OregonFluor enables quantitative intracellular paired agent imaging to assess drug target availability in live cells and tissues

Abstract

Non-destructive fluorophore diffusion across cell membranes to provide an unbiased fluorescence intensity read-out is critical for quantitative imaging applications in live cells and tissues. Commercially available small molecule fluorophores have been engineered for biological compatibility, imparting high water solubility by modifying rhodamine and cyanine dye scaffolds with multiple sulfonate groups. The resulting net negative charge, however, often renders these fluorophores cell membrane-impermeant. Herein, we report the design and development of our biologically compatible, water-soluble, and cell membrane-permeable fluorophores, termed OregonFluor (ORFluor). By adapting previously established ratiometric imaging methodology using bio-affinity agents, it is now possible to use small molecule ORFluor-labeled therapeutic inhibitors to quantitatively visualize their intracellular distribution and protein target-specific binding, providing a chemical tool kit for quantifying drug target availability in live cells and tissues.

Introduction

Intracellular drug target validation and quantification of drug target engagement are a continuous challenge to the development of successful therapeutics due to some degree of non-specific accumulation and off-target effects. This is owed to variable drug affinity, biodistribution, pharmacokinetics and metabolism. Importantly, quantification of drug distribution and binding necessitates accounting for both the drug that binds to its target as well as the drug that accumulates in the cells and tissues in a non-specific manner.¹ While fluorescence imaging has long been used to visualize biological specimens, recent years have seen significant advances in the development of synthetic dyes,² genetically encoded biosensors,³ and fluorescent nanomaterials.⁴ Indeed, fluorescent proteins (FPs) have become common imaging tools for investigating cellular structure and function, as well as the underlying mechanisms of subcellular events.^5–7 Many FPs present intrinsic sensitivities to intracellular conditions such as pH and ion concentration, which can enable environmental sensing, but hinder quantitative fluorescence imaging.^{8, 9} Moreover, FP are relatively large tags with the smallest FP ca. 17 kDa,¹⁰ making them unsuitable as fluorescent reporters to image either exogenous (e.g., drugs, glycans, etc.) or endogenous substrates (e.g., hormones, neurotransmitters, etc.). Fluorescent nanomaterials also suffer from similar size constraints for labeling small molecular sized targets.

Due to their substantially smaller size (i.e., ca. 0.5–1.5 kDa), organic small-molecule fluorophores are routinely used as labels for optical imaging, with applications spanning from single-molecule localization microscopy to whole-body in vivo imaging.^11–16 Among the small-molecule fluorophores, rhodamines have long stood out as a promising class of probes, largely owing to their favorable photophysical properties, including high molar extinction coefficient and fluorescence quantum yield, as well as photo- and chemical stability. However, like many other xanthene-based fluorophores, classic rhodamines generally have poor water solubility leading to aggregation in aqueous environments and limiting biological compatibility. Bioconjugation to proteins and peptides (i.e., antibodies, nanobodies, affibodies, etc.) is typically carried out in an aqueous environment and in vivo imaging of fluorophores often requires administration of a substantially higher concentration than in vitro applications, highlighting the importance of water-solubility. Over the past two decades, significant efforts to address the xanthene solubility issue have been put forth. One successful chemical engineering approach for enhancing solubility involves the addition of sulfonate groups, affording an array of popular commercially available small molecule fluorophores, such as the ATTO and Alexa Fluor dyes series.^17–19 While modification with sulfonate groups has resulted in water-soluble fluorophores, these probes are often cell membrane-impermeant due to their net negative charge (Supplementary Fig. 1), limiting their application to extracellular target imaging.

Additionally, classic rhodamine dyes (e.g., tetramethylrhodamine [TMR], Silicon-substituted TMR [SiR]) exhibit a fluorescence on/off switching equilibrium as a result of an intramolecular cyclization reaction between the central bridging carbon of the chromophore and the carboxylate group attached at the ortho-position of the pendant phenyl ring (Fig. 1a). This reversible spirocyclic on/off mechanism within the rhodamine scaffold has been widely adapted as a “smart” probe design strategy in developing dyes for imaging and sensing biologically relevant analytes (e.g., ions, pH, biomolecule, metabolite, etc.).^{2, 20} Notably, several groups have developed a series of fluorogenic net-neutral fluorophores that effectively allow for modulation of the spirocyclic open/close equilibrium constants, providing improved overall performance for live-cell imaging and single-molecule microscopy applications.^21–25 However, in the context of quantitative fluorescence imaging applications, such as drug uptake and distribution monitoring in cells and tissues, this fluorescence on/off switching equilibrium is particularly undesirable. Quantitative fluorescence imaging becomes especially problematic when the target of interest is bound to the fluorophore in the fluorescence-off state (i.e., xanthene fluorophore in the spirolactone form, Fig. 1a), or to a fluorophore that has high affinity towards peripheral intracellular organelles (Fig. 1b), which could lead to a false fluorescence signal read-out due to environmental effects on the fluorescence activity. Fluorophores that are water-soluble, cell membrane-permeable, and insensitive to the environment are therefore, ideal dye labels for quantitative live-cell imaging applications (Fig. 1c).

Fig. 1. Design strategies to develop water-soluble and cell membrane-permeant fluorophores for quantitative, live-cell and tissue imaging.

Schematic illustration of ORFluor structure design. a Fluorogenic silicon-substituted rhodamine (SiR) exhibits zwitterion-spirolactone equilibrium, resulting in a suboptimal dye label for drug biodistribution monitoring. b SiR-Me is a fluorescence always-on derivative of SiR, which has shown increased affinity towards intracellular organelles, resulting in misleading, off-target fluorescent labeling compared to SiR when selected as the fluorescent reporter. The blue star represents the intended biodistribution of the drug after labeling, while red star represents altered biodistribution due to the inherently high affinity for mitochondria from SiR-Me. c Structure of ORFluor as the superior fluorophore for drug labeling, shows balanced charge with improved solubility suitable for quantitative imaging in live cells. d Structures of ORFluor dyes OF650, OF550 and their rhodamine analogs SiR and TMR. e Normalized absorption (left) and fluorescence emission (right) spectra of selected dyes (SiR, OF650, TMR, and OF550). f Tabulated spectral properties of ORFluors in comparison to the TMR- and SiR-based dyes in aqueous media. g Water solubility shows ORFluor dyes OF650 (left) and OF550 (right) both exhibit improved solubility compared to their base fluorophores SiR and TMR, respectively. Each dot represents an independent experiment; n=3; data are presented as mean ± standard deviation and are marked with asterisks for statistical significance between fluorophores, where significance was evaluated using an unpaired, two-tailed, parametric t test. The confidence level was 95% for both data sets. P-values are denoted by the number of asterisks as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. All statistical analyses were completed using GraphPad Prism 7.0 or 9.0. h Fluorophore inertness comparison in response to environmental changes, where screening factors including aqueous solutions with varied polarities, pH values, viscosities, temperatures and solvent types were evaluated using measured absorbance values. Each dot represents the means of n=3 independent experimental values under each screening condition. Detailed absorbance and fluorescence spectral comparison under each tested condition are shown in Supplementary Fig. 9.

In response, we report the design and synthesis of our fully profiled water-soluble and cell membrane-permeant fluorophore pair as superior labels to the classic SiR and TMR for quantitative imaging applications (Fig. 1c and Supplementary Fig. 2). Termed OregonFluor (ORFluor), our dyes have shown substantially improved water-solubility and stability against environmental changes compared to the base fluorophores SiR and TMR (Fig. 1d). To illustrate their utility, we have developed fluorescently labeled, paired, targeted and untargeted tyrosine kinase inhibitors to assess drug target availability (DTA) for the epidermal growth factor receptor (EGFR) as proof-of-concept. This approach utilizes a dynamic ratiometric imaging platform for estimating the concentration of cell surface receptors, termed Paired Agent Imaging (PAI). PAI was originally developed for quantitative in vivo imaging, where non-specific accumulation of protein-based bio-affinity reagents (e.g., antibodies) was discovered to dominate malignant tissue signals.²⁶ The PAI approach hypothesized that such non-specific distribution of a molecular-targeted imaging agent could be corrected for by normalizing its signal to the signal of a second co-injected inactive (untargeted) control of the targeted imaging agent, thereby accounting for any non-specific uptake. The targeted and untargeted (i.e., inactive control) agent pair can be labeled with either different isotopes or spectrally-distinct fluorescent dyes (i.e., ORFluors), allowing quantification of extracellular membrane receptors.^27–32 Extending PAI to intracellular protein targets, we demonstrate that intracellular PAI (iPAI) can be used to generate maps of bound and unbound inhibitors at the single-cell level both in vitro and ex vivo, highlighting the utility of our ORFluor pair as a chemical tool for quantitative imaging in the physiological context of intact cells and tissues.

Results and discussion

Water-soluble and cell-permeant ORFluors.

The century-old xanthene-based rhodamine fluorophores continue to receive broad attention from chemists and biologists alike owing to their favorable photophysical characteristics. Introducing ionizable groups (e.g., sulfonic acid, phosphoric acid, carboxylic acid, hydroxyl, amine, etc.) to improve water-solubility is a well-established strategy with the added benefit of increasing quantum yield in aqueous media (Supplementary Fig. 1).^33–35 For optimal utility, a variety of factors must be considered when selecting or modifying a fluorescent dye for quantitative live-cell imaging. Ideally, the fluorescent label should be cell membrane-permeable with a sufficient degree of water-solubility to afford biological compatibility. In addition, optimal fluorophores would also have minimum non-specific affinity for peripheral biological targets and maintain high photo- and chemical stability. In general, a net charge of zero (e.g., SiR, TMR, and ORFluors, etc.) or uncharged small molecules, and those with charge delocalized across the molecule are able to cross the cell membrane via a passive diffusion mechanism.^36–38 While small molecules that incorporate multiple negatively charged groups (e.g., ATTO and Alexa Fluor dyes series) provide improved solubility, these probes are often cell membrane-impermeant due to their net negative charge, which limits their application to the imaging of extracellular targets in live cell and tissue studies. However, these generalized rules do not apply when transporter mediated entry (e.g., ion channels, glucose transporter, etc.) or the endocytic pathway (e.g., protein toxins, virus, etc.) are involved. Notably, the effect of net charge on fluorophore stability, both chemical and photophysical, as well as its impact on in vitro and in vivo imaging when used as fluorescent labels have been investigated.^{34, 39–41} The general consensus regarding these studies is that zwitterionic fluorophores (i.e., net charge neutral) with charges balanced both electronically and geometrically over the molecular surface act as ideal selfshielding fluorescent labels with extremely low incidence of non-specific binding.^42–48

Here, we employ a strategic structural design to engineer rhodamine dyes that are suitable for quantitative live-cell imaging (Fig. 1c and Supplementary Fig. 2). SiR-Me is a structural derivative of SiR obtained by replacing the nucleophilic group (e.g., carboxylate, hydroxyl, amine, etc.) adjacent to the central bridging carbon with a methyl group (Fig. 1b), thus preventing the intramolecular cyclization leading to the non-fluorescent spirolactone while providing steric protection to the adjacent electron-deficient carbon. SiR-Me, however, has been shown to inherently target mitochondria (Extended Data Fig. 1) owing to the cationic net charge (+1).^{39, 44, 49, 50} While useful as a mitochondrial imaging probe, there have also been consistent reports of unwanted non-specific uptake, yielding false-positive signal from the dye-ligand conjugates when cationic SiR-Me was used as the fluorescent label vs. the zwitterionic SiR (Supplementary Fig. 3).^{24, 44–48} We hypothesized that this cationic charge could be balanced by the addition of a single negatively charged sulfonate moiety, chosen for its inert quality, and would result in a stable molecule less likely to undergo post-synthetic structural modifications (Fig. 1c). This structural design strategy resulted in zwitterionic, water-soluble and cell membrane permeable ORFluor dyes, poised to withstand environmental changes while resistant towards non-specific binding with subcellular structures. To date, efforts toward tuning the silicon-substituted rhodamines for improved chemical and photophysical properties have been hampered by various synthetic challenges that were only exacerbated by the asymmetric nature of our design strategy (Supplementary Fig. 4). This prompted us to develop robust synthetic sequences for each derivative that, in turn, provided routes to both the N-hydroxysuccinimide (NHS) ester and azido reactive versions with utility for bioconjugation reactions (Extended Data Fig. 2–3).

Photophysical properties of ORFluor dyes.

In order to evaluate the photophysical properties of SiR and TMR for comparison, we also synthesized these base fluorophores alongside our ORFluor dyes, OF650 and OF550 (Fig. 1d and Supplementary Fig. 5–7). The absorption and fluorescence emission spectra were collected in a phosphate-buffered saline (PBS) solution containing 10% dimethyl sulfoxide (DMSO). As shown in Fig. 1e, both OF650 and OF550 exhibited absorption and fluorescence emission that were red-shifted by less than 10 nm from their parent fluorophores. Taken together, both the wavelengths of maximum absorbance and emission of OF650 exhibited 100 nm red-shifts relative to OF550, and gave matching Stokes’ shift values (23 nm and 19 nm for OF550 and OF650, respectively). As expected, the extinction coefficient of SiR in aqueous buffer was low (ε_{PBS buffer} = 19,200 M⁻¹ cm⁻¹), owing to an open/close equilibrium that favors the colorless spirolactone isomer at physiological pH (Supplementary Fig. 3). Conversely, OF650 exhibited an extinction coefficient value (ε_{PBS buffer} = 125,500 M⁻¹ cm⁻¹) six times higher than that of SiR in the same buffer system. Unlike SiR however, the open/close equilibrium of TMR favors the zwitterionic open-ring form over the neutral closed-ring form. This was also evident from its merely 40% lower extinction coefficient value relative to OF550. Moreover, the structural modifications developed herein further improved the quantum yields of ORFluors compared to their base fluorophores. The brightness of OF650, for example, was nearly eight times greater than SiR, owing to an improved extinction coefficient, more closely matching the brightness of OF550. The absorption and emission spectra of both OF650 and OF550 produced sharper peaks with tighter full width at half maximum (FWHM) values (Fig. 1f) in agreement with the measurably higher water-solubility of the ORFluor dyes compared to SiR and TMR (Fig. 1g). A comparison of spectral sensitivity for the five fluorescent dyes, including fluorescent on/off switchable SiR and TMR, and fluorescent always-on SiR-Me and ORFluors was made in response to changes in environmental factors including aqueous and organic solvent systems with varied polarities, pH, temperature, and viscosity (Fig. 1h, and Supplementary Fig. 8–9). Using measured absorbance and emission intensity values, inertness of the five fluorescent dyes was quantified. Notably, fluorescent always-on SiR-Me and ORFluors were highly insensitive to these environmental changes, whereas the switchable SiR and TMR showed drastic changes in both absorbance and emission intensities following solvent polarity and viscosity changes. To a lesser extent, only SiR displayed pH-dependent behavior, while all five fluorescent dyes were resistant to temperature changes within the tested range (25 – 40 °C).

ORFluor-based probes suitable for live-cell imaging.

We first tested the cell permeability of our ORFluor OF550 and OF650 in live-cell staining studies. Following 30 min incubation in live U2OS cells, confocal fluorescence images of stained cells demonstrated the ORFluors entered cells efficiently and exhibited a punctate vesicular pattern. Subsequent studies demonstrated that both OF550 and OF650 were co-localized with the commercial lysosomal probe LysoTracker Green (Fig. 2a). To evaluate the safety profile, a standard MTS assay was used to assess cell viability, where the ORFluor dyes exhibited nearly 100% cell viability, as observed with the base fluorophores (Supplementary Fig. 10).

Fig. 2. ORFluors are a class of zwitterionic, water-soluble, cell membrane-permeant fluorophores suitable for live-cell imaging.

a Confocal live-cell images of U2OS cells co-incubated with ORFluors at a concentration of 1 μM and LysoTracker Green (green) show both OF550 (red, top row) and OF650 (red, bottom row) localizing in the lysosome compartment. Presented images are autocontrast adjusted. b Structures of ORFluor dyes conjugated to the HaloTag and SNAP-tag substrates. Airyscan confocal images of U2OS TetR-ER-HaloTag expressing cells (top row) co-labeled with OF550_HALO (green), OF650_HALO (red), and corresponding correlation of OF550_HALO and OF650_HALO fluorescence signals (top row, right). Airyscan confocal images of U2OS TetR-Mito-SNAP-tag expressing cells (bottom row) co-labeled with OF550_SNAP (green), OF650_SNAP (red), and corresponding correlation of OF550_SNAP and OF650_SNAP fluorescence signals (bottom row, right). Merged images of colocalization (yellow) are shown above the dashed white line (upper right) in each image. All cells were co-stained with Hoechst (blue) and examined from a minimum of three independent staining experiments. Scale bar = 10 μm. Single focal plane. r = Pearson correlation coefficient.

To fully profile our ORFluor dyes as inert self-shielding fluorescent labels for quantitative live-cell imaging applications, we engineered two doxycycline (i.e., tetracycline derivative) inducible cell lines using a Tet-On system that utilizes doxycycline to control the expression of a gene-of-interest and the production of the encoded protein. We generated one cell line with a stable expression of endoplasmic reticulum (ER) localized HaloTag fusion protein, and the other with a stable expression of mitochondria (Mito) localized SNAP-tag fusion protein. Self-labeling protein tag substrates for HaloTag and SNAP-tag were each conjugated to SiR and TMR, providing previously reported SiR_HALO and TMR_SNAP probes^{21, 24} that were used to confirm the fusion protein expression following Dox treatment (Supplementary Fig. 11–14). We then conjugated HaloTag and SNAP-tag substrates to the ORFluors to obtain probes OF650_HALO, OF550_HALO, OF650_SNAP, and OF550_SNAP. In live-cell imaging experiments, all four probes provided positive intracellular stained signal following Dox treatment (Supplementary Fig. 15–18). To further examine the labeling specificity of the ORFluor conjugated HaloTag and SNAP-tag ligands, we performed live-cell imaging at the single cell level using confocal microscopy with Airyscan. As expected, all cells stained with either OF650_HALO or OF550_HALO showed diffuse cytoplasmic staining patterns consistent with ER-specific labeling. Similarly, cells that were stained with ORFluor conjugated SNAP-tag substrates showed highly specific mitochondria labeling as expected (Extended data Fig. 4 and Supplementary Fig. 19).

To investigate the potential utility of these ORFluors for quantitative iPAI applications, we constructed a third Dox-inducible U2OS-TetR cell line with stable co-expression of ER-localized HaloTag and Mito-localized SNAP-tag fusion proteins. In the multicolor live-cell imaging experiments, cells were labeled with two sets of probes, each set containing spectrally distinct HaloTag and SNAP-tag substrates conjugated to ORFluors along with Hoechst stain for nuclear imaging (Supplementary Fig. 20). All four probes showed comparable brightness with high labeling specificity, similar to the cells stained and imaged in their respective single channels. Importantly, there was no spectral cross talk observed between channels. Beyond cytotoxicity and brightness, another critical factor for spatial and temporal measurements is the photostability of the fluorophores used in multicolor imaging. We have successfully demonstrated these ORFluors are useful as inert fluorescent labels for real-time multicolor imaging, allowing one to monitor the dynamic changes of the labeled protein targets in live cells (Supplementary Videos 1–5). We further confirmed the photostability of fusion proteins tagged with ORFluor dyes, which were comparable to the base fluorophores SiR and TMR (Supplementary Fig. 21). We then completed a final study to validate the inertness of our ORFluor dyes required for the iPAI probe design, where equivalent intracellular biodistribution between the paired dyes (i.e., OF550 and OF650) as drug labels will be required for precise quantification in ratiometric imaging. We co-incubated OF650_HALO and OF550_HALO in the U2OS TetR-ER-HaloTag expressing cells and were able to attain high co-localization precision and density between the two HaloTag ligands within the same cell. We observed similar results in the U2OS TetR-Mito-SNAP-tag expressing cells that were co-labeled with OF650_SNAP and OF550_SNAP (Fig. 2b and Supplementary Fig. 22). The appealing features described above confirm the suitability of dye pair OF650 and OF550 for quantitative iPAI in live cells.

ORFluor-labeled small molecule inhibitor iPAI probes.

To facilitate quantification of specific and non-specific accumulation of drugs in live cells, we sought to adapt the PAI technology to image the epidermal growth factor receptor (EGFR) protein bound to tyrosine kinase inhibitor (TKI) intracellularly in live cells (Fig. 3a). EGFR is a member of the tyrosine kinase family, involved in several signal transduction pathways impacting cell survival and proliferation.^51–53 EGFR mutations and overexpression, leading to aberrant cell signaling, has been observed in a range of human cancers.^{52, 54} Erlotinib, an FDA-approved TKI, targets the intracellular EGFR tyrosine kinase domain (EGFR-TKD) and blocks signal transduction. Previous molecular docking studies of EGFR-bound erlotinib co-crystal structures have shown that several functional groups on the EGFR residue participate in a hydrogen-bonding network with the quinazoline moiety of erlotinib, either directly or indirectly (i.e., water-mediated interaction).⁵⁵ The alkyne group attached to the phenyl ring further enhances binding to the kinase domain via hydrophobic interactions. The hydrophilic alkoxy tails, however, remain in the solvent region while the drug is active.⁵⁶ Guided by the published co-crystal structure and binding network of EGFR-bound erlotinib, we envisioned a means of labeling erlotinib by replacing the alkoxy tail (Fig. 3a, black) furthest from the critical binding motif (Fig. 3a, green) with a fluorescent label that would result in minimal, if any, disruption to the binding network, providing a targeted imaging agent that competes with the parent drug for the EGFR binding site. On the other hand, by converting the alkyne to a triazole group and linking directly to a fluorophore with a spectral profile distinct from the targeted version, the highest disruption to the binding-network could be generated, providing an equivalent inactive (untargeted) control agent.

Fig. 3. Erlotinib intracellular Paired Agent Image (iPAI) probe design.

a Structures of tyrosine kinase inhibitor erlotinib and the fluorescently labeled targeted and untargeted erlotinib iPAI probe pairs. The functional moieties of erlotinib and iPAI probes that are involved in binding with the protein target are color-coded green. b Docking of iPAI targeted probes SiR_Erl(T), OF650_Erl(T), TMR_Erl(T), and OF550_Erl(T) into the human EGFR ATP pocket in complex with the parent inhibitor erlotinib (orange). The ligands are displayed as sticks and colored by atom type; the protein receptor is represented as a Connolly Surface (Molsoft ICM). c Tabulated results of molecular docking calculations for the parent drug erlotinib, four targeted iPAI probes (SiR_Erl(T), OF650_Erl(T), TMR_Erl(T), and OF550_Erl(T)), and four untargeted iPAI probes (SiR_Erl(UnT), OF650_Erl(UnT), TMR_Erl(UnT), and OF550_Erl(UnT)) bounded complexes. d Bio-layer interferometry (BLI) binding sensorgrams of ORFluor labeled targeted (OF650_Erl(T) and OF550_Erl(T)) and untargeted probes (OF650_Erl(UnT) and OF550_Erl(UnT)) binding to biotinylated recombinant EGFR tyrosine kinase domain (EGFR-TKD) immobilized on Super Streptavidin (SSA) Biosensors. Kinetic data (Colored curve: observed experimental data) were fit globally using a Langmuir 1:1 binding model (Black dashed curve: fitted experimental data) in ForteBio Data Analysis HT 10.0 evaluation software to obtain the e association rate constant (k_a), dissociation rate constant (k_d) and the equilibrium dissociation constant (K_D) for all targeted probes.

As an initial testing of the targeted and untargeted design of these iPAI probes, we computationally docked all four targeted probes (SiR_Erl(T), OF650_Erl(T), TMR_Erl(T) and OF550_Erl(T)) into the EGFR ATP-binding pocket in complex with the parent drug erlotinib (Fig. 3b). Erlotinib docked with a score of −43.18, reproducing the H-bond interaction with the amide nitrogen of the hinge residue Met769 obtained in the co-crystal structure.⁵⁷ All four docked targeted probes confirmed the presence of the H-bond between N1 of the quinazoline from the targeting moiety and Met769, with docking scores closely matched to that of parent drug erlotinib. In contrast, docking of untargeted probes showed the intended loss of this key interaction, with docking scores drastically diminished (Fig. 3c). Given the positive computational outcome, we then synthesized these four erlotinib iPAI probe pairs to further validate our structural design.

To validate the model predicted targeted and untargeted behaviors of these iPAI agents, we used a biolayer interferometry (BLI) assay to determine the binding kinetics and affinity of these probes to recombinant EGFR-TKD immobilized on a Super Streptavidin (SSA) biosensor (Fig. 3d and Supplementary Fig. 23). In agreement with the computational docking studies, all targeted probes showed binding affinity for the EGFR-TKD with concentration dependency similar to the parent drug erlotinib. Notably, ORFluor labeled targeted erlotinib probes retained higher affinity for the EGFR-TKD than the SiR and TMR labeled versions. In contrast, none of untargeted probes or the untagged fluorescent dyes exhibited binding to the EGFR-TKD (Fig. 3e and Extended Data Fig. 5).

Erlotinib iPAI probes for quantitative live-cell imaging.

To demonstrate the functionality of our targeted and untargeted probes in vitro, we performed a series of co-staining experiments in the EGFR-positive human epidermoid carcinoma cancer cell line, A431. We observed a high level of co-localization in both targeted and untargeted groups using our spectrally distinct OF650 and OF550 as drug labels (Fig. 4a). In addition to being localized on the cell membrane, marking EGFR as confirmed by immunofluorescence staining using fluorescently labeled cetuximab, both OF550_Erl(T) and OF650_Erl(T) were also found to localize in the cytosol, closely resembling the targeted and non-specific accumulation of the parent drug erlotinib. In contrast, untargeted probes OF550_Erl(UnT) and OF650_Erl(UnT) only exhibited the intended non-specific accumulation across the cytoplasm. Surprisingly, neither SiR nor TMR labeled targeted probes showed characteristic membranous EGFR staining (Fig. 4b), contradicting what was found in the computational docking and BLI binding studies. This unexpected phenomenon likely resulted directly from the spirocyclic open/close equilibrium, where SiR_Erl(T) and TMR_Erl(T) predominantly exhibited the non-fluorescent spirolactone forms when bound in the EGFR-TKD. Notably, co-localization of ORFluor labeled untargeted erlotinib derivatives OF650_Erl(UnT) and OF550_Erl(UnT) was high, while the conventionally labeled untargeted erlotinib derivatives SiR_Erl(UnT) and TMR_Erl(UnT) showed variable uptake, further demonstrating the inertness of our ORFluors even when conjugated to small molecule therapeutics (Fig. 5a). Consistent with our structural design, environmentally dependent behavior was once again observed from the iPAI probes using SiR and TMR as labels in solvent systems consisting of various percentages of dioxane in water, whereas this issue was virtually nonexistent in the ORFluor-labeled iPAI probes (Fig. 5b).

Fig. 4. Superior live-cell imaging of drug distribution with ORFluor labeled iPAI probes.

a Fluorescence images of EGFR-positive A431 cells co-incubated with OF650_Erl(T) (red) and OF550_Erl(T) (green) show ORFluor labeled targeted probes co-localizing on the cell membrane with EGFR (blue), while untargeted probes also demonstrate equivalent biodistribution across the cytoplasm but lack targeted uptake. b SiR and TMR labeled targeted probes do not provide EGFR targeted signal, revealing their inability to fully display drug distribution in cells. EGFR expression was confirmed by immunofluorescence staining with cetuximab directly conjugated to Alexa Fluor 488 (EGFR_AB AF488). Scale bar = 25 μm.

Fig. 5. Quantitative imaging of erlotinib iPAI probes in live cells.

a DTA calculation for ratiometric images. b Environmental dependent comparison in response to polarity changes using measured absorbance (left) and emission intensity (right) values of the two imaging drug pairs (iPAI probe pair 1 [OF650_Erl(T) and OF550_Erl(UnT)] and 3 [SiR_Erl(T) and TMR_Erl(UnT)]) in aqueous solvent systems containing various percentages of dioxane, demonstrating ORFluor labeled drugs are more environmentally stable. Each dot represents the means of n=4 independent experimental values under each screening condition. c Comparison of live cell erlotinib iPAI staining in targeted channel (left), untargeted channel (middle) and corresponding DTA (right) images show ORFluor dyes (top row) are superior fluorophores to the classic rhodamine SiR and TMR (bottom row) as dye labels for quantitative imaging applications. Fluorescence images are normalized to one another in each imaging channel. Scale bar = 50 μm. Color bar shows DTA values in arbitrary units (A.U.).

DTA assessment using ORFluor-labeled erlotinib iPAI probes in live-cells.

As a starting point, we incubated the EGFR mutation-positive non-small cell lung cancer cell line HCC827 with iPAI probe pair 1 (OF650_Erl(T) and OF550_Erl(UnT)), which was labeled with ORFluor dyes, and compared it to the equivalent probe pair 3 (SiR_Erl(T) and TMR_Erl(UnT)), which was labeled with conventional SiR and TMR (Fig. 5c, Supplementary Fig. 24). Using pair 1, we observed positive fluorescence signal in both targeted and untargeted channels in live cells. By comparison, both targeted and untargeted agents of iPAI probe pair 3-stained cells showed substantially lower fluorescence intensities. Similarly, in another live-cell imaging experiment, iPAI probe pair 2 labeled with OF650 in the untargeted channel and OF550 in the targeted channel showed equivalent fluorescent signal compared to probe pair 1 (Supplementary Fig. 25). These results are in line with the hypothesis that a certain population of the parent fluorophore-labeled drugs were in fact, non-fluorescent due to the spirocyclic on/off property of SiR and TMR. Furthermore, the vast difference of the spirocyclic open/close equilibrium constants between SiR and TMR could cast doubt on the accuracy of ratiometric calculation of DTA using these conventional fluorophores.

DTA assessment using ORFluor-labeled erlotinib iPAI probes in tissues.

Plasma protein binding (PPB) of iPAI probes and untagged fluorophores was quantified, where all fluorescently labeled targeted erlotinib probes were found to exhibit PPB >90% (Supplementary Fig. 26), in agreement with previously reported PPB of erlotinib, which is approximately 95%.^{58, 59} In agreement with the reported pharmacology studies for erlotinib,^{60, 61} both targeted probes OF650_Erl(T) and SiR_Erl(T) were found to primarily localized to the liver, where clearance appeared to be by hepatic metabolism and biliary excretion, while the untargeted probes mainly exhibited a non-specific distribution pattern as expected (Supplementary Fig. 27). Given the consistently superior performance of the iPAI probe pairs labeled with ORFluor dyes, we set out to demonstrate the applicability of iPAI probes for assessing DTA in vivo in cell line-derived mouse xenografts (CDX). The targeted and untargeted OF650_Erl(T)/OF550_Erl(UnT) probe pair 1 were compared to the base fluorophore-labeled SiR_Erl(T) and TMR_Erl(UnT) targeted and untargeted probe pair 3. HCC827 tumors were collected 4 hours post intravenous injection of the respective probe pairs. Resulting iPAI probe distribution was quantified on the tissue sections of the resected tumors. Similar to their in vitro performance, the iPAI probe pair labeled with ORFluors (probe pair 1) showed twice the fluorescence intensity than the iPAI probe pair labeled with SiR and TMR (probe pair 3) in both targeted and untargeted channels (Fig. 6a). The targeted and untargeted images were then used to calculate maps of DTA in tissue, correlating to localized drug distribution. We observed higher DTA values and broader DTA distribution in pixel probability histograms that were quantified for the cohort of mice injected with iPAI probe pair 1. Future utility of the iPAI methodology will be towards the assessment and quantification of treatment studies using the unlabeled parent drug (i.e., erlotinib).

Fig. 6. iPAI erlotinib imaging enables quantification of drug target availability (DTA) in tissues.

A HCC827 xenograft bearing mice were systemically administered equivalent doses of either the targeted and untargeted OF650_Erl(T)/OF550Erl_(UnT) probe pair 1 (right column) or the SiR_Erl(T)/TMR_Erl(UnT) probe pair 3 (left column). Fluorescence images of the targeted (top row) and untargeted (second row) channels for each probe were collected, enabling calculation of DTA spatial maps (third row) and pixel probabilities (bottom row). In vivo competitive binding studies were performed in A431 xenograft bearing mice, where b iPAI probe pair 3 (SiR [targeted]and TMR [untargeted] fluorophores) or c iPAI probe pair 1 (ORFluors OF650 [targeted] and OF550 [untargeted] was systemically administered alone (-Erlotinib, left column) or in combination with the parent drug (+Erlotinib, right column). Pixel probability histograms for each imaging channel provide a quantitative view of the signal changes. All targeted and untargeted images are displayed with auto-contrast for visualization while DTA spatial maps (bottom row) are displayed at equivalent contrast levels across all panels. Color bar shows DTA values in arbitrary units (A.U.). For direct comparison between probes, injection conditions and tumor types, the inset values indicate relative intensity normalized across sections for each imaging channel (i.e., targeted, untargeted and DTA channels). Scale bar = 500 μm.

To examine DTA differences between cell lines with different EGFR expression levels, and to monitor how DTA changes in response to in vivo drug treatment, we then administered probe pair 1 (Fig. 6b) and 3 (Fig. 6c) to separate A431 CDX mouse cohorts following tumor treatment with and without parent drug erlotinib. In comparison to cohorts of mice without parent drug treatment, lower overall fluorescence intensities were observed in both targeted channels from probe pair 1 and 3, but not in the untargeted channels. This observation further demonstrated the unmodified parent drug erlotinib was competing with our fluorescently labeled targeted probes (OF650_Erl(T) and SiR_Erl(T)), but not the untargeted probes (OF550_Erl(UnT) and TMR_Erl(UnT)). Notably, among the cohorts of mice with no parent drug treatment, there was no fluorescence intensity difference in the targeted channels alone when comparing tissue sections between the HCC827 and A431 tumors for both OF650_Erl(T) and SiR_Erl(T). However, higher DTA values were observed in A431 tumors from both probe pair 1 and 3 once the untargeted distributions were addressed using the corresponding untargeted probes, OF550_Erl(UnT) and TMR_Erl(UnT), recapitulating the higher EGFR expression level in the A431 cell line.⁶² In addition, due to environmentally independent properties of the ORFluor labeled iPAI probes, the targeted and untargeted tissue-specific signals were brighter than using the conventional rhodamine-based fluorophore labels. This translated to higher DTA ratios even in the presence of parent drug, thereby improving the dynamic range available for DTA evaluation and stabilizing the ratiometric imaging approach.

Conclusion

Termed OregonFluors (ORFluor), OF650 and OF550 have shown substantially improved inertness and stability against environmental changes compared to the base fluorophores SiR and TMR, respectively. In addition to matched structural features, these two fluorophores also exhibited matched photophysical properties, including net charge, molecular size, Stokes’ shift, FWHM, and brightness. These matched spectral properties and improved stability against environmentally induced changes in live cells and tissues, along with their matched structural features, effectively demonstrate the added utility of these water-soluble and cell membrane-permeable fluorophores for quantitative live-cell imaging applications. Integration of OF650 and OF550 into the iPAI platform opens up opportunities for studying numerous other therapeutics in cells and tissues, offering a comprehensive spatial view of drug distribution and binding for improved personalized therapy assessment.

METHODS

Full synthetic details of all compounds, their structural and optical characterization, as well as imaging experiments are given in the Supplementary Information. Structural and purity data were processed using MassHunter Workstation Qualitative Analysis software (B.05.00) and MestReNova software (14.1.2).

Plasmids.

Plasmids used were: paGFP-Omp25 (Addgene #69598), pENTR4-SNAPf (w878–1) (Addgene #29652), pENTR4-HaloTag (w876–1) (Addgene #29644), and pLenti-puro-CMV/TO (Addgene #17293). The In-Fusion HD Cloning kit (Takara Bio, San Jose, CA) was used to generate recombinant fusion proteins SNAP-OMP25 and sp-Halo-KDEL in the entry clone backbone. The mitochondria outer membrane protein 25 fragment was amplified from the plasmid paGFP-Omp25, and then inserted into the plasmid pENTR4-SNAPf (w878–1). The HALO-tag fragment was amplified from pENTR4-HaloTag(w876–1), and an N-terminal signal peptide sequence from human endoplasmic reticulum chaperone BiP(HSPA5) and a C-terminal ER retention signal (KDEL) were included in the amplification primers. A Gateway LR cloning kit (ThermoFisher) was used to shuttle the recombinant fragments from entry clones into the lentiviral expression vector (pLenti-puro-CMV/TO). All primer sequences are listed in Supplementary Table 1.

General cell culture.

The doxycycline-regulated U2OS-TetR cell line was generously provided by Dr. Xiaolin Nan at Oregon Health and Science University (OHSU, Portland, OR).⁶³ The human kidney cell line, HEK293T (ATCC, CRL-3216), the human osteosarcoma cell line, U2OS (ATCC, HTB-96), the human epidermoid carcinoma cancer cell line, A431 (ATCC, CRL-1555), and the human lung adenocarcinoma cell line, HCC827 (ATCC, CRL-2868), were purchased from ATCC (Manassas, VA). The cell lines were expanded in their respective optimal growth media (U2OS, U2OS-TetR, HEK293T, A431: DMEM [ThermoFisher] + 10% fetal bovine serum [FBS, VWR Scientific, Radnor, PA] + 1% penicillin/streptomycin/glutamine [ThermoFisher]; HCC827: RPMI 1640 [ThermoFisher] + 10% FBS + 1% penicillin/streptomycin/glutamine) and stored at 37 °C in a 5% CO₂ incubator. All cells were maintained mycoplasma-free at passage numbers <25 for all studies.

Stable cell line generation.

Three doxycycline inducible U2OS cell lines were generated, including (1) U2OS cells (U2OS TetR-ER-HaloTag) stably expressing endoplasmic reticulum (ER) localized HaloTag fusion protein; (2) U2OS cells (U2OS TetR-Mito-SNAP-tag) stably expressing mitochondria (Mito) localized SNAP-tag fusion protein; and (3) U2OS cells (U2OS TetR-ER-Halo-Mito-SNAP-tag) stably co-expressing ER localized HaloTag fusion protein and Mito localized SNAP-tag fusion protein. The integrated lentiviral vector was used to generate these stable cell lines. The lentiviral particles were first packaged using the ViralPower Lentiviral Packaging System as previously described.⁶⁴ Briefly, the HEK293T cells were transfected by lentiviral plasmids using X-tremeGENE^™ 9 DNA Transfection Reagent (Sigma-Aldrich). The virus-containing supernatant was collected on day 2, 3, and 4 after transfection, sterile filtered using a 0.45 μm filter unit, and further concentrated in the Lenti-X Concentrator (Takara Bio). To generate a stable doxycycline inducible cell line, U2OS-TetR cells were transduced with the respective harvested virus. 48 hours after the infection, media was removed, and cells were exposed to three rounds of puromycin selection (2 μg/mL in DMEM with 10% FBS) before being tagged with SiR_HALO or TMR_SNAP to confirm the positive expression of the target fusion proteins in each constructed cell line. To generate a stable doxycycline inducible cell line with HaloTag and SNAP-tag fusion protein coexpression, U2OS-TetR-ER-HaloTag cells with constitutive expression of ER-localized HaloTag fusion protein were transduced with the SNAP-OMP25 lenti virus. Transduced cells were subjected to co-labeling with SiR_HALO and TMR_SNAP and subsequent fluorescence-activated cell sorting (FACS) on a BD FACSymphony S6 cell sorter (BD Biosciences) to isolate single clones.

Confocal live-cell imaging.

For confocal microscopy analysis of subcellular localization for OF650 and OF550, U2OS cells were passaged in DMEM with FBS (10%) and plated in 8-well chambered coverglass with 1.5H glass bottom. 48 hours later, media was removed, and cells were incubated in fresh FluoroBrite DMEM (ThermoFisher) containing OF650 (1 μM) or OF550 (1 μM), and each co-stained with LysoTracker Green DND-26 (1 μM) for 0.5 hour at 37 °C/5% CO₂. Staining media was removed and fresh media containing Hoechst 33342 (4 μM) was added. Cells were incubated for an additional 10 minutes in the incubator, then briefly washed with FluoroBrite DMEM media before fluorescence images were acquired on a laser scanning microscope (Zeiss LSM 880, Zeiss Zen Black software) with a Plan-Apochromat 63×/1.4 DIC M27 oil objective. The following excitation (x) and emission (m) filter combinations were utilized for image collection: 405x/(408–470)m; 488x/(490–553)m; 561x/(561–630)m; 633x/(638–700)m. Images were processed using ImageJ/FIJI (NIH). For Fast Airyscan confocal microscopy analysis, doxycycline-inducible cells were passaged in DMEM with FBS (10%) containing doxycycline (50 ng/mL) and plated in 8-well chambered coverglass with 1.5H glass bottom. 48 hours later, media was removed, and cells were incubated in fresh FluoroBrite DMEM (ThermoFisher) containing ORFluor-labeled SNAP-tag (1 μM) and/or HaloTag (5 μM) substrates for 1 hour at 37 °C/5% CO₂. Staining media was removed and fresh media containing Hoechst 33342 (4 μM) was added. Cells were incubated for an additional 10 minutes in the incubator, then rinsed 3 × 5 minutes in FluoroBrite DMEM prior to image acquisition. Live-cell imaging was performed at 37 °C and 5% CO₂ in an environmental stage enclosure. Fast Airyscan confocal microscopy was performed on a laser scanning microscope (Zeiss LSM 880) equipped with an Airyscan array detector. Standard laser lines (405-, 561- and 633-nm) were used to excite Hoechst-, OF550- and OF650-labeled cells, respectively. Fast Airyscan images were taken with a 63× oil objective. Automatic 2D or 3D Airyscan processing ZEN Black software (Zeiss) was used to acquire and process images. Pearson’s correlation coefficient values (r) were quantified using ImageJ/FIJI (NIH).

Molecular modeling and docking.

The 3D-coordinates of the EGFR catalytic domain in complex with the tyrosine kinase inhibitor, erlotinib, were retrieved from the PDB (1M17).⁵⁷ The model was energetically minimized in the internal coordinate space with Molsoft ICM 3.8.⁶⁵ Molecular docking studies of parent erlotinib, SiR_Erl(T), SiR_Erl(UnT), OF650_Erl(T), OF650_Erl(UnT), OF550_Erl(T), OF550_Erl(UnT), TMR_Erl(T), and TMR_Erl(UnT) were run as previously reported.⁶⁶ Each compound was assigned a score according to its fit within the receptor; this ICM score accounted for continuum and discreet electrostatics, hydrophobicity and entropy parameter.⁶⁶ All ICM scores obtained are listed in Fig. 3b.

In vitro biotinylation of recombinant EGFR-TKD protein.

Recombinant DNA fragment encoding the EGFR kinase domain residues 672–1016 (NCBI accession NP_005219) with a N-terminal AviTag and C-terminal 6xHis purification tag, was synthetically produced after codon optimization, and assembled into baculovirus vector pFastBac1 for expression in insect cells Spodoptera frugiperda (Sf9) at GenScript (New Jersey, USA). Purified recombinant proteins (>85% purity) were stored in 50 mM Tris buffer (pH8.0), 500 mM NaCl, and 5% glycerol. Avi-tagged EGFR-TKD was buffer exchanged using a 0.5 mL, 3 kDa Amicon Ultra centrifugal filter (Sigma-Aldrich) into 10 mM Tris (pH8.0), 50 mM NaCl and biotinylated with BirA biotin-protein ligase reaction kit (BirA500, Avidity) according to the manufacturer’s protocol. To remove excess d-biotin 0.5 mL, 10 kDa Amicon Ultra centrifugal filters (Sigma-Aldrich) were used and buffer exchanged into DPBS (ThermoFisher). The purified EGFR-TKD and biotinylated EGFR-TKD (^bEGFR-TKD) were further validated by the SDS-PAGE and western blot analysis. A-431 and U2OS cell lysate as controls were prepared using Pierce^® RIPA lysis and extraction buffer (ThermoFisher) according to manufacturer’s protocol. All samples were prepared by mixing NuPAGE^® LDS sample buffer (4X, ThermoFisher) with 50 mM DTT (Sigma-Aldrich), followed by heating of the sample for denaturing electrophoresis at 95 °C for 10 min. SDS-PAGE was performed using NuPAGE^® 4–12% Bis-Tris mini protein gel (ThermoFisher) in NuPAGE^® MOPS SDS running buffer (ThermoFisher). For direct gel staining, SimpleBlue^® safestain (ThermoFisher) was used according to manufacturer’s protocol. For western blot analysis, protein samples were transferred to a 0.45 μm Immobilon^®-P PVDF membrane in NuPAGE^® transfer buffer. After washing with DPBS, the membrane was blocked for 1 h at room temperature using intercept^® PBS (LI-COR Biosciences). The membrane was then incubated with mouse anti-6x-His tag antibody (ThermoFisher, MA1–135) and rabbit anti-EGFR antibody (Abcam, ab32562) diluted 1500x respectively in blocking buffer with 0.2% Tween-20 for 1 h at room temperature. Following three washes with DPBS, the membrane was further developed with IRDye^® 680RD donkey anti-mouse IgG secondary antibody (LI-COR Biosciences, 926–68072) and IRDye^® 800CW donkey anti-rabbit IgG secondary antibody (LI-COR Biosciences, 925–32213) diluted 8000x respectively in blocking buffer with 0.2% Tween-20 for 1 h at room temperature. Following three washes with DPBS, membranes were visualized on an iBright imaging system (ThermoFisher).

Bio-layer Interferometry (BLI) for Affinity Determination.

Kinetic assays were carried out at 30 °C in 384-well tilted-bottom plates (Sartorius) with orbital shaking at 1,000 rpm on a ForteBio Octet^® RED384 instrument with ForteBio Data Acquisition software 9.0. Running buffer consisting of DPBS (pH 7.4), 1% bovine serum albumin (BSA), 0.5% DMSO and 0.1% Tween-20 was used. Super Streptavidin (SSA) biosensor tips (Sartorius) pre-equilibrated in DPBS for 10 mins were loaded with 40 μg/mL of biotinylated EGFR-TKD ligand or biocytin (Sigma-Aldrich) controls in DPBS for 1200 s. The biocytin loaded sensor was used as a parallel reference sensor to quantify background binding for each concentration of analyte. The loaded biosensor tips were then blocked with 100 μg/mL of biocytin for 600 s, and washed with running buffer for 120 s. To establish a baseline, both the EGFR-TKD loaded sensor and parallel reference sensor were equilibrated in running buffer for 600 s and then dipped into wells containing three-fold dilutions of test analytes (erlotinib, targeted and untargeted probes, fluorescent dyes) in running buffer, or buffer alone (reference sample). Increasing concentrations of analytes ranging from 12–3000 nM were allowed to associate for 90 s, followed by a 180 s dissociation step in running buffer. Binding kinetics were determined using ForteBio Data Analysis HT 10.0 evaluation software. To minimize noise from background signal, non-specific binding, and signal drifts, the raw data was processed by subtracting parallel reference biosensor (biocytin loaded) and reference well (no analyte). For each interaction pair, the association rate constant (k_a) and the dissociation rate constant (k_d) were calculated as an average of three independent assays with at least four different concentrations that were globally fit to a 1:1 Langmuir binding model. Equilibrium dissociation constant (K_D) that defines the strength of the interaction or affinity was calculated as the ratio of the k_d to k_a. The analyzed data was exported and plotted in GraphPad Prism 9.0.

In Vitro iPAI Studies.

For erlotinib iPAI probe co-localization studies, A431 cells were passaged in DMEM with FBS (10%) and plated in 8-well chambered coverglass with 1.5H glass bottom. 48 hours later, media was removed, and cells were incubated in fresh FluoroBrite DMEM (ThermoFisher) containing OF650_Erl(T) (1 μM), OF550_Erl(T) (1 μM), and EGFR_ABAF488 (cetuximab directly conjugated to Alexa Fluor 488 [AF488], 1:3 antibody to fluorophore conjugation ratio with a concentration of 1.86 mg/mL, diluted 500x) for 0.5 hour at 37 °C/5% CO₂. Staining media was removed and fresh media containing Hoechst 33342 (4 μM) was added. Cells were incubated for an additional 10 minutes in the incubator, then briefly washed with FluoroBrite DMEM media. Fluorescence images were acquired using Zeiss Zen 3.5 blue software on a Zeiss

Axio Observer.Z1 microscope with an ApoTome.3 imaging system and outfitted with a 20x/0.8 Plan-Apochromat objective. Standard laser lines (405-, 488-, 561- and 633-nm) were used to excite Hoechst-, OF550- and OF650-labeled cells, respectively. Images were processed using ImageJ/FIJI (NIH). For in vitro iPAI studies, HCC827 cells were plated in phenol-red free RPMI 1640 media (supplemented with 10% FBS + 1% penicillin/streptomycin/glutamine) at a density of 5 × 10⁵ cells/mL (100 μL/well) in 96-well glass bottom cell culture plates (Cellvis). After cells had equilibrated for 24 hours at 37 °C/5% CO₂, imaging reagents were titrated onto cells in equal molar concentrations (0 – 1 μM) and allowed to co-incubate for 1 hour at 37 °C/5% CO₂. Wells were then replaced with fresh phenol-red free media and live cells were immediately imaged using Zeiss Zen 2.6 blue software on a Zeiss AxioObserver modified with a Yokogawa CSU-X1 spinning disk confocal unit and outfitted with a 20x/0.8 Plan-Apochromat objective. The following excitation (x) and emission (m) filter combinations were utilized for image collection: AF555 – 550/25x, 605/70m; AF647 –640/30x, 690/50m. For DTA quantification, normalization between the Cy3 and Cy5 imaging channels was completed by imaging the targeted and untargeted probes at matched titrations to calculate the scaling factor. The probes were titrated into a 96-well glass bottom plate and imaged at identical exposure times as the cell images to generate the scaling factor, which was introduced into our custom MatLab code to calculate DTA images.

Animal care and use.

All animal experiments were approved by the OHSU Institutional Animal Care and Use Committee (IACUC). All mice were hosted with up to five mice per age in ventilator cages in an AAALAC certified OHSU vivarium with controlled temperature (21–23 °C) and humidity (30–70%), supplied with food, water, and on 12-hour dark-light cycle. Mice were monitored daily for pain or distress for the duration of experimentation. Mice were placed on a chlorophyll-free diet (Animal Specialties, Inc., Hubbard, OR) one week prior to tumor resection. All rodent surgical procedures described herein were performed under full anesthesia composed of a 9/1 mixture of ketamine/xylazine. Ketamine (Hospira) was administered at a dose of 100 mg/kg and xylazine (AnaSed) was administered at dose of 10 mg/kg by intraperitoneal (IP) injection. The toe pinch method was employed to verify the depth of anesthesia prior to commencement of any surgical procedures. The standard method of euthanasia for mice was inhalation of carbon dioxide under full anesthesia at the end of each experiment. Euthanasia was confirmed by physical examination to ensure cessation of heartbeat and respiration and is consistent with the recommendations of the Panel on Euthanasia of the American Veterinary Medical Association.

Mouse xenograft models.

Mouse xenograft models were established using the previously developed protocol.⁶⁷ Briefly, Mixed male and female Athymic nude mice (Homozygous 490, Charles River Laboratories) were purchased at 32–38 days old. After at least 48 hours of acclimation, a total of 10 mice were subcutaneously implanted with A431 or HCC827 cell xenografts, described briefly as follows: Cells were trypsinized, counted and resuspended in their appropriate growth media to a concentration of 2 × 10⁷ cells/mL. The mice were then implanted with cells into each rear flank at a final concentration of 1 × 10⁶ cells/flank in 50% v/v Matrigel (Corning), resulting in two tumors per mouse. A total of n = 5 nude mice each were implanted for A431 and HCC827 cell derived xenograft (CDX) models, respectively. The mice were monitored daily after implantation for tumor growth. The tumors were allowed to grow to a maximum size of 1 cm³ with growth times varying for each cell line (A431: ~2–3 weeks; HCC827: ~7–8 weeks). Mice weighed ~20–25 g at the time of iPAI agent administration prior to sacrifice and tissue collection. When tumors reached 0.8–1 cm³, the iPAI agents were co-injected via the tail vein to n = 3 of each CDX mouse model at 2.5–3.5 mg/kg for both the targeted and untargeted iPAI probes. Formulation for all systemically administered erlotinib and iPAI probes was a co-solvent mixture of 10% DMSO, 5% Kolliphor (Sigma-Aldrich), and 85% of 75/25 FBS/PBS (VWR Scientific). For all CDX models, tumors were resected four hours after systemic administration and flash frozen in optimal cutting temperature (OCT) compounds (ThermoFisher).

Competitive binding study between parent drug erlotinib and iPAI derivatives.

Competitive binding studies was completed using the previously established protocol.⁶⁷ Briefly, A431 xenograft bearing mice were systemically administered parent drug erlotinib (2.5 mg/kg) 24 hours prior to systemic administration of equivalent doses of both parent erlotinib and the erlotinib iPAI probes (all at 2.5 mg/kg) four hours before sacrifice and tumor resection. For comparison, an A431 CDX mouse was systemically administered only the erlotinib iPAI probes (2.5 mg/kg) four hours prior to sacrifice and tumor resection.

Ex vivo iPAI imaging of varied EGFR expressing xenograft tissue.

Ex vivo iPAI imaging data collection and quantification was completed using the previously established protocol with minor modification.⁶⁷ Briefly, Frozen A431 and HCC827 xenograft tissue blocks were sectioned at 10 μm thickness with one section adhered to each slide. Fluorescence images of whole tissue sections were acquired on a Zeiss AxioScan.Z1 microscope equipped with a Colibri 7 light source (Zeiss) and Orca Flash4 v.2 (Hamamatsu Photonics). Images were collected using the following filter sets: Zeiss 38HE (Cy2/AF488), Zeiss 43HE (Cy3/AF555), and Zeiss 50 (Cy5/AF647). Excitation light was filtered using the following bandpass (BP) filters: 470/40 (38HE), 550/25 (43HE), and 640/30 (50) for Cy2, Cy3, and Cy5 channels, respectively. Emission light was filtered using the following BP filters: 525/50 (38HE), 605/70 (43HE), and 690/50 (50) for Cy2, Cy3, and Cy5 channels, respectively. Images were captured at 20X (Plan-Apochromat, 0.8NA) magnification, where image tiles with 10% overlap were stitched together using the Zeiss Zen 3.5 Blue software to produce a single tissue map. Mean fluorescence intensity for each image was calculated in ImageJ through manual segmentation of the tissue. The relative fluorescence intensities for each image were calculated by dividing the mean fluorescence intensity of an image by the highest mean fluorescence intensity of that particular image channel (i.e., targeted and untargeted channels). A custom MATLAB (R2019b, Mathworks) script (https://doi.org/10.5281/zenodo.4004647) was used to calculate DTA images for the A431 and HCC827 iPAI injected tissues. Relative fluorescence intensity and DTA values along with the generation of a histogram comparing the DTA images of the displayed A431 and HCC827 tissue sections were calculated following the same previously published protocol described above.⁶²

Data Availability

All data associated with this study are present in the paper, the Source Data or the Supplementary Information. Additional data generated during the study are available from the corresponding author upon reasonable request.

Code Availability

Custom-written MATLAB code used to calculate drug target availability is available at https://doi.org/10.5281/zenodo.4004647.

Extended Data

Extended Data Fig. 1. Fluorescence live-cell images of SiR, SiR-Me and OF650.

Representative images of U2OS cells stained with a fluorescent on/off switchable rhodamine fluorophore SiR, b fluorescent always-on fluorophore SiR-Me, and c ORFluor OF650 at 30 nM for 30 min. Fluorescence images are shown normalized to one another for the three rhodamine dyes. Positive intracellular localization was observed in cells that were stained with SiR-Me, while negligible fluorescence signal was observed in the cells stained with either SiR or OF650. d Colocalization imaging showed SiR-Me preferentially localized to the mitochondria with high affinity and specificity, confirmed by co-staining with commercial mitochondria probe, MitoTracker Green (green). All cells were co-stained with Hoechst (blue) and examined from a minimum of three independent staining experiments. Scale bar, 100 μm.

Extended Data Fig. 2. Synthesis of ORFluor OF650 and its conjugatable versions.

 ORFluor OF650 was prepared in six steps with an averaged chemical yield of 68%. A Vilsmeier–Haack reaction of dimethylbromoaniline precursor 23 gave benzaldehyde intermediate 24 in 88% yield. This was then followed by the addition of pre-lithiated 2-bromo-toluene (25), giving diaryl methanol intermediate 26 in decent yield (81%). The allyl-protected N-methyl bromoaniline 27 was activated by Lewis acid, ZnCl₂, to participate in a nucleophilic substitution reaction with secondary alcohol 26, providing heterodimer 28 in reasonable yield (56%). Subsequent lithiation of heterodimeric bromide 28 using tert-BuLi and addition of dichlorodimethylsilane yielded the leuco-basic rhodamine 29 (60%). Deprotection (i.e., allyl group removal) of 29, catalyzed by a Pd(PPh₃)₄ and dimethylbarbituric acid (DMBA) system, provided leuco-base rhodamine 30 in 68% yield. Introduction of a sulfonate group to the silicon-substituted rhodamine was achieved through alkylation of intermediate 30 using excess propane-1,3-sultone, followed by direct oxidation with p-chloranil, affording the final ORFluor OF650 (2) in 55% yield. To demonstrate that our synthetic strategy can be generalized to other asymmetrical silicon-substituted rhodamines and provide access to bio-conjugatable derivatives, we modified the synthetic sequence to prepare carboxylic acid OF650_COOH and N-hydroxysuccinimide (NHS) ester OF650_NHS versions (Fig. 1d). Intermediate 31, bearing a functional handle, was metalated before reaction with compound 24, giving the diaryl methanolic intermediate 32 with a protected carboxyl group in 74% yield. The reaction protocols used to synthesize OF650 – until removal of the allylic protecting group – were again utilized to afford compound 35. Initial attempts at alkylation using propane-1,3-sultone followed directly by hydrolysis gave very poor yields. However, when an acid-catalyzed hydrolysis was performed first, releasing the carboxyl group, the free carboxylic acid intermediate (36) was given in 46% yield and subsequent alkylation and oxidation products were furnished by the aforementioned protocols. The free acid OF650_COOH (6), treated with TSTU, was transformed to NHS ester OF650_NHS (37) in 69% yield.

Extended Data Fig. 3. Extension of zwitterionic structural design strategy to the TMR analog.

ORFluor OF550 was prepared in three steps with an averaged chemical yield of 48%. Typically, an asymmetric rhodamine dye is prepared through condensation between a hydroxy benzophenone and an aniline derivative together in a one-to-one ratio at high temperature, under acidic conditions. In our case, we sought to efficiently prepare asymmetrical zwitterionic fluorophore derivatives by introducing required structural variables (i.e., the phenyl ring motifs) at a later stage in the synthetic sequence. An Ullmann cross-coupling reaction of the commercially available dimethylaminophenol 38 and aryl bromide 39, catalyzed by CuI and 2-picolinic acid under basic conditions, afforded the diaryl ether intermediate 40. Alkylation of compound 40 with an excess amount of propane-1,3-sultone gave the sulfonated intermediate 41 (70%). Because the diaryl ether derivative 41 is prone to electronic activation, an acid catalyzed reaction should readily provide the oxidized fluorophore product via the Friedel–Crafts acylation–cyclization reaction. To ease the product recovery and purification process, a catalytic system consisting of ZnCl₂ in ethanol was introduced to the condensation step. However, cyclization of compound 41 with a benzaldehyde 42 did not proceed until the temperature rose above 100 °C. A subsequent DDQ-mediated oxidation reaction was performed to push the reaction to completion in 33% yield over two steps. A bioconjugatable version OF550_NHS was prepared using the aforementioned key intermediate 41. Using this same protocol and the orthogonal protecting group-bearing intermediate 43 allowed for the introduction of a functional handle (e.g., protected carboxyl group). Deprotection under acidic conditions, followed by DDQ-mediated oxidation provided the OF550_COOH (8) as a single regioisomer. The free acid OF550_COOH (8) was then transformed into NHS ester OF550_NHS (44) via treatment with TSTU in 48% yield.

Extended Data Fig. 4. Fast Airyscan confocal fluorescence imaging with SNAP-tag and HaloTag ligands.

Intracellular localization of OF650 (red) and OF550 (green) labeled SNAP-tag and HaloTag substrates in doxycycline inducible U2OS TetR-Mito-SNAP-tag expressing cells and TetR-ER-HaloTag expressing cells. All cells were co-stained with Hoechst (blue) prior to fixation, and examined from a minimum of three independent staining experiments. Scale bar = 10 μm. Single focal plane.

Extended Data Fig. 5. BLI binding determination of erlotinib and SiR and TMR labeled iPAI probes.

Representative BLI binding sensorgrams of erlotinib, SiR and TMR labeled targeted and untargeted probes, as well as untagged fluorophores SiR and TMR, binding to biotinylated recombinant EGFR-TKD immobilized on Super Streptavidin (SSA) biosensors. BLI sensorgrams show the association of each probe (0–90 s) at the indicated concentrations ranging from 37–3000 nM, followed by the subsequent dissociation in binding buffer without analytes (90–180 s). Kinetic data (Colored curve: observed experimental data) were fit globally using a simple Langmuir 1:1 binding model (Black dashed curve: fitted experimental data) in ForteBio Data Analysis HT 10.0 evaluation software to obtain the association rate constant k_a, dissociation rate constant k_d and the equilibrium dissociation constant K_D.