DNA double strand break position leads to distinct gene expression changes and regulates VSG switching pathway choice

Antigenic variation is an immune evasion strategy used by Trypanosoma brucei that results in the periodic exchange of the surface protein coat. This process is facilitated by the movement of variant surface glycoprotein genes in or out of a specialized locus known as bloodstream form expression site by homologous recombination, facilitated by blocks of repetitive sequence known as the 70-bp repeats, that provide homology for gene conversion events. DNA double strand breaks are potent drivers of antigenic variation, however where these breaks must fall to elicit a switch is not well understood. To understand how the position of a break influences antigenic variation we established a series of cell lines to study the effect of an I-SceI meganuclease break in the active expression site. We found that a DNA break within repetitive regions is not productive for VSG switching, and show that the break position leads to a distinct gene expression profile and DNA repair response which dictates how antigenic variation proceeds in African trypanosomes.


Author summary
Crucial to triggering antigenic variation is the formation of DNA double strand breaks (DSB). These lesions have been shown to be potent drivers of variant surface glycoprotein (VSG) switching, albeit highly toxic. Trypanosomes immune evasion strategy relies on their ability to rapidly exchange the singly expressed VSG for one that is antigenically distinct. It has been previously shown that the subtelomeric ends, here the locus from which the VSG is expressed, accumulate DSBs. Using the I-SceI meganuclease system we established a series of cell lines to assess how the position of a DSB influences antigenic variation and the cellular response to a break. We show that a DSB in highly repetitive regions are poor triggers for antigenic variation. Contrastingly, a DSB that does lead to VSG

Introduction
Trypanosoma brucei exists as an extracellular parasite in the mammalian host, colonizing the blood, adipose tissue and skin [1,2]. Survival in the bloodstream is dependent on the parasite's ability to periodically exchange the expressed surface antigen, the variant surface glycoprotein (VSG), for one that is immunologically distinct [3]. A large proportion of the trypanosome subtelomeric nuclear genome encodes for VSG genes and their associated sequences, providing a vast repertoire of potential surface antigens for immune evasion [4]. The VSG is expressed from a subtelomeric locus called a bloodstream form expression site (BES) by RNA Polymerase I at an extra nucleolar transcription factory termed the expression site body (ESB) [5,6] which also recruits the VSG exclusion (VEX) complex [7][8][9]. BESs are polycistronic transcription units which contain several protein coding genes termed expression site associated genes (ESAGs) along with the VSG. Although there are approximately fifteen BESs in the nuclear genome [10], monoallelic expression of a single BES ensures that only one VSG is exposed to the immune system at a time.
The BESs are fragile, and DNA double strand breaks (DSB) in the active BES trigger VSG switching by gene conversion, using the subtelomeric VSG archive for repair [11][12][13]. VSG genes are flanked by stretches of 70-bp repeats upstream which serve as recombination substrates for repair [14], and promote access to the archival VSGs [12] allowing for a greater VSG diversity to be expressed during the course of an infection. Antigenic variation appears to follow a semi-predictable order in the selection and expression of VSG donors for recombination during the course of an infection [15][16][17][18]. The predominant mechanism for switching is via recombination-based VSG gene conversion (GC) events, where the expressed VSG is replaced with a silent VSG gene. Less frequently switching occurs via in situ switching which involves the concomitant silencing of the active ES and activation of a silent ES, without any DNA rearrangements [19]. Homologous recombination (HR) is crucial for successful antigenic variation and several factors have been implicated in its success. Factors including BRCA2 [20], RAD51, the RAD51-paralogs [21,22] and the MRN (MRE11 -RAD50-NBS1) complex [23] have been shown to be required for both efficient HR and antigenic variation. Depletion of the RNases H1 and H2, RTR complex (RECQ2-TOPO3a-RMI1) and the RECQ2 helicase leads to increased VSG gene switching by recombination [24,25]. Conversely, ATR kinase and a histone acetyltransferase HAT3, suppress recombination driven VSG switching [26,27], reinforcing the intimate link between HR and antigenic variation. Disruption of BES chromatin integrity additionally leads to increased DNA accessibility across the BES and switching of the active BES by HR [28][29][30][31][32][33][34]. However, the molecular machinery that facilitates VSG switching remains elusive.
VSG switch events are triggered by DSBs in the BES [11,13] however, little is known about how the position of the DSB influences antigenic variation or if breaks within the 70-bp repeats trigger VSG switching. Although the 70-bp repeats are dispensable [35], they have been shown to direct DNA pairing and provide homology during switching [13,36]. We sought to further understand how the position of the DSB in the active BES determined the VSG switching outcome. DSBs in the active BES, either adjacent to the telomere, immediately upstream of the active VSG or adjacent to the VSG promoter have been shown to be highly toxic, but VSG switching is only triggered when it falls immediately upstream of the active VSG [13,37]. Equally, naturally occurring DSBs have been detected in both the active and silent BESs, including the 70-bp repeats of the active BES [13,37]. This led to the intriguing hypothesis that the nature of the 70-bp repeats rendered them fragile and prone to DSBs, therefore triggering antigenic variation. But this has not been experimentally tested. Here, we have assessed the relationship between the position of a DSB in the active BES, changes in genes expression immediately following a DSB and the repair outcome. We have found that when a DSB in the active BES leads to a significant loss of VSG transcript, VSG switching is triggered by homologous recombination.

Trypanosoma brucei growth and manipulation
Lister 427, MITat1.2 (clone 221a), bloodstream stage cells were cultured in HMI-11 medium [38] at 37.4˚C with 5% CO 2. Cell density was determined using a haemocytometer. For transformation, 2.5 x 10 7 cells were spun for 10 minutes at 1000g at room temperature and the supernatant discarded. The cell pellet was resuspended in prewarmed cytomix solution [39] with 10 μg linearised DNA and placed in a 0.2 cm gap cuvette, and nucleofected (Lonza) using the X-001 program. The transfected cells were placed into one 25 cm 2 culture flask per transfection with 36 ml warmed HMI-11 medium only and place in an incubator to allow the cells to recover for approximately 6 hours. After 6 hours, the media was distributed into 48-well plates with the appropriate drug selection. Strains expressing TetR and Sce ORF and with an I-SceI recognition-sites upstream of the active VSG-ESs (VSG up ) [11] have been described previously. G418, and blasticidin were selected at 2 μg.ml -1 and 10 μg.ml -1 respectively. Puromycin, phleomycin, G418, hygromycin, blasticidin and tetracycline were maintained at 1 μg.ml -1 . Clonogenic assays were plated out with 32 cells per plate under non-inducting conditions for all cell lines and either 32 cells per plate or 480 cells per plate under inducing conditions. Plates were counted 5-6 days later and subclones selected for further analysis. Puromycin sensitivity was assessed using 2 μg.ml -1 .

Cell line set up
To construct p70int sce , we synthesised pMA-RQ-70int (Invitrogen) which contains a 315 bp block of 70-bp repeats with a XcmI site embedded within the repeats, 150 bp from either end. pMA-RQ-70int was linearized with Xcm1 (New England Biolabs) and dephosphorylated with Calf Intestinal Phosphatase (MBI Fermantas) to introduce the RFP sce PAC [40]. pARDR s P was digested with Xcm1 digestion (NEB) to release the RFP sce PAC and ligated into the linearized pMA-RQ-70int. Before transfection p70int sce was linearized with Not1 (NEB).
To construct pESAG1 sce , pMA-T-ESAG1 was synthesised (Invitrogen) to contain 300 bp of sequence homologous to the region downstream of ESAG1 with and XcmI site 150 bp from either end. The RFP sce PAC was integrated as with p70int sce and linearised by NotI. To generate the Pseudo sce cell line we used a long primer approach. Using primers PseudoRFPF (CAACA AAATTATAGCAGAATGCAACGTCGACAAAAGGCTCAAGAAATTAACGGCCTACAC GCGGGTCCCATTGTTTGCCTCT) and PseudoPACR (GTTTTGGCGCGTTGTTCCGTAT CTGCTGAGCAAACCTTTTGCGCCGGCTGCTGCGGCGGATAACTATTTTCTTTGATG AAAG) we amplified the RFP sce PAC cassette using Phusion Polymerase (ThermoFisher). 5 μg of PCR product was used for transfection using standard protocols. Individual biological clones were used for each cell line. VSG up served as a control [37].
Plasmids used to generate a double knock-out of the RAD51 gene and MRE11 are described in [23,37]. Biological replicates were used for the RAD51 gene and MRE11 analysis.

Immunofluorescence microscopy
Immunofluorescence analysis was carried out using standard protocols as described previously [37]. Rabbit anti-VSG2 was used at 1:20 000 and rabbit anti -γH2A [41] was used at 1:250. Fluorescein-conjugated goat α-rabbit and goat anti-mouse secondary antibodies (Pierce) were used at 1:2000. Samples were mounted in Vectashield (Vector Laboratories) containing 4, 6-diamidino-2-phenylindole (DAPI). In T. brucei, DAPI-stained nuclear and mitochondrial DNA can be used as cytological markers for cell cycle stage [42]; one nucleus and one kinetoplast (1N:1K) indicate G 1 , one nucleus and an elongated kinetoplast (1N:eK) indicate S phase, one nucleus and two kinetoplasts (1N:2K) indicate G 2 /M and two nuclei and two kinetoplasts (2N:2K) indicate post-mitosis. Images were captured using a ZEISS Imager 72 epifluorescence microscope with an Axiocam 506 mono camera and images were processed in ImageJ. For both γH2A and G2 counts, 2 independent inductions were performed and all counts were done by 2 individual researchers

RNA Analysis
RNA samples were taken at 0 and 6 hours post I-SceI induction in triplicate. RNA was extracted from 50 ml of culture at 1 x 10 6 cells / ml. Briefly, polyadenylated transcripts were enriched using poly-dT beads and reverse-transcribed before sequencing. RNA-seq was carried out on a BGISeq platform at The Beijing Genome Institute (BGI). Reads were mapped to a hybrid genome assembly consisting of the T. brucei 427 reference genome plus the bloodstream VSG-ESs [10,33,43]. Bowtie 2-mapping was used in the default mode with the parameters-very-sensitive-no-discordant-phred33. Alignment files were manipulated with SAMtools [44]. Per-gene read counts were derived using the Artemis genome browser [45]; MapQ, 0. Read counts were normalised using edgeR and differential expression was determined with classic edgeR. RPKM values were derived from normalised read counts in edgeR [46].

DNA analysis
Genomic DNA was extracted from 50 ml of a >1x 10 6 cells-ml culture, spun at 1000 g for 10 minutes and the pellet was resuspended in 200 μl of PBS. Genomic DNA was prepared using the DNeasy Tissue Kit (Qiagen). For Southern Blot analysis, purified DNA was digested with I-SceI (NEB) and AvrII (NEB) and run on a 0.75% agarose gel at 75V until there was adequate separation. The gels were washed in 0.25M HCl for 15 minutes followed by two 10 minutes washes in dH 2 O. The DNA was denatured for 45 minutes (1.5 M NaCl (Sigma), 0.5 M NaOH.), followed by two washes in neutralization solution (1.5 M NaCl (Sigma), 0.5 M Tris base (Sigma), p.H. to 7.0 with HCl) for 45 minutes and a final wash in in 2x SSC (3 M NaCl (Sigma), 0.3 M sodium citrate (Sigma), p.H. to 7.0 with HCl). The DNA was transferred on a Zeta-probe nylon membrane (Bio-Rad) by capillary action (according to the standard protocol) overnight. The DNA was cross-linked to the membrane by UV irradiation on a Stratalinker (Stratagene). The membrane was washed in a pre-hybridization solution (6 x SSC, 1 x Denhardt's, 100 μg/ml denatured Herring Sperm DNA (Sigma) and 0.5% SDS) at 65˚C for 2 hours, and incubated overnight in hybridization containing the 32 P probes at 65˚C. Before to exposing the membrane to a phosphor screen the membrane was washed twice in a washing solution (0.2 x SSC and 0.2% SDS). The phosphor screen was exposed for at least 48 hours before visualization on a Phosphor-imager (Amersham). The PAC probe was a 618 bp EcoRI fragment of the pPac plasmid. The VSG221 probe was a PCR product using primers VSG221F and VSG221 R (details below).

qPCR and RT-qPCR analysis
To determine the dynamics of cleavage we followed a published qPCR protocol [25]. Briefly, 1 x 10 6 cells were collected at 3 hours intervals between 0-24 hours post I-SceI induction. Genomic DNA was extracted using the Qiagen Blood and Tissue kit and quantified using a Nanodrop. The DNA was diluted to 0.2 ng.μl -1 and 1 ng was analysed by qPCR using Luna Universal qPCR MasterMix (NEB) with 500nM of primers. For each pair of primers (below), samples were run in triplicate (Hard-shell PCR Plates 96 well, thin wall; Bio-Rad), which were sealed with Microseal 'B' Seals (BioRad). All experiments were run on a CFX96 Touch Realtime Detection system with a C1000 Touch Thermal cycler (Bio-Rad), using the following PCR cycling conditions: 95˚C for 3 min, followed by 40 cycles of 95˚C for 15 sec and 60˚C for 1 min (fluorescence intensity data collected at the end of the last step). Data was then analysed by relative quantification using the ΔΔCt method (CFX Maestro software-Bio-Rad). Primer pairs CACAACGAGGACTACACCATC and CGGCCTATTACCCTGTTATCC (ESAG1 sce , Pseudo sce , 70int sce ), or GTTGTGAGTGTGTGCTTACC and ATCTAGAGGATCTGGGA CCC (VSG up ) [25]. The expression levels of ESAG6/7 and VSG2 following induction of a break were determined using the following protocol. RNA was extracted using a Qiagen RNeasy Kit and the samples were treated with DNase 1 for 1 hour according to manufactures instructions and eluted in 30 μl of RNase free water. The samples were quantified using a Nanodrop (ThermoFisher). cDNA was prepared using SuperScript IV (ThermoFisher) following the supplier instructions from 1-2 μg RNA with a polyT primer. For each pair of primers (used at 500nM), triplicates of each sample were run per plate (Hard-shell PCR Plates 96 well, thin wall; Bio-Rad), which were sealed with Microseal 'B' Seals (BioRad). All experiments were run on a CFX96 Touch Real-time Detection system with a C1000 Touch Thermal cycler (Bio-Rad), using the following PCR cycling conditions: 95˚C for 3 min, followed by 40 cycles of 95˚C for 15 sec and 60˚C for 1 min (fluorescence intensity data collected at the end of the last step). Data was then analysed by relative quantification using the ΔΔCt method (CFX Maestro software-Bio-Rad) and Cq determination regression was used. Primer pairs VSG2F (AGCTA GACGACCAACCGAAGG) and VSG2R (CGCTGGTGCCGCTCTCCTTTG) [48] or ESAG6/ 7 F (ACTGTGGATGAATTGGCGAA) ESAG6/7 R (ACTGCTACTGTGTTGGACCC). In all cases, product abundance was determined relative to an actin control locus, which was amplified with primer pairs actin F (GTACCACTGGCATTGTTCTCG) and actin R (CTTCATGAGATATTCCGTCAGGTC).

A DSB in the 70-bp repeats of the active BES does not lead to VSG switching
Natural DSBs have been detected in both silent and active BESs and within the 70-bp repeats [13,37], but whether these breaks lead to a productive VSG switching has not been directly tested. To do this we used the established tetracycline-inducible I-SceI meganuclease system [37,49,50], and integrated an RFP:PAC fusion cassette containing the I-SceI recognition site into the major block of 70-bp repeats in BES1, the active BES ( Fig 1A; 70int sce ). In parallel, we selected two additional positions to test in BES1 where the break is in close proximity to 70-bp repeats but distal to the active VSG. In addition to the 70int sce cell line we integrated an SceR between ESAG1 and the smaller block of 70-bp repeats ( Fig 1A; ESAG1 sce ), and between two blocks of 70-bp repeats in the Pseudo gene ( Fig 1A; Pseudo sce ), as has been previously described [13], in both cases using an RFP:PAC fusion cassette containing the I-SceI recognition site. An RFP:PAC fusion cassette containing the I-SceI recognition site has been used to characterise both homology-based and microhomology based repair pathways [37,47,49,50] and we do not believe that integration of the cassette per se would disrupt a specific repair pathway or influence repair pathway choice. To validate correct integration in the 70int sce cell line, we performed southern blotting and PCR assays (S1 Fig). Following digestion with I-SceI and AvrII, the parental cell line gives a 5.8 kb fragment, while the 70int sce will give a fragment of 3.9 kb (S1A Fig). Additional validation using a PCR assay, confirmed the integration of the RFP:PAC cassette in the 70-bp repeats (S1A Fig). To validate correct integration in the Pseudo sce and ESAG1 sce cell lines we used specific PCR assays (S1B and S1C Fig). Following validation, we set up survival assays to determine the effect of a DSB in the 70int sce , Pseudo sce and ESAG1 sce cell lines. As a control for VSG switching, we used a previously published cell line, where the I-SceI recognition site is immediately upstream of the active VSG and adjacent to the large block of 70-bp repeats (Fig 1A; VSG up ), and which shows high rates of VSG switching following induction of a DSB [13,37]. In this cell line, the I-SceI recognition site is on a construct that contains the PAC cassette only. To assess the effect of a DSB in these cell lines, we induced I-SceI using tetracycline. In the VSG up cell line, only 5% of the cells survive a DNA break, which agrees with previously published data [23,25,37] (Fig 1B). To our surprise, 50% of the cells in the ESAG1 sce cell line, 75% in Pseudo sce and 95% in 70int sce survived the DSB (Fig 1B). This was confirmed when looking at the effect of an I-SceI break only (Fig 1C), here the proportion of survivors was calculated by dividing the number of induced survivors by the number of uninduced survivors. This is in stark contrast to other positions tested in the active BES, where in addition to the VSG up cell line, a DNA break adjacent to the active VSG promoter or at the telomere / chromosome junction resulted in between 20-5% survival respectively [37]. It was only when an SceR site was inserted into a non-transcribed silent BES that more than 80% cell survival was recorded [37]. A DSB in the Pseudo gene of the active BES has been reported before [13], and in our study served as an additional control, however due to differences in the efficiency of cutting it is not possible to directly compare survival. From our data we conclude that an I-SceI induced DSB that is flanked by 70-bp repeats and distal to the VSG gene is not toxic to bloodstream form trypanosomes.
To assess the DNA damage response following a DSB in the BES we used the G2 cell cycle checkpoint and γH2A focal accumulation [11,41]. The number of cells in G2 and the number of nuclei with γH2A foci increased in all four cell lines between 0-12 hours following induction (Fig 1D and 1E). In the ESAG1 sce , Pseudo sce , 70int sce and VSG up cells lines we saw an increase in cells in G2 between 6 to 12 hours following induction (Fig 1D). For the γH2A focal accumulation, we saw an increase in the number of foci at 12 hours post DSB induction in the ESAG1 sce , Pseudo sce , 70int sce and VSG up respectively (Fig 1E). We noted that for both cells in G2 and γH2A focal accumulation, the weakest response was seen in the 70int sce cell line.

Repair of DSBs within or upstream of the 70-bp repeats does not result in BES sequence loss
The high rate of survival following a break in the 70int sce , Pseudo sce and ESAG1 sce cell lines suggests a mechanism of repair that occurs either in the absence of any significant deficit to the BES, or a reduction in the efficiency of I-SceI cleavage within the sequence we are targeting. The presence of both native and exogenous DNA within the BES provides a number of markers that can be used to assess both cleavage and BES reorganization following the DNA break-repair cycle. We used individual subclones to establish how the survivors repaired the DNA break. We assessed 24 surviving clones from the ESAG1 sce cell line, 25 from the Pseudo sce cell line, 25 from the 70int sce and 20 from the VSG up cell line. From previous studies we have shown that following a DSB, resection results in loss of the RFP:PAC cassette [37,50], so sensitivity to puromycin is an indicator for I-SceI cutting, while puromycin resistance suggests no cutting (Fig 1A). One ESAG1 sce and one 70int sce clone were found to be puromycin resistant which indicated they were uncut (Fig 2A). All the other clones were puromycin sensitive. This suggests that cleavage by I-SceI is equally efficient across all sites tested. We then looked at VSG2 in the ESAG1 sce , Pseudo sce , 70int sce cell lines and found that all clones were positive for VSG2 by immunofluorescence assay (Fig 2B) revealing none had undergone VSG switching. These results suggest that a DSB in or upstream of the 70-bp repeats are poor catalysts for VSG switching. Our results from the Pseudo sce cell line broadly agree with previously published data [13], where a DSB in this position has been shown to exhibit a low switching frequency [12]. The slight discrepancy between the two studies may be due to technical variations. In contrast all the VSG up clones were VSG2 negative as previously published ( Fig  2B) [13,37].
We next assessed DNA rearrangements following repair in the cloned survivors. Using primers specific to genes in BES1 or the RFP:PAC cassette we used their presence or absence to infer the mechanisms of repair. In the ESAG1 sce cell line, 23 clones retained ESAG1, pseudo and VSG2 (Figs 2C and S2). The presence of the VSG2 gene was expected as the cells were positive for VSG2 by immunofluorescence (Fig 2B). Three clones retained the RFP:PAC cassette but it was reduced in size (clones 5, A weak band was present in clone 1 that is approximately the same size as the parental clones. This clone was puromycin sensitive and no cells grew out of the treated culture suggesting they had all been cut. This band either represents a background band or the original clone divided following the DSB and one cell repaired following loss of the entire RFP:PAC cassette and the second following and MMEJ event that resulted in a small deletion that did not significantly reduce the size of the RFP:PAC cassette. To determine the extent of sequence loss in the remaining 20 clones we used primers to amplify a product from ESAG1 -Pseudogene (S3B We then assessed the Pseudo sce and 70int sce clones. In both cell lines, all the clones had retained ESAG1 and VSG2 but lost the RFP:PAC (Figs 2D and 2E and S3 and S4), apart from the single 70int sce puromycin resistant clone (clone 10) that had retained the RFP:PAC cassette as expected (Figs 2E and S4). These results agree with the immunofluorescence showing VSG2 on the surface of all the survivors (Fig 2B). In VSG up cell line, all the clones had lost VSG2, 18 had retained the Pseudo gene and 19 had retained ESAG1 (Figs 2F and S5) which agrees with previously published data [23,37].

DSB repair in the 70-bp repeats is independent of RAD51 and MRE11
In trypanosomes, RAD51 is required for HR and plays a crucial role in antigenic variation [11,21,22,50] as does MRE11 which is also required for processing of single strand (ss) DNA at a BES [23,51,52]. We generated RAD51 null mutants in the ESAG1 sce , Pseudo sce , 70int sce cell lines (S6 Fig) and MRE11 null mutants in the 70int sce cell line (S7A Fig). In the rad51 nulls the survival was reduced by approximately 80% in the ESAG1 sce , 60% in the Pseudo sce and 80% in the 70int sce cell lines (Fig 3A). The reduction in survival is in line with what has been reported previously for rad51 nulls [11]. We next looked at the cloning efficiency following induction of an I-SceI DSB. By comparing the cloning efficiency, we see that the majority of repair in the 70int sce rad51 cell line is RAD51-independent (proportion of survival 1 vs 0.8 Figs 1C and 3B). In both the ESAG1 sce rad51 and Pseudo sce rad51 null cell lines, survival was reduced compared to the 70int sce rad51 but there was no significant difference compared to the respective parental cell lines (ESAG1 sce vs ESAG1 sce rad51-0.4 vs 0.3 (Figs 1C and 3B); Pseudo sce vs Pseudo sce rad51-0.9 vs 0.5 (Figs 1C and 3B). Additionally, we studied the impact of MRE11 in the 70int sce cell line. We generated MRE11 nulls in this cell line to determine if repair by MRE11 dependent MMEJ dominated at this position. MRE11 has been implicated in repair by MMEJ in eukaryotes [53] but not in Leishmania [54]. In the 70int sce mre11 null cells only 40% of the uninduced population was able to grow ( Fig 3A). Following an I-SceI break, approximately 35% of the cells were able to survive a DSB (survival proportion of 0.9; Fig 3A and 3B) and PCR assays revealed that the cells had lost the RFP:PAC cassette but retained VSG2 and ESAG1, as has the parental 70int sce cell line, this suggests that repair at this position is MRE11-independent (S7B Fig). Taking all the data into account, we assessed repair pathway choice in these cell lines. Both the Pseudo sce and 70int sce cells lines use MMEJ to repair at these positions, whereas in the VSG up cell line a DSB is repaired by homology-based repair, as has been previously reported [13,37] (Fig 3C). In contrast, whilst the ESAG1 sce cell line repaired predominantly via MMEJ, 25% of the cells were able to repair by homology-based repair (Fig 3C).

The cleavage-repair cycle and not the timing of the DSB determines the VSG switching outcome
The striking difference between DSB repair mechanism and the VSG switching outcome at different positions led us to ask whether these differences could have arisen due to variability in the timing of I-SceI cleavage. We used a qPCR assay previously described [25] to assess cleavage in all four cell lines. Genomic DNA was prepared between 0-24 hours post induction, taking samples every 3 hours (Fig 4A) and a qPCR assay was run using primers that spanned the I-SceI recognition site. The amount of product was reduced in all cell lines as early as 3 hours following a DSB, consistent with previous reports, but most dramatically in the ESAG1 sce cell line, where cutting appears almost complete by three hours. In both the ESAG1 sce and Pseudo sce cell line, no product was detected at 24 hours suggesting I-SceI cleavage was complete (Fig 4A). In the VSG up and 70int sce cell lines, the SceR site was cleaved in more than 60% of the cells after 24 hours. Although we did not detect an increase in the amount of PCR product in the VSG up cell line as seen in Devlin, et al [25], we do detect a slight increase between 9-12 hours (Fig 4A, black line) and, the data are broadly consistent between the two studies. We next looked at VSG2 expression by RT-qPCR [48]. RNA was extracted at 6-hour intervals following I-SceI induction, and as anticipated, VSG2 expression decreased over the time course in the VSG up cell line, closely matching that of I-SceI cleavage, with a 50% reduction in VSG2 expression at 18 hours (Fig 4A and 4B). In the ESAG1 sce cell line rapid cutting resulted in an initial drop in VSG2 expression, which returned close to baseline level at 18

PLOS PATHOGENS
DNA double strand break position regulates VSG switching pathway choice hours. A similar pattern was observed in the 70int sce cell line where although more than 50% of the cells had undergone I-SceI cleavage the reduction in level of VSG2 expression returned over 80% at 18 hours (Fig 4A and 4B). The VSG2 RT-qPCR revealed an increase in VSG2

PLOS PATHOGENS
DNA double strand break position regulates VSG switching pathway choice transcript in the Pseudo sce cell line, peaking at 6 hours post DSB, which again returned wildtype levels at 18 hours (Fig 4B). These results suggest that the activated repair mechanism and subsequent VSG switching outcome were not influenced by the timing of I-SceI cleavage. However, this suggests that inhibition of VSG2 transcription for over 12 hours does have an impact on VSG switching. Values are averages of three independent replicates relative to wild-type controls. Red circles BES linked genes; yellow circles BES VSGs; white circle, VSG2; pink circles, BES ESAG6; blue circles, BES ESAG7; green circles, DNA damage linked genes (All genes are included in S1 Table).

The position of a DSB in the active BES leads to distinct gene expression changes
To explore gene expression changes following a DSB we analysed the transcriptome in the Pseudo sce and the VSG up cell line-comparing a cell line that exhibits 100% VSG switching versus one where no VSG switching was detected following a DSB (Figs 4C and S8; S1 Table). The increase in VSG2 expression in the Pseudo sce cell line led us to look first at BES linked genes. In the VSG up cell line, we noted that there were very few changes in the BES linked genes following a break (Figs 4C upper panel: VSG up and 5). In comparison, although the Pseudo sce cell line showed a more constrained change in the gene expression profile following a break than the VSG up cell line, a specific cohort of BES linked genes was upregulated in response to a break at 6 hours ( Fig 4C lower panel: Pseudo sce ). Closer analysis of this cohort of genes revealed that BES linked ESAG7 and ESAG6 are specifically upregulated in the Pseudo sce cell line (between the 16 x ESAG7 genes we see a range in the fold change (FC) from 0.48 to 2.3 and Log 10 from 1.64 to 10.64; between the 14 x ESAG6 genes we see a range in the FC from 1.62 to 0.46 and Log 10 from 0.95 to 11.32) (Fig 4C). Assessment of all BES-linked genes (i.e., from all the silent BESs and the active BES) using a generic BES-which follows the overall conservation in gene order from BES promoter to VSG [55], to compare gene expression changes revealed that ESAG7 (FC of 0.345 vs 1.224; P value = <0.0001) and ESAG6 (FC of 0.339 vs 1.100; P value = <0.0001) were significantly upregulated 6 hours following a DSB in the Pseudo sce cell line as well as ESAG 3C (FC of 0.211 vs 0.767; P value = 0.0471) and ESAG 8 (FC of 0.066 vs 0.536; P value = <0.0001) (Fig 5A). The VSG up cell line showed a significant increase in expression in ESAG 4 (FC of 0.678 vs 0.219; P value = 0.03) only. We then assessed whether there were any changes in the expression of the BES linked VSG genes (14 in total for both cell lines) and observed a small change in the Pseudo sce cell line (Fig 5A) only. Looking specifically at changes in gene expression across the active BES, a similar pattern was seen with the Pseudo sce cell line, ESAG7 and ESAG6 were upregulated (FC 0.58 vs 1.56 and 0.4 5 vs 1.6) (Fig 5B) and in the VSG up cell line, ESAG 5C and ESAG 4 were upregulated (FC 0.29 vs 1.39 and 0.39  Table). Values are averages of three independent replicates relative to wild-type controls. https://doi.org/10.1371/journal.ppat.1010038.g005

PLOS PATHOGENS
DNA double strand break position regulates VSG switching pathway choice vs 0.15, respectively) ( Fig 5B). Surprisingly, the biggest change seen in the active BES, was a 2 -fold reduction in expression of the Pseudo gene in the VSG up cell line (Fig 5B). We included in our analysis other RNA polymerase I transcribed genes, here Procyclin and procyclin associated genes (PAGs). We did not see any significant change in the expression of this cohort of genes in the VSG up cell line and only a small shift in the Pseudo sce cell line (S9 Fig). It has been reported that some PAG genes are closely related to ESAG6 and 7 [56]. This analysis confirms our findings that a break in the Pseudo gene leads to derepression of BES linked ESAG6 and 7.
To determine, which, if any, genes were upregulated following a DNA break, we took an arbitrary cut-off of the top 100 genes with the greatest fold changes in both the Pseudo sce and the VSG up cells lines. Using TritrypDB, we performed a qualitative assessment of these genes based on the Tritryp curations only. As already reported here, in the Pseudo sce cell line this list was dominated by ESAG6 and 7 (S2 Table-transferrin receptor). We observed that for the VSG up cells line, 22 genes were 'hypothetical conserved', which is not surprising given that over 60% of the trypanosome genome is unannotated (S2 Table) and of the remaining genes several of those that were annotated could be associated with a role in DNA damage and repair ( Fig 4C). These include NBS1 which forms part of the MRN complex which has been shown to promote repair by HR in trypanosomes [23]. Here we see an upregulation in the expression of NBS1 (Tb927.8.5710; FC 1.67 and Log 10 53,56). RAD5 is required for post replication repair [57] and here we see an upregulation in RAD5 expression (Tb927.1.1090; 1.86 FC and Log 10 22.19). Two putative helicase genes were upregulated Tb927.6.920; FC 1.86 and Log 10 25.88 and Tb927.9.13610; FC 1.71 and Log 10 40.15), and the role of DNA helicases in DNA repair are well described [58]. Protein kinases play a key role in the DNA damage response, coordinating the cellular response that preserves genome integrity [59]. Four kinase genes were upregulated, a MEKK-related kinase 1 (Tb927.10.14300; FC 1.33 and Log 10 27.52), a CMGC/SRPK protein kinase (Tb927.7.960; FC 1.63 and Log 10 47.94)-which was also identified in a genomewide screen for DNA damage in trypanosomes [60], a serine/threonine-protein kinase (Tb927.10.1910; FC 1.40 and Log 10 49.9) and a inositol hexakisphosphate kinases (Tb927.8.3410; FC 2.28 and Log 10 35.5), which have been shown to regulate DNA repair via conserved signalling pathways that sense damage [61]. SUMOylation and ubiquitination are both implicated in DNA repair [62] and within this cohort of genes, two proteins were annotated as SUMO-interacting motif containing proteins (Tb927.3.1660 and Tb927.8.1290) and one as a ubiquitin-conjugating enzyme (Tb927.3720). SUMOylation is also required for VSG expression in trypanosomes [48]. These results show that as early as 6 hours the position of a DSB can lead to specific gene expression changes that position the cell to either invoke a repair pathway that leads to VSG switching or partially activate silent BES promoters as a backup should repair damage the integrity of the BES.

Discussion
DSBs in the active expression site, although highly toxic, are a potent driver of antigenic variation [13,37]. What leads to the formation of these DSBs is unknown, but it is unlikely to be a process directed by an endonuclease, and rather a function of the location of the BES in the subtelomeres or a characteristic of the BES itself [63]. In this study we show that a DSB in repetitive regions in the BES are poor triggers for antigenic variation. We propose a model where MMEJ dominates as the major form of repair, silent BES promoters are partially activated as a means to facilitate rapid in situ switching should repair be deleterious to the cell. We also show that the position of the break leads to distinct gene expression changes and regulates repair pathway choice and the VSG switching outcome.

PLOS PATHOGENS
DNA double strand break position regulates VSG switching pathway choice

DNA breaks in repetitive sequence do not trigger antigenic variation
Found within the BES and associated with silent VSG genes in the subtelomeric arrays, the 70-bp repeats have been shown to facilitate HR [12,35]. It has been hypothesized that the nature of the repetitive sequence may lead to break formation thereby acting as the trigger for VSG switching. By establishing a series of cell lines, we were able directly assess the effect of a DSB embedded within or flanked by repetitive sequence, in the active BES. We then assessed the DNA damage response and repair pathway choice in these cell lines and individual clones.
Our results indicate that a DSB within the central portion of the 70-bp repeats or upstreamin the Pseudo gene or adjacent to ESAG1, do not display the same toxicity as elsewhere in the active BES [37], and are not productive for antigenic variation. In the three positions tested, we do note that all trigger the DNA damage response as indicated by an increase in γH2A foci, and cells in stalled in G2 position in the cell cycle, but the response is dampened in the 70int sce cell line. Our data also suggests that the majority of repair in the 70-bp repeats or upstream happens via RAD51 and MRE11 independent MMEJ. In contrast a DSB proximal to the VSG is repaired by homology-based mechanisms as previously shown [37]. In mammalian cells, MRE11 is critical MMEJ initiation [64], but both in Trypanosomes [23] and the related kinetoplast parasite Leishmania [54], MMEJ has been shown to be MRE11 independent. We propose that a rapid cleavage-repair cycle facilitated by MMEJ within the 70-bp repeats or in the Pseudo gene, where there are an abundance of sequence microhomologies, maintains VSG expression, BES integrity and high levels of cell survival. Previous work has shown that a DSB proximal to the active BES promoter [37] is also repaired by MMEJ without resulting in VSG switching, but with high levels of cell death. We suggest that like in the ESAG1 sce cell line, where 50% of the cells are able to repair and survive, limited microhomologies at these positions result in a lower proportion of survival.
In BES1 -the active BES in our strain, the 70-bp repeats are approximately 5 kb, so it remains possible that breaks at the immediate VSG proximal end, but still within the 70-bp repeats, may lead to VSG switching if resection extends into the VSG gene or upstream sequence. Upstream of the BES linked VSG genes is the co-transposed region (CTR), deletion of the CTR and the 5' end of VSG2 leads to rapid switching [65]. We tentatively suggest that for a DNA break to be productive in terms of VSG switching, processing by resection would have to disrupt the CTR or VSG expression. Consistently, the only position that shows high frequency VSG switching is following a break adjacent to the 70-bp repeats and up stream of the VSG [13,37].

DNA break site in the active BES gives rise to distinct gene expression changes that determine VSG switching pathway
Our data shows, for the first time, that there are distinct gene expression changes associated with the position of a DNA break. I-SceI access is broadly equivalent across all sites tested, suggesting the differences in the response to the break are not due to variations in the efficiency of cutting due to the location of the break. The reduction in VSG2 expression could be interpreted one of two ways. Either DSB repair pathway choice results in downregulation of transcription-i.e. commitment to homology-based repair results in the formation of secondary structure that's inhibit transcription, or the reduction in VSG2 transcript triggers specific repair pathway choice. Our results indicate that there are distinct gene expression changes as early as 6 hours following a DSB in the active BES and propose that the position of the DSB triggers specific VSG switching pathways which may act to preserve the integrity of the BES. Specifically, we see expression of BES promoter proximal ESAG 7 and ESAG 6 in the Pseudo sce cell line, suggesting partial derepression of silent promoters. MMEJ is a mutagenic repair mechanism which results in deletions flanking the site of the break [66], however it appears to be the favoured form of repair in the active BES (DNA break at the active BES promoter [37], ESAG1 gene, Pseudo gene and within the 70 bp repeats). Given its mutagenetic potential, in order to prevent loss of VSG transcription, or a deletion of a large section of the BES, silent BES promoters are partially activated, pre-adapting the cells to in situ switching should they need to in order to survive. In contrast, where a DSB leads to repair by HR and VSG switching, the silent BES promoters are not activated, and we report upregulation of genes required for DNA damage and repair. We note that the FC are small in these analyses, however these samples were taken 6 hours following a DSB. At this early time point only a proportion of the cells had been cut (Fig 4A, 50% cut in VSG up and 40% cut in Pseudo sce ).
It has been previously reported that chemically induced nuclear DNA damage leads to upregulation of silent BES promoter and other RNA polymerase 1 transcribed genes [67]. Our data following a break in the Pseudo sce cell line broadly agrees with this, however we do not see expression of other RNA polymerase 1 genes such as EP1 Procyclin (Tb927.10.10260). This may be due to the specificity of the DSB in this study as compared to broad chemical damage in Sheader et al. Reversible derepression of silent BES promoters is also seen when the active BES promoter transcription is blocked [68], in a proposed probing of silent BESs before the cells commit to a switch. Our data suggest that this derepression phenotype is rapid, here within 6 hours of a break and specific to where a DNA break is repaired by MMEJ.

Conclusion
The mechanisms underlying recombination-based VSG switching are not well understood. To our knowledge, this is the first report that demonstrates that the position of a DSB in the active BES can trigger distinct gene expression changes, directing the cell through a specific repair pathway. We report that a DSB in the 70-bp repeats or upstream is rapidly repaired, predominantly by MMEJ, whereas only a DSB within close proximity to the expressed VSG leads to recombination-based VSG switching. This data has now expanded our understanding on DSBs as a trigger for VSG switching. Values are averages of three independent replicates relative to wild-type controls. Yellow circles BES VSGs; pink circles, BES ESAG6; blue circles, BES ESAG7; orange circles, Procyclin and procyclin associated genes genes (All genes are included in S1 Table). (TIF) S1 Table Complete table of  Funding acquisition: Lucy Glover.