Bacteriophage T5 structure reveals similarities with HK97 and T4 suggesting evolutionary relationships.

Abstract

Evolutionary relationships between viruses may be obscure by protein sequence but unmasked by structure. Analysis of bacteriophage T5 by cryo-electron microscopy and protein sequence analysis reveals analogies with HK97 and T4 that suggest a mosaic of such connections. The T5 capsid is consistent with the HK97 capsid protein fold but has a different geometry, incorporating three additional hexamers on each icosahedral facet. Similarly to HK97, the T5 major capsid protein has an N-terminal extension, or Delta-domain that is missing in the mature capsid, and by analogy with HK97, may function as an assembly or scaffold domain. This Delta-domain is predicted to be largely coiled-coil, as for that of HK97, but is approximately 70% longer correlating with the larger capsid. Thus, capsid architecture appears likely to be specified by the Delta-domain. Unlike HK97, the T5 capsid binds a decoration protein in the center of each hexamer similarly to the "hoc" protein of phage T4, suggesting a common role for these molecules. The tail-tube has unusual trimeric symmetry that may aid in the unique two-stage DNA-ejection process, and joins the tail-tip at a disk where tail fibers attach. This intriguing mix of characteristics embodied by phage T5 offers insights into virus assembly, subunit function, and the evolutionary connections between related viruses.

Full text

Page 1

Bacteriophage T5 Structure Reveals Similarities with
HK97 and T4 Suggesting Evolutionary Relationships
G. Effantin1, P. Boulanger2, E. Neumann1, L. Letellier2
and J. F. Conway1,3⁎
1Laboratoire de Microscopie
Electronique Structurale,
Institut de Biologie Structurale
J.-P. Ebel, UMR 5075
CNRS-CEA-UJF,
38027 Grenoble, France
2Institut de Biochimie et de
Biophysique Moléculaire et
Cellulaire, UMR CNRS 8619,
Université Paris-Sud XI,
91405 Orsay, France
3Department of Structural
Biology, University of
Pittsburgh School of Medicine,
Pittsburgh, PA 15260, USA
Evolutionary relationships between viruses may be obscure by protein
sequence but unmasked by structure. Analysis of bacteriophage T5 by cryo-
electron microscopy and protein sequence analysis reveals analogies with
HK97 and T4 that suggest a mosaic of such connections. The T5 capsid is
consistent with the HK97 capsid protein fold but has a different geometry,
incorporating three additional hexamers on each icosahedral facet. Similarly
to HK97, the T5 major capsid protein has an N-terminal extension, or
Δ-domain that is missing in the mature capsid, and by analogy with HK97,
may function as an assembly or scaffold domain. This Δ-domain is
predicted to be largely coiled-coil, as for that of HK97, but is ∼70% longer
correlating with the larger capsid. Thus, capsid architecture appears likely
to be specified by the Δ-domain. Unlike HK97, the T5 capsid binds a
decoration protein in the center of each hexamer similarly to the “hoc”
protein of phage T4, suggesting a common role for these molecules. The tail-
tube has unusual trimeric symmetry that may aid in the unique two-stage
DNA-ejection process, and joins the tail-tip at a disk where tail fibers attach.
This intriguing mix of characteristics embodied by phage T5 offers insights
into virus assembly, subunit function, and the evolutionary connections
between related viruses.
© 2006 Elsevier Ltd. All rights reserved.
*Corresponding author
Keywords: bacteriophage; T5; capsid; assembly domain; cryo-electron
microscopy
Introduction
T5 is one of seven tailed bacteriophages targeting
Escherichia coli that were named Tor Type phages by
Delbrück in ∼1944. It belongs to the Siphoviridae
family that includes phages λ and HK97 and is
characterized by a long flexible non-contractile tail
attached to an isometric icosahedral capsid contain-
ing the double-stranded (ds) DNA genome. Bacter-
iophage T5 is the type-member of a genus that
includes around 20 other poorly characterized
members.1Several features make T5 an unusual
and remarkable phage.2Its 121,750 bp genome is the
largest of the so-called T-odd viruses (i.e. T1, T3, T5
and T7) and carries single-stranded interruptions at
genetically defined positions on one of the DNA
strands as well as large terminal redundancies in the
form of 10,160 direct repeats. Transport of the
genome into the host occurs by a unique two step
process: 8% of the genome is transferred initially,
allowing production of pre-early phage-encoded
proteins that are necessary for transport of the
remaining DNA.3T5 also contains some of the
strongest known prokaryotic promoters.2Remark-
ably, most of these features have been known since
the 1960s and 1970s, but in comparison to phages λ,
T4 or T7 we have only a poor understanding of the
molecular events occurring during the infection
process.
Most morphological data concerning T5 come
from negative stain electron microscopy.4,5The T5
capsid has an average diameter of ∼900 Å, a planar
icosahedral outline, and a flexible tail of 120 Å
diameter with a total length of 2500 Å. The tail ends
in a conical structure to whose tip is attached a 500 Å
long straight fiber that crosses the host envelope
upon infection.3Three L-shaped fibers, joining the
Present address: G. Effantin, Laboratory of Structural
Biology, NIAMS, National Institutes of Health, Bethesda,
MD 20892, USA.
Abbreviations used: cryoEM, cryo-electron microscopy;
ds, double-stranded.
E-mail address of the corresponding author:
jxc100@pitt.edu
doi:10.1016/j.jmb.2006.06.081J. Mol. Biol. (2006) 361, 993–1002
0022-2836/$ - see front matter © 2006 Elsevier Ltd. All rights reserved.

Page 2

distal end of tail-tube, permit attachment to the host,
and the tail-tube itself appears to be assembled from
trimeric rings6rather than hexameric rings as
observed in most other phage tails. Cryo-electron
microscopy (cryoEM) of phage particles interacting
withtheir purified host receptor, the outer membrane
protein FhuA, shows similar general features7,8such
as a well preserved and regular capsid whose shape
is consistent with icosahedral symmetry. However,
the larger diameter of the T5 capsid compared to that
of other Siphoviridae such as λ and HK97 suggests
that the subunit organization is likely to be different,
in particular with a triangulation number t>7 seen
for these other capsids.
Biochemicalstudiespreviouslyidentifiedpb8asthe
major capsid protein (730 copies/phage particle) and
pb10 as a secondary head protein (114 copies/
particle)9,10that is dispensable for infectivity.4Saigo
further proposed that pb10 is a “decoration protein”
analogous to the T4 hoc protein, since both are found
in 1:6 molar ratio with their respective major head
proteins. However, examination of the T5 genome
(NCBIEntrezID,AY692264)revealsapatternofgenes
encoding capsid proteins that is more similar to
phages λ and HK97 than to T4. Located upstream of
the major head protein are the terminases, the pb7
portal protein, pb10 and a head maturational pro-
tease, pb11. Notable are two features shared by T5
and HK97 phages: one is the absence of a gene
encoding a scaffolding protein that is commonly
employed by capsids of ∼500Å diameter orlarger for
assembly but expelled from the capsid before DNA
packaging, and the second is the proteolytic removal
ofasubstantialportionofthemajorheadproteinfrom
theNterminus.ForHK97,the102residueN-terminal
domainofthegp5capsidproteinisproposedtoaidin
assembly of the capsid and is subsequently cleaved
leaving 282 residues in the mature protein,11whereas
the T5 pb8 protein is also N-terminally trimmed from
a 50 kDa precursor to the 32 kDa mature form.9
Together, these features suggest that T5 and HK97
mayshareasimilarcapsidassemblystrategydirected
by the N-terminal “assembly” domains although
yielding differently sized structures.
We have embarked on a program of structural
analysis using cryo-electron microscopy and three-
dimensional image reconstruction, beginning with
the more straightforward targets such as the
icosahedral capsid and the helical tail, but aiming
ultimately for as complete a rendering as possible of
the entire T5 phage. Unusual and therefore interest-
ing aspects of T5 morphology are in our immediate
sights, such as the structure and assembly pathway
of the relatively large capsid compared to λ and
HK97, the location and role of the pb10 decoration
protein, the organization of the encapsidated DNA
and its effect on capsid structure, as well as the
structure of the connector–tail-tip complex and
the organization of the helical tail-tube. Recent se-
quencing of the T5 genome (NCBI Entrez IDs,
AY587007,12AY692264 and AY543070) offers con-
siderable support for the present work. By compar-
ing our new structural knowledge with data
available for other viruses, we gain an under-
standing about what is common among such
complex multi-component molecular machines,
suggesting a web of evolutionary connections, as
well as what makes T5 unique.
Results
Capsid structure by cryoEM
Cryo-electron micrographs reveal T5 phage parti-
cles with regular, angular capsids and long tails that
are generally straight, although neither tip nor fibers
are readily visible (Figure 1(a)). Dark capsids are
filled with the viral genome and light capsids appear
completely devoid of DNA. We assume that the
latter are mature capsids that have been subse-
quently emptied of their DNA since they have tails
attached, an assembly step that follows DNA
packaging.
Particles were selected and subjected to analysis
using a negative-stain reconstruction as a starting
point (see Supplementary Data), yielding a cryoEM
structure of 20 Å resolution (Figure 1(b)). The
organization of pentameric capsomers at the vertices
and hexameric capsomers elsewhere, arranged on a
t=13 lattice, is clearly visualized. Two alternative
hands are possible, one is the mirror image of the
other, but the correct one cannot be determined in
such an analysis and t=13l (laevo) has been chosen
arbitrarily, although with some evidence in its favor
from the t=13l end-caps of phage T213and the giant
head of phage T4.14The sectioned views (Figure 1(c)
and (d)) show planar walls of ∼40 Å thickness and
apparent layering of dsDNA, similar to the DNA
patterns observed in reconstructions of the T7
head,15,16as well as the mutant isometric particle
of T4 that shares the same t=13 geometry17and the
prolate T4 head.18Visualization of this layering is
consistent with the spool model established for T7
and the imposition of icosahedral symmetry.15On
the radial density profile (Figure 1(g)), ten layers are
indicated ranging over radii from 160 Å –380 Å and
yielding an average spacing of 24.4 Å.
The structure of the empty capsid obtained by
urea treatment is resolved at 19 Å resolution, similar
to that of the full head particle but with generally
improved features due to the absence of co-
projected DNA. The exterior surface (Figure 1(e))
resembles the DNA-filled head, while the interior
view (Figure 1(f)) shows cavities under each
capsomer with a central dimple and six satellite
dimples, similar to the HK97 head.19Radial profiles
of the density distributions were calculated from the
empty and full capsid structures (Figure 1(g)), and
mutually calibrated with reconstructions that were
derived from the same micrographs although to a
lower resolution than shown here (data not shown).
Unexpectedly, the radial profile of the empty capsid
has a higher average radius than the full capsid,
despite the lack of internal pressure from packaged
DNA. In particular, the capsid walls near the 2-fold
994
Phage T5 is Structurally Similar to HK97 and T4

Page 3

Figure 1. CryoEM visualization of phage T5. (a) Area from a cryo-electron micrograph of DNA-filled phage particles
and including two empty particles. Capsids viewed along symmetry axes have distinctive appearances, such as the
hexagonal views that indicate 3-fold axis (an example is marked with 3), and the chevrons of the 2-fold axis (marked with
2). The ice layer traps the tails, largely limiting the orientations of the capsids to freedom of rotation about the tail axis.
Tails show a banded pattern with repeat of 43.8 Å. Inset are images of particles taken closer-to-focus where the microscope
contrast transfer function does not interfere with the DNA duplex spacing. Various patterns of DNA are visible, including
a swirling pattern (top), striations (middle), and punctate appearance (bottom). The bars represent 500 Å. Views of the
DNA-filled head particle reconstruction along an icosahedral 2-fold axis include (b) the exterior surface, (c) the sectioned
surface view with the front half of the capsid density removed, and (d) the central section corresponding to this sectioning
plane. Corresponding views of the reconstruction from empty mature capsids obtained by urea treatment include (e) the
exterior surface and (f) the interior surface where the interior capsid surface is visualized. The bar represents 200 Å. (g)
Radial density profiles reveal the averaged layer spacing of the packaged DNA to be 24.4 Å, and that the empty capsid has
a slightly larger average radius than the DNA-filled head. Arrows in (c), (d) and (f) indicate positions on the icosahedral
2-fold axis where the DNA-filled capsid appears slightly bowed in comparison to the empty capsid.
995
Phage T5 is Structurally Similar to HK97 and T4

Page 4

axes (Figure 1(f)) appear straighter than for the full
capsid (Figure 1(c)) where they are slightly bowed in
by about 10 Å reducing the diameter to 820 Å. In
comparison we detect no change in the 940 Å
diameter across the 5-fold axes. It is likely that the
accompanying conformational changes are the con-
sequence of DNA release rather than to the urea
treatment, since they were also observed on T5
particles that had released their DNA in vitro upon
bindingtotheir hostreceptor FhuA(datanot shown).
The subunits of the hexameric and pentameric
capsomers appear similar—on the exterior, a ridge
extends along the surface radially away from the
central axis of each capsomer. Biochemical character-
ization of the capsid has indicated only a single major
capsid protein, pb8, which we propose to be the
subunit for both the pentamers and hexamers, as is
the case for other Siphoviridae such as λ (gpE) and
HK97 (gp5). Apart from the numbers of subunits, the
hexamers differ from pentamers by the presence of a
lump of density extending outwardly along the local
symmetryaxisfrom the center ofeachcapsomer. This
density is only weakly continuous with the under-
lying capsid, allowing us to estimate its volume as
being consistent with a polypeptide, or polypeptides,
totaling ∼160 amino acid residues. The lack of a
strong connection suggests that this density is not an
extension of the capsid protein, and indeed it may be
removed progressively by incubation with 4 M
guanidine–HCl (Figure 2(a)). Absence of similar
density on pentamers further supports assignment
of the lump as a decoration protein while the weak
connection to the hexamer indicates that it is not
present in a 1:1 ratio with the underlying hexamer
subunits. This is consistent with the molecule being
located on the local quasi-6-fold axis, occupying
roughlythe samevolumebutindifferentorientations
at equivalent sites on the capsid surface, and
associating with any one of the six available hexamer
binding sites in a pattern that is random over
equivalent positions within and between capsids.
A candidate protein for this decoration density is
pb10, of which a monomer, 164 amino acid residues,
would account for the volume of the lump. The
molar ratio with pb8 would be 120:780 or 1:6.5, close
to the biochemical assessment of 1:610and consistent
with there being a symmetry mismatch between the
lump and the underlying hexamer. This apparently
unlikely arrangement is not without precedent,
having been established for the hoc protein of
phage T4.20To test our hypothesis, we determined
the capsid structure of a T5 mutant (T5st amN5) that
produces empty heads lacking pb10 as shown by
SDS-PAGE analysis10(Figure 2(b)). The main differ-
ence between this structure and the mature head
capsid is the absence of the lumps, confirming our
assignment of the hexamer-associated lump as a
monomer of pb10.
Modeling the T5 capsid with the HK97 capsid
protein
The maturecapsidproteins of phagesT5and HK97
have similar lengths, 299 and 282 amino acids,
respectively, and the morphology of their capsomers
is also similar. We modeled the T5 capsid by fitting a
pentamerandahexamerofthematureHK97capsid21
as a solid body into the T5 cryoEM density map
(Figure 3). Fits of subunit atomic models into cryoEM
density maps of similar resolution have proven
accurate to within 1–2 Å deviation of the Cα
atoms.22,23The angle between the two capsomers
needed no adjustment, but a small displacement of
the hexamer by 5 Å away from the pentamer was
necessary to properly center it on the local symmetry
axis. Visually, the resulting fit is very good, with only
a small region at the outer surface unoccupied that
may be attributed to the smaller size of the HK97
protein used for fitting (see below). Several small
loops of the HK97 gp5 extend beyond the T5 capsid
envelope, but these are near the hoc-like density of T5
for which there is no analog in HK97 and thus some
divergence may be expected between the capsid
structures at this location. Similarly, the E-loop of
HK97 gp5 is only partially embedded in the T5
density map, and is again a region of divergent
function having a unique role inthe unusual covalent
crosslink that generates interlinked rings in the
mature HK97 head.24,25This is a stabilizing mechan-
ism that T5 does not employ according to an absence
of capsid protein oligomers in SDS-PAGE (data not
shown)10aswellassequencecomparison(seebelow).
Protein sequence comparison between T5 and
HK97
The above results highlight the structural compar-
ison between the T5 and HK97 mature capsids. We
extended this analysis by comparing the sequences
of their major capsid proteins. HK97 dispenses with
a separate scaffolding gene for guiding capsid
Figure 2. Localization of the gp10 decoration protein.
(a) Removal of the external decoration molecule by
treatment of wild-type phage T5 with guanidine–HCl is
revealed in a time course of capsid reconstructions, shown
sectioned through a hexameric capsomer, at time points of
0 (top), 1 h (center) and 3 h (bottom). The decoration
density (arrow) is greatly diminished after 1 h and has
disappeared after 3 h. The bar represents 100 Å. (b)
Reconstruction of mutant T5 particles lacking the pb10
protein shows hexamers absent of the central decoration
density (cf Figure 1(b) and (e)). The bar represents 200 Å.
996
Phage T5 is Structurally Similar to HK97 and T4

Page 5

assembly, unlike the related phages λ and P22.
Instead, the capsid protein has an extensive amino-
terminal Δ-domain of 102 amino acid residues, 25%
of the full-length protein, that is proteolysed after
assembly of the first complete capsid form, prohead
I, and that has been proposed to aid assembly. This
domain has been visualized in cryoEM reconstruc-
tions of prohead I as a basket hanging beneath the
capsomers on the interior of the capsid.19,26Two
regions of the Δ-domain are strongly predicted to
adopt a coiled-coil conformation.19
The search for a putative T5 gene encoding a
scaffolding protein is hampered by the lack of
sequence homology in these genes. Nonetheless,
no obvious candidate is apparent in the region of
structural genes. On the other hand, several argu-
ments suggest that the T5 capsid might mature by a
mechanism similar to that of HK97. The mature
protein has a molecular mass of 32 kDa10derived
from a larger 50 kDa protein9encoded by the pb8
gene (NCBI Entrez ID, AY692264). N-terminal
sequencing of the mature pb8 (P.B. et al., unpub-
lished results) indicates that the amino-terminal 159
residues have been removed, presumably by a
putative prohead protease identified on the genome.
This excised T5 Δ-domain is 60% longer than that of
HK97 but also has a high prediction for coiled-coil
and in particular the 102 residues that overlap with
those of HK97 have a remarkably similar prediction
pattern (Figure 4(a)). Given the apparent lack of a
scaffold gene in the T5 genome, we propose that the
pb8 Δ-domain serves an assembly role as for HK97.
Alignment of the mature T5 and HK97 capsid
protein sequences by PSI-BLAST (Figure 4(b)) yields
an amino acid identity of 20%, similar to the 17%
found between structurally equivalent regions of
HK97 and the T4 capsid vertex protein, gp24, which
adopts the same core fold.27However, several
threading algorithms score the T5 capsid protein
sequence significantly higher than T4 gp24 against
the HK97 structure, suggesting that the T5 protein
may more closely resemble the HK97 fold (see
Supplementary Data). The PSI-BLAST alignment
introduced several small gaps into the HK97
sequence, corresponding to insertions in T5 pb8.
The largest, eight residues, is at the position of helix
α5 and this region of the HK97 subunit does not fit
well into the T5 capsid density (Figure 3, pink circle).
Gaps also occur at the positions of the HK97
“chainmail” cross-linking residues: Lys169, located
in the E-loop (Figure 3, black circle); Asn356; and the
putative catalytic residue, Glu363.24,25However, no
crosslink triad is evident in T5 pb8 by alignment,
supporting biochemical evidence that the T5 capsid
protein does not crosslink, and so differences in the
folds may well be expected in these regions. Several
small insertions in and around the HK97 βF strand
map to an area adjacent to the capsomer axis (Figure
3, blue circle), where the T5 pb10 protein appears to
bind and for which HK97 has no analogue.
Discussion
The capsid structure of T5 is the first reported with
triangulation number t=13 for a wild-type single-
layered virus, but several other protein shells with
this architecture are well known. The structure most
closely related is from phage T4, where a point
mutant in the major capsid protein, gp23, results in
the assembly of isometric capsids with a t=13
lattice.17,28These mutant capsids include the hoc
and soc decoration proteins: soc binds as trimers at
the periphery of the gp23 hexamers while hoc is a
monomer located at the center of each hexamer, in
the same location as the T5 pb10 molecule observed
here. The similarity is emphasized when soc is
absent from the T4 isometric capsid (Figure 5). The
hoc molecule has a length of 377 amino acid residues,
roughly twice that of T5 pb10 at 164 residues, and
with quite different appearances in the two available
T4 structures. In a personal communication, the
Figure 3. Modeling the T5 capsid density with
capsomers of HK97. (a) A side view and (b) an oblique
view of the HK97 gp5 hexamer fit into the empty T5 head
reconstruction. The pb10 lump central to the T5 hexamer
(arrow) has no analog in the HK97 capsid and is
unoccupied, but much of the remaining density is well
fit by the HK97 subunits. Differences occur at the gp5 E-
loops and at the central domains (blue circles) adjacent to
the pb10 binding sites, as well asthe gp5 α3–α5 region that
protrudes from the T5 capsid surface. The bar represents
15 Å. Inset is theatomic modelof HK97gp5 corresponding
to the capsid subunit A position (PDB entry 1OHG21).
997
Phage T5 is Structurally Similar to HK97 and T4