Starvation-induced regulation of carbohydrate transport at the blood-brain barrier is TGF-β-signaling dependent

During hunger or malnutrition animals prioritize alimentation of the brain over other organs to ensure its function and thus their survival. This so-called brain sparing is described from Drosophila to humans. However, little is known about the molecular mechanisms adapting carbohydrate transport. Here, we used Drosophila genetics to unravel the mechanisms operating at the blood-brain barrier (BBB) under nutrient restriction. During starvation, expression of the carbohydrate transporter Tret1-1 is increased to provide more efficient carbohydrate uptake. Two mechanisms are responsible for this increase. Similarly to the regulation of mammalian GLUT4, Rab-dependent intracellular shuttling is needed for Tret1-1 integration into the plasma membrane, even though Tret1-1 regulation is independent of insulin signaling. In addition, starvation induces transcriptional upregulation controlled by TGF-β signaling. Considering TGF-β-dependent regulation of the glucose transporter GLUT1 in murine chondrocytes, our study reveals an evolutionarily conserved regulatory paradigm adapting the expression of sugar transporters at the BBB.

pathogens and other blood-derived potentially harmful substances. Protein, 43 ion and metabolite concentrations fluctuate much stronger in circulation than 44 in the cerebrospinal fluid, the brains extracellular milieu (Begley, 2006). Thus Therefore, understanding the regulatory mechanisms that govern 62 carbohydrate transport into the nervous system is of major interest. 63 Interestingly, it has been reported that endothelial GLUT1 expression is 64 increased upon hypoglycemia (Boado and Pardridge, 1993; Kumagai et al., The Drosophila nervous system, as the mammalian nervous system, is 108 protected from the effects of malnutrition through a process called brain 109 sparing. It has been shown that upon nutrient restriction neuroblasts (neural 110 stem cells) can still divide and are thus protected from the growth defects that 111 are caused by a lack of proper nutrition in other tissues (reviewed in Lanet  112 and Maurange, 2014). This protection is achieved by Jelly belly 113 (Jeb)/Anaplastic lymphoma kinase (ALK) signaling that constitutes an 114 alternative growth promoting pathway active in neuroblasts (Cheng et al., 115 2011). However, if the brain continues developing and keeps its normal 116 function, nutrient provision needs to be adapted to ensure sufficient uptake, 117 even under challenging circumstances, like low circulating carbohydrate 118 levels. How nutrient transport at the BBB is adapted to meet the needs of the 119 nervous system even under nutrient restriction has not been studied. 120 Here, we show that carbohydrate transporter expression in Drosophila as in 121 mammals adapts to changes in carbohydrate availability in circulation. Tret1-1 122 expression in perineurial glia of Drosophila larvae is strongly upregulated 123 upon starvation. This upregulation is triggered by starvation-induced 124 hypoglycemia as a mechanism protecting the nervous system from the effects 125 of nutrient restriction. Ex vivo glucose uptake measurements using a 126 genetically encoded FRET-based glucose sensor show that the upregulation 127 of carbohydrate transporter expression leads to an increase in carbohydrate 128 uptake efficiency. The compensatory upregulation of Tret1-1 transcription is 129 The Drosophila larval brain is separated from circulation by the blood-brain 138 barrier to avoid uncontrolled leakage of hemolymph-derived potentially 139 harmful substances. At the same time, the blood-brain barrier also cuts off the 140 brain from nutrients available in the hemolymph. Thus, transport systems are 141 necessary to ensure a constant supply of nutrients, including carbohydrates. 142 The trehalose transporter Tret1-1 is expressed in the perineurial glial cells of 143 the larval and adult nervous system (Volkenhoff et al., 2015). In order to better 144 understand whether carbohydrate transport at the BBB is adapted to the 145 metabolic state of the animal, we analyzed Tret1-1 dynamics under different 146 physiological conditions. In fed animals Tret1-1 can be found at the plasma 147 membrane of the perineurial glial cells, but a large portion localizes to 148 intracellular vesicles ( Figure 1A, Volkenhoff et al., 2015). We subjected wild 149 type larvae to chronic starvation applying a well-established paradigm that 150 allows 48 h of starvation without disturbing development (Zinke et al., 2002). 151 Starvation increases Tret1-1 protein levels in the perineurial glial cells (Figure  152 1A). Furthermore, an enrichment of Tret1-1 protein at the plasma membrane 153 was observed ( Figure 1A, asterisk), showing that starvation induces changes 154 in Tret1-1 levels as well as localization. 155

Intracellular trafficking of Tret1-1 is Rab7 and Rab10 dependent 156
Three mammalian Glucose transporters, GLUT4, GLUT6 and GLUT8, are 157 regulated via trafficking between storage vesicles and the plasma membrane 158 (Corvera et al., 1994;Cushman and Wardzala, 1980;Lisinski et al., 2001;159 Suzuki and Kono, 1980). Similarly, a large amount of Tret1-1 localizes to 160 intracellular vesicles ( Figure 1A). Thus, intracellular trafficking of Tret1-1 may 161 partially regulate carbohydrate uptake into the perineurial glial cells. 162 To analyze if regulation of Tret1-1 expression requires intracellular trafficking, 163 we studied the involvement of different Rab-GTPases. Utilizing an EYFP-Rab 164 library available for Drosophila (Dunst et al., 2015) we found that subsets of Hedgehog regulation in response to dietary changes, but its exact functions 172 are unclear (Çiçek et al., 2016;Pataki et al., 2010). Rab19 has been 173 described to act in enteroendocrine cell differentiation, but its role in this 174 process is unknown (Nagy et al., 2017). 175 To determine a possible functional role of these Rab-GTPases in regulating 176 Tret1-1 trafficking, we analyzed Tret1-1 localization in the background of a 177 glia-specific knockdown (or expression of dominant-negative forms) of the 178 respective Rab proteins (Figure 2A  show that the tret1-1 promotor is induced upon starvation and thus Tret1-1 237 levels are transcriptionally adapted to the animal's metabolic state. 238

Glucose uptake rate increases upon starvation 239
Tret1-1 upregulation in perineurial glial cells is most likely a mechanism that 240 ensures efficient carbohydrate uptake into the nervous system even under 241 conditions of low circulating carbohydrate levels. Therefore, we aimed to 242 study the impact of Tret1-1 upregulation on carbohydrate uptake at the BBB. 243 Kanamori et al., 2010 showed that Tret1-1 transports trehalose when 244 heterologously expressed in Xenopus laevis oocytes. Since not only trehalose 245 but also glucose and fructose are found in the Drosophila hemolymph, we 246 analyzed whether Tret1-1 also transports other carbohydrates. Therefore, we 247 expressed Tret1-1 in Xenopus leavis oocytes to study its substrate specificity. 248 The Tret1-1 antibody is specific to the Tret1-1PA isoform, and thus at least 249 this isoform is upregulated in the perineurial glial cells upon starvation.  Since Put has been implicated in regulating carbohydrate homeostasis, we 334 asked if Put-dependent TGF-β signaling could also play a role in 335 carbohydrate-dependent Tret1-1 regulation. Thus, we expressed dsRNA 336 constructs against put in a glia-specific manner and analyzed Tret1-1 levels in 337 the perineurial glial cells of fed and starved animals ( Figure 6). Indeed, 338 starvation-dependent upregulation of Tret1-1 was completely abolished upon 339 put knockdown in the glial cells using either put KK102676 or put GD2545 . 340 Quantification shows no upregulation of Tret1-1 upon starvation in put 341 knockdown animals ( Figure 6B). In contrast, knockdown of wit using 342 wit KK100911 , did not affect Tret1-1 upregulation upon starvation ( Figure 6). This 343 data suggests, that Put-dependent TGF-β signaling in glia is essential for 344 starvation-induced upregulation of Tret1-1. 345 The Activin-branch of TGF-β signaling has been shown to be important for This indicates that the BMP-branch of TGF-β signaling is implicated in tret1-1 356 regulation. To analyze its involvement, we knocked down the BMP branch-357 specific type I receptors Thickveins (Tkv) and Saxophone (Sax) (reviewed in lethal, but Tkv overexpression can rescue sax loss-of-function, thus Tkv 360 seems to be the primary type I receptor in the BMP-branch of TGF-β signaling 361 (Brummel et al., 1994). Glia-specific knockdown of sax using sax GD50 or 362 sax GD2546 did not show any differences in Tret1-1 regulation upon starvation 363 compared to control knockdown animals ( Figure 6). In contrast, knockdown of 364 tkv using tkv KK102319 abolished Tret1-1 upregulation upon starvation, 365 highlighting its importance for signaling ( Figure 6).  (Figure 7). This effect is specific to 383 Gbb, since neither GFP-expressing control animals nor Dpp-expressing 384 animals display this effect (Figure 7). This shows that Gbb-dependent 385 signaling does induce Tret1-1 upregulation. 386 Taken together, the data reported here show that, upon starvation, moderate 387 levels of Gbb are produced by an unknown source, probably locally in the 388 subperineurial glial cells. Gbb activates the BMP-branch of TGF-β signaling in induces Tret1-1 expression. Since it has been shown that mammalian GLUT1 391 is also upregulated upon hypoglycemia, it will be interesting to see whether 392 TGF-β signaling is conserved as a pathway adapting carbohydrate transport 393 to changes in nutrient availability. 394 Discussion 395 The nervous system is separated from circulation by the BBB. This separation 396 on one hand protects the nervous system form circulation-derived harmful 397 substances, but on the other hand necessitates efficient nutrient transport to 398 ensure neuronal function. Since the nervous system mainly uses 399 carbohydrates to meet its energetic demands, carbohydrates need to be 400 taken up at a sufficient rate. We previously showed that the carbohydrate 401 transporter Tret1-1 is specifically expressed in perineurial glial cells that 402 surround the Drosophila brain and that glucose is taken up into the nervous  We show that the tret1-1 promoter is induced upon starvation (Figure 3). This 431 suggests that the tret1-1 locus harbors a starvation-responsive element. in perineurial glial cells we report here is independent of insulin signaling as 446 well as AKH signaling ( Figure 5). Thus, the regulatory mechanisms reported 447 here may be conserved. This is especially interesting since aberrations in The induction of carbohydrate transport at the BBB upon hypoglycemia or 455 starvation seems to be a mechanism that is required to spare the brain from 456 the effects of malnutrition. It has previously been shown in mammals, as well 457 as in flies, that the developing nervous system is protected from such effects 458 to allow proper brain growth, while other organs undergo severe growth 459 restriction. This process is called asymmetric intra-uterine growth restriction in 460 humans or "brain sparing" in model organisms ( Interestingly, the transcription of the Drosophila sodium/solute cotransporter 495 cupcake has also been shown to be upregulated upon starvation. Cupcake is 496 expressed in some ellipsoid body neurons upon starvation and is essential for 497 the ability of the animal to choose feeding on a nutritive sugar over feeding on 498 a sweeter non-nutritive sugar after a period of nutrient deprivation. 499 Furthermore, several solute carrier family members have been shown to be 500 regulated by carbohydrate availability in mouse cortical cell culture (Ceder et 501 al., 2020). It will be very interesting to investigate whether such transcriptional 502 upregulation is also mediated by TGF-β signaling and if TGF-β-mediated 503 transcriptional regulation in the nervous system is a central mechanism that 504 allows survival under nutrient shortage. 505 In summary, we report here a potentially conserved mechanism that spares 506 the nervous system from effects of nutrient shortage by upregulation of 507 carbohydrate transport at the BBB. This upregulation renders carbohydrate 508 uptake more efficient and most likely allows sufficient carbohydrate uptake 509 even when circulating carbohydrate levels are low. In Drosophila, 510 compensatory upregulation of Tret1-1 is regulated via Gbb and the BMP 511 branch of TGF-β signaling. This mechanism is likely to be evolutionarily 512 conserved, since mammalian Glut1 has been shown to be regulated via BMP 513 signaling in other tissues (Lee et al., 2018) and thus might in the future allow 514 designing a treatment against diseases caused by non-sufficient carbohydrate 515 transport in the nervous system. 516 We are grateful to M. Brankatschk for fly stocks. We thank Astrid Fleige for 518 help with cloning and Western blots. We are grateful to C. Klämbt for 519 discussions and critical reading of the manuscript. The work was supported by 520 grants of the DFG to SS (SFB1009, SCHI 1380/2-1). 521

Declaration of interest 530
The authors declare no competing interests. 531

Materials and Methods 532
All experiments have been conducted at least 3 times independently of each 533 other to assess interexperimental variation. In each experiment several 534 animals have been used to assess variations between animals. N gives the 535 number of independent experiments; n is the total number of animals 536 analyzed. If not noted otherwise, immunostainings have been done 3 times 537 independently including several animals in each experiment. 538

Larval starvation 584
Flies were kept overnight on standard food to stage the embryos. 42 h after 585 staging similar sized larvae were collected, cleared from food and transferred 586 to different food conditions: standard food, water-soaked filter paper or 10 % 587 sucrose in PBS. They were kept for 48 h on this condition before dissecting. 588 (YFP channel). Each larval brain was imaged in a separate experiment 611 (n=10). After 2.5 minutes, HL3 buffer was exchanged for glucose buffer (HL3 612 supplemented with 10 mM glucose; pH 7.2) and replaced by HL3 again after a 613 further 7.5 minutes. 614 For data analysis, a ROI containing the entire larval brain was selected and 615 the mean grey value of all pixels minus background for each channel was 616 calculated. Values were normalized to known minimum (HL3 buffer). 617 Statistical and regression analysis of data obtained was performed using 618 SigmaPlot software (Jandel). To determine glucose uptake rates, 10 time 619 points 9 seconds after values rose above baseline levels were used to 620 calculate the linear slope of each curve. Differences were assessed by the 621 Oocytes of the stages V and VI were injected with 18 ng (for mass 645 spectrometry) to 20 ng (for scintillation analysis) of cRNA and measurements 646 were carried out three to six days after cRNA injection. 647 To analyze the transport capacity by scintillation measurements radioactive 648 sugar substrates were generated using unlabeled sugar solutions of different  Co-staining of endogenous EYFP-tagged Rab-GTPases (green) and Tret1-1 (magenta) in the surface glia of third instar larval brains. All Drosophila Rab-GTPases endogenously labeled with EYFP were tested. Tret1-1-positive vesicles show overlapping staining with Rab7 EYFP , Rab10 EYFP , Rab19 EYFP and Rab23 EYFP . (green, grey) that shows that tret1-1-driven stgGFP is specifically expressed in perineurial glial nuclei.