Oncogene (2008) 27, 308–317 & 2008 Nature Publishing Group All rights reserved 0950-9232/08 $30.00 www.nature.com/onc ORIGINAL ARTICLE p63(TP63) elicits strong trans-activation of the MFG-E8/lactadherin/ BA46 through interactions between the TA and DN isoforms

T Okuyama1,2,3, S Kurata4,5, Y Tomimori1, N Fukunishi4, S Sato6, M Osada7, K Tsukinoki5, H-F Jin8, A Yamashita8, M Ito8, S Kobayashi2, R-I Hata5, Y Ikawa1,9 and I Katoh1,8

1Ikawa Laboratory, RIKEN, Wako, Japan; 2Department of Molecular Physiology, Kyoritsu University of Pharmacy, Tokyo, Japan; 3Department of Cell Regulation, Medical Research Institute, Tokyo Medical and Dental University, Tokyo, Japan; 4Department of Redox Response Cell Biology, Medical Research Institute, Tokyo Medical and Dental University, Tokyo, Japan; 5Oral Health Science Research Center, Kanagawa Dental College, Yokosuka, Japan; 6Department of Immune Regulation, Tokyo Medical and Dental University, Tokyo, Japan; 7Human Gene Sciences Center, Tokyo Medical and Dental University, Tokyo, Japan; 8Department of Microbiology, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Yamanashi, Japan and 9Department of Hematology, Tokyo Medical and Dental University, Tokyo, Japan

We report here that human MFGE8 encoding milk fat Introduction globule-EGF factor 8 (MFG-E8), also termed 46 kDa breast epithelial antigen and lactadherin, is p63 (TP63), a member of the p53 (TP53) gene family transcriptionally activated by p63, or TP63, a p53 (TP53) (Osada et al., 1998; Yang et al., 1998), is essential for family protein frequently overexpressed in head-and-neck embryonic epithelial tissue development (Celli et al., squamous cell carcinomas, mammary carcinomas and so 1999; Mills et al., 1999; Yang et al., 1999; Pellegrini on. Despite that human MFG-E8 was originally identified et al., 2001; Brunner et al., 2002). It is frequently over- as a breast cancer marker, and has recently been reported expressed in head-and-neck squamous cell carcinomas to provide peptides for cancer immunotherapy, its trans- and in other cancers derived from epithelial basal cells of criptional control remains an open question. Observations breast, lung, cervix, urinary tracts, and so on (Hibi et al., in immunohistochemical analyses, a tetracycline-induced 2000; Quade et al., 2001; DiRenzo et al., 2002; Koga p63 expression system and keratinocyte cultures sug- et al., 2003; Massion et al., 2003) as well as in gested a physiological link between p63 and MFGE8.By keratinocyte stem cells (Pellegrini et al., 2001). reporter assays with immediately upstream regions of Reflecting the considerable structural similarity to MFGE8, we determined that the trans-activator (TA) p53, p63 also trans-activates various through isoforms of p63 activate MFGE8 transcription though a p53-binding consensus sequences (p53 motifs) and/or p53/p63 motif at À370, which was confirmed by a related sites (Osada et al., 1998, 2005b). The p63 chromatin immunoprecipitation experiment. Upon siRNA- gene is transcribed into 50-terminal RNA isoforms mediated p63 silencing in a squamous cell carinoma line, encoding either the trans-activator (TA) domain or the MFG-E8 production decreased to diminish Saos-2 cell N-terminally truncated (DN) domain, each of which adhesion. Interestingly, the DN-p63 isoform lacking the is further processed to the a, b and g variants by TA domain enhanced the MFGE8-activating function of 30-terminal splicing. Among them, TAp63g (p51A) was TA-p63, if DN-p63 was dominant over TA-p63 as first identified as a TA corresponding to tumor typically observed in undifferentiated keratinocytes and suppressor p53. In contrast, the DN-p63 isoforms were squamous cell carcinomas, implying a self-regulatory initially described as dominant-negative type mechanism of p63 by the TA:DN association. MFG-E8 against TA-p63. With the oligomerization domain, DN- may provide a novel pathway of epithelial–nonepithelial p63 binds to TA-p63. Recent studies, however, detected cell interactions inducible by p63, probably in pathological positive effects of the DN isoforms upon certain processes. promoters (Kurata et al., 2004; Wu et al., 2005; Helton Oncogene (2008) 27, 308–317; doi:10.1038/sj.onc.1210646; et al., 2006). published online 16 July 2007 Breast epithelial antigen, also termed lactadherin and human milk fat globule protein (HMFG) (GenBank Keywords: p63; MFG-E8; lactadherin; BA46; P53 S56151), was identified as a marker of breast cancers (Larocca et al., 1991; Taylor et al., 1997), and is now known as the human ortholog (GenBank NM_005928, NP_005919) to milk fat globule-EGF factor 8 protein Correspondence: Dr I Katoh, Department of Microbiology, Inter- (MFG-E8) of mouse and other mammals (Mather et al., disciplinary Graduate School of Medicine and Engineering, University 1993; Hvarregaard et al., 1996). In this report, we refer of Yamanashi, Chuo, Yamanashi 409-3898, Japan. to the human protein as MFG-E8, and the gene as E-mail: [email protected] Received 24 October 2006; revised 4 June 2007; accepted 5 June 2007; MFGE8. MFG-E8 is a membrane-associated, heavily published online 16 July 2007 glycosylated secretory protein with two distinct types of Transcriptional activation of MFGE8 by p63 T Okuyama et al 309 adhesion sites: the Arg-Gly-Asp triplet arranged to bind a TAp63γ TAp63α Np63γ Np63α to integrin avb3 or avb5; and the discoidin domains (C1 --+ + - + - + tet and C2) that interact with phosphatidylserine located on TAp63α the inner side of plasma membrane (Taylor et al., 1997; ∆Np63α Shi and Gilbert, 2003). Mouse Mfge8 produces MFG- TAp63γ E8-L and MFG-E8-S, in which the former is mouse- ∆Np63γ Western nonspecific specific, and the latter conserved among humans and blot β-actin animals (Watanabe et al., 2005). While the L form containing a Pro/Thr-enriched domain facilitates pha- MFGE8 gocytosis of apoptotic cells by macrophages (Hanayama p21WAF1 (32) et al., 2002, 2004), the S form was reported to support RT-PCR the gamete interaction upon fertilization (Ensslin and MFGE8 (26) Shur, 2003). Very importantly, recent studies deter- 18S rRNA (15) mined peptide sequences derived from human MFG-E8 as the potential targets for tumor immunotherapy b Days after Ca2+ -induction (Carmon et al., 2002; Liu et al., 2005). 0 3 7 In addition to the increased expression of MFGE8 TAp63α in lactating and transformed breast epithelial cells ∆Np63α (Peterson et al., 1990), steady-state expression also TAp63γ occurs in various tissues (Kruger et al., 2000; Watanabe Western ∆Np63γ et al blot ., 2005). However, little is known about the MFG-E8 transcriptional control of MFGE8. This study shows that p63 enhances the MFGE8 promoter and offers an β-actin explanation about the MFG-E8 augmentation in carcinomas. IVL

RT-PCR MFGE8

Results 18S rRNA Figure 1 Enhancement of MFGE8 expression by p63. (a) Induction of MFGE8 by p63 Tetracycline-inducible p63 expression system. Induced cells ( þ ) We tested whether p63 affected MFGE8 expression in and control cells (À) were analysed for the p63 and milk fat HEK293 cells, using the tetracycline-inducible p63 globule-EGF factor 8 protein (MFG-E8) proteins (upper panels). expression system (Osada et al., 2005b). After induction Positions of the p63 isoforms were indicated. Isoform-specific reverse transcription (RT)–PCR was carried out (lower panels). by tetracyclin, TAp63g (57 kDa), also termed p51A PCR cycle number is indicated for each experiment. (b) Neonatal (Osada et al., 1998; Yang et al., 1998), and TAp63a human keratinocyte (NHK) culture. Cells before (0) and after (on (85 kDa), or p51B, increased the MFGE8 transcript days À3 and À7) the Ca2 þ (1 mM) input were analysed by western approximately fivefold and threefold, respectively, blotting and RT–PCR for p63 and MFG-E8. b-Actin and 18S as determined by reverse transcription (RT)–PCR rRNA were tested to ensure equal sample loading. (Figure 1a). Neither DNp63g (47 kDa) nor DNp63a (75 kDa) altered the level of MFGE8 mRNA. Western blot analyses supported these findings. promoter/enhancer region upstream of MFGE8, encom- Both MFGE8 and p63 are endogenously expressed in passing À1890 to þ 50, nucleotide numbers relative to neonatal human keratinocytes (NHK) (Figure 1b). After the transcription start site ( þ 1), contained three full-site these cells were induced to differentiate by increasing the motifs at positions À1198, À370 and À27, and a single Ca2 þ concentration in the medium to 1 mM, p63 had half site at À742, in agreement with the decreased gradually on day 4 and day 7 at the mRNA database (Figure 2A). We constructed luciferase gene and protein levels. MFGE8 mRNA diminished concur- (luc) expression plasmids, 2K-luc, 1K-luc and S-luc, rently. In contrast, IVL mRNA coding for involucrin, having nucleotides À1890 to þ 50, À877 to þ 50, and an early differentiation marker absent in the p63- À250 to þ 50, respectively. Trans-activation assays in localized basal layer of epidermis (Li et al., 2000), HEK293 cells indicated that both p53 and TAp63g substantially increased. enhanced the 2K-luc promoter approximately threefold, and TAp63a did so 1.5-fold (Figures 2Ba and b). Presence of p53/p63 motifs in the MFGE8 Neither DNp63g nor DNp63a affected 2K-luc (Figures promoter/enhancer region 2Bc and d), supporting the results of the drug-inducible The p53 binding consensus sequences comprise a system (Figure 1a). The 1K-luc plasmid could also tandem repeat of the half-site motif, RRRC(A/T)(A/ respond to TAp63g, while S-luc failed to do so, T)GYYY, in which R and Y represent purine and localizing the main p63-responsive element between pyrimidine, respectively (el-Deiry et al., 1992). p63 uses À877 and À249 (Figure 2Ba). p53, however, appeared to the same or similar motifs for trans-activation (Osada enhance the 2K, 1K and S promoters equally, implying et al., 1998, 2005b; Yang et al., 1998). The cloned importance of the S region (Figure 2Be).

Oncogene Transcriptional activation of MFGE8 by p63 T Okuyama et al 310 A Transcription Start Site(+1) -1198 -742 -370 -27

-1890 +50 2K-luc -877 1K-luc -250 S-luc

abTAp63 γ TAp63α B 3.5 3.5 3 3 2.5 2.5 2 2 1.5 1.5 1 1 0.5 0.5 0 0 2K 1K S 2K 1K S

3.5 cde∆Np63 γ∆3.5 Np63α 3.5 p53 3 3 3

Activation (-fold) 2.5 2.5 2.5 2 2 2 1.5 1.5 1.5 1 1 1 0.5 0.5 0.5 0 0 0 2K 1K S 2K 1K S 2K 1K S

AA AA C RRRC GYYY-N -RRRC GYYY Consensus TT 0-13 TT (full-site) half-site half-site HS1 AGGCTTGCTT M1 ******T*** . HS2/3 AAGCATGCTTGGACCTGTTC M2 ******T************* M23 ******T*********T*** . . . . . HS4/5 TGGCCAGCTTGGGCGGAGCGCACGGCC M45 ******T*** ******T***

D -742 -370 -27 HS1 HS2/3 HS4/5 1K-luc S-luc M1-luc M2-luc M1-3-luc M45-luc M1-5-luc

Figure 2 The promoter/enhancer sequences upstream of MFGE8.(A) Outline of the cloned MFGE8 promoter/enhancer region. The top scheme shows location of the p53-binding consensus sequences. Separate and paired ovals indicate half sites and full sites of the consensus, respectively. Filled boxes indicate two CAT box motifs. Line drawing represents DNA segments, 2K, 1K and S, cloned into the luc vector. The 50 and 30 end points are indicated by nucleotide numbers. (B) Trans-activation assays in HEK293 cells. A p63 expression plasmid (pRcCMV-p53, -TAp63g, -TAp63a or -DNp63g) or control plasmid pRcCMV was co-transfected with a luc expression plasmid (2K-, 1K- or S-luc). Bar graphs indicate relative luciferase activities (-fold activation) in comparison with the activity obtained with pRcCMV (1-fold). Each bar represents the average of triplicates. (C) The p53-binding consensus sequences forming the full-site/half-site motifs within the 1K region (top). Nucleotide sequences at the HS1, HS23 and HS45 sites are aligned, where dotted nucleotides indicate original sequence deviations from the consensus. Altered (T) and preserved (asterisk) sequences in the mutant plasmids, M1, M2, M23 and M45 are indicated. (D) Illustration of the mutant luc plasmids. Open and filled ovals denote the original and mutated motifs. 1K-luc and S-luc are also aligned.

Oncogene Transcriptional activation of MFGE8 by p63 T Okuyama et al 311 Determination of the main p63-responsive sites in the suggesting involvement of an unidentified site between MFGE8 promoter À250 and þ 50 or an indirect mechanism. We analysed the 1K region, focusing on a half site at À742 (HS1), a full site at À370 (HS2/3) and a full site at À27 (HS4/5) for its responsiveness to p63. Mutant Stabilization of TA-p63 by DN-p63 in HeLa promoter/enhancer plasmids, MI, M2, M1-3, M45 and In HeLa cells, TAp63g enhanced 1K-driven luc expres- M1-5, were rendered by introducing a core nucleotide sion 5.3-fold. Results with the M2 and M45 mutants mutation, ‘G to T’, at the seventh position of each half- confirmed HS2/3 as the primary target site of TA-p63 site decamer in various combinations (Figures 2C and (Figures 3Ba and b). Surprisingly, DNp63g and DNp63a D). In HEK293 cells, the sensitivity of 1K to TAp63g activated 1K-luc 12-fold and 25-fold, respectively diminished when HS2 was mutated in the forms of M2- (Figures 3Bc and d). Apparently, DNp63a was fivefold luc, M1-3-luc and M1-5-luc (Figure 3Aa). In contrast, more efficient than TAp63g in activation of the 1K, M2 the M1 and M45 mutants retained the full sensitivity to and M45 promoters. TAp63g. The TA-p63 proteins including TAp63g and To gain an insight into the positive effects of the DN TAp63a (Figure 3Ab) were thought to activate MFGE8 isoforms in HeLa, we analysed several epithelial cell transcription through HS2/3. However, none of the lines for p63 expression (Figure 4a). FaDu cells derived mutations affected p53 significantly (Figure 3Ae), from a hypopharyngeal squamous cell carcinoma

A HEK293 4.5 abTAp63 γ 4.5 TAp63α 4 4 3.5 3.5 3 3 2.5 2.5 2 2 1.5 1.5 1 1 0.5 0.5 0 0 1 2 1 2 1K M M 1-3 45 1K M M 1-3 M M M1-5 M M45 M1-5

4.5 cd∆Np63 γ 4.5 ∆Np63α 4.5 e p53 4 4 4 Activation (-fold) 3.5 3.5 3.5 3 3 3 2.5 2.5 2.5 2 2 2 1.5 1.5 1.5 1 1 1 0.5 0.5 0.5 0 0 0 1 2 1 2 1 1K M M 1-3 45 1K M M 1-3 45 1K M M2 1-3 45 M M M1-5 M M M1-5 M M M1-5 B HeLa 7 abTAp63 γ 7 TAp63α 30 cd∆Np63 γ 30 ∆Np63α 6 6 25 25 5 5 20 20 4 4 15 15 3 3 10 10 2 2 Activation (-fold) 5 1 1 5 0 0 0 0 1K 1K 1K 1K M2 M45 M2 M45 M2 M45 M2 M45 Figure 3 Luciferase reporter assay with the wild type and mutated MFGE8 1 K segments. (A) Trans-activation by TAp63g (a), TAp63a (b), DNp63g (c), DNp63a (d) and p53 (e) in HEK293 cells. (B) We tested 1K-luc, M2-luc and M45-luc for activation by TAp63g (a), TAp63a (b), DNp63g (c) and DNp63a (d) in HeLa cells.

Oncogene Transcriptional activation of MFGE8 by p63 T Okuyama et al 312 a RT-PCR Western blot 20 20 15 15

HeLa HEK293FaDu HSC-1 NHK HeLa HEK293FaDu HSC-1NHK 10 10 TA (34) TA-α ∆N-α 5 5 ∆N (30) TA-γ 0 0 γ α α/β (28) ∆N-γ TA- 8 7 6 4 2 1 0 TA- 8 7 6 4 2 1 0 ∆N-γ 0 1 2 4 6 7 8 ∆N-γ 0 1 2 4 6 7 8 γ (31)

Integrin ITGA3 α (145 kD) 20 20 (26) 3

MFGE8 MFG-E8 -5 15 15 (28) 10 10 rRNA β-actin

(15) RLU x10 5 5 HeLa cells transfected b 0 0 RT-PCR Western blot TA-γ 8 7 6 4 2 1 0 TA-α 8 7 6 4 2 1 0

γ α γ α γ α γ α ∆N-α 0 1 2 4 6 7 8 ∆N-α 0 1 2 4 6 7 8 N- N- N- N- Vector TA- TA- ∆ ∆ M VectorTA- TA- ∆ ∆ FaDu TA-α 10 TA ∆ α 75 N- ∆N TA-γ γ ∆N-γ 5 α /β 31

MFGE8 MFG-E8

β 18S rRNA actin 0 TA-γ 8 764210 12345 6 TA-α 0124678 Figure 4 Endogenous p63 expression in HeLa cells and TA-p63 stabilization by transfected DN-p63. (a) Cell lines were analysed for Figure 5 Enhancement of TA-p63 by dominantly expressed p63 and the target gene expression. Numbers in the parentheses DN-p63. 1K-luc trans-activation assays by combined expression indicate amplification cycles in reverse transcription (RT)–PCR. of TA-p63 with DN-p63. TA:DN or TA:TA plasmid ratio (b) HeLa cells transfected with TAp63g-, TAp63a-, DNp63g- and employed in each experiment (a–e) is shown at the bottom. Filled DNp63a-expression vectors were analysed for p63 and MFGE8 columns represent values (RLU) after subtraction of the back- mRNAs (left) and the proteins (right). ground count obtained by co-transfection with the pRcCMV vector. Hatched columns indicate luciferase activities correspond- ing to 0.1 mg of the TA-p63 expression plasmid, provided that the (Yamaguchi et al., 2000), HSC-1 from a skin squamous DN-p63 isoforms lack an activity to trans-activate this promoter. cell carcinoma and NHK were ‘p63-positive’: the TA, DN, a/b and g mRNA sequences were detected by RT–PCR, and TAp63a (85 kDa), DNp63a (75 kDa), slight increase in TA-a and DN-a also occurred in the TAp63g (57 kDa) and DNp63g (47 kDa) by western DN-g-transfected cells (lane 4). In contrast, neither TA-a blotting. HEK293 cells were negative for p63 mRNAs nor TA-g affected other p63 isoforms (lanes 2 and 3). and proteins: the TA and DN isoform mRNAs were 4- RT–PCR analyses ruled out the possibility that one fold and 32-fold less in HEK293, respectively. Although transfected isoform induced transcription of others. HeLa cells appeared p63-negative by western blotting, a Both the TA and DN transfection increased endogenous level of the TA, a/b and g mRNAs was detectable. The MFGE8 expression 2.5-fold. The protein amount also DN form mRNA became apparent after four additional increased to the level found in FaDu cells. Thus, we PCR cycles. RT–PCR analyses for ITGA3 and MFGE8 found in HeLa (1) HS2/3 was not only responsible for (Kurata et al., 2004) suggested a relation between these the trans-activation by TA-p63, but also effective in the mRNA levels and the p63 status. The MFGE8 mRNA enhanced activation by DN-p63 and (2) when one DN- level in HSC-1 and NHK is lower than that in FaDu, p63 isoform was introduced by transfection other forms but is certainly (more than threefold) increased in including TA-a and TA-g accumulated. We therefore comparison with the level in p63-negative HEK293. speculated that transfected DN-p63 caused the MFGE8 Western blotting results were essentially consistent with promoter activation through endogenously expressed the RT–PCR analyses. We noted, however, a higher TA-p63 in HeLa cells. background of MFG-E8 in HeLa cells. In addition to the constitutive, moderate MFGE8 expression due to Isoform interactions and binding to the target site p63, HeLa has abundant expression of integrin avb5 The TA:DN combination experiment in HEK293 cells (Smith and Giorgio, 2004), a receptor of MFG-E8, indicated that DN-a strongly suppressed TA-g and TA- which may have caused longer retention of this protein. a, when the TA:DN-a ratio was between 7:1 and 4:4 When HeLa cells were transfected with the DN-a (Figures 5b and d), supporting the original character- expression vector, amounts of the DN-g and TA-a ization of DN-a as a potent dominant-negative isoform. proteins, in addition to DN-a, emerged (Figure 4b, lane However, when the DN content was increased to 2:6 and 5). A lower level of TA-g also became detectable. A 1:7, the luciferase activity normalized to 0.1 mg of the TA

Oncogene Transcriptional activation of MFGE8 by p63 T Okuyama et al 313 plasmid amount (b and d, hatched columns) markedly significantly more abundant in the immunoprecipitate increased. When the TA:DN-g ratio was between 7:1 and from the HA-TAp63g-expressing cells (Figure 6b) than 4:4, DN-g did not affect the TA-g activity (a, hatched in the precipitate from the control cells. The ITGA3 columns), but moderately suppressed TA-a (c, hatched intron-1 segment having the p63 target sites (Kurata columns). This is consistent with the initial finding that et al., 2004) also efficiently co-precipitated with HA- DN-g is a moderate suppressor of TA-p63 (Yang et al., TAp63g, whereas the TATA box region of glyceralde- 1998). TA-enhancing effects of DN-g became apparent hyde-3-phosphate dehydrogenase (GAPDH) failed to at 2:6 and 1:7 (a and c). Assays with the TA-a:TA-g (e) do so. Control of nonimmune serum did not efficiently and DN-a:DN-g (data not shown) combinations pro- precipitate the HS2/3 segment (data not shown). duced luciferase activities proportional to the isoform amounts. MFGE8 transcription controlled by p63 in carcinoma When we transfected a suboptimal amount of cells TAp63a and TAp63g in HEK293, endogenous MFGE8 FaDu cells were transfected with small interfering RNA transcription was not altered (Figure 6a, lanes 2 and 6). (p63si) targeting the DNA-binding domain-encoding Upon tetracycline-induced, massive DNp63a produc- sequences of p63. On day 2 and 4 post-transfection, tion, however, the transfected cells increased MFGE8 western blotting indicated suppression of the p63 transcription and the protein synthesis, which were protein level to 20710% (deviation depending on the accompanied by TA-p63 accumulation (lanes 3, 4, 7 and experiments) in comparison with the control siRNA 8). Without the TA-p63 transfection, chromosomal (Csi)-transfected cells (Figure 7a). The p63si-transfected MFGE8 expression remained unaltered (lanes 9 and cells showed 40% (7p10%) decrease in MFGE8 10). Thus, stabilization and activation of TA-p63 mRNA and the protein on day 4. In the immunofluor- concurrently occur by dominantly expressed DNp63a. escent staining, a 70% (7 10%) cell population We next examined p63 for interaction with the HS2/3 p displayed significant decrease in p63 on day 4. In the site on DNA (15q25) by chromatin parallel culture, the cytoplasmic and membrane label immunoprecipitation (ChIP). The generally used anti- with the anti-MFG-E8 antibody decreased in 50% p63 antibody, 4A4, was nonreactive with p63 unless (7 10%) cells. proteins were fully denatured, and therefore unsuitable p Using Saos-2 osteosarcoma cells presenting integrins for ChIP. We produced hemagglutinin (HA)-tagged a b and a b (Koistinen et al., 1999), receptors of TAp63g in the drug-inducible system (Supplementary V 3 V 5 MFG-E8, we tested FaDu culture supernatants for cell- Information 2), and analysed DNA sequences co- binding activity as described by Taylor et al. (1997). precipitated with HA-TAp63g by an anti-HA antibody. Plastic multi-well plates were coated with concentrated PCR analyses indicated that the HS2/3 segment was supernatants from serum-free cultures of FaDu trans- fected with Csi and p63si. In the wells with the Csi- transfected supernatant (12 mg/ml), approximately 40% a 12 3 4 56 7 8 9 10 of the total Saos-2 cell binding was blocked by a tet (∆N-α) - - 4h 7h - - 4h 7h - 7h monoclonal antibody reactive with native form MFG- Transfection - TA-γ TA-γ TA-γ - TA-α TA-α TA-α - - E8, corresponding to MFG-E8-dependent Saos-2 cell TA adhesion (Figure 7b). Total adherent cell count was RT- MFGE8 35% decreased by p63si-transfection, in which the PCR 18S rRNA antibody blocking caused only an 8.3% reduction. Consistent with earlier results (Taylor et al., 1997), p63 western blot analysis of the supernatants detected two Western broad bands at apparent molecular masses of 42–44 kDa blot MFG-E8 (Figure 7c), indicating secretion of MFG-E8 in poorly β-actin glycosylated forms (Oshima et al., 2002; Watanabe et al., 2005). The bands’ intensity was decreased to b MFGE8 ITGA3 GAPDH approximately 40% by the p63 suppression. Thus, we HS2/3 Intron-1 TATA confirmed that p63 influences MFG-E8-mediated cell Input DNA (33) adhesion in vitro. IP(anti-HA) (36)

HEK293 HA- HA- HA- C C C TA-γ TA-γ TA-γ Discussion Figure 6 Activation of endogenous MFGE8 expression by the p63 isoforms. (a) Cooperation of transiently expressed TA-p63 with We determined that MFGE8 is a target gene of p63 in tetracycline-induced DNp63a in HEK293. TAp63g and TAp63a are epithelial cells. Only the TA isoforms were able to marked with filled and open triangles, respectively, and DNp63a by directly enhance the promoter. However, if one DN- arrows. (b) ChIP experiment detecting an interaction between HA- isoform, DNp63a in particular, was overexpressed TAp63g and the HS2/3 site. The MFGE8 enhancer segment covering HS2/3, ITGA3 intron-1 and the glyceraldehyde-3-phos- together with a significantly lower level of the TA phate dehydrogenase (GAPDH) promoter segment spanning the isoforms, the TA proteins were stabilized to induce TATA box were amplified from the immunoprecipitates by PCR. the MFGE8 transcription. The dominant expression of

Oncogene Transcriptional activation of MFGE8 by p63 T Okuyama et al 314 a Csi p63si p63/cyt-c MFG-E8 p63/cyt-c MFG-E8

Day 2 Day 4 p63 (Western blot) (RT-PCR) Csi p63si Csi p63si Day 2 Day 4 MFGE8 Csi p63si Csi p63si TA-α 18S rRNA ∆N-α (Western blot) TA-γ ∆N-γ MFG-E8

β-actin

b 120 c

100 Super- Cell natant 80 lysate 100 1 Csi p63si 60

40

20 Saos-2 cell adhesion (%)

0

Supernatant Csi p63si Csi p63si MFG-E8 secreted protein 12 6 (µg/ml)

anti-MFG-E8 (-) (+) (-) (+) (-) (+) (-) (+)

Figure 7 p63 silencing in FaDu cells by siRNA. (a) Analyses of the siRNA-transfected cells for p63 and milk fat globule-EGF factor 8 protein (MFG-E8). For immunofluorescent staining, p63si- and Csi-transfected cells were recultured in chambers on the same slide. p63 (Alexa Fluor 488, green, left) and MFG-E8 (Alexa Fluor 594, red, right) were detected with mouse monoclonal antibodies, while control stain was with a rabbit anti-cytochrome c antibody (Alexa Fluor 594, left). Reverse transcription (RT)–PCR and western blot analyses of the whole cell lysates are shown. (b) Saos-2 cell attachment to the plates coated with supernatants of the FaDu cell cultures transfected with Csi and p63si. Assays after blocking with an anti-MFG-E8 antibody ( þ ) and control mouse IgG (À) are shown. Binding efficiency (%, 7s.d.) is indicated in relation to the control experiment (100%) with Csi-transfected cells. (c) Detection of secreted MFG-E8 by western blotting. Untransfected cell lysates (100 or 1 mg) and supernatants (10 mg; Csi, p63si) were analysed in the slots. Secreted forms (42–44 kDa) were indicated by bars.

DN-p63 in squamous cell carcinomas may contribute to using the oligomerization domain (Yang et al., 1998; the MFG-E8 abrogation by this mechanism. Kojima et al., 2001). Certainly, DN-a and DN-g were HeLa cells constitutively transcribe TA-p63 as effi- associated with HA-TAp63g (Supplementary Informa- ciently as immunologically p63-positive cells including tion 2), implying formation of mixed dimers and FaDu, HSC-1 and NHK, but fail to maintain the tetramers in p63-expressing keratinocytes. This associa- TA-p63 proteins at a detectable level. The DN type tion may serve as a flexible switch between the activation transcription seems roughly 20-fold less efficient in and suppression status depending on the TA:DN ratio HeLa than in other cells. As we reported earlier, TA-p63 (Figure 5). A recent study identified Itch, an E3 undergoes rapid degradation by 26S proteasome due to ubiquitin ligase, expressed in well-differentiated cell the TA peptides (Osada et al., 2001; Okada et al., 2002). layers of human skin, as a factor controlling the p63 Furthermore, TA-p63 and DN-p63 can form complexes stability (Rossi et al., 2006), indicating a major p63

Oncogene Transcriptional activation of MFGE8 by p63 T Okuyama et al 315 degradation/inactivation mechanism during the keratino- RT–PCR cyte differentiation. In contrast, this study dealt with the Total RNA was isolated with RNAwiz (Ambion, Austin, TX, functional p63 proteins localized to undifferentiated USA) and subjected to semiquantitative RT–PCR with the keratinocytes in the basal layer of stratified epithelia and Platinum Taq One-step RT–PCR kit (Invitrogen, Carlsbad, in carcinomas derived from the stage. CA, USA). Primer sequences are shown in Supplementary Information 3. MFG-E8-L, the mouse-specific long isoform, having a Pro/Thr-enriched domain, facilitates phagocytosis of apoptotic cells by macrophages as demonstrated by Western blotting recent studies with MFGE8-deficient mice (Hanayama We used anti-MFG-E8 (HMFG1, clone EDM45, Lab Vision, et al., 2002; Hanayama and Nagata, 2005). However, Fremont, CA, USA at a dilution of 1:1000), anti-p63 (4A4, the other isoform, MFG-E8-S, lacking the phagocytosis- Santa Cruz Biotechnology Inc., Fremont, CA, USA 1:1000), facilitating activity is more ubiquitously, abundantly anti-b-actin (A5441, Sigma-Aldrich, St Louis, MO, USA, expressed (Watanabe et al., 2005) and expected 1:5000), anti-integrin a3 (Biogenesis, Poole, UK, 1:1000) to facilitate various types of cell–cell interactions. antibodies. EDM45 IgG is specifically, strongly reactive with Importantly, human MFGE8 codes for a protein similar the major form of 46 kDa MFG-E8 identified by an anti- to mouse MFGE8-S, but not an MFG-E8-L-type MFG-E8 monoclonal antibody (MAB2767, R&D Systems, Minneapolis, MN, USA). Alexa Fluor 680 (Molecular Probes, protein. Invitrogen) and IRDye 800CW were used for quantitation For human MFG-E8, an angiogenesis-promoting using the Odyssey Infrared Imaging System (LI-COL/Aloka). activity (Silvestre et al., 2005) and an anticoagulant activity (Shi and Gilbert, 2003) have been reported. Supporting the original characterization (Taylor et al., Cloning of the promoter/enhancer region of MFGE8 1997), we have shown that MFG-E8 released from Genomic DNA was from peripheral blood cells of a healthy FaDu influences Saos-2 cell adhesion. Either or both of donor (YT) with informed consents. The 2K (À1890 to þ 50), 1K ( 877 to 50) and S ( 250 to 50) regions were the aVb3 and aVb5 integrins are expressed in particular À þ À þ lineages including vascular cells, osteoblastomas, retinal amplified by PCR with Ex Taq (Takara Bio, Kyoto, Japan), pigmented epithelial cells and fibroblasts, by which cloned into pGEM-T Easy (Promega, Madison, WI, USA), and sequenced. attachment, migration and invasion are regulated (Gailit et al., 1997; Koistinen et al., 1999; Hoffmann et al., 2005). MFG-E8 might affect these events during Transfection and luciferase assay carcinoma tissue growth and/or wound healing, since Recipient cells were plated 24 h before transfection at a density no substantial defect appeared in the corresponding of 0.4 Â 105 cells in each well on 24-well plates. Typically, a tissues of MFGE8-deficient mice. As reported recently, pRcCMV-based, p63 isoform expression vector (0.1 mg for p63 induces membrane-anchored molecules including each well) was co-transfected with a luc-expression vector the integrin a3 and b4 subunits, envoplakin, Perp and (0.1 mg) with Effectene (Qiagen, Hilden, Germany). At 40 bullous pemphigoid antigen-1/2 (Kurata et al., 2004; Ihrie (HeLa) or 48 h (HEK293) posttransfection, we measured et al., 2005; Osada et al., 2005a, b; Carroll et al., 2006) luciferase activity using the Steady-Glo Luciferase Assay for cell–cell and cell–matrix binding. Soluble protein System (Promega). We used the GeneTailor Site-Directed Mutagenesis System (Invitrogen) to render the ‘G to T’ MFG-E8 may provide a novel mechanism of epithelial– mutations and the SiLentFect Lipid Reagent (Bio-Rad, nonepithelial cell interactions around the p63-expres- Hercules, CA, USA) for siRNA transfection. sing cells. p53 seemed to act on MFGE8 indirectly in the reporter assay. In fact, the p53 pathway is frequently ChIP inactivated in carcinoma cells including FaDu by p53 Tetracycline-induced, HA-tagged TAp63g-expressing HEK293 mutations or by viral oncogene products. In normal cells and uninduced cells were crosslinked with formaldehyde epithelial cells, p53 is silenced by the intrinsic degrada- and subjected to enzymatic chromatin DNA fragmentation tion mechanism with MDM2, unless DNA damage or (ChIP-I Enzymatic Shearing Kit, Active Motif, Carlsbad, CA, other stress signals are induced. Furthermore, associa- USA). HA-p63 and associated materials were immunopreci- tion of p53 with p63 is unlikely (Kojima et al., 2001). pitated with an anti-HA monoclonal antibody (Santa Cruz, 7392 X). Contribution of p53 to the high-level expression of MFGE8 is less probable.

Saos-2 binding assay FaDu cells transfected with Csi and p63si were expanded on day 2. Cells at 80% confluency on day 4 were washed and fed Materials and methods with serum-free KGM-2 medium supplemented with transfer- rin and insulin (Lonza). The supernatants, at 24 h, were clari- Cell culture fied and concentrated with centrifugal filter units (Vivaspin The Health Science Research Resources Bank (Osaka, Japan) 500, Sartorius, Goettingen, Germany). Binding assay was provided us with HeLa, HEK293 and Saos-2 cells. We performed in wells on 24-well plates as described by Taylor obtained FaDu from American Type Culture Collection. et al. (1997). An anti-MFG-E8 monoclonal antibody (2 mg/ml, Primary human neonatal keratinocyte culture was prepared Chemicon, Temecula, CA, USA, CBL421) was used for as described by Kurata et al. (2004). blocking.

Oncogene Transcriptional activation of MFGE8 by p63 T Okuyama et al 316 Acknowledgements High-Tech Research Center Project (RH) and by Strategic Research Fund from University of Yamanashi (IK). We thank This work was supported by Grant-in Aids from MEXT N Shirato, Photography Department, Tokyo Medical and Japan, for Scientific Research on Priority Areas (YI) and for Dental University, for data file making.

References

Brunner HG, Hamel BCJ, van Bokhoven H. (2002). The p63 gene in Kojima T, Ikawa Y, Katoh I. (2001). Analysis of molecular EEC and other syndromes. J Med Genet 39: 377–381. interactions of the p53-family p51(p63) gene products in a yeast Carmon L, Bobilev-Priel I, Brenner B, Bobilev D, Paz A, Bar-Haim E two-hybrid system: homotypic and heterotypic interactions and et al. (2002). Characterization of novel breast carcinoma-associated association with p53-regulatory factors. Biochem Biophys Res BA46-derived peptides in HLA-A2.1/D(b)-beta2m transgenic mice. Commun 281: 1170–1175. J Clin Invest 110: 453–462. Kruger W, Lohner R, Jung R, Kroger N, Zander AR. (2000). Carroll DK, Carroll JS, Leong CO, Cheng F, Brown M, Mills AA Expression of human milk fat globulin proteins in cells of et al. (2006). p63 regulates an adhesion programme and cell survival haemopoietic origin. Br J Cancer 83: 874–879. in epithelial cells. Nat Cell Biol 8: 551–561. Kurata S, Okuyama T, Osada M, Watanabe T, Tomimori Y, Sato S Celli J, Duijf P, Hamel BC, Bamshad M, Kramer B, Smits AP et al. et al. (2004). p51/p63 Controls subunit alpha3 of the major (1999). Heterozygous germline mutations in the p53 homolog p63 epidermis integrin anchoring the stem cells to the niche. J Biol are the cause of EEC syndrome. Cell 99: 143–153. Chem 279: 50069–50077. DiRenzo J, Signoretti S, Nakamura N, Rivera-Gonzalez R, Sellers W, Larocca D, Peterson JA, Urrea R, Kuniyoshi J, Bistrain AM, Ceriani Loda M et al. (2002). Growth factor requirements and basal RL. (1991). A Mr 46,000 human milk fat globule protein that is phenotype of an immortalized mammary epithelial cell line. Cancer highly expressed in human breast tumors contains factor VIII-like Res 62: 89–98. domains. Cancer Res 51: 4994–4998. el-Deiry WS, Kern SE, Pietenpol JA, Kinzler KW, Vogelstein B. Li ER, Owens DM, Djian P, Watt FM. (2000). Expression of (1992). Definition of a consensus binding site for p53. Nat Genet 1: involucrin in normal, hyperproliferative and neoplastic mouse 45–49. keratinocytes. Exp Dermatol 9: 431–438. Ensslin MA, Shur BD. (2003). Identification of mouse sperm SED1, a Liu Y, Chiriva-Internati M, You C, Luo R, You H, Prasad CK et al. bimotif EGF repeat and discoidin-domain protein involved in (2005). Use and specificity of breast cancer antigen/milk protein sperm-egg binding. Cell 114: 405–417. BA46 for generating anti-self-cytotoxic T lymphocytes by recombi- Gailit J, Clarke C, Newman D, Tonnesen MG, Mosesson MW, Clark nant adeno-associated virus-based gene loading of dendritic cells. RAF. (1997). Human fibroblasts bind directly to fibrinogen at RGD Cancer Gene Ther 12: 304–312. sites through integrin [alpha]v[beta]3. Exp Cell Res 232: 118–126. Massion PP, Taflan PM, Jamshedur Rahman SM, Yildiz P, Shyr Y, Hanayama R, Nagata S. (2005). Impaired involution of mammary Edgerton ME et al. (2003). Significance of p63 amplification and glands in the absence of milk fat globule EGF factor 8. Proc Natl Ooerexpression in lung cancer development and prognosis. Cancer Acad Sci USA 102: 16886–16891. Res 63: 7113–7121. Hanayama R, Tanaka M, Miwa K, Shinohara A, Iwamatsu A, Nagata Mather IH, Banghart LR, Lane WS. (1993). The major fat-globule S. (2002). Identification of a factor that links apoptotic cells to membrane proteins, bovine components 15/16 and guinea-pig GP phagocytes. Nature 417: 182–187. 55, are homologous to MGF-E8, a murine glycoprotein containing Hanayama R, Tanaka M, Miyasaka K, Aozasa K, Koike M, epidermal growth factor-like and factor V/VIII-like sequences. Uchiyama Y et al. (2004). Autoimmune disease and impaired Biochem Mol Biol Int 29: 545–554. uptake of apoptotic cells in MFG-E8-deficient mice. Science 304: Mills AA, Zheng B, Wang XJ, Vogel H, Roop DR, Bradley A. (1999). 1147–1150. p63 is a p53 homologue required for limb and epidermal Helton ES, Zhu J, Chen X. (2006). The unique NH2-terminally deleted morphogenesis. Nature 398: 708–713. (DeltaN) residues, the PXXP motif, and the PPXY motif are Okada Y, Osada M, Kurata S, Sato S, Aisaki K, Kageyama Y et al. required for the transcriptional activity of the DeltaN variant of p63. (2002). p53 gene family p51(p63)-encoded, secondary transactivator J Biol Chem 281: 2533–2542. p51B(TAp63alpha) occurs without forming an immunoprecipitable Hibi K, Trink B, Patturajan M, Westra WH, Caballero OL, Hill DE complex with MDM2, but responds to genotoxic stress by et al. (2000). AIS is an oncogene amplified in squamous cell accumulation. Exp Cell Res 276: 194–200. carcinoma. Proc Natl Acad Sci 97: 5462–5467. Osada M, Inaba R, Shinohara H, Hagiwara M, Nakamura M, Ikawa Hoffmann S, He S, Jin M, Ehren M, Wiedemann P, Ryan S et al. Y. (2001). Regulatory domain of protein stability of human (2005). A selective cyclic integrin antagonist blocks the integrin P51/TAP63, a P53 homologue. Biochem Biophys Res Commun 283: receptors alphavbeta3 and alphavbeta5 and inhibits retinal pigment 1135–1141. epithelium cell attachment, migration and invasion. BMC Ophthal- Osada M, Nagakawa Y, Park HL, Yamashita K, Wu G, Kim MS et al. mol 5: 16. (2005a). p63-specific activation of the BPAG-1e promoter. J Invest Hvarregaard J, Andersen MH, Berglund L, Rasmussen JT, Petersen Dermatol 125: 52–60. TE. (1996). Characterization of glycoprotein PAS-6/7 from mem- Osada M, Ohba M, Kawahara C, Ishioka C, Kanamaru R, Katoh I branes of bovine milk fat globules. Eur J Biochem 240: 628–636. et al. (1998). Cloning and functional analysis of human p51, Ihrie RA, Marques MR, Nguyen BT, Horner JS, Papazoglu C, which structurally and functionally resembles p53. Nat Med 4: Bronson RT et al. (2005). Perp is a p63-regulated gene essential for 839–843. epithelial integrity. Cell 120: 843–856. Osada M, Park HL, Nagakawa Y, Yamashita K, Fomenkov A, Kim Koga F, Kawakami S, Kumagai J, Takizawa T, Ando N, Arai G et al. MS et al. (2005b). Differential recognition of response elements (2003). Impaired delta Np63 expression associates with reduced determines target gene specificity for p53 and p63. Mol Cell Biol 25: beta-catenin and aggressive phenotypes of urothelial neoplasms. Br 6077–6089. J Cancer 88: 740–747. Oshima K, Aoki N, Kato T, Kitajima K, Matsuda T. (2002). Secretion Koistinen P, Pulli T, Uitto V-J, Nissinen L, Hyypia T, Heino J. (1999). of a peripheral membrane protein, MFG-E8, as a complex with Depletion of [alpha]V integrins from osteosarcoma cells by membrane vesicles. Eur J Biochem 269: 1209–1218. intracellular antibody expression induces bone differentiation Pellegrini G, Dellambra E, Golisano O, Martinelli E, Fantozzi I, marker genes and suppresses gelatinase (MMP-2) synthesis. Matrix Bondanza S et al. (2001). p63 identifies keratinocyte stem cells. Proc Biol 18: 239–251. Natl Acad Sci USA 98: 3156–3161.

Oncogene Transcriptional activation of MFGE8 by p63 T Okuyama et al 317 Peterson JA, Zava DT, Duwe AK, Blank EW, Battifora H, Ceriani Taylor MR, Couto JR, Scallan CD, Ceriani RL, Peterson JA. (1997). RL. (1990). Biochemical and histological characterization of Lactadherin (formerly BA46), a membrane-associated glycoprotein antigens preferentially expressed on the surface and cytoplasm of expressed in human milk and breast carcinomas, promotes Arg-Gly- breast carcinoma cells identified by monoclonal antibodies against Asp (RGD)-dependent cell adhesion. DNA Cell Biol 16: 861–869. the human milk fat globule. Hybridoma 9: 221–235. Watanabe T, Totsuka R, Miyatani S, Kurata S, Sato S, Katoh I et al. Quade BJ, Yang A, Wang Y, Sun D, Park J, Sheets EE et al. (2001). (2005). Production of the long and short forms of MFG-E8 by Expression of the p53 homologue p63 in early cervical neoplasia. epidermal keratinocytes. Cell Tissue Res 321: 185–193. Gynecol Oncol 80: 24–29. Wu G, Osada M, Guo Z, Fomenkov A, Begum S, Zhao M et al. Rossi M, Aqeilan RI, Neale M, Candi E, Salomoni P, Knight RA et al. (2005). DeltaNp63alpha up-regulates the Hsp70 gene in human (2006). The E3 ubiquitin ligase Itch controls the protein stability of cancer. Cancer Res 65: 758–766. p63. Proc Natl Acad Sci USA 103: 12753–12758. Yamaguchi K, Wu L, Caballero OL, Hibi K, Trink B, Resto V et al. Shi J, Gibert GE. (2003). Lactadherin inhibits enzyme complexes of (2000). Frequent gain of the p40/p51/p63 gene locus in primary head blood coagulation by completing for phospholipid-binding sites. and neck squamous cell carcinoma. Int J Cancer 86: 684–689. Blood 101: 2628–2636. Yang A, Kaghad M, Wang Y, Gillett E, Fleming MD, Dotsch V et al. Silvestre JS, Thery C, Hamard G, Boddaert J, Aguilar B, Delcayre A (1998). p63, a p53 homolog at 3q27-29, encodes multiple products et al. (2005). Lactadherin promotes VEGF-dependent neovascular- with transactivating, death-inducing, and dominant-negative ization. Nat Med 11: 499–506. activities. Mol Cell 2: 305–316. Smith RA, Giorgio TD. (2004). Quantitation and kinetics of CD51 Yang A, Schweitzer R, Sun D, Kaghad M, Walker N, Bronson RT surface receptor expression: implications for targeted delivery. Ann et al. (1999). p63 is essential for regenerative proliferation in limb, Biomed Eng 32: 635–644. craniofacial and epithelial development. Nature 398: 714–718.

Supplementary Information accompanies the paper on the Oncogene web site (http://www.nature.com/onc).

Oncogene