Drosophila ßHeavy-Spectrin is required in polarized ensheathing glia that form a diffusion-barrier around the neuropil

In the central nervous system (CNS), functional tasks are often allocated to distinct compartments. This is also evident in the Drosophila CNS where synapses and dendrites are clustered in distinct neuropil regions. The neuropil is separated from neuronal cell bodies by ensheathing glia, which as we show using dye injection experiments, contribute to the formation of an internal diffusion barrier. We find that ensheathing glia are polarized with a basolateral plasma membrane rich in phosphatidylinositol-(3,4,5)-triphosphate (PIP3) and the Na+/K+-ATPase Nervana2 (Nrv2) that abuts an extracellular matrix formed at neuropil-cortex interface. The apical plasma membrane is facing the neuropil and is rich in phosphatidylinositol-(4,5)-bisphosphate (PIP2) that is supported by a sub-membranous ßHeavy-Spectrin cytoskeleton. ßHeavy-spectrin mutant larvae affect ensheathing glial cell polarity with delocalized PIP2 and Nrv2 and exhibit an abnormal locomotion which is similarly shown by ensheathing glia ablated larvae. Thus, polarized glia compartmentalizes the brain and is essential for proper nervous system function. In the invertebrate CNS, synapses and dendrites are clustered in distinct neuropil areas that are separated from neuronal cell bodies by ensheathing glia (EG). Here, the authors show that Drosophila EG are polarized cells that form an internal diffusion barrier. EG cell polarity requires βHeavy-Spectrin and is needed for normal locomotor behaviour.

T he complexity of the nervous system is not only defined by intricate neuronal networks but is also reflected by the cellular diversity of the different glial cells found throughout the nervous system. The number of functions attributed to the different glial cells is steadily growing and ranges from guiding and instructing functions during early brain development to synapse formation and plasticity, and general ion and metabolite homeostasis 1,2 .
One of the key functions of glial cells lies in the establishment of boundaries that help compartmentalizing neuronal computing. A very tight boundary is the blood-brain barrier. In all invertebrates as well as in primitive vertebrates this barrier is established by glial cells, whereas higher vertebrates transferred this function to endothelial cells [3][4][5] . Smaller signaling compartments are established by astrocytes that tile the synaptic areas in the brain [6][7][8] . Oligodendrocytes insulate single segments of up to 20 different axons by forming myelin sheets around them 9 . Moreover, oligodendrocyte progenitor cells act as innate immune cells and participate in scar formation separating intact brain tissue from the injured site 10 .
Invertebrates have a less complex nervous system and glial cells are clearly outnumbered by neurons. In the ant Harpegnathos saltator 27% of all brain cells are of glial nature, whereas in the fruit fly Drosophila melanogaster just 10% of all neural cells are of glial nature 11,12 . As in primitive vertebrates, glial cells form the blood-brain barrier and organize the entire metabolite traffic into and out of the nervous system and also integrate external stimuli such as circadian rhythms with general brain functions [13][14][15][16] . All neuronal cell bodies are situated beneath the blood-brain barrier and are surrounded by cortex glial cells. These cells regulate neuroblast division, nurture neurons, and phagocytose dying neurons [17][18][19][20][21][22][23][24] .
Fly neurons project their dendrites and axons into the central neuropil. This morphologically distinct structure is separated from all neuronal cell bodies by the neuropil-associated glia, which comprises astrocyte-like glial cells and ensheathing glial cells [25][26][27][28][29] . Astrocyte-like cells form numerous fine and highly branched processes that infiltrate the neuropil. As in vertebrates they tile the synaptic neuropil into discrete units. They express a number of neurotransmitter transporters to clear synaptic spillover although they do not form frequent intimate contacts with individual synapses and in addition are able to secrete gliotransmitters [29][30][31] .
The neuropil encasing ensheathing glia comprises two distinct classes that can be classified by morphological as well as molecular criteria 11,27,28,32 . In the larval nervous system, just four ensheathing glial cells are formed in each hemineuromer. Two of them embrace the neuropil and do not wrap around individual axons. In contrast, the other two larval ensheathing glial cells found in each abdominal hemineuromere exhibit a more complex morphology. In part they encase the neuropil as the other two ensheathing glial cells but in addition they also enwrap axons between the CNS/PNS boundary and the neuropil. They are thus called ensheathing/wrapping glial cells 27,28 . In the larva, the organization of the neuropil is relatively simple, but the adult CNS shows complex regional compartmentalization 33,34 . This is reflected by an increase in the number of ensheathing glial cells that are generated during pupal stages and, as in larval stages, fall into two morphological classes 32,35 .
Several studies have already shed some light on the functional roles of ensheathing glia in the fly nervous system. First, it was demonstrated that ensheathing glial cells remove neuronal debris after injury utilizing the Draper pathway 24,36,37 . In addition to this immune and surveillance function, ensheathing glia can participate in neuronal signaling. Mutant analysis indicates the sulfite oxidase Shopper needs to be expressed by ensheathing glia to regulate glutamate homeostasis in the neuropil 28 . Likewise, the Excitatory amino acid transporter 2 (Eaat2) functions in ensheathing glia to modulate sleep in the adult 38 . In addition, cell type specific knockdown experiments demonstrate that the voltage-gated potassium channel encoded by the gene seizure (sei) is required in ensheathing glia to protect flies from acute heatinduced seizures 39 .
Given the apparent role of ensheathing glia in forming neuronal compartments and modulating their function, we initiated a comprehensive analysis of this-as we found-highly polarized cell type. The plasma membrane of the ensheathing glia facing the neuropil is rich in the phospholipid PIP 2 and is supported by submembranous ß Heavy -Spectrin. Thus, it can be assumed that the apical cell domain is oriented towards the neuropil. The basolateral plasma membrane exhibits an accumulation of the Na + /K + ATPase Nervana 2 (Nrv2) suggesting that sodium ion dependent transport is polarized in these cells. In addition, basally localized Integrins anchor the ensheathing glia to a specific extracellular matrix (ECM) formed at the interface of neuropil and CNS cortex. To study the functional relevance of ensheathing glia, we ablated these cells using a split Gal4 driver that we had established. The absence of ensheathing glia triggers a compensatory growth of astrocyte-like cells which, however, does not completely restore an internal diffusion barrier normally promoted by ensheathing glia. Animals lacking ensheathing glia show abnormal larval locomotion and reduced longevity. Reduction of ß Heavy -Spectrin expression affects ensheathing glia cell morphology and leads to mislocalization of PIP 2 . Moreover, larvae with ensheathing glia specific ß Heavy -spectrin knockdown or larvae lacking ensheathing glia show similar abnormal locomotor phenotypes as ß Heavy -spectrin mutant larvae. In conclusion, we show that separation of the neuropil by polarized ensheathing glia is required for nervous system function.

Results
Ensheathing glia encase the neuropil throughout development. Four ensheathing glial cells are formed in each hemineuromere during mid embryogenesis and are associated with the neuropil 27 . Two of these cells only encase the neuropil, whereas the two other also wrap axons connecting the peripheral nerves with the neuropil (ensheathing/wrapping glial cells, Fig. 1) 27,32,40 . To target the larval ensheathing glial cells, we compared the activity of the previously used Gal4 lines (NP6520-Gal4, 56F03-Gal4, nrv2-Gal4, and 83E12-Gal4) and found the 83E12-Gal4 driver as the most specific one for both larval and adult CNS 27,28,32,41 . Cell counts and 3D reconstructions (see below) suggest that 83E12-Gal4 is active in all larval ensheathing glial cells. However, since no 3D reconstruction has been conducted for the adult TEM volume we cannot exclude that some 83E12-Gal4 negative ensheathing glial cells exist.
During embryonic stages, the ensheathing glia initially cover the dorsal domain of the neuropil (Fig. 1a, a'). In first instar larvae, thin processes of the ensheathing glia begin to encase the neuropil (Fig. 1b, b'). Notably, small gaps between individual ensheathing glial processes are still detectable in the second instar larva (Fig. 1c, c'). In the third larval instar, ensheathing glia appear to form a closed case around the neuropil. The ensheathing glia also send processes around the dorsally located neuronal cell bodies (Fig. 1d', Supplementary Fig. S1). Multi-color flipout (MCFO) labelling demonstrates that ensheathing glia processes tile the neuropil ( Fig. 1e-g, j) to separate it from the CNS cortex. The cortex glia can be visualized using the Gal4 drivers 55B12-Gal4; NP2222-Gal4 and NP0577-Gal4 28,42 . These drivers showed that no cortex glial cells reside at the dorsal surface of the ventral nerve cord (Fig. 1h-j). This notion is further  corroborated by split-GFP experiments where GFP is reconstituted at the baso-lateral boundary between cortex and neuropil but not at the dorsal surface of the ventral nerve cord ( Supplementary Fig. S1a-f'). Here, several large glutamatergic neurons are found that appear to be encased by processes of the ensheathing glia ( Supplementary Fig. S1g-l). In conclusion, the ensheathing glia can wrap axons as they connect with peripheral organs, associate with dorsal neurons, and encase the neuropil (Fig. 1j).
Electron microscopic analysis of larval ensheathing glia. To further study the morphological characteristics of the ensheathing glia, we analyzed serial section transmission electron microscopy (ssTEM) data sets of a first instar larval ventral nerve cord (L1) 43 . We annotated all neuropil-associated glial cells in the abdominal neuromeres A1-A8/9 using CATMAID 44 (Fig. 2). In total, 159 neuropil-associated glial cells were identified that based on their typical morphology could be assigned to one of the three classes of neuropil-associated glial cells 27 : astrocyte-like glial cells, ensheathing glia and ensheathing/wrapping glia (Fig. 2a, b). The morphological characteristics of the different glial subtypes are already evident in first larval instar and become more pronounced in third instar larval brain (Fig. 2c, d). Astrocyte-like glial cells were identified by a prominent process extending into the neuropil proximal to the astrocyte nucleus 29,31 . In the abdominal ventral nerve cord of the L1 volume, 96 astrocyte-like glial cells were identified matching the previously known numbers 27,29 .
In total 28 ensheathing glial cells were identified in the abdominal neuromeres 1-7 (Fig. 2a). These glial cells extend thin membrane sheaths along the face of the neuropil (Fig. 2e). The nuclei of the central ensheathing glial cell in each neuromere is usually associated with the dorso-ventral channel 25 . The other ensheathing glia is found at a ventral position at the neuropil cortex interface (see also 27 ).
In total 33 cells were annotated as ensheathing/wrapping glial cells, which encase the neuropil and wrap axons in the nerve root. However, this is not very pronounced in the first instar stage, yet (Fig. 2g). With one exception (Fig. 2b), two cells, both located at the dorsal aspect of the neuropil, are associated with each intersegmental nerve root in neuromeres 1-7 (Fig. 2b). The fused A8/A9 nerve has two nerve roots associated with two ensheathing/wrapping glial cells, each (Fig. 2b). To determine whether the overall morphology of ensheathing glial cells changes during development, we annotated the ensheathing glia in three abdominal neuromeres (A1-A3) of a third instar larval brain (L3) 45 . Here, too, two ensheathing and two ensheathing/wrapping glial cells are found in each hemineuromere (Fig. 2f, h). Note, that dorsally to the neuropil, ensheathing glia send protrusions to the blood-brain barrier and thus adopt a more complex polarized morphology (Fig. 2h). In consequence, dorsal neurons are encased by ensheathing glia processes and not by cortex glia (Fig. 1j, Fig. 2h, see below). The ventrolateral ensheathing glial cells do not encase neuronal cell bodies, but form very flat membrane sheets around the neuropil ( Fig. 1g-i, Fig. 2e, f).
The light microscopic analysis suggests that ensheathing glial cells encase the neuropil in third instar larvae. This idea is supported by the ssTEM data sets. In L1 ventral nerve cord ensheathing, ensheathing/wrapping and astrocyte-like glial cells form an almost closed structure around the neuropil (Fig. 2c). In contrast, in the L3 larval ventral nerve cord the neuropil encasement is complete (Fig. 2d). In addition, multiple layers of glial processes are found around the neuropil (Fig. 2i, j). At the ventral face of the neuropil, the glial sheath is often made of cell processes of ensheathing glial and cortex glial cells (Fig. 2i).
The complexity of ensheathing glia increases during development. About 112 ensheathing glial cells expressing 83E12-Gal4 can be detected in the third instar larva with only few ensheathing glial cells in the brain lobes (n = 5, with 110, 112, 112, 114, and 116 83E12-Gal4 positive cells, median = 112) (Fig. 3a, m), which corresponds to previously determined numbers 27,35,40,46,47 and matches the annotations made in the TEM volume mentioned above. During the subsequent pupal development, the number of the 83E12-Gal4 positive ensheathing glial cell population increases more than 1,100 ensheathing glial cells in the adult (ventral nerve cord and brain without optic lobes, number determined using Imaris) (Fig. 3c, n).
It is unclear whether all larval ensheathing glia degenerate or whether some survive to participate in the formation of adult ensheathing glia 35,46 . To test whether ensheathing glial cells are competent to divide, we expressed activated FGF-receptor which is known to trigger perineurial glial proliferation 48,49 . Interestingly, a proliferative response upon activated FGF-receptor expression is most prominently observed in thoracic neuromeres ( Fig. 3a-f, m, n, Supplementary Fig. S2). We then blocked cell division in the ensheathing glia by RNAi-based silencing of string, which encodes the Drosophila Cdc25 phosphatase homolog required for progression from G2 to M phase 50 . This expression regime did not severely affect ensheathing glia cell number in the larval CNS (Fig. 3g, h, m, 94 ensheathing glia in string knockdown larvae vs. 112 ensheathing glial cells in control animals, n = 5 with 83, 83, 110, 99, and 94 83E12-Gal4 positive cells, median = 94). In contrast, suppression of string function resulted in 38% less ensheathing glial cells in the adult nervous system (724 [n = 5, 724, 707, 769, 633, and 848 83E12-Gal4 positive cells, median = 724] instead of the expected 1136 ensheathing glial cells (Fig. 3i, n). Similarly, when we expressed the cell cycle regulator Fizzy related (Fzr) which blocks cell division upon overexpression 51-53 we noted a similar reduction in the number of adult ensheathing glia ( Fig. 3j-n). Interestingly, the block of cell proliferation in ensheathing glia mostly affects the adult brain (Fig. 3i, l). To further test whether larval ensheathing glia could divide, we analyzed the ploidy of the ensheathing glia using DAPI staining 54 . In the larval ventral nerve cord, 30 of 30 tested abdominal glial nuclei are polyploid and thus are likely not to divide (Fig. 3o, Supplementary Fig. S3a-d). In contrast, 10 out of 25 ensheathing glial cells in thoracic neuromeres are diploid and possibly could divide (Fig. 3o, Supplementary Fig. S3a-d). Phospho-histone H3 serves as a specific marker for mitosis. In the larval CNS, we only rarely detected 83E12-Gal4 and phospho-histone H3 positive cells ( Supplementary Fig. S3e). Dividing ensheathing glial cells were easily detected in early pupae but not in 42 APF old pupae (Supplementary Fig. S3f-h). In addition, we fed the thymidine analogue 5-ethynyl-2′-deoxyuridine (EdU) which allows to label DNA synthesis during larval development and frequently detected EdU staining in 83E12-Gal4 positive ensheathing glia of larval and young pupal brains ( Supplementary Fig. S3i, j). Moreover, when we labelled larval ensheathing glia using a MCFO2 strategy 55 in third instar larvae or early pupae, we found labelled ensheathing glia in the adult (Supplementary Fig. S4a-c).
In conclusion, at least some larval 83E12-Gal4 expressing ensheathing glial cells are diploid and can proliferate.
Ensheathing glial cells are not required for viability. To test the functional relevance of the ensheathing glia, we performed ablation experiments. Expression of hid and reaper (rpr) using repo-Gal4 or 83E12-Gal4 resulted in early larval lethality. Since 83E12-Gal4 is also expressed in the midgut, we used a split-Gal4 approach to restrict Gal4 expression to only the ensheathing glial cells of the CNS 56,57 . We generated a construct that allowed the expression of a Gal4 activation domain under the control of the 83E12 enhancer [83E12-Gal4 AD ]. When crossed to flies that carry an element that directs expression of the Gal4 DNA binding domain in all glial cells [repo-Gal4 DBD ], Gal4 activity will be reconstituted only in CNS ensheathing glial cells. Indeed, animals carrying both transgenes [83E12-Gal4 AD , repo-Gal4 DBD ] show the same expression domain as the original 83E12-Gal4 driver in the adult as well as in the larval nervous system (Supplementary Fig. S4d-g).
Flies expressing hid and rpr only in the ensheathing glia [UASrpr; 83E12-Gal4 AD ; repo-Gal4 DBD , UAS-hid] survive to adulthood. Possibly, in flies with the genotype [UAS-rpr; 83E12-Gal4 AD ; repo-Gal4 DBD , UAS-hid] low levels of Gal4 based hid and or rpr expression might also occur in the related astrocyte-like cells. To further validate the specificity of ensheathing glia ablation, we stained the specimens for Rumpel and Nazgul. Rumpel is a SLC5A transporter strongly expressed by the ensheathing glia and weakly by astrocyte-like glia (Fig. 4). Nazgul is a NADP-retinol dehydrogenase that is specifically expressed by astrocyte-like cells 58 . In control, third instar larval brains Rumpel is found along the entire neuropil with Nazgul expressing astrocytes positioned in a characteristic pattern on the dorsal surface of the neuropil (Fig. 4a-c). Likewise, when we ablated the ensheathing glia in a background of a UAS-CD8::GFP transgene [UAS-rpr; 83E12-Gal4 AD ; repo-Gal4 DBD , UAS-hid] no ensheathing glia cells can be detected in the CNS. Only few wrapping glial cells in the peripheral nervous system can still be detected (Supplementary Fig. S4h-j).
Astrocytes infiltrate the entire neuropil and form only few processes on the outer surface of the neuropil (Fig. 4c). Upon ablation of the ensheathing glia, only a very faint astrocytic Rumpel signal is detected in the head region (Fig. 4d). In the ventral nerve cord, astrocyte-like cells form additional large cell processes around the neuropil (Fig. 4f). In addition, in adult flies lacking ensheathing glia we noted a 20% increase in the number of Nazgul positive astrocyte-like cells (Fig. 4g-i, n = 6 for each genotype, [control: 427, 459, 379, 458, 368 and 433, median = 430; ablation: 507, 520, 469, 478, 516, and 545, median = 512; **p = 0.0022). Adult flies are fertile but lifespan is reduced by about 40% compared to control animals (Fig. 4j, n = 200 mated females, p = 4,67365E-83). In conclusion, these data show that ensheathing glial cells are not essential for viability. Upon ablation of ensheathing glia, astrocyte-like cells form additional processes encasing the neuropil. This compensatory growth of astrocyte-like cells suggests that barrier establishment is a key function of the ensheathing glia.
Ensheathing glial cells contribute to barrier formation around the neuropil. The third instar larval CNS is covered by subperineurial glial cells that do not allow penetration of labelled 10 kDa dextran. In wild type, subperineurial glial cells block paracellular diffusion across the blood-brain barrier by forming septate junctions 59 . However, in larvae lacking septate junctions, extensive interdigitations between neighboring subperineurial cells can also provide a barrier function 60 .
Extensive electron microscopic analyses failed to demonstrate septate junction like structures between different ensheathing glial cells. However, glial processes overlap extensively at the neuropil cortex interface (Fig. 2i, j), which might increase the length of the diffusion path at the boundary between CNS cortex and neuropil. To directly test whether ensheathing glial cells indeed provide a barrier function, we performed dextran diffusion assays in control brains and those lacking ensheathing glia. For this, we carefully dissected third instar larval brains including anterior cuticular structures and placed them on a coverslip. Upon injection of fluorescently labelled 10 kDa dextran directly into the neuropil of one brain lobe employing capillary normally used for DNA injections, we determined diffusion of the labelled dextran within the neuropil through the large brain commissure into the contralateral hemisphere (Fig. 5a). Diffusion of dextran was monitored under the confocal microscope and quantified for 10 minutes (Fig. 5b, d, movies S1,2). The ratio of the fluorescence measured at two identical sized ROIs at the neuropil and the cortex contralateral to the injection site was plotted against time (Fig. 5c). In control larvae, a fluorescence is mostly confined to the neuropil area and only little dye reaches the cortex area. When we performed the injection experiments in larvae lacking ensheathing glia, we noted a significantly faster increase in fluorescence signal in the cortex area relative to the neuropil (Fig. 5c). Thus, we conclude that the ensheathing glia, although they lack specialized occluding junctions, contribute to a barrier function that possibly involves the extensive overlap noted for ensheathing glia cell processes (Fig. 2).
Ensheathing glial cells are polarized. Tissue barriers are generally formed by polarized cells which are characterized by a To detect a possible differential localization of these phospholipids, we employed different lipid sensors based on distinct PHdomains. PH-PLCδ-mCherry is targeted by PIP 2 63 whereas PH-AKT-GFP preferentially binds PIP 3 64 . We focused our analysis on the ensheathing glial cells covering the dorsal aspect of the neuropil, since the ventrolateral ensheathing glia form very thin, less than 200 nm thick, sheets around the neuropil that precludes a further analysis by confocal microscopy.
We co-expressed both sensors in the ensheathing glia using 83E12-Gal4 and quantified the distribution of both sensors by determining the number of green and red fluorescent pixels in 83E12-Gal4 positive cells in orthogonal sections comparing the cell domain close to the neuropil with the one close to the blood-brain barrier (Fig. 6a-g). Here, we noted an increased PH-PLCδ-mCherry localization at the direct interface of the neuropil and the ensheathing glia (Fig. 6b-e). For quantification, we measured the mean fluorescence intensity of PH-PLCδ-mCherry and PH-AKT-GFP (30 dorsal neuropil areas of 10 larvae). This demonstrated an almost twofold enrichment of PIP 2 at the plasma membrane domain facing the neuropil. The PIP 3 detecting PH-AKT-GFP shows a complimentary distribution with an enrichment in basolateral plasma membrane domains (Fig. 6f, g). A similar polarization of the ensheathing glial cells can be detected in adult stages ( Fig. 6f; Supplementary Fig. S5, 30 neuropil areas in 10 adult brains).
Polarized epithelial cells are characterized by an enrichment of the Na + /K + -ATPase Nervana 2 (Nrv2) at the basolateral plasma membrane 65 . To study the localization of Nrv2, we employed a gene trap insertion 66 . As found for PIP 3 , Nrv2::GFP localizes predominantly at the basal side of the ensheathing glia ( Fig. 6h-k), which corresponds to the localization of Nrv2 in epithelial cells. Therefore, we conclude that dorsal ensheathing glial cells are polarized cells throughout development, facing their apical-like domain towards the neuropil (Fig. 6g). Although all polarity markers are also expressed by the ventrolateral ensheathing glia, we cannot resolve their polarity due to the very flat shape of these membrane sheets which does not allow the separation of apical and basolateral membrane domains by optical methods.
Integrin α and extracellular matrix components are expressed by adult ensheathing glia. In epithelial cells, apical-basal polarity is also reflected by a basally located integrin receptor which bind the basally located extracellular matrix (ECM) proteins. inflated encodes an alpha-subunit of the integrin receptor and inflated mRNA is expressed by adult perineurial and ensheathing glial cells 11,67 (Supplementary Fig. S6). Endogenously YFP-tagged Integrin α (if CPTI-004152 68 ) localizes at the blood-brain barrier and around the neuropil in larval brains (Fig. 7a, b, e). In the adult brain, Integrin YFP α localization around the neuropil appears more pronounced compared to abdominal neuromeres (Fig. 7c, d). Integrin receptors bind ECM components, suggesting that these proteins are also expressed within the nervous system. We thus tested expression of all genes annotated to encode proteins involved in extracellular matrix formation (FlyBase) using published single cell sequencing data of adult brains 11 ( Supplementary Fig. S6). Eight of these genes appear expressed by the ensheathing glia and GFP-protein trap insertion lines were available for trol (Perlecan), viking (Collagen IV), and dally (Glypican).
The heparan sulfate proteoglycan Dally is most strongly detected close to the cells of the larval blood-brain barrier (Fig. 7f). Within the larval CNS, Dally is found in the cortex and is enriched around the larval neuropil (Fig. 7f, g). In adults, Dally localization at the blood-brain barrier ceases but is still found in the CNS cortex. A prominent enrichment of Dally is seen around the neuropil (Fig. 7h-j). In contrast to Dally, we could not detect Trol and Viking around the larval neuropil and only detected strong signals at the blood-brain barrier (Fig. 7k, l,  p, q). However, in the adult nervous system both Trol and Viking are detected basally of the ensheathing glia facing the cortex glia with an even distribution in the different parts of the CNS (Fig. 7m-o, r-t). In conclusion, the ensheathing glia express ECM proteins which is characteristic for polarized cell types.

Polar distribution of ß Heavy -Spectrin in larval ensheathing glia.
Polarization of cells is also evident in the cytoskeleton underlying the plasma membrane. Whereas α-Spectrin is found below the entire plasma membrane, ß-Spectrin decorates the basolateral plasma membrane and ß Heavy -Spectrin (ß H -Spectrin) is located at the apical domain of the cell 69 . ß-Spectrin is found in all neural cells in the ventral nerve cord 70 and no specific localization can be resolved in the ensheathing glia due to its strong overall expression. ß H -Spectrin is encoded by the karst gene. One available MiMIC insertion allows to target two of the seven isoforms (Fig. 8a). In third instar larval brains, GFP-ß H -Spectrin localizes at the blood-brain barrier (Fig. 8c, d, f, g), and close to the neuropil (Fig. 8c, d, f), indicating expression by the ensheathing glia. Here, ß H -Spectrin is enriched close to the apical-like, PIP 2 containing membrane of the ensheathing glia facing towards the neuropil (Fig. 8c, d). To further validate the expression of Karst in the ensheathing glia, we silenced karst expression by expressing double stranded RNA. When karst expression is silenced in all glial cells using repo-Gal4 only expression in trachea as well as a weak background signal at the neural lamella around the CNS is detected (Supplementary Fig. S7a, b). Silencing of karst expression in the blood-brain barrier using moody-Gal4 did not affect localization of GFP around the neuropil (Supplementary Fig. S7c, d). When we silenced karst specifically in the ensheathing glia using 83E12-Gal4 protein localization in the blood-brain barrier is unaffected but no ß H -Spectrin protein can be detected around the neuropil (Fig. 8g).   -GFP staining). a, b, e During late larval stages, prominent Inflated expression is seen at the blood-brain barrier (arrows) and weak expression is seen around the neuropil at the position of the ensheathing glia (arrowhead). c, d In adults, Integrin α is still found at the blood-brain barrier (arrows) and becomes more prominent around the neuropil (arrowheads). Note that strongest Inflated expression is detected in larval brains. No Inflated expression can be detected in the abdominal neuromeres (asterisk). f, g During the larval stage Dally is expressed at the blood-brain barrier (arrows) and in the CNS cortex (asterisks). A slightly stronger localization of Dally can be detected at the position of basal ensheathing glia processes (arrowhead). h-j In adults, Dally is enriched at the position of the ensheathing glia (arrowheads). k, l In the larval CNS Collagen IV is found at blood-brain barrier (arrows) but not within the nervous system. m-o Collagen IV is detected at the neuropil-cortex interface in adult stages (arrowheads in m, n, boxed area is shown in (o)). p, q Trol is detected at the larval blood-brain barrier (arrows) but not around the neuropil. r-t In adults, prominent Trol localization is seen at the position of the ensheathing glia (arrowheads). The boxed area in (s) is shown at higher magnification in (t). Scale bars are: larval CNS 50 µm, adult CNS 100 µm except for e: 25 µm; j, o, t: 50 µm.
A similar ß H -Spectrin localization was detected using antisera directed against a C-terminal peptide present in all predicted ß H -Spectrin isoforms (Fig. 8a, e, f, for specificity of the generated antiserum see: Supplementary Fig. S7e, f). Furthermore, we generated a Trojan-Gal4 insertion using the same MiMIC insertion that was used to generate the Karst GFP protein trap (Fig. 8a). The kst Trojan-Gal4 allele is expected to report only the expression pattern of the proximal promoter which activates expression of two out of the seven karst splice variants (FlyBase) (Fig. 8a). Indeed, when driving a nuclear GFP reporter (UAS-Lam::GFP), expression in the ensheathing and blood-brain barrier glial cells is apparent (Fig. 8h, i). In addition, cells in  (Fig. 8i). During pupal stages, expression of Karst ceases between 72 and 96 hours after puparium formation (APF) ( Supplementary  Fig. S8). Adult ensheathing glial cells lack detectable expression of the kst MiMIC::GFP insertion and show no reactivity using the anti-Karst antisera.
ß H -Spectrin is required for ensheathing glial polarity. We next tested whether ß H -Spectrin is required for ensheathing glial cell polarity and determined the overall morphology of the ensheathing glia in the karst loss of function karst MiMIC13613 larvae using 83E12-Gal4 driving CD8::GFP. In karst mutants, larval ensheathing glial cells are still present but show an abnormal collapsed morphology. The extension of ensheathing glia cell protrusions around dorsally located neuronal cell bodies is less evident (Fig. 9a, b). When we silenced karst expression using RNAi, we noted a weaker phenotype and dorsal protrusions frequently formed (Fig. 9c). This allowed an analysis of the distribution of PIP 2 and PIP 3 (Fig. 9c-e). Whereas in ensheathing glial cells of control ventral nerve cords, PIP 2 (as detected by PH-PLCδ-mCherry) is found predominantly on the domain facing the neuropil, an almost even distribution is noted upon kst knockdown in larvae (compare Figs. 6f, 9i). Interestingly, the polar distribution of PIP 3 does not appear to be affected by kst knockdown (Fig. 9e, i). To further test whether ß H -Spectrin is needed for ensheathing glial polarization, we performed overexpression experiments. To direct expression of wild type Karst protein in ensheathing glia, we utilized the EP-element insertion P{EPgy2}EY01010. Overexpression of karst results in abnormally convoluted and partially collapsed ensheathing glial cell morphology characterized by an even distribution of both PIP 2 and PIP 3 (Fig. 9f-h, j).
We next assayed Nrv2 localization in karst null mutants as well as upon 83E12-Gal4 directed kst knockdown. In larvae lacking karst expression ensheathing glial cells show a collapsed morphology and polarized localization of Nrv2 cannot be resolved (Fig. 9l-n). Upon kst knockdown Nrv2 correctly localizes to basolateral protrusions (Fig. 9k, o-q). In kst gain of function larvae, the polarized Nrv2 distribution is lost (Fig. 9k, r-t), supporting the notion that the spectrin cytoskeleton might affect proper positioning of Nrv2 65,71-73 .
Polarized ensheathing glia is required for locomotor behavior. The above data show that larval ensheathing glia are polarized, ECM abutting cells that separate the neuropil from the CNS cortex and are required for longevity of the adult fly. To test whether the ensheathing glial cells are required for normal locomotor control during larval stages, we compared locomotion of control animals with those lacking ensheathing glia or with those with reduced ß H -Spectrin or Nrv2 expression using FIM imaging 74,75 .
Control animals move on long paths interrupted by short reorientation phases that are characterized by increased body bending (Fig. 10a). kst knockdown specifically in ensheathing glial cells [83E12-Gal4 AD ; repo-Gal4 DBD , UAS-kst dsRNA ] causes a reduction in the peristalsis efficiency during go-phases (Fig. 10a, b, f, g). Likewise, crawling velocity is reduced significantly (Fig. 10g). This suggests that the specific lack of ß H -Spectrin in ensheathing glia causes a strong locomotor phenotype. The observed locomotor phenotypes might be due to an ectopic neuronal activity of the split-Gal4 driver in addition to its activity in the ensheathing glia. Assuming that the repo promoter is indeed glial cell specific 76 , neuronal activity is not expected in the split Gal4 approach. Moreover, we have no evidence that karst is expressed by neurons. Furthermore, we analyzed kst mutant larvae to validate the RNAiinduced phenotypes. The split Gal4 driver might have ectopic activity in other glial cells, but given the specific karst knockdown only in ensheathing glia we consider this unlikely (Fig. 8g). The Trojan-Gal4 insertion in the karst MiMIC03134 insertion is expected to affect only isoforms Karst-PE and Karst-PG (Fig. 8a). As control, we used an insertion of the Gal4 element in the opposite, unproductive orientation. Similar as detected for the kst knockdown, we noted a decreased peristalsis efficiency and a reduced crawling velocity (Fig. 10c, i, j). Larvae completely lacking zygotic karst expression [karst MiMIC13613 /Df(3 L)ED2083] show a comparable larval locomotion phenotype (Fig. 10d, i, j). This larval locomotion phenotype is similar to the one observed following ablation of the ensheathing glial cells using the genotype [UAS-rpr; 83E12-Gal4 AD ; repo-Gal4 DBD , UAS-hid] (Fig. 10e). Thus, we conclude that polarized ensheathing glia that connect the bloodbrain barrier with the dorsal neuropil are required for normal locomotor behavior. In order to perform vectorial transport, the Na + /K + ATPase must act in a polarized fashion. We thus also compared larval locomotion of animals with reduced nrv2 expression to control animals. Interestingly, larvae with ensheathing glial cells lacking nrv2 expression behave opposite to larvae that lack ß H -spectrin showing an increased peristalsis efficiency as well as an increased crawling velocity (Supplementary Fig. S9). The analysis of the role of polarized Nrv2 distribution for ensheathing glia physiology will thus be an interesting topic for future research.

Discussion
Here, we present a comprehensive analysis of the ensheathing glia found in the CNS of Drosophila. We developed a split-Gal4 tool to specifically manipulate the ensheathing glia and show that ensheathing glial cells are highly polarized cells with an apical-like cell domain facing towards the neuropil. Polarized membrane lipids as well as several polar localized proteins were identified. The apical sub-membranous cytoskeleton is organized by ß H -Spectrin encoded by karst. Ensheathing glia cells contribute to the formation of an internal diffusion barrier as determined by dye penetration experiments in ensheathing glia ablated animals. Flies that lack ensheathing glia have increased numbers of astrocytes, they are viable but show reduced longevity. Moreover, we show that ensheathing glial cells lacking Karst have an altered cell polarity and kst mutants as well as cell type specific knockdown larvae show abnormal locomotor behavior.
To study ensheathing glial cell biology, we utilized the 83E12-Gal4 driver which, as described 41 , appears specific for ensheathing glia. Some of the larval ensheathing glia are diploid and block of cell proliferation reduces their number by 38%, suggesting that proliferation of these cells contributes to the increase of ensheathing glial complexity. These finding support the notion that some larval ensheathing glia can dedifferentiate, proliferate and redifferentiate to form the more complex adult ensheathing glia 46 . Ablation of ensheathing glia results in an increased number of adult astrocyte-like cells which suggests that the close interdependence of these two cell types identified in embryonic stages 27,77 persists to adulthood. Similarly, a plastic interaction Fig. 9 The ß H -Spectrin cytoskeleton is required for glial polarity. Representative images are shown. a Orthogonal view of the dorsal aspect of a third instar control ventral nerve cord stained for ensheathing glial (EG) morphology [83E12-Gal4, UAS-CD8::GFP]. b Third instar larval brain of a zygotic karst null mutant larvae with the genotype [kst MiMIC13613 / kst MiMIC13613 , 83E12-Gal4, UAS-CD8::GFP]. The absence of all ß H -Spectrin protein affects the morphology of the dorsally located ensheathing glial cells. c-e Third instar larval ventral nerve cord with reduced kst expression in ensheathing glia [83E12-Gal4, UAS-kst dsRNA , UAS-PH-AKT-GFP, UAS-PH-PLCδ-Cherry]. d PH-PLCδ-Cherry binds to PIP 2 and shows an even distribution between apical and basolateral plasma membrane domains. For quantification see (i). Quantification was performed as described in Fig. 6. e PH-AKT-GFP binds to PIP 3 and is distributed in a polarized manner. For quantification see (i). f-h Third instar larval ventral nerve cord with increased kst expression in ensheathing glia [83E12-Gal4, kst P{EPgy2}EY01010 , UAS-PH-AKT-GFP, UAS-PH-PLCδ-Cherry]. Note, the variable phenotype noted upon karst overexpression. Ensheathing glia with almost normal morphology (open arrowhead) is next to hyperconvoluted ensheathing glia (asterisk). Quantification of PIP 2 and PIP 3 localization is shown in (j). i, j Quantification was performed as described in Fig. 6. The mean and the standard deviation is shown. k Quantification of the distribution of Nrv2::GFP using an unpaired t-test, two-tailed. *p control-kstdsRNA = 0.0474; ****p control-kstEP < 0.001. The mean and the standard deviation is shown. l-n In karst mutant larvae, Nrv2 is found in ensheathing glia cell processes (open arrowhead) that face the neuropil. The dashed line indicates the neuropil ensheathing glia boundary. o-q Upon ensheathing glial specific kst knockdown, Nrv2 localizes to the basolateral cell domains, for quantification see (k). also occurs between cortex and ensheathing glia 20 . Such interactions might also contribute to the functional differences that appear likely to be associated with for example dorsal or ventral ensheathing glial cells. Unfortunately, no specific molecular tools exist that allow to test this hypothesis.
A likely role of ensheathing glial cells is to establish a diffusion barrier around the neuropil, as was demonstrated by our dye injection experiments in ensheathing glia ablated larvae. However, we cannot exclude the possibility that ensheathing glia instruct neighboring cells such as the cortex glia or astrocyte-like glia to form a barrier around the neuropil. The separation of the neuropil from the cortex in the larva might facilitate the formation of signaling compartments specified by astrocyte-like cells. The 10-fold increase in the number of ensheathing glial cells in the adult brain is likely to further organize neuronal functions into discrete domains 35,[78][79][80] .
As also observed in vertebrates, the formation of signaling compartments can be detected on a gross anatomical level as well as on a circuit level. In both cases, the neuropil-associated glia comprising ensheathing glia are at center stage 25 and complexity of regional compartmentalization increases in the adult, mirroring the increase in ensheathing glial cell number 33,34,81 . In vertebrates, a classical example for this fundamental principle can be seen at the rhombomeric organization of the vertebrate hindbrain 82 . Here, glial cells provide essential functions in boundary formation, too 83 . Thus, glia may not only structure the nervous system by compartmentalizing synapses but also act on the level of larger functional units.
The establishment and maintenance of cell polarity is a fundamental cell biological feature and is required for the development of different cell types and tissues. For instance, the mammalian nervous system is protected by the blood-brain-barrier established by polar endothelial cells whereas the blood-brain barrier around the invertebrate nervous system is comprised of polarized glial cells 4,13,[84][85][86] . Cell polarity is induced by evolutionarily conserved mechanisms 87,88 . Apical polarity regulators (APRs) comprise a diverse set of proteins centered around the two apical polarity protein complexes Bazooka/Par3 and Crumbs 89 . In addition, basolateral regulators including the Scribble complex (Scribble, Discs large and Lethal (2) giant larvae and the kinases PAR-1 and LKB1 participate in the definition of a polar cell phenotype 90 . Some of these genes (Par1, Scrib, dlg1) appear to be expressed by adult ensheathing glia 11 . For Scrib but not Par1 or dlg1 this can be confirmed protein trap insertion lines. However, RNAi mediated knockdown of these genes did not cause any morphological defects in larval ensheathing glia. Possibly, similar as in follicle cell development, different mechanisms are in place to guarantee the establishment and maintenance of the polar ensheathing glial cell phenotype 91 .
A hallmark of polarized cells is the asymmetric distribution of lipids with PIP 2 being enriched at the apical surface 62 and reduction in β H -Spectrin causes a disrupted PIP 2 localization. This suggests, that a polar spectrin cytoskeleton might orchestrate lipid composition at the apical membrane domain which subsequently affects polarized distribution of transmembrane proteins such as the Na + /K + ATPase. This might cause to a differential distribution of Na + ions which are used by antiporters involved in metabolite transport. One of them is the Excitatory amino acid transporter 2 (EAAT2) 92 . Antibodies against EAAT2 are available 27 but unfortunately do not allow the analysis of its subcellular localization in larvae. Together, the different transport systems expressed in the ensheathing glia will also affect extracellular environment and thus diffusion properties in and out of the neuropil, which needs to be addressed in further studies.
The synaptic neuropil is localized at a unique position within the nerve cord very close to the dorsal surface and thus, the hemolymph. Whereas the ventral as well as the lateral parts of the neuropil are flanked by neuronal cell bodies embedded in cortex glia, only few neuronal cell bodies are found at the dorsal surface of the nervous system. More importantly, no cortex glial cells are found dorsally to the neuropil 20,28 . Thus, metabolites can be transported from the blood-brain barrier to the neuropil only by the ensheathing glia. Upon ablation of the ensheathing glia, the blood-brain barrier forming glial cells might provide enough metabolites to ensure an almost normal function of the nervous system. In case of suppression of Nrv2 expression in the ensheathing glia, ion gradients required for directed transport are not established across the ensheathing glial plasma membrane. At the same time the barrier provided by the ensheathing glia is still in place and therefore metabolically isolates the neuropil which results in an increased mobility possibly aimed to direct the larva to new and better food sources. The neuropil itself is highly organized and its dorsal domain is dedicated to synaptic connections of motor neurons 93 . Possibly, the motor neurons are more sensitive to nutrient supply and thus a more direct metabolic support through the ensheathing glia is required.

Methods
Drosophila genetics. Flies were raised on food prepared according to a recipe of the Bloomington stock center (https://bdsc.indiana.edu/information/recipes/ bloomfood.html). Depending on the culture conditions, fly stocks were flipped every 2-4 weeks. Experimental flies were always kept at 25°C. For crosses flies were anesthetized with CO 2 . Flies of undesired genotypes were either inactivated by placing them at −30°C for 2 days or by placing them in 70% ethanol as approved by the authorities.
The following flies were obtained from public stock centers: 83E12-Gal4 Electron microscopic analysis. All ssTEM data were generated at the Howard Hughes Medical Institute Janelia Research Campus and processed as described 31,43 . The CATMAID interface 44,96 was used to annotate glial cells associated with the neuropil. Glial cell bodies were identified by characteristic color of the cytoplasm and non-neuronal morphology. A cell body node was positioned at the glial soma and annotated using the following information: hemineuromere, glial subtype, serial number for cells in one hemineuromere. The tools built into CATMAID were used to generate 3D renderings and extract image data. Cells identified in RAW data were highlighted using Adobe Photoshop CS6.
Measurement of DAPI intensity. To determine the polyploidy of ensheathing glia, we stained larval or adult brains with DAPI, Repo and Elav in order to identify glial cells and neurons. We compared DAPI stained nuclei identified by 83E12-Gal4 > UAS-Lam::GFP with DAPI staining of directly neighboring neuronal nuclei identified by anti-Elav staining. The amount of DAPI staining of neighboring glial and neuronal nuclei was determined using Fiji 1.52b 97 . In each case the entire nuclear volume was measured.
EdU incorporation assay. First instar larvae were placed on freshly prepared fly food containing 0.2 mM EdU and were kept on the food at 25°C until the third instar stage. Brains of the desired age were dissected and stained for EdU presence using the Click-iT Plus EdU detection kit of Thermo Fisher (# C10638) according to the instructions of the manufacture.
Dye injection assay. Brains of wandering L3 larvae were dissected on ice and immediately placed on Poly-L-lysine coated object slide and covered by 10 S Volatef halocarbon oil. An automated injection station (FemtoJet 4 L, Eppendorf) was used to inject 2.5 mM 10 kDa Texas-Red conjugated dextran (Cat#D1863, Invitrogen) diluted in H 2 O into the neuropil of one brain lobe in each brain. For injections, we used borosilicate glass capillaries of 1 mm outer diameter and 0.58 mm inner diameter (Harvard apparatus 30-0016 capillaries GC100-10) that were pulled on a P-1000 Sutter instrument micropipette puller to reach a tip diameter of about 10 µm as commonly used for DNA injections.
Longevity assay. Flies were raised at 25°C. The offspring of the control and the ablation experiment were collected for up to 3 days immediately after hatching and separated between males and females. 20 mated females were put in a vial and a total of 200 animals were observed over a period of time. The living flies were counted and placed on fresh food without anesthesia every two days.
Larval locomotion analysis. Larvae for behavioral experiments were raised at 25°C according. Larval locomotion of wandering third instar larvae was tracked by the FTIR-based Imaging Method (FIM) based on frustrated total internal reflection (FTIR) at room temperature 74,75 . 10-15 larvae per genotype were placed on the tracking arena and after a 60 s accommodation phase recording started for 3 minutes using FIMtrack v3.1.28.5 and analyzed suing FIMAnalytics v0.1.1.2.
Quantification, statistical analysis and reproducibility. All confocal analyses were repeated independently at least three times with at least 5 specimens each and similar results were obtained. Fiji was used to investigate polar distribution. Cells were divided in two ROIs and the fluorescence intensity ratio between apical and basal was compared. Counting of the nuclei and the astrocytes was performed in a 3D model in Imaris 8.4.1 Oxford Instruments. Fluorescence intensities for the Dextran Assay were measured using Fiji 1.52b 97 using the normalized fluorescence intensity and the ratio between neuropil and cortex. Statistical analysis and calculation were obtained by Prism 6.0 and Excel 2016.