The inability to visualize the packed DNA and the portal vertex in the virion reflects the difficulties in determining these structural features which lack icosahedral symmetry and are lost in any icosahedral averaging used in X-ray crystallography or electron cryomicroscopy (cryoEM). Using a cryoEM single particle reconstruction technique without symmetry imposition, we have been able to determine the structure of these critical features of Salmonella phage Epsilon15, some of which are unexpected.

Epsilon15 is a short tailed dsDNA bacteriophage that infects Salmonella anatum. Its genome (NCBI accession number: NC_004775) contains 39,671 base pairs with 49 open reading frames (Supplementary Fig. 1) among which six, coding for structural proteins, were resolved by SDS-PAGE and identified by tryptic-digest/mass spectrometry (Supplementary Fig. 2).

CryoEM of frozen hydrated Epsilon15 shows intact particles consisting of isometric heads and a protruding density (Fig. 1a). ~15,000 particle images were used for image reconstruction without symmetry imposition. The 20 Å resolution density map of the whole phage is shown in Fig. 1b (Supplementary Movie 1). The angular-shaped capsid has a T=7l shell, with a larger diameter (700 Å) along the icosahedral 5-fold axes and a smaller diameter (650 Å) along the 3-fold axes. The virion capsid is made up of 11 pentons and 60 hexons.

One of the 5-fold vertices is occupied by six tailspikes (red) surrounding an external tail hub (yellow). A cross-section of the reconstruction (Fig. 1c) and cut-away view (Fig. 1d) reveal, inside the capsid shell (dark green), the densities of the portal complex (purple), a protein core (light green), and the packaged viral DNA (blue). Figure 1e shows three well-resolved concentric shells representing the dsDNA packed coaxially around the 5-fold axis.

The ~40kb Epsilon15 dsDNA genome has a length of 14 μm in an extended form and needs to assume a compact arrangement when it is packaged into the capsid. This compact arrangement must also be topologically organized to facilitate efficient dsDNA release during infection. The Epsilon15 genome structure is sufficiently resolved to see the individual dsDNA strands with a ~25 Å separation for at least the 3 outermost layers (Figs. 1d,e), which are packed coaxially around the vertical axis in Figure 1c,d coincident with the capsid 5-fold/portal/tail axis. Neighboring dsDNA strands are packed hexagonally, which appears to be the most space efficient and energetically favorable packing pattern for the cylindrically shaped dsDNA strands. The coaxial dsDNA rings are better resolved in the outermost layers closest to the capsid shell than those in the inner layers (Figs. 1c,d). This suggests that the dsDNA packaging starts from the inner surface of the capsid shell (light blue in Figure 1d,e), spools towards the inner radius region and ends as a terminal shaft in the portal channel (dark blue in Figure 1d,e). This type of packing agrees qualitatively with the coaxial spooling model based on scattering¹⁰, simulations¹¹ and previous cryoEM reconstruction¹². Our observed spool axis is consistent with the T7 spooling model¹³, but orthogonal to an earlier model of T4¹⁰, and does not appear to vary in different layers¹¹.

On the surface of the capsid, six tailspikes, composed of gp20, connect to a central tail hub and extend out from one of capsid 5-fold vertices (Figs. 1b and and2a).2a). This symmetry mismatch with a six-fold tail complex located at a five-fold vertex is a characteristic feature of tailed dsDNA phages. Each tailspike consists of two structural domains separated by a bend: a capsid proximal stem bound to the central tail hub and a distal flower-like domain composed of three petals around a central knob (Fig. 2a, Supplementary Fig. 4b). The distal end of the flower-like domain is clearly trimeric, suggesting a homo-trimer of gp20. These tailspikes bind and cleave the O-antigen component of the host’s cell surface lipopolysaccharide¹. The petal-like densities are candidates for this enzymatic function. The bend may represent a functional hinge of the tailspike that transmits a recognition signal to activate the tail hub for viral genome injection.

The tail deviates from exact six-fold azimuthal symmetry (Fig. 2a, ,2c2c and Supplementary Fig. 4a). Each of the tailspikes has a slightly different orientation (Supplementary Fig. 4a) but similar conformation except tailspike 3 (Supplementary Fig. 4b). These differences among individual tailspikes may arise from the different interaction sites on the capsid shell surface and the symmetry mismatches between the strict 5-fold shell and the quasi 6-fold tail. Two neighboring tailspikes (labeled* in Fig. 2b) contact the protrusions at the local 2-fold positions of the shell, while the other 4 spikes have tenuous interactions with the shell at different locations at smaller radii. The two elevated contact points force these two tailspikes to a higher radius by 22Å than the other four tailspikes, and result in a slight overall tilt of the tail (Fig. 2b).

At the center of the six tailspikes is the tail hub of size ~170 Å (height) ×140 Å (width) (Fig. 2d). The tail hub has good 6-fold symmetry (Fig. 2a). It appears to anchor the tailspikes as well as cap the central opening of the portal to prevent premature DNA leakage. Upon host cell surface binding by the distal petal region of the tailspikes, the induced conformational changes might trigger the opening of the capped central channel (Figs. 1c,d) allowing for DNA injection into the host cell in a manner analogous to a gated ion channel.

Just below the tail hub lies the portal. While the oligomeric state of portals within intact phages has long been thought to be 12, it has not been measured until now. Our Epsilon15 density map provides unambiguous structural evidence that there are 12 densities at the portal in situ (Figs. 1c,d, Figs. 3a,b,c and Supplementary Figs. 5a,b), which is consistent with previous studies of portal complexes biochemically extracted from phage particles^{14, 15}.

Circumscribed by a well-resolved dsDNA ring, the portal has the appearance of twelve well-resolved turbine-like densities (180Å in diameter) with nearly equivalent angular spacing (Fig. 3c, Supplementary Fig. 5a). This overall feature resembles that seen in other isolated portals with negligible sequence similarity^16–19. The azimuthal density distribution of Epsilon15 portal deviates noticeably from perfect 12-fold symmetry (Fig. 3c and Supplementary Fig. 5a). The asymmetry of the portal ring may be caused by the non-equivalent interactions between the twelve portal subunits and the five surrounding shell protein subunits.

These observations require that the portal complex has the same spatial relationship with the tailspikes, tail hub, and the shell, from particle to particle, implying there are specific interactions among these components. Otherwise, the portal structure would not be resolved by averaging ~15,000 varying phage particles.

Figures 3a and 3d show an internal density (light green) organized immediately internal to the portal complex. This feature is ~200 Å in height and ~180 Å in width without obvious symmetry. We interpret this to be a protein core reminiscent of the phage T7 internal 8-fold symmetric structure observed in electron micrographs²⁰. The lack of obvious symmetry for Epsilon15 core could be authentic or result from an azimuthal variation of the core among particles. The Epsilon15 capsid volume can accommodate up to 90kb dsDNA. Since the Epsilon15 genome is only ~40kb, there is ample space for a protein core of this size in the capsid chamber. The protein core may facilitate the topological ordering of the dsDNA genome during packaging and/or release as suggested for T7 core²¹.

A long straight segment of uniform density (~310 Å in length and ~30 Å in width) extends from near the capsid center along the 5-fold capsid-tail axis. Because these densities are significantly higher than those of the surrounding protein core and portal complex (Figs. 1c,d), we interpret this segment of density to be the last-in segment of packaged dsDNA (~90 base pairs). It appears to be poised for release and injection into host cell during infection.

This dsDNA terminus passes through the central opening of the portal and ends slightly beyond the portal to where it is capped by densities closing the tail hub’s central channel (Figs. 1c,d). The observation that Epsilon15’s portal channel is filled with a straight segment of putative dsDNA terminus is consistent with the observation that phage SPP1’s terminal segment of packaged dsDNA is tightly associated with its portal protein²².

Imposing the icosahedral symmetry, a 9.5 Å density map was generated from a separate dataset imaged to give a higher resolution reconstruction (Supplementary Movie 2, Supplementary Figs. 3b, 6a,b). This map is sufficiently resolved to delineate the molecular boundaries of each capsid protein subunit and the secondary structure elements (Figs. 4a,b). In the average subunit map, three helices could be identified (Fig. 4b). Helix H1 is ~40Å long and extends toward the 3-fold axes where three neighboring capsomeres interact intimately (Fig. 4a, Supplementary Fig. 8b). Parallel to helix H1 is a large β-sheet that contributes significantly to the local 3-fold interactions. The two shorter helices (H2 and H3) are located at the interfaces of adjacent subunits within the same hexon and are nearly parallel to each other.

Supplementary Movie 1

Supplementary Movie 2