Cell lines (derived from the American Type Culture Collection (ATCC)) were cultured in DMEM (Life Technologies, cat. 41965039) supplemented with 10% FBS (Life Technologies, cat. 10270106), 20 mM L-glutamine (Life Technologies, cat. 25030024), 10 U/ml penicillin and 10 μg/ml streptomycin (Life Technologies, cat. 15140122) and MEM Non-essential amino acids solution (100x, Life Technologies, Cat. No. 11140035), and maintained at 37°C under 5% CO₂. Cell line identities were regularly confirmed by short tandem repeat (STR) profiling using the PowerPlex16 system and GeneMapper software (Promega) and cell lines were regularly screened for mycoplasma.

Cells were seeded in 384-well plates (Corning, cat. 3764) at t=0, using the seeding densities shown in Table 1 . At t=24 h trametinib or DMSO was added to a final concentration of 1 μM. Library compounds were added at final concentrations of 10 nM, 100 nM and 1 μM with appropriate solvent controls using the Sciclone ALH 3000 liquid handling robot. At t=96 h cell viability was determined using an MTT assay. The area under the curve was calculated using Graphpad Prism 6.

The Sequoia anti-neoplastic library contains 157 drugs used in cancer treatment and was purchased from Sequoia Research Products. The SCREEN-WELL^® epigenetics library contains 43 epigenetic compounds and was acquired from Enzo Life Sciences. Other targeted inhibitors used in this study were purchased from Selleck Chemicals.

Samples were harvested in CHAPS buffer and subjected to co-immunoprecipitation (co-IP) using Dynabeads magnetic protein A agarose beads (Life Technologies) according to the manufacturer’s protocol. Immunoprecipitation antibodies were crosslinked to the beads using the Pierce BS3 crosslinker (ThermoFisher, cat. 21580). 1 mg of protein lysate was used as input material for co-IP reactions. Antibodies were acquired from Cell Signaling Technology unless noted otherwise. Antibodies were used for MEK (#8727), ERK (#9102), pMEK (#9154), pERK (#4376), BCL-2 (#4223), BCL-XL (#2764), MCL-1 (#4572), BIM (#2933), α-tubulin (#3873) and β-actin (Abcam, Ab6276).

For cell viability assays, 5–20 × 10³ cells were seeded in 96-well plates, one day before treatment. Trametinib to a final concentration of 1 µM or DMSO was added to the cells and subsequently the BCL-2 family inhibitors were added in a seven-point fivefold dilution series. Cell viability was assayed by an MTT assay after 72 h (10). All experiments were performed in triplicate and values were compared to solvent-treated controls.

Bliss independence scores were calculated to measure synergy between trametinib and each of the tested concentrations of BCL-2 family inhibitors (17). Using the highest bliss score observed for each cell line, we subsequently calculated the average highest bliss score for the RAS-, NF1-, ALK, and non-mutated cell lines, respectively.

NMB cells expressing the doxycycline-inducible mutant NRAS Q61V were generated as previously described (10).

All experiments were conducted after obtaining ethical approval from the Dutch animal ethics committee of the Amsterdam UMC (SKNAS, DAG217) and the Netherlands Cancer Institute (KP-N-YN, AVD30100202011584, EGP10056). Single-cell suspensions containing 5 x 10⁶ cells of SKNAS or KP-N-YN neuroblastoma cell lines were generated in matrigel/DPBS (1:1) and transplanted subcutaneously into female NMRI nude mice (Charles River Laboratories (SKNAS), Janvier (KP-N-YN), 5-7 weeks old). Once the engrafted tumors reached a size of approximately 250 mm³, mice were randomly assigned to one of the following treatment arms (1): vehicle trametinib + vehicle venetoclax/navitoclax, (2) venetoclax + vehicle trametinib, (3) navitoclax + vehicle trametinib, (4) trametinib + vehicle venetoclax/navitoclax, (5) trametinib + venetoclax, and (6) trametinib + navitoclax. Trametinib (MedChem Express) was dissolved in 0.5% hydroxypropylmethylcellulose (Sigma) with 0.2% Tween-80 (Sigma) in MQ and dosed at 1 mg/kg/day via oral gavage. Venetoclax and navitoclax (MedChem Express) were dissolved in a vehicle consisting of 10% ethanol (Boom), 30% PEG400 (Sigma) and 60% Phosal (Lipoid). Both BCL-2 inhibitors were dosed at 100 mg/kg/day via oral gavage. Six to nine mice per treatment group received daily treatment during a period of 21 days or until humane endpoints were reached. In addition, two to three mice per group were treated for seven days and sacrificed 4 h after the last dose to harvest tumors for pharmacodynamic analysis. The number of mice per treatment arm is further specified in Supplementary Table 1 . Tumor size was monitored two to three times a week, in a blinded manner. Tumor burden was determined according to the formula (π/6)d³, where d represents the mean tumor diameter obtained by caliper measurement. Statistical analysis was performed using linear mixed effects modeling followed by a pairwise analysis to compare the effects of single compound treatment to their respective combination therapies, with asterisks indicating a p-value <0.05.

Neuroblastoma cell lines that are treated with an increasing concentration of trametinib do not show complete cell death, but instead reach a plateau phase resulting from complete growth inhibition ( Supplementary Figure 1 ). To identify compounds that are effective in combination with MEK inhibitors, we performed a large-scale drug screen. Four neuroblastoma cell lines were included based on the presence of the most common mutations affecting the RAS-MAPK pathway, RAS (KRAS in SJNB8 and NRAS in SKNAS), NF1 (SKNBE) and ALK (KCNR). In total, 210 compounds were selected: 60 chemotherapeutic compounds, 106 targeted drugs that were either approved for clinical use or far along in (pre)clinical development, and 44 epigenetic modifiers ( Supplementary Table 2 ). Cell lines were screened with 10 nM, 100 nM and 1 µM of the library compounds with and without 1 µM trametinib. Cell viability was normalized to solvent-treated cells in the DMSO assays and to the trametinib-treated cells in the trametinib assays. The area under the curve (AUC) for all compounds was determined in both assays and subsequently normalized to the AUC for DMSO-treated cells ( Supplementary Figure 2 ). Most compounds were similarly effective in the DMSO- and trametinib-treated cells as evidenced by an average slope of 1.0 (range 0.95 tot 1.13) calculated by linear regression over all points per cell line (data not shown). As a negative control, we determined the effect of combining trametinib with other MEK inhibitors. In cells treated with trametinib, increased inhibition of RAS-MAPK signaling by the MEK inhibitors present in the library, showed no additional effect ( Figure 1A , Supplementary Figure 3 ). Contrastingly, PI3K inhibitors are known to work synergistically with MEK inhibitors (18–20) and indeed had a larger effect when combined with trametinib. This shows that our methodology allows detection of synergistic and antagonistic effects.

To rank the compounds in their effectiveness, we calculated the average difference in AUC in the DMSO and trametinib assay over the four cell lines ( Figure 1A , Supplementary Table 3 ). The most effective compound in combination with trametinib was navitoclax. Navitoclax is an inhibitor of the interaction between BIM and anti-apoptotic BCL-2 family members, showing strong affinity for BCL-2, BCL-xL and BCL-W (21). Its analog compound venetoclax, which specifically targets BCL-2 (22), showed a similar effect in the SKNAS and KCNR cell lines ( Supplementary Figure 3 ). Interestingly, both navitoclax and venetoclax have previously shown effectivity in neuroblastoma cell lines and in vivo models (23–25). The combination with MEK inhibitors has also already been described as a synergistic combination in several other cancer types (26–28).

To confirm that this sensitizing effect also occurs in neuroblastoma cells with mutations in the RAS-MAPK pathway, we first evaluated the effect of trametinib in an extended panel of cell lines. Trametinib monotherapy inhibited cell growth more effectively in cell lines with RAS-MAPK mutations ( Supplementary Figure 4 ). Exposing the RAS-MAPK-mutated cell lines to the combination of trametinib and navitoclax led to a clear sensitizing effect of trametinib on navitoclax sensitivity, while in non-mutated cell lines this was less pronounced ( Figure 1B , Supplementary Figure 5 ). This suggests that the beneficial combined effect of MEK and BCL-2 family inhibition is dependent on an active RAS-MAPK pathway.

MEK inhibition is expected to decrease phosphorylated ERK (pERK) and increase feedback loop-mediated upregulation of phosphorylated MEK (pMEK) (29, 30). This decrease in pERK was indeed observed upon treatment with trametinib, proving that the compound worked on-target in our set of neuroblastoma cell lines ( Figure 1C ). The expected increase in pMEK was also detected in most cell lines, except for NRAS-mutant SKNAS cells and wild-type NMB cells ( Figure 1C ). Despite the observed results, treating RAS-MAPK-mutated neuroblastoma cells with trametinib had mainly a cytostatic effect, as evidenced by the plateau phase at higher concentrations ( Supplementary Figure 1 ). Combining trametinib with navitoclax is expected to enhance the apoptotic response. In line with this, largest increases in cleaved caspase 3 and cleaved PARP were generally observed upon combination of both compounds ( Figure 1C ). Next to testing these apoptotic markers, we also calculated the bliss independence scores as a measure for synergy. The average highest bliss score was 0.33 for the RAS-mutated lines, 0.15 for the NF1-mutated lines and 0.34 for the ALK-mutated lines. Cells without RAS-MAPK mutations, did not respond to the combination, as shown by the lack of cleaved caspase 3 and cleaved PARP in the NMB cells and an average highest bliss score of 0.08 for the wild-type cell lines ( Figure 1C , Supplementary Figure 5 ). This further supports that the beneficial effect of combining trametinib and navitoclax is more pronounced in cells with overactivation of the RAS-MAPK pathway.

In cells with an active RAS-MAPK pathway, pERK is known to phosphorylate the pro-apoptotic protein BIM. Phosphorylated BIM (pBIM) is unstable and will be targeted for degradation in the proteasome, leading to an overall decrease in total BIM levels ( Supplementary Figure 6 (31)). To test whether this effect can be counteracted by MEK inhibition and could play a role in the synergy observed between trametinib and navitoclax, we studied how trametinib affects BIM levels. Trametinib treatment induced a clear increase in BIM protein levels in cell lines with mutations in RAS, NF1 or ALK ( Figure 2A ). In line with Figure 1C , these results suggest that trametinib is capable of inhibiting pERK, thereby enhancing stabile non-phosphorylated BIM levels.

To validate whether the observed effect is dependent on an active RAS-MAPK pathway, we expressed a doxycycline-inducible mutant NRAS Q61V protein in the cell line NMB. NMB is a cell line without mutations in the RAS-MAPK pathway, causing this line to normally be insensitive to trametinib treatment. In addition, wild-type NMB cells show slight sensitivity to navitoclax treatment ( Figure 2C , Supplementary Figure 5 ). Activation of the RAS-MAPK pathway by the addition of doxycycline, leads to a decrease in BIM levels ( Figure 2B ). As expected, this causes the NRAS-mutant NMB cells to become resistant to navitoclax treatment ( Figure 2C ). This effect could be abrogated by adding trametinib, indicating that the observed effects are indeed mediated by the RAS-MAPK pathway and not by other pathways downstream of RAS.

To further dissect whether trametinib would also sensitize cells towards other BH3 mimetics, we exposed our cell line panel to venetoclax, a specific BCL-2 inhibitor, and S63845, a specific MCL-1 inhibitor. In both cases, sensitivity was tested in the absence and presence of trametinib.

Of the six cell lines with mutations in the RAS-MAPK pathway, SKNAS and KCNR were the only lines that were sensitized to venetoclax by trametinib treatment ( Figure 3A , Supplementary Figure 7 ). Subsequent co-IP validation studies revealed that BIM is bound to BCL-2 in SKNAS and KCNR ( Figure 3B ), thereby explaining why synergistic effects were only observed in these lines. In KCNR, BIM is bound to BCL-2 irrespective of the presence of trametinib, while in SKNAS BIM is only bound to BCL-2 when trametinib is present.

Most RAS-MAPK-mutated lines were also sensitized to S63845 after trametinib treatment ( Supplementary Figure 8A , Supplementary Figure 9 ). Co-IP of BIM in SJNB8 also shows an increase in MCL-1 bound to BIM when treated with trametinib ( Supplementary Figure 8B ). Overall, this suggests that increased BIM binding to anti-apoptotic proteins could be a general mechanism for the sensitizing effect of MEK inhibitors to various BH3 mimetics.

Since navitoclax and venetoclax were both already in clinical development, we continued studying the effects of these two compounds in combination with trametinib. To validate the identified combinations in vivo, we first treated SKNAS xenografts with the combination of trametinib and navitoclax. The SKNAS model was expected to be sensitive to trametinib treatment due to the presence of an NRAS Q61V mutation. Over a treatment of 21 days, trametinib indeed appeared to cause tumor growth inhibition compared to vehicle-treated xenografts ( Figure 4A ), although significance was not reached (p=0.08). Navitoclax monotherapy had no clear effect on tumor growth in SKNAS xenografts, which was in line with our in vitro results ( Supplementary Figure 5 ). Yet, combination with trametinib contributed to significant tumor regression ( Figure 4A , Supplementary Figure 10A ), with 6/6 mice showing PR as best response (32) ( Supplementary Table 4 ). This indicates that the combination could hold potential for neuroblastoma tumors with RAS mutations.

Western blot analysis showed that the beneficial effect of the combination therapy can also be detected on a protein level. In vivo navitoclax treatment caused an increase in cleaved caspase 3, while trametinib decreased ERK phosphorylation ( Figure 4B ). Upon combination of navitoclax and trametinib, both effects were observed together. These responses are in line with our in vitro results, confirming that trametinib worked on-target and that the combination enhances apoptosis.

The same experimental set-up was used to test the combination of trametinib with the more specific BCL-2 inhibitor venetoclax. The combination induced tumor growth stabilization in the first stage of treatment, after which tumors slowly started growing again. Over time, trametinib-venetoclax combination-treated tumors were significantly smaller compared to tumors treated with venetoclax alone ( Figure 4A ), but best responses (2/6 PR, 3/6 SD, 1/6 PD (32)) were less powerful compared to tumors treated with the trametinib-navitoclax combination ( Supplementary Table 4 ).

To further examine the applicability of these therapies in neuroblastoma with other RAS-MAPK mutations, both combinations were also tested in high BCL-2-expressing and NF1-mutant KP-N-YN cells. In vitro analysis of KP-N-YN showed a consistent response to trametinib monotherapy ( Supplementary Figure 11A ) and synergistic effects on cell viability after combination with either navitoclax (highest bliss score: 0.36) or venetoclax (highest bliss score: 0.40) ( Supplementary Figure 11B ). However, these in vitro results were not representative of the effects observed in vivo. Although in vivo trends were similar to the SKNAS model, results were less pronounced (trametinib + navitoclax: 4/9 PR, 3/9 SD, 2/9 PD and trametinib + venetoclax 3/8 PR, 2/8 SD, 3/8 PD) ( Figure 4A , Supplementary Figure 10B , Supplementary Table 4 ). Significance was only reached between the navitoclax and trametinib-navitoclax treatment arms ( Figure 4A ).