Cryo-EM structure of a kinetically trapped dodecameric portal protein from the Pseudomonas-phage PaP3

Abstract

Portal proteins are dodecameric assemblies that occupy a unique 5-fold vertex of the icosahedral capsid of tailed bacteriophages and herpesviruses. The portal vertex interrupts the icosahedral symmetry, and in vivo, its assembly and incorporation in procapsid are controlled by the scaffolding protein. Ectopically expressed portal oligomers are polymorphic in solution, and portal rings built by a different number of subunits have been documented in the literature. In this paper, we describe the cryo-EM structure of the portal protein from the Pseudomonas-phage PaP3, which we determined at 3.4 Å resolution. Structural analysis revealed a dodecamer with helical rather than rotational symmetry, which we hypothesize is kinetically trapped. The helical assembly was stabilized by local mispairing of portal subunits caused by the slippage of crown and barrel helices that move like a lever with respect to the portal body. Removing the C-terminal barrel promoted assembly of undecameric and dodecameric rings with quasi-rotational symmetry, suggesting that the barrel contributes to subunits mispairing. However, ΔC-portal rings were intrinsically asymmetric, with most particles having one open portal subunit interface. Together, these data expand the structural repertoire of viral portal proteins to Pseudomonas-phages and shed light on the unexpected plasticity of the portal protein quaternary structure.

INTRODUCTION

The capsid of tailed bacteriophages and herpesviruses is built by multiple copies of a coat and scaffolding proteins assembled into an icosahedral shell that incorporates one dodecameric portal protein at a unique 5-fold vertex [1]. The portal protein functions as a bidirectional gateway through which a virus can exchange DNA with the outside environment [2]. During virus morphogenesis, the portal protein recruits small and large terminase subunits (TerS, and TerL, respectively) that assemble to form a genome-packaging motor. While DNA spools inside the capsid, the portal protein likely functions as a sensor for genome-packaging, making close contact with a ring of DNA that runs around the portal perimeter [3, 4]. After packaging, the portal is sealed by tail factors that prevent leakage of genetic material and provide an attachment point for the tail apparatus [5]. Once morphogenesis has ended and infectious particles have been formed, mature virions eject their genetic material inside bacteria through the portal protein channel. This channel can be extended to cross the bacterium cell envelope either by the tail [6] or by internal virion proteins (known as ejection proteins) that are expelled into the host [7]. In herpesviruses, the portal protein is also connected to a tail-like portal-vertex-associated tegument (PVAT) that consists of several poorly characterized factors [8–12]. Thus, both in tailed bacteriophages and herpesviruses, packaging and ejection of genomes occur through the portal protein that occupies a unique 5-fold vertex of the icosahedral capsid [2, 13, 14].

Structural studies on ectopically expressed portal oligomers [3, 15–24] and in situ visualization of portal assemblies in virions [4, 25–33] have provided valuable information to decipher the structure and assembly of portal oligomers. With minimal sequence conservation and an overall size ranging from 35-90 kDa, portal proteins share a conserved fold consisting of at least four regions, known as the wing, clip, stem, and crown [20]. A C-terminal helical barrel, which is inserted into the capsid, is also present in certain P22-like phages [34]. The exact function of the barrel remains unknown, although, in P22, it was proposed to provide an attachment point for the ejection proteins [33, 35]. The assembly of portal proteins has been studied both in vitro and in vivo, revealing fundamental similarities in the way bacteriophages and herpesvirus assemble an icosahedral capsid [2]. The portal protein forms a complex with the scaffolding protein, that in some bacteriophages and herpeviruses [2] functions as a nucleator for the assembly of an icosahedral procapsid. A cryo-EM structure of P22 procapsid visualized the scaffolding protein wedged between the coat and portal protein [36]. An asymmetric reconstruction also revealed that the scaffolding C-terminus most likely interacts with the portal protein [34]. Similarly, the portal protein was chemically cross-linked to the scaffolding protein of phage phi29, confirming that the two proteins form a molecular complex [37]. More recently, it was found that the scaffolding protein can facilitate the oligomerization of the P22 portal monomers into procapsid to generate a dodecamer [38].

Pseudomonas aeruginosa (P. aeruginosa) is a biofilm-forming, opportunistic bacterium and a common pathogen infecting COVID-19 patients [39]. It is responsible for 30% of deaths caused by pneumonia and septicemia and is a significant cause of respiratory tract infections [40]. Recent progress on phage therapy [41, 42] has fueled a growing interest in understanding the biology of Pseudomonas-phages, which are less characterized than classical phage model systems infecting Enterobacteriaceae. The architecture and composition of the genome-delivery components from Pseudomonas-phages are mostly unexplored in stark contrast to their increasing biomedical relevance. In this paper, we have studied the portal protein of PaP3, a temperate phage isolated from hospital sewage [43, 44]. The complete sequence of the PaP3 genome (~45.5 kb) revealed limited homology to other phages, although the virion morphology has similarities to classical Podoviridae, like P22 or T7 [45]. The PaP3 virion consists of an icosahedral capsid made of coat and scaffolding proteins and encodes components of a short ~12 nm tail, including a portal protein and tail spikes. PaP3 packages DNA using a cos-mechanism, and both terminase subunits were recently characterized [46, 47].

Here, we present the near-atomic structure of the PaP3 portal that we determined from a helical dodecamer using cryo-EM single-particle analysis (SPA). The structure sheds light on the conformational dynamics of the portal ring and the role the barrel plays in oligomerization.

RESULTS

Purification and assembly of Pseudomonas-phage PaP3 portal protein

We expressed the full-length PaP3 portal protein (FL-portal) (residues 1-705) in bacteria and purified milligram quantities of the recombinant protein. To assemble the dodecamer in vitro [2, 48], we concentrated the FL-portal protomer and subjected it to different incubation times at room temperature, followed by size exclusion chromatography (SEC). Surprisingly, under all conditions tested, the FL-portal migrated as a broad peak eluting between 11 ml and 14 ml on an analytical SEC column (Figure 1A, green trace). Calibration with molecular weight (M.W.) markers suggested the FL-portal existed as a heterogeneous population of particles between ~370 and 100 kDa, larger than the PaP3 portal monomer (M.W. ~70 kDa) but smaller than the expected dodecamer (M.W. ~840 kDa). The position and shape of the FL-portal peak did not change in a broad concentration range, also varying the time and temperature of incubation (data not shown). We vitrified all fractions in this peak in intervals of 300 μl and found ring-like particles in the far-left portion of the peak (eluting at ~11 ml) (Figure 1C). These particles had a high background, likely from unoligomerized portal monomers, given that the FL-portal was 95% pure by SDS-PAGE analysis (Figure 1A, bottom panel). We collected a complete dataset (9,351 movies) of the FL-portal on a 300 kV Titan Krios electron microscope equipped with a Gatan K3 direct detector.

Figure 1. Purification of recombinant PaP3 portal protein.

(A) Gel filtration chromatograms of the FL-portal at 0.32 mg/ml (green line) and the ΔC-portal at 0.35 mg/ml (unoligomerized and oligomerized are colored blue and orange, respectively). Samples were separated on a Superdex 200 10/30 SEC column. The first peak of unoligomerized ΔC-portal (blue line) eluting at 10 ml was heavily DNA-contaminated. The column was calibrated using molecular weight markers, whose elution volumes and relative molecular weights are indicated. (B) SDS-PAGE analysis of the FL- and the ΔC-portal peak fractions. Representative cryo-EM micrographs of the FL-portal (C) and oligomerized ΔC-portal (D) eluted after 11 ml and 9 ml, respectively.

In analogy to phage P22 portal protein [19, 21, 49], we also generated a C-terminal mutant of the PaP3 portal (residues 1-595) lacking the helical barrel (ΔC-portal), which we identified using a coiled-coil prediction algorithm [34]. The ΔC-portal was purified in milligram quantities, but unlike the FL-portal, it displayed the expected oligomerization-dependent migration by SEC [50]. Unoligomerized ΔC-portal migrated as a broad peak corresponding to a mixture of species in a ~50-250 kDa M.W. range, contaminated by a high M.W. species enriched in nucleic acids (Figure 1A, blue trace). In contrast, oligomerized ΔC-portal eluted as a discrete species of large M.W. (Figure 1A, orange trace) that had ring-like particles when imaged by cryo-EM (Figure 1D). This sample was used to collect a complete dataset on a 200 kV Glacios electron microscope equipped with a Falcon 4 direct detector.

Single-particle analysis of the full-length portal protein

Visual inspection of FL-portal cryo-micrographs revealed exclusively top/bottom-views, easily recognizable from the characteristic ring-like appearance. Automated auto-picking of particles based on the average M.W. identified only top/bottom-views preventing meaningful structural analysis. Similarly, 2D references generated by manually picking an initial set of 2,000 particles were biased toward the portal top/bottom-views. We estimated there were only 2 to 3 side-views out of an average of 270 particles in each micrograph, which were difficult to identify due to the high background of un-oligomerized portal monomers and the low contrast of the portal barrel (Figure 2A, side view), the hallmark feature of side-views. To overcome this limitation, we devised a brute-force computational pipeline that used both SCIPION3 (50) and RELION 3.1.2 (53,54). First, we identified a handful of rare side-views using SCIPION3’s assisted auto-picking (50). A template micrograph of manual-picking was generated and applied to other micrographs. Auto-picked micrographs were then supervised to include more particles and exclude unwanted particles (e.g., small ice particles, broken particles, etc). After training for more than 20 micrographs, the auto-picking settings were applied to the rest of the micrographs, generating an initial pool of 2.5 million particles. After the first 2D classification, three subsets of 545 K top/bottom-views, 755 K barrel-focused views, and 268 K side-views were selected (Figure 2A). To best tackle particle heterogeneity, each class was subjected to two rounds of 2D classification. The particle pool was then merged from the three subsets and re-extracted with an un-binned 240-pixel box, yielding 315 K particles (Figure 2B). One round of 2D classification was applied to the un-binned pool. 55 K particles lacking the barrel-focused subset were manually selected, and an initial 3D-map calculated from these 55 K particles imposing C6 symmetry yielded a nominal FSC resolution of 5.4 Å (Figure 2C). Despite the resolution, this averaged map was featureless, as commonly seen when the rotational symmetry used for averaging is overestimated, or particles of different subunit numbers are averaged together [51]. The pool of 55 K particles was then 3D classified with C1 symmetry with three classes (Figure 2D). One of the maps from 3D classification (boxed in Figure 2D), obtained from just 15 K particles, gave a more featureful portal oligomer that, judging from the shape and resolution (~4.5 Å), was more realistic than the C6 map obtained from 55 K particles (Figure 2C). However, the number of particles (15 K) limited the resolution of this asymmetric map. We then went back to the pool of 315 K particles (Figure 2B) and, using the 15 K pool as a reference, we were able to bootstrap an additional 23 K particles containing side- and top/bottom-views of the FL-portal (Figure 2E). This larger pool of 38 K particles was used for another round of 3D classification without imposing symmetry, which gave us three 3D maps (Figure 2F). One of them, obtained from just 14 K particles, had discernable features for the portal barrel and could be auto-refined and post-processed to 4.1 Å resolution. Noticeably, nearly half of the particles in the 14 K pool consist of side-views.

Figure 2. Flowchart of cryo-EM SPA for PaP3 FL-portal protein.

(A) A pool of 2.5 million particles consisting of the FL-portal top/bottom, barrel-focused, and possible side-views were picked from 9,351 movies. (B) The three views were 2D classified several times separately and then combined into a 315 K particles pool. (C) Initial 15 Å map calculated imposing C6 symmetry from a pool of selected 55 K particles. (D) The first 3D class with just 15 K particles was refined and post-processed to 4.5 Å. (E). An expanded pool of 38 K was obtained by adding back refined side-view and top/bottom-view without barrel particles from the previous 315 K pool to be the 2^nd pool. (F) The 2^nd pool was then 3D classified, 3D refined, and post-processed 14 K particles to 4.1 Å. Particles were then imported to RELION. (G) Focused auto-picking with a single side-view 2D average identified 632 K particles in RELION. (H) The side-view particles were then 2D classified several times to 177 K particles, and a final round of 3D classification was applied. One class of 18 K particles was selected as the 3^rd pool and refined to 4.2 Å. (I) Representative 2D-class averages of the final 28 K particles used for structural analysis.

The data were then imported into RELION 3.1.2 [52, 53]. One 2D class average from SCIPION3 corresponding to a side-view was used for reference-dependent auto-picking, which identified upward of 632 K particles (Figure 2G). These particles were subjected to four consecutive rounds of 2D classification, which reduced the pool to 177 K particles. A round of 3D classification gave only one class, obtained from 18 K particles, that had the expected portal protein shape and could be post-refined to 4.2 Å resolution (Figure 2H). At this point, the 18 K particles from RELION (Figure 2H) were combined with the best 14 K particles imported from SCIPION3 (Figure 2F) and duplicates removed. One last 2D classification was performed to discard a few hundred poorly aligned particles yielding 28,062 particles as the final pool. This pool consisted of 75 % side-views, 17 % tilted-views, and 8 % top/bottom-views, corresponding to about 1.1 % of the initial 2.5 million picked projections (Figure 2A). The particles were subjected to a final 3D auto-refinement runs with CTF refinement followed by Bayesian polishing and post-processed to 3.4 - 3.8 Å applying a tight and loose mask, respectively. The local resolution of this map varied greatly and was estimated to range from 3.37 Å to 9.68 Å with a median of 4.17 Å (Figure S1A, B).

Cryo-EM analysis of PaP3 FL-portal protein reveals a helical dodecamer

The cryo-EM structure of the FL-portal resembles a horseshoe in the top/bottom-views (Figure 3A), with a cleft at the interface of subunits g and h that explains the fuzzy edge seen in certain 2D class averages (red arrow in Figure 2I). Rotating the density by 90°, in the side-view, the FL-portal density has a helical organization with a striking fracture at the junction between subunits g and h. The two protomers facing each other at the cleft have weak electron density and poor local resolution (lower than 6 Å). In contrast, the portal region opposite the cleft has the best density (subunits a and b) (Figure 3A). The 3.4 Å asymmetric map covering the reference subunit a (Figure 3B) has a continuous and interpretable density for most side chains, which allowed us to build a complete atomic model for residues 13-600 (Figure S3). The most visible regions of the PaP3 portal protomer include the N-terminal helix-loop (residues 13-63) and the various α-helices building the stem (residues 341-336), tunnel helix (residues 410-425), and three crown helixes (residues 523-544, 554-567, and 582-600) (Figure 3B). In contrast, two surface loops spanning residues 143-190 and 242-277 (circled in Figure 3B) had no discernable density in the high-resolution map. These loops, also disordered in the structure of the P22 portal protein, interact directly with dsDNA in the mature virion [3]. Likewise, most of the C-terminal barrel (residues 596-704) was also poorly visible in the 3.4 Å sharpened map (Figure 3A), although present in our 2D class averages (red arrows in Figure 2I). This prompted us to compute a low-resolution density (~8 Å resolution) and blur the density map using a B-factor of 500 Ų, which revealed features for the barrel and crown loops (Figure S2). We modeled the C-terminal barrel of chain a as a continuous α-helical stretch [3] (residues 597-647). Similarly, the upper wing loop spanning residues 143-190 was modeled in a low-resolution density using the equivalent loop from P22 portal protein (PDB: 5JJ3) as a template (Figure S2). The complete model for the reference protomer was subjected to real space refinement yielding a Correlation Coefficient (CC) of 0.82. This protomer was then used to model the remaining 11 copies. The 12^th protomer (chain g) sitting in the cleft has weak density and was modeled as a poly-alanine. We manually adjusted the C-terminal barrel helix that is less visible in the subunits near the cleft.

Figure 3. Asymmetric reconstruction of the PaP3 portal protein.

(A) The FL-portal map obtained from the last pool of polished particles in Figure 2I was post-processed to a resolution between 3.4 Å (tight mask) and 3.8 Å (loose mask) (Figure S1). The average resolution in the portal core is ~3.4 Å, while the mispaired surface between protomers g and h is ~5.5 Å or lower. The coloring scheme reflects a resolution gradient from 3.4 Å (red) to 5.5 Å (blue). The portal barrel is visible only in a low-resolution density before CTF refinement in RELION or after blurring the B-factor (Figure S2). (B) The representative density for the best resolved PaP3 protomer (chain a) overlaid to the refined model, colored according to the refined B-factor. The arrows point to magnified secondary structure elements of the portal protomer.

Overall, the final model of the FL-portal built in the 3.4 Å cryo-EM map was refined to a final CC = 0.70 (Table 2). The quaternary structure of PaP3 portal protein reveals a helical arrangement with 12 subunits spirally arranged around a central axis, with an overall diameter of about 91 Å. The FL-portal quasi-helix has an average radius of 45.8 ± 0.7 Å standard deviation. The twelve adjacent protomers have an average distance of 22.7 ± 0.7 Å from each other, suggesting the 11 rises are nearly identical. The protomers g and h, located at the cleft (colored in gray in Figure 4A), are shifted by ~49 Å. The height of a 12-turn helix is 43.4 Å, with a height per turn of ~3.6 Å, which is calculated by measuring the distance between the subunit h and a hypothetical subunit h’ vertical to h (Figure S4).

Table 2.

Map and model refinement statistics

	FL-Portal 3.4 Å (C1)	ΔC-Portal 3.5 Å (C12)	ΔC-Portal 4.0 Å (C1)	ΔC-Portal 4.8 Å (C1)	ΔC-Portal 6.2 Å (C1)
Map
EMBD code	EMD-25500	EMD-25521	EMD-25572	EMD-25560	EMD-25562
Half maps (FSC = 0.143) (Å)	3.4	3.5	4.0	4.8	6.2
d FSC model masked (0/0.143/0.5) (Å)	2.6/3.6/4.4	3.1/3.4/4.1	3.2/3.8/4.6	3.6/4.4/6.9	4.1/6.4/9.0
Map min/max/mean	−0.03/0.05/0	−0.02/0.06/0.0	−0.01/0.09/0.01	−0.03/0.10/0.00	−0.03/0.08/0.00
Model
PDB id	7XSK	7SYA	7SYA	7SZ4	7SZ6
Chain (#)	12	12	12	12	11
Atoms (#)	48937	49860	49860	48242	43229
Protein Residues (#)	6513	6180	6180	5973	5349
Water (#)	0	0	0	0	0
RMSD: Length (Å) (# > 4σ)	0.003 (0)	0.003 (0)	0.003 (0)	0.003 (0)	0.004 (0)
RMSD: Angles (°) (# > 4σ)	0.716 (4)	0.690 (46)	0.690 (46)	0.792 (42)	0.839 (45)
MolProbity score	2.44	2.40	2.40	2.62	2.75
Clash score	16.43	15.74	15.74	25.28	34.74
Ramachandran (%) favored/allowed/outliers	81.03/18.66/0.31	82.73/16.52/0.75	82.73/16.52/0.75	80.47/18.81/0.72	80.47/19.12/0.42
Rama-Z whole (N = 6435)	−3.78 (0.10)	−2.66 (0.11)	−2.66 (0.11)	−3.2 (0.11)	−2.96 (0.11)
Rama-Z helix (N = 2868)	0.16 (0.10)	0.25 (0.10)	0.25 (0.10)	−0.77 (0.10)	−0.24 (0.11)
Rama-Z sheet (N = 479)	−1.81 (0.23)	−0.53 (0.29)	−0.53 (0.29)	−0.43 (0.27)	−0.60 (0.29)
Rama-Z loop (N = 3088)	−4.07 (0.10)	−4.05 (0.09)	−4.05 (0.09)	−3.87 (0.11)	−3.91 (0.11)
Rotamer outliers (%)	0.00	0.11	0.11	0.02	0.02
Cβ outliers (%)	0.00	0.00	0.00	0.00	0.00
Twisted proline/general (%)	1.0/0.3	0.0/0.3	0.0/0.3	0/0	0.0/0.0
CaBLAM outliers (%)	11.03	8.90	8.90	10.16	8.77
Protein B-factor (min/max/mean)	64.67/799.89/191.83	81.81/720.65/176.81	81.81/720.65/176.81	171.67/999.99/515.32	160.56/999.99/665.26
Model vs. Map
CC (mask)	0.70	0.73	0.75	0.78	0.73
CC (box)	0.78	0.79	0.86	0.89	0.88
CC (peaks)	0.62	0.65	0.71	0.68	0.62
CC (volume)	0.70	0.73	0.75	0.78	0.72

Figure 4. Structure of a kinetically trapped portal oligomer.

(A) Ribbon diagram of the PaP3 portal assembly in a side- and top/bottom-view, with α-helices illustrated as cylinders. The sidechain oxygen atom of Ser82 from each subunit is shown as a red sphere to visualize the helical pitch. (B) Ribbon diagram of the PaP3 portal protomer illustrating the five regions: wing, clip, and stem, forming the portal body, and crown and barrel in the lever. (C) Superimposition of all 12 chains from the kinetically trapped oligomer reveals significant structural rearrangements in the crown and barrel helices (dashed circle). The arrow indicates the direction of the hinge axis.

Plasticity of the PaP3 portal protomer

The PaP3 portal protomer adopts a classical portal fold [2], with five distinct subregions known as the wing, clip, stem, crown, and a C-terminal helical barrel (Figure 3C). The wing, clip, and stem are structurally continuous and form the portal body, whereas the crown and barrel are detached and do not make direct contact with the rest of the protomer tertiary structure. A DALI [54] search identified the T7 portal protein as the most similar to the PaP3 portal protomer presented in this paper. The RMSD between these two proteins is 8.8 Å, and the DALI Z-score is 18 despite the limited sequence identity (less than 10 %) (Figure S5). We next analyzed the portal:portal dimerization interface focusing on chains a and b that have the best electron density. We identified 48 residues at the portal:portal binding interface that, together, account for 29 hydrogen bonds, six salt bridges, and numerous van der Waals contacts. We mapped these bonds to the four regions of the PaP3 portal protomer (Figure S6), omitting the barrel for which the density is weak. The most heavily bonded region of the PaP3 portal lies in the stem helices, which are stabilized by 11 hydrogen bonds, and four of the six salt bridges at the portal:portal interface, in addition to 37 van der Waals (hydrophobic) contacts. The lateral stacking of stem α-helices likely drives oligomerization and keeps the portal bodies bonded together as semi-rigid entities, capable of sliding down in response to the slippage of crown and barrel helices (see below). Interestingly, the crown helices are enriched in hydrophobic contacts but lack salt bridges (Figure S6).

To determine why the PaP3 portal subunits fold into a helical assembly instead of a ring with protomers related by rotational symmetry, we systematically compared the tertiary structures of all twelve protomers modeled in the cryo-EM density using the software DynDom [55]. This analysis revealed that the PaP3-portal body is nearly identical in all subunits (Figure 4C). However, the twelve subunits drastically differ for the position of the crown α-helices (Figure 4C, see arrow), which are contiguous to the barrel helices. The crown helices slide downward by ~8 Å in the 12 subunits, with a maximum displacement observed between subunits f and h (Figure 4A) (e.g., crown helices were not modeled for chain g). Together, the crown and barrel move like a lever, sliding down into the portal channel in a spiral fashion. As a result of this movement, there is a 49 Å shift between the protomers g and h (see red spheres in Figure 4A and S4). Thus, the helical arrangement of PaP3 portal protein seen in the reconstruction results from local mispairing of subunits caused by the slippage of crown and barrel helices.

Deleting the barrel promotes the assembly of polymorphic rings

To test the hypothesis that the portal barrel is responsible for subunit mispairing, we also solved the cryo-EM structure of the ΔC-portal obtained by in vitro assembly of portal protomers (Figure 1A, D). Micrographs of the ΔC-portal had less background than the FL-portal (Figure 1C), yielding ~400 particles per micrograph and a total of 1.2 million particles from 3,189 movies. Particle orientation was close to ideal without the barrel, with approximately 228 K top/bottom-views, 641 K tilted views, and 115 K side-views. However, after several rounds of 2D- and 3D-classification, we realized the ΔC-portal dataset was highly heterogeneous, preventing straightforward SPA. CryoSPARC [56] identified an initial pool of ~200 K particles with features consistent with a dodecameric ring. Further analysis using RELION’s non-sampling 3D classification identified three major classes in this pool: a proper dodecameric ring, an open dodecamer, and an open undecamer (Figure 5A–C) that we refined to 4.0 Å, 4.8 Å, and 6.2 Å, respectively (Figure S1A, B), without applying symmetry. We found the first class, obtained from 53 K particles (with approximately 24% side-views, 40% top/bottom-views, and 36% tilted views), consisted of a genuine homo-dodecameric ring, which improved to 3.5 Å after imposing C12 symmetry (Figure S7). The PaP3 protomer previously built in the FL-portal map (Figure 3B) but lacking the C-terminal barrel was readily fit into the ΔC-portal density and used to generate a homo-dodecamer (Table 2), which had a height of 120 Å and an average diameter of 185 Å. The FL- and the ΔC-portal protomers are very similar (RMSD ~1.5 Å), suggesting the barrel affects the assembly but not the overall tertiary structure of the portal protomer. The point-to-point distances between Pro582, Ala190, and Met10 (colored in blue, orange, and green in Figure 5A) in neighboring protomers were 25 Å, 39 Å, and 47 Å, respectively, dramatically smaller than the distances measured for residues at the cleft in the FL-portal helical assembly (33 Å, 60 Å, and 63 Å, respectively) (Figure 5D).

Figure 5. Cryo-EM reconstructions of the three ΔC-portal assemblies identified in a pool of 1.2 million particles.

(A) 4.0 Å symmetric dodecamer with proper rotational symmetry, (B) 4.8 Å open dodecamer, and (C) 6.2 Å open undecamer. For reference, panel (D) shows the 3.4 Å helical dodecamer of the FL-portal. The final refined model is overlaid to the cryo-EM density in all panels. Residues Pro582, Ala190, and Met10 from juxtaposing subunits at the cleft are labeled as blue, orange, and green circles, respectively, to illustrate the degree of asymmetry in each assembly.

The second 3D class of the ΔC-portal was an open dodecamer (Figure 5B), obtained from 74 K particles. The point-to-point distances between the two protomers at the cleft were 38 Å, 60 Å, and 77 Å, respectively, suggesting a lack of binding complementarity between chains g and h. Nonetheless, the twelve subunits retained quasi-rotational symmetry, unlike the FL-portal that adopts a helical arrangement (Figure 5D). Finally, the last 3D-class contained an open undecamer, which we reconstructed from 85 K particles to 6.2 Å resolution (Figure 5C). The maximum diameter of this assembly (~180.5 Å) (Figure 5C) was comparable to that of the asymmetric dodecamer (~181.5 Å) (Figure 5B). However, the cleft between subunits g and h was more pronounced in the undecamer, with a point-to-point distance between equivalent residues at the cleft of 40 Å, 78 Å, and 101 Å (Figure 5C). Also, unlike the FL-portal, all protomers in the undecamer lay on the same plane. Overall, these results strongly suggest that the portal barrel plays a role in subunit pairing, but a mutant lacking the barrel assembles polymorphically in vitro.

Anatomy of the PaP3 DNA portal DNA channel

We generated a composite model of the dodecameric PaP3 portal protein (residues 1-705) using the FL-portal protomer (from chain a, Figure 4A) and the 3.5 Å C12-symmetrised ΔC-portal ring (Figure 5B). First, we modeled the entire barrel as a continuous ~140 Å α-helix spanning Asp602 to Arg705 (Figure S3). Then, we threaded twelve FL-portal protomers into the symmetrized structure of the ΔC-portal ring, which was subjected to energy minimization to reduce steric clashes. The coulombic electrostatic potential was calculated and is displayed with surface coloring in Figure 6A. As observed in other portal proteins [57], the DNA channel appears to be mildly negatively charged, especially under the tunnel loops. Two clusters at the top (residue Arg704, Arg705) and base (residue Arg611) of the barrel have a more pronounced basic charge. Overall, the PaP3 portal DNA channel is ~292 Å long, with a diameter ranging from ~19 Å in the middle of the tunnel loops to ~45 Å in the vestibulum, at the base of the barrel, as calculated by MOLE 2.5 (Figure 6B). The barrel has a predicted diameter of ~22 Å, sufficient to fit double-stranded B-DNA (ds-DNA). A 120-mer dsDNA threaded through the channel revealed clashing at the level of the tunnel loops that protrude inside the interior of the DNA channel generating a restriction (Figure 6C). These loops may tilt aside during DNA passage, as proposed for phages T7 and P23-45 [29, 58]. The narrower diameter of the PaP3 DNA channel provides indirect evidence supporting the push through a one-way valve model [59], which predicts the portal protein channel has a DNA-retention function [60], serving as a one-way valve to prevent backward motion of DNA during packaging.

Figure 6. Composite atomic model of PaP3 portal protein DNA channel.

(A) Section view of the PaP3 portal DNA channel showing the Coulombic electrostatic potential surface calculated using ChimeraX [83]. A 120-mer dsDNA is docked inside the portal DNA channel. (B) The diameter of the DNA-tunnel diameter was calculated using MOLE 2.5. (C) A magnified view of the restriction point in the portal channel highlighting tunnel loops (red) and dsDNA (yellow).

DISCUSSION

This paper describes the structure of the PaP3 portal protein, which we solved from two independent structural snapshots, namely, a helical assembly of the FL-portal and heterogenous portal rings lacking the C-terminal barrel. Our work expands the structural repertoire of viral portal proteins to Pseudomonas-phages and sheds light on the conformational dynamics of the portal barrel. It also highlights the power of cryo-EM in exploring macromolecular assemblies otherwise intractable using other biophysical methods.

Structural polymorphism of portal assemblies

Portal proteins exist as dodecamers in virions [61] but form polymorphic oligomers in vitro [22, 48, 61, 62]. All portal protein structures solved thus far consist of oligomeric rings generated by rotational symmetry [2], mainly dodecamers, although 13-mers [63] and 11-mers [62, 64, 65] were also reported. In this study, we struggled with the severe orientation bias of the PaP3 FL-portal, which hindered SPA. We solved this limitation using a brute force approach that required extensive data collection (>9,000 movies) combined with iterative bootstrapping of side-views. We reconstructed the asymmetric structure of the FL-portal from less than 1.1% of the total particles (~28 K particles from an initial pool of ~2.5 million particles). Unexpectedly, we found the FL-portal monomers assemble into a helical oligomer, which forms in solution at a low concentration of portal protomers. Structural analysis revealed that the mispairing of portal subunits results in a helical quaternary structure, possibly consistent with a kinetically trapped oligomerization product. The involvement of the C-terminal barrel in generating a helical assembly was validated by a deletion construct lacking the barrel, ΔC-portal, which formed oligomers even at a low concentration. However, the ΔC-portal remained intrinsically heterogeneous, forming at least three assemblies. About one-quarter of particles assembled into a ring of 12 subunits related by proper rotational symmetry. The other two classes contained open rings of 12- and 11-subunits adopting a horseshoe-like structure.

Together, our data suggest that portal asymmetry is dictated by the plastic nature of the portal fold, which is highly conserved in the virosphere [61]. The wedge-shaped portal protein body (built by the wing, clip, and stem) can stack laterally, generating an oligomeric structure. The portal crown and barrel helices, detached from the portal body, do not make direct contact with the rest of the protomer tertiary structure but determine the ability of portal bodies to stack together and generate a curvature. Intriguingly, the barrel helix is only present in large portal proteins exceeding 700 amino acids [34]. We propose the barrel directs portal assembly by recruiting the scaffolding protein, one of the three structural factors expressed during capsid morphogenesis, together with coat and portal protein. A recent study found that in phage P22, the scaffolding protein triggers portal ring assembly from portal monomers [38]. It is possible that the scaffolding protein, also a α-helical protein [66], stabilizes the portal barrel helix via the formation of transient coiled-coil structures that prevent portal subunit mispairing, avoiding the formation of kinetically trapped helical oligomers.

Role of PaP3 portal plasticity in capsid assembly and genome packaging

The cryo-EM analysis of the PaP3 portal protein presented in this paper provides direct visualization of the conformational plasticity of a portal protein. We found that the PaP3 portal oligomer forms helical and horseshoe-shaped assemblies more frequently than circular assemblies related by rotational symmetry. The observed conformational plasticity provides additional evidence that the portal protein is not a mere tunnel for DNA passage but takes part in DNA translocation by undergoing conformational changes. However, the exact way the portal protein moves during DNA translocation is unknown. The original symmetry mismatch hypothesis postulated that the portal protein rotates due to the lack of binding complementarity between portal protomers and the 5-fold icosahedral vertex, which generates a 12:5 symmetry mismatch [67]. A single-molecule experiment ruled out macroscopic rotation of the phi29 portal protein during DNA-packaging [68, 69], although conformational rearrangements of portal protein subunits have been widely documented [2, 3]. For instance, the P22 portal protein has two conformational states, an asymmetric structure in procapsid and a fully symmetric conformation in the mature virion [3]. A closed and open conformation of the T7 portal protein were visualized using cryo-EM SPA [58]. Conformational changes in the central tunnel of portal protein [29] and at the interface of portal protomers [70] were also reported. Thus, the portal may follow DNA translocation by undergoing subtle and asynchronous motion of its subunits without macroscopic rotation.

Interestingly, a recent cryo-EM analysis of phi29 TerL bound to the immature capsid revealed that the TerL pentamer adopts a helical quaternary structure in complex with the portal protein and DNA [71, 72]. This work led to a model for genome-packaging that postulates the transition between cyclic to helical symmetry powers the translocation of viral DNA inside the procapsid. The helical arrangement of the PaP3 portal protein described in this paper is consistent with the observed helical structure of phi29 TerL. We speculate that the portal accompanies genome-packaging by letting its subunits alternate between a rotational to helical motion. The relative misalignment of portal protomers due to the slippage of the crown and barrel helices may allow the portal to undergo subtle conformational changes during genome packaging. This motion would not be as extreme as that seen in the kinetically trapped PaP3 oligomer presented here, but sufficient to accompany TerL asymmetry [71, 72] by generating a wave motion inside the DNA channel. We envision the portal protein works like a set of steel ball bearings that allow TerL to undergo conformational changes while pumping DNA with minimal friction. Our hypothesis can only be tested by obtaining high-resolution snapshots of the packaging motor bound to the portal protein in the presence of DNA and comparing the relative conformation of the portal protomers.

In conclusion, cryo-EM SPA has allowed us to study the complexity of a Pseudomonas-phage portal protein beyond its static image as a homo-oligomeric ring trapped inside the virion. Our work points to the built-in plasticity of the portal fold that requires a dedicated chaperone, the scaffolding protein [36, 38], for productive assembly into procapsids and TerL-dependent genome packaging.

MATERIALS AND METHODS

Biochemical techniques

A synthetic gene encoding full-length PaP3 portal protein (FL-portal) (Gene ID: 2700603) was purchased from Genewiz and ligated into a modified pET-28a vector (+) (Novagen) (plasmid pET-28a_PaP3_PP). C-terminal tail truncated PaP3 portal protein (ΔC-portal) lacking the barrel was generated by introducing a stop codon at residue Ala596 (plasmid pET-28a_PaP3_ΔC-PP). Both portal constructs were expressed in E. coli BL21 (DE3) strain in the presence of kanamycin (50 mg/L). Bacterial cultures were grown in L.B. medium at 37 ºC until A₆₀₀ = ~0.6 when the cultures were induced with 0.5 mM IPTG for 4 hours. Cell pellets expressing portal were lysed by sonication in Lysis Buffer (20 mM Tris-HCl pH 8.0, 400 mM NaCl, 1 M Urea, 5% Glycerol). The crude extract was then subjected to centrifugation at 12,000 rpm in a JA-20 rotor for 30 min, and the cleared lysate was then incubated with Nickel Agarose Beads (GenScript) for 2 hr at 4 ºC. The beads were washed and eluted with Elution Buffer (20 mM Tris-HCl pH 8.0, 400 mM NaCl, 1 M Urea, 250 mM Imidazole, 5% Glycerol). The FL- and the ΔC-portal were further purified by SEC using an analytical Superdex 200 10/30 column (Cytiva) equilibrated with Gel Filtration Buffer (20 mM Tris-HCl pH 8.0, 200 mM NaCl, 5 mM EDTA pH 8.0). The Superdex 200 column was calibrated using Molecular Weight Markers (Sigma) that included myoglobin (17 kDa), albumin (44 kDa), γ-globulin (158 kDa), thyroglobulin (670 kDa), and blue dextran (2,000 kDa). FL-portal protomers and oligomers containing the barrel migrated smaller than excepted based on M.W. calibration markers, possibly due to non-specific contacts between the flexible barrel and gel filtration beads [73]. Eluted fractions containing portal were concentrated to ~50 mg/ml using a 30 kDa Millipore concentrator. Oligomerization of FL- or ΔC-portal monomers was induced by incubating highly concentrated portal samples (~50 mg/ml) at 22°C overnight, as previously described [48], followed by a new round of gel filtration using an analytical Superdex 200 10/30 column (Cytiva).

Vitrification and data collection

2.5 μl of the FL-portal at 1.5 mg/ml were adsorbed for 7 sec to a 300-mesh copper Quantifoil R 1.2/1.3 holey carbon grid (EMS) previously glow-discharged for 60 sec at 15 mA using an easiGlow (PELCO). The sample was blotted and frozen in liquid ethane using a Vitrobot (FEI). Movies were collected on a Titan Krios microscope operated at 300 kV and equipped with a K3 direct electron detector camera (Gatan) at the National Cryo-EM Facility at the Frederick National Laboratory, MD. Micrographs were collected in regular resolution mode with an image pixel size of 1.08 Å, a nominal 105,000x magnification, total dose 50 e/Ų, and defocus range of −1 to −2.25 um. The ΔC-portal was vitrified using the same conditions described above, except using 2.5 μl at 3.5 mg/ml specimen. Micrographs of the ΔC-portal were collected on a 200 kV Glacios equipped with a Falcon4 detector at Thomas Jefferson University. EPU software was used for data collection using accurate positioning mode. Each movie had a total accumulated electron exposure of 50 e/Ų fractioned into 45 frames. Defocus was set to −0.8 to −2.2 μm. The magnification used for data collection was 150,000x with a calibrated pixel size of 0.91 Å. Further collection parameters are in Table 1.

Table 1.

Cryo-EM data collection statistics

Parameter	FL-Portal Protein	ΔC-Portal Protein
Scope	FEI Krios (300 KV)	FEI Glacios (200 KV)
Detector	K3	Falcon 4
Imaging/Camera Mode	Nanoprobe EFTEM/Counting	NanoProbe/Counting
Program	Latitude	EPU
C2 aperture (μm)	100	50
Cs	2.7	2.7
Nominal magnification	105,000 x	150,000 x
No. micrographs	9351	3189
Pixel size (Å/px)	1.08	0.91
Spot Size	7	3
Energy filter slit (eV)	20	None
Exposure (sec)	3.4	7.88
Dose (e−/px/sec)	17.2	5.7
Total dose (e−/Å2)	50	50
No. Fractions	40	45
Defocus range/step (μm)	−1.0 to −2.25 (0.25)	−0.8 to −2.2 (0.2)
Exposures per hole	1	1

Cryo-EM single particle analysis (SPA)

A total of 9,351 movies for the FL-portal and 3,189 movies for the ΔC-portal were motion-corrected using RELION’s implementation of MotionCor2 [74]. Motion correction was applied with options of dose-weighted averaged micrographs and the sum of non-dose weighted power spectra every 4 e/Å ². For the FL-portal, motion-corrected micrographs were imported to SCIPION3 [75] for initial particle analysis. For the ΔC-portal, movies were initially processed with cryoSPARC live [56]. CTF (Contrast Transfer Function) estimation was carried out using CTFFIND4 [76]. The template used for particle picking was low-pass filtered to 20 Å to remove reference bias. All steps of SPA, including 2D/3D classification, 3D refinement, CTF refinement, Bayesian polishing, post-processing, and local resolution calculation, were carried out using RELION 3.1.2 [52, 53] on a 3-GPU (Nvidia RTX 2080 Ti) and 10 cores/20 threads CPU (Intel Xeon W-2255) Linux workstation (Ubuntu 20.04 OS system). A detailed workflow of SPA for the FL-portal is shown in Figure 2.

De novo model building, oligomer generation, and refinement

The FL- and the ΔC-portal density maps were sharpened using phenix.auto_sharpen [77] and built de novo using Coot [78] and Chimera [79]. For FL-portal, the entire protomer (chain a) was subjected to several rounds of real-space using phenix.real_space_refinement [80], which yielded a final Correlation Coefficient (CC) of 0.82. The barrel residues 601-647 were modeled in a low-resolution map as a poly-alanine. The oligomer was generated by Secondary Structure Superimposition (SSM) [81] in Coot using the dodecameric model of the P22 portal protein placed inside the PaP3 density. After SSM, the position of each PaP3 protomer was adjusted against the density first using the Fit in Map command in Chimera and then manually in Coot. The barrel domain was modeled by hand in low-resolution densities (after blurring the B-factor), and it is less visible for chains e, f, g, h, and i. The 12^th subunit (chain g) is less visible and was modeled only between residues 12-535 as a poly-alanine. A complete model of the PaP3 portal oligomer was subjected to several rounds of rigid-, real-space, and B-factor refinement using phenix.real_space_refinement. The final model includes 12 protomers (CC = 0.70) and was validated using MolProbity [82] (Table 2). The same model-building strategy was used to interpret the ΔC-portal densities. We placed the P22 portal model (PDB: 5JJ3) into the density and superimposed the PaP3 chain a (lacking the barrel) onto each of the twelve chains, generating a full dodecamer. The undecamer was modeled by fitting the dodecameric model in the 6.2 Å map and manually removing the missing subunit using Coot. All atomic models of the ΔC-portal models were refined against the sharpened cryo-EM maps using phenix.real_space_refinement, which yielded a final CC = 0.75, 0.78 and 0.71 for the symmetric dodecamer, asymmetric dodecamer, and undecamer, respectively (Table 2).

Structure analysis and modeling

All ribbon and surface representations were generated using ChimeraX [83] and PyMol [84]. Structural neighbors and flexible regions were identified using the DALI server [54]. Binding interfaces were analyzed using PISA [85] and PDBsum [86]. The sequence and secondary structure alignment were also prepared using PDBsum [86]. The DNA-tunnel diameter was calculated using MOLE 2.5 webservice [87]. RELION_postprocess [52, 53] was used for local-resolution estimation, and drawings of electron density maps and local resolution maps were generated using ChimeraX [83]. DynDom [55] was used to identify domain movements. RMSD between superimposed PDBs was calculated using SuperPose Version 1.0 (superpose.wishartlab.com) [88]. The full-length barrel spanning residues 602-705 was modeled using Coot [78]. The composite model of FL-portal protein shown in Figure 6A was refined using energy minimization in Phenix [80]. Double-stranded DNA was generated using scfbio-iitd.res.in [89]. The Coulombic Electrostatic Potential was calculated and displayed with surface coloring using ChimeraX [83].

Highlights.

• First cryo-EM structure of a Pseudomonas-phage portal protein

• Naïve PaP3 portal protein assembles into a helical quaternary structure

• The C-terminal portal barrel controls quaternary structure assembly

• Profund asymmetry of portal rings built by a different number of subunits

• Tunnel loops generate a restriction in the DNA channel

Abbreviations used:

cryo-EM

cryogenic electron microscopy

P. aeruginosa

Pseudomonas aeruginosa

FL-portal

full-length portal protein

SPA

single-particle analysis

CC

correlation coefficient

M.W.

molecular weight

RMSD

room-mean square deviation

dsDNA

double-stranded DNA

SSM

secondary structure superimposition

SEC

single particle analysis

SEC

size exclusion chromatography

SDS-PAGE

sodium dodecyl sulfate-polyacrylamide gel electrophoresis

TerS

small terminase

TerL

large terminase