Evolutionarily conserved short linear motifs drive actin filament binding

Figures

• 

  Fig. 1. Structural alignment of ITPKAArg28–Ala49 and Lifeact points towards an actin filament-binding motif, prompting proteome-wide discovery using the SLiMFold pipeline. a, Structural superimposition of the ITPKAArg28–Ala49–actin complex (the two actin subunits, A0 and A−2, are shown) with the Lifeact–actin structure (PDB: 7AD9). b, A 180° rotation of the helices reveals which residues interact with actin filaments. Conserved side chains are represented as sticks and labelled P1, P4, P8 and P9. c, Domain architecture of Saccharomyces cerevisiae ABP140 and Homo sapiens ITPKA. ABP140 comprises an N-terminal ABD and a C-terminal methyltransferase (MTase) domain. ITPKA features an N-terminal ABD, a central calmodulin-binding domain (CaMBD) and a C-terminal inositol 1,4,5-trisphosphate 3-kinase (IP3K) domain. d, Sequence comparison. The conserved positions P1, P4, P8 and P9 are labelled in red. e, SLiMFold pipeline overview. The SLiMFold pipeline starts with the hypothesized SLiMs of the Lifeact and ITPKAArg28–Ala49 sequence, on which basis a PSSM was calculated to identify motif hits from NCBI databases, which were filtered by PSSM scores, IUPRED, ANCHOR and PSIPRED. This yielded 539 hits…

• 

  Fig. 2. Validation and functional analysis of the short linear actin filament-binding motif by cellular actin filament co-localization. a, Five peptides identified by the SLiMFold pipeline were aligned with the ITPKA and Lifeact sequences. The conserved positions P1, P4, P8 and P9 are highlighted in red. b, Primary human macrophages were transfected with vectors coding for the EGFP-tagged peptides (EGFP–Pep) depicted in a and stained with phalloidin-568 to visualize actin filaments. Representative cell images and magnified views of the boxed region (magnif.) showing a podosome substructure are provided. Scale bars, 10 (main images) and 5 µm (magnified views). Imaris three-dimensional reconstructions and Poji radial profiles of the fluorescence intensities, at a z plane of highest fluorescent phalloidin intensity, with the mean ± s.d. of the actin filaments and the respective peptide from 180–953 podosomes from 3 cells are shown (n = 3). The Poji radial profiles were normalized to set the intensity values from 0 to 100%. FI, fluorescence intensity. Source data

• 

  Fig. 3. Determination of SFM binding affinities to actin filaments and identification of affinity-modulating amino acids. a, Five additional peptides identified by the SLiMFold pipeline were aligned with the sequences of ITPKA and Lifeact (left). The conserved positions P1, P4, P8 and P9 are highlighted in red. Secondary structures of these peptides, from SLiMFold predictions, are shown (right). b, The peptides were co-sedimented with actin filaments, analysed by western blotting and the actin filament concentrations were plotted against the normalized band intensities (left). The Kd,appvalues for all peptides tested are provided (right). The dots represent the Kd,app values obtained from three independent experiments, the bars show the mean ± s.e.m. and red circles indicate the global-fit Kd,app values. The ITPKA and Lifeact binding curves were compared using the extra-sum-of-squares F-test. ITPKA differed significantly from CGNL1 (P = 0.0009), CEFIP (P = 0.0009), ESPNL (P = 0.0077), SHROOM3 (P < 0.0001) and USP54 (P = 0.0054) but not Lifeact (P = 0.9006). Similarly, Lifeact differed significantly from CGNL1 (P < 0.0001), CEFIP (P < 0.0001), ESPNL (P = 0.0007), SHROOM3 (P < 0.0001) and…

• 

  Fig. 4. Structural comparison of ITPKA–SFM and USP54 M1–actin complexes. a, ITPKA–SFM (cyan) bound to actin filaments shown as molecular surface (central subunit A0 in orange, neighbouring subunit A−2 in red). Two views are shown, rotated by 90°. b, Same representation for the USP54 M1 helix (blue) shown in two orientations related by a 90° rotation. c, Superposition of the ITPKA and USP54 M1 helices on the same actin pocket. Cα RMSD = 1.30 Å (least-squares fit on the helices residues), indicating a highly similar placement and orientation. d–f, Magnified views of the conserved anchor positions P1 (d), P4 and P5 (e), and P8 and P9 (f) in a 90° top view. g,h, Helical-wheel projections for ITPKA (g) and USP54 M1 (h). Circles indicate helix residues coloured by chemistry. Dotted lines denote contacts to actin; actin residues contacted by both peptides are highlighted in red.

• 

  Fig. 5. Molecular dynamics simulations reveal SFM-induced stabilization of the actin D-loop and increased filament bending flexibility. a, Time-resolved contact occupancy for SFM peptides binding to actin filament. For each SFM residue (ITPKA, left; USP54 M1, right), the fraction of residue instances in contact with actin is plotted over the 40 ns trajectory. Colour intensity indicates contact fraction (white, no contact; cyan/purple, sustained contact). b, Actin filament stiffness is decreased by SFM binding. Decay of the tangent–tangent correlation function plotted against contour separation (s) for actin alone, actin + ITPKA and actin + USP54 M1. Data points represent the mean ± s.e.m. from spatial (11 subunits) and temporal (36 ns) averaging. Dashed lines show linear fits to the initial decay region, from which the Lp is extracted as Lp = −1 / slope. c, Differential RMSF (ΔRMSF; heavy atoms), computed as (actin + SFM) − (actin alone) for both ITPKA and USP54 M1, plotted across all actin residues. The shaded bands indicate the s.d. across the 11 actin subunits. Inset: expanded view of the D-loop region (residues 42–52) to highlight the residues most affected by SFM binding. d, The…

• 

  Fig. 6. Evolutionary analysis of validated SFM-containing proteins. a, Phylogenetic tree illustrating the origin and retention of the SFM. Simplified species tree of the sampled clades (left); the MRCA symbols shown on the tree indicate the inferred first appearance of the SFM within each gene. Clade × gene matrix for validated SFM-containing genes (right); green, motif present; red, motif absent or not detected. b, Class-pooled sequence logos for the SFM across nine representative organism classes (Actinopteri, Coelacanthiformes, Ceratodontiformes, Amphibia, Mammalia, Lepidosauria, Testudines, Aves and Crocodylia). For each class, the logo is built from all instances contributed by the 13 validated SFM-containing proteins that possess a confidently detected SFM in that class. Proteins that lack a confidently detected SFM in a given class do not contribute. Key motif positions P1, P4, P8 and P9 are labelled on the x axis. The letter height reflects residue conservation—larger letters indicate higher residue conservation. c, Analysis of the dN/dS ratio of all validated SFM-containing peptides (n = 12). Data are the mean ± s.e.m. The conserved SFM residues (P1, P4, P8 and P9) are…

• 

  Fig. 7. Iterative SLiMFold pipeline expansion and functional classification of newly identified SFMf-containing proteins. a, Schematic of three iterative SLiMFold rounds (detailed description in Fig. 3) including in vitro and validation in cells. AF, AlphaFold2. b, Eukaryotic cell map illustrating the functional categories of newly identified SFM-bearing proteins. The SFM candidates are grouped according to their principal roles and subcellular localizations. *Proteins for which actin filament binding has been attributed to the SFM (for example, validated by truncation studies); †proteins that bind actin filament but lack direct evidence linking the interaction specifically to the SFM. Panel b created in BioRender: Paraschiakos, T. https://biorender.com/35bnaup (2026).

• 

  Extended Data Fig. 1. Cryo-EM structure of the ITPKA–actin complex. a,The electron density map illustrates the ITPKA–actin complex. A single ITPKA helix of residue 28–49 (cyan) is resolved, forming the primary binding interface with actin. Actin subunits are displayed with the central subunit in orange and surrounding subunits in red. Phalloidin is shown in purple, and ADP in light. b, A cartoon representation of the ITPKA–actin complex shows ITPKA bridging two adjacent actin subunits. Key interaction regions in actin helix 5 (AA 137–145), helix 16 (AA 338–348), loop 2 (AA 21–28), the D-loop (AA 38-51) and actin subdomains SD1–SD4 are labelled. c, Apolar interactions between ITPKA and actin are depicted by surface colour-coding for hydrophobicity (yellow = hydrophobic). Left panel: Zoomed-in view and a 90° rotation highlight interacting residues displayed as sticks. Right panel: Zoomed-in view and a 180° rotation; here hydrophobicity scores are shown (low: cyan, high: yellow). d, Polar contacts between ITPKA and actin are shown via electrostatic potential mapping on actin (negative: red, positive: blue). Left panel: Zoomed-in view and a 90° rotation highlight interacting residues displayed as sticks.…

• 

  Extended Data Fig. 2. Position-specific scoring matrix derived from the two validated motif sequences, ITPKA and Lifeact and optimization of AlphaFold2 parameters for the SLiMFold pipeline. a, Nine positions (P1–P9) were aligned (Fig. 2c) and scored using a BLOSUM62-based approach, generating the PSSM shown. In this heat map, rows are grouped by amino acid type: positively charged (+), negatively charged (−), polar (P), unique (U), non-polar (NP) and aromatic (A) residues; and the position-wise scores are colour-coded (red = favourable, blue = unfavourable). Higher (positive) scores indicate preferential substitutions, whereas lower (negative) values suggest disfavoured residues at a given position. This matrix, which reflects sequence conservation in the ITPKA and Lifeact SLiMs, provided the starting point for screening and iterative refinement in the SLiMFold pipeline (Fig. 2 and Methods). b, Comparison of the number of sequences identified (y axis, log scale) under different multiple sequence alignment (MSA) algorithms: MMSeq2, UniRef90 with original jackhmmer parameters, and UniRef90 with modified jackhmmer parameters. Positive controls (black) are known SLiM-containing peptides validated in this…

• 

  Extended Data Fig. 3. Orientation-based clustering of SLiMFold hits on actin. a, Reference frame used to compare peptide orientations. Actin is shown in grey; example helices (ITPKA and GPATCH2) are overlaid to illustrate the alignment procedure. After superposing each peptide–actin model onto the ITPKA–actin reference, helix direction and position are described by two angles: Δφ (rotation around the filament/barbed-end axis) and Δθ (tilt)—and a polarity term (+1, same N → C direction as ITPKA; −1, opposite). b, 3D scatter of all predicted complexes in (Δφ, Δθ, polarity) space, coloured by RMSD. c, The same embedding coloured by hdbscan cluster assignment. d, Cluster prototypes mapped onto actin (two views 90° apart). Helices are coloured by cluster: Cluster 0 (blue), Cluster 1 (green), Cluster 2 (purple), Cluster 3 (orange). Outliers are denoted Cluster −1 (grey). The four clusters occupy distinct grooves/orientations around the barbed-end surface. e, Representative models from each cluster with example protein names annotated (matching the colours in d). These exemplars illustrate the sequence and placement diversity captured by the orientation-based clustering.

• 

  Extended Data Fig. 4. Validation and functional analysis of the novel short linear actin filament-binding motif by cellular actin filament co-localization. a, Representative confocal images are shown in both top and side views (Imaris 3D reconstructions), emphasising subcellular peptide localization. Wild-type peptides generally display robust co-localization with F-actin (appearing white in merged panels), whereas alanine substitutions at key residues (P1 or P8) diminish or abolish F-actin association. b, Primary human macrophages were transfected with the vectors coding for the EGFP-tagged peptides and their respective mutants (position P1 and P8 were mutated to Glutamate E) and stained with phalloidin-568 to visualize actin filaments (magenta). Representative cell images are shown including zoomed- into one podosome substructure, scale bars: 10 and 5 µm, respectively. Co-localization is evident when EGFP-tagged peptides (green) overlap with phalloidin-stained actin (magenta). Right panels: Imaris 3D reconstructions and Poji radial profiles confirm varying degrees of peptide–actin co-localization. Shown are Poji radial profiles of fluorescence intensities, at a z plane of highest fluorescent…

• 

  Extended Data Fig. 5. Co-localization of phalloidin-stained actin and EGFP-tagged peptides and proteins in H1299 cells with accompanying intensity profiles. a, Representative confocal micrographs show EGFP-tagged peptides (green) alongside phalloidin-568 staining of F-actin (magenta). Merged images highlight varying degrees of co-localization (appearing white). Intensity profiles (right panels) were generated along the dashed white line in each cell, quantifying the overlap between F-actin and peptide signals. Wild-type peptides generally display robust co-localization, whereas alanine substitutions at key positions (P1 or P8) diminish or abrogate actin association. Scale bars, 10 µm. b, Representative confocal images show H1299 cells transfected with WT or mutant full-length (FL) constructs fused to EGFP (green), together with phalloidin-568 staining of F-actin (magenta). Merged panels highlight varying extents of co-localization (appearing white). Intensity profiles (right panels) were recorded along the dashed white lines in each cell, quantifying relative fluorescence for F-actin and the respective EGFP-FL protein. Wild-type constructs generally exhibit robust co-localization with actin filaments,…