Classification of aromatic-centered motifs into membrane inserters and non-inserters by MD simulations.

(A) Initial dataset containing 10 sequences (sequence 1 from TrkA, 2 from Drp1, 3 and 4 from CD3ε, and 5 to 10 from CD3ζ). (B) Initial snapshot, time trace of aromatic Ztip, and snapshot at 980.7 ns for sequence 1. (C) Initial snapshot, time trace of aromatic Ztip, and snapshot at 964.3 ns for sequence 6. (B) and (C) represent simulations where the peptides are inserted or not inserted, respectively, into the membrane. Insertion means at least 80% of the frames in the last 100 ns (shaded region) have aromatic Ztip at least 3.1 Å (indicated by a red dashed line) below the phosphate plane. Time traces are smoothed using the moving average in a 2.5-ns window. The membrane is displayed as surface, with phosphates and all atoms above in orange while those below in gray; the peptide is displayed as cartoon in cyan, with the central aromatic residue as sticks in magenta and red. (D) The number of replicate simulations, out of a total of 20, where the peptide is inserted into the membrane.

MD, PPM, and AroMIP results on the initial dataset

Differences and similarities in membrane insertion and cation-π interactions among F-, W-, and Y-centered peptides.

(A) Aromatic Ztip distribution for an F-centered peptide (sequence 1). Insets: snapshot with F deeply inserted to interact with lipid acyl chains (bottom), or above the membrane to form cation-π interaction with a choline group (top). (B) Aromatic Ztip distribution for a W-centered peptide (sequence 2). Insets: snapshot with W deeply inserted to interact with lipid acyl chains and form a hydrogen bond with a glycerol oxygen atom (bottom), or in the headgroup region to form cation-π interaction with a choline group (top). (C) Aromatic Ztip distribution for a Y-centered peptide (sequence 3). Insets: snapshot with Y in the headgroup region to form a hydrogen bond with a glycerol oxygen atom (bottom), or near the membrane to form cation-π interaction with a choline group (top). In (A-C), results were accumulated from the last 100 ns of 20 simulations of each peptide; a red dashed line is drawn at Ztip = -3.1 Å. (D) Illustration of stable positions of F, W, and Y with respect to the membrane. (E) Extremely strong correlation between the percentage of membrane-inserting frames in MD simulations and the percentage of membrane-inserting conformations in PPM runs (correlation coefficient = 0.995).

Three-step pathways to membrane insertion.

(A) F-centered sequence 1. Basic sidechains at positions -4, -3, and -1 first tether one end of the peptide to the membrane surface. R[+4] then tethers the peptide at the other end, situating the central F for cation-π interaction. Finally, the aromatic sidechain is inserted into the acyl chain region. (B) W-centered sequence 2. The R[+1] sidechain initiates membrane contact. With possible stabilization by the membrane penetration of an aliphatic residue at position +3 or +4, the central W forms cation-π interaction with a choline and then enters the acyl chain region. (C) Y-centered sequences. The initial membrane contact can be formed by either a flanking basic sidechain (R[-3]; via electrostatic interaction) or the central Y itself (via cation-π interaction). After stabilizing at the membrane surface by R[-3] and L[+3], the central Y lowers into the headgroup region to form a hydrogen bond with a glycerol oxygen atom. The aromatic ring occasionally dips into the acyl chain region, bringing the lipid glycerol with it and producing local curvature.

Classification of aromatic-centered motifs into membrane inserters and non-inserters by PPM runs.

(A) Disordered conformations of 9-residue motifs generated by TraDES. (B) The pose of an F-centered peptide (sequence 1) from a PPM run. The phosphate plane is represented by arrays of small salmon spheres. (C) A scatter plot of ΔG and Ztip data for a batch of 100 conformations of sequence 1. Dash lines represent cutoffs for defining inserters (ΔG < -4.1 kcal/mol and Ztip < -1.5 Å). 15 conformations satisfy the cutoffs. (D) Extremely strong correlation between the percentage of MD simulations classified as membrane-inserting and the percentage of conformations in PPM runs (correlation coefficient = 0.997).

Flanking residues as determinants of membrane insertion.

(A) Illustration of the library of 3 x 58 = 1.2 x 106 sequences. (B) Percentages of membrane inserters and prediction accuracy of AroMIP. (C) L and R are enriched in all flanking positions of membrane inserters. (D) N, E, and G are enriched in all flanking positions of non-inserters. In (C, D), an orange dashed line is drawn at a frequency of 0.2 expected of random distributions. (E) q parameters for F-, W-, and Y-centered sequences.

Implementation of AroMIP on 2.4 x 104 aromatic-centered motifs in IDRs of the human proteome.

(A) Illustration of 2.4 x 104 aromatic-centered motifs in IDRs of the human proteome. This dataset was split 80:20 for training and testing, respectively. (B) q parameters for F-, W-, and Y-centered motifs. (C) Performance of AroMIP on the training and test sets.