|Home | About | Journals | Submit | Contact Us | Français|
CUL7, a recently identified member of the cullin family of E3 ubiquitin ligases, forms a unique SCF-like complex and is required for mouse embryonic development. To further investigate CUL7 function, we sought to identify CUL7 binding proteins. The p53-associated, parkin-like cytoplasmic protein (PARC), a homolog of CUL7, was identified as a CUL7-interacting protein by mass spectrometry. The heterodimerization of PARC and CUL7, as well as homodimerization of PARC and CUL7, was confirmed in vivo. To determine the biological role of PARC by itself and in conjunction with CUL7, a targeted deletion of Parc was created in the mouse. In contrast to the neonatal lethality of the Cul7 knockout mice, Parc knockout mice were born at the expected Mendelian ratios and exhibited no apparent phenotype. Additionally, Parc deletion did not appear to affect the stability or function of p53. These results suggest that PARC and CUL7 form an endogenous complex and that PARC and CUL7 functions are at least partially nonoverlapping. In addition, although PARC and p53 form a complex, the absence of effect of Parc deletion on p53 stability, localization, and function suggests that p53 binding to PARC may serve to control PARC function.
The ubiquitin-proteasome system plays an important role in controlling diverse biological processes, ranging from signal transduction to cell cycle control. The covalent addition of ubiquitin to proteins is the result of a sequential cascade in which ubiquitin is activated by the E1 ubiquitin-activating enzyme, transferred to an E2 ubiquitin-conjugating enzyme, and transferred to a lysine residue on the substrate by an E3 ubiquitin ligase (10). Reiteration of this cascade leads to the formation of a polyubiquitin chain, targeting the substrate to the proteasome for degradation.
The substrate specificity, and thus ultimate control of the ubiquitination reaction, is determined by the E3 ubiquitin ligase. Many E3 ligase complexes have been identified, and they are generally grouped into two categories based on the presence of either a HECT domain protein or a RING protein in the complex. The canonical SCF (SKP1/CUL1/F-box) complex is a prototype RING-type E3 ligase, with the core cullin scaffold recruiting the small RING protein RBX1 in addition to SKP1 and a variable F-box (3, 5). In this complex, RBX1 recruits the E2 enzyme and SKP1 acts as a bridging factor to various F-box proteins that dictate substrate specificity for the entire complex (11). Within the context of the CUL1 scaffold, F-boxes, such as β-TRCP1, SKP2, and FBW7, target IκBα, p27, and cyclin E, respectively, for ubiquitination and subsequent degradation by the proteasome (3).
Database searches reveal eight members of the mammalian cullin family, CUL1, CUL2, CUL3, CUL4A, CUL4B, CUL5, CUL7, and PARC. With the exception of PARC, all of these cullins have previously been shown to form complexes similar to the CUL1-based SCF complex (3). Although all of these complexes contain RBX1 and are modified by NEDD8, the CUL2-, CUL3-, CUL4A-, CUL4B-, and CUL5-based complexes contain analogs of the SKP1 bridging factor and non-F-box substrate specificity factors. The CUL2 and CUL5-based complexes replace SKP1 with elongin B and elongin C, and they target substrates, such as hypoxia-inducible factor 1α and APOBEC3G, for ubiquitination through SOCS box proteins, such as VHL and the human immunodeficiency virus Vif protein (22, 34). CUL4-based complexes utilize DDB1 and DDB2 to bridge to specificity factors such as hDET1, which has been reported to target c-JUN for ubiquitination, and CUL3-based complexes use BTB proteins as both bridging proteins and substrate specificity factors (7, 8, 23, 25, 31, 33).
Although CUL7-based complexes are very similar to CUL1-based complexes based on RBX1, SKP1, and F-box binding, recent reports suggest that these two complexes bind to different accessory proteins and distinct subsets of F-box proteins (1, 6). Glomulin (FAP68) is unique to the CUL7 complex, and although overexpressed FBXW8 (FBW6, FBX29) can bind both CUL7 and CUL1, it binds more strongly to CUL7, with coexpression of SKP1 and FBXW8 increasing FBXW8 binding to CUL7 but not CUL1 (1). In addition, targeted deletion of Cul7 results in a loss of FBXW8 stability, further supporting the specificity of FBXW8 for CUL7 (1). Finally, CUL7 contains multiple domains outside of the cullin homology domain that suggest that CUL7 can form unique SCF-like complexes (1, 6).
Although the cullin complexes are diverse in structure and substrates, targeted deletions of cullins 1, 3, 4A, and 7 in the mouse have demonstrated each complex's importance in development. Deletion of Cul1 results in embryonic lethality by embryonic day 5.5, and deletions of Cul3 or Cul4a result in embryonic lethality before embryonic day 7.5 (4, 17, 26, 30). Despite displaying a milder phenotype than the other cullin knockouts, Cul7 knockout mice exhibit neonatal lethality with reduced size and possible vascular defects (1). Presumably the defects in the development of Cul1, Cul3, Cul4a, and Cul7 knockout mice result from the inability of the deleted cullins to degrade their normal substrates.
The least-characterized mammalian cullin was recently identified as a PARkin-like, Cytoplasmic, p53 binding protein (PARC) (20). Although it was suggested that PARC could serve as an E3 ubiquitin ligase, no ubiquitination activity towards p53 was demonstrated. Instead, it was proposed that PARC functioned to sequester p53 in the cytoplasm, preventing p53 from inducing growth arrest and apoptosis (20). PARC is highly homologous to CUL7 throughout its sequence, sharing up to 60% identity in several domains, although PARC differs from CUL7 in the presence of a “spacer domain” of unknown function and a C-terminal RING-IBR-RING domain. Like CUL7, PARC has several discrete domains present in other proteins found in E3 ligase complexes, such as HERC2, APC10/DOC1, and Parkin (1, 6, 9, 16, 29). Further sequence analysis suggests that PARC is the evolutionary precursor of CUL7, since only PARC, and not CUL7, sequences are present in the Danio rerio and Takifugu rubripes databases.
Here we show that CUL7 and PARC can each form homodimers, as well as forming heterodimers with each other, and that a complex containing PARC can bind p53. Furthermore, through the targeted deletion of Parc and the previously published deletion of Cul7 in the mouse, we demonstrate that the PARC complex is not required for mouse viability, and CUL7 and PARC functions are at least partially nonoverlapping. Additionally, Parc deletion has no discernible effect on p53 stability, on the induction of p53 target genes, or on p53 localization.
A partial PARC cDNA, KIAA0708, was obtained from the Kazusa DNA Research Institute. The sequence of the missing 5′ region of PARC was determined by two cycles of 5′ rapid amplification of cDNA ends, and the resulting sequence information was used to generate a full-length cDNA of PARC by PCR from a fetal brain cDNA library. The full-length cDNA was cloned into pcDNA3.0 HA using BamHI and NotI, and mutations introduced by PCR were corrected by replacing sequences with expressed sequence tags, KIAA0708 and IMAGE no. 1351950. An intron present in IMAGE no. 1351950 was removed using a fragment from the full-length PARC generated by PCR.
The short hairpin RNA (shRNA) vector pTU6IIa(−) was generated from pBS KS. A U6 promoter containing a TetO element overlapping the TATA box and containing an AgeI site at the initiating site of the U6 RNA was generated by PCR (21). Tet-responsive U6 was cloned into pBS KS using EcoRI and XhoI. A blasticidin resistance cassette driven by the simian virus 40 (SV40) promoter was inserted nondirectionally into the resulting plasmid at the SacII site. pTU6IIa(−) contains this cassette in the opposite orientation from the Tet-U6 promoter. shRNA constructs targeting PARC were cloned from annealed oligonucleotides into pTU6IIa(−) using the AgeI and XhoI sites. The shRNA used in this paper corresponds to GGAGTGTCCTGAAGTGTGGGGAAGTGGGCTATGCCCACTTCCCCACACTTCAGGACACTCC.
U-2 OS cells stably expressing an shRNA directed against PARC were constructed by cotransfection of pTU6IIa(−) sh281#2 and pcDNA3.0 TetR. Stable clones were selected in 10 μg/ml blasticidin (Invitrogen) and 1 mg/ml G418 (Cellgro). Multiple clones tested positive for PARC depletion, but no clones responded to tetracycline. TetR expression was also undetectable.
Mouse embryo fibroblasts (MEFs) were generated from embryos isolated at embryonic day 14.5. To generate polyclonal MEFs expressing SV40 large T antigen, cells were transduced with the pLB(N)CX-T retroviral vector and selected in 10 μg/ml blasticidin (Invitrogen). For irradiation experiments, fully confluent plates of MEFs were irradiated with 20 Gy from a 137Cs source and harvested at the indicated time points. All transfections were performed using Lipofectamine Plus (Invitrogen). All cells were cultured in Dulbecco's modified Eagle's medium containing 10% Fetalclone serum (HyClone).
The rabbit anti-PARC antibodies BL616, BL617, and BL920, which recognize epitopes within the N-terminal 50 residues, C-terminal 50 residues, and residues between 700 and 750 amino acids, respectively, were from Bethyl Laboratories (Montgomery, TX). Affinity-purified anti-CUL7 (h1557, BL653, and BL1541) antibodies to epitopes mapping between amino acids 1658 to 1676, 1500 to 1550, and 1648 to 1698, respectively, were also produced by Bethyl Laboratories. Additional monoclonal antibodies used for immunoprecipitation and Western blotting include anti-hemagglutinin (HA) (12CA5 and HA-11; Covance), anti-T7 (Novagen), anti-Myc (9B11; Cell Signaling Technology), anti-p53 (Ab-4/Pab122; Neomarkers), anti-MDM2 (2A10; Calbiochem), anti-p21 (HZ52/Ab-5; Neomarkers), and anti-Vinculin (hVin1; Sigma). Rabbit polyclonal anti-Lamin A (Cell Signaling Technology) and anti-PTEN (Cell Signaling Technology) were used for Western blotting. Purified, nonspecific rabbit immunoglobulin G (IgG) (Bethyl) was used as a control for immunoprecipitations. For indirect immunofluorescence microscopy, a mixture of four monoclonal antibodies to p53 (Pab122/Ab-4 [Neomarkers], DO-7/Ab-5 [Neomarkers], Pab240/Ab-1 [Neomarkers], and Pab1620/Ab-5 [Calbiochem]) was used in conjunction with rhodamine-conjugated goat anti-mouse secondary antibody (Molecular Probes).
U-2 OS cells or MEFs were washed twice in phosphate-buffered saline before lysis in either NET-N (20 mM Tris, pH 8, 100 mM NaCl, 1 mM EDTA, 0.5% NP-40) or EBC (50 mM Tris, pH 8, 120 mM NaCl, 0.5% NP-40) containing a 1× concentration of protease inhibitor cocktail set I (Calbiochem). Extracts were cleared by centrifugation at 12,000 × g for 10 min, and protein concentrations were determined by Bradford Assay (Bio-Rad). Nuclear and cytoplasmic extracts were generated using NE-PER (Pierce).
For immunoprecipitation, all antibodies were used with protein A Sepharose beads, and rabbit polyclonal antibodies were used at a concentration of 2 μg per 1 mg of total protein. Immunoprecipitations were performed for 3 h or more at 4°C, and the subsequent immunoprecipitates were washed three times with NET-N or EBC before boiling in sodium dodecyl sulfate (SDS) sample buffer. Whole-cell lysates and immunoprecipitates were separated by SDS-polyacrylamide gel electrophoresis (PAGE) and transferred to nitrocellulose membranes (Bio-Rad) for Western blotting. Membranes were blocked in 5% milk-Tween 20-Tris-buffered saline before Western blotting was performed. Detection of proteins was accomplished using the appropriate secondary antibodies conjugated to horseradish peroxidase (Pierce) at a 1:5,000 dilution in 5% milk-Tween 20-Tris-buffered saline. Western blots were developed using enhanced chemiluminescence (Super Signal; Pierce).
In the experiments leading to the identification of human PARC (hPARC) as a CUL7-interacting protein, A2P2 cells were starved in methionine- and cysteine-free medium for 1 h prior to addition of [35S]methionine at a concentration of 1 mCi per ml (1). Cells were labeled for 3 h before harvesting for immunoprecipitation using HA-11 antibody (Covance). The large-scale preparation used to identify PARC by mass spectrometry is described elsewhere (1). Mass spectrometry yielded the peptide sequence LQQETQPFLLLLR, corresponding to hPARC amino acids 988 to 1000.
A 500-bp probe from coding exon 1 of the PARC genomic sequence was generated from PCR of 129/C57BL/6 mouse genomic DNA using the PCR primers GAACGGCGTGCAGGGAACCTTCTGGTACCC and CCAGAGCCTGGAGCATCTTGCCTGCACTCC. This probe was used to screen a mouse 129/Sv/En/Tac genomic bacterial artificial chromosome (BAC) library, and four BACs containing PARC genomic DNA were identified (ResGen). PCR from the BAC templates was used to generate a 2.3-kb EcoRI-XhoI fragment containing coding exons 2 and 3, a 1.9-kb SacII-SmaI fragment containing coding exons 4 to 6 with the 3′ loxP site, and a 4.5-kb XmaI-SnaBI fragment containing coding exon 7 for cloning into the pGEM-T Easy vector (Promega). The EcoRI, XhoI, SacII, SmaI/XmaI, and 3′ loxP sites were introduced by PCR (LA Taq; Takara), and the primer sequences used in the amplification of these fragments are available upon request. The sequence of all coding exons was determined in pGEM-T Easy before the fragments were subcloned into the targeting vector pKOII (2). The fragments containing exons 2 to 3 and exons 4 to 6 were cloned with the restriction sites introduced by PCR as indicated above. The 4.5-kb fragment was cloned using the XmaI site introduced by PCR and a NotI site introduced from pGEM-T Easy. This NotI site was used to linearize the targeting vector prior to electroporation. The sequence of all exons contained in the targeting vector was reconfirmed before electroporation of embryonic stem (ES) cells and selection with G418. The correct integration of the targeting vector was confirmed in one ES cell clone by Southern blotting, PCR, and sequencing across the 3′ loxP site of the targeted allele. The PCR primers used in screening were CACAGTCAGCTGCTATGAAGCAGCCCTGTA and CTGCTAAAGCGCATGCTCCAGACTGCCTTG. To generate chimeras, correctly targeted ES cells were injected into C57BL/6 blastocysts and transferred into pseudopregnant foster mothers. The chimeric mice were backcrossed with C57BL/6 mice to demonstrate germ line transmission or with C57BL/6 mice carrying an EIIA-Cre transgene to confirm germ line transmission and generate the null allele (15). Parc+/− EIIA-Cre+/− mice were backcrossed to C57BL/6 mice to remove the Cre transgene and backcrossed once more to the C57BL/6 background before subsequent mating to generate homozygous null animals. Genotyping for EIIA-Cre was performed using the PCR primers ATGTCCAATTTACTGACCGTACACCAAAAT and ACGATGAAGCATGTTTAGCTGGCCCAAATG. Subsequent genotyping for PARC was performed by Southern blotting using the probes above and by PCR using the primers ACTGAAGAACAGGTGGCTGTCAGTTAGACA, GGAGGCAGCTACTGGTGAGTTTGGTTCTGG, and GGTAGCATGGCTTTAAGCCTGGCTCTTAAG. The DNA for genotyping was generated from tails using Purgene kits (Gentra Systems). The deleted allele is officially designated Parctm1Jdec.
For immunofluorescence experiments, cells were plated onto coverslips 24 h prior to staining. MG132-treated cells were treated with 25 μM MG132 6 h prior to processing. Cells were fixed in 10% formalin-2% sodium acetate and permeabilized in 0.5% Triton X-100-3% goat serum-PBS for staining with the indicated antibodies. For anti-BrdU staining, cells were irradiated with 7.5 Gy and labeled with BrdU for 1.5 h at 14 h postirradiation. After BrdU labeling, cells were fixed in 70% ethanol prior to denaturation of DNA in 2N HCl. Blocking was performed with 3% bovine serum albumin in 0.1 M Na2B4O7, pH 8.5, before incubation with the anti-BrdU-fluorescein isothiocyanate (FITC) conjugate or an isotype control-FITC conjugate (BD Pharmingen). Subsequently, cells were stained with propidium iodide and analyzed by fluorescence-activated cell sorting (FACS).
The previously isolated CUL7 complex contained proteins consistent with those in other SCF-like complexes, including SKP1, RBX1, and FBXW8 (1, 6). However, the presence of glomulin in this complex and the existence of multiple conserved domains outside of the cullin homology domain of CUL7 suggested that additional proteins might be part of the complete CUL7 complex (Fig. (Fig.1B)1B) (1). To identify CUL7 binding proteins, U-2 OS cells stably expressing HA-tagged human CUL7 (HA-hCUL7) (A2P2 cells) were metabolically labeled with [35S]methionine, treated with increasing amounts of the proteasome inhibitor MG132, and immunoprecipitated with anti-HA antibody (1). In addition to a 58-kDa band that contains FBXW8 and glomulin, the HA immunoprecipitation for HA-hCUL7 specifically coimmunoprecipitated a protein of approximately 280 kDa (Fig. (Fig.1A).1A). A previously described preparative scale immunoprecipitation from A2P2 lysates generated sufficient amounts of the 280-kDa protein to allow identification by mass spectrometry, which yielded one peptide (LQQETQPFLLLLR) corresponding to KIAA0708, later named hPARC (1).
Initial database searches using the PARC sequence revealed an extremely high homology to CUL7. PARC and CUL7 are homologous over their entire length, sharing approximately 60% identity and 70% homology over their shared domains (Fig. (Fig.1B).1B). Notably, PARC differs from CUL7 in the presence of a “spacer” domain of unknown function and a C-terminal RING-IBR-RING. In addition to exhibiting homology to each other, PARC and CUL7 share regions of homology with other proteins in the database, including HERC2, APC10DOC1, and the cullins (Fig. (Fig.1B)1B) (1, 6, 9, 16). Notably, all of these proteins are linked with E3 ubiquitin ligase function, suggesting that PARC and CUL7 may both function as E3 ubiquitin ligases. Although multiple proteins have been noted to contain combinations of these domains in tandem, HERC2, PARC, and CUL7 are unique in the presence of three or more such domains (9).
Subsequent searches of the genomes of lower organisms indicated that PARC is most likely the evolutionary precursor of CUL7. PARC first appears in the genomes of D. rerio, Xenopus tropicalis, and T. rubripes, while CUL7 appears only in mammalian genomes, suggesting that PARC is the evolutionary precursor for CUL7. BLAST searches of the D. rerio expressed-sequence-tag and genomic databases with human PARC revealed extensive conservation throughout the entire PARC primary sequence. These D. rerio sequences can clearly be distinguished from that of CUL7 due to characteristic insertions in the PARC N terminus (i.e., residues 96 to 102) that are not present in CUL7 and additional PARC-specific sequences, such as those of the “spacer” and RING-IBR-RING domains, that are not found in CUL7. The two genes are arranged in a head-to-head fashion on human chromosome 6 and mouse chromosome 17, and they are separated by approximately 130 kb of genomic DNA, containing several confirmed genes, including Srf, Ptk, and Knsl8 (Fig. (Fig.1C1C).
The apparent dimerization of hCUL7 with its close homolog hPARC suggested that hCUL7 could also form homodimers. To confirm the heterodimerization of hCUL7 and hPARC and to look for homodimerization of hCUL7, HA-tagged hCUL7 and HA-tagged hPARC were cotransfected with T7-tagged hCUL7 for analysis by coimmunoprecipitation and Western blotting using anti-T7 and anti-HA antibodies. Immunoprecipitations for T7-hCUL7 coprecipitated both HA-hCUL7 and HA-hPARC, while the reciprocal immunoprecipitations for HA-hCUL7 and HA-hPARC precipitated T7-hCUL7, confirming the formation of CUL7-PARC heterodimers and CUL7-CUL7 dimers (Fig. (Fig.2A).2A). The homodimerization of hCUL7 and the extensive homology between CUL7 and PARC suggested that PARC could also form homodimers. To test for hPARC homodimerization, cells were transfected with HA-tagged hPARC and Myc-tagged hPARC for analysis by coimmunoprecipitation and Western blotting. Reciprocal immunoprecipitations for Myc-hPARC or HA-hPARC coprecipitated HA-hPARC and Myc-hPARC, respectively, confirming homodimerization of PARC (Fig. (Fig.2B2B).
To confirm that the interaction between CUL7 and PARC was not the result of the transient overexpression of either protein, polyclonal, monospecific antibodies were generated against hCUL7 and hPARC to test the interaction of endogenous protein by coimmunoprecipitation and Western blotting. Immunoprecipitation of endogenous CUL7 from U-2 OS cells revealed coprecipitation of PARC, and immunoprecipitation of PARC using antibodies generated to two distinct epitopes resulted in the coprecipitation of CUL7, confirming the dimerization of endogenous CUL7 and PARC (Fig. (Fig.2C).2C). The specificity of the anti-CUL7 and anti-PARC antibodies was demonstrated by the inability of purified rabbit IgG to immunoprecipitate either protein. Additionally, a U-2 OS cell line stably expressing an shRNA directed against PARC demonstrates the specificity of the anti-PARC BL617 antibody for Western blotting. The presence of an endogenous CUL7-PARC complex was also confirmed in MEFs by immunoprecipitation with antibodies specific for each protein followed by Western blotting (Fig. (Fig.2D).2D). In MEF whole-cell lysates and immunoprecipitations, a nonspecific, high-molecular-weight band appears above PARC upon Western blotting. This band was unaffected by Parc knockout (see below), and it was immunoprecipitated nonspecifically by rabbit IgG (Fig. (Fig.2D)2D) or any other antibody tested (see below). Notably, human PARC and mouse PARC show no size difference in gel migration by SDS-PAGE, so lysates from the U-2 OS and U-2 OS short hairpin PARC (shPARC) cells were included in subsequent experiments as a reference standard for mouse PARC (see below).
To understand the function of the CUL7-PARC complex and the potential PARC-PARC complex, we generated a targeted deletion of Parc in the mouse model. Previously we reported that mice with a targeted deletion of Cul7 are inviable, exhibiting neonatal lethality with reduced size and vascular defects (1). Because all previous cullin knockouts resulted in embryonic or neonatal lethality, the Parc gene-targeting vector was designed to conditionally delete coding exons 4 to 6 when the Cre recombinase is expressed, causing exon 3 to splice into exon 7 where the message is out of frame with multiple stops (Fig. (Fig.3A3A).
Mouse ES cells were electroporated with the linearized construct and selected for G418 resistance. Correct integration of the targeting construct was confirmed by Southern blotting using a probe located 5′ to the short arm of the targeting construct and a PCR screen (data not shown). The probe used for ES cell screening was subsequently used for all genotyping Southern blots. Correct integration of the 5′ end of the targeting construct was confirmed in two ES cell clones, and the presence of the 3′ loxP site was confirmed in one clone by sequencing. Chimeric mice generated from injection of this ES cell clone into a C57BL/6 blastocyst were mated to C57BL/6 mice to demonstrate germ line transmission or to C57BL/6 mice carrying an EIIA-Cre transgene to demonstrate germ line transmission while generating the null allele (Fig. (Fig.3B)3B) (15). Parc+/− EIIA-Cre mice were backcrossed to C57BL/6 to remove the EIIA-Cre, and these Parc+/− mice were backcrossed again to the C57BL/6 background before breeding heterozygous mice to generate Parc−/− mice.
Mating of Parc+/− mice generated Parc−/− mice, as shown by Southern blotting and PCR of tail DNA from 3-week-old mice (Fig. 3C and D). The number of Parc−/− mice obtained was not significantly different from the expected Mendelian ratio (15 Parc−/− mice, 20 Parc+/− mice, and 12 Parc+/+ mice; χ2 = 1.246; P = 0.4903). Parc−/− mice exhibit no physical deformities or growth defects, and they appeared phenotypically normal when compared to wild type or heterozygous littermates. No health problems have emerged as these mice have aged to more than 400 days. Additionally, both male and female Parc−/− mice are fertile, since crosses involving knockouts yielded consistently normal litter sizes.
To confirm that Parc−/− mice were indeed null for PARC, MEFs were generated from 14.5-day embryos from Parc+/− × Parc+/− crosses. The genotypes of the MEFs were confirmed by Southern blotting and PCR (Fig. (Fig.4A;4A; also data not shown). Parc+/− and Parc−/− MEFs displayed no apparent growth defects in culture when compared to MEFs from wild-type littermates. Matched littermate MEFs were used for subsequent Western blotting for the PARC protein. Western blots of lysates prepared from Parc+/+, Parc+/−, and Parc−/− MEFs demonstrated that a null allele of Parc was generated, and Parc−/− tissues do not contain PARC protein (Fig. (Fig.4B).4B). The absence of PARC protein from Parc−/− MEFs was further confirmed by the absence of PARC from anti-PARC immunoprecipitates using antibodies against three distinct epitopes in the N terminus, C terminus, and an internal region, as well as the failure of anti-CUL7 antibody to coimmunoprecipitate PARC (Fig. (Fig.4C).4C). The previously noted, nonspecific, high-molecular-weight band migrating slower than PARC in mouse cells was present in Parc−/− MEFs, and it could be distinguished from PARC by the inclusion of lysates from U-2 OS and U-2 OS shPARC cells as a size standard. Western blotting of whole-cell extracts and anti-PARC immunoprecipitations also revealed that heterozygous MEFs contained a reduced level of PARC relative to wild-type MEFs. Western blotting for CUL7 from lysates of Parc−/− MEFs showed a slight decrease in the expression of CUL7 (Fig. (Fig.4B4B).
PARC was reported to bind p53 in a cytoplasmic complex in human cells (20). To confirm binding of PARC to p53 in mouse cells, we performed immunoprecipitations for PARC from wild-type, Parc−/−, and Trp53−/− MEFs. Human CUL7 also binds p53 (Kasper et al., unpublished data), so PARC immunoprecipitations were also performed from Cul7−/− MEF lysates to determine if PARC binding to p53 was dependent on CUL7. Western blotting of PARC immunoprecipitates from wild-type and Cul7−/− MEFs revealed that PARC bound p53, and this binding was independent of CUL7 (Fig. (Fig.5A).5A). Notably, Trp53−/− and Cul7−/− MEFs expressed lower levels of PARC than wild-type MEFs (Fig. (Fig.4B4B and Fig. Fig.5A).5A). The level of PARC expression in these MEFs was less than the levels found in Parc+/− MEFs (Fig. (Fig.4B4B).
The observation that p53 binds to PARC raised the possibility that p53 could serve as a substrate for the potential E3 ligase activity of PARC. To address this possibility, the effect of Parc deletion on p53 stability was examined. First, the steady-state levels of p53 in primary Parc+/+, Parc+/−, and Parc−/− MEFs were analyzed by Western blotting. In early-passage +/+, +/−, and −/− MEFs, p53 was nearly undetectable due to its short half-life (Fig. (Fig.5B).5B). Treatment of these MEFs with the proteasome inhibitor MG132 led to increased levels of p53. Similarly, expression of SV40 large T antigen in these MEFs led to the stabilization and increased expression of p53 in MEFs of all genotypes. In mock-treated, MG132-treated, and T-antigen-expressing MEFs, PARC deletion appeared to have no detectable effect on the steady-state levels of p53 (Fig. (Fig.5B).5B). In addition, Parc−/− MEFs did not show any changes in the steady-state expression levels of p53 downstream targets, such as MDM2 and p21, compared to +/+ or +/− MEFs.
While no differences were found in the steady-state levels of p53, PARC might control the levels of p53 upon exposure to genotoxic stresses. To test the p53 response to γ-irradiation, wild-type or Parc−/− MEFs were irradiated with 20 Gy and harvested at the indicated time points (Fig. (Fig.5C).5C). Wild-type MEFs displayed a characteristic increase in p53 levels beginning at around 45 min after γ-irradiation, followed by a decrease at 180 min. The p53 response in Parc−/− MEFs was indistinguishable from that observed in wild-type MEFs (Fig. (Fig.5C).5C). The p53 response of wild-type and Parc−/− MEFs was further confirmed using an antibody to the radiation-induced serine 15 phosphorylation of p53. Additionally, p53 was able to induce increased MDM2 expression by 90 min in both wild-type and Parc−/− MEFs, suggesting that the transactivation function of p53 was normal in Parc−/− MEFs.
Although p53 function appeared intact by the induction of MDM2, we wished to further confirm that the overall biological function of p53 was unaffected by Parc deletion. In response to γ-irradiation, MEFs arrest in G1 phase in a p53-dependent manner. To test the ability of p53 to induce this G1 arrest in Parc-deficient cells, early-passage wild-type, Parc−/− and Trp53−/− MEFs were irradiated and labeled with BrdU 14 h postirradiation before analysis by FACS (Fig. (Fig.6).6). As measured by anti-BrdU and propidium iodide staining, wild-type MEFs showed a decrease in S-phase cells (−10.9%) in response to irradiation, while Trp53−/− MEFs showed no decrease (−0.9%) in S-phase cells in response to irradiation. Despite being early-passage MEFs, the Trp53-deficient MEFs showed a significant increase in 4N and 8N cells. Similarly to wild-type MEFs, Parc−/− MEFs displayed an 11.8% decrease in S phase in response to irradiation, indicating normal p53 function in these cells.
Previously it was suggested that PARC functions as a cytoplasmic anchor for p53 and that depletion of PARC resulted in nuclear accumulation of p53 and induction of apoptosis (20). The ability to generate viable Parc−/− mice suggested that PARC was not essential to inhibition of p53's apoptotic functions. To examine the effect of Parc deletion on the cellular localization of p53, nuclear and cytoplasmic extracts were generated from wild-type and Parc−/− MEFs (Fig. (Fig.7A).7A). In wild-type and Parc−/− MEFs, p53 was found in barely detectable amounts in the nuclear compartment. MG132 treatment of cells prior to extraction increased the amount of nuclear p53 and allowed the detection of a small amount of cytoplasmic p53 in both wild-type and Parc−/− MEFs. Notably, no detectable redistribution of p53 from the cytoplasm to the nucleus was observed in either wild-type or Parc−/− MEFs. Western blotting for PTEN and nuclear Lamins A and C confirmed proper separation of the cytoplasm from the nucleus and equal loading of protein.
To further examine the effect of Parc deletion on p53 localization, immunofluorescence microscopy for p53 was performed on wild-type, Parc−/−, and Trp53−/− MEFs (Fig. (Fig.7B).7B). As in Western blots, endogenous, noninduced p53 was undetectable by immunofluorescence in MEFs of all genotypes. MG132 treatment prior to immunostaining allowed the specific detection of p53 in wild-type and Parc−/− MEFs but not the in Trp53−/− MEF control. In both wild-type and Parc−/− MEFs, p53 was detectable only in the nuclear compartment, and no redistribution of p53 from the cytoplasm to the nucleus was apparent.
Mammals encode at least eight cullins that all form structurally similar yet unique complexes, sharing some common elements. During an investigation of the CUL7 complex, PARC was found to form an endogenous complex with CUL7. Additionally, homodimers of hCUL7 and hPARC were observed in overexpression systems. The ability to form homodimers as well as heterodimers suggests that three distinct complexes containing CUL7 and/or PARC may exist in cells: CUL7-CUL7, CUL7-PARC, and PARC-PARC. This possibility is supported not only by the ability of overexpressed protein to form homodimers and heterodimers but also by data from coimmunoprecipitations of endogenous protein that show not all of PARC is in complex with CUL7 nor is all of CUL7 in complex with PARC.
The viability of Parc−/− mice contrasts with the neonatal lethality of Cul7−/− mice and suggests that the interaction between CUL7 and PARC is not essential for normal development of mice. This observation suggests that although a CUL7-PARC complex forms, only a complex containing CUL7—potentially as a homodimer—is required for development. It is interesting that although CUL7 and PARC are ubiquitously expressed, their expression levels have been reported to vary greatly in certain tissues, which could make the CUL7-to-PARC expression ratio the determining factor in the formation of functional complexes (13, 20). Indeed, the relative expression levels of CUL7 and PARC were substantially different when several different human cancer cell lines were compared (data not shown).
CUL7 binding to PARC represents the first example of cullin dimerization. Several models could explain the functional significance of this interaction. First, CUL7 could be a substrate for PARC-mediated ubiquitination and degradation, or PARC could be a substrate for CUL7-mediated ubiquitination and degradation. However, this possibility seems unlikely given the relative expression levels of each protein in the other's knockout. Parc−/− MEFs express normal or slightly reduced levels of CUL7, and Cul7−/− MEFs show decreased rather than increased levels of PARC. Other explanations for the CUL7-PARC interaction include the requirement of an external RING protein (other than RBX1) for certain cullin functions. An SCF complex containing CUL1 and FBXW7 has been reported to bind Parkin (PARK2), a PARC-related RING-IBR-RING protein, to target cyclin E for degradation in neurons, suggesting that other cullins may have RING-IBR-RING binding partners required for subsets of substrates (27). Finally, previous studies have demonstrated hetero- and homodimerization of F-box proteins. In particular, a complex containing human β-TRCP-1 and β-TRCP-2 has been shown to target IκBα for ubiquitination, and a complex containing Schizosaccharomyces pombe Pop1 and Pop2 has been shown to target Cdc18 and Rum1 for ubiquitination (14, 28, 32). These results suggest that the dimerization of CUL7 and PARC could be mediated through F-box proteins and that other cullins, such as CUL1, could form dimers. Notably, dimerization of F-box proteins has been reported only for F-box proteins containing WD-40 repeats, such as β-TRCP1, β-TRCP2, Pop1, and Pop2. The only known F-box for CUL7, FBXW8, is also a WD-40 domain-containing F-box, suggesting that FBXW8 dimerization could mediate the interaction between PARC and CUL7. Attempts to analyze the contribution of F boxes to CUL7 and PARC dimerization will require the identification of all F-boxes that bind CUL7 and PARC.
To date, all of the targeted deletions of cullins in mice have resulted in early embryonic lethality, with only the Cul7 knockout producing lethality at later stages in development (1, 4, 26, 30). In light of the previously observed cullin phenotypes, the high degree of conservation of the Parc gene from fish to human, and the inviability of the Cul7 knockout, the targeted deletion of Parc was expected to result in embryonic lethality. However, Parc−/− mice were born at a normal Mendelian ratio and exhibit no apparent growth or developmental differences when compared to littermate wild-type animals. Currently it is unclear why Parc−/− mice survive when mice lacking Cul7, an apparent evolutionary descendant of Parc, survive. The fish genome does not appear to contain Cul7, and it is likely that fish PARC fulfills a role handled jointly by mouse CUL7 and PARC. Although it is possible that CUL7 could compensate for the loss of PARC in Parc−/− mice, it is notable that CUL7 and PARC, though both ubiquitously expressed, are often not highly expressed in the same tissues, suggesting that complementation would require large upregulations of CUL7 that are not observed in knockout MEFs. Unfortunately, it is not possible to generate a Parc−/− Cul7+/− or a Parc−/− Cul7−/− embryo or mouse via mating of existing knockout mice, since the two genes are less than 130 kb apart on mouse chromosome 17 (Fig. (Fig.1).1). In order to study the ability of CUL7 to mask Parc−/− phenotypes, Parc+/− ES cells will require retargeting of Cul7.
The survival of Parc−/− mice is surprising not only given the lethal phenotypes of the other cullins and the evolutionary conservation of PARC but also because of previously published reports that depletion of PARC by small interfering RNA leads to apoptosis through p53 relocalization (20). If PARC functioned as an inhibitor of p53 activity, the Parc knockout mouse should exhibit a phenotype similar to that of the Mdm2 knockout, which exhibits early embryonic lethality due to the inability to deactivate p53 (12, 19). However, the survival of the Parc knockout mouse and the ability to establish U-2 OS lines stably expressing shRNAs against PARC suggest that the reported induction of apoptosis upon depletion of PARC was a nonspecific effect of the single small interfering RNA used (24).
Although a PARC-p53 complex was detectable in MEFs, Parc deletion failed to affect p53 stability, target gene induction, or localization. If PARC functioned as an E3 ubiquitin ligase for p53, it would be expected that Parc−/− MEFs would exhibit increased steady-state levels of p53, but, unexpectedly, the inverse was true. PARC−/− MEFs contained normal levels of p53, while Trp53−/− MEFs displayed markedly lower levels of PARC than wild-type MEFs. The ability to upregulate and later downregulate p53 in Parc−/− MEFs in response to γ- irradiation also appears intact, suggesting that even when MDM2-ubiquitinated p53 was shuttled from the nucleus to the cytoplasm, PARC has little or no role in p53 degradation following a response to ionizing radiation. Additionally, the observed induction of MDM2 demonstrates that Parc deletion has no effect on the transactivation activities of p53 in response to γ-irradiation. Previous reports suggested that PARC negatively regulated p53 by sequestration in the cytoplasm, and the loss of p53 sequestration by PARC caused increased expression of p53 target genes. However, Parc deletion had no effect on the levels of p21 and MDM2, and in Parc−/− MEFs no change in the ratio of nuclear to cytoplasmic p53 was observed compared to wild-type MEFs, even when the levels of p53 were increased by the addition of MG132. Finally, deletion of Parc did not affect the ability of p53 to induce a G1 arrest following γ irradiation, suggesting that p53 is fully functional in Parc−/− MEFs.
The absence of an effect of Parc deletion on p53 stability and localization demonstrates that PARC is not necessary for these functions. Although the high degree of homology between PARC and CUL7 and the presence of a CUL7-p53 complex in human cells suggest that CUL7 may be sufficient to properly regulate p53 levels and activity in the absence of PARC, p53 protein levels and activity appear normal in Cul7−/− MEFs (data not shown). Notably, these Cul7−/− MEFs exhibit normal p53 levels and function even though they express very low levels of PARC.
Several alternative models to the negative regulation of p53 by PARC through sequestration and ubiquitination also exist. Transcription factors such as VP16 require ubiquitination for activity, making it possible that PARC could ubiquitinate p53, thereby increasing transcriptional activity (35). However, the ability of p53 to induce MDM2 in Parc−/− MEFs suggests that such an activity is not required for Mdm2 induction. Additional models also address the apparent discrepancy of the binding of PARC, a cytoplasmic protein, to p53, a predominantly nuclear protein. Cytoplasmic p53 has been reported to translocate to the mitochondria following DNA damage to induce apoptosis (18). It is possible that p53 translocation to the mitochondria could be controlled by PARC binding and modification, since small ubiquitin-like protein modifiers often impact protein localization. Finally, the binding of p53 to an N-terminal region of PARC also suggests an alternative model to PARC-based regulation of p53. The N-terminal region of cullins dictates binding to bridging and substrate specificity proteins, such as SKP1 and F-boxes, and p53 binding to PARC may block the binding of these proteins, inactivating a PARC-based, SCF-like, E3 ubiquitin ligase. Such a model could also explain the apparent discrepancy of a cytoplasmic protein, PARC, binding a predominantly nuclear protein, p53. Under normal conditions an active PARC-based complex could form, degrading specific substrates in isolation from nuclear p53. Upon exposure to genotoxic stresses, the stabilization of p53 allows the cytoplasmic concentration of p53 to rise, acting as an inhibitor of the cytoplasmic PARC complexes. Further work to determine specific substrate adapters for PARC is required to evaluate this model and to examine other potential roles of PARC-p53 complexes.
We thank J. Horner and members of the DFCI Transgenic gene Targeting Core Facility for knockout mouse production; R. DePinho, N. Bardeesy, and A. Aguirre for the pKOII vector, EIIa-Cre mice, and general mouse advice; and E. McIntush at Bethyl Laboratories for help in antibody design and production. We also thank the Kazusa DNA Research Institute for KIAA clones, Darrell Borger for the pLB(N)CX T Ag construct, and J. Kasper for sharing unpublished data on CUL7.
J.R.S. was supported by an NSF Graduate Research Fellowship. This work was funded in part by National Cancer Institute grant RO1CA93804.