Cytosolic ribosomes on the surface of mitochondria

By electron cryo-tomography and subtomogram averaging, translation-arrested ribosomes were used to depict the clustered organisation of the TOM complex on the surface of mitochondria, corroborating earlier reports of localized translation. Ribosomes were shown to interact specifically with the TOM complex and nascent chain binding was crucial for ribosome recruitment and stabilization. Ribosomes were bound to the membrane in discrete clusters, often in the vicinity of the crista junctions. This interaction highlights how protein synthesis may be coupled with transport, and the importance of spatial organization for efficient mitochondrial protein import.


11
relationship between the import machinery and the CJ [52,[59][60][61][62]. To directly visualize 258 the spatial relationship between the TOM complex and the CJ in situ, the distance 259 between each MAR-R ribosome and its nearest CJ was calculated ( Fig EV4). This was 260 compared to previous data (now visualised differently) showing the distribution of 261 saturated TOM-TIM23 supercomplexes (Fig 4C). This analysis revealed that whilst both 262 TOM and TOM-TIM23 supercomplexes tend to cluster preferentially around CJs, the 263 TOM complex distribution is significantly broader than that of TOM-TIM23 ( Fig 4D). 264 Additional statistical analyses were performed to investigate the distribution of cluster 265 sizes. For both data sets, <15% of ribosomes existed as a single entity, and the major 266 group size was between 2-5 ribosomes per cluster (Fig 4E & F). In the MAR-M 267 population, ~5% of ribosomes existed in 'superclusters', defined as a group of >26 268 ribosomes. MAR-P clusters existed in groups of maximum 25 ribosomes, similar to that 269 reported previously for cytosolic ribosomes observed in whole cells [56]; in this case 270 'superclusters' were not seen ( Fig 4F).  Fig EV3). Visualization of the resulting average in the 3D volume also revealed 276 discrete clusters on small vesicles ( Fig 5C). However, as we only report on a small part 277 of the ER-R population, detailed statistical analysis of clustering was not carried out. A 278 small density could be observed to make a connection between ribosomes and the ER 279 membrane (Fig 5D). By docking X-ray structures of yeast ribosomes [63] into the ER-R 280 and MAR-M StA maps, the density was identified as rRNA expansion segment eS7 L a (Fig  281   5E). This is in agreement with previous reports of ER membrane-associated canine 282 ribosomes [57]. Contra to the ER-R population, at this resolution eS7 L a is not seen to 283 connect to the mitochondrial membrane ( Fig 5F). No density was observed for rRNA 284 expansion segment eS27 L in either structure (Fig 5E & F), in line with previous reports 285 of its extremely dynamic behavior [63]. 286

287
The lack of protein or rRNA density between the ribosome and the mitochondrial 288 membrane suggests that CHX-stabilized ribosomes could be tethered to the TOM 289 complex by the polypeptide chain only. Analysis of the distances between MAR-M or 290 ER-R populations and their corresponding membranes demonstrated the variability in 291 tethering between the two groups. The average distance (measured from the base of 292 the cleft between the 60S and 40S subunits to the membrane) was similar, at ~13 nm 293 and ~ 12 nm respectively ( Fig 5G & Fig EV5). The more notable difference was the 294 variation in tethering distances, with variance calculated at 8.6 nm for MAR-M and 3.2 295 nm for ER-R populations respectively ( Fig 5G, H & Fig EV5). With respect to tethering 296 distances, the ER-R group displayed a clear narrow distribution, with ~70% of 297 ribosomes within the range 10-14 nm from the membrane. The MAR-M group however 298 displayed a much wider distribution, with only ~50% within the same range. A StA 299 calculated for the MAR-M population that included only ribosomes located within the 300 10-14 nm range (240 particles, a similar number to that used in the ER-R average) did 301 not result in additional information (data not shown). Due to the extremely low 302 number of ribosomes bound in conditions without CHX stabilization, StA was not 303 possible. Such a flexible mode of tethering agrees with the observation that the MAR-304 M population exhibits a significant degree of orientational flexibility with respect to the 305 position of the polypeptide exit tunnel relative to the membrane (Fig 3H). 306 307 Discussion 308 309 13 Using cryoET, we were able to provide supportive evidence for the existence of co-310 translational import into isolated mitochondria. Using CHX-arrested RNCs bound to 311 mitochondria, we performed StA and biochemical analyses to 312 demonstrate that cytosolic ribosomes are localized at the mitochondrial outer 313 membrane due to nascent chain import. This is based on several lines of evidence 314 described as follows. Firstly, we were able to detect the ribosome-TOM complex 315 interaction, which was reversible by induction of nascent chain release. CryoET and StA 316 revealed two groups of associated ribosomes, a distinct population located at the 317 mitochondrial membrane (MAR-M), and a second group of soluble polysomes (MAR-P). 318 The MAR-M group was directionally oriented with the polypeptide exit tunnel pointing 319 towards the membrane for import and was tethered through the TOM complex by the 320 polypeptide chain. The ribosomes in the MAR-P population displayed more undefined 321 orientations. In human cells, polysomes were found to exist in various conformations, 322 ranging from unordered to helical, planar and spiral [56]. It is possible that organelle 323 isolation and thus the absence of certain cytosolic factors could result in the 324 predominantly undefined orientations described here. 325

326
The tethering distance between MAR-M and the mitochondrial membrane and ER-R 327 and the ER membrane is 12-13 nm, but the variance is approximately 3-fold more (8.6 328 nm to 3.2 nm) for MAR-M. The larger variation in tethering distance is likely due to the 329 flexibility and varying angle of attachment afforded by the connection made through a 330 nascent polypeptide chain. Interestingly, dissipation of the membrane potential by the 331 chemical uncoupler CCCP affected ribosome association with mitochondria only if CCCP 332 treatment preceded the addition of CHX. This indicates that post lysis RNC recruitment 333 to mitochondria had no significant effect on our results. These data do not exclude the 334 presence of a specific mitochondrial receptor for ribosomes that may be critical for 335 14 specific earlier steps of import, such as binding and initiation. A connection is observed 336 between ER-R and the membrane by eS7 L a, which is flexible in yeast as it is not 337 stabilized by ribosomal proteins such as L28e, found in other species [63]. This could 338 explain why eS7 L a appears to be partially twisted away in both structures, similar to 339 that observed previously [4]. 340

341
Here, we were able to locate 167 TOM complexes per m 2 outer membrane surface, 342 approximately 2-fold more than the 69 TOM-TIM23 import sites determined in the 343 previous study [39]. This is in agreement with the fact that TOM is more abundant in 344 mitochondria than TIM23 [64]. We also demonstrate that import through the TOM 345 complex occurs in the vicinity of CJs, but this distribution is significantly broader than 346 for arrested TOM-TIM23 supercomplexes. Our data therefore highlight key roles that 347 the TIM23 complex may play in the mitochondrial organizing network. Both MAR-M 348 and MAR-P were seen to associate with mitochondria in the form of clusters, also 349 observed for proteins imported by the TOM-TIM23 route [39]. Import sites were 350 observed to cluster around fusion sites [39] and in this work, around a potential fission 351 constriction. Yeast proteins that are reportedly involved in fusion and fission are 352 imported to mitochondria from cytosolic ribosomes [58,65]. This is therefore 353 consistent with the idea that import sites can redistribute to specific regions of 354 mitochondria [39]. 355

356
In conclusion, our data provides structural evidence supporting the theory that 357 nuclear-encoded mitochondrial proteins are synthesized locally at the mitochondrial 358 outer membrane. mRNA recruitment to the mitochondrial surface is a key step to sort 359 and polarize translation within the cell [22,27,66]. During ongoing translation the 360 distance between the nascent chain and protein translocase is short, increasing the 361 15 import efficiency [67]. Knowing that protein translocation is much faster than protein 362 translation, protein length may determine if the two processes will occur 363 simultaneously [68]. It is therefore no surprise that the most studied protein thought to 364 be delivered to mitochondria in a co-translational manner is Fum1, with a larger than 365 average molecular weight [69]. Nevertheless, by stalling translation with CHX, we could 366 observed different ribosome populations, including strings of polysomes present on the 367 mitochondrial surface. Thus, at any given time, only a small fraction of ribosomes are 368 seen to interact with the TOM complex, whilst many more could translate 369 mitochondrial proteins from a single mRNA molecule. The strains used in this study were derivatives of Saccharomyces cerevisiae YPH499 386 (MATa,   Mitochondrial samples at a protein concentration of ~5 mg ml -1 total mitochondrial 505 protein were mixed 1:1 with 10 nm protein A-gold (Aurion, Wageningen, The 506 Netherlands) as fiducial markers and applied to glow-discharged R2/2 Cu 300 mesh 507 holey carbon coated support grids (Quantifoil, Jena, Germany) by gentle pipetting. 508 Grids were blotted for ~4 s in a humidified atmosphere and plunge-frozen in liquid 509 ethane in a home-made device. Dose-fractionated tomograms (3-8 frames per 510 projection image) were typically collected from +60° to -60° at tilt steps of 2° and 5-8 511 21 μm underfocus with a total dose per tomogram of <140e -/Å 2 . Data collected at 42,000x 512 (corresponding to a pixel size of 3.3 Å) on the Titan Krios were used for all StA. 513 514 Electron cryo-tomography 515 Tomography was performed either using a Tecnai Polara, Titan Krios (FEI, Hillsboro, 516 USA) or JEM-3200FSC (JEOL, Tokyo, Japan) microscope. All microscopes are equipped 517 with field emission guns operating at 300 keV, K2 Summit direct electron detector 518 cameras (Gatan, Pleasanton, USA) and either a post-column Quantum energy filter 519 operated at a slit width of 20 eV (FEI microscopes) or an in-column energy filter 520 operated with a slit width of 40 eV (JEOL microscope). Dose fractionated data were 521 collected using Digital Micrograph (Gatan) with various pixel sizes (depending on the 522 microscope) per image. Tomograms were aligned using the gold fiducial markers and 523 volumes reconstructed by weighted back-projection using the IMOD software [76]. 524 Contrast was enhanced by non-linear anisotropic diffusion (NAD) filtering in IMOD [77]. 525 Segmentation was performed using AMIRA (FEI). 526 527

Subtomogram averaging 528
For the MAR-M and ER-R populations, two-point coordinates corresponding to the 529 centre of the ribosome and the centre of either the outer mitochondrial or ER-530 membrane were marked manually in IMOD [76]. Sub-volumes from twice-binned 531 tomograms were then extracted from NAD filtered data and an initial alignment and 532 averaging performed in SPIDER [78]. This average was used as a reference for 533 alignment and refinement using PEET [79]. A full 360° search was performed in Phi 534 (twist around the particle), whereas Theta and Psi (bending in the x-y plane and z 535 angles respectively) covered only +/-90°. 1215 subvolumes were used for the MAR-M 536 structure and 230 subvolumes for the ER-R calculation, using a mask to exclude the 537 22 membrane from the alignment. In the final iteration step for the MAR-M average, NAD-538 filtered tomograms were replaced by unfiltered contrast transfer function (CTF)-539 corrected data (Fig. 3d). Due to the reduced particle number for the ER-R population, 540 this final step was not performed. Resolution estimates were obtained using 541 conventional 'even/odd' Fourier shell correlation (FSC), applying the 0.5 FSC criterion, 542 using a mask to exclude the membranes from this estimate. In order to visualize the 543 distribution and orientation of the MAR-P population in 3D space, a StA was also 544 calculated. One-point co-ordinates were selected in the centre of each ribosome, and 545 subvolumes extracted for a full angular search in all three directions (Fig. 3e). All NAD-546 filtered ribosome populations were displayed in AMIRA using the PlaceObjectsInSpace 547 tool (Fig. 3). X-ray data of yeast ribosomes (PDB-4V6I with PDB-3IZD, including a model 548 of the position of eS27 L ) [63] were docked into comparably NAD-filtered 3D maps of 549 MAR-M or ER-R structures using Chimera (Fig. 5e & f), which was also used to remove 550 low contrast background noise for display using the 'hide dust' tool (UCSF, San 551 Francisco, USA). 552 553

Calculation of the number of ribosomes associated with each mitochondrion 554
In order to calculate the approximate number of ribosomes bound to mitochondria 555 during optimization of sample preparation (Fig. 1a), only side-view ribosomes were 556 counted. This is due to the 'missing wedge' of information in tomography and the 557 difficulty in identifying ribosomes bound to the upper and lower surfaces of 558 mitochondria, especially those that are large and dense (> 500 nm). These values 559 should therefore not be taken as absolute, but rather as a relative comparison between 560 all 4 sample preparation conditions. Sample sizes for side-view ribosomes (Fig. 1a) are 561 taken from 22 mitochondria in total and accumulate as follows: 30 MAR in +Mg(OAc) 2 , 562 206 MAR in +Mg(OAc) 2 +CHX and 824 MAR in +Mg(OAc) 2 +CHX +I. After further data 563 23 collection, an accurate absolute value was calculated for MAR-M and MAR-P 564 populations under final stabilizing conditions (+Mg(OAc) 2 +CHX +I in Fig. 1a), by 565 selecting only mitochondria in thin ice (< 500 nm) for the analysis, whereby ribosomes 566 could be clearly defined around the entire circumference (Fig. 4b). This was performed 567 for 923 MAR-M and 523 MAR-P data points, combined from 6 mitochondria. 568 Calculation of mitochondrial surface area was performed as previously described [39]. 569 570

Calculation of ribosome distribution and clustering 571
The distance between ribosomes, and between ribosomes and CJs, was determined 572 with a MATLAB (Mathworks, California, USA) script as previously described [39]. In