A chemigenetic endogenous Ras activity biosensor for live-cell pharmacology

Abstract

Ras is a small GTPase that regulates cell growth and proliferation. Hyperactive Ras is prevalent in cancer and has been a therapeutic target for decades. Many lines of evidence have demonstrated that Ras signals from the plasma membrane as well as non-canonical compartments such as the Golgi, making live-cell biosensors for tracking the spatiotemporal dynamics of Ras activity a valuable approach for Ras studies. However, current biosensors are limited in their quantitative capacity for measuring endogenous Ras activity. Herein, we introduce a chemigenetic biosensor design, making use of circularly permuted HaloTag labeled with the fluorophore JF₆₃₅, for detecting endogenous Ras activity. This HaloTag-based Ras Activity Reporter (HaloRasAR) revealed the spatiotemporal dynamics of Ras activity downstream of either growth factor signaling or PKC activation. In addition, live-cell characterization of a RasG12C inhibitor and a Ras-GEF inhibitor revealed subcellular-specific inhibition profiles. HaloRasAR represents an important advance in spatiotemporal interrogation of Ras signaling and live-cell pharmacology.

INTRODUCTION

Ras is a small GTPase and critical regulator of cell growth and proliferation. Since its discovery in 1964 as a sarcoma-inducing factor¹ and subsequent work showing its human origin² and prevalence in cancer,^2,3 this oncoprotein has become one of the most extensively pursued molecular targets for cancer therapy.⁴ Ras is a “molecular switch” that alternates between the active form, RasGTP, promoted through Guanine Nucleotide Exchange Factors (GEFs) and inactive form, RasGDP, promoted through inhibitory GTPase-Activating Proteins (GAPs). Mitogenic signals such as Epidermal Growth Factor (EGF) are known to stimulate the formation of the active state. Active Ras is poised for selective interaction with Ras effectors such as c-Raf, which activate the downstream kinase ERK, promoting growth and proliferation.⁵ In addition to ERK, Ras can also transmit signals through multiple different effectors, such as PI3K/RalGDS.² Moreover, Ras may be stimulated by non-canonical upstream activators, such as GPCRs⁶ and protein kinase C (PKC).⁷

Ras is activated at multiple compartments including the plasma membrane and the Golgi Apparatus,^8,9 with reports suggesting that Ras signaling at distinct subcellular locations is linked to different upstream signals¹⁰ and downstream signaling outputs.¹¹ These subcellular measurements were enabled by advances in Ras biosensors including the recently reported Ras-LOCKR-S, which incorporates a de novo-designed alpha-helical switch to convert the binding of endogenous, active Ras into changes in FRET efficiency albeit with limited dynamic range.¹² Nevertheless, our understanding of the spatiotemporal regulation of Ras signaling is still limited, largely owing to a lack of sensitive Ras biosensors capable of detecting endogenous activity.^13,14 Here, we report the development of a far-red fluorescent biosensor capable of sensing endogenous Ras activity with high sensitivity based on a chemigenetic biosensor design. We use this sensor to characterize the spatiotemporal dynamics of both the activation and inhibition of endogenous Ras in living cells.

RESULTS

Design and Validation of HaloRasAR

To develop a sensor capable of sensing endogenous Ras activity, we utilized circularly permuted HaloTag (cpHaloTag). Previous biosensor designs have implemented cpHaloTag to orient conformationally responsive sensing units proximal to the fluorophore binding site, resulting in biosensors with exceptionally high dynamic ranges.^15,16 This allows protein conformational changes in cpHaloTag-fusion proteins to induce perturbation of the equilibrium between zwitterionic (Z) and lactone (L) forms of the fluorogenic JF₆₃₅ and produce intensiometric optical reporting in response to intracellular changes in signaling activity.¹⁵ We hypothesized that a single Ras binding domain sensitive to active Ras would be sufficient to induce conformational perturbations in the fluorophore environment upon Ras binding, driving a shift in the L-Z equilibrium upon active Ras engagement (Fig. 1a–b). To test this, we fused either the wildtype c-Raf Ras-binding Domain (RBD) or a high-affinity variant, RBDv12, to the N-terminus of cpHaloTag(144,146).¹⁷ Initial variants bearing the RBDv12 showed responses, and upon screening of sensor variants with differing linker lengths, we observed clear improvement in brightness of a single variant (Extended Data Fig. 1a–b). The improved, 2 amino-acid linker variant was chosen, and to control for variable expression levels and dynamic changes in subcellular localization, mNeonGreen was attached to the N-terminus (Fig. 1a–b, Fig. S1). With this design, we used the far-red/green emission ratio as the sensor readout.

Figure 1. HaloRasAR reports Ras activity by an increase in far-red/green emission ratio.

Design and domain structure (a) of HaloRasAR, which contains the Ras binding domain (RBDv12) fused N-terminally to cpHaloTag. Ras binding to RBDv12 induces a conformational change that increases the fluorescence of JF₆₃₅. mNeonGreen is fused N-terminally, providing a ratiometric readout. (b) JF₆₃₅ exhibits an equilibrium between lactone (L) and fluorescent zwitterionic (Z) state. Unbound HaloRasAR shows relatively unbiased L-Z equilibrium that is perturbed towards zwitterionic upon RasGTP binding. (c-d) COS7 cells expressing HaloRasAR and dominant negative HRasS17N or constitutively active HRasG12V show Ras activity-dependent increase in far-red/green emission ratio (n = 3607 cells for S17N, 5502 cells for G12V, Cohen’s d = 3.18). (e) Average time courses showing responses of HaloRasAR in COS7 cells overexpressing NRas, HRas, or KRasWT and stimulated with EGF with negative control R89L (n = 64 cells for HRasWT, 77 cells for NRas, 35 cells for KRas, negative control R89L data were from 56, 67, 50 cells, respectively) (f) Pseudocolor image of a cell expressing HRasWT and HaloRasAR stimulated with EGF. All data are combined from 3 independent experiments, and pseudocolor images are representative of 3 independent experiments. For time-series traces, solid lines indicate the mean and shaded areas represent s.e.m.. All scatter plots include median value.

We next purified the sensor and performed in vitro characterization. This sensor, labeled with JF₆₃₅, showed a 161 ± 19% increase in far-red fluorescence when mixed with KRasG12V bound to guanosine 5′-[β,γ-imido]triphosphate (GppNHp), a guanosine triphosphate analogue insensitive to the GTPase activity of Ras. No effect was observed in mNeonGreen fluorescence (n = 3 samples, mean ± standard deviation, Extended Data Fig. 2a). Changes in its fluorescence emission mostly stem from changes in extinction coefficient ε (Extended Data Fig. 2b–d), whereas quantum yield φ did not change (φ = 0.83 ± 0.01 and 0.79 ± 0.02 without and with KRasG12V, respectively (n = 3 samples, mean ± standard deviation)). Using purified RBDv12 in a competition assay, we estimated a K_d of 8.6 ± 3.6 nM for the sensor binding to GppNHp-bound KRasG12V, placing its affinity in the low nanomolar range (n = 3 samples, mean ± standard deviation, Extended Data Fig. 2e–g).

To test for proper dye labeling in cells, COS7 cells expressing each biosensor variant were incubated with JF₆₆₉, which should produce bright fluorescence irrespective of biosensor conformation.¹⁸ Indeed, JF₆₆₉ labeling yielded strong fluorescence from cpHalo-RBD variants, although the intensity was lower compared to unmodified HaloTag (Extended Data Fig. 3a–b). We next coexpressed each variant with either dominant-negative HRasS17N or constitutively active HRasG12V in COS7 cells and measured the far-red/green emission ratio of each variant labeled with fluorogenic JF₆₃₅. The RBDv12 variant showed a 130% higher median far-red/green emission ratio in cells co-expressing constitutively active HRasG12V compared to cells co-expressing dominant-negative HRasS17N (Fig. 1c, Cohen’s d = 3.18, n = 3607 and 5502 cells for HRasS17N and HRasG12V, respectively). Alternatively, the wtRBD variant showed only a 47% higher median far-red/green emission ratio in HRasG12V-expressing cells versus HRasS17N-expressing cells (Extended Data Fig. 3c–d). To compare our sensor’s performance relative to Ras-LOCKR-S,¹² currently the only biosensor shown to report endogenous Ras activity, we performed the same assay for cytosolic Ras-LOCKR-S, which showed approximately a 20% increase in Y/C emission ratio in the HRasG12V expression group (Extended Data Fig. 3g).

Given the higher dynamic range of the RBDv12 variant (130.8 ± 0.7% [mean ± s.e.m.] for RBDv12, 42.4 ± 0.9% for wtRBD, n = 5504 cells, RBDv12, n = 1114 cells, wtRBD, p<0.0001), we moved forward with this variant and named it HaloRasAR. We co-transfected COS7 cells with HaloRasAR plus NRasWT, HRasWT or KRasWT and monitored the responses following treatment with EGF to acutely activate Ras. For each Ras isoform, we observed rapid increases in the HaloRasAR far-red/green emission ratio upon EGF stimulation (Fig. 1e–f, HRas: ΔR/R₀ = 21.1% ± 2.4% [mean ± s.e.m.], n = 64 cells; NRas: ΔR/R₀ = 15.9% ± 0.9%, n = 77 cells; KRas: ΔR/R₀ = 8.8% ± 0.6%, n = 35 cells), whereas cells expressing a negative-control HaloRasAR bearing a Ras-insensitive RBD R89L showed no response. Furthermore, in NRasWT-expressing cells, no detectable EGF-stimulated increases in the far-red/green emission ratio were observed when labeled with the non-fluorogenic JF₆₆₉ (Extended Data Fig. 3f). Lastly, we plotted normalized far-red and green intensity changes, verifying the ratiometric response is driven by intensiometric increases in the far-red channel (Extended Data 4a–c). Altogether, HaloRasAR, with its chemigenetic design and high-affinity sensing unit, can report Ras activity in living cells with high sensitivity.

HaloRasAR Detects Endogenous Ras Activity

Beyond its canonical plasma membrane role, Ras has also been shown to be active at the Golgi, leading to distinct functional outcomes.^8–11 Many studies overexpress Ras isoforms to permit live-cell Ras activity measurements, leaving subcellular regulation of endogenous Ras incompletely characterized. We sought to measure endogenous Ras activities in both the Golgi and the plasma membrane. EGF-treated COS7 cells showed rapid increases in the HaloRasAR far-red/green emission ratio, with a negative control RBD mutant R89L showing no responses (Fig. 2a–b), indicating that HaloRasAR is sensitive enough to detect endogenous Ras activity. Strikingly, we also observed clear compartmentalization of endogenous Ras activity in EGF-stimulated cells at the periphery and in a perinuclear compartment (Fig. 2c–d, Extended Data 5a), which colocalized with the Golgi marker mTagBFP2-Giantin (Fig. 2e). Drawing specific regions of interest (ROIs) around the cell periphery or the Golgi revealed significantly different HaloRasAR response dynamics, with Golgi-selective ROIs showing slower (t_1/2 = 1.70 ± 0.19 min) and more sustained responses (SAM15 = 0.78 ± 0.3%), as measured by Sustained Activity Metric (SAM)¹⁹ compared with peripheral ROIs (t_1/2 = 0.67 ± 0.04 min, SAM15 = 0.46 ± 0.3%, p<0.0001) (Fig. 2d and Extended Data Fig. 5b). We also observed similar transient activation in EGF-stimulated HEK293T cells (Extended Data Fig. 5g). As active Ras has low expression in cells, we tested whether ultra-low expression of HaloRasAR improved performance. With weak induction of stably expressed Tet-inducible HaloRasAR using 500 pg/mL doxycycline, we observed significantly improved endogenous Ras responses (Extended Data Fig. 6a–c, Whole cell: R_max/R₀ = 21.4% ± 0.4% [mean ± s.e.m.], n = 150 cells; Peripheral: R_max/R₀ = 29.1% ± 0.8%, n = 129 cells; Supplemental Video 1) compared with transient expression of CMV-driven HaloRasAR (Peripheral: R_max/R₀ = 11.6% ± 0.6%, n = 60 cells, [mean ± s.e.m.], p<0.0001). Western blot analysis with phospho-ERK showed that expression of HaloRasAR via transient transfection had no significant effect on downstream ERK signaling (Extended Data Fig. 7a–b).

Figure 2. HaloRasAR reports endogenous Ras activity.

(a-b) COS7 cells expressing HaloRasAR and stimulated with EGF show increases in far-red/green emission ratio (n = 60 cells for HaloRasAR, 57 cells for R89L control). (b) R_max/R₀ values are significantly elevated compared to R89L control (n = 60 cells for HaloRasAR, 57 cells for R89L control, p=1.699e-034 for whole cell Ras activity by Mann-Whitney test). (c-d) Site-selective ROIs at the periphery and Golgi demonstrate distinct kinetics of activation, with Golgi activity more sustained as calculated by SAM15 (n=48 cells for peripheral, 51 cells for Golgi, p=1.366e-010 as determined by Mann-Whitney test). (e) Partial colocalization of HaloRasAR response with Golgi marker mTagBFP2-Giantin (Pearson’s R = 0.58). (f) Golgi-HaloRasAR domain structure with Giantin localizing sequence. (g-h) Golgi-HaloRasAR response in the context of EGF stimulation, with AG1478 pretreatment results in inhibited response relative to DMSO control (n = 36 cells for DMSO control, 48 cells for AG1478 pretreatment control). (h) Golgi-HaloRasAR response is significantly higher compared to AG1478 control. R_max/R₀ values obtained are responses for t = 25 min for each group (n = 36 cells for DMSO control, 48 cells for AG1478 pretreatment control, p=2.511e-009 by Mann-Whitney test). (i) Pseudocolor image of Golgi-HaloRasAR before and after EGF stimulation. Data shown are combined results of 3 independent experiments. Pseudocolor images are representative of 3 independent experiments. For time-series traces, solid lines indicate the mean and shaded areas represent s.e.m. All scatter plots include median value. All Mann-Whitney comparisons are two-sided.

To further measure Golgi Ras activity, we constructed Golgi-targeted HaloRasAR by fusing a Giantin localization tag to the C-terminus of HaloRasAR (Fig. 2f, Figure S1). Golgi-HaloRasAR successfully co-localized with a Golgi marker (Extended Data 5d), and upon treatment with EGF, Golgi-HaloRasAR-expressing COS7 cells exhibited a slow and sustained increase in the far-red/green emission ratio, which was abolished by pretreatment with the EGFR inhibitor AG1478 (Fig. 2g–i). For plasma membrane measurements, we attached a C-terminal KRas-CAAX localization sequence (Extended Data 5c, Fig. S1). As this targeting motif localizes the sensor in proximity to membrane Ras, we found that pm-HaloRasAR containing wtRBD was sufficient for detecting endogenous activity, although a poorly expressing RBDv12-containing variant was also functional (Extended Data 5e). EGF stimulation of COS7 cells expressing pm-HaloRasAR yielded a fast increase in the far-red/green emission ratio. Like our observations with untargeted HaloRasAR using location-selective ROIs, pm-HaloRasAR responses were significantly faster (t_1/2 = 0.37 ± 0.03 min, [mean ± s.e.m.]) compared with Golgi-HaloRasAR responses (t_1/2 = 5.33 ± 0.41 min, p<0.0001; Extended Data Fig. 5f). Importantly, green fluorescence was not a driver of the acutely activated far-red/green ratiometric response in measurements for endogenous Ras activity or all other previous measurements (Extended Data Fig. 8a–f). Thus, the high dynamic range and sensitivity of HaloRasAR enable real-time monitoring of endogenous Ras activity, revealing distinct activity dynamics at the plasma membrane and Golgi.

The Spatiotemporal Dynamics of PKC Activation of Ras

Ras regulation by PKC has been well documented, though the mechanism remains unclear. PKC may activate Ras via RasGEFs of the RasGRP and SOS family^9,20 or induce translocation of KRas4B from the plasma membrane to internal membranes.²¹ PKC biology is of great interest in cancer, as early characterization of PKC as oncogenic has been challenged by increasing evidence suggesting a tumor-suppressive role.²² We hypothesize that PKC exerts spatiotemporally distinct effects on Ras activity dynamics in different contexts, such as PKC signaling strength.

To test this hypothesis, we set out to examine Ras activity dynamics in response to strong and moderate PKC stimulation. Conventional PKC isoforms are activated by diacylglycerol and Ca²⁺. Therefore, we used phorbol 12,13-dibutyrate (PDBu) to mimic the effects of diacylglycerol⁷ and ionomycin to induce Ca²⁺ influx, treating COS7 cells with 50 nM PDBu to induce moderate PKC activity and 200 nM PDBu plus 500 ng/mL ionomycin to induce strong PKC activity. As expected, cells expressing the PKC sensor ExRaiCKAR2²¹ showed a gradual, submaximal response to moderate PKC stimulation and a rapid, maximal response to strong PKC stimulation (Extended Data Fig. 9a–b). Subjecting HaloRasAR-expressing COS7 cells to moderate PKC stimulation, we observed a sustained response, which was concentrated towards the cell periphery, whereas cells pretreated with the pan-PKC inhibitor Gö6983 showed almost no response (Fig. 3a–c). Responses were observed to be particularly prominent within membrane ruffles occurring along the cell periphery (Supp. Video 2). Lower responses were observed at the Golgi apparatus, as assessed by location-selective ROIs (Fig. 3b).

Figure 3. HaloRasAR reveals PKC activation of Ras activity.

(a-c) Average time courses showing responses of HaloRasAR expressed in COS7 cells stimulated with 50 nM PDBu. Peripherally, (a) PDBu-treated cells, pretreated with Gö6983 or DMSO control (ΔR/R₀ = 9.1% ± 1.1% n = 42, ΔR/R₀ = 3.6% ± 0.7% n = 40 cells for DMSO, Gö6983, respectively, p=1.045e-005, for peripheral ROIs). Responses are significantly elevated whether at periphery or entire cell. R_max/R₀ values are t = 25 time point (Mann-Whitney test, p=1.045e-005 for cell periphery, p=0.006211 for whole cell) (b) ROIs for the periphery or Golgi in 50 nM PDBu treated cells. Area under the curve (AUC) shows significant elevation at the periphery (n = 46, 45 cells for periphery, Golgi, respectively, p=0.004309 by Mann-Whitney comparison). (c) Pseudocolor image of stimulated cells. (d) HaloRasAR expressed in COS7 cells and cotreated with 200 nM PDBu and 500 ng/mL ionomycin, as well as abrogation of the response in Gö6983 pretreatment group (ΔR/R₀ = 9.9% ± 0.9% for DMSO, n = 51, ΔR/R₀ = 0.6% ± 0.7% for Gö6983 n = 58 cells, p=6.572e-014 by Mann-Whitney test). (e) ROIs show no difference in AUC (n = 56, 55 cells for periphery and Golgi, respectively, Mann-Whitney test, p=0.9789). (f) Pseudocolor image of stimulated cells. (g-h) 10 μM BAY293, teal, pretreatment of COS7 cells expressing HaloRasAR variants compared to DMSO, black. (g) HaloRasAR response at the cell periphery in response to 50 nM PDBu shows higher AUC (n = 41 cells, DMSO, 36 cells BAY293, p=3.695e-007 by Mann-Whitney test). (h-i) Golgi-HaloRasAR shows no difference in AUC with BAY293 pretreatment (n = 45, 44 cells for DMSO, BAY293, respectively, p=0.05787 by Mann-Whitney test), as does whole cell HaloRasAR response to 200 nM PDBu and 500 ng/mL ionomycin (n = 46, 37 cells for DMSO, BAY293, respectively, p=0.6585 by Mann-Whitney test). (j) Elevation in AUC for Golgi-HaloRasAR response upon subjecting COS7 cells to strong agonism as before (n = 54, 58 cells for DMSO, BAY293, respectively, p=0.02290 by Mann-Whitney test). Panels are combined results of 3 independent experiments. Pseudocolor images are representative of 3 independent experiments. Solid lines in time-series traces indicate the mean and shaded areas represent s.e.m. All Mann-Whitney comparisons are two-sided.

Subjecting COS7 cells to strong PKC stimulation induced a strong HaloRasAR signal increase throughout the entire cell (Fig. 3d–f), with peripheral and Golgi-selective ROIs showing no significant differences (Fig. 3e). Consistent with the observation that Golgi Ras activity was significantly elevated with stronger PKC agonism, we also observed a larger response from Golgi-HaloRasAR induced by strong PKC stimulation compared with moderate PKC stimulation (Extended Data Fig. 9c–d). In contrast to moderate PKC stimulation, global Ras activity induced by strong PKC stimulation was not associated with membrane ruffling (Supp. Video. 3).

Previous studies have suggested that phorbol esters and PKC may initiate Ras activity via SOS1 RasGEFs.²⁰ To test this possible mechanism, we pretreated HaloRasAR-expressing COS7 cells with the SOS1 inhibitor BAY293, followed by moderate or strong PKC stimulation. The only significant effect of SOS1 inhibition on Ras activity was observed at the cell periphery in response to moderate PKC agonism, where BAY293 pretreatment led to a significant decrease in the response amplitude and kinetics of HaloRasAR with moderate PKC activation (Fig. 3g–h, and Extended Data Fig. 9e). With strong PKC stimulation, we observed minimal effect with BAY293 pretreatment (Fig. 3i–j, Extended Data Fig. 9f). These results support our hypothesis that PKC can induce distinct spatiotemporal Ras activity dynamics depending on the signaling strength, with moderate PKC activity leading to SOS1-dependent local Ras activation within membrane ruffles and strong PKC activity promoting global, SOS1-independent Ras responses.

Differential Effects of Ras Pathway Inhibitors on Subcellular Ras Activity

Ras is a prevalent oncogene and highly pursued drug target.⁴ Several novel Ras inhibitors have been developed in the past few years, adding to the repertoire of Ras pathway inhibitors.⁴ However, most of these inhibitors have not been assessed with assays capable of measuring spatiotemporal dynamics of Ras inhibition in single cells. We sought to characterize the newly designed Ras inhibitor RMC6291 that directly targets active KRasG12C, as well as the SOS1 inhibitor BAY293, in the context of acute inhibition in MIA PaCa-2 cells that express KRasG12C. Consistent with hyperactive Ras signaling, MIA PaCa-2 cells expressing HaloRasAR showed basally elevated far-red/green emission ratios at the cell periphery and Golgi (Fig. 4a, Extended Data Fig. 10a). Treatment with RMC6291 to inhibit KRasG12C induced a rapid decrease in the basal far-red/green emission ratio, whereas no change was observed in cells expressing HaloRasAR R89L (Fig. 5a, Extended Data Fig. 10b). Notably, analyzing the responses using location-selective ROIs revealed spatial differences in the magnitude of the emission ratio decrease, with a stronger decrease recorded at the cell periphery compared to the Golgi region (Fig. 4b–c, ΔR/R₀ = −20.2% ± 1.2%, periphery, ΔR/R₀ = −14.7% ± 0.8%, Golgi n = 36 cells for each group, p<0.0001).

Figure 4. HaloRasAR measures Ras inhibition in living cells.

(a) Pseudocolor image demonstrates localization of elevated far-red/green emission ratios before and after RMC6291 introduction. (b-c) Average time courses showing responses of HaloRasAR expressed in MIA PaCa-2 cells treated with 100 nM RMC6291. (b) Whole cell ROI shows response to RMC6291 (n = 49 cells, HaloRasAR, 48 cells, R89L control). (c) Region-selective ROIs are drawn at cell periphery or Golgi and responses are shown (n = 36 cells each). (d) Region selective ROIs result in significantly different ΔR/R₀ values obtained at t = 10 (n= 36 cells each group, p=0.0002 by Mann-Whitney test). (e-h) HaloRasAR expressed in MIA PaCa-2 cells and treated with 10 μM BAY293. (e) Response as measured by whole cell ROIs (f) ΔR/R₀ values obtained at t = 10 are significantly lower in HaloRasAR group (ΔR/R₀ = −8.5% ± 0.5%, n = 77, ΔR/R₀ = −1.5% ± 0.5% n = 66 cells for BAY293 and R89L groups, respectively, p=2.500e-016, Mann-Whitney test). (g) Region-selective ROIs at cell periphery or Golgi show distinct far-red/green emission ratio responses, and (h) these responses at t = 10 are significantly lower in the Golgi ROI group (p=5.238e-005, Mann-Whitney test). (i-k) HaloRasAR expressed in MIA PaCa-2 subjected to tandem 10 μM BAY293 treatment followed by RMC6291 treatment. (i) Responses of whole cell ROIs (n=51 cells HaloRasAR, 35 cells R89L control) plotted alongside (j) Region selective ROIs for the periphery and Golgi (n = 61 cells each). (k) ΔR_1st/ΔR_max shows a significantly larger proportional response at Golgi upon BAY293 treatment compared to the periphery (p=3.755e-005 for Golgi vs Periphery, Mann-Whitney test). (l) Cytosolic EKAR4 expressed in MIA PaCa-2 and subjected to tandem 10 μM BAY293, followed by 100 nM RMC6291 inhibition (n = 96 cells BAY293, 120 cells RMC6291, 120 cells tandem addition). All panels are combined results of 3 independent experiments. For time-series traces, solid lines indicate the mean and shaded areas represent s.e.m. All Mann-Whitney comparisons are two-sided. Pseudocolor images are representative of 3 independent experiments.

BAY293 functions by blocking Ras-SOS1 interactions. While constitutively active mutants such as KRasG12C are generally less sensitive to SOS1 GEF activity compared to WT Ras isotypes, BAY293 has been used to inhibit cells harboring oncogenic Ras mutations.^4,23 Interestingly, MIA PaCa-2 cells expressing HaloRasAR showed a rapid decrease in the basal far-red/green emission ratio upon BAY293 treatment (Fig. 4e–f). Like RMC6291 treatment, location-selective ROIs revealed subcellular differences in BAY293-induced responses; however, in contrast to the KRasG12C-specific inhibitor, greater emission ratio reductions were observed with Golgi-specific ROIs compared to the cell periphery (Fig. 4g–h, ΔR/R₀ = −7.1% ± 0.9%, periphery, n = 49 cells, ΔR/R₀ = −12.5% ± 0.9%, Golgi, n = 47 cells, p<0.0001).

Given the contrasting subcellular efficacy of these two drugs, we sought to further interrogate this phenomenon by sequential inhibition of SOS1 and Ras activity, allowing comparisons of spatiotemporal responses within individual cell populations. Cells were first treated with BAY293 followed by RMC6291. The partial inhibition with BAY293 is demonstrated by a further decrease in Ras activity with RMC6291 (Fig. 4i, Extended Data 10e). Location-selective ROIs further revealed a larger BAY293 response, calculated as the ratio of the BAY293-induced decrease to the total ratio decrease (ΔR_1st /ΔR_max), at the Golgi than at the plasma membrane (Fig. 4j–k, ΔR_1st /ΔR_max = 35 ± 2% for periphery, ΔR_1st /ΔR_max = 48 ± 2% for Golgi, p<0.0001). We then used targeted HaloRasAR to confirm the subcellular differences between BAY293 and RMC6291. Cells expressing pm-HaloRasAR showed little sensitivity to BAY293 but a pronounced reduction in their far-red/green emission ratio upon RMC6291 addition. Conversely, Golgi-HaloRasAR exhibited a clear decrease in ratio upon BAY293 treatment, compared with only a slight drop following RMC6291 treatment (Extended Data Fig. 10g–i). To further verify the preference for inhibition of Golgi Ras by BAY293, we reversed the order of dual inhibitor addition, adding RMC6291 first and then BAY293 second. We observed a significantly smaller proportional change upon the first inhibitor treatment, as measured by ΔR_1st/ΔR_max, with initial RMC6291 treatment compared to initial BAY293 treatment (Extended Data Fig. 10j–k). These data suggest that KRasG12C and SOS1 inhibition target spatially distinct pools of Ras activity.

ERK is the primary effector kinase of Ras signaling in growth and proliferation. Given the partial Ras inhibition we observed with BAY293, we asked whether ERK activity would similarly exhibit partial inhibition. To test this, we used the ERK sensor EKAR4¹⁹ to visualize downstream ERK activity in the context of sequential treatment. As with Ras, BAY293 triggered a partial reduction in the EKAR4 Y/C emission ratio, with a further decrease observed upon RMC6291 addition (Fig. 4l). Interestingly, treatment with RMC6291 alone completely inhibited ERK activity (Extended Data Fig. 10l, ΔR/R₀ = −21.0% ± 1.1% for RMC6291 treated cells, ΔR/R₀ = −18.7% ± 1.0% for BAY293 and RMC6291 treated cells, n = 120 cells each, n.s.), suggesting that the elevated ERK activity requires KRasG12C. In total, these results suggest that GEF inhibition and mutant KRas inhibition target distinct subcellular pools of Ras activity.

DISCUSSION

The field of Ras biosensors has developed for over 20 years,^8,24 yet few sensors are capable of sensitive and quantitative detection of endogenous Ras activity.¹² While translocation probes may report endogenous Ras activity, their quantitation may be impaired by intensity artifacts driven by membrane ruffling.¹³ FRET-based Raichu-Ras biosensors consist of the Ras catalytic domain fused to an RBD and sandwiched between a FRET pair. Upon Ras activation, these two modules interact to produce FRET increases.¹⁴ This design requires overexpressing modified Ras, which may lead to altered subcellular localization²⁵ and biochemical properties.²⁶ Our FRET-based RasAR was effective in the context of Ras overexpression but insensitive to endogenous Ras.⁴⁴ A more recently developed Ras sensor Ras-LOCKR-S incorporates a de novo-designed alpha-helical switch to convert the binding of endogenous, active Ras into changes in FRET.¹² However, this sensor features a bimolecular design that requires carefully matched expression of each sensor component and showed a limited dynamic range (Extended Data Fig. 3g).

To address these limitations, we have developed a chemigenetic biosensor, HaloRasAR, that is capable of robustly detecting endogenous Ras activity. HaloRasAR leverages cpHaloTag with an N-terminally fused high-affinity RBDv12 capable of interacting with endogenous Ras. Ras-RBDv12 interaction alters the chemical environment of a covalently attached far-red fluorescent dye JF₆₃₅, favoring its bright, zwitterionic state. The use of protein scaffolds to perturb the equilibrium between the zwitterionic and lactone state of JF₆₃₅ has the potential to produce large intensiometric responses, as indicated by the superior dynamic range of HaloRasAR. Furthermore, the use of far-red silicon rhodamine dyes with increased brightness, signal-to-background, and photostability allows using low concentrations of sensor in cells. HaloRasAR demonstrates low nanomolar affinity amenable to detecting low concentrations of acutely activated endogenous Ras in living cells. Given that low sensor concentrations are desirable in detecting endogenous Ras activity in cells (Extended Data Fig. 6a–c), high brightness and photostability are key factors to enhance signal-to-noise ratio (SNR).

Both Ras and PKC are longstanding molecular targets for cancer therapy.^3,22 While Ras is clearly oncogenic,³ PKC family isozymes show an ambiguous role in cancer.²² In the context of Ras-driven malignancy, conventional PKC isozymes are primarily tumor suppressive, with low cPKC expression associated with poor prognosis in multiple cancers.^27–30 In contrast, novel PKCε has been shown to synergize with Ras to promote tumorigenesis in breast cancer³¹ and non-small cell lung cancer, NSCLC.³² Phorbol esters have previously been demonstrated to activate Ras through PKC signaling^7,33, yet strong agonism of PKC has also been reported to induce KRas4B translocation through PKC phosphorylation at Ser181.²¹ We have found that different strengths of PKC input can result in distinct spatiotemporal features of Ras activity. With moderate stimulation, we observed Ras-activity-rich ruffles—perhaps indicative of previously reported metastatic/migratory roles for Ras and PKC.^34,35 Mechanistically, we showed that inhibiting SOS1 GEF significantly inhibited peripheral Ras activity in the context of moderate stimulation. This observation aligns with previously proposed models, which suggest that RasGRP-family RasGEFs may initiate a positive feedback loop of activation between Ras and SOS1.²⁰ On the other hand, a strong PKC agonism promoted pronounced Ras activation at the Golgi in a manner independent of SOS1, which is more in line with previous reports linking phorbol ester stimulation to RasGRP1 activation of Golgi Ras activity.⁹

For years, targeting Ras has been stymied by the limited topology of this small GTPase and its high affinity for guanine nucleotides.³⁶ However, recent advances have resulted in the first FDA-approved drugs to inhibit mutant Ras in its inactive state, such as Sotorasib.^37,38 Since then, additional drugs with different mechanisms of action have emerged to target Ras. The novel RMC6291 binds to KRasG12C and induces dimerization with cyclophilin A, thus blocking downstream effector signaling and enabling rapid inhibition of active Ras (Fig. 4a), compared to the slow kinetics that we observed with Sotorasib.³⁹ Other Ras signaling pathway components have also been targeted, like blocking SOS1 GEF activity towards Ras.²³ Using HaloRasAR as a tool for live-cell pharmacology, we found that the inhibitory effect of SOS1 inhibitor BAY293 was more prominent at the Golgi compared to the plasma membrane in MIA PaCa-2 cells. A possible explanation for this observation involves wildtype H/NRas isoforms that are localized to the Golgi and sensitive to GEF inhibition. H/NRas isoforms are palmitoylated, and thus dynamically traffic between the plasma membrane and the Golgi in a process regarded as the “acylation cycle”. While these wildtype H/NRas isoforms are not constitutively activated in MIA PaCa-2 cells, their activity in our experiments may be driven by the “cross-activation” mechanism described previously⁴⁰, wherein SOS1 is allosterically activated by constitutively active KRasG12C. This observation raises important questions regarding the differential cellular perturbations from direct Ras and Ras pathway inhibitors in cancers driven by different pathological ras genotypes. If spatially compartmentalized inhibitions are observed across pathway and direct Ras inhibitors, what are the downstream and functional effects of these inhibitory profiles? Future studies, enabled by HaloRasAR, EKAR and other molecular tools, will address these questions and provide new insights into drug efficacy and resistance.

HaloRasAR should enable numerous new studies. HaloRasAR has allowed definitive characterization of endogenous Ras signaling at the Golgi. Subcellular targeting of HaloRasAR should similarly enable future investigations of Ras activity at other non-canonical Ras signaling compartments, such as the mitochondria and specific signaling microdomains, as well as new insights into critical dynamic processes like cell migration. Red-shifted sensors, especially those with far-red emission, are desirable for in vivo imaging due to the reduced autofluorescence and increased tissue penetration at these wavelengths.⁴¹ While the in vivo use of the highly fluorogenic silicon rhodamine dyes such as JF₆₃₅ is hampered by poor bioavailability, HaloRasAR can be labeled with other fluorogenic dyes with greater bioavailability to explore in vivo imaging, as demonstrated in recent studies.⁴² Future development will also explore alternative designs that are compatible with less fluorogenic and more bioavailable fluorophores like JF₆₆₉.^42–44 Finally, HaloRasAR has revealed location-specific inhibition of the Ras pathway and should further facilitate the development of Ras inhibitors.⁴⁰ Recent advances in the development of Ras therapeutics have made targeting this once “undruggable” GTPase a reality. However, direct pharmacological targeting of Ras is challenged by resistance mechanisms that circumvent Ras inhibition, with wildtype Ras isoforms often playing compensatory roles.^45,46 Live-cell pharmacology using HaloRasAR allows the spatial and temporal characterization of drug-target interactions in living systems, leading to further insights into mechanisms underlying drug effects and resistance. The continued development of high-performance endogenous Ras biosensors provides useful tools to help target this notorious target for cancer therapy and better understand mechanisms of resistance.

METHODS

Materials

JF₆₃₅-CA (referred to as JF₆₃₅) and JF₆₆₉-CA (referred to as JF₆₆₉) dyes were kindly provided by Dr. Luke Lavis (Janelia Research Campus). Additional JF₆₃₅ was purchased (Promega #HT1050) and prepared as 250 μM stocks (all experiments with transiently transfected HaloRasAR) or 25 μM stocks (experiments with pLenti-TetOn integrated HaloRasAR). Epidermal Growth Factor (EGF, Sigma-Aldrich) was prepared as a 1000x stock in ultra-pure water (MilliQ). Phorbol 12,13-dibutyrate (PDBu, LC Labs, P-4833), Gö6983 (MilliporeSigma, G1918), AG1478 (Selleckchem, S2728), BAY293 (Sigma, SML2703), RMC6291 (MedChemExpress, HY-153346) were prepared as 4000x stocks in DMSO (Sigma-Aldrich). RIPA lysis buffer used is composed of 150 mM NaCl, 0.05% (v/v) NP-40, 2.5% (w/v) sodium deoxycholate, 0.1% (w/v) sodium dodecyl sulfate, 50 mM Tris, pH 8.

Gene Construction

All constructs were generated using a one-step multi-fragment Gibson Assembly reaction (NEB 2x High Fidelity Master Mix). All fragments were PCR amplified with Q5 polymerase (NEB). For HaloRasAR, PCR fragments were generated from mNeonGreen, a gift from Nathan Shaner Lab, circularly permuted HaloTag, derived from linker variants from HaloAKAR.¹⁶ RBDv12, was derived from wt-Raf1, sourced from CRY2PHP-mCherry-Raf1--a gift from the Cui Lab,⁴⁷ according to previously reported mutation scheme.¹⁷ wt-RBD used in plasma membrane HaloRasAR was sourced from the same construct. Golgi-targeting Giantin, a gift from the Maly lab,³⁹ was fused to the c-terminus of HaloRasAR. psPAX2 (#12260) and pMD2.G (#12259) were sourced from Addgene. The minimal KRas HVR responsible for membrane targeting was added using overhang extension.

Cell Culture

COS7 (ATCC CRL-1651) and HEK293T (ATCC CRL-3216) cells were cultured in Dulbecco modified Eagle medium (DMEM, Gibco) containing 4.5g/L glucose and supplemented with 10% (v/v) fetal bovine serum (FBS; Sigma) and 1% (v/v) penicillin-streptomycin (Pen-Strep, Sigma-Aldrich). MIA PaCa-2 cells (ATCC CRM-CRL-1420) were maintained containing 4.5g/L glucose and supplemented with 10% (v/v) fetal bovine serum (FBS; Sigma), 2.5% (v/v) donor horse serum (DHS; Sigma) and 1% (v/v) penicillin-streptomycin (Pen-Strep, Sigma-Aldrich). Cells were maintained in a humidity-controlled incubator at 37°C with 5% CO₂. Cells were seeded onto sterile 35-mm glass-bottomed dishes and grown to 60-70% confluency. COS7 and HEK293T cells were transfected with PolyJet (SignaGen Laboratories) 24 hr prior to imaging. MIA PaCa-2 cells were transfected with Lipofectamine (Prior to transfection the COS7 cells were washed with Dulbecco Phosphate Buffered Saline (DPBS, Gibco), and medium was changed to a serum-free DMEM to serum starve the cells for the 24 hr during transfection prior to imaging. MIA PaCa-2 dishes were maintained in 10% FBS, 2.5% DHS through the duration of cell culture until imaging. For transient transfection of COS7 and MIA PaCa-2 cell lines, 250 ng of HaloRasAR and 750 ng empty pcDNA3.1+ vector is used. For HEK293T, HaloRasAR is transfected with 50 ng HaloRasAR, 950 ng empty pcDNA3.1+ vector is used.

Lentiviral Transduction of Tet-On HaloRasAR and Induction

Lentiviral vectors coding expression of rtTA transcription factor and doxycycline inducible-HaloRasAR vectors was packaged in HEK293T cells. Specifically, HEK293T cells were cotransfected with relevant lentiviral transfer plasmids + psPAX2 + pMD2.G using PolyJet. After 48hr the supernatant was collected, filtered with a 0.45 μm syringe filter, and then concentrated using Lenti-X concentrator (Takara Bio USA, Inc.) and added to HEK293T cells in 35 mm dishes. Polybrene Transfection Reagent (TR-1003-G, Millipore) was supplemented to improve the transduction. After 48 hours, cells were passed in fresh growth medium containing puromycin (1 μg/mL) and blasticidin (2 μg/mL) to select transduced cells for at least one week. Cells were maintained in selection medium containing ½ selection concentration and passed for related experiments. For low induction of HaloRasAR, 100-500 pg/mL of doxycycline was used to treat the cells for 24 hours to modestly enhance HaloRasAR expression above “leaky” levels and removed prior to mixing with 25 nM of JF₆₃₅ dye for >2 hours. Low dye concentration was used to minimize background labeling.

Time-Lapse Widefield Epifluorescence Imaging

At least 3 hours prior to imaging, COS7, HEK293T, and MIA PaCa-2 cells are placed in serum free-DMEM (COS7, HEK293T) containing 250 nM JF₆₃₅ or JF₆₆₉, MIA PaCa-2 is maintained at 10% FBS, 2.5% DHS with 250 nM JF₆₃₅. Cells were washed three times with Hanks’s balanced salt solution (HBSS, Gibco) and incubated in a humidity-controlled incubator for 30-45 minutes. Following mounting on the imaging, stage cells were left to incubate at 37°C for 10 minutes on the imaging stage prior to imaging. Cells were imaged in HBSS in the dark at 37°C. For EKAR4-NES experiments in MIA PaCa-2, cells placed in phenyl-red free DMEM with 10% FBS 30 minutes prior to imaging and imaged in this media. All drugs added at times indicated in plots. All drug additions using DMSO-dissolved drugs were performed at .025% final DMSO upon addition—with higher DMSO loads of .1% we observed DMSO-induced artefactual drops in far-red/green emission ratio.

Cells were imaged on a Zeiss Axio Observer Z1 microscope equipped with a 20x/0.8 NA objective for data in Fig 1c, Supp. Fig 1a–d or a 40x/1.3 NA objective for all other data, and a Hamamatsu Flash 4.0 sCMOS camera, controlled using MicroManager 2.0 software. For time series, images from multiple fields of view were acquired at each time point at 30sec intervals. CY5 was imaged using a 640DF30 excitation filter, a T660lpxr dichroic mirror, and a 700DF5 emission filter. GFP was measured using a 480DF30 excitation filter, a T505DCXR dichroic mirror, and a 535DF50 emission filter. RFP was imaged using a 568DF55 excitation filter, a Q600LPXR dichroic mirror, and a 653DF95 emission filter. CFP was imaged using a 420DF20 excitation filter, a T455LPXR dichroic mirror, and a 470DF40 emission filter. CYFRET was imaged using a 420DF20 excitation filter, a T455LPXR dichroic mirror, and a 535DF25 emission filter. YFP was imaged using a 495DF10 excitation filter, a T515LP dichroic mirror, and a 535DF25 emission filter. Exposure times for HaloRasAR were 500 ms for CY5 and 150 ms for GFP, all other exposure times were 500 ms. All filter sets were controlled by a Lambda 10-2 filter-changer (Sutter Instruments). All fluorescence images were background corrected by rolling circle background subtraction in ImageJ 1.52v using a rolling-ball circle radius of 300-500 pixels. Far-red/green emission ratios were calculated by dividing CY5 channel values by GFP channel values. Cyan/yellow or yellow/cyan emission ratios were then calculated by dividing the CFP channel values by the CYFRET channel values, or the CYFRET channel values by CFP channel values, respectively. For the time-series analysis, all values were normalized to the time point prior to drug addition by dividing each value by basal ratio value prior to drug addition. Co-expression experiments were processed using Cell Profiler on background subtracted images and segmented using GFP channel images for all except HaloTag brightness segmentations, which only had far-red channels to segment. Individual t_1/2 measurements were acquired using custom script using Python.

Imaging Analysis

t_1/2 calculations from acute perturbation data:

To calculate t_1/2, normalized traces from acute perturbation experiments are used as presented, baseline is calculated as the average of all prestimulation values. To minimize the effect of artefacts from low-response value traces, maximum values are found by scanning three consecutive values outside the baseline, averaging each set of three, and finding the max value of each set. To further minimize contaminating effects of low or no response cells, t_1/2 values are calculated from maximum value calculations greater than one standard deviation higher than baseline values, with all others excluded. Precise t_1/2 value is found from linear interpolation of two data points closest to half maximum.

Sustained Activity Metric (SAM) calculations:

Sustained activity metric is defined as the ratio of the final time point (time point 15 used exclusively in this work) to the maximum time point. To calculate this from normalized traces of acute perturbation experiments, baseline is calculated as average of all prestimulation values, with maximum calculated as before with a scan of two points with the maximum average defined as the maximum. Sustained activity metric then finds the ratio of final minus the baseline over the maximum minus the baseline. Cells with maxima or final time points less than the baseline average are excluded.

Immunoblotting

For phospho-ERK measurements, HEK293T cells were transiently transfected under conditions described above. Cells were washed with ice-cold PBS and then lysed in RIPA lysis buffer containing protease inhibitor cocktail (vendor), 1 mM PMSF, 1 mM Na₃VO₄, 1 mM NaF, and 25 nM calyculin A. Total cell lysates were incubated on ice for 30 min and then centrifuged at 4°C for 20 min. Total protein was separated via 4-15% SDS-PAGE and transferred to PVDF membranes. The membranes were blocked with TBS containing 0.1% Tween-20 and 5% bovine serum albumin and then incubated with primary antibodies overnight at 4 °C. After incubation with the appropriate horseradish peroxidase-conjugated secondary antibodies, the membranes were developed using horseradish peroxidase-based chemiluminescent substrate (34579 and 34076, Thermoscientific). The intensity of the bands was quantified with ImageJ 1.52v software. The following antibodies from Cell Signaling Technology (CST) were used for immunoblotting: ERK1/2 (CST #4695), p-ERK (CST #4370), GAPDH (CST #2118). ERK1/2 and p-ERK antibodies were used at 1:5000 dilution. GAPDH was used at 1:10000 dilution. The horseradish peroxidase-labeled goat anti rabbit (PI31460) secondary antibodies were purchased from Pierce.

In Vitro Characterization of HaloRasAR

Protein production and purification:

Proteins were expressed in E. coli strain BL21. LB cultures were grown at 37 °C to an optical density (OD600nm) of 0.6, induced by the addition of isopropyl-β-D-thiogalactopyranoside (IPTG, 0.5 mM) and grown at 16 °C overnight. The cells were pelleted by centrifugation (3,100 g, 10 min, 4 °C), resuspended in lysis buffer (50 mM PBS, pH 7.4, 300 mM NaCl) and lysed by sonication (1 min sonication, 1 min rest, 3 min total, 60%, Sonics Vibra cells) on ice. The cell lysate was cleared by centrifugation (15,000 g, 30 min, 4 °C). Clarified lysate was loaded onto HisPurTM Ni-NTA resin (ThermoFisher Scientific). Unbound protein was washed away using wash buffer (50 mM PBS, pH 7.4, 300 mM NaCl, 25 mM imidazole), and bound protein was subsequently eluted using elution buffer (50 mM PBS, pH 7.4, 300 mM NaCl, 250 mM imidazole). Eluted fractions were analyzed via SDS–PAGE (coomassie staining), pooled and concentrated using Amicon Ultra-15 centrifugal filter device (30 kD cut-off, HaloRasAR, 3 kD cutoff, RBDv12, Millipore) followed by buffer exchange into 50 mM PBS, 300 mM NaCl, pH 7.4 (<0.1 mM imidazole). The proteins were frozen and stored at −80 °C.

Excitation and emission spectra:

Purified biosensor (1 μM) was incubated with JF₆₃₅ (1 μM) in modified PBS binding buffer (50 mM PBS, pH 7.4, 0.5mg/mL BSA) for >30 min. Afterwards, excitation and emission spectra of 100 μL of each solution were loaded into wells of black-walled clear bottom plates and were measured on TECANspark. Excitation scans were collected at 680 nm and excited from 550-650 nm, and emission scans were performed from 645-745 nm and excited at 615 nm. Spectra for mNeonGreen were also gathered. Excitation collected at 545 nm emission and varied from 400-510 nm. Emission spectra were excited at 485 nm and collected from 510-600 nm. Spectra were gathered for purified HaloRasAR with or without the presence of 1.5 μM KRasG12V bound to GppNHp, purchased from BPS (#101359-1).

Extinction coefficient measurements:

Purified HaloRasAR or HaloTag7 (20 μM) was incubated with JF₆₃₅ (10 μM) in PBS binding buffer (50 mM PBS, pH 7.4, 0.5 mg/mL BSA) for >30 min. Bound HaloRasAR was subsequently incubated with KRasG12V-GppNHp (BPS #101359-1) at 25 μM. Afterwards, the absorbance spectra of each solution was measured on a dual absorbance emission spectrophotometer (Duetta Horiba, EzSpec 1.3.0.63.) Data were baseline corrected, and the maximum absorbance values at 640 nm were used to calculate the extinction coefficient relative to the reported value for HaloTag7-JF₆₃₅ (81,000 M⁻¹ cm⁻¹).⁴⁸

Relative quantum yields:

In addition to measuring the above solutions for absorbance spectra (see extinction coefficient measurements), emission spectra (630-730 nm) excited at 610 nm were measured on a combined fluorescence and absorbance spectrometer (Duetta Horiba, EzSpec 1.3.0.63.). All spectra were baseline corrected and the area under the emission spectra were calculated. The measured absorbance data (A) and measured areas for fluorescence intensity (I) were used to calculate the relative quantum yields (φ) from the reported value for HaloTag7-JF₆₃₅ (0.75)⁴⁸ according to equation (1).⁴⁹ n_s and n_r are refractive indices of respective buffers, the ratio of which equal 1.

$${\mathrm{\Phi }}_{\mathrm{S}}={\mathrm{\Phi }}_{\mathrm{R}}\left(\frac{I_{\mathrm{S}}}{I_{\mathrm{R}}}\right)\left(\frac{1-{10}^{-A_{\mathrm{R}}}}{1-{10}^{-A_{\mathrm{S}}}}\right){\left(\frac{n_{\mathrm{S}}}{n_{\mathrm{R}}}\right)}^2$$ equation (1)

Competition Assay:

For the measurement of sensor K_d, a competition assay was performed. Purified Halo RasAR (500 nM), labeled with JF₆₃₅ (500 nM) for >30min, and 250 nM KRasG12V-GppNHp (BPS #101359-1) were fixed across a range of six purified RBDv12 concentrations, 0.05 nM-5.0 μM varied by one order of magnitude for each concentration. Samples are mixed in binding buffer (PBS 50 mM, 300 mM NaCl, 0.5 mg/mL BSA), and 50 μL of samples are loaded into clear-bottom, black-walled plates and measured in TECANspark. To maximize SNR of the assay, emission spectra were gathered, with excitation at 630 nm, and emission wavelengths 655-755 nm gathered. mNeonGreen emission spectra were gathered, with 480nm excitation and, emission from 510-600 nm. Spectra for each sample were analyzed for far-red/green emission ratio by calculating the area under peak emission and then dividing far-red peak emission areas under green peak emission areas. Data were curve-fit using algorithms provided by GraphPad Prism 10 (GraphPad software), and IC₅₀ values calculated. As measurements were gathered under high ligand depletion factor (δ₀) conditions, we used the full equation for Cheng-Prusoff, including Goldstein-Barrett shifts, equation (2).⁵⁰ The dissociation constant K_d(RBD) for RBDv12 is 3 nM based on previous work. L_T is total HaloRasAR concentration and K_dL is the dissociation constant of HaloRasAR. As δ₀ varies with K_dL, least-squares minimization is used to find solutions to K_dL.

$$K_{\text{d(RBD)}}=\frac{I{C}_{50}}{1+\frac{[L_{\text{T}}]({1-\delta }_0/2)}{K_{\text{d(L)}}}+\frac{{\delta }_0}{(1-{\delta }_0)}}$$ equation (2)

Statistics and Reproducibility

All imaging experiments were performed with independent experimental replicates with all replicate attempts being successful. Experiments from figures 1–4, Extended data figures 1–10 were performed with 3 independent replicates. Statistical analyses were performed using GraphPad Prism 10 (GraphPad Software). In data sets with large sample sizes (Fig. 1a and Extended Data Fig. 3a–d), small effects led to statistically significant changes in typical statistical comparisons. Due to this, we performed Cohen’s d, effect size calculations based on pairwise t-test. For smaller data sets. Mann-Whitney comparisons are performed for all tests involving two experimental groups, in cases of multiple groups we performed one-way ANOVA to compare significant difference among groups, and Dunnett’s comparison for tests of all groups to a single experimental group of interest. Averaged live-cell imaging time-courses depict mean ± s.e.m.; and scatter plots with depictions of the median and quartiles, are indicated in the figure legends.

Extended Data

Extended Data Figure 1. Linker length screening of HaloRasAR variants.

(a) (top) Linker sequence variants of HaloRasAR. Orange color denotes the terminal residues of RBDv12. (bottom) Domain structure of initial HaloRasAR variants. The linker between RBDv12 and cpHaloTag was varied in 1 amino acid (aa) increments across 10 variants depicted in the table above. (b) Far-red fluorescence of all 10 variants coexpressed with HRas dominant negative (DN) mutant HRasS17N or constitutively active mutant HRasG12V. All variants show higher fluorescence with HRasG12V coexpression, but the 2aa variant shows the most improved fluorescence. All endpoints of column bar plots indicate the median value for each data set; error bars indicate the 95% confidence interval.

Extended Data Figure 2. In vitro characterization of HaloRasAR.

(a) Fluorescence excitation and emission spectra measured of 1 μM purified HaloRasAR with or without excess KRasG12V-GppNHp. (b) Fluorescence emission and absorbance measurements of HaloRasAR (20 μM) labeled with JF₆₃₅ (10 μM), mixed with or without excess KRasG12V (>20 μM). Comparison to purified HaloTag7 (20 μM) labeled with JF₆₃₅ (10 μM) is performed in parallel. (c) Equation for relative quantum yield Φ measurements. Sensor S is calculated relative to HaloTag7 reference R, fluorescence intensity I is calculated as area under the curve of emission profiles. Absorbance A derived from peak absorbance. n is the refractive index of aqueous buffers used. (d) Table listing calculated molar absorptivity and quantum yield data for HaloRasAR. (e) Representative data of competition assay. Labeled HaloRasAR is at a fixed concentration of 500 nM and KRasG12V is at a fixed concentration of 250 nM. Purified RBDv12 is titrated from 0.075 nM to 7500 nM. far-red/green emission ratios are measured at different concentrations. (f) Equation for calculating K_d at high ligand depletion. K_d(RBD) is the reported value for RBDv12, IC₅₀ is calculated from curve fitting algorithm. L_T is the total HaloRasAR concentration and K_dL is the K_d of HaloRasAR. δ₀ is the ligand depletion factor. (g) Calculated IC₅₀ and K_dL are provided. n = 3 samples, mean ± standard deviation.

Extended Data Figure 3. Further cellular characterization of HaloRasAR.

(a) HaloRasAR, HaloRasARwt and HaloTag expressed in COS7 cells and labeled with JF₆₆₉ for 3 hours. Individual cells far-red fluorescence shown (Cohen’s f = 0.57, mean = 654, 332, 2869, n = 988, 1701, 1288 cells for RBDv12 biosensor, RBDwt biosensor, HaloTag7, respectively). (b) Responses of HaloRasAR coexpressed with HRasG12V or S17N are plotted against green fluorescence to account for possible variations in sensor expression (c) HaloRasARwt bearing RBDwt is coexpressed with HRasS17N or HRasG12V and measured for far-red/green ratios. Like HaloRasAR, elevated far-red/green ratios are observed in groups expressing constitutively active HRasG12V (n = 1551 cells, HRasS17N, n = 1114 cells, HRasG12V, Cohen’s d = 1.89). (d) Far-red/green ratios plotted against green fluorescence to account for possible variations in sensor expression. (e) K_L-Z equilibria of JF₆₃₅ and JF₆₆₉. The putative mechanism of JF₆₃₅-based sensors depends on JF₆₃₅ primarily adopting the lactone form, and perturbation by the surrounding protein shifts the equilibrium to the zwitterionic state. The equilibrium of JF₆₆₉ is predominantly in the zwitterionic state already and hence does not allow further perturbation. (f) HaloRasAR exhibits no apparent response when labeled with JF₆₆₉ (n = 24 cells, JF₆₃₅, 28 cells, JF_669, [traces show mean ± s.e.m.]). (g) Ras-LOCKR-S is coexpressed with HRasG12V or HRasS17N and Y/C emission ratios are measured (n = 4229 cells, HRasG12V, n = 2288 cells, HRasS17N, p<0.0001, Mann-Whitney test).

Extended Data Figure 4. Individual far-red and green fluorescence channels.

(a-c) Individual channels with their normalized intensities plotted for Figure 1e. Far-red fluorescence shown in black, green fluorescence represented with green. Solid lines indicate the mean and shaded areas represent s.e.m.

Extended Data Figure 5. Endogenous Ras activity detected with HaloRasAR.

(a) Pseudocolor image of far-red/green ratio of COS7 cells stimulated with EGF. (b) t_1/2 data plotted in minutes of location-selective ROIs reveal slower kinetics of activation at the Golgi (p=1.247e-007 by Mann-Whitney test). (c) Domain structure of plasma membrane HaloRasAR (pm-HaloRasAR). (d) Colocalization of Golgi-HaloRasAR with mTagBFP2-Giantin (Pearson’s R = 0.96). (e) pm-HaloRasAR-WT bearing a C-terminal KRas CAAX domain stimulated with EGF compared to AG1478 negative control (n = 33 cells, DMSO, n = 29 cells, AG1478). (f) t_1/2 comparison between pm-HaloRasAR and Golgi-HaloRasAR stimulated with EGF, Golgi t_1/2 data from Figure 2g (n = 33 cells, pm-HaloRasAR, n = 43 cells, Golgi-HaloRasAR, p=2.951e-020 by Mann-Whitney test). (g) Peripheral activity of HEK293T cells expressing HaloRasAR stimulated with EGF (n = 60 cells, HaloRasAR, n = 36 cells, R89L negative control). For time-series traces, solid lines indicate the mean and shaded areas represent s.e.m. All Mann-Whitney comparisons are two-sided.

Extended Data Figure 6. Enhanced response of HaloRasAR at low expression.

(a) HaloRasAR, under the control of a TRE3G promoter, is stably integrated into the HEK293T cell genome using lentiviral transduction. To achieve a low concentration of HaloRasAR, we use low concentrations of DOX. (b) Traces from three independent experiments (n = 150 cells for whole cell, n = 129 cells for peripheral activity, [mean ± s.e.m.]). Peripheral and whole cell activities are plotted. (c) Comparison of peripheral R_max/R₀ for low induction of stably expressed, TRE3G-driven HaloRasAR versus transiently transfected, CMV-driven HaloRasAR (R_max/R₀ = 29.1% ± 0.8% for TRE3G-driven expression, n = 129 cells, R_max/R₀ = 11.6% ± 0.6% for CMV-driven expression, n = 60 cells, [mean ± s.e.m.], p=4.986e-031). (d) Representative images from three independent experiments before and after EGF stimulation. Far-red (top) and far-red/green ratio (bottom) are included.

Extended Data Figure 7. Phospho-ERK blots of HEK293T cells expressing HaloRasAR.

(a) HaloRasAR is expressed in HEK293T cells at amounts equivalent to previous experiments. Cells are stimulated with EGF for 0, 10 or 30 minutes and then lysed. Blots for phospho-ERK are quantitated relative to total ERK (mean ± s.e.m.). (b) Representative blot from three independent experiments.

Extended Data Figure 8. Individual far-red and green fluorescence channels.

Normalized intensity of response values from endogenous Ras responses plotted. Far-red fluorescence shown in black, green fluorescence represented with green. (a-c) Cytosolic HaloRasAR detecting endogenous Ras responses from Figures 2a, c. (d-e) Targeted HaloRasAR detecting endogenous Ras responses from Figure 2g, and Extended Data Figure 5e. (f) HEK293T cells expressing low levels of HaloRasAR detecting endogenous Ras responses from Extended Data Figure 6. For time-series traces, solid lines indicate the mean and shaded areas represent s.e.m.

Extended Data Figure 9. PKC-mediated Ras activity.

(a-b) Average time courses showing responses of ExRai-CKAR2 expressed in COS7 cells and stimulated with 50 nM PDBu (moderate) or 200 nM PDBu and 670 nM ionomycin (strong) (n = 203 cells, 50 nM PDBu, n = 90 cells, 200 nM PDBu/Iono, p=2.500e-045, Mann-Whitney test). (c) Golgi-HaloRasAR stimulated with strong and moderate agonism (ΔR/R₀ = 20.9% ± 1.9% for strong, ΔR/R₀ = 7.8% ± 1.1% for moderate, n = 44 cells for each group), resulting in (d) significant elevation in response to strong agonism (p=1.433e-006, Mann-Whitney test) (e) Kinetics of HaloRasAR response at cell periphery to 50 nM PDBu is significantly slower with BAY293 pretreatment. Many cells in BAY293 pretreatment group did not exhibit significant response and are excluded (p=0.0005, Mann-Whitney test). (f) Peripheral HaloRasAR response shows no significant difference in AUC with BAY293 pretreatment (n = 45, 36 cells for DMSO, BAY293, respectively, p=0.6236 by Mann-Whitney test). For time-series traces, solid lines indicate the mean and shaded areas represent s.e.m. All Mann-Whitney comparisons are two-sided.

Extended Data Figure 10. Inhibition of endogenous Ras with BAY293/RMC6291.

(a) Colocalization of HaloRasAR expressed in MIA PaCa-2 cells with mTagBFP2-Giantin (Pearson’s correlation coefficient R = 0.59). (b) Response comparison of responses to 100 nM RMC6291 treatment in HaloRasAR expressing cells or HaloRasAR-R89L control group (ΔR/R₀ = −15.3% ± 0.6% n = 49 cells, ΔR/R₀ = −0.2% ± 0.4% n = 48 cells for HaloRasAR, HaloRasAR-R89L groups, respectively, p=1.570e-028 by Mann-Whitney test). (c) No significant difference in kinetics among region-selective ROIs in RMC6291 treatment by one-way ANOVA (n=49 cells for whole cell, 36 cells for peripheral, 36 cells for Golgi ROIs, F=1926). (d) With 10 μM BAY293 treatment, ROIs at the cell periphery (Periph.) showed slower kinetics of activation (n=77 cells for whole cell, 47 cells for peripheral, 47 cells for Golgi ROI’s, p=0.0060, Periph. vs. Whole cell ROI, p=0.0137, Periph. vs. Golgi ROI, one-way ANOVA with Dunnett’s test). (e) Tandem treatment of BAY293 followed by RMC6291, ΔR/R₀ values following BAY293 treatment is significantly higher than subsequent RMC6291 treatment (n=78 cells for each group, p=4.954e-023, Mann-Whitney test). (f) Tandem treatment of HaloRasAR expressing cells with RMC6291 initially and subsequent BAY293. (g) Golgi-HaloRasAR (n = 60 cells, drug treatment, n = 25 cells DMSO addition control) or (h) pm-HaloRasAR (n = 57 cells for drug treatment, n = 23 cells for DMSO addition control) expressed in MIA PaCa-2 cells subjected to tandem treatment of BAY293 followed by RMC6291. (i) Comparison of ΔR_1st/ΔR_max values of targeted HaloRasAR sensors following BAY293 and subsequent RMC6291 treatment (BAY293-induced change: ΔR_1st /ΔR_max = 40 ± 2% for plasma membrane, n=57 cells, ΔR_1st /ΔR_max = 64 ± 3% for Golgi, n=60 cells, p=6.385e-015, Mann-Whitney test). (j) Tandem treatment of Golgi-HaloRasAR with RMC6291, followed by BAY293 results in more prominent inhibition upon BAY293 treatment. (k) Comparison of ΔR_1st/ΔR_max values of Golgi-HaloRasAR sensors in response to tandem inhibitor treatment. (ΔR_1st/R_max = 64 ± 3% for BAY293 first, n=60 cells, ΔR_1st/R_max = 50 ± 17% for RMC6291 first, n=34 cells, p=1.428e-006, Mann-Whitney test). (l) Partial response of EKAR4-NES in MIA PaCa-2 cells to BAY293 compared to RMC6291 (left) Pairwise comparison of BAY293 and RMC6291 with separate treatments (n=96 cells for BAY293 treatment, n=120 cells for RMC6291 treatment, p=3.763e-008, Mann-Whitney test). (right) Pairwise comparison of BAY293 and RMC6291 with tandem treatment (n=120 cells for each group, p=1.204e-009, Mann-Whitney test). For time-series traces, solid lines indicate the mean and shaded areas represent s.e.m. All Mann-Whitney comparisons are two-sided.

DATA AVAILABILITY

All data supporting the findings of this study are available from the corresponding author upon request.

CODE AVAILABILITY

Custom Python analysis code used to analyze imaging data is available. https://github.com/rweeeeks/A-chemigenetic-endogenous-Ras-activity-biosensor-for-live-cell-pharmacology