HIV-1 initiates genomic RNA packaging in a unique subset of host RNA granules

How HIV-1 genomic RNA (gRNA) is packaged into assembling virus remains unclear. Here, we use biochemical and in situ approaches to identify the complex in which the capsid protein Gag first associates with gRNA, termed the packaging initiation complex. First, we show that in the absence of assembling Gag, non-nuclear non-translating gRNA is nearly absent from the soluble fraction of provirus-expressing cells, and is found instead primarily in complexes >30S. When we express a Gag mutant known to be arrested at packaging initiation, we find only one complex containing Gag and gRNA; thus, this complex corresponds to the packaging initiation complex. This ∼80S complex also contains two cellular facilitators of assembly, ABCE1 and the RNA granule protein DDX6, and therefore corresponds to a co-opted host RNA granule and a previously described capsid assembly intermediate. Additionally, we find this granule-derived packaging initiation complex in HIV-1-infected H9 T cells, and demonstrate that wild-type Gag forms both the packaging initiation complex and a larger granule-derived complex corresponding to a late packaging/assembly intermediate. We also demonstrate that packaging initiation complexes are far more numerous than P bodies in situ. Finally, we show that Gag enters the ∼80S granule to form the packaging initiation complex via a two-step mechanism. In a step that is independent of a gRNA-binding domain, Gag enters a broad class of RNA granules, most of which lack gRNA. In a second step that is dependent on the gRNA-binding nucleocapsid domain of Gag or a heterologous gRNA-binding domain, Gag enters a gRNA-containing subset of these granules. Thus, we conclude that packaging in cells does not result from random encounters between Gag and gRNA; instead our data support a fundamentally different model in which Gag is directed to gRNA within a unique host RNA granule to initiate this critical event in HIV-1 replication. Nontechnical Summary To form infectious virus, the HIV-1 capsid protein Gag must associate with and package the viral genomic RNA (gRNA) during the virus assembly process. HIV-1 Gag first associates with gRNA in the cytoplasm, forming a complex termed the packaging initiation complex; this complex subsequently targets to the plasma membrane where Gag completes the assembly and packaging process before releasing the virus from the cell. Although the packaging initiation complex is critical for infectious virus formation, its identity and composition, and the mechanism by which it is formed, remain unknown. Here we identify the packaging initiation complex, and demonstrate that it corresponds to a host RNA granule that is co-opted by the virus. RNA granules are diverse complexes utilized by host cells for all aspects of RNA storage and metabolism besides translation. Our study also defines the mechanism by which HIV-1 Gag enters this host RNA granule to form the packaging initiation complex, and reveal that it involves two steps that depend on different regions of Gag. Our finding that Gag co-opts a poorly studied host complex to first associate with gRNA during packaging provides a new paradigm for understanding this critical event in the viral life cycle.

For released HIV-1 particles to be infectious, two copies of full-length genomic RNA (gRNA) 106 must be packaged during assembly of the immature HIV-1 capsid. Capsid assembly involves 107 oligomerization of Gag in the cytoplasm followed by targeting of Gag via its N-terminal myristate to the 108 plasma membrane (PM), where ~3000 Gag proteins multimerize to form each immature capsid.
packaging. Localizing packaging to RNA granules could be advantageous to the virus in a number of 144 ways: it would sequester gRNA from the host innate immune system, concentrate Gag at the site where 145 gRNA is located, and place packaging and assembly in proximity with host enzymes that could facilitate 146 those events. In keeping with the latter possibility, two of the host proteins present in both the assembly 147 intermediates and the host RNA granules from which they are derived are cellular facilitators of HIV-1 148 capsid assembly -the ATP-binding cassette protein E1 (ABCE1) and DDX6 [11,14]. 149 Here we identified the packaging initiation complex and tested the RNA granule model of HIV-1 150 packaging. First we showed that all non-translating HIV-1 gRNAs are in large complexes, even in the 151 absence of assembling Gag. Using a Gag mutant that is arrested after packaging initiation, we 152 demonstrated that only one complex contains Gag associated with gRNA, and therefore fits the definition 153 of the packaging initiation complex. This complex is an ~80S RNA granule that also contains the cellular 154 facilitators of assembly, ABCE1 and DDX6. In situ studies confirmed these findings and revealed that 155 the packaging initiation complexes are smaller and more numerous than P bodies. Additionally, we 156 7 examined the mechanism by which the packaging initiation complex is formed, and found that Gag uses 157 both a gRNA-binding-independent and a gRNA-binding-dependent step to localize to gRNA-containing 158 RNA granules. Finally, we confirmed the physiological relevance of the RNA-granule derived packaging 159 intermediates by identifying them in a chronically infected human T cell line. Together, our data argue 160 that packaging of the HIV-1 genome is initiated only after Gag localizes to a unique and poorly 161 understood subclass of host RNA granules that contains non-translating gRNA.  165 To study gRNA packaging, we used a variety of gRNA expression systems ( Fig 1A) that 166 produce either WT Gag or Gag mutants with known phenotypes (Fig 1B). WT Gag forms the packaging 167 initiation complex and completes assembly to produce virus-like particles (VLPs) that contain gRNA. 168 Two assembly-defective Gag mutants, MACA Gag and G2A Gag, were studied because they arrest Gag 169 assembly before or after packaging initiation complex formation, respectively. The assembly-170 incompetent MACA Gag fails to form the packaging initiation complex because it lacks the gRNA-171 binding NC domain (reviewed in [13,15,16]). In contrast, the assembly-defective G2A Gag forms the 172 cytoplasmic packaging initiation complex [7] but is arrested in the cytoplasm due to a point mutation that 173 prevents the myristoylation required for PM targeting [17][18][19]. We also analyzed the assembly-competent 174 but gRNA-interaction-deficient HIV-1 GagZip chimera. In place of the gRNA-binding NC domain, 175 GagZip contains a dimerizing leucine zipper (LZ), which supports assembly but not RNA association [20- 176 24]; thus GagZip produces VLPs that lack gRNA, and would not be expected to form the packaging 177 initiation complex. GagZip is of interest since a packaging model should be able to explain how GagZip 178 successfully assembles capsids but fails to incorporate gRNA. 179 We confirmed the VLP phenotypes described above following transfection of WT and mutant  (Fig 1C, top). Additionally, the predicted packaging initiation complex phenotypes of these four Gag constructs (Fig 1B) were confirmed by quantifying gRNA copies associated with intracellular Gag, 183 using immunoprecipitation (IP) from cell lysates with antibody directed against Gag (αGag), followed by 184 RTqPCR of IP eluates (Fig 1C, bottom). with MACA provirus (Fig 2A), which expresses an otherwise full-length gRNA encoding a truncated 194 assembly-incompetent MACA Gag protein that is arrested as an ~10S complex [25]. These cells were 195 harvested under conditions that removed nuclei, kept RNPCs intact, and solubilized any membranes 196 associated with proteins. Lysates were analyzed by velocity sedimentation followed by RT-qPCR to 197 determine the approximate size (S value) of all non-nuclear RNPCs containing gRNA (or other RNAs).

198
This analysis was performed with both translating and non-translating complexes intact, and also 199 following disruption of translating ribosomes with puromycin and high salt treatment (PuroHS). PuroHS of only non-translating gRNAs, the population likely to undergo packaging. 203 We first determined the efficacy of ribosome disruption by PuroHS. To do this, we analyzed 204 MACA gradient fractions for 28S rRNA, a marker for the 60S large ribosomal subunit (Fig 2B), 205 following PuroHS treatment or in the absence of PuroHS treatment. As expected, in the absence of 28S rRNA migrated almost entirely in an ~60S peak representing the dissociated large ribosomal subunit, 209 with almost no 28S rRNA remaining in the polysome region (>150S). The near complete absence of 28S 210 rRNA in the polysome region after PuroHS treatment indicated highly effective ribosome disassembly.

211
From these data we concluded that PuroHS disassembles most 28S-containing monosomes and 212 polysomes into large and small ribosomal subunits. As a negative control, the same fractions were 213 examined for 7SL RNA, a marker for signal recognition particle (SRP), which is a ribosome-independent granule, we would expect gRNA to associate with RNA granule proteins in these fractions.

248
To help identify the packaging initiation complex, we utilized a provirus that expresses G2A Gag.   In contrast, G2A Gag protein was strongly associated with gRNA by αGag coIP in a broad RNPC that 272 peaked at ~80S, heretofore referred to as the ~80S complex (Fig 3C, graph). From these data, we 273 conclude only one RNPC -the ~80S complex -fits the definition of the packaging initiation complex in 274 that it contains G2A Gag associated with gRNA.

275
Next we asked whether this ~80S RNPC is an RNA granule, by determining if gRNA in this 276 complex is associated with the RNA granule protein, DDX6. DDX6 is a marker for P bodies [31], but is 277 also found in ~80S and ~500S RNA granules that are co-opted by Gag during assembly and contain the 278 host enzyme ABCE1 [11]. In cells expressing G2A provirus (Fig 3D), gRNA was again observed 279 primarily in the ~40-80S region of gradients ( Fig 3E). Moreover, antibody to DDX6 (αDDX6) 280 coimmunoprecipitated gRNA from a complex that peaks in the ~80S region ( Fig 3F). These findings are 281 consistent with our previous observation that DDX6 is associated with Gag in the ~80S assembly 282 intermediate by coIP [11]. Taken together, these data suggest that to form the packaging initiation 283 complex, packaging-competent G2A Gag must enter an ~80S DDX6-containing RNA granule that 284 contains non-translating gRNA. Notably, gRNA is absent from the <20S fraction of the cell lysate, with 285 or without PuroHS treatment (Fig 2C), and in the presence or absence of packaging-initiation-competent 286 Gag ( Fig 3B); thus we could find no evidence of a small complex containing only gRNA and a dimer of 287 Gag.  system also allowed us to express a single well-studied V1B genomic construct with different Gag 309 constructs. As expected, coexpression of GagGFP and V1B constructs resulted in the same VLP and 310 packaging-initiation-complex phenotypes observed for proviral constructs in Fig 1B (Fig S1A).

13
To test whether WT Gag forms the ~80S putative packaging initiation complex and larger ~500S 312 and ~750S complexes, we analyzed cell lysates coexpressing either WT Gag with V1B or G2A GagGFP 313 with the V1B genomic construct. In this experiment, we used a velocity sedimentation gradient that 314 allows resolution of early and late assembly intermediates [25,30]. Steady state levels of intracellular 315 Gag protein were similar for both WT and G2A, as were gRNA levels ( Fig 4A). G2A Gag protein 316 formed only the ~10S and ~80S intermediates (as in Fig 3B); in contrast, at steady state, WT Gag protein  00S complexes, and that gRNA in those complexes was associated with ABCE1 by coIP (Fig 4D-F).

355
Gag was also present in the ~80S and ~500S regions of the gradient, as expected (Fig 4E WB). 356 Interestingly, at steady state in H9-HIV lysate, the ratio of non-translating gRNA in 80S vs. 500S  Given the longstanding observation that the NC domain is required for association of Gag with 365 gRNA, we would expect NC to be critical for targeting Gag to gRNA-containing RNA granules. To test 366 this possibility, we took advantage of GagZip, in which NC is replaced with a heterologous leucine zipper 367 (LZ). GagZip has been used to demonstrate that NC has two functions during immature capsid assembly 368 -NC binds specifically to gRNA and also promotes oligomerization of Gag via non-specific RNA  Interestingly, we found previously that GagZip forms the ~80S and ~500S ABCE1-and DDX6-373 containing assembly intermediates [11,22], even though it produces VLPs that lack gRNA (Fig 1C).

374
These data raised the possibility that GagZip localizes to a broad class of ABCE1-and DDX6-containing 375 RNA granules, but fails to localize to the subset of these granules that contains gRNA because it lacks the 376 gRNA-binding NC domain.

377
Before testing this hypothesis, we first confirmed that GagZipGFP, like GagZip, produces VLPs 378 that lack gRNA when cotransfected with the V1B genome in trans (Fig S1A and 1C). In cells transfected 379 with WT Gag or GagZipGFP, and V1B in trans (Fig 1A, Set II constructs), both Gag proteins were 380 expressed at similar steady state levels as were their gRNAs (Fig 5A), and non-translating V1B gRNA 381 was primarily in the ~80S RNA granule in both cases (Fig 5B graph). WB confirmed that GagZipGFP 382 forms a prominent ~500S RNA granule-like assembly intermediate (Fig 5B WB). In addition, previously 383 we confirmed that GagZip also forms the ~80S RNA granule albeit at lower levels than for WT Gag (in 00S fractions of cells expressing WT GagGFP (as in Fig 4C), but αGFP failed to coIP gRNA from any 386 fraction of the GagZipGFP gradient (Fig 5C, graph). Controls showed that αGFP immunoprecipitated 387 GagZipGFP protein as effectively as WT GagGFP from ~80S and ~500S fractions (Fig S3A), so the 388 failure to coIP gRNA cannot be attributed to reduced IP efficiency. These data argue that both WT Gag 389 and GagZip localize to ~80S RNA granules, but WT Gag stably associates with a subset of these RNA 390 granules that contains gRNA, while GagZip does not stably associate with the gRNA-containing subset of 391 these granules. Thus, the NC domain is required to form the packaging initiation complex, and acts by 392 directing Gag to the gRNA-containing granule subset. Moreover, our data suggest that a different region 393 of Gag (present in GagZip but not in MACA) is responsible for bringing Gag to a broader class of RNA 394 granules, most of which lack gRNA. 395 Finally, we asked whether we could restore targeting of GagZip to a gRNA-containing RNA  Therefore, we generated a construct, here called GagZipMCP, in which MCP is fused to the C-terminus 400 of GagZip. Additionally, we showed that GagZipMCP forms VLPs that contain the V1B genome, unlike 401 GagZip ( Fig S3B). When VIB was cotransfected either with GagZip or GagZip MCP, to similar steady 402 state levels (Fig 5D), and analyzed on gradients, gRNA was found largely in ~80S granules in both cases  Gag-DDX6 and Gag-ABCE1 PLA spots relative to assembly-incompetent MACA Gag, which fails to 420 enter granules (Fig 3A-C) and does not coIP with DDX6 or ABCE1 [11,25]. To test this, 293T cells were 421 transfected with WT vs. mutant provirus (Fig 1A, Set I) and analyzed for Gag-DDX6 or Gag-ABCE1 422 colocalization by PLA. Concurrent Gag IF (Fig 6A) allowed us to confirm that the majority of PLA spots 423 were observed in Gag-expressing cells, and to choose fields for quantitation with comparable Gag levels.

424
For cells expressing WT Gag, G2A Gag, or GagZip, quantified fields contained ~50 Gag-DDX6 PLA 425 spots per cell (Fig 6 B, C), three-fold more than for cells expressing MACA. Similar results were 426 observed for Gag-ABCE1 PLA spots (Fig 7). Some PLA background was expected in cells transfected 427 with MACA provirus, given that DDX6 and ABCE1, like MACA (Fig 3B), are found in the ~10S fraction  DDX6 is also found in smaller RNA granules [11]. Given that cells typically contain fewer than ten P 437 bodies [38], our finding that each Z stack image contains ~50 Gag-DDX6 PLA spots (Fig 6) suggested 438 that granules containing Gag and DDX6 are far more numerous than P bodies. To test this possibility 439 directly, we analyzed 293T cells for G2A Gag-DDX6 PLA spots with concurrent DDX6 IF, to allow 440 detection of PLA spots and P bodies in the same fields (Fig 8A). G2A provirus was used here because 441 G2A Gag is arrested as the DDX6-containing ~80S packaging initiation complex (Fig 3B, C, F,); thus, 442 most G2A Gag-DDX6 PLA spots likely represent packaging initiation complexes. Each Z stack image 443 from G2A-expressing cells displayed an average of 56 Gag-DDX6 PLA spots, but only one P body by 444 DDX6 IF (Fig 8B, C). Thus, packaging initiation complexes are far more numerous than P bodies.

445
Interestingly, DDX6 IF with high gain revealed a diffuse, low-intensity, granular DDX6 signal (Fig 8C   446 insets), in addition to the bright DDX6 positive P body foci, in both G2A-expressing and mock cells. The Our biochemical studies showed that WT Gag remains associated with the co-opted RNA granule 452 during late stages of assembly (Fig4), suggesting that assembling Gag takes the co-opted gRNA-453 containing granule to PM sites of budding and assembly. In contrast, we found that GagZip associates 454 with RNA granule proteins at late stages of assembly, but not with the subset of RNA granules that 455 contains gRNA (Fig 5). Thus, we would expect in situ approaches to reveal DDX6-containing RNA 456 granules to be present at WT Gag or GagZip PM sites of assembly; however, the granules co-opted by 457 WT Gag should also contain gRNA, while the granules co-opted by GagZip should contain DDX6 but no 458 gRNA. Previously, we used quantitative immunoelectron microscopy (IEM) to demonstrate that RNA 459 granule proteins (ABCE1 and/or DDX6) are recruited to PM sites of WT Gag and GagZip assembly 460 [11,22,33]; however, these studies did not assess whether gRNA was associated with these granules.

461
Here we used quantitative IEM with double labeling for gRNA and DDX6 to ask whether gRNA is 462 associated with RNA granules at PM sites of assembly for WT Gag vs. GagZip in situ (Fig 9). HeLa cells 463 stably expressing MCP fused to GFP (HeLa-MCP-GFP cells) were transfected with V1B genomic 464 constructs encoding MS2 binding sites and Gag in cis (Fig 9A; Fig 1A, Set IV constructs; phenotypes 465 confirmed in Fig S1B). Sections were labeled with αDDX6 (large gold) to mark RNA granules, and with 466 αGFP to mark MCP-GFP-tagged gRNAs (small gold). PM assembly sites, defined by the presence of 467 membrane deformation consistent with budding, were quantified and scored for gRNA labeling, DDX6 468 labeling, and double labeling (Fig 9B, C; Table S1). When all sites of DDX6 labeling at PM assembly 469 sites were quantified, similar high levels of DDX6 labeling were observed at both WT and GagZip PM 470 assembly sites (56% vs. 40% of all WT vs. GagZip PM assembly sites displayed DDX6 labeling, 471 respectively, p>0.01; shown as total DDX6+ in Fig 9B; shown as D+ tot in Table S1). Notably, 472 quantitation of all sites of gRNA labeling at the PM revealed that gRNA was significantly more common 473 at WT assembly sites relative to GagZip PM assembly sites (64% vs. 20% of all WT vs. GagZip PM 474 assembly sites displayed gRNA labeling, respectively, p<0.005; shown as total gRNA+ in Fig 9B; shown 475 as g+ Tot in Table S1). Our most striking results were obtained upon quantitation of PM assembly sites 476 that were double labeled for DDX6 and gRNA. Abundant gRNA and DDX6 double labeling was 477 observed at WT PM assembly sites, but not at GagZip PM assembly sites (33% vs. 5% of all WT vs.

478
GagZip PM assembly sites, respectively, displayed double labeling, p<0.005; shown as gRNA+/DDX6+ 479 in Fig 9B; shown as g+D+ in Table S1). As expected, the assembly-incompetent MACA formed very 480 few early or late PM assembly sites, unlike WT and GagZip. The same patterns were observed when 481 early and late assembly sites were analyzed separately (Table S1). Thus, IEM analysis of PM assembly 482 sites supports our conclusion that both WT and GagZip co-opt RNA granules during packaging and 483 assembly, but only the RNA granules co-opted by WT Gag also contain gRNA. Moreover, these 484 quantitative IEM studies (Fig 9) along with our PLA studies (Fig 6-8  Gag mutant, which arrests at packaging initiation ( Fig 3C). We also demonstrated that the RNA-granule-496 derived ~80S packaging initiation complex contains the host enzymes ABCE1 and DDX6, and is found in 497 HIV-1-infected human T cells. Additionally, our studies revealed that Gag does not require a gRNA-498 binding domain to stably associate with ~80S host RNA granules, but does require a gRNA-binding 499 domain to stably associate with the subset of these granules that contains gRNA. Finally, we showed that 500 packaging initiation complexes are far more numerous than P bodies in situ; thus, packaging initiation 501 complexes are not equivalent to P bodies, consistent with our finding that packaging initiation complexes 502 correspond to smaller ~80S RNA granules.

503
Based on our findings, we propose the following model for initiation of gRNA packaging (Fig   504   10). In the cytoplasm, gRNA is either in translating complexes or in non-translating host RNPCs of 505 ~3 0S-80S. WT Gag uses a two-step mechanism to co-opt gRNA-containing granules to form the ~80S 506 packaging initiation complex. One of these two steps involves an event that is independent of gRNA 507 binding, and localizes WT Gag to a broad class of ABCE1-and DDX6-containing ~80S RNA granules.

508
The second step involves a gRNA-binding-dependent event that localizes WT Gag to the subset of these 509 granules that contains gRNA. In the case of WT Gag, targeting, packaging, and late stages of  Importantly, using PLA we showed that although the packaging initiation complex contains the P 544 body marker DDX6, it is not equivalent to a P body, and corresponds instead to granules that are much 545 22 smaller than P bodies (Fig 8). The PLA findings correspond well with our biochemical findings, which 546 suggest that the ~80S granule should be similar in size to the ~80S ribosome, which is ~25 nm in 547 diameter [47]; thus the DDX6-containing ~80S granule should be four to twelve times smaller than 548 DDX6-containing P bodies, which range from 100 -300 nm in size [48]. These findings are also 549 consistent with an earlier study showing that Gag and gRNA are not found in P bodies [49]. Note, 550 however, that the co-opted ~80S RNA granules could represent P body subunits or could exchange with P 551 bodies. Thus, we cannot rule out the possibility that the packaging initiation complex is occasionally 552 found in P bodies, perhaps explaining an earlier report [50]. Future studies will need to compare the  Our identification of the two-step mechanism for stable association of Gag with gRNA-564 containing ~80S granules resulted from the observation that an assembling Gag that lacks a gRNA-565 binding domain (GagZip) targets to RNA granules that lack gRNA, but can be redirected to gRNA-566 containing granules by addition of a heterologous gRNA-binding domain (Fig 5C, F; [11,22]). Thus, it 567 appears that most ~80S RNA granules contain cellular RNAs, with only a small subset containing gRNA.

568
Notably, a key difference between MACA, which fails to enter RNA granules, and GagZip, which enters 569 ABCE1-and DDX6-containing granules that lack gRNA, is that GagZip has a heterologous 570 oligomerization domain (LZ) that substitutes for NC, a domain in WT Gag that mediates oligomerization.

23
Together, these data argue that two events are required for stable association of Gag with gRNA-572 containing granules, with one being the aforementioned poorly understood, NC-independent step in 573 which oligomerization-competent Gag targets to a large class of ~80S ABCE1-and DDX6-containing 574 RNA granules, most of which lack gRNA; and the other being a gRNA-binding step, dependent on NC or 575 a heterologous gRNA-binding domain, that enables stable association with a gRNA-containing subset of 576 these granules.

577
Our finding that, in WT Gag, the NC domain is required for packaging initiation complex   Oligos used for site-directed mutagenesis and Gibson assemblies are available upon request.

620
H9 cells (ATCC) were used to generate H9-HIVpro-cells (Set III in Fig 1A) by infection with  RT-PCR detection system and iQ5 software (Bio-Rad). Amplicons corresponding to region amplified by 685 qPCR were used to generate standard curves. Duplicate nine-point standard curves were included on 686 every qPCR plate, ranging from 10 1 copies to 10 8 copies, with a typical efficiency of ~90% or greater and 687 an R 2 of 0.99. Standard curves were able to detect 10 copies per reaction but not 1 copy, thereby setting 688 the detection threshold at 10 copies per reaction, which was equivalent to ~1000 copies per 1000 cells for 689 inputs and total gRNA from gradient fractions, ~100 copies per 1000 cells for IPs from total cell lysates 690 or gradient fractions, and ~50 copies per 1000 cells for VLPs. Minus RNA controls were included in 691 each experiment and were always zero. RT minus controls were also included in each experiment and 692 ranged from 0 -100 copies per reaction. Mock transfected VLP controls were used to set the baseline in 693 graphs and ranged from 1 -1000 copies per 1000 cells. For IPs, nonimmune gRNA copy number was 694 analyzed in parallel and was typically 1-2 logs lower than IPs. KCl, 100 mM NaCl, 0.625% NP-40). For each sample, 120 µl lysate was layered on a step gradient. To 714 resolve complexes of ~10S to ~150S (Fig 2, 3), step gradients were prepared from 5%, 10% 15%, 20%, 715 25%, and 30% sucrose in NP40 buffer without MgCl (10 mM Tris-HCl, pH 7.9, 100 mM NaCl, 50 mM 716 KCl, 0.625% NP40) and subjected to velocity sedimentation in a 5 ml Beckman MLS50 rotor at 45,000 717 rpm (162,500 x g) for 90 min at 4°C. To resolve from ~10S to ~750S, step gradients were prepared from 718 10%, 15%, 40%, 50%, 60%, 70%, and 80% sucrose in NP40 buffer without MgCl, and subjected to 719 velocity sedimentation in a 5 ml Beckman MLS50 rotor at 45,000 rpm (162,500 x g), for 45 min at 4°C.

720
Gradients were fractioned from top to bottom, and aliquots were analyzed by WB, IP, and/or RT-qPCR as    For quantification, images were acquired for ten cells from each of the three groups, with the goal 775 being to analyze similar total PM lengths in each group. Cells were chosen randomly, but excluded for 776 the WT and GagZip groups if they had fewer than ten particles at the PM visible at low power. Images  Table 1. Each PM assembly site was scored as genome positive (g+), DDX6+ (D+), or double-782 labeled (g+D+). The following definitions were used for image analysis: early PM assembly sites were 783 defined as displaying curvature at the membrane but with < 50% of a complete bud; late assembly sites" 784 at the PM were defined as displaying curvature but with > 50% of a complete bud. If early or late sites 785 contained two or more small gold particles within the full circle defined by the bud, they were scored as 786 g+. If these sites contained one or more large gold particles within a 150 nm perimeter outside the full 787 circle defined by the bud (roughly the size of an RNA granule plus space to account for the antibodies and 788 gold particle bound to an antigen at the periphery of such a granule), then they were scored as D+. Table   789 1 shows the average number of early, late, and early+late PM assembly events per 25 µm of PM per cell 790 (n=10 cells +/-SEM), along with the breakdown of how many of these events were g+ (total vs. single-791 labeled), D+ (total vs. single-labeled), or g+D+ (double-labeled). In italics are g+, D+, and g+D+ per 25 792 µm of PM per cell as a percentage of the total for each group. Significance was determined on percentage 793 data using a two tailed t-test; not significant was defined as p > 0.01. Labeling of early+late events as a 794 percent of total early+late events is also shown in graphical form in Fig 8B. As described previously 795 [22], the sensitivity of IEM for capturing colocalization is limited by a number of factors including the 796 fact that the 50 nm sections only capture < 50% of a single capsid, which has a diameter of ~100-150 nm.