Comparative RNA-sequencing of T. congolense and T. brucei

To compare BSF T. congolense and T. brucei transcriptomes, RNAseq analysis was carried out on parasites cultured in vitro (sampled from actively dividing cultures grown to densities of 1.8–2.0 × 10⁶ cells/mL) and trypanosome samples isolated from infected mice at first peak parasitaemia (10⁷ cells/mL; ex vivo; Fig 1). T. congolense (IL3000) and pleomorphic T. brucei (STIB 247) in vitro and ex vivo samples were prepared, in order to assess similarities and differences between trypanosomes grown in culture and those from an infection (Fig 1A and 1B), and to compare and contrast the transcriptome across the species (Fig 1C and 1D). Sequencing data were aligned to the respective genome sequences with a mean overall alignment rate of 88.8 ± 6.5% and 92.6 ± 2.0% for T. brucei and T. congolense reads, respectively. The lower alignment rate for T. brucei was likely due to the TREU 927 reference genome having a different VSG repertoire than STIB 247. Resultant files were sorted and filtered for quality and read counts were normalised using transcripts per million (TPM) [56]. Orthologues were inferred between the species using Orthofinder [57], in order to directly compare TPM values for 1-to-1 orthologues, as well as sum-of-TPM values for groups containing families of paralogues (e.g. hexose transporters). These normalised read counts are henceforth referred to as orthoTPM values (S1 Table). The Orthofinder dataset (S2 Table) consisted of 6,677 orthogroups (denoted with the prefix “TbTc_”), of which 5,398 (80.84%) were 1-to-1 orthologues. The Orthofinder tool was also used to predict genes only present in one of the two species (S2 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Overview of comparative transcriptomics analysis of T. brucei and T. congolense, isolated from ex vivo and in vitro conditions.

RNAseq data from T. congolense (IL3000) and T. brucei (STIB 247) in both in vitro and ex vivo (from mouse infections) conditions were aligned to the species’ respective reference genomes and read counts were normalised by the transcripts per million (TPM) method. To directly compare the species, a pseudogenome was generated using the Orthofinder tool [57]. TPM values from the 4 sample groups were plotted against each other to analyse correlation between conditions (A and B) and between species in the same conditions (C and D). Correlation was assessed using both Pearson correlation (Pearson’s r) and Spearman’s rank correlation coefficients.

https://doi.org/10.1371/journal.ppat.1009734.g001

There are several metabolic genes that are not present in the T. congolense genome, including a putative delta-4 desaturase (Tb927.10.7100), a succinate dehydrogenase subunit (SDH11; Tb927.8.6890) and mitochondrial pyruvate carrier 1 (MPC1; Tb927.9.3780). These genes are encoded in regions exhibiting high levels of synteny between the species (S1 Fig) and thus seem likely to represent genuine deletions in T. congolense. Furthermore, guanine deaminase (GDA; Tb927.5.4560) was not identified in the T. congolense genome (S2 Table). Although the GDA locus is subtelomeric in the T. brucei genome, and these regions are not syntenic with T. congolense, sequences matching GDA could not be found either in the assembled genome or the available unassembled contigs. Therefore, the most parsimonious explanation is that GDA is also not encoded in the T. congolense genome. Based on the redundancy in trypanosomal purine salvage pathways, it is highly unlikely that GDA would be essential in BSF T. brucei [58].

Differences between four sample groups were assessed based on orthoTPM values (Fig 1; full dataset in S1 Table). There was a strong intra-species correlation between the in vitro and ex vivo transcriptomes (Pearson correlation coefficient, T. congolense (ρ) = 0.914, Fig 1A; T. brucei ρ = 0.838, Fig 1B), showing that in vitro-derived BSF T. congolense and T. brucei closely resemble parasites isolated from infections at the transcriptome level. Furthermore, although correlations between species in the same condition were also high (ex vivo: ρ = 0.622, Fig 1C; in vitro: ρ = 0.614, Fig 1D), the reduced correlation coefficient values do indicate a degree of transcriptional differences between the species.

To compare data from this study to BSF T. congolense transcriptomics data generated by Silvester et al. at ascending and peak parasitaemia [59], TPM values for each annotated T. congolense gene were compared directly (S2 Fig and S1 Table). There was good correlation between both in vitro and ex vivo T. congolense BSF datasets and the data from Silvester et al. (ρ > 0.85, S2 Fig), with highest correlation between the ex vivo and ascending data (ρ = 0.905, S2 Fig), albeit the correlation between the ‘ascending’ and ‘peak parasitaemia’ in Silvester et al. was higher (ρ = 0.988, S2 Fig).

T. congolense metabolite consumption and output

Global metabolite (metabolomics) analysis of in vitro culture supernatant samples provides a detailed insight into the metabolic inputs and outputs of cultured cells [60]. However, high levels of medium components can often mask subtle but significant changes in culture medium composition over time. To counteract this, a modified culture medium was designed for T. congolense strain IL3000, based on previously published medium formulations (Steketee’s congolense medium, SCM-3; for details see Materials and Methods) [17,18].

A BSF T. congolense IL3000 time course was initiated, with cells isolated during mid-exponential growth phase inoculated into fresh SCM-3 medium (0 h time point). Culture supernatant samples were collected at 0, 8, 24, 32, 48 and 56 hours (n = 4 at each time point) and metabolites extracted for LC-MS analysis.

A total of 290 putative metabolites were detected across all samples (207 after removing putative metabolites that did not map to metabolic pathways, e.g. peptides and medium components), of which 37 were matched to an authentic standard to confidently predict their identity (S3 Table). Of the 207 metabolites in the final dataset, 100 were putatively annotated as lipids. Annotations of putative mono- and poly-unsaturated fatty acids are of lower confidence due to the challenges of identifying these metabolite species using this LC-MS platform (see Materials and Methods for details).

80 of the 207 putative metabolites were significantly altered across the dataset (false discovery rate-adjusted p<0.05; one-way repeated measures ANOVA; Fig 2A and S3 Table). To analyse metabolites undergoing similar changes, K-means clustering with Pearson correlation coefficient as the similarity metric was used, highlighting seven clusters with two clusters of particular interest: one containing metabolites that accumulated over time, and the other containing metabolites depleted over time (Fig 2A). Log₂ fold change (log₂ FC) between the first and final time points (0 and 56 h, respectively) was also calculated for each metabolite (S3 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Analysis of supernatant metabolites after T. congolense culture.

A heatmap covering the 80 putative medium components judged to be significantly altered after 56 hours of in vitro cell culture containing T. congolense strain IL3000, as calculated by a one-way repeated measures ANOVA (P < 0.05). Peak abundances were log transformed and mean centred and metabolites were clustered based on Pearson correlation. Two clusters of interest were identified, which are shown in a larger format on the right. Metabolites in the top cluster were observed to increase significantly over time, whilst those in the bottom cluster decreased. Metabolite names followed by {*} were matched to an authentic standard and all other identifications are putative based on mass, retention time and formula. B) Comparison of metabolite changes in medium supernatants after 56 hours between T. brucei [60] and T. congolense (S3 Table). Relative changes in metabolite abundance were calculated as log₂ fold change of 56 h vs 0 h and metabolites exhibiting differences between the species are listed next to the figure.

https://doi.org/10.1371/journal.ppat.1009734.g002

Glucose, the primary energy source for T. brucei, whilst clearly consumed, was not fully depleted after 56 hours in T. congolense culture (log₂ FC: -0.76; Figs 2A and 3A), in contrast to T. brucei, where 10 mM glucose is consumed by the same time-point [60]. Ribose, glucosamine, inosine and threonine were similarly depleted in T. congolense culture (log₂ FC: -0.78, -0.97, -2.82 and -0.89, respectively).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Energy metabolism in T. congolense.

A-E) Supernatant metabolomics analysis of metabolites involved in glycolytic metabolism in T. congolense. Grey bars indicate a negative medium control incubated for 56 hours. F) A commercial kit was used to measure acetate concentration during T. congolense culture, with supernatant samples analysed at the same time points as the supernatant metabolomics experiment. G) A simplified overview of the glycolytic pathway. Numbers refer to the following proteins: 1, hexokinase; 2, glucose 6-phosphate isomerase; 3a, phosphofructokinase; 3b, fructose-1,6-bisphosphatase; 4, aldolase; 5, triosephosphate isomerase; 6, glycerol 3-phosphate dehydrogenase; 7, glycerol kinase; 8, glyceraldehyde 3-phosphate dehydrogenase; 9, phosphoglycerate kinase; 10, phosphoglycerate mutase; 11, enolase; 12, phosphenolpyruvate carboxykinase; 13, malate dehydrogenase; 14, fumarate hydratase; 15, NADH-dependent fumarate reductase; 16, pyruvate kinase; 17, alanine aminotransferase; 18, pyruvate dehydrogenase complex; 19, acetate:succinate CoA-transferase and acetyl-CoA thioesterase. H) Tracing glucose derived carbon usage through glycolytic metabolism. T. congolense were incubated with a 50:50 mix of ¹²C-D-glucose:¹³C-U-D-glucose before cell pellets were isolated for metabolomics analysis. Results were compared to those generated in T. brucei by Creek and colleagues [45]. Colours indicate the number of ¹³C atoms in each metabolite. I) Comparative analysis of transcript level activity of glycolysis in T. brucei and T. congolense from both in vitro and ex vivo conditions. Gene IDs: HK1 & 2, hexokinase, TbTc_0341; GPI, glucose 6-phosphate isomerase, TbTc_1840; PFK, phosphofructokinase, TbTc_1399; ALDA, aldolase, TbTc_0358; TPI, Triosephosphate isomerase, TbTc_1075; GPDH, glycerol 3-phosphate dehydrogenase, TbTc_2722; GK, glycerol kinase, TbTc_0392; GAPDH, glyceraldehyde 3-phosphate dehydrogenase, TbTc_0377; PGK, phosphoglycerate kinase, TbTc_6030; PGKA, phosphoglycerate kinase A, TbTc_0241; PGKB/C, phosphoglycerate kinase B & C, TbTc_0240; PGM, phosphoglycerate mutase, TbTc_5039; ENO1, enolase, TbTc_0465; ENO, putative, enolase, putative, TbTc_3614; PK1, pyruvate kinase 1, TbTc_0372; FBPase, fructose-1,6-bisphosphatase, TbTc_1967; PEPCK, phosphoenolpyrvuate carboxykinase, TbTc_0348; gMDH, glycosomal malate dehydrogenase, TbTc_0642; FH, fumarate hydratase, TbTc_0242; Frd, NADH-dependent fumarate reductase, TbTc_0141; PPDK, pyruvate phosphate dikinase, TbTc_1304; AAT, alanine aminotransferase, TbTc_0675; PDH E1α, pyruvate dehydrogenase E1 alpha subunit, TbTc_4169; PDH E1β, pyruvate dehydrogenase E1 beta subunit, TbTc_5437.

https://doi.org/10.1371/journal.ppat.1009734.g003

In contrast, a number of metabolites accumulated in the medium (Fig 2A). The most significant of these were guanine (log₂ FC: 6.34; Fig 2A), succinate (log₂ FC: 5.60; Figs 2A and 3B) and (S)-malate (malate, log₂ FC: 1.37; Figs 2A and 3B). Interestingly, pyruvate (log₂ FC: 0.24; Fig 3B) was not excreted at the high levels relative to starting concentration, in contrast to BSF T. brucei culture where pyruvate secretion is consistently observed in both HMI-11 and in Creek’s Minimal medium (CMM) [60]. Instead, succinate and malate appear to be the primary glycolytic outputs from BSF T. congolense, which is similar to PCF T. brucei. Elevated levels of 2-oxoglutarate and a metabolite putatively identified as 2-oxoglutaramate were observed, which potentially originate from alanine aminotransferase activity using L-glutamate and L-glutamine, respectively, as substrates [45,61]. Moreover, a significant build-up of N6-Acetyl-L-lysine (log₂ FC: 6.30) was observed (Fig 2B). Whilst the low molecular weight of acetate means it could not be detected by the LC-MS platform used, concentrations of this molecule were measured directly using an acetate assay in samples taken at the same time points from four independent cultures, which confirmed high levels of acetate excretion by BSF T. congolense (Fig 3F).

Other notable observations included the depletion of several putatively identified lysophosphatidylcholine species at 56 hours (Fig 2A and S3 Table), as seen in T. brucei [60], coincident with increased medium levels of sn-glycero-3-phosphocholine, choline and choline phosphate, indicating lyso-phospholipase activity where the charged headgroup moiety of a lyso-species is cleaved from its bound fatty acid [62]. However, given the putative nature of these fatty acid annotations, we could not confidently establish the origin of the elevated choline-related metabolites. In addition, tryptophan (log₂ FC: -0.74; S3 Table) was significantly consumed (p = 0.042), in contrast with cysteine (log₂ FC: -0.07; p > 0.05), despite the latter being essential to T. brucei [63] (S3 Table).

The log₂ metabolite fold changes after 56 hours of culture of T. congolense were compared to those of T. brucei grown in HMI-11 (Fig 2B) [60]. A total of 90 metabolites were identified in both datasets, with some showing divergence between the two species (Fig 2B). Several metabolites only accumulated in T. brucei supernatant, in particular pyruvate, D-glycerate, 2-oxoglutarate and hydroxydodecanoic acid (Fig 2B). Conversely, succinate, N6-acetyl-L-lysine, 4-hydroxy-4-methylglutamate, N6,N6,N6-trimethyl-L-lysine and choline only accumulated in T. congolense supernatant (Fig 2B). Whilst cystine (Fig 2B; 12) was depleted in T. brucei samples, this metabolite remained unchanged in those from T. congolense.

In summary, whilst core elements of metabolism have been conserved between BSF T. congolense and T. brucei, several pronounced differences in T. congolense metabolism were identified based solely on metabolic input and output in in vitro culture. An integrated analysis of the metabolomic and transcriptomic datasets was then undertaken in order to further define the metabolic differences between the two species.

Energy metabolism

As described above, RNA sequencing and culture supernatant metabolomics provided initial indications that BSF T. congolense energy metabolism, specifically with respect to glucose usage, diverges substantially from that of BSF T. brucei (simplified map of glycolysis depicted in Fig 3G). To dissect metabolic differences at the transcriptome level, pathway analysis was carried out using the TrypanoCyc database [64], which contains 186 manually curated pathways covering 288 genes or groups of multi-copy genes (S4 Table). These analyses showed broadly similar levels of gene expression of glycolytic components between BSF T. brucei and T. congolense (Fig 3G and 3I). However, the T. brucei ex vivo samples displayed a more distinct expression profile, with low transcript abundances for most glycolytic components compared to all sample groups. This is most likely the result of cells being sampled near peak parasitaemia, and as the pleomorphic strain STIB 247 was used, having a higher proportion of tsetse-transmissible, quiescent short stumpy forms–consistent with this there was elevated expression of stumpy markers such as the PAD array (TbTc_0074), PIP39 (TbTc_0700) and reduced expression of RBP10 (TbTc_0619) (S1 Table) [65–67].

Transcripts associated with gluconeogenesis, the succinate shunt, and the acetate generation pathway were increased in abundance in BSF T. congolense under both in vitro and ex vivo conditions compared to BSF T. brucei. Key examples of this are PPDK, PEPCK, glycosomal malate dehydrogenase (gMDH) and two subunits of pyruvate dehydrogenase (PDH) (Fig 3I). PPDK was previously reported to be expressed in BSF T. congolense, but not in BSF T. brucei [47], and it may be assumed that the enzyme serves a similar function in BSF T. congolense as it does in PCF T. brucei; primarily in a glycolytic role to maintain ATP/ADP balance in the glycosome. The high levels of gMDH expression in BSF T. congolense contrasts with BSF T. brucei, where gMDH expression is reported to be mostly absent, and cytosolic MDH (cMDH) is the major isoform [68]. The RNAseq analysis also supports a previous study showing high levels of glycerol kinase expression in BSF T. congolense [48].

To test whether elevated levels of succinate and malate seen in T. congolense supernatant medium samples originated from glucose, LC-MS analysis using ¹³C-U-D-glucose was carried out on intracellular metabolites isolated from in vitro-cultured cells. Stable isotope analysis has provided valuable insights into T. brucei central carbon metabolism [45], and generating T. congolense datasets enabled comparative analysis of glucose catabolism (albeit with an unavoidable difference in medium supplementation of goat serum for T. congolense, rather than foetal bovine serum for T. brucei).

BSF T. congolense was grown for 48 hours in a custom medium (SCM-6; S5 Table), containing a total D-glucose concentration of 10 mM in a 1:1 ratio of D-glucose:¹³C-U-D-glucose prior to LC-MS analysis. Labelling ratios of downstream metabolites were largely similar to that of intracellular glucose, and carbon labelling patterns matched those that would be expected in the BSF T. brucei glycolytic pathway (i.e. three ¹³C atoms in all metabolites downstream of glyceraldehyde 3-phosphate and glycerol-3-phosphate). Similar to T. brucei, a high percentage of 3-carbon labelled fructose-1,6-bisphosphate (FBP) (34.8%) was observed in T. congolense (Fig 3H), probably a result of the “reverse” aldolase reaction occurring in the glycosome [45]. Importantly, two-carbon labelling was observed in several acetylated compounds (N-acetylornithine & N-acetyl-L-lysine; Fig 3H), confirming that acetyl groups used to generate these metabolites originate from D-glucose. Although acetyl-CoA, the product of pyruvate oxidation, was not detected for technical reasons, the 2-carbon labelling patterns of acetylated metabolites suggests that the acetyl moieties in these compounds originate from glucose, potentially through acetyl-CoA and acetate, as previously evidenced in other trypanosomatids [45,69]. Taken together, these data indicate that the flow of carbon atoms for glycolytic components in T. congolense is very similar to that in T. brucei. However, the metabolic outputs differ drastically from BSF T. brucei and appear to be more similar to PCF T. brucei.

Metabolite labelling was corrected for the 1:1 (50%) ratio of natural glucose to ¹³C-U-D-glucose, which equated to a mean percentage labelling of 43.2% (the value is less than 50% due to D-glucose in the serum). All glycolytic metabolites up to pyruvate showed >40% labelling when corrected (for glucose 6-phosphate and fructose-1,6-bisphosphate, both 3-carbon and 6-carbon labels were taken into account), although glycerol and glycerol 3-phosphate exhibited 28.5% and 32.5% labelling, respectively, as these metabolites can also be obtained from catabolism of lipid precursors. Moreover, 42.4% (49.1% corrected) labelling was detected in L-alanine, suggesting that the alanine aminotransferase reaction that utilizes pyruvate to generate 2-oxoglutarate and L-alanine in both BSF and PCF T. brucei, also occurs in BSF T. congolense [45,70]. For both malate and succinate, 3 carbons were derived from glucose and these metabolites showed 40.2% (46.6% corrected) and 32.3% (37.7% corrected) labelling, respectively. These results suggest that glucose is not the only source of intracellular succinate and malate in T. congolense. However, these values were higher than those reported in T. brucei (35% and 26% for malate and succinate, respectively [45]).

Whilst PCF T. brucei exhibit TCA cycle activity, this pathway is not used to catabolize glucose [38,40]. No citric acid cycle intermediate isotopologues (e.g. citrate) were found when BSF T. congolense were incubated with ¹³C-U-D-glucose, indicating that like in T. brucei, T. congolense TCA metabolism is not linked to glycolysis. However, small amounts of 2-carbon labelled succinate and malate were observed (Fig 3H), which can be explained by the reversal of the glycosomal succinate shunt. Taken together, these data suggest that BSF T. congolense, both from in vitro cultures and in vivo infection, metabolically resemble an intermediate between BSF and PCF T. brucei, with moderate glycolytic capacity and significant levels of succinate shunt activity (glycosomal, rather than mitochondrial; S1 Table) as well as a highly active mitochondrial acetate generating pathway.

Previous work has shown that reduction of glucose concentrations in BSF T. brucei culture from 10 mM to 5 mM leads to decreased cellular motility, reduction in growth and cell body rounding morphology within 8 hours [71]. Given that glucose was not substantially depleted in T. congolense cultures after 56 h, we tested the effect of reduced glucose concentrations on BSF T. congolense viability. Unlike T. brucei, T. congolense was able to maintain a growth rate equal to controls at concentrations as low as 2 mM (Fig 4A) when continuously passaged with no observable change in morphology or motility. To test whether glucose-derived ATP was essential to T. congolense, cells were incubated with D-glucose in addition to varying concentrations of 2-deoxy-D-glucose (2DG), which can be internalised, but not metabolised further than 2-deoxy-D-glucose 6-phosphate. As a result, glycolysis and pentose phosphate pathway metabolism are inhibited and glycosomal ATP levels are depleted as they cannot be regenerated in the latter stages of glycolysis, leading to a drastic change in glycosomal ATP/ADP ratio [72] (Fig 4B). Incubation of T. congolense in medium supplemented with 2DG (in addition to 10 mM glucose) led to growth defects in a dose dependent manner, likely due to 2DG being outcompeted by glucose at lower concentrations (Fig 4B). Although the growth defect was minor in the presence of 1 mM 2DG, there was a more pronounced reduction with 5 mM 2DG. When equimolar concentrations of glucose and 2DG were used, growth was repressed and cell death occurred within 48 hours (Fig 4B). T. congolense viability was also tested in SCM-6 in the presence of N-acetyl-D-glucosamine (GlcNAc), a sugar that inhibits glucose uptake [73] (Fig 4C). In the presence of 60 mM GlcNAc with 10 mM glucose, there was a moderate, yet significant (p < 0.0001 at 96 h, t-test of cell densities) growth defect in T. congolense (Fig 4C). Viability was further reduced when the same concentration of GlcNAc was used alongside 2 mM glucose (p < 0.0001 at 96 h, t-test of cell densities), the lowest concentration T. congolense could tolerate (Fig 4C). The rate of glucose consumption was measured by assaying glucose concentrations in cell culture supplemented with 4 mM glucose. Doubling times (6–7 hours and 11–12 hours for T. brucei and T. congolense, respectively) were taken into account, and rate of glucose uptake was shown to be 47.17 ± 27.91 nmol^-1 min^-1 10⁸ cells in T. congolense, significantly lower than the rate (132.18 ± 16.31 nmol^-1 min^-1 10⁸ cells) in T. brucei (n = 4, p = 0.0039; t-test) which was comparable to previous studies [27,61].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. In vitro analysis of glycolytic metabolism.

A) T. congolense remains viable in reduced glucose concentrations. A growth defect was only observed when glucose concentrations were reduced to <2 mM. B) Supplementation with increased concentrations of 2-deoxy-D-glucose leads to T. congolense cell death (red dotted line indicates detection limit by haemocytometer). C) Analysis of growth in the presence and absence of N-acetyl-D-glucosamine. Parasites were cultured in SCM-6 supplemented with 10 mM or 2 mM glucose in the presence or absence of 60 mM GlcNAc and density monitored by haemocytometer every 24 hours. D) Knock-down of the entire glucose transporter (TcoHT) array does not affect in vitro cell viability. RNAi was induced in three independent clones by the addition of 1 μg/mL tetracycline (Tet), and cell densities of induced and uninduced cells were monitored daily. E) Normalised TcoHT mRNA abundance over time after RNAi induction. F) Changes in glucose uptake in RNAi-induced cells were detected via an enzyme-linked luminescence assay coupled to 2-deoxy-D-glucose uptake over a period of 30 minutes. The assay was carried out 72-hours post-induction. Of the three RNAi lines, 2 showed a significant reduction in glucose uptake capability (Student’s T-test, *P < 0.05; ***P < 0.001).

https://doi.org/10.1371/journal.ppat.1009734.g004

To further probe the essentiality of glucose uptake in BSF T. congolense, RNAi was used to knock down expression of the hexose transporter (TcoHT) array, specifically those matching the THT1 and THT2 array in T. brucei (TcIL3000.A.H_000260500, TcIL3000.A.H_000260600, TcIL3000.A.H_000794500, TcIL3000.A.H_000794600, TcIL3000.A.H_000794700), which has been shown to significantly restrict growth of BSF T. brucei [74]. Whilst growth rate was unaffected in BSF T. congolense (Fig 4D), induction of HT RNAi led to a reduction in transcript abundance at all time points (mean transcript levels of 83%, 75%, 68% and 65% compared to uninduced controls at 24, 48, 72 and 96 h post-induction, respectively; Fig 4E). Glucose uptake was decreased (mean reduction of 37% in uptake compared to uninduced controls after 72 h; Fig 4F), suggesting that either lower levels of glucose are sufficient for energy generation in T. congolense, or the parasite can utilize other carbon sources for ATP production. These alternative sources could include medium or serum components such as amino acids [75].

PCF T. brucei express most components of the electron transport chain (ETC) to generate ATP through oxidative phosphorylation. In contrast, BSF T. brucei express an F₁F_o-ATPase that functions in reverse, and alternative oxidase [76]. In addition, a recent study has suggested that complex I is expressed and functional in BSF T. brucei [77]. Transcriptomics analysis of the ETC was carried out, using a gene list generated by Zikova and colleagues [76], but no significant patterns could be discerned, and thus we were not able to draw a conclusion with regards to ETC activity in BSF T. congolense based on transcriptomics data alone (S1 Table and S3 Fig).

Nucleotide metabolism

Metabolomic analysis of BSF T. congolense culture supernatants indicated a significant uptake of exogenous ribose, a contributor to nucleotide metabolism via uptake, or via the pentose phosphate pathway (PPP; Figs 2A and 5A). Whilst guanosine was not detected in the supernatant, significant accumulation of guanine (Fig 5B) was observed, suggesting either excretion of this metabolite, or, hydrolysis of guanosine through parasite-secreted hydrolases/nucleosidases (previously identified in BSF T. brucei secretomes [78,79]). This mechanism would enable uptake of guanine and other nucleobases through nucleobase transporters, for which multiple orthologues have been identified in the T. congolense genome [21] through homology with known T. brucei nucleobase transporters TbNT8.1 and TbNBT1 [80,81]. In addition, there was an accumulation of xanthine, a product of xanthosine hydrolysis, and depletion of inosine, an important nucleoside composed of hypoxanthine and ribose (Fig 5C and 5D). The nucleoside cytidine and the nucleobase hypoxanthine were also detected, but appeared to remain unchanged during the time course, although the latter was a medium supplement potentially added in excess (S3 Table). It is noteworthy that only a single nucleoside transporter gene (TbTc_1072; T. congolense gene IDs: TcIL3000.A.H_000665800 and the pseudogene TcIL3000.A.H_000679300; S2 Table) can be identified in T. congolense, which is a syntenic homologue of TbNT10 [21], functionally characterized as a P1-type purine nucleoside transporter [82], and is thus unlikely to transport cytidine [83].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Nucleotide metabolism in T. congolense.

Supernatant analysis of T. congolense in vitro cultures showing changes in abundance of D-ribose (A), guanine (B), xanthine (C) and inosine (D) over 56 hours. Grey bar indicates a negative medium control group E) Simplified overview of purine salvage and synthesis in trypanosomatids (adapted from [148]). Numbers indicate the following enzymes: 1, APRT; 2, AD; 3, HGPRT; 4, IMPD; 5, HGXPRT; 6, GMPR; 7, GMPS; 8, HGPRT. Red cross indicates guanine deaminase, which is not encoded/annotated in the T. congolense genome (based on current assembly). F) Comparison of glucose-derived purine carbon labelling in T. congolense and T. brucei [45]. Colours indicate the number of ¹³C atoms in each metabolite. G) Comparative RNAseq analysis of T. congolense and T. brucei under both in vitro and ex vivo conditions. Gene IDs from top to bottom: PWY0-162 (pyrimidine biosynthesis): PYR1A-B, glutamine hydrolysing carbomoyl phosphate synthase, TbTc_1631; DHODH, dihydroorotate dehydrogenase (fumarate), TbTc_0620; CTPS, cytidine triphosphate synthase, TbTc_0920; PYR3, dihydroorotase, TbTc_3801; PYR2, aspartate carbamoyltransferase, TbTc_1630; CMF40a, nucleoside diphosphate kinase, TbTc_5784; OMPDC/OPRT, orotidine-5-monophosphate decarboxylase/orotate phosphoribosyltransferase, TbTc_0735. PWY0-163 (pyrimidine salvage): CMF40a, nucleoside diphosphate kinase, TbTc_5784; NDPK3, nucleoside diphosphate kinase 3, TbTc_2560; NDPK, nucleoside diphosphate kinase, TbTc_0593; CDA, cytidine deaminase, TbTc_3318; UPRT, uracil phosphoribosyltransferase, TbTc_4220; UP, uridine phosphorylase, TbTc_5794. P121-PWY (adenine/adenosine salvage): ADSL, adenylosuccinate lyase, TbTc_1986; APRT-2, glycosomal adenine phosphoribosyltransferase, TbTc_5918; GMPR, GMP reductase, TbTc_4627; IMPDH1, inosine-5’-monophosphate dehydrogenase, TbTc_1648; HGXPRT, hypoxanthine-guanine-xanthine phosphoribosyltransferase, TbTc_3696; APRT-1, cytosolic adenine phosphoribosyltransferase, TbTc_3522; HGPRT, hypoxanthine-guanine phosphoribosyltransferase, TbTc_0726; ADSS, adenylosuccinate synthetase, TbTc_1142.

https://doi.org/10.1371/journal.ppat.1009734.g005

Purine salvage is an essential process in trypanosomatids, as they lack the de novo synthesis pathway for the purine ring [84], and previous analysis of cell pellets to investigate intracellular nucleotide metabolism utilizing ¹³C-U-D-glucose in BSF T. brucei showed purine salvage pathways incorporating 5-carbon labelled ribose derived from glucose [45] (Fig 5F). Whilst the ribose incorporated into these nucleosides originates almost exclusively from glucose in T. brucei (Fig 5F), T. congolense appears to use far less glucose-derived ribose to make purine nucleosides such as adenosine, guanosine and inosine (Fig 5F).

Transcriptomics analyses indicated upregulation of genes associated with generation of adenosine nucleotides (Fig 5G; red vertical bar), especially in ex vivo T. congolense, as well as hypoxanthine-guanine phosphoribosyltransferase (HGPRT) and uracil phosphoribosyltransferase (UPRT). Upregulation of nucleoside hydrolases and phosphoribosyltransferases supports a previous suggestion based on genome content that T. congolense has a capacity for nucleobase uptake [21].

The purines guanosine and inosine, which incorporate glucose-derived ribose in T. brucei, were almost entirely unlabelled in T. congolense (Fig 5F). However, the phosphorylated nucleosides GMP, GDP and GTP all incorporate glucose-derived carbon atoms, presumably through ribose. Given the labelling patterns seen in adenosine, one possible explanation could be conversion of AMP to inosine monophosphate (IMP; adenosine monophosphate deaminase; TbTc_0145), IMP to xanthosine monophosphate (IMP dehydrogenase; TbTc_1648) and XMP to GMP (GMP synthase; TbTc_1452). However, only one of these enzymes, GMP synthase, was expressed at higher abundance in T. congolense (log₂ fold change: 1.61 and 2.06 for ex vivo and in vitro, respectively; false discovery rate-adjusted p < 0.001). Overall, incorporation of glucose-derived carbons into purine nucleosides is reduced in T. congolense compared to T. brucei. It should be noted that in both experiments, there was no ribose supplementation in the media

Of the pyrimidines, uracil and its derivatives were detected during the glucose labelling experiment (S4 Fig). Uracil is known to be the main pyrimidine salvaged by other kinetoplastids including T. brucei [85–87]. Whilst the majority of the uridine, UMP, UDP and UTP pools incorporate glucose-derived ribose (five ¹³C labels), 5-carbon isotopologues of these pyrimidines were reduced in abundance in T. congolense compared to T. brucei. Conversely, 2-carbon labelled isotopologues appeared to comprise the majority of uridine, uracil and their nucleotides (S4 Fig).

Whilst uracil biosynthesis is not essential in T. brucei [88], the uracil pool in T. congolense appears to derive almost entirely from glucose, when corrected for 50% glucose labelling (45.9% in T. congolense vs 24% in T. brucei [45]; S4 Fig), suggesting that this species predominantly synthesizes uracil from orotate to UMP (orotate phosphoribosyltransferase/orotidine 5-phosphate decarboxylase; TbTc_0735) and from UMP to uracil (uracil phosphoribosyltransferase; TbTc_4220), as can occur in T. brucei [45]. Both these genes are expressed at higher abundance in T. congolense, both in vitro and ex vivo, compared to T. brucei (Fig 5G and S1 Table), which could explain the increased isotopologue labelling.

These data indicate that, at least under the growth conditions used here, BSF T. congolense favours purine nucleoside/nucleotide synthesis from nucleobases with a reduced dependence on glucose-derived ribose 5-phosphate, in addition to de novo synthesis of orotate, uracil and uridine nucleosides. However, the difference in serum requirements for the two organisms is a confounding factor to the interpretation of this difference.

Amino acid metabolism

It is well established that trypanosomatid parasites scavenge amino acids, key nutrients for survival, from their hosts [89,90]. Therefore, comparative analyses of T. congolense and T. brucei amino acid metabolism were undertaken. Whilst the majority of amino acids were detected during the supernatant time course, relative abundances in the medium did not vary greatly after 56 hours of in vitro culture (Fig 6A–6C and S3 Table). The greatest reductions were observed in threonine (log₂ FC after 56 hours: -0.89; Fig 6A), tryptophan (log₂ FC: -0.74; Fig 6B), glutamine (log₂ FC: -0.39), asparagine (log₂ FC: -0.35) and phenylalanine (log₂ FC: -0.35). Interestingly, cysteine, an essential factor for the in vitro culture of T. brucei [63,91], was not significantly consumed by 56 hours (log₂ FC: -0.07; Fig 6C). However, at least low-level exogenous cysteine is still required to sustain parasite growth in vitro, as viability was significantly affected in the absence of cysteine (for both 1.5 mM and 1 mM vs 0 mM cysteine, p < 0.0001, t-test of cells densities at 96 h; S5 Fig). Experiments were carried out to test the essentiality of all other individual amino acids (with the exception of glutamine, known to be an important amino donor in trypanosomatid metabolism [60,89]). Using the minimal medium SCM-6, cell viability was monitored for 72 hours in the absence of specific amino acids. Removal of the following amino acids from culture medium led to defects in growth over 72 hours: asparagine, histidine, isoleucine, leucine, methionine, proline, serine, tyrosine and valine (Fig 6D–6G). Whilst aspartate appeared to be depleted in spent culture supernatants (S3 Table), this also occurred in the medium only control. Furthermore, removal of aspartate did not lead to reduced cell viability or growth rate in culture (Fig 6F). Long term culture was impossible without the addition of phenylalanine and threonine, leading to a final culture formulation, SCM-7 (S5 Table), containing a total of 14 amino acids. Therefore, BSF T. congolense appears to require a higher number of amino acids than BSF T. brucei, at least in vitro, with CMM containing only 8 amino acids in total, including cysteine and glutamine [60]. To further probe amino acid metabolism, pathway analysis was carried out on the transcriptome (S6 Fig) and metabolome (Figs 6 and S7 and S8).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Amino acid metabolism in T. congolense IL3000.

A-C) Analysis of indicated amino acids in T. congolense IL3000 culture supernatants over a 56 h time course. Grey bars indicate a negative medium control group. D-F) Growth curves in SCM-6 excluding one amino acid at a time, to determine those essential to T. congolense viability. In each experiment, full SCM-6 was used as a positive control. Legends indicate which amino acid was removed in each experiment. G) Growth analysis of SCM-6 and SCM-7, the latter containing only amino acids deemed essential, compared to HMI-93 [124]. H) Simplified map of intracellular glutamine metabolism. Numbers refer to the following enzymes: 1, glutaminase; 2, glutamate decarboxylase; 3, 4-aminobutyrate aminotransferase; 4, succinate semialdehyde dehydrogenase; 5, glutamate dehydrogenase; 6, 2-oxoglutarate dehydrogenase; 7, Succinyl-CoA synthetase; 8, isocitrate dehydrogenase; 9 & 10, aconitase. I) Carbon utilisation from L-glutamine was analysed in T. congolense (100% ¹³C-U-L-glutamine) and compared to that in T. brucei (50:50 ratio of L-glutamine and ¹³C-U-L-glutamine) [89].

https://doi.org/10.1371/journal.ppat.1009734.g006

BSF T. brucei utilizes exogenous L-glutamine as the primary source of intracellular glutamate and 2-oxoglutarate, and produces significant levels of glutamine-derived succinate [45,89] (Fig 6I). Given the high levels of succinate excreted by T. congolense, stable isotope labelling was used to determine the contribution of L-glutamine to this pool. T. congolense was incubated for 48 hours with 1 mM ¹³C-U-L-glutamine and cell pellets were analysed by LC-MS. Results indicated the presence of biochemical activities consistent with those previously observed in T. brucei [89] (Fig 6). Significant glutamine-derived carbon labelling was detected after 48 h incubation for succinate (41.3%, 48.5% corrected), glutamate (76.1%, 89.2% corrected), 2-oxoglutarate (80.5%, 94.3% corrected) and succinate semialdehyde (94.7% corrected, Fig 6I). As would be anticipated, labelling of glutathione (86.1%) and trypanothione (98.4%) from glutamine through glutamate was also observed (S7 Fig). No labelling of malate or aspartate was seen in this study, despite the use of high concentrations of ¹³C-U-L-glutamine compared to the equivalent study performed in T. brucei with a 50:50 ratio of ¹³C-U-L-glutamine [89]. Transcriptomics analysis showed high expression levels of glutamine synthetase in T. congolense, compared to T. brucei, under ex vivo conditions only, suggesting ATP-dependent generation of glutamine may occur in the parasite under these conditions (S1 Table).

The apparent essentiality of several amino acids was also investigated using stable isotope labelling. Proline is an essential carbon source for PCF T. brucei and is required for protein synthesis in BSF T. brucei, although it is not a required supplement in BSF medium [92]. In contrast, removal of proline from BSF T. congolense medium led to reduced growth (Fig 6F). RNAi-mediated knock-down of proline metabolism (specifically pyrroline-5-carboxylate dehydrogenase, TbP5CDH) in PCF T. brucei has highlighted the requirement of proline metabolism for mitochondrial function [92]. Indeed, both P5CDH (TbTc_1695) and proline dehydrogenase (TbTc_1591) were expressed at higher levels in ex vivo T. congolense, compared to T. brucei (S1 Table and S6 Fig). However, ¹³C-U-L-proline labelling showed that this amino acid did not contribute meaningfully to the biosynthesis of other metabolites (S8 Fig). Therefore, similar to BSF T. brucei, a requirement for proline in BSF T. congolense may be for the purposes of polypeptide synthesis only.

As in T. brucei [45], glucose-derived carbon usage was detected in several amino acids in T. congolense (S6A Fig). Aspartate (a precursor for pyrimidine nucleotide biosynthesis) and alanine (a by-product of a pyruvate-utilising aminotransferase reaction) (S6A Fig) exhibited 3-carbon isotopologues derived from ¹³C-U-D-glucose in both species. However, in T. brucei, a small proportion of L-asparagine labelling was observed (1.2% 3-carbon labelling) [45], whilst none was observed in T. congolense (S6A Fig). The metabolism of asparagine has not been studied in African trypanosomes; given the reduction of cell growth in the absence of this amino acid (Fig 6F), labelling with ¹³C-U-L-asparagine was performed, but no other labelled metabolites were detected (S8 Fig). This indicates that, as with proline, asparagine uptake is required principally for protein synthesis in T. congolense. The reduced expression of asparagine synthetase (TbTc_4894; TcIL3000.A.H_000497800), which converts aspartate to asparagine (S6 Fig), suggests that BSF T. congolense may rely upon scavenging of exogenous asparagine.

Serine was also shown to be essential to T. congolense (Fig 6F), in contrast to minimal culturing requirements for T. brucei [60]. ¹³C-U-L-serine labelling indicated that T. congolense L-serine metabolism mirrors that of T. brucei in several aspects, such as de novo sphingolipid biosynthesis, with 70.0% 2-carbon labelling of sphinganine and downstream labelling of ceramide and sphingomyelin species (S8 Fig). Similarly, phosphatidylserine decarboxylase activity was evidenced at both transcript and metabolite levels, with 40.1% 2-carbon labelling of glycerol-phospho-ethanolamine (S1 Table and S8 Fig). However, L-serine also has a minor role in S-adenosyl-L-homocysteine detoxification in T. congolense, where serine-derived carbon ultimately contributes to cysteine biosynthesis (S8 Fig). Serine-derived carbon labelling was detected in cystathionine (18.1%) and cysteine (16.7%), through to glutathione (4.1%) and trypanothione disulfide (3-carbon labelled, 6.8%; 6-carbon labelled, 0.02%; S7 Fig). Therefore, the inability to exclude L-serine from T. congolense in vitro culture media may primarily be attributable to lipid metabolism and an increased demand for serine-derived cysteine, potentially over exogenously obtained cysteine, depending on bioavailability. Indeed, metabolomics analysis of culture medium indicates that the ability of T. congolense to take up cysteine from its environment may be lower than in T. brucei (Fig 6C).

Although L-cysteine is primarily considered a source of sulphur for trypanosomatids [89,91,93], we also investigated the carbon contribution of this amino acid in T. congolense, and in particular, whether L-cysteine-derived carbon atoms contribute to the biosynthesis of glutathione and trypanothione. ¹³C-U-L-cysteine stable isotope labelling experiments were performed (S7 and S8 Figs). Direct replacement of the 1.5 mM L-cysteine present in SCM-6 with ¹³C-U-L-cysteine led to high levels of labelling in glutathione and trypanothione disulfide (S7B Fig). This indicates that T. congolense can readily take up and metabolise exogenous cysteine, even though abundance of the amino acid is not reduced significantly over 56 hours of parasite in vitro culture. Although no clear pattern could be observed in transcriptomic analysis of the trypanothione biosynthesis pathway, both trypanothione synthase (TRYS; TbTc_1359) and trypanothione reductase (TRYR; TbTc_4239) were expressed at high levels in in vitro T. congolense cells relative to ex vivo cells, indicating that under in vitro conditions, cells may be subjected to higher levels of oxidative stress (S7C Fig).

There were several further notable observations in the transcriptomics data regarding deamination and decarboxylation of amino acids. For example, cytosolic aspartate aminotransferase (cASAT; TbTc_0799) was expressed at similar levels in both species, indicating that the transamination of tryptophan, tyrosine and phenylalanine to generate aromatic ketoacids [94,95] likely occurs in T. congolense. The mitochondrial isoform (mASAT; TbTc_0877) was expressed at higher levels in cultured T. congolense (log₂ FC: 1.73, p < 0.001), indicative of increased mitochondrial amino acid metabolism.

Fatty acid metabolism in T. congolense

Lipids have a variety of crucial roles in trypanosomes, in particular as a major constituent of membranes. BSF T. brucei require large quantities of myristic acid in particular, for the synthesis of glycosylphosphatidylinositol (GPI) that anchors the parasite’s major surface glycoproteins [96]. To do this, BSF T. brucei both synthesises and scavenges myristic acid. Glucose labelling experiments in T. brucei have shown that myristic acid is partially synthesized from glucose-derived carbon through acetyl-CoA, using a system of fatty acid elongases [97] (Fig 7A). One previous study suggested that GPI anchors in T. congolense also incorporate myristic acid [98]. However, no saturated fatty acid carbon labelling was detected after incubation of T. congolense with ¹³C-U-D-glucose (Fig 7A), unlike T. brucei [45]. Carbon dissemination was also investigated from threonine, which is used as a source of acetate, and thus, lipids [99] (Fig 7B). Similarly, no saturated lipid carbon labelling was observed, suggesting that T. congolense either uses alternative sources of carbon for lipid biosynthesis, or does not rely on acetate as a source of lipids in the same way as T. brucei [33].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Fatty acid metabolism in T. congolense.

A) Glucose-derived ¹³C carbon labelling of saturated fatty acids in T. congolense and T. brucei [45]. Colours correspond to the number of ¹³C labels detected in each metabolite. B) L-threonine-derived saturated fatty acid ¹³C labelling in T. congolense. Fatty acid systematic names and numbers: lauric acid: dodecanoic acid, C12:0; myristic acid: tetradecanoic acid, C14:0; palmitic acid: hexadecanoic acid, C16:0; nonadecyclic acid: nonadecanoic acid, C19:0. C) Transcriptomics analysis of acetate and lipid metabolism. Gene names and IDs: Acetate metabolism (PWY1V8-8): AKCT, 2-amino-3-ketobutyrate-CoA ligase, TbTc_6236; TDH, L-threonine 3-dehydrogenase, TbTc_5991; AceCS, acetyl-CoA synthetase, TbTc_0318; PYK1, pyruvate kinase, TbTc_0372; PDHe3, pyruvate dehydrogenase E3, TbTc_4765; PDHe1β, pyruvate dehydrogenase E1 β subunit, TbTc_5437; PPDK, pyruvate phosphate dikinase, TbTc_1304; SCSα, succinyl-CoA synthetase α subunit, TbTc_0813; PDHe1α, pyruvate dehydrogenase E1 α subunit, TbTc_4169; ACH, acetyl-CoA hydrolase/thioesterase, TbTc_5515; PDHe2, dihydrolipoamide acetyltransferase, TbTc_1015. Fatty acid biosynthesis (PWY0-881): ACC, acetyl-CoA carboxylase, TbTc_0754; BKS, β-ketoacyl synthase, TbTc_3372; BKR, β-ketoacyl-ACP reductase, TbTc_1241. Sterol metabolism (PWY1V8-3): SPPS, solanesyl-diphosphate synthase, TbTc_3025; SQase, squalene synthase, TbTc_2577; CYP51A1, lanosterol 14α demethylase, TbTc_4837; SMT, sterol 24-c methyltransferase, TbTc_0387; LSS, lanosteral synthase, TbTc_4540; MVK, mevalonate kinase, TbTc_3761; FPPS, farnesyl pyrophosphate synthase, TbTc_5375; HMGCL, hydroxymethylglutaryl-CoA lyase, TbTc_6160; SM, squalene monooxygenase, TbTc_3357; MDD, mevalonate diphosphate decarboxylase, TbTc_0546; IDI, isopentenyl-diphosphate delta-isomerase, TbTc_1099; PTase, prenyltransferase, TbTc_1352; GGTase-IIβ, geranylgeranyl transferase type II β subunit, TbTc_0680; SCP2, 3-ketoacyl-CoA thiolase, TbTc_4024; PMVK, phosphomevalonate kinase, TbTc_3039; HMGR, 3-hydroxy-3-methylglutaryl-CoA reductase, TbTc_3189. Fatty acid oxidation (FAO-PWY): ECH, enoyl-CoA hydratase, TbTc_3283; ACS3/ACS4, fatty acyl-CoA synthetase 3 & 4, TbTc_0101; ACS2, fatty acyl-CoA synthetase 2, TbTc_0102; LACS5, fatty acyl-CoA synthetase, TbTc_0099; ACSL_0688, long-chain-fatty-acid-CoA ligase, TbTc_0688; ACSL_2381, long-chain-fatty-acid-CoA ligase, TbTc_2381; ACS1, fatty acyl-CoA synthetase 1, TbTc_0100; TFEα1, enoyl-CoA hydratase/enoyl-CoA isomerase, TbTc_3362; ECI_4184, 3,2-trans-enoyl-CoA isomerase, TbTc_4184; SCP2, 3-ketoacyl-CoA thiolase, TbTc_4024; ECI_0360, 3,2-trans-enoyl-CoA isomerase, TbTc_0360; ACAD, acyl-CoA dehydrogenase, TbTc_4954.

https://doi.org/10.1371/journal.ppat.1009734.g007

Several genes associated with acetate/acetyl-CoA metabolism were highly expressed in T. congolense compared to T. brucei (Fig 7C). For example, ASCT (TbTc_0236), which catalyses ATP-coupled acetate production, was higher under both conditions in T. congolense (log₂ fold changes: 2.04 and 0.46 for in vitro and ex vivo, respectively, p < 0.001; S1 Table). Conversely, acetyl-CoA hydrolase (also known as acetyl-CoA thioesterase, ACH; TbTc_5515), an enzyme involved in ATP synthesis-uncoupled acetate production in T. brucei [100] was expressed at lower levels in ex vivo T. congolense compared to T. brucei (Fig 7C). Consistent with metabolic data, expression of acetyl-CoA synthetase (AceCS; TbTc_0318), a key enzyme in lipid biosynthesis from acetate, was reduced in both ex vivo and in vitro T. congolense (Fig 7C). Other enzymes involved in fatty acid biosynthesis, namely acetyl-CoA carboxylase (ACC; TbTc_0754), β-ketoacyl-CoA synthase (BKS; TbTc_3372) and β-ketoacyl-CoA reductase (BKR; TbTc_1241), were all expressed at lower abundance in T. congolense than T. brucei, in particular in ex vivo cells (Fig 7C). Of the four elongases, ELO1 (TbTc_0159) and ELO2 (TbTc_1882) were expressed at similar levels in both BSF T. congolense and T. brucei (S1 Table). Whilst expression of ELO3 (TbTc_0235) appeared to be reduced in T. congolense (log₂ FC: -1.53 and -1.93 compared to T. brucei for in vitro and ex vivo, respectively, p < 0.001; S1 Table), T. congolense cells expressed higher levels of ELO4 (TbTc_0737) in both in vitro and ex vivo conditions, compared to T. brucei (log₂ FC: 1.43 and 1.44 for in vitro and ex vivo comparisons, respectively, p < 0.001).

Lipoic acid, a fatty acid de novo synthesised by kinetoplastids [97,101], was not detected using the LC-MS platform. However, gene expression data indicated that transcript abundance of genes involved in lipoic acid synthesis (lipoic acid synthase, TbTc_4472; dihydrolipodamide dehydrogenase, TbTc_0275 and TbTc_0276) was similar across both species (p = 0.115), suggesting T. congolense also synthesise lipoic acid de novo.

The variation in observed gene expression associated with the sterol pathway appeared to correlate with sample condition rather than species (Fig 7C). However, T. congolense transcripts for genes involved in lanosterol synthesis were reduced, especially under in vitro conditions (squalene synthase, SQase, TbTc_2577; squalene monooxygenase, SM, TbTc_3357; lanosterol synthase, LSS, TbTc_4540; Fig 7C).

Transcripts associated with fatty acid oxidation were less abundant in T. congolense compared to T. brucei under both conditions (Fig 7C). Energy generation from fatty acids has not been reported for T. brucei. In T. congolense, high abundance of transcripts in key genes such as enoyl-CoA hydratase (ECH; TbTc_3283), particularly in ex vivo samples, was observed. It would therefore be important to establish whether this species has a capacity for fatty acid oxidation.

Exploiting differences in metabolism for potential pharmacological intervention

Differences in metabolism between T. congolense and T. brucei have implications for differential drug efficacy between the two species. To validate our findings in key areas of metabolism, pharmacological inhibition was attempted for specific targets in trypanosome metabolism, in order to compare inhibitory concentrations (EC₅₀).

To assess whether areas of mitochondrial metabolism were more necessary in BSF T. congolense than in BSF T. brucei, both species were treated with FCCP, an uncoupling agent that depolarises the mitochondrial membrane. However, there was no difference in sensitivity between the species (EC₅₀: 13.0 ± 5.0 μM and 12.6 ± 5.3 μM for T. brucei and T congolense, respectively; Table 1). Given both metabolic and transcriptomic data indicated no increased electron transport chain activity, we also treated with the complex III inhibitor antimycin A, again with no significant differences observed between the species (Table 1). In addition, there was no change in sensitivity to azide, an inhibitor of ATP hydrolysis by the F₁-ATPase (Table 1). However, T. congolense appeared to be less sensitive to rotenone, a complex I NADH dehydrogenase inhibitor (Table 1). Previous data inferred complex I activity in BSF T. congolense based on nitroblue tetrazolium staining [51]. Rotenone resistance could indicate NADH dehydrogenase activity of a rotenone-insensitive NADH dehydrogenase, such as the inner membrane space-facing NDH2 [77].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Comparative analysis of sensitivity to metabolic inhibitors in T. congolense and T. brucei.

Abbreviations: FCCP, carbonyl cyanide-p-trifluoromethoxyphenylhydrazone; SHAM, salicylhydroxamic acid; TAO, trypanosome alternative oxidase; AceCS, acetyl-CoA synthetase.

https://doi.org/10.1371/journal.ppat.1009734.t001

T. congolense also showed enhanced sensitivity to salicylhydroxamic acid (SHAM), an inhibitor of the trypanosome alternative oxidase (TAO; Table 1) [102]. Sensitivity of both species to SHAM was also tested in the presence of 10 mM glycerol (Table 1). Addition of glycerol as a carbon source has previously been shown to increase T. brucei sensitivity to SHAM [103]. Both species exhibited increased sensitivity to SHAM in the presence of glycerol (EC₅₀ of 2.28 ± 3.0 μM and 5.00 ± 0.1 μM for T. congolense and T. brucei, respectively). This represented a 12.1- and 13.3-fold increase in sensitivity for T. congolense and T. brucei, respectively. Taken together, these data indicate that, like T. brucei, T. congolense does not rely on oxidative phosphorylation for ATP production, as indicated by transcriptomics analysis, and that, as previously reported, TAO is the terminal oxidase [51,54].

Metabolomics and transcriptomics data indicated that T. congolense direct pyruvate towards mitochondrial metabolism, with high transcript levels in PDH and enzymes involved in acetate generation, compared to T. brucei (Figs 3 and 7). We therefore hypothesised that T. congolense would be more sensitive to inhibition of mitochondrial pyruvate uptake, and to investigate this further, we tested drug sensitivities for UK5099, an inhibitor of mitochondrial pyruvate transport [104]. As expected, T. congolense (EC₅₀: 82.1 μM) was significantly more sensitive (p = 0.0091, unpaired t-test) to UK5099 compared to T. brucei (130.0 μM; Table 1).

Whilst acetate generation appears to be important in T. congolense, our data suggest that the acetate does not appear to be utilised for the biosynthesis of fatty acids, in contrast to what has been shown for T. brucei. To probe this further, we compared drug sensitivity of the two species with compounds targeting fatty acid synthesis (S9 Fig). Indeed, T. congolense was significantly more resistant than T. brucei to an acetyl-CoA synthetase inhibitor (AceCS inhibitor; 1-(2,3-di(thiophen-2-yl)quinoxalin-6-yl)-3-(2-methoxyethyl)urea, [105]; S9 Fig and Table 1), indicating that AceCS is not as essential to this species. AceCS is essential to both BSF and PCF T. brucei [33,69], thus indicating a key metabolic difference between the species.

We next compared drug sensitivity to Orlistat, an inhibitor of fatty acid synthase and phospholipase [35]. Here, a striking difference was found, with T. congolense exhibiting significantly less sensitivity (780-fold increase in EC₅₀) to the compound compared to T. brucei (S9 Fig and Table 1), providing further evidence that T. congolense primarily relies on fatty acid scavenging, instead of synthesis, as predicted by the combination of metabolomics and transcriptomics.

Ethics statement

All animal experiments were performed in accordance with the Animals (Scientific Procedures) Act 1986 and the University of Glasgow care and maintenance guidelines. All animal protocols and procedures were approved by The Home Office of the UK government and the University of Glasgow Ethics Committee.

Compounds and reagents

All compounds were obtained from Sigma/Merck with the exception of: Orlistat (Cambridge Bioscience), oligomycin A (VWR International), diminazene aceturate (Cambridge BioScience) and FCCP (Abcam).

Cell lines and in vitro culture

In all cases, T. congolense strain IL3000 [51] was used (originally received from Theo Baltz, University of Bordeaux). For RNAi experiments a T. congolense IL3000 single marker line, TcoSM, was used [107]. For in vitro experiments, cells were grown at 34°C, 5% CO₂ and routinely cultured in either TcBSF3 [17] or HMI-93 [18], in both cases without a serum plus supplement, with 20% goat serum (Gibco). For global metabolite analysis of culture supernatant, an experimental medium (Steketee’s congolense medium-3; SCM-3) was used with the following components: 77 mM NaCl, 1.5 mM CaCl₂, 4.5 mM KCl, 0.8 mM MgSO₄, 36 mM NaHCO₃, 25 mM HEPES, 0.05 mM bathocuproinedisulfonic acid, 0.22 mM 2-mercaptoethanol, 50 U/mL penicillin/streptomycin, 2.5 mM glucose, 1 mM pyruvate, 10% goat serum, 10% TcBSF3 [17], 1 mM each of L-cysteine and L-glutamine, and 100 μM L-tyrosine, L-phenylalanine, L-tryptophan, L-leucine, L-methionine and L-arginine. BSF T. congolense in exponential growth phase were centrifuged at 1,500 × g for 10 minutes, washed with PBS and inoculated into this medium (0 h time point).

For stable isotope labelling experiments, as well as experiments involving the removal or addition of specific medium components, a custom medium, Steketee’s Congolense Medium-6 (SCM-6) was used (S5 Table). The final medium formulation based on this study’s findings, SCM-7, is provided in S5 Table. This medium is essentially HMI-93, although i) vitamins (with the exception of folate) were removed, ii) D-glucose concentrations were modified depending on experimental procedure, but was routinely kept at 10 mM, iii) goat serum levels were reduced to 10% and, iv) of the 20 amino acids, 14 were added. Increasing the temperature to 37°C led to a detrimental effect on cell viability after several passages, as previously reported [18].

For experiments involving T. brucei, either the monomorphic Lister 427 (in vitro experiments and growth curves) or pleomorphic STIB 247 (RNAseq experiments, both in vitro and ex vivo sample groups) strains were used. Lister 427 cells were grown in HMI-11 [124], whilst STIB 247 were grown in modified HMI-9 containing 1.1% methylcellulose and 20% serum plus (Sigma) [125,126]. In both cases, cells were incubated at 37°C, 5% CO₂.

For both species, cell counts were carried out using a haemocytometer, and in the case of T. congolense, cells were mechanically detached from the culturing plasticware by pipetting prior to counting. Growth curves were routinely carried out in 2 mL samples incubated in 24-well plates, resuspended using a P1000. For detachment of cells from flasks, 10 mL plastic pipettes were used. In cases where cells were harvested for experiments other than those involving metabolomics, cells could also be detached by replacing the medium with PBS for incubating at room temperature for several minutes, prior to vigorously tapping the flask to detach parasites.

RNAi experiments using TcoSM were carried out in HMI-93 in 20 mL cultures. Cells were seeded at 7 × 10⁵ cells/mL and RNAi induction was initiated with the addition of 1 μg/mL tetracycline (Sigma) and 1 × 10⁷ cells were isolated every 24 hours for RNA analysis (outlined below) before cells were passaged.

Animal experiments

Adult female CD-1 mice (20–30 g body weight; Charles River Laboratories) were infected with 5 × 10⁴ wild-type T. brucei STIB 247 or 1 × 10⁵ wild-type T. congolense IL3000 by intraperitoneal injection. Parasitaemia was monitored daily by venesection of the lateral tail vein [127]. At first peak of parasitaemia (>10⁷ cells/mL) mice were euthanised and blood isolated. Parasites of both species were purified from blood by anion exchange using DEAE cellulose [128]. Purified cells were counted, and a total of 1 × 10⁸ cells were centrifuged for 10 minutes at 1,500 × g prior to RNA extraction.

RNA extraction

For in vitro parasite RNA isolation, samples (10⁸ cells) of both T. congolense (IL3000) and T. brucei (STIB 247) were taken from actively dividing cultures grown to densities of 1.8–2 × 10⁶ cells/mL. For ex vivo parasite RNA isolation, samples (also 10⁸ cells) were derived from mouse infections as described above. RNA was extracted using the QIAgen RNeasy kit (Qiagen) with an on-column DNase treatment step. Sample concentrations were analysed by Nanodrop and QuBit, and concentrations adjusted to 37 ng/μL of which 80 μL (2.96 μg) was submitted for RNAseq.

For RNAi time course experiments, cell pellets (10⁷ cells) were resuspended in 1 mL TRIzol (Invitrogen) and stored at -80°C. Samples were thawed, 200 μL chloroform was added, samples were shaken vigorously for 15 seconds and incubated at room temperature for 3 minutes, prior to centrifugation at 12,000 × g for 15 minutes, 4°C. The aqueous layer was transferred to a fresh tube and 500 μL isopropanol and 1 μL Glycoblue (Invitrogen) were added. Samples were mixed by inverting, incubated at room temperature for 10 minutes and centrifuged at 12,000 × g for 10 minutes at 4°C. RNA pellet was washed in ice-cold 75% ethanol and centrifuged at 12,000 × g for 10 minutes at 4°C. After air-drying, RNA was resuspended in 20 μL RNase-free water and concentration adjusted to 100 ng/μL. DNase treatment was carried out using the Ambion TURBO DNase kit (Applied Biosystems) as per manufacturer’s instructions.

Metabolomics sample preparation

For metabolomics analysis of supernatants, 10 mL T. congolense cultures were incubated in T25 flasks in relevant media. Cells were centrifuged at 1,500 × g for 10 minutes, washed with PBS, resuspended in relevant media and density adjusted to 1 × 10⁵ cells/mL. At each time-point, 500 μL medium was transferred to a 1.5 mL Eppendorf tube and briefly quenched in a dry ice/ethanol bath, before centrifuging at 1,500 × g for 10 minutes at 4°C. A 5 μL aliquot was then transferred to a new Eppendorf containing 200 μL metabolite extraction solvent (chloroform:methanol:water in a 1:3:1 ratio) and samples vortexed at 4°C for one hour. Samples were centrifuged for 5 minutes at 13,000 × g (4°C) and supernatants transferred to new Eppendorf tubes. Samples were stored at -80°C prior to analysis.

For analysis of intracellular metabolites, cells were grown to a final density of 2 × 10⁶ cells/mL and a total of 10⁸ cells isolated. Cells were quenched in 50 mL falcon tubes to 4°C using a dry ice/ethanol bath (stirred and measured by thermometer) and all subsequent steps were carried out at 4°C. Cells were centrifuged at 1,500 × g for 10 minutes and if supernatant samples were required in addition to cell pellets, 5 μL was transferred to an Eppendorf containing 200 μL extraction solvent. Cells were resuspended in residual medium before transfer to Eppendorf tubes. Cells were then centrifuged (1,500 × g, 5 minutes) and washed twice with ice-cold phosphate buffered saline (PBS) before resuspension in 200 μL extraction solvent (chloroform:methanol:water in a 1:3:1 ratio). Samples were vortexed at 4°C for 1 hour, and then centrifuged for 5 minutes at 13,000 × g. Supernatants were transferred to clean Eppendorf tubes. For all experiments, a quality control sample was generated by pooling 10 μL from each sample and samples were stored under argon gas at -80°C.

Primers and plasmids

RNAi experiments were carried out using a T. congolense single marker line, TcoSM [107] that expresses Tet repressor and T7 polymerase, maintained in 0.5 μg/mL puromycin, and gene specific RNAi constructs were introduced with a T. congolense specific plasmid, p3T7-TcoV [107]. Primers carrying a HindIII (forward) or an FseI (reverse) restriction site were used to amplify a fragment of TcoHT (fwd: AAGCTTAAACAGAGCAATGCCAGTCG; rev: GGCCGGCCTTATTACGTTTGGCATTATG; restriction sites underlined). Gene fragments were amplified using a HiFi polymerase master mix (NEB) and cloned into pGEM-T easy (Promega) and sequenced to confirm correct sequence identity of each fragment. The constructs were then digested with HindIII and FseI and ligated into the p3T7-TcoV vector [107] using T4 DNA ligase (Promega). The final plasmid was linearised with NotI before purification by ethanol precipitation prior to electroporation into TcoSM cells.

Transfections/electroporations

T. congolense IL3000 electroporation experiments and selection experiments were performed as developed by [107]. A total of 4 × 10⁷ cells were used per transfection, including a negative (buffer only) control. A transfection buffer previously published for use with T. brucei was used for T. congolense transfections [129]. Cells were centrifuged at 1,500 × g for 10 minutes, pellets resuspended in residual medium and transferred to Eppendorf tubes for a further centrifugation step. Cells were subsequently washed in transfection buffer prior to final resuspension in 100 μL buffer per transfection. Up to 12 μg linearised plasmid DNA was added to an electroporation cuvette (Sigma), and 100 μL cells were subsequently added. Electroporation was carried out using a Nucleofector II (Lonza) programme Z-001. Transfected cells were then incubated overnight in 25 mL warm medium in the absence of selective antibiotics, prior to their addition and plating out at dilutions of 1:50, 1:100 and 1:200 in 96-well plates. Antibiotics were added at the following concentrations: Puromycin: 0.5 μg/mL; Neomycin (G418): 0.4 μg/mL. Clones were retrieved after 7–10 days, and these were maintained in 0.25 μg/mL puromycin and 0.2 μg/mL G418.

Drug sensitivity assays

Drug sensitivity assays were carried out using the alamar blue method developed by Raz and colleagues [130]. Briefly, Compounds were diluted to 2× starting concentration in SCM-6 (with 10% goat serum for T. congolense IL3000 or 10% FBS for T. brucei Lister 427) and 200 μL was transferred to the first well of a solid white flat-bottomed 96-well plate. 100 μL medium was then added to 23 further wells and compounds were diluted 1:2 over this series of wells, with the exception of the last well, for a negative control. Subsequently, 100 μL cells were added at 2× starting density (4 × 10⁴ cells/mL for T. brucei and 5 × 10⁵ cells/mL for T. congolense). Plates were incubated for 48 hours (37°C or 34°c for T. brucei and T. congolense, respectively, 5% CO₂ in both cases), prior to addition of 20 μL resazurin sodium salt (0.49 mM in 1× PBS, pH 7.4) to each well. Plates were then incubated for a further 24 hours before measurements of cell viability.

Reduction of the resazurin salt was measured as a function of cell viability. Fluorescence of each plate was read using a Cytation 5 imaging reader (BioTek) and GEN5 software. Parameters were as follows: λ_excitation = 540 nm and λ_emission = 590 nm. Raw values were plotted against concentrations (converted to log₁₀ values) and normalised (0% defined as smallest mean in the data; 100% defined as largest mean in the data) using Graphpad Prism version 8.4.0. EC₅₀ values for each compound were calculated using a non-linear sigmoidal dose-response curve. Each assay was performed in duplicate and each EC₅₀ value represents a mean of three independent experiments.

Real-time quantitative PCR (RT-qPCR)

RNA was extracted as described above, and reverse transcription was carried out in 20 μL using 1 μg RNA, using a high capacity cDNA kit (Applied Biosystems). Primers for RT-qPCR analysis were designed using Primer 3 [131], and primer efficiency was tested using serial dilutions of T. congolense IL3000 genomic DNA by plotting Ct value against log₁₀(DNA concentration). Real-time PCR was carried out using the SensiFAST SYBR Hi-ROX kit (Bioline, BIO92005). Briefly, a 20 μL reaction was set up using 10 μL SYBR mix, RT template and 400 nM of each primer. Primers targeting all five HT genes were used (TcoHT_fwd: ATAGTGACGGAGCGGTTCAT; TcoHT_rev: GACGCAACGACACCAATGAT). Cycling conditions were: 96°C, 120 seconds, followed by 40 cycles of 95°C for 5 seconds, 62°C for 10 seconds and 72°C for 20 seconds. Previously published endogenous control primers for TcoTERT were used for within sample normalisation (TcoTert_fwd: TTTCGCCCTCGTTTTCCTCA; TcoTert_rev: AGAAATCACGACCACACGCT) [132], and normalised transcript level was calculated using the delta delta Ct method [133].

Glucose uptake assays

For analysis of wild-type T. congolense and T. brucei glucose uptake, cells were seeded in 10 mL cultures of SCM-6 at an initial density of 2 × 10⁵ cells/mL (four cultures per species), with 10 mM glucose added separately at the start of the experiment. Upon the addition of glucose, 1 mL supernatant was immediately centrifuged (1,500 × g, 10 minutes) and supernatant stored at -80°C. This process was repeated at 12, 15, 18, 21 and 24 h, and cell density measured by haemocytometer. A medium-only control (4 replicates) was also incubated alongside in vitro cultures. Glucose concentration of each supernatant sample was analysed using the Glucose (GO) assay kit (GAGO-20; Sigma) in a 96-well format. Briefly, 40 μL supernatant sample (diluted if necessary) was incubated with 80 μL assay reagent for 30 minutes at 37°C, after which 80 μL 12 N sulphuric acid was added and absorbance measured at 540 nm using a spectrophotometer. A standard curve was also run to calculate glucose concentration. Rate of glucose consumption was calculated using a custom script [61]. Doubling times of T. congolense IL3000 (11–12 hours) and T. brucei Lister 427 (6–7 hours) were taken into account when calculating rate of glucose consumption.

For glucose consumption of the TcoHT RNAi line, the Glucose Uptake-Glo kit (Promega) was used. RNAi was induced for 72 hours prior to carrying out the assay. Cells were centrifuged, washed in PBS and resuspended in assay buffer (77 mM NaCl, 1.5 mM CaCl₂-2H₂O, 4.5 mM KCl, 0.8 mM MgSO₄-7H₂O, 36 mM NaHCO₃, 25 mM HEPES and 0.02 mM bathocuproinedisulfonic acid), as it was determined T. congolense viability is reduced in PBS alone. Density was adjusted to 10⁸ cells/mL, and three 100 μL replicates of each sample were added to wells of a black flat-bottomed 96-well plate. The uptake reaction was started by the addition of 50 μL 1 mM 2-deoxy-D-glucose. Plate was shaken for 15 minutes at 34°C prior to addition of 25 μL stop buffer, 25 μL neutralisation buffer and 100 μL pre-prepared 2DG6P detection reagent. Plates were shaken in between addition of the buffers. Finally, the plate was read with 0.3–1 second integration on a luminometer (Cytation 5 Imaging reader, BioTek). Wild-type T. congolense, and T. congolense supplemented with glucose were used as controls, in addition to cells without 2-deoxy-D-glucose and assays in the absence of cells.

Acetate concentration assay

Acetate was not detectable by mass spectrometry and therefore, a commercial colorimetric acetate assay kit (MAK086, Merck) was used to analyse changes in supernatant acetate concentrations in trypanosome cultures over time. T. congolense IL3000 were seeded in 10 mL SCM-6 at a density of 1 × 10⁵ cells/mL as outlined in the supernatant metabolomics experiment. At each time-point, 500 μL supernatant was taken from each flask and transferred to an Eppendorf tube. Samples were centrifuged at 1,500 × g, the supernatant was transferred to a fresh Eppendorf tube, and samples were stored at -80°C until samples from all time-points had been collected. Acetate concentration assays were carried out according to the manufacturer’s instructions. Briefly, for each sample, 5 μL was added to the wells of a 96-well plate in duplicate and 45 μL assay buffer was added to each sample. Subsequently, 50 μL reaction mix was added to each well, and the plate was mixed and incubated for 40 minutes at room temperature before absorbance was read at 450 nm. Acetate concentrations were calculated using the standard curve comprised of six concentrations run alongside the experimental samples.

Metabolomics–liquid chromatography mass spectrometry

Hydrophilic interaction liquid chromatography (HILIC) was carried out by Glasgow Polyomics (Glasgow, UK), using a Dionex UltiMate 3000 RSLC system (Thermo Fischer Scientific) coupled to a ZIC-pHILIC column (150 mm × 4.6 mm, 5 μm column, Merch Sequant). The column was maintained at 30°C and samples were eluted with a linear gradient (20 mM ammonium carbonate in water and acetonitrile) over 26 minutes with a flow rate of 0.3 mL/minute.

Sample injection volume was 10 μL and samples were maintained at 4°C before injection. A Thermo Orbitrap Exactive (Thermo Fischer Scientific) was used to generate mass spectra, and was operated in polarity switching mode with the following settings: Resolution: 50,000; AGC: 106; m/z range: 70–1,400; sheath gas: 40; auxiliary gas: 5; sweep gas: 1; probe temperature: 150°C; capillary temperature: 275°C. Samples were run in both positive and negative polarity with the following ionisation: source voltage +4.5 kV, capillary voltage +50 V, tube voltage +70 kV and skimmer voltage +20 V for positive mode; source voltage -3.5 kV, capillary voltage -50 V, tube voltage -70 V and skimmer voltage -20 V for negative mode. Mass calibration was performed for each polarity immediately prior to each analysis batch. The calibration mass range was extended to cover small metabolites by inclusion of low-mass contaminants with the standard Thermo calmix masses (below m/z 1400), C₂H₆NO₂ for positive ion electrospray ionisation (PIESI) mode (m/z 76.0393) and C₃H₅O₃ for negative ion electrospray ionisation (NIESI) mode (m/z 89.0244). To enhance calibration stability, lock-mass correction was also applied to each analytical run using these ubiquitous low-mass contaminants. A set of authentic standards was run prior to the sample set for each experiment.

Metabolomics data analysis

RAW spectra were converted to mzXML files (mzML files for fragmentation data) using XCMS for untargeted peak detection [134]. The resultant files were further processed using mzMatch [135] for peak matching and annotation, resulting in a tabular output that was analysed using IDEOM with default settings [136]. Identification of metabolites was performed in Ideom (as previously described by [45]), by matching accurate masses and retention times of authentic standards (MSI confidence level 1). When standards were not available, predicted mass and retention time was calculated by a validated model [137] (MSI confidence level 2; putative annotation based on exact mass determination). Many of the metabolites names given in the datasets in this study are generated automatically as the Ideom software provides a best match to public database entries of the given mass and formula. Published literature and pathway/genome databases were considered to improve annotation in cases where isomers could not be differentiated based on accurate mass and retention time. Metabolites included in the manuscript were manually annotated using authentic standards where available. However, note that many of the metabolite names given in the IDEOM file are generated automatically as the software provides a best match to database entries of the given mass and formula (S3 Table). In the absence of additional information these must be considered as putatively-annotated hits; the confidence score in the column adjacent to putative metabolite name (S3 Table) serves as a guide to this. In addition, potential isomers matching the formula are provided in a drop down list for each metabolite (S3 Table). Clearly it is beyond the scope of any study to provide authenticated annotations to many hundreds of detected compounds, but the full datasets are included in the spirit of open access data.

For stable-isotope assisted metabolomics experiments, mzMatch output (in.peakml format) was analysed using mzMatch-ISO to extract all carbon isotopologue abundances from putative metabolites [138]. Data analysis of stable isotope-labelled metabolomics was based on a 48 hour time-point in all experiments. Data was further analysed using Microsoft Excel, R or Metaboanalyst v4.0 [139]. The raw data from all metabolomics analyses are available in Metabolights (accession number: MTBLS2372; URL: www.ebi.ac.uk/metabolights/MTBLS2372).

RNA sequencing and data processing

RNA sequencing was carried out by Edinburgh Genomics (Edinburgh, UK). Libraries were prepared from 16 samples (8× T. brucei, 8× T. congolense) using the TruSeq Stranded mRNA kit (Illumina) and 2 × 75 bp paired-end sequencing was carried out using a HiSeq 4000 system (Illumina). Sequencing reads were aligned to the corresponding genome sequence using HiSat2 (default settings with “-k 1—no-spliced-alignment” to limit multi-mapping reads to one alignment) [140]. For T. brucei, the TREU 927 reference genome sequence was used (v51.0 from TriTrypDB [141]), whilst a PacBio assembly of T. congolense IL3000 (2019, v51.0 from TriTrypDB) was used for T. congolense [142]. The resulting SAM files were converted to BAM files using samtools [143], and subsequently filtered for quality (-q 1). Read counts were extracted from the filtered BAM files using HTSeq-count (“-s reverse -f bam -t CDS -i Parent -m union -a 0—nonunique-all”). “CDS” was chosen as feature for read counting instead of “exon”, as UTRs are not annotated in the T. congolense genome, in contrast to the T. brucei TREU 927 genome, which could lead to further discrepancies during cross-species transcriptome analysis. Read counts are provided in S1 Table.

For all samples, transcripts per million (TPM) values for each gene were calculated manually from the read count files using Microsoft Excel as follows: 1) Reads per kilobase (RPK) were calculated by dividing the read counts by the length of gene in kilobases (S1 Table); 2) All RPK values in a sample were summed and divided by 1 million as a scaling factor (S1 Table); 3) Each RPK value was divided by the scaling factor to yield TPM values [56]. To compare transcript abundances between the two species, Orthofinder [57] was used to infer orthologue genes or gene groups. Default parameters were used to compare the T. brucei TREU 927 and the T. congolense IL3000 annotated proteins (S2 Table). A custom MATLAB (version R2020a) script was used to combine the Orthofinder dataset and the TPM values for 1-to-1 orthologues, as well as “sum of TPM” values for groups containing multiple genes, where TPM value for each gene was summed, resulting in a final dataset (S1 Table). Raw RNA-seq data is deposited at GEO (accession number: GSE165290). Transcriptomics data were cross-referenced with the TrypanoCyc database (vm-trypanocyc.toulouse.inra.fr/; [64]) to enable pathway analysis of the data.

Statistical analysis of the orthoTPM dataset was performed using the limma package in R [144]. Briefly, genes with an orthoTPM value of < 1 across three or more of the four sample groups were removed from the dataset (98 genes). The data set was then log-transformed (log₂[TPM+1]) prior to analysis with the eBayes() function including the parameters “trend = TRUE” and “robust = TRUE”. Statistical output is included in S1 Table.

Computation

Figures were generated using Graphpad Prism version 8.4.0 (www.graphpad.com) with the exception of scatter plots and heatmaps, which were generated using R [145]. Heatmaps were generated using the R packages pheatmap and ComplexHeatmap [146]; scatter plots were generated using GGplot2 and GGally; synteny plots were generated using the python tool MCScan [147]; pathway maps were generated with Inkscape v1.0.