The classical nuclear protein import pathway is mediated by the α and β subunits of importin. Importin α functions as an adapter molecule by binding both importin β and the NLS-bearing import substrate. While only one importin β isoform has been found in humans thus far, six human genes for importin α exist, including that for importin α7, newly described here. Some of these importin α isoforms may even occur in different versions. We found partial cDNAs in the database corresponding to importin α6 and importin α7, which may represent alternatively spliced mRNAs. Those cDNAs would code for proteins with amino termini three amino acids shorter or longer than those published, respectively. As reported previously (23
) and now confirmed in more detail here, the relative expression levels of a particular importin α form may vary between different tissues, indicating a special demand of different cell types for specific importin α’s. Nevertheless, all human importin α proteins can be found in various tissues with the exception of importin α6, which has thus far been found only in testis. Even more striking, these proteins are expressed simultaneously in many cell lines, such as HeLa cells or human umbilical vein epithelial cells (reference 23
and data not shown).
The radiation of ancestral importin α genes into different orthologues probably occurred several times during evolution (Fig. B). Most of the known importin α forms, such as the six mammalian importin α’s, belong to one of three main subgroups, which differ from one another in about 50% of their amino acids. However, several other species such as C. elegans and rice possess additional importin α-like proteins differing in more than 75% of their amino acids from those isoforms of the main subgroups. Most species examined so far possess at least one protein which belongs to the SRP1-like subgroup. Humans have three different SRP1-like proteins, namely, importin α5, importin α6, and the newly reported importin α7. In contrast to the diversity found in other organisms, Schizosaccharomyces pombe has two genes coding for importin α proteins and S. cerevisiae has only one importin α gene, SRP1. This indicates that one importin α isoform alone may be sufficient to fulfill the basic requirements of a eucaryotic cell.
Although the identity of the primary sequences of the importin α isoforms including the IBB domain varies between 50 and 85%, the proteins do not differ dramatically in their interactions with their transport receptors CAS and importin β. The human importin α proteins bind CAS with a KD
between 2 and 5 nM. These differences are small and could be caused by variations in the quality of the protein purification process. Whereas the binding affinity of Xenopus
importin α2 turned out to be in the same range as that of the human importin α forms, yeast SRP1p showed a clearly less efficient binding affinity to CAS (KD
> 20 nM). Since yeast SRP1p is less homologous to the human importin α proteins than Xenopus
importin α2 is, the weaker binding affinity of yeast SRP1p to human CAS was not surprising. No binding to Crm1/exportin was observed. These results fit with recently reported data from Herold et al., who found similar binding affinities of importin α1/Rch1, importin α5/hSRP1, and importin α4 to CAS but not to Crm1 in two-hybrid studies (18
). The KD
s for the interactions between importin β and the various importin α forms in the Biacore assay are also in a nanomolar range (between 3 and 18 nM). These differences are unlikely to have an influence on the in vitro assay, where the proteins are present in micromolar concentrations. Again, these small differences may be caused by variations in the quality of the protein purification process. In addition, the coupling of the proteins to the sensor chips may impair the importin α forms to different extents.
The simultaneous existence of several highly divergent importin α proteins in a given cell led to the question whether they might be specialized in their efficiency to transport different nuclear proteins. Several experiments clearly support this hypothesis. Sekimoto et al. recently reported that intracellular injection of antibodies against importin α5/hSRP1, but not against importin α1/Rch1, can inhibit nuclear import of the transcription factor Stat1 (43
). Fisher et al. reported that the Epstein-Barr virus protein EBNA1 interacts with importin α1/Rch1 but not with importin α5/hSRP1 in the yeast two-hybrid system (11
). By pull-down assays with different NLS-BSA conjugates, Nadler et al. showed that importin α1/Rch1 and importin α5/hSRP1 share distinct binding affinities for various NLSs (33
). Finally, Miyamoto et al. demonstrated that the efficiency of nuclear import of different NLS-reporter protein conjugates in vitro may depend on the importin α protein present in the assay (28
). Thus, earlier studies indicated that there might exist substrate specificities for the different importin α proteins. Therefore, we compared the import activities of all known ubiquitously expressed human importin α proteins on different artificial and natural substrates by using a defined in vitro import system. For comparison we also included frog importin α2, a paralogue of human importin α1/Rch1, and yeast SRP1p in our study. Most substrates were imported with about the same efficiency by all importin α isoforms, if they are added as single substrates to the assay. Nevertheless, significant differences between the different isoforms were detectable, and the nuclear import of RCC1, a protein that is strictly localized within the nucleus, showed a very strong dependence on the presence of one particular importin α isoform (importin α3). The addition of two differently labeled substrates into one import reaction mixture clearly demonstrated that those differences are unlikely to be caused by variations in the quality of purification of the recombinant proteins. For example, the very strong effect of importin α1/Rch1 on the nuclear import of hnRNP K demonstrates that its weak import efficiency on nucleoplasmin in the same assay reaction is not due to a functionally inactive protein. Only yeast SRP1p usually showed a weak import efficiency, probably because its evolutionary distance from the human proteins fostered an impaired interaction with the mammalian substrates in our import assays.
It is unlikely that differences in binding affinity between substrate and importin α are the main reason for the observed effects in the import assay. Although experiments using SPR demonstrate that RCC1 binds significantly better to importin α3 and α4 (KDs ~9 nM) than to the other isoforms (KDs ~18 to 30 nM) while nucleoplasmin shows no significant differences in its binding to the various importin α forms (KDs ~4 to 8 nM) (data not shown), these differences are too small to have an influence on the in vitro import assay where substrates and importin α’s are present in micromolar concentrations.
Some of our results disagree with data obtained by other groups. Nachury et al. (32
) reported that importin α4/hSRP1γ had a weaker efficiency to import NLS-BSA in vitro than importin α1/Rch1. We could not detect those differences. In contrast to Nachury et al. (32
), who obtained importin α4 from HeLa cells via vaccinia virus infection and importin α1/Rch1 from E. coli
, we purified all importin α proteins from the same system (E. coli
), which might explain those differences. Moreover, Miyamoto et al. (28
) reported that a CBP80-allophycocyanin fusion protein becomes imported in their in vitro import system by importin α1/Rch1 and by importin α5/SRP1 but not by importin α3/Qip1. In our hands, recombinant human cap binding protein is imported by all human importin α proteins tested (data not shown). This finding demonstrates that artificial NLS fusion proteins and their corresponding full-length proteins may behave quite differently in the import assay system and that testing the functional activity of the purified import factors is very important for comparison of their efficiencies. Furthermore, the source of the recombinant proteins might influence the result to some extent.
If one adds two substrates simultaneously, the preference of a particular importin α for a certain substrate can get more clear-cut. For example, if nucleoplasmin and P/CAF are added to one import reaction mixture at the same time, importin α3 is still able to import P/CAF very efficiently. This result is in clear contrast to those for the other importin α proteins, although all of them can still import nucleoplasmin. The situation for RCC1 is similar. This protein is transported efficiently as a single added substrate into the nucleus only by importin α3 and to a less efficient extent by importin α4, which is highly homologous to importin α3. The addition of nucleoplasmin as a second substrate enhances these differences between the members of the importin α3/α4 subfamily and members of the other subfamilies. This observation indicates that one level of the import efficiency regulation in the cell may be the competition between import substrates for their “preferred” importin α form. However, the results of our in vitro studies cannot predict to what extent this concurrence happens in living cells and whether or not different importin α proteins can substitute for each other in vivo.