Oncogene (1999) 18, 6621 ± 6634 ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $15.00 http://www.stockton-press.co.uk/onc Mmip-2, a novel RING ®nger that interacts with mad members of the oncoprotein network

Xiao-Ying Yin1, Kalpana Gupta1, Wei Ping Han3, Edwin S Levitan3 and Edward V Prochownik*,1,2,4

1Section of Hematology/Oncology, Children's Hospital of Pittsburgh, The University of Pittsburgh Cancer Institute, Pittsburgh Pennsylvania, PA 15213, USA; 2The Department of Molecular Genetics & Biochemistry, The University of Pittsburgh Cancer Institute, Pittsburgh Pennsylvania, PA 15213, USA; 3The Department of Pharmacology, The University of Pittsburgh Medical Center, The University of Pittsburgh Cancer Institute, Pittsburgh Pennsylvania, PA 15213, USA; 4The University of Pittsburgh Cancer Institute, Pittsburgh Pennsylvania, PA 15213, USA

Mad are basic ± helix ± loop ± helix ± leucine TGTG found in many myc-responsive (Black- zipper (bHLH-ZIP)-containing members of the myc wood et al., 1992; Blackwell et al., 1993; Ma et al., 1993; oncoprotein network. They interact with the bHLH- Prochownik and VanAntwerp, 1993; Henriksson and ZIP protein , compete for the same DNA binding Luscher, 1996; Facchini and Penn, 1998; Dang, 1999). sites as myc-max heterodimers and down-regulate myc- Target activation is then mediated by a transcrip- responsive genes. Using the bHLH-ZIP domain of mad1 tional activation domain (TAD) present in the amino- as a yeast two-hybrid `bait', we identi®ed Mmip-2, a terminus of each myc member (Kato et al., 1990; novel RING ®nger protein that interacts with all mad Barrett et al., 1992). c-myc is expressed primarily in members, but weakly or not at all with c-myc, max or proliferating cells, is extremely labile, and is strongly unrelated bHLH or bZIP proteins. The mad1-Mmip-2 down-regulated in response to growth arrest or terminal interaction is mediated by the ZIP domain in the former di€erentiation (Lachman and Skoultchi, 1984; St. protein and by at least two regions in the latter which do Arnaud et al., 1988; DePinho et al., 1991; Evan and not include the RING ®nger. Mmip-2 can disrupt max- Littlewood, 1993; Henriksson and Luscher, 1996; mad DNA binding and can reverse the suppressive e€ects Zhang et al., 1997; Dang, 1999). The expression of N- of mad proteins on c-myc-responsive target genes and on myc and L-myc is highly tissue-restricted, particularly in c-myc+ras-mediated focus formation in ®broblasts. the adult animal (Zimmerman et al., 1986). Therefore, a Tagging with spectral variants of green ¯uorescent signi®cant amount of the positive transcriptional protein showed that Mmip-2 and mad proteins reside in regulation of myc-responsive genes that occurs in vivo separate cytoplasmic and nuclear compartments, respec- is the result of ¯uctuations in myc protein levels or tively. When co-expressed, however, the proteins interact di€erential tissue expression of myc family proteins. and translocate to the cellular compartment occupied by Negative regulation of myc-responsive genes is the more abundant protein. These observations suggest a mediated by a di€erent arm of the myc network that novel way by which Mmip-2 can modulate the involves heterodimerization between max and at least transcriptional activity of myc oncoproteins. four related bHLH-ZIP proteins referred to as the mad family (Ayer et al., 1993; Zervos et al., 1993, 1994; Keywords: RING ®nger proteins; zinc ®nger proteins; Hurlin et al., 1995a; Schreiber-Agus and DePinho, helix ± loop ± helix ± proteins; mad; mxi1; 1998). Each mad member (mad1, mxi1, mad3, and max; myc mad4) shows distinct patterns of tissue- and differ- entiation-speci®c expression (Larsson et al., 1994; Hurlin et al., 1995b; Queva et al., 1998). Mad protein expression tends to correlate inversely with that of c- Introduction myc, being induced during terminal di€erentiation and repressed during active proliferation and in undiffer- Transcriptional regulation by members of the myc entiated cells and tissues (Larsson et al., 1994; Hurlin oncoprotein network is mediated, both directly and et al., 1995a,b; Vastrik et al., 1995; Queva et al., 1998). indirectly, by a variety of protein ± protein interactions It is believed that mad-max heterodimers compete with (Henriksson and Luscher, 1996; Facchini and Penn, myc-max heterodimers for the same DNA binding sites 1998; Schreiber-Agus and DePinho, 1998). The central and repress transcription through an active process participant is max (Blackwood and Eisenman, 1991; involving the recruitment by mad of a non-bHLH-ZIP Prendergast et al., 1991), an extremely stable and co-repressor, mSin3, as well as other proteins, ubiquitously expressed bHLH-ZIP protein that can including a histone deacetylase (Lahoz et al., 1994; associate with all three major myc proteins (c-myc, N- Schreiber-Agus et al., 1995; Ayer et al., 1995; Hassig et myc, and L-myc) to form heterodimers with high al., 1997; Laherty et al., 1997; Schreiber-Agus and anity for the consensus `myc box sequence' CAC/ DePinho, 1998). The interaction with mSin3 requires a highly conserved region, termed the SID domain, found at the extreme amino termini of all mad members (Ayer et al., 1995; Schreiber-Agus et al., *Correspondence: EV Prochownik, Section of Hematology/ Oncology, Children's Hospital of Pittsburgh, 3705 Fifth Ave., 1995; Schreiber-Agus and DePinho, 1998). Pittsburgh, PA 15213, USA The need to exert rigid control over myc oncopro- Received 12 May 1999; revised 15 July 1999; accepted 15 July 1999 teins is illustrated by the biological consequences of Mmip-2, a novel RING finger protein X-Y Yin et al 6622 their aberrant expression, which may be a primary library. A total of 46106 colonies were plated onto defect in many human cancers (Nesbit et al., 1999). SC plates lacking his, leu, and trp and containing More recently, it has been shown that mad1 or mxi1 30 mM 3-aminotriazole. Approximately 50 histidine nullizygous mice show developmental abnormalities, prototrophs were identi®ed and, of these, three tested are predisposed to spontaneous or experimentally very strongly in an in situ b-galactosidase assay. We induced neoplasms, and manifest growth and restora- designated the cDNA as `mad member interacting tive abnormalities of primary cells in vitro (Foley et al., protein-2' (Mmip-2) (see Gupta et al., 1998 for the 1998; Schreiber-Agus et al., 1998). cloning and characterization of Mmip-1). In addition to the above interactions, which all We initially de®ned the interactions between Mmip-2 occur via bHLH-ZIP domains, a number of associa- and other members of the myc network, as well as tions between myc and non-bHLH-ZIP proteins have several other unrelated proteins (Table 1). Yeast been characterized (reviewed in Facchini and Penn, expressing the original pGAD-Mmip-2 isolate were 1998). Bin1, p107, and TRRAP are three unrelated transformed with pGBT9 vectors encoding c-myc, max, proteins which interact with the TAD of c-myc and mad1, mad3, mad4, as well as with the empty pGBT9 modulate its activity (Beijersbergen et al., 1994; Gu et vector alone. As seen in Table 1 (lines 1, 2, 6, 9, 12, 15, al., 1994; Sakamuro et al., 1996; McMahon et al., 18, 22 and 26), neither the pGAD-Mmip-2 `prey' nor 1998). Several non bHLH-ZIP proteins which interact any of the `baits' was able to self-transactivate the b- with the bHLH-ZIP region of c-myc have also been galactosidase reporter gene. A strong interaction was described and include AP-2, and two zinc-®nger proteins, Miz1 and YY-1 (Shrivastava et al., 1993, 1996; Gaubatz et al., 1995; Peukert et al., 1997). At Table 1 Interaction between Mmip-2 and myc network members. least two non bHLH-ZIP proteins have been Yeast transformants expressing each of the indicated pGBT9 `baits' characterized which interact with one or more mad were selected as individual Trp prototrophs. Single colonies were family members. mSin3 has been shown to be grown, retransformed with the indicated pGAD vectors, and selected important for mad proteins to repress myc-responsive for Trp+Leu prototrophy. Five to ten individual double transfor- mants were streaked onto EGG plates and allowed to grow for an genes (Ayer et al., 1995; Schreiber-Agus et al., 1995; additional 2 ± 3 days prior to performing in situ b-galactosidase assays Schreiber-Agus and DePinho, 1998). More recently, we as previously described (Langlands et al., 1997; Gupta et al., 1998). have described a novel ZIP only-containing protein, Results are expressed as an `average' rate of color change for the Mmip-1 (`mad member-interacting protein-1'), that colonies studied interacts with the ZIP domains of all four mad In situ members (Gupta et al., 1998). Overexpression of pGBT9-expressed pGAD-expressed b-galactosidase Mmip1 reverses the suppressive e€ects of mad proteins `bait' protein `prey' protein activity on promoters containing both arti®cial and naturally- 1. None (pGBT9 only) Mmip-2 7 occurring myc binding sites. Thus, normal function 2. mad1 None (pGAD only) 7 and regulation of the myc network, in addition to 3. None (pGBT9 only) max 7 4. mad1 max ++++ requiring `classic' bHLH-ZIP associations, is likely to 5. mad1 Mmip-2 ++++ involve non-bHLH-ZIP interactions as well (Facchini 6. mxi1 None (pGAD only) 7 and Penn, 1998). 7. mxi1 max ++++ Because of the importance of mad proteins in the 8. mxi1 Mmip-2 ++ 9. mad3 None (pGAD only) 7 control of myc oncoprotein function, we have 10. mad3 max +++ continued to search for new mad-interacting partners. 11. mad3 Mmip-2 ++++ We report here the cloning and characterization of a 12. mad4 None (pGAD only) 7 novel RING ®nger-containing protein, Mmip-2. 13. mad4 max +++ Together with our previous characterization of 14. mad4 Mmip-2 ++++ 15. c-myc None (pGAD only) 7 Mmip-1 (Gupta et al., 1998), our ®ndings suggest the 16. c-myc max +++ existence of multiple proteins, of diverse structure, 17. c-myc Mmip-2 +/7 whose function it is to modify max-mad interactions 18. max None (pGAD only) 7 and thus indirectly to in¯uence myc oncoprotein 19. None (pGBT9 only) mad-1 7 20. max mad-1 ++++ transcriptional activity. Our ®nding that Mmip-2 can 21. max Mmip-2 7 alter the subcellular location of mad proteins suggests a 22. Id1 None (pGAD only) 7 novel mechanism by which myc transcriptional activity 23. None (pGBT9 only) MyoD 7 can be regulated. 24. Id1 MyoD +++ 25. Id1 Mmip-2 7 26. MyoD None (pGAD only) 7 27. None (pGBT9 only) Id1 7 Results 28. MyoD Id1 +++ 29. MyoD Mmip-2 7 30. mad1 delH1 max 7 Cloning and characterization of Mmip-2 31. mad1 delH1 Mmip-2 ++++ 32. mad1 delZIP max 7 To identify proteins which interacted with mad1, we 33. mad1 delZIP Mmip-2 7 expressed its bHLH-ZIP domain in yeast as a Gal4 ++++Intense blue color observed within 15 ± 30 min; DNA binding domain fusion in the pGBT9 vector. The +++intense blue color observed within 30 min ± 1 h; ++dark protein did not self-transactivate but interacted blue color observed within 1 ± 2 h; +light blue color seen within 2 ± strongly with a Gal4-max fusion protein (Table 1, 8h;+/7occasional faint blue color seen with overnight incubation; lines 2 and 4). 7no evidence for blue color with overnight incubation. Each experiment was performed a minimum of three times with similar Yeast harboring the pGBT-9-mad1 vector were results. Simultaneous selection for double transformants did not transformed with a d.11 mouse embryo cDNA in¯uence the observed outcomes Mmip-2, a novel RING finger protein X-Y Yin et al 6623 seen between Mmip-2 and mad1 (Table 1, line 5), thus nated portions of either the HLH or ZIP domains so as providing con®rmation of our original observation. to render them incapable of interacting with max Equally strong interactions between Mmip-2 and mad3 (Table 1, lines 30 and 32). pGBT9-mad1delH1 which and mad4 were also seen (Table 1, lines 11 and 14). Of encodes a fusion protein lacking almost the entire ®rst interest was that the interaction between Mmip-2 and helix of the mad1 HLH domain was capable of mxi1, was reproducibly weaker than that seen with any interacting strongly with Mmip-2 despite its inability other mad family member (Table 1, line 8). A very to interact with max (Table 1, compare lines 30 and weak interaction between Mmip-2 and c-myc was seen 31). In contrast, pGBT9-mad1delZIP which lacks most in some experiments (Table 1, line 17), whereas no of the ZIP domain failed to interact with Mmip-2 interaction seen with max, Id1 or MyoD, two unrelated (Table 1, lines 32 and 33). From these results, we bHLH proteins (Table 1, lines 22, 25 and 29, conclude that the interaction between mad1 and respectively). In other experiments (not shown), we Mmip-2 requires only an intact ZIP domain of the have also shown that Mmip-2 failed to interact with former protein. two other members of the Id family, Id2 and Id3, two other myogenic bHLH proteins, and , In vitro interactions mirror those seen in yeast and the bZIP proteins c-fos and fra-1. That the observed interactions or lack thereof with Mmip-2 To con®rm the in vivo interactions seen in yeast, the did not simply re¯ect large di€erences in the amounts original Mmip-2 cDNA was excised from its pGAD10 of proteins made or their improper folding was ruled vector and expressed as a GST fusion protein in the out by control experiments showing that all pGBT9 pGEX-4T vector. GST-Mmip-2, or GST alone, was and pGAD fusion proteins interacted strongly with puri®ed by standard glutathione-agarose anity previously identi®ed partners (Table 1). chromatography and tested in `pulldown' experiments Because the original `bait' vector contained only the with 35S-methionine labeled in vitro translated members bHLH-ZIP domain of mad1, it was evident that this of the myc network. As seen in Figure 1, GST-Mmip-2 region was both necessary and sucient for the Mmip- interacted quantitatively with 35S-mad1, mad3 and 2 interaction. However, we have previously shown that mad4. A variable but consistently low level of only the ZIP domain of mad proteins is required for interaction was seen with mxi1 whereas no detectable their interaction with Mmip-1, a ZIP-containing interaction was seen with either max or c-myc. We protein (Gupta et al., 1998). Therefore, we created conclude that, in general, the in vitro interactions two additional pGBT9-mad1 constructs which elimi- mirrored those seen in vivo (Table 1).

Figure 1 `GST pulldown' experiments. The original Mmip-2 cDNA isolate (encoding amino acids 2 ± 301, see Figure 2) was expressed as a GST fusion protein and puri®ed by glutathione-agarose anity chromatography. GST-Mmip-2 or GST alone were then incubated with the indicated 35S-labeled full-length myc network proteins as previously described (Gupta et al., 1998), precipitated by the addition of glutathione-agarose, washed extensively, and resolved by SDS ± PAGE. In the rightmost lane of each panel, an aliquot of 35S-labeled protein, equivalent to that used for each GST pulldown, was electrophoresed in parallel to serve as a denominator for the eciency of GST or GST-Mmip-2 capture Mmip-2, a novel RING finger protein X-Y Yin et al 6624 agreement with the calculated 35.7 kDa size of the full- Complete coding sequence of Mmip-2 length conceptual translation product (not shown). Sequencing of the original Mmip-2 cDNA in the Although searches of GenBank indicated that pGAD10 vector revealed an 898 nt insert, in-frame Mmip-2 encoded a unique protein, several signi®cant over its entire length with the Gal4 TAD. This features were identi®ed. Most notably, we observed suggested that our original cDNA clone did not that amino acids 33 ± 75 (Figure 2a underlined) encode a full-length protein. We therefore screened a domain that is in perfect agreement with the approximately 106 cDNA clones from a Friend murine previously determined consensus sequence for a RING

erythroleukemia cDNA library. Several positive clones ®nger: C-X2-C-X(9 ± 39)-C-X(1 ± 3)-H-X(2 ± 3)-C-X2-C-X(4 ± 48)-

were identi®ed and sequenced. The complete coding C-X2-C (Fremont, 1993; Saurin et al., 1996). The N- sequence of Mmip-2, compiled from overlapping terminal region of this element contains a calcium- cDNAs, is shown in Figure 2. Of note is that the calmodulin-dependent kinase II phosphorylation site longest composite cDNA is 1098 nt in length and (RKVSV: amino acids 38 ± 42) whereas the C-terminal encodes an open reading frame for a 316 amino acid half (amino acids 52 ± 76) bears homology (50% protein. Upstream of the putative initiation codon at identity; 67% similarity) to a previously cloned RING position 127, the 5'-untranslated region contains ®nger-containing protein that interacts with the numerous translation termination codons in all three cytoplasmic domain of CD40/FAS (Hu et al., 1994). reading frames. The termination codon at nt 1076 is Amino acids 204 ± 303 encode a region with distant followed by a short, 21 nt 3'-untranslated region with homology to murine Type II cytoskeletal keratin (26% no evidence of a polyadenylation signal. The original identity and 71% similarity) (Figure 2b). A cluster of Mmip-2 cDNA isolate begins at nt 132, ends at nt 1030 consensus N-glycosylation sites (NXS/T) is found at and thus encodes all but 16 amino acids of the full- amino acids 110, 135, and 138. Finally, since Mmip-2 length protein. SDS ± PAGE of the products of a contains no additional zinc ®nger domains, it does not coupled in vitro transcription/translation reaction appear to be a member of the `RBCC' or `TRAF'

showed a single band of Mr=37 kDa, in excellent subgroup of RING ®nger proteins (Saurin et al., 1996).

Tissue expression of Mmip-2 To determine the size of the endogenous Mmip-2 transcript, we performed a Northern blot on total RNA puri®ed from NIH3T3 ®broblasts and Friend murine erythroleukemia cells. This showed the presence of a single, ca. 4.5 kb transcript that was expressed at a higher level in the latter cell type (Figure 3a). To determine the tissue distribution of Mmip2, we hybridized the same cDNA probe to a `Mouse RNA Master Blot' (Clontech) that contains poly(A)+ RNA derived from 18 adult tissues and total embryos of four di€erent gestational ages. As seen in Figure 3c, Mmip- 2 was expressed at the highest level in adult testis and at signi®cantly lower but detectable levels in the adult thyroid, submaxillary gland, ovary, and epididymis. The lack of detectable transcript in any of the whole embryo RNAs was consistent with its restricted expression in adult tissues.

The Mmip-2 RING ®nger is not necessary for interaction with mad1 Having established that the ZIP domain of mad1 was essential for its interaction with Mmip-2, we performed the converse experiment. A series of amino terminal and carboxy terminal deletions of Mmip-2 were generated, expressed as fusion proteins in pGAD10, and tested for interaction with the mad1 fusion protein originally used to conduct the yeast two hybrid screen. As described previously, the fusion protein encoded by Figure 2 (a) cDNA and protein sequence for Mmip-2. The the original Mmip-2 cDNA isolate (amino acids 2 ± sequence shown was derived from a composite of several 301) interacted strongly with mad1 (Figure 5, line 2). overlapping clones isolated from the original murine embryo cDNA library in the pGAD10 vector and from a Friend murine An amino-terminal deletion which removed an erythroleukemia cell cDNA library in the lZAP vector. The additional 42 residues, including a large segment of proposed translation initiation coding at nt 127 is preceded by the RING ®nger, interacted with the mad1 bait as several termination codons in all three reading frames. The strongly as the original isolate (Figure 5, line 3). The consensus RING ®nger domain is underlined. The original cDNA removal of an additional 88 amino acids, including all isolate encoded amino acids 2 ± 301. (b) Homology to murine Type II cytokeratin. Alignment of amino acids 204 ± 303 of of the RING ®nger resulted in a considerable, although Mmip-2 with amino acids 232 ± 327 of murine cytokeratin incomplete, loss of interaction (Figure 5, line 4), Mmip-2, a novel RING finger protein X-Y Yin et al 6625 whereas a segment containing only the C-terminal 52 amino acids 44 ± 132 is primarily responsible for the amino acids was entirely inactive (Figure 5, line 5). association. From these deletions, we conclude that the RING Deletions of the C-terminus further revealed a ®nger domain is largely, if not entirely, dispensable for second region in Mmip-2 required for its interaction the mad1 interaction. Rather, the region between with mad1. As seen in Figure 5 (line 6), the removal of 52 amino acids from the C-terminus resulted in a total loss of interaction with mad1. Further deletions (Figure 5, lines 7 ± 9) also failed to interact. From these and the foregoing observations, we can conclude that Mmip-2 requires at least two distinct regions for its interaction with mad1, and presumably with all other mad members. One region localizes just down- stream of the RING ®nger in the N-terminal portion of the molecule whereas another localizes to the extreme C-terminus.

In vitro functional consequences of Mmip-2 and mad protein interactions The observation that, both in yeast and in `GST- pulldown'-type experiments, Mmip-2 bound each of the four known mad proteins but not c-myc, max, or several unrelated bHLH or bZIP proteins, not only Figure 3 Tissue expression of Mmip2. (a) A Northern blot indicated that these interactions were speci®c, but that containing total RNA from NIH3T3 ®broblasts (lane 1) or Friend murine erythroleukemia cells (lane 2) was hybridized with a they may also have functional signi®cance. To cDNA probe encoding the entire coding region of Mmip-2. A determine whether this was indeed the case, we single ca. 4.5 kb transcript was detected in both tissues. (b). The expressed mad1, mxi1, mad4, and max as hexahisti- same blot, stripped and re-probed with a b-actin cDNA as a dine fusion proteins in E. coli and puri®ed each to control for RNA loading. (c) A `Mouse RNA Master Blot' near homogeneity (Prochownik and VanAntwerp, (Clontech) containing dot-blotted poly(A)+ RNAs from a variety of adult tissues and total embryos of four di€erent gestational 1993; Zhang et al., 1997; Yin et al., 1998). Using a ages was hybridized with the Mmip-2 cDNA probe from (a). 32P-end-labeled oligonucleotide containing a consensus Tissues examined included brain (a1), eye (a2), liver (a3), lung c-myc binding site in an electrophoretic mobility shift (a4), kidney (a5), heart (b1), skeletal muscle (b2), smooth muscle assay (EMSA), we then asked whether the addition of (b3), pancreas (c1), thyroid (c2), thymus (c3), submaxillary gland (c4), spleen (c5), testis (d1), ovary (d2), prostate (d3), epididymis GST-Mmip-2 could inhibit DNA binding by mad- (d4), uterus ( d5), day 7 embryo (e1), day11 embryo (e2), day 15 max heterodimers or max-max homodimers. To embryo (e3), day 17 embryo (e4), yeast total RNA (f1), yeast eliminate any ambiguity due to similarities in the tRNA (f2), E. coli rRNA (f3), E. coli DNA (f4), poly r(A) (g1), appearance of shifted complexes that contained both Cot1 DNA (g2), mouse DNA (g3 and g4). Mmip-2 expression max and mad member proteins, we used only the was detected at highest level in testis (d1), and at lower levels in thyroid (c2), submaxillary gland (c4), ovary (d2), and epididymis shorter, 151 amino acid isoform of max [max(S)] (d4) which, as we have previously shown, binds DNA very

Figure 4 Deletion analysis of Mmip-2. The indicated deletions were generated by standard PCR-based techniques using primers with engineered 5'-EcoRI and 3'-BamHI sites to allow for directional cloning into the pGAD10 vector. Each deletion mutant was then introduced into yeast along with either pGBT9 alone or the pGBT9-mad1 construct used as the original two-hybrid bait. 5 ± 10 leucine+tryptophan prototrophs were randomly selected, grown for 2 ± 3 days on EGG plates and tested for b-galactosidase production using an in situ ®lter assay as previously described (Anand et al., 1997; Gupta et al., 1998). The scoring system for b- galactosidase expression was as described in the legend to Table 1 Mmip-2, a novel RING finger protein X-Y Yin et al 6626

Figure 5 Electrophoretic mobility shift assays. All members of the myc network were expressed as hexahistidine-tagged proteins in E. coli and puri®ed to near homogeneity by nickel-agarose anity chromatography (Prochownik and VanAntwerp, 1993). Approximately 20 ng was used in each EMSA reaction. GST and GST-Mmip-2 were puri®ed as described in the legend to Figure 1 and added at the indicated amounts. After a 20 min incubation at 408C to allow for dimer formation, approximately 0.1 ng of the 32P-labeled EO(GAC) c-myc binding site oligonucleotide plus 1 mg of poly(dI) : (dC) were added for an additional 20 min incubation followed by 4% non-denaturing polyacrylamide gel electrophoresis in 16TBE bu€er as previously described (Prochownik and VanAntwerp, 1993; Zhang et al., 1997; Gupta et al., 1998)

poorly as a homodimer (Prochownik and VanAnt- loss of DNA binding was ®rst detected with 25 ± werp, 1993; Zhang et al., 1997). To test the e€ect of 50 ng of GST-Mmip-2 and was almost completely Mmip-2 on DNA binding by max, we used the 160 abolished with 100 ng (Figure 5a,c). In contrast, amino acid isoform [max(L)] of the latter protein nearly twice as much GST-Mmip2 was required to which binds DNA4100-fold more strongly than reverse the DNA binding of mxi1-max heterodimers max(S) (Zhang et al., 1997). As seen in Figure 5a ± c (Figure 5b). This agreed well with our other (lanes 2 ± 4), neither max(S) nor any of the individual observations (Table 1 and Figure 1) that the mad member proteins bound signi®cant amounts of interaction between Mmip-2 and mxi1 was the the input probe. On the other hand, as we and others weakest of all the mad family members. have previously shown (Ayer et al., 1993; Zervos et In contrast to the e€ects that GST-Mmip-2 exerted al., 1993; Hurlin et al., 1995a; Gupta et al., 1998), the on DNA binding by mad-max heterodimers, the combination of max(S) and any mad protein resulted protein had no e€ect on DNA binding by max(L) in strong DNA binding (Figure 5a ± c, lanes 5). In all homodimers. Although some loss of DNA binding was three cases, the addition of puri®ed GST-Mmip-2 occasionally seen when 4300 ng of GST-Mmip-2 was resulted in a progressive loss of DNA binding added, similar results were obtained with control GST whereas the addition of GST alone had no e€ect. In protein (not shown) that we attribute to a non-speci®c several experiments performed with mad1 and mad4, e€ect. Thus our ®ndings are consistent with the idea Mmip-2, a novel RING finger protein X-Y Yin et al 6627 that Mmip-2 exerts a speci®c inhibitory e€ect on DNA cognate mad1, mxi1, or Mmip-2 fusion vectors, binding by mad-max heterodimers. resulted in the production of more than background amounts of CAT activity. In contrast, the combination of pSG424-Mmip-2 and pNLVP-mad1 generated a In vivo interactions between Mmip-2 and mad proteins signi®cant amount of CAT activity . Much lower, but To determine whether Mmip-2 and mad 1 were capable still signi®cant, amounts of CAT activity were seen of interacting in vivo, we performed a mammalian two- with the pNLVP-mxi1 vector, thus con®rming the hybrid assay (Finkel et al., 1993; Anand et al., 1997; relatively weak association between mxi1 and Mmip-2. Langlands et al., 1997). The plasmid pSG424-Mmip-2 As an independent means of assessing the ability of was created to allow the expression of Mmip-2 as a Mmip-2 to interact with mad family members, we Gal4 DNA binding domain fusion protein. Similarly, determined its ability to reverse the previously the plasmid pNLVP was used to create two fusion described suppressive e€ects of mad proteins on a c- proteins between the VP-16 transactivation domain myc responsive reporter gene (Ayer et al., 1993; Zervos and either mad1 or mxi1. Based on the data presented et al., 1993; Hurlin et al., 1995a; Gupta et al., 1998). above, we predicted that the Gal4-Mmip-2 prey should Expression vectors containing the complete coding interact more strongly with the VP16-mad1 bait than region of each relevant cDNA were co-transfected with the VP16-mxi1 bait. As a positive control for a into NIH3T3 cells along with a reporter construct strong interaction, we measured the association containing the c-myc-responsive ornithine decarbox- between cdk4 and cyclinD1 (Anand et al., 1997; ylase (ODC) promoter juxtaposed to a CAT gene Langlands et al., 1997). As seen in Figure 6, neither (Packham and Cleveland, 1994). As we have previously pSG424 nor pVLVP16 alone, or together with the shown (Zhang et al., 1997; Yin et al., 1998), the expression of c-myc resulted in a 4 ± 8-fold up- regulation of the ODC promoter (Figure 7). Con- current expression of any one of the four mad members resulted in a strong suppression of the c- myc e€ect to baseline levels or below (Gupta et al., 1998). This was completely reversed in a dose- dependent manner, however, by co-expression of Mmip-2. As had been seen in our other assays (Figures 1, 5 and 6), Mmip-2 was less e€ective in reversing the e€ect of mxi1 than of other mad members. Similar results were observed with an arti®cial promoter (p3xMycE1b-luciferase, Gupta et al., 1993) (not shown). As a ®nal means of assessing the ability of Mmip-2 to reverse the inhibitory e€ects of mad proteins, we performed rat embryo ®broblast transformation assays with c-myc and activated ras expression plasmids. As previously demonstrated (Gupta et al., 1998), none of the individual myc network proteins, Mmip-2, or ras alone produced transformed foci (Figure 8, columns 1 ± 6). As expected, the combination of c-myc+ras expression plasmids produced a high level of focus formation that was almost entirely inhibited by the concurrent expression of either mad1, mxi1, or mad3 (Figure 8, compare column 7 with columns 9 ± 11). In the case of mad1 and mad3, complete rescue of focus- forming ability was seen when Mmip-2 was co- expressed (Figure 8, columns 12 and 14). Although rescue of focus-forming ability was also seen in the case of mxi1, the e€ect was partial (Figure 8, column 13) and did not change signi®cantly when larger amounts of the Mmip-2 expression vector were used Figure 6 Mammalian two-hybrid analysis. (a) Schematic (not shown). Mmip-2 by itself produced a small summary of the association between the VP16-mad1 fusion enhancement of c-myc+ras focus formation (Figure protein and the Gal4 binding domain (BD)-Mmip-2 fusion protein. The interaction will permit both binding to the Gal4 8, column 8). This probably re¯ects the minor tandem repeats in the CAT reporter plasmid and its transcrip- contribution of endogenous mad proteins to modulat- tional activation. (b) Results of the two-hybrid assay. The ing exogenously supplied c-myc function. indicated amounts of each plasmid were transfected into HeLa cells using a standard calcium phosphate-based technique (Anand et al., 1997; Langlands et al., 1997). Three mg of pCMV-b-Gal In situ interaction between Mmip-2 and mad proteins (Clontech) was added to each precipitate to allow normalization results in changes in subcellular distribution of transfection eciencies. Two days later cells were lysed and assayed for b-galactosidase and CAT activities as previously To determine the subcellular distribution of Mmip-2, described. CAT conversion over background was determined by phosphorImage analysis. Values shown represent the average of we expressed it as a fusion protein with a variant form three independent experiments, each performed in duplicate plates of green ¯uorescent protein (GFP `Topaz') and +1 standard error (s.e.) transiently expressed it in NIH3T3 ®broblasts. Exam- Mmip-2, a novel RING finger protein X-Y Yin et al 6628

Figure 7 Mmip-2 reverses the suppressive e€ects of mad proteins on a c-myc-responsive gene. The c-myc-responsive ODC ± CAT reporter plasmid was transfected into NIH3T3 cells along with c-myc and the indicated amounts of expression vectors for mad1 (a), mxi1 (b), mad3 (c), mad4 (d), and Mmip-2 (a±d). Each transfection was performed by a standard calcium phosphate-based procedure and pCMVb-gal was included in each precipitate to allow normalization of transfection eciencies. Two days later, cells were harvested and assayed for bgalactosidase and CAT as previously described. The results shown represent the average of three independent experiments, each performed in duplicate+1 s.e.

ination by epi¯uorescence microscopy of either live or and DePinho, 1998), it was initially unclear how they ®xed cells showed virtually all the Topaz-Mmip-2 might be able to interact with Mmip-2. To investigate protein to be concentrated in an average of three this in more detail, we created another fusion protein (range 1 ± 5) large, well circumscribed juxtanuclear foci between mad3 and a di€erent spectral variant of GFP per cell (Figure 9c). Observations of several hundred (GFP `Sapphire'). GFP-Sapphire and GFP-Topaz can cells showed that Topaz-Mmip-2 never localized to the be distinguished from one another by di€erences in nucleus. their excitation/emission maxima (Figure 9a ± d). As Because mad proteins are thought to be con®ned to expected, Sapphire-mad3 was always localized to the the nucleus (Ayer and Eisenman, 1993; Schreiber-Agus nucleus in a distinct punctate pattern (Figure 9a). Mmip-2, a novel RING finger protein X-Y Yin et al 6629 We next co-transfected the two GFP vectors at mad1 as well as the anticipated much weaker various ratios to determine whether interactions interaction between Mmip-2 and mxi1 (not shown). between Mmip-2 and mad3 might in¯uence their distribution. When co-expressed at a 1 : 1 molar ratio, Co-immunoprecipitation of Mmip2 and mad1 little change in their respective subcellular localization patterns was seen (Figure 9e,f). This segregated pattern To provide additional and independent evidence for an was seen in 480% of the cells which expressed both in vivo interaction between Mmip-2 and mad member proteins. A much di€erent pattern was seen at a 1 : 5 proteins, we performed co-immunoprecipitation experi- molar ratio of Topaz-Mmip-2 to Sapphire-mad3. ments. An expression vector encoding a myc epitope- Under these conditions, virtually all the Topaz- tagged Mmip-2 protein was co-transfected into Cos-7 Mmip-2 protein entered the nucleus and retained its cells along with a non-epitope-tagged full-length mad1 distinctive appearance (Figure 9h). expression vector. Two di€erent ratios of plasmids Conversely, when the plasmid ratios were reversed were used based on the experiments presented in the (Topaz-Mmip-2:Sapphire-mad3=5 : 1), most of the preceding section. Two days later, the cells were Sapphire-mad3 protein left the nucleus and co- harvested and the expressed Mmip-2 protein was localized with Mmip-2 in the previously described precipitated from lysates with a monoclonal antibody juxtanuclear foci (Figure 9j ± l) Thus, although mad3 directed against the myc epitope tag. Precipitates were (and presumably other mad members) and Mmip-2 are collected, resolved by SDS ± PAGE and transferred to con®ned to distinct subcellular regions, they are a PVDF membrane which was then probed in a capable, under appropriate conditions, of in¯uencing Western type analysis with a polyclonal anti-mad1 one another's location in a co-dominant and dynamic antibody. As seen in Figure 10, none of the control manner. This pattern was seen in 490% of cells lanes showed evidence of a protein related to mad1 expressing both proteins at the above described (Figure 10, lanes 1 ± 3). This generally remained true unequal ratios. Similar experiments performed in when Mmip-2 and mad 1 plasmids were co-expressed either Rat1a cells or Cos7 cells gave identical results at 1 : 1 molar ratios (Figure 10, lane 4). In contrast, (not shown). In other experiments, we have shown an when mad1 and Mmip-2 were co-expressed at a 1 : 5 identical co-dominant interaction between Mmip-2 and molar ratio, we detected a distinct mad1-speci®c band (Figure 10, lane 5). These results are consistent with those presented in the preceding section and indicate that mad1 and Mmip-2 interact most strongly with one another only when they are co-expressed at unequal ratios.

Discussion

Mmip-2 is a novel RING ®nger protein that was identi®ed in a yeast two-hybrid screen by virtue of its ability to interact with mad1, a transcriptional repressor of myc target genes (Ayer et al., 1993; Schreiber-Agus and DePinho, 1998). Mmip-2 also interacts strongly with each of the three other known members of the mad family (Zervos et al., 1993; Hurlin et al., 1995a) but associates only very weakly or not at all with c-myc, max, or several proteins which contain only bHLH or bZIP domains. The functional consequences of the Mmip-2-mad member interaction include the loss of DNA binding by mad-max heterodimers, the restoration of transcriptional po- tency by c-myc-max heterodimers, and reversal of mad- mediated inhibition of focus formation by c-myc+ras in primary ®broblasts. Using the yeast two hybrid assay as an unbiased Figure 8 Reversal of mad-mediated inhibition of c-myc+ras ± means for assessing biologically relevant pro- dependent focus formation in primary rat embryo ®broblasts. The tein : protein interactions, we determined that the indicated DNAs were added to duplicate, 100 mm plates of Fisher rat embryo ®broblasts seeded the previous day at 56105 interaction with Mmip-2 requires only the ZIP domain cells/plate. Empty vector DNAs were used to equalize the total of mad1. The ZIP domain of mad proteins is also amounts of DNA in all transfections. In addition, each plate also required for the interaction with Mmip-1, a ZIP- received 3 mg of pCMV-CAT as a control for transfection containing protein bearing no homology to Mmip-2 eciency. Two days after transfection, one plate was harvested for a CAT assay while the other was split 1 : 5 and grown in the (Gupta et al., 1998). That two unrelated proteins presence of 5% fetal calf serum until reaching 80 ± 90% should possess similar mechanisms of mad protein con¯uency at which time the serum concentration was reduced interaction suggests that interfering with ZIP dimeriza- to 2% for the remainder of the experiment. Foci were enumerated tion may be a common means of controlling myc 14 ± 21 days later and were adjusted for transfection eciencies based on CAT assays (generally 520% among samples). The network associations. results shown represent the average numbers of foci for 3 ± 5 Similar yeast two-hybrid experiments, performed independent experiments +1 s.e. with a series of N- and C-terminal deletions of Mmip-2, a novel RING finger protein X-Y Yin et al 6630

Figure 9 Subcellular localization of Mmip-2 and mad3. Mmip-2 and mad3 were expressed as full-length fusions with the Topaz or Sapphire spectral variants of GFP, respectively, under the control of the CMV immediate-early promoter. Each vector was then transiently transfected into NIH3T3 ®broblasts, either alone or in combination. Nuclei were counterstained with DAPI and the GFPs were imaged by epi¯uorescence microscopy using either `Topaz' or `Sapphire' conditions (see Materials and methods). (a) Cells were transfected with Sapphire-mad3 alone and imaged under Sapphire conditions; (b) The same cell imaged under Topaz conditions; (c) Cells were transfected with Topaz-Mmip-2 and imaged under Topaz conditions; (d) The same cell imaged under Sapphire conditions; (e) Co-transfection with Sapphire-mad3 and Topaz-Mmip-2 at a 1 : 1 molar ratio and imaged under Sapphire conditions. Note that, as in (b), most of the Sapphire-mad3 protein is nuclear; (f) the same cell imaged under Topaz conditions showing that, as in (c), most of the Topaz-Mmip-2 is juxtanuclear; (g) Co-transfection with Sapphire-mad3 and Topaz-Mmip-2 at a 5 : 1 molar ratio and imaged under Sapphire conditions. Sapphire-mad3 is again nuclear; (h) the same cell under Topaz conditions. Note that the Topaz-Mmip-2 is now nuclear. (i) a merged image of (g) and (h). Yellow/orange indicates co-localization of Sapphire- mad3 and Topaz-Mmip-2 proteins. (j) Sapphire-mad3 and Topaz-Mmip-2 were co-expressed at a 1 : 5 ratio and imaged under Sapphire conditions. Note that most of the Sapphire-mad3 protein has left the nucleus and now resides in juxtanuclear deposits; (k) The same cell imaged under Topaz conditions. Note that, as before, Topaz-Mmip-2 is also in juxtanuclear despoits; (l) The merged image of (j) and (k) shows that Sapphire-mad3 and Topaz-Mmip-2 co-localize to the juxtanuclear region. Similar patterns of co- localization were observed in both Rat1a and Cos7 cells (not shown)

Mmip-2, revealed a more complex requirement for mad function to bind other targets, perhaps as part of a protein association. Two regions, one at the N- higher order complex in association with mad proteins. terminus, immediately adjacent to and possibly over- The ®nding that Mmip-2 localizes to a small number lapping with the RING ®nger, and another at the of discrete, juxtanuclear foci, and that this pattern is extreme C-terminus appear to be required for Mmip- dynamic, is in keeping with the tendency of several 2's interaction with mad1. Although the RING ®nger other well characterized RING ®nger proteins to has now been determined to be necessary for display related behaviors. One well known example is protein : protein interactions in a variety of cases seen in the case of PML, a protein that is fused to the (Kim et al., 1996; Bellon et al., 1997; Jensen et al., retinoic acid (RAR) in the most common, 1998), other bona ®de in vivo interactions do not t(15;17), subtype of acute promyelocytic leukemia require this motif. These include the associations (Borrow et al., 1990; Longo et al., 1990; deThe et al., between the RING ®nger-containing proteins Bmi-1 1991). The nonfusion PML protein normally localizes and dinG/RING1B and the Polyhomeotic protein to large subnuclear particles of distinctly di€erent Mph2 (Hemenway et al., 1998), the small nuclear appearance than those formed by the disease-asso- RING ®nger protein SNURF and the androgen ciated PML-RAR protein (Dyck et al., 1994). receptor (Moilanen et al., 1998), and the formation Similarly, the BRCA-1 RING-®nger-containing breast of a ternary complex between cyclinH-cdk7 and the and ovarian cancer susceptibility protein localizes to MAT1 RING ®nger protein (Tassan et al., 1995). large, discrete nuclear foci in association with the However, while the RING ®nger of Mmip-2 is Rad51 and BARD1 proteins (Wu et al., 1996; Scully et dispensable for its interaction with mad family al., 1997). This subnuclear pattern rapidly disperses members, it seems reasonable to believe that it may following DNA damage. Yet another example involves Mmip-2, a novel RING finger protein X-Y Yin et al 6631 tion of mad proteins in the cytoplasm could also explain how some proliferating cells with high c-myc levels might be able to tolerate the simultaneous expression of mad proteins (Hurlin et al., 1995b). Why Mmip-2 should enter the nucleus when mad proteins are in excess, and what function it might serve in this capacity is currently unknown.

Materials and methods

Yeast two-hybrid screening and sequencing of Mmip-2 A `bait' vector consisting of the entire bHLH-ZIP domain of mad1 was constructed by amplifying codons 68 ± 140 (Ayer et al., 1993) using the polymerase chain reaction (PCR). Ampli®cation primers consisted of the following sequences: forward: 5'-CGC GAA TTC AGA CGG GCT CAT CTT CGC TTG-3'; reverse: 5'-CGC GGA TCC TTA CCT CTC AAT GCC CAG CTT CTC-3'' where bases in italics indicate `GC' clamps and EcoRI and BamHI restriction sites, respectively, to facilitate directional cloning into Figure 10 Co-immunoprecipitation of Mmip-2 and mad1. Cos-7 EcoRI+BamHI-digested pGBT9. The plasmid was trans- cells were transfected with the expression vectors for mad1 or formed into the Y153 yeast strain as previously described Mmip-2 fused to a c-myc epitope tag at its C-terminus. A control (Anand et al., 1997; Langlands et al., 1997; Gupta et al., culture (lane 1) was transfected with each of the empty vectors 1998) and a single colony was picked. Expression of a fusion alone. Two days later, cells were harvested, lysed and precipitated protein between the yeast Gal4 DNA binding domain and the with the 9E10 monoclonal antibody directed against the epitope tag. Precipitates were collected, resolved by SDS ± PAGE, electro- mad1 bHLH-ZIP domain was con®rmed by Western blotting blotted to a PVDF membrane and subjected to Western analysis of yeast extracts using a polyclonal antibody against the Gal4 using a rabbit polyclonal anti-mad1 antibody. No mad1 protein DNA binding moiety as previously described (Langlands et was detected in the empty vector control sample (lane 1), in either al., 1997) (not shown). That the mad1 bHLH-ZIP domain of the samples transfected with only the mad1 or Mmip-2 retained its proper function was con®rmed by demonstrating expression vectors (lanes 2 and 3), or in a mad1 : Mmip-2 co- a strong interaction with a pGAD10 `prey' vector expressing transfection performed at a 1 : 1 molar ratio (lane 4). In contrast, a Gal4 activation domain-max fusion to max (Table 1, line when mad1 and Mmip-2 were co-expressed at a 1 : 5 ratio, a mad1 4). Screening of a d.11 murine embryo cDNA library in the signal was readily detected (lane 5) pGAD10 yeast vector was performed as previously described (Anand et al., 1997; Gupta et al., 1998). Brie¯y, approxi- mately 46106 transformants were plated onto his-, trp-, leu- the rag-1 and rag-2 RING ®nger proteins involved in SC plates containing 20 mg/ml adenine sulfate, 20 mg/ml V(D)J recombination and B-lymphocyte development. uracil, and 30 mM 3-aminotriazole (Sigma, St Louis, MO, In cells expressing individual proteins, rag-1 localizes to USA). A total of three histidine prototrophs were strongly regions of the nucleus which stain poorly with positive for b-galactosidase production. Recovery of cDNAs propidium iodide (`PI holes') whereas rag-2 is was performed in the JEB 181 E. coli strain as previously excluded from them (Leu and Schatz, 1995). Con- described (Anand et al., 1997; Gupta et al., 1998). DNA current expression of the two proteins results in rag-2's sequence analysis showed that all three cDNAs encoded the same open reading frame. The interaction between each of co-localization with rag-1 to the PI holes. This suggests the proteins encoded by the pGAD vector and the pGBT9- that rag-1 and rag-2 either directly or indirectly mad1 bait was con®rmed by re-transformation of yeast with associate into a common macromolecular complex the appropriate combinations of vectors (Table 1, line 5, and necessary for the proper biological function of these data not shown). proteins. Based on these observations, it seems A more complete Mmip-2 cDNA was obtained by reasonable to suggest that the large juxtanuclear screening a Friend murine erythroleukemia cDNA library Mmip-2 dots seen in Figure 9 represent a complex in the lZAP vector (Stratagene, LaJolla, CA, USA) with the multiprotein structure of which Mmip-2 is only one pGAD-Mmip-2 cDNA insert. Several hybridizing clones, component. containing coding region sequence both upstream and The observation that Mmip-2 and mad3 are co- downstream of the original cDNA isolate were identi®ed and were used to reconstruct a single cDNA containing the dominant (Figure 9) suggests a novel means of entire coding region. All DNA sequencing was performed on regulation for mad member proteins. In actively an ABI Model 373 automated DNA sequencer (Applied proliferating or undi€erentiated cells, which typically Biosystems, Inc. Foster City, CA, USA) using an ABI dye contain high levels of c-myc and low levels of mad terminator cycle sequencing kit according to the supplier's protein, transcriptional repression by any mad which is directions. expressed might be kept in abeyance by virtue of its association with Mmip-2 and its exclusion from the Yeast expression vectors nucleus. The induction of high mad protein levels, such as occurs during terminal di€erentiation (Larsson et The construction of pGBT9-based vectors containing the al., 1994; Hurlin et al., 1995a,b; Vastrik et al, 1995), following fusion proteins have been previously described (Langlands et al., 1997; Gupta et al., 1998): mxi1 (amino would result in a change in Mmip-2 : mad ratios and acids 60 ± 160), mad3 (amino acids 48 ± 150), mad4 (amino the re-entry of mad into the nucleus where it could acids 46 ± 152), max (amino acids 3 ± 160), c-myc (amino acids then serve to repress myc-responsive target genes 362 ± 439), Id1 (73 ± 138), and MyoD (amino acids 83 ± 184). (Schreiber-Agus and DePinho, 1998). The sequestra- All of the expressed proteins have also been previously Mmip-2, a novel RING finger protein X-Y Yin et al 6632 demonstrated to interact speci®cally with previously identi®ed previously described (Anand et al., 1997; Langlands et al., partners (Gupta et al., 1998). pGBT9-mad1delH1, which 1997; Gupta et al., 1998). For transient transfections, the removed the basic domain and nearly all of helix 1 from the following plasmid DNAs were used: pSVLneo-c-myc, bHLH domain was constructed by using PCR to amplify pSVLneo-max (160 amino acid isoform) (Zhang et al., mad1 codons 80 ± 140. Similarly, pGBT9-mad1delZIP, which 1997), and pRcCMV (Invitrogen) containing full-length removed most of the ZIP domain, was constructed by using cDNAs for mad1, mxi1, mad3, or mad4 (Zhang et al., PCR to amplify mad1 codons 68 ± 121. Neither of these 1997; Gupta et al., 1998; Yin et al., 1998). Mmip-2 was proteins was able to interact with a pGAD-max fusion expressed in the pSVLneo vector after altering its initiator protein in a two-hybrid assay (Table 1, lines 30 and 32), thus ATG so as to conform better to a consensus Kozak site con®rming the need for an intact HLH-ZIP domain. (GCA ATG G?ACA ATG G). The region between nt 124 and 1087, containing the entire coding region of Mmip-2 (Figure 2), was ampli®ed with primers containing unique GST pulldowns XhoI restriction sites, digested with XhoI and cloned into the The original Mmip-2 cDNA isolate, encoding amino acids XhoI site of the vector. The correct structure of the resultant 3 ± 302, was excised from the pGAD10 parental vector and construct was con®rmed by restriction mapping and DNA cloned into the pGEX-4T expression vector (Pharmacia, sequencing. Reporter constructs consisted of ODC-CAT or Piscataway, NJ, USA). GST-Mmip-2, or GST alone, was p3xMycE1bLuc (Gupta et al., 1993; Packham and Cleveland, puri®ed to near homogeneity as previously described (Anand 1994). Each transfection mix also contained 3 mgof et al., 1997; Gupta et al., 1998). 35S-methionine-labeled full- pCMVbGal to allow for normalization of transfection length c-myc, max, mad1, mxi, mad3, and mad4 were eciencies (Zhang et al., 1997). All plasmid DNAs were synthesized in a coupled in vitro transcription/translation puri®ed using Qiagen columns. Transfections were performed reaction (Promega, Madison, WI, USA) as previously at least three times in duplicate plates using a calcium described (Gupta et al., 1998). Approximately one-quarter phosphate-based procedure. b-galactosidase, luciferase, and of each reaction was incubated with 0.5 mg of GST or GST- CAT assays were performed as previously described (Zhang Mmip-2 and then precipitated by the addition of 30 mlof et al., 1997; Gupta et al., 1998; Yin et al., 1998). Focus glutathione-agarose beads. The precipitates were collected by formation assays in primary rat embryo ®broblasts were centrifugation and washed three times in wash bu€er (Gupta performed with Lipofectamine (GIBCO ± BRL, Grand Island, et al., 1998) containing 150 ± 350 mM NaCl depending upon NY, USA) as previously described (Gupta et al., 1998). the protein combination under study. The ®nal pellets were resuspended and boiled in loading bu€er, resolved by SDS- Mammalian two-hybrid assays 12% polyacrylamide gel electrophoresis (SDS ± PAGE) and processed for ¯uorography after ®xing the gel in Amplify The original Mmip-2 cDNA isolate (codons 3 ± 302) was (Amersham, Bucks, UK). In a separate lane, the same excised from the parental pGAD10 vector and recloned into amount of total in vitro translated protein was included to the pSG424 mammalian two-hybrid vector (Finkel et al., serve as the denominator for the amount of product bound 1993). The resultant vector (pSG424-Mmip-2) encodes a by GST-Mmip-3 or GST alone. fusion protein consisting of the Gal4 DNA binding domain and the nearly full-length Mmip2. The plasmid pNLVP (Finkel et al., 1993) was used to create a fusion between the Electrophoretic mobility shift assays (EMSAs) VP16 transactivation domain and full-length mad1 or mxi1. All proteins except c-myc were expressed as full-length amino Transient transfections were performed in HeLa cells as terminally-tagged hexahistidine fusions in one of the QE series previously described using the pGal5E472CAT reporter of bacterial expression vectors (Qiagen, Chatsworth, CA, plasmid (Langlands et al., 1997). 3 mg of the pCMV-b-gal USA) (Prochownik and VanAntwerp, 1993; Zhang et al., plasmid (Clontech, Palo Alto, CA, USA) was included as an 1997; Gupta et al., 1998; Yin et al., 1998). In the case of c-myc, internal control for transfection eciencies. amino acids 346 ± 439 were expressed to allow improved solubility of the recombinant product (Prochownik and Northern blotting VanAntwerp, 1993). All proteins were puri®ed to 490% homogeneity using nickel-agarose anity chromatography as To determine the size of the endogenous Mmip-2 transcript, previously described (Prochownik and VanAntwerp, 1993) total RNA (10 mg/lane) was electrophoresed in a 1% agarose- and were stored in 16 binding bu€er in small aliquots at formaldehyde gel and transferred to a Nytran membrane 7808C. EMSAs were performed as previously described (Schleicher and Schuell, Keene, NH, USA). The blot was (Prochownik and VanAntwerp, 1993; Zhang et al., 1997; hybridized sequentially with a 32P-labeled cDNA containing Yin et al., 1998) using approximately 20 ng of each myc the entire coding region of Mmip-2 under standard network protein and the indicated amounts of GST or GST- conditions and a cDNA encoding murine a-actin. To Mmip-2. Protein interactions were allowed to proceed at 408C determine the tissue distribution of Mmip-2, a `Mouse for 20 min prior to the addition of 1 mg of poly(dI):(dC) RNA Master Blot' (Clontech), containing 2 mg of `dot- (Pharmacia) and approximately 0.1 ng of the palindromic, blotted' poly(A)+ RNA from 18 di€erent adult murine tissues double-stranded E0(GAC) c-myc binding site oligonucleotide and total embryos of four di€erent gestational ages was also (Prochownik and VanAntwerp, 1993), end-labeled with 32P-g- hybridized with the Mmip-2 cDNA probe. Blots were ATP (sp. act. 43000 Ci/mmol, Amersham, Arlington, IL, exposed to BioMax MS X-ray ®lm (Kodak, Rochester, NY, USA) in a standard polynucleotide kinase reaction to a USA) with an intensifying screen at 7808C for 4 h ± 3 days. speci®c activity of 3 ± 56108 d.p.m./mg. The ®nal volume of all reactions was 20 ml. Following an additional 20 min incuba- Subcellular localization tion at room temperature, the reactions were electrophoresed at room temperature on a 4% non-denaturing polyacrylamide Two spectral variants of green ¯uorescent protein (GFP) gel, dried and processed for autoradiography (Prochownik termed `Topaz' and `Sapphire' (Packard Instruments) were and VanAntwerp, 1993; Zhang et al., 1997; Yin et al., 1998). used to study the interaction between Mmip-2 and mad3. The excitation/emission maxima for the proteins (To- paz=514 nm/527 nm; Sapphire=395 nm/510 nm) are such Cell culture and DNA transfections that they can each be separately identi®ed without signi®cant NIH3T3, Cos-7, HeLa and primary rat embryo ®broblasts interference from the other, thus making them of utility in were maintained, and DNA transfections were performed, as co-expression studies. To construct mammalian expression Mmip-2, a novel RING finger protein X-Y Yin et al 6633 vectors, the cDNAs for each GFP variant were ampli®ed by This fragment was cloned into the pSVLneo-myc tag vector PCR using the following primers: forward: 5'-CGC GCT so that the Mmip-2 coding region was in-frame with the c- AGC ACC ATG GTG AGC AAG GGC GAG GA-3' where myc epitope, EQKLISQQDL. Transfection of Cos-7 cells the italics indicate a `GC clamp' and an NheI restriction site, with this vector con®rmed the production of an appropriately and the underlined region represents an optimized Kozak sized myc-tagged Mmip-2 protein that was easily detected by element-translation initiation site; reverse: 5'-CGC AGA TCT Western blotting using the 9E10 monoclonal antibody (Santa CTT GTA CAG CTC GTC CAT GCC-3' where the italics Cruz Biotechnology, Inc., Santa Cruz, CA, USA) (not again indicate a GC clamp and a BglII restriction and the shown). For co-immunoprecipitation studies, 5 mg of the underlined sequence indicates the codon of the last amino above vector, either alone or together with the indicated acid in the GFP protein. After digestion of the ampli®ed amounts of pSVLneo-mad1 were co-transfected into Cos-7 fragments with NheI and BglII and puri®cation by agarose cells using a standard calcium phosphate transfection gel electrophoresis, they were used to replace wild-type GFP protocol (Gupta et al., 1998). Two days later, cells were coding sequences in the pEGFP-C1 mammalian expression washed twice in phosphate-bu€ered saline (PBS) and then vector (Clontech). The plasmid containing Topaz sequences lysed in 1 ml of PBS-1% NP-40 containing Complete1 was next next digested with SalI to allow for ligation of a protease inhibitor cocktail (Boehringer-Mannheim, Indiana- full-length Mmip-2 cDNA excised from the pSVLneo-Mmip- polis, IN, USA) at the concentration recommended by the 2 expression vector with XhoI. This fragment contains 6 bp supplier. After clearing of the lysate by centrifugation of 5' untranslated region derived from the original Mmip-2 (10 000 g for 10 min), 0.5 ml was precipitated by the cDNA and thus introduces only two amino acids between the addition of 9E10 antibody to a ®nal dilution of 1 : 1000 end of GFP and the beginning of Mmip-2. This vector was After overnight incubation at 48C, antigen-antibody com- designated pTopaz-Mmip-2 Similarly, a PCR-ampli®ed plexes were precipitated by the addition of 40 ml Staph A cDNA fragment containing the entire coding region of protein-agarose beads (BioRad, Hercules, CA, USA) for an mad3 and 6 bp of 5'-untranslated region was ligated into additional 2 h at room temperature. Precipitates were the Sal site of the Sapphire expression vector to create collected by centrifugation, washed ®ve times in PBS- pSapphire-mad3. The identity, orientation and reading frame 1%NP-40, boiled in standard SDS ± PAGE lysis bu€er, and of each plasmid was con®rmed by DNA sequencing using a resolved by SDS ± PAGE. The gel was then electroblotted to GFP primer. Western blots of lysates from transiently a PVDF membrane as previously described (Langlands and transfected Cos7 cells were probed with an anti-GFP Prochownik, 1997). The blot was incubated with a 1 : 500 monoclonal antibody (Clontech) to verify the expression of dilution of a rabbit polyclonal anti-mad1 antibody that had Topaz-Mmip-2 and Sapphire-mad3 fusion proteins of the been raised against the full-length hexahistidine fusion predicted size (not shown). Each of the above plasmids was protein described above. Blocking was performed in PBS-T then introduced, either individually or in combination, and at containing 5% non-fat dried milk (Langlands and Prochow- di€erent molar ratios, into NIH3T3 cells using Lipofectamine nik, 1997). Following extensive washing and incubation with (GIBCO ± BRL). The cells had been plated the day before a second horseradish peroxidase-labeled goat-anti rabbit onto glass coverslips in 6 well tissue culture plates. Following secondary antibody, the blot was developed using a a 3 h application of the DNA-Lipofectamine mixture to cells chemiluminescence kit (Renaissance, New England Nuclear, in serum-free medium, an equivalent volume of fresh medium Boston, MA, USA) according to the supplier's recommenda- containing 20% supplemented calf serum (GIBCO ± BRL) tions. was added to the cells for an additional 3 ± 4 h. Cells were ®xed in 2%-paraformaldehyde-PBS prior to viewing by epi¯uorescence microscopy. For this we used a Nikon Diaphot inverted microscope equipped with an attenuated Note added in proof xenon arc lamp, a cooled CCD camera and a Sutter Lambda The sequence of MMIP-2 has been deposited with 10-2 ®lter wheel for changing excitation ®lters (400+20 nm GenBank (Accession no. AF190166). for Sapphire, 505+5 nm for Topaz, and 340+nm for DAPI). A 535+25 nm emission ®lter was used for imaging GFP- tagged proteins while a 480+30 nm emission ®lter was used Acknowledgements for imaging DAPI-labeled nuclei. We thank Simon Watkins for advice on epi¯uorescence microscopy and for help in preparing Figure 9, Linette Grove for DNA sequencing, and members of the Co-immunoprecipitation experiments Prochownik lab for comments on the manuscript. This Standard PCR techniques were used to amplify the complete work was supported by NIH grants HL33741 to EV coding sequence of Mmip-2 minus the termination codon. Prochownik and NS32385 to ES Levitan.

References

Anand G, Yin XY, Shahidi AK, Grove L and Prochownik BlackwellTK,HuangJ,MaA,KretznerL,AltFW, EV. (1997). J. Biol. Chem., 272, 19140 ± 19151. Eisenman RN and Weintraub H. (1993). Mol. Cell Biol., Ayer DE and Eisenman RN. (1993). Genes Dev., 7, 2110 ± 13, 5216 ± 5224. 2119. Blackwood EM and Eisenman RN. (1991). Science, 251, Ayer DE, Kretzner L and Eisenman RN. (1993). Cell, 72, 1211 ± 1217. 211 ± 222. Blackwood EM, Luscher B and Eisenman RN. (1992). Genes Ayer DE, Lawrence QA and Eisenman RN. (1995). Cell, 80, Dev., 6, 71 ± 80. 767 ± 776. Borrow J, Goddard AD, Sheer D and Solomon E. (1990). Barrett J, Birrer MJ, Kato GJ, Dosoka-Akita H and Dang Science, 249, 1577 ± 1580. CV. (1992). Mol. Cell Biol., 12, 3130 ± 3137. Dang CV. (1999). Mol. Cell. Biol., 19, 1 ± 11. Beijersbergen RL, Hijmans EM, Zhu L and Bernards R. DePinho RA, Schreiber-Agus N and Alt FW. (1991). Adv. (1994). EMBO J., 13, 4080 ± 4086. Cancer Res., 57, 1±46. Bellon SF, Rodgers KK, Schatz DG, Coleman JE and Steitz DeThe H, Lavau C, Marchio A, Chomniene C, Degos L and TA. (1997). Nat. Struct. Biol., 4, 586 ± 591. Dejean A. (1991). Cell, 66, 675 ± 684. Mmip-2, a novel RING finger protein X-Y Yin et al 6634 Dyck JA, Maul GG, Miller Jr WH, Chen JD, Kakizuka A Ma A, Moroy T, Collum R, Weintraub H, Alt FW and and Evans RN. (1994). Cell, 76, 333 ± 343. Blackwell TK. (1993). Oncogene, 8, 1093 ± 1098. Evan GI and Littlewood TD. (1993). Curr. Opin. Genet. McMahon SB, Van Buskirk HA, Dugan KA, Copeland TD Dev., 3, 44 ± 49. and Cole MD. (1998). Cell, 94, 363 ± 374. Facchini LM and Penn LZ. (1998). FASEB J., 12, 633 ± 651. Moilanen AM, Poukka H, Karvonen U, Hakli M, Janne OA Finkel T, Duc J, Fearon ER, Dang CV and Tomaselli GF. and Palvimo JJ. (1998). Mol. Cell Biol., 18, 5128 ± 5139. (1993). J. Biol. Chem., 268, 5±8. Nesbit CE, Tersak JM and Prochownik EV. (1999). Foley KP, McArthur GA, Queva C, Hurlin PJ, Soriano P Oncogene, 18, 3004 ± 3016. and Eisenman RN. (1998). EMBO J., 17, 774 ± 785. Packham G and Cleveland JL. (1994). Mol. Cell Biol., 14, Fremont PS. (1993). Ann. NY Acad. Sci., 684, 174 ± 192. 5741 ± 5747. GaubatzS,ImhofA,DoschR,WernerO,MitchellP, Peukert K, Staller P, Schneider A, Carmichael G, Hanel F Buettner R and Eilers M. (1995). EMBO J., 14, 1508 ± and Eilers M. (1997). EMBO J., 16, 5672 ± 5686. 1519. Prendergast GC, Lawe D and Zi€ EB. (1991). Cell, 65, 395 ± Gu W, Bhatia K, Magrath I, Dang CV and Dalla-Favera R. 407. (1994). Science, 264, 251 ± 254. Prochownik EV and VanAntwerp ME. (1993). Proc. Natl. Gupta K, Anand G, Yin XY and Prochownik EV. (1998). Acad.Sci.USA,90, 960 ± 964. Oncogene, 16, 1149 ± 1159. Queva C, Hurlin PJ, Foley KP and Eisenman RN. (1998). Gupta S, Seth A and Davis RJ. (1993). Proc. Natl. Acad. Sci. Oncogene, 16, 967 ± 977. USA, 90, 3216 ± 3220. Sakamuro D, Elliott KJ, Wechsler-Reya R and Prendergast Hassig CA, Fleischer TC, Billin AN, Schreiber SL and Ayer GC. (1996). Genet., 14, 69 ± 77. DE. (1997). Cell, 89, 341 ± 347. Saurin AJ, Borden KLB, Boddy MN and Fremont PS. Hemenway CS, Halligan BW and Levy LS. (1998). (1996). Trends Biochem., 21, 208 ± 214. Oncogene, 16, 2541 ± 2547. Schreiber-Agus N, Chin L, Chen K, Torres R, Rao G, Guida Henriksson M and Luscher B. (1996). Adv. Cancer Res., 68, P, Skoultchi AI and DePinho RA. (1995). Cell, 80, 777 ± 109 ± 182. 786. Hu HM, O'Rourke K, Boguski MS and Dixit VM. (1994). J. Schreiber-Agus N and DePinho RA. (1998). Bioessays, 20, Biol. Chem., 269, 30069 ± 30072. 808 ± 818. Hurlin PJ, Foley KP, Ayer DE, Eisenman RN, Hanahan D Schreiber-Agus N, Meng Y, Hoang T, Hou H, Chen K, and Arbeit JM. (1995a). Oncogene, 11, 2487 ± 2501. Greenberg R, Cordon-Cardo C, Lee H-W and DePinho Hurlin PJ, Queva C, Koskinen PJ, Steingrimsson E, Ayer RA. (1998). Nature, 393, 483 ± 487. DE, Copeland NG, Jenkins NA and Eisenman RN. Scully R, Chen J, Plug A, Xiao Y, Weaver D, Feunteun J, (1995b). EMBO J., 14, 5646 ± 5659. Ashley T and Livingston DM. (1997). Cell, 88, 265 ± 275. Jensen DE, Procter H, Marquis ST, Gardner HP, Ha SI, Shrivastava A, Saleque S, Kalpana GV, Artavidi S, Go€ SP Chodosh LA, Ishov AM, Tommerup N, Vissing H, Sekido and Calame K. (1993). Science, 262, 1889 ± 1892. Y, Minna J, Borodovsky A, Schultz DC, Wilkinson KD, Shrivastava A, Yu J, Artandi S and Calame K. (1996). Proc. Maul GG, Barlev N, Berger SL, Prendergast GC and Natl. Acad. Sci. USA, 93, 10638 ± 10641. Rauscher Jr FJ. (1998). Oncogene, 16, 1097 ± 1112. St.ArnaudR,NepveuA,MarcuKBandMcBurneyWM. Kato GJ, Barrett J, Villa-Garcia M and Dang CV. (1990). (1988). Oncogene, 3, 553 ± 559. Mol. Cell Biol., 10, 5914 ± 5920. Tassan JP, Jaquenoud M, Fry AM, Frutiger S, Hughes GJ Kim SS, Chen YM, O'Leary E, Witzgall M and Bonventre and Nigg EA. (1995). EMBO J., 14, 5608 ± 5617. JV. (1996). Proc. Natl. Acad. Sci. USA, 93, 15299 ± 15304. Vastrik I, Kaipainen I, Penttila T-L, Lymboussakis A, Lachman HM and Skoultchi AI. (1984). Nature, 310, 592 ± Alitalo R, Parvinen M and Alitalo K. (1995). J. Cell. 594. Biol., 128, 1197 ± 1208. Laherty CD, Yang W-M, Sun J-M, Davie JR, Seto E and Wu LC, Wang ZW, Tsan JT, Spillman MA, Phung A, Xu Eisenman RN. (1997). Cell, 89, 349 ± 356. XL, Yang MC, Hwang LY, Bowcock AM and Baer R. Lahoz EG, Xu L, Schreiber-Agus N and DePinho RA. (1996). Nature Genet., 4, 430 ± 440. (1994). Proc. Natl. Acad. Sci. USA, 91, 5503 ± 5507. Yin X-Y, Grove LE and Prochownik EV. (1998). Oncogene, Langlands K, Anand G, Yin X-Y and Prochownik EV. 16, 2629 ± 2637. (1997). J. Biol. Chem., 272, 19785 ± 19793. Zervos A, Gyuris J and Brent R. (1993). Cell, 72, 223 ± 232. Langlands K and Prochownik EV. (1997). Anal. Biochem., Zervos A, Gyuris J and Brent R. (1994). Erratum. Cell 79, 249, 250 ± 252. 389. Larsson L-G, Pettersson M, Oberg F, Nilsson K and Luscher Zhang H, Fan S and Prochownik EV. (1997). J. Biol. Chem., B. (1994). Oncogene, 9, 1247 ± 1252. 272, 17416 ± 17424. Leu TMJ and Schatz DG. (1995). Mol. Cell Biol., 15, 5657 ± Zimmerman KA, Yancopoulous GD, Collum RG, Smith 5670. RK, Kohl NE, Denis KA, Nau MM, Witte ON, Toran- Longo L, Pandol® PP, Biondi A, Rombaldi A, Mencarelli A, Allerand D, Gee CE, Minna JD and Alt FW. (1986). LoCoco F, Direiro D, Pegoraro L, Avanzi G, Tabilio A, Nature, 319, 780 ± 783. Zangrilli D, Alcalay M, Donti E, Grignani F and Pelicci PG. (1990). J. Exp. Med., 172, 1571 ± 1575.