In our previous studies we showed that Nse1, Nse3 and Nse4 formed a sub-complex within the highly conserved SMC5-6 protein complex and that Nse3 was structurally homologous to the MAGE protein family 
. We have now refined the architectural definition of this sub-complex and focussed on the Nse3/MAGE protein. We have identified a surface on Nse3 that interacts with Nse4 and a structural domain of Nse3 that interacts with Nse1. This analysis is based on modelling the structure of Nse3 onto the structure of MAGEA4 and G1 deposited in the Protein Database. The validity of our conclusions obviously depends on the accuracy of our modelling. The high level of sequence similarity between MAGE proteins and Nse3 together with the internal self-consistency of our observations gives us confidence that our modelling is reasonably accurate. The interacting region between S. pombe
Nse1 and Nse3 that we have defined, based on our two-hybrid and modelling analysis, corresponds well with that deduced from the crystal structure of the orthologous human NSE1-MAGEG1 
. Furthermore NSE1 and the hydrophobic cleft on Nse3/MAGEG1 that we predict forms the interaction surface with Nse4 are positioned on the same face of Nse3/MAGEG1. We predict that Nse1/NSE1 and the hydrophobic cleft together form a pocket in which the N-terminus of Nse4/NSE4a/4b is located, as shown schematically in and .
We have expanded our findings into mammalian systems. We showed previously that there were two NSE4 paralogs in mammals 
. Using yeast 2-hybrid analysis and co-immunoprecipitation, we have demonstrated that mutations in MAGEG1 corresponding to those that reduced the interaction with Nse4 in S. pombe
, also reduced the interaction of MAGEG1 with NSE4b. To gain further insights into the functional significance of the NSE4b/MAGEG1 interaction, we used a transcription activation reporter system. Intriguingly, there was a synergistic interaction on transcription activation between MAGEG1 and NSE4b (though not between MAGEG1 and NSE4a – unpublished data), and it was reduced in MAGEG1 mutants that diminished the interaction between MAGEG1 and NSE4b. In our experimental system, we think that this transcriptional activation most likely results from a binary “free” complex of NSE4b and MAGEG1. However it raises the question of whether it can also occur in the context of the SMC5-6 complex. This would indicate a novel role for the SMC5-6 complex in transcriptional activation. Further studies are required to resolve this issue.
The evolutionary diversification of the MAGE protein family is remarkable. There is only a single member in fungi, insects 
, birds 
, fish and plants 
, and its most likely function is as part of the SMC5-6 complex. In non-placental mammals there is one member in platypus and two in opossum. In contrast, in placental mammals, there are 33 (+22 pseudogenes) in man, a similar number in mouse and even more in elephants (JP, unpublished data). We showed previously that MAGEG1 is the only MAGE protein detected in the SMC5-6 complex, and that MAGEF1 could not be integrated into the complex 
. This is consistent with our finding that MAGEF1 does not interact with NSE4a or b (). Instead MAGEF1 protein can form complexes with EID proteins (which lack the C-terminal WHD domain essential for binding to the SMC5 head domain). Remarkably we found that most of the MAGE proteins that we examined interacted with both NSE4a and NSE4b (). However, with the exception of the MAGEG1 interactions, the MAGE-NSE4 interactions do not take place in the context of the SMC5-6 complex, since neither NSE1 nor SMC6 is found in the immunoprecipitates (, data not shown). Consistent with our results, Doyle et al. found that most of the MAGE proteins that they examined were unable to interact with NSE1 
. We have shown that Nse1 stabilizes the interaction between S. pombe
Nse4 and Nse3 (, 
), and the same is probably the case for the human orthologs. Without NSE1, it is likely that the MAGE-NSE4 subcomplexes are not able to bind to the SMC6-SMC5-NSE2 subcomplex (). Furthermore, we previously showed that not only Nse4 but also Nse3 (as well as Nse5 and Nse6 in S. pombe
) bound to the head domain of Smc6 (
; K. Bednarova unpublished data). We speculate that the MAGE proteins (other than MAGEG1), have lost their ability to bind to the SMC6 head domain and to NSE1. The evolutionary diversification of such a binding surface(s) then resulted in a gain of new binding partners and the formation of novel MAGE complexes with RING-finger proteins () (
; our unpublished data).
Remarkably, the EID family shows a similar pattern of evolutionary diversification to the MAGE family, albeit to a less dramatic extent, namely a single member (Nse4) in most eukaryotes up to non-placental mammals (although there are two in the plant Arabidopsis thaliana
) and four members in placental mammals. The fifth member, EID2b, is found only in rodents and primates. Our finding that, of the pairs that we examined, most MAGE proteins interacted with most of the EID proteins () suggests that the diversification of these two protein families may be connected.
Interestingly, two tumour-related mutations in MAGE proteins were described recently. In MAGEA1 Glu217 (corresponding to Phe235 in yeast Nse3, ) was mutated to Lys in a melanoma sample 
. We speculate that this change could disturb the MAGEA1 binding to NSE4/EID partners. Similarly in MAGEC1 Ile1001 (corresponding to Met214 in yeast Nse3, ) was mutated to Phe in glioblastoma multiforme cells 
. Although this change is less severe, it could change the affinity and/or specificity of the binding of MAGEC1 to its putative NSE4/EID partner.
The physical interaction between the MAGE and EID proteins raises the question of their functional significance. In contrast to the broadly similar physical interactions between members of the two families, their effects in the transcriptional activation reporter system were quite different. In the EID family, only EID1 repressed transcription in the Gal4-SF1 system in HEK293 cells. Of the MAGE proteins examined, MAGEA1 and D4b were strong transcription co-activators, whereas several other MAGE proteins had little effect. There are various reports in the literature on the effects of MAGE proteins on transcription systems. MAGEA1 represses transcription mediated by Ski interacting protein 
, whereas Wilson and co-workers reported that MAGEA11 increased the transcriptional activity of the androgen receptor 
via an interaction with p300 
. MAGED1 was shown to be a co-activator of the RORα and RORγ proteins, but this co-activation did not require the MHD of MAGED1 
. We found that, when EID and MAGE proteins were co-expressed, EID1 reversed the co-activation mediated by MAGEA1 and MAGED4b, whereas it had no effect on the much lower activation in the presence of necdin. The latter result agrees with the finding of Bush and Wevrick 
. Our results suggest a relatively specific functional interplay between MAGE and EID proteins which contrasts with the general physical interactions that we have observed. It is evident that other proteins interacting with these partners may influence the transcription level. Much more detailed studies need to be carried out in future work in order to unravel the nature of these complex interactions and to understand the functions of these two protein families in their normal cellular contexts.
In conclusion, we have found that, despite the evolutionary diversification of the MAGE family, the characteristic hydrophobic surface shared by all MAGE proteins from yeast to humans mediates its binding to NSE4/EID proteins.