Background

In the past few years, several new therapies have contributed to the treatment of various human cancers. In addition to the classical complex surgical, radio- and chemotherapy, the emergence of novel targeted [1, 2] and immunotherapies [3] resulted in longer progression-free and overall survival [3, 4]. In hormone-receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2−) breast cancer CDK4/6 inhibitors and mammalian target of rapamycin (mTOR) inhibitors are the most widely used targeted therapies, adding significant benefit to baseline endocrine therapy [4, 5].

A subset of patients receiving targeted therapies observe disease progression [6, 7]. Recent progress indicates that tumor heterogeneity and subclonal evolution can be key features contributing to drug resistance [8,9,10,11]. Following clonal expansion, acquired mutations in cancer cells give rise to different subclones, populations of distinct geno- and phenotypic characteristics and provide a basis for adaptive evolution of the tumor mass [8, 10]. In the case of selective pressure, resistant subclones can exhibit a relative proliferative advantage compared to sensitive cells, resulting in resistant cells becoming the predominant subclones, eventually overtaking the entirety of the tumor mass [8]. These resistant subclones can be therapy-induced (i.e. they have not been present as a population before the start of therapy); however, a growing body of evidence confirms that in several cases pre-existing resistant subclones are being selected for during the course of treatment [8, 10, 12,13,14].

Most current standard-of-care therapy regimens are altered only when chemoresistance renders the tumor mass unresponsive to the drug, resulting in progression or relapse [15,16,17]. Previously effective treatments lose their ability to control the tumor burden and because cross-resistance renders several secondary drug classes ineffective, efficacious second-line treatments can be difficult to find [17, 18]. Some of these resistance traits include rewiring key pro-proliferative pathways which can create acquired and targetable sensitivities [19].

Therapeutic approaches could benefit from taking into account evolutionary processes in cancer to develop new tools to postpone or overcome drug resistance. Adaptive therapy aims to exploit the changing proliferative advantage between resistant and sensitive cells. This approach succeeds when resistant cells are more fit compared to sensitive cells when drug pressure is on, while when no treatment is present sensitive cells are more fit [20,21,22]. Another approach in treating both sensitive and resistant cells without providing relative proliferative benefit to either cell type is the application of collateral sensitivity. Collateral sensitivity is the acquired vulnerability of a resistant cell against a second drug, which was not applied previously when resistance for the preceding drugs was generated [23, 24]. Exploiting collateral sensitivity aims to control the tumor burden through a combination of drugs by targeting sensitive cells with the standard-of-care primary drug while targeting the acquired sensitivities of resistant cells with a secondary drug [17, 23, 24]. Recent clinical trials targeting frequent resistance mechanisms up-front revealed a clear advantage over only blocking the primary target in EGFR-mutant non-small-cell lung cancer and BRAF-mutant melanoma [25, 26]. In addition to cancer treatments, collateral sensitivities of antibiotic-resistant bacteria are also highly sought after to propose novel, more effective antibacterial treatment regimens in an era of emerging antibiotic resistance [27].

Here, we developed a coculture system to study the effects of collateral sensitivity on the growth of spheroids containing cells sensitive and resistant to CDK4/6 inhibitors. Transcriptional profiling of sensitive and resistant cells revealed druggable acquired vulnerabilities of the resistant cells. By labeling the sensitive and resistant cells with different fluorescent proteins we were able to track their proliferation under drug pressure, mimicking the population dynamics of sensitive and resistant subclones. Our results show that coculture spheroids of sensitive and resistant cells under the selective pressure of ribociclib selects for the ribociclib-resistant cells. Comparing the transcriptional differences of sensitive and resistant cells revealed the downregulation of G2/M checkpoint in resistant cells, upon which collateral sensitivity against the Wee-1 inhibitor adavosertib (AZD1775) was confirmed in the resistant cells. The combination of ribociclib and adavosertib outperformed the antiproliferative effect of mono-treatments in the cocultures but not in the monocultures, while also decreasing the selection for resistant cells. Our results promote a phenotype-driven optimization of evolutionary antiproliferative therapy as a model for further assessment in the pre-clinical and clinical setting.

Methods

Cell lines and reagents

The estrogen-receptor-positive (ER+), HER2- CAMA-1 breast cancer cell line was maintained in DMEM+ 10% FBS+ 1% antibiotic–antimycotic solution. CAMA-1 cell line was authenticated by the ATCC cell line authentication service. Cells were continuously treated with ribociclib (Selleck Chemicals, Cat. No: S7440) at 1 µM concentration for 1 month followed by 250 nM for 4 months to develop resistance. Ribociclib-resistant CAMA-1 cells (CAMA-1_ribociclib_resistant) were further maintained in complete culture medium supplemented with 250 nM ribociclib. Resistance and collateral sensitivity against adavosertib (Selleck Chemicals, Cat. No.: S1525) were detected by the alteration of the dose–response curve measured using CellTiterGlo Chemoluminescent Kit (Promega Corporation, Cat. No.: G7573). Cell lines were confirmed to be mycoplasma-negative using the Mycoalert PLUS Mycoplasma detection kit (Lonza, Cat. No.: LT07-703).

Sequencing and bioinformatic analysis

Parental sensitive CAMA-1 and CAMA-1_ribociclib_resistant cell lines were plated 500,000 cells/well in a 6-well plate in triplicates. 24 h after plating 1 µM ribociclib or vehicle (dimethyl sulfoxide, DMSO) treatment were applied for 12 h, after which cells were trypsinized, washed and the pellet was frozen at − 80C for subsequent RNA isolation. RNA was isolated using the RNeasy Plus Mini Kit (Qiagen, Cat. No.: 74136) following the manufacturer’s protocol.

RNA-seq libraries were prepared using Illumina TruSeq Stranded Total RNA library Prep Ribo-zero Gold following manufacturer’s protocol. Libraries were sequenced with biological triplicates on an Illumina NovaSeq6000 instrument with 2 × 150 paired-end reads resulting in an average of 25 million reads per sample. Samples were aligned to the human reference genome (hg19) using the STAR (v2.7.0) aligner [28]. Transcripts were quantified by RSEM (v1.3.1) followed by differential expression analysis using DESeq2 (v1.26) and GSEA pathways analysis using TPM normalized values and the R packages GSVA (v1.30.0) [29,30,31]. Genes with at least a twofold change in expression with FDR < 0.05 were considered statistically significant. Signature scores were generated using the Molecular Signatures Database (v6) Hallmark signature sets. Pathway enrichments with global p-value < 0.05 and FDR < 0.25 were considered statistically significant. Differentially expressed genes were also subjected to pathway analysis regarding the Biocarta pathways using DAVID Bioinformatics Resources [32]. In this analysis an FDR-corrected p-value < 0.05 was considered statistically significant.

DNA was isolated from CAMA-1 and CAMA-1_ribociclib_resistant cells using DNeasy Blood & Tissue Kit (Qiagen, Cat. No.: 69504) according to the manufacturer’s protocol. For whole-exome sequencing, libraries were prepared using Agilent SureSelect XT human All Exon v7 following the manufacturer’s protocol. Libraries were sequenced on a NovaSeq6000 with 2 X150 paired-end reads to a sequencing depth of 285X (CAMA-1) and 300X (CAMA-1_ribociclib_resistant). Reads were trimmed with Trimmomatic prior to alignment to hg19 using BWA-mem [33, 34]. Genome Analysis Tool Kit (GATK) best practice guidelines were then followed including the use of Picard and Samtools for PCR duplicate removal and bam manipulation, and GATK for indel realignment and base recalibration [35, 36]. Variant calling was performed using an n − 1 consensus approach using three somatic variant callers: Mutect2, Strelka and Varscan2 [35, 37, 38]. This approach was chosen to reduce caller specific false positives. Variant annotations were generated using SNPeff, and Annovar [39, 40]. All bioinformatic analysis utilized the BETSY workflow manager [41].

Lentiviral labeling of sensitive and resistant cells

Lentiviruses incorporating Venus (LeGO-V2) and mCherry (LeGO-C2) fluorescent proteins were generated using Lipofectamine 3000 reagent (Thermo Fisher Scientific) according to the manufacturer’s instructions. LeGO-V2 and LeGO-C2 vectors were gifts from Boris Fehse (Addgene plasmids #27340 and #27339) [42]. CAMA-1 and CAMA-1_ribociclib_resistant cell lines were transduced with Venus- and mCherry-containing lentivirus, respectively using reverse transduction, resulting in CAMA-1_V2 and CAMA-1_riboR_C2 cell lines. Briefly, 1 ml of polybrene-containing cell suspension of 200,000 cells were plated in a well of a 6-well plate, where 0.5 ml of viral aliquot was previously dispensed. Cells were incubated for 48 h at 37 °C and 5% CO2, after which cells were washed and fresh regular culture medium was applied. Fluorescently labeled cells were selected using fluorescence-activated cell sorting after further subculture of transduced cells to attain homogeneously labeled cell populations.

Coculture experiments

2000 cells were plated in different proportions (100% CAMA-1_V2, 50% CAMA-1_V2—50% CAMA-1_riboR_C2, 100% CAMA-1_riboR_C2) in 96-well round-bottom ultra-low attachment spheroid microplate (Corning, Cat. No.: 4520). 24 h later, spheroids were washed and fresh medium including treatment drugs was applied. Spheroids were treated for a total of 21 days with imaging and media change been performed at every 4th and 7th day of the week. Imaging was performed using Cytation 5 imager (Biotek Instruments) gathering signal intensity from brightfield, YFP (for Venus fluorescence) and Texas Red (for mCherry fluorescence) channels. Raw data processing and image analysis were performed using Gen5 3.05 software (Biotek Instruments). Briefly, the stitching of 2 × 2 montage images and Z-projection using focus stacking was performed on raw images followed by spheroid area analysis. On the 21st day of treatment, spheroids were harvested, trypsinized, washed, resuspended in 2 µg/ml DAPI containing flow cytometry buffer (PBS + 5% FBS). Samples were subjected to flow cytometry analysis using Fortessa X20 flow cytometer (BD Biosciences) to assess the relative proportion of sensitive (Venus-labeled) and resistant (mCherry-labeled) cells. Venus fluorescence was excited with 488 nm laser and was separated by a 505 long-pass filter and detected through a 530/30 bandpass filter. mCherry fluorescence was excited with 561 nm laser and was separated by a 595 long-pass filter and detected through a 610/20 bandpass filter. DAPI fluorescence was excited with 355 nm laser and was separated by a 410 nm long-pass filter and detected through a 450/50 bandpass filter. Flow cytometry data were analyzed using FlowJo software (FlowJo, LLC version 10.5.3). Following the exclusion of doublets, fluorescence-positive live cells were analyzed in order to determine the proportion of sensitive and resistant cells within the spheroids. All coculture experiments were performed in triplicates.

Statistical analysis

Dose–response curves were generated using GraphPad Prism 7.02 software. Differences in dose–response curves were compared using extra sum-of-squares F-test. Differences in spheroid areas and cell proportions were analyzed using Student’s independent samples T-test. Unless otherwise stated a p-value < 0.05 was considered statistically significant.

Results

Long-term ribociclib treatment results in ribociclib-resistant cell line

After months of continuous ribociclib treatment, resistance to ribociclib emerged (CAMA-1_ribociclib_resistant cell line), as demonstrated with the different dose–response curves (p < 0.0001 Fig. 1a). Short-term ribociclib treatment resulted in 151 and 69 differentially expressed genes in the sensitive (CAMA-1) and resistant (CAMA-1_ribociclib_resistant) cell lines, respectively, with 68 genes downregulated in both cell lines (Fig. 1b–d, Additional file 1: Tables S1, S2). Using both Hallmark and Biocarta pathway gene sets, we find that “CDK regulation of DNA replication” was significantly altered in the resistant cells (untreated vs. ribociclib treated CAMA-1_ribociclib_resistant cells, FDR-corrected p-value: 7.2 × 10−6) while the two significantly dysregulated pathways between untreated and treated sensitive CAMA-1 cells were “CDK regulation of DNA replication” (FDR-corrected p-value: 6.6 × 10−10) and “cyclins and cell cycle regulation” (FDR-corrected p-value: 3.0 × 10−2), which is consistent with the mode of action of ribociclib.

Fig. 1
figure 1

Transcriptional response to ribociclib in sensitive and ribociclib-resistant cells. a Dose–response curves of CAMA-1 and CAMA-1_ribociclib_resistant cells under different concentrations of ribociclib treatment. Cells were treated with increasing concentration of ribociclib for 96 h, after which viability was measured using CellTiterGlo Chemiluminescent kit. The measured luminescence was normalized to the average of the lowest applied concentration (0.01 nM). Data points show the average of three replicates, error bars show standard deviation if it is larger than the size of the data point. b Venn diagram demonstrating the number of significantly differentially expressed genes in response to 12 h of 1 µM ribociclib treatment in CAMA-1 and CAMA-1_ribociclib_resistant cells. Blue circle incorporates differentially expressed genes in CAMA-1, while the red circle incorporates differentially expressed genes in CAMA-1_ribociclib_resistant cells. Red numbers demonstrate the number of upregulated, while blue numbers demonstrate the number of downregulated genes in response to ribociclib treatment. c, d Heatmaps demonstrating the expression of significantly differentially expressed genes in response to ribociclib treatment in CAMA-1_ribociclib_resistant (c) and CAMA-1 (d) cells

Full size image

We performed whole-exome sequencing on DNA isolated from the sensitive and resistant cells to detect acquired mutations in resistant cells. We found 9 high impact mutations mainly resulting in early stop codon occurrence and 34 moderate mutations resulting in single amino acid changes, while 18 mutations did not change the amino acid sequence of the coded proteins (Table 1). Although no detected mutations are directly associated with resistance against CDK4/6 inhibition, we found several mutations in cancer-associated genes possibly contributing to the acquired resistance against CDK4/6 inhibition.

Table 1 SNPs of high and moderate predicted impact in CAMA-1_ribociclib_resistant cells compared to parental CAMA-1 cells
Full size table

Ribociclib treatment selects for resistant cells in mixed cultures of sensitive and resistant cells

To discriminate sensitive and resistant cells in a coculture system, cell lines were labeled using lentiviral gene transfer with Venus and mCherry fluorescent proteins, respectively. Positively labeled cells were sorted using fluorescence-activated cell sorting to retain the homogenously labeled populations and were further subcultured to be utilized in the spheroid experiments. To analyze the long term effects of ribociclib treatment in 3D cocultures of sensitive and resistant cells, 21-day-long experiments using spheroids of different compositions (100% sensitive, 50% sensitive − 50% resistant, 100% resistant) were initiated (Fig. 2a).

Fig. 2
figure 2

Proliferation of mono- and coculture of sensitive and resistant cells under ribociclib treatment. a Representative images of time course spheroid growth of spheroids of different compositions. Images present the merged images of the Venus and mCherry channels. White bars correspond to 1000 µm. b Time course growth of spheroids presented on a. Data points show average spheroid areas of three replicates, error bars show standard deviation if it is larger than the size of the data point. Vertical lines represent statistical comparisons of the spheroid area between untreated sensitive vs. mixed (red), untreated sensitive and untreated resistant (brown), treated sensitive and mixed (purple) and treated sensitive and resistant (orange) spheroids on Day 21. Asterisks mark statistical significance (p < 0.05). Further statistical analysis including all time points and raw data are presented in Additional file 1: Table S1. c Representative FACS analysis of mixed spheroids on Day 21 on untreated (left) and ribociclib treated (right) spheroids initiated in a composition of 50% sensitive and 50% resistant cells. Samples are identical to the mixed samples on a (second and fifth rows). d Proportion of sensitive cells in untreated and ribociclib treated cocultured spheroids based on FACS analysis. Bars show the average of three replicates, error bars show standard deviation. Asterisk marks statistical significance (p < 0.05)

Full size image

Under no treatment, the growth of all spheroids was remarkable, reaching a 15- to 17-fold increase in spheroid area during the 21-day treatment (Fig. 2b, Additional file 1: Table S3). In the coculture setting, sensitive cells exhibited a relative proliferative advantage, as the FACS analysis confirmed that the majority of cells were Venus-labeled sensitive cells (Fig. 2c, d). Continuous ribociclib treatment of 400 nM principally inhibited the growth of the spheroids that were started with 100% sensitive cells (only threefold increase in spheroid area over 3 weeks). In contrast, treatment only had a modest effect on the growth of spheroids that were started with 100% resistant cells (12-fold increase in spheroid area over 3 weeks). Resistant cells proliferated more effectively under drug pressure in the coculture as well, resulting in a spheroid area 10% smaller than the treated 100% resistant monoculture. Flow cytometry analysis confirmed that ribociclib treatment selected for the resistant cells (Fig. 2c, d).

Transcriptional profiling of sensitive and resistant cells reveal key acquired vulnerability of resistant cells

To optimize the treatment of mixed spheroids of sensitive and resistant cells, we aimed to detect acquired vulnerabilities in the phenotype of the resistant cells, upon which collateral sensitivity of potential combination drugs can be applied. We compared the transcriptional profiles of untreated sensitive and resistant cells (untreated CAMA-1 vs. untreated CAMA-1_ribociclib_resistant) and found that 625 genes were significantly dysregulated (fold change > 2 and false discovery rate < 0.05, Additional file 2: Figure S1, Additional file 1: Table S4). Among these genes, several cell cycle (CDK6, RUNX2, CCNB3) and antiapoptotic (BCL2) genes were found to be overexpressed in resistant compared to sensitive cells. Pathway analysis using the Hallmark pathways provided 7 pathways significantly dysregulated between sensitive and resistant cells (Fig. 3a, b, Additional file 3: Figure S2, Additional file 1: Table S5).

Fig. 3
figure 3

Comparing the transcriptional program of CAMA-1 and CAMA-1_ribociclib_resistant cells reveals collateral sensitivity to Wee-1 inhibition. a Schematic representation of significantly altered Hallmark pathways between untreated CAMA-1 and CAMA-1_ribociclib_resistant cell lines. Positive normalized enrichment scores (NES) corresponds to enriched pathways in CAMA-1 (blue circles), while pathways with negative NES values (red circles) are enriched in CAMA-1_ribociclib_resistant cells. Additional enrichment scores for Hallmark pathways can be found in Additional file 1: Table S5. b Heatmap of all genes included in the Hallmark G2/M pathway in untreated CAMA-1 and CAMA-1_ribociclib_resistant cells. Representative genes of the pathway are labeled. Heatmap with all genes labeled can be found in Additional file 3: Figure S2. c Dose–response curves of CAMA-1 and CAMA-1_ribociclib_resistant cells under different concentrations of Wee-1 inhibitor adavosertib treatment. Cells were treated with increasing concentration of adavosertib for 96 h, after which viability was measured using CellTiterGlo Chemiluminescent kit. The measured luminescence was normalized to the average of the lowest applied concentration (0.01 nM). Data points show the average of three replicates, error bars show standard deviation if it is larger than the size of the data point

Full size image

The loss of the G2/M checkpoint renders cells more susceptible to mitotic catastrophe in the absence of sufficient quality control. Since the CAMA-1_ribociclib_resistant cells are more resistant to the G1 arrest caused by ribociclib and quality control at the G2/M checkpoint is also diminished in these cells based on the transcriptional profiling, Wee-1 inhibition stimulating resistant cells to enter mitotic catastrophe can be more toxic to these cells. Dose–response experiments with the Wee-1-inhibitor, adavosertib, on sensitive and resistant cells showed the IC50 value for adavosertib drop** by 42% (504.2 nM in CAMA-1 cells compared to 291.1 nM in CAMA-1_riboR cells, p-value: 0.0497), confirming the acquired sensitivity of resistant cells against adavosertib (Fig. 3c).

Leveraging collateral sensitivity controls spheroid growth without selecting for resistant cells

In an effort to overcome the selection of resistant cells in mixed spheroids, while still taking advantage of the antiproliferative effects of ribociclib treatment, we designed a follow-up experiment leveraging the collateral sensitivity of resistant cells in response to adavosertib. The same coculture system was used to investigate the long-term effects of to ribociclib, adavosertib and the combination of these two drugs. While the results of ribociclib treatment replicated our previous experiment, long-term adavosertib treatment proved to be more effective on resistant cells, having only a slight antiproliferative effect on sensitive and mixed spheroids (Fig. 4a, b, Additional file 1: Table S6). In line with the acquired adavosertib sensitivity in resistant cells, the proportion of sensitive cells was higher in adavosertib treated mixed spheroids, even compared to untreated control mixed spheroids, although no drug pressure also selected for the sensitive cells (Fig. 4c, d). The combination of ribociclib and adavosertib achieved the strongest growth-limiting effect in the spheroids of various composition (100% sensitive, 50% sensitive − 50% resistant, 100% resistant); however, it only added additional amount of growth inhibition compared to single-agent therapy in the case of the mixed spheroids. The growth of 100% sensitive spheroids under ribociclib pressure with or without the addition of adavosertib as well as the growth of 100% resistant spheroids under adavosertib pressure with or without ribociclib was quite indistinguishable, concluding that the combination adds considerable antiproliferative effect only in the mixed spheroids. The combination treatment also resulted in a larger proportion of sensitive cells compared to ribociclib mono-treatment as it selected for neither the sensitive nor the resistant cells.

Fig. 4
figure 4

Effects of individual and combination treatments of ribociclib and adavosertib on spheroid growth and composition. a Representative images of time course spheroid growth of spheroids of different compositions. Images present the merged images of the Venus and mCherry channels. White bars correspond to 1000 µm. b Time course growth of spheroids presented on Panel A. Data points show average spheroid areas of three replicates, error bars show standard deviation if it is larger than the size of the data point. Vertical lines represent statistical comparisons of the spheroid area between spheroids on Day 21. Asterisks mark statistical significance (*p < 0.05, **p < 0.01). Further statistical analysis including all time points and raw data are presented in Additional file 1: Table S6. c Representative FACS analysis of mixed spheroids on Day 21 on untreated, ribociclib, adavosertib and ribociclib + adavosertib treated spheroids initiated in a composition of 50% sensitive and 50% resistant cells. Samples are identical to the mixed samples on Panel A (second, fifth, eighth and eleventh rows). d Proportion of sensitive cells under different treatments based on FACS analysis. Bars show the average of three replicates, error bars show standard deviation. Asterisks mark statistical significance (p < 0.05)

Full size image

Discussion

Leveraging collateral sensitivity aims to control the growth of the tumor mass for a longer period by maintaining a delicate balance between sensitive and resistant cells without providing an unequivocal advantage to either cell type [17, 23]. Since rapid advancements in single-cell sequencing technologies enable us to track each cancer’s subclonal architecture throughout time and treatments [8, 10], personalized clinical management of various tumors in the future might reside on the observed phenotypic differences of resistant subclones and their druggable acquired vulnerabilities.

In the present work, we aimed to create a 3D spheroid system as an in vitro model of a heterogeneous tumor that incorporates both sensitive and resistant cells, followed by analysis of the antiproliferative relevance of specifically targeting both cell types. 3D spheroids recapitulate more closely the transcriptional program of in situ tumor cells compared to monolayer cultures and thus represent an emerging tool to better mimic the in vivo phenotype of tumors affected by cell–cell, cell–matrix interactions [43, 44].

As CDK4/6 inhibitors including ribociclib, abemaciclib, and palbociclib start to occupy a key role in the treatment of HR+, HER2− breast cancer [1], our interest focused on creating a ribociclib-resistant cell line from the widely used and characterized ER+, HER2− CAMA-1 cell line [45, 46]. Long-term exposure of ribociclib resulted in CAMA-1_ribociclib_resistant, a ribociclib-resistant CAMA-1 cell line. Sixty-nine and 151 transcripts were differentially expressed in response to short-term ribociclib treatment in the resistant and the sensitive cell lines, respectively. Although many fewer genes were dysregulated in response to ribociclib treatment in the resistant cell line, Biocarta pathway analysis on the dysregulated genes revealed an association with “CDK regulation of DNA replication”. The pathway analysis, as well as the abundance of key members of the cell cycle-dependent transcriptional program (members of the E2F transcription factor family, RRM2, cyclins, MCM2) [47, 48] underlined the clear cell cycle arrest ribociclib achieved in both cell lines. By comparing the baseline transcriptional profile of sensitive and resistant cells, we found CDK6 overexpression in resistant compared to sensitive cells which potentially provides a significant contribution to ribociclib resistance. Yang et al. also found that CDK6 overexpression is responsible for acquired resistance to CDK4/6 inhibitor abemaciclib in the in vitro setting [49], while other recent studies further confirmed resistance mechanisms facilitating G1/S transition despite CDK4/6 inhibition such as cyclin E1 amplification or Rb1 loss [50, 51]. In our results, the overexpression of additional key cell cycle regulators (RUNX2, CCNB3) may also facilitate cell cycle progression in response to CDK4/6 inhibition [52]. On the DNA level, we found several acquired mutations in the resistant cells. Although no specific mutated gene can be directly linked to resistance against CDK4/6 inhibitors, high impact mutations in several benign prognostic factors (SEMA6A, ALDH5A1, ARHGDIA) might also contribute to unrestricted proliferation even under drug pressure [53,54,20], with the possibility that further treatment with this combination might be durable.

It is important to underline the limitations of our study. Our study is not able to directly translate to in vivo behavior of heterogeneous tumor proliferation, rather it represents a clinically relevant in vitro model system.

Conclusions

We created an in vitro 3D spheroid coculture system modeling tumor heterogeneity dynamics with respect to sensitive and resistant cells towards the primary CDK4/6-inhibitor treatment. Following transcriptional profiling to detect acquired, collateral sensitivity of resistant cells, we show that an integrative approach selectively targeting sensitive and resistant cells is needed to optimally restrict spheroid growth of cocultures.