Introduction

Alternative splicing greatly expands proteome diversity and can impact brain development and disease (Vuong et al., 2016). As most metazoan proteins are represented by multiple isoforms, defining the functional significance of individual splicing events is a vast challenge. Mouse studies do not have the throughput to assess the neural role of 10s or 100s of isoforms, and cell culture does not recapitulate the complex development of an animal brain.

One class of particularly conserved and developmentally regulated spliced exons is microexons or mini-exons. These small exons, defined here as 3-30 nucleotides or 1-10 amino acids in length, are either gained or lost in transcripts in the brain compared to other tissues. While most are in-frame, this pathway can also trigger nonsense-mediated decay to influence protein levels (Lin et al., 2020). Although these neuron-specific exons were recognized almost 40 years ago (Brugge et al., 1985; Martinez et al., 1987), their discovery and annotation have continued into recent years (Parada et al., 2021). Thus far, hundreds have been identified. Although more than one splicing regulator has been implicated in the alternative inclusion of microexons in neurons (Ciampi et al., 2022; Y. I. Li et al., 2015), many are controlled by the vertebrate-specific protein SRRM4 (Irimia et al., 2014; Quesnel-Vallières et al., 2015). Srrm4 interacts with Srsf11 and Rnps1 at polypyrimidine sequences upstream of microexons to regulate their inclusion (Gonatopoulos-Pournatzis et al., 2018).

Proteins that regulate and contain microexons are involved in neurodevelopmental disorders. In the brains of individuals with autism spectrum disorder (ASD), the level of SRRM4 is reduced, leading to the misregulated splicing of multiple microexons (Irimia et al., 2014). Mice haploinsufficient for Srrm4 display reduced social interaction, sensory hypersensitivity, and impaired synaptic transmission (Quesnel-Vallières et al., 2015). Furthermore, Srrm4 is involved in regulating the ASD-relevant KCC2-dependent GABAergic excitatory to inhibitory switch (Nakano et al., 2019), and a mouse model of ASD with disrupted Pten has decreased Srrm4 expression and microexon inclusion (Thacker et al., 2020). Loss-of-function variants in its interacting partner SRSF11 have been implicated in ASD in humans (C Yuen et al., 2017). While the behavioral roles of only a few microexons have been studied in detail (Gonatopoulos-Pournatzis and Blencowe, 2020), the elimination of one from Eif4g1 altered synaptic plasticity and social behavior in mice (Gonatopoulos-Pournatzis et al., 2020). Many genes containing microexons also lead to neurodevelopmental disorders when mutated in humans. Examples include CASK-mediated microcephaly and cerebellar hypoplasia, SPTAN1-mediated childhood-onset epilepsy, and MAPK8IP3-mediated intellectual disability (Giacomini et al., 2021; Platzer et al., 2019; Syrbe et al., 2017). Taken together, this splicing program, the genes with microexons, and microexons themselves are important to enable proper neural development.

Although most microexons are unstudied, detailed structural and functional characterization of a small subset has revealed involvement in synapse development. The most well-studied are found in the receptor-type tyrosine phosphatase (e.g., ptprf) and teneurin (e.g., tenm4) gene families. Recognized in the 1990s (O’Grady et al., 1994; Pulido et al., 1995), two microexons in presynaptic PTPδ (Ptprd), PTPσ (Ptprs), and LAR (Ptprf) modulate trans-synaptic interfaces and determine selective interaction with postsynaptic partners (Lin et al., 2018; Y. Li et al., 2015; Park et al., 2020; Um et al., 2014; Yamagata et al., 2015; Yim et al., 2013). Similarly, differential inclusion of two conserved microexons in teneurins can modulate trans-synaptic homophilic or heterophilic interactions through splicing-dependent, large conformational changes (Berns et al., 2018; Gogou et al., 2024; Li et al., 2020). While studying these cell adhesion molecules has highlighted how microexons can trigger differential protein-protein interactions, little is known about the developmental consequences of these biochemical switches in other protein classes.

Identifying the neurobiological roles of many individual microexons is a challenge. Studies of splicing regulators such as srrm4 impact the entire splicing program, making it impossible to determine the importance of individual microexons to protein function. Furthermore, microexons could still be differentially included in a srrm4 regulatory mutant via compensation by other splicing factors, such as its paralogue srrm3, which is responsible for microexon inclusion in photoreceptor transcripts in zebrafish (Ciampi et al., 2022). A vertebrate model is necessary for these studies; although there is a similar splicing mechanism in Drosophila melanogaster, there is no overlap of the individual microexons (Torres-Mendez et al., 2021). Thus, we used larval zebrafish, a vertebrate that supports more high-throughput studies than mammalian models (Thyme et al., 2019), to assess the brain activity, brain structure, and behavior of zebrafish larvae mutant for 45 microexons (Figure 1A).

Figure 1.

Results

A recently developed computational tool identified microexons that are differentially spliced during mouse brain development and conserved in zebrafish (Parada et al., 2021). Based on the intersection of these two sets, 95 microexons fulfilled both parameters (Table S1). Of these 95 microexons, 42 exist in a canonical layout in the zebrafish genome, with both a UGC and UC repeat – or similar polypyrimidine tract – directly upstream of the alternatively spliced exon (Gonatopoulos-Pournatzis et al., 2018) (Table S1, Figure S1), indicating that Srrm4 likely controls their inclusion. Of the remaining microexons, 44 are organized similarly to the canonical layout, typically with either a UGC or UC repeat. Thus, they may also be regulated by Srrm4. The protein sequences of the 95 microexons are highly conserved between mouse and zebrafish (Figure 1B), more so than the full-length protein sequences (Figure 1C). The genes containing microexons are enriched for those involved in neuronal development as well as the cytoskeleton (Figure 1D). The majority of these microexons are included in transcripts as the brain develops, rather than lost, in both mouse and zebrafish (Figure 1E), making CRISPR/Cas9 removal of the microexon an ideal approach for studying the function of the added amino acids.

Using CRISPR/Cas9, we generated lines that removed 45 conserved microexons (Table S2) and assayed larval brain activity, brain structure, and behavior (Figure 1A). Four guide RNAs were used, two on each side of the microexon (Table S2, Figure S1). For microexons with upstream regulatory elements that are likely important for splicing, these elements were also removed (Figure S1). While we were most interested in those microexons only present after 24 hours post-fertilization (Figure 1E), as this is when the brain is forming, we also removed microexons from several genes with an earlier expression of the isoform. For eight mutant lines, we confirmed that the microexon was eliminated from the transcripts as expected (Figure S2). Although our genomic deletion did not yield unexpected isoforms, qRT-PCR on these eight lines revealed significant downregulation for the homozygous vav2 mutant (Figure S2), indicating possibly complex genetic regulation. Protein-truncating mutations in eleven additional genes that contain microexons revealed developmental and neural phenotypes in zebrafish (Figure S3, Figure S4), indicating that the genes themselves are involved in biologically relevant pathways. Three of these genes – tenm4, sptan1, and ppp6r3 – are also in our microexon line collection.

To uncover neural phenotypes, we assessed baseline and stimulus-driven movement. Homozygous mutant and control sibling larvae were subjected to a behavioral pipeline in 96-well plates from 4-7 days post-fertilization (dpf) (Figure 2A). Only a small number of strong, repeatable phenotypes were observed either at baseline or in response to acoustic or visual stimuli (Figure 2B, Figure 2C, Figure S5, Figure S6, Figure S7). Baseline behavioral differences were observed in multiple movement categories. For example, rapgef2 mutants showed an increased movement frequency (Figure 2B, Figure 2D) but no changes to the characteristics of these movements (magnitude) or location in the well. In contrast, dop1a mutants had reduced daytime movement and an increased well-edge preference specifically at night (Figure S6, Figure S8, Figure 2B, Figure 2D). Other baseline movement phenotypes included a well-center preference for vav2 mutants, a decreased movement magnitude for ptprd-2 (also referred to in the literature as meA) mutants (Figure S8), and an increased movement magnitude for ptprd-1 (also referred to in the literature as meB) mutants (Figure 2D). Dark-flash response phenotypes included a reduced latency for eif4g3b mutants and a reduced response frequency for ptprd-1 (meB) and dop1a mutants (Figure 2C, Figure 2D, Figure S8). Acoustic response phenotypes included an increased frequency of response to strong stimuli for ppp6r3 mutants and a reduced frequency of response to strong acoustic stimuli for ptprd-2.

Figure 2.

To detect changes to brain development and function, we collected phospho-Erk (pErk) brain activity maps under unstimulated conditions at 6 dpf. Large morphological differences were identified using the Jacobian matrix derived from the image registration process (Jefferis et al., 2007; Rohlfing and Maurer, 2003). Similar to the behavioral findings, few phenotypes were observed in brain activity (Figure 3A, Figure S9, Figure S10) or brain structure (Figure 3B, Figure S11, Figure S12) in homozygous mutants compared to wild-type control siblings when quantified by region (Figure 3C). Repeatable phenotypes were mild (Figure 3D) compared to a mutant with a moderate phenotype and both increased and decreased activity and structure (see control-tcf7l2) (Capps et al., 2024). Only one mutant, ppp6r3, displayed substantial and repeatable differences in brain morphology (Figure 3B, Figure 3E), with reduced size observed mainly in the thalamus and neighboring pretectum. While the loss of both copies of microexons was required for most phenotypes, brain activity differences were observed in vti1a and sptan1 heterozygotes (Figure 3F) as well as vav2 and mapk8ip3 (Figure S9, Figure S10). While repeatable phenotypes in brain activity and structure were observed in a small subset of mutants, other zebrafish studies using identical methods have typically uncovered brain activity phenotypes for approximately half of the lines that are stronger than any we observed (Figure 3G) (Capps et al., 2024; Thyme et al., 2019; Weinschutz Mendes et al., 2023).

Figure 3.

Discussion

Our study provides a broad overview of the larval zebrafish neural phenotypes resulting from removing developmentally regulated microexons. Here, we leveraged our previously successful brain activity mapping and behavioral profiling strategy, which has high sensitivity and has uncovered phenotypes for dozens of mutants (Thyme et al., 2019). While only a few zebrafish lines had neural phenotypes, most of the microexons with behavioral or brain activity differences were previously unstudied and found in diverse protein classes.

The small number of observed phenotypes was unexpected based on conservation of the microexons (Figure 1B, Figure 1C) and the discovery rates of previous screens using similar methods (Capps et al., 2024; Thyme et al., 2019; Weinschutz Mendes et al., 2023). Although most of these studies focused on protein-truncating alleles, we have found phenotypes in lines with missense mutations (Capps et al., 2024). However, this outcome is consistent with a recent, similar study of microexons in zebrafish (Lopez-Blanch et al., 2024). Although only three of the 18 microexons this group studied are in the same genes as our set (ap1g1, vav2, and vti1a), the phenotypes for individual mutants are similar and this work shares the same overall finding of minimal phenotypes in zebrafish larvae. Importantly, the pErk brain activity mapping method we used is highly sensitive, significantly minimizing the likelihood that substantial zebrafish larval phenotypes from individual microexon removal are being missed. In our published work (Capps et al., 2024; Thyme et al., 2019), we showed that brain activity can be drastically impacted without manifesting in differences in those behaviors assessed in a typical larval screen such as was completed by Lopez-Blanch et al.. Furthermore, the genes that contain microexons are developmentally important (Figure S3, Figure S4). For instance, the caska loss-of-function mutant displayed substantial changes to brain activity, a smaller brain size, and multiple behavioral phenotypes. Loss-of-function mutants for sptan1 and ppp6r3 are stronger than lines that only remove the microexon (Figure S3, Figure S4).

The limited effects of individual microexon deletions suggest that stronger phenotypes may require disrupting multiple microexons or examining later stages of development. It is possible that perturbation of multiple microexons simultaneously, as was observed in the brains of autistic individuals (Irimia et al., 2014), is necessary to produce pronounced effects on early neurodevelopment. Zebrafish homozygous for srrm4 deletion did not display any overt developmental phenotypes (Ciampi et al., 2022), in contrast to a previous morpholino study (Calarco et al., 2009). More recently, it was revealed that loss of srrm4 has minimal impacts on behavior (Lopez-Blanch et al., 2024) and brain structure (Gupta et al., 2025), while srrm3 mutants have significant neural phenotypes and early mortality (Ciampi et al., 2022; Lopez-Blanch et al., 2024). This finding indicates that Srrm3 loss is necessary for complete elimination of these microexons, but studying microexons in the context of this line is also complicated by possible downstream consequences of severe vision loss. Alternatively, although the microexons are present at larval stages, they may only become important during later stages of neuronal maturation that support more complex behaviors. Mouse models were studied as adults and for disruptions to social interaction, learning, and memory (Gonatopoulos-Pournatzis et al., 2020; Han et al., 2024), which larvae are not yet capable of (Dreosti et al., 2015; Valente et al., 2012).

Our study extends on findings for previously studied genes containing microexons. We discovered increased brain activity (Figure 3D) and an altered dark flash response in mutants for the eif4g3b (Figure 2D), whereas a mouse model with the microexon removed was described as having no behavior phenotypes (Gonatopoulos-Pournatzis et al., 2020). That study, however, focused on social behavior, learning, and memory and may not have included stimuli analogous to dark flashes. Alternatively, microexon removal in zebrafish could have differential behavioral impacts than in rodents. Mice mutant for the ptprd-2 (meA) microexon have increased movement and interrupted sleep (Park et al., 2020). Both zebrafish mutants in ptprd demonstrate the opposite, with reduced daytime movement and increased sleep during the day for ptprd-2 (Figure S8). While the underlying disrupted circuitry could differ between the species, the daytime sleep increase raises the possibility of a rebound response to undetected nighttime sleep disruptions. In other cases, the gene that contains the microexon has been extensively studied, but research on the microexon itself is absent. For example, the function of SNARE protein Vti1a in neural development and dense core vesicle biogenesis has been studied in great depth (Bollmann et al., 2022; Emperador-Melero et al., 2018; Kunwar et al., 2011; Sokpor et al., 2021; Walter et al., 2014). Its brain-specific microexon, however, was recognized over twenty years ago (Antonin et al., 2000), and its specific importance has been largely ignored. We discovered reduced brain activity mainly in the telencephalon in both homozygous and heterozygous lines compared to wild-type siblings (Figure 3F), highlighting the relevance of considering multiple conserved isoforms when characterizing protein function in the brain.

Mutants for unstudied microexons with the strongest phenotypes come from diverse protein classes (Gonatopoulos-Pournatzis and Blencowe, 2020), indicating this splicing program impacts neurodevelopmental processes beyond trans-synaptic partner selection. The only mutant with a substantial brain size difference was ppp6r3, which encodes a regulator of protein phosphatase 6 (PP6), and this reduced size was present and more severe in the predicted protein-truncating ppp6r3^sa16892 line (Figure S3). Although little is known about this protein, it has been linked to cell cycle (Heo et al., 2020) and cell death (Wu et al., 2022), nominating hypotheses for determining future research on the source of the reduced size. Dctn4 is part of the dynactin complex, Mapk8ip3 is involved in Kinesin-1-Dependent axonal trafficking (Platzer et al., 2019; Watt et al., 2015), and Sptan1 is a cytoskeleton scaffold protein (Huang et al., 2017); the number microexon-containing genes related to the cytoskeleton (Figure 1D) and also trafficking (e.g., vti1a), suggests that microexons may be important to the specialized cytoskeletal needs of developing neurons. The functions of microexons in cytoskeletal proteins are beginning to be discovered (Poliński et al., 2023). Transcriptional regulation is represented by the meaf6 mutant, while Rapgef2, which has the strongest behavioral phenotypes upon microexon removal (Figure 3D), is a signaling protein.

Although our list is not comprehensive, as many microexons remain unstudied, we have prioritized several microexons for future study from the 45 tested. For seven mutants, we confirmed that the microexon was cleanly removed without unanticipated effects on isoforms or transcript levels (Figure S2), including for some of the most interesting that are entirely unstudied (ppp6r3, sptan1, meaf6). For those with phenotypes that are not yet confirmed by qRT-PCR (rapgef2, dctn4, dop1a, mapk8ip3), studies of gene loss-of-function in zebrafish and mice reveal far stronger phenotypes and lethality, suggesting that our lines impact only the microexon (Abeler-Dörner et al., 2020; Drerup and Nechiporuk, 2013; Satyanarayana et al., 2010; Tuttle et al., 2019). In future work, proximity-dependent ligation in zebrafish lines with and without the microexon could reveal differential binding partners in vivo (Rosenthal et al., 2021), as microexons are often found in surface-accessible protein domains and can regulate protein-protein interactions (Dergai et al., 2010; Irimia et al., 2014), The discovery of microexon mutants with neural phenotypes lays the groundwork for future investigation of impacted neurodevelopmental pathways and how interactomes differ between isoforms.

Materials and Methods

Zebrafish husbandry

Zebrafish experiments were approved by the UAB Institutional Animal Care and Use Committee (IACUC protocols 22155 and 21744) and UMass Chan Institutional Animal Care and Use Committee (IACUC protocol 202300000053). Mutants were generated in an Ekkwill-based strain using CRISPR/Cas9 as previously described (Capps et al., 2024; Thyme et al., 2019) (Table S2). Both adults and larvae were maintained on a 14 h/10 h light/dark cycle at 28°C. Experimental larvae were maintained at a density of less than 160 per dish in 150 mm petri dishes in fish water with methylene blue, and debris was removed at least twice prior to experimentation. Visibly unhealthy larvae and those without inflated swim bladders were excluded from experiments. Control animals were always siblings from the same clutch and derived from a single parental pair, and all larvae were genotyped after experimentation. Even with using sibling controls and collecting multiple biological replicates from individual parents, the possibility remains that linked genetic variation may have contributed to the mild phenotypes we observed, as only a single line was generated.

Brain activity and morphology

Phosphorylated-ERK (pErk) antibody staining was conducted as previously described (Capps et al., 2024; Thyme et al., 2019). The total ERK antibody (Cell Signaling, #4696) was used at 3:1000 and pErk antibody was used at 1:500 (Cell Signaling, #4370). Typically, the primary antibody exposure was for 2-3 days and secondary for one day. Image stacks were collected with a Zeiss LSM 900 upright confocal microscope using a 20x/1.0 NA water-dipping objective. These stacks were registered to the standard zebrafish Z-Brain reference using Computational Morphometry Toolkit (CMTK) (Jefferis et al., 2007; Randlett et al., 2015; Rohlfing and Maurer, 2003; Thyme et al., 2019). The significance threshold for MapMAPPING (Randlett et al., 2015; Thyme et al., 2019) was set based on a false discovery rate (FDR) where 0.05% of control pixels would be called as significant for both the brain activity and structural measurements.

Larval behavior

Larval behavior assays were conducted and analyzed as previously described (Capps et al., 2024; Joo et al., 2020; Thyme et al., 2019). The pipeline begins on the evening of larval stage 4 dpf and includes the following stimuli: 5 dpf light flashes (9:11-9:25), mixed acoustic stimuli (prepulse, strong, and weak, from 9:38-2:59), and three blocks of acoustic habituation (3:35-6:35); the night between 5 and 6 dpf mixed acoustic stimuli (1:02-5:00) and light flashes (6:01-6:20); 6 dpf three blocks of dark flashes with an hour between them (10:00-3:00) and mixed acoustic stimuli and light flashes (4:02-6:00). Stimulus responses were quantified from high-speed (285 frames-per-second) 1-second-long movies. All code for behavior analysis is available at https://github.com/thymelab/ZebrafishBehavior. Frequency of movement, magnitude of movement, and location preferences were calculated for the baseline data. An example of a “frequency” measure would be the active seconds / hour. An example of the “magnitude” measure would be the movement velocity of a bout. An example of a “location” measure would be the fraction of the bout time spent in the center of the well. Frequency, response magnitude parameters, and latency to response are also calculated for each stimulus response from the high-speed movies. The stimulus responses are also broken up into subsets, such as the dark flashes that occur at the beginning of the dark flash block and those that occur at the end. Multiple measures are calculated for the experiment, and both summarized and raw measures are shown in Figure 2 and the Supplementary Material. The summarized data is generated by merging related terms (e.g., all the “frequency” measures) together. The size of the bubble represents the percent of significant measurements in the summarized category, and the color represents the mean of the strictly standardized mean difference (SSMD) of the significant assays in that category. It is possible for some measures in the merged group to be increased (purple) and some decreased (orange), such as if a behavioral change occurs between 4 and 6 dpf, which is why it is possible to have two offset bubbles in each square. The scripts for merging these measurements and generating the summarized bubble plot graphs are available at https://github.com/thymelab/DownstreamAnalysis.

Reverse Transcription PCR (RT-PCR and qRT-PCR)

RNA was extracted from zebrafish embryos using the E.Z.N.A. MicroElute Total RNA Kit (Omega Bio-Tek R6834-02), and cDNA synthesis was performed with iScript Reverse Transcription Supermix (Bio-Rad #1708840). Following PCR with GoTaq polymerase (Promega M7123) and primers (Table S1), samples were run on a 4% agarose gel and the band intensities were quantified using Fiji (Schindelin et al., 2012). For qRT-PCR, we combined 3-4 zebrafish heads at 6 dpf for each biological replicate. For RT-PCR, these separate biological samples were combined into one per genotype. We homogenized the heads and performed RNA extraction and cDNA synthesis as described above. We then performed qRT-PCR with SsoAdvanced Universal SYBR Green Supermix (Bio-Rad #1725270EDU) according to the manufacturer’s protocol and using primers in Table S1.

Data and materials availability

The mutants described in the paper are available from ZIRC. Code is available from https://github.com/thymelab. Processed behavioral and imaging data is available from Zenodo under the following DOIs: 10.5281/zenodo.13137486 (behavior output files), 10.5281/zenodo.13138646 (behavior input files), and 10.5281/zenodo.13138766 (brain activity mapping stacks).

Acknowledgements

This research was supported by a Mallinckrodt Award from the Edward Mallinckrodt Jr. Foundation (SBT). Guillermo Parada was instrumental to this work by sharing his list of conserved and developmentally regulated microexons while his study was still a preprint under review. We thank the UAB fish facility staff, UMass Chan fish facility staff, the Zebrafish International Resource Center (ZIRC), particularly Andrzej Nasiadka, the Research Computing team at UAB, and the UAB Department of Neurobiology for supporting this study. We also thank the following Thyme lab members for experimental assistance or training that supported this study: Brandon Bastein, Anna Moyer, Lynne Zhou, Vaishnavi Balaji, Alexia Barcus, and Sahil Malhotra.

Additional information

Author contributions

SBT conceived of the study, analyzed imaging and behavioral data, and wrote the manuscript. VM, JW, and CLC generated the mutant lines. CCSC and KM initiated the project and CCSC collected most of the experimental data. MESC, WCG, MCK, CLC, and EGT contributed to the experimental data in this work by genotyping and propagating zebrafish lines, running larval behavior experiments, and collecting brain imaging data.

Funding

Edward Mallinckrodt Jr. Foundation

Additional files

Table S1

Table S2

Supplementary Figures S1-S12