Egress of human cytomegalovirus through multivesicular bodies 1

33 34 Human Cytomegalovirus (HCMV) can infect a variety of cell types by using virions of varying 35 glycoprotein composition. It is unclear how this diversity is generated, but spatiotemporally 36 separated envelopment and egress pathways might play a role. So far, one egress pathway has 37 been described in which HCMV particles are separately enveloped into individual vesicles and 38

particles in large structures resembling multivesicular bodies (MVBs) of unknown origin (Bughio et

123
To get an overview of HCMV envelopment and egress routes, we initially used live-spinning disk 124 fluorescence microscopy and followed the fate of capsids and viral membranes with an HCMV 125 mutant expressing EGFP-labeled capsid-associated tegument protein pp150 and mCherry-labeled 126 glycoprotein gM (HCMV-TB40-pp150-EGFP-gM-mCherry) (Sampaio et al., 2013) using single-127 particle tracking. Despite considerable effort and computational filtering of thousands of analyzed 128 capsid tracks, we were unable to identify more than a few events in which diffraction-limited capsid 129 and membrane signals merged and were co-transported subsequently. While we assumed that this might be the result of either endo-or exocytosis of EVAs (two slices from a volume are shown in Fig. 1F).

152
Next we sought to investigate the source of EVAs. To our surprise, we also found large intracellular 168 microscopy but could also be found in the SBF-SEM data (Sup. Fig. 2). We also regularly found 169 vMVBs in cells infected with wild-type HCMV (Sup. Fig. 4).

171
Pulses of bulk release lead to viral extracellular accumulations at the plasma membrane

173
To illuminate the fate of vMVBs, we used two modalities of live-cell fluorescence microscopy. First,

174
we utilized inverted lattice light-sheet microscopy to acquire 3D-volumes of infected cells at high 175 temporal resolution for 15-45 minutes with minimal phototoxicity and photobleaching. We found 176 that vMVBs traveled from the assembly complex to the plasma membrane, where they seemed to  To confirm that the observed bulk release events were indeed induced by fusion with the plasma 194 membrane, we used the pH-sensitive fluorescent protein super-ecliptic pHluorin as a biosensor to 195 detect exocytosis events. We created a cell line stably expressing a CD63-pHluorin fusion construct 196 (Bebelman et al., 2020) as our data indicated that CD63 is enriched on vMVBs membranes but not points towards the luminal side in multivesicular structures and to the extracellular environment rendering the construct almost non-fluorescent. However, upon fusion with the plasma membrane, 201 pHluorin gets exposed to the pH-neutral extracellular milieu, and fluorescence recovers rapidly.

202
The increase in fluorescence intensity provides an easily detectable and quantifiable indicator of

206
For imaging of potential fusion events, we picked cells that had not yet accumulated EVAs on the 207 outside of the basolateral cell surface and used live-cell total internal reflection microscopy (TIRF) 208 to image fusion events for several hours without phototoxicity. We took images every 1.5-2 seconds 209 for 60 minutes since we predicted that actual membrane fusion and pH equilibration might be very 210 rapid. We found that vMVBs came into the TIRF-field and relaxed into EVAs shortly after arrival at

239
CD9 and CD81, in our hands, localized to the AC but not specifically to vMVBs (Sup. Fig. 6A-B).

240
On the other hand, exocytosed material in EVAs did not show any CD63, implying that CD63 is

260
Here, we were intrigued by the occurrence of EVAs in the majority of infected cells and investigated 261 their formation. We found that HCMV can form virus particles by budding into vMVBs by using a 262 novel 3D-CLEM workflow that combines dynamic information from spinning disk fluorescence 263 microscopy with high-resolution information from serial block-face scanning electron microscopy.

313
We found colocalization between the tetraspanin CD63 and gB, gM, and pp150. However, CD81 314 and CD9, which are also associated with exosomes, did not, in our hands, colocalize with the viral 315 markers as strongly. Since EVAs were negative for CD63, this marker might be excluded during

352
Future work needs to focus on characterizing the particle populations that are exocytosed by these 353 pathways with regards to their glycoprotein content and define their role in potentially divergent 354 egress routes. We used the HCMV strain TB40, which can produce two virus populations on HFF 355 cells that are endothelial-cell and fibroblast-topic (Scrivano et al., 2011). The EVAs that we found 356 were largely static during live-cell imaging and might represent a cell-associated viral population. the only ones that carry exosome-markers, then this would suggest that it is unlikely that they stay 362 cell-associated and play a role in cell-to-cell spread.

363
In summary, our data, combined with published studies, suggest a model in which membranes 364 originating from a fusion of both the endosomal and trans-Golgi network are used for either 365 individual envelopment or capsids or to generate vMVBs in spatiotemporally separated processes.

366
vMVBs are then transported to the plasma membrane, where fusion results in bulk pulses of virus 367 particle exocytosis and formation of EVAs (Fig. 6).

409
For imaging, the sample stage was biased with a 500V positive charge to account for sample 410 charging during the scanning process. For the acquisition, 3x3 nm pixel size images were scanned, 411 followed by the repeated ablation of 50 nm sections. The acquisition was controlled by the Gatan

417
For TEM, cells were fixed and processed as described for SBF-SEM up to the embedding step.

418
The cells were embedded in Epon without fillers, sectioned to 50 nm on a Leica Ultracut Microtome, 419 (Leica), and subsequently transferred to copper mesh grids. Electron microscopy was performed 420 on an FEI Tecnai G20 (FEI/ Thermo Fisher Scientific) and images were acquired on an Olympus 421 Veleta side mounted camera (Olympus).

499
For the weighted colocalization heatmaps, pixel intensities were calculated, taking into account the 500 absolute intensities in both channels, as well as the ratio between the intensities. The calculation 501 was performed by first normalizing each channel to relative intensity. In the following, the relative

518
The resuspended virus was centrifuged at 18000xg for 1 min at 4°C to remove large aggregates 519 and then loaded over a continuous gradient made from 15% sodium tartrate with 30% glycerol 520 (w/w) and 35% sodium tartrate (w/w) in 40 mM sodium phosphate pH 7.4 (Talbot and Almeida, and pelleted at 14000xg for 1.5 h at 4°C. The purified virus pellet was resuspended overnight in 524 PBS and then stored at -80°C.

528
The

544
Peptides resulting from trypsinization were analyzed on a Synapt G2-Si QToF mass spectrometer 545 connected to a NanoAcquity Ultra Performance UPLC system (both Waters Corporation). The data