Intrinsic fluorescence of the clinically approved multikinase inhibitor nintedanib reveals lysosomal sequestration as resistance mechanism in FGFR-driven lung cancer

The human lung cancer cell lines NCI-H1703, NCI-H520 (non-small cell lung cancer, NSCLC) and DMS114 (small cell lung cancer, SCLC) were purchased from American Type Culture Collection (Manassas, VA, USA) and grown in RPMI-1640 supplemented with 10% fetal calf serum (FCS, PAA, Linz, Austria) at 37 °C and 5% CO₂. Cells were authenticated by array comparative genomic hybridization (aCGH) and checked for Mycoplasma contamination (Mycoplasma Stain kit, Sigma, St. Louis, Missouri, USA) on a regular basis.

Nintedanib, elacridar and chloroquine were purchased from Selleckchem (Munich, Germany). LysoTracker® Red was obtained from Thermo Fisher Scientific (Waltham, MA, USA), bafilomycin A1 was purchased from Sigma.

Three dimensional-fluorescence spectra were recorded on a Horiba FluoroMax®-4 spectrofluorometer (Kyoto, Japan) and processed using the FluorEssence v3.5 software package. Stock solutions of nintedanib-ethanesulfonate in dimethylsulfoxide (DMSO) were diluted with phosphate-buffered saline (PBS) (10 mM, pH 7.4) to 15 μM (final DMSO concentration 1%) and the fluorescence spectra were measured at excitation wavelengths from 220 nm to 420 nm while the emission was within the range of 240–700 nm. Scans were run at room temperature with excitation and emission slit widths of 5 nm.

To determine cell viability upon inhibition of FGFR1, 3 × 10³ cells were seeded in 96-well plates and incubated overnight. Cells were exposed to the indicated concentrations of nintedanib in the presence or absence of the indicated concentrations of elacridar, bafilomycin A1 or chloroquine. After 72 h, cell survival was determined using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT)-based vitality assay (EZ4U, Biomedica, Vienna, Austria). Dose-response curves were plotted using GraphPad Prism software (La Jolla, CA, USA). IC₅₀ values were determined from non-linear regression curve-fitting (sigmoidal dose-response with variable slope) in GraphPad Prism and indicate drug concentrations that resulted in a 50% reduced cell viability in comparison to untreated controls. Drug synergism was determined using Calcu Syn software (Biosoft, Ferguson, MO, USA) according to Chou-Talalay and expressed as combination index (CI) [33]. A CI value of <0.9 was considered a synergistic effect, a CI value between 0.9–1.1 indicates additivity and a CI value greater than 1.1 was considered an antagonistic effect.

5 × 10⁵ cells were resuspended in serum-free RPMI medium containing 2.09 mg/ml 4-morpholine-propanesulfonic acid (MOPS, Sigma) and 15 mM 4-(2-hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES, Sigma). Following a 1 h preincubation with 10 μM elacridar or 1 μM bafilomycin A1, cells were treated with the indicated concentrations of nintedanib. Intracellular drug accumulation was measured on a LSRFortessa flow cytometer (BD Biosciences, East Rutherford, NJ, USA) at the indicated time-points. Compound fluorescence was detected using 405 nm and 488 nm laser excitation wavelengths, and Horizon V450 (450/40 nm) and FITC (530/30 nm) bandpass emission filters, respectively. Data were analyzed using Flowing Software (University of Turku, Finland) and are depicted as relative increase in fluorescence intensities (arbitrary units, a.u.) compared to untreated controls.

5 × 10⁴ NCI-H1703 cells were seeded in 8-well chamber slides (Ibidi, Martinsried, Germany). After 24 h, cells were treated with 10 μM nintedanib and intracellular drug accumulation was imaged at the indicated time-points on a live cell microscope (Visitron Systems, Puchheim, Germany) using a 40× oil immersion DIC objective and VisiView® software. LEDs were used for widefield DIC and fluorescence (395/25 nm excitation and 460/50 nm bandpass filter for blue (DAPI) fluorescence and 475/34 nm excitation and 525/50 nm bandpass filter for green (FITC) fluorescence) illumination (Visitron Systems). Images were taken with a sCMOS 4.2MPxl digital camera. Increase in nintedanib fluorescence intensity over time was quantified using ImageJ software. The fluorescence signals of individual cells from at least three independent microscopic images were determined and values are depicted as relative fluorescence increase as compared to the 0 min control. To investigate the impact of lysosomal pH on the intralysosomal accumulation of nintedanib, cells were pretreated with 1 μM bafilomycin A1 and 1 μM LysoTracker® Red prior to exposure to 10 μM nintedanib. Cell-free nintedanib fluorescence properties were determined by imaging of the drug in its crystalline form, afterwards dissolved to 1 mM in DMSO and recrystallized after DMSO evaporation in the DIC, FITC and DAPI channels at 40× magnification.

Total cell protein extracts were separated by sodium dodecyl sulfate- polyacrylamide gel electrophoresis (SDS-PAGE) and blotted onto polyvinylidene difluoride membranes (PVDF, Thermo Fisher Scientific). Anti-ß-actin (AC-15) was purchased from Sigma, anti-ABCB1 (C219) from BioLegend (San Diego, CA, USA). Horseradish peroxidase-conjugated secondary antibodies were obtained from Santa Cruz Biotech (Dallas, TX, USA).

5 × 10³ cells were seeded in 8-well spot slides (Thermo Fisher Scientific). The next day, cells were coincubated for 1 h with 10 μM nintedanib and LysoTracker® Red. Alternatively, DMS114 cells were incubated for 1 h with 10 μM nintedanib in the presence or absence of 10 μM of the ABCB1 inhibitor elacridar [34]. Cells were fixed with 4% paraformaldehyde (PFA) for 15 min. Images were acquired on an inverted point scanning confocal microscope with PMTs (LSM700, Zeiss, Jena, Germany) using a 63× Plan-Apochromat 63×/1.4 oil immersion objective with Zen2010® software (Zeiss) using 405 nm (nintedanib blue), 488 nm (nintedanib green) or 555 nm (LysoTracker® Red) solid state laser lines for excitation and 420 nm longpass, 556 nm shortpass and 559 nm longpass emission filters, respectively. Colocalization signals derived from nintedanib and LysoTracker® Red were calculated in ImageJ software using the thresholded Manders‘ Colocalization Coefficient (MCC), which corrects for differences in mean signal intensities of the two channels (a MCC of 0 means no colocalization and 1 means perfect colocalization, see [35]). The mean thresholded MCC was derived from ten to twenty individual cells analyzed from at least three independent microscopic images. Statistical significance of overlapping pixel intensities was calculated by ImageJ software using Costes Colocalization Test [36]. A p-value of 0.95 or greater was considered to be statistically significant.

For detection of nintedanib in subcutaneous tumor allografts, 5 × 10⁵ CT26 mouse colon carcinoma cells were injected subcutaneously into the right flank of four 8 weeks old male BALB/c mice. Animal experiments were authorized by the Ethics committee at the Medical University of Vienna and carried out in accordance with the guidelines for the welfare and use of animals in cancer research, as well as meeting the Federation of Laboratory Animal Science Associations (FELASA) guidelines’ definition of humane endpoints and the Arrive guidelines for animal care and protection [37]. Upon tumor formation, two mice received a single oral dose of 100 mg nintedanib per kg bodyweight dissolved in H₂O containing 0.5% Klucel and 0.1% Tween80 18 days post-engraftment. Two mice received solvent only. Two hours after administration, mice were sacrificed and tumors were embedded and frozen in OCT medium (Sakura Finetek, Staufen, Germany). 5 μm sections were sliced on a cryomicrotome (Thermo Fisher Scientific), fixed with 4% PFA on Superfrost Plus microscope slides (Thermo Fisher Scientific) and counterstained with 4′,6-diamidine-2′-phenylindole dihydrochloride (DAPI, Sigma). Tumor endothelium was visualized on consecutive sections using an anti-MECA-32 antibody (ab27853, Abcam, Cambridge, United Kingdom). Images of overlapping regions of consecutive cryosections were acquired on a Zeiss LSM700 confocal laser scanning microscope using 40× and 63× oil immersion objectives.

10⁶ cells were plated in 6-well plates. The next day, cells were preincubated with 1 μM bafilomycin A1 for 1 h. Subsequently, 10 μM nintedanib was added to cells for 10 or 60 min. Cells were washed with PBS, trypsinized, washed again twice and lysed by repeated freeze/thaw cycles. Samples were analyzed by HPLC using a Dionex UltiMate 3000 system equipped with an L-7250 injector, an L-7100 pump, an L-7300 column oven (set at 35 °C), a D-7000 interface and an L-7400 UV detector (Thermo Fisher Scientific) set at a wavelength of 287 nm. Separation of nintedanib was carried out at 35 °C using a Hypersil BDS-C18 column (5 μm, 250 × 4.6 mm I.D., Thermo Fisher Scientific), preceded by a Hypersil BDS-C18 precolumn (5 μm, 10 × 4.6 mm I.D.). The mobile phase consisted of a continuous gradient mixed from ion pair buffer, pH 3.0 (50 mM potassium phosphate with phosphoric acid and 5 mM heptane sulfonic acid) (mobile phase A) and acetonitrile (mobile phase B). Mobile phase was filtered through a 0.45 μm filter (HVLP04700, Millipore, Billerica, MA). Mobile phase B linearly increased from 10% acetonitrile (0 min) to 95% at 7 min, at which point it was kept constant until 20 min. The percentage of acetonitrile was decreased within 1 min to 10% to equilibrate the column for 14 min before application of the next sample. Linear calibration curves were performed by spiking drug-free cell culture medium with standard solutions of nintedanib to give a concentration range from 0.05 to 10 μg/ml (average correlation coefficients: >0.999). The limit of quantification (LOQ) for nintedanib was 0.03 μg/ml. Coefficients of accuracy and precision for these compounds were <9%.

Data were analyzed using GraphPad Prism software. If not stated otherwise in the figure legends, one out of at least three independent experiments in triplicate is depicted. So each data point represents the mean ± SD of triplicate values. Statistical evaluation was performed using student’s t-test as well as 1-way or 2-way analysis of variance (ANOVA). Tukey’s post-test was performed to examine differences between drug treatment effects. P-values below 0.05 were considered statistically significant. * p < 0.05; ** p < 0.01; *** p < 0.001.

With the intention to dissect the intracellular distribution of nintedanib, we investigated whether this multikinase and FGFR-targeting inhibitor might display intrinsic fluorescence activity. Indeed, full excitation-emission 3-dimensional (3D) fluorescence plots revealed a distinct fluorescence pattern of this compound. Under cell-free conditions, nintedanib exhibited an emission maximum at 482 nm when excited at 390 nm (Fig. 1a). Consequently, we investigated whether the observed intrinsic fluorescence activity under cell-free conditions could be utilized to detect this compound label-free in treated cancer cells. To this end, the three FGFR1-driven lung cancer cell lines DMS114, NCI-H520 and NCI-H1703 were used. These cell lines all bear oncogenic amplification of the FGFR1 gene and are hypersensitive towards FGFR1 inhibitors [38]. Accordingly, also in our analyses IC₅₀ values for nintedanib were found in the low-micromolar to nanomolar range (Fig. 1b; 1.97 ± 0.31 μM, 6.67 ± 4.20 μM and 0.98 ± 0.33 μM for DMS114, NCI-H520 and NCI-H1703, respectively). Flow cytometric analysis showed significantly increasing fluorescence signals in the blue portion of the spectrum (450/40 nm bandbass filter) upon nintedanib treatment when excited with the 405 nm laser (Fig. 1c). Interestingly and unexpectedly from cell-free conditions, also 488 nm laser excitation (530/30 nm bandpass filter) of cells treated with nintedanib yielded strong fluorescence activity in the green emission range (Fig. 1c; Additional file 1: Table S1). This effect was also observed in live-cell fluorescence microscopy of treated NCI-H1703 cells (Fig. 1d). This suggests that, based on specific (sub)cellular conditions, the nintedanib molecule or a derivative/state had adopted novel fluorescence properties. Interestingly, fluorescence intensity was distinctly higher when excited with the 488 nm as compared to the 405 nm laser suggesting that either this fluorescence is stronger per se or that an accumulation of the green-fluorescent nintedanib derivative/state had occurred.

The observed fluorescence activity of nintedanib prompted us to investigate whether flow cytometry is a feasible approach to measure intracellular drug accumulation in a dose- and time-dependent manner. Indeed, exposure to nintedanib resulted in a distinct dose- and time-dependent increase in fluorescence intensity in all tested cell lines (Fig. 2a-c). Of note, increasing nintedanib fluorescence intensity over time suggested steady uptake kinetics at least until 2 h after drug administration with the exception of NCI-H520 cells, where the signal peaked after 30 min. We further investigated whether -additionally to flow cytometry- fluorescence microscopy might be another useful approach to monitor the uptake kinetics of the investigated TKI in treated cells. Thus, we performed live cell imaging of NCI-H1703 cells treated with 10 μM nintedanib. Indeed, weak fluorescence was observed already 5 min after drug exposure. At 10 min exposure, distinct fluorescent foci became apparent which increased in intensity in a time-dependent manner (Fig. 2d, e). These data prove that intracellular accumulation of nintedanib and its dose and time-dependent dynamics can be measured by both flow cytometry and fluorescence microscopy. Additionally, the observed foci suggested that the green-fluorescent nintedanib derivative/state accumulated in a yet undefined cell organelle.

We further wanted to verify whether the detected nintedanib fluorescence enables the visualization of decreased intracellular drug accumulation mediated by active drug efflux mechanisms. To this end, we used DMS114 cells and their isogenic subline DMS114/NIN previously selected in our group for acquired nintedanib resistance [39]. This resistance phenotype is mediated by overexpression of the ATP-binding cassette drug efflux transporter B1 (ABCB1, P-glycoprotein), which actively pumps nintedanib out of the cell, thus minimizing its cytotoxic efficacy (Fig. 3a, b). In line, ABCB1 blockade by the allosteric inhibitor elacridar completely restored nintedanib sensitivity (Fig. 3b). In parallel to viability assays, DMS114 and DMS114/NIN cells treated with 10 μM nintedanib for 1 h in the presence or absence of an equimolar concentration of elacridar were also investigated by confocal fluorescence microscopy of PFA-fixed cells. The green fluorescence signal of nintedanib and its spotted pattern were well preserved in fixed cells. In contrast, the fluorescence signal in DMS114/NIN cells was, indeed, virtually absent. However, coincubation with elacridar completely restored nintedanib fluorescence intensity in DMS114/NIN cells while having no significant effects in the parental cell line lacking ABCB1 expression (Fig. 3c, d). In line with this, flow cytometric analysis of cells treated with 10 μM nintedanib confirmed a lack of nintedanib fluorescence in DMS114/NIN cells and accumulation restoration by coincubation with elacridar (Fig. 3e).

The focal appearance of nintedanib fluorescence (compare Fig. 2d) prompted us to further analyze the underlying intracellular drug compartmentalization. The fact that nintedanib is a weakly basic compound which can be fully protonated in a low pH environment led us to hypothesize that preferred sites of intracellular accumulation might be acidic organelles such as lysosomes. Thus, we performed confocal microscopy of nintedanib-treated cells that were simultaneously labelled with LysoTracker® Red (Fig. 4a). Corroborating our hypothesis, spatial overlap of signals originating from nintedanib with LysoTracker® Red was found, suggesting selective nintedanib accumulation in lysosomes. To analyze whether nintedanib signal indeed colocalized with LysoTracker® Red, correlations of overlapping pixel intensities were calculated using thresholded Mander’s colocalization coefficient (MCC). This analysis yielded high correlation coefficients of 0.87, 0.92 and 0.72 for NCI-H1703, DMS114 and NCI-H520, respectively, indicating substantial colocalization between nintedanib- and LysoTracker® Red-derived signals (Fig. 4b). Scatter plots showing drug/LysoTracker® Red pixel intensity correlations for nintedanib are shown in Fig. 4c-e. Additionally, Costes test of overlapping pixel intensities yielded a p-value of 1.00, providing statistical evidence for significant colocalization. Thus, we conclude that nintedanib selectively accumulates in lysosomes of treated cells and hence represents a lysosomotropic agent.

To test whether nintedanib is also detectable in tumor tissue in vivo, we generated subcutaneous allograft models of the highly tumorigenic mouse colorectal carcinoma cell line CT-26. Indeed, oral treatment resulted in the accumulation of detectable drug levels adjacent to terminal microvasculature of the tumor tissue 2 h post-administration (Fig. 4f). Importantly, the focal appearance of nintedanib in the fixed tumor sections seemed to recapitulate the subcellular, dotted distribution of the drug in cell culture. This illustrates that the intrinsic fluorescence activity might enable sensitive detection of nintedanib not only in vitro, but also in vivo in the relevant target tissue following oral administration.

The identified lysosomotropism of nintedanib prompted us to ask whether its cytotoxic potential might be enhanced upon disruption of the lysosomal integrity. Weakly basic lipophilic agents are commonly believed to become fully protonated in the acidic environment of lysosomes. This protonation renders them incapable of freely permeating lipid bilayers, thus trapping them in the lysosomal lumen [32]. To find out whether decreased lysosomal accumulation of nintedanib results in enhanced cytotoxicity, we exposed cells to the lysosomal vacuolar H⁺-ATPase (V-ATPase) inhibitor bafilomycin A1. This compound specifically blocks H⁺-ion influx and thus results in lysosomal alkalization [40]. Importantly, flow cytometric analysis as well as live cell microscopy revealed that bafilomycin A1 treatment completely abrogated green fluorescence of nintedanib in the investigated cell lines. Also for the blue nintedanib fluorescence, a moderate reduction was observed, albeit at a less pronounced level (1.3-fold in NCI-H1703, 1.7-fold in DMS114 and 1.5-fold in NCI-H520 cells), reaching significance only in DMS114 cells (Fig. 5a; Additional file 2: Figure S1A-C and Additional file 3: Table S2). Slightly reduced blue nintedanib fluorescence in the presence of Bafilomycin A1 was also detected by fluorescence microscopy (Additional file 2: Figure S1C). This indicated a trend towards reduced intracellular nintedanib levels in case of lysosomal alkalization. Direct quantification by HPLC indeed confirmed distinctly lower total intracellular nintedanib concentrations in the presence of bafilomycin A1 (1.8-fold and 3.4-fold after 10 min and 60 min, respectively) (Fig. 5b). In live cell microscopy, the blue nintedanib fluorescence exhibited a more diffuse intracellular distribution as compared to the green, exclusively intralysosomal signal (Additional file 2: Figure S1C). Furthermore, imaging of NCI-H1703 cells showed that lysosomal nintedanib accumulation was completely abolished in cells pretreated with bafilomycin A1 (Fig. 5c; Additional file 2: Figure S1C). This suggests that the observed green fluorescence activity originates from a specific state of nintedanib present exclusively in lysosomes with low pH. Interestingly, coincubation with non-toxic concentrations of bafilomycin A1 (Additional file 4: Figure S2A) led to a distinct sensitization of NCI-H1703 and NCI-H520 cells towards nintedanib (Fig. 5d, e and Additional file 4: Figure S2B, C, respectively). The slightly lower sensitization of NCI-H520 as compared to NCI-H1703 cells at both bafilomycin A1 concentrations (1.7-fold versus 2-fold sensitization, respectively) is likely to originate from the inherently lower sensitivity of the former cell line towards FGFR inhibition. Similarly, nintedanib co-incubation with subtoxic doses of chloroquine, another inhibitor of lysosomal acidification based in this case on proton capturing, also exerted synergistic effects, albeit to a slightly lesser extent (Additional file 4: Figure S2D-F). Together, this suggests that lysosomal alkalization suppresses protonation-based sequestration of nintedanib and, hence, loss of the respective green fluorescence emission. The lysosome exclusivity of the green nintedanib fluorescence likely arises from drug accumulation to such a high extent that results in compound precipitation. This is substantiated by the observation that crystalline nintedanib in cell-free conditions exhibited both green and blue fluorescence activity (Additional file 5: Figure S3). However, nintedanib dissolution resulted in complete and selective quenching of the green fluorescence, whereas the blue one was retained.

In conclusion, despite the observed decrease in overall intracellular nintedanib levels upon raising of the lysosomal pH, a concomitant increase of its cytosolic concentration leading to enhanced target interaction efficacy might be causative for the distinctly increased cytotoxic potential.