Functional Interaction of BRCA1-Associated BARD1 with Polyadenylation Factor CstF-50

See allHide authors and affiliations

Science  03 Sep 1999:
Vol. 285, Issue 5433, pp. 1576-1579
DOI: 10.1126/science.285.5433.1576


Polyadenylation of messenger RNA precursors requires a complex protein machinery that is closely integrated with the even more complex transcriptional apparatus. Here a polyadenylation factor, CstF-50 (cleavage stimulation factor), is shown to interact in vitro and in intact cells with a nuclear protein of previously unknown function, BRCA1-associated RING domain protein (BARD1). The BARD1-CstF-50 interaction inhibits polyadenylation in vitro. BARD1, like CstF-50, also interacts with RNA polymerase II. These results indicate that BARD1-mediated inhibition of polyadenylation may prevent inappropriate RNA processing during transcription, perhaps at sites of DNA repair, and they reveal an unanticipated integration of diverse nuclear events.

Almost all eukaryotic mRNAs have a polyadenylated [poly(A)] tail at the 3′ end. Formation of this structure involves endonucleolytic cleavage of the mRNA precursor coupled with poly(A) synthesis, a reaction that requires a complex set of protein factors (1). Although polyadenylation can be reconstituted in vitro with purified components and a synthetic mRNA precursor, considerable evidence now exists that, in the cell nucleus, 3′ end formation is normally tightly coupled to transcription by RNA polymerase II (RNAP II). Polyadenylation factors associate with RNAP II at the promoter (2) and appear to remain with it during elongation (3). When RNAP II reaches the site of polyadenylation, it participates directly in the processing reaction together with the other polyadenylation factors (4); it also receives a signal needed for subsequent transcription termination (5). How these complex interactions are orchestrated is not known. However, polyadenylation can be regulated, for example, during the cell cycle (6) and cellular differentiation (7).

Cleavage stimulation factor (CstF) is a polyadenylation factor that helps specify the site of processing (8, 9). It is a heterotrimeric protein with subunits of 77, 64, and 50 kD (CstF-77, -64, and -50) that recognizes the G+U–rich element, a sequence located downstream of the cleavage site. RNA binding is mediated by CstF-64 (10). CstF-77, or Suppressor-of-forked inDrosophila, bridges CstF-64 and -50 (11) and interacts with another multisubunit factor, the AAUAAA-binding cleavage-polyadenylation specificity factor (CPSF), to define the poly(A) site (12). CstF-50 contains seven WD-40 repeats (13), is required for CstF activity in vitro (11), and interacts with the COOH-terminal domain of the RNAP II largest subunit (CTD) (3).

To identify additional CstF-50–interacting proteins, we performed a yeast two-hybrid assay (14); for bait we used CstF-50 fused to the LexA DNA binding domain (15). Among the strongest interacters recovered was BARD1 (16), a protein known to associate with the breast cancer–associated tumor suppressor BRCA1 in vivo. Both BARD1 and BRCA1 possess NH2-terminal RING motifs and COOH-terminal BRCT domains, with the former responsible for the BARD1-BRCA1 interaction (17). Although the function of BARD1 is unknown, inhibiting its expression in cultured cells results in changes that suggest a premalignant phenotype (18). BARD1 likely plays a role in BRCA1-mediated tumor suppression and may itself be a target for tumorigenic mutations in some cancers (19). BARD1, in association with BRCA1, can colocalize with the DNA replication and repair factors PCNA and Rad51, perhaps participating in the cellular response to DNA damage (20).

The BARD1 cDNAs obtained from the two-hybrid screen encoded amino acids 457 to 777 (Fig. 1A), which includes two of three ankyrin repeats and the BRCT domain. We used essentially full-length BARD1 (amino acid residues 13 to 777), which interacted as efficiently as the smaller fragment, to confirm the interaction in yeast. CstF-50 (amino acid residues 92 to 431), which lacks the NH2-terminal region of CstF-50, also interacted strongly with BARD1. As observed with BRCA1 (21), yeast strains expressing BARD1 displayed a slow-growth phenotype (22).

Figure 1

Interaction of BARD1 and CstF-50 in yeast. (A) Two-hybrid screening identifies a CstF50–BARD1 interaction. LexA–CstF50 fusion proteins are indicated on the left and BARD1 derivatives are shown below. Features of proteins are indicated and numbers represent amino acid residue. Interactions were detected by LEU2 and LacZ expression. (B) Diagram of BARD1 and CstF-50 derivatives used in in vitro experiments. Proteins were produced by in vitro translation or as NH2-terminal GST fusions inE. coli.

We next characterized the BARD1–CstF-50 interaction in vitro. The two proteins were translated separately in vitro, mixed, and subjected to immunoprecipitation with antibodies to BARD1 (anti-BARD1 antibodies) (23) (Fig. 2A). CstF-50 was precipitated in the presence (lane 3), but not in the absence (lane 2), of BARD1, which strongly suggests a specific interaction between the two proteins. In another approach, a glutathione S-transferase (GST)–BARD1 fusion protein (BARD1 residues 13 to 777) was purified from Escherichia coli and used in protein interaction assays, again with in vitro translated CstF-50 (Fig. 2B). A significant fraction of the input CstF-50 bound to GST-BARD1 (lane 3) but not to GST alone (lane 2). COOH-terminal truncations of CstF-50 (see Fig. 1B) eliminated binding (lanes 4 to 7), consistent with the requirement of at least the seventh WD-40 repeat for interaction. To elucidate the region of BARD1 required, we used additional GST-BARD1 derivatives (Fig. 1B) in binding assays with CstF-50 (Fig. 2C). A COOH-terminal truncation removing most of the BRCT domain only slightly reduced binding (compare lanes 3 and 4), indicating that this region is not essential. However, two different derivatives lacking the ankyrin repeats failed to bind CstF-50 (lanes 5 and 6). In combination, the data suggest that the two COOH-terminal ankyrin repeats, the region just downstream, or both constitute the CstF-50 interaction domain.

Figure 2

Interactions of BARD1 and CstF-50 in vitro. (A) Coimmunoprecipitation. BARD1 and CstF-50 were produced by in vitro translation in separate reaction mixtures (only CstF-50 was labeled). Equivalent amounts were mixed, incubated, and then immunoprecipitated with anti-BARD1 antibodies. Pellets were washed, dissolved in sample buffer, and resolved by SDS-PAGE. Lane 1, 5% of CstF-50-containing lysate used for immunoprecipitation. Lane 2, immunoprecipitation of reaction mixture lacking in vitro translated BARD1. Lane 3, immunoprecipitation of reaction mixture containing BARD1. (B) Interaction of GST-BARD1 and CstF-50. Full-length CstF-50 (lanes 1 to 3) and the two indicated COOH-terminal truncations (lanes 4 to 7) were produced by in vitro translation and incubated with purified GST (lane 2) or GST-BARD1 (lanes 3, 5, and 7). Bound proteins were eluted and resolved by SDS-PAGE. Ten percent of each of the CstF-50 derivatives used in binding reactions is shown in lanes 1, 4, and 6. (C) Requirement of BARD1 ankyrin repeats for CstF-50 interaction. GST (lane 2) and the indicated GST-BARD1 derivatives (lanes 3 to 6) were used in in vitro binding reactions with full-length CstF-50 produced by in vitro translation. Lane 1 contains 10% of the CstF-50 used in binding reactions. (D) Interaction with RNAP II CTD and with PCNA. BARD1 and CstF-50 were produced by in vitro translation (lanes 1 and 2) and incubated with 1 μg of either GST (lane 3), GST-CTD (lanes 4 to 6), or GST-PCNA (lanes 7 to 9). Binding reactions were performed and analyzed as in (B).

To determine whether CstF-50 interacts with BARD1 in vivo, we analyzed HeLa cell nuclear extracts (NE) by cofractionation and coimmunoprecipitation (24). Because CstF-50 is part of the CstF heterotrimer, we used an antibody directed against another subunit, CstF-64, which ensured that any interactions detected were with intact CstF. We first tested whether the proteins could be coimmunoprecipitated from NE. The results (Fig. 3A) indicate that antibodies against either protein coprecipitated a small but significant fraction of the other, as judged by immunoblots of the immunoprecipitates (IPs) (lanes 1 to 5). Extraneous antibodies did not immunoprecipitate either protein (lanes 6 and 7), nor did either antibody cross-react with the other protein (22). We also followed the proteins during the early steps of CstF purification, which includes (NH4)2SO4 fractionation followed by Superose-6 gel filtration (25). This produces a fraction, cleavage-specificity factor (CSF), that contains all the factors essential for 3′ cleavage. Immunoblotting of the Superose fractions (Fig. 3B) revealed, as expected, significant amounts of CstF coeluting with CSF activity. However, large amounts were also detected in more rapidly eluting fractions. BARD1, on the other hand, was found predominantly in the heavier fractions, with smaller amounts in the CSF active fractions. To determine whether BARD1 and CstF were associated in these fractions, we again performed coimmunoprecipitation experiments. Given the interaction between BARD1 and BRCA1 (16,20), we also examined the possible presence of BRCA1 in CstF-containing complexes. We obtained similar results with all fractions that contained the three proteins; Fig. 3C presents results obtained with fraction 22. Strikingly, all three antibodies precipitated not only their cognate protein but also significant amounts of the other two. These findings indicate that the CstF50-BARD1 interaction occurs in the nuclei of HeLa cells and is likely to be physiologically relevant.

Figure 3

BARD1, BRCA1, and CstF associate in vivo. (A) BARD1 and CstF-50 coimmunoprecipitate from HeLa cell NE. Aliquots of NE (lane 1) were immunoprecipitated with anti-CstF64 (lanes 2 and 3), anti-BARD1 (lanes 4 and 5) or preimmune (lanes 6 and 7) antibodies. Equivalent amounts of the pellets (IP; lanes 2, 4, and 6) and supernatants (SUP; lanes 3, 5, and 7) were resolved by SDS-PAGE and proteins were detected by immunoblotting with anti-BARD1 or CstF-64 antibodies. Positions of BARD1 and CstF-64 are indicated. Asterisk indicates a species detected by the anti-BARD1 mAb. It can be separated from authentic BARD1 [see (B)], and its relationship to BARD1 is currently unknown (35). However, the fact that it can be coimmunoprecipitated not only with the anti-BARD1 polyclonal but also with the anti-CstF64 mAb and the anti-BRCA1 mAb (22) strongly suggests that it is related to BARD1. (B) Cofractionation of BARD1 and CstF. NE was subjected to 20% to 40% ammonium sulfate fractionation and Superose-6 gel filtration. Aliquots from NE and fractions 22 to 30 of the column were resolved by SDS-PAGE, and proteins were detected by immunoblotting with anti-BARD1 and anti-CstF64 antibodies. Positions of BARD1 and CstF-64 are indicated. Asterisk denotes the putative BARD1-related species [see (A)]. CSF activity was present in fractions 26 to 28. (C) Association of CstF, BARD1, and BRCA1 in a high molecular weight Superose-6 fraction. Aliquots of a fraction equivalent to fraction 22 in (B) were subjected to immunoprecipitation exactly as in (A), except that an anti-BRCA1 antibody was used in both the immunoprecipitation and immunoblotting (lanes 6 and 7) in addition to the anti-BARD1 (lanes 4 and 5) and anti-CstF64 (lanes 2 and 3) antibodies. An anti-p53 mAb was used as an irrelevant control (lanes 8 and 9). Positions of proteins are indicated on the left and sizes are on the right.

We next determined whether BARD1 affects polyadenylation in vitro. During preliminary experiments aimed at developing conditions for immunodepletion of NE, we observed that the cleavage step was enhanced by addition of an anti-BARD1 monoclonal antibody (mAb), EE6 (16), and we developed conditions that allowed this activation to be observed reproducibly (26) (Fig. 4A). We monitored cleavage of a simian virus 40 late pre-mRNA substrate in a limiting amount of NE that had been preincubated with no additions (lane 1) or with increasing amounts of anti-BARD1 (lanes 2 to 4) or anti-p53 (lanes 5 to 7). The anti-BARD1 mAb significantly and specifically enhanced cleavage, which suggests that BARD1 may antagonize 3′ end formation. Inclusion of increasing amounts of purified GST-BARD1 in the preincubation mixture prevented antibody activation and completely inhibited cleavage at the highest concentrations (Fig. 4B, lanes 1 to 5). GST-BARD1 also was sufficient to block cleavage (Fig. 4C, lanes 1 to 4). Essentially identical results, with both antibody activation and inhibition by GST-BARD1, were observed with a second pre-mRNA, adenovirus L3 (22). These results indicate that BARD1 inhibits the first step of mRNA 3′ end formation in vitro and may explain why the active CSF fractions from the Superose column contained relatively little BARD1 (Fig. 3B).

Figure 4

BARD1 inhibits pre-mRNA 3′ cleavage in vitro. (A) Activation of cleavage by anti-BARD1 antibodies. HeLa NE was preincubated with no addition (lane 1) or with increasing amounts (100, 200, or 300 ng) of a purified anti-BARD1 (lanes 2 to 4) or anti-p53 (lanes 5 to 7) mAb. After 15 min, pre-mRNA was added and incubation continued for 90 min. RNAs were purified and analyzed by denaturing PAGE. Positions of pre-mRNA and the 5′ cleavage product are indicated. (B) GST-BARD1 reverses activation by anti-BARD1 antibodies. NE was preincubated with no addition (lane 1); with 200 ng of anti-BARD1 (lanes 2 to 8); and with 50, 100, or 200 ng of GST-BARD1 (lanes 3 to 5) or 100 or 200 ng of GST (lanes 6 and 7). Pre-mRNA was added and reaction mixtures were processed as in (A). (C) Inhibition of 3′ processing by GST-BARD1. NE was preincubated with no additions (lanes 1 and 5) or with 50, 100, or 150 ng of GST-BARD1 (lanes 2 to 4) or GST (lanes 6 to 8). Pre-mRNA was added and reaction mixtures were processed as in (A). (D) Domain requirements for inhibition of 3′ processing by GST-BARD1. Processing reactions were performed as in (C), except that 50, 100, or 150 ng of the indicated GST-BARD1 derivative was used.

We next examined regions of BARD1 required for inhibition. The results, shown in Fig. 4D, correlate well with the binding data in Fig. 2C. GST-BARD1 (residues 13 to 638), which lacks the BRCT domain, inhibited processing, although somewhat less efficiently than did GST-BARD1 (compare lanes 2 to 4 with lanes 5 to 7). In contrast, the two derivatives lacking the ankyrin repeats were inactive (lanes 8 to 13). These results indicate that the same region of BARD1 required for binding CstF-50 is necessary for inhibiting 3′ pre-mRNA cleavage.

We suggest two related models in which the in vivo function of BARD1 is to prevent premature or incorrect 3′ end formation. The first stems from the previously documented interaction of CstF, as well as BRCA1, with the RNAP II holoenzyme (3, 27). The tight association between BRCA1 and BARD1 suggests that BARD1 may also be associated with RNAP II. Indeed, we found that in vitro translated BARD1 bound to a GST-CTD fusion protein with the same efficiency as CstF-50 (Fig. 2D, lanes 4 to 6) (3). BRCA1, in contrast, did not bind GST-CTD (22), and it is conceivable that the BARD1-CTD interaction contributes to the presence of BRCA1 in the holoenzyme. Our data suggest that, because of BARD1, the polyadenylation apparatus associated with the holoenzyme is inactive, which would prevent inappropriate processing of nascent transcripts—for example, at sites of RNAP II pausing or at cryptic or pseudo-poly(A) signals within the body of the transcript. By this view, the absence of BARD1 could lead to prematurely terminated transcripts and production of truncated, potentially deleterious proteins.

A related possibility is that BARD1 prevents premature polyadenylation of nascent transcripts associated with polymerases stalled at sites of DNA damage. Several observations support this hypothesis. First, BRCA1, likely with BARD1, colocalizes with Rad51 (28), a protein implicated in DNA recombination and repair (29). Rad51 has also been identified as a component of the RNAP II holoenzyme, supporting the link between transcription and DNA repair (30); genetic studies have also implicated BRCA1 in transcription-coupled repair of oxidative DNA damage (31). Second, BRCA1, BARD1, and Rad51 colocalize with the DNA replication and repair factor PCNA to sites of DNA damage during S phase (20). Remarkably, the strongest CstF-50–interacting protein isolated in our two-hybrid screen was PCNA (22). The interaction required the CstF-50 NH2-terminus, which contains a sequence similar to the consensus PCNA interaction motif (32), and was observed in vitro with GST-PCNA and in vitro translated CstF-50 (Fig. 2D, lanes 7 to 9; these data also show an interaction between BARD1 and PCNA, consistent with their known colocalization). Together, these results suggest a model in which BARD1, as part of the RNAP II holoenzyme, senses sites of DNA damage and repair, and the inhibitory interaction with CstF ensures that nascent RNAs are not erroneously polyadenylated at such sites.

  • * To whom correspondence should be addressed. E-mail: jlm2{at}


Stay Connected to Science

Navigate This Article