Functional Interaction between the Fanconi Anemia D2 Protein and Proliferating Cell Nuclear Antigen (PCNA) via a Conserved Putative PCNA Interaction Motif*

Fanconi Anemia (FA) is a rare recessive disease characterized by congenital abnormalities, bone marrow failure, and cancer susceptibility. The FA proteins and the familial breast cancer susceptibility gene products, BRCA1 and FANCD1/BRCA2, function cooperatively in the FA-BRCA pathway to repair damaged DNA and to prevent cellular transformation. Activation of this pathway occurs via the mono-ubiquitination of the FANCD2 protein, targeting it to nuclear foci where it co-localizes with FANCD1/BRCA2, RAD51, and PCNA. The regulation of the mono-ubiquitination of FANCD2, as well as its function in DNA repair remain poorly understood. In this study, we have further characterized the interaction between the FANCD2 and PCNA proteins. We have identified a highly conserved, putative FANCD2 PCNA interaction motif (PIP-box), and demonstrate that mutation of this motif disrupts FANCD2-PCNA binding and precludes the mono-ubiquitination of FANCD2. Consequently, the FANCD2 PIP-box mutant protein fails to correct the mitomycin C hypersensitivity of FA-D2 patient cells. Our results suggest that PCNA may function as a molecular platform to facilitate the mono-ubiquitination of FANCD2 and activation of the FA-BRCA pathway.

A central step in the activation of the FA-BRCA pathway is the mono-ubiquitination of the FANCD2 and FANCI proteins, catalyzed by the core FA E2/E3 holoenzyme complex (5,6). The mono-ubiquitination of FANCD2 and FANCI signals their translocation to discrete nuclear foci, where they co-localize with the BRCA1 and RAD51 DNA repair proteins, as well as the major cellular DNA polymerase processivity factor PCNA (3,4,(7)(8)(9). Several studies have suggested an important role for the FA-BRCA pathway in a DNA replication-associated DNA repair process, e.g. homologous recombination (HR), and/or translesion DNA synthesis (TLS) (3, 4, 10 -12). Accordingly, additional proteins with established roles in the DNA replication stress response, including ATR, CHK1, HCLK2, and RPA, modulate DNA damage-inducible FANCD2 mono-ubiquitination (13)(14)(15). Our understanding of the regulation of this critical post-translational modification, however, is incomplete.
We, and others (4,7) have previously reported an association between FANCD2 and PCNA. FANCD2 and PCNA co-localize in nuclear foci following treatment with agents that inhibit DNA replication. Like FANCD2, PCNA is mono-ubiquitinated following exposure to DNA-damaging agents (16,17). While FANCD2 and PCNA are mono-ubiquitinated by different E3 ubiquitin ligases, FANCL and RAD18 (16 -19), respectively, both proteins are de-ubiquitinated by the USP1 enzyme (20,21). The functional significance of the FANCD2-PCNA interaction, however, has not been determined.
In addition to its role as a DNA polymerase processivity factor, PCNA interacts with many DNA repair proteins, e.g. MSH3, XPG, and p21 Cip1/Waf1 (22). These interactions typically occur in a hydrophobic pocket of the PCNA homotrimer, termed the interdomain connecting loop (ICL). Proteins that interact with the PCNA ICL harbor a highly conserved PCNAbinding motif called the PIP-box, defined by the amino acid sequence QXXhXXaa, where h represents amino acids with moderately hydrophobic side chains, e.g. leucine, isoleucine, or methionine (L, I, M), a represents amino acids with highly hydrophobic, aromatic side chains, e.g. phenylalanine and tyrosine (F, Y), and X is any amino acid (23).
Here, we describe an important functional interaction between FANCD2 and PCNA. We have identified a highly con-served putative PIP-box in FANCD2, and demonstrate that mutation of this motif disrupts the FANCD2-PCNA interaction, and precludes both the spontaneous and DNA damageinducible mono-ubiquitination of FANCD2. Consequently, the FANCD2 PIP-box mutant fails to correct the MMC hypersensitivity of FA-D2 patient-derived cells. However, the mutant protein retains the ability to localize to chromatin, interact with FANCE, and undergo DNA damage-inducible phosphorylation. Our results suggest that PCNA may act as a molecular platform for the mono-ubiquitination of FANCD2 and for the activation of the FA-BRCA pathway.
DNA Damage Assays-MMC sensitivity and chromosome breakage assays were performed as described previously (4,26). Statistical analysis of total chromosome gaps and breaks data were carried out using the Student's t-test for equal or unequal variance. Variance for each data set was determined using the sample variance F-test. Cells were X-irradiated using a Philips 250 orthovoltage unit at ϳ2 Gy/min in the Irradiation Core of the University of Michigan Cancer Center.

RESULTS
Further Characterization of the FANCD2-PCNA Interaction-Using immunoprecipitation approaches, we sought to expand upon previous findings of an association between FANCD2 and PCNA in DNA damage-inducible nuclear foci (4,7). First, whole-cell lysates were prepared from HeLa cells stably expressing HA-tagged PCNA, untransfected or transiently transfected with a positive control FLAG-tagged p21 construct. HA-PCNA co-immunoprecipitated with FLAG-p21 as well as both non-and mono-ubiquitinated FANCD2 (Fig. 1A). Next, we generated a FA-D2 (FANCD2-null) patient cell line stably expressing 6xHis/V5-tagged FANCD2 (Fig. 1B). Expression of 6xHis/V5-FANCD2 rescued the MMC hypersensitivity of these cells (results not shown). Using a Ni-nitrilotriacetic acid pull-down approach, endogenous PCNA co-immunoprecipitated with 6xHis/V5-FANCD2 from these cells (Fig. 1B, lower panel, lane 3). To analyze the interaction between endogenous FANCD2 and PCNA in chromatin, HCT116 cells were irradiated with 40 J/m 2 UV-C, fractionated in low and high salt lysis buffers, and sonicated briefly to release chromatin-bound proteins (24). We observed an enrichment for mono-ubiquitinated FANCD2 in the sonicated high salt cellular fraction, as expected (27) (Fig. 1C, upper panel, lanes 3 and 6). A mouse monoclonal anti-PCNA antibody, but not mouse IgG, exclusively co-immunoprecipitated FANCD2 from both high salt fractions (Fig. 1C, middle panel, lanes 5 and 6). The specificity of the anti-PCNA mouse monoclonal antibody for PCNA immune complexes was confirmed by immunoblotting with an anti-PCNA rabbit polyclonal antibody (Fig. 1C, lower panel). In a reciprocal approach, a rabbit polyclonal anti-FANCD2 antibody, but not rabbit serum, preferentially immunoprecipitated PCNA from chromatin fractions prepared from untreated and UV-irradiated HCT116 cells (supplemental Fig. S1, lower panel, lanes 3 and 4). In support of a direct interaction between FANCD2 and PCNA, GST-PCNA, and not GST alone, bound to 6xHis/V5-FANCD2 overexpressed in COS-7 cells (Fig. 1D, lanes 6 and 8). For this experiment, COS-7 cells overexpressing FLAG-p21 was used as a positive control (Fig. 1D, lane 8).
The FANCE protein has been proposed to form a molecular bridge between the core FA complex and FANCD2, and has been demonstrated to bind directly to FANCD2 (28). To analyze the effect of mutation of the FANCD2 PIP-box on the FANCD2-FANCE interaction, we transiently transfected PD20F pMMP-Empty, -FANCD2 wt, -FANCD2 K561R, and -FANCD2 PIP⌬ cells with a FLAG-tagged FANCE construct, and examined the association between FANCD2 and FLAG-FANCE (Fig. 2D). The use of an epitope-tagged FANCE construct was required for this experiment because of the absence of a suitable commercially available FANCE antibody. When normalized for total FANCD2 protein levels, approximately equal amounts of FANCD2 wt, FANCD2 K561R, and FANCD2 PIP⌬ co-immunoprecipitated with FLAG-FANCE, indicating that the FANCD2 PIP⌬ mutant protein retains the ability to associate with FANCE ( Fig. 2D and supplemental Fig. S2C). We also sought to determine if mutation of the FANCD2 PIP-box disrupted the interaction between FANCD2 and FANCI. However, we were unable to co-immunoprecipitate FANCD2 wt and FANCI, precluding us from addressing this question. Collectively, these results suggest that the interaction between FANCD2 and PCNA is mediated by a putative FANCD2 PIPbox that is dispensable for FANCD2-FANCE binding.
We next examined the ability of the FANCD2 PIP-box mutant to undergo UV-inducible, ATR-mediated phosphorylation (13,29). Whole-cell lysates prepared from untreated and UV-irradiated PD20F pMMP-FANCD2 wt and -FANCD2 PIP⌬ cells were incubated in the absence or presence of phosphatase. As expected, phosphatase treatment accelerated the migration of FANCD2 wt (Fig. 3D, lanes 1-4). Similarly, phosphatase treatment accelerated the migration of both untreated and UV-irradiated FANCD2 PIP⌬ (Fig. 3D, lanes  5-8). Collectively, these results indicate that the FANCD2 PIP-box mutant fails to undergo both spontaneous and DNA damageinducible mono-ubiquitination, yet remains competent for ATR-mediated phosphorylation.
The FANCD2 PIP-box Mutant Localizes to Chromatin Yet Fails to Form Nuclear Foci-We next examined the subcellular localization of the FANCD2 PIP⌬ mutant protein. Cells were lysed in a hypotonic buffer to extract cytoplasmic proteins (S1). The remaining pellets were lysed in a high-salt buffer to extract soluble nuclear proteins (S2). Salt-extracted pellets were subsequently treated with micrococcal nuclease to release chromatin-associated proteins, characterized by the presence of the nucleosomal histone H2A (S3) (30). We detected mono-ubiquitinated FANCD2 in the S2 and S3 fractions, in the absence and presence of MMC, as expected (Fig. 4A, upper  panel, lanes 9 -12). However, we also detected both non-ubiquitinated FANCD2 K561R and FANCD2 PIP⌬ in all three fractions, in the absence and presence of MMC (Fig. 4A), indicating that the FANCD2 PIP⌬ mutant protein remains competent for chromatin localization. A similar observation has previously been described for FANCD2 K561R (31).
We next examined the ability of the FANCD2 PIP-box mutant protein to assemble into discrete nuclear foci. Following exposure to APH, FANCD2 wt formed discrete nuclear foci as previously reported (3,4) (Fig. 4B). However, like FANCD2 K561R (3), FANCD2 PIP⌬ failed to form nuclear foci in the absence or presence of APH. Instead, we consistently observed non-punctate staining patterns, including diffuse nuclear as well as concentrated nucleolar staining (Fig. 4B).
FANCD2 Mono-ubiquitination Promotes Its Stabilization in Chromatin-We next determined the protein half-lives of FANCD2-Ub, FANCD2 wt, FANCD2 K561R, and FANCD2 PIP⌬. Cells were exposed to the translation inhibitor cycloheximide and protein levels monitored over a 24 h period. FANCD2 K561R and FANCD2 PIP⌬ exhibited significantly shorter half-lives than both FANCD2-Ub and FANCD2 wt. For example, 3 h following exposure to cycloheximide FANCD2 wt, K561R, and PIP⌬ protein levels had decreased by ϳ23, 38, and 81%, respectively (Fig. 5, A and B). In contrast, wild-type FANCD2-Ub protein levels increased ϳ28%, and remained elevated 24 h following the addition of cycloheximide. These results indicate that the mono-ubiquitination of FANCD2 promotes its stabilization in chromatin, and   (Fig. 6A). We also measured the frequency of chromosome gaps, breaks, and radial formations in metaphase spreads prepared from cells treated with APH or MMC. As expected, exposure to APH and MMC induced significantly elevated levels of chromosome aberrations in PD20F pMMP-Empty cells, 6.2-fold (p Ͻ 0.0001) and 2.1-fold (p ϭ 0.0093), respectively, compared with untreated cells (Fig. 6B). Expression of FANCD2 wt corrected the elevated APH-(p ϭ 0.0315) and MMC-(p ϭ 0.001) induced chromosome aberrations of PD20F cells. However, expression of either FANCD2 K561R or FANCD2 PIP⌬ failed to correct the elevated frequency of APH-and MMC-induced chromosome aberrations. Like PD20F pMMP-Empty, a 6.0-fold (p Ͻ 0.0001) and 2.4-fold (p ϭ 0.0026) increase in chromosome aberrations was observed for PD20F pMMP-FANCD2 PIP⌬ cells following treatment with APH and MMC, respectively, compared with untreated cells (Fig. 6B). Finally, using flow cytometry, we examined cell cycle progression in the absence or presence of MMC. Like PD20F pMMP-Empty and -FANCD2 K561R cells, PD20F pMMP-FANCD2 PIP⌬ cells displayed a prolonged accumulation in G2 phase of the cell cycle following exposure to MMC. For example, a ϳ2-fold greater proportion of PD20F pMMP-FANCD2 PIP⌬ cells FIGURE 3. The FANCD2 PIP-box motif is required for both spontaneous and DNA damage-inducible FANCD2 mono-ubiquitination. A, cells were untreated (NT) or exposed to 1.0 M APH for 24 h (APH), and WCE were immunoblotted with the indicated antibodies. The catalytic subunit of DNA Pol ␦ was used as a protein loading control. B, cells were untreated (NT) or exposed to 60 J/m 2 UV irradiation (UV), and 4 h later whole-cell lysates were prepared and immunoblotted with the indicated antibodies. C, cells were untreated (NT) or exposed to 15 Gy X-irradiation (X-IR). Whole-cell lysates were prepared 3 h post-irradiation and immunoblotted with the indicated antibodies. D, PD20F pMMP-FANCD2 wt and -FANCD2 PIP⌬ were untreated (NT) or exposed to 60 J/m 2 UV irradiation (UV), and 4 h later, whole-cell lysates were prepared and incubated in the presence or absence of phosphatase (-P Pase), prior to immunoblotting with the indicated antibodies. accumulated in G2/M following exposure to MMC, compared with PD20F pMMP-FANCD2 wt cells (Fig. 6C). Collectively these results strongly suggest that the FANCD2 PIP-box is essential for the function of FANCD2 in the cellular DNA damage response.

DISCUSSION
In support of an important role for the FA-BRCA pathway in a S-phase DNA repair process, both spontaneous and DNA crosslink-induced FANCD2 mono-ubiquitination and nuclear foci formation occur almost exclusively during S-phase (9,32,33). Thus, it is likely that a stalled or collapsed DNA replication fork is a major stimulus for the mono-ubiquitination of FANCD2. The mechanism by which FANCD2 is recruited to damaged DNA replication forks has not, however, been determined. Here, we describe an important interaction between the FANCD2 and PCNA proteins. We demonstrate that FANCD2 and PCNA associate in chromatin, and that this association is dependent on a highly conserved FANCD2 PIP-box motif. Furthermore, disruption of this motif abrogates both spontaneous and DNA damage-inducible FANCD2 mono-ubiquitination. Our results suggest that the association between PCNA and FANCD2 may be an important prerequisite for the mono-ubiquitination of FANCD2 and, consequently, its DNA repair function. A similar auxiliary role for PCNA, also mediated via PIP-box interactions, has been established for several DNA repair enzymes including MSH3, XPG, and XPV/Pol (34 -37). The association of FANCD2 with PCNA in a DNA damage surveillance capacity represents an attractive model for the localization of FANCD2 to sites of DNA damage encountered by the DNA replisome.
That the mutation of the FANCD2 PIP-box completely abrogated both spontaneous and DNA damage-inducible FANCD2 monoubiquitination was unexpected. To our knowledge, only one other FANCD2 mutation, that of the site of ubiquitin conjugation itself, Lys-561, has been described that leads to a complete loss of FANCD2 mono-ubiquitination (3). We could not rule out the possibility that a change in protein  Error bars represent the S.E. from triplicate measurements, and are smaller than the dimensions of the symbols for several points. This experiment was performed three times with similar findings. B, same cells were untreated (white bars) or exposed to 0.2 M APH (gray bars) or 30 nM MMC (black bars) for 24 h, and chromosome aberrations quantified. At least 30 metaphases were scored for each treatment. Error bars indicate the S.E. This experiment was performed three times with similar findings. C, for the G2 accumulation assay cells were untreated (NT) or exposed to 75 nM MMC for 48 h, fixed, stained with propidium iodide, and analyzed by flow cytometry.
folding underlies the inability of the FANCD2 PIP-box mutant to undergo mono-ubiquitination, as the PIP-box is proximal to Lys-561 (3). Thus, a localized alteration in protein folding may preclude access of the FA E2/E3 holoenzyme to Lys-561. Several lines of evidence, however, argue against a gross structural alteration. First, like FANCD2 wt, the FANCD2 PIP⌬ mutant retained the ability to interact with FANCE, a core FA complex protein previously demonstrated to bind directly to FANCD2 (28). Second, the FANCD2 PIP⌬ mutant protein underwent DNA damage-inducible phosphorylation (13,29). Third, the FANCD2 PIP⌬ mutant protein localized to chromatin. These results suggest that the mutant protein retains structural integrity and at least partial functionality. Nevertheless, we cannot rule out the possibility of a localized effect, or that mutation of the PIP-box precludes the interaction of FANCD2 with a protein, other than PCNA, required for its efficient mono-ubiquitination.
The mono-ubiquitination of FANCD2 has previously been hypothesized to be required for its targeting to chromatin (27). However, in this study, albeit using a different subcellular fractionation protocol, we detected non-ubiquitinated FANCD2 K561R and FANCD2 PIP⌬ proteins in chromatin. We propose that mono-ubiquitination stabilizes FANCD2 in chromatin, and that non-ubiquitinated FANCD2 fails to be retained in chromatin. In support of a chromatin-dependent FANCD2 mono-ubiquitination model, the FANCD2 E2 ubiquitin-conjugating enzyme UBE2T (38) has been demonstrated to be constitutively chromatin-localized, independent of DNA damage or cell cycle stage (31). Furthermore, the FA core complex localizes to chromatin, and its E3 ubiquitin ligase activity is dependent on its DNA damage-induced chromatin localization (31,39). Thus, we hypothesize that the association between FANCD2 and PCNA may facilitate the FA E2/E3 holoenzymemediated mono-ubiquitination of FANCD2, and its stabilization and retention in chromatin until USP1-mediated de-ubiquitination and chromatin release (40). Consistent with this, we have also demonstrated that mono-ubiquitinated FANCD2 displays a longer protein half-life than FANCD2 K561R and, in particular, FANCD2 PIP⌬. Mono-ubiquitin may serve as a molecular adaptor between PCNA-associated FANCD2 and an as yet unidentified ubiquitin-binding domain (UBD)-containing chromatin-associated protein. In support of this hypothesis, Matsushita et al. (41) demonstrated that a chicken DT40 FANCD2-Ub chimeric protein harboring an I44A mutation in the essential hydrophobic patch of ubiquitin (that disrupts ubiquitin-UBD binding) could only weakly complement the cisplatin sensitivity of fancd2 cells. Several candidate chromatin-associated ubiquitin receptors exist, including XPV/Pol , REV1, or one of the core FA complex proteins.
An important FANCD2 mono-ubiquitination-independent function for the FA core complex in REV1-mediated TLS has been demonstrated (11): FANCA and FANCG, yet not FANCD2, were demonstrated to be required for the efficient assembly of the REV1 protein into nuclear foci (11). It has recently been proposed that the mono-ubiquitination of PCNA may be required for a post-replicative DNA repair process, for example daughter-strand gap (DSG) repair (12,42,43). As DSG repair is thought to involve both TLS and a HR-mediated tem-plate switching process (12,43), we propose a model whereby FANCD2 and PCNA, through a mono-ubiquitination-dependent process, may function in the HR repair of DSGs (Fig. 7). This is consistent with several studies indicating a role for the FA proteins in an undefined HR process, as well as earlier observations that the major HR proteins BRCA1, FANCD1/BRCA2, and RAD51, co-localize with PCNA and FANCD2 in nuclear foci during S-phase (44,45).
In summary, our findings strongly suggest that the FANCD2 protein associates with PCNA in a DNA damage surveillance capacity via a highly conserved PIP-box. Upon encounter of a lesion-damaged template during DNA replication, PCNA-associated FANCD2 becomes mono-ubiquitinated via the concerted activity of the core FA complex, the FANCL ubiquitin ligase, and the UBE2T E2 ubiquitin-conjugating enzyme, thus facilitating its function in DNA repair. The erroneous or untimely repair of damaged DNA replication forks could lead to the generation of the complex, unbalanced chromosomal abnormalities frequently observed in FA cancer cells. Our findings have significant implications for our understanding of the DNA damage surveillance and repair mechanisms activated upon DNA replication fork arrest, and lend further insight into the molecular etiology of a rare, yet highly important, cancer susceptibility syndrome.