The role of PDE6D in trafficking KRAS

1 Introduction

Membrane-associated lipidated proteins can bind reversibly to lipid bilayers, and their steady state localization reflects a dynamic equilibrium between cytosolic and membrane-bound pools (Yuan et al. 2024). This distribution is governed by multiple parameters on both sides of the interaction: on the membrane side by lipid composition, biophysical properties, curvature and geometry, and on the protein side by the nature, accessibility, and combination of membrane-binding motifs (Bigay and Antonny 2012). Changes in any of these parameters, for example membrane remodeling, post-translational modification, or conformational switching of the protein, can shift the steady-state distribution and redistribute proteins between compartments. In addition to steady-state control, the kinetics with which these parameters change can be equally important, enabling rapid spatial reorganization in response to signaling cues.

Several molecular mechanisms exist to regulate these equilibria. For small GTPases of the Rho and Rab families, GDP dissociation inhibitors (GDIs) act as soluble lipid chaperones that sequester the geranylgeranyl moiety in a hydrophobic pocket, thereby shifting the equilibrium toward the cytosol. GDI binding is bipartite, involving both the lipid modification and the GTPase switch regions. Because the conformation of these switch regions is determined by the bound nucleotide, the nucleotide state directly regulates the affinity of the GTPase for GDIs and consequently its partitioning between membranes and the cytosol (Cherfils and Zeghouf 2013).

For myristoylated and farnesylated proteins, membrane affinity is typically weak when mediated by the lipid alone (Peitzsch and McLaughlin 1993; Silvius and l’Heureux 1994). As a result, most proteins rely on a second targeting signal such as a polybasic region (e,g, src and KRAS4B) or a palmitoyl group (N and HRAS) to achieve stable membrane association, consistent with the “two-signal” hypothesis (Resh 2006). In this context, regulated exposure or masking of the lipid moiety (as in the Arf myristoyl switch), reversible palmitoylation/depalmitoylation cycles, or phosphorylation-dependent modulation of electrostatic interactions provide a means to regulate membrane binding (Bivona et al. 2006; Pasqualato et al. 2002; Rocks et al. 2005).

For many years, KRAS, NRAS and HRAS proteins were thought to lack a dedicated GDI-like system. Nevertheless, currently there are at least three soluble prenyl-binding proteins, including smgGDS-558, calmodulin, and PDE6D, that are proposed to act as lipid chaperones for farnesylated Ras proteins, and facilitating their cytosolic trafficking (Schaffner-Reckinger et al. 2025). Here, we will highlight the role of PDE6D in trafficking Ras.

2 Phosphodiesterase 6 solubilising protein (PDE6D)

In rod and cone photoreceptors, light-activated rhodopsin stimulates the heterotrimeric G protein transducin, which in turn activates phosphodiesterase-6 (PDE6), a prenylated effector enzyme whose catalytic subunits are anchored to photoreceptor disc membranes. Activation of PDE6 leads to rapid hydrolysis of cGMP, closure of cGMP-gated ion channels, and conversion of photon absorption into an electrical signal (Cote 2021).

Phosphodiesterase-6 delta (PDE6D) is a small protein (∼150 amino acid) which was originally identified in complex with the soluble fraction of rod PDE6 (Gillespie et al. 1989). Binding depends on the prenylated C-termini of the catalytic PDE6 subunits and occurs with low-nanomolar affinity, allowing PDE6D to solubilize and transport PDE6 (Cook et al. 2000; Florio et al. 1996; Zhang et al. 2012). Although first characterized in the retina, PDE6D is broadly expressed in many tissues, suggesting functions beyond photoreceptors (Florio et al. 1996; Zhang et al. 2012).

The photoreceptor outer segment itself represents a highly specialized form of a primary cilium, a non-motile, antenna-like organelle that can be formed by most mammalian cells and that concentrates signaling proteins. This relationship provided an early conceptual link between PDE6D and more general mechanisms of prenylated protein trafficking in cilia. In addition to binding prenylated cargo proteins, PDE6D also participates in prenylation-independent interactions, including with the small GTPases ARL2 and ARL3, whose GTP-bound forms bind PDE6D, which acts as their effector (Linari et al. 1999a,b).

3 PDE6D as a GDI like solubilising factor for RAS

The discovery that PDE6D solubilizes prenylated PDE6, together with the established role of GDIs in extracting and recycling prenylated Rho and Rab GTPases between membranes and the cytosol, while no analogous GDI was known for Ras, and a report that PDE6D can act in a GDI-like manner for Rab13, motivated the search for a comparable system for Ras proteins (Florio et al. 1996; Marzesco et al. 1998).

Guided by this idea, two independent groups reported that PDE6D interacts with several Ras family members (Hanzal-Bayer et al. 2002; Nancy et al. 2002). These conclusions were based on yeast two-hybrid screens using panels of small GTPases and on structural homology: in one case derived from the crystal structure of the PDE6D–ARL2 complex, which revealed similarity to RhoGDI, and in the other from independent homology modeling that also identified RhoGDI as a structural analogue (Hanzal-Bayer et al. 2002; Nancy et al. 2002).

Because of the difficulty in producing fully modified prenylated full-length Ras proteins, direct structural validation of these interactions was initially not possible. This limitation was overcome by the Waldmann and Wittinghofer groups, which enabled the preparation of fully modified Ras proteins and led to the first crystal structures of PDE6D in complex with prenylated Rheb, a prenylated, RAS-like protein (Figure 1A) (Ismail et al. 2011). These structures showed, in agreement with earlier biochemical data, that PDE6D binding does not involve the Ras switch regions, unlike the situation for Rho and Rab GDIs (Figure 1B). A subsequent structure of full-length farnesylated KRAS bound to PDE6D was in agreement with this binding mode (Figure 1C) (Dharmaiah et al. 2016). Quantitative binding studies further demonstrated that, in contrast to the nanomolar affinity of PDE6 for PDE6D, Ras proteins bind PDE6D with affinities in the low micromolar range (Dharmaiah et al. 2016).

Figure 1:

Structural basis of PDE6D-mediated sequestration and release of prenylated Ras proteins. (A) Structure of farnesylated Rheb–GDP bound to PDE6D (PDB 3T5G), showing insertion of the farnesyl group into the hydrophobic prenyl-binding pocket of PDE6D (surface representation). (B) Structure of CDC42–GDP bound to RhoGDI (PDB 1DOA) for comparison. RhoGDI engages both the geranylgeranyl group and the GTPase switch regions via a regulatory arm, in contrast to the predominantly lipid-mediated interaction of PDE6D. (C) Structure of farnesylated KRAS–GDP in complex with PDE6D (PDB 5TB5), illustrating the conserved prenyl-dependent mode of Ras binding. (D) Electrostatic surface representation of PDE6D showing the hydrophobic prenyl-binding cavity. The structure of PDE6D in complex with ARL2·GTP (PDB: 1KSH) is shown and sliced to visualize the closed pocket conformation. To illustrate steric occlusion of the binding site, the farnesylated cysteine was modeled from the PDE6D–Rheb complex (PDB: 3T5G) after superposition of the PDE6D structures. Only the farnesyl moiety from the Rheb complex is displayed. The arrow marks the site at which the ARL2-bound conformation closes off access to the prenyl-binding pocket, thereby preventing accommodation of the lipidated cargo. (E) Superposition of the PDE6D–Rheb complex (PDB: 3T5G) and the ARL2·GTP–PDE6D complex (PDB: 1KSG). PDE6D from the ARL2·GTP complex is shown in yellow, while PDE6D from the Rheb complex is shown in grey. Rheb·GDP is shown in cyan and ARL2·GTP in magenta. The farnesyl group of Rheb (red) occupies the hydrophobic pocket of PDE6D. Structures were aligned via PDE6D to illustrate that ARL2 binding is sterically compatible with prenylated cargo bound to PDE6D, consistent with a ternary complex and supporting an allosteric mechanism of cargo release.

Finally, comparison of the PDE6D–ARL2 and PDE6D–Rheb structures showed that the prenyl-binding pocket of PDE6D becomes occluded upon ARL2 binding (Figure 1D) (Ismail et al. 2011). Because ARL2, PDE6D, and prenylated cargo can form a sterically compatible ternary complex, this indicated that ARL2 does not compete directly with the cargo but instead acts as an allosteric release factor by inducing pocket closure (Figure 1E). This mechanism was later confirmed biochemically. ARL3, which shares ∼52 % sequence identity with ARL2, operates through the same mechanism but with greater efficiency (Ismail et al. 2011).

Quantitative binding studies further demonstrated that, in contrast to the nanomolar affinity of PDE6 for PDE6D, Ras proteins bind PDE6D with affinities in the low micromolar range (Chen et al. 2010; Dharmaiah et al. 2016; Florio et al. 1996; Yelland et al. 2022). To date, numerous potential prenylated PDE6D cargo proteins have been reported (for a comprehensive overview see Zhang et al. (2012)). These cargoes can broadly be divided into low- and high-affinity binders. Low-affinity interactions include prenylated Ras proteins (NRAS, HRAS, and KRAS) as well as Rheb, whereas high-affinity cargoes include PDE6 and INPP5E (Fansa et al. 2016).

4 A model for PDE6D-mediated Ras trafficking

At the time these studies emerged, it was already established that HRAS and NRAS undergo dynamic cycles of palmitoylation and depalmitoylation that continuously redistribute them between intracellular membranes. Palmitoylation occurs primarily at the Golgi apparatus, promoting stable membrane association and forward trafficking to the plasma membrane. In contrast, depalmitoylation, mediated by cytosolic thioesterases, returns Ras to a lower–membrane-affinity state and increases its mobility, thereby enhancing the probability of re-encountering the Golgi apparatus where repalmitoylation can occur (Rocks et al. 2005; Vartak and Bastiaens 2010). This ongoing acylation cycle enforces selective enrichment of HRAS and NRAS at the Golgi and plasma membrane (steady state distribution) (Rocks et al. 2005; Vartak and Bastiaens 2010). On the other hand, the dynamic nature of KRAS4B membrane association was already evident from studies of its polybasic–prenyl membrane-targeting motif. Binding of Ca²⁺/calmodulin and phosphorylation of Ser181 by protein kinase C, and later also by PKG2, promote redistribution of KRAS between membrane compartments (Bivona et al. 2006; Cho et al. 2016; Fivaz and Meyer 2005).

Building on this view of regulated and dynamic trafficking of KRAS, NRAS, and HRAS, PDE6D was proposed as a cytosolic factor that couples these processes to spatial Ras organization. Live-cell imaging experiments using fluorescently tagged Ras showed that PDE6D binds and solubilizes farnesylated KRAS, NRAS, and HRAS, thereby capturing Ras molecules that transiently dissociate from membranes and enabling their redistribution between cellular compartments (Chandra et al. 2012). In the case of KRAS, such dissociation occurs as part of its spatial cycle, in which endocytic trafficking and the lower negative charge of endomembranes weaken electrostatic interactions with the polybasic targeting motif (Schmick et al. 2014). For NRAS and HRAS, depalmitoylation similarly reduces membrane affinity and removes the palmitoyl group that sterically prevents insertion of the farnesyl moiety into the PDE6D prenyl-binding pocket, thereby enabling interaction with the chaperone (Chandra et al. 2012; Dharmaiah et al. 2016). Additional regulatory inputs can further perturb this spatial cycle. For example, phosphorylation of the KRAS polybasic region reduces its electrostatic interaction with membranes and has also been reported to decrease its affinity for PDE6D, indicating that this modification modulates multiple steps of the spatial cycle (Dharmaiah et al. 2016; Schmick et al. 2014).

Together, these observations led to a model in which PDE6D and ARL2 cooperate to maintain plasma-membrane enrichment of KRAS. KRAS molecules that escape the membranes are captured by cytosolic PDE6D and trafficked to perinuclear endomembranes, where ARL2-dependent cargo release has been proposed to occur (Schmick et al. 2014). Electrostatic interactions at these membranes, including the recycling endosome, promote membrane trapping, after which vesicular trafficking pathways return KRAS to the plasma membrane, thereby sustaining spatially organized Ras signaling (Figure 2) (Schmick et al. 2014). However, the precise mechanism underlying ARL2 activation at these membranes remains unclear and represents an important open question. In particular, no dedicated guanine nucleotide exchange factor for ARL2 has yet been identified. Notably, ARL2 displays relatively high intrinsic nucleotide exchange compared with ARL3, raising the possibility that activation may occur without a canonical GEF, although additional layers of regulation cannot be excluded (ElMaghloob et al. 2021).

Figure 2:

Model for PDE6D–ARL2-dependent KRAS trafficking. Schematic illustrating the dynamic cycle that maintains KRAS enrichment at the plasma membrane. Farnesylated KRAS associates with the plasma membrane through combined prenylation and electrostatic interactions mediated by its polybasic C-terminal region. Upon dissociation or endocytic trafficking, KRAS can redistribute to endomembranes. Cytosolic PDE6D sequesters farnesylated KRAS from endomembranes by binding the prenyl group, allowing diffusion through the cytosol. At perinuclear membranes, particularly recycling endosomes, GTP-bound ARL2 promotes KRAS release from PDE6D, enabling membrane capture through electrostatic interactions. Vesicular recycling then restores KRAS to the plasma membrane, sustaining spatially organized KRAS signaling.

5 Inhibition of KRAS through PDE6D

KRAS is one of the most frequently mutated oncogenes in human cancer, with activating mutations present in approximately 90 % of pancreatic ductal adenocarcinomas. For decades, direct targeting of KRAS proved extremely challenging due to the absence of obvious druggable pockets beyond the nucleotide-binding site, which is difficult to target because of the high affinity of KRAS for GDP and GTP and their high intracellular concentrations. The identification of PDE6D as a KRAS trafficking factor, together with its well-defined hydrophobic prenyl-binding cavity, opened a new therapeutic avenue. Structure-guided inhibitor development subsequently yielded high-affinity PDE6D ligands that block prenyl binding and disrupt the interaction between PDE6D and farnesylated cargo. In cellular systems, these compounds mislocalize KRAS from the plasma membrane and modulate downstream signaling outputs, resulting in reduced proliferation of KRAS-dependent cancer cells, while early studies also reported inhibitory effects in xenograft models (Zimmermann et al. 2013).

Following the initial demonstration that PDE6D inhibition can mislocalize KRAS, extensive medicinal-chemistry efforts were undertaken to improve the affinity and selectivity of PDE6D inhibitors. This work resulted in compounds with nanomolar and even subnanomolar affinities for the prenyl-binding pocket of PDE6D (for a comprehensive overview of PDE6D inhibitors, see Schaffner-Reckinger et al. 2025). However, in cellular assays these compounds typically produced only micromolar effects on KRAS–PDE6D engagement, KRAS localization, and downstream signaling (Schaffner-Reckinger et al. 2025). One important factor is that PDE6D has a broad cargo spectrum, so inhibitors must compete not only with KRAS but with many prenylated proteins for pocket occupancy, which might contribute to the relatively high concentrations required to observe cellular effects. Moreover, because PDE6D interacts with numerous prenylated cargo proteins, perturbation of PDE6D function may affect multiple clients, making it difficult to attribute the observed cellular phenotypes solely to disruption of KRAS trafficking. In addition, quantitative cell-based BRET measurements showed that only ∼26 % of functional KRAS membrane organization depends on PDE6D, whereas blocking prenylation itself reduces KRAS membrane association by ∼50 %, indicating that substantial PDE6D-independent trafficking routes exist, likely involving alternative chaperones such as calmodulin or vesicular recycling pathways (Kaya et al. 2024).

Because ARL2 and ARL3 induce closure of the hydrophobic prenyl-binding pocket of PDE6D, concerns were raised that small-molecule inhibitors targeting this site might be allosterically displaced, limiting their efficacy. This prompted the development of covalent PDE6D inhibitors as well as PDE6D-directed PROTACs by the Waldmann group (Winzker et al. 2020; Zhang et al. 2026). Although these PROTACs still bind the prenyl pocket, their mechanism is event-driven rather than occupancy-driven, such that transient engagement is sufficient to trigger ubiquitin-mediated degradation of PDE6D (Winzker et al. 2020; Zhang et al. 2026). Quantitative proteomics showed that treatment with PDE6D PROTACs induced upregulation of sterol-regulatory-element-controlled enzymes of the mevalonate pathway and led to accumulation of cholesterol biosynthetic intermediates, as measured by SRE reporter assays and metabolomics (Winzker et al. 2020). While these observations indicate that perturbation of PDE6D function can influence cellular pathways beyond prenyl-protein trafficking, such effects may arise indirectly through mislocalization of prenylated cargo proteins, several of which, including RAS and Rheb, are known regulators of cellular metabolism. More recently, alternative inhibitor design strategies have been proposed that take inspiration from high-affinity natural PDE6D cargo proteins, in which interactions not only within the hydrophobic prenyl-binding pocket but also at the pocket entrance contribute to binding affinity, thereby offering a potential route to develop higher-affinity PDE6D inhibitors (Gómez-Mulas et al. 2025).

6 Concluding remarks

In summary, PDE6D has emerged as a key solubilizing factor for prenylated proteins that contributes to the spatial organization of KRAS and other lipid-modified cargo proteins. Structural and biochemical studies have revealed how PDE6D recognizes prenylated substrates through its hydrophobic binding pocket and how cargo release is regulated by the small GTPases ARL2 and ARL3. Cellular studies further suggest that PDE6D participates in KRAS trafficking between intracellular membranes and the plasma membrane, although its contribution appears to represent only one component of a broader network of trafficking pathways. These insights have stimulated the development of small-molecule inhibitors, covalent ligands, and PROTAC-based strategies targeting PDE6D, highlighting the possibility of indirectly modulating oncogenic KRAS localization. At the same time, the broad cargo spectrum of PDE6D indicates that its cellular functions likely extend beyond KRAS trafficking and involve the spatial regulation of numerous prenylated proteins.

To date, multiple lines of evidence suggest that PDE6D may be particularly critical for ciliary trafficking. Mutations in PDE6D cause Joubert syndrome, a ciliopathy, indicating an essential role in ciliary biology (Thomas et al. 2014). Moreover, there is a strong correlation between PDE6D binding affinity and ciliary localization: INPP5E, a well-established ciliary protein, binds PDE6D with nanomolar affinity, whereas KRAS and many other cargos bind only in the micromolar range, and it has been proposed that cargo sorting into cilia depends on affinity for PDE6D (Fansa et al. 2016; Humbert et al. 2012; Peeters and Ismail 2025). In addition, the PDE6D release factors ARL2 and ARL3 display distinct cargo specificities: ARL2 releases low-affinity cargos, whereas ARL3 is capable of releasing both low- and high-affinity cargos (ElMaghloob et al. 2021; Fansa et al. 2016; Gotthardt et al. 2015). Because the ARL3 guanine nucleotide exchange factor ARL13B is highly enriched in primary cilia, ARL3 activation is thought to occur preferentially within this compartment (Gotthardt et al. 2015). A model therefore emerges in which low-affinity cargos can be released by ARL2 at extraciliary membranes, whereas high-affinity cargos are preferentially released within the cilium by ARL3 (Peeters and Ismail 2025). Whether PDE6D should ultimately be viewed as a general prenyl-protein chaperone or as a specialized factor for ciliary targeting and sorting high-affinity cargos remains an open and important question.