We recently reported the first purifications of PRAME-containing protein complexes by epitope-tagged immunoprecipitations and mass spectrometry, which revealed that PRAME is a BC-box substrate receptor component of Cullin2-based E3 ubiquitin ligases 
. In the present study, we identified novel interactions of Cul2-PRAME ligases with the human EKC/KEOPS complex, we discovered that PRAME recruits ligase components to EKC, and we showed that these proteins associate with the same promoter regions, strongly suggesting a functional link between these complexes in the same pathway.
EKC is an ancient and extremely conserved complex with orthologues in Archaea and Eukarya. It has been implicated in telomere maintenance, transcriptional regulation, and t6
A modification of tRNAs in yeast 
. Our interaction data support a linear architecture LAGE3-OSGEP-TP53RK-TPRKB for the human EKC complex and dimerization around the LAGE3 subunit, with two OSGEP moieties per complex. Our results are fully consistent with recently reported crystal structures of archaeal EKC proteins 
Consistent with the notion that all subunits contribute to the activity of the complex, yeast EKC was reported to constitute a stable entity. Downey et al. reported that deletion of either Bud32p, Cgi121p, or Gon7p resulted in shorter telomeres. An independent study found that Pcc1p was required for efficient transcription of several yeast genes induced by pheromones or galactose, and that deletion of Pcc1 inhibited the recruitment of the SAGA and Mediator co-activators 
. Dimerization of the Pcc1p and Kae1p subunits was found to be necessary for cell viability 
, and an intact interaction surface between Kae1p and Bud32p was required for both the transcriptional and telomere maintenance functions 
. Most recently, the yeast EKC complex was shown to be essential for a universal chemical modification of tRNAs, called threonyl carbamoyl adenosine (t6
. This modification occurs at position 37 of tRNAs decoding ANN codons, including the first ATG of all protein-coding mRNAs, and is required for efficient translation 
. The t6
A modification activity was dependent on Pcc1p, the ATPase activity of Kae1p, and the kinase activity of Bud32p, but not on Cgi121p 
In our experiments, all human EKC orthologues co-purified together and with similar abundance both in K562 and HeLaS3 cells, demonstrating for the first time that a complete and stable EKC complex is present in human cells. However, we consistently found that PRAME complexes contained primarily OSGEP and LAGE3, and very low or undetectable levels of the other two subunits TP53RK and TPRKB. This might reflect a PRAME-mediated disruption of an otherwise stable EKC complex, or a preferential binding of PRAME to an already existing (and possibly dynamic) EKC submodule that was not previously detected in yeast. It is tempting to speculate that PRAME could mediate ubiquitination of EKC components via recruitment of the Cul2 E3 ligase complex. In an attempt to test this hypothesis, we performed ubiquitination assays upon transient transfections in HEK293 cells in presence or absence of the proteasome inhibitor MG132, and in vitro assays with purified proteins. As expected for BC-box proteins, we detected polyubiquitin chains associated with PRAME itself. However, the high background signals in these assays did not allow us to detect a putative ubiquitination of EKC subunits. Hence, our assays could not rule out that EKC might be ubiquitination targets of Cul2-PRAME, and more efforts are needed to test this hypothesis experimentally.
Interestingly, dynamic regulation of EKC components has been reported: the kinase activity of Bud32p was repressed by association with Kae1p 
, but promoted by binding to Cgi121p 
. Moreover, we found that the protein levels of OSGEP can be significantly modulated by association with other proteins (), with LAGE3 exerting the most pronounced positive effect, followed by TP53RK and PRAME. Interestingly, the effect of LAGE3 correlates with the formation of tetrameric complexes in which an LAGE3 dimer binds two OSGEP moieties (), which are likely not in direct contact with each other 
The results of our transient transfection experiments and those obtained using the baculovirus system suggested (i) that PRAME is able to interact independently with LAGE3 and OSGEP and (ii) that LAGE3 enhances the association of OSGEP with PRAME. Considering the dimerization properties of the LAGE3-OSGEP submodule, we therefore considered the possibility that multiple PRAME moieties might be present in the same protein complex. In this case, both epitope-tagged PRAME and endogenous PRAME should be present in the same protein complex in K562-TAG-PRAME cells, since the expression level of the two proteins were similar. Contrary to this hypothesis, however, TAG-PRAME did not co-immunoprecipitate endogenous PRAME in our experiments (see lane 4). Although we cannot formally exclude the possibility that TAG-PRAME might compete with the endogenous protein, our data are most consistent with the idea that there is only one PRAME molecule per complex. Notably, our data suggest that the interaction between PRAME and LAGE3-OSGEP is very stable, since it was efficiently detected in all assays and despite the long incubations times in our protein-complex purification protocols. However, we cannot exclude that high levels of PRAME might be required for the interactions to take place in cells. Taking into account the published crystal structures of archaeal EKC, we propose the models depicted in for the architecture of the human EKC complex and the relationship between EKC and Cul2-PRAME ligases. Further experiments are needed to confirm the stoichiometry of PRAME-EKC complexes, and to characterize the interaction surfaces involved.
Besides the functional experiments in yeast, no information is available yet about EKC functions in other organisms. At present, only studies focusing on single subunits of the human complex have been reported. LAGE3 appears to be ubiquitously expressed, while the two closely related LAGE1 and LAGE2 are cancer-testis antigens with high expression confined to healthy testis and cancer tissues 
. Interestingly, expression of the N-terminal domain of LAGE3 fused to the C-terminus of Pcc1p could partially rescue the growth defects of a Pcc1 N-terminal mutant yeast strain, while full length LAGE3
could not substitute for Pcc1 
. The OSGEP
gene was found expressed in all somatic tissues tested and it lies in a head-to-head orientation with the tightly regulated APEX
gene, which encodes a multifunctional DNA repair enzyme 
. Analyses of their bidirectional promoter identified a CCAAT box crucial for OSGEP
transcription. Futhermore, OSGEP
could partially rescue the growth defects of the yeast kae1
Δ mutant 
, underlining degrees of functional conservation. Human TP53RK
was first cloned by cDNA subtraction as an interleukin-2 upregulated gene in cytotoxic T-cells, transcripts were detected in a number of normal tissues and cancer cell lines, and it was shown to phosphorylate the human oncoprotein p53 at Serine 15 
. Recombinant TP53RK expressed in bacteria was found catalytically inactive, but incubation with COS-7 cell lysates was sufficient to activate the enzymatic activity, suggesting the need for activators and potentially protein interactors. The kinase Akt/PKB was found to play a role in the activation of TP53RK through phosphorylation of the Ser250 residue 
. Underscoring a high level of functional homology, TP53RK could partially rescue the growth phenotype of Bud32 deletion in yeast, and Bud32 could interact with and phosphorylate human p53 in vitro 
, although yeast does not possess a bone fide
p53 homologue. Human TPRKB
was first isolated by a two-hybrid screen using TP53RK as a bait, and was suggested to inhibit the binding of TP53RK to p53 
. Expression of TP53RK
was detected in the high-PRAME expressing cell line K562 as well as in the low-PRAME expressing cell line KG1 
. Gene expression data obtained from the GeneSapiens database further suggests that EKC subunits are expressed in most if not all human tissues (Figures S2, S3, S4).
Considering the extreme conservation of EKC components during evolution, it can be expected that at least some of the functions and pathways already identified for this complex in yeast could be valid for human cells as well. However, it cannot be excluded that human EKC acquired additional, yet to be identified functions during evolution. The absence of efficient knock-down or knock-out models have so far hindered advanced functional analysis of EKC in higher eukaryotes. Future studies will have to address these issues, and in particular to understand the connections between apparently different functions ascribed to this complex, like phosphorylation of the p53 oncogene, chemical modification of tRNAs, and physical association with promoter regions.
Kae1p and Pcc1p have been found to associate with active chromatin and to regulate expression of some of the yeast promoters to which they are recruited 
. A pcc1-4
mutation prevented activation of GAL1
by impairing recruitment of the SAGA histone acetyltransferase and mediator without affecting recruitment of the activator Gal4p. In contrast, while Pcc1p was recruited to the constitutively active PMA1
and the heat-shock inducible HSP104
genes in a transcription-dependent manner, transcription of these genes was not affected in the pcc1-4
Our ChIP assays revealed that LAGE3 and OSGEP are also present on chromatin and, in particular, that they colocalize with PRAME at many of its target sites, consistent with our protein-protein interaction data. Since neither PRAME nor EKC proteins possess known DNA-binding domains, it is not clear whether they can bind DNA directly. Genome-wide ChIP-seq experiments have recently revealed that PRAME associates with transcriptionally active NFY-regulated promoters and nearby enhancers 
. Therefore, the transcription factor NFY might play a direct or indirect role in the targeting of a Cul2-PRAME-EKC complex to chromatin. It is also not clear whether there is a hierarchy in the association of PRAME and EKC proteins with chromatin. In order to address this point, we established stable cell lines with downregulation of endogenous PRAME using the published retroviral pRetroSuper system in either K562 wild type or K562-TTE-LAGE3 cells and we performed ChIP experiments. Unfortunately, despite a 70% downregulation of PRAME mRNA levels, our preliminary results did not support a solid conclusion on the hierarchy of binding to chromatin for PRAME and EKC. More experiments, and likely in a different cell and knockdown system, are required to address this point.
Importantly, EKC proteins have an ancient evolutionary origin and seem to be ubiquitously expressed in adult tissues, while PRAME appeared more recently in evolution, and its expression seems to be tightly confined to particular developmental stages and tissues and to be particularly deregulated in oncogenesis. These characteristics are consistent with a model in which EKC can associate with chromatin independently of PRAME. Only when PRAME is expressed would a Cul2-PRAME ubiquitin ligase be recruited to chromatin by virtue of specific protein-protein interactions between LAGE3-OSGEP and PRAME.
In an attempt to study the putative function of EKC subunits on target gene transcription, we established K562 cells with stable downregulation of OSGEP or LAGE3 using pRetroSuper vectors and performed genome-wide mRNA-seq analyses. Despite a 50–80% knockdown efficiency for OSGEP and LAGE3, we did not detect genes with significant expression changes when compared to a GFP knockdown control. Our inconclusive data could of course be the result of inefficient downregulation of the target genes. Although the K562 cell line was successfully used to dissect the protein-protein interaction network, we cannot exclude that this cell line might not be suitable for the study of the finer putative chromatin functions and signaling of EKC and PRAME.
The development of ChIP-grade antibodies against EKC components will be a crucial step to address important questions as to whether PRAME target sites are a subset of EKC sites (or viceversa), and to address the functional relationships between EKC, Cul2-PRAME ligases and NFY on transcriptional regulation of target genes, and potentially other pathways like the recently described modification of tRNAs.
Taken together, our data reveal a novel link between the human oncoprotein PRAME and the ancient EKC complex. Our data clearly show that PRAME can recruit Cullin2 ubiquitin ligases to EKC, and that these complexes associate with the same genomic regions, supporting a functional link in the same pathways. Our results add a new twist to both PRAME and EKC biology and provide an important contribution to understanding the molecular mechanisms in which these proteins are active, both in healthy and cancer cells.