Identification of raw as a regulator of glial development

Glial cells perform numerous functions to support neuron development and function, including axon wrapping, formation of the blood brain barrier, and enhancement of synaptic transmission. We have identified a novel gene, raw, which functions in glia of the central and peripheral nervous systems in Drosophila. Reducing Raw levels in glia results in morphological defects in the brain and ventral nerve cord, as well as defects in neuron function, as revealed by decreased locomotion in crawling assays. Examination of the number of glia along peripheral nerves reveals a reduction in glial number upon raw knockdown. The reduced number of glia along peripheral nerves occurs as a result of decreased glial proliferation. As Raw has been shown to negatively regulate Jun N-terminal kinase (JNK) signaling in other developmental contexts, we examined the expression of a JNK reporter and the downstream JNK target, matrix metalloproteinase 1 (mmp1), and found that raw knockdown results in increased reporter activity and Mmp1 levels. These results are consistent with previous studies showing increased Mmp levels lead to nerve cord defects similar to those observed upon raw knockdown. In addition, knockdown of puckered, a negative feedback regulator of JNK signaling, also causes a decrease in glial number. Thus, our studies have resulted in the identification of a new regulator of gliogenesis, and demonstrate that increased JNK signaling negatively impacts glial development.


Introduction
The precise interaction between neurons and glia is essential for the establishment and maintenance of nervous system structure and function. A failure of neurons and glia to interact properly or a breakdown in interactions can result in neuropathy or neurodegeneration [1][2][3]. Neurons play a critical role in sensing environmental stimuli and conveying responses to these stimuli. However, their function depends on their close association with different types of glia. Glia perform a variety of roles in the nervous system, from regulating blood brain barrier (BBB) formation and synapse structure to insulating neurons, clearing debris, and providing trophic support (reviewed in [4]).
Drosophila melanogaster has proven a fruitful organism for studying the roles of different glial subtypes and the molecular mechanisms that regulate their development and function, a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 and morphogenesis of the salivary glands, hindgut, and malpighian tubules during embryogenesis, as well as larval dendritic patterning [18][19][20]. In addition, Raw is necessary for the ensheathment of germ cells (GCs) by somatic gonadal precursor cells (SGPs) in the developing gonad, a process required for proper GC proliferation and gene expression [21][22][23]. Structurally, Raw is a transmembrane protein with an N-terminal extracellular domain and a C-terminal intracellular domain that lacks additional domains [19]. Analysis of the molecular context of Raw function during embryogenesis revealed that Raw negatively regulates Jun N-terminal Kinase (JNK) signaling and positively regulates cell adhesion through Drosophila E-Cadherin [20,22,24]. More recent studies have shown that Raw also interacts with the Nuclear dbf2-related (NDR) family kinase, Tricornered (Trc), in dendrites [19].
The requirement of Raw for the ensheathment of GCs by SGPs in the developing gonad and the previously described role of Raw in embryonic nervous system development led us to ask if Raw may play a specific role in glia to promote nervous system development. In addition, we were curious if Raw may function as a negative regulator of JNK signaling in the glia as it does in other contexts. This study demonstrates that Raw is required for proper glial development and the establishment of nervous system structure and function. Reducing raw levels in glia results in abnormal brain and VNC structure, as well as a reduced number of glia along peripheral nerves. Further exploration of these phenotypes reveals that Raw is required for glial proliferation, and negatively regulates JNK signaling to promote proper glial development.
In order to quantitate relative changes in Mmp1 and TRE-GFP reporter levels, average pixel intensity was measured over a region of the VNC and over neighboring muscle tissue, which is not expected to undergo any changes in Mmp1 or GFP reporter levels when raw levels are altered in glia. Average pixel intensity in the VNC was normalized to average pixel intensity in the muscle. These values were graphed and the change relative to controls was determined. In the case of Mmp1 staining, data was analyzed using GraphPad Prism 7 by oneway ANOVA with post-hoc Tukey test. In the case of the TRE-GFP reporter, data from the control versus the raw knockdown sample was analyzed using GraphPad Prism 7 by the Mann-Whitney test.

Morphological analysis
For morphological analysis of the VNC, wandering third instar larvae were imaged using a Zeiss SteREO Discovery.V12 or a Zeiss Axio Zoom.V16 microscope, and their length was measured using Zeiss Zen software. The brain and VNC were dissected in 1x PBS and imaged on a Zeiss AxioImager.M2 or Olympus DSU spinning disc microscope, and the length of the VNC was measured using Zeiss Zen or Olympus Metamorph software, respectively. The length of the VNC relative to body length was calculated for each sample, as previously described, given that VNC length has been demonstrated to scale with overall body length in other insects [32,33]. All data was analyzed using GraphPad Prism 7 by one-way ANOVA with post-hoc Tukey test to assess statistical significance.

Locomotor analysis
Locomotor function of third instar larvae was assessed by a crawling assay. Third instar larvae were placed on a slab of 2% agarose brought to a temperature of 25˚C. The agarose was placed on a piece of plexiglass and dropped from a distance of~1-2 inches to stimulate larval movement. Larvae were imaged for 50-60 seconds with an Apple iPhone Camera. Videos were converted to avi format by ffmpeg. The resulting avi videos were analyzed from the 15-45 second mark using ImageJ software with the wrmtrck plugin [34]. Specifications were adjusted for tracking of Drosophila larvae. Data on path length and average speed was analyzed by Mann-Whitney test using GraphPad Prizm 7 Software.

raw knockdown results in reduced glial cell number
Strong hypomorphic alleles of raw lead to defects in embryonic nervous system development and are embryonic lethal [18,21]. Therefore, to further examine Raw function in the developing nervous system beyond embryogenesis, the GAL4/UAS system was utilized to knockdown raw levels via RNA interference (RNAi). Previous studies have demonstrated that reversed polarity (repo) is expressed in all glia with the exception of the midline glia [35]. Therefore, repo-Gal4 was crossed to two UAS-raw-RNAi lines with overlapping raw target regions, one inserted on the second chromosome and one inserted on the third chromosome, referred to as UAS-raw-RNAi 2 and UAS-raw-RNAi 3 , respectively [36,37]. In addition, raw levels were reduced specifically in neurons, using elav-Gal4. raw knockdown in neither neurons nor glia affected viability. However, the co-expression of UAS-dcr2 in the presence of UAS-raw-RNAi 2 in glia led to pupal lethality, suggesting a critical requirement for Raw in these cells.
To explore how Raw functions in glia during nervous system development, immunohistochemistry was performed on dissected third instar larvae and glial nuclei were visualized using the anti-Repo antibody. Immunostaining revealed a striking decrease in the number of glial nuclei along the NER of peripheral nerves upon raw knockdown. Therefore, the number of glial nuclei along the NER of the A8/9 peripheral nerves was counted in controls and raw knockdown larvae (Fig 2). While control larvae (repo-Gal4 and repo-Gal4>dcr2) had~60-65 nuclei/nerve, this number was reduced to~35-40 glia/nerve upon raw knockdown (repo-Gal4>raw-RNAi) using two different UAS-raw-RNAi lines (Fig 2). This number was further reduced to an average of less than 10 glia/nerve upon raw knockdown in the presence of overexpressed dcr2 (repo-Gal4>raw-RNAi 2 , dcr2 flies) (Fig 2). While use of the UAS-raw-RNAi 3 line had previously been reported [36,37], the UAS-raw-RNAi 2 line had not been previously used in published work. Therefore, the specificity of this RNAi line was confirmed by testing the ability of overexpressed raw to rescue the repo-Gal4>raw-RNAi 2 phenotype. Co-expression of raw significantly rescued the raw knockdown phenotype, increasing the average number of glial nuclei along the A8/9 nerve from 37 to 53 (Fig 2G and 2I). Overexpression of raw did not result in a significant change in glial number as compared to controls (repo-Gal4) ( Fig  2F). Thus, our results demonstrate that raw is required for the population of nerves with the appropriate number of glial cells during development.

Raw is required for glial proliferation and to prevent cell death along peripheral nerves
As raw knockdown results in a reduced number of glia compared to controls, we hypothesized that reduced glial number could result from decreased proliferation or increased cell death along peripheral nerves in third instar larvae. Cell death was examined in raw knockdown and control larvae using an antibody directed against the activated form of the Drosophila caspase, Death caspase-1 (Dcp-1). While apoptosis of midline glia has been observed during embryogenesis and metamorphosis, apoptosis of peripheral glia has not previously been reported [38][39][40][41]. Therefore. we did not expect to observe glial cell death along nerves in controls. Consistent with this expectation, activated Dcp-1 was not observed in repo-Gal4 controls and cell death was observed in only one out of nine repo-Gal4>dcr2 samples. In contrast, raw knockdown in the presence of overexpressed dcr2 (repo-Gal4>raw-RNAi, dcr2 flies) resulted in the presence of Dcp-1 immunostaining along a subset of peripheral nerves in eight out of nine samples analyzed (Fig 3). This cell death only occasionally localized around Repo-positive nuclei, suggesting that reduced levels of raw in glia can lead to neuronal cell death. Surprisingly, cell death was not observed along nerves in repo-Gal4>raw-RNAi larvae (n = 10). Thus, while cell death may contribute to the decreased number of glia in third instar larvae, this does not appear to be the predominant reason for the reduction in glial number.

raw knockdown affects CNS morphology
In addition to defects in glial number, raw knockdown in the presence of overexpressed dcr2 (repo-Gal4>raw-RNAi, dcr2) also caused significant defects in brain and VNC morphology. Specifically, the brain appeared smaller and the VNC appeared thin and elongated (Fig 5A-5D). Therefore, VNC length was quantified relative to total body length, as previously described, to account for age-related differences in size [33]. Quantitation revealed that nerve cords of the repo-Gal4 control larvae were an average of 10% of total body length ( Fig 5E). Reducing raw levels by RNAi resulted in VNC lengths that were~12% of total body length and this went up to~14% of total body length when raw was knocked down in the presence of overexpressed dcr2 (Fig 5E). Overexpression of dcr2 in glia alone did not cause a significant change in the relative length of the VNC (~10% of total body length; Fig 5E). These results demonstrate a requirement for Raw in the morphological changes of the VNC during development. We also observed significant morphological differences in the brains of repo-Gal4>raw-RNAi, dcr2 larvae compared to controls. In order to determine if brain size was reduced, the diameter of the brain was measured. A significant reduction in brain diameter was observed in repo-Gal4>raw-RNAi, dcr2 relative to repo-Gal4>dcr2 alone (d = 0.224mm versus d = 0.276mm, p<0.005), but repo-Gal4>raw-RNAi larvae did not exhibit a significant change in brain size relative to repo>Gal4 controls (d = 0.245mm versus d = 0.245mm, not significant). These results suggest that raw knockdown results in significant defects in the establishment of proper nervous system morphology.

raw knockdown results in increased JNK signaling
Previous work has shown that increased levels of Matrix metalloproteinase 2 (Mmp2) in glia cause defects in VNC morphology, similar to the defects we observed upon raw knockdown [33]. Given that Raw has previously been demonstrated to negatively regulate JNK signaling in other contexts, and activation of JNK signaling results in increased transcription of matrix metalloproteinase 1 (mmp1) [22,24,43], we hypothesized that raw knockdown may lead to increased Mmp1 levels. In the peripheral nerves, Mmp1-positive punctae were observed upon raw knockdown in the presence of overexpressed dcr2, but were absent from controls, demonstrating that Mmp1 localizes differently in nerves upon raw knockdown (Fig 6). In the VNC, immunostaining for Mmp1 upon raw knockdown revealed increased Mmp1 levels in the VNC relative to controls (Fig 6A‴ and 6C‴). These levels appeared further increased when raw was knocked down in the presence of overexpressed dcr2 (Fig 6A‴-6D‴). Quantitation of normalized Mmp1 levels revealed significantly higher Mmp1 levels in the VNC of raw knockdown (repo-Gal4>raw-RNAi 2 , dcr2) samples compared to controls (repo-Gal4 and repo-Gal4>dcr2) (Fig 6E). From these results we conclude that raw knockdown leads increased JNK signaling and upregulation of the JNK target, mmp1.  While increased Mmp1 levels in raw knockdown samples suggests increased levels of JNK signaling, this increase could also occur indirectly through a different signaling pathway. Therefore, to demonstrate more directly that reducing raw levels increases JNK signaling, a JNK reporter (TRE-GFP) containing four binding sites for AP-1 (a heterodimeric Jun/Fos transcription factor targeted by JNK) upstream of GFP was utilized to examine JNK activity [27,44,45]. While GFP expression in the VNC and peripheral nerves was low in control samples (repo-Gal4>dcr2), a significant increase in GFP expression was observed when raw was knocked down in the presence of overexpressed dcr2 (repo-Gal4>raw-RNAi 2 , dcr2) (Fig 7A-7D"). Quantitation of normalized expression levels revealed significantly higher GFP expression in the VNC in raw knockdown (repo-Gal4>raw-RNAi 2 , dcr2) samples as compared to the control (repo-Gal4>dcr2) (Fig 7E). From these results we conclude that raw knockdown leads to increased in JNK signaling and upregulation of downstream transcriptional targets.

Increased JNK signaling results in reduced glial number
As JNK signaling is increased upon raw knockdown and glial cell number is reduced along peripheral nerves, it was hypothesized that simply increasing JNK signaling in glia could reduce glial number. Puckered (Puc) is a negative feedback regulator of JNK. Therefore, puc levels were reduced by RNAi to increase JNK signaling, and the number of glia along the A8/9 peripheral nerve was determined. Knockdown of puc resulted in an average of~30 glia along the peripheral A8/9 nerve as compared to~63 glia in the repo-Gal4 control (Fig 2H and 2I). These results suggest that increased JNK signaling negatively impacts glial cell number, and that that effect of raw knockdown on glial number could occur as a result of a failure of Raw to inhibit JNK signaling in glia.

raw knockdown in glia affects locomotor function
The defects observed in glial number and nervous system morphology raised the possibility that overall nervous system function may be disrupted. In order to examine nervous system function, larval crawling assays were performed. The average speed and distance traveled by larvae over a period of 30 seconds were analyzed. Control larvae (repo-Gal4) crawled at an average speed of .11cm/s, whereas repo>raw-RNAi larvae crawled at an average speed of 0.066cm/s (Fig 8A; S1 and S2 Movies). The average distance covered by controls was also significantly greater at 3.34cm, as compared to 1.99cm for repo>raw-RNAi larvae (Fig 8B). These results demonstrate that reduced raw levels not only affect glial cell number, but also result in impaired locomotor activity.

Discussion
Our studies have resulted in the identification of Raw as a regulator of glial development. We show that reducing raw levels in glia affects the morphology of the CNS, resulting in a smaller brain and an elongated VNC. In addition, examination of glial cell number along the peripheral nerves revealed that raw knockdown led to a reduction in the number of glia along the A8/9 peripheral nerves, which is due in part to reduced glial proliferation. Previous studies demonstrated that Raw functions as a negative regulator of JNK signaling in other developmental contexts. Therefore, we examined the expression of a JNK reporter and Mmp1, a knockdown of raw in the presence of overexpressed dcr2 results in further VNC elongation relative to total body length (repo-Gal4>mRFP, raw-RNAi 2 , dcr2; n = 15). (E) Quantitation of ventral nerve cord length as a percentage of total body length. Scale bars are 100μm. Ã p < .005 and ÃÃ p<0.0001 based on one-way ANOVA with Tukey's post hoc test. https://doi.org/10.1371/journal.pone.0198161.g005 Regulation of glial development of third instar larvae stained for Mmp1. All images acquired with 60x oil objective with same confocal settings. Scale downstream target of JNK signaling. We observed increased reporter expression and Mmp1 levels in the VNC and increased reporter expression and Mmp1-positive punctae along peripheral nerves upon raw knockdown in the presence of overexpressed dcr2. Supporting the hypothesis that increased JNK signaling can lead to defects in glial development, knockdown of puc, a negative feedback regulator of the JNK pathway, also resulted in a decrease in the number of glia along peripheral nerves. Finally, larvae with reduced raw levels had reduced crawling ability, suggesting that reduced glial number negatively impacts motor neuron function. These results suggest that Raw functions as a negative regulator of JNK signaling in the context of glial development, thereby affecting glial number and leading to morphological defects that impair nervous system function.

Molecular role of Raw and JNK signaling in gliogenesis
Embryonic studies have provided insight into how Raw regulates JNK signaling, although our understanding remains far from complete. Recent work revealed that Raw regulates expression of the long non-coding RNA acal, which functions to negatively regulate expression of Connector of kinase to AP-1 (Cka), a protein that acts as a scaffold for JNK and Drosophila Jun, known as Jun-related antigen (Jra) [46,47]. These results are consistent with earlier studies, suggesting that Raw functions at the level of the JNK, Basket (Bsk), and limits accumulation of Jra in the nucleus [20,22,24]. Thus, Raw may function to reduce efficiency of JNK signaling by limiting scaffold production in glia as well.
While previous studies have demonstrated that JNK signaling is upregulated in glia in response to neuronal injury and plays a role in promoting glial phagocytosis of neurons during development, the role of JNK signaling in gliogenesis has been largely unexplored [48][49][50]. Studies in the embryo showed that mutation of zinc finger homeodomain 1 (zfh1) promoted apoptosis of ePG10 in a JNK-dependent manner, although Zfh1 was expressed in all ePGs [51]. Our results demonstrate that raw knockdown increases expression of a JNK reporter and the JNK target, mmp1, and that increased JNK signaling in glial cells results in a significant decrease in the number of glia along peripheral nerves, similar to that observed upon raw knockdown. These results suggest that it is altered JNK signaling that is responsible for the reduction in glial number observed upon raw knockdown rather than regulation of another signaling pathway.
However, Raw may also function with other molecular players to regulate glial development, as JNK signaling is not the only pathway with which Raw has been shown to interact. In class IV dendritic arborization neurons, Raw appears to function independently of JNK signaling and instead cooperates with Trc, a NDR family kinase, to regulate dendritic patterning [19]. NDR family kinases include Trc and Warts (Wts), which each belong to different NDR kinase subfamilies, and have been demonstrated to function downstream of the protein kinase Hippo (Hpo) [52,53]. Wts negatively regulates activity of the transcription factor Yorkie (Yki), and thereby limits cell proliferation [52,54]. Interestingly, activation of Yki is required for glial proliferation in the context of the brain and developing eye in Drosophila [55]. In contrast to Wts, Trc appears to function through downstream effectors other than Yki [52,54]. In fact, the mammalian homolog of Trc, Ndr1, promotes progression through the cell cycle via  phosphorylation of the cell cycle inhibitory protein p21 [56]. These results suggest that Wts and Trc have distinct functions downstream of Hpo. Examination of the roles of Wts and Trc in dendrites supports this conclusion, as Trc regulates dendritic tiling, while Wts functions in dendritic maintenance [53]. Studies in the context of dendrite morphogenesis suggest Trc functions via Rac GTPase to regulate the actin cytoskeleton, suggesting a mechanism by which Raw and Trc could function to promote changes in nervous system morphology during development [57]. Given the proliferation defects observed upon raw knockdown, and the previously characterized interaction of Raw and Trc, it raises the possibility that Raw may be functioning in cooperation with NDR family kinases to regulate glial proliferation and nervous system morphology.
While the above results suggest Raw may function in the context of Hpo signaling, it is important to note that the Hpo and JNK signaling pathways have also been observed to regulate each other's function. In Drosophila models of tumorigenesis, activation of Hpo has been observed to increase JNK signaling to promote cell invasion [58], while JNK signaling has been demonstrated to enhance Yki activation by inhibition of Wts [59,60]. However, in normal epithelium, JNK signaling has been observed to inhibit Yki activation [60]. These results suggest the mechanisms by which these pathways influence each other is context-dependent. Given that Yki is a positive regulator of glial proliferation, the relationship between JNK signaling and Yki activation in normal epithelium would be consistent with a model in glia where raw knockdown results in increased JNK signaling, leading to a decrease in Yki activation and reduced glial proliferation. [55].

Regulation of glial development and function
In addition to understanding the molecular context of Raw function, it is also important to identify the specific glial subtypes that require Raw activity, and to understand the role of this activity in cell-cell and cell-matrix interactions. Previous work demonstrated that expression Regulation of glial development of Mmp2 in all glia (excluding the midline glia) or the perineurial glia can lead to an elongated VNC, while panglial overexpression of Mmp1 was larval lethal [12,33]. Our findings that raw knockdown results in increased levels of Mmp1 and an elongated VNC are consistent with a potential requirement for Raw function in mediating the interaction of perineurial glia with the surrounding neural lamella. Raw has been demonstrated to regulate dendrite-ECM adhesion in cooperation with Trc, again suggesting a similar regulatory mechanism could exist in glia [19]. In mammals, loss of cell-ECM interactions results in increased Hpo signaling and inhibition of YAP, the mammalian homolog of Yki [61]. Thus, the increase in Mmp1 levels observed upon raw knockdown could reduce glia-matrix interactions, leading to reduced perineurial glial proliferation. Future exploration of the relationship of Raw to components of the Hpo signaling pathways is critical for our understanding of how Raw regulates glial development.
While these data suggest a cell autonomous role for Raw in perineurial glia, is does not exclude the possibility that Raw could function cell non-autonomously to regulate perineurial glia proliferation, or that Raw may have additional cell autonomous functions in the subperineurial and wrapping glia. The fact that larvae have motor defects also supports the possibility that Raw functions in the glia that are more tightly associated with motor neurons to ensure their proper function. Thus, knockdown of raw in each glial subtype is critical for addressing these possibilities. Examination of the ability of Raw to interact with other signaling pathways during development will increase our understanding of the mechanisms regulating gliogenesis, and may also provide insight into the underlying causes of neuropathy and neurodegenerative disease.