 mM NH4Cl due to cellular toxicity ( Supplementary  Fig. 6d). Importantly, TRIM46L and TRIM46S were not differentially affected and neither protein increased abundance (Fig. 3g), suggesting that TRIM46 protein is unlikely to be regulated by the lysosomal pathway. Overall, these results show TRIM46S proteins were less stable than TRIM46L proteins due to higher sensitivity to UPS-mediated protein degradation.
Ablation of exon 10 results in loss of TRIM46 protein and reduced AnkG localization at the proximal axon. To test the functional significance of exon 10 alternative splicing during axon formation, we generated an exon 10 deletion mutant using CRISPR/Cas9 and tested its effect on TRIM46 protein and AnkG localization to the proximal axon. We designed two sequencespecific guide RNAs targeting intronic sequences flanking exon 10. We then used the CRISPR/Cas9 system to knock out exon 10 (Trim46 ΔE10/ΔE10 or E10-KO) in 46C ES cells (Fig. 4a) so that the derived neurons would constitutively express only the exonskipped E10− variant, allowing the study of loss of TRIM46L function during axon formation.
Targeted exon 10 deletion in WT ES cell line was verified by PCR genotyping (Fig. 4b). This E10-KO line contains two slightly different KO alleles as demonstrated by Sanger sequencing (Fig. 4c). We performed Western Blot analysis for endogenous TRIM46 protein using WT and E10-KO ESC-derived neurons on DIV 5 and confirmed the loss of 83 kDa TRIM46L proteins (Fig. 4d). The TRIM46S protein was not detected despite the loss of exon 10 ( Fig. 4d), which is consistent with previous results showing lower stability of TRIM46S proteins (Fig. 3). We also tested pellets from E10-KO protein lysate and, again, confirmed the absence of TRIM46S protein ( Supplementary Fig. 5b).
The loss of TRIM46 proteins was not due to transcriptional defects in E10-KO because there was no significant difference in total Trim46 mRNA levels between control and mutant neurons ( Fig. 4e). We observed a general upregulation of total Trim46 mRNA levels from EB D8 to DIV 7 in both WT and E10-KO neurons, indicating transcriptional activation. The lack of mRNA expression differences between control and mutant exon 10-KO neurons supports that transcriptional regulation alone is not responsible for the change in TRIM46 protein expression in the mutant neurons. Indeed, the lack of functional TRIM46 proteins in exon 10 KO demonstrates that exon 10 splicing regulation is an essential gatekeeper of TRIM46 protein function.
The major functional outcome of TRIM46 is to influence AnkG recruitment to the proximal axon (AIS) 4,22 . Therefore, we examined AnkG localization in E10-KO ESC-derived neurons. E10-KO and control parental WT ESCs were differentiated into neurons in parallel and stained with TRIM46 and AnkG (Fig. 4f). By DIV 7, WT ESC-derived neurons exhibited TRIM46 accumulation at the proximal axon as expected. By contrast, E10-KO neurons universally lacked the prominent TRIM46 staining in the corresponding region (Fig. 4f), consistent with loss of TRIM46 proteins revealed by Western immunoblots (Fig. 4d). Note that the TRIM46 antibody exhibited nonspecific somatic staining ( Fig. 4f and Supplementary Fig. 1c).
Since no TRIM46 protein was detected in E10-KO neurons (Fig. 4d), exon 10 KO is essentially a null for TRIM46 function. We observed a significant decrease in the percentage of AnkGpositive neurons when comparing E10-KO (51%) vs. WT (86%) (Fig. 4g). Surprisingly, about half of E10-KO neurons retain the capacity for generating AnkG clusters at the AIS site. For those displaying AnkG staining at the proximal axon, we did not observe any significant difference in AnkG length or AnkG intensity between WT and E10-KO neurons (Figs. 4h and 4i). Therefore, TRIM46 is a significant regulator but not absolutely required for AnkG clustering at AIS (see more in Discussion). Together, these results showed that in the absence of exon 10 inclusion, the TRIM46 protein was not detectable and consequently AnkG was less effectively localized to the proximal axon.
We investigated whether neuronal polarity was impaired in E10-KO neurons, which could contribute to the AnkG phenotype. WT and E10-KO neurons were co-stained with axonal marker Tau1 and dendritic marker MAP2 at DIV 5 (Supplementary Fig. 8a). We then quantified the average dendritic intensity (I d ) and average axonal intensity (I a ) of Tau1. We calculated the Tau1 axonal polarity index P ¼ IaÀId IaþId . P > 0 means polarized axonal distribution; P = 0 means uniform distribution; P < 0 means polarized dendritic distribution. As shown in Fig. 2 Presumptive endogenous TRIM46S protein is undetectable. a Immunoblot for TRIM46 protein in the supernatant and pellet fraction from total lysate of WT ESC-derived neurons on DIV 3. Whole cell lysates from N2a cells expressing FLAG-TRIM46L and FLAG-TRIM46S proteins were used as a positive control. Asterisk (*) denotes nonspecific band. b Immunoprecipitation (IP) using protein A magnetic beads incubated with TRIM46 antibody or CTRL IgG to pulldown FLAG-TRIM46L and FLAG-TRIM46S proteins expressed in N2a cells. c P10 cortices (CTX) and d E12.5 mouse brains used in TRIM46-IP experiment to detect endogenous TRIM46 protein isoforms. CTRL-IP used IgG for immunoprecipitation. The two lanes on the right are positive controls showing the position of the expected protein isoforms ran on the same gel: N2a cells expressing FLAG-TRIM46L and FLAG-TRIM46S. e CRISPR/ Cas9 strategy to generate 3xFLAG-knock-in:Trim46 (3xFKI:Trim46) cell line from WT ESC using single-stranded oligo DNA (ssODN) along with exon 1-sgRNA (E1 sgRNA) targeting exon 1. After homology-dependent repair, 3xFLAG is introduced to the endogenous Trim46 locus. F2 and R2 indicate positions of genotyping primers annealing to 5'UTR and exon 1 of Trim46. f Schematic CRISPR/Cas9-based process in producing gene edited cell lines through lipofectamine-based co-transfection of Cas9-sgRNA plasmid, ssODN, and a GFP reporter plasmid. CRISPR/Cas9-edited cell lines were generated by selecting GFP+ colony and diluting cells into 96-well plates for screening. Identification of single-cell derived colonies was performed by genotyping and verified by Sanger sequencing. g PCR results showing genomic DNA band representing the 3xFLAG knock-in element of ES cell line, 3xFKI:Trim46 using primers shown in e. Expected size for band of knock-in is 741 bp. h Sanger sequencing confirming 3xFLAG DNA sequence in 3xFKI:Trim46 cell line. i Western blot using FLAG antibody for endogenously tagged FLAG-TRIM46 protein from 3xFKI:Trim46. Expected protein sizes of 3xFLAG-TRIM46L and 3xFLAG-TRIM46S protein isoforms are 86 kDa and 63 kDa, respectively. The right two lanes show the positive control (but with 1XFLAG): FLAG-TRIM46L and FLAG-TRIM46S expressed in N2a cells. j FLAG-IP experiment to detect 3xFLAG-TRIM46 protein isoforms using FLAG magnetic beads against lysate from 3xFKI: There is essentially no difference among WT and mutants regarding the percentage of neurons with 2 axons (7.8% and 8.6% for control and exon 10 KO, respectively, p = 0.83, two-tailed, unpaired t-test) or without axons (5.8%, 5.5% for control and exon 10 KO respectively, p = 0.87, two-tailed unpaired t-test). Since Tau1 polarity and axon number are unaffected, the exon 10 knockouts do not appear to alter the initial step of axonal polarization but the subsequent formation of AnkG cluster. Therefore, temporal control of Trim46 splicing and expression regulates certain but not all aspects of axon formation.
We also investigated whether exon 10 knockout affected neuronal subtype identity to cause the AnkG phenotype. We stained WT and E10-KO neurons with glutamatergic marker (vGLUT1) and GABAergic marker (GAD67). We found no differences in population composition between WT and E10-KO ESC-neurons: the overwhelming majority are 97% glutamatergic neurons for both ( Supplementary Fig. 8d and 8e). Given similar percentage of vGLUT1+ and GAD67+ cells between control and exon 10-KO neurons this shows the LOF phenotype is not due to alteration in neuronal populations. Exon 8 inclusion leads to nonsense-mediated mRNA decay of Trim46 transcripts and is developmentally downregulated. We next studied another alternative exon (exon 8) of Trim46. The TRIM46 protein is translated from the exon 8 exclusion (E8−) isoform with a stop codon in exon 11 (Fig. 5a). Exon 8 inclusion results in frameshift and termination of the reading frame at a premature stop codon (PTC) in exon 9, making the inclusion isoform a possible target of NMD for degradation without productive translation (Fig. 5a). We observe the inclusion of exon 8 declined dramatically from EB day 8 to DIV 1 and showed continuous downregulation in subsequent stages until DIV 7 (Fig. 5b). The ENCODE data show exon 8 inclusion decreases from E11 to E18 in developing cortices (Supplementary Fig. 2c). Therefore, exon 8 is developmentally regulated during axon formation.
To evaluate whether the exon 8 inclusion isoform is targeted by NMD, we treated cortical neurons with CHX to inhibit NMD. Compared to DMSO control, there was a significant (5.0-fold) increase in the E8+ variant in CHX-treated neurons (Figs. 5c and 5d, p = 1.09 × 10 −10 ). To further confirm NMD regulation of Trim46 transcripts, we also treated N2a cells with CHX. N2a showed a similar upregulation of the E8+ NMD isoform upon CHX treatment (Supplementary Fig. 9a and 9b).
We found the premature stop codon of the exon 8+ isoform is highly conserved across a multitude of species (including all annotated mammals), suggesting that NMD regulation of Trim46 is under selection pressure and is biologically important (Fig. 5i). Given that NMD targets are translationally repressed, exon 8 splicing is a major mechanism restraining TRIM46 protein expression in neural progenitors and developing neurons [37][38][39] . This restraint is alleviated as exon 8 splicing decreases during development, resulting in increased TRIM46 protein expression. Consistent with this idea, the TRIM46 protein is up-regulated during differentiation of ESC-derived neurons and in developing neocortices (Figs. 1c, 5j, and Supplementary Fig. 1b). The above data, taken together, supports the conclusion that a decrease in exon 8 inclusion or a switch from the NMD-sensitive to the NMD-insensitive isoform of Trim46 transcripts allows the increase in TRIM46 protein expression during axon formation.
Ablation of Trim46 exon 8 precociously induces TRIM46 proteins and enhances AnkG localization to the proximal axon.
To examine the functional significance of AS-NMD regulation on Trim46, we deleted exon 8 and its flanking introns from the genome to analyze the effects on TRIM46 expression and subsequent AnkG localization. We designed sequence-specific guide RNAs targeting intronic sequences flanking exon 8. We used CRISPR/Cas9 to knock out exon 8 (Trim46 ΔE8/ΔE8 or E8-KO) in ES cells so that the derived neurons would constitutively express only the E8-skipped isoform, abolishing developmental use of the E8+ variant and NMD regulation (Fig. 6a). Targeted exon 8 deletion in ES cells was verified by PCR genotyping (Fig. 6b) and Sanger sequencing (Fig. 6c).
To investigate the effect of exon 8 KO on TRIM46 expression, we collected protein lysates from WT and E8-KO ESC-derived neurons on DIV 3 and DIV 5. Western blot analysis showed TRIM46L proteins in E8-KO were substantially upregulated at both stages (Figs. 6d-6g). This is consistent with the prediction that exon 8 deletion de-represses TRIM46 protein expression. Despite upregulation of TRIM46L proteins, the TRIM46S protein isoform was again undetectable.
At DIV 1 and DIV 7 there was no obvious difference in total Trim46 mRNA levels between control and E8-KO mutant neurons (Fig. 6h). The general lack of mRNA expression differences between control and exon 8-KO neurons supports that transcriptional regulation alone is not responsible for the change in TRIM46 protein expression in mutant neurons. However, we note on EB D8 there was significant mRNA upregulation in E8-KO neurons compared to control. This is because in WT neurons exon 8 is largely included at EB day 8 leading to NMD of Trim46 transcripts. Exon 8 deletion prevents NMD of Trim46 transcripts, effectively increasing its steady-state abundance. Therefore, this data confirms the significant role of exon 8 splicing and NMD in regulating Trim46 transcript abundance.
To understand the functional consequence of precocious TRIM46 induction by exon 8 deletion, we stained WT and E8-KO ESC-derived neurons on DIV 3 for TRIM46 and AnkG (Fig. 6i). We found that the TRIM46 length at the proximal axon in E8-KO neurons was significantly longer than in WT (14.7 μm vs. 9.1 μm, Fig. 6j). In addition, the normalized mean TRIM46 intensity was significantly higher in E8-KO than in WT neurons, showing more TRIM46 proteins clustered at the proximal axon (Fig. 6k). This also indicates that the TRIM46 amount localized to the future axon is correlated with total TRIM46 expression.
We further analyzed whether neuronal cell types or neuronal polarity are affected in E8-KO. We found loss of exon 8 influenced neither Tau1 distribution nor polarization of neurons ( Supplementary Fig. 10a and 10b). Axon numbers in E8-KO neurons were likewise unaffected ( Supplementary Fig. 10c). We found no differences in population composition between mutant E8-KO ESC derived-neurons (97% glutamatergic and 3% GABAergic, Supplementary Fig. 10d and 10e). Together, the data show neuronal polarity and identity do not contribute to the AnkG phenotype in E8-KO neurons.
Relationship between exon 8 and exon 10 splicing. Since temporal splicing of exons 8 and 10 is concurrent during axon formation, we investigated whether regulation of the two exons was interconnected. If splicing of exons 8 and 10 are independent, neurons can express four Trim46 isoforms (E8+E10−, E8+E10+, E8−E10−, E8−E10+). If one exon splicing fully determines the other, neurons would express only two of these four isoforms (E8+E10− and E8−E10+), since exon 8 inclusion is correlated with exon 10 skipping (Figs. 1f and 5b). To distinguish these two possibilities, we designed primers to amplify regions from exon 7 to exon 11. As shown in Supplementary  Fig. 11a, we observed all four isoforms in all developmental stages.
To directly test causality, we asked whether exon 8 knockout affects exon 10 splicing, and vice versa. We found exon 8 inclusion did not change in exon 10-KO neurons in comparison to control ES-derived neurons during differentiation (Supplementary Fig. 11c and 11d). Similarly, exon 10 inclusion was not significantly affected in E8-KO neurons throughout neuronal differentiation from EB stage to maturing neurons at DIV 7 ( Supplementary Fig. 11e and 11f). We also note that in PTBP2-cKO brains, exon 10 splicing changes (Figs. 1g-1i), but exon 8 splicing does not ( Supplementary Fig. 12a and 12b), showing that exon 10 splicing does not affect exon 8 splicing. Therefore, exon 10 splicing itself is not a determinant of exon 8 splicing, and vice versa. Our data suggest that exons 8 and 10 splicing are likely coordinated by the differentiation signal upstream of their transregulators. The signal (either intrinsic or extrinsic or both) influences two different (possibly overlapping) sets of trans regulators that control exon 8 and exon 10 splicing.
Exon 8 and Exon 10 splicing enforces tissue-specific TRIM46 expression. Since TRIM46 is a significant determinant of axon formation and only neuronal tissues can generate axons, TRIM46 expression may be subject to stringent tissue-dependent regulations. We therefore examined RNA-seq data of adult mouse tissues generated by the ENCODE project. The steady-state Trim46 mRNA levels are substantially higher in the brain tissues than non-neural tissues (mean TPM: 68.5 vs 8.0; Fig. 7a). Notably, this difference does not solely reflect differential transcriptional regulation, given NMD targeting of Trim46 exon 8+ isoforms. The exon 8+ isoform was roughly 7.7-fold more efficiently degraded than the exon 8− isoform based on Upf2-cKO analyses (Figs. 5e and 5f). The more inclusion of exon 8, the more underestimation of Trim46 transcription activity using steady-state mRNA levels. Indeed, exon 8 is overwhelmingly included in non-neural tissues even at the steady-state level (mean PSI: 15; Fig. 7b). After excluding the NMD effect, the mean Trim46 TPM levels were estimated to be 288 in the brain tissues and 56 in the non-neural tissues, indicating about a 5-fold difference in transcriptional outputs.
The steady-state Trim46 mRNA levels do not precisely forecast the expression output of functional TRIM46 proteins. Under AS-NMD control, Trim46 exon 8+ isoforms are degraded without productive translation; only the exon 8 skipping isoform is translationally productive. The expression difference between brain and non-brain tissues of the exon 8− isoform is substantially larger (mean TPM: 35.6 vs 0.9, or about 40-fold; Fig. 7c) than that of the total Trim46 mRNA level. Besides exon 8, splicing of exon 10 also influences the final output of functional TRIM46 proteins, assuming TRIM46S remains unstable in nonneural tissues. Indeed, exon 10 is differentially spliced between neural and non-neural tissues (mean PSI: 96 vs 27, Fig. 7d). TRIM46L expression measured by the exon 10+ isoform is substantially larger in the brain than in non-neural tissues (mean TPM: 66 vs 1.5, or 44-fold; Fig. 7e).
Therefore, while estimated Trim46 transcription exhibits on average 5-fold difference (up to 8-fold by steady-state mRNA levels) between the brain and non-neural tissues, exon 8 and exon 10 splicing amplifies the difference of effective expression outputs to 40-and 44-fold, respectively. To further quantify this divergent expression, we calculated index τ, a robust metric of tissue specificity, where 1 means expression in a single tissue, and 0 means equal expression among all tissues 73 . The τ values for the estimated transcriptional output, steady-state mRNA level, exon 10+ isoform, and exon 8− isoform are 0.38, 0.56, 0.80, and 0.84, respectively (Fig. 7f). In summary, exon 8 and exon 10 splicing, as well as NMD control, are essential parts of the regulatory mechanisms enforcing brain-specific expression of TRIM46.

Discussion
Here, we report the gene regulatory mechanisms controlling temporal and tissue-specific expression of TRIM46, one of the earliest proteins localized to prospective axons and postulated to be a determinant of axonal fate. We found the TRIM46 protein is undetectable prior to axon formation and is induced at the onset of axonogenesis. Surprisingly, Trim46 mRNA is well transcribed before axon formation. The developmental up-regulation of the TRIM46 protein is orchestrated by alternative splicing of exons 8 and 10. Exon 8 splicing determines mRNA stability of Trim46 transcripts, whereas exon 10 splicing influences protein stability of TRIM46 protein variants.
Before axon formation, Trim46 pre-mRNA is alternatively spliced to generate isoforms that restrain the expression of Around the time when NPCs differentiate into post-mitotic neurons, these restraints are removed to allow the TRIM46 protein to accumulate due to splicing regulation of exons 8 and 10. We found exon 10 is repressed by PTBP2 during early neuronal differentiation but exon 8 is not, meaning they are separately regulated. Therefore, a multi-layered control, encompassing two independently-regulated alternative splicing events, NMD, and protein stability regulation coordinates expression of TRIM46 proteins to promote axon specification.
To what extent and which AS-NMD controls have biological function remains mostly unknown. While AS-NMD regulation appears pervasive and has gained appreciation for its role in finetuning gene expression, genetic testing for the necessity of a specific AS-NMD control is rarely reported. As NMD isoforms are often minor isoforms, most AS-NMD controls only modestly affect gene output 74 . We generated exon 8 knockout cells and demonstrated the biological significance of exon 8 AS-NMD regulation. As exon 8 does not participate in encoding a functional TRIM46 protein, the impact of its splicing on the TRIM46 protein level further showcases the functional significance of RNA processing and control elements outside of protein-coding sequences.
Alternative splicing coupled with differential stability control of protein isoforms represents another mechanism of adjusting gene output through alternative splicing. To our knowledge, no precedent other than Trim46 has been reported for this mechanism. Our study comprehensively demonstrated that endogenous TRIM46S proteins encoded by the E10− isoform are unstable compared to the E10+ TRIM46L proteins. First, two authenticated TRIM46 antibodies detected no endogenous TRIM46S proteins in total lysates or insoluble fractions of WT and E10-KO neurons. If TRIM46S and TRIM46L had similar stability, E10-KO neurons would express TRIM46S at the same level as WT neurons express TRIM46L. Second, enrichment through immunoprecipitation was still unable to detect TRIM46S proteins. Third, endogenous FLAG-tagged TRIM46 proteins were detected by the anti-FLAG antibody for the long isoform but not the short isoform. Fourth, TRIM46S protein had a substantially lower half-life than TRIM46L. Fifth, TRIM46S was more up-regulated than TRIM46L upon inhibition of the proteasome degradation pathway. Therefore, the regulation of exon 10 splicing mainly controls the total output of TRIM46 expression rather than generating proteins of varied functions.
Trim46 would be the first example where two alternative exons coupled with two different downstream mechanisms converge to influence the total gene output. This integrated regulation is effective not only during axon formation but also for neuralspecific expression of TRIM46 proteins. The difference in expression output between neural and non-neural tissues is substantially augmented by splicing of either exon (Fig. 7) and may be further enhanced by compounding their splicing effects. Absolute exon 8 inclusion or exon 10 exclusion can result in natural "knockout" of the Trim46 gene product. However, alternative splicing outcomes are rarely all or none (i.e., PSI equal to 100 or 0), as seen for exon 8 and exon 10. They appear as double safeguards constituting a stringent mechanism to assure neuralspecific expression of TRIM46. The ramification of TRIM46 expression in non-neural tissues is unclear but may relate to unique functions of TRIM46 for axonogenesis.
The exon 8 and exon 10 knockout alleles shed new light on TRIM46's role in axon specification. Exon 8 knockout precociously induces TRIM46 expression and consequently TRIM46 localization to future AIS sites. The increase in TRIM46 length at the AIS site lengthens the AnkG clustering along the axon, suggesting TRIM46 plays an instructive role for AnkG localization and AIS formation. Interestingly, the increase in the mean TRIM46 intensity at the AIS site did not change the mean AnkG intensity, indicating that TRIM46 functions to qualitatively mark the AIS location (and boundary) during the seeding step instead of to quantitatively determine the AnkG amount to be recruited to the site.
Exon 10 knockout is essentially Trim46 null, expressing little functional TRIM46 proteins. Consistent with the instructive role of TRIM46, the percentage of AnkG+ neurons dropped substantially in the absence of TRIM46. However, about half of neurons still retained AnkG+ at the AIS with normal AnkG length and intensity. Therefore, TRIM46 is not always required for AnkG localization or clustering. This insight implies the need to search for additional TRIM46-like proteins that could compensate for the loss of TRIM46 proteins. The result is also consistent with the notion that TRIM46 and the yet-to-be-identified TRIM46-like proteins largely function to seed the AIS site rather than stoichiometrically determine AnkG density at the site. The overall reduction of the seeding proteins reduces the probability of marking an AIS site. However, once a site is tagged for AIS, the amount of AnkG recruited to the site is independent of the level of seeding proteins at that site. The exon 10 knockout phenotype appears less pronounced than previously reported Trim46 knockdown neurons, which could be due to the timing and kinetics of TRIM46 deletion. van Beuningen et al.'s isolated hippocampal cultures and cortical cultures from E18 rat brains, and transfected the cultures with shRNA at DIV 1 and cortical cultures at DIV 4 4 . Because shRNA typically takes 2 or more days to deplete proteins (depending on shRNA efficiency and protein stability), it is likely TRIM46 is already well expressed and assumes functions of controlling AIS formation before being knocked down. In other words, van Beuningen's method acutely cuts off the supply of TRIM46 in the midst of exerting its cellular function. In E10-KO neurons, TRIM46 is not detectable at any time. Although this may sound worse than the knockdown, the absence may trigger compensatory mechanisms earlier, which provides neurons more time to adjust to the loss of TRIM46, e.g., by inducing the expression and accumulation of TRIM46-like molecules earlier. Given that TRIM46 is not absolutely required for AnkG accumulation, functionally redundant proteins likely exist to compensate in the absence of TRIM46. Second, RNAi-mediated knockdown is well known to have secondary effects including off-targets, which could worsen the phenotype or sensitize neurons to the depletion of TRIM46.
Accumulating evidence has demonstrated the pivotal role of alternative splicing control for mammalian axonogenesis 28 . Programming a large set of neural-specific alternative splicing events accompanies early axon formation 32 . Many of these are enriched in genes encoding cytoskeleton-associated proteins. These proteins presumably modify their enzymatic activities or their interactions with cytoskeletons through alternative splicing to regulate axon formation, as exemplified by Shtn1 and Cdc42 that control axonal growth 75,76 . Many AIS genes, including AnkG, are subject to RBFOX-mediated alternative splicing control during AIS formation 33 . Trim46 exons were not reported to alter splicing in the Rbfox triple knockout. We found Trim46 exon 10 is repressed by PTBP2. Notably, the PTBP2-mediated splicing program generally precedes the RBFOX-mediated splicing program 33 . Accordingly, TRIM46 expression and localization at AIS precedes the expression and localization of AnkG. These results strengthen the notion that developmental splicing programming is an integrated part of a larger gene expression regulatory network that coordinates axonogenesis. Future work should continue to elucidate the functional significance of these alternative splicing events for axon biology.
RNA extraction, cDNA synthesis, RT-qPCR, and QIAxcel quantitative capillary electrophoresis analysis. Total RNA from mouse embryonic and postnatal neocortices or N2a samples were extracted using TRIzol reagent (Life Technologies, 15596-018) according to the manufacturer's instructions. Isolated RNA was treated with 4 units of Turbo DNase I (Ambion, AM2239) at 37°C for 35 min to degrade any remaining DNA before cDNA synthesis. After the DNase treatment, RNA was purified using phenol-chloroform (pH 4.5, 97064-744). RNA concentrations were measured using Nanodrop 2000c (Thermo Fisher). RNA conversion to first-strand cDNAs were synthesized in 20  Plasmid Constructs. For this study, pCAGIG and pEGFP-C1 were used as base vectors. Briefly, the desired Trim46L or Trim46S isoform sequences were inserted into pCAGIG or pEGFP-C1 vectors using SacI and ApaI sites or NotI sites, respectively. Plasmids were propagated in DH5-α E. coli and prepared using Miniprep kits (Qiagen). Prior to experimentation, constructs were sequenced to ensure no mutations were introduced during the cloning process.
CRISPR/Cas9 generation of knock-out and knock-in ES cell lines. Mouse 46 C cells are transfected with Lipofectamine 2000 (Invitrogen) at a density of 15,000 in a 1.5 ml Eppendorf tube. For knock-out mutants, two pX330 plasmids coexpressing Cas9 enzyme and sgRNA at 0.2 µg each and 0.1 µg GFP reporter plasmid (CAG-GFP) was used to ensure exon deletion. For knock-in mutant, a single sgRNA at 0.4 µg, 1 µg of ssODN, and 0.1 µg CAG-GFP were used. Six hours post-transfection, cells are transferred to a 10-cm 0.1% gelatin-coated dish for small edited colonies to form. After 3-4 days, when colonies derived from gene editing are sufficiently large, GFP+ colonies are visualized by using fluorescent stereomicroscope to pick~100 cell-sized colonies, treated with TrypLE Express for 3 min to dissociate into single cells, and diluted to 1 cell per well in a 96 multi-well plate to generate and screen for single-cell derived mutant colonies. Around 6-8 days later, genomic DNA is isolated and PCR genotyping along with Sanger sequencing (Eton Bioscience) is performed to confirm targeted exon deletion or knock-in of FLAG sequence in the cell line and frozen down for storage and downstream experiments.
Quantitative imaging analysis. Images for ESC-derived neurons were acquired on the Nikon Eclipse Ci microscope at 40x magnification and were analyzed with NIS-Elements BR4.5 software. Singular neurons within the field of view (FOV) with uninterrupted and non-overlapping processes are quantified. Clustered or interconnected neurons are ignored. TRIM46 and AnkG quantification are based on their respective fluorescence intensity profiles. Neurons with no detectable TRIM46 or AnkG immunostaining at the proximal axon or AIS were categorized as TRIM46 negative (TRIM46−) or AnkG negative (AnkG−), respectively. For WT and E8-KO AnkG analysis, only TRIM46+ neurons were included in the calculation of % AnkG+ neurons. To analyze for TRIM46 specific effect on AnkG, TRIM46− (negative) cells are not included in the count. For quantification of % AnkG-positive neurons, all cells were manually counted and tallied in Excel. The same set of neurons used for measuring % AnkG+ neurons was also used to measure AnkG length and expression. To measure AnkG length, neurons must have clearly identifiable start and endpoints. The start point is the region with significant fluorescent intensity detected away from the cell body and must be continuous along the projection; the endpoint is weakened signal to the point where it could no longer be discerned from background level. AnkG length was measured by using the line tool from the NIS-Elements software and by carefully tracing along the projection of the AnkG+ neuron. For AnkG intensity, an outline for region of interest (ROI) is used to trace the proximal axon or AIS region for measurement. The ROI fluorescent AnkG intensity is subtracted from background to get the corrected value and this corrected value is used as normalized mean AnkG intensity. TRIM46 length and normalized mean intensity for WT and E8-KO neurons were measured and quantified the same as AnkG. For E10-KO analysis, AnkG length and intensity were measured under the same procedure. For vGLUT1 and GAD67 analysis, only DAPI+ neurons were included in calculation. vGLUT1+ neurons exhibited easily identifiable and distinct signal whereas GAD67+ neurons must be non-vGLUT1+ and show specific and elevated staining above background levels.
Axonal polarity and axon number. For Tau1-based polarity analysis, the mean dendritic intensity (I d ) in dendritic regions and mean axonal intensity (I a ) in axonal regions was used to calculate the axonal polarity index P ¼ IaÀId IaþId . P > 0 means polarized axonal distribution; P < 0 means polarized dendritic distribution; P = 0 (I a = I d ) means uniform distribution. The number of axons were defined based on Tau1 staining. ESC-neurons lacking specific Tau1 staining in the distal axon were categorized as 0 axon whereas single axon (1) and double axons (2) were identified by distinct Tau1 signal in projections.
Immunoprecipitation. To detect endogenous TRIM46 or FLAG-TRIM46 protein isoforms by immunoprecipitation, E12.5 brain, P10 cortices, or ESC-derived neuron lysate are collected with IP buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.10% Triton X-100, 1 mM PMSF, 1x protease inhibitor cocktail (Roche), 100 U/ml Turbonuclease (Sigma)). Lysate is cleared through centrifugation at 18,800 g for 15 min in 4°C. Protein A magnetic beads (Biorad, 1614013) or FLAG magnetic beads (Sigma-Aldrich, M8823-5ML) are washed three times with IP buffer. After washing, beads are incubated at 20 rpm at RT with either rabbit anti-TRIM46 antibody (Proteintech, 21026-1-AP) or rabbit anti-IgG control antibody (Proteintech, 30000-0-AP) for 20 min to form bead-antibody complexes. For FLAG beads, no incubation is needed. After the bead-antibody incubation step, complexes are washed three times (same as lysis buffer) to remove unbound antibodies. After centrifugation, cleared lysate is split equally half and half to control IgG-IP and TRIM46-IP. Incubation is performed at 25 rpm in 4°C for 4 h with the first flowthrough (FT) for FLAG-IP is collected and beads are washed three times. Bead bound proteins were denatured by boiling in 2x Laemmli buffer with β−ME at 95°C for 10 min. Using a magnetic plate, immunoprecipitated proteins are separated from beads and ready for western blot analysis. The IP fractions were resolved by SDS-PAGE, transferred to Immobilon-FL PVDF membrane, and probed with primary antibodies that include polyclonal rabbit anti-TRIM46 (Proteintech, 21026-1-AP, 1:1000) or polyclonal rabbit anti-FLAG (Sigma-Aldrich, F7425, 1:2000). Primary antibodies were detected by appropriate Alexa Fluor conjugated secondary antibodies (Thermo Fisher; A21206, A10042, A31573, 1:2000) and blots were visualized using Typhoon FLA 9000.
Mouse care. Mice were maintained in accordance with the guidelines of the Institutional Animal Care and Use Committees at the University of California, Riverside (UCR). Under routine monitoring and checking by veterinary and mouse room staff, all animals used in this study were kept healthy in a mouse room with a stable temperature at 22 ± 2°C and in a 12 h light/dark cycle. Mice were maintained in ventilated cages and clean bedding was provided every two weeks. Food and water were provided ad libitum. No more than 5 mice were allowed in each cage. Mice of both sexes were used. The Ptbp2 loxp/loxp mice 32 and Upf2 loxp/loxp mice 40,55 used in this study were previously described. Briefly, Ptbp2 conditional KO mice (Ptbp2-cKO) were generated by breeding Ptbp2 loxp/loxp with Ptbp2; loxp/+ Emx-Cre. Additionally, Upf2 conditional KO mice (Upf2-cKO) were generated by breeding Upf2 loxp/loxp with Upf2; loxp/+ Emx-Cre mice. All animal procedures were approved by the UCR Institutional Animal Care and Use Committee.
Mouse primary neuron culture. Prior descriptions of culturing primary cortical neurons have been reported 32,40 . Mouse neocortices were dissected at targeted E13.5 stage and dissociated in 0.25% trypsin (Thermo Scientific) comprising of 1 mM EDTA (Sigma) in 1x HBSS (Invitrogen) at 37°C for approximately 8 to 10 min. Subsequently, dissociated neurons are plated in fresh media containing 25 mM glucose (Sigma), 1x GlutaMAX (Thermo), and 20% heat-inactivated horse serum (Thermo) in MEM (Invitrogen). After 6 h, plating media is replaced with neurobasal-based feeding media consisting of 1x B-27 (Invitrogen) and 1x Gluta-MAX. For CHX treatment (Fisher Scientific), neurons one day later (E13.5-DIV 1) were treated at final concentration of 0.5 mg/ml for 5 h and collected in TRizol for RNA extraction.
RNA-Seq data analysis. Mouse adult tissue RNA-seq data (generated by Thomas Gingera's lab) was downloaded from ENCODE. We mapped the RNA-seq data using STAR based on the mouse GENCODE annotation (mmv20). We obtained exon junction reads for Trim46 exon 8 and exon 10 and calculated inclusion ratio (PSI) as the following: 0.5 x inclusive junction reads / (0.5 x inclusive junction reads + exclusive junction reads) x 100. Exclusion ratio (%) is defined as 100−PSI. Exon 8− isoform expression level is determined as steady-state expression TPM x exon 8 exclusion ratio. Exon 10+ isoform expression level is determined as steady-state expression TPM x exon 10 inclusion ratio. The Gingera dataset contains two biological replicates for each tissue. PSI and TPM values are presented as the average of two biological replicates. Tissue specificity index τ ¼ n is the number of tissues and x i is the expression of the gene in tissue i, x i ¼ log 2 ðTPM i þ 1Þ. We took the average of log 2 ðTPM i þ 1Þ values from three brain tissues to obtain one value for the brain tissue.
Statistics. All statistical analyses were performed using Microsoft Excel with Student's t-test under paired or unpaired parameters and under two-tailed or onetailed conditions as indicated with the exception of half-life calculation performed using GraphPad Prism 8 software.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.