ICP0 Dismantles Microtubule Networks in Herpes Simplex Virus-Infected Cells

Infected-cell protein 0 (ICP0) is a RING finger E3 ligase that regulates herpes simplex virus (HSV) mRNA synthesis, and strongly influences the balance between latency and replication of HSV. For 25 years, the nuclear functions of ICP0 have been the subject of intense scrutiny. To obtain new clues about ICP0's mechanism of action, we constructed HSV-1 viruses that expressed GFP-tagged ICP0. To our surprise, both GFP-tagged and wild-type ICP0 were predominantly observed in the cytoplasm of HSV-infected cells. Although ICP0 is exclusively nuclear during the immediate-early phase of HSV infection, further analysis revealed that ICP0 translocated to the cytoplasm during the early phase where it triggered a previously unrecognized process; ICP0 dismantled the microtubule network of the host cell. A RING finger mutant of ICP0 efficiently bundled microtubules, but failed to disperse microtubule bundles. Synthesis of ICP0 proved to be necessary and sufficient to disrupt microtubule networks in HSV-infected and transfected cells. Plant and animal viruses encode many proteins that reorganize microtubules. However, this is the first report of a viral E3 ligase that regulates microtubule stability. Intriguingly, several cellular E3 ligases orchestrate microtubule disassembly and reassembly during mitosis. Our results suggest that ICP0 serves a dual role in the HSV life cycle, acting first as a nuclear regulator of viral mRNA synthesis and acting later, in the cytoplasm, to dismantle the host cell's microtubule network in preparation for virion synthesis and/or egress.


Introduction
Herpes simplex virus 1 (HSV-1) is a large, dsDNA virus that is capable of alternating between two programs of gene expression that lead to a productive or silent infection. One of HSV's immediate-early (IE) proteins, ICP0, is a positive regulator whose synthesis represents a key step by which HSV ''decides'' whether or not an infection is likely to culminate in the production of new infectious virus (reviewed in Ref. [1]). For example, ICP0 is sufficient to trigger HSV reactivation in latently infected trigeminal ganglion neurons [2,3]. At multiplicities of infection (MOI) above 1 pfu per cell, HSV ICP0-deficient (ICP0 2 ) viruses replicate to nearly wild-type levels. In contrast, at MOIs below 0.1 pfu per cell, the same ICP0 2 viruses establish quiescent infections in 99% of the cells they infect [4,5,6].
Point mutations in ICP0's RING-finger domain (amino acids 116 to 156) destroy ICP0's E3 ligase activity and destroy ICP0's capacity to promote HSV replication [11]. It remains unclear which substrate(s) explain how ICP0's E3 ligase activity promotes HSV replication and spread. Although ICP0 triggers the efficient dispersal of PML nuclear bodies, ICP0 does not ubiquitinate the PML protein in an in vitro E3 ligase assay [17].
Like many laboratories, we have been interested in learning how ICP0 influences outcomes of HSV infection. Our most recent study clarifies that ICP0 physically interacts with HSV's major transcriptional regulator, ICP4, and suggests that ICP0 influences whether ICP4 functions predominantly as an activator or a repressor of HSV mRNA synthesis [9]. However, ICP0's interaction with ICP4 does not explain 1. how ICP0 triggers the dispersal of PML nuclear bodies and centromere proteins [14,18], nor does it explain 2. why synthesis of ICP0 causes cells to arrest in the G2/M phase of the cell cycle [19,20]. These latter observations suggest that ICP0 must interact with at least one cellular protein. Rather than interrogate specific proteins for their capacity to interact with ICP0, we chose to use live-cell imaging to determine if new clues might be obtained by tracking the distribution of a green fluorescent protein (GFP)-tagged form of ICP0 in HSV-infected cells over time.
Three ICP0 + viruses were constructed that bore an ,750 bp insertion of GFP coding sequence inserted in the ICP0 gene. The resulting recombinant viruses, HSV-1 0 + GFP 12 , HSV-1 0 + GFP 24 , and HSV-1 0 + GFP 105 each synthesized a 3.5 kb ICP0 +GFP mRNA and a 140 kDa protein. Each GFP-tagged ICP0 protein retained much of ICP0's activity, and was visible in HSV-1 infected cells by fluorescent microscopy. Contrary to our initial expectations, the majority of GFP-tagged ICP0 was observed in the cytoplasm of HSV-1 infected cells. Subsequent tests verified that wild-type ICP0 also accumulated to much higher levels in the cytoplasm than in the nucleus of virus-infected cells.
Since the discovery that ICP0 potently stimulates HSV mRNA synthesis [7,8], sporadic reports have documented the presence of ICP0 in the cytoplasm of HSV-infected cells, or have described ICP0 translocating from the nucleus to the cytoplasm of HSVinfected cells [21,22,23,24,25]. In recent years, ICP0's presence in the cytoplasm of HSV-infected cells has become undeniable, and it is now well established that ICP0 is incorporated into the tegument of HSV-1 virions [26,27,28,29]. However, it remains unclear if ICP0 mediates a specific function in the cytoplasm of HSV-1 infected cells, or if ICP0 translocates to the cytoplasm for the sole purpose of its incorporation into HSV-1 virions [26,27,28,29]. Therefore, most studies of ICP0 continue to focus on deciphering how this E3 ligase functions in the nucleus to stimulate HSV mRNA synthesis [9,30,31,32,33,34].
The results of the current study clarify that ICP0 is an exclusively nuclear protein during the IE phase of HSV infection, but translocates to the cytoplasm during the early (E) phase. Once in the cytoplasm of HSV-infected cells, ICP0 efficiently bundles and disperses host cell microtubules. Microtubule networks are known to be disrupted in HSV-infected cells [35], but the effectors that mediate this process are unknown. The HSV-1 tegument protein VP22 bundles microtubules when overexpressed in transfected cells [36], but this is not observed in HSV-infected cells [37]. Our results demonstrate that ICP0 is necessary and sufficient to trigger the complete disassembly of the host cell microtubule network. This finding adds to a growing list of proteins encoded by plant and animal viruses that reorganize microtubules [38,39,40,41]. This is the first report of a viral E3 ligase that regulates microtubule stability. Intriguingly, cellular E3 ligases such as the anaphase-promoting complex and cullin 3 orchestrate massive reorganizations of microtubules during mitosis [42,43,44]. ICP0like E3 ligases are not unique to herpes simplex virus, but are encoded by at least 20 other a-herpesviruses that infect primates, pigs, dogs, kangaroos, and other species [45,46]. The results of the current study suggest that these a-herpesvirus-encoded E3 ligases fulfill a dual function of 1. immediately stimulating viral gene expression in the nucleus, and 2. acting later, in the cytoplasm, to dismantle the host cell's microtubule network in preparation for virion synthesis and/or egress.
These observations raised questions both about the requirements for ICP0's nuclear-to-cytoplasmic translocation, and what function ICP0 might fulfill in the cytoplasm of HSV-1 infected cells. Further experiments were performed to address these questions.  Cytoplasmic translocation of ICP0 is an early event in the HSV-1 replication cycle Although HSV-1 ICP4 2 null viruses overexpress viral IE proteins, a considerable leak of E proteins such as the large subunit of HSV's ribonucleotide reductase (ICP6) occurs in the absence of ICP4 [48,49]. To determine if HSV-1 IE proteins were indeed sufficient to promote GFP-tagged ICP0's translocation to the cytoplasm, chemical inhibitors were used to unambiguously separate HSV-1 protein synthesis into its IE and E phases. Vero cells were treated with cycloheximide to block protein translation, and cells were inoculated with 5 pfu per cell of HSV-1 0 + GFP 105 . After allowing 6 hours for the accumulation of viral IE mRNAs, cycloheximide was removed and cultures were released into medium containing either the mRNA synthesis inhibitor actinomycin D [50], the DNA synthesis inhibitor acyclovir [51] or no inhibitor (vehicle).
In HSV-1 0 + GFP 105 -infected cells released into medium containing actinomycin D, ICP0 +GFP-105 remained in the nuclei of .90% of cells at 1 and 4 hours post-release from a cycloheximide block (Fig. 4A). Thus, translocation of ICP0 +GFP-105 failed to occur when only viral IE proteins were efficiently expressed. When HSV-1 0 + GFP 105 -infected cells were released into medium containing acyclovir (IE and E proteins expressed), ICP0 +GFP-105 was initially observed in nuclei, but translocated to the cytoplasm of most cells by 4 hours post-release (Fig. 4A). Likewise, when cycloheximide treatment was followed with vehicle, ICP0 +GFP-105 translocated to the cytoplasm with similar kinetics (Fig. 4A). Parallel tests with HSV-1 0 + GFP 12 , 0 + GFP 24 , and 0 + GFP 24 -D4 yielded equivalent results (not shown). These results suggested that the nuclear-to-cytoplasmic translocation of ICP0 +GFP-105 occurred during the E phase of HSV-1 replication.
To determine if these results were relevant to wild-type ICP0, similar tests were performed with wild-type HSV-1 strain KOS [52]. When cycloheximide treatment was followed with actinomycin D, wild-type ICP0 was retained in the nuclei of KOSinfected cells at 4 hours post-release (Fig. 4B). In contrast, when cycloheximide treatment was followed with acyclovir or vehicle, wild-type ICP0 translocated to the cytoplasm by 4 hours postrelease and was observed in globular and linear structures that encircled the nuclei of .90% of KOS-infected cells (Fig. 4B). Therefore, both GFP-tagged ICP0 and wild-type ICP0 were predominantly nuclear during the IE phase of viral infection, but rapidly translocated to the cytoplasm when viral E protein synthesis was allowed to occur. The cytoplasmic structures in which ICP0 accumulated did not appear to be mitochondria, lysosomes, Golgi apparatus, or endoplasmic reticulum based upon the failure of ICP0 to colocalize with MitoTracker dye [53], LysoTracker dye [54], Golgi marker b-COP [55], or the endoplasmic reticulum marker calreticulin [56]. Moreover, in early attempts to isolate the fluorescent-ICP0 labeled cytoplasmic bodies by density gradient sedimentation, it was noted that the bodies rapidly dispersed upon homogenization of cells in ice-cold buffer (not shown). To gather further clues about the nature of these structures, live-cell imaging was used to study the dynamics of accumulation of ICP0 +GFP-105 in the cytoplasm of HSV-1 0 + GFP 105 -infected cells.
We considered the possibility that ICP0 +GFP-105 might trigger the dispersal of cytoplasmic structures in much the same manner that ICP0 triggers the dispersal of PML nuclear bodies [13,14]. To test this hypothesis, a recombinant virus was constructed, HSV-1 0DRING, which encoded a GFP-tagged ICP0 DRING protein that was deleted of amino acids 105-221, and thus lacked ICP0's RING-finger domain [10]. When cycloheximide-release experiments were performed in cells inoculated with HSV-1 0DRING, the ICP0 DRING protein accumulated in linear cytoplasmic structures between 2 and 6 hours post-release (Table 1, Fig.  S1B). Intriguingly, ICP0 DRING was rarely observed in globular bodies, but rather the protein was observed almost exclusively in linear cytoplasmic structures (Table 1, Fig. S1B). These findings suggested that ICP0 DRING stably accumulated in linear cytoplasmic structures, whereas ICP0 +GFP-105 triggered the dispersal of these structures upon accumulating at these sites.
To test this hypothesis, live-cell imaging was used to track the fate of linear structures in which ICP0 +GFP-105 accumulated. Cells were inoculated with HSV-1 0 + GFP 105 and time-lapse photography was conducted between 4.0 and 4.5 hours post-release from a cycloheximide block. Linear, cytoplasmic arrays of ICP0 +GFP-105 that initially encircled the nucleus were found to be unstable, and dispersed into smaller, globular bodies over a 10-to 20-minute time frame (Fig. 5A, Video S2). In contrast, experiments with HSV-1 0DRING demonstrated that the ICP0 DRING protein stably accumulated in linear cytoplasmic structures which grew visibly in length and thickness between 4.0 and 4.5 hours post-release ( Fig. 5B, Video S3). This trend continued until 6 hours postrelease, at which time the experiment was terminated (not shown). Therefore, ICP0 +GFP-105 translocated to the cytoplasm, accumulated in linear cytoplasmic structures, and then triggered the dispersal of these structures in a RING-finger-dependent manner.

ICP0 co-localizes with disrupted microtubule networks in HSV-1 infected cells
The linear cytoplasmic structures in which ICP0 +GFP-105 transiently accumulated were reminiscent of microtubule bundles formed by HSV-1's VP22 protein [36]. Therefore, we considered the possibility that ICP0 +GFP-105 and ICP0 DRING proteins might associate with host cell microtubules. To test this hypothesis, cycloheximide-release experiments were performed in cells inoculated with HSV-1 0DRING, HSV-1 0 + GFP 105 , or wild-type HSV-1 KOS. At 4 hours post-release, cells were fixed and immunofluorescently stained for a-tubulin and ICP0. In uninfect-ed Vero cells, a-tubulin staining revealed a normal network of microtubules that radiated from the microtubule-organizing center at the periphery of the nucleus (Fig. 6). In cells inoculated with HSV-1 0DRING, a-tubulin staining revealed extensive thickening of microtubules into elongated bundles that circumscribed the nucleus, and co-localized with the ICP0 DRING protein (Fig. 6). In cells inoculated with HSV-1 0 + GFP 105 , a-tubulin staining revealed that the microtubule network had dispersed into small a-tubulin + , globular bodies that co-localized with ICP0 +GFP-105 (Fig. 6). In cells inoculated with wild-type HSV-1 KOS, the host cell microtubule network was also disrupted, and a-tubulin was dispersed into a-tubulin + globular bodies that co-localized with ICP0 (Fig. 6). Although the ICP0 DRING protein bundled microtubules, it failed to trigger their dispersal into globular bodies (Fig. 6). These results raised the possibility that HSV's E3 ligase, ICP0, might contribute to microtubule disassembly in HSV-infected cells [35]. Further experiments were conducted to test the validity of this hypothesis.
ICP0 DRING protein accumulates in linear structures that are nocodazole-sensitive If microtubules did indeed provide the underlying scaffold on which ICP0 DRING protein accumulated in linear arrays (Fig. 5, 6), then inhibition of microtubule polymerization with nocodazole [57] should recapitulate the effect of ICP0's RING finger domain. Specifically, nocodazole treatment, like wild-type ICP0, should  trigger microtubule disassembly and cause linear accumulations of ICP0 DRING protein to disperse into smaller bodies. To test this prediction, cells were infected with HSV-1 0DRING in the presence of cycloheximide for 10 hours, and were released into medium containing no inhibitor. Four hours later, the stability of linear arrays of ICP0 DRING protein was compared in the presence or absence of nocodazole.
In the absence of nocodazole, ICP0 DRING protein stably accumulated in linear structures that encircled the nucleus of HSV-1 0DRING-infected cells between 4.0 and 4.5 hours postrelease (Fig. 7A, Video S3). In contrast, nocodazole treatment caused linear accumulations of ICP0 DRING protein to disperse into small globular bodies within 10 to 30 minutes, and the time required for dispersal varied in proportion to the thickness of each bundle (Fig. 7B, Video S4). Nocodazole-induced, ICP0 DRING+ globular bodies were identical in appearance to a-tubulin + globular bodies that co-localized with ICP0 and ICP0 +GFP-105 (Fig. 6). Independent tests verified that nocodazole-induced, ICP0 DRING+ globular bodies were also a-tubulin + (not shown).
These observations indicated that the linear cytoplasmic structures in which ICP0 DRING protein accumulated were indeed microtubule bundles. This finding was also consistent with our prior observation that ICP0 +GFP-105 -containing bodies rapidly dispersed upon homogenization in ice-cold buffer; a condition known to cause depolymerization of microtubules [58].

ICP0 is necessary for efficient dispersal of microtubules in HSV-1 infected cells
It is well established that the host cell microtubule network is disrupted during the course of HSV-1 infection [35,59,60]. However, it is unclear what process triggers microtubule reorganization in HSV-1 infected cells. We questioned whether this event might be ICP0-dependent. To test this hypothesis, atubulin and ICP0 staining were compared in Vero cells that were uninfected (UI) or were inoculated with HSV-1 KOS, HSV-1 0 + GFP 105 , or an ICP0 2 null virus, HSV-1 0 2 GFP. Cells were inoculated with 5 pfu per cell of each virus in the presence of cycloheximide for 10 hours to allow high and equivalent levels of ICP0 + and ICP0 2 mRNAs to accumulate [9]. At 1 and 4 hours post-release, a normal distribution of a-tubulin staining was noted in uninfected Vero cells; specifically, microtubules radiated from the microtubule-organizing center on one side of the nucleus (Fig. 8A). At 1 hour post-release, ICP0 and ICP0 +GFP-105 were observed in the nuclei of cells infected with HSV-1 KOS and 0 + GFP 105 (Fig. 8A). Importantly, a-tubulin staining was always normal when ICP0 or ICP0 +GFP-105 were confined to the nucleus (Fig. 8A). However, at 4 hours post-release, ICP0 and ICP0 +GFP-105 had translocated to the cytoplasm and this event coincided with dispersal of microtubules in KOS-and 0 + GFP 105 -infected cells, respectively (Fig. 8B). Specifically, microtubule-organizing centers were no longer discernible, and a-tubulin was dispersed into globular bodies that co-localized with ICP0 or ICP0 +GFP-105 (Fig. 8B). In contrast, at 1 and 4 hours post-release, cells inoculated with HSV-1 0 2 GFP retained an intact microtubule network that radiated from a perinuclear microtubule-organizing center (Fig. 8A, 8B). Likewise, at 6 and 8 hours post-release, microtubule networks remained intact in HSV-1 0 2 GFP-infected cells (Fig. S2). Therefore, synthesis of ICP0 appeared to be necessary for the efficient dismantling of microtubule networks that normally occurs in HSV-1 infected cells [35].

ICP0 is sufficient to dismantle microtubule networks in transfected cells
The HSV-1 proteins US11 [61], UL34 [62], VP26 [63] and VP22 [36] interact with host cell microtubules. Thus, ICP0 might be one of many HSV proteins that is required to disperse the intracellular 'freeway system' that is the cell's microtubule network [64]. Alternatively, synthesis of ICP0 alone might be sufficient to disperse cellular microtubules. To differentiate between these possibilities, Vero cells were transfected with pICP0, p0 + GFP 105 , or p0DRING to determine if synthesis of ICP0 proteins affected the subcellular distribution of a-tubulin, one of the principal structural subunits of the hollow, polymeric tubes known as microtubules [65].
A normal distribution of a-tubulin staining was noted in mocktransfected Vero cells at 12 and 24 hours post-transfection. Specifically, microtubules radiated from the organizing center on one side of the nucleus, and a-tubulin staining was not observed in the nuclei of mock-transfected Vero cells (Fig. 9A, 9B). Consistent with prior results ( Fig. 2A-2C), wild-type ICP0 and ICP0 +GFP-105 were observed in the nuclei of cells transfected with pICP0 and p0 + GFP 105 at 12 hours post-transfection (Fig. 9A). At this early time, a normal distribution of a-tubulin staining was noted in the cytoplasm, but to our surprise significant amounts of a-tubulin were present in the nuclei and co-localized with wild-type ICP0 and ICP0 +GFP-105 Figure 6. ICP0 co-localizes with reorganized microtubules in HSV-1 infected cells. Immunofluorescent staining of Vero cells that were uninfected (no virus) or that were inoculated with 5 pfu per cell of HSV-1 0DRING, HSV-1 0 + GFP 105, or wild-type HSV-1 strain KOS. Cells were inoculated in the presence of 200 mM cycloheximide from 20.5 to 10 hours p.i., and were released into medium containing no drugs. At 4 hours post-release, uninfected cells and KOS-infected cells were fixed and stained with antibodies against a-tubulin (rabbit IgG, Alexa Fluor 594) and ICP0 (mouse IgG, fluorescein). Nuclei were counterstained with the DNA-binding dye Hoechst 33342. HSV-1 0DRING and 0 + GFP 105 -infected cells were fixed and stained for a-tubulin, and the ICP0 DRING and ICP0 +GFP-105 proteins were visualized using their GFP fluorophores. The scale bar denotes a distance of 10 mm. doi:10.1371/journal.pone.0010975.g006 (Fig. 9A, Fig. S3). Control staining with individual antibodies verified that the appearance of co-localization was not an unintended consequence of bleedover between the red and green channels (not shown). In cells transfected with p0DRING, the ICP0 DRING protein was also observed in the nuclei of cells at 12 hours post-transfection, and a normal distribution of a-tubulin staining was noted in the cytoplasm. However, large amounts of a-tubulin were present in the nuclei of transfected cells, and a-tubulin co-localized with the ICP0 DRING protein (Fig. 9A, Fig. S3).
At 24 hours post-transfection, wild-type ICP0 and ICP0 +GFP-105 were observed in the nuclei and cytoplasm of cells transfected with pICP0 and p0 + GFP 105 , respectively (Fig. 9B). Translocation of ICP0 or ICP0 +GFP-105 to the cytoplasm consistently correlated with 1. dissolution of the host cell microtubule-organizing center and microtubule network, and 2. co-localization of a-tubulin with ICP0 or ICP0 +GFP-105 in small, globular bodies in the cytoplasm ( Fig. 9B; Fig. S3). Likewise, 24 hours after transfection with p0DRING, ICP0 DRING protein co-localized with a-tubulin in elongated microtubule bundles ( Fig. 9; Fig. S3). Therefore, synthesis of the ICP0 DRING protein was sufficient to trigger bundling, but not dispersal of microtubules. In contrast, wild-type ICP0 was sufficient to trigger a complete dispersal of host cell microtubules.

ICP0 translocation and microtubule dispersal in a single HSV-1 plaque
High MOIs and cycloheximide-release experiments allowed robust visualization of ICP0 and co-localization of ICP0 with dispersed a-tubulin. These tests left unaddressed the question of whether or not ICP0 actually translocates and/or dismantles microtubule networks during the normal progression of HSV-1 infection. To address this issue, Vero cells were inoculated with wild-type virus at an MOI of 0.0001 pfu per cell, and the distribution of ICP0 and a-tubulin was analyzed in isolated HSV-1 plaques at 40 hours p.i. A single, representative plaque is considered to illustrate our findings (Fig. 10, Fig. S4).
With the enhanced sensitivity of immunofluorescent staining (versus direct imaging of GFP fluorescence, Fig. 2), low levels of ICP0 were detectable in the nuclei of cells at the outermost, advancing edge of HSV-1 plaques (N symbols, Fig. 10A, Fig. S4A). In those HSV-1 infected cells in which ICP0 was observed solely in the nucleus, the perinuclear pattern of microtubule staining remained normal (N symbols, Fig. 10B, Fig. S4B). Behind this advancing front, a second row of HSV-1 infected cells was observed in which ICP0 was far more abundant and predominantly localized to the cytoplasm of HSV-1 infected cells (X symbols, Fig. 10A, Fig. S4A). When ICP0 accumulated in the cytoplasm of HSV-1 infected cells, gross reorganization of the cellular microtubule network was observed (X symbols, Fig. 10B, Fig. S4B). Moreover, a-tubulin was frequently observed reorganized into linear structures or globular bodies that colocalized with wild-type ICP0 (white arrows, Fig. 10, Fig. S4). Therefore, we conclude that regardless of MOI, a nuclear-tocytoplasmic translocation of ICP0 routinely occurs in HSV-1 infected cells, and the timing of ICP0's translocation to the cytoplasm coincides with a massive reorganization of the host cell's microtubule network.

ICP0 serves distinct roles in the nucleus and cytoplasm of HSV-infected cells
Studies of ICP0 have focused on the protein's role in the nucleus [18,30,31,32]. Cytoplasmic ICP0 has been repeatedly observed [21,22,23,24,25], but its significance remains obscure for two reasons. First, the kinetics of ICP0's nuclear-to-cytoplasmic translocation has been poorly defined. Thus, it has not been apparent that ICP0's function in the cytoplasm is kinetically delayed and is only manifest once ICP0 fulfills its IE function in the nucleus [7,8,9]. Second, the role that ICP0 fulfills in the cytoplasm of HSV-infected cells has not been clearly articulated.
The results of the current study address these gaps in knowledge.
The results of the current study demonstrate that once E proteins accumulate, ICP0 translocates to the cytoplasm (Fig. 4) and there dismantles the host cell's microtubule network. Five observations support this conclusion: 1. Microtubule dispersal is ICP0-dependent in HSV-infected cells (Fig. 8); 2. Synthesis of ICP0 is sufficient to trigger microtubule dispersal (Fig. 9); 3. Dispersal of microtubule bundles is dependent upon ICP0's RING finger domain (Fig. 6); 4. The timing of microtubule dispersal coincides with ICP0's translocation to the cytoplasm (Fig. 8); and 5. dispersed a-tubulin co-localizes with ICP0 (Fig. S3).

GFP-tagged ICP0 yields new insights into the biology of ICP0
HSV-1 viruses that encoded chimeric ICP0 +GFP proteins formed plaques with ,66% efficiency relative to wild-type HSV-1 (Fig. 1E). Using this new tool, we were able to observe GFP-tagged ICP0 in the act of bundling and/or dispersing microtubules in the cytoplasm of HSV-infected cells (Videos S2, S3). These findings would be difficult to ascertain from analysis of fixed cells because ICP0's cytoplasmic pattern is pleomorphic in HSV-infected cells. However, direct observation of a single HSV-infected cell in real time revealed the fluid process by which GFP-tagged ICP0 triggered the disassembly of subcellular structures that proved to be bundled microtubules (Figs. 5 and 6; Table 1; Video S2).

Virus-induced reorganization of host cell microtubules
Many animal and plant viruses use host cell microtubule networks during their replication cycle [39,40,41]. For example, movement protein of tobacco mosaic virus [66] and transmission factor of cauliflower mosaic virus [67] modify microtubules in plant cells. The human immunodeficiency virus Rev protein forms dimers that bind microtubule ends, and inhibit their polymerization [68]. The 3C protease of foot-and-mouth disease virus cleaves microtubule-associated protein 4, and thus excludes microtubules from cytoplasmic replication compartments [38,69]. Expression of Epstein-Barr virus's LMP-1 protein or SV40's large T antigen causes formation of aberrant microtubule structures via the respective modulation of microtubule-stabilizing proteins RASSF1 A [70] and TACC2 [39]. While many viruses modify the microtubule network during their replication cycles, the functional significance of these interactions is often unclear. Thus, there is no clear virological precedent that explains why HSV should encode a protein, ICP0, that dismantles the microtubule network.

ICP0: a viral E3 ligase that orchestrates microtubule disassembly
The current study provides the first example of a viral E3 ligase that triggers microtubule disassembly. The genomes of at least 16 a-herpesviruses encode RING-finger E3 ligases, and these proteins constitute a family of ICP0-like proteins that are functional homologs of one another [45]. Therefore, it is reasonable to expect that the ICP0-like proteins of all a-herpesviruses should likewise fulfill dual roles in the nucleus and cytoplasm of virus-infected cells [9], and thus should trigger disassembly of microtubule networks upon their translocation to the cytoplasm. Further testing will be required to validate this prediction.
The a-herpesviruses are not alone in their use of E3 ligases as regulatory molecules. The Rta protein of Kaposi's sarcoma herpesvirus possesses intrinsic E3 ligase activity despite the absence of a canonical RING finger domain [71]. Smaller RNA and DNA viruses exploit ubiquitination as a regulatory mechanism by encoding adapter proteins such as adenovirus E1b, papillomavirus E6 and E7, or paramyxovirus V proteins that redirect cellular E3 ligases to substrates that benefit the virus [71,72,73,74,75]. It will be of interest to determine in coming years if any of these other viral E3 ligases reorganize host cell microtubule networks.
Many cellular E3 ligases play a prominent role in coordinating the complex events that transpire during mitosis, and which include disassembly of the microtubule network [43]. For example, cullin 3 is an E3 ligase that regulates the activity of a microtubule- severing protein known as katanin (i.e., 'sword' in Japanese; Ref. [43,44,76]). Given the propensity of viruses to steal useful cellular functions [71,77], our results raise the possibility that ICP0 may be functionally homologous to cellular E3 ligases that influence cellcycle progression by regulating the stability of microtubule networks [43]. Further investigation will be required to explore this important hypothesis.
Is a-tubulin a substrate of ICP0's E3 ligase activity?
The discovery of ICP0's E3 ligase activity by Everett and colleagues [10,78] arose from a need to explain why PML nuclear bodies were dispersed shortly after ICP0 arrived in these subnuclear structures [13,14,15]. Although it seemed likely that PML would prove to be a direct substrate of ICP0's E3 ligase activity, the development of a robust in vitro assay for ICP0's E3 ligase activity demonstrated that this was not the case [10,17].
It is possible that a-tubulin may be a direct substrate of ICP0's E3 ligase activity, but the history of ICP0 research suggests that such speculation is premature. What should be considered, however, is the extraordinarily robust co-localization of a-tubulin and ICP0. For example, in transfected cells, large amounts of atubulin co-localized with ICP0 in the nucleus, which is not normally observed in uninfected cells (Fig. 9, Fig. S3). These results suggest that newly synthesized ICP0 may be tethered to a-tubulin so rapidly in the cytoplasm that the ICP0-containing complexes that are imported into the nucleus may also contain sizable amounts of a-tubulin. Moreover, ICP0 co-localized with dispersed a-tubulin + structures for hours after disassembly of the microtubule network ( Fig. 6, 8, 9). In contrast, ICP0's nuclear interactions with PML [13,15] and ICP4 [9,79] appear to be far more transient in nature. Clearly, further work is needed to define how ICP0's E3 ligase activity triggers microtubule disassembly in host cells, and to define why ICP0's co-localization with a-tubulin appears to be so much more stable than any other cellular protein identified to date.

ICP0-induced disassembly of microtubules versus G2/M cell cycle arrest
Synthesis of ICP0 causes dividing cells to arrest in the G2/M phase of the cell cycle [18,19,20]. Two explanations for this phenomenon have been offered including ICP0-induced degradation of centromere protein-C [18], and ICP0-induced activation of a DNA damage response pathway [80]. The current study suggests a third possibility; ICP0-induced disassembly of microtubules produces the same type of G2/M arrest that is caused by nocodazole [81]. Like, nocodazole, ICP0 triggers the rapid disassembly of microtubule networks in cells. Nocodazole causes cells to arrest in the G2/M phase of the cell cycle by blocking mitotic spindle formation, which is needed to align chromosomes on the metaphase plate prior to their segregation into daughter cells [81]. Likewise, synthesis of ICP0 has nocodazole-like effects on cell division [19,20]. Specifically, chromosome condensation occurs normally in cells treated with ICP0-expressing viruses [18] or nocodazole [82], but in both cases the condensed chromosomes fail to align on metaphase plates [18,82]. Further work will be required to test the validity of this interpretation, and determine if ICP0-induced disassembly of microtubules explains why synthesis of ICP0 causes cell division to stall during mitosis.

Conclusion
In recent years, the list of viruses that interact with cellular microtubules has grown in length, but the general significance of the phenomenon remains unclear. Herpesviruses rely on micro-tubules as a 'conveyor belt' to carry their incoming virions from the cell membrane to the nuclear pore [64,83]. Therefore, it is conceivable that disruption of this conveyor belt may be critical for virion egress, such that newly formed herpesvirus virions may flow efficiently in the reverse direction, from the nucleus back to the cell membrane. Clearly, further studies will be required to determine precisely why ICP0 translocates to the cytoplasm and destabilizes the host cell microtubule network during the E phase of HSV infection.

Cells, viruses, and plasmids
Vero cells, ICP0-complementing L7 cells [84], and ICP4complementing E5 cells [85] were propagated in Dulbecco's Modified Eagle's medium (DMEM) containing 0.15% HCO 3 2 supplemented with 5% fetal bovine serum (FBS), penicillin G (100 U/ml), and streptomycin (100 mg/ml), hereafter referred to as ''complete DMEM.'' ICP4-complementing E5 and ICP0complementing L7 cell lines were kindly provided by Neal Deluca (University of Pittsburgh; [84,85]). HSV-1 strain KOS was used as the parent wild-type virus in this study. KOS and all of the HSV-1 recombinant viruses used in this study were propagated in Vero cells, L7 cells, or E5 cells cultured in complete DMEM. KOS-GFP is a recombinant virus derived from HSV-1 strain KOS that expresses GFP from a CMV promoter cassette inserted in the intergenic region at the 39 ends of the UL26 and UL27 genes [47,86]. To construct HSV-1 recombinant viruses, a 7.3 kb DNA fragment which encompasses the entire LAT-ICP0 locus was first subcloned from HSV-1 strain KOS into a pCRII plasmid vector (Invitrogen Corporation, Carlsbad, CA). Mutations were introduced into the ICP0 gene of this plasmid, as described below.
i. p0+GFP12. A GFP coding sequence was amplified from plasmid peGFP-N1 (Clontech Laboratories, Mountain View, CA) using PCR primers that added NcoI and Bsu36I restriction sites to the 59 and 39 ends of the GFP coding sequence. Following PCR amplification with a high fidelity mixture of thermostable DNA polymerases, the GFP sequence was subcloned into NcoI and Bsu36I sites that occur in codons 1 and 11 of the ICP0 coding sequence. The resulting plasmid, p0 + GFP 12 , encoded ICP0 +GFP-12 protein. The genetic identity of this and all GFP insertions in the ICP0 gene was confirmed by DNA sequencing. Homologous recombination between HSV-1 KOS and the p0 + GFP 12 plasmid yielded the recombinant virus HSV-1 0 + GFP 12 .
ii. p0+GFP24. The plasmid p0 + GFP 24 was created by PCR amplification of three DNA fragments that were ligated in series to create a GFP insertion at the 59 end of exon 2 of the ICP0 gene (Fig. 1A). Specifically, 1. the 39 end of intron 1 of the ICP0 gene; 2. a GFP coding sequence; and 3. the 59 half of exon 2 of the ICP0 gene were PCR-amplified, ligated together, sequenced, and the resulting BamHI -XhoI fragment was subcloned into the BamHI -XhoI sites in the ICP0 gene. The resulting plasmid, p0 + GFP 24 , encoded ICP0 +GFP-24 protein in which GFP is inserted between amino acids 23 and 24 of ICP0. Homologous recombination between HSV-1 KOS and the p0 + GFP 24 plasmid yielded the recombinant virus HSV-1 0 + GFP 24 .
iii. p0-GFP. The plasmid p0 2 GFP gene was created by inserting a polylinker, GFP coding sequence, and stop codon derived from the plasmid peGFP-N1 (Clontech Laboratories) into a XhoI site in codon 104 of the ICP0 gene. The resulting plasmid, p0 2 GFP, encoded the N-terminal 104 amino acids of ICP0, a 14 amino acid linker, and GFP. Homologous recombination between HSV-1 KOS and the p0 2 GFP plasmid yielded the recombinant virus HSV-1 0 2 GFP. iv. p0+GFP105. The plasmid p0 + GFP 105 was derived by subcloning a dsDNA linker containing BsrGI and XhoI sites (tgtaca agatat ctcgag) in the 39 end of the GFP coding sequence in p0 2 GFP. Thus, a TAA terminator codon was removed and the GFP and ICP0 coding sequences were placed in the same openreading frame. The resulting plasmid, p0 + GFP 105 , encoded ICP0 +GFP-105 protein in which GFP is inserted between amino acids 104 and 105 of ICP0. Homologous recombination between HSV-1 KOS and p0 + GFP 105 yielded the virus HSV-1 0 + GFP 105 .
v. p0DRING. The plasmid p0DRING was derived by deletion of codons 105-229 of the ICP0 gene from p0 + GFP 105 . Homologous recombination between HSV-1 KOS and the p0DRING plasmid yielded the recombinant virus HSV-1 0DRING.

Construction of recombinant HSV-1 viruses
Infectious HSV-1 DNA was prepared by a protocol that relies upon dialysis to minimize shearing of genome-length HSV-1 DNA; this is a modification of a protocol that was generously provided by Karen Mossman (McMaster University, Hamilton, Ontario). Five 100 mm dishes of Vero cells (3610 7 cells) were inoculated with 5 pfu per cell of HSV-1 strain KOS. After 24 hours, cells were scraped, centrifuged, rinsed with PBS, resuspended in 7.0 ml of 200 mM EDTA pH 8.0, and transferred into a 15 ml conical. Proteinase K (75 ml of 10 mg/ml) and 375 ml of 10% SDS were added to virus-infected cells, and the tube was incubated in a rotisserie oven with slow rotation at 50uC for 16 hours. Proteins were removed by phenol : chloroform extraction, and DNA was transferred into a 0.5-3.0 mL Slide-alyzer cassette (10,000 MW cutoff; Pierce Chemical Co., Rockford, IL) and dialyzed against 0.16 standard saline citrate for 24 hours. Following dialysis, infectious HSV-1 DNA was aliquoted and frozen at 280uC.
Recombinant HSV-1 viruses were generated by co-transfection of 2 mg infectious HSV-1 KOS DNA and 1 mg each plasmid into a 60 mm dish containing 8610 5 ICP0-complementing L7 cells. After 12 hours, medium was replaced with complete DMEM containing 1% methylcellulose and GFP + plaques were selected on a Nikon TE2000 fluorescent microscope (Nikon Instruments, Lewisville, TX). GFP + recombinant viruses were repeatedly passed in ICP0-complementing L7 cells until a uniform population of viruses was obtained that produced 100% GFP + plaques, at which time Southern blot analysis was used to confirm that the anticipated mutation had been introduced into the ICP0 gene of HSV-1.

Southern blot analysis
Vero cell cultures were established at a density of 1.5610 6 cells per 100 mm dish and were inoculated with viruses at an MOI of 5 pfu per cell. DNA was harvested from virus-infected cells or uninfected controls at 24 hours p.i. using a standard DNA extraction procedure and Southern blot analysis was performed as previously described [87,88]. Oligonucleotide probes specific for intron 1 of the ICP0 gene (59-cccctagatgcgtgtgagtaaggggggcctgcgtatgagt-39) were used to probe restriction fragments of HSV-1 viruses.

Northern blot analysis of ICP0 GFP mRNA
Vero cell cultures were established at a density of 1.5610 6 cells per plate in 60 mm dishes. Vero cells were inoculated with viruses at an MOI of 10 pfu per cell. RNA was isolated from each treatment group at 12 hours p.i. using Ultraspec RNA isolation reagent (Biotecx Inc., Houston, TX), and Northern blot analysis was performed as previously described [2,9]. The oligonucleotide probes used were specific for either exon 3 of ICP0 mRNA (59ggagtcgctgatcactatggggtctctgttgtttgcaagg-39) or the GFP coding sequence (59-atagacgttgtggctgttgtagttgtactcc-agcttgtgc-39) [9,89].

Western blot analysis of ICP0 GFP reporter proteins
Vero cell cultures were established at a density of 3610 5 cells per well in 12-well plates, and were infected at an MOI of 10 pfu per cell. After 18 hours, proteins were harvested using mammalian protein extraction reagent (Pierce Chemical Co., Rockford, IL) supplemented with 1 mM dithiothreitol and protease inhibitor cocktail set I (Calbiochem, La Jolla, CA). After heat denaturation, 20 mg of each protein sample and MagicMark TM XP protein MW markers (Invitrogen Corporation, Carlsbad, CA) were resolved in a 10% polyacrylamide gel with a 4% stacking gel, and were transferred to nitrocellulose membranes. Protein blots were blocked in phosphate-buffered saline (PBS) containing 5% nonfat dry milk, and were incubated overnight at 4uC in PBS+0.1% Tween-20+5% nonfat dry milk containing a 1:1000 dilution of mouse monoclonal H1083 antibody specific for amino acids 395 to 775 of ICP0 [79,90,91] (EastCoast Bio, North Berwick, MA) and a 1:500 dilution of rabbit polyclonal anti-GFP antibody (Clontech Laboratories Inc.). Following incubation with primary antibodies, membranes were washed four times with PBS+0.1% Tween-20 (PBS-T), and were incubated for 1 hour with 1:20,000 dilution of goat anti-rabbit IgG and goat anti-mouse IgG conjugated, respectively, to the infrared fluorescent dyes IRDyeH 680 and IRDyeH 800CW (LI-COR Bioscience, Lincoln, NE). Protein blots were washed three times in PBS-T, and were scanned for two-color fluorescence using the Odyssey Infrared imaging system (LI-COR Bioscience). Data were analyzed using Odyssey application software version 3.0.16 (LI-COR Bioscience).

Immunofluorescent staining in Vero cells
Immunofluorescent staining in Vero cells was performed using an adaptation of a staining protocol that was generously provided by Roger Everett (MRC Virology Unit, Glasgow, Scotland). Glass coverslips were placed on the bottom of 6-well dishes and Vero cells were seeded at a density of 1610 6 cells per well. After allowing 8 hours for cell attachment, cultures were inoculated with HSV-1 viruses using the conditions defined in each Figure Legend and Results sub-section. At the indicated time of harvest, glass coverslips were removed from culture wells with fine forceps, and were fixed with PBS containing 1.9% formaldehyde and 2% sucrose for 10 minutes, followed by permeabilization with 90% methanol for 10 minutes. HSV-1 Fc-c receptors (glycoprotein E-I heterodimers [92]) were blocked along with all other non-specific protein-binding sites by incubating fixed cells in a solution of PBS containing 0.5% fetal bovine serum (FBS), as well as 10 mg/ml each of the c-globulin fractions of human, donkey, and goat serum (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA). Fixed cells were incubated in PBS+FBS+c-globulin block solution for 10 minutes, and the same solution was used as the diluent for primary and secondary antibodies. Each protein of interest was labeled by incubating fixed cells in a 1:1000 dilution of primary antibody for 16 hours. Excess primary antibody was removed by washing with PBS+FBS+c-globulin block solution, and cells were then incubated for 1 hour in PBS+FBS+c-globulin block containing a 1:1000 dilution of each secondary antibody as well as 10 ng/ ml of Hoechst 33342 to apply a blue label to the nuclei of cells (Calbiochem, La Jolla, CA). Excess secondary antibody was removed by washing with the PBS+FBS+c-globulin block solution, and coverslips were mounted on glass slides using FluorSave Reagent (Calbiochem, Gibbstown, NJ). Cells on coverslips were photographed using an Olympus BX41 microscope equipped with Olympus DP70 digital camera (Olympus America, Center Valley, PA). Images of green, red, and blue fluorescence were captured at exposure times of 500, 200, and 20 ms, respectively.
Live-cell imaging and video analysis of GFP-tagged ICP0 in HSV-1 infected cells Live-cell imaging was performed by placing a Nikon TE2000 microscope in a 37uC warm room and culturing Vero cells in HEPES-buffered (pH 7.9) RPMI medium containing 5% fetal bovine serum (Gibco BRL, Gaithersburg MD). Time-lapse photography was made possible by outfitting the microscope with an Olympus DP72 digital camera (Olympus America, Center Valley, PA) and a computer containing Olympus DP2-BSW microscope digital camera software which controlled the operation of a Lambda SC SmartShutter TM control system (Sutter Instrument, Novato, CA). Images were captured at 30-second intervals to create Videos S1, S2, S3, S4, or were captured at 10minute intervals to create static Figures 5 and 7. Time-lapse videos were set to play at a rate of 5 frames per second (2.5 minutes per second of video), such that the motion of GFP-tagged ICP0 shown in videos occurs 150 times faster than real time.

Transfection of Vero cells
Vero cells were transfected with plasmid DNA (0.75mg/well in 12-well plates) using Lipofectamine 2000 (Invitrogen Corporation, Carlsbad, CA) in complete DMEM at 37uC. Medium was replaced at 4 hours post-transfection to avoid toxicity associated with the transfection reagent, and cells were incubated at 37uC in complete DMEM until cells were fixed for immunofluorescent staining. Figure S1 ICP0 +GFP-105 disperses linear cytoplasmic structures in HSV-1 0 + GFP 105 -infected cells. Vero cells were inoculated with 5 pfu per cell of (A) HSV-1 0 + GFP 105 or (B) HSV-1 0DRING in the presence of 200 mM cycloheximide from 20.5 to 10 hours p.i., and were released into medium containing no drugs. Photographs of (A) ICP0 +GFP-105 or (B) ICP0 DRING between 1 and 6 hours after release from the cycloheximide block. The scale bar denotes a distance of 10 mm. Found at: doi:10.1371/journal.pone.0010975.s001 (6.55 MB TIF) Figure S2 Microtubule networks remain intact in cells infected with an ICP0 2 null virus, HSV-1 0 2 GFP. Vero cells were uninfected or were inoculated with 5 pfu per cell of inoculated with HSV-1 0 2 GFP in the presence of 200 mM cycloheximide from 20.5 to 10 hours p.i., and were released into medium containing no drugs. At 4, 6, or 8 hours post-release, cells were fixed and stained with rabbit antibody against a-tubulin and the truncated ICP0 2GFP peptide was directly visualized. The scale bar denotes a distance of 10 mm. Found at: doi:10.1371/journal.pone.0010975.s002 (5.81 MB TIF) Figure S3 Co-localization of ICP0 and a-tubulin in cells transfected with ICP0-expressing plasmids. Merged images of the a-tubulin and ICP0 staining shown in Figure 9 in cells that were mock transfected, or which were transfected with pICP0, p0 + GFP 105 , or p0DRING. The scale bar denotes a distance of 10 mm. Found at: doi:10.1371/journal.pone.0010975.s003 (7.18 MB TIF) Figure S4 Enlarged image of one portion of an HSV-1 plaque in which ICP0 translocation, a-tubulin dispersal, and co-localization of ICP0 and a-tubulin are observed. Enlarged photographs of (A) ICP0-staining and (B) a-tubulin staining from Figure 10. Found at: doi:10.1371/journal.pone.0010975.s004 (7.30 MB TIF) Video S1 Rapid movement of ICP0 +GFP-24 -labeled globular bodies in HSV-1 0 + GFP 24 -D4-infected cells. Time-lapse photographs of ICP0 +GFP-24 protein in a representative field of Vero cells between 6.0 and 6.5 hours after inoculation with 5 pfu per cell of HSV-1 0 + GFP 24 -D4. Cultures were continuously incubated at 37uC before and during time-lapse photography. Photographs were captured at a rate of twice per minute for 30 minutes, and are shown in this video at an elapsed rate that is 150 times faster than normal. Video S3 ICP0 DRING accumulates in linear cytoplasmic structures in HSV-1 0DRING-infected cells. Time-lapse photographs of ICP0 DRING protein in a representative field of Vero cells between 4.0 and 4.5 hours post-release from a cycloheximide block. Vero cells were inoculated with 5 pfu per cell of HSV-1 0DRING in the presence of 200 mM cycloheximide from 20.5 to 10 hours p.i., and were released into medium containing no drugs. At 4.0 hours post-release, a field of cells was chosen where ICP0 DRING was present in linear cytoplasmic structures. Cultures were continuously incubated at 37uC before and during time-lapse photography. Photographs were captured at a rate of twice per minute for 30 minutes, and are shown in this video at an elapsed rate that is 150 times faster than normal. Found at: doi:10.1371/journal.pone.0010975.s007 (9.37 MB MOV)

Supporting Information
Video S4 Bundles of ICP0 DRING disperse following nocodazole treatment of HSV-1 0DRING-infected cells. Vero cells were inoculated with 5 pfu per cell of HSV-1 0DRING in the presence of 200 mM cycloheximide from 20.5 to 10 hours p.i., and were released into medium containing no drugs. At 4.0 hours postrelease, a field of cells was chosen and a 36solution of nocodazole was added to culture medium to achieve a final concentration of 10 mg/ml. Time-lapse photography of ICP0 DRING protein was commenced 1 minute later. Cultures were continuously incubated at 37uC before and during time-lapse photography. Photographs were captured at a rate of twice per minute for 30 minutes, and are shown in this video at an elapsed rate that is 150 times faster than normal. Found at: doi:10.1371/journal.pone.0010975.s008 (5.42 MB MOV)