Cryo-EM of human Arp2/3 complexes provides structural insights into actin nucleation modulation by ARPC5 isoforms

Abstract

The Arp2/3 complex regulates many cellular processes by stimulating formation of branched actin filament networks. Because three of its seven subunits exist as two different isoforms, mammals produce a family of Arp2/3 complexes with different properties that may be suited to different physiological contexts. To shed light on how isoform diversification affects Arp2/3 function, we determined a 4.2 Å resolution cryo-EM structure of the most active human Arp2/3 complex containing ARPC1B and ARPC5L, and compared it with the structure of the least active ARPC1A-ARPC5-containing complex. The architecture of each isoform-specific Arp2/3 complex is the same. Strikingly, however, the N-terminal half of ARPC5L is partially disordered compared to ARPC5, suggesting that this region of ARPC5/ARPC5L is an important determinant of complex activity. Confirming this idea, the nucleation activity of Arp2/3 complexes containing hybrid ARPC5/ARPC5L subunits is higher when the ARPC5L N-terminus is present, thereby providing insight into activity differences between the different Arp2/3 complexes.

Keywords: Actin; Arp2/3; Cryo-EM; Cytoskeleton; Isoforms; Nucleation.

Fig. 1.

The cryo-EM structure of the human Arp2/3 ARPC1B-ARPC5L complex. (A) Left, overview of the cryo-EM reconstruction of Arp2/3-C1B-C5L with the docked model in the density, viewed and coloured as originally presented by Robinson et al. (2001): Arp2: red; Arp3: orange; ARPC1B: green; ARPC2: cyan; ARPC3: magenta; ARPC4: light blue; ARPC5L: yellow; right, same view of the reconstruction with a ∼8 Å low-pass filter applied showing more flexible regions of the complex at lower resolution; (B) left, 180° rotated view compared to A of the Arp2/3-C1B-C5L reconstruction and model; right, same view of the reconstruction with a ∼8 Å low-pass filter applied showing more flexible regions of the complex at lower resolution, which includes flexible connectivity between Arp2 and Arp3 (red arrow) and parts of ARPC5L. sd, subdomains of Arp2.

Fig. 2.

Nucleotide binding sites of Arp3 and Arp2 in Arp2/3-C1B-C5. (A) Cryo-EM reconstruction and model of nucleotide binding pocket of Arp3 with density corresponding to bound ATP indicated (dotted black oval); (B) ribbon depiction of the Arp3 model with density corresponding to bound nucleotide shown in surface representation. This density is the calculated difference between our cryo-EM reconstruction and simulated density of the atomic model without nucleotide at equivalent resolution, calculated using Chimera. This supports the conclusion that ATP is bound to Arp3; (C) conformation of the nucleotide binding pocket of ATP-bound Arp3 in Arp2/3-C1B-C5L aligned (on subdomain 3) with a previously determined structure of nucleotide-free Arp2/3 (PDB 1K8K), showing closure of the pocket in the presence of bound nucleotide; (D) conformation of the nucleotide binding pocket of ATP-bound Arp3 aligned (on subdomain 3) with a previously determined structure of ATP-bound Arp2/3 (2P9S; Nolen and Pollard, 2007), showing equivalent closure of the pocket in the presence of bound nucleotide compared to the absence of nucleotide; (E) cryo-EM reconstruction and model of nucleotide binding pocket of Arp2 with visible subdomain regions labelled and density corresponding to bound ATP indicated (dotted black oval); (F) ribbon depiction of the Arp2 model with density corresponding to bound nucleotide shown in surface representation. This density is the calculated difference between our cryo-EM reconstruction and simulated density of the atomic model without nucleotide at equivalent resolution, calculated using Chimera. This supports the conclusion that ATP is bound to Arp2; (G) 2D class averages of Arp2/3-C1B-C5L showing views corresponding to Fig. 1A (left panels) and Fig. 1B (right panels) illustrating the variable density corresponding to subdomain 2 of Arp2 (red arrows) and to ARPC5L (yellow arrows). ARPC2 is also labelled for reference (blue arrows). (H) Arp2 in the GMF-inhibited Arp2/3 (PDB 4JD2, in tan) (Luan and Nolen, 2013) is shown aligned with Arp2 (red) in Arp2/3-C1B-C5L within the low-pass filtered cryo-EM density. As previously shown, in the Arp2/3-C1B-C5L reconstruction Arp2 subdomain 2 is flexibly connected to Arp3 helix-α9 (orange arrow), whereas the well-defined structure of the Arp2 subdomain 2 in the GMF-inhibited structure adopts a different conformation which protrudes from the EM density (tan asterisk). For clarity, other subunits within the GMF-inhibited complex are not shown. sd, subdomains of Arp2.

Fig. 3.

The cryo-EM structure of the human Arp2/3-ARPC1A-ARPC5 complex. (A) Left, cryo-EM reconstruction of Arp2/3-C1A-C5; right, same view with a ∼8 Å low-pass filter applied to potentially reveal more flexible regions of the complex at lower resolution. The docked model is coloured as in previous figures: Arp2: red; Arp3: orange; ARPC2: cyan; ARPC3: dark pink; ARPC4: blue, except that ARPC1A is dark green and ARPC5 is pale yellow; (B) left, 180° rotated view compared to (A) of the Arp2/3-C1A-C5 reconstruction and model; right, same view of the reconstruction with a ∼8 Å low-pass filter applied. sd, subdomains of Arp2.

Fig. 4.

Isoform-mediated differences in human Arp2/3 complexes. (A) Left, location of non-conserved sequence variation between human ARPC1A and ARPC1B (green spheres) and location of disease-causing ARPC1B point mutations (purple spheres) mapped onto ARPC1B; right, cross section through ARPC1B; (B) left, density corresponding to ARPC5L, showing the incomplete density for this subunit apart from helix-α7 adjacent to ARPC4; right, density corresponding to ARPC5, showing the near complete density for this subunit (C) left, 90° rotated view compared to B, left of ARPC5L showing the incomplete density for this subunit and lack of connectivity to Arp2; right, 90° rotated view compared to B, right of ARPC5, showing the clear density for most of the subunit, including its N-terminal tether to Arp2.

Fig. 5.

Activity of ARPC5/C5L hybrid complexes support a role for ARPC5/C5L N-terminus in defining functional differences between Arp2/3 subunit isoforms. (A) Schematic and nomenclature of ARPC5/C5L hybrids. (B) Coomassie-stained gel of purified recombinant Arp2/3 complexes containing ARPC1A together with ARPC5, ARPC5L or their hybrids. Gel band quantification of ARPC5, ARPC5L and the hybrids normalised to ARPC2 showed the same ratio in all cases, consistent with equivalent subunit occupancy [ARPC5/ARPC2=0.52±0.04; ARPC5L/ARPC2=0.52±0.09; C5C5L/ARPC2=0.46±0.04; C5LC5/ARPC2=0.46±0.04 (mean±s.d., n=3, technical replicates)]; (C) immunoblot analysis of purified recombinant Arp2/3 complexes used in this study. (D) In vitro polymerisation of 2 µM pyrene-actin (5% labelled), either alone (black curve) or in the presence of 5 nM VCA and 1.25 nM of Arp2/3-C1A with the indicated ARPC5 isoforms or hybrids (named as in panel A) shows differences in actin assembly according to the ARPC5/5L N-terminal region present. The curves shown here come from one representative experiment, which was repeated four times, giving similar results. The time at half-maximum, normalised to that of the ARPC5L isoform, is 1.27±0.06 for ARPC5, 1.32±0.10 for ARPC5/C5L, and 1.01±0.06 for ARPC5L/C5 (average±s.e., n=4, technical replicates).