The RNA-binding protein Caper is required for sensory neuron development in Drosophila melanogaster

Abstract

Background

Alternative splicing mediated by RNA-binding proteins (RBPs) is emerging as a fundamental mechanism for the regulation of gene expression. Alternative splicing has been shown to be a widespread phenomenon that facilitates the diversification of gene products in a tissue specific manner. Although defects in alternative splicing are rooted in many neurological disorders, only a small fraction of splicing factors have been investigated in detail.

Results

We find that the splicing factor Caper is required for the development of multiple different mechanosensory neuron subtypes at multiple life stages in Drosophila melanogaster. Disruption of Caper function causes defects in dendrite morphogenesis of larval dendrite arborization neurons, neuronal positioning of embryonic proprioceptors, as well as the development and maintenance of adult mechanosensory bristles. Additionally, we find that Caper dysfunction results in aberrant locomotor behavior in adult flies. Transcriptome-wide analyses further support a role for Caper in alternative isoform regulation of genes that function in neurogenesis.

Conclusions

Our results provide the first evidence for a fundamental and broad requirement for the highly conserved splicing factor Caper in the development and maintenance of the nervous system and provide a framework for future studies on the detailed mechanism of Caper mediated RNA regulation.

Introduction

RNA regulatory mechanisms such as alternative splicing, transcript localization, and local translation have recently emerged as critical processes in controlling gene expression during neural development. The sum of these events increases the complexity of gene products within the cell and provides multiple points of regulation to fine-tune gene expression within a neuron in both a spatial and temporal manner. Additionally, mutations in genes encoding RNA-Binding Proteins (RBPs), which mediate these RNA regulatory events, have been implicated in myriad neurological diseases including amyotrophic lateral sclerosis (ALS), spinal muscular atrophy (SMA), epilepsy, schizophrenia, and autism (Zhou et al., 2014; Vuong, Black, and Zheng, 2016; Scotti and Swanson, 2016). This highlights the importance of understanding the basic mechanisms of RNA regulation within neurons, yet a paucity of information exists about the precise molecular mechanisms in which specific RBPs participate and the identities of their target RNAs.

In addition to the high level of conservation of RBPs across metazoa, the use of sophisticated genetic techniques in highly tractable model systems, such as Drosophila and Caenorhabditis elegans, provides unique opportunities to study the roles of RBPs within the nervous system in vivo. To better understand the function of RBPs in neural morphogenesis, we recently identified a subset of conserved RBPs that are required for dendrite development in both Drosophila and C. elegans (Olesnicky et al., 2014; Antonacci et al., 2015). Among these RBPs is the splicing factor Caper, which is an RNA Recognition Motif (RRM) containing protein that has been shown to co-localize with spliceosomal proteins in human cells and regulate a large number of alternative splicing events in human breast adenocarcinoma cell lines (Huang et al., 2012; Dowhan et al., 2005; Brooks et al., 2015; Stepanyuk et al., 2016; Mai et al., 2016). A recent RNA interference (RNAi) screen of splicing factors in Drosophila S2-DRSC cells showed that Caper is a critical regulator of over 400 splicing events (Brooks et al., 2015). A detailed characterization of the protein composition of both Drosophila and human spliceosomal complexes showed that Caper/RBM39 is a component of large conserved spliceosomal RNP complexes (Herold et al., 2009). Additionally, human Caper orthologs have been shown to participate in hormone receptor mediated alternative splicing and transcriptional regulation, suggesting that Caper may function at multiple levels of regulation of gene expression (Dowhan et al., 2005).

caper orthologs in C. elegans and human cell lines have been shown to have a conserved function in energy homeostasis and as regulators of mitochondrial metabolism (Kang et al., 2015). Yet although the human caper ortholog, RBM39, is expressed in the nervous system (Liu et al., 2008; Antonacci et al., 2015), surprisingly little is known about the function or target RNAs of Caper or its orthologs within neurons. Recently, the mouse caper ortholog RBM39 was shown to interact with the RBP ZFP106, which has been implicated in neuromuscular degeneration. Nonetheless, the role of RBM39 itself in neural development and maintenance of neuromuscular junctions has not been explored (Anderson et al., 2016).

In this study, we sought to determine the extent to which Caper is required for the development of mechanosensory neurons. To this end, we analyzed the phenotypic consequences of reduced Caper function in multiple mechanosensory neurons by using a combination of RNA interference (RNAi) mediated knockdown of Caper function and a caper mutant genetic background. Additionally, we performed transcriptomic analyses to determine the effects of Caper dysfunction on gene expression during embryogenesis. The combination of these approaches confirmed a broad and fundamental role for Caper in the morphogenesis of sensory neurons within the peripheral nervous system throughout embryonic, larval and adult stages of development. While splicing defects have been implicated in many neurological disorders, only a small fraction of RBPs have been investigated for a specific role in neurogenesis. Our study lays the foundational work for the conserved splicing factor Caper as a new and fundamental player in the regulation of neurogenesis.

Results

caper mutant Class IV da neurons form supernumerary dendrites

Neurons elaborate distinct dendritic patterns to facilitate the formation of synaptic connections and the innervation of epithelial tissues by sensory neurons. Drosophila Class IV dendrite arborization (da) neurons serve as an excellent model for dendrite morphogenesis as they develop hundreds of dendritic branches to establish a sensory field for mechanosensation, as well as thermal and chemical nociception (Grueber et al., 2003; Grueber and Jan, 2004). To better understand the contribution of RBPs to dendrite morphogenesis, we previously utilized Class IV da neuron specific RNA interference (RNAi) to knock down the majority of RBPs encoded by the Drosophila genome and qualitatively screen for dendritic defects. caper was among the 63 RBP-encoding genes that resulted in dendritic defects upon RNAi-mediated knockdown (Olesnicky et al., 2014). To confirm that Caper deficient neurons undergo aberrant dendrite morphogenesis and to quantify the precise defects that result from Caper dysfunction, we examined dendrite morphology in a caper genetic mutant line. The caper^CC01391 allele was generated using transposition mutagenesis with a GFP protein trap construct (Buszczak et al., 2007). In the caper^CC01391 genetic background, the transposable element bearing the EGFP open reading frame, flanked by splice acceptor and splice donor sequences inserted within the intron residing between the first two protein coding exons. To determine if the protein trap generates a full-length fusion protein or instead results in aberrant or truncated spliceforms, we performed a Western Blot analysis using an antibody against GFP on larval and adult lysates derived from animals that were homozygous for the caper^CC01391 allele and OreR as a control. Caper:GFP products were detected at a lower molecular weight (70 kDa) than expected for the full-length fusion protein (100 kDa) in larval lysates (Fig. 1A). However, we did not detect any Caper:GFP in the adult, which we hypothesize is due to inappropriate/cryptic splicing and transcript degradation. We therefore conclude that the protein trap allele is likely a hypomorphic allele.

Figure 1. caper regulates dendrite morphogenesis of Class IV da neurons.

(A) Caper:GFP truncated protein (70 kDa) is detected in the larvae caper^CC01391. Coomassie brilliant blue-stained membrane (top panel) and immunoblot (bottom panel) of cell lysates from larvae and adult D. melanogaster caper^CC01391 and oreR control lines. While Western blot analysis of the caper^CC01391 GFP protein trap allele shows a 70 kDA truncated Caper:GFP fusion product in larval lysates, no GFP fusion protein is detected in adult lysates. Three immunoblots were performed, one representative image is shown. (B) Control Class IV ddaC neurons expressing td-Tomato driven by ppkGal4 elaborate hundreds of dendritic branches to cover their sensory field. (C) caper^CC01391 (designated by caper ^−/−) da neurons show defects in field coverage and dendrite morphogenesis. (D) Quantification of dendritic termini reveals that caper dysfunction leads to the formation of supernumerary dendritic termini, compared to controls (p=0.03). (E) However, individual branches are significantly shorter in caper^CC01391 da neurons as compared to controls (p=0.0068). (F) Thus caper^CC01391 da neurons form significantly more branch points per dendrite length but fail to cover their field, as compared to controls (p=0.0068). Scale bar = 50 μm.

To confirm a role for Caper in dendrite morphogenesis and characterize the precise dendritic defects associated with caper dysfunction, Class IV da neurons were visualized using the ppk-CD4-tdTomato reporter (Han et al., 2012) in the caper^CC01391 mutant background. Live imaging of caper^CC01391 homozygous mutant Class IV ddaC neurons confirmed that reduced Caper function results in abnormal coverage and distribution of the dendritic receptive field (Fig. 1B,C). Quantification of dendritic termini number revealed that caper deficient neurons have a statistically significant increase in the number of terminal dendritic branches, as compared to wild type (Fig. 1D). However, caper mutant neurons have significantly decreased branch length and a higher number of branch points per dendrite length compared to control neurons (Fig. 1E,F). Thus, caper is required for da neuron dendrite morphogenesis, and reduced function of Caper leads to the formation of shortened supernumerary branches resulting in incomplete coverage of the dendritic field.

Caper is widely expressed during embryogenesis

The caper^CC01391 allele was generated using a GFP protein trap (Buszczak et al., 2007), enabling us to utilize this line to investigate caper expression throughout embryonic development. Anti-GFP immunofluorescence reveals that Caper:GFP is expressed throughout the embryo during development. Moreover, Caper:GFP expression is pronounced within the brain and ventral nerve cord during embryogenesis (Fig. 2 A–D). Co-immunofluorescence of Caper:GFP and DAPI, to stain DNA within nuclei, shows that Caper is localized to the nuclear compartment and is excluded from the cytoplasm (Fig 2. E–G). We next examined the expression of Caper relative to markers of the nervous system. Co-immunofluorescence of Caper:GFP and the peripheral nervous system (PNS; marked by DHSB m22C10 and m21A6 antibodies) in caper^CC01391 embryos reveals nuclear localization of Caper throughout cells of the PNS including da neurons and chordotonal neurons (Fig. 3 A–F). Caper is also expressed along the ventral portion of the embryo but is excluded from the axons of the central nervous system (as marked by DHSB BP102 antibody; Fig. 3 J–L). In all tissues examined, Caper:GFP is localized to the nucleus, and excluded from the cytoplasm, consistent with its role as a splice factor (Fig. 2E–G; Fig. 3 A–F).

Figure 2. Caper is widely expressed throughout embryogenesis.

Caper expression was visualized by confocal microscopy by performing anti-GFP immunofluorescence in the caper^CC01391 GFP protein trap genetic background. Caper is expressed early in embryogenesis prior to gastrulation (A) and is seen throughout all cells of stage 4 blastoderm embryos including within the pole cells. (B) In stage 9 embryos, Caper is seen throughout the mesoderm and ventral ectoderm as neurogenesis begins. In later embryonic stages, (C, stage 13; D, stage 15) Caper is enriched within the ventral nerve cord (vnc) and central brain lobes (cb). Scale bar = 100 μm. (E,G) Higher magnification images reveal that Caper is localized to the nucleus during embryogenesis as determined by co-immunofluorescence with (F,G) DAPI. Scale bar = 10 μm.

Figure 3. caper regulates development and positioning of the lch5 organ.

Caper:GFP (B,E) is expressed in da neurons (A) and lch5 organs (D) of the embryonic PNS. Overlays of the PNS (marked by DHSB m22C10 and m21A6 antibody stains) and Caper:GFP are shown in (C,F). (G,I) Control lch5 organs form a stereotyped array in the abdominal segments. (H,I) caper^CC01391 embryos exhibit dorsal mispositioning and disorganization of lch5 clusters, as marked by an asterisk. (I) A schematic of the lch5 dorsal mispositioning phenotype as compared to control lch5 organs. Caper:GFP is expressed in the embryonic ventral midline (K), but is excluded from the CNS axons (marked by DHSB BP102 antibody stain; J). Overlay is shown in (L). Scale bar = 25 μm.

Loss of Caper function results in defects in lch5 organ positioning

To determine if caper is broadly required for neuronal morphogenesis, we analyzed the consequences of Caper dysfunction on the development of additional sensory neuron subtypes. Chordotonal (ch) neurons function as proprioceptors within the Drosophila PNS. The lateral ch neurons (lch5) are a group of five ch neurons that display a very distinct morphology and are positioned laterally along abdominal segments A1–A7. During embryogenesis lch5 precursor cells begin in a dorsal position with dendrites pointing ventrally. ch cells then migrate and stretch ventrally to occupy their final lateral position within each segment. Concomitantly, the entire ch organ undergoes a synchronized rotation to assume the mature stereotyped configuration, with the five dendrites of the organ aligned and pointing dorso-posteriorly (Figure 3 D,G,I; (Inbal, Levanon, and Salzberg, 2003)).

The majority (86%) of caper^CC01391 homozygous mutant embryos have at least one segment with aberrant lch5 development, including disorganization of the neurons within the organ, mispositioning of the entire cluster relative to other segments and/or fewer ch neurons within the organ. In contrast to normal lch5 clusters whose dendrites are evenly spaced with each tip pointing in parallel, 15.7% of mutant lch5 clusters contain neurons with dendrites that are mispositioned relative to other neurons in the organ. For example, in some instances, caper mutant ch neuron dendrites overlapped one another. 6.0% of segments also contained entire lch5 organs that were dorsally mispositioned relative to clusters in other abdominal segments. Dorsally mispositioned organs often had neurons oriented such that their dendrites point ventrally, indicating a failure of the organ to migrate and rotate during neurogenesis. In addition to lch5 organ positioning defects, 14.0% of lch5 clusters contained a reduced number of neurons within the organ. While a wild type organ typically has 5 neurons with 1 dendrite per neuron, lch5 clusters in caper mutant embryos often contained only one, two, three, or four neurons with dendrites per cluster (Table 1; Figure 3).

Table 1.

Chordotonal defects in caper deficient neurons

OreR control
Normal	Disorganized	Dorsal MP	Fewer Cells	More Cells	Total # of embryos (or segments)
33 (258)	7 (11)	0 (0)	3 (4)	1 (1)	44 (273)
94.5%	4.0%	0.0%	1.5%	0.3%
caper CC01391/ caper CC01391
Normal	Disorganized	Dorsal MP	Fewer Cells	More Cells	Total # of embryos (or segments)
5 (161)	23 (37)***	10 (14)**	19 (33)***	0 (0)	37 (235)
68.5%	15.7%	6.0%	14.0%	0.0%
elavGal4 control
Normal	Disorganized	Dorsal MP	Fewer Cells	More Cells	Total # of embryos (or segments)
14 (123)	6 (10)	0 (0)	3 (4)	4 (4)	20 (140)
87.9%	7.1%	0.0%	2.9%	2.9%
elavGal4; caperRNAiTRiP
Normal	Disorganized	Dorsal MP	Fewer Cells	More Cells	Total # of embryos (or segments)
12 (124)	6 (10)	1 (2)	1 (1)	1 (1)	20 (138)
89.9%	7.2%	1.4%	0.72%	0.72%
elavGal4; caperRNAiNIG
Normal	Disorganized	Dorsal MP	Fewer Cells	More Cells	Total # of embryos (or segments)
10 (125)	8 (11)	1 (2)	4 (6)	0 (0)	20 (140)
89.3%	7.9%	1.4%	4.3%	0.0%
ActinGal4 control
Normal	Disorganized	Dorsal MP	Fewer Cells	More Cells	Total # of embryos (or segments)
16 (129)	2 (5)	0 (0)	4 (4)	1 (1)	138 (21)
93.5%	3.6%	0.0%	2.9%	0.7%
ActinGal4; caperRNAiTRiP
Normal	Disorganized	Dorsal MP	Fewer Cells	More Cells	Total # of embryos (or segments)
12 (116)	6 (9)	2 (2)	5 (11)	2 (2)	19 (33)
87.2%	6.8%	1.5%	8.3%	1.5%

^*Note that some segments are disorganized or mispositioned and have an aberrant number of neurons. elav and Actin experiments were conducted at 29°C. All other experiments conducted at 25°C. Percentages are listed for the proportion of segments affected in each category. p<.0001***; p<.001**

To independently confirm the caper mutant lch5 phenotype, we knocked down caper expression using two different UAScaper-dsRNA hairpin lines. RNAi mediated knockdown of caper either ubiquitously using actinGal4 or pan-neuronally using elavGal4, results in a small number of embryos with dorsally mispositioned lch5 organs; dorsally mispositioned organs are never observed in gal4 control lines or wild type (OreR) lines (Table 1). We note that although disorganized clusters and organs containing fewer neurons are observed using RNAi mediated knockdown, these phenotypic categories are not significantly different from the controls. Since caper is maternally expressed, it is likely that RNAi efficacy is greatly reduced at early embryonic stages. Indeed, quantitative PCR analysis shows that although RNAi mediated knockdown of caper is effective at adult stages, a significant knockdown is not achieved during embryonic stages (see Experimental Procedures). Nonetheless, taken together, these results support the hypothesis that caper is required for the development and positioning of lch5 organs.

The dorsal mispositioning of lch5 organs is a very unique and striking phenotype that has been previously described in embryos mutant for the transcription factor encoding gene ventral veinless (vvl) (Inbal, Levanon and Salzberg, 2003). To determine if caper and vvl might function in a pathway that regulates lch5 positioning, we performed a genetic interaction analysis using hypomorphic alleles for both caper and vvl. Embryos that are homozygous mutant for both caper and vvl, have the same proportion of dorsally mispositioned lch5 organs (24%; n=41) as embryos that are homozygous mutant for caper^CC01391 alone (27%; n=37). These results do not support a genetic interaction between vvl and caper, but instead suggest that caper functions in a novel parallel pathway for lch5 positioning.

Caper regulates the development and maintenance of adult mechanosensory organs

Bristles are another well-studied mechanosensory organ of the adult PNS that can be categorized into two distinct subtypes: the microchaetes, which are smaller and variable in pattern among individuals, and the macrochaetes, which are larger and have a constant number and positioning across individuals. Sensory bristle organ progenitors are specified within imaginal discs during larval stages and ultimately develop a mature adult bristle organ comprising four cells including the externally visible shaft and socket cells, as well as the internal sheath cell and sensory neuron. The relative simplicity of the bristle organ has facilitated studying the genetic mechanisms that regulate the formation of these four cells from a single sensory organ precursor cell (Furman and Bukharina, 2008; Hartenstein and Wodarz, 2013). To determine whether caper is also required for the development of adult mechanosensory organs, we examined the effects of caper dysfunction on macrochaete bristle morphology in one-day old and seven-day old flies.

All one-day old control yw females (n=122) and males (n=122) showed a full complement of macrochaetes arranged in a stereotyped pattern. By contrast, 8.5% of caper^CC01391 homozygous mutant females (n=117) and 21% of caper^CC01391 homozygous mutant males (n=87) examined were missing some macrochaetes (Figure 4). When these same animals were examined 7 days post eclosion, we found that while only 0.8% female controls 0% of male controls were missing bristles on day 7, the number of missing bristles had increased to 18% in caper^CC01391 females and 32% in caper^CC01391 males. These results suggest that caper is required for the maintenance of bristle organs and caper dysfunction leads to the progressive loss of mechanosensory organs in adults. Moreover, this maintenance phenotype is more pronounced in males, as compared to females. The exact location and number of bristles missing was variable between individuals. These results suggest that caper functions in the development and maintenance of adult mechanosensory organs.

Figure 4. caper is required for bristle development and locomotor behavior in adults.

(A) yw and (C) RNAi control flies exhibit a stereotyped pattern of macrochaete bristle organs. On day one post-eclosion (B) caper^CC01391 and (D) caper RNAi knockdown adults are often missing bristle organs (arrows). (E) This missing bristle phenotype is significant among both caper^CC01391 males and females, but males are more affected than females (p= 0.013). (F) caper RNAi knockdown specifically affects males bristle development but not female bristle development. (G) caper^CC01391 adults are slower to climb up a cylinder after being startled, as compared to control flies; males are significantly more affected than females (ANOVA: sex, p= 0.002; genotype, p< 0.0001; sex × genotype interaction, p= 0.007). P-values reported on the figure are from post-hoc pairwise comparisons using Sidak’s adjustment for multiple testing. (H) caper RNAi knockdown also results in slower climbing speeds after startling as compared to controls but there was no difference in the magnitude of the effect in males versus females (ANOVA: sex, p= 0.617; genotype, p< 0.0001; sex x genotype, p= 0.366). Means and 95% confidence intervals (error bars) are back-transformed.

We next used RNAi mediated knockdown to confirm a role for caper in bristle development. While no actinGal4 control males (n=106) and only 0.8% control females (n=115) were missing macrochaetes on day 1 post-eclosion, 36% of caper-RNAi males (n=104) and 1% of caper-RNAi females (n=102) were missing macrochaetes on either the head or thorax. Similar results were obtained when caper was knocked down specifically in the nervous system using the elavGal4 driver (Fig. 4 A–F; Table 2).

Table 2.

Mechanosensory bristle organ defects in caper deficient adults

OreR
	Shortened Bristles	Missing Bristles	Total Flies
Female Day 1	0	0	111
Female Day 7	1	1	105
Male Day 1	0	0	103
Male Day 7	0	0	101
yw
	Shortened Bristles	Missing Bristles	Total Flies
Female Day 1	1	0	122
Female Day 7	97	0	107
Male Day 1	0	0	122
Male Day 7	93	1	116
Male Day 7	39	27	84
caperCC01391/caperCC01391
	Shortened Bristles	Missing Bristles	Total Flies
Female Day 1	1	10	117
Female Day 7	45	20	110
Male Day 1	2	18	87
Male Day 7	39	27	84
elavGal4 control
	Shortened Bristles	Missing Bristles	Total Flies
Female	0	0	172
Male	0	0	166
elavGal4; UAScaperRNAi
	Shortened Bristles	Missing Bristles	Total Flies
Female	0	0	100
Male	0	32	150
actinGal4 control
	Shortened Bristles	Missing Bristles	Total Flies
Female	1	1	115
Male	0	0	106
actinGal4;UAScaper RNAi
	Shortened Bristles	Missing Bristles	Total Flies
Female	2	1	102
Male	1	37	104

Interestingly, bristle patterning defects were significantly more pronounced in males than females in the genetic caper^CC01391 mutant background (p= 0.013; Fig. 4E). Additionally, while RNAi-mediated knockdown did not affect bristle formation in females, a significant proportion of males were missing bristles (Fig. 4F; Table 3). In fact, a subset of these males showed a more severe phenotype than caper^CC01391males and did not develop any macrochaetes. The stronger phenotypes seen with caper RNAi knockdown as compared to the caper genetic mutant background are consistent with the caper^CC01391 allele being hypomorphic. Taken together, these findings support a role for caper in bristle development and suggest that there may be sex specific requirements for caper in this process.

Table 3.

Sex-Specific Differences in Adult Phenotypes

	caperCC01391 mutants		caperRNAi
	Males	Females	Males	Females
Bristle Organ Defects	++*	+	+	−
Gravitaxis	++*	+	+	+

^*Although both sexes exhibit the mutant phenotype, the males are significantly more affected than females.

Loss of Caper disrupts negative gravitaxis behavior

Drosophila serves as an excellent model for studying animal behavior in response to gravity. When startled, adult flies exhibit a negative geotactic behavior, in which flies crawl up, away from the gravitational pull. In Drosophila, mechanosensory bristles and a subset of ch neurons located in the Johnston’s Organ of the adult antennae regulate sensory perception of the Earth’s gravitational field (Desroches et al., 2010). Since Caper dysfunction results in defects in various mechanosensory neurons, we used a well-established gravitaxis assay, which quantifies climbing speeds after the flies are startled, to determine if Caper dysfunction also results in behavioral defects.

Reduced Caper function, either through genetic mutation or RNAi-mediated knockdown, resulted in gravitaxis defects when compared to same-sex controls (Fig. 4 G–H). Both males and females in a homozygous caper^CC01391 background were slower in the gravitaxis assay after being startled. Males were, however, more strongly affected than females in the caper^CC01391 background (ANOVA: sex x genotype interaction, p = 0.007; Table 3). RNAi mediated knockdown of caper also resulted in slower climbing speeds, as compared to actinGal4 controls, but there was no difference in the severity of the defect between the sexes (ANOVA: genotype, p< 0.0001; sex x genotype interaction, p= 0.366). In addition to exhibiting slower climbing speeds, caper deficient flies displayed uncoordinated movements and tended to pause and rotate towards and against the gravitational field rather than moving up the walls continuously. Indeed, many caper deficient flies remained at the bottom of the cylinder and did not attempt to walk upwards after startling.

Caper functions outside the nervous system during development

While our study has largely focused on the role of caper in neurogenesis, it is likely that caper functions in other tissue types during embryogenesis. For example, we found that 28% of homozygous caper^CC01391 adult males (n=180) and 21% of homozygous caper^CC01391 adult females (n=227) had segmentation defects ranging from mild fusions of one or more abdominal segments (Fig. 5B), to missing a large portion of the thorax (Fig. 5C). By contrast only 1% of yw control males (n=101) and 0% of control females (n=96) exhibited segmentation defects (Fig. 5D). However, we did not see a significant difference in segmentation defects when caper was knocked down using RNAi. This is likely due to inefficient RNAi knockdown during early embryogenesis, when segmentation is established.

Figure 5. Caper dysfunction results in segmentation defects.

(B) caper^CC01391 adults exhibit abdominal segmentation and (C) thoracic segmentation defects in an incompletely penetrant manner (D) as compared to control adults (A). (D) Segmentation defects are significantly affected in both sexes in a caper^CC01391 background, as compared to controls.

Transcriptome-wide analyses confirm a broad role for caper in neurogenesis

To gain insight into the developmental processes regulated by caper, we performed transcriptome-wide analyses on caper^CC01391 homozygous mutant embryos as compared to control embryos and performed gene ontology (GO term) enrichment analyses on differentially expressed genes. caper^CC01391 embryos show differential expression of 583 genes, as compared to wild type control embryos (Table S2). GO term analysis shows that caper dysfunction affects the expression of genes associated with many different cellular and developmental processes. Importantly, a large number of significantly enriched GO terms from the differential expression analysis are related to nervous system development, dendrite morphogenesis, RNA processing, vesicular transport/synaptic signaling and cytoskeletal organization (Figure 6; Table S2). Caper has been previously shown to have separable molecular functions in splicing and in transcriptional regulation (Dowhan et al., 2005). Thus the changes in gene expression resulting from Caper dysfunction could be the result of direct Caper-mediated transcriptional regulation, or could be secondary effects associated with mis-splicing of transcriptional regulators. Nonetheless, these results confirm our phenotypic analyses and support a role for caper in neurite morphogenesis, as well as sensory neuron development.

Figure 6. Transcriptome-wide bioinformatics analyses implicate caper in the regulation of neurogenesis, endocytosis, RNA processing and cytoskeletal organization.

caper^CC01391 mutant embryos show significant differences in the expression of genes involved in many developmental and cellular processes. A subset of the significant GO terms are specifically relevant to neurogenesis and include (A) endocytosis, (B) nervous system development, (C) RNA processing and (D) cytoskeletal organization. Genes were clustered into functional groups based on GO terms for biological process and molecular function. Each node represents the most significantly enriched GO term for that cluster (FDR cut-off = 0.01). Green nodes represent clusters in which the majority of genes are significantly upregulated in caper^CC01391 mutant embryos, while red nodes represent clusters in which the majority of genes are significantly downregulated in caper^CC01391 mutant embryos. Color intensity is positively correlated with the proportion of genes in the cluster that are up- or downregulated. Grey nodes represent cases where upregulated and downregulated genes occurred in equal proportions. Node size correlates with statistical significance, where larger nodes are more significant than smaller nodes. (A,B) The GO term for each node is numbered and listed in the panel legend.

Since Caper has been established to function as a splice factor, we next used JunctionSeq to analyze our RNA-Seq datasets to detect differential splicing and other forms of alternative isoform regulation, such as alternative promoter usage or polyadenlyation (Hartley and Mullikin, 2016). Specifically, JunctionSeq identifies differential usage of exons or splice junctions by comparing relative expression of each sub-feature of a gene with the overall relative level of expression for the gene as a whole. This analysis identified 1541 genes that displayed alternative isoform regulation in the caper mutant background as compared to controls. Overall, this included differential usage of 2433 exons, 1626 known splice junctions, and 47 novel splice junctions. GO term enrichment analyses show that alternative exon usage is found within genes that regulate a variety of developmental and cellular processes including, but not limited to cytoskeletal organization, neurogenesis, as well as embryonic pattern specification. For example, we find that multiple genes that regulate dendrite number and dendrite length undergo alternative isoform regulation in caper^CC01391 embryos (Figure 7; Table S3). alan shepard (shep) is a conserved RBP that regulates dendrite length in Class IV da neurons and neuron number and organization in lch5 neurons (Schachtner et al., 2015; Olesnicky et al., 2014; Antonacci et al., 2015). In caper^CC01391 embryos, shep transcripts contain three specific exons that are statistically underrepresented as compared to those expressed in controls. Additionally, trio, fry and cut, each of which functions in Class IV da neuron dendrite number and/or length (Iyer et al., 2012; Emoto et al., 2004; Grueber, Jan and Jan, 2003), show alternative last exon and/or alternative first exon usage in caper^CC01391 embryos compared to controls (Figure 7). While this lends support to the hypothesis that Caper may function co-transcriptionally to influence promoter usage, future studies that identify direct Caper targets will be needed to adequately support this hypothesis. In sum, our bioinformatics analyses, combined with our detailed phenotypic analyses, implicate caper as an important regulator of Drosophila sensory neuron development.

Figure 7. JunctionSeq analysis in caper^CC01391 mutant animals compared to controls reveals differences in alternative isoform regulation of genes that are known regulators in da neuron development dendrite.

caper^CC01391 mutant embryos use alternative first exons and alternative stop sites for (A) cut mRNA. (B) fry mRNA also shows alternative first exon usage in caper^CC01391 mutant embryos. (C) shep mRNA shows alternative usage of internal exons in caper^CC01391 mutant embryos (D) trio mRNA shows alternative start and alternative internal exon usage in caper^CC01391 mutant embryos. Exons and splice junctions (brackets) colored in pink are those in which the relative level of expression between caper^CC01391 and control embryos differs from the levels observed for the gene overall (depicted in the right panel). Exon expression levels are plotted above the gene model with control reads in blue and caper^CC01391 reads in red. Alternative start and stop sites are depicted by a curved line underneath the alternative first exon or alternative stop site.

Discussion

Alternative splicing has emerged as a critical step in gene regulation and a fundamental mechanism for substantially increasing the complexity of the proteome despite genomes having a comparatively small number of protein coding genes. Interestingly, the incidence of alternative splicing is much higher in multicellular eukaryotes compared to unicellular eukaryotes, and in vertebrates compared to invertebrates. This suggests that rates of alternative splicing may be an important factor in evolution (Kelemen et al., 2013; Zaghlool et al., 2014). Indeed, it has been postulated that differences in alternative splicing patterns among human and nonhuman primate brain may be a contributing factor in human evolution (Lin et al., 2010; Merkin et al., 2012; Gueroussov et al., 2015).

The brain and nervous system have been found to have some of the highest levels of alternative splicing (Mohr and Hartmann, 2014; Pan et al., 2008; Zaghlool et al., 2014). Given this observation, it is not surprising that mutations in genes that encode regulators of splicing have been implicated in many neurological disorders. Moreover, a number of single nucleotide polymorphisms that are thought to result in aberrant splicing of the gene product are significantly associated with disease (Scotti and Swanson, 2016). Despite the fundamental importance of splicing to the development, function and maintenance of the nervous system, relatively few splicing factors have been studied in detail, especially with regard to their neural functions.

In this study, we find that the splice factor-encoding gene caper is widely expressed during Drosophila embryogenesis and that reduced caper function affects the development of various cell types within the nervous system. We extend our previous studies to show that caper regulates multiple aspects of dendrite development in Class IV dendrite arborization (da) neurons of the Drosophila peripheral nervous system. Additionally, we show that caper is required for the development of proprioceptive neurons during embryonic development and that adult caper mutant flies exhibit sensory organ defects and behavioral phenotypes. The phenotypes for various neuronal cell types, including bristle sensory organs and chordotonal neurons, are incompletely penetrant and vary in expressivity. It is likely that the incomplete penetrance is due to incomplete RNAi knockdown and residual Caper function in the caper hypomorphic allele. However, it remains possible that Caper function is redundant with other splicing factors that can compensate for Caper in its absence. Moreover, we find that phenotypes in adult gravitaxis, as well as mechanosensory bristle development and maintenance, significantly affect males more than females in the caper^CC01391 genetic mutant background. RNAi mediated knockdown of caper also results in sex-specific differences in mechanosensory bristle development in adults. These results suggest that caper may regulate splicing in a sex-specific manner (Table 3). Although we did not detect differences in splicing of fruitless, a master regulator of sex determination, our differential isoform analyses in caper^CC01391 mutant animals, as compared to controls, did detect an enrichment of genes associated with male genitalia development and male reproductive system development.

Differential gene expression analyses also confirm that loss of caper significantly affects the expression of genes that have been implicated in processes related to dendrite morphogenesis, organization of the cytoskeleton, nervous system development and RNA processing. These results suggest that although caper is widely expressed, it likely has specific requirements in neurogenesis. Indeed, while many splicing factors are widely expressed, even within the nervous system, studies have shown that specific neuronal subtypes are often more sensitive to the dysfunction of specific splice factors (Vuong, Black, and Zheng, 2016). In fact, we have previously shown that dendrite development within da sensory neurons in Drosophila and PVD neurons in C. elegans is particularly sensitive to splicing factor dysfunction (Olesnicky et al., 2014; Antonacci et al., 2015). The idea that certain cell types may be more susceptible to dysfunction of ubiquitously expressed RBPs likely stems from the observation that multiple RBPs may function combinatorially in large macromolecular complexes, termed ribonucleoprotein particles (RNPs), to regulate target RNAs in a cell-type or tissue-specific manner. Thus, dysfunction of a ubiquitously expressed RBP may contribute to tissue-specific pathologies if the overall function of the RNP complex is compromised (Brooks et al., 2015; Scotti and Swanson, 2016).

Our analysis of exon usage revealed 1541 genes that were differentially spliced in caper^CC01391 embryos as compared to wild type embryos. As this represents only ~10% of the Drosophila protein-coding genome and only specific exons within a given gene were often affected, these results support a role for Caper in alternative splicing and alternative isoform regulation, as opposed to a more general role as a core splicing factor. Additionally, many of the differences in splicing resulted in an overall shift in the prevalence of isoform usage, rather than the formation of novel transcript isoforms or transcripts encoding truncated protein products. Few studies have assessed the specific contributions of isoform usage and the importance of maintaining the appropriate ratio of spliceforms within cells. Nonetheless, it has been established that a strict balance in isoform ratios for certain genes is critical for cellular function. Indeed some neurological disorders, such as SMA and Frontotemporal dementia and Parkinsonism linked to Chromosome 17, are linked to an imbalance of alternative isoform usage (Faustino and Cooper, 2003).

Notably, among the genes that showed alternative isoform regulation within caper^CC01391 mutant embryos were known regulators of Class IV da neuron dendrite development including shep, trio, fry and cut. Previous studies have shown that similar to caper^CC01391 mutant Class IV da neurons, trio dysfunction results in the formation of shortened dendritic branches, more branch points per dendrite length, and an overall reduction in field coverage (Iyer et al., 2012). Additionally, disruption of either shep or cut function is associated with shortened dendritic branches and field coverage defects. Moreover, cut and shep mutant phenotypes are also highly variable and not fully penetrant, similar to those observed for various caper^CC01391 mutant neuronal phenotypes (Schachtner et al., 2015; Grueber, Jan and Jan, 2003). In addition to these regulators of branch length, caper^CC01391 mutant embryos show aberrant alternative isoform regulation for the gene fry. Similar to caper deficient Class IV da neurons, fry dysfunction results in supernumerary dendrite branch formation in Class IV da neurons (Emoto et al., 2004).

In addition to its role in alternative splicing, as defined by the use of alternative internal exons, our results suggest that Caper may also regulate the usage of alternative first exons. For example, both fry and trio mRNAs show alternative first exon usage in caper^CC01391 embryos as compared to controls (Fig. 7B,D). These results are consistent with mounting evidence that splicing predominantly occurs co-transcriptionally and that splicing machinery may contribute to the regulation of transcription (Brooks et al., 2015; Vuong, Black, and Zheng, 2016; Scotti, and Swanson, 2016; Mathieu, and Bouché, 2014; Jangi, and Sharp, 2014; Kelemen et al., 2013; Dowhan et al., 2005).

Studies have shown that alternative mRNA isoforms are differentially localized within cells. Indeed, mRNA localization has been shown to be important for the morphogenesis of many cell types, including neurons, and aberrant mRNA localization has been implicated in various neurological diseases (Tolino, Köhrmann and Kiebler, 2012; Holt and Schuman, 2013; Job and Eberwine, 2001; Doyle and Kiebler, 2011; Schuman, Dynes, and Steward, 2006; Brechbiel and Gavis, 2008; Mikl et al., 2010; Vessey et al., 2006). For example, alternative brain-derived neurotrophic factor (Bdnf) mRNA isoforms show distinct localization patterns in neurons, where a short isoform is restricted to the cell body and a long isoform is instead localized to the dendritic compartment. Importantly, these different isoforms regulate distinct aspects of dendritic spine morphogenesis, lending support to the idea that different mRNA isoforms can have distinct functions even within the same cell (Orefice et al., 2013).

A recent study has shown that alternative last exon usage dictates subcellular mRNA localization, where isoforms bearing the distal alternative exon are preferentially localized to neurites. Moreover, this likely represents a widespread occurrence within neurons, given that hundreds of genes showed isoforms with differential subcellular localization patterns that are correlated with alternative exon usage (Taliaferro et al., 2016). Interestingly, the results from our JunctionSeq analysis suggest that Caper may also regulate alternative last exon usage. Future studies will be required to elucidate if Caper directly influences alternative exon usage or if it is a secondary consequence of Caper regulation of other RBPs. Nonetheless, our results show that Caper dysfunction ultimately results in alternative isoform regulation and differential expression of many genes that have well-established roles in neurogenesis. This, combined with our phenotypic analyses of caper deficient animals, provides evidence for Caper as a new and critical player in the regulation of neurogenesis.

Experimental Procedures

Fly Strains

The following fly strains were used: ppkGal4,UAS-CD4-tdTomato (Han et al., 2012); elav-GAL4, UAS-mCD8:GFP (Lee, and Luo, 2001). The following stocks were obtained from the Bloomington Stock Center: {w[+mC]=Act5C-GAL4}y[1] w[*]; P17bFO1/TM6B, Tb[1]; caper^CC01391 (Buszczak et al., 2007) and UAScaperRNAi^TRIP (Caper^HMC03924) (Ni et al., 2011). UAS-caperRNAi^NIG was obtained from the National Institute of Genetics in Japan (Caper^NIG.11266R) (Ashton-Beaucage et al., 2014).

For RNAi experiments, UAS transgenes were used in single copy: virgin females from each Gal4 driver strain were crossed to UAS-RNAi males to generate larvae expressing RNAi hairpins. For each GAL4 driver, control larvae expressing GAL4 were generated by outcrossing to OreR. Animals were maintained at 25°C for all experiments using the caper^CC01391 mutant strain. RNAi experiments for chordotonal neuron defects were performed at 29°C to increase GAL4 efficiency. Stage 15–17 embryos were analyzed for chordotonal neurons. All other RNAi experiments were performed at 25°C using the UAScaperRNAi^TRIP, as high levels of lethality in caperRNAi experiments at 29°C precluded the ability to collect knockdown animals at adult stages.

Since the caper^CC01391 allele was generated in a yw background, OreR and yw were used as controls to ensure that genetic background was not influencing the phenotypic analyses. No significant phenotypic differences were noted between OreR or yw for any of the behavioral analyses, or for embryonic and larval sensory neuron analyses. We do note however that yw flies have shortened macrochaetes, as do caper^CC01391 adults (Table 2); we therefore do not attribute the shortened bristle phenotype to Caper dysfunction.

qPCR to assess efficacy of RNAi knockdown

Quantitative PCR (qPCR) was used to verify RNAi knockdown of caper expression in embryos and adult flies. RNA was extracted from pools of embryos or single adult male files following a standard Trizol protocol (5–7 biological replicates for control and RNAi lines). Any residual DNA was removed with the Turbo DNA-free kit (Thermofisher Scientific) and RNA was then converted into cDNA using the High Capacity RNA-to-cDNA kit (ThermoFisher Scientific). Primers were designed to amplify a 150 bp region of caper present in all annotated transcripts. selD was used as an endogenous control gene, given its highly stable expression across conditions in the RNA-seq experiment. Primers targeted a 132 bp region present in all annotated selD transcripts. Standard curves were generated for each primer set using a 1:5 dilution series, which indicated primer efficiencies were sufficiently high (caper= 95.5%; selD= 91.2%). Quantitative PCR was conducted using the PerfeCTa SYBR Green Supermix Rox kit with three technical replicates per biological replicate. Melt curves were analyzed and each primer set consistently produced a peak indicative of a single amplicon. One-sided students t-tests comparing comparing ΔCt values between control and RNAi lines were used to assess statistical significance of differences in expression, under the a priori hypothesis that RNAi lines would exhibit reduced expression (Yuan et al., 2006). Relative quantification was estimated using the efficiency corrected ΔΔCt approach.

There was no evidence for differences in caper transcript levels in embryos between RNAi and control lines (Student’s t-test, P= 0.997), suggesting that RNAi was not effective at this developmental stage. In contrast, RNAi knockdown in adults was effective, with the RNAi line exhibiting a 46% reduction in average caper transcript levels relative to the control line (Student’s t-test, P= 0.039).

Dendrite arborization neuron analyses

Dendrite morphology was examined in live dendrite arborization neurons of wandering larvae. Wandering larvae were placed on a microscope slide in a droplet of phosphate buffered saline (PBS). Two 22×22 mm No 1.5 glass coverslips were placed on either side of the larva to act as spacers; each spacer was held in place using a droplet of PBS. A 22×40 mm No 1.5 glass coverslip was then placed on top of the larva to immobilize it for confocal imaging. Only ddaC neurons from abdominal segments A4-A6 were visualized. No more than two ddaC neurons were imaged per larva. Branch number and branch length were quantified in Z-series projections of live ddaC neurons imaged on a Leica SP5 confocal microscope, using a Å~40/1.25 NA oil objective. Confocal images were taken with the ddaC neuron cell body of each neuron in the same position to ensure consistency in the field of view across all neurons as previously performed (Schachtner et al., 2015). The number of dendritic branches visible within this standardized field of view and dendrite branch length were quantified using NeuronJ (Meijering et al., 2004). Statistical significance was determined by performing the Mann-Whitney test, using OreR as a control.

Immunofluorescence

Embryos were dechorionated using a 50% bleach solution and fixed in a 1:4 solution of 4 % paraformaldehyde/heptane for 20 min. Embryos were subsequently devitellinized using methanol. For immunofluorescence, embryos were blocked using Image iT FX Signal Enhancer (Life Technologies) for 30 min and incubated overnight at 4 °C with the following primary antibodies in PBS/0.1 % Triton X-100/5 % normal goat serum (NGS): 1:500 mouse anti-Futsch (m22C10, (Developmental Studies Hybridoma Bank [DSHB]) and 1:200 m21A6 (DSHB) to stain chordotonal neurons; 1:350 Alexa Fluor 488 rabbit anti-GFP (Invitrogen) to stain for Caper:GFP, and 1:1000 mouse anti-central nervous system (CNS) axons BP102 (DHSB). The following secondary antibodies (Invitrogen) were incubated overnight at 4°C in PBS/0.1 % Triton X-100/5 % NGS: 1:500 Alexa Fluor 546 goat anti-mouse, 1:500 Alexa Fluor 488 goat anti-mouse. Chordotonal neurons were imaged at the same settings on a Leica SP5 confocal microscope using a Å~63/1.4 NA oil objective.

SDS-PAGE and Immunoblotting

D. melanogaster cleared cell lysates were prepared by crushing larvae and adults with a pestle in a 1.5 ml microfuge tube in 200 μl of 1X dPBS followed by centrifugation at 13,000 rpm for 10 minutes to remove cell debris. Cell lysates were fractionated in Novex® Pre-Cast 4–20% Tris-Glycine gels (Invitrogen) and transferred to Immobilon-P PVDF membranes (Millipore) by electroblotting at 30 mA overnight at 4 °C. Caper:GFP proteins were detected with a mouse anti-GFP monoclonal antibody (Sigma G6539) at 1:2000 dilution followed by peroxidase conjugated goat anti-mouse IgG, Fcγ fragment specific (Jackson ImmunoResearch 115-035-008) at a 1:20,000 dilution in 1X dPBS, tween 20 0.1% and 5% non-fat dried milk. The membranes were developed by chemiluminescence using the ECL Prime Western Blotting detection system (Amersham Biosciences) on a Li-COR C-DiGit.

Negative Geotaxis Behavioral Assay

The climbing assay was performed with one-day old flies, which were tested in groups of 10 animals per trial (Vrailas-Mortimer et al., 2011). Males and females were separated for trials as they exhibit distinct speeds dependent on sex. Flies were briefly tapped 3 times to the bottom of a graduated cylinder and the time it took for 50% of flies to walk upward to a marked line, was documented. Each trial was video recorded and the time for 50% of the flies to climb to the mark was determined from video recordings to ensure accuracy. For yw controls, 47 trials were conducted with males and 49 with females. 46 trials were conducted with males and 59 trials were conducted for females in a caper^CC01391 background. caper RNAi was activated using actinGal4; 19 trials were conducted for caper RNAi males and 32 trials were conducted for caper RNAi females. As a control, actinGal4 was outcrossed to OreR. Forty-four trials were conducted on the males and 50 trials were conducted on the female progeny of this cross. All behavioral data was analyzed using a Factorial Analysis of Variance (ANOVA), with model factors including genotype, sex, and a genotype by sex interaction. Data was reciprocally transformed prior to analysis in order to normalize residuals. In cases where the interaction term was significant, the nature of the interaction was dissected using post-hoc tests following Sidak’s procedure to adjust for multiple testing.

RNA isolation and Sequencing

RNA was isolated from OreR control and caper mutant Drosophila embryos using TRIzol Reagent (Invitrogen), as per manufacturer’s protocol. Three biological replicates were used for each genotype. After DNase I (Roche) treatment, RNA was purified with a phenol:chloroform extraction and resuspended in RNAse-free water. RNA concentrations were determined by absorbance at 260 nm.

RNA libraries were prepared and sequenced by the University of Colorado Genomics and Microarray core using the Illumina TruSeq Stranded RNA library preparation kit. Paired-end 2×125bp sequencing was performed in a single lane using an Illumina HiSEQ 2500.

RNA-seq Data Analysis

Derived sequences were analyzed by the Anschutz Genomics Core Facility by applying a custom computational pipeline consisting of the open-source gSNAP, Cufflinks, and R for sequence alignment and ascertainment of differential gene expression. In short, reads generated were mapped to the Drosophila genome (Bdgp5) by gSNAP, counts (FPKM) were derived by Cufflinks, and differential expression analyzed with ANOVA in R (Baird et al., 2014; Bradford et al., 2015; Maycotte et al., 2015; Wu and Nacu, 2010; Trapnell et al., 2010). A Q value of 0.01 was used as the cut-off for differential expression. Gene Ontology (GO term) enrichment analysis and pathway annotation network analysis was performed using the ClueGO plugin for Cytoscape (Bindea et al., 2009). Genes were clustered using GO terms for biological process and molecular function at GO tree levels from 5–11. Enrichment was tested using a right-sided hypergeometric test, and P-values were adjusted to control the false-discovery rate (FDR) using the Benjamini-Hochberg procedure (cut-off significance was 0.01). Analysis of differential splicing and novel spliceforms was performed using JunctionSeq with a FDR cut-off for significance of 0.01 (Hartley and Mullikin, 2016).

Statistical Analyses

Defects in bristle development, chordotonal neuron development and segmentation were analyzed using the Pearson Chi-square test or Fisher’s exact test (when expected cell counts were less than 5). Bristle defect comparisons were performed in day-old flies. For bristle development and segmentation, comparisons between mutant or RNAi lines and controls were carried out separately for males and females. If significant defects were detected in mutant or RNAi lines, sexes were subsequently compared to test whether the severity of defects differed between males and females. To analyze potential defects in chordotonal neuron development, comparisons between mutant or RNAi lines and controls were performed separately for four distinct phenotypes: (1) disorganization (2) dorsal mispositioning (3) presence of more cells, and (4) presence of fewer cells. Specifically, the number of embryos displaying an aberrant phenotype in at least one segment were compared between mutant or RNAi lines and controls. Since many embryos displayed defects in multiple segments, the total number of defective segments are also reported.