Cut, via CrebA, transcriptionally regulates the COPII secretory pathway to direct dendrite development in Drosophila

Neurons are complex, highly polarized cells that come in an astonishing array of shapes and sizes, attributable largely to their dendritic branching patterns. As dendrites are primarily specialized to receive and process neuronal inputs, the specific morphology of the dendrite can govern neuronal function, signal integration and circuit assembly (London and Häusser, 2005). Thus, understanding the biological mechanisms regulating the development of dendritic arbors is of particular importance for an understanding of nervous system organization and function.

Owing to their stereotyped and class-specific dendrite branching patterns, Drosophila dendritic arborization (da) neurons provide an excellent model system to investigate the cellular and molecular mechanisms that regulate the acquisition of distinct dendritic architectures. Several studies have focused on the identification and characterization of transcription factors that function to specify and control dendritic growth, branching and cytoskeletal rearrangements (Jan and Jan, 2010; Corty et al., 2009; Ye et al., 2011; Iyer et al., 2012; Nagel et al., 2012). For example, members of the Cut/Cux1/Cux2 family of homeodomain transcription factors have been shown to be multi-level regulators of synaptogenesis and dendritic spine morphology in the brain cortex (Cubelos et al., 2010; Li et al., 2010) in the acquisition of class-specific dendritic arbor complexity among Drosophila da neurons as well as (Grueber et al., 2003). In other cases, the combinatorial action of multiple transcription factors operates to specify dendritic arbor shape in individual subtypes (Jinushi-Nakao et al., 2007). Although it is known that transcriptional programs function to generate the array of neuronal shapes that arise, what remains to be understood is the downstream implementation of these programs and what cellular pathways and biological processes are recruited to enable these changes in dendrite morphology as well as maintain class-specific dendritic homeostasis.

Directed secretory flux is a core cellular process that is known to play a crucial role in many dynamic cellular events through directional addition of plasma membrane (Kupfer et al., 1982; Kupfer et al., 1985; Prigozhina and Waterman-Storer, 2004). Morphological changes that occur during neuronal differentiation, particularly dendritic branching, are highly dynamic involving cyto-architecture and signaling events that require the coordinated addition of membrane to particular regions. Given the extremely large areas and distances over which secreted molecules and membrane must be transported in neurons to promote growth, the secretory pathway has specialized functional significance in providing the major source of plasma membrane to promote dendritic growth and facilitate branching (Cui-Wang et al., 2012). Neurons have evolved a rather novel spatial organization of the endoplasmic reticulum (ER) and Golgi compared with non-neuronal cells: they have somatic and satellite dendritic ER and Golgi compartments known as ER or Golgi outposts (Aridor et al., 2004; Horton and Ehlers, 2003; Ye et al., 2007). Polarized secretory trafficking is a conserved process that is preferentially implicated in the specification and asymmetric growth of dendrites versus axons (Horton and Ehlers, 2003; Horton et al., 2005; Ye et al., 2007; Satoh et al., 2008; Zheng et al., 2008). Although such a distinct involvement in different neuronal processes suggests that the secretory pathway is probably under the regulation of genetic or transcriptional programs linked to dendritic patterning, little is known about how these transcriptional programs might regulate the secretory pathway to mediate this biological process.

Here, we demonstrate that the COPII coat protein Sec31 functions cell-autonomously in promoting dendritic growth and branching. We provide evidence to show that Sec31 and other COPII secretory pathway components are transcriptionally regulated by Cut to mediate changes in dendritic complexity. These transcriptional changes also translate into a concomitant increase in the number of satellite secretory outposts that colocalize largely with the branch initiation sites. Moreover, we reveal a novel functional role for CrebA in directing proper dendrite development. Our analyses indicate that CrebA functions as a downstream effector of Cut-induced dendrite elaboration that coordinately regulates expression levels of the COPII secretory pathway components. Collectively, these studies provide evidence of a novel transcriptional regulatory program governing COPII-mediated secretory activity in modulating dendrite arborization.

From an ethyl methanesulfonate (EMS) mutagenesis screen, we isolated a hypomorphic recessive lethal mutation [initially designated l(2)0805] that produced strong defects in dendritic growth and branching at the late embryonic stage (supplementary material Fig. S1A,B). To genetically map the l(2)0805 mutation, we conducted complementation analyses between this allele and molecularly defined genomic deficiencies that mapped the allele between cytological bands 44E3 and 44F7. Complementation analyses within this cytological region ultimately revealed that this mutation represented a sec31 allele, hereafter referred to as sec31⁰⁸⁰⁵. Rescue experiments performed using Sec31 overexpression rescued both lethality and phenotypic defects exhibited by homozygous sec31 mutant larvae, clearly indicating that the reported phenotypes are specifically due to the sec31⁰⁸⁰⁵ mutation (supplementary material Fig. S1F,G). Immunohistochemistry (IHC) analyses reveal that Sec31 is expressed in a punctate fashion in all da neurons (supplementary material Fig. S1C–E). To confirm antibody specificity and validate the efficacy and specificity of UAS-sec31^RNAi transgenes, we performed Sec31 IHC analyses on class IV da neurons expressing RNAi transgenes, and compared the results with expression levels in adjacent class III da neurons. These analyses revealed specific knockdown of Sec31 in class IV neurons (∼43%) relative to class III neurons within the same cluster, supporting the specificity and efficacy of the sec31 reagents (supplementary material Fig. S1H–J).

To investigate potential cell-autonomous roles of sec31⁰⁸⁰⁵, we conducted systematic mosaic analysis with a repressible cell marker (MARCM) analyses in each of the four da neuron subclasses. Relative to wild-type (WT; Fig. 1A,A′), sec31 mutant class IV neurons (Fig. 1B,B′) displayed significant reductions in both terminal dendritic branching (Fig. 1G) and total length (Fig. 1H). Class III da neurons are distinguished by the presence of numerous actin-rich dendritic spikes emanating from the primary branches. The reduced number of terminal dendritic filopodia in sec31 mutant class III neurons (Fig. 1D′D) relative to WT (Fig. 1C′C), supports a role for sec31 in the development of these actin-rich dendritic processes (Fig. 1G,H). Similarly, sec31 class II da neurons displayed defects in terminal dendritic branching and elongation (Fig. 1F,F′) compared with WT (Fig. 1E,E′). A previous study (Ye et al., 2007) reported similar dendritic branching phenotypes caused by sec23 and Sar1 mutations in class IV neurons. In contrast to class II–IV da neurons, sec31⁰⁸⁰⁵ mutant class I da neuron clones did not show any major quantitative defects in either dendritic branching or growth, compared with WT (Fig. 1G,H). Collectively, these analyses indicate that sec31 functions cell-autonomously in promoting higher order terminal dendritic branching and growth in order to achieve dendritic architecture homeostasis.

We next conducted overexpression studies using two independent UAS-sec31 transgenic strains. These studies revealed that altering Sec31 levels induces varying dendrite morphogenesis phenotypes in the different da neuron subclasses. Intriguingly, we observed significantly increased dendritic branching and growth, in class I neurons, relative to controls, particularly proximal to the cell body (Fig. 1I,J,M,N). In contrast, Sec31 overexpression in class IV neurons resulted in a significant reduction in dendritic complexity both proximal and distal to the cell body (Fig. 1K,L,O,P) as well as a proximal shift in branch order distribution, giving rise to a higher percentage of lower order branching relative to control (Fig. 1Q). These results further confirm roles for Sec31 in regulating dendritic homeostasis and suggest that proper levels of Sec31 are required for this process with respect to dendritic growth and branching among distinct da neuron subclasses.

The Cut homeodomain transcription factor has been demonstrated to selectively control da neuron dendrite morphogenesis on the basis of varying levels of expression in different neuron subclasses. Because Cut protein immunoreactivity is normally not detectable in class I da neurons (Grueber et al., 2003), ectopic Cut expression in these neurons provides an excellent sensitized genetic background for dissecting the molecular mechanisms by which Cut regulates dendrite morphogenesis. Ectopic Cut expression in class I da neurons exerts dramatic effects on dendritic complexity, shifting their relatively simple dendrite morphology to a highly complex branching pattern similar to that observed in class III and IV da neurons (Grueber et al., 2003) (Fig. 2A,A′,B,B′). Given that sec31 MARCM analyses revealed roles in promoting dendritic growth, branching and the development of dendritic filopodia, we hypothesized that Sec31 may be functioning downstream of Cut to promote dendritic branching complexity. Moreover, comparative microarray analyses of class I da neurons ectopically overexpressing Cut revealed an ∼36-fold increase in sec31 expression levels over WT levels (data not shown), further leading us to investigate the potential regulatory relationship between Cut and Sec31.

To test our hypothesis, we phenotypically examined the requirement of Sec31 for Cut-mediated development of dendritic filopodia in class I neurons. Class I vpda neurons ectopically overexpressing Cut with a simultaneous sec31 RNAi knockdown (Fig. 2C,C′), showed a strong suppression of Cut-induced dendritic filopodium formation and growth (Fig. 2E,F) as compared with control neurons overexpressing Cut alone (Fig. 2B,B′). Similarly, Cut overexpression in a sec31⁰⁸⁰⁵ mutant background also suppressed dendritic branching (Fig. 2E) and growth (Fig. 2F). We confirmed these observations using a second independent sec31 RNAi transgene and similar phenotypes were also observed in the other two class I neurons, ddaD and ddaE (data not shown). Simultaneous overexpression of Cut and Sec31 led to increased dendritic complexity in class I neurons, as compared with those overexpressing Cut alone (Fig. 2B,B′,E,F), indicating a synergistic interaction in inducing dendritic branching (2D,D′,E) and growth (Fig. 2F). To rule out that the observed suppression of the UAS-cut phenotype could be due to titration of GAL4 caused by the presence of an additional UAS sequence, we introduced a neutral UAS line (UAS-mCD8::RFP) as a control in co-expression experiments. Quantitative analyses revealed that the presence of an additional neutral UAS transgene fails to produce any abnormal dendritic phenotype as compared with Cut overexpression alone (supplementary material Fig. S2A,B), thereby confirming the specificity of the suppression phenotype produced by co-overexpression of Cut and sec31^RNAi.

To assess potential Cut transcriptional regulatory effects on sec31 expression levels, we isolated pure populations of class I neurons from WT and ectopic-Cut-expressing genetic backgrounds using magnetic bead based sorting (Iyer et al., 2009; Sulkowski et al., 2011; Iyer et al., 2012). Total RNA was analyzed by qRT-PCR for the relative expression levels of sec31 and cut mRNAs between genetic backgrounds. These analyses revealed significant upregulation of sec31 (4±1.1-fold) and cut (14±4-fold) mRNAs, indicating that ectopic Cut upregulates sec31 expression (Fig. 2G). Corroborating this observation, IHC analyses showed a modest, but significant, increase in Sec31 protein expression (20–25%) in Cut-overexpressing neurons (Fig. 2I,I′,J), relative to WT levels (Fig. 2H,H′,J). This transcriptional regulatory relationship also holds true in other da neuron subclasses where Cut is substantially expressed. Cut overexpression in class IV da neurons (4.4±1.2-fold) also caused a significant increase in sec31 expression levels (3±0.3-fold; Fig. 2L). Moreover, targeted knockdown of cut in class III da neurons leads to a significant reduction in sec31 expression relative to controls (Fig. 2K). IHC analyses confirmed significant reduction in Cut expression levels with cut^RNAi knockdown (supplementary material Fig. S2D). These analyses reveal that Cut regulates Sec31 expression and that Sec31 functions as a downstream effector of Cut-mediated dendrite morphogenesis.

To investigate whether other known COPII components were required for facilitating Cut-driven dendritic elaboration, we expressed sec13^RNAi, sec23^RNAi, sec24^RNAi and Sar1^RNAi in class I neurons that were simultaneously overexpressing Cut and compared the results to those in neurons ectopically expressing Cut alone (Fig. 3A,A′–F,F′). In all cases, the predominant phenotypic effect was a significant reduction in the number of dendritic branches with COPII component knockdown (Fig. 3C,C′–F,F′). Quantitative analyses confirmed statistically significant reductions in the number of dendritic terminals (Fig. 3G) and total dendritic length (Fig. 3H). We confirmed these observations through the use of a second independent RNAi transgene against sec13, sec23, sec24 and Sar1. We examined the effects of concurrent overexpression of Cut and Sec23 in class I neurons that likewise led to a dramatic enhancement of dendritic growth and/or branching (Fig. 3G,H) as with Cut and Sec31 (Fig. 2D,D′). Moreover, qRT-PCR analyses revealed that all COPII component gene expression levels were significantly upregulated with Cut overexpression, including sec13 (2.5±0.5-fold), sec23 (2.05±0.2-fold), sec24 (5.5±1.2-fold) and Sar1 (5±1.2-fold) relative to WT levels (Fig. 3I). Cut-mediated transcriptional regulation of the COPII machinery was also independently verified in class IV da neurons overexpressing Cut (4.4±1.2-fold), where relative to WT, expression of these genes was likewise upregulated as follows: sec13 (3.2±0.2-fold), sec23 (2.6±0.3-fold), sec24 (2±0.3-fold) and Sar1 (2.1±0.2-fold) (supplementary material Fig. S2C). These results indicate that COPII machinery is transcriptionally regulated by Cut to induce changes in dendritic complexity.

The CrebA/Creb3-like family of transcription factors have been implicated in the regulation of secretory function in the salivary gland (Abrams and Andrew, 2005), and other non-secretory cell types (Fox et al., 2010). Drosophila CrebA protein binds directly to the enhancers of genes in the general secretory machinery as well as cell-type-specific secreted cargo (Fox et al., 2010). As our results indicate that Cut regulates the secretory pathway to influence dendrite morphology, we hypothesized that CrebA may function as an intermediate transcription factor in this gene expression cascade.

To test this hypothesis, we compared the phenotype of class I neurons ectopically overexpressing Cut with class I neurons ectopically expressing Cut and simultaneously expressing CrebA^RNAi. These analyses revealed strong suppression of the Cut-mediated formation of dendritic filopodia in the presence of CrebA^RNAi (Fig. 4C,C′) compared with control cells (Fig. 4B,B′) resulting in a significant reduction in the number of dendritic terminals (Fig. 4E) and total dendritic length (Fig. 4F). Similarly, Cut overexpression in a CrebA⁰³⁵⁷⁶ mutant background also suppressed dendritic branching (Fig. 4E) and growth (Fig. 4F), which could be rescued by simultaneous overexpression of UAS-CrebA in a CrebA mutant background (Fig. 4E,F). Synergistic interaction between CrebA and Cut by simultaneous overexpression (Fig. 4D,D′) resulted in enhanced dendritic branching (Fig. 4E) and growth phenotypes (Fig. 4F). Moreover, CrebA overexpression was able to rescue the loss of dendritic terminals caused by knockdown of cut in class III da neurons (Fig. 4G–I,L), further indicating that CrebA is an essential downstream effector of Cut-induced dendritic elaboration. We also examined whether Sec31 overexpression could rescue cut knockdown effects on class III dendritic branching, and although Sec31 partially rescued reductions in dendritic terminal branching, overexpression of Sec31 alone, unlike CrebA, was insufficient to produce statistically significant effects (Fig. 4L).

To examine potential Cut transcriptional regulation of CrebA, we performed qRT-PCR analyses and, relative to WT, Cut overexpression in class I neurons significantly upregulated CrebA expression by ∼4±1.1-fold (Fig. 4J). Moreover, IHC analyses showed a significant increase in CrebA protein expression levels (20–25%) in Cut-overexpressing neurons (Fig. 4N; supplementary material Fig. S2G,H), relative to WT (Fig. 4N; supplementary material Fig. S2E,F). Cut-mediated transcriptional regulation of CrebA was also independently verified in class IV da neurons overexpressing Cut, where, relative to WT, expression of CrebA was enhanced by 5.4±1.2-fold (Fig. 4M). In class III da neurons, RNAi-mediated knockdown of cut resulted in a significant decrease in CrebA expression levels (Fig. 4P,P′,K), relative to WT (Fig. 4O,O′,K). Analyses of cut null MARCM clones revealed that CrebA expression is not solely dependent upon Cut for expression in da neurons (supplementary material Fig. S2I–K). Interestingly, Cut overexpression in class I neurons also causes a significant upregulation in forkhead (fkh) expression by approximately twofold (Fig. 4J). fkh has been demonstrated to have a crucial role in the morphogenesis of the salivary gland where it maintains the expression of CrebA (Abrams et al., 2006), suggesting that a similar transcriptional cascade may operate in da neurons for Cut-mediated modulation of CrebA levels.

To further analyze CrebA, we conducted analyses using two independent CrebA^RNAi transgenes. We first examined the effects of CrebA knockdown in class III neurons, which resulted in a substantial reduction in the formation dendritic filopodia (Fig. 5A,A′,B,B′) together with reductions in the number of dendritic terminals (Fig. 5E) and total length (Fig. 5F). Similarly, CrebA knockdown in class IV ddaC neurons affected dendritic terminals (Fig. 5E) and total length (Fig. 5F) causing an overall reduction in branch complexity (Fig. 5C,C′,D,D′), indicating that CrebA functions in promoting higher order branching and dendritic growth in both class III and IV neurons.

CrebA overexpression in class IV neurons lead to reduced dendritic field coverage with rudimentary dendritic branching both proximal and distal to the cell body (Fig. 5J–L), in contrast to class I neurons in which CrebA overexpression resulted in overall increased dendritic branching and growth (Fig. 5H,K,L) relative to WT (Fig. 5G,I,K,L). Thus, consistent with observations for Sec31 (Fig. 1) and Cut (Grueber et al., 2003), altering CrebA expression levels results in varying effects on class-specific dendrite arborization.

Furthermore, IHC analysis of CrebA revealed its nuclear and cytoplasmic localization in da neuron cell bodies (Fig. 5M–O). This cytoplasmic and nuclear expression pattern is consistent with that previously observed with the CrebA human orthologues, Creb3L1 and Creb3L2, in which the full-length protein colocalizes with the ER and the truncated form is predominantly nuclear (Fox et al., 2010). Interestingly, we also observed CrebA localization in a punctate pattern along dendritic branches (Fig. 5P–R). Previous studies of mammalian CREB have also revealed dendritic localization (Crino et al., 1998), however, CrebA is more closely related to the Creb3-like transcription factors. Finally, staining of CrebA⁰³⁵⁷⁶ homozygous mutant larvae with CrebA antibodies failed to reveal any specific staining, thereby confirming the antibody specificity (Fig. 5S–U).

To investigate the potential regulation of the COPII secretory pathway by CrebA in dendrite development, we employed a phenotypic suppression assay in which we performed targeted RNAi knockdown of individual COPII components including sec31, sec13, sec23, sec24 and Sar1 in class I neurons simultaneously overexpressing CrebA and compared them with neurons expressing CrebA alone (Figs 6B,B′–G,G′). These analyses revealed that RNAi against each of the COPII components produced phenotypic suppression of CrebA-mediated increases in vpda dendritic branching and growth (Fig. 6C,C′–G,G′,I,J). We also examined the effects of concurrent overexpression of CrebA and Sec31 (Fig. 6H,H′,I,J) and CrebA and Sec23 (Fig. 6I,J) in class I neurons. These analyses revealed synergistic enhancement in dendritic branching and growth in both cases relative to that observed with CrebA overexpression alone. Thus, consistent with Cut, CrebA also functions cooperatively with Sec31 and Sec23 in promoting dendritic branching and growth. Moreover, Sec31 overexpression was able to rescue the loss of dendritic terminals (Fig. 6M) caused by knockdown of CrebA in class IV da neurons (supplementary material Fig. S2L–N), further indicating that Sec31 functions downstream of CrebA to modulate dendrite architecture.

We assessed the transcriptional effect of CrebA on COPII mRNA expression levels, by isolating total RNA from WT and CrebA-overexpressing da neurons using the pan-da neuron marker Gal4^21-7,UAS-mCD8::GFP. qRT-PCR analyses revealed that the mRNA expression levels of each COPII component was significantly upregulated with CrebA overexpression including sec31 (10.2±4.3-fold), sec13 (11.1±4.8-fold), sec23 (3.5±1.9-fold), sec24 (7.4±3.6-fold) and Sar1 (9.8±4.1-fold) relative to WT expression levels (Fig. 6L). Moreover, we verified CrebA mRNA levels were upregulated (19.2±7.8-fold) relative to WT expression levels (Fig. 6L; supplementary material Table S1). These results, together with the phenotypic assays, indicate that CrebA exerts transcriptional regulation over the COPII transport machinery in da neurons.

To determine whether CrebA is an essential intermediate in the transcriptional cascade involving Cut-mediated upregulation of COPII mRNA expression levels, we isolated total RNA from Cut-overexpressing class I neurons and class I neurons expressing Cut with simultaneous knockdown of CrebA. qRT-PCR analyses revealed that with CrebA knockdown, Cut-mediated transcriptional upregulation of the COPII components, relative to that in class I neurons ectopically expressing Cut alone, was significantly reduced. Specifically, we observed significant percentage reductions in the expression of all COPII components: sec31 (56.14±13.15%), sec13 (65.2±16.2%), sec23 (87.17±11.79%), sec24 (64.4±14.13%) and Sar1 (53.05±17.37%) (Fig. 6K). These results indicate that CrebA is essential for Cut-mediated regulation of COPII transport machinery. Moreover, we found that the CrebA^RNAi transgene is effective in knocking down CrebA levels (Fig. 6K). Collectively, these analyses reveal that Cut acts through CrebA to regulate COPII gene transcription.

To determine whether Cut-induced regulation of COPII secretory pathway gene expression also translated into regulation of the distribution and/or number of the satellite secretory outposts in dendritic processes, we first used UAS-Sten-GFP, which marks ER exit sites (Förster et al., 2010), in class I neurons (Fig. 7A–C) and also class I neurons expressing UAS-cut that exhibit de novo dendritic branching (Fig. 7D–F). Quantitative analyses showed that the total number of ER exit sites significantly increases with Cut overexpression (130±45) compared with WT (55±9) (Fig. 7M). Moreover, ER exit sites are localized to the branch initiation sites (Fig. 7D–F), which was also confirmed by quantitative analysis (Fig. 7N). Similarly, we examined the distribution of Golgi outposts using the UAS-ManII-eGFP transgenic reporter and verified the exclusion of these satellite Golgi outposts from the axonal compartment of class I neurons (supplementary material Fig. S3D–F) consistent with what has previously been shown for class IV neurons (Ye et al., 2007). Moreover, the presence of Golgi outposts in class I neurons was independently confirmed by IHC staining with the Golgi marker, GM130 (supplementary material Fig. S3A–C). As with ER exit sites, we observed Golgi outposts along the dendrite and highly coincident with branch initiation sites (Fig. 7J–L) in class I neurons overexpressing Cut. This was also reflected in the quantitative analysis where greater than 90% of the Golgi outposts colocalized with branch initiation points (Fig. 7O). We likewise performed these analyses for ER exit sites in class IV da neurons. Interestingly, although Cut overexpression in class IV neurons led to reductions in dendritic complexity (supplementary material Fig. S4A–D) (Grueber et al., 2003), when the number of ER exit sites is normalized to total dendritic length there is a highly significant increase in ER exit site puncta (supplementary material Fig. S4E) consistent with upregulation of the COPII machinery. These results together indicate that Cut-mediated effects on dendritic complexity may function through a mechanism consisting of changes in satellite ER–Golgi complexity, as a result of an increase in expression levels of the COPII secretory apparatus, that act as focal points for branch point initiation.

Dendrite morphogenesis is a growth-intensive process that requires cytoskeletal reorganization and the addition of enormous amounts of plasma membrane (Bradke and Dotti, 1997; Lecuit and Pilot, 2003), which is principally provided by the secretory pathway through the ER, Golgi complex and post-Golgi intermediates (van Vliet et al., 2003). Recent findings have demonstrated the importance of polarized post-Golgi trafficking in dendritic specification and compartmentalization (Ye et al., 2007, Horton et al., 2005) strongly indicative of regulatory programs that control the secretory pathway. Here we present evidence for a novel transcriptional regulatory program mediated by Cut through CrebA that directs expression of the COPII secretory machinery to drive dendrite morphogenesis.

Although the importance of Cut/Cux1/Cux2 has been well established in neuronal specification and dendrite development (Grueber et al., 2003; Iulianella et al., 2009; Cubelos et al., 2010; Li et al., 2010) it is only recently that studies have begun to unravel the machinery underlying Cut-mediated regulation of dendritic elaboration in da neurons (Sulkowski et al., 2011; Iyer et al., 2012; Nagel et al., 2012). Our results show that gene expression cascades initiated by Cut regulate a specific sub-cellular function, COPII transport, as one important means of mediating large-scale changes in cellular morphology. Recent studies in different cellular contexts indicate that transcription factors often organize and regulate complex cellular functions through streamlining gene expression cascades and subsets of cellular physiology. For example, PGC-1α, a master regulator that plays a crucial role in many physiological stimuli including oxidative stress, functions primarily through upregulating a battery of genes related to mitochondrial biogenesis and controlling several key metabolic hepatic pathways (Lehman et al., 2000; Lin et al., 2005; García-Giménez et al., 2011), whereas DIMM transcriptionally directs the regulated secretory pathway in neurons (Hamanaka et al., 2010).

CrebA has been shown to effect secretory function through direct binding to a consensus motif sequence in the enhancer regions of the secretory pathway control genes (Fox et al., 2010). We found a novel function for CrebA in dendrites where it is required to promote high order dendritic branching complexity by regulation of the COPII secretory pathway genes. Microarray analysis of CrebA targets identified more than 383 targets that encode not only general secretory machinery, but also other genes encoding secreted cargo cytoskeletal related proteins, ion channels, transporter proteins and receptors among others (Fox et al., 2010). Thus, it is possible that CrebA may function to influence dendrite morphology not only directly by regulating the secretory aspect, but also by altering the expression levels of other proteins that intrinsically effect dendrite shape. It is also interesting that CrebA has been shown to exert distinct effects through combinatorial interaction with tissue-specific transcription factors, resulting in upregulated expression of tissue-specific secreted cargo along with the general secretory machinery (Fox et al., 2010). Given that secretory cargo vary considerably between cell types, it is possible that Creb3-like transcription factors work cooperatively with tissue-specific factors to effect secretion of cell- or tissue-specific cargo (Murakami et al., 2009; Saito et al., 2009). This could be especially important for neuronal morphogenesis because the repertoire of transcription factors expressed in different neurons has been demonstrated to contribute to generation of unique morphologies (Parrish et al., 2006; de la Torre-Ubieta and Bonni, 2011).

Interestingly, we observed that the overexpression of Cut (Grueber et al., 2003), CrebA and Sec31 all cause an increase in dendritic branching density and growth in class I neurons, while causing a decrease in branching complexity in class IV neurons, indicative of cell-type-specific effects in distinct da neuron subclasses and suggesting that a common molecular program involving these proteins may be in effect to regulate class-specific dendrite morphology. In the case of class IV neurons, both loss-of-function and gain-of-function studies involving Cut, CrebA and Sec31 reveal similar phenotypic defects in dendritic growth and branching indicating that proper levels of these molecules are required for normal class-specific dendritic homeostasis such that over- or under-expression of these factors disrupts dendritogenesis in a similar manner. This type of effect is consistent with previous reports on Cut (Grueber et al., 2003) and Flamingo (Gao et al., 2000) whereby proper levels of these molecules are required for normal class-specific dendritogenesis.

In fact, the secretory pathway is not only a source of plasma membrane, but also affects protein composition in local domains (Eberwine et al., 2001; Raab-Graham et al., 2006). Thus disruption or alterations in the dendritic secretory pathway could possibly produce different effects in different classes of neurons, depending on altered micro-composition of proteins that are trafficked from these sites. Intriguingly, upregulation of a coat protein such as Sec31 or Sec23 would not necessarily be expected to be rate limiting for the formation of COPII vesicles; however, there is evidence to show that Sec31 directly interacts with Sar1 GTPase to stimulate the GTP hydrolysis rate, which in turn affects vesicle budding and size of the vesicle (Bi et al., 2007; Fromme et al., 2007; Fromme et al., 2008). Moreover, the Sec23–Sec24 complex functions as an adaptor protein that provides an additional level of dynamic control in cargo selection and recruitment (Gürkan et al., 2006). In addition, cargo composition can modulate the rate of coat assembly and budding, thereby affecting the rate of COPII vesicle formation from the ER (Aridor et al., 1999). Thus elevated levels of Sec31 or Sec23 alone may be capable of influencing the rate of COPII vesicle formation, which in turn could modulate dendrite development.

Moreover, Golgi outpost function is required for dendritic branch dynamics where disruption of secretory function results in defective extension as well as aberrant retraction (Ye et al., 2007), suggesting that secretory trafficking may be important not only in growth, but also in maintenance of branch stability. Golgi outposts were found to be enriched in the terminal branches and branch points of class IV da neurons (Ye et al., 2007), which could explain why we found that alterations in ER to Golgi transport mainly affected the formation of new branches while leaving the primary branches relatively unaffected. Moreover, we demonstrate that Cut-mediated increases in dendritic branching complexity of class I neurons is accompanied by a significant increase in the numbers of satellite ER and Golgi, where again outposts are highly coincident with branch points, suggesting that formation of new branch sites may be linked to these satellite secretory outposts. This potentially explains why the relatively simple WT class I da neurons, with little or no higher order dendritic branching, exhibit different requirements for the secretory pathway. Alternatively class I neurons, because of their relatively simple dendritic branching architecture, may complete dendritogenesis more quickly than complex da neuron subclasses and as such the hypomorphic sec31 mutants may produce sufficient Sec31 levels to enable class I dendrite development, but these levels are insufficient for proper class II–IV dendritogenesis.

Coordination between cytoskeletal dynamics and membrane dynamics is emerging as an important theme in various biological processes, and has been demonstrated in endocytosis as well as cell membrane invagination (Itoh et al., 2005; Mattila et al., 2007; Tsujita et al., 2006; Duleh and Welch, 2010). Dendritic development involves dynamic rearrangements of the actin and microtubule cytoskeleton, which provide the framework necessary for directional intracellular motility. Local control of ER complexity has also been shown to spatially control secretory trafficking within elaborate dendritic arbors by acting as focal points of ER export and sites of new dendritic branch formation (Cui-Wang et al., 2012). In addition, the ER–Golgi bi-directional secretory trafficking network interacts closely with the actin cytoskeleton. A recent study has uncovered an Arp2/3 nucleation promoting factor that also binds to microtubules and functions in ER to Golgi transport (Campellone et al., 2008). Our results demonstrate that upregulating the COPII secretory machinery leads to de novo formation of dendritic branches as well as increases in dendritic ER and Golgi outposts primarily to branch points and termini. Previous evidence in da neurons has shown that branch points and termini are primarily actin-rich processes and are subject to regulation by Rac1 (Lee et al., 2003; Andersen et al., 2005). Moreover, previous studies have demonstrated Cut acts through the RhoGEF Trio and Rac1 and Rho1 in regulating da neuron dendritic complexity (Iyer et al., 2012; Jinushi-Nakao et al., 2007). Thus, transcriptional control of ER–Golgi distribution as well as cytoskeletal dynamics may be one of the conserved mechanisms used to modulate dramatic dendritic arbor changes.

The following fly strains were obtained from Bloomington Stock Center. Deficiency stocks used for complementation analyses and P-element mapping of sec31⁰⁸⁰⁵ were Df(2R)ED1742/SM6a, Df(2R)ED1770/SM6a, Df(2R)ED1791/SM6a, Df(2R)Exel7098/CyO and sec31^c02461/CyO. P{PZ}CrebA⁰³⁵⁷⁶/TM3-Sb¹,Ser¹; P{UAS-CrebA.R}Tr11;CrebA^rR9/TMB6 Tb¹; UAS-sec31^EY19759; UAS-sec23^EY06757; (UAS-RNAi lines: cut^HMS00924,CrebA^JF27648; sec13^HMS00468; sec23^HMS00356; sec31^HMS00666; Sar1^HMS00355); nompC-GAL4 (Petersen and Stowers, 2011); Gal4^109(2)80,UAS-mCD8::GFP; Gal4⁴⁷⁷,UASmCD8::GFP; w,elavC155-Gal4,UASmCD8GFP,hsFLP; FRT^G13,tubP-GAL80. A second UAS-RNAi line for each of the following genes was obtained from Vienna Drosophila RNAi Center (VDRC) (CrebA^KK110650, sec31^GD35867, sec13^KK110428, sec23^GD24452, sec24^KK107154 and Sar1^KK108458) (Dietzl et al., 2007). Additional stocks included: UAS-CrebA (Abrams and Andrew, 2005); UAS-sec31; UAS-Sten-GFP (Förster et al., 2010); UAS-ManII-eGFP (Ye et al., 2007); Gal4^ppk.1.9,UASmCD8::GFP (Grueber et al., 2007); ppk-Gal4,UASmCD8::GFP; ppk-Gal4,UASmCD8::RFP; UAS-Cut; Gal4²²¹,UASmCD8::GFP; Gal4²²¹,UASmCD8::RFP; yw,tub-Gal80,hsFLP,FRT^19A; Gal4^109(2)80,UAS-mCD8::GFP; w,ct^c145,FRT^19A/y⁺,ct⁺,Y (Grueber et al., 2003); Gal4^21-7,UAS-mCD8::GFP (Song et al., 2007); ppk-Gal4,UAS-mCD8::GFP,ppk-Gal80 (labels class III da neurons); ppk-Gal80 (Yang et al., 2009). Oregon-R was used as a wild-type strain.

Ethyl methanesulfonate (EMS) mutagenesis and live imaging were carried out as described by Gao et al. (Gao et al., 1999). Late stage 17 mutant embryos were screened and select homozygous lethal mutations producing strong dendritic phenotypes were recombination mapped using the S¹, wg^Sp-1, amos^Tft, nw^B, Pin^Yt line, followed by chromosomal deficiency mapping and P-element complementation analyses using recessive lethal mutant alleles in known genes within the smallest non-complementing deficiency. For MARCM studies, the sec31⁰⁸⁰⁵ allele was recombined onto the FRT^G13 chromosome and analyses were performed as previously described (Sulkowski et al., 2011). Third instar larvae with GFP-labeled neurons were subjected to live-cell confocal microscopy. The same procedure was used for generating ct^c145 MARCM clones.

Dissection, staining, mounting and confocal imaging of third instar larval filets were performed as previously described (Sulkowski et al., 2011). Primary antibodies used in these studies included: rabbit anti-Sec31 (1:1000; a gift from Dr F. Gorelick, Yale University), mouse anti-CD8 (1:100; Life Technologies, Carlsbad, CA, USA) and rabbit anti-CrebA (1:200; a gift from Dr D. Andrews, Johns Hopkins University), anti-HRP (1:200; Jackson Immunoresearch, West Grove, PA, USA), mouse anti-cut (1:100; DSHB, Iowa City, Iowa, USA), rabbit anti-GM130 (1:100; Abcam, Cambridge, MA), chicken anti-GFP (1:1000; Abcam, Cambridge, MA). Donkey anti-rabbit, anti-mouse and anti-chicken secondary antibodies (Jackson Immunoresearch) were used at 1:200. IHC slides were then mounted in 70% glycerol:PBS and imaged at room temperature on a Nikon C1 Plus confocal system with either a 20× (0.75 NA), 40× (1.3 NA) or 60× (1.4 NA) oil immersion objective. For live confocal imaging, third instar larvae were immersed in a few drops of 1:5 (v/v) diethyl ether:halocarbon oil. Three dimensional z-stacks were then volume rendered into a two-dimensional maximum projection and resultant images were processed for quantitative neuronal reconstruction analyses. Image processing subsequent to data acquisition was performed using Adobe Photoshop 6.0.

Quantitative analyses of neuronal reconstructions were performed as previously described (Iyer et al., 2012). Quantification of dendritic arbor complexity was performed by importing confocal images into Neuromantic to generate the reconstruction file, which was then imported into L-Measure for assessment of parameters including total dendritic length, number of terminals and dendritic branch order. Statistical analyses of neurometric quantitative data, including means ± s.d., were performed by importing data into SigmaPlot (Systat Software) in which pairwise Student's t-tests were conducted to determine statistical significance. The branch order distributions were computed from whole neuron reconstructions using the branch order function of L-Measure and fitted curves were generated using the Weibull continuous probability distribution function. Analyses of protein expression levels were performed by quantifying pixel intensity using the polygon method as previously described (Sulkowski et al., 2011). For analysis of class IV neurons, not generated by MARCM, 200×300 µm boxes were placed over representative neuronal images (longer side oriented dorsoventrally), where one corner of the box coincided with the cell body of the neuron. One to two boxed regions for each representative neuron were reconstructed and analysis performed by comparison with parallel regions from control class IV neurons.

Neuronal subpopulations were isolated using magnetic-bead-based cell sorting as previously described (Iyer et al., 2009). Following RNA purification from isolated cells, qRT-PCR and data analyses were performed as previously described (Sulkowski et al., 2011; Iyer et al., 2012). The following pre-validated Qiagen QuantiTect Primer Assays (Qiagen, Germantown, MD, USA) were used: CrebA (QT00964901), cut (QT00501389), Sar1 (QT00979790), fkh (QT00984508), sec13 (QT00501263), sec23 (QT00968877), sec24 (QT00943243), and sec31 (QT00942298). Expression data were normalized using primers for GAPDH2 (QT00922957) and RpL32 (QT00985677) and are reported as means ± s.d.

We thank Y. N. Jan, W. Grueber, D. Andrew, F. S. Gorelick and S. Luschnig for providing invaluable reagents used in these studies. We also thank the TRiP at Harvard Medical School (NIH/NIGMS R01-GM084947), Bloomington, and VDRC for providing RNAi transgenes. We thank L. Sullivan, M. Garland, W. Osman, V. Thota, S. Prakash and Y. Lau for assistance with neural reconstructions.