Structural basis for spectrin recognition by ankyrin

Abstract

Maintenance of membrane integrity and organization in the metazoan cell is accomplished through intracellular tethering of membrane proteins to an extensive, flexible protein network. Spectrin, the principal component of this network, is anchored to membrane proteins through the adaptor protein ankyrin. To elucidate the atomic basis for this interaction, we determined a crystal structure of human betaI-spectrin repeats 13 to 15 in complex with the ZU5-ANK domain of human ankyrin R. The structure reveals the role of repeats 14 to 15 in binding, the electrostatic and hydrophobic contributions along the interface, and the necessity for a particular orientation of the spectrin repeats. Using structural and biochemical data as a guide, we characterized the individual proteins and their interactions by binding and thermal stability analyses. In addition to validating the structural model, these data provide insight into the nature of some mutations associated with cell morphology defects, including those found in human diseases such as hereditary spherocytosis and elliptocytosis. Finally, analysis of the ZU5 domain suggests it is a versatile protein-protein interaction module with distinct interaction surfaces. The structure represents not only the first of a spectrin fragment in complex with its binding partner, but also that of an intermolecular complex involving a ZU5 domain.

Figure 1

Structure of the human β I-spectrin/ZU5-ankyrin R complex. (A) HEβ −1315 (blue) comprises 3 tandem canonical spectrin repeats (teal, blue, and dark blue from N- to C-terminus), each folded into a 3-helix bundle with the repeats connected by α-helical linkers. ZU5-ANK (gold), the spectrin binding domain of ankyrin R, maintains a compact β-sandwich fold connected by extended loops. The main interacting surfaces of the complex are formed by helices A and C of repeat 14 and the B/C loop of repeat 15 in β-spectrin and by the 2 strands from the β-core as well as 2 loops of ZU5-ANK. The 2 views are related by a 90° rotation. (B) Three different views of the spectrin/ankyrin interaction, that is, along the C-terminus of helix C of repeat 14 of spectrin (top), near the B/C loop of repeat 15 (middle), and near the N-terminus of helix C of repeat 14 (bottom).

Figure 2

Binding surfaces of the spectrin-ankyrin interaction reveal a bimodal interaction. In all panels the molecules of the complex are opened and rotated such that the interacting regions of both face the reader. (A) The spectrin residues involved in ankyrin binding are found principally along repeat 14. The interacting surfaces present many charged residues along the N-terminal portion of repeat 14, whereas more hydrophobic residues are localized near the spectrin B/C loop. (B) Surface footprints of spectrin (green/blue) and ankyrin (red/gold) illustrate the shape complementarity. The interacting residues in spectrin and ankyrin are shown in green and red, respectively. (C) The electrostatic surface of the molecules show significant charge interactions involving a negatively charged region along repeat 14 of spectrin and in a positively charged patch on the ankyrin fragment. The molecular surface of the molecules is shown with the equipotential electrostatic surface mapped onto them at ±15 k_bT/e_c with red corresponding to negative and blue corresponding to positive charges.

Figure 3

Clinical mutations in the ankyrin binding region of β-spectrin disrupt protein stability rather than binding affinity. (A) The characterized human β-spectrin mutants can be classified into 4 categories: those from clinical sources or functionally based experiments (orange), those along the helix A of repeat 14 interface (green), those along helix C of repeat 14 (blue), and those in the B/C loop of repeat 15 (purple). The ankyrin mutations coloring is assigned on the basis of the position of their spectrin binding partner where applicable. (B) Surface plasmon resonance measurements of the binding affinity (color-coded as in panel A) illustrate that structure-guided mutations disrupt binding significantly, whereas many clinical mutations do not. The inset depicts the fit of each mutant's sensorgram upon injection of ZU5-ANK at 200nM. (C) In contrast, a similar analysis of thermal stability demonstrated that clinical mutations have a significant destabilizing effect, whereas structure-based mutations do not. Bars indicate average values with error bars corresponding to the SE. The insert presents the circular dichroism thermal denaturation data (fraction unfolded vs temperature) and 2-state fits for each mutant.

Figure 4

Comparison of structural models of spectrin di-repeats suggest a binding mechanism aided by a specific bend angle between repeats 14 and 15. (A) Superposition of multiple structures of repeats 14 and 15 of β-spectrin^,, in the unbound state (gray) onto repeats 14 and 15 in the complex (blue) show marked structural similarities. In total, 5 models from 3 independent crystal structures of unbound β-spectrin repeats 14 and 15 were superimposed. The structures are virtually identical, showing the same relative orientation of the 2 repeats (RMSD values of ∼ 1.5 Å between all of the structures). The similarity among the structures suggests that the presence of a specific bending angle is important in recognition. (B) Alignment of 3 different spectrin di-repeat structures to repeat 14 of the complex demonstrate how the overall bend angle between β-spectrin repeats 14 and 15 (blue; HEβ1415) generates a close-fitting and matching interaction with ZU5-ANK (gold; ZU5-ANK). In contrast, the shallower bend angles seen in human erythroid β-spectrin repeats 8 and 9 (gray; HEβ89) and chicken brain α-spectrin repeats 15 and 16 (purple; CBα1516) would reduce the interacting surface or induce steric clashes, respectively. The dashed blue outline completing repeat 15 was modeled from the structure of brain β-spectrin repeats 14 to 16.

Figure 5

The ZU5 subdomain is capable of forming multiple, specific interactions with distinct surfaces. (A) Structure of the complex between spectrin (blue) and the ZU5-ANK domain of ankyrin (gold). (B) Structure of the cytoplasmic portion of the netrin receptor UNC5b (ZU5-2 domain in magenta, UPA domain in green, DD domain in purple) in the same orientation as panel A. (C) Superposition of the ZU5 domains of the 2 structures demonstrates that ZU5 domains have been adapted to bind multiple ligands with the use of different molecular surfaces. Note that recent sequence analysis of ankyrin R has revealed 2 tandem ZU5 domains: ZU5-ANK, which binds to spectrin, and a second ZU5 domain (ZU5-2). On the basis of this alignment, the ZU5 domain in the UNC5b structure correlates with the ZU5-2 domain of ankyrin. The superimposition of the models shown in panel C therefore does not reflect a predicted macromolecular assembly; it serves only to demonstrate that distinct interaction surfaces of ZU5 domains have now been observed.