Functional redundancy revealed by the deletion of the mimivirus GMC-oxidoreductase genes

Abstract The mimivirus 1.2 Mb genome was shown to be organized into a nucleocapsid-like genomic fiber encased in the nucleoid compartment inside the icosahedral capsid. The genomic fiber protein shell is composed of a mixture of two GMC-oxidoreductase paralogs, one of them being the main component of the glycosylated layer of fibrils at the surface of the virion. In this study, we determined the effect of the deletion of each of the corresponding genes on the genomic fiber and the layer of surface fibrils. First, we deleted the GMC-oxidoreductase, the most abundant in the genomic fiber, and determined its structure and composition in the mutant. As expected, it was composed of the second GMC-oxidoreductase and contained 5- and 6-start helices similar to the wild-type fiber. This result led us to propose a model explaining their coexistence. Then we deleted the GMC-oxidoreductase, the most abundant in the layer of fibrils, to analyze its protein composition in the mutant. Second, we showed that the fitness of single mutants and the double mutant were not decreased compared with the wild-type viruses under laboratory conditions. Third, we determined that deleting the GMC-oxidoreductase genes did not impact the glycosylation or the glycan composition of the layer of surface fibrils, despite modifying their protein composition. Because the glycosylation machinery and glycan composition of members of different clades are different, we expanded the analysis of the protein composition of the layer of fibrils to members of the B and C clades and showed that it was different among the three clades and even among isolates within the same clade. Taken together, the results obtained on two distinct central processes (genome packaging and virion coating) illustrate an unexpected functional redundancy in members of the family Mimiviridae, suggesting this may be the major evolutionary force behind their giant genomes.


Introduction
Mimivirus is the inaugural member of the family Mimiviridae , part of the Nucleocytoviricota phylum encompassing large and giant DNA viruses infecting eukaryotes (Koonin et al. 2020 ).Members of the family Mimiviridae infecting amoeba have dsDNA genomes up to 1.5 Mb encoding more than 1000 proteins, including a complete gl ycosylation mac hinery (Raoult et al. 2004, Aber gel et al. 2007, Abrahão et al. 2018, Notaro et al. 2021, 2022, Speciale et al. 2022 ).Mimivirus virion penetrate the cell through phagocytosis, and the acidic vacuole mediates opening of the stargate structure at one vertex of its icosahedral capsid (Zauberman et al. 2008, Sc hr ad et al. 2020 ).The internal membrane unwraps and fuses with the phagosome membrane, allowing transfer of the nucleoid compartment into the host cytoplasm, while empty capsids remain in the vacuole (Raoult et al. 2004, Zauberman et al. 2008, Clav erie and Aber gel 2010 ).The infectious cycle occurs in the cytoplasm where a large viral factory is developed (Raoult et al. 2004, Suzan-Monti et al. 2007, Claverie and Abergel 2010, Mutsafi et al. 2010 ).At the late stage of the cycle, neo-synthesized virions bud at the periphery of the viral factory where they are filled with the genome.Finall y, a gl ycosylated layer of fibrils composed of proteins and two large polysaccharides synthesized by the vir all y encoded machinery is added to the capsids (Kuznetsov et al. 2013, Notaro et al. 2021, de Aquino et al. 2023 ).As a result, the 750nm diameter virions resemble Russian dolls made of the external layer of reticulated glycosylated fibrils (referred to as the "layer of fibrils") decorating the surface of the icosahedral capsids.Underneath the capsid shell, the nucleoid compartment encases the 1.2 Mb dsDNA genome organized into a 30-nm large nucleocapsidlike structure (referred to as the "genomic fiber").The genomic fiber is made of a protein shell internally lined by the folded DNA and a central channel that can accommodate large proteins such as the viral RNA polymerase (Villalta et al. 2022 ).Three independent genomic fiber structur es hav e been determined by cryo-electr on micr oscop y (cry o-EM): tw o compact 5-and 6-start DNAcontaining helices, and a 5-start relaxed helix, without DNA (Villalta et al. 2022 ).Unexpectedl y, the pr otein shell was found to be composed of two glucose-methanol-c holine (GMC) oxidor eductases sharing 69% sequence identity, with a ratio of 5 between qu_946 and qu_143 according to the protein composition of the purified genomic fiber (Villalta et al. 2022 ).The resolution of the reconstructions (3.7 and 4.4 Å) pr e v ented us from determining whether each helix contained a single paralog or a mixture of both.Inter estingl y, one of the two GMC-oxidoreductases, qu_143 (R135 in mimivirus prototype), was known to compose the external layer of fibrils at the surface of mimivirus capsids and it was hypothesized that it was the major target for glycosylation (Klose et al. 1993, Bo y er et al. 2011, Notaro et al. 2021, Speciale et al. 2022 ).
Ho w e v er, despite its involvement in both the mimivirus genomic fiber inside the nucleoid, and the layer of fibrils at the periphery of the icosahedr al ca psid, GMC-oxidor eductase homologs ar e absent in the labor atory-e volv ed mimivirus M4 str ain (Bo y er et al. 2011 ).M4 also lacks the glycosylation machinery, described for members of different clades of the subfamily Megamimivirinae and responsible for synthesizing and br anc hing the pol ysacc harides on the ca psids (Notar o et al. 2021(Notar o et al. , 2022 ) ).In the present study, we used in-house de v eloped tools (Philippe et al. 2024 ) to delete mimivirus GMC-oxidoreductase genes.We then assessed the fitness cost associated with these deletions and investigated their impact on the formation of the genomic fiber and the protein and glycan composition of the layer of fibrils.Cry o-EM w as used to determine the structure of the KO_946 genomic fiber made of qu_143, the less abundant in the wild-type (wt) genomic fiber.Nuclear magnetic resonance (NMR) and gas chromatography mass spectrometry (GC-MS) were used to analyze the compositions in glycans and their structures for each mutant and to compare them with the wt layer of fibrils.We used mass spectrometry (MS)-based proteomics to analyze, for each of the three mutants, the protein composition of their layer of fibrils and extended the study to members belonging to B and C clades (moumouviruses and megaviruses) known to glycosylate their layer of fibrils with differ ent gl ycans using clade-specific gl ycosylation mac hineries (Notar o et al. 2022, Speciale et al. 2022 ).While confirming the non-essentiality of the two GMC-oxidor eductases, our r esults document the unexpected resilience of mimivirus to the deletion of these two genes through the use of alternative proteins to compensate their disruptions.

Cloning of DNA constructs
A detailed protocol for gene manipulation of giant viruses and their host is provided in Philippe et al. ( 2024 ).

Gene knock-out vectors
The plasmid for gene knock-out was generated by sequential cloning of the 3' UTR of mg_18 (megavirus chiliensis), the promoter of mg_741 (megavirus chiliensis) and a nourseothricin N-acetyl tr ansfer ase (NAT) or a neomycin (NEO) selection cassette (Fig. 1 A).Each cloning step was performed using the Phusion Taq pol ymer ase (ThermoFisher) and InFusion (Takar a).Finally, 500-bp homology arms were introduced at the 5' and 3' end of the cassette to induce homologous recombination with the viral DNA (Philippe et al. 2024 ).Befor e tr ansfection, plasmids were digested with EcoRI and NotI.All primers are shown in Fig. S1 .

Gene knock-out
Gene knoc k out str ategy was performed as pr e viousl y described for pandoravirus (Bisio et al. 2023 ).Briefly, 1.5 ×105 Acanthamoeba castellanii cells were transfected with 6 μg of linearized plasmid using Polyfect (QIAGEN) in phosphate saline buffer (PBS).One hour after tr ansfection, PBS was r eplaced with PPYG, and cells were infected with 1.5 × 10 7 mimivirus reunion particles for 1 h with sequential washes to r emov e extr acellular virions.Next, 24 h after infection, the new generation of viruses (P0) was collected and used to infect new cells.An aliquot of P0 viruses was utilized for genotyping to confirm the integration of the selection cassette.Primers used for genotyping are shown in Table S1 .A new infection was allowed to proceed for 1 h, then washed to r emov e extracellular virions and nourseothricin was added to the media.Vir al gr o wth w as allo w ed to pr oceed for 24 h.This pr ocedur e was repeated one more time before removing the nourseothricin selection to allow viruses to expand more rapidly.Once the viral infection was visible, the selection procedure was repeated one more time (Fig. 1 B).Viruses produced after this new round of selection were used for genotyping and cloning (Philippe et al. 2024 ).Double knoc k out of the GMC-oxidoreductases was obtaining by using a clonal population of qu_143 knoc k out viruses as par ental str ain.The locus of qu_946 was r e placed by a neom ycin resistance cassette .T he tr ansfection and selection of r ecombinant viruses' pr ocedur e was performed identical to the process to generate single knockout, but replacing nourseothricin by geneticin.

Cloning and genotyping
Next, 150 000 A. castellanii cells were seeded on 6-well plates with 2 mL of PPYG.After adhesion, viruses were added to the well at a multiplicity of infection (MOI) of 1. One-hour post-infection, the w ell w as w ashed five times with 1 mL of PPYG, and cells wer e r ecov er ed by well scr a ping.Amoebas wer e then diluted until obtaining a suspension of 1 amoeba/ μL.Then 1 μL of such suspension was added in each well of a 96-well plate containing 1000 uninfected A. castellanii cells and 200 μL of PPYG.Wells wer e later monitor ed for cell death and 100 μL collected for genotyping (Philippe et al. 2024 ).Genotyping was performed using Terr a PCR Dir ect Pol ymer ase Mix (Takar a) following the manufacturer's specifications.Primers used for genotyping are shown in Table S1 .

Competition assay and quantitative PCR analysis
Equal infectious particles of wild-type and recombinant mimivirus reunion were mixed and used to infect A. castellanii at an a ppr oximate MOI of 0.8.Viruses were allowed to grow overnight in the presence or absence of nourseothricin.Subsequent vir al pr ogenies wer e used to infect ne w A. castellanii cells in r eiter ativ e passa ges.A fr action of eac h passa ge was collected for genomic DNA extraction.
Vir al genomes wer e purified using Wizard genomic DNA purification kit (PROMEGA).To determine the amplification kinetic, the fluorescence of the EvaGreen dye incorporated into the PCR product was measured at the end of each cycle using SoFast Eva-Green Supermix 2 × kit (Bio-Rad, F rance).A standar d curve using the gDNA of purified viruses was performed in parallel with each experiment.For each point, a technical triplicate was performed.Quantitativ e r eal-time PCR (qRT-PCR) anal yses wer e performed on a CFX96 Real-Time System (Bio-Rad) (Fig. 1 C).S1 .(B) Gr a phic depicting the strategy for the selection of recombinant viruses.Viral infection was performed 1-h post-transfection; Ntc: Nourseothricin; P, passage.(C) Growth competition assays revealed no significant defects in the lytic cycle of deletion strains .T he competition was also performed in the presence of Nourseothricin, which allows the outcompetition of the recombinant strains due to the expression of a Nourseothricin selection cassette.Measurements were performed by qPCR of an endogenous locus (present in wt and recombinant strains) and the Nourseothricin selection cassette (only present in recombinant viruses).

Genome sequencing and assembly of mutants' genomes
Genomic DN A w as extr acted fr om 10 10 virus particles using Pur e-Link TM Genomic DNA mini kit (Invitrogen) according to the manufactur er's pr otocol.Clones of individual m utants and wt wer e sequenced on an illumina platform (Novogen To confirm the deletion of eac h m utant, the r eads wer e ma pped on the wt genome resulting in homogeneous cov er a ge along the genome, except for the qu_143 and qu_946 centr al positions, whic h ar e cov er ed.In addition, we used the central part of the GMC-oxidoreductase genes (deleted in mutants) as blast queries against the reads of each m utant genome, whic h also confirmed the absence of the central region.

Extraction and purification of the qu_946 and qu_143 mutants' genomic fiber
The genomic fiber of the mimivirus reunion single mutants of qu_946 (KO_qu946) and qu_143 (KO_qu143) were extracted as described in Villalta et al. ( 2022 ) for the wt virus .T he genomic fiber was extracted from 12 mL of purified single deletion mutant virions at 2 × 10 10 particles/mL, split into 12 ×1 mL samples processed in parallel.Trypsin (Sigma T8003) in 40 mM Tris-HCl pH 7.5 buffer was added at a final concentration of 50 μg/mL and the virus-enzyme mix was incubated for 2 h at 30 • C in a heating dry block (Grant Bio PCH-1).DTT was then added at a final concentration of 10 mM and incubated at 30 • C for 16 h.Finally, 0.001% Triton X-100 was added to the mix and incubated for 4 h at 30 • C. Each tube was centrifuged at 4000 x g for 5 min to pellet the opened capsids .T he supernatant was recovered and concentrated by centrifugation at 15 000 x g for 4 h at 4 • C. Most of the supernatant was discarded, leaving 12x ∼200 μL of concentr ated br oken pieces of genomic fiber that wer e pooled and layer ed on top of ultracentrifuge tubes of 4 mL (polypropylene centrifuge tubes, Beckman Coulter) containing a discontinuous cesium chloride gradient (1.4,1.3, 1.2 g/cm 3 in 40 mM Tris-HCl pH 7.5 buffer).The gradients were centrifuged at 200 000 x g for 16 h at 4 • C. Because no visible band was observ ed, successiv e 0.5 mL fractions wer e r ecov er ed fr om the bottom of the tube.Each fraction was dialyzed using 20 kDa Slide-A-Lyzers (ThermoFisher) against 40 mM Tris-HCl pH 7.5 to r emov e the CsCl.These fr actions wer e further concentrated by centrifugation at 15 000 x g, at 4 • C for 4 h, and most of the supernatant was r emov ed, leaving ∼100 μL of sample at the bottom of each tube.At each step of the extraction procedure, the sample was imaged by negative staining transmission electr on micr oscopy (NS-TEM) to assess the integrity of the genomic fiber.Each fraction of the gradient was finally controlled by NS-TEM.

Negati v e staining TEM
The 300-mesh ultra-thin carbon-coated copper grids (Electron Microscop y Sciences) w ere prepared for negative staining by adsorbing 4-7 μL of the sample for 3 min, follo w ed b y blotting excess liquid and staining for 2 min in 2% ur an yl acetate to image the genomic fiber.For fibrils and mutant virions, staining was performed with a drop of 1% ur an yl follo w ed b y blotting after 10-15 s, and a dr op of ur an yl acetate coupled with methylcellulose (2% and 0.2%, r espectiv el y) was added twice and left for 5-10 s before blotting.
The grids wer e ima ged either on an FEI Tecnai G2 microscope operated at 200 keV and equipped with an Olympus Veleta 2k camer a (IBDM micr oscopy platform, Marseille, France); an FEI Tecnai G2 microscope operated at 200 keV and equipped with a Gatan OneVie w camer a (IMM, micr oscop y platform, F rance) (Fig. 2 ).

Sample prepar a tion
F or the K O_qu946 mutant, 3 μL of the purified sample was applied to glow-disc har ged Quantifoil R 2/1 Cu grids, blotted for 2 s using a Vitrobot Mk IV (Thermo Scientific) and a ppl ying the following parameters: 4 • C, 100% humidity and blotting force 0 then plunged frozen in liquid ethane/propane cooled to liquid nitrogen temper atur e.

Data acquisition
Grids wer e ima ged using a Titan Krios (Thermo Scientific) microscope operated at 300 keV and equipped with a K3 direct electron detector and a GIF BioQuantum energy filter (Gatan).Next, 2224 movie frames were collected using EPU software (Thermo Scientific) at a nominal magnification of 81 000x with a pixel size of 1.0859 Å and a defocus range of −0.6 to −2.8 μm.Micr ogr a phs wer e acquir ed using EPU (Thermo Scientific) with 2.3 s exposur e time, fractionated into 40 frames and 18.25 e − /pixel/s (total fluence of 35.597 e − / Å 2 ).

2D classification and clustering of 2D classes
All movie frames were aligned using MotionCor2 (Zheng et al. 2017 ) and used for contrast transfer function (CTF) estimation with CTFFIND-4.1 (Rohou and Grigorieff 2015 ).Helical segments of the purified genomic fiber, manually picked with Relion 3.1.0,wer e extr acted with 400-pixel box sizes (decimated to 100 pixels) using a rise of 7.93 Å and a tube diameter of 300 Å. Particles were subjected to reference-free 2D classification in Relion 3.1.0(Sc her es 2012 , He and Sc her es 2017 ).We then performed additional cluster analysis of the initial 2D classes provided by Relion to aim for more homogeneous clusters (Villalta et al. 2022 ), e v entuall y corr esponding to differ ent states (Figs 3 and 4 ).

Identification of candidate helical parameters
Fourier tr ansform anal ysis methods wer e used to confirm the helical parameters were the same as in wt (Diaz et al. 2010, Sachse 2015, Coudray et al. 2016, Villalta et al. 2022 ) for the Cl1 (Cl1a in wt) and Cl2.

Cryo-EM data processing and 3D reconstruction
After determining the helical par ameters, segments wer e extracted with a box size of 400 pixels (decimated to 100 pixels) using the proper rises for the Cl1 (7.93 Å, cylinder 300 Å, 442 237 segments) and the Cl2 (20.47 Å, cylinder 330 Å, 172 431 segments).A dedicated 2D classification protocol was performed independentl y on eac h extr action.For the Cl2, one r ound of 50 expectation-maximization (E-M) iterations was sufficient to produce 133 homogeneous 2D classes submitted to cluster analysis and 44 2D classes were selected (21 186 segments).For the Cl1 extr action, thr ee iter ativ e 2D classification/selection rounds were performed (25, 50 and 100 E-M iterations) producing 61 classes, fr om whic h 30 (149 593 segments) wer e finall y selected based on the cluster analysis.
Values of the helical parameters, rise and twist (Cl1: 7.9475 Å, 138.921 • ; Cl2: 20.48 Å, 49.43 • ), were then used for Relion 3D classification (Sc her es 2012 , He and Sc her es 2017 ), with a + / −0.5 units fr eedom searc h r ange, using a featur eless cylinder as initial r eference (diameter of 300 Å and C1 symmetry for Cl1 and 330 Å and C3 symmetry for Cl2).In real space, the helical symmetry w as sear ched using 50% of the central part of the box for both structures .T he number of asymmetrical units in each segment box was set to one for Cl1 and to six for Cl2.The entire helical reconstruction and protein shell dimensions were obtained using an in-house de v eloped pr ogr am.
The superimposable 3D classes (same helical parameters, same helix orientation) were then selected, reducing the dataset to 98 882 segments for the 5-start helix (Cl1) and 20 899 segments for the 6-start helix (Cl2).After r e-extr action of the selected segments without scaling, a first unmasked 3D refinement was performed with a rescaled 3D classification output low pass filtered to 15 Å as r efer ence, follo w ed b y a 3D r efinement step using solv ent flattened, Fourier shell correlation (FSCs) and CTF refinement using the standard pr ocedur e described in Relion.To further impr ov e the resolution of the maps , Ba yesian polishing was applied using 10 000 segments for training and default values for polishing.A last round of 3D refinement with solvent flattening was applied to the pr e vious r efined ma p using the polished particles.At that point the maps were sufficiently resolved (Cl1: 4.3 Å, Cl2: 4.2; FSC threshold 0.5) to identify secondary structure elements (with visible cylinders corresponding to the helices) and were used to compute local resolution.

Structures refinement
The best resolution Cl2 map was used to fit the qu_143 dimeric structure (PDB 7YX3) using UCSF ChimeraX 1.5 (Pettersen et al. 2004 ).Each monomer was then rigid-body fitted independently into the map.The entire protein was then inspected within coot 0.9.7 (Emsley et al. 2010 ) to fix local inconsistencies and was further r efined a gainst the ma p using the r eal-space r efinement pr ogram in PHENIX 1.20.1 (Liebschner et al. 2019 ).The protein was  F igure 4. Workflo w of the 5-and 6-start helices reconstruction processes.Segment extraction was performed with a box size of 400 pixels (pix) binned (box size 100 pix, 4.3436 Å/pix).The distance between consecutive boxes was equal to the axial rise calculated by indexation of the power spectrum.After clustering, 2D classes were selected for Cl1 and Cl2 and 3D classification was carried out using the selected segments, helical symmetry par ameters fr om the po w er spectrum indexation and a 300 Å or 330 Å featureless c ylinder as 3D r efer ence for Cl1 and Cl2, r espectiv el y. 3D r efinement of the two boxed 3D classes was ac hie v ed using one low pass filtered 3D class as reference on the unbinned segments.A first 3D refinement was performed with solvent flattening follo w ed b y CTF r efinement and polishing of the selected segments.A last 3D r efinement was ac hie v ed with solv ent flattening.The EM maps colored by local resolution from 5 Å (blue) to 3 Å (red) with Euler angle distribution of the particles used for the 3D r econstruction ar e pr esented.
submitted to five cycles of rigid body refinement (with each chain defined) follo w ed b y twice 10 c ycles of refinement with the default PHENIX options .T he r esulting structur e was manuall y corrected within coot.The resulting protein model was submitted to the same steps refinement in PHENIX.This final model was then fitted into the Cl1 map, inspected with coot and refined using five cycles of rigid body refinement and simulated annealing, follo w ed b y twice 10 cycles of refinement with the default PHENIX options.Validations were performed with PHENIX using the compr ehensiv e v alidation pr ogr am ( Table S3 ).RMSDs between different structures (monomers and dimers, Table S4 ) was computed using the align pr ocedur e in Pymol suite (The PyMOL Molecular Gr a phics System, Version 3.0 Schrödinger, LLC. as recommended by the authors: https:// pymol.org/support.html ) ( PyMOL ).

Extraction and purification of the mutants' fibrils
To analyze the glycan composition and polysaccharides structures of the three mutants and wt mimivirus reunion strain, we applied an already described protocol (Notaro et al. 2021 ).Briefly, 4 × 10 11 viral particles were centrifuged at 14 000 g for 10 min, the supernatant w as discar ded and the pellet was re-suspended in 10 ml of 50 mM DTT. Fibril extraction was performed at 100 • C under stirring.After 2 h, the tube was centrifuged at 14 000 g for 15 min, at 4 • C, and the fibrils r ecov er ed with the supernatant.The fibrils were then dried and purified on Biogel P10, follo w ed b y subsequent NMR analysis of each mutant and wt.
We also de v eloped a softer defibrillation pr otocol to r ecov er the fibrils without contaminating them with pr oteins r eleased by damaged virions in order to analyze the fibrils' protein composition by MS-based proteomics.Purified virions (1.5 × 10 10 ) were incubated in Eppendorf tubes in 500-μL 40 mM Tris-HCl pH 7.5 buffer, 500 mM DTT for 2 h at 30 • C. Tubes were then centrifuged at 14 000 g for 10 min.The supernatants containing the fibrils wer e r ecov er ed and concentr ated on viv aspin ® 3 KDa (Sartorius, VS04T91) at 3000 g.The pellet w as w ashed twice with 40 mM Tris-HCl pH 7.5 buffer and centrifuged at 14 000 g for 10 min and finally resuspended in the same buffer.Intact virions, pellets and fibrils wer e ima ged by NS-TEM to assess the integrity of the defibrillated virions in the pellet and the presence of fibrils in the supernatant.For the Nqu143-GFP mutants, defibrillated virions were also observ ed by fluor escence micr oscopy, whic h confirmed the absence of fluorescence due to the removal of the GFP together with the layer of fibrils.

Sugar composition of viral particles of Mimivirus reunion wt and mutants
Monosaccharide composition analyses as acetylated methyl glycoside (AMG) were performed on the intact viral particles (1.25 × 10 10 , ∼250 μl) of mimivirus reunion wt and mutants, following the pr ocedur e r eported by De Castr o et al. ( 2010).The obtained AMG was analyzed via gas c hr omatogr a phy-mass spectr ometry (GC-MS) on an Agilent instrument (GC instrument Agilent 6850 coupled to MS Agilent 5973) equipped with a SPB-5 capillary column (Supelco, 30 m × 0.25 i.d., flow rate, 0.8 mL min -1 ) and He as the carrier gas .T he identification of the monosaccharides derivatized as AMG was obtained by studying the fr a gmentation pattern corresponding to each peak of the chromatogram and by comparison with suitable standards.

Purification and 1 H NMR analysis of the fibrils
The fibrils of mimivirus reunion wt and m utants, extr acted as r eported abo ve , were purified to remove the DTT used for the extr action pr ocedur e.
Briefly, the gl ycopr oteins (pr otein/s carrying the pol ysacc harides) were precipitated with cold acetone at 80%, at 4 • C, for 16 h, twice .T he supernatant containing DTT and salts was discarded, while the precipitate was dissolved in water and freeze-dried.Then the precipitate was purified by size exclusion c hr omatogr a phy (Biogel P10, flow: 12 ml/h) to completel y r emov e the DTT.The eluted fractions were checked by 1 H NMR, revealing that the glycan-containing material was eluted at one-third of the column volume (full spectra are shown in Fig. 5 B).
The 1 H NMR measur ements wer e carried out on a 600-MHz Bruker instrument, equipped with a CryoProbe™ at 310 K.The intensity of the solvent signal was reduced by measuring a monodimensional DOSY spectrum, setting δ and to 2.4 and 100 ms, respectiv el y, and the v ariable gr adient to 50% of its maximum po w er.Spectr a wer e pr ocessed and anal yzed using the Bruker TopSpin 4.0.9pr ogr am.

Mass spectrometry-based proteomic analyses
Pr oteins wer e solubilized with Laemmli buffer (four volumes of sample with one volume of Laemmli 5X-125 mM Tris-HCl pH 6.8, 10% SDS, 20% gl ycer ol, 25% β-merca ptoethanol and tr aces of br omophenol blue) and heated for 10 min at 95 • C. The extracted proteins were stacked in the top of an SDS-PAGE gel (4-12% NuPAGE, Life Technologies), stained with Coomassie blue R-250 (Bio-Rad) before in-gel digestion using modified trypsin (Promega, sequencing grade) as previously described (Casabona et al. 2013 ).The resulting peptides were analyzed by online nanoliquid chromatogr a phy coupled to tandem MS (UltiMate 3000 RSLCnano and Q-Exactive Plus or Q-Exactive HF, Thermo Scientific).Peptides were sampled on a 300 μm x 5 mm PepMap C18 precolumn and separated on a 75 μm x 250 mm C18 column (Reprosil-Pur 120 C18-AQ, 1.9 μm, Dr. Maisch, except for KO_qu143, KO_qu946 and Nqu143-GFP m utant samples separ ated on Aur or a, 1.7 μm, IonOpticks) using a 140-min gradient (except for fibrils from Nqu143-GFP m utant, for whic h a 60-min gradient was used).MS and MS/MS data were acquired using Xcalibur 4.0 (Thermo Scientific).Peptides and proteins were identified using Mascot (version 2.8.0, Matrix Science) through concomitant searches against a homemade A. castellanii protein sequence database, homemade virusspecific protein sequence databases, and a homemade database containing the sequences of classical contaminant proteins found in proteomic analyses (human keratins, trypsin…).Trypsin/P was chosen as the enzyme and two missed cleavages were allo w ed.Precursor and fragment mass error tolerances were set at 10 and 20 ppm, r espectiv el y.P eptide modifications allo w ed during the search were: Carbamidomethyl (C, fixed), Acetyl (Protein Nterm, variable) and Oxidation (M, variable).Proline software version 2.2.0 (Bouyssié et al. 2020 ) was used for the compilation, grouping and filtering of the r esults: conserv ation of r ank 1 peptides, peptide length ≥ 6 amino acids, peptide-spectrum-match identification false discovery rate < 1% (Couté et al. 2020 ) and a minimum of one specific peptide per identified protein group.Proline was then used to perform a MS1-based quantification of the identified protein groups.Intensity-based absolute quantification (iB AQ) (Schw anhäusser et al. 2011 ) values w ere calculated for each protein group from the MS intensities of razor and specific peptides ( Table S2 ).The r elativ e abundance of individual proteins in virions and fibrils was calculated as the ratio of the individual pr otein iBAQ v alues to the sum of the iBAQ values of all proteins in each sample .T he relative enrichment of individual proteins between virions and fibrils was calculated as the ratio of their relative abundances in each fraction (Villalta et al. 2022 ).

Neither of the two GMC-oxidoreductases is essential
We used our r ecentl y de v eloped pr otocol (Philippe et al. 2024 ) combining homologous recombination with the introduction of a NAT selection cassette to delete each of the two genes encoding the GMC-oxidoreductases (qu_946 and qu_143).We selected recombinant viruses that were cloned to obtain homogeneous populations (Fig. 1 B) (Philippe et al. 2024 ).Eac h m utant was easil y produced and genotyped to confirm the mutation ( Fig. S1 ).Using a second resistance gene (NEO selection cassette) we were able to delete both genes ( Fig. S1 ), demonstrating that the two GMC-oxidor eductases wer e not essential.The absence of additional mutations in every mutant was confirmed by genome sequencing.

Mutants' fitness
To assess whether a fitness cost was associated with the mutations, we performed competition assays against wt mimivirus reunion strain by measuring the abundance of eac h m utant ov er se v er al cycles in the presence and absence of selection.In contrast to the wt, each mutant presents a common nourseothricin resistance gene .T he double deletion mutant (2KO) encodes for an additional geneticin resistance gene.As a result, in the presence of nourseothricin, each mutant ratio increased, with the disappearance of the wt virus after five passages (Fig. 1 C).These data allo w ed competition assays to be validated as an effective tool to assess the fitness of recombinant viruses.In the absence of selection, the r elativ e abundance of the m utants compar ed with wt r emained ∼0.5 ov er fiv e passa ges, supporting the absence of a fitness cost, e v en for the double deletion m utant (Fig. 1 c).

Composition of the genomic fiber of single mutants
To determine the composition of the genomic fiber of each single m utant we extr acted and purified their genomic fiber.MS-based proteomics confirmed that the most abundant protein was the remaining GMC-oxidor eductase.NS-TEM highlighted sur prising differences between the two structures (Fig. 2 ), despite the fact that the two proteins (qu_946 and qu_143) share 69% sequence identity (81% similarity).Specifically, the genomic fiber extracted from the KO_qu143 mutant and made of qu_946 (the most abundant in the wild type genomic fiber) was mostly in the unwound state (Fig. 2 B), while the one made of qu_143 (KO_qu946) resulted in very long and stable helices that did not unwind (Fig. 2 c).T hus , the use of both GMC-oxidoreductases in the wt genomic fiber could contribute to a fine tuning of its biophysical properties, with an intermediate state between the wt and each deletion mutant (Fig. 2 A).
In the case of the double m utant, the pr otocols for ca psid opening to extract the genomic fiber of wt or single mutants did not work pr operl y.An optimized pr otocol allo w ed the extraction of a possible thinner genomic fiber, but in poor yield, pr e v enting its purification and compositional c har acterization (Fig. 2 D).

Cryo-EM single particle analysis of the qu_143 genomic fiber
To determine the contribution of eac h GMC-oxidor eductase to the wt genomic fiber we performed cryo-EM single-particle analysis on the most stable genomic fiber composed by qu_143 (mutant KO_qu946).As for the wt, the 2D classification r e v ealed an heterogeneity of the sample, and the 2D classes were sorted by a ppl ying our already described clustering protocol (Villalta et al. 2022 ).
The two main clusters corresponding in width to the compact (Cl1a) and (Cl2) structures of the wt genomic fiber were respectiv el y named Cl1 and Cl2 (Fig. 3 ).For each main cluster, we confirmed that the helical symmetry par ameters wer e the same as the wt genomic fiber and proceeded to structure determination and refinement (Fig. 4 ).For the less populated smaller cluster (Cl0, ∼25 nm), absent from the wt genomic fiber 2D classes, we failed to identify its helical parameters due to the lo w er number of segments and the resulting lo w er r esolution.After 3D r efinement, we obtained a 4.3 Å r esolution helical ∼29 nm diameter structure (FSC threshold 0.5, masked) for Cl1.This structure corresponds to the same 5-start left-handed helix as the wt (Cl1a), made of an ∼8-nm thic k pr oteinaceous external shell with five dsDNA strands lining the interior of the shell and an ∼9-nm wide central channel (Fig. 4 , Fig. S2 ).The 4.2 Å resolution Cl2 map obtained after 3D refinement (Fig. 4 , Fig. S2 ) corresponds to the same ∼32-nm diameter 6-start left-handed helix as the wt, with six dsDNA strands lining the external shell and an ∼12-nm wide inner channel (Fig. 4 ).
Because the helical parameters between the wt Cl1a and the mutant Cl1 are the same, we used the qu_143 dimeric structure refined in the Cl1a focus r efined ma p for r efinement into the mutant maps (Materials and Methods section and Table S3 ).The den-sity that can be attributed to the FAD cofactor was present in both maps ( Fig. S3 ) and the models of Cl1 and Cl2 dimers are superimposable with a core RMSD of 0.467 Å based on C α atoms ( Table S4 ).As the qu_143 helices are more stable than the wt (composed of both GMC-oxidoreductases), the relaxed Cl3 cluster was never observed with this mutant.

Model explaining the co-existence of 5-and 6-start helices
As for the wt, the genomic fiber of KO_qu946 is composed of a mixture of five and six strands of DNA, despite the presence of a single GMC-oxidoreductase in the shell.We estimated the ratio of 5start and 6-start from the clustering (Fig. 3 ) and can now propose a model that reconciles the co-occurrence of the two structures.In this model, the whole genome would be folded into six parallel str ands, fiv e longer than the sixth one .T he helix would then be formed initially as a 6-start helix until the sixth strand ends and, from that point, becomes a 5-start helix (Fig. 6 ).According to this model, assuming the length of the genomic fiber is limited by the size of mimivirus nucleoid compartment, we can estimate that the maximum genome length would be ∼1.4Mb for a full 6start helix with an 8-nm thick protein shell.We hypothesize that the last cluster (Cl0, Fig. 3 ) could correspond to a 4-start with an additional shorter DNA strand.

The N-terminal cys-pro-rich domain is an addressing domain to the virion surface
The sequence of the cys-pr o-ric h N-ter domain of the GMCoxidoreductases is not covered by proteomic analysis of the purified genomic fiber but is cov er ed by peptides in the purified fibrils forming the external layer at the surface of the capsids (Villalta et al. 2022 ).To assess whether this cys-pr o-ric h N-ter domain could be a structural signature used to address proteins to the layer of fibrils, we replaced the second GMC-oxidoreductase (qu_143) in the genome of the mutant KO_qu946 by the sequence of the GFP in fusion with the sequence of the qu_143 N-terminal domain (Nqu143-GFP).We then analyzed the resulting virions by MS-based proteomics (Table 1 and Table S2 ) and fluorescence microscop y.Purified Nqu143-GFP virions sho w ed a str ong fluor escence supporting the incor por ation of the c himeric pr otein into the viral particles (Fig. 7 ).Moreover, after defibrillation of the virions, the GFP fluorescence was lost.In addition, Nqu143-GFP was identified in purified virions (ranked 97th in terms of relative abundance, Table S2 ) and was found enriched 12 times in the fraction containing the external fibrils (ranked 20th, Table 1 and Table S2 ) with peptides covering the N-terminal domain ( Fig. S5 ).Taken together, these data indicate that the N-terminal cys-prorich domain of the GMC oxidoreductase is sufficient to direct the proteins to the layer of fibrils at the surface of mimivirus particles.is conserved in all members of the A clade, it is absent in the orthologous proteins in members of the B and C clades .Moreo ver, homologs of the two GMC-oxidoreductases are pseudogenized in tupan viruses , suggesting that differ ent pr oteins should compose their fibrils ( Fig. S6 ).We thus conducted a systematic proteomic analysis of the purified fibrils of different members of the three clades, in addition to the mutants.For the double mutant (2KO) we identified a group of proteins as the most abundant in the fibrils fraction (qu_734, qu_757, qu_384 and qu_482, Table 1 and Table S2 ).These proteins were also highly ranked in the Nqu143-GFP double mutant fibrils and the orthologous protein of qu_734 was identified as the most abundant protein in the laboratorye volv ed M4 m utant fibrils (696-L688, T able 1 and T able S2 ).This mutant does not encode a glycosylation machinery and is not glycosylated (Bo y er et al. 2011, Notaro et al. 2021 ), thus its capsid lac ks the lar ge r eticulated layer of fibrils decor ating the mimivirus ca psid ( Fig. S8f).Inter estingl y, the qu_734 pr otein also possesses a cys-pr o-ric h N-ter domain that is conserved in all clades ( Fig. S6 ).

Protein composition of the layer of fibrils in additional members of the family Mimiviridae
It was not possible to determine if one of these proteins was also the building block composing the 2KO genomic fiber, given the difficulty in opening the 2KO capsids .T hus we concluded that the c hange in pr otein composition of the fibrils led to capsids with different stability properties .While , as for the wt, the most abundant protein in KO_qu946 remains qu_143, the lack of qu143 in KO_qu143 does not lead to its replacement by the second GMCoxidoreductase.Instead, it is replaced by a group of proteins, with qu_465 (predicted as a thioredoxin domain containing protein) as the most abundant, follo w ed b y qu_757, the second most abundant in the wt fibrils (Table 1 and Table S2 ).
The pr oteomic anal ysis of fibrils purified from two isolates of the B clade, moumouvirus australiensis and maliensis, sho w ed similar protein compositions, but with slight differences in their r elativ e abundances (Table 1 and Table S2 ).The moumouvirus maliensis protein mm_751, ranked first in its fibrils, is absent from A clade fibrils and is in the top three in the fibrils of members of the B and C clades.Cystein and proline amino acids are present in the N-ter domain of mm_751, but less abundant than in the GMCoxidoreductases.For ma_195 (qu_738 in mimivirus reunion, Table 1 and Table S2 ), ranked first in the fibrils of moumouvirus australiensis, the N-ter domain is not cys-pro-rich.The best ranked in the fibrils of members of the C clade are the same as for members of the B clade .T he first ranked in megavirus chilensis is mg749 (qu_657 in mimivirus r eunion), whic h pr esents a cys-pr o-ric h Nter domain, while the first ranked in megavirus vitis, mvi_646 (qu_600 in mimivirus r eunion), lac ks a cys-pr o-ric h N-ter domain.
Yet, for each virus in each clade, there is at least one protein with a cys-pr o-ric h N-terminal domain among the most abundant in the fibrils.

Glycan composition of the fibrils in mimivirus mutants
We r ecentl y established that the fibrils wer e coated with gl ycans (Notaro et al. 2022, Speciale et al. 2022 ) and demonstrated that for mimivirus, the prototype of the A clade, they were made of two distinct large molecular weight pol ysacc harides (Notar o et al. 2021 ).Because the cluster of 12 genes responsible for the biosynthesis of the pol ysacc harides is conserv ed in the A clade (Notar o et al. 2022 ), we hypothesized that the fibrils of the mimivirus reunion str ain would hav e the same composition as mimivirus.To assess whether the knoc k out of the most abundant protein composing the fibrils could also affect the br anc hing of these pol ysacc harides and their composition, we analyzed the sugar content of the viral particles of all three mimivirus reunion mutants (KO_qu946, KO_qu143 and 2KO) together with the wt strain (as reported in Notaro et al. 2021Notaro et al. , 2022 ) ).The chemical characterization revealed for each of them the presence of sugars (Fig. 5 A), with rhamnose (Rha), viosamine (V io), 2-OMe-V io, glucose (Glc) and glucosamine (GlcN) (Fig. 5A ), confirming that mimivirus reunion strain and all mutants had the same glycan composition as mimivirus prototype (Notaro et al. 2022 ).We then analyzed the fibrils of the mutants and wt by 1 H NMR spectroscopy (Speciale et al. 2022 ) and compared them with the mimivirus prototype, which confirmed that the different sugars were assembled the same way as in the r efer ence mimivirus to produce the same two polysaccharides (Fig. 5 , Fig. S7 ) (Notaro et al. 2021 ).NS-TEM images of the virions of the different mutants were obtained after methylcellulose staining to assess whether the m utations c hanged the layer of fibrils a ppear ance, whic h would suggest that the change in protein im-pacted their le v el of gl ycosylation (Hac ker et al. 2016 ).All m utants' virions sho w ed the same cross-linked outer lay er, suggesting that despite their differences in protein composition, their glycosylation was not affected ( Fig. S8 ).
It was hypothesized that mimivirus R135 GMC oxidoreductase (qu_143 in mimivirus reunion) was the major target for glycosylation in the surface fibrils of mimivirus (Bo y er et al. 2011, Notaro et al. 2021 ).The results presented here indicate that the mutations, including the deletion of the two GMC-oxidoreductase genes, did not affect the surface glycosylation.These data suggest that the GMC-oxidor eductases ar e not the main target for glycosylation.Alternativ el y, it is also possible that the glycosylation machinery was able to use the qu_734 protein as support for the two pol ysacc harides.As expected, no glycans were identified in the labor atory-e volv ed M4 mutant layer of fibrils.

Discussion
Genome pac ka ging is essential for the pr opa gation of viruses.For instance, the pac ka ging ATP ase of poxviruses (Cassetti et al. 1998 ) or the histones of marseilleviruses (Liu et al. 2021, Valencia-Sánchez et al. 2021 ) are essential for the productive infection by their virions.In addition, proteins at the surface of vir al ca psids usually play a central role in interacting with the host cell surface and the initial steps of infection (Sobhy 2017 ).The two processes, genome encapsidation and viral entry, are thus essential.As GMC-oxidoreductases compose both the genomic fiber and the surface fibrils (Villalta et al. 2022 ), we could have expected them to be essential.Ho w e v er, these two enzymes are absent in the labor atory-e volv ed mimivirus M4 str ain that also lost the gl ycosylation machinery (Bo y er et al. 2011 ).Our recent implementation of genetics tools for cytoplasmic giant viruses provided the first opportunity to dir ectl y assess the impact of the GMCoxidoreductases deletions in mimivirus r eunion str ain (Philippe et al. 2024 ).Given the homology between the two enzymes, the deletion of the most abundant one in the wt genomic fiber induced its replacement by the second one, with no apparent cost for the virus, under laboratory conditions.Yet, the single mutant genomic fibers were significantly different compared with the wt (Fig. 2 ), indicating a certain degree of functional specialization for each GMC-oxidoreductase.While the wt genomic fiber was highl y heter ogeneous, with full y r elaxed helices having lost DNA (Villalta et al. 2022 ), the KO_qu946 genomic fiber a ppear ed more stable, with no occurrence of fully relaxed structures.By contrast, the deletion of qu_143 seemed to increase the genomic fiber instability, suggesting a stabilizing role for this second GMCoxidoreductase (Fig. 2 ).The cryo-EM analysis of the KO_qu946 genomic fiber structure, as for the wt, produced the same two 5and 6-start helices, but composed of a single GMC-oxidoreductase.This led us to propose a model to explain the co-existence of both structures.In this model, the genome is folded into six strands befor e assembl y into a 6-start helix.The first and the last strand can be shorter than the other strands, leading to a 5-start helix with five DNA strands and a 4-start helix with four DNA strands.Finally, the deletion of both GMC-oxidoreductases with no significant loss of viral fitness demonstrated that they were not essential.
The qu_143 GMC oxidoreductase is five times less abundant than qu_946 in the genomic fiber but it is ranked first in the layer of fibrils and is three times more abundant than the next one, qu_757.It was pr e viousl y pr edicted as the main target for glycosylation (Klose et al. 1993 S2 ).MS-based pro-teomics r e v ealed that the cys-pr o-ric h N-ter domain of the GMCoxidor eductase is pr esent in the fibrils and cleaved in the genomic fiber (Villalta et al. 2022 ), but the protease and precise cleavage site remains to be identified.Interestingly, an additional mutant, in whic h this cys-pr o-ric h N-ter domain was fused to GFP, produced virions in which the GFP was identified at the surface of the virions and is enriched in the surface layer of fibrils.Further studies will be needed to elucidate the mechanism behind the recognition of this domain and its addressing to the fibrils.In addition, the virions of all mutants were decorated by long, reticulated fibrils, as evidenced by TEM ( Fig. S8 ).Except for the KO_qu946 mutant, MS-based proteomic analysis of the fibrils r e v ealed that they were composed of other proteins, some of which with a cys-prorich N-ter domain ( Fig. S6 ).Finally, the glycan analysis of the GMCoxidoreductases single and double mutants confirmed that their fibrils were still glycosylated by the same two pol ysacc harides as the wt virions.
In a pr e vious study, it was r eported that the glycosylation machinery was clade-specific and produced different glycans (Notaro et al. 2022 ).In the present study, the analysis of the purified fibrils of members of the B and C clades r e v ealed they were also composed of different proteins, inter-and even intra-clade, and that B and C clade virions' fibrils present a protein composition closer to each other than that of the A clade.
The glycosylated layer of fibrils bears gl ycans ec hoing bacterial ones, recognized by the amoeba that feeds on bacteria (Rodrigues et al. 2015, Notaro et al. 2022 ).They thus appear to be key for productive infection.In a given environment, the variable composition of the fibrils could be used to favor the engulfment of a given virion by a given amoeba, compared with other virions and even bacteria.The protein content of the layer of fibrils appears to have been optimized in a given clade, but their complex composition suggests that they can be made from a large set of div erse pr oteins .Ha ving a flexible toolbox for building the external layer of fibrils would reflect the need to constantly adjust the capsid composition to outcompete other parasites and secure infection.We can thus hypothesize that in the population of virions resulting from a single infection, this composition might also be variable, helping out the virus to ensure the pr oductiv e infection of at least one of the possible hosts present in that environment by at least one virion.
The question whether it is always the same protein that makes both the layer of fibril and the genomic fiber, e v en when other proteins than the GMC-oxidoreductases are used, remains unanswered.Yet the most abundant protein composing the external layer of fibrils of the 2KO and M4 corresponds to the 222 aminoacids qu_734 pr otein, whic h, according to the high confidence al-phaFold (Jumper et al. 2021 ) prediction, could be almost twice as small ( ∼4.5 nm) than the GMC-oxidor eductase.Fr om the pr e vious study, the central part of the genomic fiber has to correspond to a central channel of at least 9 nm to accommodate proteins such as the RNA polymerase (Villalta et al. 2022 ).There is an additional 4 nm for the DNA lining the protein shell, and the height of the GMC-oxidoreductase is 8 nm, which makes the helical shell 16 nm, leading to an ∼29-nm helix.The protein making the shell is ther efor e defining the final dimension of the genomic fiber.For the 2KO, if this is the same protein that makes the external layer of fibrils and the shell of the genomic fiber, we expect a genomic fiber of 22 nm, compatible with the thinner structur e observ ed in Fig. 2 D ( ∼20-25 nm).
We have previously proposed an increased genome redundancy as a contributing factor to the a ppear ance of viral gigantism (Bisio et al. 2023 ).The data pr esented her e v alidate suc h a pr emise and extend these predictions outside of the Pandoraviridae .Overall, our r esults r e v eal the r esilience of mimivirus, with r edundant solutions securing essential functions such as infectiousness and genome pac ka ging.Functional r edundanc y, w ell documented in the cellular world as a way to pr eserv e essential function such as cell division (Erickson and Osawa 2010 ), and until now a hallmark of the optimized micr oor ganisms , ma y thus also be at work in the viral world.

Figure 1 .
Figure 1.Mimivirus m utants' gener ation and phenotypic c har acterization.(A) Sc hematic r epr esentation of the v ector and str ategy utilized for deletion of qu_143 (KO_qu143) and qu_946 (KO_qu946) in mimivirus reunion strain; goi : gene of interest.Selection cassette was introduced by homologous recombination and recombinant viruses were generated, as shown in (B).Primer annealing sites are also shown and the sequence of the primers is included in TableS1.(B) Gr a phic depicting the strategy for the selection of recombinant viruses.Viral infection was performed 1-h post-transfection; Ntc: Nourseothricin; P, passage.(C) Growth competition assays revealed no significant defects in the lytic cycle of deletion strains .T he competition was also performed in the presence of Nourseothricin, which allows the outcompetition of the recombinant strains due to the expression of a Nourseothricin selection cassette.Measurements were performed by qPCR of an endogenous locus (present in wt and recombinant strains) and the Nourseothricin selection cassette (only present in recombinant viruses).

Figure 5 .
Figure 5. Compositional and NMR analysis of the fibrils of mimivirus reunion strain wt and mutants.(A) GC-MS chromatogram profiles of the sugars composing the fibrils of wt (a), KO_qu946 (b), KO_qu143 (c) and 2KO (d).(B) Comparison of the 1 H NMR spectra of mimivirus reunion strain wt and r elated m utants with that of mimivirus pr ototype str ain.The anomeric signals r elated to pol y_1 (C and D units) ar e in blac k, while those of pol y_2 ar e in red.(C) Structures of mimivirus polysaccharides as reported in Notaro et al. ( 2021 ).

Figure 6 .
Figure 6.Model explaining the transition from a 6-to a 5-start helix.(A) Flat models of the transition from a 6-to a 5-start involving a decrease of the helix diameter by ∼3 nm.The small cluster could thus correspond to a 25-nm diameter 4-start helix.(B) Model of the different helices.A longitudinal section of the GMC-oxidoreductase shell is represented around the central channel and each helix was positioned to produce a certain continuity of the DNA strands.
, Bo y er et al. 2011 , Notaro et al. 2021 ), while qu_946 is only ranked 13th in the fibrils ( Table

Table 1 .
Rank of the most abundant proteins in the purified fibrils of members of clade A, B and C. : not applicable, the corresponding gene is absent in the corresponding genome.Sequences of the r ele v ant orthologues of mimivirus reunion in other viruses are provided in Fig.S6. NA