Identification of phage encoded family of tubulins, PhuZ
We surveyed the genomic database and found a number of tubulin-like protein sequences encoded within several different phage genomes. A phylogenetic tree demonstrates that these phage-encoded tubulins are extraordinarily diverse and are only distantly related to the cell division protein FtsZ and the plasmid segregation protein TubZ (). All of the phage genomes encoding these tubulins are very large, ranging from 186 to 316 kb and they infect both Gram negative and Gram positive bacteria. Although many of these phage have been characterized previously, no evidence for a tubulin-based cytoskeletal polymer has been reported.
Phylogenetic relationship and conserved sequences of PhuZ and other distantly related tubulins
PhuZ forms dynamic filaments in vivo and in vitro
To study one of these novel tubulin-like proteins and determine if it had the ability to polymerize in vivo
, we generated a GFP fusion to the phuZ
) from phage 201ϕ2-1 and expressed it from the arabinose promoter (Qiu et al., 2008
) on a plasmid in Pseudomonas chlororaphis
. When expressed at 30°C at low levels (0.15% or 0.25% arabinose), fluorescence from GFP-PhuZ was uniform throughout the cell (). As the arabinose concentration increased, a threshold concentration (0.4%) was reached where the majority of cells (82%) spontaneously assembled filaments (). Quantitation of GFP-PhuZ expression using in-gel fluorescence demonstrated that the fusion protein was full length and that expression increased linearly with arabinose concentration (Figure S1
). When expressed just above the threshold inducer concentration for assembly (0.4%), cells contained multiple filaments () that moved rapidly throughout the cell (Movie S1
). At higher expression levels (1%), GFP-PhuZ filaments extended the entire length of the cell ().
PhuZ polymer assembly in P.chlororaphis
To gain insight into the nature of PhuZ filaments observed in vivo
, PhuZ was expressed recombinantly, purified (Figure S2A
) and its in vitro
polymer growth kinetics examined by right-angle light scattering. As with other tubulins, PhuZ polymerizes in a GTP-dependent manner, with no polymerization observed in the presence of GDP (). It also displays a lag phase characteristic of a nucleation-extension mechanism of polymer growth. As expected, the length of the lag phase and maximum signal at the plateau are proportional to the concentration of PhuZ, from which we determine the critical concentration to be 2.8 ± 0.1 µM (Figure S2B
To assess the morphology of PhuZ filaments, PhuZ was polymerized in vitro
in the presence of the non-hydrolyzable GTP analog GMPCPP and examined by negative-stain EM. shows a representative micrograph of individual PhuZ filaments together with high magnification views (also Figure S2
). The morphology of PhuZ filaments is distinct from microtubules formed by tubulin and single-stranded filaments formed by FtsZ, but is reminiscent of the two-stranded, helical filaments formed by TubZ (Aylett et al., 2010
; Chen and Erickson, 2008
). Although crossover events are observed in the PhuZ filaments, they appear at irregular intervals, and their architecture is not immediately evident. Further EM analysis will be required to determine the detailed structure of the filaments.
Structure of the PhuZ-GDP monomer
To better understand the similarities and differences between PhuZ and other tubulins, crystals of wild-type and Se-Met PhuZ were grown in the presence of GDP and the structure of Se-Met PhuZ was solved by MAD phasing (Table S1
) and subsequently refined using the 1.67 Å resolution, wild-type PhuZ-GDP data (). The initial 2Fo
electron density maps at 1.67Å showed strong density for GDP in the nucleotide-binding pocket (). As in other tubulins, the structure of PhuZ consists of two domains: an N-terminal domain containing the nucleotide binding pocket () and C-terminal domain, bridged by a long, central helix, H7. Notably, a short loop replaces the normally highly conserved tubulin inter-domain helix, H6. The structure of PhuZ represents the first tubulin homologue in which all of the C-terminal residues have been observed, revealing that they form an extended C-terminal tail (residues 295–315) appended to a long helix, H11. Overlays of the PhuZ backbone structure with those of α-tubulin (Nogales et al., 1998b
), Aquifex aeolicus
FtsZ (Oliva et al., 2007
), and TubZ (Aylett et al., 2010
) result in calculated RMSD values of 2.9, 2.6, and 2.9 Å, respectively (Figure S3
). The PhuZ structure is too divergent from αβ-tubulin to unambiguously determine if this represents a straight or curved conformation.
Structure and nucleotide binding of PhuZ
Although the tertiary structure of PhuZ is highly consistent with the structure of other tubulin family members, there are several notable differences. As in tubulins, PhuZ lacks the N-terminal extension present in both FtsZ and TubZ. Surprisingly, H6, which makes key longitudinal contacts in forming tubulin and FtsZ protofilaments (Downing and Nogales, 1998
) is missing in PhuZ. The absence of H6 leaves an acidic surface patch in its place (Figure S3D
). The PhuZ C-terminal domain is smaller in size than in other tubulin family members due to smaller loops and helices, especially H10. Like TubZ, PhuZ contains a long helix, H11, after the conserved C-terminal domain (Aylett et al., 2010
; Ni et al., 2010
The nucleotide-binding pocket is conserved and contains key catalytic residues required for polymer dynamics
Within the highly conserved nucleotide-binding pocket (), backbone nitrogen atoms in loops L1 and L4 coordinate the phosphates. While the GQxG motif in L1 is conserved across eukaryotic tubulins and TubZ, in PhuZ, L1 contains the sequence 11-GGTG-14, replacing the conserved Gln with a Gly, as in FtsZ (). As a consequence, the L1 loop has a tighter turn leading into H1. How this affects nucleotide binding is unclear. The GGGTGT/SG tubulin consensus sequence, or G-box, in L4 is also slightly varied to 91-GGGSGSV-97 in PhuZ (), although the presence of the Val side chain does not appear to affect the conserved structural elements (Figure S3
). As expected, the nucleotide base pi-stacks with Y161 and hydrogen-bonds with N16, N155, and N165 (). Residues S89, E122, N126, and Y161, plus two water molecules, all provide hydrogen-bonding interactions with the sugar. These interactions are consistent with the highly conserved nucleotide binding-mode of other tubulins.
The catalytic loop, T7, which normally inserts itself into the nucleotide-binding pocket of the preceding longitudinal monomer to aid in GTP hydrolysis, is modified from the GxxNxDxxD/E tubulin/FtsZ consensus sequence to NxxRxDxxD, although the key catalytic DxxD residues are conserved. To confirm the functional assignment of these aspartates, they were separately mutated to alanines. Similar mutations in TubZ and FtsZ compromise GTP hydrolysis but not GTP binding and as a result, the mutant proteins form long static polymers (Larsen et al., 2007
; Lu et al., 2001
). GFP-PhuZD187A (Figure S4A
) and D190A () mutants were expressed from the arabinose promoter in P. chlororaphis
. Both of the mutant proteins behaved similarly and were dramatically different from wild type. When expressed at low arabinose concentrations at 30°C (0.15%), both of the mutant proteins assembled short polymers in approximately 80% of the cells, suggesting that they had a lower threshold concentration for assembly (, S3
). There was also no detectable accumulation of diffuse fluorescence in the background of the cells and a significant percentage of cells (10% for D187A and 18% D190A) assembled filaments even when no arabinose was present, suggesting that even the smallest amount of expressed protein assembled polymers (, S3
). When expressed at higher levels (0.4% arabinose), both mutants formed long filaments that often chained the cells together (). In contrast to wild type filaments, the mutant filaments appeared relatively immobile in time-lapse experiments (Movie S2
). We used FRAP to quantitate turnover dynamics within these mutant filaments. Unlike wild-type filaments (Figure S4B
), no recovery or movement of the bleached zone over time (; Movie S3
) was observed in the mutants, even after extended periods, indicating that the filaments are completely static. These results are consistent with an essential role for these two residues for PhuZ GTP hydrolysis and polymerization dynamics.
PhuZ forms a filament in the crystal with longitudinal spacing consistent EM observations
shows PhuZ surrounded by its four symmetry-related molecules in the crystal, revealing two parallel protofilaments, related by 21
crystallographic symmetry. We propose that the intermolecular contacts, especially the longitudinal contacts, observed in the crystal are informative for those made within a PhuZ filament. Like all other tubulin-like proteins, the nucleotide resides at the longitudinal monomer-monomer interface. However, the longitudinal spacing of 47 Å between monomers observed in the crystal lattice is 3–7 Å longer than that of α/β-tubulin (Nogales et al., 1998b
), FtsZ (Oliva et al., 2004
; Oliva et al., 2007
), or TubZ (Aylett et al., 2010
), resulting in the smallest longitudinal interface seen among tubulins. In PhuZ, terminal side chain atoms of only ten residues in the longitudinal interface lose solvent accessible surface area due to crystal packing, burying only 188 Å2
/monomer as compared with typical values of 1666 and 1034 Å2
for α/β-tubulin and FtsZ respectively. While the gap between monomers is highly solvated, no single water directly contacts the two monomers, and no waters which bridge the two are resolved. This interaction mode is related to, but more extreme than the one observed for A. aeolicus
FtsZ, which has a smaller interaction surface than other FtsZs at 655 Å2
(Oliva et al., 2007
). In both of these cases, interactions between the intermediate domain of one monomer and H10 of the other are missing. In PhuZ, the residues of the catalytic T7 loop do not appear to be positioned correctly for hydrolysis, with the catalytic Asp side chains being displaced by more than 3 Å from where expected if they were to be able to interact with the γ-phosphate. This is likely a consequence of the structure containing GDP, and there may well be changes in local conformation or monomer packing upon GTP binding.
Crystal lattice contains filament-like contacts with the C-terminal tail providing most of the contact surface
Unlike eukaryotic tubulins, PhuZ does not make canonical lateral interactions. Instead, each PhuZ is rotated by 180° about an axis parallel to the filament and translated by 23.5 Å, resulting in interdigitated corner contacts defining a flat ribbon. Although roughly analogous to interactions in TubZ, the TubZ translation is significantly smaller and the rotation angle is ~190° (Aylett et al., 2010
), resulting in a helical filament reminiscent of actin-like polymers. More precisely, the lateral corner contacts between PhuZ monomers are defined by interactions of H3 on one monomer with H4 and H5 on another (), with this interaction occurring twice so as to interact with two lateral PhuZ monomers, burying a total surface area of 476 Å2
per monomer, stabilizing the connection between the two longitudinal protofilaments.
To assess the relevance of the putative filaments formed within the crystal lattice we compare more closely to the filaments seen by negative-stain EM (). Two-dimensional averages of 1000 defined segments were generated to produce a reliable view of monomer packing. Although the filament architecture is not yet clear, the average reveals a PhuZ filament morphology similar to that observed in the crystals. Of particular importance, the longitudinal spacing derived from EM (~45 Å) is consistent with the 47 Å spacing in the crystal, supporting our hypothesis that the crystal lattice provides a suitable model for interactions stabilizing filament formation. This observed spacing is longer than observed for other tubulins, 40–42 Å, and would require a compaction of the lattice in order for the catalytic residues of the T7 loop to come into position for nucleotide hydrolysis. It is possible that the crystal lattice represents an expansion of the filament lattice that would occur after GTP is hydrolyzed to GDP.
The unique PhuZ C-terminal tail makes extensive interactions required for polymer formation
By contrast with the minimal direct longitudinal interface, the 21 C-terminal residues make extensive contacts with the neighboring longitudinal monomer, with a total buried surface area of 1226 Å2/monomer. Many of these contacts are driven by either electrostatic or polar interactions (). The 13 most C-terminal residues of the protein contain six acidic residues (D303, D305, D306, D309, D311, E310) forming an acidic “knuckle” that is inserted into a basic patch of the longitudinal symmetry mate formed by helices H3, H4, and H5, containing R60, R68 and K135. Non-polar residues of the knuckle interact with L64, L104, and I140 on the symmetry mate providing further stabilization. While the most C-terminal residues make the most extensive contacts, significant interactions are also provided by the extended 8 residues that lie between Helix-11 and the knuckle, including significant interactions of R298 with E138, Q297 with Q207, and F295 with I227. These residues, especially the aspartic acids of the acid knuckle, are also conserved in the other PhuZ sequences (), suggesting that other phage tubulins may contain a similar C-terminal tail.
Given the extensive interactions contributed by the C-terminal tail, and the otherwise rather limited interactions that stabilize the longitudinal interface, the tail is likely quite important for polymer formation. To test this, we made two point mutants (R298A, D311A) to disrupt salt bridges as well as a truncation mutant, ΔI302 that removed the last 13 residues (knuckle region) and examined the functional consequences in vitro by light scattering (). Even at a concentration of 20 µM and in the presence of 1 mM GTP, PhuZ-ΔI302 was unable to form detectable polymer in vitro, whereas wild-type PhuZ polymerizes efficiently at concentrations above 5 µM (). Similarly, both the R298A and D311A mutants, which disrupt salt bridge and H-bond formation with H5 and H11 and H3, respectively, also compromise in vitro filament formation, with no detectable polymerization at 20 µM ().
These mutants were also tested for their ability to form polymers in vivo
by expressing them in P. chlororaphis
. GFP fusion proteins containing point mutations (D311A and R298A) in the tail were severely impaired for assembly and only formed polymers at the highest expression levels (1% or 2% arabinose, ). The C-terminal tail truncation (ΔI302) completely abolished filament formation in vivo
at all expression levels (). Using in-gel fluorescence we demonstrated that all of the C-terminal fusions were stably produced at the expected levels in vivo
). These findings demonstrate the importance of the C-terminal 21 residues of PhuZ in polymerization.
PhuZ assembles a dynamic spindle-like array during phage lytic growth
No specific role in the life cycle of a phage has ever been ascribed to a tubulin cytoskeletal protein. One anticipated function for PhuZ is as a DNA segregation system during lysogeny if 201ϕ2-1 replicates separately from the chromosome like a plasmid. While an attractive hypothesis, so far we have been unable to obtain lysogens of 201ϕ2-1 in P. chlororaphis
. Therefore, we sought to determine if PhuZ was expressed and assembled polymers during lytic growth. First, we used RT-PCR to show that phuZ
mRNA accumulates at two hours post infection (Figure S5A
). Second, we devised a microscopic single cell assay to determine if PhuZ assembles polymers during an infection cycle. To accomplish this, P. chlororaphis
cells were grown on an agarose pad at 30°C and GFP-PhuZ was expressed from the arabinose promoter below its critical threshold for assembly (0.15% arabinose at 30°C). We then infected cells with phage and performed time-lapse microscopy in which GFP-PhuZ assembly, phage production and phage-mediated cell lysis were simultaneously monitored. Since GFP-PhuZ does not spontaneously assemble polymers at this expression level, polymers would only be observed if additional PhuZ (or a regulator of PhuZ assembly) was expressed by the phage. By including DAPI and DNaseI in the pad, the release of phage upon cell lysis could be visualized. DNaseI degrades any remaining cellular DNA but not DNA packaged within viral capsids. At the terminal time point, after cells had lysed, we captured images of DAPI fluorescence, allowing the number of released phage particles to be counted. Cell lysis was detected using the membrane dye FM 4–64, which only faintly stains wild type P. chlororaphis
but intensely stains cell debris.
In the first example (), GFP-PhuZ formed diffuse fluorescence at the beginning of the experiment (15 min after the addition of phage). Within 56 minutes, GFP-PhuZ assembled a polymer that extended from pole to pole (). This cell maintained at least one polymer for the next 175 minutes, at which point the cell lysed. DAPI staining alongside DNaseI treatment confirmed that this cell had released phage particles, indicating that lysis was phage induced.
A single cell assay for phage infection reveals that PhuZ assembles filaments in vivo during infection of the host cell with 201ϕ2-1
In the example shown in , GFP-PhuZ was expressed at a slightly higher level (0.25% arabinose) to allow for brighter images and more frequent time-points. At early time-points (18 min after phage addition) fluorescence was uniform, but over time, one (39 min) and then multiple (59 min) filaments formed (). Filaments were very dynamic (, Movie S4
), undergoing cycles of assembly and disassembly, with at least one filament always assembled until the cell ultimately lysed.
To gain additional insight into the role of PhuZ polymers during lytic growth, we simultaneously visualized GFP-PhuZ and DNA in fixed cells that had been stained with DAPI. During lytic growth, cells became elongated and formed an unusual bulge at the cell midpoint (). DAPI staining revealed that the central bulge contained a high concentration of DNA, which we refer to as the "infection nucleoid", while the rest of the cell contained very little DNA (). In comparison, uninfected cells contained one or two vegetative nucleoids that filled the majority of the cytoplasm (Figure S5B
). Quantitation showed that most cells contained just a single infection nucleoid () that was located within 5% of the middle of the cell in 80% of cells and within 10% of the middle in 98% of cells (). PhuZ filaments frequently appeared to make contact with the edge of the infection nucleoid, forming an array on either side of this structure (). Multiple filaments of various lengths (ranging from 0.2 to 2µm) were observed in fixed cells (), as might be expected for a population of cells containing dynamic polymers trapped in various states of polymerization.
To determine if the centrally located DNA masses contained phage encapsidated DNA, we digested them with DNAseI. As shown in the examples in , upon DNaseI treatment, much of the centrally located DNA was degraded, leaving DNAseI resistant foci indicative of phage encapsidated DNA. PhuZ filaments were extended on each side of these DNA foci (). Optical sectioning revealed that these phage encapsidated DNA molecules occurred in a rosette-like structure at the edges of the digested nucleoid (; Movie S5
), suggesting that phage DNA occurs in an organized structure at the cell mid point.
Dynamic filaments are required for phage positioning and maximal burst size
Since the majority of phage infected cells contained PhuZ polymers that extended on each side of the centrally located infection nucleoid, we speculated that PhuZ participates in DNA organization or positioning. To test this idea, we examined DNA positioning in cells expressing either wild type GFP-PhuZ or a mutant version (GFP-PhuZD190A) that we demonstrated () assembles static polymers in vivo
. In other tubulins, including TubZ (Larsen, et. al. 2007
) and FtsZ (Lu, et.al. 2001
), catalytic mutants defective in GTP hydrolysis co-assemble with the wild type and behave as dominant negatives. Positioning of the nucleoid during infection was severely affected by expression of GFP-PhuZD190A; only 39% of mutant cells positioned the infection nucleoid within 5% of the middle (compared to 80% for wild type; ). In many cells, the PhuZD190A filaments appeared to make contact with the edge of an infection nucleoid that was mispositioned close to the cell pole (). Significant mispositioning occurred in the mutant cells regardless of their size (). In addition, while 94% of cells infected in the presence of wild type PhuZ had a single large nucleoid, more than a third of infected cells expressing PhuZD190A contained either two or three nucleoids (), typically present at random positions, further indicating disruption of DNA localization. Taken together, these results suggest that PhuZ assembles a dynamic cytoskeletal element that functions to position phage at the cell midpoint during phage lytic growth.
To assess the importance of PhuZ to phage yield, we attempted population based phage growth curves, but phage infections rates were too low to make the results interpretable. We therefore performed single cell infection assays and found a significant decrease in burst size when cells expressed the PhuZD190A catalytic mutant, from an average of 16 phage per cell for wild type (n=25) to an average of 7 (n=25) for the mutant (p=0.0001). Proper phage centering by PhuZ thus contributes significantly to the efficiency of phage production. Such a 50% reduction in yield would be a significant evolutionary disadvantage.