Gluconeogenesis using glycerol as a substrate in bloodstream-form Trypanosoma brucei

Bloodstream form African trypanosomes are thought to rely exclusively upon glycolysis, using glucose as a substrate, for ATP production. Indeed, the pathway has long been considered a potential therapeutic target to tackle the devastating and neglected tropical diseases caused by these parasites. However, plasma membrane glucose and glycerol transporters are both expressed by trypanosomes and these parasites can infiltrate tissues that contain glycerol. Here, we show that bloodstream form trypanosomes can use glycerol for gluconeogenesis and for ATP production, particularly when deprived of glucose following hexose transporter depletion. We demonstrate that Trypanosoma brucei hexose transporters 1 and 2 (THT1 and THT2) are localized to the plasma membrane and that knockdown of THT1 expression leads to a growth defect that is more severe when THT2 is also knocked down. These data are consistent with THT1 and THT2 being the primary routes of glucose supply for the production of ATP by glycolysis. However, supplementation of the growth medium with glycerol substantially rescued the growth defect caused by THT1 and THT2 knockdown. Metabolomic analyses with heavy-isotope labelled glycerol demonstrated that trypanosomes take up glycerol and use it to synthesize intermediates of gluconeogenesis, including fructose 1,6-bisphosphate and hexose 6-phosphates, which feed the pentose phosphate pathway and variant surface glycoprotein biosynthesis. We used Cas9-mediated gene knockout to demonstrate a gluconeogenesis-specific, but fructose-1,6-bisphosphatase (Tb927.9.8720)-independent activity, converting fructose 1,6-bisphosphate into fructose 6-phosphate. In addition, we observed increased flux through the tricarboxylic acid cycle and the succinate shunt. Thus, contrary to prior thinking, gluconeogenesis can operate in bloodstream form T. brucei. This pathway, using glycerol as a physiological substrate, may be required in mammalian host tissues.

Introduction T. brucei is the causative agent of human and animal African trypanosomiases, devastating but neglected tropical diseases. The mammalian-infective form of the parasite, typically referred to as the bloodstream form (BSF), lives in blood of mammalian hosts and enters the central nervous system (CNS), leading to a fatal disease if not treated. In addition, trypanosomes were recently detected in adipose tissue in a mouse model [1] and the skin of both humans [2] and mice [3]. Tsetse flies transmit the parasites; these procyclic forms (PCF) grow in the insect mid-gut, differentiating through other adaptive life-cycle stages, and later migrating to the salivary glands, for transmission in saliva as metacyclic forms. Each of the parasite's stages is morphologically and metabolically adapted to the respective environmental conditions. Nutrient availability is variable in the tsetse mid-gut and PCF trypanosomes can utilize proline, and generate the majority of their ATP in a reticulated mitochondrion containing canonical functions; although the tricarboxylic acid (TCA) cycle appears to operate in a non-canonical manner [4]. On the other hand, BSF trypanosomes in the bloodstream grow in a stable, nutrient rich environment, with a constant and abundant glucose supply, producing ATP from glycolysis; the mitochondrion, the electron transport chain and the TCA cycle are substantially reduced in BSF cells [5]. The recent identification of trypanosomes in adipose tissue [1] and in the skin of humans [2] and mice [3], indicated that the 'bloodstream forms' should now be considered as bloodstream-resident, CNS-resident, adipose-resident or skin-resident forms, potentially with differing metabolic capacities.
Glycolysis is the metabolic pathway with the highest flux in BSF T. brucei grown in culture medium; this pathway has been thoroughly studied and has long been considered a promising potential drug target [6]. The majority of the glycolytic enzymes are localized inside glycosomes, specialized peroxisomes harboring glycolysis and additional metabolic pathways [7]. The glycosome membrane is semi-permeable; hence only smaller metabolites can pass freely [8], while ADP/ATP or NAD + /NADH regeneration must be balanced inside the organelle. It has been proposed that compartmentalized glycolysis emerged to facilitate adaptation to different environmental conditions [9]. Other metabolic pathways that are compartmentalized inside glycosomes include the pentose phosphate pathway (PPP), nucleotide sugar biosynthesis, nucleotide biosynthesis and salvage, lipid synthesis, and fatty acid β-oxidation, probably due to their connection to glycolysis [10].
Gluconeogenesis (GNG) is a metabolic pathway that results in the generation of glucose from non-carbohydrate carbon substrates such as glycerol, lactate or glucogenic amino acids. In principle, it is the reverse of glycolysis, as many glycolytic enzymes are reversible, the direction depending on substrate and product concentrations. Only two steps are thought to be unique to GNG in protozoa; requiring fructose-1,6-bisphosphatase (FbPase) and phosphoenolpyruvate carboxykinase activity [11]. PCF T. brucei display GNG capacity fed by proline [12] but GNG was thought to be absent from BSF T. brucei, since FbPase activity was not detected [13]. Metabolomic analysis with labelled glucose also failed to reveal evidence for GNG [11]. Consequently, it is often stated that BSF trypanosomes depend 'exclusively' [14,15,16,17] or 'entirely' [18,19] on glycolysis, using glucose as a substrate, for ATP production (reviewed in [20,21]). For this reason, the glycolytic pathway has been considered to be a promising target for antitrypanosomal drug discovery [6]. On the other hand, Ryley [22] reported utilization of glycerol for respiration by both BSF and PCF T. b. rhodesiense in the early 1960s. Furthermore, glycerol has been used routinely to sustain trypanosomes in glucose-free media for several hours in radiolabelling experiments [23]. We also recently reported the use of glycerol for ATP production by BSF T. brucei [24].
There are thought to be five copies of each trypanosome hexose transporters gene, THT1 (Tb927.10.8440-8480) and THT2 (Tb927.10.8490-8530), arranged in an array in the T. b. brucei strain 927 reference genome, however the number of copies is variable across different strains [25]. The two gene types are similar, containing some identical domains. THT1 transcripts are substantially more abundant in BSF cells relative to PCF cells, while in PCF cells, THT2 transcripts are substantially more abundant than THT1 transcripts [26,27]. Specific regions in the 3'-untranslated regions contribute to this stage-specific expression pattern [28], while only THT2 transcripts are upregulated after glucose depletion [27]. The THTs comprise twelve trans-membrane domains and a cysteine-rich loop [26]. They are closely related to mammalian hexose transporters, although some substrate selectivity was observed [29]. Overall, the substrate selectivity of THT1 and THT2 is similar, but THT1 is a high capacity, low affinity transporter, whereas THT2 is a lower capacity, higher affinity transporter; this may reflect the conditions under which each of the proteins is expressed (reviewed in [26]).
Here, we explore glycerol utilization for GNG in T. brucei depleted of THTs, and find GNG operating even in wild-type cells that have access to glycerol. Carbons from stable isotope labelled glycerol are detected in sugar phosphates, PPP intermediates, VSG glycans and other metabolites. We also detect robust FbPase activity, even after deletion of the annotated FbPase gene. Thus, contrary to prior thinking, GNG is available to mammalian stage T. brucei and may operate in tissue environments where glycerol is available. This metabolic flexibility may be essential for adaptation to environmental conditions and survival in mammalian host tissues.

THT1 and THT2 are plasma membrane-localised and life cycle stageregulated
T. brucei hexose transporter 1 (THT1) transcripts are substantially more abundant than THT2 transcripts in the bloodstream stage, whereas the situation is reversed in the insect-stage [27]. Both proteins are thought to be localized to the plasma membrane. To determine whether this is indeed the case, we assembled strains that express an N-terminal mNeonGreen (mNG) tagged copy of each gene type; tags were inserted at the native loci. Both proteins were detected by direct fluorescence microscopy. mNG THT1 was localised on the surface plasma membrane of bloodstream form cells, while mNG THT2 was on the surface of insect stage cells (Fig 1A). To assess transcript levels, we analyzed our recently published transcriptome data from bloodstream and insect life cycle stages [30]. The region encompassing the five copies of THT1 and the five immediately adjacent copies of THT2 is shown in Fig 1B and confirms higher THT1 transcript expression in the bloodstream stage ( Fig 1B, top panel) and almost exclusive THT2 expression in the insect stage ( Fig 1B, lower panel). As noted above, specific regions in the 3'untranslated regions contribute to this stage-specific expression pattern [28].
To explore developmental controls at the protein level, we assembled an additional BSF strain that expressed mNG THT2. Bloodstream strains expressing either mNG THT1 or mNG THT2 were differentiated to insect stage cells. Subsequent protein blotting revealed an expression pattern that mirrored the transcript levels seen above. Specifically, mNG THT1 was expressed primarily in bloodstream form cells, while mNG THT2 was expressed primarily in insect stage cells (Fig 1C and  1D). Thus, we demonstrate plasma membrane localization of both hexose transporters and confirm the expected life-cycle stage regulated controls at both the transcript and protein levels.
The hexose transporters are required for glucose uptake and viability THT1 and THT2 are closely related but contain multiple diverged domains (Fig 2A). For functional assessment of THT1, we assembled an RNA interference (RNAi) construct targeting the unique portions of the THT1 genes (Fig 2A) and used this construct to assemble knockdown strains that also expressed a native tagged copy of mNG THT1. We then induced specific knockdown of THT1, which revealed a substantial growth defect; efficient mNG THT1 knockdown was confirmed by protein blotting (Fig 2B). Although growth was perturbed by THT1-specific knockdown, cell growth continued, and increased over time (Fig 2B), possibly due to loss of the RNAi construct [31] or low-level THT2 expression (see Fig 1B and S1A Fig). We, therefore, assembled a second RNAi construct targeting the shared portions of both THT1 and THT2 simultaneously, and used this construct to assemble knockdown strains that again expressed a native tagged copy of mNG THT1. We induced simultaneous knockdown of THT1 and THT2, which in this case revealed a severe and more pronounced growth defect; efficient mNG THT1 knockdown was once again confirmed by protein blotting (Fig 2C). We also demonstrated the capacity for mNG THT2 knockdown using this approach in insect stage T. brucei (S1B Fig). Failure to grow after simultaneous THT1 and THT2 knockdown ( Fig 2C) indicated that the hexose transporters are essential for continued growth in BSF trypanosomes. This is consistent with the idea that these transporters are essential uptake routes for glucose and also that glycolysis is required for ATP production in these cells [4].
To directly measure glucose uptake in cells depleted for THT1 or for both THT1 and THT2, we assessed the accumulation of the radiolabelled glucose analogue, 2-14 C(U)-deoxyglucose, in wild-type and knockdown strains; knockdown was carried out in the presence of 5 mM glycerol (see Fig 3 below). We observed no uptake of label in cells incubated at 4˚C but robust accumulation of label in wild-type cells incubated at 37˚C over a 2 min time-course; 428 ± 211 pmol/min/10 8 cells (Fig 2D). In contrast, accumulation of label was 20-fold lower in knockdown cells; THT1-knockdown yielded only marginally (1.5-fold) higher labelling relative to THT1/THT2 knockdowns ( Fig 2D).

Glycerol supplementation rescues the THT-knockdown defect
The growth medium typically used to propagate BSF T. brucei in culture is rich in (25 mM) glucose [32]. This HMI-11 growth medium may fail to reflect the full range of metabolic pathways available to these parasites in their natural host environment. We were particularly interested in the possible utilization of glycerol as a carbon source, since T. brucei express aquaglyceroporins (AQPs) [24,33]. These AQPs appear to be important to remove glycerol from the cells, a product of glycolysis under anaerobic conditions [34]. On the other hand, we recently demonstrated that  [66]; raw transcriptomic data was derived from Hutchinson et al [30]. (C) Western blot analyses indicating stage specific expression of in situ tagged mNG THT1 and mNG THT2 in BSF (left) and PCF and cells differentiated in DTM medium (right). WT, wild-type; EF1α serves as a loading control. (D) Differentiation in DTM medium of mNG THT expressing cells ( mNG THT2 cells shown) was validated by immunofluorescence microscopy with stage specific surface markers, VSG-2 for BSF and EP procyclin for PCF. In A and D, DNA was counter stained with DAPI; Scale bars 5 μm. T. brucei can use glycerol as an energy source, in an AQP-dependent manner [24]. In addition, T. brucei have recently been shown to occupy adipose tissue in a mouse model [1], a tissue where the glycerol concentration is four-fold higher than in blood (~200 μM) [35,36].
We, therefore, exploited the THT-knockdown strains to explore the ability of T. brucei to use and to grow on glycerol as a carbon source. Supplementation of the standard HMI-11 growth medium with 5 mM glycerol had little impact on T. brucei growth ( Fig 3A). Notably though, the same supplementation, of cells in which THT1 and THT2 had been simultaneously knocked down, substantially rescued the severe growth defect observed above ( Fig  3B). These results indicate that glycerol can serve as an alternative carbon source for BSF T. brucei. It was previously observed that replacement of glucose with glycerol triggered differentiation of BSF into PCF [37], however we did not observe such a phenotype with THT knockdown cells; all cells continued to express variant surface glycoprotein 2 (VSG-2) and no cells expressed EP procyclin following 5 days of knockdown (S1C Fig).

Exogenous glycerol is a substrate for gluconeogenesis
We next employed a liquid chromatography-mass spectrometry (LC-MS) metabolomics approach to explore the utilization of stable heavy isotope labelled glycerol as a carbon source in bloodstream form trypanosomes. For these experiments, wild-type and THT1/THT2 knockdown cells were grown in HMI-11 medium supplemented with 5 mM U-13 C 3 -glycerol (glycerol with heavy isotope labelled carbon in all three positions) for three days; knockdown was induced over the three-day time-course in the latter case. Both samples cultured without any 13 C label, but with 5 mM supplementary glycerol in the case of the knockdowns, were grown and analysed in parallel. Subsequent metabolomic analyses allowed us to detect a number of intermediates of metabolism and also to trace the 13 C-labelled metabolites. Fig 4A indicates key steps in glycolysis and in the reverse pathway of gluconeogenesis, including the metabolites that we detected and analysed by LC-MS. In terms of the exogenous 13 C 3 labelled glycerol, we see that this is taken up by both wild-type and THT1/THT2 knockdown cells (Fig 4B; glycerol panel, green); two thirds (67%) of the glycerol in the knockdown cells was 13 C 3 labelled. Notably, the glycerol content was increased in all three cases where exogenous glycerol was supplied, particularly following knockdown ( Fig 4B). Next, in the knockdown cells, we can see that the labelled glycerol was phosphorylated by glycerol kinase to form glycerol 3-phosphate ( Fig 4B; glycerol-P panel, green); half (50%) of the glycerol-phosphate in the knockdown cells was 13 C 3 labelled. Labelled glycerol-phosphate was then channelled to further gluconeogenic intermediates. For fructose 1,6-bisphosphate ( Fig 4B; fructose 1,6bP panel), almost a third (31%) was 13 C 3 labelled (green) and more than half (59%) was fully 13 C 6 labelled (red). For hexose 6-phosphates ( Fig 4B; hexose 6P panel), one fifth (20%) was 13 C 3 labelled (green) and almost half (49%) was fully 13 C 6 labelled (red). For 2/3-phosphoglycerate (Fig 4B; P-glycerate panel, green), almost three quarters (74%) was 13 C 3 labelled, and similar patterns were observed in pyruvate ( Fig 4B; pyruvate panel, green) and phosphoenolpyruvate (S2 Fig). These labelled intermediates shown in Fig 4B were also detected, albeit at reduced levels ( 13 C 3labelled part comprised 10, 6, 7 and 16%, respectively), in wild-type cells. Thus, even wild-type T. brucei take up glycerol, when available, and use it as a substrate for gluconeogenesis.
T. brucei appear to prefer to use glucose, when available, as a carbon source. The cells analysed here that use mostly glycerol instead of glucose do grow at a slower rate ( Fig 3B). When glycerol is used, pathway intermediates drop to approximately one third of the levels observed in wild-type cells grown on glucose; fructose 1,6-bisphosphate drops to 32% (

De novo synthesised glucose 6-phosphate is utilised in other metabolic pathways
We find that glucose 6-phosphate (G6P) synthesised from 13 C 3 -glycerol is utilised in additional metabolic pathways, such as the pentose phosphate pathway (PPP) and for protein (A) A scheme of glycolysis and gluconeogenesis with highlighted fructose-1,6-bisphosphatase. PPP, pentose phosphate pathway; 1, hexokinase; 2 glucose-6-phosphate isomerase; 3, phosphofructokinase; 4, fructose-bisphosphate aldolase; 5, triose-phosphate isomerase; 6, glycerol-3-phosphate dehydrogenase; 7, glycerol kinase; 8, mitochondrial glycerol-3-phosphate dehydrogenase; 9, glyceraldehyde-3-phosphate dehydrogenase; 10, phosphoglycerate kinase; 11, phosphoglycerate mutase; 12, enolase; 13, pyruvate kinase. (B) Metabolites from glycolysis/gluconeogenesis as detected in the LC-MS analysis. The size of the bars represents the total abundance, and coloured parts indicate 13 C labelling as depicted in the key. The samples are from WT cells, WT grown in 13 C-glycerol, THT1/THT2 RNAi grown in 12 C-glycerol and THT1/THT2 RNAi grown in 13 C-glycerol. Glucose was also present for all experiments. Natural abundance of 13 C is 1%, hence the 1C labelling in 'un-labelled' samples that is proportional to the number of carbons in each metabolite. � identity of these metabolites was confirmed using a match with a standard. glycosylation. The oxidative branch of the PPP is indispensable in BSF T. brucei, being the main production route of reduced NADPH, a component essential for oxidative stress defence [38]. G6P is used as a substrate and since the non-oxidative branch of the pathway is missing in BSF T. brucei, the final end product is ribose 5-phosphate [39]. 25% of pentose phosphate is fully 13 C 5 labelled in trypanosomes grown on glycerol (Fig 5A), and ribose 5-phosphate is further incorporated into ATP and GTP, as demonstrated with 23% and 22%, respectively, being 13 C 5 labelled ( Fig 5A). We did not detect sedoheptulose 7-phosphate, octulose 8-phosphate or erythrose 4-phosphate in the LC-MS data suggesting that transketolase activity is absent, as reported previously for BSF T. brucei [13,40]. Hence, the non-oxidative branch of the PPP is not activated under hexose-transporter knockdown and glycerol-replete conditions and the labelling pattern of the glycolytic end products is consistent with this activity being absent, i.e. all three or no carbons are labelled in phosphoglycerate, phosphoenolpyruvate and pyruvate; carbon shuffling would have been indicative of the non-oxidative branch of the PPP (Fig 4, S2 Fig, S1 Table). 13 C carbons from glycerol are incorporated into numerous additional metabolic intermediates under hexose-transporter knockdown and glycerol-replete conditions (S1 Table). Alanine is produced as an end product from pyruvate, and 13 C 3 -alanine comprises 63% of the total following THT1/THT2 knockdown; 17% in wild-type cells (S2 Fig, S1 Table). Orotate, an intermediate of pyrimidine synthesis, comprises 50% fully labelled following THT1/THT2 knockdown (S2 Fig), and the label is further incorporated into pyrimidines. For example, UDP contains a substantial proportion of 13 C 2 from orotate (32%) and 13 C 5 from ribose 5-phosphate (14%), resulting in 13 C 7 (10%, S2 Fig). Glycerol is also incorporated into glycerophospholipids, e. g. glycerol-3-phosphoinositol, glycerol-3-phosphocholine and glycerol-3-phosphoethanolamine (S2 Fig). We also detected a substantial proportion of N-acetylglucosamine with two, three, six and eight carbons labelled following THT1/THT2 knockdown (S2 Fig).  Fig 4B. (B) VSG was purified from wild-type and THT1/ THT2 RNAi knockdown parasites grown with and without 13 C glycerol, as indicated, and subjected to monosaccharide compositional analysis by GC-MS. Following correction for natural isotope abundance, the proportion of VSG-derived mannose and galactose residues containing 13 C carbon from glycerol were determined. For all analyses, n = 3; � p < 0.0001, Student's t-test.
https://doi.org/10.1371/journal.ppat.1007475.g005 are the Gal and Man donors for the glycosyltransferases involved in protein N-glycosylation and glycosylphosphatidylinositol (GPI) anchor addition and processing in the parasite, including for the major cell surface component, the variant surface glycoprotein (VSG). Thus VSG is a convenient terminal metabolite with which to assess GNG. Soluble form VSG was purified from wild-type and THT1/THT2 knockdown parasites grown in the presence of 13 C glycerol and subjected to monosaccharide composition analysis by GC-MS, following methanolysis and trimethylsilyl-(TMS-) derivatization. Electron impact mass spectra of methyl-glycoside TMS derivatives produce an intense [(CH 3 ) 3 Si-O-CH = CH-O-Si(CH 3 )] + fragment ion at m/z 204, where the sugar-derived atoms are indicated in bold. The natural relative (to m/z 204) abundance of m/z 206 from this composition is 8.6%. We therefore interpret relative abundance values of m/z 206 of >8.6% as being synonymous with the appearance of 13 C from glycerol into the respective monosaccharides (S3 Fig). Our sample analyses indicated that 26 ± 7% of Man and 21 ± 5% of Gal was 13 C labelled following THT1/THT2 knockdown in the presence of 13 C glycerol, whereas labelling in all the other samples was not statistically significant ( Fig 5B). The labelling of Man and Gal in the VSG N-linked and GPI glycans from 13 C glycerol shows that G6P from GNG is used to drive terminal metabolite biosynthesis, as well as intermediate metabolism.

GNG-specific fructose-1,6-bisphosphatase activity from T. brucei
Most of the enzymatic steps of glycolysis are reversible and used in GNG. The fructose 6-phosphate to fructose 1,6-bisphosphate step, catalyzed by phosphofructokinase, is often irreversible, however, such that the reverse reaction requires a GNG-specific fructose-1,6-bisphosphatase (FbPase, EC 3.1.3.11, see Fig 4A) activity. Thus, utilization of glycerol to generate hexose 6-phosphates typically requires FbPase activity, and our metabolomics data above suggest the presence of such an activity in T. brucei grown on glycerol and even in wild-type cells. It was failure to detect this activity that previously suggested the absence of GNG in T. brucei, however [13]; with a limit of detection, 0.5 nmol/min/mg protein.
We measured FbPase activity using a coupled enzymatic assay (Fig 6A). FbPase enzyme activity was indeed detected in both wild-type cells grown in the absence of glycerol and hexose transporter knockdown cells grown in the presence of glycerol, at 10.6 ± 4.5 nmol/min/mg protein (n = 3) and 15.6 ± 3.5 nmol/min/mg protein (n = 3), respectively (Fig 6B). These measurements correspond with the metabolomic data, since FbPase activity is required for the incorporation of carbons from glycerol into G6P. Thus, BSF T. brucei has the enzymatic capacity to perform GNG, and the activity of the pathway appears to be specifically increased under hexose-transporter knockdown and glycerol-replete conditions.
T. brucei express a glycosomal FbPase orthologue (Tb927.9.8720) [7] in both the BSF and PCF life-cycle stages [44]. To further explore the source of the FbPase activity detected above, we generated Tb927.9.8720-null cells in which the protein coding sequences were precisely removed (Fig 6C), using a Cas9-based editing approach [45]. An FbPase activity at 11.2 ± 7.1 nmol/min/mg protein (n = 2 replicates of 2 populations) was also detected in these null cells grown in the absence of glycerol (Fig 6D). Thus, T. brucei bloodstream form cells express a Tb927.9.8720-independent source of FbPase activity.

Metabolic compensations for gluconeogenesis
The metabolomic analysis indicated other notable labelled metabolites under hexose-transporter knockdown and glycerol-replete conditions, possibly indicating metabolic compensations for GNG (Fig 7A). Specifically, the total amount of fumarate and malate is two-fold increased under these conditions (Fig 7A), suggesting higher flux in both the TCA cycle and the succinate shunt. Between 2-5% of malate, fumarate, and aspartate are 13 C 2 labelled in wild-type cells, which increases three fold (to 8-10%) under hexose-transporter knockdown and glycerol-replete conditions (Fig 7A and 7B). This may be due to increased activity of the tricarboxylic acid (TCA) cycle, since 13 C carbons from glycerol are passed into 13 C 2 labelled acetyl-CoA, which is then incorporated into TCA cycle intermediates (Fig 7D). Acetyl-CoA was not detected in the LC-MS analysis, but N-acetyl-lysine, GlucNAc (S2 Fig, S1 Table) and other intermediates containing 13 C 2 indicate labelling from acetyl-CoA. Operation of the TCA cycle was also suggested recently for adipose tissue resident T. brucei [1]. coupled enzymatic assay used for fructose-1,6-bisphosphatase (FbPase) activity detection. F1,6bP is added to cell extract, and if FbPase is present, it converts F1,6bP into F6P. F6P is converted into G6P by glucose-6-phosphate isomerase (PGI) and further into 6-phosphogluconolactone by glucose-6-phosphate dehydrogenase (G6PDH) producing NADPH. Further, 6-phosphogluconate is made by 6-phosphogluconolactonase, followed by a 6-phosphogluconate dehydrogenase reaction resulting in ribulose 5-phosphate and additional NADPH. PGI and G6PDH were added to the reaction, while the latter two enzymes are present in the cell extract. NADPH production is detected spectrophotometrically. (B) FbPase activity detected in WT (no glycerol in growth medium) and THT1/THT2 knockdown cells induced with tetracycline (plus 5 mM glycerol) for 3 days, and a negative control where F1,6bP was omitted. The lines show means of replicates (n = 3) and shaded areas indicate SD. The slopes measured for WT and THT1/THT2 knockdown cells are significantly different Under glycolysis, dihydroxyacetonephosphate (DHAP) is converted into glycerol 3-phosphate (Gly 3-P) by glycosomal glycerol-3-phosphate dehydrogenase (G3PDH), producing oxidised NAD + , but during GNG this source of NAD + is lost and must be compensated for. Since the glycosomal membrane is impermeable to large metabolites such as NAD(H), NAD/H, recycling has to be balanced within the glycosome [46]. Upregulation of the glycosomal succinate shunt could compensate for this deficiency through malate dehydrogenase activity and NADH dependent fumarate reductase activity. Consistent with this view, we observe similar labelling for malate, fumarate and aspartate (Fig 7A), with relatively small proportions (8-11%) of these metabolites containing 13 C 3 label (Fig 7C) in wild-type T. brucei in the presence of glycerol, which increases under hexose-transporter knockdown and glycerol-replete conditions; approximately three-fold, and reaching 36-47% (Fig 7C). These metabolites are likely products of phosphoenolpyruvate (PEP) being converted into oxaloacetate, and further into malate, fumarate and succinate. This suggests upregulation of the glycosomal succinate shunt (Fig 7D) to compensate for NAD/H balance in the glycosome when glycosomal G3PDH is  Fig 4B. (B) Proportion of these metabolites with two 13 C labelled carbons in WT and THT1/THT2 RNAi cells, � p < 0.001. (C) Proportion of these metabolites with three 13 C labelled carbons in WT and THT1/THT2 RNAi cells, � p < 0.001. (D) Scheme indicating potential explanation for the labelling patterns observed. The 13 C 2 part may be produced by the TCA cycle activity fed by 13 C 2 labelled acetyl-CoA, advantageous for additional ATP production. The 13 C 3 part could be produced in the glycosomal succinate shunt, which is activated in order to maintain the NAD + /NADH balance inside the organelle.
Thus, since use of glycerol produces only a single molecule of ATP per molecule of glycerol when GNG is operative, the cells may compensate by producing ATP in the mitochondrion, either by the TCA cycle activity or by acetate:succinyl-CoA shunt via succinyl-CoA synthetase converting succinyl-CoA into succinate [47].

Discussion
It has been thought for decades that mammalian forms of T. brucei, the causative agents of African sleeping sickness in humans and nagana in cattle, are exclusively dependent on glycolysis, using glucose as a substrate, for ATP production. We now provide evidence that these cells can utilise glycerol for gluconeogenesis (GNG) and for ATP production. Thus, metabolism in these parasites is not as simplified and reduced as had been thought. Indeed, we demonstrate some GNG even in wild-type cells grown on glycerol, even when glucose is also available. Metabolism may indeed be simple and dependent on glycolysis in blood, but it now seems likely that GNG can be activated in different environments.
Creek and colleagues [11] utilised metabolomic analyses to demonstrate extensive utilization of glucose in a wide range of metabolic pathways. We now report the utilization of glycerol for many of those same pathways. FbPase is considered to be one of the rate-limiting activities of GNG in other systems [48], but this activity has not previously been detected in T. brucei and therefore kinetic parameters have not been established. An annotated FbPase (Tb927.9.8720) is present in trypanosome glycosomes [7] and is expressed at similar protein levels in both bloodstream and insect life-cycle stages [44], the transcript levels are similarly not significantly different between culture, blood, or adipose tissue derived BSF [1]. We now show that FbPase activity is present in BSF T. brucei cells and increases when GNG is activated. We also show, however, that this activity is Tb927.9.8720-independent. Thus, the dephosphorylation of F1,6bP to F6P in T. brucei may require reversal of the phosphofructokinase reaction or sedoheptulose-1,7-bisphosphatase (Tb927.2.5800) activity [49].
In addition to FbPase, phosphoenolpyruvate carboxykinase (PEPCK) is the other enzyme specific for GNG, and responsible for its regulation in mammals [48]; this enzyme is also known to be present and active in T. brucei [11]. As expected, a third GNG-specific enzyme in animals, glucose-6-phosphatase, is not present in T. brucei, nor in other protozoa, since loss of the phosphate group would allow free glucose to diffuse out of the cells [48]. Previous modelling and metabolomics of glycolysis suggested high reverse flux of aldolase [11], however, similar information is lacking for the other enzymes. T. brucei glycerol kinase is unique in its bidirectional activity, and is known to lack the classical allosteric regulation [50]. According to canonical biochemistry, glycolysis and GNG cannot operate simultaneously, and the exclusive regulatory mechanisms are well known in mammalian systems. However, this classical regulation is missing in the T. brucei enzymes [34,51]. Compartmentalization in glycosomes would not present a solution if both glycolysis and FbPase activity are localised inside glycosomes [7] and we cannot exclude futile cycling in T. brucei, which has been suggested in Toxoplasma gondii [52]. However, since we now demonstrate Tb927.9.8720-independent FbPase activity, further work will be required to determine whether this FbPase activity is indeed compartmentalized within glycosomes in T. brucei.
GNG is essential in other unicellular pathogens. It is vital for the mammalian stage of Leishmania due to the need for mannogen biosynthesis [53]. GNG is also vital for Toxoplasma gondii, regardless of whether infected host cells are rich or poor in glucose [52]. Mycobacterium tuberculosis is also dependent on GNG and harbours two independent FbPase genes [54].
GNG may be particularly important in mammalian form T. brucei to feed the essential pentose-phosphate and glycoprotein glycosylation pathways, since even if the cells produce ATP from the second half of glycolysis when glucose is limiting, G6P is required as a substrate for these other essential pathways.
Aquaglyceroporins (AQPs) transport glycerol, but have generally been considered important for glycerol efflux rather than acquisition [33]. AQPs can be used for glycerol uptake to fuel T. brucei metabolism, however [24]. Indeed, our current data are consistent with a report from 1962 on trypanosomes using glycerol as a substrate for respiration [22], and a report from 1977 on pyruvate production from glycerol [55]. Is glycerol the only GNG substrate in T. brucei? Leishmania can also use aspartate, alanine or lactate to feed GNG [56]. Although the same enzymatic repertoire is theoretically available in T. brucei [4], our findings suggest that amino acid supply in growth medium fails to fuel GNG in cultured bloodstream form T. brucei. On the other hand, glycerol may activate GNG, thereby allowing other substrates to be used.
Futile cycling between glycolysis and GNG would be energetically disadvantageous, but it is consistent with our observation that GNG is associated with a fitness cost. Depletion of triosephosphate isomerase, leading to production of one ATP per molecule of glucose, caused a severe growth defect in BSF T. brucei [57]. Similarly, one molecule of ATP is produced per molecule of glycerol when GNG is operative, while two molecules of ATP are produced per molecule of glucose in glycolysis. However, this may be compensated for by upregulating glycerol uptake. Decrease in glycolytic intermediates indicates decrease of flux in glycolysis, which may be a consequence of futile cycling. In addition, glycerol must be phosphorylated in the initial step of GNG, potentially inhibiting glycolysis by depleting the ATP required to drive the hexokinase and the phosphofructokinase reactions.
In parallel with GNG, we do see evidence for metabolic compensation to balance ATP production. Additional ATP may be produced in the mitochondrion in the acetate:succinyl-CoA shunt, the TCA cycle directly by succinyl-CoA synthetase, or indirectly by feeding the electron transport chain with reduced cofactors from the TCA cycle. Specifically, labelling patterns observed in fumarate, malate, and aspartate support activity of the TCA cycle, as proposed previously in adipose tissue trypanosomes [1]; pyruvate is converted into acetyl-CoA, which then introduces two labelled carbons into the TCA cycle. An alternative explanation for this labelling pattern is that the enzymes, malate dehydrogenase and fumarase, are working in both directions and shuffling carbons, as a consequence of GNG disrupting the usual metabolic steady state. The glycerol-derived 13 C 3 component of fumarate, malate, and aspartate; comprising about 8% in wild-type, increased three-fold following hexose transporter knockdown. This may reflect glycosomal succinate shunt activity or the reverse action of TCA cycle enzymes. Notably, increased activity of the glycosomal succinate shunt would also serve to regenerate NAD + and maintain NAD + /NADH balance inside the glycosome.
Various mammalian host tissues may provide glycerol in quantities sufficient for GNG in trypanosomes. Infection in the cerebrospinal fluid is well known. However, little is known about parasite metabolism in the central nervous system and glycerol concentration is lower than in blood [36]. Adipose tissue contains glucose at concentrations about seven fold lower than in plasma [58] and glycerol at concentrations about four fold higher than in plasma [35,36], although these levels may be variable. Adipose tissue form trypanosomes may possess metabolically active or upregulated pathways, which are silent in the BSF, i.e. the TCA cycle and fatty acid β-oxidation [1], and we show here that GNG is active in the presence of glycerol. T. brucei are also present in skin, but their metabolism has yet to be scrutinised in this tissue [2,3]. Trypanosomes may never encounter environments completely lacking glucose under physiological conditions in mammalian hosts. Notably, in this regard, GNG is operative when both glucose and glycerol are present. Incomplete labelling in our glycerol-fed metabolomics analyses suggests that, although glucose uptake was below the detection limit of our assay in the absence of hexose transporters, some glucose was likely still imported by endocytosis. An ability to use both substrates may be crucial for adaptation to particular tissue environments or during transitions between tissues.
Our metabolomics analysis indicates that, in the presence of a suitable substrate, GNG does operate in mammalian form T. brucei. The pathway even operates, albeit at a relatively low level, in wild-type cells in the presence of glycerol. Cells that have limited access to glucose, in this case following hexose transporter knockdown, display a major increase in flux through GNG. We conclude that GNG in T. brucei, using glycerol taken up via aquaglyceroporins [24], could be important for colonization of, and survival in, different host tissue environments.

Microscopy and immunofluorescence microscopy assay
For microscopy, cells were washed with 1 x PBS and fixed in methanol-free 3% formaldehyde (Thermo Scientific) for 15 min at RT. Following two washes in PBS, cells were resuspended in 1% bovine serum albumin (Sigma-Aldrich) and allowed to dry on microscopy slides. For direct fluorescence microscopy, slides were immediately mounted with Vectashield with DAPI (Vector Laboratories). For immunofluorescence microscopy, slides were blocked with 50% FBS in PBS for 15 min at RT and, after two washes with PBS, primary α-VSG-2 (1:10,000; [62]) or α-EP1 (1:1,000; Cedarlane) antibodies were applied for 1 h at RT. Following 3 washes in PBS, the secondary antibodies, α-rat IgA-rhodamine (1:1,1000; Sigma-Aldrich) and αmouse Alexa 568 (1:1,000; Life Technologies), respectively, were applied for 1 h at RT, followed by a further three washes in PBS. Images were captured using an Axiovert 200 epifluorescence microscope and processed using Zen imaging software (Zeiss).

2-14 C(U)-deoxyglucose uptake assay
The uptake assay was performed as described previously [24] with minor modifications. Briefly, 10 8 cells were harvested (1000 g, 10 min 4˚C), washed twice in ice-cold transport buffer without glucose (33 mM HEPES, 98 mM NaCl, 4.6 mM KCl, 0.55 mM CaCl 2 , 0.07 mM MgSO 4 , 5.8 mM NaH 2 PO 4 , 0.3 mM MgCl 2 , 23 mM NaHCO 3 , pH 7.3) and resuspended to 10 8 cells/ml in the same buffer. Uptake was initiated by adding 100 μl of cell suspension to 100 μl of transport buffer containing 0.25 μCi 14 C-2-deoxyglucose (PerkinElmer) layered over 100 μl of dibutyl phthalate (Sigma-Aldrich). After incubation at 37˚C or 4˚C for the appropriate time, transport was stopped by centrifugation through the oil layer (16,000 g, 1 min). Microcentrifuge tubes were flash frozen in liquid nitrogen and the bottoms of the tubes, containing the cell pellets, were snipped into scintillation vials. Pellets were solubilised overnight in 150 μl of 1 M NaOH before mixing with 2 ml of scintillation fluid and radioactivity was measured on a scintillation counter (Beckman LS 6500) for 1 min.

LC-MS metabolomic analysis
For the liquid chromatography-mass spectrometry (LC-MS) metabolomic analysis the sample extraction was performed as described previously [11]. THT1/THT2 knockdown cells were grown in the presence of tetracycline and 5 mM 13 C 3 -U-glycerol (Sigma-Aldrich) three days prior to sample preparation. Briefly, 5 x 10 7 cells were used for each final 200 μl sample. Cells were rapidly cooled in a dry ice/ethanol bath to 4˚C, centrifuged at 1,300g, 4˚C for 10 min, washed with 1 x PBS, and resuspended in extraction solvent (chloroform:methanol:water, 1:3:1 volume ratio). Following shaking for 1 h at 4˚C, samples were centrifuged at 16,000g at 4˚C for 10 min and the supernatant was collected and stored at -80˚C. The analysis was performed using separation on 150 x 4.6 mm ZIC-pHILIC (Merck) on Dionex UltiMate 3000 RSLC (Thermo Scientific) followed by mass detection on an Orbitrap QExactive mass spectrometer (Thermo Fisher) at Glasgow Polyomics. Analysis was operated in polarity switching mode, using 10 μl injection volume and a flow rate of 300 μl/min. The samples were run alongside 170 authentic standards. The data were processed and analyzed using mzMatch software [63] and mzMatchISO [64]. The analysis was performed in 4 replicates, means of which are indicated, non-labelled samples were run in parallel. Metabolites were identified based on matches with standards or were predicted based on mass and retention time. Metabolomics data have been deposited to the EMBL-EBI MetaboLights database (https://www.ebi.ac.uk/ metabolights/index) with the identifier MTBLS706.

VSG glycosylation analysis
The VSG isolation was performed as described previously with minor modifications [65]. The cells were cultured in 5 mM 13 C 3 -U-glycerol (Sigma-Aldrich) for three days prior to sample preparation, and unlabelled samples were prepared alongside. 10 8 cells were harvested (1,300 g, 4˚C, 10 min), washed twice in ice-cold PBS and resuspended in 300 μl of 10 mM Na 2 HPO 4 pH 8.0 in MS grade water, containing 0.1 mM TLCK (Sigma-Aldrich), 1 μg/ml leupeptin (Sigma-Aldrich), 1 μg/ml apoprotinin (Sigma-Aldrich), and 10 mM PMSF (Sigma-Aldrich). Following 5 min incubation at 37˚C, the samples were cooled on ice and centrifuged at 14,000 g, 4˚C for 5 min. The samples were applied on chromatography columns containing 400 μl of bead mixture (50:50 volume ratio, Anion Exchange Cellulose DE52 (Whatman) in 10 mM Na 2 HPO 4 pH 8.0 buffer) and eluted with 800 μl of 10 mM Na 2 HPO 4 pH 8.0. The obtained eluate was concentrated and diafiltered against water using Millipore Amicon Ultra-0.5 Centrifugal Filter Devices (Merck) following the manufacturer's instructions. Carbohydrate compositional analysis was performed by GC-MS. Samples (6-10 μg) were mixed with 2 nmol scyllo-inositol internal standard and dried using vacuum centrifugation. Dried samples were then subjected to methanolysis by adding 50 μl of 0.5 M HCl in dry methanol and incubating at 85˚C for 4 h. Methanolysates were re-N-acetylated by the addition of 10 μl pyridine and 10 μl acetic anhydride and incubating at RT for 30 min. The samples were dried under vacuum and derivatised with 15 μl trimethysilylation (TMS) reagent at RT for 30 min. 1 μl aliquots of each sample was injected in GC-MS (Agilent Technologies, 7890B Gas Chromatography system with 5977A MSD) equipped with Agilent J&W HP-5ms GC Column (30 m X 0.25 mm, 0.25 μm) with He carrier gas at 0.5 ml/min. The temperature program used was run over 32.5 min with 95˚C (for 1 min) -140˚C (30˚C/min) to 265˚C at 5˚C/min (for 5 min). The mass spectra were collected from linear scanning over m/z 50-650, and quantification was based on the integration of the extracted ion-current chromatograms and empirically determined molar response factors. The size of the bars represents the total abundance, and coloured parts indicate 13 C labelling as depicted in the legend. The samples are from WT cells, WT grown in 13 C-glycerol, THT1/THT2 RNAi grown in 12 Cglycerol and THT1/THT2 RNAi grown in 13 C-glycerol. Natural abundance of 13 C is 1%, hence the 1C labelling in 'un-labelled' samples. � the identity of these metabolites was confirmed using a match with a standard. (PDF)

S3 Fig. Measurement of incorporation of 13 C glycerol into the mannose and galactose residues of VSG.
Total ion chromatograms of the methyl-glycoside TMS derivatives from wildtype (panels A-B) and THT1/THT2 knockdown trypanosomes (panels C-D) grown in the absence and presence of 13 C glycerol, respectively. The peaks due to mannose (Man), galactose (Gal) and the scyllo-inositol internal standard (s-I) are indicated. The insets show a detail of the electron impact mass spectra of the main Man peak, illustrating the natural abundance of m/z 206 relative to m/z 204 (panel A) compared to a sample where 13 C has been incorporated into the VSG sugar residues (panel D). One representative replicate, n = 3. (PDF) S1 Table. LC-MS metabolomics data with 13 C 3 -glycerol. The samples are from WT cells, WT grown in 13 C-glycerol, THT1/THT2 RNAi grown in 12 C-glycerol and THT1/THT2 RNAi grown in 13 C-glycerol, n = 4.