Crystal structure of the nuclear effector of Notch signaling, CSL, bound to DNA

Abstract

Notch signaling is a conserved pathway of communication between neighboring cells that results in cell fate specification, and CSL is the universal transcriptional effector of Notch signaling. The Notch intracellular domain translocates to the nucleus after proteolytic release upon Notch extracellular engagement, and there it displaces corepressors from DNA-bound CSL and recruits activators of Notch target genes. Here we report the 2.85 A crystal structure of CSL with a target DNA. CSL comprises three structurally integrated domains: its amino (NTD)- and carboxy (CTD)-terminal domains are strikingly similar to those of Rel transcription factors, but a surprising beta-trefoil domain (BTD) is inserted between them. CSL-bound DNA is recognized specifically by conserved residues from NTD and BTD. A hydrophobic pocket on BTD is identified as the likely site of Notch interaction with CSL, which has functional implications for the mechanism of Notch signaling.

Figure 1

Sequence conservation of CSL core regions and Notch RAM domains. (A) Sequence alignment of CSL orthologs from C. elegans, Halocynthia, human, D. melanogaster, and Xenopus laevis. The domain boundaries are colored blue for NTD, green for BTD, and orange for CTD. Strand βC4, which bridges all three domains, is colored magenta. Identical residues are highlighted and colored according to domain location. Secondary structural elements derived from the refined structure are noted above the sequence, in which α=helix, β=strand, T=turn. Open circles above sequence alignment denote two disordered loop regions that are not observed in the electron-density map. Naming conventions for NTD and CTD secondary structural elements are consistent with the Rel proteins, and the BTD secondary structure naming uses the interleukin-1α convention. Residue side chains that make DNA contacts are denoted by colored triagonals below the sequence alignment, red for specific DNA interactions (Asn226, Arg234, Lys368, Gln401) and cyan for DNA backbone contacts. Sequences were aligned with CLUSTALW (Thompson et al, 1994) and the figure was created with ESPript (Gouet et al, 1999). Secondary structure was determined by the DSSP (Kabsch and Sander, 1983) implementation in ESPript and modified where necessary to conform to aforementioned conventions. (B) Sequence alignment of amino-terminal portions of NotchIC RAM domains from various organisms. Highlighted is the conserved nonpolar ΦWΦP motif (Φ=hydrophobic) that likely interacts with the hydrophobic pocket of BTD.

Figure 2

Ribbon representation of the CSL–DNA complex and domain organization. Orthogonal views are shown. NTD is colored blue, BTD green, and CTD orange. Secondary structural elements are labeled where visible. The boundaries of the main-chain Cα atoms that were not modeled in the structure, residues 262, 280, 304, and 319, are labeled. The piece of DNA used in the structural determination is shown and approximately aligned with the protein–DNA complex on the left. A schematic representation of the domain arrangements is also shown. The figure was created with Molscript (Kraulis, 1991) and Povscript (Fenn et al, 2003).

Figure 3

Structural comparison of CSL with the Rel proteins and interleukin-1α. Individual domains of CSL were submitted to the DALI server for structural similarity analysis. The figure was created with Bobscript (Esnouf, 1997). (A) Stereo diagram of the Cα overlay of the NTD of CSL (black) with the RHR-N domain of NFAT (blue) from the NFAT–Fos/Jun–DNA (1A02) complex. A total of 97 Cα atoms from CSL and NFAT were overlaid with an r.m.s.d. of 2.7 Å. (B) Stereo diagram of the Cα overlay of the BTD of CSL (black) with interleukin-1α (2ILA) colored in green. Strands B2 and B3 from interleukin-1α are colored red and represent the strands that are absent in the CSL beta-trefoil fold. A red sphere highlights residue Pro467 of CSL, which denotes where strands B2 and B3 would be inserted into the structure. For the structural comparison, 100 Cα atoms were overlaid with an r.m.s.d. of 2.2 Å. (C) Stereo diagram of the Cα overlay of the CTD of CSL (black) with the RHR-C domain of p65 (orange) from the NF-κB–IκB complex (1NFI). A total of 78 Cα atoms were overlaid with an r.m.s.d. of 2.1 Å.

Figure 4

Arrangement of CTD and RHR-C with respect to NTD and RHR-N. The orientation of CTD and RHR-C in relation to NTD and RHR-N and DNA were compared for CSL and Rel proteins. The NFAT/Fos-Jun/DNA (1A02) structure is depicted on the left, the p52 homodimer–DNA structure (1A3Q) is displayed on the right, and the CSL–DNA structure is in the middle. In each case, RHR-N or NTD is colored blue and RHR-C or CTD is orange. For the comparison, the NFAT and p52 RHR-N domains were overlaid with the NTD of CSL. The p52 protomer that is not used for the overlay is lightly shaded. The orientation of RHR-N and NTD with respect to the DNA is similar in all cases; however, the resulting orientation of the RHR-C domain is different in all three examples, and unlike the RHR-C domains of NFAT and p52, the CSL CTD does not contact DNA, nor is it involved in any protein–protein interactions that promote dimerization. The figure was created with Molscript and Povscript (Kraulis, 1991; Fenn et al, 2003).

Figure 5

Molecular surface and B-factor analysis. The molecular surfaces in panels A–D were calculated in GRASP (Nicholls et al, 1993) and rendered in Povscript (Fenn et al, 2003). (A) Electrostatic surface representation of CSL bound to DNA. Negative electrostatic potential is red, positive is blue, and white regions are neutral (nonpolar). A prominent region of positive charge on the protein interacts with DNA. (B) 180° view from (A) which shows the nonpolar surface, top left, located on a face of the BTD that is opposite of that used for DNA binding. The approximate location of the hydrophobic pocket is denoted by a black oval. (C) Sequence conservation is mapped to the molecular surface in a color gradient manner, such that dark red, orange, yellow, and white represent regions of absolute identity, high and moderate similarity, and regions of no conservation, respectively. The orientation is as in (A). The protein–DNA interface is entirely conserved. (D) View of mapping of (C) in the orientation of (B). The hydrophobic pocket is also highly conserved. (E) Worm representation of atomic mobility along the CSL backbone. The view and orientation is similar to (B). BTD is colored green, NTD blue, and CTD orange. A schematic representation of the DNA backbone is colored gray. Crystallographic B-factors from the Cα backbone of the refined structure represent atomic mobility and are mapped to the backbone worm with a worm radius proportional to the B-factor; regions of low mobility have a thinner backbone worm and regions of high mobility have a thicker backbone worm. Overall B-factors are higher in the BTD than in the NTD and the CTD domains. The figure was created with SPOCK (Christopher, 1998).

Figure 6

BTD analysis. (A) RAM peptide binding to the isolated BTDs analyzed by native gel electrophoresis. The Lin-12 RAM peptide (RMINASVWMPPME) and the reverse-sequence control peptide (EMPPMWVSANIMR) were incubated with the isolated BTDs from C. elegans Lag-1 (375–575) and murine RBP-Jκ (203–393). BTD gel mobilities increase for the RAM peptide/BTD complexes (lanes 2, 4 and 6, 8) relative to the peptide-free BTDs (lanes 1 and 5). BTD mobilities are unchanged when only in the presence of the control peptides (lanes 3 and 7). The gel shift is less with RBP-Jκ because this protein has higher net negative charge than Lag-1. (B) Previously reported CSL mutations are mapped onto the Cα-trace of BTD. The view is similar to (C). Mutations in residues that affect specifically NotchIC RAM domain interaction are colored purple. These include murine CSL mutations KV212GS (KV391GS), RK291GS (RK472GS), FY314GS (FQ498GS), R218H (R397H), and ΔD254 (ΔE433) (Sakai et al, 1998; Fuchs et al, 2001). Mutations in residues that affect NotchIC interaction and interactions with the corepressors SMRT or CIR are colored gold. These include human CSL mutations EEF233AAA (DNF440AAA) and KLV249AAA (KLV456AAA), and murine mutations F261L (F442L), K275M (K456M), and A284V (A465V) (Hsieh et al, 1999; Fuchs et al, 2001). The original mutations and numbering are denoted with the equivalent residues from Lag-1 in parentheses. Overall, the mutations roughly map to two regions on BTD. The first region is proximal to the hydrophobic pocket. This region is populated by purple residues that affect NotchIC RAM interaction but not corepressor interaction. The second region maps to areas adjacent to the hydrophobic pocket and to an extended loop of BTD. This region is populated by gold-colored residues that affect both NotchIC RAM domain and corepressor interactions. The figure was created with Povscript (Fenn et al, 2003). (C) BTD surface of molecular curvature, colored green for regions of positive curvature and gray for regions of negative curvature. A distinct pocket is visible in the region where strands B2 and B3 would be in a prototypical beta-trefoil fold, and is denoted by a black oval. The figure was created with GRASP and Povscript (Nicholls et al, 1993; Fenn et al, 2003).

Figure 7

CSL–DNA interactions. (B–D) were created with Molscript and Povscript (Kraulis, 1991; Fenn et al, 2003). (A) Schematic representation of all protein–DNA interactions in the CSL–DNA complex. Specific interactions with the DNA bases are shaded in gray and nonspecific interactions are clear boxes. Hydrogen-bonding or salt-bridge interactions are denoted as an arrow and Van der Waals interactions are depicted as closed circles. (B) Stereo view of the major-groove protein–DNA interactions of Arg234 and Asn26 from CSL with Gua8 and Gua9. The DNA is colored atom-specifically: C yellow, N blue, O red, and P gray. The protein is colored blue with interacting loops solid and other parts transparent. (C) Stereo view of the interaction of Lys368 with Gua10 and Thy7′. Coloring is as in (B). (D) Stereo view of the minor-groove interactions of the side chain of Gln401 with Ade13′ and the backbone carbonyl of Ser400 with Gua6. Coloring is as in (B).