Herpes simplex virus 1 ICP0 co-localizes with a SUMO-specific protease

The herpes simplex virus 1 (HSV-1) immediate early regulatory protein ICP0 functions as a broad-spectrum transactivator of viral gene expression (reviewed in Roizman & Sears, 1996 ). ICP0 is required for normal levels of expression of all classes of HSV genes and appears to play an important role in virus reactivation from latency (Cai et al., 1993 ; Clements & Stow, 1989 ; Harris et al., 1989 ; Leib et al., 1989 ; Sacks & Schaffer, 1987 ; Stow & Stow, 1986 ; Zhu et al., 1990 ). Whilst the precise mechanism of ICP0 transcription transactivation remains unclear, several important features have been established. Current evidence suggests that the interaction between ICP0 and the cellular compartments known as PML nuclear bodies (NBs) plays a key role. The protein PML is the defining component of PML NBs, discrete domains of the nucleus that are enriched in a number of additional proteins but appear to be devoid of DNA and RNA. The exact cellular function of PML NBs has been the subject of much speculation, with roles proposed in transcriptional regulation, apoptosis and genomic stability. With regard to virus replication, sites for the initiation of herpesvirus DNA synthesis have been reported to lie specifically adjacent to PML NBs (Ishov & Maul, 1996 ) and additional involvement of NBs in aspects of the replication of adenovirus, EpsteinBarr virus and human cytomegalovirus (HCMV) replication have been reported (Everett & Maul, 1994 ; Ishov & Maul, 1996 ; Leppard & Everett, 1999 ; Maul, 1998 ).

NBs are also enriched with proteins that are modified at specific lysine residues by a small ubiquitin-like molecule, known as SUMO-1 [also termed Sentrin (Okura et al., 1996 ), PIC1 (Boddy et al., 1996 ), or GMP-1 (Mahajan et al., 1997 ; Matunis et al., 1996 )]. It has been suggested that SUMO-1 modification of PML itself may be important for recruiting PML to the NB structure (Kamitani et al., 1998b ; Müller et al., 1998 ), although for a number of other NB proteins the SUMO-1 modification does not appear to be essential for localization.

ICP0 promotes the loss of PML NBs in a proteasome-dependent manner (Everett & Maul, 1994 ). ICP0 also promotes the proteasome-mediated degradation of other cellular proteins, including CENP-C, CENP-A and the regulatory subunit of the cellular DNA-dependent protein kinase (Everett et al., 1998 , 1999a ; Lomonte et al., 2001 ; Parkinson et al., 1999 ). The ability of ICP0 to promote targeted NB disruption and protein degradation is dependent on the conserved RING finger domain. Earlier work had established that the ICP0 RING domain is essential for transactivation, providing robust evidence that NB disruption, protein degradation and promotion of gene expression are linked (Everett et al., 1995 ; ORourke & OHare, 1993 ).

Emerging evidence indicates that RING finger domains may act as ubiquitin E3 ligases (for reviews see Aravind & Koonin, 2000 ; Freemont, 2000 ; Tyers & Jorgensen, 2000 ). ICP0 has also now been shown to function as a ubiquitin E3 ligase, although the precise mechanism has been disputed (Boutell et al., 2002 ; Van Sant et al., 2001 ) and direct evidence of ICP0-mediated conjugation of ubiquitin to specific cellular targets has yet to be established. However, consistent with the role of an E3 ligase function, ICP0 has been reported to induce the accumulation of conjugated ubiquitinated protein species (Everett, 2000 ) and to interact with a ubiquitin-specific protease (USP7/HAUSP) that is also present within the PML NB (Everett et al., 1997 , 1999b ).

In order for PML degradation to occur during virus infection, there may be a requirement for the removal of SUMO-1 from the SUMO-1-modified protein isoforms prior to degradation of PML or other target protein(s). Recent work has identified a number of SUMO-1-specific proteases that reverse the covalent modification of SUMO-1 from the target protein (Gong et al., 2000 ; Kim et al., 2000 ; Li & Hochstrasser, 1999 ; Suzuki et al., 1999 ), although little is understood of their regulation and activity. Some evidence of substrate specificity for SUMO-1 deconjugation has been obtained from yeast. Yeast has two proteases (Ulp1 and Ulp2) and mutants which lack either protease show the accumulation of different Smt3 (yeast homologue of SUMO-1)-conjugated species, suggesting that the enzymes may exhibit selectivity in controlling the conjugation profile (Li & Hochstrasser, 1999 , 2000 ). This has not been reported for the mammalian system, nor is it known if a spatial regulation of SUMO-1 conjugation/deconjugation occurs.

In this work we examine the effects of ICP0 on the SUMO-1 modification status of the cell, with particular emphasis on the association of ICP0 with a SUMO-specific protease (SENP1). To facilitate analyses of the SUMO-1 pathway, we constructed a cell line that constitutively expresses epitope-tagged SUMO-1. Our results demonstrate that ICP0 promotes the generalized loss of SUMO-1 from the nucleus. We also demonstrate that SENP1 localizes to a number of subnuclear domains and also promotes the loss of SUMO-1 from the nucleus and from PML. Interestingly, PML remains in NB structures suggesting that removal of the SUMO-1 moiety alone is not sufficient for NB dispersal. We observe that co-expression of ICP0 with SENP1 results in the recruitment of SENP1 into ICP0 subnuclear structures, and that this occurs during early viral infection. The relevance of these results to the loss of SUMO-1 modified PML during virus infection is discussed.

Cells and viruses.
HEp2, Vero and HeLa cells were grown on Dulbeccos modified Eagles medium, supplemented with 10% foetal calf serum. HEp2-SUMO and HeLa-SUMO cell lines (see below) were cultured under similar conditions with the addition of 2 µg/ml puromycin to maintain the integrated SUMO-1. Virus stock used was wild-type HSV-1 strain 17syn⁺.

Construction of the SENP1 expression vector.
The ∼1·9 kb SENP1 ORF was amplified from a human testis cDNA library (Clontech #7414-1) using the sense (5' GAGCTCGCTAGCATGGATGATATTGCTGATA 3') and antisense (5' CTCGAGTCTAGATCACAAGAGTTTTCGGTGG 3') primers, based on published sequence (Gong et al., 2000 ). The PCR product was inserted into pCR-II-Topo, then excised as an NheIScaI fragment and inserted into a vector containing an in-frame Flag-HA-tag. The Flag-Ha-SENP1 EcoRIKpnI fragment was then excised and inserted into pcDNA3 for expression of Flag and HA epitope-tagged SENP1. Sequencing of the SENP1 clone demonstrated that the clone differed from the published sequence due to the presence of an additional glutamic acid residue at aa 593. Comparison with database sequences indicated that this sequence was identical to the orf predicted by the ENSEMBL project (ID: ENSG00000079387).

Construction of stable cell lines expressing His-Myc-tagged SUMO-1.
A His-Myc-tagged SUMO-1 clone (hmSUMO-1) was constructed by inserting a 6xHis tag linker (sense: 5' AGCTTATGCATCATCATCATCATCA 3'; antisense: 5' CATGTGATGATGATGATGATGCATA 3') and a ∼1·1 kb NcoIXbaI Myc-tagged SUMO-1 fragment from pMLV-Myc-PIC1 (Boddy et al., 1996 ) in a three-way ligation with a HindIII/XbaI-digested pcDNA3 backbone, to generate pcDNA3-Myc-SUMO-1. The His-Myc SUMO-1 fragment was then subcloned into a vector, pIRES-P, which had been digested with XhoI, blunt-ended with Klenow DNA polymerase and then digested with XbaI. pcDNA3-Myc-SUMO-1 was digested with HindIII, filled-in with Klenow DNA polymerase and further digested with XbaI. This generated a 1·1 kb fragment that was then inserted into the cleaved pIRES-P backbone to produce vector pIRES-HIS-SUMO-1. The pIRES-HIS-SUMO-1 vector was transfected into HEp2 cells and stable line clones selected using 2 µmg/ml puromycin. Following selection, 12 independent HEp2 cell colonies were isolated and analysed for His-Myc-tagged SUMO-1 expression. Immunofluorescence analysis demonstrated that 100% of cells from all the independent colonies showed approximately similar levels of nuclear Myc/SUMO-1 expression. Cells derived from a colony (HEp2-SUMO) were used in these studies. Similar Myc/SUMO-1 modification profiles of cellular proteins were observed when analysed by Western blot, with some variation in expression levels between clones (data not shown). Similar results were also obtained when generating the tagged SUMO-1-expressing stable lines independently in HeLa cells (HeLa-SUMO).

Transfections.
Transfections were performed using the calcium phosphate precipitation procedure modified by the use of BES-buffered saline (pH 7·06) as previously described (Batchelor & OHare, 1992 ). The total amount of DNA was equalized to 2 µg with pUC19 DNA.

Immunofluorescence studies.
Approximately 40 h post-transfection cells, plated on glass coverslips, were washed in PBS and fixed with ice-cold methanol. Primary antibodies, diluted in PBS10% FCS, were anti-c-Myc 9E10 (1:400, Boehringer Mannheim) for the Myc-tag; anti-GMP-1 (1:1000, Zymed) or anti-PIC1 (1:200) for SUMO-1; anti-11060 (1:1000), R191 and Rb95 (1:200) for ICP0; anti-Flag M2 (1:200, Stratagene) and anti-OctA (1:10, Santa Cruz) for the Flag-tag. A rabbit polyclonal antibody against a purified GSTPML bacterial expression product was generated by standard methods. Specificity was verified by comparing against an existing anti-PML monoclonal antibody (5E10). Secondary antibodies conjugated to fluorochromes Alexa 488 or Alexa 543 dyes (Molecular Probes) were diluted 1:200 in PBS10% FCS. After washing in PBS, cells were visualized using a Zeiss LSM 410 confocal microscope. Images for each channel were captured sequentially with 8-fold averaging. Composite illustrations, representative of numerous images gathered for each test construct and condition, were prepared using Adobe software.

Western blot analysis.
After separation by SDSPAGE, proteins were transferred to nitrocellulose membranes which were blocked with PBS0·05% Tween 20 (PBST) containing 5% dried non-fat milk. The membranes were incubated (1 h) with primary antibody in PBST5% dried milk, washed three times in PBS1% Triton X-100 and incubated for a further 1 h in PBST5% dried milk containing the appropriate horseradish peroxidase-conjugated secondary antibody. After further washing in PBS1% Triton X-100, membranes were processed with chemiluminescence detection reagents (Pierce). Primary antibodies used for immunoblot were anti-c-Myc 9E10 (1:400, Boehringer Mannheim), anti-SV5 (1:5000, kindly supplied by R. Randall, University of St Andrews) and anti-LP1 (1:4000, kindly supplied by T. Minson, University of Cambridge) and anti-GMP-1 (1:1000, Zymed).

Comparison of SUMO-specific proteases
The schematic diagram in Fig. 1(a) illustrates recently identified SUMO-specific proteases with confirmed functional activity, two from Saccharomyces cerevisiae and two from mammalian cells (Gong et al., 2000 ; Kim et al., 2000 ; Li & Hochstrasser, 1999 , 2000 ). The proteases differ significantly in size, but all possess a conserved core domain. This core domain incorporates residues predicted to make up the putative catalytic triad (histidine, aspartate and cysteine), and an invariant glutamine residue predicted to aid in the formation of the active site (Fig. 1b). Although the existence of a number of other SUMO-specific protease homologues has been predicted from sequence alignments (Kim et al., 2000 ; Li & Hochstrasser, 1999 ), confirmation of proteolytic activity has not been demonstrated. We wished to examine the SENP1 SUMO-specific protease, its effects upon the cellular SUMO-1 modification pathways and its possible involvement in ICP0 mediated events.

(34K): Fig. 1. Comparison of SUMO-1-specific proteases and expression of SUMO-1. (a) Schematic illustration of the two human SUMO-specific proteases (SENP1 and SUSP1) and the two yeast homologues (ScUlp1 and ScUlp2). The conserved core domain is shaded, whilst the arrows indicate the positions of conserved residues of the putative catalytic triad (histidine, aspartate and cysteine) and an additional glutamine residue predicted to aid in the formation of the active site. (b) Amino acid sequence alignments illustrating regions of the conserved core domain. (c) An N-terminal Flag and haemagglutinin (HA) epitope-tagged SENP1 clone was expressed from a plasmid containing the CMV promoter (pcDNA3-FH-SENP1). COS cells were transfected with 0·5 µg of pcDNA3-FH-SENP1. Cell extracts were harvested 40 h later, fractionated by electrophoresis and analysed by Western blot, with anti-Flag M5 (Sigma). Mock- and pcDNA3-FH-SENP1-transfected lanes are shown. The epitope-tagged SENP1 protein had a predicted molecular mass of approximately 76 kDa.

Expression of epitope-tagged SENP1 and its effects upon endogenous SUMO-1
We designed primers based on sequence data (Gong et al., 2000 ), to amplify the SENP1 clone from a human testes cDNA library. The PCR product was inserted into a vector to create an N-terminal Flag-HA epitope-tagged SENP1 clone. The SENP1 protein product was subsequently expressed under the control of the constitutive CMV promoter following transient transfection into mammalian cells. Western blot analysis for SENP1 protein expression demonstrated expression of a single epitope-tagged protein migrating at a position consistent with the predicted molecular mass (76 kDa) of SENP1 (Fig. 1c). The results of immunofluorescence analysis demonstrated that at low levels of expression SENP1 was present in discrete foci in the nucleus, with some nuclear rim staining (Fig. 2a, upper panel). At higher levels of expression the SENP1 foci increased in number, and SENP1 was also seen with a more nuclear diffuse appearance, but excluded from the nucleolus (Fig. 2a, lower panels). Similar results were observed in several cell types, including HeLa and HEp2 cells.

(35K): Fig. 2. Expression of SENP1 in transient transfection. (a) HeLa and HEp2 cells were transfected with 0·3 µg of pcDNA3-FH-SENP1. Cells were fixed approximately 40 h later, and immunofluorescence was performed using anti-Flag M2 antibody as the primary antibody. At low levels of expression SENP1 was present within discrete foci within the nucleus, and there was also some nuclear rim staining. At higher levels of expression SENP1 was present with a more diffuse nuclear appearance. (b) HeLa and HEp2 cells transfected with SENP1 were co-stained for SUMO-1 with an anti-PIC1 antibody (kindly supplied by P. Freemont, Imperial College of Science, Technology and Medicine). SUMO-1 was observed to be lost from the nucleus in SENP1-transfected cells, as indicated by the arrows.

In order to examine the effect of SENP1 expression on the localization of endogenous SUMO-1, cells were co-stained for SUMO-1 after transfection of SENP1. The results demonstrated that in cells expressing SENP1, SUMO-1 appeared to be totally lost from the nucleus (Fig. 2b). Although the low levels of the endogenous SUMO-1 species make these studies difficult, the effect of SENP1 was clear and reproducible.

Construction of stable cell lines expressing epitope-tagged SUMO-1
Nevertheless, the technical difficulties in detecting the very low levels of endogenous SUMO-1 limited ready analysis of effects of SENP1 in both immunofluorescence and Western blotting applications. To facilitate the analysis of the activity of SENP1 upon SUMO-1, we established cell lines (HEp2 or HeLa cells) maintaining expression of an epitope-tagged SUMO-1 (hmSUMO-1) using a bicistronic vector system and puromycin selection (Fig. 3a). In these cells hmSUMO-1 was present in predominantly diffuse nuclear staining (Fig. 3a), with a subset of the protein frequently accumulating in foci that co-localized with PML (data not shown). To confirm that the introduced hmSUMO-1 acted as a substrate for conjugation into high molecular mass conjugates, we performed Western blot analysis on the parental cells or the hmSUMO-1-expressing cell lines with antibodies to SUMO-1 or an anti-Myc monoclonal antibody to specifically detect hmSUMO-1. In the control HeLa cells a major band was detected with the anti-SUMO-1 antibody (Fig. 3b, left-hand panel) which from previous reports most likely represents Ran-GAP (Matunis et al., 1996 ; Saitoh & Hinchey, 2000 ). The detection of additional SUMO-1-modified species in the normal HeLa cells requires long exposures of the blot (data not shown). In the hmSUMO-1-expressing cells, numerous high molecular mass conjugates were observed and these species were also detected with the anti-Myc antibody (Fig. 3b, right-hand panel), confirming that the hmSUMO-1 was indeed a substrate for conjugation. The majority of hmSUMO-1 within these cell lines appeared to be conjugated in high molecular mass species with comparatively minor amounts of free SUMO-1 observed, indicating that the SUMO-1 conjugation machinery was not limiting in the established lines. We noted that the band around the size of Ran-GAP1 migrated as a doublet, most likely representing modification of Ran-GAP1 with the endogenous SUMO-1 or the hmSUMO-1, which would result in a slightly slower migration.

(33K): Fig. 3. Expression of SENP1 in the cell lines stably expressing SUMO-1. (a) A bicistronic vector containing the N-terminal 6xHis & Myc epitope-tagged SUMO-1 clone was transfected into HEp2 cells. Stable cell line clones expressing the epitope-tagged SUMO-1 were created under puromycin selection. Immunofluorescence staining to detect the hmSUMO-1 was performed using anti-Myc as the primary antibody (9E10). SUMO-1 expression was mainly nuclear and diffuse, although some dense nuclear domains were also observed. (b) HeLa cell and HeLa-SUMO extracts were subjected to electrophoresis and Western blot analysis with an anti-GMP-1 antibody to detect SUMO-1 or with 9E10 to detect the Myc epitope. Typical results demonstrated a range of high molecular mass isoforms, most evident in the HeLa-SUMO cell line. The major SUMO-1-modified isoform in HeLa extracts is presumed to be Ran-GAP1. (c) HEp2-SUMO cells were transfected with 0·3 µg of pcDNA3-FH-SENP1; approximately 40 h post-transfection cells were fixed with paraformaldehyde. Immunofluorescence staining was performed with rabbit anti-OctA polyclonal antibody to detect the epitope-tagged SENP1, and with a monoclonal anti-Myc antibody for SUMO-1. SUMO-1 was observed to be lost from the nucleus in SENP1-transfected cells, indicated by the arrows.

Analysis of SENP1 expression on hmSUMO-1-derived cell lines
We next transfected the SENP1 construct into our cell lines in order to analyse the effect of SENP1 expression upon the SUMO-1 localization within these cell lines. The ability of SENP1 to induce the loss of SUMO-1 from the nucleus could now be clearly observed, with residual SUMO-1 sometimes co-localizing with the punctate SENP1 domains within the nucleus. Typical examples are shown in the three panels (Fig. 3c, arrowed cells).

Effect of SENP1 expression upon PML and high molecular mass SUMO-1 conjugates
The majority of SUMO-1 in the cell is reported to be present in high molecular mass conjugates, both from previous reports (Desterro et al., 1998 ; Matunis et al., 1996 ) and from analysis of our hmSUMO-1 stable cell lines (Fig. 3b). To obtain biochemical confirmation that the SENP1 clone was active in promoting the loss of SUMO-1, the epitope-tagged SENP1 clone was co-transfected into COS cells with an hmSUMO-1 plasmid. Cells were harvested as previously described and analysed by Western blot. In the presence of SENP1 the general profile of the SUMO-1 modified species was significantly reduced (Fig. 4a), consistent with the generalized loss of SUMO-1 from the nucleus as observed by immunofluorescence. We noted also that levels of the unconjugated SUMO-1, migrating at approx. 18 kDa (Fig. 4a, free SUMO-1), were increased in the presence of SENP1, consistent with deconjugation from target proteins. To specifically examine the status of the modified forms of PML in the presence of SENP1, an SV5 epitope-tagged PML560 construct was transfected into COS cells, either in the presence or absence of the SENP1 construct. Expression of PML alone resulted in a major species migrating at approx. 80 kDa, with the presumptive SUMO-1 modified forms of PML migrating within the range 90200 kDa (Fig. 4b). The expression of SENP1 in conjunction with PML abolished the majority of the modified high molecular mass forms of PML (Fig. 4b, long arrows, bands 13), despite similar levels of expression of the unmodified PML species. Interestingly, some high molecular mass isoforms of PML were unaffected by SENP1 (Fig. 4b, short arrows, bands 4, 5), indicating the possibility that these represent other post-translational modifications of PML (or perhaps stable dimers). In control experiments, levels of expression of a control protein (VP16), expressed from a control plasmid, were unaffected by SENP1 expression (Fig. 4b, right-hand panel). These results, taken together, suggest that SENP1 promotes the deconjugation of SUMO-1 from PML and other high molecular mass conjugates.

(39K): Fig. 4. Biochemical analysis of SENP1 activity upon SUMO-1 and PML. (a) COS cells were co-transfected with 500 ng His-Myc-SUMO-1 plasmid and either 500 ng of the pcDNA3-FH-SENP1 clone or a non-specific control plasmid (pcDNA3-His-Lacz). Extracts were prepared as previously described and fractionated by SDSPAGE. A 38 % gradient gel was used in order to resolve the high molecular mass SUMO-1-conjugated species, whilst a 10% SDSPAGE gel was used to resolve the lower molecular mass unconjugated SUMO-1 species. Western blot analysis performed with an anti-Myc antibody was used to detect the transfected SUMO-1 species. (b) COS cells were transfected with 500 ng of either SV5-PML560 clone or VP16 control plasmid in the presence or absence of the SENP1 construct. Western blot analysis was performed with an anti-SV5 antibody to detect PML or with anti-LP-1 antibody to detect VP16. Long arrowheads indicate PML species that are lost, whilst short arrowheads indicate unaffected high molecular mass PML species. The asterisk indicates non-specific reactive bands. Expression of the control (VP16) was unaffected by SENP1.

Effect of SENP1 expression upon the PML nuclear bodies
It has previously been suggested that SUMO-1 modification of PML is critical for the formation of the nuclear body structure (Ishov et al., 1999 ; Kamitani et al., 1998a ; Zhong et al., 2000 ). As SENP1 expression was able to promote the loss of SUMO-1, we were interested in examining how SENP1 expression affected the integrity of the PML bodies. To address these observations HEp2 cells were transfected with SENP1 and analysed for PML localization. Little significant difference in the distribution of PML could be observed, even at relatively high levels of SENP1 expression (Fig. 5a, arrowed cells). Similar results were also observed using HeLa cells (data not shown). To further examine the localization of SENP1 and PML, co-transfection experiments were performed in HEp2 cells. The immunofluorescence analysis demonstrated that the PML nuclear domains were, as expected, enlarged in the presence of exogenous PML. Consistent with our observations above, the exogenous PML domains were not significantly affected by the presence of exogenous SENP1 (Fig. 5b, arrowed cells), when compared to cells transfected with PML alone (data not shown). We noted that at low levels of expression SENP1 was frequently observed in domains adjacent to or at the periphery of endogenous PML domains, seen more clearly in the high magnification image (Fig. 5c, left-hand panel). Similarly, in cells co-transfected with PML, SENP1 was also observed localized to the periphery or juxtaposed to these enlarged PML domains. Again this was most notable at low levels of SENP1 expression (Fig. 5c, right-hand panels). In both cases, SENP1 appeared to localize to other domains that did not contain PML. At higher levels of expression SENP1 was found throughout the nucleus in a pattern similar to that described above (Fig. 2), although no significant recruitment to the PML domains was observed. These observations of PML nuclear body integrity, taken together with the Western blot analysis, imply that loss of SUMO-1 from PML is not sufficient for PML dispersal and that additional events may be required to mediate this process.

(15K): Fig. 5. Expression of SENP1 and effects on endogenous and exogenous PML nuclear bodies. Transfected HEp2 cells were fixed in methanol and stained with an anti-Flag M2 antibody to detect SENP1. PML was visualized using an anti-PML polyclonal antibody (#75) raised against recombinant GSTPML. (a) HEp2 cells transfected with epitope-tagged SENP1 alone to examine endogenous PML; no significant difference in PML NB accumulation was observed. Cells expressing SENP1 are highlighted by the arrows. (b) HEp2 cells transfected with epitope-tagged SENP1 and with 100 ng SV5-epitope-tagged PML560 clone and 300 ng pcDNA3-FH-SENP1. Cells transfected with exogenous PML were readily identified on the basis of the enlarged PML domains, relative to control transfected cells (not shown). Arrows indicate SENP1 and PML co-transfected cells. (c) High magnification image of SENP1 localization with and adjacent to endogenous and exogenous PML NBs respectively.

ICP0-mediated loss of SUMO-1
The HSV-1 ICP0 protein has been reported to disrupt nuclear bodies in a proteasome-dependent manner, and this is associated with the loss of several SUMO-1-modified species (Everett et al., 1998 ). Having established a cell-line containing stable integrated hmSUMO-1, we examined the effect of ICP0 expression upon SUMO-1 localization by immunofluorescence. In cells transfected with ICP0 the results clearly showed a significant loss of SUMO-1, seen by the decreased Myc-tag staining for SUMO-1 (Fig. 6, arrowed cells). In some cases an almost total loss of SUMO-1 was observed, while in other cases a partial loss was observed, although it was difficult to correlate this with expression level. In situations where SUMO-1 was almost completely lost from the nucleus, a modest accumulation of SUMO-1 in the cytoplasm could also be sometimes seen (Fig. 6, arrowed cell, top panel). Our Western blot analysis on the SUMO-1 cell lines demonstrated that the majority of SUMO-1 was in the conjugated form (Fig. 3b), suggesting that the loss of the SUMO-1 species in the presence of ICP0 was likely to be due to loss of high molecular mass SUMO-1 conjugates. We next examined hmSUMO-1 cells transfected with an ICP0 RING finger mutant. In contrast to the wild-type ICP0 construct, no significant loss of SUMO-1 was observed (Fig. 6, arrowed cell, bottom panel). These results on the effect of ICP0 on SUMO-1 localization and the requirement for the RING finger are consistent with previous biochemical analysis showing loss of SUMO-1 conjugates in infected cells, and the requirement for the RING finger of ICP0 (Everett et al., 1998 ). We also noted that the ICP0 RING finger mutant actually promoted the accumulation of some SUMO-1 to the ICP0 domains, although whether the recruited SUMO-1 represents a subset of SUMO-1-modified target proteins remains unclear.

(76K): Fig. 6. Loss of SUMO-1 in ICP0-transfected cells. Upper two panels; HeLa-SUMO or HEp2-SUMO cells were transfected with 200 ng pDR27. Polyclonal antibody Rb95 was used to detect ICP0-expressing cells; anti-Myc antibody was used to detect the epitope-tagged SUMO-1 protein. Cells expressing ICP0 are indicated with an arrow. Lower panel; HeLa-SUMO cells transfected with pDR33, a RING deletion mutant of ICP0; note that SUMO-1 was not lost but instead accumulated at ICP0 domains.

Localization of ICP0 and SENP1
In order to determine the potential involvement of SENP1 in the process of SUMO-1 loss by ICP0, HeLa cells or HEp2 cells were co-transfected with SENP1- and ICP0-expressing plasmids. The results show a high degree of co-localization between SENP1 and ICP0 (Fig. 7). Co-localization occurred even at relatively low levels of SENP1 expression. Furthermore, the nuclear diffuse staining of SENP1 was reduced, consistent with SENP1 active recruitment to ICP0 domains. This result contrasts quite strongly with the localization profile of SENP1 in the presence of exogenous PML, which did not significantly recruit SENP1 to the nuclear bodies (Fig. 5b). These results suggest that ICP0 may be recruiting SENP1 to participate in the loss of SUMO-1. In order to determine whether this observation could occur during virus infection, we infected cells that had been transiently transfected with SENP1 and analysed them by immunofluorescence.

(11K): Fig. 7. Co-localization between ICP0 and SENP1. HeLa cells were co-transfected with SENP1- and ICP0-encoding plasmids. Immunofluorescence was performed as described in the previous figure. ICP0 and SENP1 were visualized with a rabbit polyclonal antibody (Rb95) and monoclonal anti-Flag M2 antibodies respectively. Cells expressing both SENP1 and ICP0 have been indicated with an arrow.

SENP1 and ICP0 localization during virus infection
Early in infection, ICP0 localizes to a number of discrete foci within the nucleus, during which time PML NB and SUMO-1 loss occurs. However, late in infection when the SUMO-1-modified forms of PML are lost, the ICP0 foci are generally dispersed and ICP0 is found in diffuse nuclear and cytoplasmic locations (Everett & Maul, 1994 ; Kawaguchi et al., 1997 ). We have compared the localization of ICP0 and SENP1 at early and late time points of HSV-1 infection. Our results are consistent with the results from the co-transfection assays above, and demonstrate that during early HSV-1 infection, SENP1 was localized with ICP0 to discrete foci within the nucleus (Fig. 8). Interestingly, late in infection when ICP0 was dispersed with a more diffuse nuclear and cytoplasmic appearance, SENP1 was no longer co-localized with ICP0, but instead was observed in numerous accumulations within the cytoplasm. Thus SENP1 is recruited at the time when ICP0-mediated PML/SUMO-1 loss is occurring, and together with the results above, indicates that SENP1 recruitment plays an important role in this SUMO-1 loss and NB disruption. Late in infection, when the PML NB and SUMO-1 loss has occurred, SENP1 and ICP0 no longer co-localize, suggesting the interaction, whether direct or not, may be regulated by some mechanism and allowing the possibility that SENP1 relocation to the cytoplasm could have an additional role in virus infection.

(20K): Fig. 8. Localization of SENP1 and ICP0 during infection. HEp2 cells or Vero cells were transiently transfected with pcDNA3-FH-SENP1 and 24 h later infected with HSV-1 strain 17syn+ at an m.o.i. of 10. Cells were fixed and assayed by immunofluorescence at either 4 or 16 h post-infection. ICP0 expression in infected cells was visualized with polyclonal antibody (R191), whilst SENP1-expressing cells were visualized with anti-Flag M2 antibody. Cells expressing ICP0 and SENP1 are indicated with an arrow.

Current evidence indicates that the mechanism of action of the HSV regulatory protein ICP0 is directly connected with its recruitment to PML-containing NBs and its effect on the modification of PML (and other proteins) by SUMO-1, resulting in loss of SUMO-1 conjugation, proteasome-dependent degradation and disruption of the PML NBs. Modification by SUMO-1 is a topic of much current interest and one aspect of SUMO-1 modification and metabolism, as for the ubiquitin pathway, is SUMO-1 deconjugation from its substrates catalysed by SUMO-specific proteases. The first such enzymes were characterized in yeast and, while a number of mammalian SUMO-specific proteases have been predicted based on sequence similarity, there are currently few reports characterizing their activity. One enzyme confirmed to act as a SUMO-specific protease is SENP1, a 76 kDa protein containing the core catalytic consensus motif (Gong et al., 2000 ). In this report we have examined the activity of the SENP1 in relation to the activity of ICP0 and its effects on SUMO-1 modification of PML and NB disruption. To facilitate our analysis, we also established a cell line constitutively expressing an epitope-tagged version of SUMO-1, hmSUMO-1. We demonstrate that SENP1 localizes to the nucleus mainly in discrete subdomains, a subset of which co-localizes with the PML bodies. SENP1 promoted the generalized loss of SUMO-1 from the nucleus. Consistent with this result, biochemical analysis showed SENP1 promoted loss of high molecular mass SUMO-1-conjugated species, including the modified species of PML. These results are broadly in agreement with the initial results on SENP1 (Gong et al., 2000 ), although in our hands SENP1 shows more pronounced subnuclear localization in foci, particularly at lower levels of expression. Finally, in co-transfection experiments we found that ICP0 promotes the recruitment of the SENP1 protease to nuclear domains, a result which we also observe early during infection. Taken together the results indicate that SENP1 and its recruitment by ICP0 into PML NBs play an important role in effecting ICP0-induced loss of SUMO-1 from target species, and dispersal of the PML NBs.

Certain previous reports have suggested that SUMO-1 modification of PML is an important targeting signal for PML recruitment to the NB and for NB formation (Ishov et al., 1999 ; Kamitani et al., 1998a ; Lallemand-Breitenbach et al., 2001 ; Zhong et al., 2000 ), and it could therefore have been that SENP1-mediated loss of SUMO-1 from PML was sufficient to disrupt the NBs. However, we show that in the absence of ICP0, the SENP1 deconjugation of SUMO-1 from PML did not disrupt PML localization from the NBs. Our results are consistent with a more recent report showing maintenance of NBs despite the SENP1-induced loss of SUMO-1 modification on PML (Gong et al., 2000 ). Moreover in unpublished work, we have shown that a mutated version of PML, which could not be modified by SUMO-1, was nevertheless recruited to NBs with similar efficiency to the parental PML species. Although the activity of PML within the NBs could certainly be affected by its conjugation status, our data imply that removal of SUMO-1 is not sufficient to disperse PML and that other functions may be required to accelerate this process.

In cells constitutively expressing the epitope-tagged version of SUMO-1, hmSUMO-1 was distributed in a diffuse nuclear pattern with some accumulation in punctate foci. Western blot analysis demonstrated that the vast majority of hmSUMO-1 was also present in high molecular mass conjugates, indicating, perhaps not unexpectedly, that the diffuse nuclear SUMO-1 pattern represented numerous conjugated species. The restriction of SUMO-1 to the nucleus is consistent with previous results indicating that SUMO-1 conjugation is either a nuclear event or coupled to entry. While there may well be shuttling of SUMO-1-modified species from the nucleus, in the equilibrium position the majority of SUMO-1-modified species appears to be in the nucleus.

Several non-mutually exclusive proposals can be envisaged relating SENP1 recruitment by ICP0 to NBs, with the loss of SUMO-1-conjugated species and of NBs. It is possible that upon SENP1 recruitment the key target for deconjugation is PML itself. Having lost SUMO-1, PML would then be degraded, with this event requiring additional distinct events (see above), possibly ubiquitin conjugation catalysed by the ICP0 RING finger (Boutell et al., 2002 ; Everett, 2000 ). After the disruption of the NBs, it could then be that generalized loss of SUMO-1-conjugated species is a downstream event, a consequence of the loss of the major site, rather than SENP1 actively deconjugating multiple numerous species. It could also be, however, that with the recruitment of SENP1 to the NBs, SENP now catalyses the loss of SUMO-1 from the resident PML and likewise from numerous species which are trafficking through the NBs. In either model though, the PML NBs act as a central processing centre, regulating SUMO-1 status of substrates which dynamically traffic through the sites by conjugation/deconjugation. Clearly, it is possible that ICP0, in addition to a generalized loss of SUMO-1-modified species, may promote the loss and degradation of specific protein conjugates. Indeed, during HSV-1 infection the loss of modified PML was accelerated compared to the generalized loss of SUMO-1-modified isoforms (Everett et al., 1998 ). Furthermore, analysis of the activity of other herpesvirus homologues of ICP0 on the NB protein Sp100 suggested that loss of SUMO-1 occurred first, prior to a general increase in unconjugated proteins (Parkinson & Everett, 2000 ).

Whilst significant evidence has accumulated indicating that HSV-1 promotes the loss of PML NBs by promoting the degradation of PML via the proteasome-dependent pathway (Everett, 2000 ; Everett et al., 1998 ), it has proved difficult to demonstrate that PML is ubiquitinated (even in the presence of ICP0). Nevertheless, PML (and other SUMO-1-modified proteins) are likely to be modified by ubiquitin conjugation. It may be that PML is co-modified by ubiquitin and SUMO-1, and that in order for PML degradation to occur SUMO-1 has to be first removed. Alternatively, SUMO-1 may have to be removed prior to ubiquitin conjugation on the same (or different lysines). SENP1-deconjugation activity could participate in either of these models. Our preliminary analysis with certain ICP0 mutants which lack the RING finger, an important component in their ubiquitin-ligase activity (Boutell et al., 2002 ), indicates that they are still able to recruit SENP1 (unpublished data). Previous data showed that these mutants were unable to promote degradation of PML or loss of NBs (Everett et al., 1998 ; ORourke et al., 1998 ), suggesting that recruitment of SENP1 by itself is not sufficient for degradation of the deconjugated species, and that additional factors linked to the ubiquitination pathway are involved.

There is presently little information on the activity of the mammalian SUMO-specific proteases. One recent paper (Hang & Dasso, 2002 ) examined SENP2 with results that, in contrast to SENP1 (this work, see also Gong et al., 2000 ), indicated that SENP2 was specifically located to the inner side of the nuclear membrane. This localization required an N-terminal domain not conserved in SENP1 which was shown to confer binding to a nuclear pore component. Interestingly, while native SENP2 showed little activity in vivo, a deletion mutant which lacked the nuclear rim-targeting region showed significantly increased activity in deconjugating SUMO-1 from target proteins. Therefore, both SENP1 and SENP2 show activity on SUMO-1 conjugates in vivo, but it is currently unclear for either protein whether they exhibit target selectivity with regard to the protein or the particular SUMO-1 species. In conclusion, while further work remains to be done on this class of proteins, our results indicate that SENP1 recruitment may be an important (though not sufficient) role in the modulation of SUMO-1, PML and NBs by ICP0. The cell line constitutively expressing epitope-tagged SUMO-1 should also prove useful in further characterization of SUMO-1, its regulation by SENP1 and its interaction with ICP0. The identification of target species and an understanding of how modification is regulated by specific signals should aid in understanding events occurring during virus infection.

We are grateful to Roger Everett for supplying antibodies for the detection of ICP0 and a plasmid containing Myc-tagged SUMO-1 and Paul Freemont for originally supplying the Myc-tagged SUMO-1 clone (pMLV-Myc-PIC1) and for supplying the anti-PIC-1 antibody. This work is supported by Marie Curie Cancer Care.