ARTICLE https://doi.org/10.1038/s41467-020-16952-9 OPEN Multiple low dose therapy as an effective strategy to treat EGFR inhibitor-resistant NSCLC tumours João M. Fernandes Neto 1, Ernest Nadal 2,11, Evert Bosdriesz 1,10,11, Salo N. Ooft 3,11, Lourdes Farre 2,4, Chelsea McLean 3, Sjoerd Klarenbeek5, Anouk Jurgens1, Hannes Hagen1, Liqin Wang1, Enriqueta Felip6,7, Alex Martinez-Marti6,7, August Vidal2, Emile Voest 3, Lodewyk F. A. Wessels1, Olaf van Tellingen 8, Alberto Villanueva 2,9 & René Bernards 1 ✉ 1234567890():,; Resistance to targeted cancer drugs is thought to result from selective pressure exerted by a high drug dose. Partial inhibition of multiple components in the same oncogenic signalling pathway may add up to complete pathway inhibition, while decreasing the selective pressure on each component to acquire a resistance mutation. We report here testing of this Multiple Low Dose (MLD) therapy model in EGFR mutant NSCLC. We show that as little as 20% of the individual effective drug doses is sufficient to completely block MAPK signalling and pro- liferation when used in 3D (RAF + MEK + ERK) or 4D (EGFR + RAF + MEK + ERK) inhibitor combinations. Importantly, EGFR mutant NSCLC cells treated with MLD therapy do not develop resistance. Using several animal models, we find durable responses to MLD therapy without associated toxicity. Our data support the notion that MLD therapy could deliver clinical benefit, even for those having acquired resistance to third generation EGFR inhibitor therapy. 1 Divisionof Molecular Carcinogenesis and Oncode Institute. The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands. 2 Group of Chemoresistance and Predictive Factors, Subprogram Against Cancer Therapeutic Resistance (ProCURE), ICO, Oncobell Program, IDIBELL, L’Hospitalet del Llobregat, Barcelona, Spain. 3 Division of Molecular Oncology and Immunology, The Netherlands Cancer Institute, 1066 CX Amsterdam, The Netherlands. 4 Institute Gonçalo Moniz, Fundaçao Oswaldo Cruz (FIOCRUZ), Rio de Janeiro, Brasil. 5 Experimental Animal Pathology, The Netherlands Cancer Institute, 1066 CX Amsterdam, The Netherlands. 6 Department of Medical Oncology, Vall d’Hebron University Hospital and Vall d’Hebron Institute of Oncology (VHIO), Barcelona, Spain. 7 Autonomous University of Barcelona (UAB), Barcelona, Spain. 8 Division of Clinical Pharmacology, The Netherlands Cancer Institute, 1066 CX Amsterdam, The Netherlands. 9 Xenopat S.L., Business Bioincubator, Bellvitge Health Science Campus, Barcelona, Spain. 10 Department of Computer Science, Faculty of Science, Vrije Universiteit, Amsterdam, The Netherlands. 11These authors contributed equally: Ernest Nadal, Evert Bosdriesz, Salo N. Ooft. ✉email: r.bernards@nki.nl NATURE COMMUNICATIONS | (2020)11:3157 | https://doi.org/10.1038/s41467-020-16952-9 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16952-9 I nhibition of signalling pathways that are activated by onco- if we could further reduce the drug concentrations, we diluted the genic mutations elicit therapeutic responses due to “addiction” 4D combination. When the drugs were reduced to half of the IC20 of the cancer to the activated pathway1. However, in advanced concentrations, the 4D combination was no longer able to achieve cancers, development of resistance is practically inevitable due to complete inhibition of proliferation and was similarly unable to secondary mutations that restore signalling through the drug- mediate complete MAPK pathway inhibition, indicating that inhibited pathway. Such acquired resistance mutations affect there is a threshold that limits efficacy (Supplementary Fig. 1a, b). either the drug target itself or components that act upstream, Based on this, we continued our MLD studies using the IC20 downstream or parallel to the activated signalling component2,3. concentrations as “Low Dose”. To make sure our findings were In BRAF mutant melanoma and non-small cell lung cancer not drug-specific, we tested the MLD approach using different (NSCLC), inhibition of two components of the same oncogenic inhibitors for each of the nodes in the MAPK pathway (erlotinib pathway (BRAF + MEK, referred to as “vertical targeting”) has as EGFRi, BGB-283 as RAFi, selumetinib as MEKi and LY- been shown to provide more lasting clinical benefit compared to 3214996 as ERKi). Supplementary Fig. 1c, d show that we inhibition of only BRAF4,5. More recently, both clinical6,7 and obtained essentially the same effect with these drugs in 3D and pre-clinical8 studies have shown that inhibition of three compo- 4D combinations. This, together with the notion that each drug is nents of the same oncogenic pathway further increases ther- used at low dose, makes it very unlikely that off target effects of apeutic benefit. In these scenarios the drugs are usually the four drugs are responsible for the observed effects. administered at maximum tolerated dose (MTD). The increase in the number of drugs being used in combination is often MLD therapy minimises therapeutic resistance. Next, we tested accompanied by an increase in toxicity and to this date virtually how MLD therapy compares to standard-of-care high dose no studies have been done to assess the efficacy of using drugs therapy in terms of resistance development. To mimic high dose below-MTD. In a preclinical model, multiple drugs used at low therapy, we treated PC9 cells with a concentration of EGFR dose also demonstrated promising activity in ovarian clear cell inhibitor gefitinib that inhibited cell viability by ~99% in a 5-day carcinoma9. In this study, we explore the use of a Multiple Low culture assay – henceforth referred as High Dose (HD). We found Dose (MLD) strategy in EGFR mutant NSCLC. In this approach, that 3D and 4D combinations inhibit cell proliferation and induce multiple drugs that act in the same oncogenic signalling pathway apoptosis at comparable levels to cells treated with HD of gefi- are combined at low concentration. We hypothesised that this tinib (Fig. 2a and Supplementary Fig. 2a, b). The level of pathway might add up to complete pathway inhibition without causing inhibition is also similar between cells treated with 3D and 4D prohibitive toxicity. Further, by using low drug concentrations, combinations and HD of gefitinib (Fig. 2d). In addition, we the pressure exerted on each node of the pathway should greatly performed RNA-Seq transcriptome analyses in cells treated with diminish, reducing the selective pressure on each node and 4D combination (Supplementary Fig. 2c, d). These data showed therefore diminishing the chances of acquiring resistance. that 4D combo treated cells displayed a significant down- regulation of MYC and E2F target genes, as well as cell cycle genes. Moreover, MAPK activity markers15 were significantly Results downregulated and several pro-apoptotic genes were found to be MLD therapy blocks MAPK pathway and proliferation in PC9 upregulated, while anti-apoptotic genes were downregulated. To cells. The mechanisms of resistance to EGFR inhibition (stan- study how MLD therapy compares to HD therapy regarding dard-of-care) in EGFR mutant NSCLC are well understood. We resistance, we treated PC9 cells with 3D or 4D combinations and therefore compared the efficacy of MLD therapy to standard-of- with HD of gefitinib or osimertinib for one month (Fig. 2b). As care MTD therapies in this indication. We used PC9 NSCLC cells, seen by others previously16,17, cells treated with HD of gefitinib or which harbour an activating mutation in the gene encoding osimertinib quickly developed resistance, but the cells treated EGFR10. We used four drugs, each inhibiting a different node in with 3D or 4D combinations did not. In addition, we treated PC9 the MAPK pathway: gefitinib (EGFR inhibitor), LY3009120 (pan- cells for 16 days with high dose of gefitinib or with 3D or with 4D RAF inhibitor11), trametinib (MEK inhibitor) and SCH772984 MLD combinations; we then either removed the drugs, continued (ERK inhibitor12), as shown schematically in Fig. 1a. We estab- to treat with the original drug, or treated with 4D MLD combi- lished dose-response curves for each of the four drugs using 5-day nation for another 16 days (Supplementary Fig. 2e). We observed culture assays (Fig. 1b). From these data, we inferred for all 4 resistant colonies after 32 days of gefitinib treatment, but not in inhibitors the IC20 dose, i.e., a drug concentration that inhibits the cells treated with 3D or 4D combinations. Apparently, 16 day- cell viability by 20%—henceforth referred as Low Dose (LD). To treatment with 3D or 4D combinations had killed all cells, as assess the efficacy of the MLD strategy we then tested the impact continued culturing for another 16 days in media without drugs of all possible drug combinations of the 4 drugs at LD on cell did not yield any colonies. Importantly, PC9 cells that had viability (assessed by CellTiter-Blue® assay), on cell proliferation developed resistance to high dose EGFR inhibitor, were still (assessed by long-term colony formation assay) and on pathway responsive to 4D MLD combination. This striking result indicates activity (measured by p-RSK levels13 using Western Blotting) that EGFR inhibitor-resistant cells remain sensitive to 3D and 4D (Fig. 1c-e). The expected viability and the synergy scores were combinations. This suggests that MLD therapy might be an calculated using the Bliss independence model14. We found that option for patients having developed resistance to standard-of- PC9 cells treated with the single drugs at low dose were only care EGFR inhibitor therapy. minimally affected, as expected. However, some of the drug combinations showed a striking combination effect, much higher than expected based on drug additivity. In particular, the com- MLD therapy is effective in EGFRi-resistant PC9 cells. To study bination of RAF + MEK + ERK inhibitors at low dose (hence- further if EGFRi-resistant cells are indeed sensitive to 3D and 4D forth called 3D combination) and the combination of EGFR + combinations, we generated PC9 cells resistant to clinically-used RAF + MEK + ERK inhibitors at low dose (henceforth called 4D EGFR inhibitors. We cultured PC9 cells in the presence of gefi- combination) showed an almost complete inhibition of cell via- tinib (PC9-GR) or osimertinib (PC9-OR) until cells were no bility and proliferation, along with a complete blockade of MAPK longer responsive to the inhibitors (see methods). We performed pathway signalling. Due to these notable findings we pursued the exome sequencing of the two resistant cell populations to gain MLD study focusing on the 3D and 4D combinations. To address insight into the mechanisms of acquired resistance. These data 2 NATURE COMMUNICATIONS | (2020)11:3157 | https://doi.org/10.1038/s41467-020-16952-9 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16952-9 ARTICLE a Plate cells & add drugs Assess cell viability & low dose Test drug combinations c 1.00 Viability (relative to DMSO) EGFR 1 2 3 4 EGFRi RAFi MEKi ERKi 100 0.75 Gefitinib 80 Low dose (LD) 1 2 1 3 1 4 Cell viability (%) RAS 60 2 3 2 4 3 4 0.50 RAF LY3009120 40 1 2 3 1 2 4 MEK Trametinib 20 0.25 0 1 3 4 2 3 4 ERK SCH772984 0.001 0.01 0.1 1 Drug concentration [μm] 0.00 0.00 RSK 1 2 3 4 –0.25 b Score EGFRI (gefitinib) RAFi (LY3009120) 120 120 Observed –0.50 Expected (bliss independence) 100 100 Synergy (bliss independence) Cell viability (%) 80 IC20 ≈ 7 nm Cell viability (%) 80 IC20 ≈ 250 nm –0.75 4 2 FRi SO 2 1+ 4 3 4 1+ 4 1+ 3 4 3 2+ 4 4 EF D)] K i )] )] 2+ 4 3 AFi D)] 1+ 1+ 2+ 2+ 2+ 3+ 2+ 3+ 1+ 1+ 3+ 3+ R LD D 60 60 G DM [M (L (L [R (L [E i ( 40 40 [E 1 20 20 0 0.001 0.01 0.1 1 0 0.01 0.1 1 10 d )] LD )] )] )] Drug concentration, μM LD D Drug concentration, μM LD i( (L FR i( 4 i( 3+ Ki EF AF 4 SO 4 4 3 G 3+ R 3+ 2+ 2+ 2+ [E 3 [M [R 4 4 [E 4 2 M 3 2+ 2+ 1+ 2+ 1+ 1+ 1+ 3+ 1+ 1+ 1+ D 1 2 3 4 MEKi (trametinib) ERKi (SCH772984) 120 120 100 100 Cell viability (%) Cell viability (%) 80 IC20 ≈ 8 nm 80 IC20 ≈ 250 nm e )] LD R D)] )] EF )] D LD i( L 1+ i (L 60 60 FR i( 4 i( 3+ K AF 4 SO 4 4 3 G 3+ 3+ 2+ 2+ 2+ [E 3 [M [R 4 4 [E 4 2 M 3 2+ 2+ 1+ 2+ 1+ 1+ 1+ 3+ 1+ 1+ 40 40 D 1 2 3 4 RSK 20 20 0 0 pRSK 0.001 0.01 0.1 1 0.01 0.1 1 10 Drug concentration, μM Tubulin Drug concentration, μM Fig. 1 Multiple low dose therapy blocks MAPK pathway and proliferation in PC9 cells. a Schematic of the Multiple Low Dose (MLD) efficacy determination. After plating, cells are treated with increasing drug concentrations. Four days later cell viability is measured and the low dose (LD) is assessed. At last, the efficacy of all the possible combinations at LD is determined. b Dose-response curves of EGFR, RAF, MEK and ERK inhibitors in PC9 cells. PC9 cells were cultured with increasing concentrations of EGFRi Gefitinib, RAFi LY3009120, MEKi Trametinib or ERKi SCH772984 for 4 days, after which cell viability was measured using CellTiter-Blue®. Standard deviation (SD) from 3 biologically independent replicates (each with 3 technical replicates) is plotted. Low doses (IC20s) were then determined: gefitinib=7 nM, LY3009120 = 250 nM, 292 trametinib = 8 nM and SCH772984 = 250 nM. c–e Determination of the efficacy of all the possible combinations of EGFR, RAF, MEK and ERK inhibitors at LD in PC9 cells. PC9 cells were cultured with all possible drug combinations of EGFR, RAF, MEK and ERK inhibitors at the low doses determined in b. In c cell viability from 3 biologically independent replicates (each with at least 3 technical replicates) was measured by CellTiter-Blue® assay after 4 days of treatment; In black the observed experimental viability; In dark-grey the expected viability and in light-grey the synergy scores, calculated using the Bliss independence model, are plotted. In d cells were treated for 10 days, after which plates were stained and scanned; A representative image from the 3 biologically independent replicates performed is displayed. In e protein for western blotting was harvested after 24 h of treatment; The level of pathway inhibition was determined by examining pRSK protein levels in the western blot. Tubulin was used as loading control. showed acquisition of the well-known T790M mutation in the amplified breast cancer, in which 4D combination consisted of PC9-GR cells and a number of mutations in the PC9-OR cells, HER2, PI3K, AKT and mTOR inhibitors). When treated with 4D none of which has been previously associated with resistance to combination, proliferation of all cell lines was inhibited, regard- osimertinib (Supplementary Table 2). We then tested the sensi- less of the tumour type/driver/genotype, pointing towards a broad tivity of the resistant lines to 3D and 4D combinations. In both applicability of the MLD treatment strategy (Supplementary resistant cell populations, we saw an almost complete inhibition Fig. 3d). of cell viability after only 4 days of MLD therapy treatment and a complete MAPK pathway signalling blockade (Fig. 2c, d). MLD therapy is tolerated by non-tumorigenic cell lines. One of the major concerns when using multiple drugs in combination is MLD therapy is effective in multiple tumour models. We then the possible toxicity to normal tissues18. To test the effect of the tested if the MLD strategy would also be effective in additional MLD strategy on “normal” (non-tumorigenic) cell lines we used in vitro tumour models. After low dose determination (Supple- primary human BJ (fibroblast) and RPE1 (retinal pigment epi- mentary Fig. 3a-c and Supplementary Table 1) we tested the MLD thelium) cells. Upon 3D and 4D MLD drug combination treat- strategy in patient-derived (colorectal and NSCLC) organoids. ment, cell viability was reduced, but to a much lesser extent than Treatment with 3D and 4D combinations resulted in a major in cancer cells. This indicates that the MLD strategy might be reduction in cell viability (Fig. 3a). In addition, we tested 6 dif- tolerated by normal tissues (Fig. 3b). Since the MAPK pathway is ferent MAPK pathway addicted cell lines: HCC827 and H3255 rich in cross-talk and feedback control circuits19,20, we also tested (EGFR mutant lung cancer), H2228 and H3122 (EML4-ALK how a pulse of signalling through the EGFR pathway would be translocated lung cancer, in which EGFRi was replaced with ALK affected by 3D or 4D MLD treatment. We serum-starved BJ cells inhibitor crizotinib in the 4D combination), DiFi and Lim1215 overnight and then incubated with 3D or 4D MLD drug com- (EGFR dependent colorectal cancer) and in 2 different PI3K binations for 2 h. After this, cells were stimulated with 100 ng/mL pathway addicted cell lines: SKBR3 and HCC1954 (HER2 of EGF in the presence of 3D or 4D drug combinations. Twenty NATURE COMMUNICATIONS | (2020)11:3157 | https://doi.org/10.1038/s41467-020-16952-9 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16952-9 a 100 Start of treatment Treatment refresh 0.20 Start of treatment d ) ) ) 4D om (HD 4D om (HD 4D om (HD DMSO 3D erti ) 3D erti ) 3D erti ) m (LD m (LD m (LD C ib C ib C ib bo bo bo SO o SO o bo n n n M b M b 80 Gefitinib (LD) Caspase 3/7 [AU] O nib O nib O nib D om D om om G O 0.15 Confluence (%) iti iti iti S C C C M LY3009120 (LD) ef ef ef si si si D G G + 60 0.10 Trametinib (LD) EGFR 40 SCH772984 (LD) 0.05 3D pEGFR 20 4D 0 0.00 Gefitinib (HD) 0 24 48 72 96 120 0 24 48 72 96 ERK Time (h) Time (h) pERK b ) ) n D n D ) ) tio tio RSK D D (H (L ) D (L (L na ) na D b 84 (H 0 ib ni bi (L bi 12 29 in ti om ib om ib er et in SO 09 77 in C m it C am it 30 H pRSK ef si M 3D ef 4D SC G O Tr LY D G Vinculin Tubulin 7 days culture 1 month culture PC9 PC9OR PC9GR c PC9 PC9-OR PC9-GR 120 120 120 100 100 100 80 80 Cell viability (%) Cell viability (%) 80 Cell viability (%) 60 60 60 40 40 40 20 20 20 10 10 10 5 5 5 0 0 0 ) ) SO ) SO 4 4 ) ) 4 4 ) ) ) ) ) ) SO ) 4 4 ) D ) D ) D D D D D D D D D D 3+ 3+ D D 3+ 3+ D 3+ 3+ (L am 0 (L (H (H (L (L (L (L (L (L (L (H (L (L (L M M M 2+ 2+ 2+ 2+ 2+ 2+ 20 84 D b 84 D ib b ib 20 ib b 84 D b ib b 2 ni 1+ ni 1+ ni ni 1+ in in ni tin 91 91 in 29 29 91 29 iti iti iti iti et et iti et er ef ef 00 00 ef 77 77 ef 00 ef 77 am am m G G G G 3 G 3 H H 3 H si LY LY Tr Tr 1. 1. LY Tr SC SC 1. SC O 3. 3. 3. 2. 2. 2. 4. 4. 4. Fig. 2 MLD therapy minimises therapeutic resistance and is effective in EGFRi-resistant PC9 cells. a MLD therapy abrogates cell proliferation and induces apoptosis in PC9 cells. PC9 cells were plated and incubated overnight to allow attachment to the plate. Cells were then treated with DMSO, with EGFR, RAF, MEK, ERK inhibitors at low dose, with 3D Combo (RAF + MEK + ERK inhibitors at LD) or with 4D Combo (EGFR + RAF + MEK + ERK inhibitors at LD) and placed in the IncuCyte®. Confluence (left) and caspase 3/7 activation (right) over time was measured by the IncuCyte®. Standard error of the mean (SEM) from 3 replicates is plotted. b MLD therapy prevents the acquisition of drug resistance in PC9 cells. PC9 cells were cultured with DMSO, with EGFR, RAF, MEK and ERK inhibitors at low dose (for 7 days) and with high dose (HD) of Osimertinib (200 nM), HD of Gefitinib (280 nM) and with 3D and 4D Combinations (for 1 month), after which plates were stained and scanned; A representative image from 3 biologically independent replicates is displayed. c EGFRi-resistant PC9 cells remain sensitive to MLD therapy. PC9, PC9-OR (Osimertinib-resistant) and PC9-GR (Gefitinib- resistant) cells (see methods) were cultured with DMSO, with low doses of EGFR, RAF, MEK or ERK inhibitors, with 3D or 4D combinations or with HD of Gefitinib or Osimertinib for 4 days, after which cell viability was measured using CellTiter-Blue®. Standard deviation (SD) from 3 biologically independent replicates is plotted. d MLD therapy blocks MAPK pathway in EGFRi-resistant PC9 cells. PC9, PC9-OR and PC9-GR cells were cultured with DMSO, HD of Osimertinib, HD of Gefitinib or with 3D or 4D combinations. Protein for western blotting was harvested after 24 h of treatment; The level of pathway inhibition was measured by examining pERK and pRSK protein levels and the level of EGFR inhibition was measured by examining pEGFR protein levels in the western blot. Tubulin and Vinculin were used as loading control. minutes after EGF stimulation, a significant amount of p-RSK lives) of the published maximum tolerated dose (MTD) in mice was detected, which was no longer detected at 4 h post EGF sti- for each of the individual drugs11,12,23,24. Osimertinib-resistant mulation (Fig. 4c). These data suggest that the efficient inhibition PDX-1 was implanted subcutaneously and orthotopically in the of MAPK signalling exerted by 3D and 4D MLD treatment is the lungs. In both models, treatment with 3D or 4D combination result of an effect of these drugs on homoeostatic feedback/cross- resulted in a reduction in tumour volume, without associated talk signalling19,21, as pulsatile signalling through the MAPK toxicity (Fig. 4a-d). Interestingly however, treatment with 4D pathway seems to be much less affected than persistent signalling combo was slightly more effective than 3D combo. Due to this through an oncogene-activated MAPK pathway. finding we focused the in vivo studies that followed on the 4D combination. In all PDX models tested we observed similar results to PDX-1, i.e., a reduction in tumour volume, without MLD therapy induces tumour regression without toxicity significant toxicity (Fig. 4e, f and Supplementary Fig. 4e). In in vivo. To address if the MLD strategy is effective in vivo, we addition, in gefitinib-resistant models PDX-2 and PDX-3 we used patient derived xenograft (PDX) tumours from four differ- tested if it would be possible to acquire resistance to the 4D MLD ent patients who had developed resistance to first-line or second- combination therapy during a drug holiday. In both PDX models, line therapy with EGFR inhibitors erlotinib or osimertinib22 in re-starting of 4D MLD therapy after a drug holiday resulted in a the clinic by acquiring EGFR T790M mutation, KRAS mutation second response to the drug combination, indicating that overt or MET amplification (Supplementary Table 3). For the in vivo resistance had not developed in vivo (Fig. 4e, f). studies we defined LD as 20% (for gefitinib and trametinib) and We also implanted PC9 cells in nude mice and treated them 30% (for LY3009120 and SCH772984—due to the shorter half- with vehicle, with EGFR, RAF, MEK and ERK inhibitors 4 NATURE COMMUNICATIONS | (2020)11:3157 | https://doi.org/10.1038/s41467-020-16952-9 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16952-9 ARTICLE a b TUM09 BJ RPE1 120 120 120 100 100 100 Cell viability (%) Cell viability (%) Cell viability (%) 80 80 80 60 60 60 40 40 40 Colorectal cancer organoids 20 20 20 0 0 0 C o C o bo bo C ) C ) Ki ) Ki ) ) ) ) ) EK D) EK D) FR D) FR D) K ) K ) EK D) EK D) FR O FR O D D ER (HD ER (HD R (LD R (LD R (HD R (HD ER LD ER (LD 4D mb 4D mb EG MS EG MS om om 3D (H 3D (H H H EG i (L EG i (L L L TUM10 i( i( i( i( i( o o i i i D D i i i i AF AF AF AF M M M M 120 100 c No EGF EGF (100 ng/mL) Cell viability (%) 80 0H 20 min 4H 60 UN S 3D 4D UN 1 2 3 4 S 3D 4D UN 1 2 3 4 S 3D 4D 40 EGFR 20 pEGFR 0 ITO034T ERK 120 pERK 100 Lung cancer organoid Cell viability (%) 80 RSK 60 pRSK 40 20 Tubulin 0 UN = Untreated 2 = RAFi (LD) 4 = ERKi (LD) 3D = 2+3+4 bo bo ) ) SO ) ) 1 = EGFRi (LD) 3 = MEKi (LD) S = ERKi (HD) 4D = 1+2+3+4 D LD LD LD om om (L M i( i( i( Ki D C C FR EK AF ER 4D 3D EG R M Fig. 3 MLD therapy is effective in patient-derived organoids and is tolerated by normal cell lines. a MLD therapy is effective in several colorectal and lung cancer patient-derived organoids. Organoids were cultured with DMSO, with EGFR, RAF, MEK and ERK inhibitors at LD and with 3D and 4D combos. After 5 days of drug treatment cell viability was measured using CellTiter-Glo®. Standard deviation (SD) from 3 biologically independent replicates is plotted. b Cell viability of normal cells is much less affected by MLD therapy than tumour cells. BJ and RPE1 cells were treated with DMSO, with EGFR, RAF, MEK and ERK inhibitors at low and high doses and with 3D and 4D Combos (using the LD and HD concentrations determined for PC9 cells). After 4 days of drug treatment cell viability was measured. SD from 3 replicates is plotted. c MLD therapy allows pulsed signaling in normal cells. BJ cells, after overnight starvation, were treated with the indicated inhibitors/concentrations for 2 h, after which EGF (100 ng/mL) was added. Cells were harvested before, 20 min and 4 h after EGF stimulation. The level of pathway inhibition was measured by examining pERK and pRSK protein levels. The level of EGFR inhibition was measured by examining pEGFR protein levels in the western blot. Tubulin was used as loading control. individually at low dose and with 4D combination. The use of low outgrowth of resistant cells, unlike the mice treated with 4D dose regimens was inadequate to suppress PC9 tumour growth combination, where a sustained tumour control was observed when used as single agents, but when used in combination we (Supplementary Fig. 4d). observed a sustained reduction in the tumour volume of PC9 Despite the significant tumour regressions observed in the xenografts over a period of 70 days, which was associated with an in vivo experiments none of the mice were fully cured, unlike in extended survival (Supplementary Fig. 4a, b). These observations the in vitro data where all the cells were killed by the 3D or 4D are also supported by immunohistochemical staining of the combinations. To study why this is the case we studied the tumours, which show decreased Ki67 (a proliferation marker) pharmacokinetics and pharmacodynamics of the four drugs and pERK (MAPK activation) levels in the tumours treated with in vivo over time. We found that drug plasma concentrations of 4D combination (Fig. 4g). Significantly, mice treated with 4D gefitinib and trametinib dropped relatively slowly (T1/2 8 h), but combination did not show any significant signs of toxicity, the pan-RAF and ERK inhibitors were less stable in plasma (T1/2 assessed by the weight of the mice over time and by the of 5 and 4 h, respectively). A similar difference was seen for intra- morphology of the GI tract and bone marrow (Supplementary tumoural drug concentrations (Supplementary Figs. 5a, b). Figs. 4c, f). In the clinic, the T790M mutation is already present Consistent with this, we observed a complete inhibition of pRSK (at very low percentages) in the majority of the tumours before in tumour biopsies 2 h after 4D combination drug administration, undergoing anti-EGFR treatment25,26. To mimic this scenario, we which progressively decreased after 8 and 24 h (Supplementary implanted in nude mice a mix of PC9 cells and PC9-GR cells Fig. 5c). These data indicate that, unlike in the in vitro (which are T790M positive) in a 9:1 ratio, respectively. Mice were experiments, two of the four drugs were not present at a treated with vehicle, with MTD of gefitinib and with 4D significant concentration during at least 12 h of the 24-h combination. Treatment with MTD of gefitinib resulted in a treatment cycle. As a result of this, a sustained MAPK pathway quick reduction of tumour volume which was followed by inhibition was not achieved in vivo, possibly explaining why we NATURE COMMUNICATIONS | (2020)11:3157 | https://doi.org/10.1038/s41467-020-16952-9 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16952-9 a PDX-1 (SC) c PDX-1 (ORT) e PDX-2 (SC) 400 1400 Tumor size (mm3) at sacrifice 600 1200 500 Stopped treatment Tumor volume (%) Tumor volume (%) 300 1000 400 800 200 600 300 Drug holiday 400 200 100 200 100 0 0 0 3 3 3 4 0 7 14 21 28 0 7 14 21 28 35 42 49 56 63 70 77 84 91 = = = = ,N N ,N ,N Days in treatment Days in treatment ib le bo bo tin ic om om h er Ve Vehicle, N = 4 3D Combo, N = 5 Vehicle, N = 3 4D Combo (group A), N = 3 m C C si 3D 4D O Osimertinib (5 mg/kg), N = 4 4D Combo, N = 6 Gefitinib (80 mg/kg), N = 4 4D Combo (group B), N = 3 b PDX-1 (SC) d PDX-1 (SC) f PDX-3 (SC) 150 150 400 125 125 Tumor volume (%) 300 100 Weight (%) 100 Weight (%) Drug holiday 75 75 200 50 50 100 25 25 0 0 0 0 7 14 21 28 0 7 14 21 28 0 7 14 21 28 35 42 49 56 63 70 Days in treatment Days in treatment Days in treatment Vehicle, N = 4 3D Combo, N = 5 Vehicle, N = 3 3D Combo, N = 3 Vehicle, N = 4 4D Combo, N = 5 Osimertinib (5 mg/kg), N = 4 4D Combo, N = 6 Osimertinib (5 mg/kg), N = 4 4D Combo, N = 4 Gefitinib (80 mg/kg), N = 5 4D Combo DH, N = 4 g Gefitinib LY3009120 Trametinib SCH772984 h MLD therapeutic strategy MTD therapeutic strategy Vehicle (10 mg/kg) (6 mg/kg) (0.5 mg/kg) (15 mg/kg) 4D Combo H&E EGFR EGFR EGFR EGFRi EGFRi EGFRi RAS RAS RAS RAF RAFi RAF Time RAF Ki67 MEK MEKi MEK MEK ERK ERKi ERK ERK pERK Sustained pathway inhibition Pathway inhibition Pathway reactivation Apoptosis Proliferation inhibition Resist cell death Resistance prevention Apoptosis Drug resistance Fig. 4 MLD therapy induces tumour regression without toxicity in vivo. a–f Patient derived xenografts (PDX) are sensitive to MLD therapy. PDX tumours (see Supplementary Table 3) were implanted subcutaneously (a, e, f) or orthotopically in the lungs (c) of Crl:NU-Foxn1nu mice. PDX1 was implanted both subcutaneously (a) and orthotopically in the lungs (c). We defined the in vivo LD as 20–30% of the MTD for each of the individual drugs–gefitinib (10 mg/kg), LY3009120 (6 mg/kg), trametinib (0.5 mg/kg) and SCH772984 (15 mg/kg). In a–d after tumour establishment, mice were treated 5 days/ week with vehicle, with osimertinib (5 mg/kg) and with 3D or with 4D Combos for 26 days, after which mice were sacrificed. In a tumour volume percentages ±SEM is shown, in c tumour size (mm3) at sacrifice ±SEM is shown and in b and d the mice weight percentages ± SEM is shown. e After tumour establishment, mice were treated 5 days/week with Vehicle (N = 3), with gefitinib (80 mg/kg) (N = 4) or with 4D Combo for 6 weeks (group A, N = 3) or with 4D Combo for 8 weeks (group B, N = 3). Mice treated with vehicle and gefitinib were sacrificed when tumours reached ~2000 mm3. After 6 weeks, Group A was taken off treatment and mice were sacrificed when tumours reached ~2000 mm3. After 8 weeks Group B was taken off treatment and was given 3 weeks of drug holiday. Mice were then treated for another 2 weeks with 4D combo, after which they were sacrificed. Tumour volume percentages ±SEM is shown. f After tumour establishment, mice were treated 5 days/week with vehicle (N = 4), with gefitinib (80 mg/kg) (N = 5) or with 4D Combo (N = 9) for 4 weeks, after which mice were sacrificed, except for 4 animals from the 4D Combo group. These 4 mice were spared and were given 3 weeks drug holiday (4D Combo DH group), followed by another 3 weeks of treatment, after which they were sacrificed. Tumour volume percentages ±SEM is shown. g H&E, Ki67 and pERK stainings from tumour sections of PC9 xenografts. A representative staining image from each cohort (N ≥ 6) is displayed. Scale bars 100 µm. h Schematic representation of the MLD therapy for the treatment of EGFR mutant NSCLC. didn’t achieve full tumour regressions. We tested this hypothesis there is not any apparent drug-drug interaction, except for in vitro, by removing RAF and ERK inhibitors from the treatment LY3009120, for which the half-life increases when given in the for approximately 8 h every day. We found that, as hypothesised, 4D Combo. when the drugs in the 3D or 4D combination were not present continuously the MLD therapy became less effective (Supple- mentary Fig. 5d). Finally, we tested whether there was any drug- Discussion drug interaction by measuring the half-lives of the drugs when We report here that treatment of EGFR mutant NSCLCs with given alone or in combination (Supplementary Fig. 5e). Overall, MLD therapy effectively suppresses development of drug 6 NATURE COMMUNICATIONS | (2020)11:3157 | https://doi.org/10.1038/s41467-020-16952-9 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16952-9 ARTICLE resistance, without associated toxicity. As such, our data challenge without major toxicity, as judged by lack of weight loss and lack the common paradigm that patients should be treated with the of toxicity to gut epithelium and bone marrow. However, mice MTD of a targeted agent. Our data are consistent with a model in and human are fundamentally different with respect to drug which diffuse inhibition of an oncogenic pathway at multiple toxicity. Especially skin toxicity following MAPK inhibition nodes reduces selective pressure on each of the nodes to mutate therapy is often underestimated in mice. Therefore, only a phase I and thereby increase response time (Fig. 4h). Our findings also clinical trial will be able to fully assess the toxicity of this strategy challenge the current model for MAPK pathway signalling, which in humans. postulates that the MAPK kinase cascade functions to amplify Extrapolation of dose from animals to humans based only on signals. Such amplification cascade model is clearly at odds with mg/kg conversion is difficult, since body surface area and dif- the data obtained here in which a very partial inhibition of each ferences in pharmacokinetics should also be taken into con- of 4 nodes in this cascade adds up to complete pathway inhibi- sideration. To convert the animal dose in mg/kg to human tion. Further mechanistic studies are required to better under- equivalent doses (HED) in mg/kg, it is recommended to divide by stand the efficacy of the MLD strategy. 12.329. If we estimate the HED based on the low-doses used in Importantly, we show that tumours having the most common our in vivo experiments for Gefitinib and Trametinib (where mechanisms of clinically-observed resistance to high dose stan- dosing in humans is known) using this approach then Gefitinib dard of care EGFR inhibitors still respond to MLD therapy. (10 mg/kg in mice) corresponds to 57 mg once daily in patients, Therefore, MLD treatment strategy appears especially promising which is approximately one quarter of the dose used in patients for patients that have already developed resistance to all clinically (250 mg qd). And Trametinib (0.5 mg/kg in mice) corresponds to used EGFR inhibitors, including osimertinib. In such resistant 2,8 mg once daily in patients, which is a bit higher than the dose tumours, multiple metastases may be present having different used in patients (2 mg qd). However, we also we performed an resistance mutations. In this study we have shown that MLD in vivo experiment using lower concentrations of Gefitinib (1 mg/ therapy is effective in PDX models having diverse EGFR inhibitor kg) and Trametinib (0.1 mg/kg) (Supplementary Fig. 4d). These resistance mechanisms, including EGFR T790M mutation, KRAS drug concentrations correspond to 2,5% of the human daily mutation, MET amplification and even SCLC transformation, dose for gefitinib and 28% of the daily human dose of trametinib, highlighting that MLD therapy could apply to a diverse range of using the calculation method of Nair mentioned above. These EGFR TKI resistant tumours. However, not all the resistance data indicate that with these further reduced concentrations of mechanisms have been tested and it is possible that some might Gefitinib and Trametinib we still have a significant anti-tumor not respond to MLD therapy. Indeed, in clinical practice, an MLD effect in vivo. Due to the difficulty in translating drug doses from treatment strategy can only be tested in patients having developed mice to human we feel that only a well-designed phase 1 trial can resistance to standard-of-care EGFR inhibitors. We find in PDX help assess the potential clinical utility of the MLD strategy models that 4D MLD is consistently somewhat better than 3D proposed here. MLD, which may relate to the notion that not all EGFR alleles in Even though we focused mostly on EGFR mutant NSCLC, we the tumour may have acquired resistance mutations to the EGFR have also shown that the MLD strategy can potentially be effec- inhibitor therapy. Furthermore, it will be important to maintain tive in other tumour types. Overall, our findings challenge the osimertinib in an experimental MLD therapy trial, as this drug current paradigm of using the maximum tolerated dose of single crosses the blood-brain barrier, and such late-stage patients may targeted cancer drugs and suggest that, instead, it might be have (latent) brain metastases. We therefore suggest that clinical more beneficial to use a combination of multiple drugs that target testing of the MLD strategy should include osimertinib. the oncogenically activated pathway using sub-optimal drug While we never observed development of resistance to MLD concentrations. therapy in vivo, even after long drug exposure, we did not achieve complete tumour regressions. This is most likely due to the short Methods half-lives of the RAF and ERK inhibitors used in this study, which Cell lines culture and drug-response assays. The PC9 cell line was obtained resulted in a situation in which we did not achieve a continuous from ATCC. PC9OR (osimertinib-resistant) and PC9GR (gefitinib-resistant) cells pathway blockade. This may be improved by using continuous were made by continuous (2 months) drug exposure of PC9 cells to 1 µM osi- mertinib (AZD9291) and to 2 µM gefitinib, respectively. Exome sequencing was release formulations of these drugs, or by using drugs with longer performed to determine if any de novo genetic alterations had occurred (Supple- half-lives. mentary Table 2). The HCC827, H3255, H3122, H2228, SKBR3, HCC1954, BJ and The MLD therapy described here is fundamentally different RPE1 cell lines were obtained from ATCC. And DiFi and Lim1215 cell lines were a from metronomic chemotherapy27,28. In this latter scenario, low gift from A. Bardelli (Torino, Italy). BJ and RPE1 cells were cultured in DMEM (Gibco 41966029). SKBR3 and HCC1954 which were cultured in DMEM/F-12 doses of chemotherapy are given at high frequency with the aim medium (Gibco 31331028). All the other cell lines were cultured in RPMI medium to suppress division of endothelial cells of the tumour vasculature. (Gibco 21875034). All the cell lines media were supplemented with 10% FBS In the present MLD schedule, we target the MAPK pathway of (Serana), 1% penicillin/streptomycin (Gibco 15140122) and 2 mM L-glutamine the tumour itself, as growth inhibition in all cases parallels (Gibco 25030024). All cell lines were cultured at 37 °C and with 5% CO2. All cell lines were validated by STR profiling and mycoplasma tests were performed every inhibition of the MAPK pathway (as judged by pRSK). Three- 2–3 months. drug combinations given at MTD have been used before in pre- All drug-response assays were performed in triplicate, using black-walled 384- clinical8 and clinical studies6,7 for BRAFV600E mutant tumours, well plates (Greiner 781091). Cells were plated at the optimal seeding density showing clear therapeutic benefits, but such regimen have an (Supplementary Table 1) and incubated for approximately 24 h to allow attachment to the plate. Drugs were then added to the plates using the Tecan associated cost of toxicity. D300e digital dispenser. 10 µM phenylarsine oxide was used as positive control (0% The lack of significant toxicity of the MLD therapy in mice cell viability) and DMSO was used as negative control (100% cell viability). Four may be explained by the fundamentally different nature of MAPK days later, culture medium was removed and CellTiter-Blue (Promega G8081) was pathway signalling between normal and EGFR mutant cancer added to the plates. After 1–4 h incubation, measurements were performed cells. In the former, signalling is transiently activated when according to manufacturer’s instructions using the EnVision (Perkin Elmer). growth factors are present. In the latter, oncogenic mutations result in persistent activation of the pathway. Importantly, we Organoid culture and drug-response assays. Colorectal (CRC) and non-small cell lung cancer (NSCLC) organoids were established and handled as previously show here that transient signalling in normal cells is, at least described30. All drug-response assays were performed in replicate, each by inde- initially, not interrupted by MLD treatment (Fig. 3c). This may pendent researchers. PDOs were mechanically and enzymatically dissociated into explain why long-term exposure of mice to MLD treatment is single cells, pipetted through a 40 µM cell strainer, and re-plated to allow for NATURE COMMUNICATIONS | (2020)11:3157 | https://doi.org/10.1038/s41467-020-16952-9 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16952-9 organoids formation. At day 4 PDOs were collected, Cultrex was removed by Mouse xenografts studies. All animal experiments were approved by the Animal incubation of the cell pellet in 1 mg/mL dispase II (Sigma D4693) for 15 min. Ethics Committee of the Netherlands Cancer Institute or by the Animal Ethics Whole organoids were counted using a hemocytometer and trypan blue. PDOs Committee of the Institut Català d’Oncologia and performed in accordance with were resuspended in 1:3 Advanced Dulbecco’s Modified Eagles Medium with institutional, national and European guidelines for Animal Care and Use. Nutrient Mixture F-12 Hams (Ad-DF) (Invitrogen 12634), supplemented with 1% PC9 cell line xenografts: One million PC9 cells were resuspended in PBS and penicillin/streptomycin (Invitrogen 15140122), 1% HEPES (Invitrogen 15630056) mixed 1:1 with matrigel (Corning 354230). Cells were injected subcutaneously into and 1% GlutaMAX (Invitrogen 35050) (Ad-DF+++):Cultrex at a concentration of the posterior flanks of 7-week-old immunodeficient BALB/cAnNRj-Foxn1nu mice 20 organoids/µL. Five µl/well was dispensed in clear-bottomed, white-walled 96- (half male and half female; Janvier Laboratories, The Netherlands). Tumour well plates (Greiner Bio-One 655098) and overlaid with 200 µL CRC or NSCLC formation was monitored twice a week. Tumour volume, based on calliper culture medium. We generated 10-step dose response curves using the Tecan measurements, was calculated by the modified ellipsoidal formula (tumour volume D300e digital dispenser, interpolated IC20 values and re-screened organoids in = 1/2(length × width2)). When tumours reached a volume of approximately 200 presence of a range of concentration around the IC20 of each drug separately and in mm3, mice were randomized into the indicated treatment arms. Vehicle, gefitinib, 3D and 4D Combos. In addition, we re-performed the dose-response curves to LY3009120, trametinib, SCH772984 or the combination of the 4 inhibitors were control for variation between experiments. Read-out was performed at day 10 in formulated in DMSO: Kolliphor EL (Sigma 27963): Saline solution, in a ratio of the positive control (10 µM phenylarsine oxide), negative control (DMSO), and the (1:1:8). Mice were treated 5 days a week (Monday to Friday) at the indicated doses drug-treated wells. Quantification of cell viability was done by replacing the by intraperitoneal injection. CRC medium with 50 µL Cell-TiterGlo 3D (Promega G9681) mixed with 50 µL Patient-derived xenografts (PDX) and orthotopic xenograft (PDOX): Primary Ad-F+++. Measurements were performed according to manufacturer’s instruc- tumours were obtained from Bellvitge Hospital (HUB) and the Catalan Institute of tions on an Infinite 200 Pro plate reader (Tecan Life Sciences) with an integration Oncology (ICO) with approval by the Ethical Committee. Ethical and legal time of 100 ms. protection guidelines of human subjects, including informed consent from the patient to implant the tumour in mice, were followed. PDX-1 was generated from a lung adenocarcinoma biopsy from a patient who was treated with Erlotinib (first Compounds, reagents and antibodies. Gefitinib (100140), LY3009120 (206161), line), Gefitinib + Capmatinib (second line) and Cisplatin+Pemetrexed (third line). trametinib (201458), SCH772984 (406578), osimertinib (206426), crizotinib This tumour has an EGFR mutation (del19) and MET amplification. PDX-2 was (202222), lapatinib (100946), BKM120 (204690), MK2206 (201913) and AZD8055 generated from a lung adenocarcinoma biopsy from a patient who was treated with (200312) were purchased from MedKoo Biosciences. Erlotinib (S7786), BGB-283 Erlotinib (first line), Gefitinib + Capmatinib (second line) and Carboplatin (S7926), selumetinib (S1008) and LY-3214996 (S8534) were purchased from Sell- +Gemcitabine and Nivolumab (third line). This tumour has an EGFR mutation eckchem. Annexin V-FITC Apoptosis Staining Detection Kit was purchased from (L858R) and MET amplification. PDX-330 was generated from a lung Abcam (ab14085). adenocarcinoma biopsy of a brain metastasis from a patient who was treated with Antibodies against Tubulin (T9026) and Vinculin (V9131) were purchased Erlotinib (first line) and Osimertinib (second line). PDX-4 was generated from a lung from Sigma; antibodies against EGFR (4267), pERK (4377), ERK (9102) and RSK adenocarcinoma biopsy from a patient who was treated with Afatinib (first line) and (8408) were purchased from Cell Signalling; antibody against pRSK (04-419) was CBDCA + pemetrexed (second line). This tumour has a germline p53 mutation and purchased from Millipore; antibody against pEGFR (ab5644) was purchased from an EGFR mutation (del19). Tumours were isolated and implanted subcutaneously Abcam. Secondary antibodies Goat Anti-Rabbit IgG (H + L)-HRP Conjugate (or orthotopically, in the lungs, in the case of PDX-3) into Crl:NU-Foxn1nu mice by (1706515) and Goat Anti-Mouse IgG (H + L)-HRP Conjugate (1706516) were following previously reported procedures22,31. In the subcutaneous models, tumour purchased from Bio Rad. volume was monitored twice a week by a digital caliper. When tumours reached a volume of approximately 200–600 mm3, mice were randomized into the indicated treatment arms. In the orthotopic model, tumours were left to grow for 2 weeks, followed by 26 days of treatment. Vehicle, gefitinib, osimertinib or the 3D and 4D Colony formation and IncuCyte cell proliferation assays. Cells were seeded in Combos were formulated in DMSO: Kolliphor EL (Sigma 27963): Saline solution, in the appropriate density (Supplementary Table 1) in 6-well plates. Cells were a ratio of (1:1:8). Mice were treated 5 days a week (Monday to Friday) at the incubated for approximately 24 h to allow attachment to the plates, after which indicated doses by intraperitoneal injection. drugs were added to the cells using the Tecan D300e digital dispenser as indicated. The culture media/drugs were refreshed every 3/4 days. When control wells (DMSO) were confluent (unless otherwise stated in the text) cells were fixed using a In vivo pharmacokinetics and pharmacodynamics studies. Plasma and tumour solution of 2% formaldehyde (Millipore 104002) diluted in phosphate-buffered samples were assayed by liquid chromatography triple quadrupole mass spectro- saline (PBS). Two hours later, they were stained, using a solution of 0.1% crystal metry (LC-MS/MS) using an API4000 detector (Sciex) for the simultaneous violet (Sigma HT90132) diluted in water. Not more than 10 min later the staining determination of Gefitinib (MRM: 447.4/128.1), LY3009120 MRM: 425.5/324.2), solution was removed, plates were washed with water left to dry overnight. Finally, Trametinib (616.3/491.2) and SCH772984 (MRM: 588.4/320.2). Gefitinib-d8 plates were scanned and stored. (MRM: 455.4/136.3) was used as internal standard. LC separation was achieved For IncuCyte proliferation assays, cells were seeded in 96-well plates and using a Zorbax Extend C18 column (100 × 2.0 mm; ID). Mobile phase A and B incubated overnight to allow attachment to the plates. Drugs were added to the comprised 0.1% formic acid in water and methanol, respectively. The flow rate was cells using the Tecan D300e digital dispenser. Cells were imaged every 4 h in the 0.4 ml/min and a linear gradient from 20%B to 95%B in 2.5 min, followed by 95%B IncuCyte ZOOM (Essen Bioscience). Phase-contrast images were collected and for 2 min, followed by re-equilibration at 20%B for 10 min was used for elution. analysed to detect cell proliferation based on cell confluence. For cell apoptosis, Sample pre-treatment was accomplished by mixing 5 μL (plasma) or 25 μL (tumour IncuCyte® Caspase-3/7 green apoptosis assay reagent (Essen Bioscience 4440) was homogenate) with 30 or 150 μL, respectively, of formic acid in acetonitrile (1 + 99) also added to culture medium and cell apoptosis was analysed based on green containing the internal standard. After centrifugation, the clear supernatant was fluorescent staining of apoptotic cells. diluted 1 + 4 with water and 50 μL was injected into the LC-MS/MS system. The plasma/tumour samples were harvested at the time points indicated in Supplementary fig. 5. Blood samples were obtained by tail cut (at 2 h and 8 h time points) and by cardiac puncture at the 24 h time point. Samples were collected on Western Blots. After the indicated culture period, cells were washed with chilled ice in tubes containing potassium EDTA as anticoagulant. The tubes were PBS and then lysed with RIPA buffer (25 mM Tris-HCl pH 7.6, 150 mM NaCl, 1% immediately cooled in melting ice and centrifuged (10 min, 5000×g, 4 °C) to NP-40, 1% sodium deoxycholate, 0.1% SDS) containing protease inhibitors separate the plasma fraction, which was transferred into clean vials. For the (Complete (Roche) and phosphatase inhibitor cocktails II and III). Samples were tumours samples, the mice were sacrificed by cervical dislocation, the tumour was then centrifuged for 10 min at 14.000 rpm at 4 °C and supernatant was collected. dissected and frozen at −80 °C. Half of the tumour was then lysed mechanically Protein concentration of the samples was normalised after performing a Bicinch- with RIPA buffer and lysates were analysed by Western blot. The other half was oninic Acid (BCA) assay (Pierce BCA, Thermo Scientific), according to the weighed and homogenised in 1 mL of ice-cold 1% of BSA in water and stored at manufacturer’s instructions. −20 °C until further analysis. Protein samples (denatured with DTT followed by 5 min heating at 95 °C) were then loaded in a 4–12% polyacrylamide gel. Gels were run (SDS-PAGE) for approximately 60 min at 165 V. Proteins were then transferred from the gel to a Reporting summary. Further information on research design is available in the polyvinylidene fluoride (PVDF) membrane, using 330 mA for 90 min. Nature Research Reporting Summary linked to this article. After the transfer, membranes were placed in blocking solution (5% bovine serum albumin (BSA) in PBS with 0,1% Tween-20 (PBS-T). Subsequently, Data availability membranes were probed with primary antibody in blocking solution (1:1000) and All relevant data are available from the corresponding author upon request. Full scans of left shaking overnight at 4 °C. Membranes were then washed 3 times for 10 min the gels and blots are available in Supplementary Fig. 7. All the other data supporting the with PBS-T, followed by 1 h incubation at room temperature with the secondary findings of this study are available within the article and its supplementary information antibody (HRP conjugated, 1:10000) in blocking solution. Membranes were again files. A reporting summary for this article is available as a Supplementary Information washed 3 times for 10 min in PBS-T. Finally, a chemiluminescence substrate (ECL, file. RNAseq data can be accessed with the GEO assession GSE144258. Most raw data can Bio-Rad) was added to the membranes and the Western Blot was resolved using be assessed at https://doi.org/10.6084/m9.figshare.12408803.v1 the ChemiDoc (Bio-Rad). 8 NATURE COMMUNICATIONS | (2020)11:3157 | https://doi.org/10.1038/s41467-020-16952-9 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16952-9 ARTICLE Received: 28 October 2019; Accepted: 4 June 2020; (del19) mutations and treated with first-generation TKIs. Clin. Cancer Res. 25, 4280–4289 (2019). 27. Kerbel, R. S. & Kamen, B. A. The anti-angiogenic basis of metronomic chemotherapy. Nat. Rev. Cancer 4, 423–436 (2004). 28. Scharovsky, O. G., Mainetti, L. E. & Rozados, V. R. Metronomic chemotherapy: changing the paradigm that more is better. Curr. Oncol. 16, References 7–15 (2009). 1. Weinstein, I. B. Addiction to oncogenes-the Achilles heal of cancer. Science 29. Nair, A. B. & Jacob, S. A simple practice guide for dose conversion between 297, 63–64 (2002). animals and human. J. Basic Clin. Pharm. 7, 27–31 (2016). 2. Konieczkowski, D. J., Johannessen, C. M. & Garraway, L. A. A convergence- 30. Dijkstra, K. K. et al. Generation of rumor-reactive T cells by co-culture of based framework for cancer drug resistance. Cancer Cell 33, 801–815 (2018). peripheral blood lymphocytes and tumor organoids. Cell 174, 1586–1598.e12 3. Sun, C. & Bernards, R. Feedback and redundancy in receptor tyrosine kinase (2018). signaling: relevance to cancer therapies. Trends Biochem. Sci. 39, 465–474 31. Ambrogio, C. et al. Combined inhibition of DDR1 and Notch signaling is a (2014). therapeutic strategy for KRAS-driven lung adenocarcinoma. Nat. Med. 22, 4. Flaherty, K. T. et al. Combined BRAF and MEK inhibition in melanoma with 270–277 (2016). BRAF V600 mutations. N. Engl. J. Med. 367, 1694–1703 (2012). 5. Khunger, A., Khunger, M. & Velcheti, V. Dabrafenib in combination with trametinib in the treatment of patients with BRAF V600-positive advanced or Acknowledgements We thank A. Bardelli (Torino, Italy) for gift of cell lines, the facilities of The Neth- metastatic non-small cell lung cancer: clinical evidence and experience. Ther. erlands Cancer Institute: Mouse Clinic–Intervention Unice (Natalie Proost, Bjorn Adv. Respir. Dis. 12, 1753466618767611 (2018). Siteur, Bas van Gerwen, Charlotte Baak, Renske Grimmerink, Rebecca Theeuwsen and 6. Kopetz, S. et al. Encorafenib, binimetinib, and cetuximab in BRAF V600E- Marieke van de Ven), Robotics and Screening Center (Ben Morris and Roderick Bei- Mutated colorectal cancer. N. Engl. J. Med. https://doi.org/10.1056/ jersbergen), Clinical Pharmacology (Levi Buil and Artur Burylo), Experimental Animal NEJMoa1908075 (2019). Pathology, Flow Cytometry and Sequencing. We also thank Richard Marais for dis- 7. Corcoran, R. B. et al. Combined BRAF, EGFR, and MEK inhibition in patients cussions. This work was supported by a grant from the Dutch Cancer Society through with BRAF(V600E)-Mutant colorectal cancer. Cancer Discov. 8, 428–443 the Oncode Institute. Al.V. was supported by the Fondo de Investigaciones Sanitarias, (2018). FIS (PI16-01898, and by the Spanish Association Against Cancer, AECC 8. Xue, Y. et al. An approach to suppress the evolution of resistance in BRAF (CGB14142035THOM) and Ideas Semilla project (IDEAS098VILL-IDEAS16) and (V600E)-mutant cancer. Nat. Med. 23, 929–937 (2017). Generalitat de Catalunya (2014SGR364). L.F. received a European Union’s Horizon 9. Caumanns, J. J. et al. Low-dose triple drug combination targeting the PI3K/ 2020 research and innovation programme under the Marie Sklodowska-Curie, grant AKT/mTOR pathway and the MAPK pathway is an effective approach in agreement number 799850. E.N. was funded by Instituto Carlos III through the project ovarian clear cell carcinoma. Cancer Lett. 461, 102–111 (2019). PI18/00920. We thank CERCA Program/Generalitat de Catalunya for their institu- 10. Sharma, S. V. et al. A chromatin-mediated reversible drug-tolerant state in tional support and grant 2017SGR448. cancer cell subpopulations. Cell 141, 69–80 (2010). 11. Peng, S.-B. et al. Inhibition of RAF isoforms and active dimers by LY3009120 leads to anti-tumor activities in RAS or BRAF mutant cancers. Cancer Cell 28, Author contributions 384–398 (2015). R.B. supervised all research. J.N., Al.V., O.v.T. and R.B., wrote the manuscript. J.N., E.N., 12. Morris, E. J. et al. Discovery of a novel ERK inhibitor with activity in models S.O., E.B., L.F., C.M., S.K., A.J., H.H., L.W. and Au.V. designed, performed and analysed of acquired resistance to BRAF and MEK inhibitors. Cancer Discov. 3, the experiments. E.V., Al.V., E.F., A.M.-M. and Lo.W. provided advice for the project. All 742–750 (2013). authors commented on the manuscript. 13. Anjum, R. & Blenis, J. The RSK family of kinases: emerging roles in cellular signalling. Nat. Rev. Mol. Cell Biol. 9, 747–758 (2008). 14. Russ, D. & Kishony, R. Additivity of inhibitory effects in multidrug Competing interests combinations. Nat. Microbiol 3, 1339–1345 (2018). Al.V. is co-founder of Xenopat S.L. The remaining authors declare no competing 15. Brant, R. et al. Clinically viable gene expression assays with potential for interests. predicting benefit from MEK inhibitors. Clin. Cancer Res. 23, 1471–1480 (2017). Additional information 16. Hata, A. N. et al. Tumor cells can follow distinct evolutionary paths to become Supplementary information is available for this paper at https://doi.org/10.1038/s41467- resistant to epidermal growth factor receptor inhibition. Nat. Med. 22, 020-16952-9. 262–269 (2016). 17. Eberlein, C. A. et al. Acquired resistance to the mutant-selective EGFR Correspondence and requests for materials should be addressed to R.B. inhibitor AZD9291 is associated with increased dependence on RAS signaling in preclinical models. Cancer Res. 75, 2489–2500 (2015). Peer review information Nature Communications thanks Keith Flaherty, Nicola 18. Park, S. R., Davis, M., Doroshow, J. H. & Kummar, S. Safety and feasibility of Normanno and the other anonymous reviewer(s) for their contribution to the peer targeted agent combinations in solid tumours. Nat. Rev. Clin. Oncol. 10, review of this work. Peer reviewer reports are available. 154–168 (2013). 19. Mendoza, M. C., Er, E. E. & Blenis, J. The Ras-ERK and PI3K-mTOR Reprints and permission information is available at http://www.nature.com/reprints pathways: cross-talk and compensation. Trends Biochem. Sci. 36, 320–328 (2011). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 20. Fey, D., Croucher, D., Kolch, W. & Kholodenko, B. Crosstalk and signaling published maps and institutional affiliations. switches in mitogen-activated protein kinase cascades. Front. Physiol. 3, 355 (2012). 21. Lake, D., Correa, S. A. L. & Muller, J. Negative feedback regulation of the Open Access This article is licensed under a Creative Commons ERK1/2 MAPK pathway. Cell. Mol. Life Sci. 73, 4397–4413 (2016). 22. Martinez-Marti, A. et al. Dual MET and ERBB inhibition overcomes Attribution 4.0 International License, which permits use, sharing, intratumor plasticity in osimertinib-resistant-advanced non-small-cell lung adaptation, distribution and reproduction in any medium or format, as long as you give cancer (NSCLC). Ann. Oncol. 28, 2451–2457 (2017). appropriate credit to the original author(s) and the source, provide a link to the Creative 23. Hayakawa, H. et al. Lower gefitinib dose led to earlier resistance acquisition Commons license, and indicate if changes were made. The images or other third party before emergence of T790M mutation in epidermal growth factor receptor- material in this article are included in the article’s Creative Commons license, unless mutated lung cancer model. Cancer Sci. 104, 1440–1446 (2013). indicated otherwise in a credit line to the material. If material is not included in the 24. Tricker, E. M. et al. Combined EGFR/MEK inhibition prevents the emergence article’s Creative Commons license and your intended use is not permitted by statutory of resistance in EGFR-mutant lung cancer. Cancer Discov. 5, 960–971 (2015). regulation or exceeds the permitted use, you will need to obtain permission directly from 25. Ichihara, E. et al. Clinical significance of repeat rebiopsy in detecting the EGFR the copyright holder. To view a copy of this license, visit http://creativecommons.org/ T790M secondary mutation in patients with non-small cell lung cancer. licenses/by/4.0/. Oncotarget 9, 29525–29531 (2018). 26. Vendrell, J. A. et al. Ultra-sensitive EGFR (T790M) detection as an independent prognostic marker for lung cancer patients harboring EGFR © The Author(s) 2020 NATURE COMMUNICATIONS | (2020)11:3157 | https://doi.org/10.1038/s41467-020-16952-9 | www.nature.com/naturecommunications 9