We employed three distinct experimental approaches for the purification and identification of c-Myc complexes. All were based on the ectopic expression of a TAPMyc fusion protein in a human fibroblast cell line. We verified that the TAPMyc fusion protein we constructed was biologically active and that it conferred the expected phenotype in the cell line used for purification. Our first approach (designated two-step) employed tandem affinity purification on IgG beads followed by Calmodulin beads. In the second approach (designated label-free) we performed only one round of purification using IgG beads. In the third approach (designated SILAC) we added metabolic labeling with 13C lysine and arginine prior to the single-step purification on IgG beads. Four independent biological replicates were performed in both the label-free and SILAC series of experiments. All purifications were processed using MudPIT followed by MS/MS analysis.
We found 48 proteins to be specifically enriched in the two-step experiment (; also see Materials and Methods for definitions of enrichment), 320 proteins in the label-free experiments (Sup. Table 1
) and 87 proteins in the SILAC experiments (). Removing Myc from these lists provides 418 non-redundant proteins as the combined total of putative c-Myc interactors identified in this study (Sup. Table 3
). Our combined dataset of 418 proteins shares 27 proteins with the 220 protein dataset of Koch et al.9
(12% overlap) and six proteins with the 71 protein HPRD dataset of known c-Myc interactors.8
(8% overlap; Sup. Table 3
). The overlaps between our three experimental datasets range from 8% to 29% () with 5 proteins being common to all three datasets. Two (40%) of these proteins (MAX, TUBA4A) are found in both the HPRD and Koch et al. datasets. Three proteins (ERBB2IP, PRKCDBP, UBC) are novel interactors. Twenty-four additional proteins are shared by two datasets ( and Sup. Table 3
), and among these are twenty novel interactors. Another useful way to evaluate the data is to consider the proteins that scored consistently in the biological replicates: 33 interactors were present in at least three out of four biological replicates in the label-free experiments (Sup. Table 1
), and 31 proteins scored in that category in the SILAC experiments (). Of the 29 interactors shared between two (or more) experimental approaches, 14 (50%) were present in three (or more) biological replicates in the label-free or SILAC experiments, indicating that these top scoring hits are being repeatedly and consistently identified.
Figure 6 Venn diagram of the relationships between the two-step, label-free and SILAC series of experiments. The overlap between the two-step and label-free datasets is 9 proteins (19% of two-step), between the two-step and SILAC datasets 7 proteins (15% of two-step), (more ...)
Of the novel proteins present in all three experimental approaches, UBC encodes one of the polyubiquitin proteins encoded by the human genome. Rather than an interactor, it is likely that ubiquitin was identified by virtue of its covalent modification of a subset of c-Myc proteins. The ability to repeatedly obtain a signal from this small and transient pool of c-Myc molecules underscores the sensitivity of our methodologies. PSMD2 (a 26S proteasome regulatory subunit) was present among the proteins shared between two experimental approaches, and PSMA3 (a proteasomal subunit) and RNF130 (a E3 ubiquitin ligase) were identified in three out of four label-free and SILAC biological replicates, respectively. Koch et al.9
also identified a significant number of components in the proteasomal pathways, suggesting that TAP-tagged Myc proteins, like native c-Myc, are being targeted for degradation.
Of the remaining two novel interactors present in all three experimental approaches, ERBB2IP, also known as Erbin, is a protein containing 17 leucine-rich repeats and one PDZ domain that was identified as an interactor and regulator of the receptor tyrosine kinase ErbB2/Her2.26
Subsequently it was shown to affect the Ras signaling pathway by disrupting Ras-Raf interaction,27
and of relevance to c-Myc, interacting though its PDZ domain with several transcriptional regulators including β-catenin, c-Rel, Smad3 and HTLV-1 Tax.28–30
Little is known about the second novel interactor, PRKCDBP. This protein, also known as SRBC, was identified as a protein kinase C (PKC) δ binding protein and in vitro substrate that contains a leucine zipper-like motif and two PEST sequences.31
The most abundant literature on SRBC implicates it as a tumor suppressor gene frequently methylated in a variety of human cancers.32–36
Recently, SRBC was identified as a caveolin-1 adapter molecule.37
Interestingly, SRBC was represented by multiple peptides in all four SILAC biological replicates, and PTRF, a protein related to SRBC and likewise a caveolin-1 adapter, was present in two of our three experimental approaches as a c-Myc interactor (Sup. Table 3
). It is tempting to speculate that intracellular trafficking mediated by caveolae membranes is involved in regulating c-Myc activity; along similar lines, caveolin-1 was recently implicated in activating p53.38
Of the 29 c-Myc interactors identified in two (or more) of our three experimental approaches (Sup. Table 3
), 8 (28%) are cytoskeletal proteins. Three major categories of filamentous proteins were present: actins, lamins and tubulins. Multiple tubulins are documented as c-Myc interactors in HPRD, and association with lamin A (LMNA) and beta actin (ACTB) has been reported in the literature.9,39
We found TUBA4A in all three of our experimental approaches, and LMNA and ACTC1 in two. LMNA and ACTG1 were present in three SILAC biological replicates, and TUBB6 in three replicates of the label-free experiments. In addition, numerous cytoskeleton associated proteins, such as filamin (FLNA), microtubule associated proteins (CEP55, CEP170, MAP1A) and others were identified. LMNA is known to functionally interact with several transcription factors and chromatin associated proteins,40
and many of the other identified proteins have nuclear as well as cytoplasmic localizations and roles. However, a number of reports over the years have implied cytoplasmic roles for the c-Myc protein, and recently a calpain protease generated fragment of c-Myc, Myc-Nick, has been identified as a microtubule-binding protein involved in promoting tubulin acetylation and cytoplasmic reorganization.41
In light of this discovery, our finding of several endoplasmic reticulum and vessicle-associated proteins, including several RAB GTPases involved in intracellular vessicular transport, is of heightened interest.
Not surprisingly, a major category of c-Myc interactors were transcription factors and chromatin-associated proteins, including ADNP, CDYL, ETV3, FOXP1, NONO, PTRF, RUNX1, TCF12, YBX1 (7 of 29 or 24%, identified in two or more of our experimental approaches, Sup. Table 3
). Of the novel interactors identified here, RUNX1 (also known as Acute Myeloid Leukemia 1 Oncogene, AML1), is important in coordinating the expression of numerous cell cycle regulators. RUNX1 interacts with several transcription cofactors, including the histone acetylase p300 (EP300) and the CREB-binding protein (CREBBP, also known as CBP), both of which have also been reported to interact with c-Myc.42,43
Furthermore, the RUNX1 regulatory subunit CBFB was identified in two of our experimental approaches, making this an interesting topic for further studies.
We also performed a Gene Ontology (GOstat) analysis44
on the entire list of 418 non-redundant proteins identified in this study (Sup. Tables 4–6
). In the Cellular Compartment ontology (Sup. Table 4
) nucleus was most highly enriched, but was followed closely by the cytoplasm, and interestingly, the mitochondrion (the preparation of nuclei we used, namely gentle lysis of cytoplasmic membranes with NP40, only provides an enrichment, and hence some cytoplasmic components were present in the extracts used for purification). Both the microtubule cytoskeleton and actin cytoskeleton were significantly enriched. A variety of cytoplasmic vesicles were also enriched. In the Biological Process ontology (Sup. Table 5
), transcription was the most highly enriched, followed by the cell cycle. Also highly enriched were DNA processes such as replication, repair and chromatin modification, followed by cytoskeletal processes such as cellular trafficking and localization, and cell stress and death responses including apoptosis. In the Molecular Function ontology (Sup. Table 6
), the most highly enriched categories were nucleotide/ATP binding. Also highly enriched were nucleic acid/RNA/DNA binding, transcription factor biding and a variety of transcription associated activities. Cytoskeletal protein binding, helicase activity and a variety of metabolic activities were also significantly enriched. This analysis is highly consistent with c-Myc being predominantly a nuclear protein that interacts with numerous proteins involved with transcription and chromatin organization. The strong enrichment for cell cycle and apoptosis is consistent with those known functions of c-Myc. The interaction with the cytoskeleton is also prominent and recently supported by direct experimental evidence.41
Not documented to date but of note are the enrichments for the mitochondrion and cytoplasmic vessicles, the latter of which may be relevant to the interaction of c-Myc with PRKCDBP (SRBC, above), a component of caveolae. We also performed an enrichment analysis for interacting protein domains,45
in which the PDZ domain was the top hit (Sup. Table 7
). Interaction of c-Myc with PDZ domains has not been previously noted, and may be relevant for its interaction with ERBB2IP (above), a candidate found in all three of our experimental approaches.
Considering our most highly purified c-Myc complexes (two-step purification, ) it is interesting to note that no interacting protein, other than Max, was as abundant as c-Myc (in Max is obscured by the contaminating IgG light chains, but immunoblotting experiments indicate that it co-purifies with c-Myc in roughly stoichiometric amounts). Koch et al.9
observed a similar lack of equivalently staining bands in two separate purifications. Taken together, these data suggest that c-Myc is not found predominantly in a single, multi-component complex, such as, for example, the SAGA complex,46
components of which have been shown to interact with c-Myc.6
The patterns of purification observed by us and Koch et al.9
are more consistent with c-Myc molecules being dispersed among multiple macromolecular complexes. A corollary of this interpretation, also consistent with the very low expression levels of c-Myc, is that the macromolecular complexes that interact with c-Myc contain the protein in very small amounts.
Size exclusion chromatography ( and
) also indicated that c-Myc is found predominantly in large (>1 mDa) macromolecular complexes. The resolution of this method in this size range is not adequate to determine the number of distinct complexes that may be present. A minor peak was seen in the 200–300 kDa range, and essentially no signal was seen in the region of Myc/Max dimers. Max co-migrated with Myc under all conditions. Koch et al.9
used sucrose gradient sedimentation, and reported c-Myc in a large number of protein complexes with a broad range of molecular weights. They also observed c-Myc in a size range consistent with Myc/Max dimers. The latter discrepancy could be due to several factors. First, we prepared nuclear extracts whereas Koch et al. used whole cell extracts. Second, the levels of c-Myc overexpression were likely higher in the cell lines used by Koch et al. Third, the presence in c-Myc complexes of fibrous proteins with extended conformations, such as actin, tubulin or lamin, could anomalously affect our gel filtration elution profiles. Finally, Koch et al. used three cycles of freeze-thawing to promote the lysis of cells, whereas we observed that any freezing-thawing (of either cell pellets or extracts) caused considerable disruption of complexes in our hands. More importantly, however, both studies documented the presence of significant amounts of very large Myc-containing complexes that were relatively stable during the TAP procedure.
Given the involvement of c-Myc in diverse biological processes, and the several already reported interactions with large, multicomponent protein complexes, it is widely believed that the c-Myc interactome is vastly larger than currently documented. Only one prior report has used an unbiased, genome-wide proteomic approach for the identification of c-Myc-interacting proteins.9
Given the importance of c-Myc in human cancer, the paucity of such efforts reflect the biochemical challenges in working with c-Myc: its low abundance in the cell, its highly unstable nature, and its many interacting partners. The main advances of the study reported here are the use of high resolution, high mass accuracy LC/MS methods for the detection of peptides from c-Myc interacting proteins, and the application of SILAC labeling of proteins in vivo. To our knowledge this is the first study to employ SILAC to obtain quantitative data of the c-Myc interactome coupled with a stringent cut-off of 2% false discovery rate (FDR) for the identification of interacting proteins estimated by a decoy database search. The datasets presented here are highly consistent with prior studies, and greatly expand our knowledge of c-Myc interacting protein partners. This new information should significantly advance our understanding of this interesting and important master regulator.