Drug-adapted cancer cell lines reveal drug-induced heterogeneity and enable the identification of biomarker candidates for the acquired resistance setting

Survivin is a drug target and the survivin suppressant YM155 a drug candidate for high-risk neuroblastoma. Findings from one YM155-adapted subline of the neuroblastoma cell line UKF-NB-3 had suggested that increased ABCB1 (mediates YM155 efflux) levels, decreased SLC35F2 (mediates YM155 uptake) levels, decreased survivin levels, and TP53 mutations indicate YM155 resistance. Here, the investigation of ten additional YM155-adapted UKF-NB-3 sublines only confirmed the roles of ABCB1 and SLC35F2. However, cellular ABCB1 and SLC35F2 levels did not indicate YM155 sensitivity in YM155-naïve cells, as indicated by drug response data derived from the Cancer Therapeutics Response Portal (CTRP) and the Genomics of Drug Sensitivity in Cancer (GDSC) databases. Moreover, the resistant sublines were characterised by a remarkable heterogeneity. Only seven sublines developed on-target resistance as indicated by resistance to RNAi-mediated survivin depletion. The sublines also varied in their response to other anti-cancer drugs. In conclusion, cancer cell populations of limited intrinsic heterogeneity can develop various resistance phenotypes in response to treatment. Therefore, individualised therapies will require monitoring of cancer cell evolution in response to treatment. Moreover, biomarkers can indicate resistance formation in the acquired resistance setting, even when they are not predictive in the intrinsic resistance setting.

A number of studies have investigated the potential of YM155 against neuroblastoma cells [13][14][15][16]. Neuroblastoma is the most common extracranial solid childhood tumor. Treatment outcomes in high-risk neuroblastoma patients remain unsatisfactory. About 50% of these patients relapse and have a 5-year-survial rate below 10% [17][18][19][20]. We have recently shown that suppression of survivin expression is the main mechanism through which YM155 exerts its anti-neuroblastoma effects [15]. Notably, the New Drug Development Strategy (NDDS, a project of Innovative Therapies for Children with Cancer, the European Network for Cancer Research in Children and Adolescents, and the International Society of Paediatric Oncology Europe Neuroblastoma) has categorized survivin as a high priority drug target in neuroblastoma and YM155 as a high priority drug [21].
The formation of acquired resistance is a central problem in (metastasised) cancer diseases that need to be treated by systemic drug therapy. Although many cancers initially respond well to therapy, resistance formation is common and cures are rare [22]. Hence, biomarkers that indicate early therapy failure are needed to adapt therapies if resistance emerges. Liquid biopsies (e.g. circulating tumor cells) enable the monitoring of cancer cell evolution in patients with ever more detail [23]. However, the translation of the resulting information into improved therapies is hampered by a lack of understanding of the processes underlying acquired resistance formation and, in turn, a lack of biomarkers.
Most studies focus on biomarkers that indicate whether a certain cancer cell (population) is likely to respond to a certain treatment, but not on biomarkers that indicate early that a current therapy has stopped working. This also applies to the previous studies that investigated the efficacy of YM155 in neuroblastoma [13,14,16].
However, it is known that intrinsic and acquired resistance mechanisms may substantially differ [24][25][26]. Using a single YM155-adapted neuroblastoma cell line, we identified increased ABCB1 (also known as P-glycoprotein or MDR1) expression, decreased SLC35F2 (Solute Carrier Family 35 Member F2) expression, decreased survivin expression, and loss-of-p53-function as potential markers of resistance formation to YM155 [15]. Given the tremendous (intra-tumour) heterogeneity in cancer [27], it is likely that the processes, which result in acquired resistance formation, are equally complex. If so, then a larger number of models of acquired resistance to a certain drug will be needed to adequately address the complexity of the resistance formation process.
To test this hypothesis, we here established and characterised 10 further YM155-adapted UKF-NB-3 neuroblastoma cell lines. To see whether we can obtain information from our acquired resistance models that cannot be identified from traditional approaches using non-adapted cell lines, we also analysed YM155 response data from the two large pharmacogenomics screens Genomics of Drug Sensitivity in Cancer (GDSC) and Cancer Therapeutic Response Portal (CTRP) [28,29]. We found a remarkable heterogeneity between the individual sublines, although they all had been derived from the same parental cell line. An increase in cellular ABCB1 levels and/ or a decrease in SLC35F2 levels indicate resistance formation to YM155, although the ABCB1 and/ or SLC35F2 levels cannot be used to infer YM155 sensitivity in YM155-naïve cell lines. The use of the panel of YM155adapted cell lines further enabled us to show that the cellular survivin levels and the TP53 status do not reliably indicate resistance formation.

Acquired YM155 resistance is associated with decreased sensitivity to survivin depletion
Our previous findings had suggested that YM155 predominantly exerts its antineuroblastoma effects via suppression of survivin expression [15]. Seven of the ten YM155-adapted UKF-NB-3 sublines (I, III, V, VII, VIII, IX, X) displayed decreased sensitivity to siRNA-mediated survivin depletion. Two sublines were similar sensitive as parental UKF-NB-3 cells (II, VI), and one subline (IV) was more sensitive (Figure 4, Figure S4). This shows that a majority of the YM155-resistant cell lines have developed on-target resistance. It also indicates that the YM155 resistance mechanisms differ between the individual UKF-NB-3 sublines.

Relevance of cellular ABCB1 and SLC35F2 levels in the context of YM155 resistance
Increased cellular ABCB1 (mediates YM155 efflux) levels and decreased SLC35F2 (mediates cellular YM155 uptake) levels have previously been identified as important YM155 resistance mechanisms [13,15,16,34]. To further investigate the relationship between ABCB1 and SLC35F2 levels and YM155 sensitivity, we compared the YM155 sensitivity in cell lines that displayed low or high expression of the respective genes using GDSC and CTRP data. In agreement with previous data, high ABCB1 expression ( Figure 5) and low SLC35F2 expression ( Figure 6) were associated with reduced YM155 sensitivity. When we used transcriptomics data from the GDSC and CTRP to correlate the expression of all genes with YM155 sensitivity, ABCB1 ranked as the gene whose expression was most strongly correlated to the YM155 AUC (area under the curve, unit used to quantify drug response) (Table 1) in the GDSC and CTRP. SLC35F2 expression was most strongly inversely correlated to the YM155 AUC (Table 2) in both data sets. There were no further overlaps among the top 10 genes between the two databases ( Table 1, Table 2). However, the YM155 sensitivity of a certain cell line could not be reliably predicted based on the cellular ABCB1 and/ or SLC35F2 levels ( Figure 5, Figure 6).
All YM155-adapted UKF-NB-3 sublines displayed increased ABCB1 levels relative to UKF-NB-3 ( Figure 7, Figure S5). Acquired YM155 resistance was also generally associated with decreased SLC35F2 levels, in particular in the sublines I, IV, VI, and X ( Figure 7, Figure S5). This indicates that increased ABCB1 levels and decreased SLC35F2 levels have potential as biomarkers indicating YM155 resistance formation in response to YM155-based therapies, although cellular ABCB1 and SLC35F2 levels do not enable the prediction of YM155 sensitivity in YM155-naïve cells.

YM155-adapted UKF-NB-3 cells remain sensitive to DNA damage caused by irradiation and cytotoxic drugs
YM155 has been proposed to exert its anti-cancer effects via the induction of DNA damage in some experimental systems [3,5,34,35]. To study whether the acquisition of YM155 resistance was associated with a generally increased resistance to DNA damage, UKF-NB-3 and its YM155-adapted UKF-NB-3 sublines were irradiated at a dose range of one to five Gy. None of the YM155-adapted UKF-NB-3 sublines displayed substantially reduced sensitivity to irradiation relative to UKF-NB-3 ( Figure 8). Moreover, none of the YM155-resistant UKF-NB-3 sublines displayed reduced sensitivity to cisplatin (causes DNA crosslinks) or topotecan (topoisomerase I inhibitor), which cause DNA damage by different mechanisms (Figure 8, Table S2).
There was also no coherent increase in resistance to the nucleoside analogue gemcitabine ( Figure 8, Table S2). These data do not suggest a dominant role of DNA damage induction in the course of the anti-neuroblastoma activity of YM155 in UKF-NB-3 cells.

Heterogeneity among YM155-adapted UKF-NB-3 sublines
While the YM155-adapted UKF-NB-3 sublines displayed limited heterogeneity in response to treatment with cisplatin and topotecan, remarkable differences in the gemcitabine IC50s were detected ( Figure 8, Table S2). The fold difference between the YM155-adapted subline with the lowest gemcitabine IC50 (V, 0.12ng/mL) and the subline with the highest IC50 (VIII, 0.65ng/mL) was 5.4-fold. This heterogeneity is in agreement with the up to 29-fold difference observed in cell viability in response to BIRC5/ survivin depletion between the most sensitive (II) and the most resistant (VII) subline ( Figure 4). Resistance profiles to the destabilising tubulin-binding agent vincristine also revealed a substantial heterogeneity between the YM155-resistant UKF-NB-3 sublines (Figure 9, Table S2), resulting in a fold difference of 127 between subline VI (vincristine IC50: 714ng/mL) and subline IX (vincristine IC50: 5.6ng/mL) (Table S2).

Discussion
In a previous study, a YM155-adapted subline of the neuroblastoma cell line UKF-NB-3 was characterised by increased cellular ABCB1 levels, decreased SLC35F2 and survivin levels, and a TP53 mutation [15]. Here, we systematically investigated the relevance of cellular ABCB1, SLC35F2, and survivin levels as well as the TP53 status as potential biomarkers of YM155 resistance formation in the intrinsic resistance setting, using data derived from the GDSC and CTRP databases [28,29], and in the acquired resistance setting, using an additional set of ten YM155-adapted UKF-NB-3 sublines, which were established in independent experiments.
Increased ABCB1 expression (mediates YM155 efflux) and decreased SLC35F2 expression (mediates cellular YM155 uptake) were identified as YM155 resistance mechanisms in panels of YM155-naïve cell lines that displayed varying levels of these proteins and in functional studies [13,15,16,34], which was further supported by our analysis of GDSC and CTRP data [28,29]. Despite their roles in determining YM155 resistance, however, cellular ABCB1 or SLC35F2 levels did not enable the prediction of whether an individual cell line would be sensitive to YM155 or not. The YM155-adapted UKF-NB-3 cell lines generally displayed elevated cellular ABCB1 levels and reduced SLC35F2 levels relative to UKF-NB-3. Hence, an increase in the cellular ABCB1 levels and/ or a decrease in the SLC35F2 levels have potential as biomarkers that indicate resistance formation, even though the respective cellular levels do not reliably predict YM155 response in YM155-naïve cells.
Initially, the TP53 status was reported not to influence the anti-cancer effects of YM155 [30], which was further supported by our analysis of GDSC and CTRP data. In neuroblastoma cells, however, YM155 induced p53 signalling, p53 depletion reduced YM155 sensitivity, and a YM155-adapted UKF-NB-3 subline harboured a TP53 mutation [15]. Here, all ten YM155-adapted UKF-NB-3 sublines retained wild-type TP53. Thus, the role of p53 seems to depend on the individual cellular context. Neither the cellular TP53 status nor the formation of TP53 mutations can currently be considered as valid biomarkers for YM155 therapies.
The relevance of cellular survivin levels for cancer cell sensitivity to YM155 is not clear [15, [30][31][32][33]. Our analysis of GDSC and CTRP data indicated that high survivin (BIRC5) expression was associated with increased YM155 sensitivity. However, it was not possible to infer the YM155 sensitivity of a particular cell line based on its survivin status. Reasons for this may include that survivin is not in all cell lines the major therapeutic target of YM155 as it is in neuroblastoma cells [1][2][3][4][5][6][7][8][9][10][11][12]15,35] and/ or that off-target resistance mechanisms such as ABCB1 and SLC35F2 expression may affect YM155 efficacy independently of the survivin status [13,15,16,34].
The YM155-adapted UKF-NB-3 sublines displayed various survivin levels, demonstrating that resistance formation to YM155 is also not associated with a consistent change in cellular survivin levels. Seven of the YM155-adapted cell lines displayed on-target resistance as indicated by reduced sensitivity to RNAi-mediated BIRC5/ survivin depletion relative to parental UKF-NB-3 cells, further confirming that survivin is a target of YM155 in neuroblastoma cells. However, cellular survivin levels do not represent a reliable biomarker of resistance formation to YM155.
While YM155 was described to act via the induction of DNA damage in some cancer types [3,5,34,35], our previous results did not indicate a causative role of DNA damage induction in the anti-cancer effects of YM155 against neuroblastoma cells [15]. YM155 resistance formation in the YM155-adapted neuroblastoma cell lines was also not associated with generally decreased sensitivity to radiation or DNA damage caused by cisplatin (causes DNA crosslinks), gemcitabine (nucleoside analogue), or topotecan (topoisomerase I inhibitor). This indicates that YM155 resistance formation in neuroblastoma cells is not generally associated with an increased resistance to DNA damage induction.
In this study, the use of multiple models of acquired resistance enabled insights that could not be gained from just one drug-adapted subline. The previous investigation of one YM155-resistant UKF-NB-3 subline had suggested that changes in the cellular TP53 status and survivin levels indicate resistance formation [15], which was not confirmed in our current panel of ten YM155-adapted UKF-NB-3 sublines.
Moreover, the use of multiple sublines provided a pioneering glimpse onto the remarkable heterogeneity of the resistance formation process, even though all resistant sublines were derived from the same parental cell line. Only seven of the ten sublines developed on-target resistance mechanisms as indicated by reduced sensitivity to survivin depletion. The sublines also showed substantial variation in their sensitivity to irradiation (up to 7-fold difference at 5Gy), gemcitabine (up to 5-fold), and vincristine (up to 127-fold). Notably, a much higher heterogeneity would be expected in the clinical situation, in which tumours are already characterised by much higher heterogeneity than cancer cell lines and in which combination therapies are common.
In conclusion, our data reveal a high phenotypic heterogeneity among a panel of ten YM155-resistant sublines of the neuroblastoma cell line UKF-NB-3. This heterogeneity is of conceptual importance, because it shows that even a defined cancer cell population of limited intrinsic heterogeneity can develop various resistance mechanisms and phenotypes in response to treatment. From a clinical perspective, this means that the close monitoring of cancer cell evolution in response to therapy will have to become an essential part of the design of individualised therapies. Notably, such insights can only be gained from preclinical model systems such as drug-adapted cancer cell lines, which enable the repeated adaptation of a given cancer cell population to the same treatment, but not from clinical material as every patient can only be treated once.
Our findings also demonstrate that biomarkers can indicate resistance formation, even when they do not enable the prediction of drug sensitivity in therapynaïve cancer cells. Hence, the use of biomarkers differs between the intrinsic and the acquired resistance setting, and pre-clinical models of acquired drug resistance are needed for the identification of such biomarkers that herald resistance development.

Cells
The and designated as UKF-NB-3 r YM155 20nM I to UKF-NB-3 r YM155 20nM X. All cells were propagated in IMDM supplemented with 10 % FBS, 100 IU/ml penicillin and 100 µg/ml streptomycin at 37°C. Cells were routinely tested for mycoplasma contamination and authenticated by short tandem repeat profiling.

Analysis of data derived from large pharmacogenomic studies
All data (including drug response area under curve (AUC) data for YM-155treated cancer cell lines, basal gene-expression for ABCB1, BIRC5 (the gene that encodes survivin), and SLC35F2, and genomic alterations of p53) in this study were obtained from two online resources: Version 2 of the Cancer Therapeutics Response Portal (CTRP v2) data [28,40] were obtained from the Cancer Target Discovery and Development (CTD 2 ) data portal (ocg.cancer.gov/programs/ctd2/data-portal). The The GDSC contains ABCB1, BIRC5, and SLC35F2 expression data for 1019 cell lines and YM155 AUC data for 945 cell lines. Expression data and WES data were available for all 945 cell lines with YM-155 sensitivity data (including 30 neuroblastoma cell lines).
Data processing was performed using Perl version 5.26.0, and R statistical packages version 3.3.2. Cell lines were determined to display either high or low expression for each gene using the median gene expression as a threshold (i.e. low expression <= median expression, high expression > median expression). Box plots indicating YM-155 sensitivity in cell lines that display low or high expression of a certain gene or wild-type or mutant TP53 were produced using the ggplot2 package [43] in R.
Statistical tests were carried out in R and included Wilcoxon rank-sum test [44] and Pearson's correlation [45]. Correction for multiple comparisons was performed using the Benjamini-Hochberg procedure [46].

Statistics
Results are expressed as mean ± S.D. of at least three experiments.
Comparisons between two groups were performed using Student's t-test. Three or more groups were compared by ANOVA followed by the Student-Newman-Keuls test.
The work was supported by the Hilfe für krebskranke Kinder Frankfurt e.V. and the Frankfurter Stiftung für krebskranke Kinder.