The WD40 Repeats of FANCL Are Required for Fanconi Anemia Core Complex Assembly*

Fanconi anemia (FA) is an autosomal recessive disorder characterized by aplastic anemia, cancer susceptibility, and cellular sensitivity to mitomycin C. Eight of the 11 cloned Fanconi anemia gene products (FANCA, -B, -C, -E, -F, -G, -L, and -M) form a multisubunit nuclear complex (FA core complex) required for monoubiquitination of a downstream FA protein, FANCD2. FANCL, which possesses three WD40 repeats and a plant homeodomain (PHD), is the putative E3 ubiquitin ligase subunit of the FA complex. Here, we demonstrate that the WD40 repeats of FANCL are required for interaction with other subunits of the FA complex. The PHD is dispensable for this interaction, although it is required for FANCD2 mono-ubiquitination. The PHD of FANCL also shares sequence similarity to the canonical RING finger of c-CBL, including a conserved tryptophan required for E2 binding by c-CBL. Mutation of this tryptophan in the FANCL PHD significantly impairs in vivo mono-ubiquitination of FANCD2 and in vitro auto-ubiquitination activity, and partially impairs restoration of mitomycin C resistance. We propose a model in which FANCL, via its WD40 region, binds the FA complex and, via its PHD, recruits an as-yet-unidentified E2 for mono-ubiquitination of FANCD2.

M) form a core complex required for the mono-ubiquitination of FANCD2 (15).
Ubiquitination has many functional roles. While it was first described for its involvement in proteasome-mediated degradation, it has since been shown to regulate a diverse set of processes, including cellular trafficking and DNA repair (16). Ubiquitination is carried out by a cascade of three enzymes: the E1 activates a ubiquitin through ATP hydrolysis and then transfers the moiety, via a thioester linkage, to an E2 ubiquitin-conjugating (UBC) enzyme. The E3 ubiquitin ligase mediates transfer of the ubiquitin to the substrate and, thereby, confers specificity. Consistent with this role, there are several classes of E3s, each with a distinct mechanism of ubiquitin transfer.
HECT domain E3s bind ubiquitin via a thioester bond and then transfer it directly to the substrate (16). Alternatively, RING domain E3 ligases recruit both the E2 and the substrate, bringing them into close proximity and allowing direct transfer of the ubiquitin from the E2 to an isopeptide linkage with an internal lysine of the substrate. The E3/E2 interaction is mediated by the RING domain, while the E3/substrate interaction is mediated by a distinct substrate-recognition domain that varies with each ligase (17).
Several lines of in vivo evidence strongly suggest that the FA complex is the E3 ligase for FANCD2. Specifically, FANCD2 mono-ubiquitination is (a) lost in patient-derived cell lines lacking any single subunit of the complex (1); and (b) reduced in "wild-type" cell lines (HeLa, U20S, 293T) following RNA interference knockdown of various members of the FA complex (11)(12)(13).
Additionally, FANCL was recently identified as the putative catalytic E3 ubiquitin ligase subunit of the FA core complex (11). It possesses three WD40 repeats and a plant homeodomain (PHD). In general, WD40 repeats mediate protein-protein interactions, while PHD fingers have been ascribed various functions, including phosphoinositide binding (18), chromatin association, and ubiquitin ligase activity (19). There is some debate, however, as to whether PHD E3s are, in fact, better categorized as RING finger variants (20). In vitro assays demonstrated auto-ubiquitination activity of FANCL in the presence of E1, E2, ATP, and ubiquitin, and this activity is lost upon mutation of a conserved zinc-coordinating cysteine that forms part of the PHD consensus sequence (11). This mutant also fails to restore FANCD2 mono-ubiquitination in vivo. It has also been suggested that the FANCL PHD is required for complex stability (11,21). No study has shown in vitro ubiquitination of FANCD2 by FANCL.
The precise roles of each domain of FANCL are poorly understood. Therefore, this study carries out structure/function analysis of the domains of FANCL. Through mutagenesis, we show that the FA complex is bound and stabilized by the WD40 repeats of FANCL and that the PHD is dispensable for this interaction. We also demonstrate that a tryptophan conserved in PHD and RING-variant E3s is required for full activity of FANCL, both in vivo and in vitro. We propose a model in which FANCL, via its WD40 repeats, binds the FA complex and, via its FA-L lymphoblasts were retrovirally transduced with pMMP-puro-FLAG-FANCL or pMMP-puro-vector and then selected with puromycin. Following treatment with 0.48 M (160 ng/ml) MMC or 2 mM hydroxyurea for 24 h, the cells were lysed and analyzed by Western blot for FLAG-FANCL and for the mono-ubiquitination of FANCD2, seen as an upper band, FANCD2-L, that is 7 kDa larger than the unmodified form, FANCD2-S. B, complemented FA-L lymphoblasts are resistant to MMC. Standard error of the mean was calculated from three independent experiments. Cells were treated with the indicated concentration of the DNA cross-linking agent, MMC, for 5 days. Cell survival was determined by CyQuant cell proliferation assay kit. OE, EUFA868 (FA-L) ϩ vector; ࡗ, EUFA868 (FA-L) ϩ FLAG-FANCL; F, GM02254 (wild-type). C, complemented FA-L lymphoblasts exhibit reduced radial chromosomes following DNA cross-linker damage. Cells were treated with 60 nM (20 ng/ml) MMC for 48 h and then treated with 10 g/ml colcemid for 1 h. Metaphase spreads were then scored for chromosomal aberrations; a cell containing a radial form was counted as a positive hit. Standard error of the mean was calculated from three independent experiments.
Netherlands) was subcloned into the retroviral vector pMMP-puro. pMMP-puro-FLAG-FANCL was generated by adding the FLAG tag (DYKDDDDK) at the amino terminus of FANCL. Production of pMMP retroviral supernatants was performed as previously described (24). FLAG-FANCL point mutants were generated through directed PCR mutagenesis (QuikChange Kit, Stratagene).
Retroviral Infection-FA-L lymphoblasts underwent three rounds of infection with pMMP supernatants: each round lasted 24 h in the pres- . The WD40 repeats of FANCL, but not the PHD, are required for interaction with the FA complex. A, schematic of FLAG-FANCL WD40 deletion and truncation mutants with a summary of FANCD2 mono-ubiquitination status, D2-Ub, and stabilization of the FA core complex. Each WD40 repeat, designated WD40-1, -2, and -3, was deleted individually. The region of residues from 275-306, between WD40-3 and the PHD, is designated as the linker region. B, FLAG-FANCL WD40 repeat deletions do not bind FANCA. FLAG-FANCL was immunoprecipitated with anti-FLAG M2 agarose from lysates of retrovirally transduced EUFA868 (FA-L) lymphoblasts. The immunoprecipitation was then analyzed by Western blot for FLAG-FANCL and FANCA. C, the linker, but not the PHD, is required for binding of FANCA and FANCG. FLAG-FANCL was immunoprecipitated and analyzed by Western blot. Asterisks indicate nonspecific bands. D, the WD40 repeats and PHD are required for mono-ubiquitination of FANCD2. Cells were treated with 0.48 M (160 ng/ml) MMC for 24 h and then lysed for analysis by Western blot. Asterisks indicate nonspecific bands. ence of 8 g/ml polybrene (Sigma) and was followed by incubation for 24 h in regular RPMI (15% fetal bovine serum). After the final round, infected cells were washed free of viral supernatant and resuspended in growth media. After 48 -72 h, the cells were transferred to media containing 1 g/ml puromycin. Dead cells were removed over Ficoll-Paque Plus (Amersham Biosciences, Uppsala, Sweden) cushion after 5 days, and the surviving cells were grown under continuous selection in puromycin.
MMC Sensitivity and Chromosomal Breakage Assays-MMC sensitivity assays for lymphoblasts were performed essentially as described previously (25) but with a CyQuant cell proliferation assay kit (Invitrogen). Chromosome breakage analysis was performed by the Cyto- genetics Core Facility of the Dana-Farber Cancer Institute as described previously (26).
Immunoprecipitation-Whole-cell extracts were prepared in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% (v/v) Triton X-100) supplemented with protease tablets (Roche Applied Science). Each lysate was normalized to contain 2 mg of total protein in 1 ml of lysis buffer, and FLAG-FANCL was immunoprecipitated with 50 l of packed M2 anti-FLAG agarose (Sigma) for 24 h at 4°C. The agarose was then washed three to four times with chilled lysis buffer. FLAG-FANCL was eluted with either 150 ng/ml FLAG peptide or with SDS sample buffer (Bio-Rad) containing 5% ␤-mercaptoethanol.
In Vitro Auto-ubiquitination Assay-We carried out in vitro autoubiquitination assays essentially as described previously (11,31). The Boxed residues are required for zinc ion coordination. A tryptophan, marked with an arrow, that is implicated in E2 binding by c-CBL is also conserved in several RING variant domains of E3 ligases. Alignment carried out manually. B, species alignment of FANCL. The boxed and shaded region represents conserved residues. Trp-341, marked with an arrow, is 100% conserved, despite degeneracy of flanking sequences in fruit fly and mosquito. Alignment was adapted from MegAlign. C, Trp-341 is required for full monoubiquitination of FANCD2. FA-L lymphoblasts retrovirally transduced with wild-type or mutant FLAG-FANCL were treated with MMC for 24 h and analyzed for status of FANCD2-L. D, Trp-341 is not required for FA complex assembly or stabilization. FLAG-FANCL and mutants were immunoprecipitated from lysates of undamaged FA-L lymphoblasts. No difference was observed in binding of wild-type or point mutant to FANCA, -F, or -G. An asterisk marks a nonspecific band. Tubulin is shown as a loading control for input and negative control for the immunoprecipitation.
Generation of DNA Damage-For mitomycin C and hydroxyurea (Sigma) treatment, cells were continuously exposed to the drug for the indicated time.

pMMP-FLAG-FANCL Functionally Corrects FA-L Lymphoblasts-
EUFA868 lymphoblasts transcribe FANCL cDNA with an insertion at a splice junction between intron 10 and exon 11, resulting in the removal of the PHD and part of the third WD40 repeat from FANCL (11). Therefore, no wild-type FANCL is detectable in EUFA868 lymphoblasts. We complemented these cells with retroviral pMMP-FLAG-FANCL or pMMP-vector only. To confirm that the cells were functionally corrected, we assayed for FANCD2 mono-ubiquitination and resistance to the DNA cross-linker, MMC. The upper band of FANCD2, representing the mono-ubiquitinated protein, was restored following treatment with MMC or hydroxyurea (Fig. 1A), and cell survival of complemented EUFA868 lymphoblasts in response to MMC treatment was similar to that of GM02254 wild-type lymphoblasts (Fig. 1B). Additionally, the cells complemented with FLAG-FANCL had significantly reduced radial chromosomes compared with FA-L lymphoblasts infected with vector control (Fig. 1C) following treatment with MMC.
The WD40 Repeats of FANCL, but Not the PHD, Are Required for Interaction with the FA Complex-We characterized various FANCL domains for their ability to promote FA complex assembly. WD40 repeats form ␤-propeller structures that mediate protein-protein interactions (32). Since FANCL is part of a large multisubunit protein complex, we hypothesized that its WD40 repeats mediate interaction with the FA complex. Therefore, we generated deletion mutants that lack each of these WD40 repeats ( Fig. 2A). Unlike wild-type FLAG-FANCL, none of the deletion mutants co-immunoprecipitated with FANCA ( Fig. 2B) or corrected FANCD2 mono-ubiquitination (Fig. 2D).
We also generated two C-terminal truncations of FANCL ( Fig. 2A). FLAG-FANCL 1-306 terminates immediately before the first conserved cysteine of the PHD; FLAG-FANCL 1-275 terminates immediately after the third WD40 repeat and, therefore, is missing the linker region between the third WD40 and the PHD.
Level of Interaction between FANCL Point Mutants and the FA Complex Correlates with FANCD2 Mono-ubiquitination-The loss of interaction observed in the WD40 deletion mutants could be explained by gross structural changes that disturb the overall fold of FANCL. To test whether the interaction with the FA complex was affected by less disruptive mutations, we altered individual residues in the FANCL WD40 region. We identified residues in FANCL that were completely conserved through Drosophila melanogaster (fruit fly) and carried out either charge reversal or alanine substitutions as indicated (Fig. 3A). FLAG-FANCL R226E did not bind FANCA (Fig. 3B) and had no activity in FANCD2 mono-ubiquitination (Fig. 3C). In contrast, FLAG-FANCL Y111E , FLAG-FANCL W201A , and FLAG-FANCL W275A bound FANCA weakly (Fig. 3B), with a corresponding decrease in activity, as determined by the reduced levels of D2 upper band (Fig. 3C). These data indicate that the WD40 region of FANCL interacts with the FA complex and that the strength of this interaction correlates with the level of FANCD2 mono-ubiquitination.
A subset of WD40 repeat ␤-propellers, such as the ␤-TRCP subunit of SCF ligases, recognize phospho-proteins (33,34). Since at least one subunit of the FA complex, FANCM, is hyperphosphorylated following DNA damage (13), we investigated whether FANCL association with the FA complex was enhanced in cells treated with MMC. We saw no change in FANCA co-immunoprecipitation following DNA damage (data not shown). Furthermore, FANCL did not co-immunoprecipitate with its putative substrate, FANCD2 (data not shown), which is hyperphosphorylated following MMC-treatment (35).
Trp-341 of FANCL Is Required for Efficient Mono-ubiquitination of FANCD2-Next, we investigated the function of the FANCL PHD. Based on sequence, structural, and functional similarity, it has been postulated that PHD E3 ubiquitin ligases are variants of RING finger E3s (20,36). To determine whether these two types of E3 domains share sequence similarity, we carried out a sequence alignment comparing the PHD of FANCL, the RING variant (RINGv) domains of several E3 ligases, and the well characterized RING of c-CBL (Fig. 4A), an E3 that ubiquitinates receptor tyrosine kinases.
Based on these alignments, we hypothesized that the PHD of FANCL plays a role similar to that of c-CBL by recruiting an E2 and that Trp-341 is involved in this interaction. To test this possibility, we expressed FLAG-FANCL W341G in EUFA868 lymphoblasts. As a control, we carried out a parallel infection of lymphoblasts with FLAG-FANCL C307A , which previously has been shown to be inactive and should be unable to coordinate a zinc ion, a key structural feature of PHD and RING fingers. FLAG-FANCL W341G had significantly reduced activity in comparison to the wild-type protein. In the absence of DNA damage, there was a lower level of mono-ubiquitinated FANCD2 compared with cells corrected with wild-type FANCL (Fig. 4C, compare lane 3 with lane 7). This modification could not be further activated with MMC treatment To confirm that this mutation did not lead to gross structural instability that disrupts the association of FANCL with the FA complex, we immunoprecipitated FLAG-FANCL W341G and wild-type FLAG-FANCL. For both proteins, we observed normal binding to FANCA, FANCF, and FANCG, as compared with wild-type protein (Fig. 4D). Furthermore, both proteins stabilized the subunits of the FA complex to a similar extent (Fig. 4D). Tubulin is shown as a loading control for the input and as a negative control for binding. These data confirm that Trp-341 is necessary for full FANCL activity and are consistent with the hypothesis that the PHD plays a role independent of FA complex binding.
Trp-341 Is Required for in Vitro Auto-ubiquitination of the FANCL PHD-To test whether reduced FANCD2 mono-ubiquitination in FA-L lymphoblasts expressing FLAG-FANCL W341G is due to a reduction in E3 ligase activity, we carried out in vitro ubiquitination with GST fusions of the FANCL PHD (GST-PHD). After incubation of GST-PHD WT with E1, E2, His-ubiquitin, and ATP, we observed an increase in higher molecular weight species by Coomassie stain (Fig. 5, top panel,  compare lane 1 with lanes 2-5). These bands were immunoreactive with anti-GST antibody (Fig. 5, lower panel) and anti-ubiquitin antibody (Fig. 5, middle panel), indicating auto-ubiquitination of GST-PHD WT .
The negative control, GST-PHD C307A , was inactive (Fig. 5, lane 6), indicating that auto-ubiquitination is mediated by an intact PHD. GST-PHD W341G was also inactive (Fig. 5, lane 7). This result is consistent with the hypothesis that Trp-341 is required for full E3 ubiquitin ligase activity of FANCL and mono-ubiquitination of FANCD2.
FLAG-FANCL W341G Partially Rescues MMC Sensitivity and Chromosomal Breakage in EUFA868 Lymphoblasts-We tested several mutants for their functional effects downstream of FANCD2 ubiquitination. FLAG-FANCL constructs individually lacking each WD40 repeat displayed levels of MMC sensitivity similar to that of vector controls (

DISCUSSION
Many WD40 proteins have been shown to mediate protein-protein interactions (32)(33)(34). Similarly, our study demonstrates that FANCL binds and stabilizes the FA complex through its three WD40 repeats and linker region. This interaction does not change with DNA damage, suggesting that the complex is constitutively assembled, an observation that is in agreement with previous studies (27,30,37).
Of note, ␤-propeller structures formed by WD40 domains typically contain seven repeats, while FANCL possesses only three. FANCL may contain cryptic repeats between the identifiable WD40 repeats that cannot be detected by current algorithms. There are enough residues between each repeat for such a possibility. Similarly, BUB3p, a protein that mediates the spindle checkpoint, was thought to possess only four WD40 repeats until a crystal structure revealed seven authentic repeats (38). Alternatively, FANCL may oligomerize to complete the full set of repeats.
Surprisingly, the PHD domain is not required for complex assembly, indicating a role distinct from FA complex stabilization. Sequence alignment reveals similarity of E3 PHD and RING variant domains to the canonical RING finger of c-CBL, an E3 ubiquitin ligase. This observation is in agreement with other published studies concluding that PHD domains of E3 ligases are, in fact, RING finger variants (20,36).
In particular, we observed conservation of a tryptophan that, in c-CBL, was shown in a crystal structure to interact with its E2, UBCH7 (17). Mutation of this tryptophan in the c-CBL RING abrogates in vitro E2 binding as well as in vitro and in vivo ubiquitin ligase activity (39). Similarly, mutation of a conserved tryptophan, Trp-341, in the FANCL PHD disrupts its activity both in vivo and in vitro. This mutation does not, however, affect the assembly of the FA complex in FA-L lymphoblasts. Based on these observations, we predict that the FANCL PHD is structurally similar to RING finger E3 ligases and recruits an unidentified E2 UBC.
These results also suggest that the E2 is not required for complex FIGURE 7. Model of FANCD2 ubiquitination by FA complex. FANCL interacts with the FA complex via its WD40 repeat region. The PHD is not required for this interaction but is required for mono-ubiquitination of FANCD2, likely through recruitment of an E2 ubiquitin-conjugating enzyme. FANCL does not interact with its substrate, FANCD2, and therefore, the substrate may be recruited by a different subunit of the complex. Based on past studies (41,43,49), FANCE is a plausible candidate for FANCD2 recruitment.
stabilization. Interestingly, the FA core complex is intact in cells belonging to the FA-I complementation group despite the absence of FANCD2 mono-ubiquitination (40). The FANCI gene, which has not yet been cloned, could, in principle, encode for the E2 UBC of the FA pathway. Surprisingly, we did not observe an interaction between FANCL and the FA substrate, FANCD2. It is possible that the recruitment of FANCD2 is carried out by another member of the complex. A rough interaction map of the FA complex has emerged through numerous co-immunoprecipitation and yeast two-hybrid studies. FANCA and FANCG interact and each requires the other for stability (37). Other interacting FA pairs include FANCF and FANCG (22,37), FANCE and FANCC (22,41), and FANCB and FANCL (12). Yeast two-hybrid studies have also characterized FANCE interactions with both FANCC and the N-terminal of FANCD2 (42). A C-terminal truncation of FANCE interacts with FANCC but not with FANCD2 (42,43). These data suggest that the C-terminal one-third of FANCE recruits FANCD2 to the FA complex, while the N-terminal two-thirds interact with the FA complex.
Based on the data presented here, we propose a model in which the binding of the substrate and the E2 is mediated by different subunits (Fig. 7). This mechanism is similar to that of multisubunit E3s, such as SCF ligases (44). In this model, FANCL is functionally analogous to the RING-box protein (Rbx) and recruits an E2, while FANCE may be analogous to the F-box protein that binds both the substrate, conferring substrate specificity, and the multisubunit SCF scaffold.
Consistent with this hypothesis, we were unable to ubiquitinate FANCD2 by FANCL alone in vitro (data not shown). It is possible that multiple FA complex subunits, perhaps even the entire complex, will be necessary for FANCD2 ubiquitination in vitro. Similarly, omission of any one of the three core components of SCF HOS -ROC1 E3 ligase results in loss of in vitro ubiquitination of its substrate, IB␣ (45). It is also possible that the unidentified FA-specific E2 UBC is required for in vitro mono-ubiquitination of FANCD2. Nevertheless, the current data do not exclude the possibility that FANCD2 is not the direct substrate for FANCL and the FA complex.
Our results also predict additional substrates of the FA complex. A recent study demonstrated that FANCD2 fused at its C terminus to ubiquitin was able to complement DT40 chicken cells lacking FANCD2 (21). This fusion, however, could not complement DT40 cells expressing a FANCL PHD mutant in place of the wild-type protein. Based on our studies, we predict that these chicken FA-L cells have an intact FA complex that lacks E3 ligase activity. The observed hypersensitivity to MMC might therefore be the result of other substrates that, in response to genotoxic stress, are active only after modification by the FA complex.
We also tested chromosomal aberrations in FA-L lymphoblasts expressing various FLAG-FANCL mutants. Cells expressing WD40 deletion mutants were similar to vector controls in terms of MMC hypersensitivity. Similarly, FLAG-FANCL 1-306 , which binds and stabilizes the complex, also failed to correct the cells functionally. FLAG-FANCL W341G partially corrected both MMC hypersensitivity and radial chromosomes, consistent with the reduced ubiquitin ligase activity as observed in vitro and by FANCD2 Western blot. These data indicate that FANCD2 mono-ubiquitination is a sensitive assay for detecting defects in the FA pathway and correlates with the functional status of a cell. Therefore, FANCD2 mono-ubiquitination may be a useful screen for 1) detecting a subset of FA patients who have FA pathways with partial activity (46,47) and 2) detecting FA pathway defects in human tumors (48).
The structure/function analysis carried out by this study raises new questions. What is the identity of the E2 in the FA pathway? Does this UBC regulate cellular processes other than DNA damage and, if so, what is their relation to FA? Which mechanistic features dictate transfer of mono-ubiquitin, rather than poly-ubiquitin, to FANCD2? Finally, what is the identity of the other substrates of the FA pathway? The answer to these questions may shed further light on the mechanism of the FA pathway in DNA cross-link repair.