To address this question, we first examined the importance of residues L86, L87, and F121 for the dimerization of full-length UL44 in living cells. To this end, we tested the ability of full-length forms of UL44 bearing the L86A/L87A or F121A substitutions to relocalize into the cell nucleus green fluorescent protein (GFP)-UL44ΔNLS, a mutant that, lacking a functional NLS, localizes exclusively in the cytoplasm (2). Entry clones pDNR207-UL44-P85G, pDNR207-UL44-L86A/L87A, pDNR207-UL44-F121A, pDNR207-UL44-I135A, pDNR207-UL44ΔNLS-L86A/L87A, and pDNR207-UL44ΔNLS-F121A were generated by site-directed mutagenesis of pDNR207-UL44 and pDNR207-UL44ΔNLS (2, 3) using the QuikChange mutagenesis kit (Stratagene) with the appropriate primers (Table 1, footnote a). Entry clones were then used to perform recombination reactions with the Gateway system (Invitrogen) plasmids pEPI-GFP (24) and pBkCMV-DsRed2 (28) to generate GFP and DsRed2 mammalian cell expression constructs, respectively. The subcellular localization of GFP-UL44ΔNLS, expressed either alone or in the presence of wild-type or mutant DsRed2-UL44 fusions, in COS-7 cells transfected with these plasmids was analyzed by confocal laser scanning microscopy (CLSM) as described previously (3). We have previously shown that this assay can efficiently detect the dimerization of UL44 in living cells and that a large deletion of UL44, as expected, impairs its dimerization (3). When expressed alone, GFP-UL44ΔNLS localized exclusively in the cytoplasm, while its coexpression with DsRed2-UL44 resulted in a marked relocalization of GFP-UL44ΔNLS to the nucleus (Fig. (Fig.1A).1A). DsRed2-UL44-P85G, a mutant which dimerizes in vitro (6), and DsRed2-UL44-I135A, a mutant bearing a substitution that in vitro prevents the binding of UL44 to UL54 (16) but not UL44 dimerization (unpublished data), also colocalized with GFP-UL44ΔNLS into the nucleus (Fig. (Fig.1A).1A). In contrast, coexpression with DsRed2-UL44-F121A or DsRed2-UL44-L86A/L87A did not result in a marked relocalization of GFP-UL44ΔNLS into the nucleus (Fig. (Fig.1A),1A), suggesting that these mutants cannot efficiently dimerize with GFP-UL44ΔNLS.

To quantify the effects of the F121A and L86A/L87A substitutions on UL44 dimerization, we first determined the levels of nuclear accumulation of GFP-UL44ΔNLS when expressed alone or in the presence of wild-type or mutant DsRed2-UL44 fusions in experiments such as those for which the results are depicted in Fig. Fig.1A.1A. Measurement of the nuclear/cytoplasmic fluorescence ratio (Fn/c) values as described previously (3, 4) revealed that coexpression with DsRed2-UL44, DsRed2-UL44-P85G, or DsRed2-UL44-I135A significantly enhanced the nuclear accumulation of GFP-UL44ΔNLS (Fig. (Fig.1C1C and Table 1). In contrast, coexpression with DsRed2-UL44-F121A or DsRed2-UL44-L86A/L87A resulted in a significantly lower nuclear accumulation of GFP-UL44ΔNLS, with the L86A/L87A substitutions having the stronger effect (Fig. (Fig.1C1C and Table 1).

We then wished to examine the effects of the F121A and L86A/L87A substitutions when present in both UL44 monomers, as this would better mimic the situation during the infection with virus containing these mutations. Thus, we introduced the F121A or L86A/L87A substitutions in the GFP-UL44ΔNLS fusion and determined the levels of nuclear accumulation of the mutants when expressed alone or in the presence of DsRed2-UL44 fusions bearing the same mutations. As expected, when expressed alone, GFP-UL44ΔNLS, GFP-UL44ΔNLS-F121A, and GFP-UL44ΔNLS-L86A/L87A localized mainly in the cytoplasm (Fig. (Fig.1A1A and data not shown). Quantitative analysis revealed that both mutant proteins partially entered the nucleus, and did so more efficiently than GFP-UL44ΔNLS (Fn/c of 0.42 and 0.59, respectively, versus 0.20; Table 1), perhaps due to monomers of the UL44 dimerization-defective mutants passively diffusing through nuclear pores due to their smaller molecular mass (3). Importantly, measurement of the Fn/c values upon the coexpression of GFP-UL44ΔNLS-F121A with DsRed2-UL44-F121A and of GFP-UL44ΔNLS-L86A/L87A with DsRed2-UL44-L86A/L87A showed that when the substitutions are present in both monomers, the F121A mutant can dimerize to a larger extent than the L86A/L87A mutant (Table 1), which is consistent with previous biochemical data (6). Nevertheless, the Fn/c values were significantly reduced compared to those of the wild-type constructs (Table 1). Similar results were obtained in quantitative yeast two-hybrid system assays (data not shown). Thus, altogether these results clearly indicate that both the F121A and L86A/L87A substitutions inhibit the dimerization of full-length UL44 in living cells, with L86A/L87A having a more marked effect.

To examine the possibility that the observed effects of the L86A/L87A and F121A substitutions could be due to misfolding, we tested the ability of the DsRed2-UL44 fusions to interact with GFP-UL54(1213-1242), a construct containing the UL54 binding domain for UL44 (4, 15). When expressed individually, GFP-UL54(1213-1242) localized both in the nucleus and in the cytoplasm (Fig. (Fig.1B),1B), due to its small molecular mass and lack of a functional NLS (4). Coexpression with the wild-type DsRed2-UL44 fusion as well as with the DsRed2-UL44-L86A/L87A, DsRed2-UL44-F121A, and DsRed2-UL44-P85G mutants resulted in the extensive colocalization of GFP-UL54(1213-1242) with the UL44 fusions into the nucleus, whereas coexpression with DsRed2-UL44-I135A, as expected (16), did not (Fig. (Fig.1B).1B). The yeast two-hybrid system assays demonstrated that the L86A/L87A and F121A substitutions also do not interfere with the interaction between UL44 and full-length UL54 (data not shown). Thus, we conclude that the defect in the dimerization of the L86A/L87A and F121A mutants is not due to the global misfolding of the protein. Furthermore, these data demonstrate that UL44 dimerization is not required to bind UL54 in cells.

To further examine the behavior of the UL44-F121A and UL44-L86A/L87A mutants in a cellular context, we analyzed the subcellular localization of the GFP-UL44-F121A and GFP-UL44-L86A/L87A fusions, which bear functional NLS's (2), when transiently expressed in COS-7 cells compared to that of GFP-UL44 and the GFP-UL44-P85G and GFP-UL44-I135A mutants as controls. As expected, all UL44 mutants localized into the nucleus like the wild-type protein (Fig. (Fig.2A).2A). However, in most cells, GFP-UL44-L86A/L87A exhibited a diffuse intranuclear pattern, whereas GFP-UL44-F121A, GFP-UL44-P85G, and GFP-UL44-I135A formed nuclear speckles typical of UL44 (2) (Fig. 2A and B). Several UL44 homologues localize with a typical punctate pattern within the cell nucleus, which is believed to be the consequence of their ability to interact with DNA or other nuclear components (1-3, 14, 29). These results are therefore consistent with the finding that the L86A/L87A mutant exhibited a 100-fold lower affinity for DNA in vitro than wild-type UL44 (6). Furthermore, quantification of the levels of nuclear accumulation relative to the GFP-UL44 fusions (Fig. (Fig.2C)2C) revealed that both GFP-UL44-F121A and GFP-UL44-L86A/L87A (but not GFP-UL44-P85G and GFP-UL44-I135A) accumulated into the nucleus slightly but significantly less efficiently than GFP-UL44. This might be explained by UL44 dimers being translocated to the nucleus more efficiently than monomers as a consequence of the presence of two NLS's rather than only one or by UL44 dimers being better retained in the nucleus than monomers due to stronger binding to DNA.

Finally, we investigated the effects of substitutions that impair UL44 dimerization on DNA replication in cells. To this end, the ability of the UL44-L86A/L87A and F121A mutants, and that of UL44-I135A and UL44-P85G as controls, to support HCMV oriLyt-dependent DNA replication was tested by the means of a cotransfection-replication assay as described previously (22, 23). The transfection reactions contained the pSP50 plasmid with the HCMV oriLyt DNA replication origin (5), a plasmid expressing wild-type or mutant UL44 (pSI-UL44, pSI-UL44-P85G, pSI-UL44-L86A/L87A, pSI-UL44-F121A, or pSI-UL44-I135A), and a set of plasmids expressing the remaining HCMV proteins essential for oriLyt-dependent DNA replication (22). The pSI-UL44 plasmid was created by cloning the XhoI/MluI fragment of pD15-UL44 (15) into the XhoI/MluI sites of pSI (Promega). The pSI-UL44-P85G, pSI-UL44-L86A/L87A, pSI-UL44-F121A, and pSI-UL44-I135A plasmids were generated by amplifying pSI-UL44 with mutagenic primers (Table 1, footnote a). The replication of pSP50 was detected by the treatment of transfected cell DNA with DpnI, which cleaves only unreplicated Dam-methylated input DNA (22, 23). As shown in Fig. 3A, a DpnI-resistant replication product was detected in the presence of wild-type UL44, as expected (23); a similar pattern was observed with the P85G mutant, suggesting that it could also support oriLyt-dependent DNA replication. In contrast, oriLyt-mediated DNA replication was not detected in the presence of either UL44-L86A/L87A or UL44-F121A. As expected, the UL44-I135A mutant, which does not interact with UL54 and does not stimulate UL54-mediated DNA synthesis in vitro (16), also failed to complement oriLyt-dependent DNA replication. Western blot analysis showed that all mutants are expressed at levels similar to that of wild-type UL44, suggesting that the inability of the UL44-L86A/L87A, -F121A, and -I135A mutants to support DNA replication is not due to reduced protein expression (Fig. (Fig.3B3B).

Thus, our study strongly suggests that UL44 dimerization is a prerequisite for supporting DNA replication in cells, as mutations that specifically affect the dimerization of UL44 also impair its ability to complement oriLyt-mediated DNA synthesis. It is noteworthy that although the F121A mutation still permits a certain degree of dimerization, this appears to be insufficient to support DNA replication. Since it has previously been shown that UL44 mutants that are impaired in dimerization are also impaired in binding to DNA in vitro (6), our results suggest that in order for UL44 to fulfill its processivity function, it must homodimerize to bind DNA and thereby tether UL54 on the template. Similarly, the DNA-binding activity of herpes simplex virus UL42 is required for the stimulation of its cognate DNA polymerase catalytic subunit (11, 27) and efficient viral DNA replication (10), and the dimerization of the processivity factor PF-8 of human herpesvirus 8 is critical for DNA binding and stimulation of the catalytic subunit Pol-8 (7). However, alternative hypotheses cannot be excluded. Like PCNA, UL44 has been reported to interact with a wide array of other proteins in addition to the DNA polymerase catalytic subunit, including HCMV-encoded UL84, UL97, and UL114 (9, 13, 21, 25). Given its structural similarity to PCNA, the dimeric nature of UL44 could allow for the second subunit to act as a scaffold to recruit other viral or cellular proteins during viral DNA replication. Thus, the mutations that impair UL44 dimerization could affect its binding to proteins directly or indirectly involved in HCMV oriLyt-dependent DNA replication. However, UL44 mutants bearing the F121A or L86A/L87A substitutions can still interact with UL54, and preliminary data suggest that these mutants also still bind certain other viral proteins (unpublished data). The impairment of unknown interactions with viral or cellular proteins cannot, of course, be excluded.

Whatever the mechanism may be, our results strongly argue that compounds that interfere with UL44 dimerization could block viral DNA synthesis and thereby effectively prevent HCMV replication. Thus, it is our hope that this will lead to the discovery of clinically useful anti-HCMV drugs and will also suggest similar strategies for inhibiting other herpesvirus polymerases or, more generally, other targets (17, 18).