Two-hybrid screen for BLM-interacting proteins
A YTH screen was performed to search for human proteins capable of interacting with the C-terminal 647 amino acids of BLM as a bait among >3 000 000 yeast transformants (Fig. A). Thirteen independent clones specifically interacting with the bait were isolated from a random-primed human peripheral blood cDNA library, among which a truncation (amino acids 198–756) of the human MMR protein MLH1 was found. The specificity of this interaction was also confirmed with full-length hMLH1 (amino acids 1–756) and by switching the hybrid partners: a fusion of hMLH1 to the LexA DNA-binding domain (LexAdbd) interacted with full-length BLM fused to the Gal4 activation domain (Gal4ad) (Fig. B).
Figure 1 YTH interactions. (A) Schematic representation of BLM. The two acidic domains (striped), the helicase domain (black), the HRDC domain (stippled) and the two putative nuclear localisation signals (arrow) are indicated. The portion of BLM used as bait (more ...)
Sgs1p, the S.cerevisiae RecQ homologue, interacts with yMlh1p
Since both RecQ helicases and MLH1 have been conserved throughout evolution, and given that RecQ helicase mutants in yeast and humans affect genomic stability, we asked whether Sgs1p, the S.cerevisiae homologue of BLM, interacts with yeast Mlh1p (yMlh1p) in the two-hybrid system. Although the C-terminal domains of Sgs1p and BLM share little sequence homology, the C-terminal domain (amino acids 784–1447) of Sgs1p was found to specifically interact with yMlh1p. Analogous to the BLM–hMLH1 interaction, the YTH interaction between Sgs1p and yMlh1p was confirmed for full-length yMlh1p and Sgs1p (Fig. C).
Different studies have shown that expression of the BLM
gene can partially complement both the hyper-recombination phenotype (31
) and the reduced lifespan of sgs1
). We wondered whether there is an interspecies interaction between BLM and MLH1, i.e. whether Sgs1p interacts with hMlh1p and BLM with yMlh1p. As shown in Figure C, we were able to detect an interspecies interaction via YTH between Sgs1p (amino acids 784–1447) and hMLH1 as well as between yMlh1 and BLM (Fig. C). The result was also confirmed for full-length Sgs1p after switching the hybrid partners (data not shown).
BLM and hMLH1 exist as a complex in human cells
Given that BLM and hMLH1 interact in the YTH assay, we wanted to test if BLM forms a complex with hMLH1 in human cells in vivo. To this end, co-immunoprecipitation experiments were performed, where an anti-MLH1 monoclonal antibody was used to precipitate its cognate protein from human nuclear cell extracts (Fig. A). BLM could be immunoprecipitated with hMLH1 from extracts of the MMR-proficient TK6 cells (lane 4), but not from extracts of hMLH1-deficient HCT116 cells (lane 3). In addition, the inverse co-immunoprecipitation experiment was carried out, in which an anti-BLM polyclonal antibody was used to immunoprecipitate hMLH1 from nuclear extracts of the human BJAB cell line. As seen in the Figure B, hMLH1 could be specifically co-immunoprecipitated with anti-BLM (lane 3), but not with the control IgG antibody (lane 2).
Figure 2 BLM and hMLH exist as a complex in human cells. (A) Co-immunoprecipitation of BLM with hMLH1. BLM could be immunoprecipitated with an anti-hMLH1 antibody from 200 µg nuclear extract of TK6 (wt) cells (lane 4), but not from HCT116 (BLM+ (more ...)
BLM and hMLH1 interact directly in vitro
We next wanted to determine whether the interaction between BLM and hMLH1 was direct, rather than being mediated via an accessory protein. Far western analysis was therefore performed to determine whether purified recombinant BLM and hMLH1 could interact directly in vitro
. To this end, full-length recombinant BLM protein (24
) was immobilised on a nitrocellulose filter, which was then incubated either with purified MutLα, a heterodimer of hMLH1 and hPMS2, or with extracts of Sf9 cells expressing either hMLH1 or hPMS2 (25
). The filter was then washed to remove the unbound proteins and the presence of BLM was detected using conventional western blotting with anti-hMLH1 and/or anti-hPMS2 antibodies. As controls, the membrane also contained BSA and BLM alone. Figure shows that BLM protein could be detected with antibodies against both hMLH1 (lane 2) and hPMS2 (lane 3) after incubation with the purified MutLα complex, but only hMLH1 alone (lane 5), and not hPMS2 (lane 7), could bind to BLM. The interaction between hMLH1 and BLM appeared to be specific, because MutLα and hMLH1 failed to bind the control protein, BSA (lanes 4 and 6), which was loaded on the same blot. To re-confirm these data, the reciprocal far western experiment was carried out, wherein human BLM protein was used to probe nitrocellulose-bound hMLH1. In this experiment, the anti-BLM antibody revealed a specific band at ~80 kDa, the position of migration of hMLH1 (lane 12). This band was due to BLM binding to hMLH1, as its position was identical to the specific hMLH1 band detected when purified MutLα complex was probed with anti-hMLH1 antibody (lane 9), and not to the band specific for PMS2 (lane 10). Again, the interaction between BLM and hMLH1 was specific, because no signal was detected with the control BSA protein (lane 11). We conclude that purified recombinant BLM and hMLH1 interact directly in vitro
and that BLM interaction with PMS2 is mediated via hMLH1.
Figure 3 BLM and hMLH interact directly. (A) Purified human MutLα complex (1.5 µg), BSA (1 µg) and BLM (1 µg) were subjected to SDS–PAGE and stained with Coomassie blue. (B) Far western analysis. The proteins were transferred (more ...)
Mapping of the BLM and hMLH1 interaction domains
To investigate which region of BLM protein was responsible for mediating the interaction with hMLH1, a series of BLM deletion mutants was generated and tested for their ability to interact with full-length hMLH1 in the YTH assay. As negative controls, the empty vector and a fragment of hMLH1 (amino acids 1–388) that did not bind to BLM were included. The results of these experiments indicated the presence of three independent hMLH1-interacting regions in BLM comprising residues 1–131, 448–572 and 1034–1417 (Fig. A).
Figure 4 Interaction domain mapping of BLM and hMLH1. (A and B) Yeast two-hybrid assays. The sequence boundaries of the deletion mutants tested in a β-galactosidase filter assay are shown with the corresponding amino acid positions indicated on the left. (more ...)
A similar approach was used to map the BLM interaction domain of hMLH1. A series of N- and C-terminal deletions of hMLH1 was generated and tested for interaction with full-length BLM (Fig. B). The results clearly indicated that the N-terminus of hMLH1 was dispensable for interaction with BLM, while the C-terminus (amino acids 396–756), a region similar to that involved in the interaction with hMLH3, hPMS1 and hPMS2 (33
), was essential. A deletion of as little as 60 amino acids from the C-terminus of hMLH1 was sufficient to destroy the interaction with BLM (Fig. B).
In addition, the YTH interaction domain mapping of both proteins was further confirmed using an in vitro binding approach, in which one of the interacting proteins was immobilised on a nitrocellulose membrane and probed with several 35S-labelled in vitro transcribed and translated deletion mutants of the other protein. Different in vitro transcribed and translated BLM fragments were incubated with membrane-bound purified recombinant MutLα heterodimer and BSA, included as a negative control (Fig. C). The weak binding of the BLM fragments might be due to the low accessibility of hMLH1 due to its heterodimerisation with hPMS2. Nevertheless, all the fragments containing one of the three domains identified in the YTH screen were able to bind to MutLα, but not to BSA, thus confirming the YTH mapping. Moreover, we could further narrow down the C-terminal interacting domain to a region between amino acids 1109 and 1378, as both C-terminal fragments tested clearly interacted with MutLα. In vitro transcribed and translated full-length hMLH1 as well as different deletions thereof were incubated with immobilised purified recombinant BLM and BSA (Fig. D). While the N-terminal half of the protein did not interact with BLM, the C-terminal half bound to BLM as strongly as the full-length protein. This interaction was completely abolished when the C-terminal half was further divided. Additionally, in vitro transcribed and translated hMLH1 was shown to bind to immobilised N-terminally truncated purified BLM protein (amino acids 212–1417; data not shown).
BLM and hMLH1 co-localise in the nucleus
The co-immunoprecipitation of BLM and hMLH1 from human cell extracts, as well as the evidence of a direct interaction between BLM and hMLH1, is consistent with these proteins forming a complex in vivo and in vitro. To provide additional evidence for the existence of this interaction, we wanted to see whether BLM and hMLH1 co-localise within the nucleus of intact human cells. Indirect immunofluorescence of exponentially growing human WI-38/VA-13 cells, using either anti-BLM or anti-hMLH1 antibodies, revealed BLM and hMLH1 to localise to prominent nuclear foci (Fig. ). Merging the fluorescent signals for BLM and hMLH1 showed a clear concordance in their localisation, thus strengthening the notion that the two proteins may function in a common biochemical pathway. A similar co-localisation pattern was obtained following aphidicolin treatment of the cells (data not shown).
Figure 5 Co-localisation of BLM and hMLH1 in the nucleus of WI-38/VA-13 cells. Indirect immunofluorescence of BLM (green) and hMLH1 (red) is shown in WI-38/VA-13 cells. The yellow colour results from overlap of the red and green foci. Nuclear (more ...)
BS cell lines are MMR proficient
No DNA helicase activities have so far been found to be associated with the MMR process. To address the question whether BLM is acting together with hMLH1 in MMR, the MMR proficiency of extracts of two lymphoblastoid (GM03403 and GM09960) and one fibroblast (GM08505) cell lines derived from BS patients were analysed. The absence of BLM protein in nuclear extracts of the BS cell lines was confirmed by western blot analysis, using antibodies raised against both the N- and C-termini of BLM (Fig. A and data not shown). Extracts derived from the lymphoblastoid TK6 and fibroblastoid MRC5 cells, HCT116 cells, and PSNF5 cells (GM08505 containing the BLM cDNA) were used as controls. As indicated in Figure B, all BS-derived nuclear extracts were MMR proficient using a substrate containing a single G·T mismatch and a strand discrimination signal (a nick) upstream (5′) from the mispair. The lower repair efficiency of the nuclear extract derived from the two lymphoblastoid cell lines GM03403 and GM09960, as compared to the MMR-proficient TK6 cells, is due to their high genomic instability and consequently to their reduced viability. Addition of recombinant purified BLM to the GM03403 or GM09960 cell extracts failed to influence the efficiency of MMR (Fig. B), which implies that the extracts had an intrinsically lower MMR capacity rather than being MMR deficient due to lack of BLM. Similar results were obtained using cytoplasmic extracts of the BLM cell lines and a template with the nick downstream (3′) from the mismatch (data not shown).
Figure 6 In vitro MMR efficiency of BS cell lines. (A) Western blot showing the absence of BLM protein in the BS cell lines. Aliquots of 25 µg of the indicated nuclear extracts were probed with anti-BLM antibody (IHIC33) and anti-XPB antibody [TFIIH (more ...)