Research ArticleBiochemistry

Calmodulin disrupts plasma membrane localization of farnesylated KRAS4b by sequestering its lipid moiety

See allHide authors and affiliations

Science Signaling  31 Mar 2020:
Vol. 13, Issue 625, eaaz0344
DOI: 10.1126/scisignal.aaz0344

How Ca2+ inhibits KRAS

The RAS protein family regulates cell proliferation, differentiation, and survival. Mutations in RAS proteins are frequently implicated in driving tumor growth. Grant et al. resolved the structure of the C terminus of the RAS isoform KRAS4b in complex with the Ca2+-sensing protein calmodulin (CaM) and developed FRET sensors to explore the interaction in cells. They found that a farnesylated region of KRAS4b was bound in a hydrophobic pocket of CaM in response to increased intracellular Ca2+, thereby sequestering KRAS4b away from the membrane and dampening its output. Exploiting this sequestration mechanism might yield a new way to inhibit KRAS signaling.


KRAS4b is a small guanosine triphosphatase (GTPase) protein that regulates several signal transduction pathways that underlie cell proliferation, differentiation, and survival. KRAS4b function requires prenylation of its C terminus and recruitment to the plasma membrane, where KRAS4b activates effector proteins including the RAF family of kinases. The Ca2+-sensing protein calmodulin (CaM) has been suggested to regulate the localization of KRAS4b through direct, Ca2+-dependent interaction, but how CaM and KRAS4b functionally interact is controversial. Here, we determined a crystal structure, which was supported by solution nuclear magnetic resonance (NMR), that revealed the sequestration of the prenyl moiety of KRAS4b in the hydrophobic pocket of the C-terminal lobe of Ca2+-bound CaM. Our engineered fluorescence resonance energy transfer (FRET)–based biosensor probes (CaMeRAS) showed that, upon stimulation of Ca2+ influx by extracellular ligands, KRAS4b reversibly translocated in a Ca2+-CaM–dependent manner from the plasma membrane to the cytoplasm in live HeLa and HEK293 cells. These results reveal a mechanism underlying the inhibition of KRAS4b activity by Ca2+ signaling pathways.


RAS-family small guanosine triphosphatases (GTPases) are key regulators of signal transduction pathways in cells, controlling proliferation, differentiation, survival, and apoptosis (1, 2). RAS proteins act as binary switches in response to external stimuli, cycling between an inactive, guanosine 5′-diphosphate (GDP)–bound, and active, guanosine 5′-triphosphate (GTP)–bound form, catalyzed by guanine nucleotide exchange factors and GTPase-activating proteins (3, 4). When localized to the inner leaflet of the plasma membrane (PM), RAS-GTP can bind to and activate several effector proteins involved in many cellular response pathways (1, 35). RAS genes (HRAS, NRAS, and KRAS) are mutated in roughly 30% of all human cancers (3, 4, 6), and KRAS is the most frequently mutated isoform, driving roughly 95% of pancreas, 45% of colon, and 31% of lung adenocarcinomas (7, 8). KRAS encodes two splice variants, KRAS4a and KRAS4b, of which the latter is more ubiquitously expressed and more extensively characterized.

RAS proteins share a highly conserved G-domain but are divergent at the C-terminal ~15 amino acids, designated the hypervariable region (HVR) (1, 3). The HVR of each isoform contains a unique combination of membrane localization signals, believed to be largely responsible for differential signaling and oncogenic capacity (4, 6, 8, 9). All isoforms contain a C-terminal CaaX box motif (C, cysteine; a, aliphatic; X, any residue), which is a target for irreversible farnesylation of cysteine, subsequent cleavage of the aaX residues, and carboxymethylation of the newly exposed C terminus (3, 10). HRAS, NRAS, and KRAS4a contain additional cysteines in the HVR that are reversibly palmitoylated, whereas KRAS4b membrane localization is enhanced through electrostatic interactions between the negatively charged membrane surface and a lysine-rich sequence in the HVR called the polybasic region (Fig. 1A and fig. S1) (1114). Because KRAS4b lacks palmitoylation sites, its affinity for the PM is lower than that of the other RAS isoforms (14).

Fig. 1 CaM interacts strongly with farnesylated KRAS4b and farnesyl compounds.

(A) Schematic of farnesylated KRAS4b purified from High Five insect cells engineered to enhance farnesylation capacity. The aaX (VIM) is cleaved off after farnesylation at Cys185, and the exposed terminal cysteine is carboxymethylated. The polybasic region of the HVR is highlighted in blue. (B) 1H-15N HSQC of uniformly 15N-labeled CaM alone (red) and in the presence of fourfold excess KRAS4b-FMe-GDP (blue). The presence of KRAS4b-FMe induces CSPs and extensive signal broadening. Unbroadened signals that experience perturbations are highlighted (arrows) and used as probes for the interaction. (C) Chemical structures of tested farnesyl-derived compounds. (D) 1H-15N HSQC of uniformly 15N-labeled Ca2+-bound CaM alone (red) and in the presence of 11-fold excess E,E-farnesol (blue). Probe peaks from (B) are marked with arrows in the larger, full spectrum. Regions of the larger spectrum were extracted for a closer view in the boxes marked 1 to 5. All spectra were collected in NMR buffer containing 50 mM Hepes (pH 7.3), 100 mM NaCl, 10 mM Ca2+, and 1 mM TCEP at 25°C and 600 MHz for eight scans. ppm, parts per million.

RAS proteins are well-validated targets in oncology; however, they lack classic druggable pockets and have thus presented a major challenge to drug discovery efforts. A variety of indirect approaches to restrain RAS signaling have been pursued (15), largely unsuccessfully. Recognizing that prenylation of RAS is required for signaling, inhibitors of farnesyl transferases were developed; however, these ultimately failed in the clinic because of alternative prenylation of KRAS4b by geranylgeranyl transferases (16, 17). KRAS4b association with the PM is a particularly dynamic process that can be regulated by phosphodiesterase-δ (PDEδ) (18), a noncatalytic subunit of the photoreceptor PDE6. PDEδ binds the farnesylated HVR of cytoplasmic KRAS4b and delivers it to back to the PM through an active mechanism via recycling endosomes (19). Furthermore, a series of reports suggest that KRAS4b localization can be modulated by calmodulin (CaM) (2024) and phosphorylation of Ser181 in the HVR (2529). A need for novel approaches to inhibit KRAS4b signaling has generated interest in manipulating these systems to perturb the localization of KRAS4b. For example, inhibitors of the KRAS4b interaction with PDEδ have been shown to impair its oncogenic signaling (30, 31). Similarly, modulation of the CaM interaction with KRAS4b could potentially dampen its oncogenic properties, although a better understanding of this interaction is first required.

CaM is a ubiquitous calcium-sensing master regulator that binds a large and diverse group of targets, some in the presence and others in the absence of Ca2+, controlling many important biological pathways. CaM is a small (16.7 kDa), highly flexible protein characterized by two similar globular domains (lobes), each composed of two Ca2+-binding EF hands, separated by a flexible linker (3234). Ca2+ binding to CaM induces a conformational change that exposes a methionine-rich hydrophobic pocket within each lobe (3235). Canonical binding targets of Ca2+-CaM are helical, basic peptides with two hydrophobic “anchor” side chains separated by 2 to 18 residues, although CaM is remarkably plastic and has been shown to bind a variety of noncanonical targets as well (3234). The KRAS4b HVR is highly basic with propensity to form a helix but lacks hydrophobic anchor residues characteristic of a canonical CaM target.

Several groups have reported that CaM interacts with KRAS4b, but not with KRAS4a, HRAS, or NRAS, in a Ca2+-dependent manner, both in cells and in vitro, and that this modulates cell signaling (2024). In addition, it has been suggested that CaM binding extracts KRAS4b from the PM and is responsible for a relocalization to the cytoplasm and potentially internal membranes, altering signal output (21, 23, 25). However, there are contradictions in the literature regarding many of the details of this interaction. Various studies have reported that the interaction is GTP dependent (20, 3638) and nucleotide independent (21, 23, 39, 40); farnesylation dependent (21, 23, 36, 3840) and independent (20, 36, 37); and that the G-domain is either involved (20, 36, 38) or not (21, 23, 37, 39, 40). The effects of the interaction on KRAS4b localization and signaling have also been disputed (21, 23, 36, 39, 40). These inconsistencies have made it difficult for the field to reach a consensus about the nature and role of the interaction. Because of the plasticity of CaM and the lack of canonical CaM binding motifs in KRAS4b, very little is currently understood about the structural and biophysical properties of this complex.

In the first part of our study, we sought to investigate the structure and biophysical properties of the KRAS4b-CaM interaction in vitro. In particular, we examined Ca2+ and nucleotide dependence, identified interacting elements on both proteins, and assessed the ability of CaM to extract KRAS4b from lipid bilayers using a combination of nuclear magnetic resonance (NMR) and x-ray crystallography. Our NMR data show that CaM binding to KRAS4b is dependent on Ca2+ and KRAS4b farnesylation but found no evidence for involvement of the KRAS4b G-domain and no preference for the nucleotide bound to KRAS4b. A crystal structure of Ca2+-CaM in complex with a fragment of the processed HVR, farnesyl cysteine methyl ester (FCME), showed the farnesyl moiety buried deeply in the C-lobe hydrophobic pocket of CaM, providing a structural rationale for extraction of KRAS4b from the membrane by CaM. To validate this mechanistic model, we used live-cell microscopy using a chimeric CaM-KRAS4b fluorescence resonance energy transfer (FRET) probe designed on the basis of our structural data. This FRET-based biosensor enabled us to simultaneously monitor the interaction of CaM with KRAS4b and intracellular translocation of the complex upon stimulation of a cellular Ca2+ signal in both HeLa and human embryonic kidney– 293 (HEK293) cells. We demonstrate that increased intracellular calcium stimulates KRAS4b translocation from the PM to the cytoplasm, which is spatially and temporally regulated via Ca2+-CaM in mammalian cells.


KRAS4b-CaM interaction is dependent on farnesylation

Some reports have described an interaction between CaM and unfarnesylated KRAS4b (20, 36, 37); thus, we began to test the interaction using unprocessed KRAS4b expressed in Escherichia coli (residues 1 to 185, representing the full-length protein after -aaX cleavage but lacking farnesylation and carboxymethylation). We collected 1H-15N heteronuclear single quantum correlation (HSQC) NMR spectra of 15N-labeled KRAS4b or 15N-CaM and analyzed chemical shift perturbations (CSPs) upon the addition of the other unlabeled protein. These experiments were performed with KRAS4b in both GDP- and GTPγS-loaded states and in the presence and absence of Ca2+. In the presence of Ca2+, a few minor CSPs were observed in both proteins independent of which nucleotide was bound; however, no large CSPs that would be indicative of a strong or specific interaction were detected (fig. S2, A to D). Some of these small CSPs were likely because of nonspecific electrostatic interactions between the basic lysine residues in the HVR and the acidic surface of CaM. Notably, the CSPs in both proteins were more appreciable when KRAS4b was GTP-loaded. Binding to GTP primes RAS for interaction with its binding partners, which may likewise make it more prone to nonspecific interactions as well. Nevertheless, we do not consider these minor changes relevant in comparison to the large CSPs typically observed upon binding of CaM to canonical targets such as protein kinases and or channels (4143). Hence, we conclude that CaM only makes weak transient contacts with unfarnesylated KRAS4b in solution.

To determine whether CaM binds to farnesylated KRAS4b, we examined the interaction of CaM with full-length, fully processed KRAS4b (designated KRAS4b-FMe, which is farnesylated and has the C-terminal tripeptide removed and the Cys185 carboxymethylated, as illustrated in Fig. 1A) expressed in and purified from insect cells using a baculovirus expression protocol developed by National Cancer Institute (NCI) Frederick (44). In contrast to the unfarnesylated protein, addition of KRAS4b-FMe to 15N-CaM in the presence of Ca2+ induced widespread changes in the 15N-CaM spectrum, affecting most of the cross-peaks (Fig. 1B). Specifically, KRAS4b-FMe induced severe line broadening of ~80% of the CaM resonances and CSPs in several of the peaks that were not broadened beyond detection, indicating a farnesyl-dependent interaction between KRAS4b-FMe and CaM. The peak broadening is likely caused by several factors, including the increased molecular mass of the complex, transient exchange between bound and unbound states, and chemical exchange from nonspecific contacts between the G-domain of KRAS4b and CaM. The broadening of NMR signals upon titration of KRAS4b-FMe into CaM has been corroborated (40). Several residues within the C-lobe, Ser81, Leu116, Ala128, Ile130, and Ala147, exhibited chemical shift changes, indicative of a more stable and specific interaction (Fig. 1B, arrows). Although the CSPs provide probes of the interaction, the extensive signal broadening limited the structural information that could be extracted from these spectra. Therefore, to improve spectral quality and gain insights into the structural mechanism, we sought to identify the minimal binding elements that mediate the interaction between KRAS4b-FMe and CaM.

Farnesyl moiety is sufficient for CaM binding

With the knowledge that farnesylation is necessary for the KRAS4b-CaM interaction, we first questioned whether the farnesyl moiety alone is sufficient for CaM binding. Using a series of farnesyl-derived compounds (Fig. 1C), we assessed CSPs in the 1H-15N HSQC spectra of CaM. First, (E,E)-farnesol, an alcohol derivative of the primary stereoisomer present in mammalian cells, was solubilized in dimethyl sulfoxide (DMSO) and titrated into 15N-CaM in the presence of Ca2+. The resulting 1H-15N HSQC spectra exhibited extensive CSPs of almost all cross-peaks, indicating that CaM undergoes a substantial structural rearrangement upon addition of (E,E)-farnesol (Fig. 1D). During titration of farnesol, CaM peaks did not exhibit the broadening caused by KRAS4b-FMe titration, and the five probe peaks identified above exhibited similar chemical shift changes in both spectra. A previous titration of Ca2+-CaM with a farnesylated KRAS4b C-terminal 6-mer peptide (KSKTKC-FMe) produces a CaM spectrum nearly identical to ours in the presence of farnesol alone (40), suggesting that the farnesyl moiety may recapitulate the KRAS4b-CaM binding mode. The spectrum also resembles those of Ca2+-CaM in complex with myristoylated polybasic peptides (fig. S3) (45, 46). Farnesol titration does not cause peak broadening due to its smaller size and the absence of chemical exchange between CaM and a dynamic G-domain. The chemical shift changes have been plotted by residue and mapped onto the surface of CaM (fig. S4). In the absence of Ca2+, no CSPs were apparent, indicating that this interaction is fully Ca2+ dependent (fig. S5), consistent with every previous report of the KRAS4b-CaM interaction.

To investigate the stereospecificity of this lipid:CaM interaction, we titrated the natural, rare stereoisomer (Z,Z)-farnesol (Fig. 1C) into 15N-CaM. The CaM spectra in the presence of the two stereoisomers were very similar (Fig. 1E), suggesting that structural plasticity of CaM can accommodate either conformation of the lipid by virtue of the flexible linker and malleable methionine-rich hydrophobic pockets (33, 47, 48). We titrated 15N-CaM with (E,E)-FCME (Fig. 1C), comprising the fully processed C-terminal residue of KRAS4b, which showed that extension of farnesol to include the carboxymethylated cysteine moiety caused minor changes in the pattern of CSPs (Fig. 1E) and promoted saturation at a lower molar ratio, suggesting that this amino acyl component contributes to binding affinity. Last, all assigned CaM peaks could be traced to saturation throughout the titrations with farnesyl compounds (fig. S6) and were found to overlay with the subset of CaM peaks that remained visible upon titration with KRAS4b-FMe (fig. S7), suggesting that CaM interaction with the lipid comprises the core of the KRAS4b-FME–CaM complex (schematic in Fig. 2A). This finding is supported by a previous crystal structure of Ca2+-CaM in complex with the myristoylated N-terminal polybasic peptide from CAP23/NAP22, in which the myristoyl moiety is directly sequestered (49), but the relative contribution of the peptide versus lipid moieties to CaM binding was not assessed.

Fig. 2 Crystal structure of CaM in complex with FCME.

(A) Schematic diagram of the CaM-KRAS4b-FMe interaction based on the NMR data presented in Fig. 1; the hydrophobic core of CaM sequesters the farnesyl moiety of KRAS4b-FMe. (B to E) X-ray crystallography structure of Ca2+-CaM bound to FCME, resolved to 1.8 Å (PDB: 6OS4). The cartoon diagram of the CaM-FCME structure (B) shows a compact CaM conformation with significant interlobe contacts. FCME is buried in the CaM C-lobe hydrophobic pocket with the aminoacyl (C185) group exiting through a gap between α-helices 1, 6, and 7. The surface electrostatic representation of CaM (C) shows the channel formed by α-helices 1, 6, and 7 through which the cysteinyl moiety of FCME exits [rotated 45° relative to (B)] and the highly acidic surface accessible to an adjacent polybasic HVR. Electrostatics were modeled with the APBS PyMol plugin. Panel (D) shows a zoomed perspective of FCME in the CaM C-lobe hydrophobic pocket [rotated 90° relative to (B), into the interlobe plane]. Ten residues involved in hydrophobic contacts with the ligand are depicted as sticks and labeled. Panel (E) shows a close-up surface representation of FCME in the C-lobe hydrophobic pocket [oriented as in (D)] with interacting residues colored gray. (F) A ligand interaction map highlighting the hydrophobic interactions between FCME and the CaM C-lobe pocket.

Crystal structure of CaM in complex with FCME reveals C-lobe binding

To obtain an atomic-resolution picture of the interaction, we used x-ray crystallography. Multiple crystallization trials (~300 conditions) of Ca2+-CaM complexed with (E,E)-farnesol were unsuccessful; however, the FCME complex produced crystals in 66 conditions, suggesting that the cysteine moiety in FCME may be involved in the specificity of binding, thus reducing structural heterogeneity. One of these conditions was optimized (see Materials and Methods), and the best crystal was diffracted to 1.8 Å and 2.0 Å with anomalous data, using a synchrotron x-ray source. Initially, density maps were generated separately by (i) molecular replacement using a structure of CaM bound to the myristoylated CAP23/NAP22 peptide [Protein Data Bank (PDB): 1L7Z] (49) as a search model and (ii) a priori with experimental phasing from anomalous data, primarily scattered by Ca2+ ions. Both approaches yielded highly similar structures. The two datasets were combined in HKL2000 and refined in Phenix to yield a high-resolution map of the CaM-FCME complex (crystallographic statistics found in Table 1).

Table 1 Crystallographic statistics reported for deposited CaM-FCME complex structure (PDB: 6OS4).

This crystal was grown at room temperature in 0.2 mM sodium acetate, 0.1 mM sodium cacodylate (pH 6.7), and 28% PEG 8000, with 1.25 mM CaM and 5 mM FCME, using the sitting drop-vapor diffusion method. The protein crystal was diffracted on the Canadian Macromolecular Crystallography Facility 08ID-1 Beamline at the Canadian Light Source at two wavelengths to collect native (n = 1) and anomalous scattering (n = 1) datasets, which were processed using HKL2000 and refined with Phenix. Values in parentheses are for the highest resolution shell. a, b, c, unit cell measurements; α, β, γ, unit cell angles; RMS, root mean square.

View this table:

In this structure, the N- and C-lobes of CaM come together to form a compact globular conformation with a hydrophobic core and acidic surface (Fig. 2, B and C), akin to many reported structures of CaM-target complexes (50). Our structure is globally similar to that of CaM in complex with the myristoylated CAP23/NAP22 peptide; however, several differences are apparent (fig. S8A). In the CAP23/NAP22 structure, the saturated, unbranched myristoyl moiety extends through a hydrophobic channel formed by both lobes, whereas our structure shows the farnesyl moiety of FCME, which is unsaturated and branched, buried deep within the C-lobe hydrophobic pocket of CaM (Fig. 2, D and F), and the N-lobe pocket is minimally exposed (fig. S8B). It is clear from the electron density that the farnesyl moiety is exclusively anchored within the C-lobe pocket, although multiple conformations likely coexist. In the major conformation, the cysteine moiety of FCME protrudes through an opening between α-helices 1, 6, and 7 (Fig. 2B and fig. S9A), whereas weak additional electron density suggests that FCME may exist in an alternate low-occupancy conformation in which the aminoacyl end exits the hydrophobic core from the opposite side. The major conformer exhibits strong, well-defined electron density (fig. S9, B and C), and the C-terminal cysteine residue occupies a position equivalent to that of the N-terminal myristoylated glycine in CAP23/NAP22; thus, the adjacent polybasic sequences would be tethered in the same region in both complexes, albeit with opposite strand directions.

This structure indicates that CaM sequesters the farnesyl moiety in its hydrophobic core and suggests that electrostatic interactions may form between the acidic surface of CaM (accessible acidic residues include Asp7, Asp11, Asp14, Asp114, Asp120, Asp123, and Asp127 and the linker, Asp82, Asp83, and Asp84) and the basic HVR of KRAS4b (Fig. 2C), whereas this interaction does not directly involve the G-domain. To gain more insight into this structure in solution, we performed NMR experiments using a paramagnetic relaxation enhancement (PRE) probe in which a 2,2,6,6-tetramethylpiperidine 1-oxyl (TEMPO) spin-labeled molecule was conjugated to the free-amine group of FCME (51, 52). TEMPO-FCME was added to 15N-CaM in solution, and 1H-15N HSQC spectra were recorded (fig. S10A). Many resonances were broadened by the spin label, but the subset of peaks that were not broadened overlaps the spectrum of CaM bound to untagged FCME, indicating that the TEMPO tag has minimal impact on the interaction of CaM with TEMPO-FCME. The residues associated with PRE-induced peak broadening were mapped onto our crystal structure (fig. S10, B and C). The largest PRE effect was observed for those residues near the α1,6,7 exit site, and another less intense PRE cluster was observed around the putative alternate exit site. This PRE broadening around α1,6,7 involves both lobes, consistent with our crystal structure and supporting the presence of a compact CaM configuration with interlobe contacts in solution.

Membrane and CaM binding of KRAS4b are mutually exclusive

Because the farnesyl group and the HVR both contribute to KRAS4b membrane localization, we sought to determine whether KRAS4b binding with CaM and with lipid bilayers is compatible or mutually exclusive. We used 15N-lysine–labeled KRAS4b-FMe to probe the HVR, as it contains nine lysine residues in the terminal 15 amino acids. As previously described, we performed NMR experiments to investigate the KRAS4b interaction with membranes using a synthetic lipid bilayer nanodisc (ND) composed of 80% dioleoylphosphatidylcholine (DOPC) and 20% dioleoylphosphatidylserine (DOPS) (53). Five signals from lysine backbone amides in the KRAS4b-FMe HVR are well resolved and assignable from previous work (Fig. 3A) (54). Upon binding to ND (1:5 ND: KRAS4b-FMe), these signals disappear, presumably due to broadening from rapid chemical exchange as a result of transient electrostatic interactions between KRAS4b-FMe (farnesyl and the HVR) and the negative surface of the ND (Fig. 3B). Subsequent addition of Ca2+-CaM to this sample restored the intensity of these HVR lysine signals (Fig. 3C), and the resulting spectrum overlays with that of 15N-lysine KRAS4b-FMe complexed with Ca2+-CaM in the absence of NDs (Fig. 3D), which is distinct from the spectrum of KRAS4b-FMe alone in solution (Fig. 3E). These data demonstrate direct competition between NDs and CaM for KRAS4b-FMe binding, consistent with the fact that they share the same interacting elements. This is also consistent with the reported affinity of CaM for KRAS4b-FMe [dissociation constant (Kd) of 0.3 to 0.4 μM] (40), which is an order of magnitude higher than the KRAS4b-FMe affinity for the lipid bilayer (Kd of 4 μM) (44).

Fig. 3 Alternative isotopic labeling schemes of KRAS4b-FMe show CaM binding is mutually exclusive with membrane association.

(A to E) Fully processed KRAS4b-FMe, specifically labeled with 15N-lysine in insect cells, provides 5 NMR probes in the HVR and 11 probes in the G-domain. The chosen window highlights the lysine residues in the HVR (in italics). Samples contain 20 μM acetyl glycine (AcGly) as an internal standard. Shown are 1H-15N HSQC spectra for (A) KRAS4b-FMe alone; (B) KRAS4b-FMe after addition of NDs composed of 80% DOPC and 20% DOPS (fivefold excess KRAS4b-FMe); (C) KRAS4b-FMe in the presence of NDs and Ca2+-CaM (1:1 KRAS4b-FMe molar ratio); (D) KRAS4b-FMe and Ca2+-CaM in the absence of NDs, overlaid with (C), in the presence of NDs; and (E) KRAS4b-FMe and Ca2+-CaM overlaid with (A), KRAS4B-FMe alone. Data are representative of two experiments each. (F) Model of the predicted interactions between KRAS4b-FMe, membranes (NDs), and CaM in the presence and absence of Ca2+ using paramagnetic relaxation enhancement (PRE). Without Ca2+, KRAS4b-FMe is associated with the membrane through HVR electrostatics and farnesyl insertion and experiences significant peak broadening by membrane embedded PRE-active Gd3+. After addition of Ca2+ and activation of CaM, KRAS4b-FMe is extracted from the membrane by CaM and is thus unaffected by PRE-broadening at the membrane. (G) Experimentally derived probe peaks for PRE interactions (Ile24, Val45, Ile55, and Leu113) within KRAS4b-FMe, specifically methyl-labeled with 13C-Thr, Ile, Leu, Val, and Met (TMILV) in an insect cell expression system, in the presence or absence of 5 mM Ca2. A PRE reading of 1 means complete loss of signal, and a reading of 0 signifies no change in peak intensity relative to pre-Ca2+. Spectra were normalized by 0.6 mM sodium trimethylsilylpropanesulfonate (DSS) internal standard peak intensities. Error bars represent spectral noise. Data are representative of two experiments.

To further validate this observation, we used PRE experiments to directly track KRAS4b-FMe membrane association. As previously described (54, 55), we monitored the signals of 13C-methyl–labeled KRAS4b-FMe (a total of 92 methyls from Thr/Ile/Leu/Val/Met are labeled) in the presence and absence of NDs containing an embedded Gd3+-chelating lipid. This scheme enabled us to identify, through PRE-induced line broadening, the extent of the interaction of KRAS4b-FMe with the membrane surface (schematic in Fig. 3F). In the presence of NDs and Ca2+-free CaM, peaks from several residues—especially Ile24, Val25, Ile55, and Leu113—which are proximal to the membrane surface in the GDP-bound state (54, 55), were strongly broadened. Addition of Ca2+ alleviated the PRE effect on KRAS4b-FMe, restoring peak intensities, consistent with a loss of membrane association of KRAS4b-FMe in the presence of Ca2+-CaM (Fig. 3G; full plots in fig. S11, A and B). Together, these results indicate that Ca2+-CaM effectively extracts KRAS4b-FMe from membranes via the sequestration of the farnesyl group in a Ca2+-dependent manner.

CaM-KRAS4b chimera construct displays Ca2+-dependent FRET in vitro

To observe the interaction between KRAS4b-FMe and CaM and monitor the localization of the proteins in cells, we prepared constructs encoding these proteins as fusions with FRET donor/acceptor fluorophores—mTurquoise2 (mTq2) and super yellow fluorescent protein 2 (SYFP2) (i.e., mTq2-CaM and SYFP2-KRAS4b)—and transiently coexpressed them in HeLa and HEK293 cells. To increase the expression of KRAS4b to levels achieved for CaM, it was necessary to optimize the codon usage by removing rare codons (56, 57). Overexpression of CaM is toxic to cells due to its perturbation of death-associated pathways (58); thus, coexpression of KRAS4b and CaM resulted in morphological changes in HeLa cells, including cell rounding, detachment, and cell death 1 to 2 days after transfection. These effects were mitigated by fusing the proteins into a chimeric construct (mTq2-CaM-SYFP2-KRAS4b). Because both KRAS4b and CaM have multiple interaction partners, linking the two proteins may favor the intramolecular interaction over other binding partners while also ensuring equimolar expression of KRAS4b and CaM. We verified that RAS functions including nucleotide exchange, sensitivity to SoS, and effector binding are preserved in this fusion construct (fig. S12). Our model (Fig. 4A) predicts that this chimera would localize to the PM through a farnesylated HVR under the resting state, whereas elevated Ca2+ would lead to binding of the CaM domain to the farnesyl moiety. The structural rearrangement from an unbound state on the membrane to a KRAS4b:CaM interaction in the cytosol would perturb the orientation and proximity of the two fluorophores and potentially alter FRET efficiency. We have named this chimeric construct “CaMeRAS” for its function as a “camera” to probe the CaM and KRAS4b interaction.

Fig. 4 CaMeRAS FRET construct demonstrates Ca2+- and farnesyl-dependent FRET signal change in vitro.

(A) Schematic of chimeric construct expressed in HEK293 and HeLa cells and bimodal use for CaM-KRAS4b binding model validation. When transiently expressed in mammalian cells, the chimeric construct is natively processed at the C-terminal CaaX box, allowing tethering of the construct to the PM. Upon Ca2+ influx, Ca2+-CaM extracts KRAS4b from the membrane, visible with fluorescence imaging, and the structural rearrangement of the chimera leads to detectable FRET. (B) FRET signal (SFYP2/mTq2 emission ratio) changes during in vitro Ca2+ titration into purified CaMeRASWT (black) CaMeRAS bearing a mutation that prevents farnesylation (CaMeRAS185A:FLAG-mTq2-CaM-SYFP2-KRAS4b-C185Astop, orange) or mutations that block Ca2+ binding (CaMeRAS1234Q:FLAG-mTq2-CaM[E31Q, E67Q, E104Q, E140Q]-SYFP2-KRAS4b, red). Data are presented as means ± SEM of three independent measurements. FRET signal was calculated by using the peak intensities of mTq2 (at 472 nm) and SYFP2 (at 522 nm) in emission spectra as shown in fig. S12. Data were fit with sigmoidal dose-response equation on Prism8 software (GraphPad). (C) In vitro Ca2+-dependent CaMeRAS FRET changes in the presence of 0.5 mM Ca2+ and after addition of 2.5 mM EDTA. CaMeRASWT, CaMeRAS185A, and CaMeRAS1234Q are described in (B). Data are presented as means ± SEM of three independent measurements.

To test whether CaMeRAS exhibits a FRET response to changes in Ca2+ concentration, a FLAG-CaMeRAS construct was expressed in HeLa cells, purified, and titrated with Ca2+ in vitro (fig. S13). CaMeRAS exhibited a small (5.5%) but statistically significant and reliable increase in FRET efficiency in a dose-dependent manner upon addition of Ca2+. In vitro Ca2+ titrations of a farnesylation-deficient CaMeRAS variant, i.e., KRAS4b-C185Astop (CaMeRAS185A) elicited a small FRET response, likely induced by compaction of CaM upon binding Ca2+. Consistently, a calcium-binding deficient CaMeRAS mutant with an E to Q mutation in each EF hand, i.e., CaM E1234Q [CaMeRAS1234Q, EF1-E31Q, EF2-E67Q, EF3-E104Q, and EF4-E140Q (59, 60)] exhibited no change in FRET in response to Ca2+. These data demonstrate that CaMeRAS senses a Ca2+- and farnesyl-dependent, specific interaction between KRAS4b-FMe and CaM (Fig. 4B), which is reversible upon chelation of Ca2+ by EDTA (Fig. 4C). The maximal FRET drops below the Ca2+-free baseline after chelation for both CaMeRASWT and CaMeRAS185A, likely because there is residual Ca2+ in buffers before Ca2+ titration. The Ca2+-insensitive CaMeRAS1234Q did not exhibit a change upon addition of Ca2+ or EDTA. Having developed and validated this FRET construct in vitro, we sought to validate our binding model in vivo.

Ca2+ stimulation induces reversible KRAS4b-FME–CaM interactions and membrane extraction in cells

To fully assess the behavior of the KRAS4b-CaM interaction in mammalian cells, we used live-cell fluorescence imaging using the CaMeRAS constructs (Fig. 5A) transiently expressed in HeLa cells. As expected, the farnesyl-containing constructs CaMeRASWT and CaMeRAS1234Q were concentrated on the PM in resting cells, with a fraction visible in the cytoplasm or on endomembranes, whereas the non-farnesylated CaMeRAS185A exhibited constitutive cytoplasmic localization (Fig. 5B). Induction of Ca2+ influx by addition of the calcium ionophore ionomycin (fig. S14, A and B) or histamine, which produces a physiologically relevant Ca2+ cascade, caused rapid relocalization of CaMeRASWT off the PM, consistent with CaM sequestration of the farnesyl moiety (Fig. 5C). CaMeRAS1234Q, which is defective for Ca2+-binding, retained PM localization following Ca2+ stimulation, and the cytoplasmic localization of CaMeRAS185A was not affected by Ca2+ influx (Fig. 5C). The maintenance of PM localization of CaMeRAS1234Q after Ca2+ influx indicates that the observed internalization of CaMeRASWT is not due to charge-shielding effects of Ca2+ ions at the PM. Representative graphs of CaMeRAS construct localization over time, based on the fluorescence intensity of mTq2 in line-scan cross sections of HeLa cells after histamine stimulation (Fig. 5D), illustrate time-dependent translocation of CaMeRASWT from the PM to the cytoplasm in response to histamine.

Fig. 5 Tracking cellular localization of CaMeRAS constructs in HeLa cells following histamine stimulation.

(A) Schematic of CaMeRAS constructs transiently expressed (CaMeRASWT, CaMeRAS185A, and CaMeRAS1234Q), with a red “X” indicating where the mutants are deficient. The farnesyl moiety is represented by the zig-zag line. (B to D) Live-cell imaging tracking the cellular localization (mTq2) of constructs before (B) and 60 s after (C) 100 μM histamine stimulation. Panel (D) shows representative mTq2 intensity distributions at the indicated cross sections [yellow lines in (B)] over time, from 0 s [as shown in (B)] to 160 s; the 120-s time point is as shown in (C); histamine was added at 60 s (yellow arrowheads) after starting imaging. Scale bars, 10 μm. Data are representative of five experiments, each analyzing five to six cells.

Having demonstrated the Ca2+-induced translocation of CaMeRASWT in live cells, we sought to further examine whether this phenomenon is synchronized with Ca2+ signaling and how translocation may be coupled to the Ca2+-dependent FRET change observed in vitro. Using Calbryte 630, a fluorescent Ca2+ indicator with an absorbance spectrum discrete from mTq2 and SYFP2, we were able to extend our in-cell fluorescence system to simultaneously coimage Ca2+, mTq2, and SYFP2 in response to histamine stimulation. Imaging mTq2 confirmed the CaMeRAS localization and translocation described above (Fig. 6A), whereas concurrent Ca2+ imaging showed a cytosolic increase of Ca2+ after histamine stimulation of these HeLa cells (Fig. 6B). In cells expressing CaMeRASWT, histamine stimulation produced a large and sustained Ca2+ signal, possibly amplified by interactions between CaM and ion channels on the membrane, which was reversible with the addition of the cell permeable Ca2+ chelator BAPTA-AM. Cells expressing CaMeRAS185A or CaMeRAS1234Q exhibited a more physiological Ca2+ response to histamine, possibly due to lack of membrane localization or Ca2+ binding, respectively. Segmenting the images into PM and cytosolic areas to assess localization showed internalization of CaMeRASWT, which was reversible with the addition of BAPTA-AM, whereas CaMeRAS185A retained cytoplasmic localization and CaMeRAS1234Q remained localized on the membrane (Fig. 6C), consistent with our previous results.

Fig. 6 CaMeRAS demonstrates reversible Ca2+-/farnesyl-dependent internalization and FRET change in live cells.

(A) Cellular localization (mTq2) of CaMeRAS constructs in HeLa cells before and after 100 μM histamine stimulation. BAPTA-AM (10 μM) treatment was added to HeLa cells expressing CaMeRASWT to examine reversibility. Regions of interest on the PM or in the cytosol were drawn manually as shown with white outlines. Data are representative of five experiments, each analyzing 5 to 10 cells. (B and C) In the cells described in (A), cytoplasmic Ca2+ concentration [assessed as normalized Calbryte 630 intensity tracking; (B)] and cytoplasmic and PM localization of CaMeRAS constructs (C) were traced over time after histamine stimulation and, for CaMeRASWT cells, BAPTA-AM treatment (red trace). In these traces, the bold lines denote the mean, and the shadows represent the SEM from five independent cells pooled from a single representative experiment. (D) Histograms of FRET readouts (ratio of SYFP2/mTq2 emissions) of the localization of CaMeRAS constructs [as aligned under (A)] at the PM and cytoplasm before (Cyt-Pre) and after (Cyt-Post) histamine stimulation (and for the WT construct, BAPTA-AM; “BAPTA”) in the marked regions of interest (see fig. S14). Data fitting of the normal distributions was performed using Gaussian functions, and data were pooled from 15 to 20 cells across three independent experiments.

In the same cells, FRET was measured by comparing the ratio of mTq2 to SYFP2 within many small segments along the PM and in the cytoplasm, before and after histamine stimulation (fig. S15A). Peak FRET values taken from Gaussian-modeled histograms of these data showed that CaMeRASWT exhibited a 7% reversible FRET increase upon Ca2+-induced translocation from the membrane to the cytoplasm (Fig. 6D and fig. S15B). A minor (<2%) FRET change was observed in the cytosol for CaMeRAS185A, consistent with our in vitro titration data, likely due to the conformational change of CaM in response to Ca2+ binding. No detectable FRET change was observed on the PM for CaMeRAS1234Q (Fig. 6D and fig. S15C). Statistics for the FRET histogram analyses are summarized in tables S1 and S2, and videos show each CaMeRAS construct’s response to histamine stimulation (Ca2+, mTq2, and SYFP2 coimaging) (movies S1 to S5). To further validate the observed in vivo FRET, the CaMeRAS were visualized under the same conditions using fluorescence-lifetime imaging, a sensitive measure of the rate at which excited fluorophores decay to their ground states, which is affected by changes in structure, dynamics, and fluorescent energy transfer. After histamine stimulation, the cytoplasmic CaMeRASWT displayed a statistically significant decrease in fluorescence lifetime relative to the membrane-localized protein before stimulation (fig. S16). The reduction in fluorescence lifetime, indicative of a change in distance and relative orientation between mTq2 and SYFP2, combined with the reversible internalization of the sensor, strongly supports our model of Ca2+-dependent sequestration of KRAS4b farnesyl by CaM, which outcompetes PM binding. Stimulation of HEK293 cells with 100 μM adenosine 5′-triphosphate (ATP) to elicit a Ca2+ signal produced similar results (fig. S17), confirming the robustness of the Ca2+-dependent translocation of CaMeRAS.


The interaction between KRAS4b and CaM has been studied for decades, with increasing intensity in recent years as therapeutic potential was recognized. However, many of the reported details of the interaction have been inconsistent. Here, we combined biophysical studies on fully processed full-length KRAS4b with observations of an engineered biosensor in live cells and demonstrated a Ca2+-dependent, nucleotide-independent interaction mediated by hydrophobic sequestration of the KRAS4b farnesyl moiety within the CaM C-lobe. This interaction directly competes with membrane association and causes relocalization of KRAS4b within cells. Our atomic-resolution structure of Ca2+-CaM complexed with the farnesyl cysteine moiety revealed intimate interactions between the hydrophobic acyl chain and hydrophobic residues of the C-terminal lobe (Phe92, Ile100, Leu105, Met109, Met124, Ile125, Ala128, Val136, Phe141, and Met144). Using this structure, we designed the biosensor CaMeRAS, which enabled visualization of Ca2+-dependent, reversible translocation of KRAS4b from the PM to the cytoplasm in HeLa and HEK293 cells, supporting the notion that canonical plasma membrane KRAS4b signaling is dampened by Ca2+-CaM. Furthermore, this competition between membrane anchoring and CaM binding only occurs upon elevation of cytoplasmic Ca2+ concentrations.

Sequestration of KRAS4b farnesyl in the CaM C-lobe hydrophobic pocket forms the “core” of the KRAS4b-CaM interaction, which is likely further stabilized by electrostatic contacts between the KRAS4b HVR and the acidic CaM surface. However, previous biophysical studies demonstrated that isolated CaM N- and C-lobes are each capable of binding KRAS4b-FMe, with the C-lobe exhibiting three- to eightfold higher affinity (40), consistent with our finding that CaM mutants with N- or C-lobe Ca2+-binding deficiencies still bind farnesyl-derived compounds through the intact lobe. The higher-affinity C-lobe binding has also been predicted by molecular dynamics studies of the farnesylated KRAS4b HVR in complex with CaM (39, 61). Among many models, the two highest confidence interaction modes involved the farnesyl group buried deeply within either the N-lobe or C-lobe hydrophobic pocket and electrostatic interactions between the KRAS4b polybasic region and the surface of CaM. The C-lobe binding model is markedly similar to our crystal structure and has nearly twice the calculated binding free energy of the N-lobe binding model (39), consistent with biophysical results (40). Despite this difference in affinities, a 2:1 stoichiometry for the interaction of KRAS4b-FMe with intact CaM has been reported, which is reduced to 1:1 with the truncation of CaM N- or C-lobes (40). We believe that our crystal structure exhibits exclusive C-lobe binding primarily due to its higher affinity and the FCME concentration used in crystallization, whereas 2:1 binding becomes feasible under more saturating conditions.

An analagous CaM-lipoprotein interaction was previously observed in a crystal structure of N-terminally myristoylated polybasic peptide (CAP23/NAP22) in complex with CaM, in which the myristoyl was sequestered within the hydrophobic core (PDB: 1L7Z) (49). This CaM-myristoyl complex somewhat resembles our CaM-farnesyl structure, but there are two significant differences: (i) the orientation of the lipid within the hydrophobic core and (ii) the position of α4 (fig. S8A). The myristoyl group is found in a linear orientation, interacting with hydrophobic surfaces of both lobes, but not deeply inserted in the pockets, whereas FCME is bound in the C-lobe pocket only, with the N-lobe pocket minimally exposed (fig. S8B). These differences likely stem from the unsaturated and branched farnesyl moiety being shorter and bulkier than myristoyl. In our CaM:FCME complex, the farnesyl moiety adopts a bent conformation, similar to that observed with PEX19 (62) and aristolochene synthase but differing from the extended conformations observed with farnesyltransferase (63) and PDEδ (64). The C-terminal cysteine residue occupies a position equivalent to that of the N-terminal myristoylated glycine in CAP23/NAP22, thus tethering their adjacent polybasic sequences in the same region, albeit with opposite strand directions.

The myristoyl/farnesyl-bound CaM conformations are distinct from CaM in complex with canonical peptides but are remarkably similar to CaM in complex with the unusual target myristoylated alanine-rich C-kinase substrate (MARCKS) (PDB: 1IWQ). This internal MARCKS peptide, which is not myristoylated, assumes an extended conformation with a short helix presenting a Phe side chain that is buried deeply in the C-lobe and a Leu two residues away that contacts a closed N-lobe (65).

CaM interactions with other myristoylated polybasic peptides similar to CAP23/NAP22, namely, the vSRC N-terminal peptide pp60v and the HIV1 Nef1 peptide, were also previously studied by NMR (45, 46). Additions of these peptides to 15N-CaM induced CSPs extremely similar to those induced by FCME (fig. S3). Other small GTPases with geranylgeranylated C-terminal polybasic sequences have also been reported to interact with CaM (21, 6668), suggesting that singly lipidated, terminal polybasic sequences may represent a general noncanonical CaM-binding motif.

Unlike KRAS4b, the HRAS, NRAS, and KRAS4a isoforms are all subject to enzymatic cycles of palmitoylation/depalmitoylation, which modulate affinity for the membrane and control signal output (6971). The high affinity of these isoforms for the membrane, C-terminal bulkiness, and minimal charge all likely contribute to their incompatibility with CaM binding. KRAS4b and other GTPases that lack regulated palmitoylation may have evolved other cellular mechanisms to regulate membrane attachment and signal output. We believe that CaM serves this vital role in response to Ca2+ signaling, and the chimeric FRET construct CaMeRAS offers a glimpse at this regulation in vivo.

In addition to the Ca2+-dependent internalization of KRAS4b by CaM, which abrogates canonical KRAS4b signaling, there are other postulated consequences of this interaction, including the sequestration of Ca2+-bound CaM by KRAS4b, which reduces CaMKII activation and suppresses noncanonical Wnt/Ca2+ signaling, thereby promoting tumorigenicity (22). Furthermore, relocalization of KRAS4b to endomembranes may enable interactions with other proteins, such as Bcl-XL, which occurs upon relocalization of KRAS4b to mitochondria in response to phosphorylation of Ser181 by protein kinase C (2527).

These reports also demonstrate that phosphorylation at Ser181 and CaM binding are mutually exclusive, hinting at a signal-dependent interplay between multiple regulatory mechanisms (25). The direct and ill-defined competition between CaM binding, phosphorylation, and PDEδ cycling, which returns internalized KRAS4b to the PM from endosomes (18, 64), indicates the existence of a complex regulatory network. PDEδ and CaM would compete for binding to the farnesyl group of KRAS4b: CaM exhibits ~5-fold higher affinity (40, 64), but its interaction is spatiotemporally limited to a Ca2+ signaling event.

Collectively, our data provide new insights into the regulation of KRAS4b by CaM, supply tools for this growing field, and solidify the farnesyl-dependent membrane extraction and relocalization paradigm for this interaction. Rigorous biophysical studies of Ca2+-CaM regulation of KRAS4b and the cross-talk between MAPK and Ca2+ signaling lead to approaches to exploit this interaction for new therapeutic avenues.


Bacterial protein expression and purification (CaM/KRAS 1–185/BRAF RBD/SoScat)

E. coli BL21 CodonPlus cells were transformed with pET28 plasmids containing CaM or KRAS4b sequences and grown at 37°C to an optical density of 0.6 to 0.8. Temperature was reduced to 15°C before induction of protein expression with 0.25 mM isopropyl-β-d-thiogalactopyranoside. Cells were grown in Luria-Bertani media with 1 mM kanamycin for unlabeled protein production or M9 [biotin (1 mg/liter), thiamine (1 mg/liter), 0.3 mM CaCl2, 1 mM kanamycin, 0.2% (w/v) glucose, trace elements (25 μM EDTA, 30.83 μM ferric chloride, 3.67 μM zinc chloride, 0.74 μM cupric chloride, 0.42 μM cobalt chloride, 1.62 μM boric acid, and 127.19 μM manganese chloride), and 15N-NH4Cl (1 g/liter)] for uniform 15N labeling. Cells were lysed by sonication in buffer containing 50 mM tris-HCl (pH 8), 150 mM NaCl, 0.1% (v/v) NP-40, 10% (v/v) glycerol, 10 mM imidazole, 5 mM MgCl2, 10 mM β-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride, lysozyme, and deoxyribonuclease. CaM and unprocessed KRAS were purified by Ni–nitrilotriacetic acid His-tag affinity chromatography followed by gel filtration. A thorough purification protocol can be found here (72). N-terminally glutathione S-transferase (GST)–tagged B-Raf proto-oncogene serine/threonine kinase (BRAF) RAS-binding domain (RBD, residues 150 to 233) and His-tagged catalytic domain of SoS (SoScat, residues 564 to 1049) were cloned into pGEX-4T2 (GE Healthcare) and pET15b (Novagen/EMD Biosciences), respectively, expressed in E. coli BL21 (DE3), and purified as previously described (73).

Insect cell culture (isotope labeling)

High Five insect cells (Thermo Fisher Scientific) were transfected with bacmids containing farnesyl transferases 1 and 2 and His-MBP (maltose-binding protein)–KRAS4b (gift from NCI Frederick). Expression and purification of these constructs followed a previously reported protocol (44). 15N-lysine labeling of KRAS4b was performed in ESF 921 Delta Series media (ΔKRWYFH) (Expession Systems) supplemented with Arg (800 mg/liter), Trp (200 mg/liter), Tyr (360 mg/liter), Phe (1000 mg/liter), and His (200 mg/liter) (Sigma-Aldrich). 15N-lysine (Cambridge Isotope Laboratory) was added (100 mg/liter) 16 hours after infection. 13C-methyl labeling of KRAS4b was performed in ESF 921 Delta Series media (ΔTMILVA) supplemented with Ala (400 mg/liter). 13C-methyl–labeled Thr (75 mg/liter), Met (100 mg/liter), Ile (150 mg/liter), Leu (100 mg/liter), and Val (50 mg/liter) were added 16 hours after infection based on a protocol adapted from Gossert et al. (74, 75).

1H-15N HSQCs

Samples were prepared for NMR by mixing the appropriate labeled proteins and substrates in NMR buffer containing 50 mM Hepes (pH 7.4), 100 mM NaCl, 1 mM tris(2-carboxyethyl)phosphine (TCEP) and 10% D2O. Most experiments also included 5 mM MgCl2 and/or 10 mM CaCl2. NMR experiments were primarily conducted on a 600-MHz Bruker Avance magnet with 1.7-mm cryoprobe and 40 μl of sample volumes. Experiments conducted on an 800-MHz magnet had 450 μl of samples in 5-mm tubes.

NMR titrations

(E,E)-farnesol, (Z,Z)-farnesol, and (Z,Z)-FCME were solubilized in DMSO to 100 mM, appropriately diluted in NMR buffer, and added directly to separate NMR samples to a final concentration of 2.5% DMSO. Titration HSQCs were collected with 216 μM CaM and 0-, 0.35-, 0.7-, 1-, 1.4-, 2.9-, 5.8-, 11.5-, 14.3-, and 29-fold excess farnesyl compound in an autosampler connected to a 600-MHz Bruker Avance magnet at 25°C. KRAS4b-FMe titrations were collected at 0.1-, 0.4-, 0.8-, 1-, 1.3-, 2-, and 4-fold excess over 0.1 mM CaM under the same conditions.

Crystallization/structure determination

CaM [1 mM; 20 mM Hepes (pH 7.4) and 10 mM Ca2+] was screened for crystallization with 5 mM FCME at room temperature using Protein Complex (Qiagen), Wizard Classic (Rigaku), and Crystal Screen HT (Hampton Research) commercial screening kits. A successful condition from Crystal Screen HT containing 0.2 mM ammonium acetate, 0.1 mM sodium cacodylate (pH 6.5), and 30% (w/v) polyethylene glycol (PEG) 8000 was further optimized for pH and PEG concentration with 1 and 1.25 mM CaM. The diffracted crystal was grown in 5 mM FCME, 0.2 mM sodium acetate, 0.1 mM sodium cacodylate (pH 6.7), and 28% PEG 8000, with 1.25 mM CaM. Glycerol was used as a cryoprotectant. Diffraction data from the Canadian Macromolecular Crystallography Facility 08ID-1 Beamline at the Canadian Light Source were processed with HKL2000 (76) and modeled and refined using COOT (77) and Phenix (78).

ND preparation

All lipids were purchased from Avanti Polar Lipids, Inc. NDs were prepared according to previous protocols (55). 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-diethylenetriaminepentaacetic acid (gadolinium salt) [PE-DTPA(Gd3+)] was added to the lipids at a molar ratio of 20 DOPC:5 DOPS:1 PE-DTPA (Gd3+). The organic solvents were then removed using gentle nitrogen flow followed by vacuum, and the dried lipid film was solubilized in aqueous buffer containing detergent [20 mM tris (pH 7.4), 100 mM NaCl, and 100 mM sodium cholate]. This mixture was subjected to three freeze/thaw cycles and vortexed until clarified, and then membrane scaffold protein [MSP1D1, prepared as described previously (79)] was added at a 1:40 molar ratio relative to the lipids. After 1-hour incubation with mild rotation at 20°C, sodium cholate was removed from the MSP-lipid mixture with Bio-Beads SM-2 adsorbents (Bio-Rad) using a batch method with 2-hour incubation at room temperature with mild rotation. The ND particles were then purified via size exclusion chromatography using a 26/60 Superdex 200 column equilibrated with buffer A without TCEP. The particle size was analyzed with dynamic light scattering and corresponded with a 10-nm-diameter particle.

ND-binding PRE

Baseline 1H-13C HSQC spectra of 200 μM 13C-TILVM KRAS4b-FMe were collected on a 600-MHz Bruker Avance magnet in the presence and absence of unlabeled CaM, with and without Ca2+, at 25°C. PRE experiments were initiated by the addition of 1:20 molar ratio Gd3+-containing ND to similarly prepared NMR samples and collecting corresponding HSQCs. Peak intensity ratios were calculated for ND versus baseline for CaM-containing samples, either in the presence of Ca2+ [20 mM Hepes (pH 7.3), 100 mM NaCl, 1 mM TCEP, 5 mM Mg2+, and 5 mM Ca2+] or absence of Ca2+ [20 mM Hepes (pH 7.3), 100 mM NaCl, 1 mM TCEP, and 0.5 mM Mg2+] by normalizing intensities to a 0.6 mM sodium trimethylsilylpropanesulfonate internal standard.

Plasmid constructions for FRET chimeras

The plasmid containing a codon-optimized KRAS4b (KRAS4b-opt) was purchased from Invitrogen GeneArt Gene Synthesis (Thermo Fisher Scientific). pmTurquoise2-C1 (80) and pSYFP2-C1 (81) were gifts from D. Gadella (Addgene plasmid nos. 60560 and 22878, respectively). pEZYmyc-His (82) was a gift from Y.-Z. Zhang (Addgene plasmid no. 18701). A complementary DNA (cDNA) encoding KRAS4b-opt amplified by polymerase chain reaction (PCR) was cloned into pENTR1A (Invitrogen) between Bam HI and Not I sites (plasmid #1). Then, a cDNA encoding CaM amplified by PCR was cloned into Bam HI site of plasmid #1 (plasmid #2). Furthermore, to insert mTq2 and SYFP2 into plasmid #2, two short linkers including restriction enzyme site(s) were inserted at the 5′-end of CaM and between CaM and KRAS4b-opt using Q5 site-directed mutagenesis (New England Biolabs Inc.) (plasmid #3). Then, cDNAs of mTq2 and SYFP2 amplified by PCR using pmTurquoise2-C1 and pSYFP2-C1 as templates were cloned into Sph I and Kpn I sites and Spe I and Bam HI sites of plasmid #3, respectively, producing mTq2-CaM-SYFP2-KRAS4b, identified as CaMeRASWT. For making a construct of mTq2-CaM1234Q-SYFP2-KRAS4b, called CaMeRAS1234Q, a cDNA of CaM1234Q (E31Q, E67Q, E104Q, and E140Q) amplified by PCR was used to replace wild-type CaM. For mTq2-CaM-SYFP2-KRAS4b-Cys185Astop or CaMeRAS185A, Cys185 in KRAS4b was mutated to Ala, and TAA stop codon was inserted after Ala using Q5 site-directed mutagenesis. To insert FLAG tag at the N terminus, a cDNA encoding FLAG-mTq2 was amplified by PCR using a forward primer harboring a FLAG sequence at the 5′-end and replaced with nontagged mTq2. mTq2-CaM-SYFP2-KRAS4b and mutants made in pENTR1A were transferred into pEZYmyc-His by Gateway LR Clonase II (Invitrogen). All plasmids were confirmed by DNA sequencing. The sequences of all primers used (table S3) and the full amino acid sequence for the chimeric construct (fig. S18) are provided in the Supplementary Materials.

Expression and purification of FLAG-mTq2-CaM-SYFP2-KRAS4b and mutants

HeLa cells were transfected with plasmids containing FLAG-mTq2-CaM-SYFP2-KRAS4b and mutants using a transfection reagent, jetPRIME (Polyplus-transfection). Cells were harvested at ~48 hours after transfection. For purification, cells were resuspended with lysis buffer [50 mM tris, 300 mM NaCl, 2 mM dithiothreitol (DTT), and 1 mM EDTA (pH 7.4) supplemented with cOmplete EDTA-free protease inhibitor cocktail (Roche)] and lysed using an EmulsiFlex-C3 (Avestin). Cell lysate was centrifuged at 45,000g for 30 min to remove cell debris. The supernatant was loaded to anti-FLAG M2 agarose (Sigma-Aldrich) and washed with the same buffer for lysis. FLAG-mTq2-CaM-SYFP2-KRAS4b was eluted with 3×FLAG peptide (0.15 mg/ml; APExBIO) in the buffer [50 mM tris, 300 mM NaCl, and 2 mM DTT (pH 7.4)].

In vitro FRET-based calcium titration

Purified FLAG-mTq2-CaM-SYFP2-KRAS4b was dialyzed with the buffer [50 mM tris, 300 mM NaCl, and 2 mM DTT (pH 7.4)] in which residual calcium had been chelated with Chelex 100 resin (Bio-Rad). Fluorescence measurement was made on a Shimadzu RF-5301PC fluorimeter (Shimazu Corp.), and emission spectra of FLAG-mTq2-CaM-SYFP2-KRAS4b excited by 405-nm light in the presence of various concentrations of CaCl2 were scanned from 420 to 600 nm at 20°C.

Biolayer interferometry

We investigated effector binding, intrinsic nucleotide exchange, and SoScat-dependent nucleotide exchange of the CaMeRAS fusion protein using an Octet RED 384 biolayer interferometry instrument equipped with the Octet Data Acquisition software (Pall FortéBio). Initially, GST-BRAF-RBD was immobilized to anti–GST-conjugated biosensors (Pall FortéBio) by incubating the sensor in a well containing GST-BRAF-RBD (5 μg/ml) for 10 min. To test the intrinsic exchange of CaMeRAS, we incubated 1.5 μM purified protein and 10 mM GTPγS in a 96-well plate. The impact of SoScat on the nucleotide exchange was assessed by addition of SoScat (1:600 molar ratio) to 1.5 μM CaMeRAS and 10 mM GTPγS. In both cases, nucleotide exchange was assessed by immersing the BRAF-RBD–conjugated sensors into the wells containing CaMeRAS at different time intervals. The time-dependent enhancements of the BLI signal induced by CaMeRAS binding the RBD indicated that nucleotide exchange occurred. All proteins were reconstituted in 20 mM Hepes buffer (pH 7.4) containing 100 mM NaCl, 2 mM MgCl2, 1 mM TCEP, 0.01% BSA, and 0.005% Tween-20. The data were analyzed with FortéBio Data analysis software (Pall FortéBio). Sensors dipped into wells containing 1.5 μM CaMeRAS supplemented with 10 mM GDP were used as a reference.

FRET and calcium imaging and data analysis

HeLa or HEK293 cells were transfected with plasmids containing the mTq2-CaM-SYFP2-KRAS4b or its mutants using jetPRIME (Polyplus-transfection). The cells were subjected to imaging at 18 to 24 hours after transfection. Hank’s balanced salt solution (Gibco) supplemented with 1 mM CaCl2, 0.5 mM MgCl2, and 0.5 mM MgSO4 was used as an imaging buffer. After loading the cells with 10 μM Calbryte 630AM (AAT Bioquest), imaging was performed at room temperature using a Leica SP8 confocal microscope with a 63×/1.4 numerical aperture (NA) oil immersion objective lens and high-sensitivity HyD detectors. For simultaneous visualization of mTq2, SYFP2, and Calbryte 630, a 440-nm solid state laser and a white light laser tuned to 608 nm were used for excitation, and fluorescence emission was collected at 455 to 490 nm, 525 to 560 nm, and 620 to 750 nm, respectively. Images were acquired at 3.4 to 4.0 s per frame. Regions of interest on the PM or in the cytosol were selected for mTq2 or SYFP2 fluorescence measurement (Fig. 6A and fig. S15), and quantification was performed using Fiji software (83). Histogram analysis and Gaussian distribution fitting were done using Igor Pro 8 (WaveMetrics Inc.).

Fluorescence lifetime imaging microscopy and data analysis

Fluorescence lifetime images were acquired using a Leica SP8 confocal microscope equipped with a time-correlated single-photon counting module (PicoHarp 300, PicoQuant). The donor (mTq2) was excited using a 405-nm solid-state laser pulsed at a 40-MHz repetition frequency. Fluorescence signal in the emission range of 455 to 480 nm was collected through a 63×/1.4 NA oil immersion objective lens using a cooled single-photon counting HyD SMD detector. Fluorescence lifetime traces were analyzed by using SymPhoTime 64 software (PicoQuant). Statistical difference between pre- and post-histamine stimulation was analyzed by using Wilcoxon signed-rank test.


Fig. S1. RAS isoform-specific HVRs.

Fig. S2. There is no major interaction between unfarnesylated KRAS4b (1–185) and CaM, independent of bound nucleotide.

Fig. S3. A new class of noncanonical CaM binding targets: Singly lipidated polybasic peptides.

Fig. S4. Chemical shift analysis of CaM residues upon farnesol binding.

Fig. S5. Farnesol and CaM do not interact in the absence of Ca2+.

Fig. S6. Farnesyl compound titrations into CaM.

Fig. S7. KRAS4b-FMe and farnesol perturbations of 15N-CaM share a characteristic pattern.

Fig. S8. Structure comparison of CaM conformations bound to FCME and CAP23/NAP22 polybasic myristoylated peptide.

Fig. S9. Lipid orientation in CaM:FCME complex structure.

Fig. S10. Solution NMR PRE validation of CaM-FCME crystal structure.

Fig. S11. Membrane-based PRE probe for KRAS4b-FMe interaction with CaM in solution.

Fig. S12. KRAS4b retains function in the CaMeRAS chimera.

Fig. S13. Calcium-dependent change in emission spectra of CaMeRAS chimera FRET protein.

Fig. S14. Tracking cellular localization of CaMeRASWT in HeLa cells after ionomycin treatment.

Fig. S15. Cellular image segmentation for FRET histogram analysis by Gaussian modeling.

Fig. S16. Fluorescence lifetime of mTq2 of CaMeRASWT is reduced by histamine stimulation in HeLa cells.

Fig. S17. ATP-evoked Ca2+ signal induces CaMeRASWT internalization and FRET change in HEK293 cells.

Fig. S18. The nucleotide and amino acid sequences of CaMeRASWT chimeric FRET construct.

Table S1. Peak values of FRET histograms and the numbers of the ROIs analyzed.

Table S2. P values from Student’s t tests comparing Gaussian curve models of FRET histogram data.

Table S3. Primer sequences used for plasmid construction.

Movie S1. Co-imaging CaMeRASWT and calcium response to histamine stimulation in HeLa cells.

Movie S2. Slow motion highlight: Co-imaging CaMeRASWT and calcium response to histamine stimulation in HeLa cells.

Movie S3. Coimaging CaMeRASWT and calcium response to histamine stimulation in HeLa cells and calcium chelation.

Movie S4. Coimaging CaMeRAS185A and calcium response to histamine stimulation in HeLa cells.

Movie S5. Coimaging CaMeRAS1234Q and calcium response to histamine stimulation in HeLa cells.

Data file S1. CaM-farnesol binding CSP by residue.

Data file S2. CaM-FCME PRE.

Data file S3. KRAS4b-ND PRE.

Data file S4. In vitro CaMeRAS Ca2+ titration.

Data file S5. Relative fluorophore intensity in vitro as a function of Ca2+ titration for CaMeRAS.

Data file S6. In vivo FRET from CaMeRAS contructs in response to histamine stimulation and Ca2+ influx.


Acknowledgments: We thank the staff at the Advanced Optical Microscopy Facility, University Health Network for the help with this project. Research described in this paper was performed using beamline 08ID-1 at the Canadian Light Source, which is supported by the Canada Foundation for Innovation (CFI), Natural Sciences and Engineering Research Council of Canada, the University of Saskatchewan, the Government of Saskatchewan, Western Economic Diversification Canada, the National Research Council Canada, and the Canadian Institutes of Health Research. We thank D. Gadella and Y.-Z. Zhang for the donations of plasmids and D. Esposito at the NCI for providing the baculovirus expression system for production of fully processed KRAS. We thank G. Gasmi-Seabrook for assistance and management of the Princess Margaret NMR facility funded by the CFI and the Princess Margaret Cancer Foundation (PMCF). We also thank K. Hueniken, Research Analyst at the University Health Network in the Department of Biostatistics, for consultation on statistical analysis. Funding: This work was supported by grants from the Canadian Institutes of Health Research (410008598), the Canadian Cancer Society Research Institute (703209 and 706696), and the Heart and Stroke Foundation of Canada (G-16-00014035). In addition, we thank the CFI, PMCF, and Canada Research Chairs for funding and support. Author contributions: B.M.M.G., C.B.M., M.E., and M.I. conceived the presented ideas. B.M.M.G., C.B.M., and M.E. developed and planned the experiments. B.G., M.E., S.-I.B., and T.G. conducted the experiments, processed and analyzed the data, and produced the figures. K.-Y.L., T.G., and N.I. assisted with conceptualization and advanced protocols. C.B.M. and M.I. supervised the project. B.G. drafted the initial manuscript and edited it with the help of C.B.M., M.E., and M.I. All authors provided useful critiques and feedback on the project and manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The structure of calcium-bound CaM in complex with FCME has been deposited in PDB (PDB ID: 6OS4). Data and materials availability: All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article