Beyond immune escape: a variant surface glycoprotein causes suramin resistance in Trypanosoma brucei

Suramin is one of the first drugs developed in a medicinal chemistry program (Bayer, 1916), and it is still the treatment of choice for the hemolymphatic stage of African sleeping sickness caused by Trypanosoma brucei rhodesiense. Cellular uptake of suramin occurs by endocytosis, and reverse genetic studies with T. b. brucei have linked downregulation of the endocytic pathway to suramin resistance. Here we show that forward selection for suramin resistance in T. brucei spp. cultures is fast, highly reproducible and linked to antigenic variation. Bloodstream‐form trypanosomes are covered by a dense coat of variant surface glycoprotein (VSG), which protects them from their mammalian hosts' immune defenses. Each T. brucei genome contains over 2000 different VSG genes, but only one is expressed at a time. An expression switch to one particular VSG, termed VSGSur, correlated with suramin resistance. Reintroduction of the originally expressed VSG gene in resistant T. brucei restored suramin susceptibility. This is the first report of a link between antigenic variation and drug resistance in African trypanosomes.


Introduction
Sleeping sickness, also called human African trypanosomiasis (HAT), is still a prevalent disease in sub-Saharan Africa. It is transmitted by the blood-feeding tsetse flies and caused by two subspecies of Trypanosoma brucei: T. b. gambiense causes the chronic form, T. b. rhodesiense the acute form of HAT. Untreated, both typically end with coma and death. Among the four currently available drug therapies, suramin is the oldest one and is used for the treatment of the first, hemolymphatic stage of T. b. rhodesiense HAT. Suramin is a big and highly charged molecule (Supporting Information Fig.  S1) and therefore not able to passively diffuse through biological membranes. It is supposedly taken up by receptor-mediated endocytosis (Fairlamb and Bowman, 1980) with an involvement of low-density lipoprotein (Vansterkenburg et al., 1993). More recent highthroughput RNAi-screens (Alsford et al., 2012) have shown a variety of endosomal and lysosomal genes to render trypanosomes less susceptible to suramin when knocked-down, which is in agreement with an uptake via endocytosis. Furthermore, knock-down of the invariant surface glycoprotein ISG75 led to a 50% reduction of suramin susceptibility. ISG75 was therefore proposed to be the binding partner of suramin on the cell surface (Alsford et al., 2012(Alsford et al., , 2013. The mode of action of suramin has been elusive since it was shown to have diverse intracellular targets (Wang, 1995).
Despite its use for about a hundred years now, there have been no reports of suramin resistance in human pathogenic trypanosomes. The cure-rates of first stage T. b. rhodesiense HAT are typically high. Suramin treatment failures were mainly observed in context of a missstaging of the disease (P epin and Milord, 1994), since suramin has a low brain permeability and is ineffective once the trypanosomes have crossed the blood-brain barrier. However, suramin resistance is prevalent among animal pathogenic trypanosomes, for example in T. evansi isolates from Sudanese camels (El Rayah et al., 1999) and Chinese buffaloes and mules (Zhou et al., 2004). Selection for suramin resistance is feasible under laboratory conditions. After 550 days of in vitro selection T. evansi were 1800-fold less susceptible to suramin (Fang et al., 1994), and T. brucei showed a resistance factor of 20-140 after in vivo selection with subcurative doses (Scott et al., 1996). The suramin resistance phenotype was lost after transformation of the trypanosomes to insect-stage, procyclic forms.
We have observed the appearance of strong drug resistance in T. brucei bloodstream forms after exposure to high suramin concentrations for only a few days. Here, we use transcriptomics and reverse genetics to elucidate the genetic mechanism behind this phenomenon.

Results
In vitro selection for suramin resistance monitored in real-time Isothermal microcalorimetry allows to monitor the growth of axenic T. brucei cultures in real-time based on the heat emitted by the cells' metabolism (Wenzler et al., 2012). Here, we use the technology to track the emergence of drug resistance. Incubated at presumably lethal concentrations of suramin, i.e., the 5-fold and 25fold of the IC 50 , T. b. rhodesiense bloodstream-form cultures appeared to be dead after two days, started to regrow after three, and had fully recovered by day five (Fig. 1). This phenomenon was consistently observed: we have performed the same experiment with T. b. brucei, which also recovered from suramin pressure within one week. In all cases, the recovered trypanosomes were suramin-resistant by factors between 30 and 100 as compared to the starting population. No such phenomenon was observed with other drugs: melarsoprol or pentamidine completely killed the T. brucei cultures at concentrations of 5-and 25-fold their IC 50 (Wenzler et al., 2012) (Fig. 1).

Rapid emergence of resistance also in clonal populations
The microcalorimetric experiment was performed with an inoculum of 10 5 cells. To rule out that a subpopulation of suramin-resistant trypanosomes had been present all along and was simply selected for under drug pressure, we repeated the experiment with fresh T. b. rhodesiense STIB900 clones. Incubated at concentrations between 500 and 1700 nM suramin (i.e., 5-80 fold their respective IC 50 ), drug-resistant progeny was again obtained within one week for all tested clones [n 5 11]. Table 1 summarizes the results for one clone (named T. b. rhodesiense STIB900_c1) and four independent suramin-resistant derivatives thereof (named T. b. rhodesiense STIB900_c1_sur1 to _sur4), selected at 1050 nM suramin (80-fold the IC 50 ). All four derivatives were at least 90-fold resistant to suramin as compared to the parent clone STIB900_c1 (p < 0.0001 by One Way ANOVA with Tukey's multiple comparisons test). While their sensitivity to melarsoprol and pentamidine was unchanged (Table 1), the suramin-resistant lines were cross-resistant to the structurally related trypanocidal dye trypan blue (p < 0.001, Ehrlich, 1904; Supporting Information Fig. S1). The cross-resistance phenotype was stable in the absence of drug pressure for several weeks. After 80 days of cultivation, however, the resistance factor had decreased to 10-to 50-fold in the four lines (Fig. 2). Cumulative growth curves (Supporting Information Fig. S2) demonstrate slightly longer population doubling times of the resistant derivatives (8.2 h for STIB900_c1_sur1 and 8.3 h for STIB900_c1_sur4) as compared to the suramin sensitive parent clone (7.7 h).

VSGs are differentially expressed in resistant and sensitive parasites
The finding that suramin resistance reproducibly developed even in cultures from freshly cloned trypanosomes, indicated that the underlying mutation -whatever its nature -had occurred during the selection period. However, the reproducible acquisition of a particular point mutation is unlikely to happen within days. We, therefore, reckoned the mechanism of suramin resistance more likely to be reflected at the level of gene expression than gene sequence. Total RNA from T. b. rhodesiense STIB900_c1 and the four selected lines, grown in the absence of suramin, was subjected to quantitative transcriptomics by mRNA-Seq. Over 50 million highquality reads were obtained for each line (European Nucleotide Archive accession PRJEB20650). As expected given the short interval between suraminsensitive parent and suramin-resistant derivatives, there were hardly any differences in transcript abundance (Supporting Information Fig. S3). Genes encoding proteins that had been shown to be involved in suramin resistance in T. brucei (Alsford et al., 2012), i.e., the invariant surface glycoprotein ISG75, the putative major facilitator superfamily transporter MFST or the lysosomal proteins CatL, CBP1 and GLP-1, were not differentially expressed in the suramin-resistant lines. The only hits that were differentially expressed mapped to VSG genes (Supporting Information Fig. S3). Thus all four selected lines had undergone an expression switch to a new VSG.

The suramin-resistant lines have all switched to the same VSG
The sequence of this newly expressed VSG was reconstructed by de novo assembly from the Illumina reads. The reconstructed sequence was verified by reverse transcriptase PCR followed by direct sequencing, making use of the fact that all T. brucei mRNAs possess the same spliced leader sequence at the 5 0 end (Parsons et al., 1984). Three of the four resistant lines had switched to the identical VSG gene, which was termed VSG Sur (GenBank accession MF093647). T. b. rhodesiense STIB900_c1_Sur2 expressed a gene that was almost identical to VSG Sur (GenBank accession MF093648), differing only in four SNPs (two of which were non-synonymous) and a different 3 0 UTR. The sequence of the VSG expressed in the parental clone STIB900_c1 was reconstructed in the same way and termed VSG 900 (GenBank accession MF093646). Remapping of the Illumina reads upon inclusion of these two VSG sequences resulted in more than 10% of reads mapping to VSG 900 in the sensitive parent clone, and more than 10% of reads mapping to VSG Sur in all the resistant derivatives (Fig. 3). Reverse-transcriptase qPCR confirmed the switch from VSG 900 to VSG Sur in the resistant lines as they all showed a more than 10,000-fold upregulation of VSG Sur (Fig. 4). We also analyzed the suramin-selected derivatives of another clone of T. b. rhodesiense STIB900 (STIB900_c2), and they had switched to VSG Sur as well (Fig. 4). Moreover, the suramin-resistant population derived from T. b. brucei strain BS221 had undergone a switch to a VSG with 97% nucleotide sequence identity (Needleman-Wunsch global alignment) to VSG Sur ( Fig. 4; GenBank accession MF093649).
Suramin resistance correlates with expression of VSG Sur but not ESAG7 VSG Sur encodes a bona fide variant surface glycoprotein of 491 amino acids with an N-terminal export signal and a C-terminal GPI-anchoring signal as predicted by Sig-nalP (Nielsen et al., 1997) and GPI-SOM (Fankhauser and M€ aser, 2005), respectively. Profile scans against the Procite and Pfam databases of sequence motifs did not return significant hits beside the entry for the T. brucei VSG conserved C-terminal domain (PF10659; E-value of 0.0002). A Blastp search with VSG Sur as the query against the non-redundant protein sequence database at NCBI returned significant hits on variant surface glycoproteins, the best one on VSG 522 (GenBank: AGH61081.1) with a sequence identity of 72%. The same search against all organisms other than Trypanosoma spp. did not return a single hit. In the absence of any particular feature that would distinguish it from other VSGs, we initially presumed that it is not VSG Sur itself that caused drug resistance but one of the expression site associated genes (ESAGs), which are polycistronically co-transcribed with each VSG and which could have failed to correctly map due to sequence divergence between STIB900 and the reference sequence used. If an antigenic switch results from activation of a new VSG expression site, new alleles of ESAGs will be expressed as well. If the switch is due to replacement of the active VSG by gene conversion, the expressed ESAGs will remain the same. ESAG7 encodes the non-anchored part of the heterodimeric transferrin receptor and differs in sequence among the T. brucei VSG expression sites ( van Luenen et al., 2005;Hertz-Fowler et al., 2008). For each of the five sequenced T. b. rhodesiense lines, the whole transcriptome was assembled de novo from the Illumina reads and the expressed ESAG7 was extracted using a blast search (Tb927.7.3260, 2175 nt). Three of the suramin-resistant lines expressed exactly the same ESAG7 gene as the parental T. b. rhodesiense STIB900_c1 (GenBank accession MF093650); the fourth expressed an ESAG7 that differed in five nucleotides, two of which were non-synonymous (GenBank accession MF093651). Thus in one of the suraminselected lines, the switch to VSG Sur appears to have occurred by activation of a new expression site, while in the remaining three it happened by gene conversion, replacing the originally expressed VSG 900 gene while maintaining the ESAGs. This strongly suggests that it is the expression of VSG Sur itself, and not of the ESAGs, that causes suramin resistance. This hypothesis was corroborated by reverse genetic manipulation of VSG expression. suramin-sensitive parental T. b. rhodesiense STIB900 clone as analyzed with DESeq2. Illumina reads were mapped to the TREU927 reference genome supplemented with Lister 427 bloodstream expression sites and with the sequences of VSG Sur and VSG 900 . The x-axis displays transcript abundance (mean read count over all samples normalized for library size) and the y-axis displays the logarithmic fold-change of transcript abundance (all four resistant derivatives vs. the sensitive parent clone). Genes with a significant (p < 0.01) fold-change are colored in red if overexpressed, and in blue if underexpressed in the resistant derivatives.

Reversal of the VSG 900 -VSG Sur switch restores suramin sensitivity
We made a genetic construct (Fig. 5A) to target the active VSG expression site of the suramin-resistant T. b. rhodesiense STIB900_c1_sur1. In order to exchange the expressed VSG Sur with VSG 900 , the construct was framed with the 5' and 3' UTR of VSG Sur and contained the coding sequence of VSG 900 together with a blasticidin resistance gene. Positive transfectants were selected with blasticidin and cloned by limiting dilution. The construct had integrated in the genomes of all the analyzed clones [n 5 6] as determined by PCR using primers binding to the 5 0 and 3 0 UTR of VSG Sur as the adjacent genomic context was unknown. Reversetranscriptase qPCR confirmed a high expression of the re-introduced VSG 900 (Fig. 5B). Subsequent drug sensitivity assays with three of the transfectants showed a complete loss of resistance in all of them, with IC 50 values between 10 and 12 nM (Fig. 5C). One of the transfectants was once more subjected to suramin pressure. It quickly reestablished suramin resistance with a resistance level similar to STIB900_c1_sur1 (Fig. 6). Since these re-selected parasites could have undergone either expression site switch or gene conversion in order to reexpress VSG Sur , we investigated three lines selected from one transfectant by PCR using primers binding to the 5 0 and 3 0 UTR of VSG Sur . Two of them (c1_sur1_900a_sur1 and c1_sur1_900a_sur2) had lost the construct, thus had undergone a gene conversion in order to switch back to VSG Sur . In the third line (c1_sur1_900a_sur3) the construct was still present, therefore it had undergone an expression site switch in order to express VSG Sur . We subsequently exposed the three lines to blasticidin; as expected the two lines that had undergone gene conversion were killed. The line that had undergone an expression site switch recovered under blasticidine pressure and the obtained parasites (c1_sur1_900a_sur3_bl1, _bl2 and _bl3) had completely lost the suramin resistance (Fig. 6). illustration of VSG Sur (top) and the construct (bottom) used for reintroduction of VSG 900 into the active bloodstream expression site of the resistant derivative STIB900_c1_sur1. The construct is framed with the 5 0 and 3' untranslated region (UTR) of VSG Sur , it contains a blasticidin resistance gene (BSD), an ab tubulin splice site (ab tub) and the coding sequence (cds) of VSG 900 . Binding sites of primers used for the verification of the genomic integration of the construct are indicated with black arrows. B) qRT-PCR, normalized with TERT. Expression of VSG Sur (red) and VSG 900 (blue) are shown for STIB900_c1_sur1 and six transfected clones thereof, with VSG 900 introduced into the active bloodstream expression site. C) Suramin dose-response curves for STIB900_c1_sur1 and three transfected clones expressing VSG 900 .

Binding and uptake of trypan blue
We exploited the facts that the suramin-selected lines were cross-resistant to trypan blue, and that trypan blue becomes fluorescent when bound to protein (Avelar-Freitas et al., 2014), to further examine the observed phenomenon of drug resistance. The fluorescence of trypanosomes was quantified by flow cytometry after incubation for two hours with different concentrations of trypan blue (Fig. 7). Since dead cells are strongly perfused by trypan blue, we first compared dead and living cells. The dead cells showed a much higher fluorescence and could be clearly separated from the living cells by setting the gate accordingly in a cell-size versus fluorescence plot (Supporting Information Fig. S4). Fluorescence of living cells incubated at physiological temperature is derived from both, surface bound and endocytosed trypan blue. At 378C the two tested resistant derivatives (STIB900_c1_sur1 and STIB900_c1_ sur4) showed significantly lower fluorescence levels when compared to the sensitive parent clone (STIB900_ c1) over all concentrations (Fig. 7A). As endocytosis in trypanosomes is inactivated at low temperatures (Brickman et al., 1995), we performed the same experiment at 48C to quantify the fluorescence of surface-bound trypan blue (Fig. 4B). The resistant parasites showed similar levels of fluorescence as the sensitive clone at concentrations of 5-200 mg ml 21 but significantly differed at 390 mg ml 21 , where the sensitive clone showed a strong increase in fluorescence. If the fluorescence at 378C represents binding plus uptake, and the fluorescence at 48C represents binding, then the curve obtained by subtraction (Fig. 7C) represents the trypan blue which had been endocytosed during the two hours of incubation. These values were lower for the resistant parasites than for the sensitive clone at concentrations up to 200 mg ml 21 . There was no difference at 390 mg ml 21 trypan blue, possibly due to toxicity and inhibition of endocytosis in the sensitive clone.

Discussion
The famous parasitologist Frank Hawking (father of the even more famous physicist Stephen Hawking) stated that 'Strains of trypanosomes may be made suraminresistant in the laboratory, but this resistance takes much longer to develop than with arsenicals' (Hawking, 1963). Here we show that > 90-fold suramin resistance emerges in T. brucei spp. bloodstream forms after only few days of in vitro selection at high drug pressure. In contrast, it had taken us almost two years to select for melarsoprol-resistant derivatives of T. b. rhodesiense STIB900 with resistance factors of the same magnitude (Bernhard et al., 2007). Comparative analysis of gene expression showed an association of a particular VSG with suramin resistance. Our first suspicion, that VSG Sur was only a marker and the actual cause of resistance was one of the ESAGs, was rebuted by the finding that three of the four resistant lines had switched their VSG gene while maintaining the VSG expression site, as indicated by perfect conservation of expressed ESAG7. The finding that one of the resistant lines expressed a slightly different ESAG7 indicated that VSG Sur in Significantly different fluorescence levels between the sensitive and the resistant lines are indicated (*, p < 0.05; **, p < 0.01; **, p < 0.001; n 5, pairwise t-test); there were no significant differences between the two resistant lines. (A) Cells incubated at 378C: the fluorescence levels represent surface bound plus endocytosed trypan blue. (B) Cells incubated at 48C: the fluorescence levels represent surface-bound trypan blue. (C) The difference between the fluorescence at 378C and 48C therefore represents an estimate of the endocytosed trypan blue. parental T. b. rhodesiense STIB900 localized to a (silent) telomeric VSG expression site rather than to the chromosome-internal or minichromosomal VSG repository (El-Sayed et al., 2000). The suramin resistance phenotype was slowly reversible over time, which might be explained by the slightly longer population doubling times of the resistant derivatives.
The causality between VSG Sur expression and suramin resistance was confirmed by reverse genetic reintroduction of the originally expressed VSG into the resistant parasites, which completely restored susceptibility. The finding that the expression of VSG Sur causes suramin resistance, explains the fast emergence of the resistance and the high reproducibility of the in vitro selection. It also explains the discrepancy with Hawking's observation cited above, because he was using subcurative treatment of infected mice to select for resistance (in vitro cultivation of T. brucei was not possible at the time).
In vivo, parasites that are bound to express a particular VSG under suramin pressure will be eliminated by host antibodies. This might be the reason why suramin, after one century of use, is still effective in the treatment of early-stage east-African sleeping sickness. The commonly used regimen is five injections of 20 mg kg 21 , leading to suramin plasma levels over 100 mg ml 21 for several weeks (Burri et al., 2004), which is clearly above the in vitro IC 50 of trypanosomes expressing VSG Sur . The situation would be different if lower doses of suramin were administered. One could speculate that such a mechanism could have facilitated the emergence of suramin resistance in animal-pathogenic trypanosomes, as the usual treatment regimen in horses and camels is a single dose of 10 mg kg 21 suramin, and the disease typically leads to immunosuppression (Giordani et al., 2016). Thus if the trypanosomes escaped the effects of suramin in the short term by switching to a certain VSG, this would give them more time to develop other resistance mechanisms. It has been reported that a functional immune system is important for effective suramin treatment of animal trypanosomiasis (Leach and Roberts, 1981).
VSGs could be either directly or indirectly involved in binding and uptake of suramin, and VSG Sur might be non-functional in this respect. Previous studies which proposed an uptake of suramin via endocytosis (Fairlamb and Bowman, 1980;Vansterkenburg et al., 1993) are in accordance with this, as VSG in T. brucei bloodstream forms is endocytosed at a very high rate (Engstler et al., 2004). An involvement of VSG in suramin binding or uptake could also explain the insensitivity of procyclic trypanosomes to suramin (Scott et al., 1996) as they do not express VSG genes.
Binding of trypan blue at 48C to suramin-sensitive trypanosomes suggests a saturable high-affinity component plus a low affinity component that becomes apparent at 390 mg ml 21 trypan blue (testing of higher concentrations was not possible because of toxicity and solubility problems). This implies the presence of more than one binding partner on the cell surface of the sensitive clone. The low affinity component appeared to be missing in the suramin-resistant clones (Fig. 7B). However, the uptake of trypan blue at 378C by the suramin-resistant lines was impaired at concentrations between 5 and 200 mg ml 21 (Fig. 7C), indicating that the low-affinity binding component may not be responsible for drug resistance. Thus while the genetic basis for fast suramin resistance in T. brucei is an expression switch to VSG Sur , the biochemical basis remains to be elucidated.

T. brucei isolates and cultivation
The T. b. rhodesiense strain STIB900 is a derivative of STIB704, which was originally isolated from a male patient at St. Francis Hospital in Ifakara, Tanzania in 1982. It was passaged several times in rodents and once in a tsetse fly (Glossina morsitans morsitans) before a cloned population was adapted to axenic in vitro culture. T. brucei BS221 was received from Dr. Sam Black, International Laboratory for Research on Animal Diseases (ILRAD, Kenya). Parasites were cultivated in Iscove's Modified Dulbecco's Medium supplemented according to Hirumi (Hirumi and Hirumi, 1989) with 15% of heat-inactivated horse-serum. The parasites were grown at 378C, 5% CO 2 and diluted three times a week.

Microcalorimetry with T. b. rhodesiense
T. b. rhodesiense strain STIB900 were cultivated in Minimum Essential Medium with Earle's salts, supplemented according to Baltz et al. (Baltz et al., 1985) with minor modifications: 0.2 mM 2-mercaptoethanol, 1 mM sodium pyruvate, 0.5 mM hypoxanthine and 15% heat-inactivated horse serum. Trypanosome cultures in the late exponential growth phase were diluted with fresh culture medium to a density of 10 5 cells ml 21 . Ampoules were filled, in triplicate, with 3 ml cell suspension and supplemented with suramin, pentamidine or melarsoprol at concentrations corresponding to 5 3 IC 50 or 25 3 IC 50 in the identical medium (Wenzler et al., 2012). Culture medium without trypanosomes served as baseline control. The heat flow was measured continuously over a period of 8 days at 378C in a TAM III isothermal microcalorimeter (thermal activity monitor, model 249; TA instruments, New Castle DE, USA).

Cloning of T. b. rhodesiense
A microdrop of a suspension with 10 4 cells ml 21 was made using a gilded paperclip and inspected by two persons under an inverted microscope. After confirmation of the Suramin resistance in T. brucei 63 presence of a single cell, the microdrop was supplemented with 50 ml of cloning medium. This consisted of 59% Iscove's modified Dulbecco's Medium, which had been supplemented according to Hirumi (Hirumi and Hirumi, 1989) for a total volume of 100%, 1% M€ aser-mix (M€ aser et al., 2002), 20% inactivated horse-serum and 20% of preconditioned medium. The pre-conditioned medium was prepared by centrifugation of a dense trypanosome culture (10 min at 1840 g) and subsequent sterile filtering of the supernatant. After four days of incubation at 378C, 5% CO 2 , the plate was inspected under an inverted microscope and 50 ml of cloning medium was added to the wells with a growing clonal colony. After another 3 days the clones were passaged to new plates and cultivated as described above.
In vitro selection for drug resistance and determination of growth rates A fresh clone of T. b. rhodesiense STIB900 in the exponential growth phase was selected with 1050 nM suramin in four wells each containing 1 ml medium with 5 3 10 4 cells. After 6 days the parasites were removed from drug pressure by washing three times with fresh culture medium and cultured in drug-free medium as described above. For the BS221 strain 10 5 cells were selected with 1070 nM (BS221_sur1) and 4270 nM (BS221_sur2) suramin and removed from drug pressure after 5 days. Cell growth was measured between day 15 and day 26 after selection. Cumulative growth curves were determined by dilution of the cells to 10 5 ml 21 , incubation for 24 h, and measurement of cell density with a CASY Cell Counter (OMNI Life Science, Bremen). This procedure was repeated for four days. Cumulative cell numbers and the population doubling times were calculated and statistically evaluated with t-tests applying Benjamini-Hochberg correction using R Studio (R version 3.3.3).

Alamar blue assay
Suramin sodium salt and pentamidine isethionate were obtained from Sigma (S2671 and P0547 respectively), Trypan Blue from Fluka (93590) and melarsoprol from Sanofi-Aventis/WHO. Alamar blue assays (R€ az et al., 1997) for determination of in vitro drug sensitivities were carried out at least three times independently. Serial drug dilutions were prepared on a 96 well plate in technical duplicates and cells were added in a concentration of 2 3 10 4 . Parasite-free, drug-free wells were used for determination of background fluorescence. After an incubation time of 69 h, plates were inspected microscopically and resazurin was added to a final concentration of 36 mM. Fluorescence intensity was measured after 3-4 h incubation using a SpectraMax (Molecular Devices) and SoftMax Pro 5.4.5 software. The 50% inhibitory concentrations (IC 50 ) were calculated with GraphPad Prism 6.00 using a non-linear regression model to fit the dose-response curve (variable slope; four parameters, bottom value set to zero). Statistical analysis was performed with GraphPad Prism 6.00 using an ordinary one-way ANOVA with Tukey's multiple comparisons test.

Extraction of RNA, mRNA-seq and comparative transcriptomics
Cultures in the exponential growth phase were centrifuged and washed once with PBS before total RNA was isolated using the RNeasy Mini Kit (Qiagen) including an on-column DNase treatment. RNA was shock-frozen in liquid nitrogen and stored at 2808C. Five microgram of RNA from two independent isolations were pooled for each sample. Library preparation and sequencing was performed at the Quantitative Genomics Facility (QGF) of the ETH Zurich in Basel. Sequencing libraries were prepared using the Tru-Seq Stranded mRNA Library preparation kit (Illumina). Single-end sequencing of 125 nucleotides was carried out on an illumina HiSeq 2500. The bioinformatic analysis was carried out on the sciCORE cluster of the University of Basel. Quality of the sequencing reads was controlled with fastqc software (Babraham Bioinformatics -FastQC A Quality Control tool for High Throughput Sequence Data. Available at: http://www.bioinformatics.babraham.ac.uk/projects/ fastqc/. Accessed: 6th April 2017). Illumina adaptors, low quality ends and reads with a length of fewer than 36 nucleotides were removed with trimmomatic (Bolger et al., 2014). Sequencing reads were mapped to the TREU927 reference genome (version TriTrypDB-28) complemented with bloodstream expression sites from Lister427 (Hertz-Fowler et al., 2008). Mapping was performed with bwa (Li and Durbin, 2009) using the default parameters (seedlength of 19 nucleotides, matching score of 1, mismatch penalty of 4, gap open penalty of 6 and a gap extension penalty of 1). The alignment files in the SAM format were converted to binary files using samtools  and sorted and indexed using picard tools (Broad Institute. Available at: http://broadinstitute.github.io/picard/. Accessed: 6th April 2017). Read counts per transcript were determined with the Python package HTSeq (Anders et al., 2015) using a combined gff file from TREU927 (version TriTrypDB-28) and bloodstream expression sites from Lister427 (Hertz-Fowler et al., 2008), default parameters with a minimal mapping quality of 10 were used. Raw counts were analyzed with the R package DESeq2 (Love et al., 2014) (version 1.14.1), independent filtering was disabled.
The de novo assembly of transcripts was carried out using trinity (Grabherr et al., 2011) with the default parameters. The de-novo assembled transcripts were screened with blastn for ESAG7-like genes using the sequence of gene Tb927.7.3260 as query.

Reverse transcription, PCR and quantitative PCR
Complementary DNA (cDNA) was synthesized with Super-Script III Reverse Transcriptase (Invitrogen via Thermo Fisher Scientific, Waltham MA, USA) according to the manufacturer's protocol using oligo(dT) primer and 1-5 mg of total RNA. One microliter of the 1:100 diluted cDNA was used for PCR. Amplification of the VSGs was performed with one primer binding to the spliced leader sequence (Parsons et al., 1984) at the 5 0 UTR and the other primer binding to the 3 0 end of the gene (Supporting Information Table S1). The sequences of the 3 0 ends were derived from the de-novo assembly. For amplification of the VSG expressed in the BS221 strain, a primer binding unspecifically to the 3 0 ends of all VSGs (Aline et al., 1985) was used (Supporting Information Table S1). The PCR was carried out with Taq polymerase (Solis BioDyne, Estonia) or with KAPA HiFi HotStart DNA Polymerase (Kapa Biosystems, Cape Town, South Africa). PCR products were run on a 1% agarose gel, which was subsequently stained with GelRed Nucleic Acid Gel Stain (Biotium, Fremont CA, USA). The PCR products were purified with the Wizard SV Gel and PCR Clean-Up System (Promega, Madison WI, USA) or the NucleoSpin Gel and PCR Clean-up (Macherey-Nagel, D€ uren, Germany) and sent to Microsynth (Balgach, Switzerland) for sequencing. Primers for sequencing were the same as used for PCR, additional internal primers were designed for completion of the sequences. Quantitative PCR was carried out with Power SYBR Green (Applied Biosystems via Thermo Fisher Scientific) on a StepOnePlus real-time PCR System (Applied Biosystems) in technical duplicates. StepOne Software v2.3 was used for analysis of the amplification plots and calculation of the C T values, which were normalized to the housekeeping gene telomerase reverse transcriptase (Brennd€ orfer and Boshart, 2010) (TERT; D C T ). All primers (Supporting Information Table  S1) were designed using the Primer3Plus software (Available at: http://primer3plus.com/cgi-bin/dev/primer3plus.cgi. Accessed: 6th April 2017)

Reverse genetics with T. brucei
For the gene exchange of VSG Sur to VSG 900 a construct was designed (Fig. 5A) containing the 3 0 and 5 0 UTRs of VSG Sur , the coding sequence of VSG 900 , an ab tubulin splice site and a blasticidin-resistance gene (BSD). The DNA was synthesized by GenScript (Piscataway Township NJ, USA) and integrated between the HindIII and the EcoRI-site of a pUC57 vector. For amplification of the insert a PCR was performed using two flanking primers. Three microgram of the purified PCR product were used for the transfection of 4 3 10 7 STIB900_c1_sur1 cells in 100 ml Tb-BSF buffer (Schumann Burkard et al., 2011) with an Amaxa Nucleofector using program Z-001. Positive transfected clones were obtained by limiting dilution in standard culture medium. After 24 h of incubation, blasticidin was added to a total concentration of 5 mg ml 21 . The construct integration was verified by PCR on genomic DNA: As the flanking genomic sequences upstream and downstream of VSG Sur are unknown, primers binding to the 5 0 and 3 0 UTR of VSG Sur were used for the PCR (Fig. 5), which resulted in a band of 1624 nucleotides derived from the wildtype sequence of VSG Sur and a second band of 2356 nucleotides in the presence of the construct. Genomic DNA was isolated from 10 ml dense culture in the following manner: The culture was centrifuged at 1800 g for 5 min and washed with TE buffer (10 mM TrisHCl pH 8, 1 mM EDTA) once. Cells were eluted in 100 ml TE buffer and lysed by addition of 1 ml of 10% SDS and incubation for 10 min at 558C. Subsequently 30 ml of 5 M potassium acetate were added to the lysate and incubated on ice for 5 min to precipitate SDS and proteins. The lysate was spun down for 5 min at full speed in a microcentrifuge and the supernatant transferred to a new tube. The DNA was precipitated with absolute ethanol, washed with 70% ethanol and eluted in 20 ml H 2 O. PCR was carried out as described above. The expression of VSG Sur and VSG 900 was measured by quantitative PCR as described above.

Trypan blue binding and uptake
A total of 2 3 10 5 parasites per sample was exposed to trypan blue (5,20,50,200 or 390 mg ml 21 ) in 1 ml standard growth medium and incubated for two hours at 378C or 48C. After incubation, the parasites were centrifuged (5 min, 3000 g) and washed twice with 1 ml PBS (Phosphate-buffered saline; 137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , pH 7.4) at room temperature or 48C respectively. The cells were resuspended in 250 ml PBS followed by the addition of 250 ml 10% formalin for fixation. The fluorescence intensity of the fixed cells was measured with a BD FACSCaliburTM flow cytometer (BD Biosciences, San Jose, CA, USA) using the FL3 channel (670 nm long pass) upon excitation at 488 nm. Ten thousand events were acquired within the parasite gate based on size versus granularity (FSC x SSC). Dot plots and their fluorescence were evaluated in FL3 histograms as trypan blue has its emission peak at 660 nm (Avelar-Freitas et al., 2014). The measured fluorescence intensity was analyzed with Flowing Software (Version 2.5.1) by using a living parasite gate based on size versus fluorescence intensity (FSC x FL3) and statistically evaluated with R Studio, applying pairwise t-tests (Version 1.0.143, R version 3.3.3).

Supporting information
Additional supporting information may be found in the online version of this article at the publisher's web-site.
Suramin resistance in T. brucei 67