Diminazene resistance in Trypanosoma congolense is linked to reduced mitochondrial membrane potential and not to reduced transport capacity

Trypanosoma congolense is one of the principal agents causing livestock trypanosomiasis in sub-Saharan Africa. This wasting disease is costing these developing economies billions of dollars and undermining food security. Only two old drugs, the diamidine diminazene and the phenanthridine isometamidium are regularly used, and resistance is widespread but poorly understood. We induce diminazene resistance in T. congolense laboratory strain IL3000 in vitro. Resistance was stable and not deleterious to in vitro growth. There was no cross-resistance with the phenanthridine drugs, with melaminophenyl arsenicals, with two promising new oxaborole trypanocides, nor with other diamidine trypanocides such as pentamidine, except the close structural analogues DB829 and DB75. Fluorescence microscopy showed that accumulation of DB75 was inhibited by folate. Uptake of [3H]-diminazene was also partly inhibited by folate, as well as by competing diamidine drugs, albeit at quite high concentrations, and uptake of tritiated diminazene and pentamidine was slow and low affinity. Uptake of [3H]-folate was in turn partly inhibited by diminazene, and inhibition of diminazene uptake by folate and pentamidine appeared to be additive, indicating multiple low affinity transport mechanisms for the drug. Expression of the T. congolense folate transporters TcoFT1-3 in diminazene-resistant T. b. brucei significantly sensitized the cells to diminazene and DB829, but not to oxaborole AN7973. However, [3H]-diminazene uptake was not different in T. congolense IL3000 and its diminazene resistant clones and RNAseq and whole-genome sequencing of multiple resistant clones did not reveal major changes in folate transporter sequence or expression. Instead, flow cytometry revealed a strong and stable reduction in the mitochondrial membrane potential Ψm in all resistant clones. We conclude that diminazene uptake in T. congolense proceed via multiple low affinity mechanisms including folate transporters and that resistance is the result of a reduction in Ψm that limits mitochondrial accumulation of the drug.


Introduction
Animal African Trypanosomiasis (AAT), also called nagana, is an infection of livestock and wild animal species with African trypanosomes, principally Trypanosoma congolense, T.
brucei sspp, T. vivax and T. suis, which is spread by the bite of infected tsetse flies. It has a very severe impact on livestock farming in sub-Saharan Africa, costing billions of dollars in lost agricultural production. As such it is has knock-on effects on issues such as food security, diet and rural development, and the livelihood of rural African communities in endemic areas [1]. Although Human African Trypanosomiasis (HAT), or sleeping sickness, is in rapid decline in most of Africa due to active case finding and rapid treatment [2,3], AAT remains a very serious problem throughout the continent. This is due in no small measure because just two old drugs have been the mainstay of AAT treatment for decades: diminazene aceturate (DA) and isometamidium chloride (ISM). Resistance to each has been reported from most countries in which the disease is prevalent [1] and no new drugs have been introduced for over half a century. Yet, while there are many studies documenting treatment failures in the field, relatively little progress has been made in understanding the mechanisms of resistance in the clinically most relevant animal trypanosome species Trypanosoma congolense and T. vivax.
In contrast, much progress has been made in understanding resistance to the suite of drugs used to treat the various manifestations of HAT, caused by the T. brucei subspecies T. b. gambiense in West and Central Africa, and T. b. rhodesiense in East and Southern Africa.
Most of the resistance mechanisms uncovered for HAT drugs involve aspects of drug uptake by the parasite. For instance, pentamidine (used for early stage gambiense HAT) and melarsoprol (now mostly used for late stage rhodesiense HAT), are internalized by two transport proteins in the trypanosome's plasma membrane. The first of these is an aminopurine transporter called P2 [4] encoded by the gene TbAT1 [5], the deletion of which has been shown to cause a loss of sensitivity to these drugs [6]. Subsequently, it was shown a second and more important determinant for pentamidine-melarsoprol cross-resistance is the High Affinity Pentamidine Transporter (HAPT1) [7,8], which was more recently shown to be the aquaglyceroporin TbAQP2 [9,10]. Rearrangements in the TbAQP2 locus were clearly linked to treatment failures, particularly with melarsoprol [11,12]. Similarly, resistance to eflornithine, used against late-stage gambiense HAT, is associated with the loss of an amino acid transporter, TbAAT6 [13,14], and resistance to suramin, the first-line drug against earlystage rhodesiense HAT, is the result of mutations that interfere with its uptake by receptormediated endocytosis [14,15].
In T. congolense, resistance to isometamidium has also been attributed to reduced accumulation of the drug [16][17][18][19], possibly in part linked to reduced mitochondrial membrane potential [20,21], which diminishes sequestration of cationic drugs into the mitochondrion [22][23][24] -the likely site of drug action. For DA, it has been shown in the T. b. brucei model that it is almost exclusively taken up by the P2/TbAT1 transporter [25] as it is a very poor substrate for the HAPT1/TbAQP2 transporter [26,27]. It was subsequently proposed that the same mechanism would apply in T. congolense, with an adenosine transporter designated TcoAT1 mediating DA uptake, and DA resistance being the result of specific single nucleotide polymorphisms (SNPs) that could be used to screen for resistance [28][29][30].
However, expression of the putative TcoAT1 in a multi-drug resistant T. b. brucei clone, B48, showed that this gene encoded a P1-type broad specificity purine nucleoside transporter with no ability to transport diminazene [31], and genomic analysis has shown that there is no direct orthologue of the P2/TbAT1 transporter in the genome of T. congolense. Regrettably, this has left us still without any insights into the urgent problem of DA resistance in animal trypanosomiasis. One urgent question beyond that of the resistance mechanism is that of cross-resistance with isometamidium, which has been reported repeatedly from the field [32,33], but inconsistently, as others report no cross-resistance and it used to be a rare occurrence [1,34,35]. Clearly, it is possible that trypanosomes may have developed resistance to both drugs independently over the decades and are now resistant to both by separate mechanisms. Thus, it is currently unclear whether other diamidine drugs should be considered for novel drug development for AAT because of the risk of cross-resistance with DA. This adds to the many reasons why it is important to understand the mechanism of DA resistance in animal-infective trypanosomes, alongside the need to identify screening tools for AAT drug resistance to assess the real levels of resistance in the field.
Here, several of these issues are addressed in an attempt to establish a hypothesis for DA resistance in T. congolense, and we report that DA resistance is not linked to reduced drug accumulation in T. congolense, nor necessarily to cross-resistance with isometamidium.
However, we present evidence of stable mitochondrial changes as the basis for DA resistance. In sharp contrast to T. brucei spp, uptake of [ 3 H]-DA by T. congolense was low affinity and inefficient and involves, at least in part, folic acid transporters.

In vitro induction of diminazene aceturate resistance in T. congolense.
Four cultures of T. congolense TcoIL3000 were grown and passaged in parallel; a control without added drug and three separate, independent cultures in the presence of DA for adaptation to increasing concentrations, from which six clonal lines were obtained. In this study we used one clone from original line 4 (4C2), and two clones each from lines 5 (5C1, 5C2) and 6 (6C1, 6C3). Resistance was very slow to develop, as documented in S1 Fig. At the end of the adaptation, after 227 passages, the control culture tolerated 50 nM DA, whereas all the adapted cultures tolerated a maximum level of 800 nM DA, in what appeared to be at least two steps, with an initial plateau at 550 nM, where the strains were assessed for their level of DA resistance, and cross-resistance pattern.
At the point of adaptation to 550 nM, the level of DA resistance and the extent of any cross-resistance with the diamidine drugs pentamidine and DB829, and with oxaboroles AN7973 and SCYX7158, was tested using an assay based on the viability dye Alamar blue (resazurin), in which live trypanosomes (but not dead ones) reduce blue and non-fluorescent resazurin to pink, fluorescent resorufin. This assay yields a fluorescent output that is proportional to the number of cells in the well [36]. The results, summarized in Fig 1A, show that all the DA-Res strains display highly similar drug sensitivity patterns. Resistance to DA, as measured in this assay, was ~2.3-3-fold, and highly significant (Fig 1A for fold differences and P values, Table 1 for EC 50 values). No cross-resistance with pentamidine was observed, but the level of resistance to DB829, a structurally closer analogue of DA, mirrored the resistance profile of DA very closely, demonstrating that cross-resistance to diamidines is not inevitable, but observed only with close structural analogues. Importantly, there was no cross-resistance with isometamidium, nor with either of the oxaboroles tested, AN7973 and SCYX7158. To further investigate the limits of cross-resistance with other diamidines, a series of bis-benzofuran analogues [37] were tested (Table 1). These analogues were chosen to explore the effect of variable linker length between the benzofuramidine end groups, varying inter-amidine distance as well as the flexibility of the molecule. All the benzofuramidines displayed low micromolar activity (Table 1) and there was no crossresistance with DA ( Fig 1B). These results clearly show that the cross-resistance phenotype is very narrow and that many diamidines are still potentially useful for the treatment of DAresistant AAT.
The DA-resistance of the five DA-resistant clones was then increased via further in vitro exposure until they could tolerate growth at 800 nM DA. The in vitro growth rates of these strains consistently appeared to be slightly slower than that of the drug-sensitive parental strain (Fig 2), although this reached statistical significance only for the 48 h and 64 h points of clones 5C2 and 6C1 (P<0.05, unpaired student's t-test). The EC 50 values, at ~2.2 µM, were similar for all clones but significantly higher than when they were measured before the cloning, at 550 nM DA medium concentration. This level of resistance, approximately 9-fold, was stable for at least 3 months of in vitro culturing in the absence of drug pressure (Fig 3).
Cross-resistance to DB829 was again observed in all clones (~5-fold), whereas sensitivity to pentamidine was identical in all clones including wildtype IL3000. Interestingly, sensitivity to AN7973 was slightly but significantly higher in most DA-Res clones (Fig 3). Crossresistance with phenanthridines isometamidium and ethidium bromide, and with the melaminophenyl arsenical cymelarsan was also tested for selected clonal lines. A slight loss of sensitivity to phenanthridines was consistently observed and reached significant (but <2fold) levels in some clones. No change in cymelarsan sensitivity was observed (Fig 4).

In vivo confirmation of cross-resistance profile.
In order to confirm that the in vitro pattern of cross-resistance held up for in vivo infections, including for a field-derived DA-resistant strain, we infected groups of 6 mice that were treated i.p. with either DB829 (Fig 5A) or AN7973 (Fig 5B), using single doses of 5, 10 or 20 mg/kg bw (two separate independent experiments with each drug, each experiment n=6). For the resistant filed isolate we used stain KONT2/151 [38]. IL3000 was more virulent than KONT2/151, as is often observed with cultured trypanosome strains, and was sensitive to both DB829 and AN7973. DB829 delayed parasitemia and death with a dose of 10 mg/kg and was curative at 15 and 20 mg/kg, whereas AN7973 delayed parasitemia at 5 mg/kg and cured with 10 or 20 mg/kg. The DA-resistant field isolate was equally sensitive to the oxaborole but resistant to DB829, as even the highest dose, 20 mg/kg, merely delayed the onset of parasitemia.
Screen for potential diminazene transport inhibitors by DB75 fluorescence microscopy.
In T. b. brucei, fluorescent diamidines such as DB75, a close structural analogue of DA and DB829, have previously been used to monitor uptake and cellular distribution of such compounds [22]. In those experiments, the rate at which the kinetoplast (first) and the nucleus (second) became visible in the blue wavelengths (λ=405/435 nm for excitation and emission, respectively) was much delayed in resistant strains, and in the presence of transport inhibitors [22]. Here, the technique was used to screen for potential transport substrates that competitively inhibit the uptake of the DB75 in animal trypanosome species. The utility of the technique was first confirmed in ex vivo T. b. brucei in mouse blood. Murine blood cells did not appear to significantly take up 10 µM DB75, ensuring a low background, but in T. b.
brucei, parasite outlines would become visible after 4 min, followed by kinetoplasts at 5-6 min and finally nuclei were faintly stained by 8 min and bright by 15 min. These stages were all observed and scored in the presence of potential inhibitors. As expected, based on current T. brucei models [25,26], the polyamines putrescine, spermine and spermidine (500 µM) had no discernible effect on DB75 staining, although the polyamine transporters of Leishmania spp and T. cruzi have been implicated in possible pentamidine uptake [39][40][41]. In contrast, DA and the shorter pentamidine analogue propamidine both dose-dependently delayed DB75 fluorescence (S1 Table); both are known substrates of the TbAT1/P2 aminopurine transporter [7,42] that is the sole uptake mechanism for DB75 in T. brucei [43].
Because of the involvement of the TbAT1/P2 purine transporter in DA uptake by T. b.
brucei, we tested a range of purine and pyrimidine nucleosides and nucleobases for delaying DB75 cell entry in T. congolense. Adenine and adenosine did not appear to slow DB75 fluorescence up to 1 mM but were toxic to the cells at concentrations of 200 µM and above.
Glycerol was also tested at up to 20 mM because some diamidines including pentamidine are taken up by an aquaglyceroporin in T. brucei [9,27], but the data indicated that glycerol also does not compete for the T. congolense DB75 transporter. However, the diamidines DA and, particularly, propamidine and pentamidine, dose-dependently (100 µM -500 µM) delayed DB75 fluorescence, and so did folic acid, from 200 µM to 500 µM (Fig 6). T. vivax was also assayed by this method and similar results were found, with no effect from a series of purines and pyrimidines, glycerol, and the cationic amino acids lysine and arginine, but dosedependent inhibition by pentamidine and DA. For this parasite, there appeared to be some minor fluorescence delay in the presence of 500 µM folic acid or biopterin, but not at lower concentrations, and possibly in the presence of the polyamines spermidine and putrescine (S1 Table).

Uptake of [ 3 H]-Diminazene by DA-sensitive and -resistant T. congolense clones
In T. brucei, resistance to DA is linked to loss of the TbAT1/P2 transporter [6,11,25]. While T. congolense does not have an equivalent transporter [44], drug resistance in trypanosomatids has very often been associated with the loss of transporters at the plasma membrane [10,12,45] and we thus investigated whether uptake of experiments, which still shows a non-significant downward trend and no difference between the slope of the sensitive and resistant lines. In order to better understand why DA-resistance in T. congolense, in contrast to T. b. brucei, is not associated with loss of import or gain of efflux function, the mechanism by which DA is internalized by T. congolense was further investigated.
Uptake of 0.1 µM [ 3 H]-DA was poorly saturable by even a large excess of unlabeled DA (Fig 7C), indicating that DA enters T. congolense by a low affinity and/or non-saturable process. Indeed, out of several potential inhibitors or substrates of the DA uptake, the process was consistently (but always partially) inhibited by 100 µM pentamidine, 1 mM folate and 100 µM DB829 (Fig 7D). This is in sharp contrast to DA uptake in T. brucei which is completely saturable and 100% inhibited by a competitive substrate of the TbAT1 transporter ( Fig 7E). The inhibition with folate was confirmed and quantified in a series of experiments with added time points, and found to be highly significant ( Fig 7F). Moreover, in a separate experiment it was tested whether the inhibition by folate and pentamidine was additive and the data indicated that this was the case (Fig 7G), suggesting that in T. congolense DA is taken up by at least two separate low affinity transporters, one sensitive to folate and another that is sensitive to pentamidine. However, it must be noted that the inhibition by both compounds was modest in this experiment and that the joint administration still did not inhibit [ 3 H]-DA uptake by even 50%. The complete lack of cross-resistance with pentamidine, despite high concentrations of pentamidine partially inhibiting DA uptake, is consistent with DA resistance not being the result of changes in its uptake rate. Indeed, the overlap in transport mechanisms of DA and pentamidine was even more clear from the inhibition of [ 3 H]-pentamidine uptake by DA in wild-type IL3000 (Fig 7H). To further investigate whether (a) folate transporter(s) might contribute to DA uptake in T. congolense, we measured the uptake of [ 3 H]-folate and found this to be partially inhibited by 1 mM DA ( Fig 7I).

T. congolense folate transporters are engaged in diminazene uptake
Three genes encoding putative folate transporters of T. congolense were identified via homology searches in TriTrypDB in the original T. congolense IL3000 genome and amplified by PCR with Phusion proof-reading polymerase from IL3000 and from the DA-resistant isolates KONT2/133 and KONT2/151 (sequences, GeneIDs and GenBank accession numbers in Table 2); all three strains are Savannah-type T. congolense. The genes were cloned into pGEM-T Easy for amplification in E. coli and for each gene at least 4 independent clones were sequenced. The gene sequences from the two KONT strains, both from Cameroon, were almost identical to each other, but displayed some SNPs relative to IL3000 (Table 3). with the percentage identity and similarity between these genes is included in S2 Table. The genes were cloned into the pHD1336 expression vector and transfected into T. b. brucei strain B48. This clonal line lacks both the TbAT1/P2 transporter and a wild-type TbAQP2 gene, and is thus highly resistant to DA, the furamidines, pentamidine and melaminophenyl arsenicals [8,27,43]. As the sensitivity to these drugs is completely restored upon re-introduction of the transporters [27,44] this is an ideal system to test potential diamidine transporters from other kinetoplastids. Clonal lines were obtained by limiting dilution; the correct integration of the expression constructs was verified by PCR.
Expression of the T. congolense folate transporters in T. b. brucei B48 confirmed that these transporters have some capacity to take up diamidines, but not the oxaborole AN7973 ( Fig 8). Only FT2 of IL3000 did not significantly sensitize to any of the drugs tested, although its diamidine EC 50 values also trended to be lower than the control (B48 transfected with the 'empty vector'). Sensitization to DA and DB829 was almost identical, but only the FT2 transporters of the KONT strains significantly sensitized to pentamidine. In all cases the sensitization was modest, reaching ~50% with DA and DB829, and <40% for pentamidine, which is consistent with the transport experiments with TcIL3000, in that the transporters appear to display significant and yet very limited diamidine uptake capacity. For comparison, the expression of TbAT1 in B48 cells, using the same expression vector and assays, resulted in ~200-fold sensitization to DA and pentamidine instead of 2-fold [44]. The data presented in Fig 8 do not suggest that the two KONT strains are DA resistant because of changes in DA uptake -or at least not by their folate transporters. With respect to DA and DB829, the IL3000 and KONT FT1 and FT3 performed identically, but the KONT FT2 clearly sensitized stronger than the IL3000 FT2.

clones.
As DA resistance is apparently not linked to changes in DA uptake, it was hypothesized that an intracellular rather than a cell surface change must be responsible for the resistance phenotype. Moreover, DA is a DNA minor grove binder and resistance arising due to mutations in a target protein are therefore not expected. Our observations with DB75 fluorescence microscopy showed that such drugs accumulate first in the kinetoplast, as reported for other species [22,46]. This requires rapid uptake into the mitochondrion, which is driven by the mitochondrial membrane potential Ψm as DA, DB75 and other diamidines are dications. We therefore investigated whether Ψm was altered in the resistant cell lines.
Fluorescence microscopy after DAPI staining confirmed that the kinetoplast was present and did not appear altered or damaged in the DA-resistant clones (S3 Fig). It is already well established that (di)cationic trypanocides that accumulate in the trypanosome's mitochondrion cause a reduction in Ψm, both through the very fact that they are cations, and through disruption of mitochondrial processes (e.g. [23,24,47,49]). However, the more important question here was whether, as (part of) the adaptation to DA, the cells had permanently lowered their Ψm, as previously reported for isometamidium [20,21]. The DA-Res parasites did present flow cytometry profiles that had shifted significantly to lower fluorescence and were broader than the control IL3000 cells (Fig 9A). For quantification, the peak of the IL3000 profile was set at 80 Arbitrary Units (AU). Taking the percentage of cells with fluorescence >80 AU as a measure for Ψm [47,48], a highly significant decrease was found in all four DA-Res cell lines tested, although the extent of the Ψm decrease was not identical in all cell lines ( Fig 9B).

Genomic analysis of DA resistance
DA-resistant T. congolense were analyzed for SNPs and indels by Illumina paired-end whole genome sequencing. Across the 4 samples (wild type and DA-resistant clones 4C2, 5C1 and 6C3) there were a total of 158,396 SNPs and 28,640 indels (see S1 File and S2 File for the statistics on SNPs and indels, respectively). Of the indels, 20,107 were insertions, whilst the remaining 8,533 were predicted deletions. Both datasets were filtered, firstly to remove mutations occurring in all four samples, as these are clearly not related to drug resistance, and subsequently to obtain lists of high, moderate and low impact (as described in the methods) mutations in protein encoding regions (5,263 SNPs and 541 indels) (S3 Table, worksheets "SNPs-HIGH-MODERATE-LOW" and "Indels-HIGH-MODERATE-LOW"). Once mutations that occurred in genes predicted to encode variant surface glycoproteins (VSGs), expression site-associated genes (ESAGs) and retrotransposon hot spot (RHS) proteins were removed, there were 19 high impact SNPs (S3 Table, worksheet "SNPs-Hi-filtered") and 192 high impact indels (S3 Table, worksheet "Indels-Hi-filtered"), in addition to 1,693 moderate impact SNPs (S3 Table, worksheet "SNPs-Mod-filtered") and 42 moderate impact indels (S3 Table, worksheet "Indels-Mod-filtered") predicted to result in non-synonymous mutations for the former, and in-frame deletions/insertions for the latter. Due to the previously described correlation between resistance to the related compound isometamidium chloride and reduced mitochondrial membrane potential [20], as well as reduced potential observed in the DAresistant cell lines, data were mined to identify genes that could be involved in maintaining or generating this membrane potential. Two copies of a vacuolar-type Ca 2+ -ATPase (TcIL3000.A.H_000569200 and TcIL3000.A.H_000569400) harbored mis-sense mutations in all three resistant lines (TcIL3000.A.H_000569200: A109T in clones 4C2 and 6C3; TcIL3000.A.H_000569400: G855E in clones 4C2 and 5C1). However, no further candidates were found. Two mis-sense mutations were found in folate transporter TcIL3000.A.H_000597400 (A132S and N131Y, both in clone 4C2 only, both heterozygous) and another two in folate transporter TcIL3000.A.H_000597700 (L122R in 4C2 and F371C in clone 6C3; both heterozygous). No mutations were found in the other folate or pteridine transporters. Mis-sense mutations were also found in several copies of an amino acid transporter (TcIL3000.A.H_000388400, TcIL3000.A.H_000388500 TcIL3000.A.H_000388600), a putative plasma-membrane choline transporter (TcIL3000.A.H_000774600) and two copies of a mitochondrial chaperonin HSP60 gene (TcIL3000.A.H_000771700 & TcIL3000.A.H_000772600) (S3 Table, worksheet "SNPs-HIGH-MODERATE-LOW").

Transcriptomics analysis of diminazene resistance
To ascertain whether differential gene expression could explain increased levels of resistance to diminazene, three replicates of the wild-type parental line, as well as three replicates of each independent resistant clone, were subjected to transcriptomic analysis by Illumina paired-end sequencing. Data was processed as described in the methods, resulting in transcript abundances for 9,360 T. congolense genes (S3 Table, worksheet "RNAseq_all").
Data was filtered to obtain differentially expressed genes relative to wild-type (qval ≤ 0.05), resulting in 274, 247 and 265 genes identified as being significantly differentially expressed in DA-resistant clones 4C2, 5C1 and 6C3, respectively, with 120 of these genes being common to all three clones (Fig 10). Genes were then filtered to remove VSGs, ESAGs and RHS proteins, resulting in a final list of 61 genes, of which 15 were annotated as hypothetical (S3 Table, worksheet "RNAseq_filtered").
Among the most significant changes was downregulation of an array of H4 histone encoding genes (mean Log 2 fold change: -7.842) (S3 Table,  There were only 13 significantly differentially expressed genes exhibiting a Log 2 fold change of >1.0 in all three resistant clones, and these were identified as primarily RNA helicases, in addition to copies of cysteine peptidase and several hypothetical proteins. As the data for the RNA helicases have different significance and expression values, it is unlikely that the identification of this gene family is due to multimapping as in the Histone H4 genes, in this case leading to a firmer conclusion that all individual genes identified are differentially expressed (S3 Table, worksheet "RNAseq_UP"). In some cases, there was significant differential expression in only two of the DA-resistant clones, for example, a predicted long chain fatty acid-CoA ligase (TcIL3000.A.H_00065500.1; mean q-value: 0.0356; mean Log 2 fold change: -0.297).

Discussion
In this manuscript we have studied the phenomenon of diminazene resistance in T. based on loss of TbAT1 function [11,25], is not applicable to T. congolense.
We found that it was possible to induce a substantial level of diminazene resistance in T. congolense by slowly increasing the drug concentration in the culture medium. The level of DA resistance was entirely stable in all clones, after months of in vitro passage in the absence of drug pressure, and repeated cycles of stabilating and reculturing. Remarkably, all the independently generated DA-resistant clones displayed the same level of resistance as well as cross-resistance patterns and growth rates, which could indicate that they had developed very similar adaptations.
One of the main aims of this study was to establish whether DA resistance in T.
congolense necessarily leads to cross-resistance to other (potential) AAT drugs. Co-resistance with isometamidium chloride has been reported in the field [52][53][54] but it is not clear whether this constitutes genuine cross-resistance. This conclusion can be hard to reach in the field as cattle are often treated with both drugs and independently-developed resistance to both drugs is therefore a real possibility.
However, there have been reports of single resistance to diminazene and to isometamidium, and more recently to both drugs [1,32,55]. This clearly shows that multipleresistance is not automatic, as it seems to be for pentamidine and melarsoprol in T. brucei sspp. [56]. Our observation that induced DA resistance in T. congolense did not change isometamidium sensitivity in any of the independent clones is certainly a strong confirmation of that hypothesis and to the best of our knowledge the first such evidence from a controlled laboratory study since experiments in 1963 by Frank Hawking with multiple T. congolense strains in guinea pigs and mice, which also found no evidence of cross-resistance between DA and phenanthridines [57]. We have also attempted to similarly induce isometamidium resistance in T. congolense, but this was not successful in vitro although it has been achieved in vivo [19]. Another important observation on the nature of DA resistance in T. congolense is that it was remarkably stable, which, considering the many reports of DA resistance from all over Africa, has important implications for the future utility of the drug. In T. brucei, drug resistance is mostly linked to changes in drug accumulation, particularly through mutations in transporter genes [12]. In T. congolense, isometamidium resistance has also been linked to reduced accumulation [16,17,19,53], but no drug transporters have been identified in this species. We therefore investigated the uptake of [ 3 H]-DA in wild-type IL3000 and found it to be very slow and, apparently, low affinity. This indicates that uptake of diminazene could be limiting to its efficacy, whereas in T. brucei sspp uptake of diminazene, the furamidines and pentamidine is very fast, and high affinity [7,25,26,43]. We found no consistent indication that DA resistance in T. congolense is linked to reduced cellular accumulation of the drug, as the accumulation rate was similar in the parental and resistant clones.
Several compounds were identified as inhibitors of both [ 3 H]-diminazene and fluorescent DB75 uptake in T. congolense, particularly pentamidine, propamidine and folate.
The latter observation indicates the involvement of folate transporters in diminazene uptake, although high concentrations were required for significant inhibition and even then diminazene uptake was only partially inhibited. Clearly, a so-far unidentified, folateinsensitive uptake mechanism is additionally involved in diminazene uptake. The notion of multiple low-affinity entry routes was further strengthened by the demonstration that inhibition by folate and pentamidine was additive, and yet still incomplete. To further investigate whether T. congolense folate transporters may contribute to diminazene uptake, all three then known FTs were cloned and expressed in the diamidine-and melaminophenylresistant clone T. brucei B48 [8]. The expression of IL3000 transporters FT1 and FT3 appeared to induce a small but significant sensitization to diminazene and its structural analogue DB829, but not to pentamidine or oxaborole AN7973; IL3000 FT2 trended in the same direction as the other two transporters but the effect did not reach statistical significance. The folate transporters of two DA-resistant isolates from Cameroon displayed several SNPs relative to the IL3000 sequences but if anything displayed higher levels of sensitization than the homologues from the sensitive strain, and FT2 from both resistant isolates significantly sensitized to all three diamidines including pentamidine, but not to AN7973. These observations show that these folate transporters have a minor capacity for diamidine transport, but also that SNPs observed in these genes are not the cause of DA resistance, either in the field or in our laboratory-adapted clones. The lack of genuine resistance mutations in the folate transporters is presumably linked to the observation that at least two of those transporters have this capacity, and that the FTs mediate only a proportion of DA uptake, considering the modest effect of 1 mM folate on [ 3 H]-diminazene uptake.
Based on the above reasoning we must conclude that diminazene resistance in T.
congolense is not principally the result of a reduced rate of cellular diminazene accumulation.
The alternative cause of resistance would be in changes of the target but in the case of diminazene and other cationic drugs this is the kinetoplast DNA [22, [46][47][48], located in the mitochondrion. While dyskinetoplastic T. brucei have been described, including as part of the adaptation to isometamidium [21], and T. evansi is considered a dyskinetoplastic T. brucei [61], dyskinetoplastic T. congolense have not been described, and our DA-resistant clones displayed normal kinetoplasts as observed by DAPI staining. Thus, the DA resistance phenotype cannot easily be attributed to changes in the target either. However, in order to reach and disable the kinetoplast, the drug must accumulate inside the mitochondrion -a process that is dependent on the mitochondrial membrane potential Ψm. We observed, in all four DA-resistant clones investigated, a strong and statistically significant reduction in Ψm, which would serve to diminish the entry of diminazene into the mitochondrion, where the drug would otherwise be strongly accumulated, bound to kDNA. This is particularly important as the entry of the drug into T. congolense is low affinity and inefficient, as opposed to the energy-dependent, high affinity process in T. brucei. Diminazene uptake in T.
congolense is thus very likely equilibrative rather than concentrative, implying that a failure of the drug to accumulate in the mitochondrion will eventually result in a reduced cellular uptake as well.
Genomic Altogether, we conclude that a clinically relevant level of DA-resistance in T.
congolense can be induced in vitro and is stable, and not primarily due to reduced DA accumulation but instead linked at least in part to a deficient accumulation into the mitochondrion due to reduced mitochondrial membrane potential. DA uptake is low affinity and partially mediated by the T. congolense folate transporters, but also by separate transporters that are sensitive to pentamidine. These results constitute the first systematic analysis of DA-transport and resistance mechanisms in T. congolense and provide some insights important for the development of new drugs against AAT.

In vitro culture of T. congolense.
Cultures of bloodstream forms of Savannah-type strain IL3000 and the adapted clones derived from this strain were cultured exactly as described by Coustou et al. [62],

Resazurin assay with T. congolense.
This drug susceptibility assay was performed essentially as described [63] and is based on For non-inhibited cells 10 µM of DB75 was added and fluorescence observed using a Zeiss Axioscope fluorescence microscope, using the DAPI filter set (λ exc = 330 nm, λ em = 400 nm) as well as brightfield. Images were obtained with Openlab imaging software (Improvision, Coventry, UK). For inhibited cells, the test inhibitor was added to the desired concentration (see S1 Table) just before the addition of 10 µM DB75 and observation of fluorescence as before. Each potential inhibitor was tested at least twice on parasites from different mice, those which inhibited up to 4 times. Every experiment had cells with no inhibitor added visualized in parallel to control for slight variation in timing of appearance of fluorescence in kinetoplasts and nuclei between batches of ex vivo parasites.
Transport assays for T. congolense bloodstream forms were performed exactly as described for T. b. brucei bloodstream forms [66] and Leishmania major promastigotes [67].
Depending on the assay, parasites were either purified from infected blood using a DEAEcellulose column and a 6:4 ratio PSG buffer as described [68], or collected from culture.  [8] using an Amaxa Nucleofector, program X-01, and TbBSF buffer as described [31,69].
Transfectants were cloned out, by limiting dilution, in standard HMI-11 medium [70] containing 5 µg ml -1 blasticidin S and clones screened for correct integration of the cassette by PCR. All clones were maintained in HMI-11 using standard growth conditions, 37 ºC and 5% CO 2 .
Drug sensitivity of the T. brucei cell lines was assessed using the standard resazurin assay, essentially as described for T. congolense, above and as described previously [71].
Briefly, the assay was performed in 96-well plates with 11 doubling test compounds dilutions, highest concentration 10 µM, and the 12 th well containing cells in drug-free medium. 2 × 10 4 bloodstream form trypanosomes were added to each well, and the plates were incubated for 48 h at 37 °C/5% CO 2 after which 20 µL of resazurin solution was added to each well and the plates incubated for another 24 h.

Mitochondrial membrane potential (Ψm) assay
Changes in Ψm were determined by flow cytometry using the indicator dye tetramethylrhodamine ethyl ester (TMRE) as the probe, exactly as previously described [47,63] Briefly, bloodstream forms of 1×10 7 T. congolense IL3000 and several IL3000- strain IL3000 [72] and gene abundances quantified, in both cases using Kallisto with default parameters (-b 100, -t 8) [73]. Statistical analyses of differential expression between the wildtype parental strain and each of the three resistant clones were carried out using the R-based package Sleuth, with default parameters (for each transcript, a second fit was performed to a reduced model that presumes abundances are equal in the two conditions -wild-type and DAresistant) as outlined in the manual [73], and figures were generated with Matlab (The MathWorks, Inc.). Genes exhibiting significant (q-value <= 0.05) differential expression between wild-type and resistant parasites were tabulated and filtered to remove VSGs, ESAGs and retrotransposon hot spot proteins, and subsequently sorted to obtain genes with significance based on q-value (i.e. p-value adjusted for multiple testing by false discovery rate). The full Sleuth output is available in S3 Table, worksheet "RNAseq_all").

Whole genome analysis
For whole genome analysis, DNA was extracted from in vitro cultures (3×10 6 cells) of the T.
congolense IL3000 wild type line, as well as the 3 DA-resistant lines (4C2, 5C1 and 6C3), using the QIAgen DNeasy kit. Libraries were prepared using the QIAseq FX DNA library kit (QIAgen, cat: 180475). The libraries were quantified using Qubit high sensitivity reagents (Thermo), and average library size was determined using DNA high sensitivity reagents for the Agilent bioanalyser 2100. The four samples were subsequently paired-end (2×75 bp) sequenced using Illumina NextSeq500 to an average of 15 million reads per sample. SNP and indel analysis were carried out on the resulting data using the GATK pipeline [74]. Reads were first trimmed using cutadapt (parameters: -q 30,30 --minimum-length=70 --pair-filter=any) [75] and subsequently aligned to the PacBio T. congolense IL3000 assembly    In vitro cross-resistance pattern of clones from three independent diminazene-adapted cultures with EC 50 values given relative to that for the control strain, IL3000, adapted to growth in 550 nM DA. A. Cross-resistance to DB829 but not pentamidine, isometamidium, AN7973 or SCYX7158 was observed (n = 5 -9). B. Cross-resistance to a series of bisbenzofuramidines was investigated -none of the strains were more than marginally resistant or sensitized to these compounds and the average EC 50 of the 5 resistant strains was in all cases very close to the value for the IL3000 control. All structures are displayed in Table 1.
S2 Table. Identification of amino acid differences in folate transporters of three different T. congolense strains.
S3 Table. RNAseq and genomic sequence data of T. congolense IL3000 and DA-Res