Cells require a constant energy supply to function. Metabolic activity is in particular high in the nervous system, where large amounts of ATP is needed to maintain synaptic transmission and cope with the resulting changes in membrane potential. This is reflected by energy consumption as the mammalian brain accounts for 20% of the total resting oxygen consumption although comprising only 2% of the body's weight (Karbowski, 2007; Mink et al., 1981; Nortley and Attwell, 2017). This energy demand is even greater in young brains and is similarly found in the invertebrate nervous system (Harris et al., 2012; Laughlin et al., 1998; Mink et al., 1981; Tsacopoulos et al., 1988).

In the vertebrate nervous system, glucose is the predominant metabolite supplying the brain with energy. Glucose circulates in the blood stream and is delivered to the different organs. The brain is metabolically separated from circulation by the blood-brain barrier, which is comprised of endothelial cells that form occluding tight junctions (Abbott et al., 2006; Tam and Watts, 2010; Tietz and Engelhardt, 2015; Zlokovic, 2008). Endothelial cells take up glucose from the blood stream via the Glut1 transporter. From the endothelial cells, glucose is then shuttled to astrocytes and neurons by different glucose transporters. While endothelial cells and astrocytes express differentially glycosylated forms of the glucose transporter Glut1, neurons predominantly express Glut3 (Barros et al., 2007; Vannucci et al., 1997). To match fluctuating neuronal energy demands, glial cells are able to sense synaptic activity. The Astrocyte Neuron Lactate Shuttle (ANLS) hypothesis, initially established for the mammalian brain provides an elegant model explaining how the flux of small C3 metabolites is regulated in the brain (Magistretti and Allaman, 2018; Pellerin and Magistretti, 1994; Pellerin et al., 2007).

In contrast to vertebrates, invertebrates do not have a vascular system. Instead, the hemolymph, the blood equivalent tissue of invertebrates, is found in all body cavities and immerses the entire nervous system. The predominant sugar in hemolymph is trehalose, a non-reducing disaccharide composed of two glucose molecules linked in an α,α-1,1-glycosidic manner, acting as the prime energy source (Wyatt and Kalf, 1957). As in vertebrates, the nervous system is metabolically separated from the remaining body by the blood-brain barrier (Carlson et al., 2000; Limmer et al., 2014; Mayer et al., 2009). In Drosophila the blood-brain barrier is established by perineurial and subperineurial glial cells (Stork et al., 2008). Perineurial cells express trehalose transporters and participate in maintaining the energy homeostasis of the brain (McMullen et al., 2020; Volkenhoff et al., 2015). The subperineurial glial cells block paracellular diffusion by interdigitating cell–cell processes and the formation of septate junctions (Babatz et al., 2018; Bundgaard and Abbott, 2008; Schwabe et al., 2005; Stork et al., 2008). Trehalose is taken up from the circulation by the Tret1-1 transporter, which is expressed by perineurial glial cells. In addition, MFS3 and Pippin are involved in carbohydrate transport in the perineurial glia. Interestingly, MFS3 or Pippin null mutants are rescued via compensatory upregulation of Tret1-1, another blood-brain barrier carbohydrate transporter, while RNAi-mediated knockdown of Mfs3 and pippin is not compensated for (McMullen et al., 2020). Trehalose is subsequently metabolically processed through glycolysis. Lactate and alanine are then delivered to neurons by as yet poorly characterized transport mechanisms (Delgado et al., 2018; González Gutiérrez et al., 2019; Volkenhoff et al., 2015).

In general metabolite transport is mediated by members of the solute carrier protein (SLC) family, which allow either facilitated, or active transport into the cell. The SLC superfamily constitutes approximately 400 genes grouped into more than 50 families and many of its members are expressed in the brain (Bai et al., 2017; Morris et al., 2017). Two of these transporter families have been identified to be involved in glucose transport. The solute carrier proteins of the SLC2A family (Glut1-14) mediate facilitated glucose diffusion across the plasma membrane, whereas members of the SLC5A family (SGLT1-5) can transport glucose, fructose, lactate or pyruvate in a sodium gradient-dependent manner (Mueckler and Thorens, 2013; Wright, 2013; Wright et al., 2011).

In Drosophila, it is long known from deoxyglucose labeling experiments that glucose can be taken up by neurons in the brain (Buchner et al., 1979). Moreover, recent experiments using a FRET-based glucose sensor expressed in neurons of the Drosophila central nervous system (CNS) demonstrated that neurons are able to take up glucose in the same manner as glial cells (Volkenhoff et al., 2018). The fly orthologue of the mammalian Glut1 transporter is expressed exclusively in neurons and is not expressed in the blood-brain barrier (Volkenhoff et al., 2018).

The cellular route and transporters involved in delivery of trehalose, as well as glycolytic derived products to neurons remains elusive. The Drosophila nervous system comprises a relatively small set of well-defined glial cells, which establish a glial network that connects glial cells of the blood-brain barrier with the synaptic neuropil (Freeman, 2015; Yildirim et al., 2018). The cortex glial cells engulf neuronal cell bodies (Coutinho-Budd et al., 2017; Spéder and Brand, 2018). Axons and dendrites are located in the neuropil, which is infiltrated by numerous fine cell processes of the astrocyte-like glial cells (MacNamee et al., 2016; Peco et al., 2016; Stork et al., 2014). These cells modulate synaptic activity by participating in neurotransmitter homeostasis and the secretion of additional modulatory factors (Liu et al., 2014; Ma et al., 2016; Sengupta et al., 2019). The cell bodies of the astrocyte-like glia cells are found at the boundary of the neuropil, next to the ensheathing glial cell bodies (Peco et al., 2016). Ensheathing glia encase the entire neuropil and also participate in the modulation of locomotor activity, as well as in the regulation of sleep (Otto et al., 2018; Stahl et al., 2018).

Assuming that trehalose is taken up from the hemolymph at the blood-brain barrier, we hypothesize that further transport of its metabolic products (glucose, pyruvate or lactate) within the brain must be coordinated by other still elusive transporters. Here we uncover such transporters. We report the identification of three SLC5A family members [rumpel, bumpel (for brother of rumpel) and kumpel (for kin of rumpel)] that are specifically expressed by inner CNS glial cells and act in highly redundant manner to support neuronal function. Loss-of-function mutants of rumpel, bumpel or kumpel cause only very subtle behavioral phenotypes, whereas double and triple mutants showed behavioral phenotypes as well as female sterility demonstrating redundant gene functions.

Energy homeostasis in the brain is mediated by carbohydrate provision. Sugars are taken up from the hemolymph at the blood-brain barrier forming glial cells and then must be shuttled to neurons by other glially expressed transporters. The Drosophila genome encodes 15 predicted glucose and monocarboxylate transporter proteins of the SLC5A family, which are strong candidates to organize sugar distribution in the nervous system (Featherstone, 2011) (Fig. 1). We thus searched for SLC5A members that are expressed by glial cells inside the CNS.

Fig. 1.

The expression of SLC5 family members in the adult brain. (A–N) Single cell RNA sequencing data in the SCENIC representations of the 57 K scRNA seq data set (Davie et al., 2018). SCope analysis for the genes indicated in each bottom right corner is shown. Each dot represents a single cell. The color coding indicates the expression level. Red: strong expression, black: low expression. Grey: no expression. (A,B) repo expression marks glial cell clusters that can be assigned as perineurial (yellow), subperineurial (blue), cortex (grey), ensheathing (green) or astrocyte-like glial cells (orange) according to marker gene expression as shown in (C–H). (I–N) Expression of SLC5 family members that show expression in Drosophila glia. (O) Dendrogram of the evolutionary relationships of the different SLC5 family members of Drosophila. The color shading indicates expression in the respective glial cell type (see B). The scale bar represents 2×105 years of evolutionary distance.

Fig. 1.

The expression of SLC5 family members in the adult brain. (A–N) Single cell RNA sequencing data in the SCENIC representations of the 57 K scRNA seq data set (Davie et al., 2018). SCope analysis for the genes indicated in each bottom right corner is shown. Each dot represents a single cell. The color coding indicates the expression level. Red: strong expression, black: low expression. Grey: no expression. (A,B) repo expression marks glial cell clusters that can be assigned as perineurial (yellow), subperineurial (blue), cortex (grey), ensheathing (green) or astrocyte-like glial cells (orange) according to marker gene expression as shown in (C–H). (I–N) Expression of SLC5 family members that show expression in Drosophila glia. (O) Dendrogram of the evolutionary relationships of the different SLC5 family members of Drosophila. The color shading indicates expression in the respective glial cell type (see B). The scale bar represents 2×105 years of evolutionary distance.

Using recent single cell RNA sequencing data (Davie et al., 2018) expression of all predicted SLC5A sugar transporters can be traced to specific glial cell types (Fig. 1A–I). Expression of the glial cell marker Repo defines all glial cells in the adult fly brain (Halter et al., 1995). The different glial subtypes are characterized by expression of specific genes [perineurial glial cells: tret1-1 (Volkenhoff et al., 2015), subperineurial glial cells: gliotactin (Auld et al., 1995; Babatz et al., 2018), cortex glia: zydeco (Melom and Littleton, 2013), astrocyte-like glial cells: GAT and nazgul (Ryglewski et al., 2017; Stork et al., 2014) and ensheathing glial cells: EAAT2 (Peco et al., 2016)]. The different glial subtypes cluster in distinct groups of cells (Davie et al., 2018) (Fig. 1A–I).

In the adult brain, CG9657 is expressed most strongly in the ensheathing glia cluster but in addition some cortex glia and astrocyte-like glial cells express CG9657 (Fig. 1A–I). CG9657 was also identified in an RNAi-based screen for adult locomotor deficits using a construct without any predicted off-target (Dietzl et al., 2007; Schmidt et al., 2012). Due to an adult paralysis phenotype the gene was named rumpel, in honor of the slow-moving character of the Sesame Street.

In addition, CG6723 and CG42235 encode highly related proteins that are expressed in very similar set of glial cells in the adult CNS. We thus named CG6723 as brother of rumpel (bumpel) and the gene CG42235 as kin of rumpel (kumpel).

The rumpel gene is situated on the X-chromosome and encodes a predicted sugar transporter protein of the sodium solute symporter 5A (SLC5A) family with 13 transmembrane domains (Fig. 2A,B; Fig. S1). To further identify the cells expressing rumpel we dissected the rumpel promotor region. A 1.1 kb long enhancer fragment designated as rumpelPF1 (Fig. 2A) directs specific expression in the nervous system only in cells that are Repo positive (Fig. 2C,D). Based on their location around the neuropil, the rumpel expressing cells may correspond to ensheathing glial cells and/or astrocyte-like glial cells. To further test which cell type activates the rumpel enhancer we crossed the rumpelPF1-stGFP construct into a genetic background directing the expression of a red nuclear marker in the ensheathing and cortex glial cells (rumpelPF1-stGFP, nrv2-Gal4; UAS-stRed) (Fig. 2E). Most rumpel expressing neuropil-associated cells also show nrv2-Gal4 activity, suggesting that rumpel positive cells are expressed in ensheathing glial cells. This notion is corroborated by split-Gal4 experiments where we co-expressed the Gal4-DNA-binding domain in the rumpel pattern and the Gal4 activation domain in the nrv2 pattern (rumpelPF1-Gal4DBD, nrv2PF4-Gal4AD) (Fig. 2F). To test whether Rumpel is also expressed by astrocyte-like glial cells, we analyzed animals expressing GFP under the control of the rumpel enhancer and dsRed under the control of the alrm enhancer, which is active in astrocytes. In the larval central nervous system of such animals, we noted frequent coexpression (Fig. 2G), suggesting that Rumpel is also expressed by astrocyte-like glial cells. Similarly, when we stained rumpelPF1-stGFP larval brains with anti-Nazgul antibodies we noted a partial overlap (Fig. 2H). The rumpelPF1 fragment overlaps with the enhancer fragment 56F03 generated by the FlyLight project (Jenett et al., 2012; Li et al., 2014), which is reported to direct expression in ensheathing glia (Li et al., 2014; Otto et al., 2018; Peco et al., 2016) (Fig. 2A). This indicates that the critical enhancer elements are located in the 700 bp overlap of the two enhancer fragments.

Fig. 2.

rumpel-PF1 induces an expression in the neuropil-associated glial cells. (A) Schematic representation of the rumpel (CG9657) locus on the X-chromosome. Exons are shown in boxes, rumpel coding exons are in dark blue, 56F03 and rumpel PF1 denote enhancer elements that direct expression in ensheathing glia. The position of the CRISPR-induced premature stop codon in amorphic allele (rumpelC40) and the rumpel locus replacement with attP-loxP-Cherry-loxP in (rumpelΔ+Cherry) is indicated. (B) The Rumpel protein is predicted to have 13 membrane (light yellow) spanning domains. The peptide sequence used to immunize rabbits is highlighted in dark blue. o, outside; i, inside. (C–E,G,H) Specimens are stained for promoter fragment induced expression of StingerGFP (stGFP, green). (E,G) RedStinger (stRed, red). (F) LaminGFP (lamGFP, green). (C,D) Glial nuclei are stained for Repo protein localization (red). (H) Astrocyte-like glial cells are stained for Nazgul protein localization (red). Neuronal membranes are shown in blue (HRP staining). (C) rumpel promoter fragment PF1 (rumpelPF1) induces stGFP expression in Repo positive cells in the third instar larval brain. White dashed line indicates the position of the orthogonal section shown in D. (D) Glial cells in the position of ensheathing glia are indicated by arrows. No expression is observed in surface associated glial cells. (E) rumpelPF1 induced stGFP expression overlaps with the nrv2 induced RedStinger expression. (F) Split Gal4 directed expression of LamGFP is found in ensheathing glial cells [rumpelPF1-Gal4DBD, nrv2PF4-Gal4AD, UAS-lamGFP]. (G) rumpelPF1 induced stGFP expression is found in some astrocyte-like glial cells labelled by alrm induced stRed expression (compare arrows with arrowheads). (H) rumpelPF1 induced stGFP expression in Nazgul positive astrocyte-like glial cells (arrows). The asterisk denotes ensheathing glial nuclei, the arrowhead denotes astrocytes not activating the rumpelPF1 enhancer. Scale bars: 50 µm.

Fig. 2.

rumpel-PF1 induces an expression in the neuropil-associated glial cells. (A) Schematic representation of the rumpel (CG9657) locus on the X-chromosome. Exons are shown in boxes, rumpel coding exons are in dark blue, 56F03 and rumpel PF1 denote enhancer elements that direct expression in ensheathing glia. The position of the CRISPR-induced premature stop codon in amorphic allele (rumpelC40) and the rumpel locus replacement with attP-loxP-Cherry-loxP in (rumpelΔ+Cherry) is indicated. (B) The Rumpel protein is predicted to have 13 membrane (light yellow) spanning domains. The peptide sequence used to immunize rabbits is highlighted in dark blue. o, outside; i, inside. (C–E,G,H) Specimens are stained for promoter fragment induced expression of StingerGFP (stGFP, green). (E,G) RedStinger (stRed, red). (F) LaminGFP (lamGFP, green). (C,D) Glial nuclei are stained for Repo protein localization (red). (H) Astrocyte-like glial cells are stained for Nazgul protein localization (red). Neuronal membranes are shown in blue (HRP staining). (C) rumpel promoter fragment PF1 (rumpelPF1) induces stGFP expression in Repo positive cells in the third instar larval brain. White dashed line indicates the position of the orthogonal section shown in D. (D) Glial cells in the position of ensheathing glia are indicated by arrows. No expression is observed in surface associated glial cells. (E) rumpelPF1 induced stGFP expression overlaps with the nrv2 induced RedStinger expression. (F) Split Gal4 directed expression of LamGFP is found in ensheathing glial cells [rumpelPF1-Gal4DBD, nrv2PF4-Gal4AD, UAS-lamGFP]. (G) rumpelPF1 induced stGFP expression is found in some astrocyte-like glial cells labelled by alrm induced stRed expression (compare arrows with arrowheads). (H) rumpelPF1 induced stGFP expression in Nazgul positive astrocyte-like glial cells (arrows). The asterisk denotes ensheathing glial nuclei, the arrowhead denotes astrocytes not activating the rumpelPF1 enhancer. Scale bars: 50 µm.

To determine the localization of the Rumpel protein, we generated an anti-peptide antiserum directed against the C-terminal most amino acids (Fig. 2B). The specificity of the antiserum is demonstrated following pan-glial silencing of rumpel expression using RNAi (Fig. 3A,B). In the third instar larvae, no expression is discernible outside the CNS, which matches RNAseq expression data (Brown et al., 2014; Graveley et al., 2011). Within the nervous system, Rumpel localizes to cell membranes of neuropil associated cells in the developing brain lobes as well as in the ventral nerve cord (Fig. 3A–F, arrows). In addition, some Rumpel protein can be found within the neuropil (Fig. 3E,F, arrowheads). Very low levels of Rumpel protein are detected along the peripheral abdominal nerves that connect the CNS with the periphery. In adults, Rumpel expression is also found prominently in the ensheathing glial cells (Fig. 3G,H).

Fig. 3.

Rumpel protein is expressed in the neuropil-associated glial cells. All specimens are stained for Repo localization to define glial nuclei (magenta), for N-Cadherin localization to visualize axonal and dendritic cell membranes (blue) and for Rumpel protein localization (green/grey). (A–F) Third instar larval brains and (G,H) adult brain. (A) In control animals [repo-Gal4, UAS-GFPdsRNA] Rumpel protein localizes around the neuropil. (B) Upon expression of rumpeldsRNA in the all-glial cells [repo-Gal4, UAS- rumpeldsRNAv43922] no Rumpel protein can be detected, demonstrating the specificity of the anti-Rumpel antibody. (C,D) Rumpel localization is observed surrounding the neuropil (arrows) in a position of the ensheathing glial cells. Very little Rumpel protein is found along larval nerves (asterisks). (E) Image of a single confocal plane through a third instar larval ventral nerve cord. Rumpel localizes to ensheathing glial cell membrane (arrow) and to cell processes of astrocyte-like glial cells (arrowhead). The dashed line indicates the position of the orthogonal section shown in F. (F) Rumpel localizes to ensheathing glial cells (arrows) and astrocytic processes in the neuropil (arrowhead). Note, the pronounced cortex-glial cell like ramifications of the ensheathing glia dorsally to the neuropil (asterisk). (G,H) Rumpel localizes around the neuropil in adult brains at a position of the ensheathing glia (inset: antennal lobe). Scale bars: 50 µm.

Fig. 3.

Rumpel protein is expressed in the neuropil-associated glial cells. All specimens are stained for Repo localization to define glial nuclei (magenta), for N-Cadherin localization to visualize axonal and dendritic cell membranes (blue) and for Rumpel protein localization (green/grey). (A–F) Third instar larval brains and (G,H) adult brain. (A) In control animals [repo-Gal4, UAS-GFPdsRNA] Rumpel protein localizes around the neuropil. (B) Upon expression of rumpeldsRNA in the all-glial cells [repo-Gal4, UAS- rumpeldsRNAv43922] no Rumpel protein can be detected, demonstrating the specificity of the anti-Rumpel antibody. (C,D) Rumpel localization is observed surrounding the neuropil (arrows) in a position of the ensheathing glial cells. Very little Rumpel protein is found along larval nerves (asterisks). (E) Image of a single confocal plane through a third instar larval ventral nerve cord. Rumpel localizes to ensheathing glial cell membrane (arrow) and to cell processes of astrocyte-like glial cells (arrowhead). The dashed line indicates the position of the orthogonal section shown in F. (F) Rumpel localizes to ensheathing glial cells (arrows) and astrocytic processes in the neuropil (arrowhead). Note, the pronounced cortex-glial cell like ramifications of the ensheathing glia dorsally to the neuropil (asterisk). (G,H) Rumpel localizes around the neuropil in adult brains at a position of the ensheathing glia (inset: antennal lobe). Scale bars: 50 µm.

To further characterize the Rumpel expressing glial cells, we performed glial cell type specific silencing experiments. Following suppression of rumpel using nrv2-Gal4, which strongly suppresses in ensheathing glial cells and less so in cortex and astrocyte-like glial cells, we noted a complete lack of Rumpel protein localization (Fig. S2A,D). Following suppression in ensheathing glial cell using 83E12-Gal4, weak Rumpel expression can be detected in the cortex and neuropil, reflecting processes of the astrocyte-like glial cells (Fig. S2B,E). Following suppression of rumpel expression, mostly in astrocyte-like glial cells, using alrm-Gal4, Rumpel protein can still be detected in the ensheathing glia (Fig. S2C,F).

In conclusion, throughout development of the central nervous system of Drosophila, the SLC5A member Rumpel is expressed specifically in glial cells and is most prominently found in ensheathing glial cells with some expression in cortex and astrocyte-like glial cells.

RNA interference-based knockdown of rumpel caused paralysis of the adult flies upon mechanical stress (Schmidt et al., 2012). Since rumpel null mutant flies fail to show any of these responses and behaved as wild-type flies this initial observation is either due to off target effects or due to genetic plasticity induced upon systemic removal of the gene (Rossi et al., 2015; Sztal and Stainier, 2020). To better quantify behavioral phenotypes, we turned to larval locomotion. Third instar larvae of mutant rumpelC40 showed slight differences, when comparing unconstrained locomotion at 25°C and at 32°C. A heat map representation of control and rumpel mutant larvae shows that at 25°C both control as well as rumpel mutants spread evenly across the tracking arena (Fig. 4A,B). In contrast at 32°C, rumpelC40 null mutants do not explore the tracking plate as intensively as control larvae (Fig. 4C,D). This reduced exploratory locomotion phenotype is reflected in the mean distance to origin of the mutant animals (Fig. 4E, n=150 larvae, 3 min tracking, P=0.023). Interestingly, when we tested rumpelΔ+cherry, that was backcrossed ten times against a w1118 background, we noted no significant change in distance to origin at elevated temperatures (Fig. S3).

Fig. 4.

Behavioral analysis of rumpel. (A–E) 150 third instar larvae of the respective genotypes were recorded in g