Epsin and Sla2 form assemblies through phospholipid interfaces

Abstract

In clathrin-mediated endocytosis, adapter proteins assemble together with clathrin through interactions with specific lipids on the plasma membrane. However, the precise mechanism of adapter protein assembly at the cell membrane is still unknown. Here, we show that the membrane-proximal domains ENTH of epsin and ANTH of Sla2 form complexes through phosphatidylinositol 4,5-bisphosphate (PIP2) lipid interfaces. Native mass spectrometry reveals how ENTH and ANTH domains form assemblies by sharing PIP2 molecules. Furthermore, crystal structures of epsin Ent2 ENTH domain from S. cerevisiae in complex with PIP2 and Sla2 ANTH domain from C. thermophilum illustrate how allosteric phospholipid binding occurs. A comparison with human ENTH and ANTH domains reveal only the human ENTH domain can form a stable hexameric core in presence of PIP2, which could explain functional differences between fungal and human epsins. We propose a general phospholipid-driven multifaceted assembly mechanism tolerating different adapter protein compositions to induce endocytosis.

Fig. 1

Native MS of ENTH/PIP2 complexes suggests an allosteric binding mechanism. a For analyzing lipid binding to ENTH domains, proteins were measured in presence of PIP2 and cytochrome c as reference. Raw spectra show free cytochrome c and unspecific attachment of 1 PIP2 (gray), while ENTH domains (here C. thermophilum) bind 0–3 PIP2. b Signal intensities from MS were summed over all charge states (back) and corrected for unspecific PIP2 clustering based on the ratio of bound/unbound reference protein (front). Data of at least three independent measurements were normalized to the corrected signal of unbound ENTH and the averages of the relative signal intensities and their standard deviations were plotted. The signal for ENTH with three PIP2 observed in raw spectra disappears after correction. c Schematic illustration of microscopic and macroscopic dissociation constants of two PIP2 molecules (orange) binding independently to ENTH (blue). For the first binding event of PIP2, two pathways with the microscopic dissociation constants k_d,1 and k_d,2 are available, leading to one apparent species of ENTH+PIP2. Combined, they account for the macroscopic dissociation constant K_D,1. The second macroscopic dissociation constant K_D,2 describes PIP2 binding to the thus far unoccupied binding site that yields the product ENTH+2PIP2. Again, this binding event can be partitioned into two pathways with the microscopic dissociation constants k_d,1 and k_d,2. If k_d,1 and k_d,2 are unaltered in the first and second binding event, binding sites are independent. Binding sites are represented by rounded rectangles

Fig. 2

Crystal structure of the ENTH2/PIP2 complex reveals an allosteric-binding mechanism. a Ribbon diagram of the ENTH2_α0 (cyan) and ENTH2_Noα0 dimer (blue) building block. N-terminal regions are colored (ENTH2_α0, yellow; ENTH2_Noα0, orange). PIP2 sitting in the interface between ENTH2 molecules is shown as spheres. b Superposition of the ENTH/IP3 epsin-1 complex (dark blue) and the ENTH2/PIP2 complex (yellow). The N-terminal α0 is colored (ENTH2/PIP2, yellow; ENTH/IP3, orange), the inositol head groups are shown as sticks. c Superimposition of the ENTH2/PIP2 dimer (blue/cyan) and the ENTH1 dimer (green/palegreen) shown as a ribbon diagram. The α0 helix of ENTH1 is oriented similarly to α0 of ENTH2_α0. d Surface presentation of the ENTH2/PIP2 complex showing that unfolded N-termini of ENTH2 are in plane with the lipid tail of the PIP2 molecules (two ENTH2_α0/ENTH2_Noα0 building blocks, cyan/blue and magenta/violet; PIP2, spheres; N-termini for empty ENTH2, yellow and orange; Tyr 16, Arg 24, Arg 62 and His 72 of ENTH2_Noα0 with empty PIP2-binding pocket, red)

Fig. 3

Evolutionarily conserved features in Hip1R subfamily of ANTH domains. a Native MS was used to investigate PIP2 binding to several members of the ANTH family: ANTH_Sla2, ANTH_Sla2,mut (four residues to Ala in the canonical PIP2-binding site, see Constructs) and CALM from indicated species. Signal intensities from mass spectra of at least three independent measurements were summed over all charge states, normalized to the corrected signal of the unbound protein and the average of the relative signal intensities and their standard deviations were plotted (back). After correction for unspecific clustering, relative signals of unbound and PIP2-bound ANTH domains are obtained (front). Signal for proteins with three attached PIP2 molecules observed in raw spectra disappear after correction. b Composite model of the ENTH and ANTH_Sla2 complex based on the low-resolution EM structure and the high-resolution X-ray crystal structures presented here (ENTH2/PIP2, yellow, and superimposed ANTH domains of Sla2 (violet), AP180 (cyan), and CALM (pale cyan). Several elements contributing to the interface are highlighted, including the inserted α helix in the ANTH_Sla2 domain, the proximate Arg 37 of ANTH_Sla2 and Thr 104 from ENTH, and the location of PIP2 from the secondary site of the ENTH2/PIP2 crystal structure. c Growth defects of Sla2(1–360) strains mutated in Sla2/Hip1R ANTH-specific features. Tenfold serial dilutions of sla2Δ strains expressing Sla2(1–360) wt, Sla2(1–360) NHL with α8–α9 loop replaced by residues NHL as occurring in Yap1802, Sla2(1–360) dYL deletion of conserved Tyr 252 and Leu 253 and Sla2(1–360) R29A were incubated on SC-Ura plates at 30, 34, and 37 °C for 2 days

Fig. 4

ENTH/ANTH/PIP2 complex formation in fungi 6:6:~18 and 8:8:~24 ENTH/ANTH/PIP2 complexes are the observed stoichiometries in native MS measurements. A cartoon of the most prominent complex is shown with ANTH in green and ENTH in blue. Complexes from C. thermophilum (green) and S. cerevisiae, containing ENTH1 (orange) or ENTH2 (red), show the same stoichiometries, signal ratios and dissociation pathways in collision-induced dissociation (CID) MS/MS. Here, the dissociation of the +40 charged 8:8:25 ± 3 ENTH1/ANTH_Sla2/PIP2 complex into partially unfolded ANTH_Sla2 (top left, cartoon shows green ANTH domain) and a residual 8:7:23 ± 1 complex (top right, showing also a cartoon of the remaining complex) is depicted. The annotation shows the stoichiometry (ENTH:ANTH:average PIP2 number), charge state of the main peaks, and average experimental masses. Ranges of PIP2 numbers, statistical errors, and an average FWHM value rating the MS resolution of all complexes can be found in Supplementary Table 2. ANTH dissociation in CID MS/MS measurements of 6:6:18 ± 1 ENTH2/ANTH_Sla2/PIP2 complexes from S. cerevisiae is presented in Supplementary Fig. 4b

Fig. 5

Dynamics of ENTH/ANTH/PIP2 complex formation in fungi and human assemblies. a Time course of ENTH2/ANTH_Sla2/PIP2 (S. cerevisiae) complex formation. Components were mixed, injected into the electrospray capillary and the spectra monitored over time. Relative signal intensities for 6:6 (green), 8:8 (dark blue), and dimers of 8:8 complexes (light blue) were determined and plotted against time (n = 3). The signal of the 6:6 complex drops within 2 min after mixing the complex components, while the signal of the 8:8 complex increases, suggesting a transition between these forms. The signal of the 8:8 dimer remains constant, ruling out aggregation effects. Average data of three independent measurements, error bars (standard deviation) are shown for data points with N = 3 b Native MS shows oligomerization of human epsin-1 ENTH and Hip1R ANTH in various stoichiometries, ranging from 6:0 to 8:8 with 6:6 being the main species. c Human epsin-1 ENTH domain forms hexamers (6:0) with at least six PIP2 molecules also in absence of Hip1R

Fig. 6

Human ENTH forms a thermally stable hexamer with a predicted membrane-binding interface. a SAXS modeling of human ENTH in the presence of PIP2. SAXS data recorded for human ENTH in solution are shown (gray circles) along with a fit to the rigid-body model refined against the SAXS data with P3 symmetry using CORAL (orange solid line). Experimental errors are from counting statistics on the Pilatus 2M detector and propagated through the data reduction process as standard errors in the scattering intensities. The χ² for the fit is 1.05. The inset shows the dimensionless Kratky plot representation of the SAXS data and the same fit. b Backbone and surface representation of the SAXS refined rigid body model. C-terminal (Ct) residues not observed in the crystal structure of the tandem domains are modeled as dummy residues by CORAL. P3 symmetry was enforced and the tandem ENTH domains with bound PIP2 used as rigid bodies. c 90° rotation of the model, demonstrating that the bound PIP2 molecules are all located on one side of the protein, consistent with a membrane-binding interface. The Thr 104 residues involved in ENTH/ENTH homodimerization are colored pink, whereas the Thr 104 residues present on the surface of the hexamer are colored green. d Hydrodynamic radius as a function of temperature measured by DLS (ENTH, S. cerevisiae, orange; ENTH/ANTH complex, S. cerevisiae, green; hexameric ENTH core, H. sapiens, blue)

Fig. 7

Schematic models of PIP2 binding initiating clustering of ENTH and ANTH domains. a In S. cerevisiae and C. thermophilum ENTH (blue) and ANTH (green) domains bind PIP2 (orange) in the membrane and cluster to hetero 12-mers with the stoichiometry 6:6 that consequently can be transformed to more stable hetero 16-mers with the stoichiometry 8:8. b Human ENTH domains (yellow) bind PIP2 (orange) and cluster to homo 6-mers. ANTH domains (red) bind in different stoichiometries. 6:6 hetero 12-mers are the most abundant species, but a transition to larger complexes, up to hetero 16-mers can be observed. Symbols assigning the different complex stoichiometries are chosen as in the mass spectrum of Fig. 5b. Dashed arrows indicate that not all complex stoichiometries are represented in this model