Histone divergence in trypanosomes results in unique alterations to nucleosome structure

Abstract Eukaryotes have a multitude of diverse mechanisms for organising and using their genomes, but the histones that make up chromatin are highly conserved. Unusually, histones from kinetoplastids are highly divergent. The structural and functional consequences of this variation are unknown. Here, we have biochemically and structurally characterised nucleosome core particles (NCPs) from the kinetoplastid parasite Trypanosoma brucei. A structure of the T. brucei NCP reveals that global histone architecture is conserved, but specific sequence alterations lead to distinct DNA and protein interaction interfaces. The T. brucei NCP is unstable and has weakened overall DNA binding. However, dramatic changes at the H2A-H2B interface introduce local reinforcement of DNA contacts. The T. brucei acidic patch has altered topology and is refractory to known binders, indicating that the nature of chromatin interactions in T. brucei may be unique. Overall, our results provide a detailed molecular basis for understanding evolutionary divergence in chromatin structure.


INTRODUCTION
Nucleosomes are the basic unit of chromatin and contr ol chr oma tin-associa ted processes in eukaryotic genomes. The nucleosome core particle (NCP) is composed of DNA wrapped around an octamer of four histone proteins (H2A, H2B, H3 and H4) and serves as a DNA compaction unit ( 1 , 2 ), inhibitor of transcription ( 3 ) and dynamic molecular interaction platform ( 4 ). In line with their role as architectural proteins, histone sequences are highly conserved, especially in well-studied eukaryotes ( 5 , 6 ). NCP structures from v ertebrates ( 7 , 8 ), inv ertebrates ( 9 ) and unicellular eukaryotes such as yeasts ( 10 , 11 ) hav e re v ealed remar kab le conservation in global histone ar chitectur e and key sites of histone-DNA interactions. Howe v er, e v en small changes to histone primary sequences can have large structural and functional consequences on NCP structure ( 8 , 12-14 ).
Histone sequence variation and nucleosome structure in highly di v ergent eukaryotes ar e r elati v ely understudied. The group Kinetoplastida is ranked amongst the most evolutionarily ancestral groups of parasitic protists and was estimated to have split from other eukaryotic lineages around 500 million years ago ( 15 ). Kinetoplastida includes multiple patho gens, particularl y those belonging to the Trypanosoma and Leishmania species. Of these, Trypanosoma brucei is a major clinical target, causing both human and animal trypanosomiasis (16)(17)(18). In T. brucei, chromatin accessibility has a direct effect on antigenic variation, a key immune evasion mechanism contributing to its pathogenicity ( 19 ). Further r esear ch on understanding how the trypanosome genome is organised and responds to stimuli could therefore have direct clinical and economic benefits.
Trypanosome chromatin has a number of unusual features. The T. brucei genome is organised into 11 large megabase chromosomes, a small, varying number of intermediate-sized chromosomes, and ∼100 minichromosomes ( 20 ). Unlike in Metazoa, mitotic chromosome compaction le v els in T. brucei are low ( ∼1.2-fold) ( 21 ). Most genes are arranged in intron-less polycistronic transcription units ( 20 ) and the chromatin context of transcription initia tion and termina tion is in part defined by histones (22)(23)(24). T. brucei histones have been found to be highly div ergent ( 25 ), and nov el trypanosome-specific histone posttransla tional modifica tions ( 24 , 26 ) and chromatin interactors (27)(28)(29)(30) have been identified. Howe v er, the molecular details of how local nucleosome-le v el chromatin structure and molecular pathways in the nucleus intersect have been largely unexplored.
Her e, we pr esent the cryo-EM structur e of the T. brucei NCP. The structure re v eals altered histone-DNA contact sites , histone-histone interactions , and exposed histone surfaces compared to well-studied model eukaryotes. Globally, T. brucei NCPs are unstable and have reduced DNA binding. Howe v er, this instability is partly compensated by drama tic altera tions in the electr ostatic pr operties of T. brucei NCPs at the H2A-H2B binding interface. Furthermore, the surface topology, charge distribution, and binding properties of the T. brucei acidic patch ar e alter ed. Our phylo genetic anal ysis of kinetoplastid histones and their predicted structur es r e v eals tha t these dif ferences identified for T. brucei NCPs are likely conserved across the kinetoplastids. Overall, our study provides a molecular basis for understanding and further exploring how DNA compaction and chromatin interactions occur in T. brucei .
Histone protein mass was confirmed by 1D intact weight ESI mass spectrometry (SIRCAMs, School of Chemistry, Uni v ersity of Edinburgh) (Supplementary Figure S2A). Concentrations were determined via absorbance at 280 nm using a Nanodrop One spectrophotometer (Thermo Scientific).

NCP reconstitution
NCPs were reconstituted essentially as described ( 33 , 34 ) with some alterations to improve stability. Briefly, DN A for wrapping all NCPs except for hydroxyl radical footprinting assays was generated by isolating large-scale quantities of the plasmids pUC57 8 × 145 bp Widom-601 DNA or 32 × 147 bp alpha-satellite DN A by m ultiple rounds of MaxiPrep Kit purifications (Qiagen). The 145 bp fragments were digested and extracted from the plasmid using EcoRV digestion and subsequent PEG and ethanol precipitation steps. For hydroxyl radical footprinting, linker DNA was r equir ed to avoid initial undigested signal and the 175bp Widom-601 sequence was used. Fluorescently-tagged DNA was generated by PCR essentially as described ( 35 , 37 ) . T. brucei 147 bp centromer e-associated r epeat DNA and 177 bp minichromosome DNA sequences were obtained as gblock fragments (Integrated DN A Technolo gies), cloned into pUC57 plasmid vectors, amplified by PCR, and purified as described ( 35 , 37 ). All primers and sequences used can be found in Supplementary Table S2.
Purified octamers were wrapped with the DNA using an 18 h exponential salt reduction gradient. The extent and purity of NCP wrapping was checked by nati v e PAGE and SDS-PAGE analysis (Supplementary Figure S2B Where necessary, NCPs were purified on HiLoad 16 / 600 Super de x 200 size exclusion column (GE Healthcare) in 20 mM HEPES pH 7.5, 150 mM NaCl, 1 mM DTT to enrich for NCP only fractions. NCPs were then dialysed for 3 h into a customised Storage Buffer (25 mM NaCl, 2.5% glycerol (v / v), 15 mM HEPES pH 7.5, 1 mM DTT), concentra ted, and stored a t 4 • C for maximum of 1 month. For all biochemical experiments, H. sapiens and T. brucei NCPs wer e tr eated identically and processed concurr ently.

Cryo-EM grid pr epar ation and tr ansmission electron microscopy
For cryo-EM grid preparation, NCPs were diluted to a final DNA concentration of 110 g / ml (DNA concentration) and NaCl concentration of 50 mM. Glutaraldehyde crosslinking agent was added (0.05%) and incubated on ice for 5 min. The reaction was quenched with excess ammonium bicarbonate and Tris pH8. NCPs were concentred through a 100 kDa spin concentrator column (Amicon ® Devices) and loaded on an HiLoad 16 / 600 Super de x 200 size exclusion column (GE Healthcare) in 20 mM HEPEs pH 7.5, 150 mM NaCl, 1mM DTT. Fractions enriched for NCPs were pooled and concentrated (Supplementary Figure S3C).
Monodispersity of the sample was confirmed by negati v e staining as described ( 38 ). Briefly, 5 g / ml NCPs were applied to 300 mesh copper-grids with continuous carbonfilm (C267, TAAB) and stained with 2% uranyl acetate for 2 min prior to washing. Grids were loaded and imaged in F20 TEM operated at 200 kV. Images were collected manually using the EMMENU software (TVIPS) on a TemCam F816 camera (TVIPS) (Uni v ersity of Edinburgh, Transmission EM facility) (Supplementary Figure S3A).
For single particle cryogenic electr on micr oscopy, 3.5 l of freshly purified and crosslinked NCPs were applied to glo w dischar ged holey carbon quantifoil R 2 / 2 grids at a concentration of 2.2 M. Grids were incubated and blotted at 100% humidity and 4 • C in a vitrobot mark IV, prior to vitrification in liquid ethane and storage in liquid nitrogen.
Grids were screened for ice quality and a small dataset was collected and processed to 2D classes on a TF20 microscope (Uni v ersity of Edinburgh, Cryo-transmission EM facility). Data collection was then performed on a Titan Krios opera ted a t 300 kV equipped with a Ga tan K3, opera ting in correlated double sampling mode. 4193 Lzw compressed tiff movies were obtained using automated serialEM software ( 39 ) using a pixel size of 0.829 Å and a total dose of 45.7 electrons / Å 2 (Supplementary Table S1).

Cryo-EM image processing
All micro gra phs wer e motion-corr ected using MotionCor2, removing the first fr ame. CTF par ameters were estimated using patch CTF in cryoSPARC ( 40 ) and poor micro gra phs were discarded. ∼1000 particles were picked manually and 2D classified to produce templates for template-based picking in cryoSPARC. Two rounds of 2D classification were performed to discard poor ly aver aged particles and discernible secondary-structure features were pooled. The selected classes were used for ab-initio reconstruction and separated into two ab-initio classes. The best class comprising 306 475 particles were re-extracted with a 316 pixel v o xel size and subjected to local CTF refinement and homogeneous refinement with a dynamic mask starting at a resolution of 20 Å and yielding a final map at 3.28 Å resolution. This map was used for all model building and figure pr eparation. Non-uniform r efinement ( 41 ) was performed, removing some noise and yielding a GS-FSC map of 3.22 Å . C2 symmetry was applied in homogenous refinement. These maps were used only to aid map interpretability during model building. Map quality and anisotropy were assessed manually in Chimera ( 42 ) and using 3D-FSC ( 43 ). 3D classification was performed using multiple starting classes using the heterogeneous refinement job in cryoSPARC (Figure 2 D).

Model building
The crystal structure of the 145 bp Widom-601 DNA was used from PDB: 3LZ0 ( 44 ) in the most lo gicall y fitting orientation based on Widom 601 DNA asymmetry and best model to map fits. Initial models for T. brucei H2A and H2B histones were generated as a dimeric assembly and T. brucei H3 and H4 as a tetrameric assembly with glycine linkers in Alphafold2 ( 45 ) and docked in the EM map using UCSF ChimeraX ( 46 ). The model was adjusted using Coot ( 47 ) and ISOLDE ( 48 ). Sequences outside of the density were removed manually in Coot and refined using using Phenix real space refinement ( 49 ). Protein geometry was assessed with MolProbity ( 50 ). Model fit was assessed using mapto-model cross correlations ( 48 ) and EMringer ( 51 ). Models hav e reasonab le stereochemistry and are in good agreement with the EM density maps. Figur es wer e pr epar ed in UCSF Chimera ( 42 ) and ChimeraX ( 46 ).

Small angle X-ray scattering (SAXS)
SEC-SAXS experiments were performed at Diamond Light Source on the B21 beamline. Freshly prepared H. sapiens and T. brucei NCPs were loaded on column at 2.4 mg / ml Nucleic Acids Research, 2023, Vol. 51, No. 15 7885 (quantification based on total NCP concentration) and separated by an S200 Increase 3.2 size exclusion column in 20mM HEPEs pH 7.5, 150 mM NaCl, and 1 mM DTT prior to injection into the beamline and recorded with 3 s exposure. Data was reduced and analysed using Sc Å tter IV ( 52 ).

Protein structure analysis
Structural alignments between T. brucei and H. sapiens histones were generated using UCSF Chimera or ChimeraX ( 42 , 46 ). Hydrogen bonding interactions between the histones and DNA in the T. brucei structure were calculated using PISA ( 53 ). The overall molecular dipole moments of T. brucei and H. sapiens histone octamers were predicted using the Protein Dipole Moments Server ( 54 ). The N-and C-terminal tails of H. sapiens histones were truncated for this purpose based on the T. brucei structure. p K a values of residues in both T. brucei and H. sapiens H2A-H2B dimers at SHL3.5 were predicted using PROPKA ( 55 ) (Supplementary Data File 1). The surface area of the acidic patch was estimated using PyMOL.

Protein sequence and phylogenetic analysis
To generate phylogenetic trees for each histone, sequences from 20 different organisms sampling kinetoplastids and other eukaryotes were collected from T riT rypDB ( 56 ) or NCBI Protein, respecti v ely ( 57 ). Multiple sequence alignments were performed with MAFFT ( 58 ) and visualized with Jalview ( 59 ). Maximum likelihood phylogenetic trees were estimated using IQ-TREE ( 60 ), rooted at midpoint, and visualised with iTOL ( 61 ). Heatmaps showing percentage identity to T. brucei were generated with iTOL using percentage identity matrices calculated by MUSCLE ( 62 ). P airwise per centage identities and similarities between H. sapiens and T. brucei histones were computed using EM-BOSS Needle ( 63 ). Isoelectric point (p I ) values were obtained from Protparam ( 64 ).
To obtain larger percentage identity matrices, the TriT-rypDB ( 56 ) and NCBI Protein ( 57 ) databases were mined for histone sequences from 41 organisms (of which 22 were primarily from kinetoplastid r efer ence genomes) and the ma trices calcula ted using MUSCLE. The r esults wer e then displayed as correlation maps coloured from 40-100% sequence identity using ggplot2.

Thermal denaturation assays
50 l reactions with 0.5 M NCPs, 20 mM HEPES pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM DTT and 5 × SYPRO orange (Life Technologies) were set up in a 96-well pla te forma t and heated from 45 • C to 95 • C with 0.5 • C increments on a Biometra TOptical RT-PCR de vice (e xcitation / emission = 490 / 580 nm). Relati v e fluorescence intensity was normalized as (RFU-RFUmin) / (RFUmax-RFUmin). Tetrasome data was normalized from 60 • C due to inherent background signal at lower temperatures. Results from three independent experiments, each with two technical r epeats wer e used to calculate T m values for H. sapiens and T. brucei NCPs (Supplementary Data File 2).

Salt stability assays
250 ng of T. brucei and H. sapiens NCPs (A260 DNA-based quantification) were incubated at various NaCl concentrations (0.5, 1.0, 1.5 and 2M) for 1 h in 10 l reactions on ice (2.5% glycerol, 15 mM HEPES pH 7.5, 1 mM DTT). After 1 h, NaCl concentrations were normalized to 0.15 M and 22 ng of each sample was loaded onto a 5% Trisgl ycine pol yacrylamide non-denaturing gel. A control sample kept in Storage Buffer (see 'NCP Reconstitution') was adjacently loaded. Percentage of wrapped NCPs vs. DNA was quantified using the BioRad Image Lab software by comparing the relati v e contribution of the NCP band and DN A band (summed to gether as 100%) in each lane. The assay was performed in triplicate for each type of NCP using independentl y-wra pped NCPs. Quantification can be accessed in Supplementary Data File 3.
Hydro xyl r adical f ootprinting 10 l samples containing 500 ng of fluorescently labelled nucleosomes (5 6-FAM labelled re v erse strand, 5 -TAMRA labelled forward strand) were set up in reaction buffer (15 mM HEPES pH 7.5, 25 mM NaCl, 1 mM EDTA, 1 mM DTT). 2.5 l each of 2 mM ammonium iron (II) sulfate / 4 mM EDTA, 0.1 M sodium ascorbate, and 0.12% H 2 O 2 were pipetted onto the sides of the reaction tube, mixed together and added to the sample. The reaction was stopped after 4 min by the addition of 100 l STOP buffer (100 mM Tris pH 7.5, 1% glycerol, 325 mM EDTA, 0.1% SDS, 0.1 mg / ml ProteinaseK [Thermo]). The stopped reaction was then incubated for 20 min at 56 • C to allow ProteinaseK digestion to occur. Fragmented DNA was purified by ethanol pr ecipitation and r esupended in 10 l HiDi Formamide. 0.5 l of GeneScan 500 LIZ size standard (Thermo) was added as a size marker. The resuspended DNA was run on either a 3130xl Genetic or 3730xl DN A Anal yzer, operated using the G5 dye filter set. Peaks were analysed using Thermofisher Connect Microsatellite analysis software. Peak size in base pairs were called by the Global southern method. (1 × NEB CutSmart Buffer, 5 U enzyme / g DNA) for 1 h a t 37 • C . Restriction enzymes were hea t inactiva ted a t 80 • C for 2 min. Widom 601 171 bp DNA was generated using PCR as described in the NCP reconstitution section. The final ladder comprised fiv e larger fragments (171, 145, 135, 124 and 114 bp). The fragments were mixed at an equal DNA mass ratio and an aliquot of the final mixture ( ∼20 ng of each fragment) was loaded on a non-denaturing 5% polyacrylamide gel.

Exonuclease III assays
Exonuclease assays were preformed essentially as described ( 32 , 67 ). Briefly, 2 Units of Exonuclease III (Takara) were added to 1 ug of NCPs in ExoIII digestion buffer (50 mM Tris-HCl (pH 8.0), 5 mM MgCl 2 , 150 mM NaCl and 1 mM DTT). The reaction was incubated at 25 • C and samples quenched in stop buffer (20 mM Tris pH 8, 200 mM NaCl, 0.5% SDS, 25 mM EDTA) at regular intervals. DNA products were deproteinized by digestion with 30 ug of proteinase K followed by ethanol precipitation. Samples were processed identically and resuspended in equal volumes of HiDi Formamide, prior to running on a denaturing urea 10% polyacrylamide gel and stained with Diamond DNA stain (Promega). The experiment was repeated in triplicate and quantified using BioRad Image Lab software, quantifying the disappearance of full-length uncut band versus the 0 timepoint taken prior to addition of Exonuclease.

Fluor escence anisotrop y peptide binding assays
NCP-binding FITC-labelled peptides deri v ed from K aposi's Sarcoma Herpesvirus La tency Associa ted Nuclear Antigen (LANA) ( 68 ) and PFV-GAG ( 36 ) were synthesised to > 95% purity by BioMatik, Canada. The LANA peptide with mutations L8A R9A S10A ('LANA LRS') and the PFV-GAG peptide with the mutation R540Q ('PFV-GAG RQ') were also synthesized and used as non-binding controls. The peptide sequences used are gi v en below: LANA Fluor escence anisotrop y assays wer e performed essentially as described ( 69 ). 50 nM of peptide tracer was incubated with an increasing concentration of T. brucei or H. sapiens NCPs from 12.5 nM to 2.4 M (corresponding to 25 nM to 4.8 M NCP binding sites) in 20 mM HEPEs pH7.5, 150 mM NaCl, 0.5 mM EDTA, 1 mM DTT and 0.05% Triton X-100. Assays were performed in 25 l reactions in a 384-well plate and incubated covered at 20 • C for 30 min. Fluorescence polarization was measured in an M5 multimode plate reader (Molecular Devices), with 480 nm excitation and 540 nm polarised filters (cut-off at 530 nm). Anisotropy ( r ) was calculated as below, where the grating factor ( G ) was approximated to 1, based on FITC alone measurements: The resulting values were background subtracted (no NCPs) and plotted against the number of binding sites on the NCPs (2 protein-binding faces per NCP). The experiment was performed in triplicate (Supplementary Data File 5). The resulting graph was fitted using GraphPad Prism with a non-linear r egr ession, sa tura tion binding curve assuming one site with total and non-specific binding. Binding affinity measurement was estimated using the equation: analysed on a non-denaturing, 5% polyacrylamide TBE gel. Gels were stained with Diamond DNA stain (Promega) and the disappearance of the NCP band quantified using Image-Lab. Quantifica tion da ta can be accessed in Supplementary Data File 5.

Kinetoplastid histones are highly divergent
Pr evious r eports have shown that a subset of trypanosomatid histones are highly di v ergent compared to model eukary otes ( 25 ). We perf ormed an extended analysis of histone sequences from 22 distinct kinetoplastid genomes including multiple clades, such as those from the Trypanosoma, Leishmania, Endotrypanum, Crithidia, Angomonas and Perkinsela species ( 70 ). This analysis re v ealed a clear e volutionary divide between the kinetoplastids and a wide sample of eukaryotic taxa (Supplementary Figure S1A). The divide was apparent across all histones, particularly H2A and H2B (Figure 1 A). Within the kinetoplastids, our analysis pointed to multiple sub-groups of conserved sequences including segregation between the Trypanosoma and Leishmania spp. (Supplementary Figure S1A). Aligned to Homo sapiens histones, the sequence identities of T. brucei histones was low, ranging from 40 to 60% ( Figure S1A). Low conservation was apparent e v en when the predicted unstructured (and more commonly di v ergent) N-and C-terminal tails were excluded (Figure 1 A). This prompted us to investigate whether histone sequence di v ergence in T. brucei leads to functional differences in nucleosome structure, assembly, and function.

Reconstituting the T. brucei nucleosome core particle (NCP)
To understand the basic unit of chromatin in T. brucei, we used an in vitro reconstitution approach to assemble recombinant T. brucei nucleosome core particles (NCPs). The four cor e histones wer e expr essed and purified from Esc heric hia coli (Figure 1 B & Supplementary Figure S2A), refolded into octamers, and wrapped with the strong positioning Widom-601 DNA sequence using standard salt dialysis protocols ( 31 , 33 , 34 ). The process of wrapping octamers with DNA required optimisation due to the presence of a soluble higher molecular weight species (Supplementary Figure S2B-D, see Materials and Methods). We found that the thermal stability of T. brucei NCPs was significantly reduced compared to that of H. sapiens NCPs (Figure 1 C). Wrapping T. brucei histone octamers with 147 bp alpha-satellite DNA, another 'strong' positioning sequence ( 7 ), also led to more unstable NCPs (Supplementary Figure S2E). On a coarse le v el, H. sapiens and T. brucei NCPs appeared structurally similar. By negati v e stain transmission electr on micr oscop y, the T. brucei NCPs appear ed as ∼10 nm disk shapes, reminiscent of NCPs from other species (Supplementary Figure S3A). Furthermore, similar dimensions for T. brucei and H. sapiens NCPs were obtained by in solution small angle X-ray scattering (SAXS) (Supplementary Figure S3B), suggesting that despite its differences in stability, at low resolutions, T. brucei nucleosome structure is maintained.

The cryo-EM structure of T. Brucei NCP reveals compressed histone ar chitectur e
In order to investigate the molecular details of di v ergence in T. brucei nucleosomes, we determined the structure of T. brucei NCP at a global resolution of 3.3 Å by single particle cryogenic electr on micr oscop y (cryo-EM) (Figur e 1 D, Supplementary Figure S3, Supplementary Movie S1). The low stability of the NCPs r equir ed us to perform mild chemical crosslinking prior to purification and sample preparation (Supplementary Figure S3C). The resolution was sufficient to allow us to build a model into the EM density for the core of the NCP (Figure 2 A, Supplementary Figure S4A, Supplementary Table S1).
The T. brucei NCP forms a characteristic coin-like shape, compacting and bending DNA around the core of the histone octamer (Figure 1 D). Although the histone secondary structure is largely preserved (Supplementary Figure S4B Figure S5B).
At SHL2, we observed tighter packing of the ␣1 helix and loop 1 region (the 'H3 elbow' ( 4 )) of histone H3 to the ␣2 helix of H4 (Figure 2 B, Supplemental Movie 2). This likely occurs due to overall reduction in hydrophobicity and the loss of a bulky aromatic residue ( H. sapiens ( 'Hs ') Phe78 versus T. brucei ( 'Tb ') Gln75) that shifts the ␣1-helix of H3 inw ards tow ards H4, increasing DNA bending in order to maintain arginine-phosphate interactions (Supplementary Figure S5C). Combined, these variations in histone architectur e r esult in an oval-shaped NCP with alter ed DNA binding.

The T. brucei NCP has more flexible entry / exit DNA
The ends of the wrapped DNA at the entry / exit site of the T. brucei NCP are poorly ordered in the cryo-EM density (Figure 1 D). We could reliably model only 126 bp of the 145 bp pr esent, pr esuming that the missing ends of the DNA are fle xib ly tether ed (Figur e 2 C). Indeed, during image processing, distinct 3D classes of DNA-ordered states were obtained. Roughly one third of the da ta indica ted a full y wra pped conf ormation, albeit with a bulged f orm of DNA, that was previously observed in Xenopus laevis NCPs as a precursor to unwrapping ( 71 ).The remaining two thirds lacked density around the DNA entry / exit sites ( Figure  2 D), indicating flexibility of DNA ends.    The pseudo-symmetry in the T. brucei NCP is broken, with one end of the DN A being considerabl y more disordered. This is likely due to the asymmetry of the Widom 601 sequence ( 44 , 72 ) and is consistent with partial asymmetric unwrapping observed previously (71)(72)(73)(74)(75). The inwards compression of DNA observed at SHL2 and 6 and the more oval shape of the T. brucei NCP (Figure 2 B) may contribute to this splaying of DNA ends. Despite DNA entry / exit site flexibility, we note that the histone-DNA register at the core of the histone octamer is maintained overall, both from comparison to other structures and a hydroxyl radical footprinting assay (Supplementary Figure S5D).
Unsurprisingly, the density for T. brucei histone tails was low, arising from a high degree of disor der ( 76 ). Howe v er, compared to H. sapiens and X. laevis EM density maps at similar r esolutions ( 71 , 77 ), ther e is poor er ordering for the C-terminal tail of H2A and the N-terminal tail of H3, which engage the final ∼13 bp of straight entry / exit DNA on the NCP ( 7 ). T. brucei has a number of amino acid substitutions in histones H2A and H3 in these terminal regions (Supplementary Figure S4B). Notably, Hs Arg53 is altered to Tb Gln50 in the ␣N helix of histone H3 (Supplementary Figure S5E and S5F). Substitutions at this position were previously shown to be critical for destabilization of entry / exit DNA binding in nucleosomes containing human H3 variants ( 14 , 78 ) DNA end accessibility was also observed in T. brucei NCPs using an exonuclease assay (Supplementary Figure S5G). Overall, the oval morphology and poor ordering of H3 / H2A tails likely leads to flexibility of entry / exit DNA in the T. brucei NCP.

Alterations in histone-histone interfaces in T. Brucei NCP lead to instability
Despite the overall conservation of the fold of the T. brucei histone octamer, notable changes are present at histonehistone interfaces in the T. brucei NCP. Within the H3-H4 tetramer, amino acid substitutions affecting steric packing between H3-H3 and the H3-H4 interface likely alter the stability of the T. brucei tetramer and octamer. Electrostatic interactions and the hydrophobic core of the H3-H3 fourhelix bundle are disrupted ( Hs H3-Ala111, Hs H3-Ala114 and H3-Leu109 to Tb H3-Cys108, Tb H3-Ser111 and Tb H3-Arg106, respecti v ely) (Figure 3 A). Single substitutions in this region hav e pre viously been shown to destabilize NCPs ( 12 , 79 ). At the H3-H4 interface , altered hydrophobic interactions arise due to the substitution of Hs H3-Phe104 to Tb H3-Leu101 (Supplementary Figure S6G). Interestingly, substitutions at this position were also reported in unstable nucleosomes ( 13 , 80 ). Multiple sites that are critical in imparting stability to the H3-H4 tetramer are therefore altered in T. brucei. Intriguingly, the H3-Cys110 pair that has been implicated in redox sensing ( 81 ) is not structurally conserved. Instead, it is replaced by two cysteine pairs that are spatially separated ( Tb H3-Cys108, Tb H3-Cys126).
At the dimer-tetramer interface, the T. brucei H2B-H4 four-helix bundle exhibits a loss of aromatic and hydrogen bonding interactions. A -stacking network mediated by Hs H2B-T yr83, H4-T yr72 and H4-T yr88 is absent due to the loss of two of the three tyrosines in T. brucei (Figure 3 B). A similar loss has been documented in the unstable nucleosome-like structures formed by Melbournevirus histone doublets ( 82 ). Additionally, a multiv alent hy drogen bonding network centred on H2B-His75 in H. sapiens NCPs is disrupted through substitution to H2B Tyr-73 in T. brucei (Figure 3 C). Conversely, a reinforcement of binding likely occurs at the H2A-H2B dimer-dimer interface, where Tb H2A-Tyr39 and Tb H2B'-Arg70 likely form a cationinteraction that is lacking in H. sapiens (Figure 3 D).
These observations agree well with our attempts to reconstitute hybrid T. brucei / H. sapiens histone octamers. Octamers with Hs H3 + Tb H2A, H2B and H4 could assemble due to the compatibility of Hs H3-Hs H3 and minor changes at the H3-H4 interface described (Supplementary Figure S2D; Supplementary Figure S6F). NCPs re-constituted with this ' Tb + Hs H3' octamer had slightly higher thermal stability (Supplementary Figure S6H), indica ting tha t the Tb H3-H3 interface has lower cohesion.
Howe v er, more e xtensi v ely altered histone-histone interfaces pre v ented the assemb ly of hybrid octamers containing Hs H2A + Tb H2B, H3 and H4 (Supplementary Figure  S6D) or Hs H2A-H2B + Tb H3-H4 ( Supplementary Figure S6E). Although heterodimerization of Hs H2A and Tb H2B itself was possible (Supplementary Figure S6D), the formation of a stable H2A-H2B dimer-dimer interface in the context of an assembled octamer was prevented.
The altered pr otein-pr otein interactions in T. brucei NCPs dri v e not only lower thermal stability but also result in an altered NCP disassembly pathway (Figure 3 E and F). As previously reported ( 83 , 84 ), H. sapiens NCPs exhibit biphasic disassembl y, w hereby the first step involves H2A-H2B disengagement from the H3-H4 tetramer follo wed by breakdo wn of the tetramer itself (Figure 3 E). Unusually, T. brucei NCP disassembly is a single e v ent (Figure  3 F). The major dri v ers are probably the instability of the dimer-tetramer interface and the instability of the tetramer itself, causing both NCP disassembly at a lower temperature and the disassembly of reconstituted T. brucei H3-H4 tetrasomes at a similar temperature to Tb NCPs (Figure 3 E).
Nucleic Acids Research, 2023, Vol. 51, No. 15 7891 Despite their slight increase in stability, the Tb + Hs H3 NCPs also exhibit monophasic disassembly (Supplementary Figure S6H), indicating inefficient integration of the Hs H3 into the altered histone configuration of the T. brucei octamer. Interestingly, the majority of the amino acid substitutions that introduce instability in T. brucei are conserved in the Trypanosoma spp. and would be expected to produce comparable effects in other Trypanosoma NCPs (Supplementary Figure 1B).

The T. brucei NCP has r educed o ver all protein-DNA interactions
Alterations in histone-DNA contacts also contribute to the instability of T. brucei NCPs. Similarly to H. sapiens NCPs, the structure of the histone folds maintains the global alignment of helix dipoles to the DNA phosphate backbone. The total number of histone-DN A hydro gen bonding interactions in the two NCPs is near-equivalent (Figure 4 A). Howe v er, the distribution of these bonds is altered across the four histone types (Figure 4 A) and changes in electrostatic interactions at specific DNA contact points lead to an overall reduction of DNA binding in the T. brucei NCP.
At SHL 0 (the dyad), two lysines from histone H3 ( Hs Lys115) are substituted for glycine ( Tb Gly112) (Figure 4 B). At SHL1.5, a critical arginine in histone H3 ( Hs Arg63) ( 7 ) is substituted for Tb Gln60 (Figure 4 C). Interestingly, SHL2.5 features a case of histone co-evolution, where histones H2B and H4 in T. brucei have a spatially coincident arginine-to-lysine and lysine-to-arginine substitutions respecti v ely (Figure 4 D). This pre v ents steric clash compared to a single substitution and allows for extended interactions with Tb H2B-Glu81. Howe v er, the ne w interaction occurs at the expense of the proximity of Tb H2B-Lys74 to DNA (Figure 4 D).
The N-terminal tail of H2A, which typically straddles the minor groove at SHL4.5 ( 7 , 34-35 , 85 ) is poorly ordered in T. brucei (Figure 4 E) and a phosphate-interacting residue, Hs Arg11, is replaced by Tb Ala11 (Figure 4 E and F). The C-terminal helix of histone H2B is displaced and is shorter in length by two residues. The lysine that normally anchors the helix to DNA is ther efor e lacking ( Supplementary Figure S7A). These changes in H2A and H2B are conserved across the kinetoplastid species (Figure 4 F, Supplementary  Figure S7B).
Cumulati v ely, the loss of DNA contacts described above suggest weaker DNA binding by the T. brucei NCP and this is supported by both lower thermal stability (Figure 1 C) and lower resistance to increasing salt conditions compared to the H. sapiens NCP (Figure 4 G). Interestingly, the Tb + Hs H3 chimeric NCPs were also more salt labile (Supplementary Figure 7C). This highlights that the highly altered histone configuration in T. brucei octamers dominates over integration of new DNA contacts by Hs H3.
We also investigated the DNA binding properties of NCPs wrapped with sequences deri v ed from the T. brucei genome, namely the 147 bp centromere-associated repeat sequence ( 86 ) and the 177 bp minichromosome repeat sequence ( 87 ). NCPs could be formed using these nati v e sequences, but T. brucei NCPs were harder to reconstitute as a single species compared to H. sapiens NCPs (Supplementary Figure S7D). These nati v e sequences were also less well bound than the optimised Widom-601 wrapped NCPs in salt stability assays (Supplementary Figure S7E-G). Furthermore, T. brucei NCPs wrapped with nati v e sequences were less stable than H. sapiens NCPs, suggesting that the global lower DNA interaction is inherent to T. brucei histones rather than due to DNA sequence. Overall, our experiments point towards global reduction of histone-DNA contacts in the T. brucei NCP.
A concentrated cluster of positi ve charge dri ves DNA binding by H2A-H2B at SHL3.5 While overall DNA-protein contacts are reduced in the T. brucei NCP (Figure 4 , Supplementary Figure S5D), a unique exception occurs at SHL3.5, where a concentrated cluster of positi v ely charged residues reinforces DNA binding at the dimer-dimer interface of histones H2A and H2B. Compared to H. sapiens, T. brucei H2A and H2B have higher calculated net positi v e charge and this holds true e v en when the disordered histone tails are discounted (Figure 1 A). The structure shows higher electrostatic surface potential at SHL3.5 ( Figure 5 A) and the interface features net gain of positi v ely-charged side chains from two charge swapping e v ents ( Hs H2A-Glu41 to Tb H2A-Arg41, Hs H2B-Glu35 to Tb H2B-Arg23), two additional arginine residues ( Tb-H2B Ar g70, Tb-H2B Ar g75), and a lysine-toarginine substitution ( Hs H2A-Lys36 to Tb -H2A-Arg36; Figure 5 B). The dense cluster of positi v e charge dramatically alters the overall predicted dipole moment of T. brucei and H. sapiens histone octamers, where the direction of the dipole is oriented opposite to the dyad at SHL 3.5 ( Figure  5 C). Interestingly, this may be a kinetoplastid-specific adaptation based on conservation at the sequence le v el (Supplementary Figure S8A). Furthermore, this interface is similarly charged in models of histone octamers generated using AlphaFold2 ( 45 ) from fiv e r epr esentati v e kinetoplastid species, but not other NCP structures (Supplementary Figure S8B).
The observation of locally r einfor ced DNA interactions by the basic cluster in H2A and H2B correlates well with micrococcal nuclease (MNase) digestion assays. We observed a characteristic digestion band for T. brucei NCPs that was observed only weakly in H. sapiens NCPs (Figure 5 D). We hypothesise this corresponds to the MNase pausing at a stall point, prior to recovering and continuing to digest the DNA. This pattern of digestion was largely independent of the over all r ate of MNase digestion (Supplementary Figure S8C).
The same experiment repeated with NCPs reconstituted with tailless T. brucei histones did not change the strong stall pattern (Supplementary Figure S8D and S8E). Similarly, the chimeric Tb + Hs H3 NCPs did not affect the stall (Supplementary Figure S8E). Howe v er, the stall was reproduced in H. sapiens NCPs by mutating Hs H2A-H2B such that the SHL 3.5 contact interface mimics T.    Figure S8G). In contrast to H. sapiens , further digestion products can barely be detected (Peak 1 in Figure 5F; Supplementary Figure S8G). This is in good agreement with the presence of the highly basic region in the T. brucei NCP at SHL 3.5 and may serve as a unique kinetoplastid mechanism to anchor DN A firml y to the base of the nucleosome, despite globally reduced overall DNA binding.

Altered octamer surface shape and charge in the T. Brucei NCP leads to an atypical acidic patch
The acidic patch is a negati v ely charged region formed between histones H2A and H2B that serves as an interaction platf orm f or man y chroma tin-associa ted proteins ( 4 ). It has also been implicated in the formation of higher order chro-matin structure ( 7 , 88 ). Remar kab ly, gi v en the high sequence di v ergence, the negati v e charge of the canonical acidic patch residues in the T. brucei are conserved (Figure 6 A). Howe v er , the coulombic surface potential of the T. brucei acidic patch e xhibits star k differences to H. sapiens (Figure 6 B).
The overall shape of the T. brucei acidic patch appears narrower and the estimated surface area of the patch is smaller 7894 Nucleic Acids Research, 2023, Vol. 51, No. 15   (4498 Å 2 versus 4702 Å 2 ; Figure 6 B). This is likely driven by the two-residue insertion in Tb H2A Loop 2 that allows an inwards shift of the ␣2 helix of H2A at SHL6 (Figure 2 B). A number of other substitutions are present in this region, particularly in histone H2A (Figure 6 B). While three of the canonical residues have undergone mild glutamateto-aspartate substitutions ( Tb H2A-Asp93, Tb H2A-Asp94 and Tb H2B-Asp93), a lysine substitution ( Tb H2A-Lys68 from Hs H2A-Asn68) is in close proximity to the patch (Figure 6 C). This substitution is conserved across the kinetoplastids (Supplementary Figur e S9A) Inter estingl y, m utation of Hs H2A-Asn68 was recently shown to reduce binding to multiple chroma tin-associa ted proteins in a proteome-wide screen ( 89 ). Both the surface charge and shape alterations are also conserved in our predicted models of fiv e other kinetoplastid histone octamers (Supplementary Figure S9B).
These differences in the T. brucei acidic patch likely have direct functional consequences for chromatin reading. A prototypical acidic patch binder, the Kaposi's Sarcoma Herpesvirus Latency Associated Nuclear Antigen (LANA) peptide ( 68 ), binds well to the H. sapiens NCP but showed no detectable interaction to the T. brucei NCP in a fluorescence polarization assay (Figure 6 D). Similar results were obtained for an acidic patch binder with an altered binding mode (Supplementary Figure 9C), the Prototype Foam y V irus GA G peptide (PFV-GA G) ( 36 ), using both fluorescence polarization (Supplementary Figure 9D) and electrophoretic mobility shift assays ( Supplementary Figure S9E). Discrete mutations to individual components of the T. brucei acidic patch were not sufficient to rescue the binding of the PFV-GAG and LANA peptides (Supplementary Figure S9A, S9E and S9F). These results re v eal e xtensi v e di v ergence of this common protein interacting hub and indicate that the kinetoplastids likely have altered chr omatin-pr otein and higher order chromatin interactions.

DISCUSSION
In this study, we present the cryo-EM structure and biochemical characterisation of the nucleosome core particle from the parasitic protist T. brucei . Despite e xtensi v e di v ergence of T. brucei histone sequences (Figure 1 ), the structure e xhibits remar kab le conservation of ov erall histone fold architecture. This is consistent with structures from other diver gent or ganisms such as the parasite Giardia lamblia ( 90 ), the archaeon Methanothermus fervidus ( 91 ), or the Marseilleviridae ( 82 , 92 ) giant viruses. Howe v er, subtle differences in histone packing and specific histone amino acid substitutions gi v e rise to properties unique to the T. brucei NCP. The T. brucei NCP has a compressed shape (Figure 2 , Supplementary Movie S2) and four out of se v en DNA contact points at half-integral SHLs have lost critical histone-DNA interactions, leading to fle xib le DNA ends and low stability in vitro (Figure 2 -4 ). A striking compensation mechanism occurs via H2A-H2B dimers at SHL3.5, w hereby DN A binding is increased by a concentrated cluster of basic residues ( Figure 5 ). Furthermore, kinetoplastidspecific alterations lead to an altered topology and charge of common protein interaction interface, the acidic patch (Fig-ure 6 ). By comparing the molecular differences in nucleosomes between trypanosomes and conventional eukaryotic systems, we can hope to understand both the evolutionary constraints and di v ersity tha t directs DNA-associa ted mechanisms.
Early studies of trypanosome chromatin showed that T. brucei chromatin is less stable ( 93 ) and has reduced higher order compaction than is observed in model eukaryotes ( 21 , 94 , 95 ). Howe v er, at the sequence le v el, the T. brucei genome is not unusual in base composition nor less enriched for nucleosome positioning featur es ( 96 , 97 ). Our r esults indica te tha t chroma tin instability in T. brucei is inherent a t the mononucleosome le v el via weakened histone-histone interfaces , histone-DNA contacts , and fle xib le entry / e xit DNA (Figure 2 -4 ). Indeed, NCPs could be wrapped using nati v e T. brucei DNA sequences, but these were also less stable compared to H. sapiens NCPs wrapped using the same sequences (Supplementary Figure 7).
Chromatin arrays constructed from other nucleosomes with increased DNA end flexibility ( 80 , 90 , 98 ) were previously found to favour more open chromatin conformations, particularly due to alterations in inter-nucleosomal DNA path ( 98 ). Future experiments on chromatin arrays would help explore if entry / exit DNA flexibility described here could explain the more open chromatin in T. brucei .
The end flexibility we observed was reminiscent of nucleosomes reconstituted with the centromeric H3 variant CENP-A ( 67 , 72 ). Interestingly, this variant is missing from T. brucei and how centromer e r egions ar e specified is currently unclear ( 99 ). Six of the T. brucei megachromosomes contain a 147 bp centromere-associated repeat sequence ( 86 ), which we show can be wrapped into NCPs (Supplementary Figure 7D). Investigating epigenetic inheritance and the role of chromatin in centromere-associated processes will be of interest in the future.
Howe v er, despite ov erall lower DNA binding of the T. brucei NCP, histones H2A and H2B in T. brucei were previously shown to be bound to DNA more persistently than H3 and H4 ( 100 ). This is consistent with our finding that H2A-H2B r einfor ce DNA binding at SHL3.5 via a large positi v e dipole ( Figure 5 ). The significance of tight DNA binding by H2A-H2B r equir es further investigation but we specula te tha t it may af fect a variety of chroma tin-based processes. For example, RN A pol ymer ase II tr anscription is known to be stalled at defined points corresponding to DNA-histone contacts while transcribing a nucleosome template (101)(102)(103)(104). Transcription in T. brucei is expected to be highly atypical ( 105 ) and future studies could re v eal whether the altered DNA binding properties of their nucleosomes affect this and other processes.
Recent work has shown the complexity of histone posttranslational marks in T. brucei ( 24 , 26 , 27 ) and T. cruzi ( 106 , 107 ) . Although most of the N-and C-terminal histone tails are not resolved in our structure, we could map 47 reported histone marks onto our unmodified structure ( 24 , 26 ) (Supplementary Figure S10A). Intriguingly, the kinetoplastid-specific H2A-Lys68, which alters the local environment of the acidic patch (Figure 6 B), has been reported to be trimethylated (Supplementary Figure  S10A). Concei vab ly, this and other modifications may act as switches to alter chromatin binding properties. We also note a cluster of phosphoryla tion / acetyla tion sites a t the N-terminal end of H2A (Supplementary Figure S10A) that could serve to further destabilise DNA binding at SHL 4.5 (Figure 4 E). Beyond histone post translational modifications, histone variants in T. brucei play a key role in modulating transcription (22)(23)(24) and controlling the parasite's variant surface gl ycoprotein imm une evasion system ( 19 ). Further work comparing the canonical structure of T. brucei nucleosomes to variants would be of great interest to help explain the biology behind antigenic variation and transcription initia tion / termina tion.
Of the four common interaction interfaces commonly coopted by chromatin-binding proteins ( 4 ), all display both sequence and structural variation in T. brucei . For example, the acidic patch in T. brucei has altered topology and is refractory to known binders ( Figure 6 ). Interestingly, the narrow topology of the patch accompanied by an insertion in H2A in T. brucei is inverse to the widening of the patch with an insertion in H2B in the nucleosome structure of the parasite G. lamblia ( 90 ) (Supplementary Figure S9A). The di v ergence of this interface suggests that trypanosome chromatin interactions may also have divergent properties. Since the acidic patch has been identified as a common interaction site in other species ( 4 ), we expect that other identified factors in T. brucei involved in regulating histone post-transla tional modifica tions ( 27 ), chroma tin remodelling ( 28 , 29 ), antigenic variation ( 30 ), or yet unidentified pathways may use this mechanism.
Howe v er, a mechanistic understanding for chromatin interactions in T. brucei is largely missing and our search for acidic-patch binders in T. brucei was challenging due to low conservation or poor annotation of potential homologs. Despite this, promising candidates could include T. brucei Dot1A and Dot1B, homologs of the Dot1 histone methyltr ansfer ase in higher eukaryotes (108)(109)(110). The catalytic fold of T. brucei Dot1A / B seems to be conserved ( 31 ) when modelled using AlphaFold2 ( 45 ) (Supplementary Figure S10B), but the loop that engages the acidic patch in human Dot1L ( 110 ) (Supplementary Figure S10C) differs in both sequence and predicted structure in T. brucei (Supplementary Figure S10D and E). This suggests that the local binding interactions to chromatin by T. brucei Dot1A / B may differ. It will be a fascinating avenue of study to probe the effects of di v ergence on chromatin recognition in kinetoplastid parasites.
Remar kab ly, our e xtensi v e phylogenetic analysis and modelling re v ealed that a majority of our findings are conserved within the kinetoplastids, including known pathogens (Supplementary Figures S1A, S8A, B, S9A, S8B). For example, conservation of the altered acidic patch includes clinically relevant targets from the Trypanosoma and Leishmania species (Supplementary Figure S9A, B) . This opens a possible therapeutic avenue for targeting the acidic patch ( 111 , 112 ) allowing specific targeting of kinetoplastids over their human or animal hosts. A global chromatin disruption mechanism would have high utility for combating diseases such as animal trypanosomiasis where the genetic di v ersity of Trypanosoma species has hindered drug de v elopment ( 17 ) and drug r esistance is a curr ent challenge ( 16 , 113 ).

DA T A A V AILABILITY
The cryo-EM density map and associated meta data for the T. brucei NCP have been deposited at the Electron Microscopy Data Bank under accession number EMD-16777. Raw micro gra phs have been uploaded to EMPIAR-11454. The atomic coordinates of the T. brucei NCP have been deposited in the Protein Data Bank under accession number 8COM. The MNase sequencing results have been deposited to the Gene Expression Omnibus NCBI database under accession number GSE226029.