In this study, we have described the structure of lysenin, a sphingomyelin-specific pore-forming protein, in its apo form and bound to POC and SM. By structural analysis we have shown that lysenin is related to pore-forming proteins from across the biosphere. It is particularly striking that the homologous N- and C-terminal domains of fungal LSL (Mancheño et al., 2005
) and annelid lysenin are in alternative positions in their respective sequences. Therefore, during their divergent evolution from a common ancestral protein, there has most likely been a genetic swap of the two domains with respect to each other. In addition, the SM-binding region of lysenin is an added edge to the PFM sheet compared to other pore-forming proteins.
We suggest that lysenin interacts with the membrane initially by binding of POC in lipids such as SM through its β-trefoil, after attraction to the membrane surface through charge-charge affinity; it will then bind the full length of SM. The insight that lysenin interacts with membranes in a two-stage process is strengthened by previously published work: removal of the β-trefoil domain reduces the protein’s affinity 100-fold (Kiyokawa et al., 2005
), showing that the C-terminal domain is needed in membrane binding even though it is the N-terminal PFM that binds SM specifically. The positively-charged patch on the C-terminal domain would make for interactions with negative charges at the membrane surface, as for example found at in the case of the sulphates in proteoglycans such as heparin sulphate and chondroitin sulphate, and may help in attracting or guiding the approach of lysenin to the membrane surface.
The structure of the SM/lysenin molecular complex represents, to our knowledge, the first crystal structure showing a direct and specific SM/protein contact, where not only the headgroup but also the acyl chain of the lipid is recognized simultaneously. The interaction with the acyl chains of SM by lysenin was already shown by differential scanning calorimetry experiments (Yamaji-Hasegawa et al., 2003
). The SM/protein complex is also in agreement with recently published results on the molecular recognition of sphingolipids by the protein transmembrane domain (TMD) of COPI (coat protein) machinery protein p24 (Contreras et al., 2012
) where Förster resonance energy transfer (FRET), alanine scanning and molecular dynamic demonstrated a direct and highly specific interaction of sphingomyelin species with the TMD. Strikingly the interaction depends on both the head-group and the backbone of the sphingolipid, as in our structure, and on the presence of a signature sequence (VXXTLXXIY) within the TMD. The acyl chain of SM appears to pack in the groove between Val13, Thr16, and Leu17 of the p24 TMD. One acyl chain of SM also occupies a groove on the lysenin PFM domain directly interacting with tyrosines 24 and 26, although it is defined by β strands and not as in p24 α helices.
To date, interaction with SM has been documented in actinoporins, which are SM-dependent pore-forming toxins from sea anemones, only via the SM headgroup and chemical moieties immediately beneath it (Mancheño et al., 2003
; Bakrac et al., 2008
). SM occupies a similar position to the lipids found in the aquaporin-0 2D crystal structure (Hite et al., 2010
), and Kir2.2 potassium channel (Hansen et al., 2011
). Thus, like the cholesterol-dependent cytolysins (CDCs), in lysenin specific lipid/protein interactions lead to membrane disruption as a function of both the protein inserting into the membrane and the lipid reorganization induced (Gilbert, 2010
). We suggest that binding of SM in one leaflet of the targeted membrane would result in its reconfiguration during oligomerization of lysenin, to disrupt the membrane and form a pore. In this way, by directly binding to a particular lipid component of the membrane, lysenin can be specifically targeted and can couple oligomerization to both its own refolding and the reorganization of the membrane. The specific binding of SM over its whole length would give oligomerizing lysenin sufficient purchase to disrupt the energetically stable lipid bilayer.
In order to investigate lysenin in its oligomeric state, we collected electron crystallography images of lysenin in liposomes containing SM. These data allowed us to identify the trigonal symmetry of the oligomer and estimate its dimensions. Lysenin pores are known to be small, with an approximate hydrodynamic diameter of 3 nm (Yamaji-Hasegawa et al., 2003
). The trigonal lattice contains a well-defined trimeric unit, in which the lysenin monomers appear to lie flat. The shape of the protomeric unit in this lattice does not look the same as lysenin in any projection and we cannot say therefore whether they are monomers or dimers of the protein. The absence of any extensive hydrophobic regions on lysenin’s surface, as found in other pore-forming proteins, means that it is likely similarly to deploy a β-hairpin across the membrane. A lysenin trimer could only supply three hairpins, or six β strands that is not enough to form a β-barrel and in itself suggests that the pore forming state may be a hexamer (a trimer of dimers). We believe that the structure observed in our 2D crystals is in fact a pre-pore state. The structures described in this study suggest that lysenin interacts with POC—not necessarily only that of SM but also from phosphocholine lipids—via its lectin domain prior to pre-pore assembly and pore formation. Subsequently, it associates with the head group and one full-length aliphatic tail of the SM molecule, which would serve to deform the membrane bilayer on the path to pore formation. Overall, our structures provide a rationale for the further development of lysenin as a tool for studying the role of SM in membrane structure, dynamics and function, while explaining the molecular basis of its dependence on SM binding for full activity (Bruhn et al., 2006