Bioluminescence-based assays for quantifying endogenous protein interactions in live cells

Abstract

Protein–protein interactions (PPIs) are integral to cellular signaling networks and are frequently disrupted in cancer, neurodegeneration, inflammation, and metabolic disorders. Targeting dysregulated PPIs presents a promising strategy for the development of therapeutic compounds. However, traditional drug discovery platforms often rely on plasmid-driven overexpression models that fail to replicate the complexity and dynamics of PPI in native cellular contexts. This study aims to evaluate the use of NanoLuc Binary Technology (NanoBiT) and NanoLuc Bioluminescence Resonance Energy Transfer (NanoBRET) for quantifying interactions of endogenously regulated proteins in live cells. To achieve this, CRISPR-mediated genome engineering was used to integrate NanoBiT and NanoBRET fusion tags at the loci for EGFR/GRB2 and KRas/CRAF in DLD-1 and HCT 116 cell lines. Assays using the engineered cell lines were then conducted in monolayer cultures using endpoint and kinetic measurements, as well as luminescence imaging. The approach was further expanded to investigate PPI in cancer-associated isogenic cell lines and 3D spheroid models that better preserve additional aspects of cellular organization. Collectively, these findings establish a robust and modular workflow for generating endogenously regulated PPI reporter cell lines to improve the relevance and predictive power of live-cell assays. By capturing interaction dynamics in a more representative background, this approach offers a potentially valuable tool for elucidating signaling mechanisms and characterizing therapeutic compounds targeting PPIs.

Protein–protein interactions (PPIs) control essential biological processes that govern the life and death of a cell. These interactions are highly orchestrated events and essential for initiating, maintaining, and terminating specific signal transduction pathways (1, 2, 3). Even minor changes to proteins, such as amino acid substitutions or post-translational modifications, can disrupt normal PPI networks, leading to aberrant signaling and disease (4, 5). Single-residue mutations in EGFR and components of the Ras/RAF pathway, for example, can alter interaction profiles and drive oncogenesis in cancers such as non-small cell lung cancer, colorectal cancers, and pancreatic ductal adenocarcinoma (6, 7, 8, 9). Given the central role of dysregulated PPIs in diverse pathologies, considerable effort has been devoted to understanding these interactions and developing strategies to pharmacologically modulate them.

Recent conceptual and technical advances have reinvigorated drug discovery efforts targeting disease-associated PPIs. Rational drug design, aided by computational modeling and structure-activity relationship (SAR) analyses, has made previously undruggable targets viable therapeutic options (3, 10, 11). Further, targeted protein degradation strategies, such as proteolysis targeting-chimeras (PROTACs) and molecular glues, have emerged as promising alternatives to conventional small-molecule inhibitors (12, 13, 14). To fully leverage these innovations, there is a growing need for assay systems that can capture the influence of cellular context, such as subcellular localization, expression levels, regulatory feedback, and competing interactions, on PPI behavior.

The binding kinetics and thermodynamic parameters of PPIs can be studied using a variety of biochemical and biophysical techniques, including surface plasmon resonance, isothermal titration calorimetry, biolayer interferometry, and stopped-flow analysis (15, 16). While powerful, these techniques typically involve purified proteins or cell lysates and thus lack the spatial and regulatory context of living cells (17). In contrast, live-cell assays allow dynamic monitoring of PPIs within intact cellular environments and capture the influence of cellular localization, abundance, and regulatory signals (18, 19).

Among the most commonly employed approaches for studying PPI in live cells are biomolecular fluorescence complementation (BiFC) and resonance energy transfer (RET). Complementation-based methods, such as BiFC, work by attaching two non-fluorescent fragments of a reporter protein to the proteins of interest (POI). When the POI interacts, the reporter segments come together to form a functional, fluorescent reporter (20). These systems provide high sensitivity and are particularly useful for imaging subcellular localization. However, BiFC tends to suffer from slow fluorophore maturation, irreversibility, and increased background caused by non-specific fragment association (21, 22, 23). RET-based formats, including fluorescence resonance energy transfer (FRET) and bioluminescence resonance energy transfer (BRET), detect PPI via energy transfer between a donor and an acceptor fluorophore when fusion proteins of interest are in proximity (24, 25). In FRET, the donor is a fluorescent protein excited by external light, while in BRET, the donor is a luciferase that generates light in the presence of substrate. In both cases, the acceptor is either a fluorescent protein or a synthetic fluorophore that receives energy and emits a detectable signal. RET formats enable real-time monitoring of interaction kinetics and are well-suited for measuring both association and dissociation. However, these systems can be limited by low signal-to-noise ratios, spectral overlap, and the need to optimize expression levels and orientation of fusion constructs.

Two reporter systems based on the 19-kDa NanoLuc (Nluc) luciferase have been optimized to quantify PPI behavior in live cells with high sensitivity and broad dynamic range (26, 27, 28, 29). The first, NanoLuc Binary Technology (NanoBiT), is a reversible enzyme complementation system optimized for brightness and low affinity (30). It consists of the LgBiT polypeptide (18 kDa) and the SmBiT peptide (1.3 kDa) expressed as genetic fusions with the proteins of interest. When the target proteins interact, the reporter subunits reassemble into an active luciferase that emits luminescence in the presence of substrate. The second system, NanoLuc Bioluminescence Resonance Energy Transfer (NanoBRET), employs Nluc as an energy donor to a fluorescent energy acceptor (31, 32, 33). A wide range of fluorescent proteins and synthetic dyes have been used as BRET acceptors with Nluc in published studies, including RFP, YFP, and Oregon Green. In the NanoBRET configuration described here, Nluc is paired with a HaloTag (HT) protein labeled with a 618 nm fluorescent ligand.

Despite improvements in live-cell PPI technologies, many assays still rely upon plasmid-derived overexpression, which can produce artifacts due to artificial expression levels (20, 34, 35). Constitutive promoters may disrupt protein stoichiometry, alter mass-action dynamics, skew interaction kinetics, and interfere with normal cellular processes (36, 37). Genome editing technologies, however, have transformed cell biology by enabling the study of proteins at endogenous levels within their native cellular environment. CRISPR-mediated knock-in (KI) of fluorescent and luminescent reporters at target loci enables the generation of cell lines that express reporter-tagged protein under control of their natural regulatory elements (38, 39, 40, 41). Compared to ectopic systems, these models generally offer more accurate insights into protein activity and regulation (41). Previous studies have demonstrated the feasibility of fully endogenous NanoBiT assays by sequentially knocking in LgBiT and SmBiT tags to generate homozygous or heterozygous cell lines (42, 43).

The current study builds upon this foundation by establishing a generalized workflow for generating NanoBiT and NanoBRET PPI cell lines with reporter tags expressed under endogenous regulatory control. Notably, this work expands the application of these models beyond standard monolayer culture to include disease-relevant isogenic mutations and 3D growth conditions that more closely replicate features of the tumor microenvironment. Recent strategies to improve KI efficiency, such as chemical inhibition of the non-homologous end joining (NHEJ) and microhomology-mediated end joining (MMEJ) DNA repair pathways, are leveraged to facilitate efficient integration of Nluc and HT sequences (44, 45). Collectively, these efforts broaden the utility of Nluc-based technologies for studying protein interactions in the context of endogenous expression levels, signaling feedback, and complex cellular architecture.

Results/discussion

Generation of model cell lines for endogenous PPI

EGFR/GRB2 and KRas/CRAF interactions were selected as biological models for the evaluation of endogenous NanoBiT and NanoBRET PPI assays. These protein pairs are well-characterized and play roles in a shared signaling cascade responsible for regulating proliferation, differentiation, and metastasis. Given the role of EGFR, RAF, and Ras in oncogenesis, the colorectal cancer cell lines DLD-1 and HCT 116 were selected as parental lines to establish the endogenous PPI models. These primarily diploid cell lines harbor a glycine-to-aspartic acid mutation at amino acid 13 of KRas (KRas^G13D) and express wild-type BRAF and EGFR (46). To generate the endogenous PPI cell lines, sequential or simultaneous CRISPR-mediated editing was performed in the presence of inhibitors targeting DNA-dependent protein kinase (DNA-PK) and DNA polymerase theta (Polθ). These inhibitors combine to decrease activity of non-homologous end joining (NHEJ) and microhomology end-joining (MMEJ) pathways for double-strand break repair, favoring homology-directed repair (HDR) for tag integration. Editing was followed by single-cell sorting to isolate homozygous clones, with zygosity and sequence integrity confirmed by droplet digital PCR (ddPCR) and Sanger sequencing (Figs. S1 and S2) (45). For the EGFR/GRB2 model, DLD-1 cells were edited to express EGFR-LgBiT and GRB2-SmBiT (NanoBiT) or EGFR-Nluc and GRB2-HT (NanoBRET). For the KRas/CRAF model, HCT 116 cells were edited to express LgBiT-CRAF and KRas-SmBiT (NanoBiT) or Nluc-CRAF and KRas-HT (NanoBRET).

Evaluation of endogenous PPI assays using the EGFR/GRB2 model

The performance of the NanoBiT and NanoBRET cell lines was initially assessed in an EGFR/GRB2 induction model. Epidermal growth factor (EGF) binding to the extracellular domain of EGFR leads to receptor dimerization and phosphorylation of the cytosolic domain. Phosphorylated EGFR recruits the GRB2 adapter protein and triggers RAS/MAPK/ERK signaling. To determine if the NanoBRET and NanoBiT assays could accurately measure the EGFR/GRB2 interaction, cells were serum-starved and then stimulated with serial dilutions of EGF. For NanoBiT assays, luminescence in the presence of enzyme substrate is a direct readout for the PPI. For NanoBRET assays, the PPI is detected in the presence of Nluc substrate and fluorescent HT ligand and reported as the ratio of the liganded-HT acceptor signal to Nluc donor signal. Both NanoBRET and NanoBiT produced near-concordant EGF EC₅₀ potencies of 14 versus 11 ng/ml, respectively, but displayed differences in the magnitude of the response ratio (Fig. 1, A and B). In subsequent experiments, both PPI assay formats produced data sufficiently robust for high-throughput screening as evidenced by Z′ values of greater than 0.5 in both 96- and 384-well formats (Fig. S3) (47).

Figure 1.

EGF response magnitude and PPI kinetics.A, NanoBiT and B, NanoBRET EGFR/GRB2 cell lines were serum-starved for 4 h and then treated for 90 s (NanoBRET) or 120 s (NanoBiT) with the indicated concentration of EGF before acquiring NanoBRET or luminescence signals. Each point represents the average fold above the untreated control (n = 4, technical replicates). Data were fit using a variable slope (four-parameter) model in GraphPad Prism. EC₅₀ and Hill Slope (h) are shown on each graph. Kinetics of association and dissociation for (C) NanoBiT and (D) NanoBRET were measured immediately after the addition of 300 ng/ml EGF. All data are plotted as average luminescence or NanoBRET signal with background subtracted (n = 4, technical replicates). The vertical dashed line indicates the time of maximum signal. For all panels, error is represented as SD.

The generation of KI PPI clones through a sequential editing workflow can be both time-consuming and resource-intensive. However, incorporating DNA-repair inhibitors during the CRISPR editing process significantly enhances reporter KI efficiency while minimizing indel formation. This improvement opens the possibility of simultaneous editing and initial evaluation in pooled populations rather than individual clones. In the EGFR/GRB2 NanoBiT test case, CRISPR-edited pools demonstrated dose-dependent EGF responsiveness that was consistent with data from EGFR/GRB2 NanoBiT clones (Fig. S4). This suggests that simultaneous editing of pooled populations could offer a more streamlined workflow for identifying optimal protein-fusion tag orientations, which are typically determined by testing multiple combinations. Use of pools could also be beneficial in situations where time restrictions prohibit isolation of clones. However, pools also contain an unknown proportion of cells with single edits. As such, the NanoBRET format could exhibit a reduction in BRET ratio relative to clones as a result of Nluc incorporation without the HT acceptor. Finally, the use of extensively propagated pools necessitates further evaluation of KI durability and growth rate fitness relative to unedited parental populations to ensure that edited cells are still present.

After establishing EGF-dependent induction of the interaction between EGFR and GRB2 for both assay formats, association and dissociation kinetics were assessed. Temporal dynamics are a critical regulatory feature of PPI in general and have been demonstrated to regulate functional outcomes for the EGF pathway (48, 49). Our data revealed the canonical fast onset PPI kinetics after EGF stimulus, followed by a less transient but pronounced dissociation response (Fig. 1, C and D). Slight temporal differences in the initial responsiveness between the two reporter systems were observed. NanoBRET achieved maximum signal more rapidly than NanoBiT signal following stimulation (∼60 s and ∼90 s, respectively). However, both reporter systems were able to capture the slower rate of signal decay equally well in real time.

The observed differences in initial signal kinetics are likely caused by the biophysical differences between the NanoBRET and the NanoBiT systems. RET is not dependent on direct physical contact and is therefore capable of reporting changes in proximity without any delay. In contrast, all protein complementation reporters must undergo both structural and functional reconstitution, which imposes limits on the temporal resolution of signal generation. Although the NanoBiT system was optimized for rapid reconstitution, it might still exhibit a slight delay in signal generation, which could be detectable when measuring fast PPI kinetics (<1 min) (30). Nevertheless, the brief delay observed with NanoBiT represents a considerable improvement over alternative protein complementation reporter systems, such as BiFC, which require hours of maturation time (20).

As previously noted, PPI overexpression models are often valuable but may be encumbered by artifacts associated with altered mass balance, kinetics, and reproducibility. To assess whether the EGFR/GRB2 model was susceptible to these experimental limitations, our investigation focused on comparing an overexpressed NanoBRET system to a CRISPR-based endogenous system. In the overexpressed NanoBRET PPI model, dose-dependent BRET increases were observed after induction, showing strong concordance with the endogenous BRET model (Fig. S5A). However, in comparison to the endogenous NanoBRET assay, the overexpression format resulted in a compressed dynamic range with significant data variability owing to high uninduced background, differential cell-to-cell transfection efficiencies, and subsequent protein expression ratios. These confounding factors were notably absent with the CRISPR-edited system. These detriments limit the utility of overexpression systems for screening applications. Additionally, the overexpression format failed to capture the native EGFR desensitization dynamic following EGF stimulation. The rapid increase in the BRET ratio following EGF addition indicates inducible interaction between EGFR and GRB2. Contrary to the endogenous format, the overexpression format shows no change of signal following the initial signaling event which implies a stable EGFR/GRB2 complex (Fig. S5B). The lack of robustness, poor dynamic range, and altered interaction kinetics persisted at different EGFR/GRB2 expression plasmid ratios. These results align with previously described biological deviations in overexpression models stemming from altered protein stoichiometry and disruptions in protein internalization and recycling pathways (50).

Real-time bioluminescent imaging experiments were conducted to verify the proper spatial distribution of pre- and post-stimulated EGFR/GRB2 fusion protein pools. For the NanoBRET system, the donor signal (indicative of EGFR) was primarily observed at the cell surface in untreated cells and early after EGF exposure. However, prolonged treatment with EGF led to a punctate pattern across the cell, which is indicative of endosomal uptake of the receptor signaling complex (Fig. 2A). The acceptor channel reported the consequences of EGFR-Nluc and GRB2-HT association via energy transfer (Fig. 2B). As such, the basal level of EGFR/GRB2 association was noted at time zero, followed by a rapid increase in BRET signal (peak at 60 s). The acceptor signal declined (relative to the donor signal) over time, which suggests dissociation of GRB2 from EGFR. This was accompanied by the same punctate pattern as observed for the donor signal. These observations are consistent with published reports of EGF-induced EGFR/GRB2 association, dissociation, and internalization and mirror immunofluorescence results obtained using the unedited parental cell line (Fig. S6) (51, 52). The NanoBiT time course followed a similar kinetic and morphological distribution of the PPI but manifested itself as only relative intensity of bioluminescence during the initiating and internalization phases of EGF-induced EGFR-GRB2 interaction (Fig. S7).

Figure 2.

Bioluminescent imaging of the EGFR/GRB2 NanoBRET response.A, EGFR-Nluc donor signal and (B) GRB2-HT acceptor signal were captured on an LV200 bioluminescence imager. Images were captured 0, 66, and 957 s post-treatment with 300 ng/ml EGF. Basal association of EGFR/GRB2 is represented at 0 s, while recruitment of intracellular GRB2 to EGFR at the cell membrane surface and compartmentalization of the complex are shown at 66 and 957 s, respectively. Images were acquired using a 60×/1.35 NA objective with 460/80 bp and 590 lp filters. EM gain was set to 1200, and exposures were set to 3 s for the donor channel and 5 s for the acceptor channel. Images are a mean projection of three images. Scale bar indicates 100 μm.

Dysregulation of EGFR signaling is a well-established driver in oncogenesis, which makes EGFR a target for the development of therapeutics, including monoclonal antibodies and small-molecule receptor tyrosine kinase inhibitors (RTKis). To determine if the endogenous PPI assays could measure the effect of these modulators on EGFR/GRB2 interactions, the assays were performed in the presence of EGF and different RTKis. Following pre-treatment with serial dilutions of gefitinib (first-generation RTKi), osimertinib (third-generation RTKi), and cetuximab (therapeutic anti-EGFR mAb), cells were exposed to EGF. NanoBRET- and NanoBiT-based EGFR-GRB2 assays produced concordant dose-dependent inhibition profiles with all compounds (Fig. S8, A–C).

While it remains possible that endogenous tagging could alter protein expression or turnover, the data presented thus far align with previously published findings, suggesting that the tagging has not introduced detectable artifacts. To further validate the accuracy of the inhibition potencies observed with the NanoBiT CRISPR-engineered cell lines, an orthogonal method using the unedited DLD-1 parental cell line was employed. In this assay, the inhibition profile was measured with a phospho-EGFR (Y1068) immunoassay that combines monoclonal antibodies against both phospho-specific and non-phosphorylated epitopes of EGFR, along with NanoBiT-labeled detection antibodies, enabling quantification of the relative phosphorylation status (53). The results obtained with osimertinib demonstrated strong agreement in potency between the two approaches (EC₅₀ = 276 versus 216 nM) (Fig. S9). The alignment in both relative inhibition profiles and absolute potency supports the validity of the endogenous measurements. Moreover, the use of any immunoassay in the same parental background but via a distinct detection modality reinforces the robustness and accuracy of the NanoBiT cell lines.

Clinical RTKi treatment can produce dramatic responses and prolong disease-free progression for patients with non-small cell lung cancer (NSCLC). However, secondary point mutations in the ATP binding pocket of the EGFR tyrosine kinase domain may lead to selective resistance for first-generation RTKi (e.g. erlotinib and gefitinib) (54, 55). To demonstrate the ability of the cell lines to detect RTKi potency difference in the presence of disease-associated mutants, the T790M point mutation (EGFR^T790M) was engineered into the NanoBiT EGFR/GRB2 clone (Figs. S2 and S10). The impact of RTKi potency on clones with EGFR^T790M was dramatic. Erlotinib and gefitinib produced respective IC₅₀ values of 25 nM and 10 nM in the WT background but incomplete inhibition for the EGFR^T790M mutant cell line (Fig. 3, A and B). Conversely, osimertinib and WZ8040 demonstrated low potencies (516 nM and 512 nM) in the WT background but enhanced potencies (409 nM and 43 nM) with the EGFR^T790M mutant cell line. These potency observations are consistent with previous RTKi reports and have since been translated into enhanced clinical efficacy (56, 57, 58). Therefore, the generation of specific, disease-associated mutations from an existing EGFR/GRB2 reporter background via CRISPR appears to provide a straightforward means of modeling RTKi resistance while allowing for proof-of-concept sensitivity screening for next-generation inhibitors. Although demonstrated here with the EGFR/GRB2 interaction, this strategy of generating isogenic panels with disease-relevant mutations could, in principle, be applied to other endogenous PPI reporter cell lines to assess the impact of mutations on target biology and drug response.

Figure 3.

Differential potencies of RTKi in the presence of EGFR^WT and EGFR^T790M. RTKi were serially diluted and applied to either (A) EGFR^WT/GRB2 or (B) EGFR^T790M/GRB2 engineered cells for 1 h. 300 ng/ml EGF was added, and luminescence was measured in the presence of substrate. Plots represent the average luminescence (n = 4, technical replicates). Curves were fit with the sigmoidal dose-response (variable slope) model in GraphPad Prism. Error is represented as SD. For the EGFR^WT cells used in panel A, the following IC₅₀ values were obtained: Gefitinib (9.59 nM), Osimertinib (516.2 nM), WZ8040 (512.4 nM), and Erlotinib (25.3 nM). For the EGFR^T790M cells used in panel B, the following IC₅₀ values were obtained: Gefitinib (112.0 μM), Osimertinib (409.4 nM), WZ8040 (42.7 nM), and Erlotinib (436 μM).

Assessment of assay performance using the KRas/CRAF model

The interaction between KRas and CRAF, an event that occurs downstream of EGFR/GRB2 signaling, was selected as an additional model by which to assess the performance of endogenous NanoBRET and NanoBiT cell lines. Upon EGFR activation, GRB2 recruits the guanine nucleotide exchange factor SOS to the plasma membrane, where it catalyzes the exchange of GDP for GTP on KRas. The GTP-bound KRas then associates with RAF family members, including ARAF, BRAF, and CRAF, to initiate a signaling cascade that drives cellular proliferation, survival, and differentiation (59). In KRas-mutant cancers, this pathway becomes dysregulated, often rendering tumors resistant to EGFR-targeted therapies due to constitutive KRas activation (60). As a result, therapeutic strategies have shifted towards targeting KRas directly or inhibiting downstream effectors such as RAF, MEK, and ERK. However, the inhibition of these proteins presents challenges due to their complex structural features, heterogeneous mutational landscape, feedback loops, and interactions with other signaling pathways (61). One particular challenge, known as the “RAF paradox,” occurs when inhibitors designed to target mutant RAF proteins (e.g., BRAF^V600E) unexpectedly promote dimerization and activate downstream MEK/ERK signaling in cells. This paradoxical activation does not occur in RAF mutant cells themselves, where the proteins signal as monomers and are effectively inhibited by monomer-selective RAF inhibitors. Instead, it is observed in cells with wild-type RAF proteins, particularly when RAS is activated, either through oncogenic KRAS mutations or upstream receptor tyrosine kinase signaling. In these contexts, RAF inhibitors can stabilize RAF dimers, inadvertently enhancing MAPK signaling (62).

To explore this effect and quantify endogenous KRas/CRAF interactions, HCT 116 cells harboring a heterozygous KRas^G13D mutation were engineered to express either LgBiT-CRAF/SmBiT-KRas (NanoBiT) or NL-CRAF/HT-KRas (NanoBRET). KRas^G13D is a constitutively active variant with impaired GTP hydrolysis that promotes persistent engagement with RAF and renders cells susceptible to paradoxical signaling. The engineered cell lines were treated with RAF inhibitors known to induce Ras/RAF interactions, including GDC0879, LY3009120, AZ628, and Dabrafenib (63). Dose-dependent increases in signal were observed following 4 h treatment, consistent with drug-induced stabilization of KRas/CRAF complexes (Fig. 4, A and B). Although the calculated potencies vary among different cell lines due to genetic variations, these values show reasonable concordance with previous reports in the literature (64). Cells were also treated with the novel BRAF inhibitor PLX8394, a “paradox breaker” capable of inhibiting BRAF without activating the Ras/RAF signaling pathway (65). Consistent with previous studies, no KRas/CRAF dimerization was detected following PLX8394 treatment, confirming its distinct mechanism of action (66).

Figure 4.

Measuring inhibitor-induced KRas/CRAF dimerization.A, NanoBRET and B, NanoBiT assays measuring CRAF dimerization with KRas following 4 h treatment with various inhibitors. Data represent the average NanoBRET signal (n = 6) or NanoBiT signal (n = 8) and were fit using an asymmetric (five parameter) dose response curve in GraphPad Prism. For NanoBRET, the following inhibitor EC₅₀ values were obtained: AZD7648 (0.0815 μM), Dabrafenib (0.731 μM), GDC0879 (1.016 μM), LY3009120 (0.166 μM), and PLX8349 (not calculated). For NanoBiT, the following EC₅₀ values were obtained: AZD7648 (0.257 μM), Dabrafenib (2.183 μM), GDC0879 (1.476 μM), LY3009120 (0.246 μM), and PLX8349 (not calculated). C, NanoBRET and (D) NanoBiT assays in which cells were treated with LY3009120 (1 μM) or GDC0879 (3 μM). Data represent average NanoBRET signal (n = 4) or NanoBiT signal (n = 5). E, NanoBRET and F, NanoBiT assays in which cells were pre-treated with LY3009120 (1 μM) or GDC0879 (3 μM) for 4 h followed by inhibitor washout by media exchange and measurement of signal for an additional 3.5 h. Data represent the remaining average BRET (n = 4) or luminescent signal (n = 4) relative to no-washout control. In all panels, error is displayed as SD and each data point represents average of technical replicates.

The assays were subsequently conducted in a kinetic format to capture the real-time association of CRAF and KRas (Fig. 4, C and D). To maintain a stable signal over an extended duration, a protected substrate that is steadily released by cellular esterases (Vivazine) was used. Following pre-treatment with the protected substrate, either GDC0879 (10 μM) or LY3009120 (3.3 μM) was added to the edited cell lines, and signal was measured every 2.4 min for 3.5 h. In the case of LY3009120, both NanoBRET and NanoBiT cell lines registered comparable association rates throughout the time course. For GD0879, both cell lines reported similar initial association rates. However, the NanoBRET signal plateaued after 60 min while the NanoBiT signal continued to rise. It is unclear why the discrepancy exists, but it is likely an assay artifact, as applying the unprotected NanoBRET Substrate at set intervals in a non-continuous format generated an association curve similar to that of NanoBiT (Fig. S11).

To demonstrate that these reporter cell lines could also monitor dissociation events, the NanoBRET and NanoBiT cell lines were pre-treated with GDC0879 or LY3009120 to induce the KRas/CRAF interaction. After treatment, a media exchange was performed to wash out the compounds and initiate KRas/CRAF dissociation. Both assay formats showed rapid and nearly complete dissociation of the target proteins initially exposed to GDC0879 (Fig. 4, E and F). In contrast, LY3009120-treated cells showed minimal dissociation within the same time period. These results align with the distinct mechanistic profiles of the two RAF inhibitor classes to which these compounds belong: Type I (GDC0819) and Type II (LY3009120). GDC0879, a Type I inhibitor, binds the RAF kinase domain in an active-like DFG-in/helix αC-in conformation, while LY3009120, a Type II inhibitor, binds in an inactive-like DFG-out/helix αC-out mode. Type II inhibitors, such as LY3009120, typically display slower off-rates, sometimes exceeding 10 h, due to extensive molecular interactions (67). These findings illustrate that NanoBRET and NanoBiT assays effectively report the dynamics of KRas/CRAF interactions and distinguish between the potencies and kinetics of various inhibitors.

Measuring CRAF/KRas interaction in isogenic NanoBiT cell lines

In the present study, the NanoBiT CRAF/KRas PPI assay was designed with LgBiT attached to CRAF by a15 amino acid linker and SmBiT attached to KRas with a smaller four amino acid linker. This raised the question whether the four amino acid linker was of sufficient length to effectively complement the NanoBiT reporter tags during CRAF/KRas interaction. To evaluate if a longer linker would enhance signal amplitude, the KRas/CRAF NanoBiT cells were engineered with extended linker lengths of 10 or 15 amino acids between SmBiT and KRas (Fig. S12). Cells that were homozygous for integration of these linkers and that retained the KRas^G13D/WT heterozygosity of HCT 116 cells were expanded for further analysis. These cells were treated with GDC0879 to assess any enhancement in assay signal due to the extended linkers. The longer linkers did not significantly alter the overall signal-to-background ratio (Fig. S13), suggesting that for this PPI, the four amino acid linker was sufficient for efficient complementation.

While validating the sequence of the isolated NanoBiT clones, it was observed that a fraction of the isolated clones contained two copies of the G13D mutation (KRas^G13D/G13D), rather than maintaining the KRas^G13D/WT heterozygosity of HCT 116 cells (Fig. S14). This was not entirely surprising given that the CRISPR donor DNA template coded for the G13D mutation in the distal homology arm (Fig. S12). An additional subset of clones was devoid of KRAS^G13D but had one copy of KRAS^WT. In this scenario, it is possible that the inherent instability of the KRas allele, combined with the Cas9-mediated cut, contributed to the loss of the KRAS^G13D allele. In otherwise genetically identical cells, it is intriguing to consider the impact of these single amino acid mutations on cellular physiology. To assess whether NanoBiT could detect differential inhibitor responses in cells with different KRas mutational states, the newly generated cell lines that were homozygous for G13D (KRas^G13D/G13D), heterozygous for G13D (KRas^G13D/G13WT), or strictly wild-type at G13 with one allele lost (KRas^G13WT/-) were analyzed for responsiveness to the natural pathway stimulant, EGF. Among these, the KRas^G13D/G13D cell line exhibited the highest uninduced basal signal, suggesting that the activating mutation was driving KRas/CRAF interaction even in the absence of EGF (Fig. 5, A and B). As expected, the KRas^G13WT/- cell line had the lowest basal signal, likely reflecting that KRas was predominantly in an inactive state without inhibitor. The basal signal in the KRas^G13D/G13WT cell line was intermediate between the fully mutated and wild-type cell lines. When stimulated with EGF, the largest response was observed in the KRas^G13WT/- line, while the lowest response was observed in the KRas^{G13D/G13D line}. This is consistent with previous reports that cells expressing KRas with activating mutations are uncoupled from EGFR and therefore less responsive to stimulation with EGF compared to cells expressing wild-type KRas (68). Use of isogenic cell lines, such as those generated here, could aid in the development of modulators of Ras/RAF signaling that elicit a desired response in the presence of specific genetic mutations.

Figure 5.

Analyzing effect of isogenic cell lines on endogenous KRas/CRAF dynamics.A, uninduced signal of NanoBiT cells re-engineered to contain the 10-amino acid linker and 0, 1, or two copies of the KRas G13D. Bars represent average luminescence (n = 3) with each replicate indicated as a closed circle. B, NanoBiT cells containing the 10-amino acid linker and either 0, 1, or two copies of the KRas G13D mutation were treated for 15 min with increasing concentrations of rhEGF. Data represent signal/background (n = 3) and were fit with an asymmetric (five parameter) dose response curve in GraphPad Prism. EC₅₀ values are as follows: KRas^WT/- (7.304 ng/ml), KRas^G13D/WT (3.146 ng/ml), and KRas^G13D/G13D (1.326 ng/ml). For all panels, error is represented as SD of technical replicates.

Measuring PPI in 3D models

In addition to introducing point mutations to generate cell lines with disease-specific mutations, another approach to improving model accuracy is the use of 3D cultures. This method more closely replicates the complex environment of living tissues, offering insights that traditional 2D cultures cannot provide (69, 70, 71). However, 3D cultures can be technically challenging to implement. Transient overexpression is insufficient due to the short half-life and the artificial levels of the fusion proteins, while traditional stable cell lines often lack control over expression, potentially disrupting cellular biology. CRISPR-integrated reporters overcome these limitations by being stably integrated and controlled by cellular regulatory elements. As such, the engineered cells developed in this study provide a path toward creating 3D models for investigating PPI. Among the assay platforms described, the ratiometric nature of BRET assays is particularly well-suited for 3D systems, as it minimizes potential confounding effects from light absorption and scattering inherent to intensity-based formats. To demonstrate the feasibility of using edited cells for measuring PPI in 3D, KRas/CRAF NanoBRET HCT 116 cells were grown in ultra-low attachment plates to form spheroids. Following treatment with RAF inhibitors, dose-dependent increases in NanoBRET signal were observed (Fig. 6A). Supporting these plate assay results, imaging of the spheroids showed increased BRET signal after compound treatment. While bioluminescence imaging could not resolve individual cells in the spheroids, monolayer imaging confirmed both increased BRET and proper membrane localization of the interaction (Figs. 6B and S15). While further work is needed to refine the technical aspects of 3D endogenous PPI assays, this study establishes a foundation for plate-based formats that better capture the complexities of cellular environments, including critical cell-cell and cell-matrix interactions essential for signaling, differentiation, and response to external cues.

Figure 6.

Detecting KRas Interaction with CRAF in Spheroids with NanoBRET.A, spheroids generated from HCT116 cells endogenously expressing Nluc-CRAF and dimerizing with HT-KRas following 4 h treatment with various inhibitors. Data represent the average NanoBRET signal (n = 6, technical replicates) with variability expressed as SD. EC₅₀ values are shown in legend. B, panels showing pseudo-colored images of the donor (cyan) and acceptor (orange) channel of individual spheroids that were untreated or treated for 4 h with LY3009120 (1 μM). Imaging parameters: 40×/0.9 NA objective, EM600, 3 s at exposure for both channels (donor: 460/80BP, acceptor: 590LP). Images are a mean projection of 10 images. Scale bar = 100 μm.

Conclusion

This study highlights the utility of CRISPR-based genome editing to integrate NanoBiT and NanoBRET reporters into cell lines, enabling the analysis of PPI under cellular regulatory conditions. The interactions between EGFR/GRB2 and KRas/CRAF were selected to establish this framework, as these signaling pathways are implicated in disease and represent critical therapeutic targets. Importantly, this work also demonstrates how standard monolayer assays can serve as a foundation for more advanced models, including 3D cultures and mutational panels with an isogenic background.

These extended endogenous PPI formats may provide additional biological insight by capturing aspects of cellular organization and signaling dynamics not achievable in traditional 2D systems. For example, 3D cultures, such as spheroids, replicate features of tissue architecture, including spatial organization and microenvironmental gradients (70, 71). These models support the study of drug diffusion, penetration, and signaling dynamics within a multicellular context, capturing features that are often absent in 2D cultures (72). Isogenic cell lines further strengthen this platform by enabling precise studies of disease-associated mutations in a consistent genetic background. For some applications, the generation of KI cell lines may be more time-consuming and resource-intensive than overexpression models, and the advanced models may necessitate the need for specialized instrumentation. However, these limitations are offset by the enhanced reproducibility, dynamic response, and contextual accuracy provided by this approach.

Collectively, these advancements establish a flexible and informative platform for exploring PPI and provide a foundation for developing next-generation cellular models to support therapeutic drug discovery. Coupled with advances in AI-driven compound screening and predictive modeling, the platform has the potential to accelerate the identification and characterization of therapeutics targeting complex cellular networks.

Experimental procedures

Cell lines and growth conditions

HCT 116 (ATCC CCL-247) and DLD-1 (ATCC CCL-221) cell lines were maintained in McCoy’s 5A (Modified) Medium (Gibco) supplemented with 10% fetal bovine serum (Avantor Seradigm) at 37°C and 5% CO₂. These human cell lines were purchased directly from ATCC and STR validated. CRISPR-modified cell lines were also expanded and maintained in the same growth medium. Opti-MEM I Reduced Serum Medium or CO₂-Independent Medium (Gibco) was used for pre-assay serum starvation, compound treatment, and substrate dilutions.

Compounds

The DNA-PK inhibitor AZD7648 was purchased from Selleckchem, and the Polϑ inhibitor PolQi3 was provided by AstraZeneca. The EGFR inhibitors gefitinib, erlotinib, osimertinib, WZ8040, and cetuximab were acquired from Selleckchem. Small-molecule modulators of KRas/CRAF interactions (AZ628, LY3009120, GDC-0879, Dabrafenib, and PLX8394) were purchased from Selleckchem. The recombinant human EGF (rhEGF) used was purchased from R&D Systems.

CRISPR-mediated genome editing

Alt-R S.p. Cas9 Nuclease V3, Alt-R CRISPR RNA (crRNA), Alt-R CRISPR-Cas9 tracrRNA, Nuclease-Free Duplex Buffer, Ultramer DNA Oligonucleotides (single-stranded DNA donors, ssDNAs), and pUCIDT Amp cloning vectors (plasmid DNA donors) were purchased from Integrated DNA Technologies (IDT). crRNA and donor DNA sequences can be found in the Supporting Data File. crRNAs were designed for the cut site to be within 15 nucleotides of the start (N-terminal fusions) or stop (C-terminal fusions) codons. Plasmid donors were designed with 400 nucleotide homology arms on each side of the tag sequence, and ssDNA donors were designed with 50 nucleotide homology arms on each side of the tag sequence. Guide RNA (gRNA) was prepared by incubating 1 nmol crRNA with 1 nmol tracrRNA in Nuclease-Free Duplex Buffer (50 μl final volume) for 5 min at 95 °C, cooled to 20 °C, and stored at −20 °C. To enhance integration efficiency, cells were pre-treated with 1 μM AZD7648 and 0.3 μM of PolQi3 in growth medium for 24 h before editing. To edit the cells, ribonucleoprotein (RNP) complexes were assembled by incubating 100 pmol Cas9 and 120 pmol gRNA for 10 min at ambient temperature in a total volume of 10 μl Nuclease-Free Duplex Buffer. 1 to 2 x 10⁶ cells were resuspended in 100 μl Ingenio Electroporation Solution (Mirus), and the RNP was added to cells in the presence of either 300 pmol ssDNA or 4 μg plasmid donor. The entire mixture was transferred to a 0.2 cm electroporation cuvette (Mirus) and electroporated at 150 V with the Ingenio EZporator (Mirus). Electroporated cells were returned to growth medium containing 1 μM AZD7648 and 0.3 μM of PolQi3 for 24 h. Medium was then replaced with growth medium without inhibitors, and the edited pools were allowed to recover an additional 3 to 5 days before isolating clones. Unless noted, cells were sequentially edited for integration of each tag, with the second round of editing occurring only on sequence-validated homozygous clones. To generate the NanoBiT EGFR^T790M cell line, the homozygous EGFR-LgBiT/GRB2-SmBiT DLD-1 clonal cell line underwent an additional round of editing using a donor template coding the mutation, followed by clonal isolation, Sanger sequencing, and ddPCR validation with probes that recognized either EGFR^WT or EGFR^T790M (73).

Clone isolation

CRISPR-edited pools were single-cell sorted into 96-well plates (Corning) using a Wolf Cell Sorter and G1 Single-Cell Dispenser (NanoCellect Biomedical). Colonies were established over 2 to 3 weeks and then replica plated. For NanoBRET cell lines, Nluc clones were first isolated. To identify Nluc expressing clones, an equal volume of Nano-Glo Luciferase Assay Substrate (diluted 1:50 in Opti-MEM I) was added to each well of the replicate plate, and luminescence was measured with a GloMax Discover multi-mode plate reader. Clones with more than 100-fold luminescence over background were expanded. Genomic DNA was isolated from the clones and then validated by ddPCR and Sanger sequencing. Homozygous clones with perfect sequence at the integration site were further CRISPR-edited for HT integration, followed by a second round of single-cell sorting and validation via ddPCR and Sanger sequencing. For NanoBiT cell lines, LgBiT positive clones were first isolated by single-cell sorting. To identify LgBiT expressing clones, HiBiT control peptide was diluted to 400 nM (2×) in Nano-Glo HiBiT Lytic Buffer and added to the cells. Luminescence was measured on a GloMax Discover, and clones with at least 100-fold luminescence above background were subjected to ddPCR and Sanger sequencing. Homozygous LgBiT clones with perfect sequence around the integration site were then edited for SmBiT integration, single-cell sorted and validated for SmBiT using ddPCR and Sanger sequencing.

ddPCR

All primers and probes for ddPCR were synthesized by IDT (Supporting Data File). Genomic DNA from all clones was isolated from 5 × 10⁴ to 5 × 10⁵ cells using the Maxwell RSC Culture Cells DNA Kit and the Maxwell RSC 48 Instrument (Promega) according to the manufacturer’s protocol. DNA was quantitated on the NanoDrop 8000 Spectrophotometer (ThermoFisher Scientific) and diluted to 2 to 10 ng/μl in nuclease-free water (Promega).

ddPCR reactions were prepared with 1× ddPCR Supermix for Probes (no dUTP, Bio-Rad), 900 nM of each primer, 250 nM of each probe, 2 to 10 ng of genomic DNA, and nuclease-free water to a final volume of 20 μl. Reactions were run on a QX ONE ddPCR System (Bio-Rad) with the following cycling conditions: 25 °C for 3 min, 95°C for 10 min, 40 cycles of 94 °C for 30 s and 60 °C for 1 min, 98 °C for 10 min, and 25 °C for 1 min. The complete set of reactions run is listed in Fig. S2. All reactions were multiplexed for the target gene and the RPPH1 reference gene. Zygosity was calculated as the number of positive droplets for the target reaction divided by the positive droplets for the RPPH1 reference reaction. For the mostly diploid DLD-1 and HCT 116 cell lines, a ratio of 1.0 indicated homozygous KI, a ratio of 0.5 indicated heterozygous KI, and 0.0 indicated no KI. An additional ddPCR reaction was run on samples generated using plasmid donors to screen for the backbone of a randomly integrated plasmid. For this purpose, reactions were designed to detect the ampicillin (Amp) resistance gene of the plasmid donor. Only clones that lacked the Amp resistance gene were selected for expansion.

Sanger sequencing

All primers were synthesized by IDT (Supporting Data File). Sequencing was performed using the same genomic DNA that was prepared for ddPCR. For sequencing of SmBiT fusions, primers were designed to amplify across the entire region of the donor template. For HT, Nluc, and LgBiT, primers were designed to amplify the region outside of the N- or C-terminal homology arm into the tag. To generate amplicons for sequencing, 40 ng of genomic DNA was combined with 500 nM of each primer, Q5 High-Fidelity 2× Master Mix (New England BioLabs), and nuclease-free water to a final volume of 50 μl. Amplification reactions were performed with a Veriti Thermal Cycler (Applied Biosystems) and the cycling conditions recommended for Q5 High-Fidelity 2× Master Mix: initial denaturation of 98°C for 30 s, 35 cycles (98 °C for 10 s, 50–72 °C for 20 s, 72 °C for 10 s), final extension of 72 °C for 2 min, and hold at 4 °C. Amplification reactions were then sequenced by Functional Biosciences.

EGFR activation and inhibitor studies

Clones that were homozygous for tag integration at both the EGFR and GRB2 loci were harvested by trypsinization, washed, and replated at a density of 20,000 cells/well in solid white 96-well plates and allowed to attach in complete McCoy’s 5A medium overnight. For EGF stimulation studies, the medium was removed, wells washed with DPBS, and replaced with a half volume of either Opti-MEM I (NanoBiT) or Opti-MEM I containing NanoBRET HaloTag 618 ligand (NanoBRET) for a period of 4 h. Five min prior to rhEGF stimulation, a 4× volume of NanoBRET Nano-Glo Substrate diluted in Opti-MEM I was delivered to each assay well and allowed to equilibrate with the cells. rhEGF was diluted in Opti-MEM I and added in the final one-quarter volume. Luminescent or luminescent donor and acceptor ratios (BRET) were measured using either a GloMax Discover or BMG CLARIOstar.

For inhibitor studies, the medium exchange, serum starvation, and HaloTag ligand labeling (NanoBRET only) were achieved as described above, but in a quarter volume. One hour prior to assaying, 4× serial dilutions of EGFR tyrosine kinase inhibitors were added in a second one quarter volume. Again, 5 min prior to assay a third one quarter volume of 4× NanoBRET Nano-Glo Substrate was delivered to the wells. Finally, rhEGF was delivered to each well for a final concentration of 100 ng/ml. Luminescence or NanoBRET ratios were measured as above.

Orthogonal pathway inhibition studies were conducted using the Lumit p-EGFR (Y1068) Immunoassay as a measure of EGFR activation. Unedited parental DLD-1 cells were compared to the EGFR-LgBiT/GRB2-SmBiT cell line. Parental cells were serum-starved and exposed to osimertinib as described above. Briefly, the cells were lysed 5 min after rhEGF stimulus, then exposed to an antibody solution containing total and phospho-EGFR mAbs, as well as an anti-mouse LgBiT and anti-rabbit labeled secondary antibodies. After incubation, the NanoBRET Nano-Glo Substrate was added and luminescence was measured.

Bioluminescent imaging

The DLD-1 cell line, edited to express EGFR-Nluc and GRB2-HT fusions, was seeded into 8-chamber borosilicate glass slides at a density of 40,000 cells per chamber in McCoy’s 5A medium with 10% FBS. After 16 h of incubation, the medium was removed from each chamber and exchanged with DPBS. DPBS was aspirated and replaced with a half volume of Opti-MEM I containing the NanoBRET HaloTag 618 Ligand. After 4 h, a quarter volume of 4× NanoBRET Nano-Glo Substrate was added. Brightfield and bioluminescent acquisition parameters (donor and acceptor channels) were established for the Olympus LV200 at 40× magnification. Last, a one-quarter volume of rhEGF (100 ng/ml final concentration) was added and images collected in real-time.

KRas/CRAF NanoBiT and NanoBRET assays

NanoBRET or NanoBiT KRas/CRAF clones were plated at 2 × 10⁴ cells/well in 96-well solid white microplates and incubated overnight. For NanoBiT endpoint assays, cells were treated with inhibitor dilutions in Opti-MEM I for 4 h. A 5× Nano-Glo Live Cell Substrate was prepared (1:20 dilution in Opti-MEM I) and added to the cells. Plates were shaken at 300 rpm for 5 min at room temperature before measuring luminescence using a GloMax Discover.

For NanoBRET endpoint assays, inhibitors were diluted in Opti-MEM I containing HT NanoBRET 618 Ligand (1:1000). After washing with DPBS, cells were treated with inhibitors for 4 h. A 5× NanoBRET Nano-Glo Substrate (1:100 dilution in Opti-MEM I) was added, and plates were shaken at 300 rpm for 5 min at ambient temperature. Donor luminescence (450 nm bp filter) and acceptor fluorescence (600 nm lp filter) were measured using a GloMax Discover.

For NanoBiT kinetic association assays, cells were washed with DPBS, then pre-incubated for 2.5 h with Nano-Glo Vivazine Substrate (1:50 dilution in CO₂-Independent Medium). Inhibitors (2× concentration) were added, and luminescence was measured every 2.4 min at 37 °C. For NanoBiT dissociation, cells were pre-treated with inhibitors for 4 h, washed, and incubated with fresh CO₂-Independent Medium containing Vivazine Substrate and either inhibitor or DMSO control. Luminescence then was measured every 2.2 min at 37 °C for 3.5 h.

For NanoBRET kinetic association assays, cells were pre-incubated with CO₂-Independent Medium containing Nano-Glo Vivazine Substrate (1:50) and HaloTag NanoBRET 618 Ligand (1:1000). Inhibitors were added, and donor luminescence and acceptor fluorescence were measured every 2.4 min at 37 °C. For NanoBRET dissociation, cells were pre-treated with inhibitors, washed, and incubated with fresh CO2-Independent Medium containing Vivazine Substrate, HaloTag NanoBRET 618 Ligand, and either inhibitor or DMSO control. Donor luminescence and acceptor fluorescence were measured as described above.

Data availability

The data that support this study are available from the corresponding author upon request.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Karin Musier-Forsyth