Temozolomide (Sigma-Aldrich, catalog number T2577), trametinib (MedChem, catalog number HY-10999/CS-0060), epidermal growth factor human (MedChem, catalog number HY-P7109), PD184352 (Sigma-Aldrich, catalog number PZ0181), DMEM low glucose (Gibco, catalog number 31600034), DMEM high glucose (Gibco, catalog number 1210046), DMEM without phenol red (Gibco, catalog number 11054001), fetal bovine serum (Gibco, catalog number 12657029), penicillin streptomycin (Gibco, catalog number 15140122), MycoAlert™ Plus Buffer (Lonza, catalog number LT27-286), ERK-KTR lentiviral plasmid (pLentiPGK Puro DEST ERKKTRClover, Addgene #90227), Apple-53BP1trunc (Apple-53BP1trunc, Addgene #69531), polybrene (EMD Millipore, catalog number TR1003G), PEI (ethylenediamine branched) (Sigma-Aldrich, catalog number 408719), puromycin (Merck, catalog number P8833).

In this study, we used A172 glioblastoma cell line (ATCC, catalog number CRL-1620), U-251 MG glioma cell line (STR validated in October 2018 by the Banco de Células do Rio de Janeiro), U138 MG glioblastoma cell line (ATCC HTB-16), and MRC5 lung fibroblast cell line (ATCC, catalog number CCL-171). A172, U-251 MG, and U138 MG glioma cell lines do not carry EGFR amplification or EGFRvIII mutation or other mutations in the MAPK pathway [40]. U-251 MG carries PTEN and TP53 loss-of-function mutations [40]. All cell lines were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 2 mM L-glutamine, 100 U/ml penicillin, 100 mg/ml streptomycin, and 10% fetal bovine serum. Hek-293 (ATCC, catalog number CRL-1573) cell line was cultured in DMEM High glucose (4.5 g/L) supplemented with 2 mM L-glutamine, 100 U/ml penicillin, 100 mg/ml streptomycin, and 10% fetal bovine serum. Cells were maintained in an incubator with constant temperature (37 °C), CO2 (5%), and humidity. Cell cultures were frequently tested for Mycoplasma contamination using MycoAlert kit following manufacturer’s instruction.

A172, U-251 MG, and U138 MG glioma cells and MRC5 lung fibroblast cells stably expressing ERK-KTR fluorescent reporter [25] and/or the nuclear marker Apple-53BP1trunc [41] were generated via lentiviral transduction. Live cell measurement of ERK activity was performed on an IncuCyte S3 or SX1 device (Sartorius). Whenever indicated, cells were stimulated with EGF or treated with temozolomide (TMZ) or TRAM at specific doses. At least 1 h prior to imaging, media were changed to imaging media (DMEM without phenol red with 10% or 0.5% of FBS). Cells were imaged in green and red fluorescence every 10 min for 3 h with imaging starting 30 min prior to EGF stimulation or 3 h prior to TRAM and TMZ treatment (10× magnification). Neither the average of ERK activity nor the phenotypic space changed if using cells from image fields or the whole well (Fig. S1A). For colonies’ growth, plates were imaged every 24 h in red fluorescence. The number of live cells was daily manually determined for fitness and fractional killing quantification. The identity of the colony was determined by its location on the plate using ImageJ software [42].

A172 and MRC5 image analysis and ERK-KTR fluorescence quantification were determined using forNMA, an in-house pipeline written in Python (version 1.0, unpublished). For each cell, the Apple-53BP1 red fluorescence channel was used for nuclear segmentation, and a cytoplasmic ring was determined by increasing the nuclear area in 3 pixels. Mean fluorescence intensity values on the ERK-KTR green channel for each cell (nuclear and cytoplasmic ring) were extracted and used to calculate ratios, meaning that the ERK signaling activity was quantified through the relative cytoplasmic to nuclear fluorescence (C/N) ratio [25]. U-251 MG image analysis and ERK phenotype classification were conducted through visual assessment of ERK-KTR fluorescence by four independent evaluators. Each cell was categorized as very inactive, inactive, active, or very active based on its nuclear fluorescence (Fig. 1A), and the distribution of cells across these categories was subsequently used to quantify phenotypic heterogeneity using the Shannon Index.

Phenotypic space of ERK activity in glioblastoma cells. A ERK activity was measured by the cytoplasmic/nuclear (C/N) ratio of the green fluorescence of cells expressing ERK-KTR and 53BP1-Apple. B Four levels of ERK activity were defined as very inactive (VI), inactive (I), active (A), and very active (VA) based on the distributions of ERK activity of A172 cells grown in 10% FBS (orange line) or 0.5% FBS (blue line) for 48 h and C 20 nM trametinib for 2 h (green line) or 2.5 ng/ml of EGF for 15 min (red line) after 48 h of serum deprivation. Relative Shannon Index of 4 groups (rSI4 or SI4ERK) calculations are indicated. D Average ERK activity and E phenotypic space (rSI4) occupied by cells treated as in B and C. Each dot represents an image field with at least 20 cells (10% n = 2085 cells; 0.5% n = 4466 cells; TRAM n = 362 cells). F Representative image field of A172 glioma cells and I MRC5 fibroblast cells before (left) and after (right) 15 min of 2.5 ng/ml of EGF treatment (20× magnification). The numbers in the image represent the C/N ratio of each cell. The distribution of cells’ phenotypes and SI4ERK for this group of cells is shown in the graphs below the images. G Distribution of cells treated with EGF after 48 h of serum deprivation in A172 glioma cells and J MRC5 cells. H SI4ERK occupied by A172 glioma cells and K MRC5 cells after 15 min of EGF treatment. Each dot represents an image field with at least 20 cells (A172 n = at least 350 cells per EGF dose; MRC5 n = at least 186 cells per EGF dose) One-way ANOVA. EGF, epidermal growth factor; TRAM, trametinib; rSI4 or SI4ERK, relative Shannon Index of 4 groups of ERK activity); *p < 0.05; **p < 0.01; ***p < 0.001

For ERK activity measurements (average, phenotypes identification and SI4ERK), A172_ERK-53bp1, U-251 MG_ERK, and MRC5_ERK-53bp1 cell lines were seeded at a density of 10,000 cells onto a 6-well plate. The next day, media were exchanged to 10% or 0.5% FBS for 48 h prior to stimulation or imaging. Whenever indicated, cells were stimulated with EGF or treated with TRAM at specific doses. At least 1 h prior to imaging, the media were changed to imaging media (DMEM without phenol red with 10% or 0.5% FBS). For the phenotype identification and SI4ERK determination, cells were transferred to an IncuCyte S3or SX1 (Sartorius) and imaged in green and red fluorescence once at the end of specific treatments (TRAM; 10% or 0.5% FBS) (10× magnification). Each treatment condition had at least 6 image fields per well and at least 3 experimental replicates.

For the Intermitotic time (IMT), A172_ERK-53bp1 cells were imaged in green and red fluorescence every 30 min, and imaging tracking of individual cells was performed manually. IMT was calculated in minutes from the birth of a cell to its next division. Average ERK activity of the last 5 h prior to mitosis was considered ERK activity of the mother cell.

For dynamics analysis of average ERK activity and SI4ERK, cells were imaged in green and red fluorescence every 10 min for 3 h with imaging starting 30 min prior to EGF stimulation or 3 h prior to TMZ treatment (10× magnification). Each treatment condition had at least 6 image fields per well. EGF stimulation had 1 experimental replicate, and the TMZ treatment condition had 3 experimental replicates.

A172_ERK-53bp1 cells were plated onto 24-well plates at a density of 100,000 cells. Plates were kept on an IncuCyte device and imaged every 24 h in red fluorescence (4× magnification). Untreated wells were passaged every 3 days to keep the exponential growth of cells and each time the same number of cells was reseeded. The number of live cells was daily manually counted for PD determination as follows:

$$PD=(}_\left(fN\right)-}_\left(iN\right))$$

where fN is the number of cells in the well at the day of image, and iN is the number of cells in the well at the day before [36]. The cumulative values of PD (CPD = sum of all PDs) were plotted versus days of culture.

For colonies’ growth, a mix of A172 and A172_ERK-53bp1 cell lines was plated at a density of 200 cells (proportion of 198:2) in a 96-well plate. Plates were kept on an IncuCyte device and imaged every 24 h in red fluorescence (4× magnification). Single cells expressing the nuclear marker fluorescence were identified on day 1, and their growth to colonies was manually tracked for 14 days using images. The number of live cells was daily manually determined for fitness and fractional killing quantification, and the identity of the colony was determined by its location on the plate. TMZ (100 µM for 3 h), TRAM (20 nM for 24 h), or a combination of TMZ (100 µM for 3 h) and TRAM (20 nM for 24 h) treatment was added on day 4 of growth for treated colonies. Colonies that did not reach a minimum growth of 4 cells at day 4 were excluded from analysis. For ERK activity (average and SI4ERK) measurement, colonies were imaged in green and red fluorescence every 10 min for 3 h on day 4 (before treatment addition) and day 7 of growth (3 days after treatment removal) (10× magnification).

Fitness of colonies was determined by total cell number at a specific time point, colony size (log2 of cell number), or expressed as cumulative population doubling (CPD) [35]. For colony size and cumulative population doubling analysis, the value of 0.5 was used to denote the total elimination of cells to avoid undefined log values. The growth rate of colonies after treatment was calculated as:

$$GR=}_\left(CSa\right)-}_\left(CSb\right)/days$$

where CSa is the colony size after treatment, and CSb is the colony size before treatment [13].

Fractional killing of colonies after treatment was determined as inversely proportional to lethal fraction (LF) [38]. LF was calculated as LF = 1 − (Nalive cells/(Ndead cells + Nalive cells). The number of live cells in each colony was daily counted manually and used to infer the occurrence of death among cells (every time a colony had fewer alive cells than previously, the missing cells were considered dead).

The Shannon’s diversity index (H or SI) was used to quantify the distribution of cells across ERK activity levels and the distribution of colonies across lethal fraction (LF) induction groups. SI is calculated as:

$$H(SI_}} ) = - \sum\limits_^ }} } \log_}} p_}} }} ) \times 100$$

where pi is the proportion of cells in each ERK level/LF groups, and S is the number of groups. SI ranges from 0 to 100, implying that a value of 0 means that all cells are in the same group while a value of 100 means that cells are equally distributed across all groups.