Compartmentalization is a well-established mechanism for regulating protein function (14
). Clearly, this critical relationship between cellular targeting and function goes beyond nuclear-versus-cytoplasmic accumulation. For instance, a formidable number of nuclear proteins target specific subnuclear domains, many of which denote specific cellular functions (13
). It is also understood that since localization is dynamic (for instance, proteins may exhibit compartmental “shuttling”), the stoichiometry is critical to overall protein function and an ideal target for protein regulation. In the case of the HTLV-1 Tax protein, localization is influenced by the presence of both NLS and NES sequences, as well as by the action of posttranslational modifications such as sumoylation and ubiquitylation (1
). Our results from the current study provide an additional regulatory sequence, the TSLS, which allows for accumulation of Tax within the subnuclear structure TSS. We have further shown that dimerization of Tax is required for overall nuclear accumulation of this important viral protein.
Although nuclear localization is delineated from cytoplasmic localization by a clear nuclear membrane, subnuclear localization is largely dictated by interactions with protein complexes present in the nucleus, with membership in these complexes defining the specific structure. All of the well-known subnuclear structures, represented by nuclear speckles (36
), paraspeckles (36
), Cajal bodies (36
), gems (20
), and ND10/PML bodies (4
), are dynamic structures that contain characteristic sets of nuclear proteins and protein complexes that reside in distinct subnuclear regions (46
). Nearly all of these nuclear structures contain subpopulations of nuclear factors, but each is distinguished by the presence of a nuclear protein(s) unique to each structure (36
). One specific class of structures, called nuclear speckles, are interchromatin granule clusters that contain the pre-mRNA splicing machinery, including small nuclear ribonucleic proteins, non-small nuclear ribonucleic protein splicing factors, and spliceosome subunits such as SC35 (59
). We previously observed that Tax localized to specific TSS, so named because of their overlap with nuclear speckles, colocalized with SC35, did not contain promyelocytic leukemia protein, and did not colocalize with nucleoli (3
). However, TSS are distinct from nuclear speckles in that they contain various other nonsplicing cellular proteins. These proteins include DNA protein kinase, Chk2, 53BP1, and γH2AX, proteins involved in DNA damage recognition and repair (15
). The colocalization of Tax with cellular machinery for transcription/splicing, DNA damage response, and checkpoint activation suggests that targeting to TSS is an important integrating event for the varied functions of Tax.
Subnuclear protein targeting to nuclear speckles appears to be mediated by at least two separable trafficking signals: one for nuclear import (the NLS) and one for mediating inclusion into speckles (12
). As might be expected, the identified domains are structurally variable and include examples such as ankyrin repeats in the IκBL, FF domains in CA150, or interaction with RNA and are believed to represent modalities for protein-protein interactions. In our study, we found that the TSLS that is required for localization to TSS was physically distinct from the NLS and could function independently to direct Tax to its specific intranuclear site. There are no remarkable features of the 25-amino-acid TSLS except that it is enriched for proline residues, a characteristic shared by many nuclear speckle proteins, but it does not contain an arginine-serine (RS) motif common to targeting signals of splicing speckle components (8
). This may reflect the unique nuclear address of TSS and the ability of Tax to form novel protein complexes.
It is also important to point out that the 25-amino-acid sequence comprising the TSLS is sufficient for targeting to TSS when fused to either the native Tax NLS or a heterologous NLS. This implies that sumoylation of Tax is not a requirement for either nuclear localization or the formation of TSS. Likely, the role for sumoylation is in the context of the whole protein and may involve the removal of a block to entry or facilitate protein-protein interactions that enable nuclear accumulation. Such a model is reminiscent of that established for the tumor suppressor Wilms' tumor gene (WT1) in which modification by SUMO proteins is separable from nuclear speckle localization (57
). As with Tax, although sumoylation is required for nuclear entry of WT1, an independent sequence is sufficient for targeting to nuclear speckles in the absence of a SUMO protein target. It may also be that sumoylation serves to mask the function of the NES, as has been suggested previously (37
Our results also demonstrate that the dimerization of Tax is a necessary prerequisite for nuclear localization/accumulation. There are numerous examples of proteins that must dimerize or oligomerize prior to nuclear translocation, including the human cytomegalovirus protein ppUL44 and the cellular protein p53 (2
). Other proteins, such as those of the AP-1 family, including c-Jun, JunD, JunB, and c-Fos, enter the nucleus as monomers but require heterodimerization in order to remain in the nucleus (40
). In our studies, we were able to link the inability of Tax mutants to dimerize with a failure to accumulate in the nucleus in spite of a competent NLS. In fact, the addition of a heterologous NLS failed to rescue this defect. The relationship between dimerization efficiency and nuclear accumulation is linear in that weakly dimerizing proteins were able to weakly accumulate in the nucleus. Specifically, we generated a deletion of two subdomains within the Tax dimerization domain as well as single-subdomain mutations. The single-subdomain Tax mutant retained a slight dimerization ability, whereas the Tax mutant in which the larger dimerization domain is removed showed no dimerization. The rescue of single-subdomain dimerization mutants by wild-type Tax protein supports a complementation phenotype, and our observation that none of the subdomain mutants could complement distal subdomain mutants implies a requirement for a dimer interface. Finally, we showed that induction of dimerization in a previously cytoplasm-restricted Tax dimerization subdomain mutant resulted in restored nuclear accumulation. Although our studies definitively link dimerization as a prerequisite to nuclear accumulation, additional studies are required to determine if this dimerization occurs in the cytoplasm as a prerequisite to nuclear entry or within the nucleus as a means of retaining the protein within its subnuclear address.
Previous studies on the dimerization of Tax have suggested that the N-terminal zinc finger domain (51
) is important for Tax self association (5
). These findings were based on yeast two-hybrid assays where an N-terminal deletion of Tax failed to interact with wild-type Tax. In mammalian cells, as used in our system, deletion of the zinc finger region (Δ30-52) was rescued for nuclear localization by the coexpression of wild-type protein, demonstrating the ability to interact. Our functional complementation assay between the NLS mutant, the Δ30-52 mutant, which is missing the zinc finger domain, and the TSLS mutant, the Δ50-75 mutant, demonstrated that the zinc finger domain is not required for dimerization, since a complementation phenotype requires interaction. Likewise, the fusion of a heterologous NLS to the Δ30-52 mutant was able to restore nuclear localization and subsequent TSS accumulation, providing further evidence that the zinc finger domain is dispensable for Tax self association in our system.
Our results also allow for some inferences regarding the stoichiometry of Tax self association. The complementation assays between the NLS mutant and TSLS mutant suggest that at the minimum, a single copy of each, NLS or TSLS, is required for nuclear accumulation of Tax in TSS. Thus, Tax exists at least in a one-to-one dimer state. In addition, the inability of the dimerization subdomain mutants to complement distal subdomain mutants implies that at least two intact molecules of Tax are needed to form the required dimer interface. This finding is supported by findings in the induced dimerization assay. The design of the heterologous “dimerizer” domain is such that only dimers, and not nonpaired higher-order oligomers (for instance, trimers), are formed upon the addition of the chemical dimerizer. Although induced dimerization was able to increase the nuclear accumulation of the mutant protein, it did not result in a total restoration of nuclear accumulation to wild-type levels. This suggests the necessity of the presentation of specific dimer interfaces for normal Tax nuclear accumulation. Perhaps these specific interfaces mediate interaction with specific proteins. Interestingly, the ability of wild-type Tax to rescue dimerization subdomain mutants suggests the existence of a higher-order (>2) oligomeric structure. This conclusion derives from the reasoning that if dimer interfaces are required for nuclear accumulation and the mutants are incapable of forming interfaces, then wild-type Tax must provide them. In order for this to occur, the rescued Tax complex must contain at least 2× molar equivalents of the wild-type protein. Thus, it is likely that Tax is capable of forming a homotetrameric complex. Higher-ordered Tax oligomerization could provide for more-complex regulation of Tax functions. Oligomerization could generate new intermolecular interfaces to improve stability, control the accessibility and specificity of active sites, and increase the number of cellular binding partners for Tax. The complexity of such a model helps explain how a single protein displays such a wide range of activities.