|Home | About | Journals | Submit | Contact Us | Français|
Multidrug resistance protein 1 (ABCC1) is a member of the ‘C’ class of ATP-binding cassette transporters, which can give rise to resistance to chemotherapy via drug export from cells. It also acts as a leukotriene C4 transporter, and hence has a role in adaptive immune response. Most C-class members have an additional NH2-terminal transmembrane domain versus other ATP-binding cassette transporters, but little is known about the structure and role of this domain. Using electron cryomicroscopy of 2D crystals, data at 1/6 per Å−1 resolution was generated for the full-length ABCC1 protein in the absence of ATP. Analysis using homologous structures from bacteria and mammals allowed the core transmembrane domains to be localised in the map. These display an inward-facing conformation and there is a noteworthy separation of the cytoplasmic nucleotide-binding domains. Examination of non-core features in the map suggests that the additional NH2-terminal domain has extensive contacts on one side of both core domains, and mirrors their inward-facing configuration in the absence of nucleotide.
Overexpression of several ATP-binding cassette (ABC) transporters can give rise to the phenotype of multidrug resistance (MDR) in cancer cells (Szakacs et al., 2004, 2006). One of these is multidrug resistance protein 1 (MRP1/ABCC1) (Cole et al., 1992, 1994; Deeley and Cole, 2006; Haimeur et al., 2004). Clinical studies have shown that MRP1 is expressed in a range of solid and haematological malignancies and MRP1 expression has been correlated with negative response to treatment and disease outcome (Deeley and Cole, 2006). MRP1 is expressed in most tissues throughout the body; in polarized cells, MRP1 is found on the basolateral cellular surface. It is thought to provide protection from xenobiotics in these tissues.
Gene transfection studies have confirmed the role of MRP1 in conferring resistance to natural product chemotherapeutic agents and genotoxins by mediating cellular efflux of these substrates (Cole et al., 1994; Peklak-Scott et al., 2005). MRP1 also transports organic anions conjugated to glutathione (GSH) (e.g., cysteinyl leukotriene C4) (Loe et al., 1996) as well as glucuronate and sulfate conjugates (Leslie et al., 2005). Studies on mrp1−/− mice suggest that the transport of leukotriene C4 by MRP1 may play a significant role in the mediation of inflammatory responses (Mueller et al., 2009). MRP1 can also transport unconjugated organic anions such as methotrexate and GSH (Cole and Deeley, 2006). GSH has been shown to stimulate the transport of some conjugated compounds, and a GSH-induced change in the conformation of the transporter may be involved (Rothnie et al., 2006).
ABC transporters typically consist of two transmembrane domains (TMDs), each composed of six predicted transmembrane α-helices, and two nucleotide-binding domains (NBDs) (Hyde et al., 1990). The TMDs probably form the translocation pathway for substrates and appear to be major determinants of substrate specificity. Unlike most ABC transporters, MRP1, contains three rather than two TMDs (Bakos et al., 1996; Cole et al., 1992; Haimeur et al., 2004; Hipfner et al., 1999) as do several other ABCCC family members (Deeley and Cole, 2006). The additional TMD (TMD0) is predicted to contain five transmembrane α-helices and a glycosylated extracellular N-terminal region. TMD0 is connected to TMD1 by a large cytoplasmic loop of ~120 amino acids.
In the ABC transporters, which work against a concentration gradient, transport is coupled to the binding and hydrolysis of ATP by the NBDs, which appear to undergo a reversible dimerisation during the transport cycle. Accumulating structural data from X-ray crystallography of bacterial ABC transporters are mostly consistent with such a transport model (Dawson and Locher, 2007; Gerber et al., 2008; Hollenstein et al., 2007; Khare et al., 2009; Locher, 2004; Oldham et al., 2008).
So far, structural data for MRP1 has been limited and of relatively low resolution (22 Å), and consequently, there was no clear resolution of the location of TMD0 in these studies (Rosenberg et al., 2001). The TMD0 region, including the extended cytoplasmic loop, displays no homology with any other structure deposited in the protein data bank. Hence homology models may be constructed only for the core four domains of the protein (DeGorter et al., 2008). Here, we have determined the 3D structure of MRP1 by electron cryomicroscopy of 2D crystals with an in-plane resolution of 1/6 per Å−1 and with a missing cone of data of only 26° giving a resolution perpendicular to the crystal plane of about 1/10 per Å−1. At this combined resolution, long transmembrane α-helices may be resolved as cylindrical regions of high density in the map (Ford and Holzenburg, 2008).
Wild-type MRP1 with hemagglutinin and hexa-histidine tags at the COOH-terminus was expressed in Pichia pastoris as described previously (Cai et al., 2001; Wu et al., 2005). A construct lacking TMD0 (MRP1204–1531) was generated and expressed in a similar manner. Membranes were solubilized in LPG, then diluted in DDM and subjected to a two-step purification procedure utilizing Co2+-IMAC resin and DE52 anion chromatography (Wu et al., 2005). Protein purity was judged to be greater than 90% by silver staining (Fig. 1A).
To determine whether the MRP1 was active after purification, its ability to bind the high affinity MRP1 substrate ([3H]LTC4) and nucleotide (8-azido-[α-32P]ATP), as well as its ATPase activity, were determined. As shown in Fig. 1B, DDM-solubilized, and purified MRP1 as well as MRP1204–1531 were readily photolabeled with [3H]LTC4. Purified MRP1 constructs were also photolabeled with 8-azido-[α-32P]ATP (Fig. 1C). No significant photolabeling of other protein bands was detected. Finally, the ATPase activity was determined for both constructs (Fig. 1D). These observations indicate that the full-length MRP1 protein used for the structural analysis described below had retained its function, and also that removal of the N-terminal TMD0 domain had no significant impact on the activity of the purified material.
2D crystals were generated on the carbon surface of an electron microscope (EM) grid (Auer et al., 1999). 2-D crystals were usually observed in the square grid windows in a peripheral zone ~2 µm across which was visible at low magnification as shown in Fig. 2A (arrows). This zone presumably reflects optimal conditions for 2D crystal formation because of a local gradient of protein/precipitant within each window of the grid. A mosaic of 2D crystals is found in this zone, suggesting that individual crystals (typically 0.5–2 µm across) nucleate in this area and grow outwards until meeting another area. At strong defocus, the unstained 2-D crystals could be identified (Fig. 2B, arrows). A similar method of 2-D crystal formation was used by Auer et al. (1999) and successes have been achieved for other ABC transporters and other membrane proteins using this method (Auer et al., 1998; Awayn et al., 2005; Rosenberg et al., 2004, 2005; Scarborough, 1994). The MRP1 2D crystals show a p1 two-dimensional plane group. Other plane group symmetries were discounted by examining phase relationships for untilted 2D crystals using the program ALLSPACE (Valpuesta et al., 1994) (see Table 1), as well as by comparison of the unit cell dimensions of the 2D crystals versus the dimensions of other ABC proteins.
Electron cryomicroscopy of individual crystals of MRP1 yielded well-sampled structure factors to about 1/6 per Å−1 resolution in the plane of the crystal and 1/10 per Å−1 resolution perpendicular to the plane (Supplementary data, Fig. 1A). After merging data from several untilted crystals, the interimage phase errors of multiply-sampled reflections with an arbitrary cut-off of 6 Å resolution were plotted (Supplementary data, Fig. 1B). The well-sampled data with low phase errors extend to approximately 10 Å resolution, whilst data to about 6 Å resolution are sampled less completely, and with larger phase errors. The overall interimage phase residual for the merging of the MRP1 data was 38° (Table 1), where a phase residual of 90° would correspond to random data. The distribution of amplitudes and phases along lattice lines showed the data extending to about 10 Å resolution perpendicular to the crystal plane (Supplementary data, Fig. 2A–D).
The 3D map generated from the MRP1 2D crystals is displayed in Fig. 3. At a high density threshold (2.0σ above the mean density level, left panel), a 50 Å – thick band of density is observed that is composed of closely packed cylindrical features, probably transmembrane α-helices. Below, and above this band of high density are the cytoplasmic and extracellular regions of the protein, which are ~60 Å and ~15 Å wide respectively. Two orthogonal views of the map, with a lower density threshold (1.25σ above the mean density level) are shown in Fig. 3, panels b and c. The map has been coloured according to the assignments discussed later: Yellow corresponds to regions assigned to TMDs 1 and 2; orange TMD 0; blue and purple NBDs and green to unassigned regions, mostly on the upper (extracellular) surface of the map. The NBDs of MRP1 are considerably further apart than in the ‘sandwich dimer’ configuration, consistent with the nucleotide-free conditions for crystal production (Aller et al., 2009; Hollenstein and Dawson, 2007; Oldham et al., 2008). A central slice through the map (Fig. 3d) provides evidence for an inward-facing configuration of the TMDs. The two halves of the transporter appear to lean towards each other, with an angle of approximately 20° (arrows). This angle is lower than that observed for the Escherichia coli MsbA structural model (Ward et al., 2007) but is similar to that displayed by the inward-facing murine P-glycoprotein structure (Aller et al., 2009).
Structures of homologues of MRP1 (Sav1866 and P-glycoprotein) were firstly compared to, and docked within the MRP1 map using the Chimera ‘fit model to map’, and Situs ‘cores’ routines. Secondly, cylindrical regions of continuous density were traced by hand. Thirdly, NBDs were initially fitted by hand as a dimer, and then the local position of each NBD was individually refined as above.
The overall positions of the docked Sav1866 and P-glycoprotein structures within the MRP1 map were similar (data for Sav1866 shown in Supplementary data, Fig. 3), but the sandwich dimer of the Sav1866 NBDs was clearly different to the MRP1 map. The result of the fitting of the nucleotide-free P-glycoprotein structure is displayed in Fig. 4, with the MRP1 map viewed as in Fig. 3c. A good fit was apparent for the C-terminal half of P-glycoprotein: The rightwards slope of the transmembrane helices was reproduced in the MRP1 map, as was the positioning of the ‘L’ shaped NBD(2). At the centre of the map (Fig. 4b), the pseudo-twofold symmetry of the P-glycoprotein structure is matched as is the curved ‘V’ shape formed by TMD helices 6 and 12 of P-glycoprotein. On the cytoplasmic side, only one of the P-glycoprotein NBDs lies well within the MRP1 map. The N-terminal P-glycoprotein NBD (arrows) does not, as further shown by the rearmost slice through the map (Fig. 4c). A shift and rotation of the N-terminal NBD would be required to place it within the MRP1 map. The leftwards tilt of the TMD helices in P-glycoprotein is matched in the rearmost slice of the MRP1 map (Fig. 4c), and the extension of density into the extracellular region (red ellipse) probably explains why the N-terminal half of P-glycoprotein is favoured for this region of the map. The fitting exercise gave an indication of the regions containing density for TMD0 (Fig. 4, white dashed outlines).
The manual tracing of cylindrical regions of density (red cylinders) in the MRP1 map (blue/yellow mesh) are displayed in Fig. 5. The position of the fitted P-glycoprotein structure is superimposed (purple ribbon trace). Panel a shows a view onto the TMDs from the extracellular surface. A cluster of cylinders (above the dashed line) shows some twofold symmetry as indicated by the triangular and oval dashed outlines, and where the curved arrows indicates the opposing directions of tilt of cylinders on each side of this region (see also Fig. 3a and c). This region of twofold symmetry roughly coincides with the fitted P-glycoprotein TMDs (ribbon trace). On the periphery of the diamond-shaped region (below the dashed line) lie several further cylindrical regions of density. Four of these (A–E) are candidates for TMD0 helices. A side-view of this peripheral region (Fig. 5b) reveals that two (C and D) extend into the extracellular region, whilst a fifth (B) may extend partially into the lipid bilayer (boundaries of which, as estimated by the fitted P-glycoprotein model, are indicated by the dashed lines, Fig. 5b). Hydropathy plots suggest that TMD0 will have a glycosylated N-terminal region of ~36 residues, and a third extracellular loop of ~18 residues. Interestingly, the fitted cylinders in the putative TMD0 region mirror the inverted ‘V’ shape of the rest of the TMD region, with the A,C pair leaning inwards to contact the D,E pair at the extracellular surface. A lozenge-shaped region of density (dashed ellipse) sits directly below the D,E cylinders in the cytoplasmic region of the map (L1) and likewise, a cylindrical region (L2) extends downwards from the vicinity of the A,B,C cylinders.
Fitting of individual NBDs to the map (Fig. 5, panels c–e), yielded positions and orientations that were roughly consistent with previous ABC transporter structures (Aller et al., 2009; Hollenstein et al., 2007; Khare et al., 2009; Kadaba et al., 2008). One of the fitted NBDs was close to the C-terminal NBD P-glycoprotein model (panels d,e and Fig. 4), but the other was located differently, as anticipated (see Fig. 4c). The overall impression gained was of asymmetry, a 45° clockwise rotation of the latter NBD around the viewing direction in panel e.
Studies of MRP1 have shown that residues in the TMDs 1 and 2 are important for binding and transport of a number of MRP1 substrates (Campbell et al., 2004; Conseil et al., 2006; Deeley and Cole, 2006; Haimeur et al., 2002, 2004; Zhang et al., 2004). Partial deletions and mutations of specific amino acids in TMD0 can interfere with protein processing and activity (Gao et al., 1998; Ito et al., 2003; Leslie et al., 2003). These latter mutations may interfere with folding of the protein, since complete removal of TMD0 does not prevent trafficking to its normal location nor transport action (Bakos et al., 1998; Westlake et al., 2003). However TMD0 and a small portion of the cytoplasmic loop connecting it to TMD1 are required for retention of MRP1 at the cell surface, and hence they contain at least partially redundant elements that contribute to the processing and trafficking of the protein (Westlake et al., 2005). MRP1 lacking TMD0 has similar ATPase and LTC4-binding activities to the full-length protein (Fig. 1). Previous structural studies of MRP1 in detergent showed monomers of 8–10 nm diameter and with a favoured projection of roughly pentagonal shape (Rosenberg et al., 2001). Interestingly, after reconstitution into lipid bilayers, the protein produced small and poorly ordered 2D crystals with a p2 plane group symmetry (Rosenberg et al., 2001). The (better ordered) 2D crystals in this study were generated with the monomeric detergent-solubilised protein, although the overall dimensions of the unit cell (12 × 8 × 7 nm) are roughly consistent with the previous studies (Rosenberg et al., 2001). Moreover, viewed along its long axis, the MRP1 map shows a roughly pentagonal outline, again consistent with the earlier low resolution data for the detergent-solubilised protein.
Long transmembrane α-helices should give rise to discernible density in the present MRP1 map. MRP1 likely has 17 transmembrane α-helices, five in TMD0 and six each in TMDs 1 and 2. The region of high density in the MRP1 map (see Fig. 3a) contains tightly packed cylindrical densities, consistent with this number of transmembrane α-helices. The TMDs of MRP1 seem to be in an inward-facing conformation and there is a concomitant separation of the NBDs. This is in agreement with the nucleotide-free P-glycoprotein structure (Aller et al., 2009) as well as for MetNI, Mal- FGK and ModBC in the nucleotide-free states (Gerber et al., 2008; Kadaba et al., 2008; Khare et al., 2009; Ward et al., 2007). The unusual relative orientation of one of the NBDs in MRP1 may reflect the relative specialisation of the two NBDs in MRP1 (Gao et al., 2000). A more prosaic explanation could be that the latter NBD is partly disordered in the crystals. Neither explanation will be unprecedented (the E. coli MsbA structure has only partial density for the NBDs and the V. cholera MsbA has an unusual rotation of the NBDs away from the usually adopted configuration such as that found for S. typhimurium MsbA) (Gerber et al., 2008; Kadaba et al., 2008; Khare et al., 2009; Ward et al., 2007).
Additional density in the TMD region could be the location of TMD0, on one side of the core ABC transporter structure. Unequivocal assignments as to which segment of TMD0 interact with which segments in TMDs 1 and 2 are not possible. If the fit shown in Fig. 4 is correct, and the core TMDs of MRP1 are similar to other ABC transporters, then TMD0 would interact with TM helices 14,15 and 16 in TMD2 and 7 in TMD1 (see also Supplementary data, Fig. 4). TMD0 would come into contact with helices crossing over from one side of the protein to the other and domain-swapped intracytoplasmic loops. The putative TMD0 region emulates the inward-facing core TMDs. Hence, TMD0 may participate in large scale conformational changes. Whether TMD0 is an active participant, or passive passenger, is debatable. Evidence from TMD0-lacking MRP1 suggests a passive role in the mechanism of action is more likely.
TMD0 not only consists of transmembrane helical segments, but also contains cytoplasmic and extracellular loops. Extracellular loops in ABC proteins are generally short connecting turns, but longer loops (>5 residues) have been found to be simple extensions of membrane-spanning helices (e.g. in P-glycoprotein (Aller et al., 2009)) or in one example, a separate domain (MalF) (Oldham et al., 2007). The data presented here suggest that the TMD0 extracellular loops are more likely to be extensions of the membrane-spanning helices or short turns (Fig. 5b, top). In contrast, the cytoplasmic region close to the putative TMD0 helices displays a larger lozenge-shaped density, L1 and a smaller cylindrical density L2 (Fig. 5b, bottom). TMD0 is connected to TMD1 in MRP1 by a cytoplasmic loop of ~120 residues. This loop displays no significant homology to any known protein structures, and may be large enough to form a discrete domain. We would propose that the L1 density in the MRP1 map could be the site of such a domain.
The plasmid pHIL-MRP1cHA-His6 for expression of full-length MRP1 in Pichia pastoris has been described previously (Cai et al., 2001; Wu et al., 2005). To express an NH2-terminally truncated MRP1 lacking TMD0, a plasmid containing cDNA coding for MRP1 amino acids 204–1531 was created as follows. pHIL-MRP1cHA-His6 was digested with SacII and BamHI, and the resulting vector fragment containing MRP1 cDNA coding for amino acids 281–1531 was gel purified. A MRP1 cDNA fragment coding for amino acids 204–300 was then generated by PCR using the plasmid pcDNA3.1(−)MRP1K as template (Leslie et al., 2003). The upstream primer (5′-AGG CCG CGG AAA AAA ATG GAC CCT AAT CCC TGC CCA-3′) contained a SacII site immediately followed by a spacer of six adenosines, an ATG start codon, and a sequence coding for six amino acids starting from position 204. The downstream primer (5′-GAC GAT CAA AGC CTC CAC C-3′) contained a DNA sequence at a position 42 nucleotides beyond the BamHI site in the MRP1 cDNA. The PCR product (approximately 0.3 kb) was digested with SacII and BamHI, gel purified, and ligated to the pHIL vector fragment containing sequence encoding residues 281–1531. The fidelity of the PCR product was confirmed by sequencing. The resulting construct was designated pHIL-MRP1204–1531cHA-His6.
Crude membranes from cells expressing either plasmid pHIL-MRP1cHA-His6 or pHIL-MRP1204–1531cHA-His6 were prepared using a French Pressure Cell (Thermo Electron Corporation, Waltham, MA). Recombinant MRP1 and MRP1204–1531 proteins were solubilized with LPG (4–6 mg ml−1) and then diluted with DDM (Sigma–Aldrich) to a final concentration of 0.02%. The solubilized proteins were purified by Co2+-IMAC (BD BioSciences Clontech) and DE52 anion chromatography. Protein purity and identification were confirmed by silver staining and by immunoblotting with MRP1-specific monoclonal antibody QCRL-1 (Cai et al., 2001) as before.
Purified MRP1 and MRP1204–1531 were photolabeled with 8-azido-[α-32P]ATP and [3H]LTC4. For 32P-azidoATP labelling, purified MRP1 and ‘short’ MRP1204–1531 (0.8–1.5 µg) in buffer containing 50 mM Tris–HCl, 250 mM sucrose, 0.187 mg ml−1 DDM, pH 7.4, were incubated with 5 µM (2.3 µCi) 8-azido-[α-32P]ATP (Affinity Labeling Technologies, Inc., Lexington, KY) and 5 mM MgCl2 on ice for 30 min and then irradiated at 302 nm for 8 min (CL-1000 Ultraviolet Crosslinker, DiaMed). Protein samples (Conseil and Deeley and Cole, 2006) were subjected to SDS–PAGE and the dried gels were exposed to film. For photolabeling with the high affinity MRP1 substrate LTC4, purified MRP1 and MRP1204–1531 (0.8–1.5 µg) were incubated with [3H]LTC4 (200 nM; 0.13–0.15 µCi) (Perkin-Elmer Life Sciences) and irradiated at 302 nm as before (25,41). [3H]LTC4-labeled proteins were then resolved by SDS–PAGE and processed for autoradiography. The ATPase activity of purified MRP1 and MRP1204–1531 was determined using a colorimetric assay essentially as described (Mao et al., 2000). Briefly, 0.25–0.75 µg of MRP1 and MRP1204–1531 were incubated at 37 °C in a reaction mixture containing 50 mM Tris–HCl (pH 7.4), 10% glycerol, 0.02% DDM, 5 mM ATP, and 5 mM MgCl2. Reactions were stopped at various times over a 5 h period by the addition of 18% SDS. The amount of inorganic phosphate released was determined immediately using a colorimetric method adapted from Chifflet et al. (1988) and Doige et al., 1992. Data were corrected for ATP hydrolysis in the absence of protein.
2D crystals were grown on carbon films, supported on gold-plated copper 400 mesh electron microscope grids (Agar) from a solution containing 0.2 mg ml−1 MRP1, 2.8 mg ml−1 dodecymaltoside, 110 mM ammonium sulphate and 17% (w/v) PEG 4000 (Hampton Research) with the pH being adjusted to 8.0 with 50 mM Tris–HCl. Carbon-coated grids were heated to 160 °C for 1 h prior to the experiment. A 10 µl drop was placed in the well of a Hampton Research plate (Crystal Quick) equilibrated against 150 µl of 1M MgCl2. The grid was placed onto the drop with the grid bars facing the solution. The plate was sealed with Crystal Clear® film (Hampton Research) and left at 4 °C for 17 h in an incubator (Rumed, Jencons).
Grids were picked up from the surface of the drops and either negatively stained with 2% (w/v) uranyl-acetate (Rosenberg et al., 2003) and viewed on a Tecnai 10 electron microscope, or blotted for five to ten seconds with filter paper (Whatman No. 4) and then frozen in liquid ethane using an FEI Vitrobot. The grids were transferred at liquid-nitrogen temperature into a Gatan cryotransfer specimen holder (Gatan, Pleasanton, CA) and observed at 95 K in a Philips CM200 FEG or a Tecnai 20 FEG operated at 200 kV and equipped with a Gatan 4096 × 4096 CCD camera with a pixel size of 15 µm. Images were collected at a magnification of 50,000 and 80,000×, and with a calibrated electron dose of 10–12 e/Å2. Some images were recorded on Kodak SO-163 film and developed for 12 min in a full-strength Kodak D19 developer. Selected films were subjected to densitometry using a Zeiss SCAI scanner set at 7 µm pixel size, corresponding to 1.14 Å at the specimen level. We did not observe any significant difference in the quality of the data recorded on film or on the CCD camera. Crystals were located in a fringe region around the edge of the grid windows (see Fig. 2A). Images were recorded with defocus varying between −0.2 and −3.0 µm, the mean defocus employed was −0.5 µm.
Images were processed using the MRC-LMB image processing suite. After three cycles of lattice unbending to correct for lattice distortions (Crowther et al., 1996), correction of the contrast transfer function was carried out after determining defocus with CTF-FIND2 (Henderson et al., 1986). Structure factors were merged by refining to a common phase origin using ORIGTILTK. Amplitudes and phases were averaged as described previously (Henderson et al., 1986). The image amplitudes were scaled as a function of resolution, with bacteriorhodopsin as a reference. Density maps were calculated with CCP4 software (Collaborative Computational Project No. 4, 1994) as described previously (Rosenberg et al., 2005).
Homologues of MRP1 (Sav1866 and P-glycoprotein) were docked within the MRP1 map using the Chimera ‘fit model to map’, and Situs ‘cores’ routines (Pettersen et al., 2004; Wriggers et al., 1999), yielding similar positions for each algorithm. Docking of individual NBDs within the MRP1 map was carried out using an initial search with the widely-separated MetNI NBD dimer (with regulatory domains removed) (Kadaba et al., 2008). The position of each NBD was then individually refined over a smaller search space as above (Pettersen et al., 2004). A search using individual NBDs versus the entire MRP1 map did not yield a consistent docking position, probably because of the relatively small size and roughly globular shape of a single NBD relative to the map and the limited resolution of the latter.
We thank Drs. Neil Ranson and Peiyi Wang for facilitating access to electron microscopes at the Universities of Leeds and Sheffield, UK; and Dr. William Nicholson (University of Leeds) and Mr. Simon Oliver (University of Manchester) for assistance with software. We would also like to thank Drs. Daniel Levy and Jean-Louis Rigaud (Institut Curie, Paris) for microscopy assistance and the British Council for a grant to MFR to visit their laboratory). The work is supported by the Leukaemia Research Fund (London, UK) and the Canadian Institutes of Health Research (Grant MOP-10519). SPCC is a Canada Research Chair in Cancer Biology and Bracken Chair in Genetics and Molecular Medicine.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.jsb.2010.01.011.