Differences in structure and hibernation mechanism highlight diversification of the microsporidian ribosome

Assembling and powering ribosomes are energy-intensive processes requiring fine-tuned cellular control mechanisms. In organisms operating under strict nutrient limitations, such as pathogenic microsporidia, conservation of energy via ribosomal hibernation and recycling is critical. The mechanisms by which hibernation is achieved in microsporidia, however, remain poorly understood. Here, we present the cryo–electron microscopy structure of the ribosome from Paranosema locustae spores, bound by the conserved eukaryotic hibernation and recycling factor Lso2. The microsporidian Lso2 homolog adopts a V-shaped conformation to bridge the mRNA decoding site and the large subunit tRNA binding sites, providing a reversible ribosome inactivation mechanism. Although microsporidian ribosomes are highly compacted, the P. locustae ribosome retains several rRNA segments absent in other microsporidia, and represents an intermediate state of rRNA reduction. In one case, the near complete reduction of an expansion segment has resulted in a single bound nucleotide, which may act as an architectural co-factor to stabilize a protein–protein interface. The presented structure highlights the reductive evolution in these emerging pathogens and sheds light on a conserved mechanism for eukaryotic ribosome hibernation.


Introduction
Ribosome biogenesis and functionality are energyintensive processes accounting for 80% of a cell's ATP usage in nutrient-rich conditions [1,2]. When nutrients are scarce, many cells transition to a dormant state typified by low metabolic activity [3]. In this state, energy is conserved by inhibiting ribosomes via a diverse family of proteins known as hibernation factors [4]. These factors play an essential role in both the inactivation of ribosomes, as well as their recovery post-dormancy [5][6][7]. Several novel hibernation factors have been recently identified [7][8][9], including the conserved microsporidian dormancy factor (Mdf1) and the species-specific factor Mdf2 [9].
Microsporidia are obligate intracellular parasites that infect organisms as evolutionarily divergent as protists and mammals [10]. Wide-spread microsporidian infections have particularly deleterious effects on immunodeficient patients [11] and commerciallyimportant silkworms and honeybees [12]. Despite their perniciousness, recent work has shown that certain microsporidia can impair the transmission of Plasmodium falciparum, the causative agent of malaria, thereby providing a potential mechanism to control this highly infectious human pathogen [13].
The microsporidian life cycle alternates between multiple, proliferative, intracellular stages, and a sporous, metabolically inactive, environmental stage [10]. Due to extreme genome compaction and loss of seemingly essential pathways such as nucleotide synthesis [14] and ATP generation via oxidative phosphorylation [15], the intracellular stages heavily depend on host cells for their energy requirements [14]. In the spore stage, the limited availability of nutrients and the requirement for rapid reactivation of essential cellular processes post-host infection, necessitates efficient reversible hibernation mechanisms.
The extreme genome compaction in microsporidia, which has resulted in the smallest eukaryotic genomes [16,17], has not only affected metabolic genes but also ancestral macromolecules like the ribosome [9]. In most eukaryotes, ribosomes contain expanded rRNA elements, termed expansion segments (ESs), and numerous eukaryotic-specific proteins [18]. While most of these ESs stabilize the additional layer of proteins [19], it is suggested that some aid in ribosome biogenesis [20], and others extend the functional repertoire of the ribosome by providing additional interaction interfaces for regulatory factors [21]. Interestingly, microsporidia have reversed this evolutionary expansion and removed previously acquired elements as a part of their continued genomic compaction [9,22]. In the extreme case of Vairimorpha necatrix, this has created one of the smallest eukaryotic cytoplasmic ribosomes [9]. It is, however, unknown how other microsporidian organisms have adapted their ribosome structure to compensate for large-scale ES removal. Therefore, microsporidia are ideal model organisms to study rRNA evolution, as well as ribosomal hibernation due to their conspicuous dormancy.
Here, we use cryo-electron microscopy (cryo-EM) to solve the structure of the ribosome from Paranosema locustae, a species of microsporidia used as a biological insecticide to control grasshopper and locust pests [23,24]. A comparative analysis of the ribosomal expansion segments present in P. locustae, with those of the related Saccharomyces cerevisiae (yeast) and V. necatrix (microsporidia) structures, gives insights into the reductive evolution of the ribosome in microsporidia. A single structural nucleotide, discovered at the interface of two ribosomal proteins, serves as the remaining element of a removed rRNA segment and acts as the most minimal version of an ES. Furthermore, we identify a nonribosomal protein bound to the P. locustae spore ribosome as Lso2 (late-annotated short open reading frame 2), a recently discovered hibernation and recovery factor [8]. We present the first structural description of this factor in microsporidia and propose a conserved functional role in other eukaryotic organisms. Together, these results provide insights into the reductive nature of microsporidian evolution and unravel a novel mechanism of translational shutdown during the extracellular stage of these emerging pathogens.

Results
The cryo-EM structure of the ribosome from P. locustae To study the microsporidian ribosome and its interaction partners during the dormant extracellular stage, we isolated ribosomes from P. locustae spores and analyzed them using cryo-EM (Fig 1, S1 Fig). A consensus refinement resulted in a cryo-EM map at an overall resolution of 2.7 Å, with a well-resolved large subunit (LSU) and a dynamic, and less resolved small subunit (SSU)-head region. A 3D classification focused on the mobile SSU-head was performed to improve this region, resulting in two states with either a rotated (State 1, 37.7 %) or non-rotated (State 2, 39.6 %) conformation (S1B Fig). The non-rotated State 2 contains additional, but poorly resolved, density for an E-site tRNA (Fig 1). Multibody refinement of State 2 improved the local resolution for the LSU (2.83 Å), the SSU-body (3.03 Å), and SSU-head (3.27 Å) (S1B Fig, S2 Fig). The improved resolution allowed for model building of the P. locustae State 2 ribosome structure, using the S. cerevisiae ribosome as a starting template [19] (S1 Table, S2  Table). The L1 stalk, L7/L10 stalk, and parts of the SSUbeak, were not resolved and therefore not included in the final model. We observed an additional A/P-site helical density, spanning from the SSU to the LSU central protuberance (Fig 1). The high occupancy (92%; S1B Fig) and resolution of this density allowed for its unambiguous assignment as the eukaryotic hibernation and recovery factor Lso2 [8]. Both conformations of the SSU-head contain Lso2 density, suggesting it neither stabilizes one particular state, nor binds in concert with the E-site tRNA. In addition, we discovered density for a nucleotide in the interface between uL6 and eL20 (Fig 1), acting as a remnant of a removed expansion segment. The complete ribosome is shown in the center, while the SSU (left) and LSU (right) are depicted in isolation on both sides. The SSU is colored in shades of yellow (RNA in gold, proteins in light-yellow), while the LSU is colored shades of blue (RNA dark blue, proteins light blue) with selected ribosomal proteins labeled and colored in shades of green. The hibernation and recovery factor Lso2 is highlighted in red. The inset showcases the nucleotide-binding site (purple) at the interface between the two LSU proteins uL6 and eL20 (shades of green), displayed by superimposing the cryo-EM map with the molecular model.

The ribosome hibernation and recovery factor Lso2 blocks key catalytic sites
The microsporidian homologue of Lso2 is bound to the central cavity of the P. locustae ribosome (Fig 1, Fig 2). In yeast, Lso2 has been recently identified as a hibernation and recovery factor that binds to ribosomes during starvation [8]. A BLAST search allowed us to verify the presence of Lso2 in almost all sequenced microsporidia (S3A Fig). In P. locustae, Lso2 is a 77 amino acid protein comprising two consecutive alpha helices (α1, α2), separated by a small hinge region (Fig 2,  S3A Fig). Residues 2-76 are well resolved and included in our model. The N-terminal extension preceding α1 contacts the SSU mRNA binding channel between helices h24, h28, and h44 (Fig 2D). The first helix (α1), connects this binding site with ribosomal protein uL5 at the central protuberance of the LSU (Fig 2E). Several positively charged and conserved residues (K9, K10, K17) anchor the beginning of α1 to the decoding center in the SSU rRNA (Fig 2D). While spanning the central cavity, Lso2 anchors to the 25S rRNA backbone of helix-69 using R16, and stacks W40 between R55 and R60 from uL5 (Fig 2E). After a short hinge region, a second helix (α2) extends from the LSU P-site to the A-site by fitting into the major groove of H38A (Fig 2F). Here, a patch of conserved positively charged residues interact with the rRNA backbone and anchor the C-terminal end of α2 to uL16. Functionally, Lso2 blocks the binding sites of three essential components of the translational machinery. While the N-terminal extension occupies the mRNA channel in the SSU, α1 superimposes with the D-arm of the P-site tRNA, and α2 sterically blocks the T-arm of both P-site and A-site tRNAs (Fig 2B and 2C). Extensive binding-site overlap supports the role of Lso2 as a hibernation factor in microsporidia and indicates that its removal is required for reactivation of protein synthesis upon infection of a host. The general conservation of SSU-and LSU-interacting residues suggests that Lso2 would adopt a similar binding mechanism in other microsporidia as well as other eukaryotes (S3 Fig).

A bound nucleotide as evidence for adaptation to ES loss
A comparison of the P. locustae ribosome to the yeast and V. necatrix structures (Fig 3) demonstrates that microsporidia commonly reduce protein size and remove expansion segments during genome compaction. Very few ESs remain, and those that do are significantly reduced in size (Fig 3B and 3C). A previous analysis of microsporidian SSU rRNA revealed that microsporidia are evolving towards rRNA reduction [9]. Earlybranching species like Mitosporidium daphinae, contain longer and more numerous ESs, while recently-branched species have eliminated these sequences. V. necatrix has lost nearly all eukaryotic-specific ESs and several additional helices of the universal core [9]. In contrast, rRNA removal has not progressed to the same extent in P.
locustae, and remnants of ancestral elements are still present. One such example is the functionally important region surrounding the polypeptide exit tunnel in the LSU, where H7, H19, and H24 share a high structural similarity with yeast and form a narrow channel (Fig 3,  S4A Fig). Contrastingly, V. necatrix has removed these elements and features a broad and open tunnel exit. In the SSU, the two large expansion segments es6 and es3 are entirely absent in V. necatrix, while they are still partially present in the P. locustae structure (Fig 3, S4B Fig). A section of P. locustae es6 (es6C) again superimposes well with the yeast counterpart, whereas the short es6D and the three larger segments es6A, es6B, and es6E have been eliminated (S4B Fig). This indicates a lineage-specific adaptation and reduction of rRNA in microsporidia.
A notable example of adaptation to ES loss can be seen in the P. locustae structure, where the elimination of ESs may have resulted in poorly stabilized interactions between ribosomal proteins (Fig. 4). In yeast and many other eukaryotic ribosomes, a nucleotide from expansion segment 39 (A3186 in yeast) is inserted into a crevasse between uL6 and eL20 (Fig 4A and 4C). Although the P. locustae rRNA does not contain this ES (Fig 4B), extra density between uL6 and eL20 is consistent with a free nucleotide (Fig 4D). This suggests that ES39 served a vital role in the stabilization of the protein-protein interface, and the nucleotide-binding site has been retained despite the loss of ES39. The V. necatrix ribosome, on the other hand, does not contain this nucleotide-binding site and is potentially demonstrative of a later evolutionary state post-ES loss (Fig 3C). Consistently, only some of the earliest diverging microsporidian species, like M. daphinae, encode ES39.
Although previous work has suggested that the loss of ESs necessitates a loss of protein binding sites on the ribosome [22], our data indicate that this is not the case. In particular, removal of the majority of the eL8 and eL27 ES binding sites does not lead, as proposed [22], to the loss of these two proteins. Both proteins are bound to the P. locustae ribosome at high occupancy (Fig 1A), indicating that a small number of important and conserved interaction loci are sufficient for binding. On the other hand, the ribosomal proteins eL38 and eL41 of the LSU are absent in our P. locustae structure, as has been noted for V. necatrix [9], suggesting absence in most microsporidian species. Finally, no density was visible for the microsporidian-specific ribosomal protein (msL1) in P. locustae.

Discussion
Microsporidia are divergent pathogens that infect essentially all animal species [25]. Despite their broad host range and recognition as emerging pathogens of high priority [26], little is known about the evolution of their molecular machinery, which has been shaped by extreme genome compaction and an accelerated evolutionary rate [27]. Previous work has shown that the reductive nature of these pathogens has resulted in one of the most compacted eukaryotic cytosolic ribosomes [9]. The significant sequence divergence between microsporidian species suggests variability in microsporidian adaptation to genome compaction and nutrient limitation. These differences can be visualized by comparing ribosome structure, composition, and hibernation mechanisms.
While most eukaryotic ribosomes contain extensive expansion segments to stabilize ribosome structure and facilitate interactions with various ribosome-associated proteins, a previous study on the microsporidian ribosome of V. necatrix has shown that nearly all of these expansion segments are eliminated during genome compaction. In this study, we provide the first structural analysis of the P. locustae ribosome and demonstrate that it occupies an intermediate state of rRNA reduction between yeast and V. necatrix. One intriguing example of rRNA reduction is ES39, which is lost in both V. necatrix and P. locustae. In yeast, ES39 contacts several ribosomal proteins in the LSU by inserting a flipped-out base (A3186) into a binding site between uL6 and eL20. Interestingly, the deletion of ES39 in yeast is known to perturb ribosome assembly and leads to accumulation of assembly intermediates [20]. In the presented cryo-EM map, we observe clear density for a free nucleotide that superimposes well with yeast A3186 (Fig 4). It is surprising that a nucleotide-binding site would be conserved after the expansion segment was eliminated, especially since no nucleotide density was visible in the V. necatrix structure. One explanation is that V. necatrix represents a later evolutionary state in rRNA compaction, and that alterations in uL6 and eL20 have rendered the nucleotide-binding site unnecessary. In this case, the bound nucleotide in P. locustae would be a relic evincing the comparatively recent evolutionary loss of ES39.
Interestingly, although the specific functional role of most expansion segments is largely unknown, ES27 has been implicated in translation fidelity [21]. Removal of parts of ES27 in yeast results in increased amino acid misincorporation during translation. The lack of ES27 in microsporidia suggests that microsporidia either encode a separate means to ensure translational fidelity, or that they utilize a more error-prone system. Previous work supports the latter hypothesis, and one notable study demonstrated that microsporidian translation is surprisingly editingdeficient [28]. Although some misincorporation was compellingly linked to incorrect loading by amino-acyl tRNA synthetases, we hypothesize that the elimination of ES27 contributes to the low fidelity of microsporidian translation.
Translation is an energy-intensive process, requiring the hydrolysis of an estimated 30 nucleotides for the biosynthesis and polymerization of each amino acid [29]. Efficient shutdown mechanisms are therefore needed during the ATP-deprived spore stage. Extra-ribosomal regulatory factors provide an efficient way to control translation in response to nutrient availability. Ribosome hibernation factors (also known as ribosome dormancy factors [9] or silencing factors [30]) are a class of proteins that bind to and inactivate ribosomes. They function via three main methods: (1) promotion of ribosome 100S dimers by joining small subunits in a head-to-head fashion, as seen with the hibernation promoting factor (HPF) [31,32]; (2) inhibition of 70/80S ribosome formation, as with ribosomal silencing factor A (RsfA) [30]: or (3), stabilization of 70S/80S ribosomes and blocking binding sites for various translational machinery, as with interferon-related developmental regulator 2 (IFRD2) [7]. Recently discovered hibernation factors in V. necatrix function via the third method, where the conserved factor Mdf1 occupies the E-tRNA binding site in the SSU while a species-specific factor, Mdf2, blocks the polypeptide transferase center and exit tunnel [9]. In a similar fashion, Lso2 interferes with key binding sites in the translation apparatus (Fig 2B and 2C).
In yeast, Lso2 has been identified as a protein required for translation recovery after glucose starvation [8]. The purification of the P. locustae ribosome from mature dormant spores and the high occupancy of Lso2 in our structure suggests that the hibernation function is important in the extracellular spore stage of microsporidia. Similar to the other two identified microsporidian dormancy factors, Mdf1 and Mdf2 [9], the presence of Lso2 is incompatible with active translation (Fig 2B and 2C). Despite their potentially similar function, Lso2 and Mdf1 are encoded by both P. locustae and V. necatrix, suggesting different modes of ribosome hibernation are required. Based on an overlapping binding site on uL5, we speculate that only one of the two factors can bind at a time. It is also possible that Mdf1 or Lso2 is involved in removing the other factor from dormant ribosomes, i.e. hibernation and recovery functions are performed separately. This cryo-EM structure serves as a model for the efficient shutdown of a mechanistically complex macromolecular machine using a small protein, and sheds light on the reductive characteristics of a unique and emerging pathogen.

Cultivation of P. locustae and isolation of spores
The microsporidium P. locustae was propagated in a constant laboratory culture of the migratory locust Locusta migratoria (Orthoptera: Acrididae). P. locustae spore suspensions were mixed with small pieces of fresh maize leaves and fed to third instar locust nymphs at an infection ratio of 10 6 spores/nymph. Nymphs were starved for 24 hours before infection. After complete consumption of contaminated forage, nymphs were maintained at 30 °C in wooden cages with metal grids and provided constant light and fresh maize foliage. After 40-60 days, surviving insects were dissected. Swollen adipose tissue, tightly packed with spores, was homogenized in a glass vial with a Teflon pestle. The homogenates were filtered using a syringe plugged with cotton, centrifuged at 1000 x g for 5-10 minutes, and washed with distilled water [33,34]. The final spore pellet was stored at 4 °C prior to experiments.

Purification of the Paranosema locustae ribosome
Spores were resuspended in EM buffer (30 mM Tris-HCl, pH 7.5, 25 mM KCl, 5 mM magnesium acetate, 1 mM DTT, 1 mM EDTA) in a 2 ml microcentrifuge tube. To liberate ribosomes, 0.5 g of 0.5 mm zirconia beads were added to the tube, and the spores were lysed by bead beating for 30 seconds. The lysed solution was centrifuged for 15 minutes at 10,000 x g to pellet the insoluble fraction. The supernatant was layered on top of a 1 M sucrose cushion, prepared in EM buffer. Ribosomes were then pelleted by centrifugation at 105,000 x g for 4 hours at 4 °C. After removing the supernatant, the pellet was resuspended in 20 μl of EM buffer. To determine purity and concentration, 5 μl were added to 95 μl of EM buffer and absorption was measured between 240 and 300 nm.

Cryo-EM grid preparation and data collection
Sample quality and homogeneity were analyzed by cryo-EM. A Quantifoil 1.2/1.3 Cu 300 grid (Quantifoil Micro Tools GmbH, Prod. No. 658-300-CU) was glowdischarged for 30 seconds at 50 mA prior to the addition of a 3.5 µl ribosome sample (A260 4 mAU). Grids were then flash-frozen in liquid ethane using an FEI Vitrobot (ThermoFisher Scientific) set to 100% humidity, 4 °C, blot force of -5, waiting time of 1 second, and blot time of 4.5 seconds. Micrographs were collected at the Umeå Core Facility for Electron Microscopy on a Titan Krios (ThermoFisher Scientific) operated at 300 kV, equipped with a Gatan K2 BioQuantum direct electron detector (Gatan Inc.). EPU (Thermo Fisher Scientific) was used for the automated data collection of a total of 5,332 movies with 40 frames at a total dose of 28.6 e -/Å 2 , a pixel size of 1.041 Å, and a defocus range between 0.7-2 μm (S1 Table).

Cryo-EM image processing
MotionCor2 [35] was used for the initial movie alignment, drift correction, and dose-weighting. The contrast transfer function was determined using the CTFFIND-4.1.13 [36]. Micrographs with poor CTF fits or drift were removed after manual inspection, resulting in a total of 5,274 micrographs. Particle picking was performed with the deep learning object detection system crYOLO [37] using a trained model based on 2781 manually picked particles. A total of 318,301 particles were initially picked. After manual inspection of all coordinates and micrographs, 320,669 particles were extracted with a box size of 400 pixels (416 Å) and subjected to an initial 2D classification to remove remaining picking contaminants. Well-resolved 2D class averages (314,213 particles) were used alongside a 60 Å lowpass-filtered initial cryoSPARC [38] model for refinement in RELION-3.1 [39].
Consensus refinement of all particles resulted in a map of 3.04 Å, which was further improved by per-particle CTF refinement to an overall resolution of 2.73 Å. Despite well-resolved density for Lso2, a focused classification with 3 classes suggests that ~92% of all particles are bound by this protein (S1B Fig). Weak density for an E-site tRNA was observed and conformational heterogeneity in the SSU-body and head region resulted in less well-resolved small subunit density. To further improve the density for the small subunit head region, an SSU-head and E-site tRNA focused 3D classification without image alignment was performed using 3 classes (S1B Fig). Two of these classes displayed an improved overall resolution for the SSUhead and tRNA site. Class 1 contained 37.7% (118,600 particles) and Class 2 contained 39.6% (124,947 particles) of the total particles mass. An overlay of both classes suggests that they adopt different rotational states (S1B Fig). The particles of Class 2 were selected and refined to an overall resolution of 2.93 Å (S2A Fig). To improve resolution of the distinct subdomains in State 2, a multibody refinement was performed focusing on the SSU-head, SSU-body and LSU regions separately. This resulted in resolutions of 3.28Å (SSU-head), 3.04Å (SSUbody) and 2.83Å (LSU) (S2B Fig). The resulting focused refined maps and their half maps were combined using PHENIX [40] to generate a map at 2.9 Å (S2B Fig).

Model building, refinement, and validation
We used three available, but non-annotated, P. locustae genomes [41][42][43] as a database to identify ribosomal protein and rRNA sequences (S2 Table). Yeast or Vairimorpha necatrix protein and rRNA sequences were used with tblastn [44] or blastn against the in-house genome database to identify the respective homologs in P. locustae. The high-resolution crystal structure of the yeast ribosome (PDB 4V88, [19]) was used as initial template for modelling the P. locustae ribosome in Coot [45]. Lso2 was built de-novo in Coot. In the overall structure, a small number of surface-exposed cysteines showed additional density close to the thiol groups indicating a low level of oxidation. Model refinement was performed against the combined map of State 2 (2.9 Å) with PHENIX [40] using phenix.real_space_refine (version 1.14-3260) and manually defined zinc coordination and rRNA restraints. Model validation was performed according to [46]. All atomic coordinates were randomly displaced by 0.5 Å, followed by refinement against half map 1. The Fourier-Shell Correlation coefficient of the resulting refined model, and half map 1 or half map 2 was calculated to evaluate the model for overfitting. Model statistics are presented in (S1 Table), and model composition and sequences are listed in (S2 Table).

Map and model visualization
Structure analysis and figure preparation were performed using PyMOL (Schrödinger) [47] and UCSF ChimeraX [48].