Crizotinib sensitizes the erlotinib resistant HCC827GR5 cell line by influencing lysosomal function

Abstract In non‐small cell lung cancer, sensitizing mutations in epidermal growth factor receptor (EGFR) or cMET amplification serve as good biomarkers for targeted therapies against EGFR or cMET, respectively. Here we aimed to determine how this different genetic background would affect the interaction between the EGFR‐inhibitor erlotinib and the cMET‐inhibitor crizotinib. To unravel the mechanism of synergy we investigated the effect of the drugs on various parameters, including cell cycle arrest, migration, protein phosphorylation, kinase activity, the expression of drug efflux pumps, intracellular drug concentrations, and live‐cell microscopy. We observed additive effects in EBC‐1, H1975, and HCC827, and a strong synergism in the HCC827GR5 cell line. This cell line is a clone of the HCC827 cells that harbor an EGFR exon 19 deletion and has been made resistant to the EGFR‐inhibitor gefitinib, resulting in cMET amplification. Remarkably, the intracellular concentration of crizotinib was significantly higher in HCC827GR5 compared to the parental HCC827 cell line. Furthermore, live‐cell microscopy with a pH‐sensitive probe showed a differential reaction of the pH in the cytoplasm and the lysosomes after drug treatment in the HCC827GR5 in comparison with the HCC827 cells. This change in pH could influence the process of lysosomal sequestration of drugs. These results led us to the conclusion that lysosomal sequestration is involved in the strong synergistic reaction of the HCC827GR5 cell line to crizotinib–erlotinib combination. This finding warrants future clinical studies to evaluate whether genetic background and lysosomal sequestration could guide tailored therapeutic interventions.

However, individual resistance to these targeted therapies occurs on average within a year, with a mechanism involving acquired mutations, such as the common EGFR-T790M mutation. The thirdgeneration EGFR-TKI osimertinib has been approved for the treatment of T790M positive NSCLC (Jänne et al., 2015) and is recently approved as first-line treatment (Passiglia, Raez, & Rolfo, 2018).
Amplification of cMET is another genetic aberration that leads to resistance to EGFR-TKIs (Engelman et al., 2007;Ou, Agarwal, & Ali, 2016). Besides functioning as a resistance mechanism, cMET is also known as an oncogenic driver itself. Hereby, amplification of the cMET gene or exon 14 skipping serve as good biomarkers for cMET-TKIs (Camidge et al., 2014;Lutterbach et al., 2007;Van Der Steen et al., 2016). The dual ALK-cMET inhibitor crizotinib has been approved for the treatment of ALK-rearranged NSCLC and is currently undergoing trials as a treatment for cMET-driven NSCLC (Bahcall et al., 2016).
The physicochemical properties of a drug can influence its activity. This is the case for hydrophobic drugs with a pKa of around 9 (Da Silva, Honeywell, Dekker, & Peters, 2015;De Klerk, Honeywell, Jansen, & Peters, 2018). Most of these drugs can diffuse freely through the plasma membrane and the membranes of intracellular compartments, but several ATP binding cassette (ABC) transporters, such as P-glycoprotein (PgP) and breast cancer resistance protein (BCRP) may mediate efflux from the cytosolic compartment into cellular organelles such as lysosomes (De Klerk et al., 2018). Since the lysosomes have a very acidic environment, with a pH of around 5, these drugs are protonated and cation-trapping takes place, also known as lysosomal sequestration (Zhitomirsky & Assaraf, 2015b).
The acidity of the lysosomes is caused by the activation of the vacuolar-type H + -ATPase. In model systems, this proton pump can be inhibited by bafilomycin A1. If the intracellular drug concentrations decrease significantly after the addition of bafilomycin A1, then lysosomal sequestration is involved (Da Silva et al., 2015;Honeywell et al., 2010). Since most of the targets of TKIs are situated at the plasma membrane, lysosomal sequestration prevents their inhibitory role (Zhitomirsky & Assaraf, 2016).
ABC transporters, such as PgP, multidrug resistance-associated proteins (MRPs) and BCRP can also mediate resistance by effluxing drugs out of the cells. These efflux pumps are located on the cellular membranes, have high expression in gut epithelium and their primary role is to remove xenobiotics from the body (Sharom, 2008). However, many tyrosine kinase inhibitors are also substrates for transport by PgP (Da Silva et al., 2015;De Klerk et al., 2018). There is a clear need to clarify the role of these pumps in limiting the efficacy of these therapies, as well as to generate new strategies, such as lysosomal photo-destruction of MDF cells overexpressing PgP (Nowak-Sliwinska et al., 2015).
Here, we investigated the combined effect of erlotinib and crizotinib on various cell lines with different properties and aimed to unravel new mechanisms of synergy in the HCC827GR5 cells. The parental HCC827 cell line is a NSCLC adenocarcinoma cell line harboring an EGFR exon 19 deletion, which is a sensitizing mutation of EGFR. This cell line has been made resistant to gefitinib through prolonged exposure with increasing dosages, which resulted in cMET amplification (Engelman et al., 2007), and combined treatment with the EGFR inhibitor gefitinib and the cMET inhibitor PHA-665752, which resulted in substantial growth inhibition (Engelman et al., 2007). In the present study, we have shown a strong synergism with two compounds commonly used in the clinical setting (erlotinib and crizotinib), and determined, for the first time, the potential role of lysosomes in this pharmacological interaction.

| Cell lines and drugs
The HCC827 and H1975 cell lines were purchased from ATCC. The  Table 1.

| Sulforhodamine B assay (SRB) and combination index
The sulforhodamine B assay was used to determine growth inhibition after mono-and combination therapy (Bijnsdorp, Giovannetti, & Peters, 2011;Sciarrillo et al., 2016). Cells were plated in a 96-well plate (Greiner Bio-One GmbH), using 2000-5000 cells per well. Cells were treated the next day with either 0.1% DMSO as control, erlotinib (0-10 µM), crizotinib (0-5 µM) or their combination. To evaluate if the pharmacological inhibition of MRP1 might affect the sensitivity to crizotinib in HCC827 cells, we performed further experiments with the MRP1 inhibitor MK571, at 20 µM, as described previously (Assaraf et al., 2003). All cells were treated for 72 hr and all experiments were repeated at least three times.
After the determination of the IC 50 values for erlotinib and crizotinib monotherapy, cells were treated with fixed concentrations (IC 20 and IC 40 ) of crizotinib in combination with a range of erlotinib (0-10 µM). The combination index (CI) was calculated with Calcusyn (Biosoft, Cambridge, UK), based on the method of Chou and Talalay as described previously (Bijnsdorp et al., 2011). This method takes into account the curves of the monotherapies at the respective concentrations to determine the effect of the combination. Fraction affected (FA) was determined at each dose and only values with an FA ≥ 0.5 were used to calculate an average combination index (CI). A CI < 0.8 represents a synergistic combination, 0.8 < CI < 1.2 is additive and CI > 1.2 is antagonistic.

| Wound healing assay
Cells were seeded in a confluent layer in 96-well plates and allowed to attach. The scratch was made using the sterile scratch tool (Peira Scientific Instruments, Belgium). Detached cells were washed away and new medium was added to the wells. Plates were photographed with the Acumen eX3 laser scanner imaging cytometer (TTP-LabTech Lts, UK). Cells were treated immediately after imaging with 10 µM erlotinib, 5 µM crizotinib or their combination. Plates were incubated at 37°C and periodically photographed to monitor the scratch area.
The experiment was terminated after 8 hr for HCC827 and 16 hr for

| Cell cycle analysis
Cells were seeded in a six-well plate and treated with drugs as previously described. After 24 and 48 hr cells were collected in round-bottom FALCON tubes (BD, Franklin Lakes, NJ) and redissolved in 0.5 ml of a propidium iodide solution (50 µg/ml PI, 0.1% sodium citrate, 0.1% Triton X-100 and 0.1 mg/ml ribonuclease-A). The analysis was performed on a BD FACS calibur and data were analyzed with CELLQuest™-Pro software.

| Spheroid assay
Flat-bottom 96-well plates were coated with sterile unsupplemented heated medium containing 1.5% agarose and allowed to dry for 20 min. Cells were seeded at 5-20 × 10 4 cells/well depending on the optimal spheroid size and given 3 days to form three-dimensional structures, that we named spheroids. Next, these spheroids were treated with 5 µM crizotinib, 10 µM erlotinib or their combination. To investigate the effects of these drugs on cells organized in the spheroids, we evaluated the amount of light passing through, as recently described (Sciarrillo et al., 2019). Images of spheroids were taken with an automated phase-contrast microscope (Acumen eX3 laser-scanner). Pixel intensities of 8-bit black/white-converted images were calculated using the ImageJ software and expressed as mean gray value (i.e., the sum of all gray values of the spheroid selection divided by the pixels of that selection). The "cell aggregation" for each drug-treated spheroid was then calculated by normalizing for the mean gray values of the sum control spheroids. Note: To determine the sensitivity to crizotinib and erlotinib cells were treated during 72 hr with several concentrations of crizotinib (0-5 µM) or erlotinib (0-10 µM). Curves were plotted in GraphPad prism v5 and IC 50 dosages were determined. VAN DER STEEN ET AL.

| 8087
2.6 | Pathscan intracellular signaling array (fluorescent read-out) The PathScan sandwich enzyme linked immunosorbent assay (ELISA) was purchased from Cell Signaling Technology (Leiden, The Netherlands) and used according to manufacturer's instructions. Cells were seeded and treated for 24 hr and lysed in lysis buffer with 1 mM phenylmethylsulfonyl fluoride (PMSF). The glass slide was blocked and the lysate was added to the wells. The slide was washed thoroughly and incubated with the detection antibody cocktail. The fluorescent signal was determined using the LI-COR Biosciences Odyssey imaging system.
The ArrayVision software was used to determine pixel intensity.

| Peptide tyrosine kinase activity array
The kinase activity of 144 kinases was determined with the PamGene tyrosine kinase activity array. Cells were seeded in 25 cm 2 flasks and treated with erlotinib, crizotinib or their combination for 24 hr. Cells were lysed and scraped in M-PER buffer containing phosphatase and protease inhibitors (Thermo Scientific, Rockford, IL). Samples were centrifuged and the supernatant was stored at −80°C. Control sample mix was prepared using ABL buffer (Westburg), 100 µM adenosine triphosphate (Sigma-Aldrich), and fluorescein-labeled antibody PY20 (Exalpha, Maynard, MA).
Five µg of lysate protein was used for the analysis. The arrays were blocked with 2% BSA before sample loading. The activity was measured at 30°C over 60 cycles using a Pamstation12 (PamGene, 's-Hertogenbosch, The Netherlands). Imaging was performed after each cycle with a 12-bit charge-coupled device (CCD) camera so as to monitor the fluorescence intensities in real-time. Intensities were fit to calculate the final phosphorylation by Bionavigator software version 6.1.

| Intracellular drug concentration determination
The intracellular drug concentration was measured as previously described (Honeywell et al., 2010). Cells were treated with the drugs for 24 hr. Pellets were thawed by adding 160 µl of cold phosphate buffer and homogenized. Samples were vortexed and 100 µl was mixed with 400 µl acetonitrile. Samples were incubated for 20 min on ice, vortexed, and centrifuged at 21,000g at 4°C for 10 min. Next, 100 µl of the sample was transferred to a 96-well plate for LC injection and analyzed (Honeywell et al., 2010).

| Western blot analysis
Cells were seeded and treated with the drugs for 24 hr. Cells were lysed in lysis buffer (Cell Signaling Technology) supplemented with 1 mM PMSF on ice for 5 min. Next, cells were dislodged using a cell scraper, lysates were sonicated three times for 10 s and spun down for 10 min at 4°C, 14,000g. Supernatants were transferred and either used immediately or stored at −80°C.
Samples were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis at 100 V for 1 hr using a TGX-precast gel (BioRad, Veenendaal, The Netherlands). Wet transfer to a PVDF membrane was performed at 200 mA for 2 hr.

| Live cell fluorescence microscopy
Cells were seeded in Lab-Tek II Chambered coverglasses grade 1.5 (Thermo Scientific, Rockford, IL) and allowed to attach overnight. Cells FIJI software was used for image analysis (Schindelin et al., 2012;Schindelin, Rueden, Hiner, & Eliceiri, 2015), importing z-stacks with Bio-Formats. Ten representative cells were selected per sample ( Figure 1a,b). In each z-plane, the lysosomes were located in the Lysotracker Red channel (Figure 1a), which was converted to binary and signals were automatically traced by the "analyze particle" tool with the threshold "triangle." The selected regions were overlayed with the pHrodoGreen image (Figure 1b,c). First, the selected regions were deleted from the pHrodoGreen image and the remaining intensity was determined ( Figure 1d). Secondly, the outer regions were deleted and the intensity of the lysosomes was measured ( Figure 1e). These analyses were repeated for each of the Z-planes per sample. The intensities were summed and corrected for the area of the cells/lysosomes and the number of z-planes.

| Statistics
Statistical analysis was performed with Graphpad Prism v5 using a one-way analysis of variance with Tukey's post hoc testing, with p < .05 considered as significant. All tests were performed in triplicate unless mentioned otherwise. Values represent mean ± standard error of mean.

| Sensitivity and drug interaction
The effect of the combination of the EGFR-TKI erlotinib and the cMET-TKI crizotinib was determined on different cell lines

| Effect on protein phosphorylation of EGFR and cMET
The influence of erlotinib and crizotinib on the phosphorylation of their respective targets EGFR and cMET was further evaluated using western blot analysis. In the HCC827, erlotinib alone was sufficient to inhibit phosphorylation of EGFR, while in the HCC827GR5 cells, phosphorylation of EGFR was not completely inhibited after erlotinib monotherapy ( Figure 2b). However, the combination of erlotinib with crizotinib resulted in complete inhibition of EGFR-phosphorylation.
Crizotinib monotherapy of both cell lines led to complete inhibition of cMET-phosphorylation, which was also found after combination therapy. Of note, in addition to blocking cMET phosphorylation, treatment of HCC827 and HCC827GR5 cells with crizotinib resulted in a decrease in phosphorylation of EGFR. Moreover, the treatment of both cell lines with erlotinib also resulted in a marked reduction in cMET phosphorylation. These results are in agreement with several previous studies which suggest that cMET associates with EGFR either directly, or via adapter molecules, and this association between receptors might also enable TGFα or EGF to phosphorylate cMET through EGFR. Interestingly this crosstalk might have important implications for tumorigenesis, as well as for motogenic signaling and response to therapies (Guo et al., 2008;Jo et al., 2000;Stolz & Michalopoulos, 1997;Van Der Steen et al., 2018).

| Cell cycle and cell death analysis
The effect of the drugs on cell cycle distribution was studied at equimolar concentrations in the two cell lines, showing similar effects. Erlotinib induced a G1 arrest (Figure 2c) from 29% to 65% in the parental cell line and from 32% to 58% in the GR5 clone, respectively. Crizotinib, on the other hand, caused a G2 arrest from 21% to 57% in the HCC827 and from 26% to 71% in the HCC827GR5. The G2/M arrest was diminished in the parental cell

| Cell migration
For this purpose, the effect of this combination therapy on cell migration was investigated by the wound-healing assay ( Figure 2a). In the parental cell line (Figure 2b), 69 ± 10% of the scratch area was closed when treated with 0.1% DMSO (control).
After crizotinib treatment 76 ± 8% of the scratch was closed, while erlotinib induced a significant reduction of migration, with 32 ± 6% of the scratch closed. Similarly, the combination of both drugs resulted in only 32 ± 5% gap closure. In the HCC827GR5 clone ( Figure 2c) the control showed 80 ± 3% gap closure.
Crizotinib or erlotinib treatment resulted in comparable closure with 67 ± 3% and 69 ± 2%, respectively, whereas the combination resulted in a significant reduction of migration, with 30 ± 5% closure.

| Spheroid assay
Since previous studies illustrated that 3D cultures are generally more Treatment with crizotinib alone or in combination with erlotinib led to a diminished phosphorylation of ERK1/2 (30%), PRAS40 (15%), and GSKβ (30% and 38% in HCC827 and HCC827GR5, respectively) in both cell lines as compared to the untreated control, whereas phosphorylation was generally less decreased after erlotinib monotherapy. However, differences in phosphorylation levels between the HCC827 and the HCC827GR5 cell lines did not seem sufficient to explain the synergy in the HCC827GR5 cell line after combination therapy. Also, the phosphorylation in the crizotinib treated cells and in the cells treated with the combination therapy did not show statistically significant differences.

Control E C Combo
HCC827 Control E C Combo To further investigate the intracellular concentration, we measured crizotinib trapping in the lysosomes. A concentration of 50 nM of bafilomycin A1, which was previously reported to cause lysosomal dysfunction (Bowman, Siebers, & Altendorf, 1988;Yoshimori, Yamamoto, Moriyama, Futai, & Tashiro, 1991),  The difference in crizotinib uptake between both cell lines could be due to a difference in expression of ATP binding cassette transporters, of which PGP and MRP1 are the most common ones. However, the Western Blot analysis (Figure 2b) showed that PGP was not expressed in both cell lines. A difference in expression of MRP1 was found, showing a slightly higher expression of MRP1 in the HCC827GR5 cells. However, this slight difference is unlikely to cause a difference in intracellular crizotinib levels. Indeed our experiments with the MK571 inhibitor of MRP1 did not impact drug sensitivity ( Figure S2).
F I G U R E 5 Effect of erlotinib on the intracellular and lysosomal accumulation of crizotinib and on the role of pH. Cells were treated with 0.1% dimethyl sulfoxide as control, 10 µM erlotinib, 5 µM crizotinib or their combination for 24 hr. Bafilomycin (50 nM) was used to perturb the lysosomal function.  Table S1. (d, e) Intracellular effect of drugs on HCC827 and HCC827GR5, respectively, cells were stained for 1 hr with 5 µM sunitinib and 0.5 µM Lysotracker Red, and for 30 min with pHrodoGreen. Each sample was divided in multiple focus planes (z-stack). Z-stacks were imaged using a Leica TCS SP8 STED 3× microscope. Image panels of d from left to right: sunitinib staining of HCC827 control cells; sunitinib staining of HCC827 treated with 10 µM erlotinib and 5 µM crizotinib; Phrodogreen and Lysotracker red staining of HCC827 control cells; Phrodogreen and Lysotracker red staining of HCC827 treated with 10 µM erlotinib and 5 µM crizotinib. Image panels of e from left to right: sunitinib staining of HCC827GR5 control cells; sunitinib staining of HCC827GR5 treated with 10 µM erlotinib and 5 µM crizotinib; Phrodogreen and Lysotracker red staining of HCC827GR5 control cells; Phrodogreen and Lysotracker red staining of HCC827GR5 treated with 10 µM erlotinib and 5 µM crizotinib. Control: 0.1% DMSO, E: 10 µM erlotinib, C: 5 µM crizotinib, Combo: 10 µM erlotinib + 5 µM crizotinib, B: 50 nM Bafilomycin A1. *p < .05 as compared to control, **p < .01; ***p < .001 as compared to control VAN Table S1.
The higher the intensity of the pHrodoGreen, the lower the pH. The

| DISCUSSION
The present study supports a potential role for both genetic properties and lysosomal function in the synergistic interaction of erlotinib and crizotinib in NSCLC cells.
This novel finding was obtained by evaluating the HCC827 cell line, which harbors an EGFR exon 19 deletions, and the HCC827GR5 cells, which also harbors the cMET amplification as a resistance mechanism to gefitinib. Combined treatment with the EGFR-TKI erlotinib and the cMET-TKI crizotinib resulted in a very strong synergy in the HCC827GR5 cells. In contrast, no synergy was observed in the cMET amplified EBC-1 or the EGFR mutated H1975 (L858R + T790M) cell lines.
The signaling pathways of cMET and EGFR are strongly intertwined (Breindel et al., 2013;Guo et al., 2008) which could result in synergism between erlotinib and crizotinib. However, the EBC-1 cell line which is wild-type for EGFR and harbors a cMET amplification only showed an additive effect of the combination. An activating EGFR mutation seems therefore essential for synergism.
Altogether these data prompted us to focus on the HCC827GR5 cells and compare all experiments to the parental HCC827 cell line.
The synergy between erlotinib and crizotinib was also reflected in the wound-healing assay, where only the combination was able to inhibit cell migration in the HCC827GR5 cell line. Activation of cMET has been shown to stimulate cell migration (Birchmeier, Birchmeier, Gherardi, & Vande Woude, 2003;Bladt, Riethmacher, Isenmann, Aguzzi, & Birchmeier, 1995) and the same holds true for EGFR (Carcereny et al., 2015;Ramis, Thomàs-Moyà, de Mattos, Rodríguez, & Villalonga, 2012). However, it is remarkable that the cell line harboring both aberrations migrates twice as slow as compared to the parental cell line, which is reflected in the time it takes for the control wound to close (16 vs. 8 hr).
Analysis of the sub-G1 fraction is also in line with the found synergy in the HCC827GR5 cell line, whereas in the parental cell line the effect of erlotinib was slightly stronger than the combination.
Further, the cell cycle analysis showed a G1 arrest after erlotinib Huether, Höpfner, Sutter, Schuppan, & Scherübl, 2005) and a G2/M arrest after crizotinib (Megiorni et al., 2015), which is in accordance with the literature. However, after the combination, the effect of crizotinib was dominant but weaker than after the monotherapy.
Remarkably, this synergy was further reflected when comparing 3D cultures of both cell lines. Through the analysis of cell aggregation of the parental HCC827 and the HCC827GR lines, which had a different growth pattern in 3D, with a recently validated method, we indeed found that the combination significantly reduced cellular aggregation of the spheroids in the HCC827GR, while only a slight reduction was detected in the HCC827 cells.
To determine whether the synergy is due to the combined inhibition of cMET and EGFR downstream signaling or if other mechanisms play an important role, we compared downstream phosphorylation of the parental cell line and the HCC827GR5 clone.
Overall, both cell lines reacted very similarly to the different treatment conditions, not giving a satisfactory explanation for the synergy in the HCC827GR5 cells. The highest differences in basal activity, around 600%, were observed for several tyrosine receptor kinases. However, in the HCC827 parental cell line, the activity of these kinases was below the baseline, whereas in the HCC827GR5 activity of these kinases was just above the baseline. This makes the comparison between these values very difficult. Moreover, in both the HCC827 and HCC827GR5 cells, the activities of kinases after treatment with crizotinib monotherapy or with the combination were approximately the same.
We next focused on the cellular pharmacokinetics of the drugs.
Crizotinib and gefitinib are both known as hydrophobic weak-base compounds (Da Silva et al., 2015). These weak bases are able to freely cross the plasma membrane and intracellular membranes. The compounds might also accumulate in the lysosomes, and this accumulation is driven by pH partitioning between the cytoplasm and the lysosomes (Zhitomirsky & Assaraf, 2015b). The low pH in the lysosomes causes protonation of several weak-base TKIs, thus preventing the membrane-crossing of these bases and leading to accumulation in the lysosomes (Gotink et al., 2011). This results in lysosomal stress and activation of the coordinated lysosomal expression and regulation (CLEAR-pathway) by the TFEB transcription factor. This accumulation may also trigger the activation of lysosomal exocytosis, whereby lysosomes travel over the microtubuli, fuse with the plasma membrane and release their cargo into the extracellular environment (Zhitomirsky & Assaraf, 2015a, 2017. Erlotinib, on the other hand, is associated with the cytoskeleton (Da Silva et al., 2015).
Of note, the intracellular concentration of crizotinib varied significantly between treatment conditions in the HCC827GR5, with a significant increase in concentration after combination treatment as compared to crizotinib monotherapy. Although we can only hypothesize that this increase would have clinically relevant effects, a similar increase in the concentration of crizotinib was associated with relevant synergistic interaction in our in vivo models of pancreatic cancer (Avan et al., 2013).
By the addition of bafilomycin, the intracellular concentration of crizotinib decreased to 1 pmol/µg protein in all treatment conditions. This demonstrates that crizotinib is sequestered in the lysosomes (Da Silva et al., 2015) and that the combination possibly influences the process of lysosomal sequestration.
This effect might be caused by an alteration of the pH partitioning between the cytoplasm and the lysosomes (Zhitomirsky & Assaraf, 2015b), as has been shown in the MCF-7 cell line (Zhitomirsky & Assaraf, 2015b). Another explanation of our results might be the influence of erlotinib on autophagy. Several studies have shown the influence of erlotinib, and EGFR-TKIs in general, on the induction of autophagy (Han et al., 2011;Li, Lam, Mak, Zheng, & Ho, 2013;Nihira et al., 2014). This induction of autophagy was observed both in EGFR mutated and EGFR-independent cells (Han et al., 2011). In the process of autophagy (White, 2012), cell compartments destined for breakdown are enveloped by a membrane and fuse with the lysosomes. Next, the proteinases in the lysosomes break down the cell compartment. It has been shown that upon the accumulation of lysosomotropic compounds, the pH of the lysosomes is increased (Lu, Jessen, Strock, & Will, 2012). In turn, this increase in lysosomal pH can prevent the optimal functioning of lysosomal proteases, ultimately leading to impairment of autophagy. Moreover, lysosomal accumulation by itself is enough to activate autophagy. The transcription factor TFEB, which is activated upon lysosomal accumulation, translocates to the nucleus and activates the expression of genes involved in lysosomal biogenesis, autophagy, and endocytosis (Settembre et al., 2011). The combination of erlotinib with crizotinib treatment might thus lead to enhancement of the impaired autophagy process, resulting in increased cell death. Of note, previous studies showed that enhancement of autophagy is associated with an increase in p-AMPK (Thr172) and a decrease in p-mTOR (Ser2448; Li et al., 2013). However, our data on phosphorylation show no increase in p-AMPK in the HCC827GR5 cell line after combination therapy as compared to the monotherapies, nor a large decrease in p-mTOR, contradicting this hypothesis.
A third explanation might be that the association of erlotinib with the cytoskeleton interferes with lysosomal exocytosis over the microtubule trackers (Zhitomirsky & Assaraf, 2017), leading to an increase in intracellular crizotinib after combination therapy as compared to monotherapy. However, there are no studies determining in detail the interaction of erlotinib with the cytoskeleton and its influence on the microtubule racks.

| CONCLUSION
Both genetic characteristics and lysosomal function seem to contribute to the synergistic interaction between crizotinib and erlotinib in the HCC827GR5 cell line, and more research is

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author.