* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Dynamic in vivo interactions among Myc network members
Survey
Document related concepts
Protein phosphorylation wikipedia , lookup
Extracellular matrix wikipedia , lookup
Cellular differentiation wikipedia , lookup
Magnesium transporter wikipedia , lookup
Endomembrane system wikipedia , lookup
Protein moonlighting wikipedia , lookup
Nuclear magnetic resonance spectroscopy of proteins wikipedia , lookup
Signal transduction wikipedia , lookup
Cell nucleus wikipedia , lookup
Western blot wikipedia , lookup
Intrinsically disordered proteins wikipedia , lookup
Protein–protein interaction wikipedia , lookup
Green fluorescent protein wikipedia , lookup
Transcript
ã Oncogene (2001) 20, 4650 ± 4664 2001 Nature Publishing Group All rights reserved 0950 ± 9232/01 $15.00 www.nature.com/onc Dynamic in vivo interactions among Myc network members Xiao-ying Yin1, Melanie F Landay1, Weiping Han2, Edwin S Levitan2, Simon C Watkins3,5, Richard M Levenson6, Daniel L Farkas6,7 and Edward V Prochownik*,1,4,5 1 Section of Hematology/Oncology, Department of Pediatrics, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania, PA 15213, USA; 2Department of Pharmacology, The University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania, PA 15213, USA; 3Department of Cell Biology and Physiology, The University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania, PA 15213, USA; 4Department of Molecular Genetics and Biochemistry, The University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania, PA 15213, USA; 5The University of Pittsburgh Cancer Institute, Pittsburgh, Pennsylvania, PA 15213, USA; 6Center for Light Microscope Imaging and Biotechnology, Carnegie-Mellon University, Pittsburgh, Pennsylvania, PA 15213, USA; 7 Department of Bioengineering, The University of Pittsburgh, Pittsburgh, Pennsylvania, PA 15213, USA Members of the Myc oncoprotein network (c-Myc, Max, and Mad) play important roles in proliferation, dierentiation, and apoptosis. We expressed chimeric green ¯uorescent protein (GFP) fusions of c-Myc, Max, and three Mad proteins in ®broblasts. Individually, c-Myc and Mad proteins localized in subnuclear speckles, whereas Max assumed a homogeneous nuclear pattern. These distributions were co-dominant and dynamic, however, as each protein assumed the pattern of its heterodimeric partner when the latter was co-expressed at a higher level. Deletion mapping of two Mad members, Mad1 and Mxi1, demonstrated that the domains responsible for nuclear localization and speckling are separable. A non-speckling Mxi1 mutant was also less eective as a transcriptional repressor than wild-type Mxi1. c-Myc nuclear speckles were distinct from SC-35 domains involved in mRNA processing. However, in the presence of co-expressed Max, c-Myc, but not Mad, co-localized to a subset of SC-35 loci. These results show that Myc network proteins comprise dynamic subnuclear structures and behave co-dominantly when co-expressed with their normal heterodimerization partners. In addition, c-Myc-Max heterodimers, but not Max-Mad heterodimers, localize to foci actively engaged in pre-mRNA transcription/processing. These ®ndings suggest novel means by which Myc network members promote transcriptional activation or repression. Oncogene (2001) 20, 4650 ± 4664. Keywords: Max; Mad Mxi1; green ¯uorescent protein (GFP); SC-35; transcription *Correspondence: EV Prochownik, Section of Hematology/ Oncology, Department of Pediatrics, Children's Hospital of Pittsburgh, 3705 Fifth Avenue, Pittsburgh, PA 15213, USA; E-mail: [email protected] Received 16 January 2001; revised 3 May 2001; accepted 9 May 2001 Introduction c-Myc, a basic ± helix ± loop ± helix ± leucine ± zipper (bHLH ± ZIP) transcription factor, plays critical roles in proliferation, dierentiation, transformation, and apoptosis (Henriksson and LuÈscher, 1996; Facchini and Penn, 1998; Nesbit et al., 1999). c-Myc is highly expressed during proliferation, in contrast to the low levels associated with cellular quiescence and terminal dierentiation (Henriksson and LuÈscher, 1996; Facchini and Penn, 1998). Following removal of growth factors, c-Myc mRNA and protein rapidly decline (Blanchard et al., 1985; Rabbits et al., 1985). Mitogenic stimulation of quiescent cells is accompanied by a rapid re-induction of c-Myc mRNA and protein with immediate-early response kinetics (Kelly et al., 1983; Campisi et al., 1984). In common with certain other bHLH, bZIP, and bHLH ± ZIP proteins, c-Myc neither homodimerizes nor binds DNA well in vitro (Smith et al., 1990), and cMyc homodimers have not been detected in vivo. Rather, the biologically active form of the protein exists in heterodimeric association with another nuclear bHLH ± ZIP protein, Max (Blackwood and Eisenman, 1991; Prendergast et al., 1991; Amati et al., 1992; Blackwood et al., 1992). Because Max is an extremely stable protein (Blackwood et al., 1992; Wagner et al., 1992; Dunn et al., 1994), the relative abundance of cMyc ± Max heterodimers is determined to a large degree by the absolute amount of c-Myc. In addition to c-Myc's bHLH ± ZIP domain, which promotes dimerization with Max and DNA binding (Blackwell et al., 1990; Blackwood and Eisenman, 1991; Van Antwerp et al., 1992), the protein contains a nuclear localization signal (NLS) and a transcriptional activation domain (TAD) (Henriksson and LuÈscher, 1996; Facchini and Penn, 1998). These regions are required for all known c-Myc functions (Henriksson and LuÈscher, 1996). Whereas c-Myc ± Max heterodimers activate c-myc target genes, transcriptional repression is mediated by heterodimers between Max and at least four other related bHLH ± ZIP family members known as Mad Myc network interactions X-y Yin et al proteins (Ayer et al., 1993; Zervos et al., 1993; Hurlin et al., 1995a). Repression by Mad proteins is an active process, requiring a highly conserved amino-terminal motif (the SID domain) (Ayer et al., 1995; SchreiberAgus et al., 1995). Mad members also show highly variable levels of tissue-speci®c expression and regulation (Larsson et al., 1994; Hurlin et al., 1995a,b; VaÈstrik et al., 1995; Queva et al., 1998). Studies in mad1 and mxi1 nullizygous mice have demonstrated distinct roles for the encoded proteins in tissue-speci®c negative control of cellular proliferation (Foley et al., 1998; Schreiber-Agus et al., 1998). Despite the well-described interactions of Myc network proteins, little is known about their distribution within the nucleus, or how they seek out one another, and dimerize in an appropriate, timely, and controlled manner. Such studies are hindered by the fact that no cells express all the proteins simultaneously, and by the often low abundance of those proteins which are expressed. In the current study, we have used green ¯uorescent protein (GFP) fusions of cMyc, Max, and three Mad proteins (Mad1, Mxi1, and Mad3) to study their subnuclear locations and interactions. Our ®ndings indicate that, depending upon the relative abundance of their heterodimeric partners, Myc network proteins may either occupy distinct compartments (`speckles') or be diusely distributed throughout the nucleus. We show that for Mad proteins, speckling requires a speci®c domain within the polypeptide, distinct from the NLS, and necessary for eective transcriptional repression. Furthermore, c-Myc nuclear speckles overlap a subset of sites engaged in pre-mRNA transcription and processing. Our studies suggest that dynamic subnuclear tracking of Myc network proteins is intimately associated with their function as transcriptional activators or repressors. Results Myc network members show distinct patterns of subnuclear localization To localize each Myc network protein within the nucleus, the complete coding sequences of c-Myc, Max(L), Max(S), Mad1, Mxi1, and Mad3 were expressed as GFP fusions in three dierent ®broblast lines. So as to avoid over-expression artifacts (Hedley et al., 1995), short transfection times (2 ± 4 h) and posttransfection incubation periods (3 ± 6 h) were used. Cells were then ®xed and examined by epi¯uorescence microscopy (Figure 1). Transfection with the pEGFPC1 parental vector alone produced a diuse, primarily cytoplasmic ¯uorescence (not shown), whereas each of the Myc network fusions localized exclusively to the nucleus in all three cell lines. In the case of GFP ± cMyc, the fusion protein assumed a speckled subnuclear appearance that was most readily appreciated in Cos-7 cells (Figure 1). In contrast, both GFP ± Max(L) and GFP ± Max(S) produced an extremely homogeneous nuclear staining. Pronounced speckling pattern were also seen with Mad1, Mxi1, and Mad3. Mad1 and Mad3 fusion proteins tended to form the most pronounced speckles, whereas those formed by GFP ± Mxi1 tended to be somewhat smaller and more numerous. Examination of several hundred cells transfected with each construct con®rmed the nuclear patterns shown in Figure 1. From the above experiments, we conclude that each of the GFP ± Myc network proteins was expressed in distinct subnuclear patterns in three dierent cell types. Experiments in several other ®broblast cell lines have shown similar patterns of expression (not shown). In order to rule out the possibility that the fusion of GFP to the amino terminus of Myc network members was responsible for their distinct subnuclear patterns, each protein was expressed in the pEGFP ± N3 vector as a carboxy terminal fusion with GFP. Expression of these constructs in all three cell lines produced patterns identical to those obtained with the pEGFP ± C1-based amino terminal fusions depicted in Figure 1. In still other experiments, each protein was epitope tagged at the carboxy terminus and detected by indirect immuno¯uorescence. Although staining was less intense than with GFP fusions owing, at least in part, to dierences in promoter strength and the method of detection, the subnuclear patterns observed were identical to those seen with GFP fusions (not shown). Finally, we addressed the possibility that the patterns shown in Figure 1 were the result of ®xation or overexpression artifacts. Several dierent methods of ®xation, performed at various times after transfection, were employed to examine this. In all cases, identical subnuclear localization patterns were seen. In other studies, time-lapse video UV-microscopy was used to follow the synthesis and nuclear targeting of GFPtagged molecules in live cells as soon as the proteins became observable. In each case, we were able to identify newly synthesized GFP fusion proteins in the cytoplasm, and to follow their immediate transport to the nucleus where they rapidly assumed the patterns shown in Figure 1 (not shown). From the above studies, we conclude that the patterns of expressed proteins were not the artifactual result of GFP tagging, of cellular ®xation, or of excessive overexpression, but rather re¯ected intrinsic properties of the Myc network proteins themselves. 4651 Some Myc network proteins co-localize and act co-dominantly The localization of c-Myc and mad family members to the nucleus in discrete, speckled patterns raised the question of whether these proteins occupied the same subnuclear space. To address this, we created new fusions between certain Myc network proteins and mutant forms of GFP that possess dierent excitation/ emission spectra (Tsien, 1998). These mutants have been dubbed `Sapphire' (Sap) and `Topaz' (Top) (Yin et al., 1999). In initial experiments shown in Figure 2, Oncogene Myc network interactions X-y Yin et al 4652 Figure 1 Nuclear patterns of Myc network proteins. The indicated cell lines were transfected with Myc network-GFP expression plasmids and visualized 3 ± 6 h later after ®xing in PBS ± 2% paraformaldehyde. Max(L) and Max(S) indicate the 160 and 151 amino acid isoforms of the Max protein, respectively (Blackwood and Eisenman, 1991; Prendergast et al., 1991). All photos:61000 transfection of Cos-7 cells with only the single Sap ± Mad1 construct produced the expected speckled pattern when examined under optimal Sap conditions (Figure 2a). In contrast, minimal or no ¯uorescence was detected using optimal Top conditions (Figure 2b). Similarly, under Top conditions, we observed bright speckled ¯uorescence following single transfections with Top ± c-Myc or Top ± Mad 3 (Figure 2c,e) whereas minimal or no ¯uorescence was detected using Sap detection conditions (Figure 2d,f). The above vectors were next used to determine whether the subnuclear compartments of co-expressed proteins were identical or spatially distinct. In the ®rst case, Cos-7 cells were co-transfected with expression vectors for Sap ± Mad1 and Top ± Mad3 at a 1 : 1 molar ratio. Examination of single Sap- and Top-speci®c images, as well as the digitally superimposed images, showed that the two proteins co-localized (Figure 2g ± i). Similarly, a transfection in which Sap ± Mad1 and Top ± c-Myc were co-expressed indicated that these Oncogene proteins also localized to identical sub-nuclear compartments (Figure 2j ± l). Co-localization of Top ± cMyc and Sap ± Mxi1 was also seen (not shown). Because c-Myc and Mad proteins neither homodimerize nor heterodimerize with one another, we conclude that their co-localization does not require their direct interaction. Max, diusely expressed in the nucleus (Figure 1), is the only member of the Myc network that heterodimerizes with each of the other members. It thus performs a unique and central role in the control of both transcriptional activation and repression (Henriksson and LuÈscher, 1996). To determine whether the expression of Max could in¯uence the subnuclear location of c-Myc and Mad proteins and vice versa, we constructed a Sap ± Max(L) vector and co-expressed it at dierent ratios together with Top-c-Myc, Top ± Mad1, or Top ± Mad3 in Cos-7 cells. As with single transfection studies using other Sap constructs (Figure 2), Sap ± Max was readily detectable under optimal Sap Myc network interactions X-y Yin et al 4653 Figure 2 Nuclear co-localization of Myc and Mad member proteins. Two spectral variants of GFP, Sapphire (Sap) and Topaz (Top) (Tsien, 1998) were fused to each of the indicated Myc network members. NIH3T3 cells were transfected with the appropriate 1 : 1 molar ratio of the plasmids and then ®xed as described in the legend to Figure 1. Individual GFP-tagged proteins were then visualized under Sapphire or Topaz conditions. (a) Cells expressing only Sap-Mad1 only and visualized under Sapphire conditions; (b) the same ®eld visualized under Topaz conditions; (c) cells expressing Top ± c-Myc only and visualized under Topaz conditions; (d) the same ®eld visualized under Sapphire conditions; (e) cells expressing Top ± Mad3 only and visualized under Topaz conditions; (f) the same ®eld visualized under Sapphire conditions; (g) cells expressing with Sap ± Mad1 and Top ± Mad3 and visualized under Sapphire conditions; (h) the same ®eld visualized under Topaz conditions; (i) digitally superimposed image of g and h; (j) cells expressing Sap ± Mad1 and Top ± c-Myc and visualized under Sapphire conditions; (k) the same ®eld visualized under Topaz conditions; (l) digitally superimposed image of j and k. (61000) conditions whereas it was barely detectable under Topaz conditions (compare Figure 3a and b). We then co-expressed Sap ± Max with either Top ± cMyc, Top ± Mad1, or Top ± Mad3 at 5 : 1 molar ratios. Under these conditions of relative Max excess, its nuclear pattern remained diuse and homogeneous as previously demonstrated in Figure 1 (Figure 3c,e,g). Under Top detection conditions, however, the patterns of c-Myc, Mad1 and Mad3 now appeared homogeneous rather than punctate (Figure 3d,f,h). When the ratios of co-transfected plasmids were reversed, SapMax assumed a speckled pattern (Figure 3i,k,m). As expected, the patterns of c-Myc, Mad1, and Mad3, when viewed under Top conditions, remained speckled Oncogene Myc network interactions X-y Yin et al 4654 Figure 3 Co-dominant behavior of Myc network proteins. NIH3T3 cells were co-transfected with the indicated ratios of plasmids encoding Sapphire- or Topaz-tagged members of the Myc network as described in the legend to Figure 2. Proteins were then visualized under Sapphire or Topaz conditions. (a) Cells transfected with Sap ± Max(L) only and visualized under Sapphire conditions; (b) the same ®eld visualized under Topaz conditions; (c) cells transfected with Sap ± Max(L)+Top-c-Myc (5 : 1 molar ratio) and visualized under Sapphire conditions; (d) the same ®eld visualized under Topaz conditions; (e) cells transfected with Sap ± Max(L) and Top ± Mad1 (5 : 1 ratio) and visualized under Sapphire conditions; (f) the same ®eld visualized under Topaz conditions; (g) cells transfected with Sap ± Max(L) and Top ± Mad3 (5 : 1 ratio) and visualized under Sapphire conditions; (h) the same ®eld visualized under Topaz conditions; (i), cells transfected with Sap ± Max(L)+Top ± c-Myc (1 : 5 ratio) and visualized under Sapphire conditions; (j) the same cells visualized under Topaz conditions; (k) cells transfected with Sap ± Max(L)+Top ± Mad1 (1 : 5 ratio) and visualized under Sapphire conditions; (l) the same ®eld visualized under Topaz conditions; (m) cells transfected with Sap ± Max(L)+Top ± Mad3 (1 : 5 ratio) and visualized under Sapphire conditions; (n) the same ®eld visualized under Topaz conditions. 61000 (Figure 3j,l,n). These experiments establish that Max and its normal heterodimerization partners act in a codominant manner, with each assuming the pattern of Oncogene whichever co-expressed protein is the more abundant. These experiments also indicate that appropriate heterodimeric interactions between the HLH ± ZIP Myc network interactions X-y Yin et al domains of Max and other Myc network members occur despite the presence of GFP tags (Blackwood and Eisenman, 1991; Ayer et al., 1993; Zervos et al., 1993; Hurlin et al., 1995a). Mad nuclear and subnuclear targeting sequences are separable The observation that, in three dierent cell lines, Mad member proteins localized in a speckled pattern following their transport into the nucleus (Figure 1) aorded the opportunity to determine whether the sequences responsible for imparting these attributes were distinct and genetically separable. Accordingly, we constructed a series of deletion mutants of Mad1 and Mxi1 and expressed each in the pEGFP-C1 vector as amino terminal-linked GFP fusions. Most mutants, along with their full-length counterparts, were tested in more than one cell line and gave consistent results. Only the results with NIH3T3 cells are shown here. In the case of Mad1 (Figure 4), two deletions from the carboxy terminus of the protein (Mad11 ± 178 and Mad11 ± 141) retained both the ability to be targeted to the nucleus and to be expressed, like the full-length protein, in a speckled pattern. Further deletion from the carboxy terminus (Mad11 ± 124), caused loss of the speckled pattern, although the ability to localize to the nucleus remained intact. Deletion beyond this point (Mad11 ± 91) also showed loss of the speckled pattern and a partial loss of nuclear retention. Deletions from the amino terminus of Mad1 were also revealing. Thus, elimination of the ®rst 42 amino acids (Mad143 ± 221) did not aect nuclear localization but instead caused a marked loss of speckling. Further deletion from the amino terminus (Mad191 ± 221) caused a substantial, although once again incomplete, loss of nuclear targeting. From these experiments, we conclude that nuclear targeting is most in¯uenced by residues centered around amino acid 91 of Mad1. In contrast, signals for speckling appeared to be bipartite, consisting of residues between amino acids 1 ± 42 and 124 ± 141. Although both of these regions were necessary for the protein to assume a speckled appearance, neither appeared to be sucient. Thus, in the case of Mad1, the nuclear localization and speckling signals map to distinct and separable domains. Similar experiments performed with Mxi1 demonstrated that a construct containing signi®cant deletions of both the amino and carboxy termini (Mxi138 ± 192) was targeted exclusively to the nucleus and assumed the typical speckled pattern of the wild-type protein (Figure 5). In contrast, a construct containing the complete amino terminus but with a longer C-terminal deletion (Mxi11 ± 152), while also targeted to the nucleus, failed to localize in a speckled pattern. The construct Mxi138 ± 175 also failed to form a speckled pattern. This suggested that at least one region of Mxi1 responsible for speckling was located between residues 175 and 192. Consistent with these ®ndings, two additional mutants, Mxi138 ± 147, and Mxi138 ± 131 were also targeted to the nucleus but in a diuse pattern. Additional deletions of the C-terminus provided evidence for the location of the nuclear localization and speckling signals. Although nuclear localization was retained with the Mxi170 ± 228 deletion, a substantial, although incomplete, loss of speckling was observed. Mxi181 ± 228 showed a signi®cant amount of cytoplasmic staining. Although the nuclear pattern was somewhat obscured by overlapping cytoplasmic staining, examination of multiple individual cells indicated that a speckled appearance was not generally maintained, despite the presence of the previously identi®ed speckling sequences between residues 175 and 192. An even more profound eect on nuclear localization was seen with Mxi1131 ± 228. These results suggested that a strong nuclear localization signal in Mxi1 resides between residues 81 and 131 with perhaps an additional weaker component centered around or just upstream of residue 81. Thus, as in the case of Mad1, the speckling signal in Mxi appears bipartite with important residues residing in the regions 38 ± 81 and 175 ± 192. 4655 Speckling signals are necessary for normal Mad protein function To determine whether the ability to localize in a speckled pattern was important for Mad protein function, we created a mammalian expression vector, pSVLneo ± Mxi11 ± 152 that encoded a C-terminally truncated deletion mutant of the Mxi1 protein. Because this mutant does not aect the bHLH ± ZIP or SID domains, it retains the ability to dimerize with Max and recruit the mammalian co-repressor mSin3 and its associated proteins (Ayer et al., 1995; Schreiber-Agus et al., 1995). However, based upon the ®ndings presented in Figure 5, we predicted that this protein would assume a diuse rather than a speckled pattern upon its translocation to the nucleus. We compared the ability of Mxi11 ± 152 and Mxi1(WT) to suppress a cMyc-responsive reporter construct. As seen in Figure 6a, a c-Myc expression vector stimulated the reporter about fourfold in comparison to the empty vector alone. As we and others have previously shown (Hurlin et al., 1995a; Gupta et al., 1998; Yin et al., 1999), coexpression of Mxi1(WT) resulted in a dose-dependent suppression of reporter activity, consistent with the prevailing model that Mad ± Max heterodimers compete with c-Myc ± Max heterodimers for the same binding sites, and mediate transcriptional repression. In contrast, the truncated Mxi11 ± 152 protein showed signi®cantly less ability to repress the reporter. This was not the result of less protein being produced in the case of Mxi11 ± 152 as indicated by Western analysis of the same cells using an anti-Mxi1 antibody (Figure 6B). Finally, as expected, both bacterially expressed full-length Mxi1 and Mxi11 ± 152 dimerized with Max and bound an E-box-containing oligonucleotide (Figure 6c). Taken together, these studies suggest that proper subnuclear localization is necessary for members of the Mad family to ful®l their roles as transcriptional repressors. Oncogene Myc network interactions X-y Yin et al 4656 Figure 4 Deletion mapping of Mad1. PCR products encoding the indicated amino acids were cloned into the pEGFP ± C1 vector to allow expression of the appropriate fusion protein. That each vector was in-frame with the GFP coding sequence and encoded the indicated residues was veri®ed by DNA sequencing. WT indicates the full-length Mad1 protein (Figure 1). The bHLH ± ZIP domain is indicated by the open box. Relevant restriction sites, used in the construction of some fusions, are also noted. Each plasmid DNA was then transiently expressed in Cos-7 cells. (a) Fusion proteins expressed in a nuclear or speckled pattern, respectively, are indicated by (+); those proteins expressed in a partially nuclear or partially speckled pattern are expressed as (+), whereas those expressed in neither a nuclear nor speckled pattern are denoted (7). (b) Representative images of cells expressing the indicated fusion protein.61000 Western blot confirmation of GFP fusion protein expression To demonstrate that all of the foregoing GFP constructs expressed the predicted fusion proteins, Western analyses were performed using lysates from transfected Cos-7 cells. Fusion protein expression was then detected using an anti-GFP antibody. As seen in Oncogene Figure 7, each construct encoded a chimeric GFP protein of the predicted size. Association of c-Myc ± Max complexes with sites of mRNA processing Some pre-mRNA transcription and splicing occurs in distinct subnuclear domains that can be identi®ed with Myc network interactions X-y Yin et al 4657 Figure 5 Deletion mapping of Mxi1. Fusion constructs were created and veri®ed as described in the legend to Figure 4 and transiently expressed in Cos-7 cells. Relevant restriction sites, used in the construction of some fusions, are also noted. (a) (+), (+) and (7) indicate protein patterns as described in the legend to Figure 4. (b) Representative images of cells expressing the indicated fusion proteins.61000 Oncogene Myc network interactions X-y Yin et al 4658 Figure 6 Comparison of Mxi1(WT) and Mxi11 ± 152 function. (a) Transient transfections were performed in NIH3T3 cells as previously described (Zhang et al., 1997; Gupta et al., 1998) using the indicated amounts of each plasmid. Dierences in amounts of input DNAs were adjusted for by the addition of the empty pSVLneo parental vector. Following transfection, cells were grown for an additional 2 days at which time they were harvested and assayed for b-galactosidase and luciferase activities (Zhang et al., 1997; Gupta et al., 1998). The results shown are the average of three separate experiments, each performed in duplicate plates,+1 s.e. (b) In separate transfections, 5 mg each of the parental pSVLneo vector, pSVLMxi1(WT) or pSVLMxi1(1 ± 152) were transiently expressed in NIH3T3 cells. Cells were lysed after 2 days, total lysates were resolved by SDS ± PAGE, and the Western blot was probed with an anti-Mxi1 polyclonal antibody (Gupta et al., 1998) (upper panel) or an anti-actin polyclonal antibody to serve as a control for protein loading (lower panel). (c) Each of the above proteins, along with full-length Max(S) was expressed in E. coli, puri®ed by nickel-agarose anity chromatography and tested for its ability to bind to a double-stranded 32P-labeled oligonucleotide containing a c-Myc binding site (Gupta et al., 1998) the SC-35 spliceosome assembly factor monoclonal antibody (Fu and Maniatis, 1990). These domains are of approximately the same size and abundance as Mmyc and Mad protein speckles (Huang and Spector, 1991; Spector et al., 1991; Hemavathy et al., 2000; Smith et al., 1999). To determine whether SC-35 domains co-localize with c-Myc and Mad speckles, Cos-7 cells were transfected with individual GFP ± cMyc or GFP ± Mad3 expression vectors, ®xed, and immunostained with the SC-35 monoclonal antibody. Nuclei were then visualized by laser confocal microscopy. As seen in Figure 8 (rows 1 and 2) GFP ± cMyc and GFP ± Mad3 localized to the previously described nuclear speckles. SC-35 domains were also detected (row 2) but were spatially separate from those of GFP ± c-Myc and GFP ± Mad3 (merged images, row 3). The examination of several hundred transfected cells in each experiment revealed no instances of co-localization of GFP ± c-Myc or GFP ± Mad3 with SC-35 domains. The absence of even a random pattern of overlap between c-Myc or Mad and SC-35 domains indicated an active process serving to maintain strong segregation between these structures. Because c-Myc and Mad3 must heterodimerize with Max in order to activate or repress transcription, we investigated the possibility that the exclusion of GFP ± c-Myc and GFP ± Mad3 from SC-35 domains re¯ected Oncogene their excess concentration relative to that of endogenous Max. GFP ± c-Myc and GFP ± Mad3 vectors were co-transfected with a non-GFP-tagged Max expression plasmid (Zhang et al., 1997). The latter was added at a molar ratio (generally 2 : 1 ± 3 : 1) that allowed for the incipient transition of GFP ± c-Myc or GFP ± Mad3 speckles into the more homogenous, previously described, `Max-like' pattern (Figures 1 and 3). The cells were then ®xed, stained for SC-35 domains and examined for GFP and SC-35 co-localization. Under these conditions, we observed that a subset of the speckles now co-localized with SC-35 domains (Figure 8, row 3). In most cases, the overlap occurred at the periphery of the SC-35 domains, which are the regions within which actual mRNA transcription has been proposed to occur (Xing et al., 1993, 1995; Smith et al., 1999). Examination of several hundred SC-35 domains from several independent experiments indicated that approximately 15% of them co-localized with GFP-cMyc signals. In contrast to the ®ndings above, co-expression of GFP ± Mad3 and Max failed to demonstrate any overlap between with SC-35 domains in several experiments (Figure 8, row 4). These studies suggested that, unlike the co-localization of GFP ± c-Myc and GFP ± Mad (Figure 2) c-Myc ± Max and Mad-3 ± Max complexes occupy distinct nuclear compartments. These experiments also further indicated that the Myc network interactions X-y Yin et al tion. In the cases of c-Myc and all three members of the Mad family studied, these patterns consist of dots or speckles, whereas the distribution of Max is homogeneous. A second conclusion is that c-Myc and Mad members co-localize. Although these proteins are generally expressed exclusively of one another, clear exceptions exist (Larsson et al., 1994; Zervos et al., 1993; Hurlin et al., 1995b). Since none of these proteins has been shown to homodimerize or to heterodimerize with one another (Henriksson and LuÈscher, 1996), it seems unlikely that a direct interaction is necessary for their proper subnuclear targeting. This co-localization may therefore indicate a common nuclear compartment in which these monomeric proteins, whether co-expressed or not, are sequestered prior to their association with Max. Depending upon the proliferative state of the cell, its tissue of origin, and its degree of dierentiation, this compartment may contain only c-Myc, only one or more Mad proteins, or a combination of c-Myc and Mad proteins. Following their association with Max, these proteins may then be dispersed throughout the nucleus if Max is in excess, or remain within the nuclear speckles if Max is less abundant. Thus, a third conclusion is that Myc network proteins are co-dominant, with the less abundant member of any given pair assuming the pattern of its more abundant natural partner. We also conclude that, at least in the case of Mxil, its ability to localize in a speckled pattern is closely linked to its function as a transcriptional repressor of c-Myc target genes. 4659 Distinct nuclear localization and `speckling' signals Figure 7 Western blot analysis of GFP fusion proteins. Plasmids encoding each of the indicated full-length (a), or various Mad1 or Mxi1 deletions (b and c) was transiently expressed in Cos-7 cells. Two days later, 25 mg of total cell lysate was resolved by standard 10% SDS ± PAGE and Western blotted. GFP proteins were detected with a polyclonal anti-GFP antibody c-Myc ± Max and SC-35 overlap patterns observed above did not simply re¯ect random events. Discussion Subnuclear localization of Myc network members One conclusion of this study is that Myc network proteins possess distinct patterns of subnuclear localiza- The c-Myc NLS has been previously shown to be bipartite (Dang and Lee, 1988). However, these regions, when separated from other regions of the molecule, are unable to impart a speckled appearance to the nuclearly localized protein. These observations are consistent with the conclusions of this study, namely, that the signals responsible for nuclear targeting and subsequent subnuclear localization of Mad family members comprise distinct elements. Based on the deletions depicted in Figure 4, the NLS of Mad1 appears to be centered around amino acid 91, which is within the loop region of the bHLH domain. Since both Mad11 ± 91 and Mad191 ± 221 retained signi®cant nuclear localization capability, it is possible that, like c-Myc, the NLS of Mad1 is bipartite. Indeed, within the region spanning amino acids 43 ± 124 are several potential NLSs. These include the region K45RRNK49, originally proposed as the NLS (Ayer et al., 1993), K49SKK52, E65KNRR69, E77KLKG81, and K100AKLHIKK107. Despite this ambiguity, it is clear from our studies that none of the constructs containing these regions was able to recapitulate the speckled patterns of the wild-type protein unless additional amino terminal and carboxy terminal regions were included. The NLS of Mxi1 also localizes to the loop region and appears bipartite. One NLS is encoded by the 81 ± 131 region whereas a weaker one is encoded by the 70 ± 81 region. Homologies to the putative NLS's of Oncogene Myc network interactions X-y Yin et al 4660 Figure 8 Co-localization of c-Myc ± Max complexes and SC-35 domains. Cos-7 cells were transfected with 1 mg of either the GFP ± -c-Myc expression vector (row 1) or the GFP ± Mad3 expression vector (row 2). A soon as the GFP signal became detectable, the cells were ®xed and stained with the anti-SC-35 monoclonal antibody and examined by confocal microscopy. The merged images show a complete lack of c-Myc or Mad3 and SC-35 co-localization. In rows 3 and 4, the above experiments were repeated except that a non-GFP-tagged Max expression vector (Zhang et al., 1997) was added at a 2 ± 3-fold molar excess. This amount was sucient to initiate the early transition of c-Myc and Mad3 from a speckled to the more diuse pattern assumed by Max. Staining for SC-35 revealed its co-localization with c-Myc ± Max in a subset of SC-35 domains (row 3, arrows). Co-localization was still not observed with SC-35 and Mad3-Max complexes (row 4) Mad1 are seen in both segments. For example, the sequences E76KNRR80 and K111AKAHIKK118 in Mxi1 are identical or nearly identical to the sequences E65KNRR69 and K100AKLHIKK107, respectively, in Mad1. Of note is that these sequences are also well conserved in Mad3 and Mad4. In contrast, the originally proposed Mad1 NLS (K45RRNK49, Ayer et al., 1993) has no counterpart in other Mad proteins (Hurlin et al., 1995a). Additional regions in Mad1 and Mxi1 serve to direct the proteins to speci®c subnuclear speckles. As with the NLS's, these regions appear to be bipartite. In the case Oncogene of Mad1, two regions, between amino acids 1 ± 42 and 124 and 141 are essential for speckling. Similarly, the regions of Mxi1 necessary for speckling also localize to separable regions of the molecule between amino acids 38 ± 81 and 175 ± 192 (Figure 4). A large number of nuclear proteins assume a speckled or dot-like pattern, and often physically interact with and co-localize with other proteins with which they share common functions (Spector, 1993; Leonhardt and Cardoso, 1995; Mortillaro et al., 1996; Kim et al., 1997; Cardoso and Leonhardt, 1998; Scully and Livingston, 2000). In the case of some Myc network interactions X-y Yin et al splicing factors, a common Ser/Arg-rich motif is necessary for subnuclear localization and is distinct from the NLS (Li and Bingham, 1991; Hedley et al., 1995). Similarly, unrelated proteins such as DNA methyltransferase (Leonhardt et al., 1992; AML ± CBFa (Zeng et al., 1997), and DNA ligase (Cardoso et al., 1997) also contain separable NLS's and subnuclear targeting sequences. Our ®ndings with Mad1 and Mxi1 are thus in keeping with this overall theme. Despite the presence of homologous subnuclear targeting signals in splicing factors, most other proteins that co-localize to nuclear speckles share neither these nor other recognizable motifs to explain their common topographical distribution. This suggests that only a small number of the proteins comprising a speckle contain an actual subnuclear targeting sequence. These proteins could serve as nucleation centers for other factors that would then be passively recruited to the speckles. Indeed, such interactions have been observed among splicing factors (Hedley et al., 1995). Such a hierarchy of dierentially interacting Myc network proteins might explain how unrelated factors such as mSin3, histone deacetylase and other proteins that constitute the Mad ± Max transcriptional repressor complex, and that interact through dierent domains, might co-localize without actually possessing distinct speckling signals (Alland et al., 1997; Heinzel et al., 1997). Depending upon the level of Max expression, both cMyc and Mad nuclear speckles appear to be quite ephemeral. Similar behavior is seen with the PML RING-®nger `POD' domains, which are disrupted in acute promyelocytic leukemia but reassembled in the presence of retinoic acid (Dyck et al., 1994; Weiss et al., 1994; Shulman and Evans, 1997). In another example, DNA methyl-transferase localizes to replication foci during S phase but is dispersed during G1 and G2/M (Leonhardt et al., 1992). Finally, the BRCA1 tumor suppressor protein localizes to a small number of nuclear dots during S phase. DNA damaging agents cause loss of BRCA1 from these foci and its reaccumulation at sites of DNA replication (Chen et al., 1996; Scully et al., 1997a,b). Myc network proteins and sites of transcription SC-35 domains are thought to represent sites at which factors necessary for pre-mRNA transcription, splicing, and polyadenylation are allowed to accumulate at locally high concentrations (Spector et al., 1991; Blencowe et al., 1994; Bregman et al., 1995; Smith et al., 1999). The edges of these domains also represent sites of active transcription of some, but not all, genes (Xing et al., 1993, 1995; Moen et al., 1995; Clemson and Lawrence, 1996; Smith et al., 1999). In contrast, the observation that class I and II histone deacetylases, along with the nuclear co-repressors N-CoR and SMRT, localize to nuclear bodies distinct from SC-35 domains (`matrix-associated deacetylase bodies' (Downes et al., 2000), suggests that transcription and repression may be segregated. Our ®nding that cMyc ± Max, but not Mad ± Max heterodimers, colocalize to the periphery of a subset of SC-35 domains, is in keeping with the above observations. This raises the possibility that the transcription of multiple c-Myc-regulated genes may occur in close physical proximity to one another within a subset of SC-35 domains, but that these domains exclude Mad proteins. It is possible to suggest a model that accounts for our observations with myc network proteins. In this model, c-Myc ± Max heterodimers localize to sites of ongoing transcription of c-Myc-target genes. Mad ± Max dimers localize to dierent areas, most likely in association with factors devoted to transcriptional repression. This need not occur concurrently; indeed, c-Myc and Mad proteins tend to be inversely regulated (Hurlin et al., 1995a,b; Henriksson and LuÈscher, 1996). The notion that c-Myc ± Max is concentrated at sites of active transcription is consistent with previous ®ndings showing c-Myc's co-localization with small ribonucleoprotein particles (Spector et al., 1987). Max would be drawn into nuclear speckles and co-localize with c-Myc only when the level of the latter protein exceeded a certain threshold (Ayer and Eisenman, 1993; Baudino and Cleveland, 2001). Mad proteins, generally expressed at much lower levels during such times, would associate with Max in a more diuse and transcriptionally ineectual pattern. The net result would be the formation of the desired, transcriptionally active c-Myc ± Max heterodimers at sites of active c-Myc target gene expression. During times when c-Myc levels are low and Mad levels are high, the opposite pattern would be observed with c-Myc becoming distributed homogeneously throughout the nucleus and Mad proteins co-localizing with Max to form the heterodimeric associations necessary to eect gene repression. Under these circumstances it would be expected that other proteins necessary to eect transcriptional repression would also localize to these same regions (Alland et al., 1997; Heinzel et al., 1997). Proteins unable to concentrate in speckles would be predicted to be less eective as transcriptional regulators, as we have indeed shown (Figure 6). Such regulation might provide a means by which Max, in addition to its previously described functions, could serve to disperse c-Myc and Mad proteins from their sites of transcriptional regulation at times when they were least needed. In summary, our results show that c-Myc network members constitute a group of proteins that participates in the formation of distinct subnuclear structures. The relative concentrations of c-Myc and Mad proteins, relative to that of free Max, is at least one of the factors dictating which types of heterodimers will be localized to these sites. In the most extreme circumstance, the nuclear speckles may represent sites of mutually exclusive gene transcription or repression. These sites are thus both physically and functionally dynamic. 4661 Oncogene Myc network interactions X-y Yin et al 4662 Materials and methods Plasmids and recombinant DNA techniques Two GFP expression vectors were employed to localize individual members of the Myc network. The ®rst, pEGFPC1 (Clontech, Palo Alto, CA, USA), allowed for the creation of a fusion between GFP and the amino terminus of each protein, whereas the second, pEGFP ± N3 fused GFP to the carboxy terminus. The Myc network proteins used consisted of c-Myc (Zhang et al., 1997), Max (both 151 and 160 amino acid isoforms, referred to as Max(S) and Max(L), respectively) (Blackwood and Eisenman, 1991), Mad1, Mxil, and Mad3 (Ayer et al., 1993; Zervos et al., 1993; Hurlin et al., 1995a). The entire coding region of each cDNA was ampli®ed using the polymerase chain reaction (PCR) with speci®c primers designed to allow inframe fusion of the PCR product with the GFP coding sequence. In the case where cDNAs were cloned into the pEGFP ± C1 vector, forward primers included the initiator methionine codons, and reverse primers included the naturally occurring termination codons. For cloning into pEGFP ± N3, the same forward primers as above were used, whereas the reverse primers without termination codons were used to allow the reading frame to extend into the GFP coding region. Each vector was analysed by DNA sequencing to con®rm the identity of the cloned fragment, its orientation, and its reading frame in relation to that of GFP. For deletion analysis of Mad1 and Mxi1, identical procedures were used except that the PCR primers were designed to allow for ampli®cation of those regions of the proteins indicated. Although only results with the pEGFP-C1 constructs are shown here, identical results were obtained with fusions in the pEGFP-N3 vector (not shown). For co-localization studies, we used variant forms of GFP (`Topaz' [ExcitationMax=514 nm; EmissionMax=527 nm] and `Sapphire': [ExcitationMax=395 nm; EmissionMax=508 nm]) (Packard Instruments, Meriden, CT, USA) (Tsien, 1998) in the `C1' vector (Yin et al., 1999). In some experiments, we used epitope tagged Myc network proteins to con®rm the results obtained with GFP vectors. Each of the above Myc network cDNAs was again ampli®ed by PCR using the same reverse primers lacking termination codons described above. After digestion with XhoI and/or SalI, the fragments were cloned into the XhoI site of the PSVLMTneo vector. This vector was constructed from the previously described pSVLneo vector (Zhang et al., 1997) by inserting a coding sequence for the Myc epitope EQKLISQQDL into its unique XhoI cloning site (not shown). Expression of an appropriately sized, epitope-tagged protein was con®rmed by transient transfection in Cos-7 cells followed by Western blotting using the anti-c-Myc epitope 9E10 monoclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA). Immunostaining was then performed as described below. Cell lines, transfections and Western blotting NIH3T3 ®broblasts, Rat- 1a ®broblasts, and Cos-7 cells were grown under standard conditions in Dulbecco'smodi®ed Eagle's minimal essential medium (D-MEM) containing 10% supplemented calf serum (GIBCO ± BRL, Grand Island, NY, USA), 2 mM glutamine; 100 mg/ml streptomycin and 100 units/ml penicillin G. Transfections was performed by either a standard calcium phosphate procedure or with Lipofectamine (GIBCO-BRL). In brief, Oncogene cells were seeded onto glass coverslips in 12-well tissue culture plates 1 ± 2 days prior to transfection and allowed to achieve logarithmic growth. Plasmid DNAs were applied and allowed to remain on the cells for 2 ± 4 h. Coverslips were then ®xed for microscopy at a time previously determined to be the earliest point at which GFP could be reproducibly observed (generally 3 ± 6 h after removal of the transfection mix), thus avoiding overexpression artifacts (Hedley et al., 1995). Coverslips were routinely ®xed in 2% paraformaldehyde ± PBS, although experiments using other protocols or in live cells indicated that none of the procedures altered the distribution of the proteins under study. For immunolocalization using the 9E10 antibody, cells were ®xed in 2% paraformaldehyde ± PBS. Antibody was applied at a 1 : 500 dilution for 1 h followed by application of FITC-conjugated polyclonal rabbit antimouse IgG (1 : 1000, Santa Cruz). The procedure for SC35 immunostaining was essentially identical except that the primary antibody (Sigma) was used at a dilution of 1 : 1000 and the secondary antibody was conjugated with biotin (Sigma). Detection was with avidin-conjugated Texas Red (GIBCO ± BRL). To con®rm the expression of GFP plasmids in vivo, Cos-7 cells were transfected with the indicated vectors using Lipofectamine. The cells were harvested 2 days later and prepared for Western analysis as previously described (Nesbit et al., 1999). Membranes were incubated with a rabbit polyclonal anti-GFP antibody (1 : 1000 dilution; Clontech) followed by a horseradish peroxidase-labeled mouse antirabbit seconday antibody. Blots were developed using an enhanced chemiluminescence system (Renaissance Plus, New England Nuclear, Boston, MA, USA). The method of transfection of the p3xmycE1b-luciferase vector, along with pSVLneo, and pSVLMxil1 ± 152 into NIH3T3 cells has been previously described (Zhang et al., 1997; Gupta et al., 1998). The pSVLMxil1 ± 152 vector was created by amplifying the relevant codons from a full-length Mxil cDNA using standard PCR techniques and cloning into the pSVLneo vector. All plasmid DNAs were puri®ed on Qiagen columns (Qiagen, Chatsworth, CA, USA). The plasmid pCMV ± b-gal (Clontech) was included in all experiments to allow normalization of transfection eciencies. Expression of Mxil (WT) and Mxi1 (1 ± 152) in NIH3T3 cells was detected by Western blotting using a polyclonal anti-Mxi1 antibody (Gupta et al., 1998). Recombinant protein expression and electrophoretic mobility shift assays (EMSAs) Max(S), full-length Mxi1 [mxi1(WT)] and Mxi11 ± 152 were cloned into the QE10 expression vector and expressed in E. coli as His6-tagged recombinant proteins (Gupta et al., 1998; Yin et al., 1999). Proteins were puri®ed by nickel-agarose anity chromatography to 490% purity (not shown). 20 ng of each protein were used for EMSAs along with 10 000 d.p.m. (approx. 20 ng) of the 32P-labeled `EO(GAC)' E-boxcontaining double-stranded oligonucleotide (Prochownik and Van Antwerp, 1993). Microscopy and subcellular localization Epi¯uorescence microscopy was performed with Nikon Diaphot inverted and Eclipse 800 upright microscopes equipped with xenon arc lamps for excitation, a cooled CCD camera and a Sutter Lambda 10-2 ®lter wheel for changing excitation ®lters (400+20 nm for Sapphire, 505+5 nm for Topaz). Light intensity was regulated by use Myc network interactions X-y Yin et al of a neutral density of ®lters. A 535+25 nm emission ®lter was used for imaging GFP-tagged proteins. Laser confocal microscopy was performed with a Leica TCS NT confocal LSM microscope (Rockleigh, NJ, USA) and images were collected using the 406objective. 0.5 mm thick optical sections were typically obtained (usually 8/cell) and a minimum of 20 ± 30 cells were scanned. The images presented are representative of those obtained. Acknowledgments We thank Ciprian Almonte and Sean Albert for confocal microscopy, to Dorothea Becker for comments on the manuscript, and to Ron Berezney and Suri Somanathan for helpful advice and discussions. This work was supported by NIH grants HL33741 and CA 78259 to EV Prochownik and by an NCI Cancer Center Core Grant to The University of Pittsburgh Cancer Institute. 4663 References Alland L, Mahle R, Hou H, Potes J, Chin L, Schreiber-Agus N and DePinho R. (1997). Nature, 387, 49 ± 55. Amati B, Dalton S, Brooks MW, Littlewood TD, Evan GI and Land HT. (1992). Nature, 359, 423 ± 426. Ayer DE and Eisenman RN. (1993). Genes Dev., 7, 2110 ± 2119. Ayer DE, Kretzner L and Eisenman RN. (1993). Cell, 72, 211 ± 222. Ayer DE, Lawrence QA and Eisenman RN. (1995). Cell, 80, 767 ± 776. Baudino TA and Cleveland JL. (2001). Mol. Cell Biol., 21, 691 ± 702. Blackwell TK, Kretzner K, Blackwood EM, Eisenman RN and Weintraub H. (1990). Science, 250, 1149 ± 1151. Blackwood EM and Eisenman RN. (1991). Science, 251, 1211 ± 1217. Blackwood E, LuÈscher B and Eisenman RN. (1992). Genes Dev., 6, 71 ± 80. Blanchard J-M, Piechaczyk M, Dani C, Chambard J-C, Franchi A, Pouyssegur J and Janteur P. (1985). Nature, 317, 443 ± 445. Blencowe BJ, Nickerson JA, Issner R, Penman S and Sharp PA. (1994). J. Cell Biol., 127, 593 ± 607. Bregman DB, Du L, van der Zee S and Warren SL. (1995). J. Cell. Biol., 129, 287 ± 298. Campisi J, Gray HE, Pardee AB, Dean M and Sonenshein GE. (1984). Cell, 36, 241 ± 247. Cardoso MC, Joseph C, Rahn HP, Reusch R, Nadal-Ginard B and Leonhardt H. (1997). J. Cell Biol., 139, 579 ± 587. Cardoso MC and Leonhardt H. (1998). J. Cell. Biochem., 70, 222 ± 230. Chen YM, Farmer AA, Chen CF, Jones DC, Chen PL and Lee WH. (1996). Cancer Res., 56, 3168 ± 3172. Clemson CM and Lawrence JB. (1996). J. Cell. Biochem., 62, 181 ± 190. Dang CV and Lee WHF. (1988). Mol. Cell. Biol., 8, 4048 ± 4054. Downes M, Ordentlich P, Kao H-Y, Alvarez JGA and Evans RM. (2000). Proc. Natl. Acad. Sci. USA, 97, 10330 ± 10335. Dunn BK, Cogliati T, Cultrano CM, Bar-Ner M and Segal S. (1994). Cell Growth Dier., 5, 847 ± 854. Dyck GA, Maul GG, Miller WH, Chen JD, Kakizuka A and Evans RM. (1994). Cell, 76, 333 ± 343. Facchini LM and Penn LZ. (1998). FASEB J., 12, 633 ± 651. Foley KP, McArthur GA, Queva C, Hurlin PJ, Soriano P and Eisenman RN. (1998). EMBO J., 17, 774 ± 785. Fu XD and Maniatis T. (1990). Nature, 343, 437 ± 441. Gupta K, Anand G, Yin XY and Prochownik EV. (1998). Oncogene, 16, 1149 ± 1159. Hedley ML, Amrein H and Maniatis T. (1995). Proc. Natl. Acad. Sci. USA., 92, 11524 ± 11528. Heinzel T, Lavinsky RM, Mullen T-M, SoÈderstroÈm M, Laherty CD, Torchia J, Yang WM, Brard G, Ngo SD, Davie JR, Eisenman RN, Rose DW, Glass CK and Rosenfeld MG. (1997). Nature, 387, 43 ± 48. Hemavathy K, Guru SC, Harris J, Chen JD and Ip IT. (2000). Mol. Cell. Biol., 26, 5087 ± 5095. Henriksson M and LuÈscher B. (1996). Adv. Cancer Res., 68, 109 ± 182. Huang S and Spector DL. (1991). Genes Dev., 5, 2288 ± 1302. Hurlin PJ, Queva C, Koskinen PJ, Steingrimsson E, Ayer DE, Copeland NG, Jenkins NA and Eisenman RN. (1995a). EMBO J., 14, 5646 ± 5659. Hurlin PJ, Foley KP, Ayer DE, Eisenman RN, Hanahan D and Arbeit JM. (1995b). Oncogene, 11, 2487 ± 2501. Kelly K, Cochran BH, Stiles CD and Leder P. (1983). Cell, 35, 603 ± 610. Kim E, Du L, Bregman DB and Warren SL. (1997). J. Cell Biol., 136, 19 ± 28. Larsson L-G, Petterson M, Oberg F, Nilsson K and Luscher B. (1994). Oncogene, 9, 1247 ± 1252. Leonhardt H, Page AW, Weier HU and Bestor TH. (1992). Cell, 71, 865 ± 873. Leonhardt H and Cardoso MC. (1995). Int. Rev. Cytol., 162B, 303 ± 335. Li H and Bingham PM. (1991). Cell, 67, 335 ± 342. Moen PT, Smith KP and Lawrence JB. (1995). Hum. Mol. Genet., 4, 1779 ± 1789. Mortillaro MJ, Blencowe BJ, Wei X, Nakayasu H, Du L, Warren SL, Sharp PA and Berezney R. (1996). Proc. Natl. Acad. Sci. USA, 93, 8253 ± 8257. Nesbit CE, Tersak JM and Prochownik EV. (1999). Oncogene, 18, 3004 ± 3016. Prendergast GC, Lawe D and Zi EB. (1991). Cell, 65, 395 ± 407. Prochownik EV and Van Antwerp ME. (1993). Proc. Natl. Acad. Sci. USA, 90, 960 ± 964. Queva C, Hurlin PJ, Foley KP and Eisenman RN. (1998). Oncogene, 16, 967 ± 977. Rabbits PH, Watson JV, Lamond A, Forster A, Stinson M, Evan G, Fischer W, Atherton E, Sheppard R and Rabbits TH. (1985). Nature, 314, 363 ± 366. Schreiber-Agus N, Chin L, Chen K, Torres R, Rao G, Guida P, Skoultchi AI and DePinho RA. (1995). Cell, 80, 777 ± 786. Schreiber-Agus N, Meng Y, Hoang T, Hou H, Chen K, Greenberg R, Cordon-Cardo C, Lee H-W and DePinho RA. (1998). Nature, 393, 483 ± 487. Scully R, Chen J, Plug A, Xiao Y, Weaver D, Feunteun J, Ashley and Livingston DM. (1997a). Cell, 88, 265 ± 275. Scully R, Chen J, Ochs RL, Keegan K, Hoekstra M, Feunteun J and Livingston DM. (1997b). Cell, 90, 425 ± 435. Oncogene Myc network interactions X-y Yin et al 4664 Oncogene Scully R and Livingston DM. (2000). Nature, 408, 429 ± 432. Shulman IG and Evans RM. (1997). Leukemia (Suppl.) 3, 376 ± 377. Smith MJ, Charron-Prochownik DC and Prochownik EV. (1990). Mol. Cell Biol., 10, 5333 ± 5339. Smith KP, Moen PT, Wydner KL, Coleman JR and Lawrence JB. (1999). J. Cell Biol., 144, 617 ± 629. Spector DL. (1993). Ann. Rev. Cell Biol., 9, 265 ± 315. Spector DL, Fu XD and Maniatis T. (1991). EMBO J., 10, 3467 ± 3481. Spector DL, Watt RA and Sullivan NF. (1987). Oncogene, 1, 5 ± 12. Tsien RY. (1998). Ann. Rev. Biochem., 67, 509 ± 544. Van Antwerp ME, Chen DG, Chang C and Prochownik EV. (1992). Proc. Natl. Acad. Sci. USA, 89, 9010 ± 9014. VaÈstrik I, Kaipainen A, Penttila T-L, Lymboussakis A, Alitalo R, Parvinen M and Alitalo K. (1995). J. Cell Biol., 128, 1197 ± 1208. Wagner AJ, LeBeau MM, Diaz MO and Hay N. (1992). Proc. Natl. Acad. Sci. USA, 89, 3111 ± 3115. Weiss K, Rambaud S, Lavau C, Jansen J, Carvalho T, Carmo-Fonseca M, Lamond A and Dejean A. (1994). Cell, 76, 345 ± 356. Xing Y, Johnson CV, Dobner PR and Lawrence JB. (1993). Science, 259, 1326 ± 1330. Xing Y, Johnson CV, Moen PT, McNeil JA and Lawrence JB. (1995). J. Cell. Biol., 131, 1635 ± 1647. Yin X-Y, Gupta K, Han WP, Levitan ES and Prochownik EV. (1999). Oncogene, 18, 6621 ± 6634. Zeng C, Van Wijnen AJ, Stein JL, Meyers S, Sun W, Shopland L, Lawrence JB, Penman S, Lian JB, Stein GS and Hiebert SW. (1997). Proc. Natl. Acad. Sci. USA, 94, 6746 ± 6751. Zervos A, Gyuris J and Brent J. (1993). Cell, 72, 223 ± 232. Zhang H, Fan F and Prochownik EV. (1997). J. Biol. Chem., 272, 17416 ± 17424.