Module 5: Regulation & Signaling

The Rho-family small GTPases (20 members in humans; Ras-superfamily) are 20-25 kDa soluble P-loop NTPases that cycle between an active GTP-bound form and an inactive GDP-bound form. Three prototypes dominate cytoskeletal signaling:

• RhoA: drives stress fibers and focal adhesions; downstream effectors include ROCK and mDia1.
• Rac1: drives lamellipodia and membrane ruffles; downstream effectors include WAVE regulatory complex (WRC) and PAK.
• Cdc42: drives filopodia and cell polarity; downstream effectors include WASP, N-WASP, and MRCK.

The landmark work of Ridley & Hall (1992, Cell) established these three phenotypic axes: microinjection of constitutively active RhoA into quiescent fibroblasts produced thick actomyosin stress fibers; active Rac1 produced flat lamellipodia; active Cdc42 produced thin finger-like filopodia. The canonical signal-to-phenotype map of the actin cytoskeleton was born.

A Rho GTPase does not do anything by itself; it is a slow enzyme with intrinsic GTPase activity of ~0.01–0.05 s^-1 and GDP-release rate ~0.01 s^-1. Cells accelerate each direction of the cycle by dedicated regulators:

• Guanine nucleotide exchange factors (GEFs): catalyze GDP release and GTP binding, flipping the switch to “on”. The Dbl-homology (DH) domain is the canonical catalytic fold (Rossman, Der & Sondek 2005 Nat. Rev. Mol. Cell Biol.). Examples: Vav1/2/3, Tiam1 (Rac), LARG (Rho), Intersectin (Cdc42). DOCK-family GEFs use an alternative catalytic fold (DOCK180 for Rac).
• GTPase-activating proteins (GAPs): insert an “arginine finger” into the GTPase active site, accelerating hydrolysis by ~10⁴-fold and flipping the switch to “off”. Examples: p190RhoGAP, ARHGAP24, chimaerin, BCR.
• Guanine nucleotide dissociation inhibitors (GDIs): bind the prenylated C-terminus of GDP-bound Rho GTPases and extract them from membranes into the cytoplasm, creating a soluble reservoir. Delivery to membranes is release from GDI (Garcia-Mata, Boulter & Burridge 2011 Nat. Rev. Mol. Cell Biol.). RhoGDI is ubiquitous; Ly/D4-GDI is hematopoietic.

Mathematically the cycle is a two-state stochastic switch where the steady-state GTP-fraction is

Each active Rho-GTPase binds a characteristic suite of effector proteins through its switch regions. The key cytoskeletal effectors are:

• ROCK (Rho kinase): ~160 kDa serine/threonine kinase activated by RhoA-GTP. ROCK phosphorylates myosin light chain (MLC) directly and also phosphorylates/inhibits MLC phosphatase; the net effect is MLC phosphorylation on Ser19 and activation of myosin II contractility. Y-27632 (an ATP-competitive ROCK inhibitor) abolishes stress fibers and cortical tension.
• mDia1 / Dia formins: Rho-GTP binds and activates the N-terminal GBD of Dia1, relieving autoinhibition and exposing the actin-nucleating FH2 domain. mDia1 nucleates straight, unbranched actin filaments at the plus end and processively elongates them while delivering profilin-bound G-actin through its FH1 domain (Pruyne, Evangelista, Yang, Bi, Zigmond, Bretscher & Boone 2002 Science; Goode & Eck 2007).
• WAVE regulatory complex (WRC): Rac1-GTP activates the WAVE (WASF) heterpentamer (WAVE, CYFIP, NAP1, HSPC300, Abi1). WRC binds and activates the Arp2/3 complex at the leading edge, driving lamellipodial branching (Chen et al. 2010 Nature).
• WASP / N-WASP: Cdc42-GTP binds the GBD of WASP, relieving intramolecular autoinhibition and exposing a verprolin-central-acidic (VCA) domain that activates Arp2/3 branching at specific foci. N-WASP drives filopodial and invadopodial actin networks.
• PAK (p21-activated kinase): activated by Rac1 and Cdc42; phosphorylates LIM kinase, which in turn phosphorylates and inhibits cofilin (see Section 5). PAK therefore stabilizes the actin network when Rac/Cdc42 are active.
• MLCK (myosin light-chain kinase): Ca²⁺/calmodulin-activated; directly phosphorylates MLC. Works in parallel with ROCK to activate myosin II.

The effectors are a combinatorial switchboard: one Rho-GTP can bind multiple effectors simultaneously, and each effector can integrate signals from multiple GTPases. This is why the Ridley-Hall phenotype is a clean textbook picture but reality is messier: Rho, Rac, and Cdc42 cross-regulate each other extensively, and their spatiotemporal dynamics determine cell shape, not their absolute levels.

The formin family (15 human genes: DIAPH1-3, FMN1-2, INF1-2, DAAM1-2, FHOD1/3, FMNL1-3) nucleates actin filaments from G-actin monomers without requiring Arp2/3. The catalytic domain is the FH2 (formin-homology 2) domain, a ring-shaped homodimer that caps the fast-growing plus end of a nascent filament, allowing new subunits to slip in between the ring and the filament tip in a processive “stair-stepping” manner (Pruyne et al. 2002 Science).

Upstream of the FH2 is the FH1 domain: a proline-rich tract that binds multiple profilin-actin complexes. FH1 tethers profilin-actin to the plus end and delivers fresh G-actin to the growing filament, accelerating elongation 3-5 fold above the bulk diffusion rate (\(v \sim k_{\text{on}}[\text{G-actin}]\cdot\delta\)) and explaining why formins build long, unbranched cables in stress fibers and cytokinetic rings.

Most formins are autoinhibited: the C-terminal DAD (diaphanous autoregulatory domain) binds the N-terminal GBD/DID, folding the molecule into a closed state. Active Rho binding to the GBD releases DAD and exposes FH1-FH2 for nucleation (Li & Higgs 2003 Curr. Biol.; Higgs 2005 Trends Biochem. Sci.).

Different formins produce different filament architectures. mDia1 (DIAPH1) drives RhoA- dependent stress fibers. mDia3 drives microtubule-actin coupling at the cell cortex. DAAM1 is part of the Wnt/planar-cell-polarity pathway. FMN2 nucleates the cytokinetic ring in female meiosis (oocyte polar body extrusion). Formin mutations cause a wide range of human disease, including deafness (DIAPH1), neural-tube defects (DAAM1), and macrothrombocytopenia.

Cofilin and ADF (actin-depolymerizing factor)are small (~19 kDa) universally conserved actin-binding proteins that both sever filaments and promote dissociation at the minus end. Cofilin binds preferentially to ADP-actin on the filament, changes the twist of the filament locally (supercoiling it by ~5\(^\circ\) per monomer), and weakens lateral contacts at the boundary between cofilin-bound and bare segments, driving severing (Bernstein & Bamburg 2010 Trends Cell Biol.; McCullough et al. 2008).

Cofilin is a classic example of cytoskeletal regulation by two independent signals:

• Phosphorylation: LIM kinase (LIMK1, LIMK2) phosphorylates Ser3 of cofilin, abolishing its actin binding (inactive). LIMK is activated by Rac1-PAK and by ROCK. Dephosphorylation by the slingshot phosphatase (SSH1-3) or chronophin (CIN) restores activity.
• pH: cofilin is more active at higher pH (~7.5) than lower pH (~6.8); acidification can therefore reduce severing during certain contexts of cell cycle or migration (Pope et al. 2004 J. Biol. Chem.).

The net effect of cofilin is to accelerate actin turnover. In a lamellipodium, cofilin severs aged (ADP-actin) filaments at the rear, creating new ends that the capping protein CapZ terminates and whose monomers profilin recycles to the leading edge. This is the molecular substrate of treadmillingat the cellular scale (Pollard & Borisy 2003 Cell).

The Arp2/3 complex is a seven-subunit ~225 kDa machine (Arp2, Arp3, and ARPC1-5) that binds the side of an existing (“mother”) actin filament and templates a daughter filament at a characteristic 70\(^\circ\) branch angle (Mullins, Heuser & Pollard 1998 PNAS; Rouiller et al. 2008 J. Cell Biol.). The complex is intrinsically inactive and must be activated by a nucleation promoting factor (NPF), typically a member of the WASP/WAVE family bearing a VCA domain at its C-terminus.

The VCA domain has three parts:

• Verprolin-homology (or WH2): binds G-actin.
• Connecting region: binds Arp2/3.
• Acidic: binds Arp2/3.

VCA delivers one G-actin monomer to Arp2, positions the Arp2/3 complex on a mother filament, and triggers a conformational change that brings Arp2 + Arp3 into a short-pitch helical arrangement. The Arp2-Arp3 dimer mimics the plus end of an actin filament, so new G-actin monomers can bind and elongate a daughter filament whose axis is angled at 70\(^\circ\) off the mother (the critical geometric constraint that determines the mesh topology).

The lamellipodium is the prototypical dendritic network: the leading edge is continuously nucleating branches, each ~70\(^\circ\) off its mother, at a rate that scales with WAVE/WASP activation by Rac1-GTP or Cdc42-GTP. As the network ages, CapZ caps the plus ends, cofilin severs and disassembles the rear, and profilin recycles monomers back to the leading edge.

Arp2/3 inhibitors (CK-666, CK-869; Nolen et al. 2009) are widely used research tools and are in development for treating breast cancer metastasis, where invadopodia are critically Arp2/3-dependent.

Beyond nucleation and severing, actin-network architecture is tuned by proteins that cap, bundle, crosslink, or stabilize filaments:

• CapZ (capping protein): a αβ-heterodimer that binds the plus end of F-actin and blocks subunit addition. Sets filament length at the leading edge by terminating Arp2/3-nucleated daughters and is itself regulated by PIP2. Found in every dendritic actin network.
• Tropomodulin: caps the minus end in sarcomeric thin filaments, stabilizing the precise length required for muscle.
• Tropomyosin: a coiled-coil dimer that runs along the long-pitch groove of F-actin, stabilizing it against cofilin severing and regulating myosin binding (troponin-tropomyosin regulation in muscle).
• α-actinin: a 100 kDa antiparallel homodimer that crosslinks parallel actin filaments at Z-disks in muscle and along stress fibers in non-muscle cells. Ca²⁺-regulated in non-muscle isoforms; a mechanoresponsive element in focal adhesions.
• Filamin A/B/C: immunoglobulin-repeat crosslinkers that form orthogonal actin gels; filamin A mutations cause periventricular heterotopia and skeletal dysplasia.
• Fascin, fimbrin, espin, villin: tight bundling proteins that assemble parallel actin into filopodia, microvilli, stereocilia, and brush borders.
• Spectrin: long tetramers that form the cortical cytoskeletal lattice of erythrocytes and underlie axonal periodic actin rings (Xu, Zhong, Babcock, Boyden & Zhuang 2013 Science).
• Gelsolin: Ca²⁺-activated severing and capping protein; six gelsolin-homology domains; activated by ~10 μM Ca²⁺. Releases fragments as intermediates for remodeling; a prototype for calcium-activated cytoskeletal remodeling (Sun, Yamamoto & Yin 1999).
• Profilin: tiny 15 kDa G-actin-binder; suppresses spontaneous nucleation but permits barbed-end addition; loads G-actin onto formin FH1. Profilin-actin is the main cellular G-actin pool.

Integrins are αβ-heterodimeric transmembrane receptors (24 combinations of 18 α and 8 β subunits in humans) that link the extracellular matrix (fibronectin, laminin, collagen) to the actin cytoskeleton through a chain of adaptor proteins: talin, vinculin, paxillin, zyxin, α-actinin. A focal adhesion is a ~0.5-2 μm dense plaque at the ventral cell surface that both anchors the cell and senses substrate stiffness (Geiger, Spatz & Bershadsky 2009 Nat. Rev. Mol. Cell Biol.; Kanchanawong et al. 2010 Nature).

Focal adhesions mature under tension: low-tension nascent adhesions near the leading edge recruit paxillin and talin; higher-tension “classical” focal adhesions recruit vinculin (force-activated through a talin stretch-opening event), zyxin, and eventually VASP for actin polymerization.

Downstream of integrin engagement, focal adhesion kinase (FAK) is autophosphorylated on Y397, creating a docking site for Src-family kinases. Src phosphorylates paxillin, p130Cas, and activates Rho-family GEFs. The net effect is activation of Rho, Rac, and Cdc42 at the adhesion and initiation of the local actin response. FAK/Src inhibitors are in clinical development for metastatic cancer.

Focal adhesions are also the site of stiffness sensing: cells on stiff substrates mature larger and more stable focal adhesions, build more stress fibers, and exhibit higher nuclear YAP/TAZ activity (Dupont et al. 2011 Nature). This is the molecular substrate of mechanotransduction linking tissue stiffness to gene expression and differentiation (Engler, Sen, Sweeney & Discher 2006 Cell).

Calcium is the master second messenger of cytoskeletal regulation. Resting cytosolic [Ca²⁺] is ~100 nM; stimulated cells reach 1-10 μM transiently. Multiple cytoskeletal components are gated by Ca²⁺:

• Gelsolin activates at ~0.5-1 μM Ca²⁺, severing and capping F-actin.
• Calmodulin binds 4 Ca²⁺ and activates MLCK (directly phosphorylates MLC), CaMKII (phosphorylates many substrates including F-actin bundlers), and calcineurin (dephosphorylates NFAT to drive gene expression).
• Troponin C in muscle sarcomeres binds Ca²⁺ and displaces tropomyosin to expose myosin-binding sites, triggering contraction.
• Myosin II in non-muscle is activated by Ca²⁺-calmodulin-MLCK in parallel with ROCK, combining acute and sustained contraction.
• Fimbrin, villin, α-actinin (non-muscle isoforms) have EF-hand Ca²⁺-binding sites that modulate their actin affinity.

Spatial Ca²⁺ gradients are therefore read as spatial cytoskeletal reorganization patterns—this is one of the oldest substrates of cell polarization, evident in everything from chemotaxing neutrophils to growth cone pathfinding to fertilization.

The cytoskeleton is not just a passive load-bearer—it regulates gene expression. Two major mechanosensitive transcriptional systems read cytoskeletal state:

• YAP/TAZ: transcriptional co-activators of the Hippo pathway. On soft substrates and in confluent monolayers, YAP/TAZ are phosphorylated by LATS1/2 and sequestered in the cytoplasm. On stiff substrates, or when F-actin is cross-linked with mature stress fibers, YAP/TAZ translocate to the nucleus and drive TEAD-dependent transcription of proliferation and anti-apoptotic genes (Dupont et al. 2011).
• MRTF-A/B (myocardin-related transcription factor): bound to G-actin in the cytoplasm and sequestered there. When G-actin is depleted by polymerization (e.g., Rho-driven formin activity), MRTF is released and translocates to the nucleus where it activates SRF (serum-response factor) and transcribes cytoskeletal genes (Miralles et al. 2003 Cell). This is a feedback loop: cytoskeletal polymerization drives transcription of cytoskeletal components.

Together, these systems close the loop from integrin-sensed ECM mechanics, through Rho-driven actomyosin contractility, to nuclear mechanotransduction (Modules 3, 6) and ultimately transcriptional output. The cytoskeleton is a control system, not a passive scaffold.

Dysregulated cytoskeletal signaling is a feature of most cancers and many other diseases: