Bromodomain factor 5 is an essential transcriptional regulator of the Leishmania genome

Leishmania are unicellular parasites that cause human and animal disease. Alongside other organisms in kinetoplastida, they have evolved an unusual genome architecture that requires all RNA polymerase II transcribed genes to be expressed constitutively, with transcriptional start regions denoted by histone variants and histone lysine acetylation. However, the way these chromatin marks are interpreted by the cell is not understood. Seven predicted bromodomain factors (BDF1-7), the reader modules for acetyl-lysine, were identified across Leishmania genomes. Using L. mexicana as a model, Cas9-driven gene deletions indicate that BDF1-5 are essential for promastigote survival, whilst DiCre inducible gene deletion of the dual bromodomain factor BDF5 identified it to be essential for both promastigotes and amastigotes. ChIP-seq assessment of BDF5s genomic distribution revealed it as highly enriched at transcriptional start sites. Using an optimised proximity proteomic and phosphoproteomic technique, XL-BioID, we defined the BDF5-proximal environment to be enriched for other bromodomain factors, histone acetyltransferase 2, and proteins essential for transcriptional activity and RNA processing. Inducible deletion of BDF5, led to a disruption of pol II transcriptional activity and global defects in gene expression. Our results indicate the requirement of Leishmania to interpret histone acetylation marks for normal levels of gene expression and thus cellular viability.


122
Bromodomain factors in Leishmania 123 124 Although bromodomain factors BDF1-5 were readily identifiable in Leishmania genomes 30 , 138 further PFAM and HMM searching identified another two potential bromodomain-containing 139 proteins BDF6 and BDF7 (Fig. 1A insert in the bromodomain region and BDF7 lacks the conserved tyrosine and asparagine 152 residues, suggesting that they may be divergent, noncanonical bromodomains or pseudo-153 bromodomains 35 . BDF6 has a C-terminal bromodomain and is predicted to have an N-terminal 154 signal peptide when analysed using SignalP4.1. BDF7 is the largest of the BDFs and contains a 155 bromodomain in the C-terminal region of the protein preceded by a predicted ATPase and 156 AAA domain. The bromodomain does not appear to have the conserved tyrosine and 157 asparagine residues that are important for acetyl-lysine binding. However, by HMMER 158 analysis and alignment, it appears that BDF7 might be a homologue of the ATAD2 factor found 159 in many other eukaryotes [36][37][38] . All of the predicted BDFs, apart from BDF6, contain 160 K[K/R]x[K/R] motifs that can act as a monopartite nuclear localisation signal 39 . The BDFs of 161 Leishmania have orthologs in other parasitic and free-living kinetoplastid organisms, except 162 for BDF1 in Bodo saltans 40,41 . BDF8 was identified by HHPred analysis of a hypothetical gene 163 identified using BDF5-proximity proteomics (this study), it may represent another pseudo-164 bromodomain. 165 166 To assess the essentiality of the seven BDFs in Leishmania promastigotes we used Cas9-167 targeted gene deletion. sgRNAs and repair templates were generated to target the CDS of 168 each gene and replace it completely (Fig. S1A) 42 . Two independent experiments were 169 performed, using either blasticidin (BSDI or BSD and neomycin (NEO) drug resistance repair 170 cassettes, leading to three semi-independent selections. Consistently, BDF6 and BDF7 null 171 mutants could be isolated (Fig. 1B, Fig. S1B, S1C, S1D). For BDF1-5 only heterozygote mutants 172 were ever isolated, indicating that although the Cas9 system was functioning, a copy of the 173 gene was required for promastigote survival and thus null mutants could not be generated. 174 175 LmxBDF5, while distinct from BET bromodomains, shares a characteristic tandem 176 bromodomain arrangement reminiscent of human BRD2 and BRD4, or the yeast RSC4 177 protein 4 . These proteins have all been identified to play roles in regulating transcription, so 178 due to this interesting feature of BDF5 we decided to investigate it in greater detail. BDF5 179 homologs are identifiable in all the kinetoplastid genomes available in TriTrypDB. The level of 180 amino acid conservation across the first bromodomain (BD5.1) is higher than the second 181 (BD5.2), but in all cases the conserved tyrosine and asparagine residues are retained in both 182 bromodomains (Fig. S2A), these correspond to Y40, N90, Y201 and N247 in LmxBDF5. Both 183 bromodomains have x-ray crystal structures available in the PDB (PDB ID: 5TCM, 5TCK),  184 confirming the bromodomain structural fold and positioning of conserved residues (Fig. S2B).

185
BDF5 was endogenously tagged using a Cas9 targeted approach to append a 3xMyc epitope 186 and the green fluorescent protein mNeonGreen to the N-terminus 42 to generate mNG:: BDF5. 187 This modification preserves the 3' UTR, which is necessary for regulating endogenous mRNA 188 levels in Leishmania allowing for native expression levels and dynamics through growth and 189 lifecycle stages. Live-cell widefield deconvolution microscopy of promastigotes identified that 190 mNG::BDF5 localised to the nucleus (Fig. 1C). The distribution of mNG::BDF5 within the 191 nucleus was heterogeneous, with foci found around the periphery of the nucleus and 192 excluded from the nucleolus (Fig. 1D). The expression of mNG::BDF5 persisted in amastigotes 193 where it was visualised in a structure consistent with the nucleus of intramacrophage 194 amastigotes (Fig. 1E) To gain a higher quality validation of BDF5 essentiality 44 and to investigate the phenotypes 229 resulting from loss of BDF5 in promastigotes, an inducible knockout strain was generated 230 using the DiCre system 45,46 (Fig. S2A, S2B). An L. mexicana strain expressing dimerisable, split 231 Cre recombinase was modified to carry a single, 6xHA epitope-tagged allele of BDF5 flanked 232 by loxP sites giving L.mx::DiCre ∆bdf5::BDF5::6xHA flox /BDF5. The second copy of BDF5 was 233 then deleted using a HYG resistance cassette giving the strain 234 L.mx::DiCre∆bdf5::HYG/∆bdf5::BDF5::6xHA flox , referred to as BDF5::6xHA -/+flx . In the absence 235 of rapamycin, this strain grew normally as per the parental DiCre strain. However, following 236 the addition of rapamycin, there was a marked reduction in the parasite growth ( Fig. 2A). 237 Rapamycin was added to cultures at 300 nM for 48 h at which point the cultures were diluted 238 to 1 x 10 5 cells per ml. Rapamycin was then added at 100 nM to suppress escape mutants 239 and the growth phenotype observed. At the 144 h time point, the rapamycin-treated flasks 240 contained ~98% fewer cells than the controls. Rapamycin did not affect the parental DiCre 241 strain, indicating that the effect was specific to the floxed strain where BDF5 could be deleted. 242 This phenotype was reproducible and observed in an independent, clonal cell line ( Fig. 2A).

243
PCR analysis of these populations at 72 h after rapamycin addition revealed that the 244 BDF5::6xHA flx allele had been excised (Fig. 2B). Some leaky excision of the BDF5::6xHA flx allele 245 was detectable in the untreated control samples. The levels of BDF5::6xHA protein at 72 h 246 were assessed by western blot, revealing a 90% reduction in the rapamycin-treated sample 247 compared to the control samples (Fig. 2C). Total protein Stain-Free technology was used to 248 provide loading controls, due to the potential for BDF5 deletion to impact on transcription of 249 housekeeping genes. To demonstrate that the deletion of BDF5 was essential for cellular 250 survival a clonogenic assay was applied to characterise the cells resulting from BDF5 excision 251 (Fig. 2D). A 98% reduction in survival of the BDF5::6xHA -/+flx strain was observed when cloned 252 in the presence of 100 nM rapamycin, moreover, those cells that survived retained the 253 BDF5::6xHA flox allele. 254 255 To ensure the phenotype was specific to BDF5 deletion and not due to off-target effects, an 256 allele of BDF5::GFP was re-introduced to the BDF5::6xHA -/+flx strain using the pNUS episome 47 257 ( Fig. S4) Clonal survival experiments were performed in media lacking drug selection for the 258 episome allowing for its loss if it confers no selective advantage. Clonal survival of the BDF5-259 complementation strain was ~50% after rapamycin addition, this is 25-fold higher than the 260 non-complemented, induced samples. While not 100% complementation it reflects the 261 potential for parasites to lose the episome (Fig. 2D), demonstrating the requirement for BDF5 262 for cellular survival. This experimental approach also allowed us to explore the essentiality of 263 the individual bromodomains by making point mutations at the conserved asparagine 264 residues in each bromodomain, N90 and N257 in BD5.1 and BD5.2 respectively (Fig. S4). These 265 were mutated to phenylalanine in anticipation that the bulky sidechain would displace any 266 binding peptide from the hydrophobic pocket 2,48,49 . Clonal survival was restored to similar 267 levels as those observed for the pNUS BDF5 complementation strain by the BDF5 N90F and 268 BDF5 N257F mutants (Fig. 2D), indicating that either these mutations are not disruptive, or that 269 any disruption due to mutation of a single BDF5 BD is tolerated by the cell. Three attempts 270 were made to generate double mutations in N90F/N257F but no viable populations of cells 271 were isolated, suggesting the BDF5 N90F/N257F is not tolerated by the cells. In light of this, we 272 used a DiCre inducible system 50 to flip-on expression of an extra BDF5 N90F/N257F ::GFP mutant 273 allele to look for dominant-negative phenotypes (Fig. S5A, Fig.S5B). Promastigote cultures 274 induced to express BDF5 N90F/N257F ::GFP exhibited a significant growth defect (Fig. S5C, S5D), 275 whereas those induced to express the BDF5::GFP protein did not exhibit this phenotype. 276 These experiments demonstrate that individually both bromodomains are redundant, but 277 that together they are required for the essential function of BDF5. 278 279 The ability to use the DiCre strains to validate target genes in Leishmania amastigotes is 280 restricted due to the toxicity of rapamycin to amastigotes and its immunomodulatory effect 281 in mammals 45 . Therefore, mid-log promastigote cultures of Lmx::DiCre, BDF5::6xHA -/+flx or 282 BDF5::6xHA -/+flx ::pNUS BDF5::GFP were treated with 500 nM rapamycin or DMSO for 72 hours 283 allowing them to induce deletion of BDF5 but still allow infectious, metacyclic promastigotes 284 to accumulate in culture. Excision of the BDF5 gene was verified by PCR ( Fig. S6A) and the 285 stationary cultures were used to infect BALB/c mice by a subcutaneous route into the rear 286 footpad. No apparent differences were observed in the size of the resulting footpad lesions 287 over the 8-week infection period ( Fig. S6B), however, there was a 50-fold reduction in the 288 parasite burden of the footpads when infected with BDF5::6xHA -/+flx rapamycin treated cells 289 compared to the BDF5::6xHA -/+flx DMSO treated cells or the parental strain (Fig. 2E). The 290 presence of the pNUS BDF5::GFP episome restored parasite burden in the rapamycin-treated 291 strain to a level not significantly different to that observed in its uninduced control (Fig. 2E).

292
The median parasite burdens of the BDF5::6xHA -/+flx strain rapamycin-treated strain in the 293 popliteal lymph nodes was ~10-fold lower than the control strain, but this difference was not 294 statistically significant (Fig. S6C). DNA extracted from footpads and lymph nodes, including 295 both host and amastigote DNA, was subjected to PCR analysis which detected non-excised 296 BDF5::6xHA -/+flx consistent with BDF5 being essential for amastigote survival as well as 297 promastigote survival (Fig. S6D). Clonal promastigote lines lacking the BDF5::6xHA -/+flx allele 298 could only be derived from the populations containing an addback copy of BDF5 (Fig. S6E). 299 We conclude that BDF5 is essential for successful infection of the mammalian host and is likely 300 to be essential for amastigote survival too. 301 302 ChIP-seq reveals BDF5 genomic distribution 303 304 305 Figure  Due to the importance of BDF5 for the survival of Leishmania parasites and the demonstration 315 that it is a nuclear protein, we sought to identify where it might be found in the context of 316 genomic architecture. The BDF5::6xHA protein expressed by the BDF5::6xHA -/+flx strain was 317 analysed by chromatin immunoprecipitation sequencing (ChIP-Seq). We identified 175 318 regions where BDF5 was determined to be enriched on the genome; these peaks were 319 distributed across all the 34 chromosomes and could be correlated with specific genomic 320 features (Fig. 3, Fig. S7A, S7B). Of the total peaks, 56 (32%) were associated with TSSs in 321 divergent strand switch regions (dSSRs). A further 30 (17%) were in subtelomeric regions likely 322 to be transcriptional start sites based on the orientation of the polycistronic transcription 323 unit. Forty-seven peaks (27%) were identified in internal regions of polycistronic transcription 324 units (PTUs) and a further 11 peaks overlapped with isolated tRNA genes (6%). Intriguingly, 325 31 peaks (18%) were found at convergent strand switch regions, which are likely 326 transcriptional termination sites. The size of the regions determined to be enriched for BDF5 327 varied, with the mean of peaks found at dSSRs encompassing ~10 kb (Fig. S7C). The shape of 328 BDF5 peaks over divergent strand switch regions tended to be broad and even, without 329 exhibiting the "twin-peaks" pattern seen for histone H3 acetylation in L. major 14 (Fig. S7A, 330 S7D). Peaks at both divergent and convergent SSRs tended to be symmetrical although they 331 were narrower and weaker at convergent SSRs (cSSRs) (Fig. S7D). Peaks found in PTUs were 332 asymmetric, rising steeply to a peak with a shallow decay in the direction of the PTU 333 transcription. The PTU peak enrichment levels were equivalent to those at dSSRs (Fig. S7E).

334
The finding that BDF5 predominantly localises to divergent SSRs and other TSSs suggests it 335 plays a role in polymerase II transcription. However, as a number of termination sites and 336 other classes of small RNA genes were also enriched for BDF5 this indicates it could also play 337 a generalised role in a range of transcriptional processes. Therefore, we sought to analyse the 338 protein complexes associated with BDF5 to give insight into its potential function. To identify the functional properties of the environment proximal to BDF we applied an in-371 situ proximity labelling technique, cross-linking BioID (XL-BioID) 51 . The promiscuous biotin 372 ligase BirA*, which generates a locally reactive (~10 nm) biotinoyl-5'-AMP 52 , was fused to the 373 N-terminus of BDF5 by endogenous tagging. The resultant parasites were incubated with 150 374 µM biotin for 18 h to permit labelling of proteins in proximity to BirA*::BDF5. The parasites 375 were then treated with a limited amount of dithiobis(succinimidyl propionate) (DSP) chemical 376 cross-linker, to increase the capture of proximal proteins which enriched with streptavidin, 377 trypsin digested and processed for LC-MS/MS analysis. Importantly, a control cell line was 378 treated in the same way to provide a control dataset of spatially segregated, nuclear proteins. 379 The nuclear-localised protein kinase KKT19 53 was chosen as it is expressed at similar levels to 380 BDF5 and localised to a distinct structure, the inner-kinetochore 54 . This provided a way to 381 subtract common background proteins labelled during the synthesis and trafficking of BDF5 382 to the nucleus as well as endogenously biotinylated cellular proteins. Following SAINTq 383 interaction scoring, 156 proteins were determined to be enriched at 1% FDR (Fig. 4, Table  384 S2.). A subset of these proteins was selected for endogenous tagging with 3xHA::mCherry in 385 mNG::BDF5 (which also contains a 3xmyc epitope) expressing strain to allow reciprocal co-386 immunoprecipitation and verification of the XL-BioID dataset (Fig. S8, Table S1). This also 387 served to confirm there was no co-localisation of BDF5 and KKT19, and thus it was an 388 appropriate control protein.

390
The BDF5 proximal proteins were assessed for potential function (Fig. 4) and assembled into 391 a loose network. We identified a core set of 11 highly enriched proteins (>10-fold), including 392 the bait protein BDF5. Also identified were BDF3, BDF4 and Histone Acetyltransferase 2 393 (HAT2), along with several hypothetical proteins LmxM. 35 (Table S2). However, we were again unable to detect any acetylated peptides 461 derived from histones. 462 463 The capacity of XL-BioID to enrich large amounts of proximal material allows it to be combined 464 with other methods, such as phosphoproteomics 51 . We engineered a cell line to carry 465 BDF5::miniTurboID for faster labelling kinetics and higher temporal resolution, allowing us to 466 explore BDF5-proximal phosphorylation events across the cell cycle of hydroxyurea 467 synchronised cultures. Following synchronisation release, 30 minute biotinylation timepoints 468 were carried out 0 , 4 and 8 h corresponding to G1/S, S and G2/M phase respectively.. 469 Samples were processed using the XL-BioID workflow, then proximal phosphopeptides were 470 enriched using Ti-IMAC resin prior to LC-MS/MS analysis. The resulting dataset was compared 471 to a reference phosphopeptide dataset derived from the kinetochore protein KKT3 51 . Two 472 BDF5-proximal phosphopeptides were identified in early-S phase which then rose to 19 and 473 13 as the cells progressed through the S and G2/M phases respectively (Fig. S9, Table S3). Of 474 these, 14 unambiguous phosphosites were detected in total for proteins in proximity to BDF5, 475 including several for BDF5 itself, pS135, pS133, pS317 and pS330. S135 and S133 are located 476 between the two bromodomains, while S317 and S330 are located after the second 477 bromodomain. LmxM. 35.2500, which was highly enriched in the original XL-BioID and 478 identified to contain a putative FHA domain, was itself found phosphorylated at S202, S208 479 and S545. LmxM.33.2300 (BDF8) was found to contain an ambiguous phosphosite at one of 480 six sites in the region of S50-S61 (Fig. S9) to ERCC spike-in controls for PTUs (n=120), on a scale of 0-100%. F. Metaplot of reads mapping 505 to the + and -strands, normalised to ERCC control at convergent SSRs (n=40) of DMSO treated 506 or rapamycin-treated BDF5 -/+flx cultures. Metaplot data is from 1 representative of the three 507 replicate RNA-seq datasets. 508 509 510 As most of the BDF5 enriched regions of the genome corresponded to transcriptional start 511 sites, and the proximal proteome contained factors associated with the transcription and 512 maturation of various classes of RNA we sought to assess the effect of BDF5 depletion on 513 cellular RNA levels. Promastigote cultures were stained for total RNA content using SYTO 514 RNASelect fluorescent stain at 24, 48 and 72 h timepoints and measured by flow cytometry 515 (Fig. 6A). For Lmx::DiCre strain, the addition of rapamycin caused no changes in the levels of 516 total RNA staining. SYTO RNASelect staining increased as cells progressed through log phases 517 of growth at 48-72 h time points. However, once BDF5 was deleted from the BDF5::6xHA -/+flx 518 cell line by the addition of rapamycin, there was a pronounced increase in the number of 519 cellular events containing very low levels of RNA staining, such that the profile at 72 h overlaps 520 with that for with unstained control cells. This result suggested that total levels of 521 transcription were reduced upon BDF5 deletion from cells. We investigated this in more detail 522 by using total, stranded RNAseq that included External RNA Controls Consortium (ERCC) 523 Spike-in controls 20 . Cultures of BDF5::6xHA -/+flx treated with Rapamycin or DMSO were 524 harvested, then RNA extraction buffer spiked with the 92 synthetic ERCC RNAs was used to 525 lyse the parasites for RNA purification. Following sequencing and read mapping these RNAs 526 were then used to provide a normalisation channel. Overall, a >50% reduction in the median 527 read depth was observed across protein-coding genes on all chromosomes (Fig. 6B). When 528 normalised read depths were compared using metaplots of divergent SSRs, this ~50% 529 reduction in transcriptional levels was reflected (Fig. 6C). However, no positional effects were 530 observed on transcriptional start sites (Fig. 6D). The 50% reduction in read depth was 531 reflected across PTUs (Fig. 6E) and at convergent SSRs (Fig. 6F). Strand-specific read depth at 532 cSSRs did not indicate any increase in transcriptional readthrough in BDF5-induced null cells 533 (Fig. 6F), suggesting the BDF5 located at these termination sites is not playing a role in 534 transcriptional termination. Overall, these results indicate BDF5 is important for global pol II-535 dependent gene transcription. 536 537 Transcriptionally active regions of kinetoplastid genomes often accumulate DNA damage 538 which occurs due to the formation of DNA-RNA hybrids (R-loops) 60,64 . As we detected proteins 539 involved in co-ordinating DNA repair in the BDF5 proximal proteome, and that this appears 540 to be a broader feature of BDF protein networks 65,66 , we examined if there was a link between 541 BDF5 and the DNA damage response in Leishmania. BDF5-induced-null promastigotes cease 542 growing quickly, whereas parasites deficient for genome-stability factors often die slowly 50 , 543 suggesting maintaining genome integrity is not the primary role of BDF5. Indeed, after using 544 western blotting to detect gH2A phosphorylation 67 , a sensitive marker for the cellular 545 response to DNA damage, we could not detect any increase in gH2A signal in BDF5-depleted 546 cells, nor was there any detectable difference in the gH2A response of these cells to a non-547 specific DNA damaging agent, phleomycin (Fig. S10). This indicates that there is no direct or 548 secondary role for BDF5 in DNA damage response. Despite enrichment in the BDF5 proximal 549 proteome for mRNA splicing factors, we did not find evidence to support trans-or cis-splicing 550 defects in BDF5 induced-null mutants using a qualitative RT-PCR assay. This assay was capable 551 of detecting splicing defects caused by inhibition of an analog-sensitised CRK9 by the bulky 552 kinase inhibitor 1NM-PP1 (Fig. S11, S12) 68 . 553 554 Discussion 555 Kinetoplastid parasites have evolved a genomic architecture that requires them to conduct 556 most gene transcription constitutively, in an apparently simplified manner and deal with 557 consequences of this using post-transcriptional regulation and specialised solutions to genes 558 requiring a "high-dose" 8 . Pol II transcriptional start sites may simply be maintained as open 559 chromatin. However, recent evidence has indicated these regions are actively regulated, 560 particularly through histone acetylation. How the cell interprets these marks is not completely 561 understood. Bromodomains are clearly critical components of this process in Leishmania; we 562 were unable to generate null mutants in five of the seven bromodomain encoding genes, also 563 implying there is no redundancy in their individual functions. Although failure to generate a 564 null is the most basic standard of genetic evidence for essential genes 44 , we were able to 565 generate high-quality, genetic target validation for BDF5, using inducible DiCre both in the 566 promastigote stage and during murine infections. BDF5 expression was confirmed in both 567 stages and expression levels were correlated with cellular growth rate in promastigote stages. 568 Combined with the rapid cytostatic phenotype occurring upon BDF5 inducible deletion, 569 followed by cell death, this identifies BDF5 as a regulator of cell growth and survival. This 570 finding demonstrates that the interpretation of histone acetylation is important for cellular 571 survival (Fig. 7), although for Leishmania the specific histone PTMs found at TSSs are not 572 currently defined. 573 574 When functionally characterising BDF5, our starting hypothesis was that BDF5 would localise 575 to polymerase II transcriptional start sites, so it was surprising to find BDF5-enriched peaks 576 associating with many other sites such as rRNA genes, tRNA genes and convergent strand 577 switch regions. This suggests BDF5 plays a multipurpose or generalist role in recruiting or 578 regulating chromatin to promote transcription by multiple polymerase complexes. This was 579 further emphasised by the proximity proteomics dataset, which revealed BDF5 to be close to 580 proteins involved in different processes linked to transcriptionally active chromatin, in 581 particular RNA maturation factors, DNA repair factors and polymerase associated complexes. 582 Our phenotypic analysis appeared to rule out roles for BDF5 in influencing the DNA damage 583 response and cis-and trans-splicing of mRNA but did demonstrate that it was required for 584 normal transcription of polymerase II PTUs. Due to the rRNA depletion method used and low 585 coverage over tRNA genes, we could not assess if pol I or pol III transcripts levels were 586 reduced. This could be determined using qPCR in future studies. Spike-in controlled total RNA 587 seq was previously used to study the influence of HAT1 and HAT2 on transcription in T. brucei 588 20 . It is striking that BDF5 knockout in L. mexicana phenocopies HAT1 knockdown in T. brucei, 589 both resulting in an overall reduction in transcription levels. In T. brucei, HAT1 is required for 590 acetylation of H2A.Z and H2B.V. Depletion of HAT1, and thus H2A.Z levels, leads to a 10-fold 591 decrease in the amount of chromatin-bound pol II, resulting in 50% reduced transcriptional 592 activity. Intriguingly, pol II levels at TSSs were not affected by this; the authors suggested 593 H2A.Z acetylation is required for optimal transcription of bound pol II. As BDF5 knockout 594 phenocopies HAT1 depletion and results in lower pol II activity, it might therefore be involved 595 in reading or applying acetylation of H2A.Z. Surprisingly, although we find HAT2 proximal to 596 BDF5, HAT1 was neither enriched nor detected in our XL-BioID dataset. This suggests that 597 there is a distinct spatial separation between BDF5, HAT2 and HAT1 in Leishmania (assuming 598 there is no technical reason HAT1 cannot be labelled by XL-BioID). BDF5 also did not co-599 precipitate very strongly with HAT2 (Fig. S8), suggesting any interaction between them might 600 be transient or indirect. We did not observe changes in the positioning of transcription 601 initiation, suggesting the HAT1/BDF5 phenotype over-rides any effect on HAT2 dysfunction if 602 this is indeed a BDF5 complex member. Purified L. donovani HAT2 has been shown to 603 acetylate H4K10 and it appears to be essential as only heterozygotes can be generated using 604 traditional knockout strategies 15,18,69 . L. donovani HAT2 -/+ heterozygotes grow slowly and 605 display a cell-cycle defect. No effect was determined on transcriptional initiation positioning 606 at TSSs but an S-phase cell cycle-dependent reduction on CYC4 and CYC9 genes was 607 reported 18 . Future work could investigate the requirement of BDF5 for cell cycle dependent 608 gene transcription in Leishmania, which is an interesting observation given the lack of obvious 609 gene specific promoters. in our proximity proteome (Fig. 4). We propose that these proteins represent a Conserved 615 Kinetoplastid Regulator of Transcription (CRKT) Complex, that is recruited to acetylated 616 histones at TSSs (Fig. 7). The association of BDF5 at transcriptional termination regions, albeit 617 in lower amounts, could indicate that BDF5 is included in a mobile complex that can progress 618 along chromatin and accumulates at start and termination sites. One complex could be the 619 PAF1 complex, a multifunctional complex associated with pol II initiation, elongation, pausing 620 and termination 70 . However, PAF1-complex is poorly characterised in kinetoplastids and the 621 PAF1 protein itself lacks identifiable orthologs in these organisms. T. brucei BDF5 has been 622 suggested as a potential component of a transcription initiation complex due to its dual 623 bromodomains and the interaction with proteins containing homology to TFIID TAF1 (which 624 also contains 2 bromodomains as well as protein kinase and acetyltransferase function). 625 626 Our findings further illustrate the strength of proteomic approaches to studying chromatin 627 regulation in kinetoplastids where the large TSSs allow plentiful material to be derived 20 . 628 Combined with XL-BioID this allowed for the enrichment of PTMs to be determined for many 629 of the complex members.  Table  733 S1. Thirty residue homology flanks were identified adjacent to the CDS and appended to 734 oligonucleotides designed to amplify drug resistance markers from blasticidin and neomycin 735 drug resistance plasmids pGL2208 and pGL2663 respectively. After amplification of the sgRNA 736 and resistance marker the PCR mixes were pooled and precipitated using standard ethanol 737 precipitation, resuspended in sterile water and added to a transfection mix with 1 x 10 7 mid-738 log promastigotes. The cell line used was L. mexicana T7/Cas9::HYG::SAT 42 . Transfection was  739  performed with an Amaxa Nucleofector 4D using program FI-115 and the Unstimulated  740 Human T-Cell Kit. The mix was resuspended in 10 ml HOMEM 20% FCS and immediately split 741 in two 5 ml aliquots. Following 6-18 h of recovery time the parasites were plated at 1:5, 1:50 742 and 1:500 dilutions in media containing the selective drug blasticidin or G418. Endogenous 743 tagging was performed using the pPLOT 3xMYC::mNG BSD donor vector to install N-terminal 744 tags to BDF5, preserving the 3' UTR for native mRNA regulation (Oligonucleotides defined in 745 Table S1).

747
DiCre 748 DiCre strains for BDF5 were generated as previously described 46 . Briefly the BDF5 CDS and 749 flanking regions were assembled into floxing or knockout plasmids using Gateway cloning, 750 BDF5 was cloned into pGL2314 to fuse a 6xHA C-terminal tag and flank with loxP sites 751 (Oligonucleotides defined in Table S1). Inducible deletion of BDF5 flx in DiCre cell lines was initiated by the addition of 300 nM 759 rapamycin (Abcam) to promastigotes cultures at 2 x 10 5 cells ml -1 . Cells were grown for 48 h 760 then passaged into new media at a concentration of 2 x 10 5 cells ml -1 ; induction was 761 maintained by the addition of 100 nM of rapamycin to suppress escape mutants. 762 763 Clonogenic assays 764 For clonogenic assays, mid-log cells were counted and then diluted to 1 cell per 800 µl and 765 plated out into 200 µl volumes in 3 x 96-well plates to yield approximately 100 clones. Cells 766 were plated in media ± 100 nM rapamycin and incubated at 25 o C for 3 w before counting of 767 viable colonies by both visual screening and microscopic analysis 768 769 770 Addback strains 771 To generate episomal addbacks the BDF5 CDS was amplified from L. mexicana genomic DNA 772 and cloned into the pNUS C-Ter GFP NEO (pGL1132) using HiFi Assembly (NEB) to generate a 773 complementation vector. This was used as a base for site-directed mutagenesis using the Q5 774 Mutagenesis product (NEB) to generate mutations in the conserved asparagine residues N90 775 (OL9577 and OL10352), N257 (OL9579 and OL10353) to phenylalanine in BDF5 BD5.1 and 776 Figure S2: CLUSTAL alignment of kinetoplastid BDF5 proteins. A. Amino acid sequences of BDF5 951 syntenic orthologues were aligned using the Clustal Omega plugin for CLC. Domains that were 952 readily identifiable using the PFAM search plugin are annotated by shaded boxes, conserved 953 tyrosine and asparagine residues are annotated by red lines within the shaded BD5.1 and 954 BD5.2 domains. B. X-ray crystal structures of LdBDF5 bromodomains generated by the SGC and 955 deposited at the PDB, conserved tyrosine residues coloured orange and conserved asparagine 956 residues in yellow. 957 958 Figure S3: Characterisation of BD5 using DiCre Inducible gene deletion in promastigotes. using rapamycin to dimerise the split Cre recombinase, exemplifying the ability to introduce 962 add-back alleles for functional genetics. C. Flow cytometry of methanol fixed, RNAse A treated, 963 propidium iodide stained promastigote cultures to characterise the effects of BDF5 knockout 964 on the cell cycle over a 72 h timecourse N=20,000 events. D. Live/dead analysis using flow 965 cytometry of non-fixed, propidium iodide treated promastigote cultures following BDF5 966 knockout. A 1 µg/ml phleomycin control was included. Points and error bars indicate mean ± 967 standard deviation, N=20, 000 events. 968