KKU-055 and KKU-213B cell lines were obtained from the Japanese Collection of Research Bioresources (JCRB) Cell Bank (Osaka, Japan). The culture methods were previously described (10). Cell lines were routinely tested for mycoplasma. Palbociclib-resistant cell lines and single-resistant clones were generated as previously described (16, 17).

Palbociclib, ribociclib, abemaciclib, oxaliplatin, cisplatin, and actinomycin D were purchased from Selleck Chemicals (Houston, TX, USA). Phenanthriplatin was generously provided by Professor Stephen J Lippard of the Department of Chemistry, Massachusetts Institute of Technology, Boston, MA, USA. Drug testing in resistant cell lines and 3-demensional (3D) spheroids was performed as previously described (10).

The whole-cell lysates were prepared by lysis in 0.5% NP-40 buffer (NP-40, 1 M HEPES pH 7.4, 5M NaCl, 0.5M EDTA pH 8) supplemented with a protease inhibitor and a phosphatase inhibitor (Thermo, 78420, USA). The nuclear-cytoplasmic fractionation lysates were prepared using cytoplasmic buffer (10 mM HEPES, 1.5 mM MgCl2, 10 mM KCl, 0.5 mM DTT, 0.05% NP-40, pH 7.9) and nucleoplasmic buffer (5 mM HEPES, 1.5 mM MgCl2, 0.2 mM EDTA, 0.5 mM DTT, 26% glycerol, pH 7.9). Equal amounts of lysate were used for Western blotting as previously described (10). Immunoprecipitation was performed according to standard protocol (18).

Cells were seeded in six-well plates at a 2,000 cell/well density and incubated for 24 hours. The next day, the culture medium was removed and new culture medium treated with 5 μM oxaliplatin, 0.5 μM cisplatin, or Vehicle was added into different wells and reincubated for 24 hours (10).

Two palbociclib-resistant clones (KKU-055R29 and KKU-055R30) and the KKU-055wt were selected for 3-plex dimethyl labeling proteomic analysis. The cells were lysed in 8 M urea lysis buffer containing protease and phosphatase inhibitors (Thermo, 78420, USA). Equal amounts of whole-cell lysate were digested, labeled, fractionated, and subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) using the Q Exactive™ Plus Hybrid Quadrupole-Orbitrap™ mass spectrometer as previously described (19). The raw data were analyzed for protein identification and relative quantification using MaxQuant software (Max Planck Institute of Biochemistry, Planegg, Germany). The files were searched against the human UniProt database (August 2019) using previously described search parameters (19). Proteins matched the reverse database and contaminants were excluded, and imputed missing values were derived from the normal distribution (width: 0.3, downshift: 0.5). The pathway analysis was performed and the statistics were calculated using Perseus software (version 1.6.14.0, Max Planck Institute of Biochemistry).

Gene expression profile from the Kyoto Encyclopedia of Genes and Genomes (KEGG) ribosome signature gene set (88 genes by gene family) analysis was performed as previously described (10). The normalized enrichment score was calculated, and the genes with a positive rank metric score were specified.

The protocols for all in vivo experiments in this study were approved by the Mahidol University – Institute Animal Care and Use Committee. Experiments were performed as previously described (10). A total of 10 × 10⁶ of KKU-055 cells was subcutaneously injected into female 2-month-old non-obese diabetic/severe combined immunodeficiency (NOD/SCID) mice, and the resulting tumor was allowed to grow to 0.5 cm in diameter. After that, 75 mg/kg palbociclib was administered daily via oral gavage until the tumor re-grew during treatment to obtain a palbociclib-resistant tumor. The palbociclib-resistant tumor was then excised and engrafted into a new cohort of study mice. After confirmation of tumor engraftment, 75 mg/kg palbociclib daily via oral gavage, 5 mg/kg oxaliplatin weekly via intraperitoneal injection, or the combination of both was given to tumor engrafted study mice. Hematoxylin and eosin (H&E) staining of mouse tumor tissues was performed according to standard protocol (10).

The study protocol was approved by the Institutional Review Board for Human Research (SI494/2019 and NCI006/2020). Intrahepatic CCA tissues were pasted using surgical blades and then washed with 12 ml of wash media (advanced DMEM/F12 containing 1 x Glutamax™ (Thermo, 10565018, USA), 10 mM HEPES, and 100 U/ml Penicillin/Streptomycin). The tissue paste was collected by centrifugation at 400 x g, 4°C for 5 min, then digested with 5 ml of the wash media containing 2 mg/ml collagenase D at 37°C for 30 min. Undigested tissues were filtered out with cell strainers. The cells were embedded in 70% matrigel and cultured in organoid culture media (20) until organoids formed. The organoids were dissociated into single cells by incubating with 1 ml of TrypLE^TM Express (Gibco, 12604021, USA) at 37°C for 5 min. In a 384-well plate format, the cells suspended in organoid culture media containing 5% matrigel were plated at 1000 cells/well, and incubated for 72 h. Eight concentrations of each drug with the following final concentrations: Palbociclib at 10, 5, 1, 0.5, 0.25, 0.125, 0.062, and 0.031 mM; Oxaliplatin at 100, 50, 10, 5, 2.5, 1.25, 0.62, and 0.31 mM; Palbociclib plus Oxaliplatin combination at 1:1 ratio, were added and incubated for 5 days. Cell viability was measured using ATPlite™ Luminescence Assay System (Perkin Elmer, Waltham, MA, USA) following the manufacturer’s instruction, and the cell viability percentages were calculated by normalizing with non-treated control (0.05% DMSO).

Unless otherwise stated, comparison was made and statistical significance was determined between groups using a two-sided Student’s t-test. A p-value of less than 0.05 was considered statistically significant for all tests. Data are shown as the mean ± standard deviation (SD) of at least 3 experiments. Analysis of variance (ANOVA) or Kruskal-Wallis test was used for comparisons among 3 or more groups.

To investigate acquired vulnerability in CCA cells that proliferate despite CDK4/6 inhibition, we developed the CDK4/6 inhibitor-resistant CCA cell lines KKU-055R and KKU-213BR from parental KKU-055 (KKU-055wt) and KKU-213B (KKU-213Bwt) cells by culturing the parental cells with the CDK4/6 inhibitor palbociclib using a step-wise dosing protocol ( Figure 1A ). The KKU-055R and KKU-213BR palbociclib GR50s were 18.57- and 15.53-fold higher than that of the parental cells, respectively ( Figure 1B ). We isolated and expanded up to 29 single cells (clones) from KKU-055R, and 13 clones from KKU-213BR to study various mechanisms of CDK4/6 inhibitor resistance ( Figure 1A ). We validated that these drug-resistant clones were resistant to the all of the FDA-approved CDK4/6 inhibitors (i.e., palbociclib, ribociclib, and abemaciclib) ( Figures 1C, D , and Supplementary Figures 1A, B ). The resistant clones divided under CDK4/6 inhibitor treatments with varied rates and doubling times, which suggests heterogeneous behavior among the cells ( Figures 1E, F ). We then explored the mechanisms of drug resistance in 12 randomly selected KKU-055R clones by immunoblotting ( Supplementary Figure 1C ). Virtually all of the representative clones demonstrated altered levels of proteins previously reported to be associated with CDK4/6 inhibitor resistance (21). More specifically, clones 8, 18, and 30 showed pRB downregulation; clones 12, 15, 18, 27, 28, and 29 showed CDK4-Cyclin Ds downregulation and cyclin E upregulation; and, clones 12, 15, 27, 28, and 30 showed CDK6 overexpression that correlated with slight upregulation of phospho-S6 kinase. These results indicated that we had established CDK4/6 inhibitor-resistant CCA clones that had developed varied adaptive molecular profiles to survive under CDK4/6 inhibitor treatment.

We used the IncuCyte^® Live-Cell Analysis System to perform acquired vulnerability screening by comparing the 5-day growth curves of twenty-nine resistant KKU-055R clones to the growth curves of KKU-055wt while under treatment with individual drugs from a small cancer drug library that covered 25 cancer pathways plus commonly used chemotherapies. We defined acquired vulnerability as positive when the average (n=4) area under the curve (AUC) of a resistant clone was smaller than the AUC of the parental cell by at least two fold ( Figure 2A ). As expected, we found all of the KKU-055R clones to universally resistant to all 3 of the CDK4/6 inhibitors (i.e., palbociclib, ribociclib, and abemaciclib) [ Figure 2B (blue color indicates less sensitivity to a drug compared to the sensitivity observed in KKU-055wt]. None of the resistant clones acquired vulnerability to the standard first-line CCA chemotherapy gemcitabine or cisplatin, except clone 12, which developed sensitivity to cisplatin (shown in red). We also found that several resistant clones became hypersensitive to mTOR inhibitors [i.e., everolimus (13/29 clones), and rapamycin (6/29 clones)] ( Figure 2B ), which is consistent with previous findings from breast cancer studies that investigated whether targeting activated PI3K/mTOR signaling may overcome acquired resistance to CDK4/6-based therapies (21–23). Unexpectedly, we found that most of the resistant clones (25/29 clones, 86.2%) had acquired sensitivity to oxaliplatin, which is a platinum agent that is used in combination with fluoropyrimidine, or gemcitabine as one of the standard second-line CCA treatment ( Figure 2B , Supplementary Figure 2A ). To validate the acquired vulnerability to oxaliplatin, we generated oxaliplatin dose-response curves for several KKU-055R and KKU-213BR clones, as well as for pools of KKU-055R, and KKU-213BR cells. We found a consistent reduction in oxaliplatin GR50s in all the clones and pooled resistant cells ( Figures 2C–E ). Oxaliplatin is an atypical platinum agent, which unlike the prototypical cisplatin, kills cancer cells by interfering with ribosome biogenesis (24). Clonogenic survival assay confirmed that the resistant clones were more sensitive to oxaliplatin compared to KKU-055wt cells ( Supplementary Figure 2B ). We then tested the relative sensitivity of the resistant clones to other drugs known to interfere with ribosomal biogenesis, such as phenanthriplatin (24) and actinomycin D. We found most of the KKU-055R and 213BR clones tested and the pooled resistant clones to be vulnerable to anti-ribosome biogenesis drugs ( Supplementary Figures 2C–H ). To further confirm the target of oxaliplatin in the ribosomal pathway, we also treated CCA cells with omacetaxine, which is a ribosome inhibitor whose anti-cancer function relies on intact ribosome function. We found that increasing the dose of oxaliplatin antagonized the anti-cancer activity of omacetaxine, which suggests that both drugs exploit a similar biological pathway to inhibit cancer cells ( Supplementary Figure 3A ). Of note, the observed antagonism was more noticeable in resistant KKU-055R cells compared to the parental cells, which suggests a more pronounced ribosomal dependency in the resistant cells. In addition, to exclude the possibility that resistant cells acquired a vulnerability to DNA damage-induced killing, we treated KKU-055R clones with various doses of cisplatin. We found that none of the resistant clones acquired sensitivity to cisplatin compared to the KKU-055wt ( Supplementary Figure 3B ). We, therefore, concluded that our acquired vulnerability screen had uncovered an acquired vulnerability to ribosomal biogenesis stress that may be targeted by oxaliplatin, an anti-ribosomal biogenesis drugs.

To understand the acquired sensitivity to anti-ribosome biogenesis agents, we quantitatively analyzed the proteomic profiles of resistant clones. We selected KKU-055R30 with a low level of pRB, and KKU-055R29 with cyclin E overexpression to compare with KKU-055wt cells. We found quite similar proteomic changes in both resistant clones. Top changes that were shared between the two resistant clones were cell cycle (p=0.0000017 vs. p=0.0000013), amino acid and glucose metabolism (p=0.0000014 vs. p=0.0000085), phagocytosis (p=0.0000031 vs. p=0.000023), regulation of actin (p=0.00000078 vs. p=0.0000058), and ribosomal proteins (p=0.0000044 vs. p=0.000013), all respectively ( Figure 3A ). As a result of the acquired vulnerability to anti-ribosomal biogenesis agents observed in the resistant cells, we focused on common changes in ribosomal proteins. We found significant enrichment of 15, and 10 ribosomal proteins in KKU-055R29, and KKU-055R30, respectively ( Figure 3B ). Individually, Ribosomal Protein (RP) Large subunit 29 (RPL29) was the most significantly upregulated in both resistant cells compared to KKU-055wt ( Figure 3C ).

Interestingly, we found the levels of RPL29 in resistant cells to be associated with drug treatment. More specifically, RPL29 was upregulated under palbociclib treatment, but it was downregulated under oxaliplatin treatment ( Figure 3D and Supplementary Figures 4A, B ). This finding suggests a functional role of RPL29 in CCA survival under these two treatments. Interestingly, the level of RPL29 was also suppressed in oxaliplatin treatment in spite of palbociclib treatment ( Supplementary Figure 4B ). The drug-induced ribosomal changes were specific to RPL29, but not to the other key RPs in the ribosome checkpoint pathway, such as RPL5, RPL11, and RPS14 ( Figure 3D ). Of note, cisplatin treatment did not cause any noticeable change in RPL29 expression ( Figure 3E ). Interestingly, we also observed drug-induced changes in RPL29 in the CDK4/6 inhibitor sensitive KKU-055wt cells, in a lesser extent ( Supplementary Figure 4B ), which suggests that drug-induced changes in RPL29 are a consequence of molecular reprofiling in resistant cells. RPL29 is a long half-life protein that was stable for 96 hours under cycloheximide treatment. We found that oxaliplatin significantly shortened the half-life of RPL29 to 20 hours ( Figures 3F, G ). Oxaliplatin had no effect on the half-life of RPL5 or RPL11 ( Figure 5F ). Therefore, to survive CDK4/6 inhibition, CCA cells reprofiled their molecular networks as reflected by drug-induced changes in RPL29. To further investigate whether RPL29 plays an essential role in the survival of the resistant clones, we partially depleted RPL29 by 2 independent sequences of RPL29 siRNA ( Supplementary Figure 4C ). We found that RPL29 siRNA-a, and b, which depleted 50% of the endogenous RPL29, significantly decreased the survival of the resistant clones, and the addition of palbociclib further decreased the survival of the resistant clones ( Figures 3H-J ).

Ribosomal biogenesis is tightly linked to cellular activities, such as growth and cell cycle progression. Perturbation of ribosomal biogenesis can cause nucleolar stress. The process through which RPs transmit nucleolar stress signals via MDM2-p53 has been described as a crucial tumor suppression mechanism (25). Downregulation of RPL29 has been linked to p53 activation by RPL5/11 sequestration of MDM2 in the nucleus (18, 26–28). Since we found that oxaliplatin promotes downregulation of RPL29, we hypothesized that oxaliplatin triggers RPL5/11-mediated MDM2 sequestration and p53 activation in CDK4/6 inhibitor-resistant cells.

As shown by the immunoblots in Figure 4A , oxaliplatin downregulated RPL29 in all of the resistant clones tested (to a lesser extent in wild-type cells), and promoted relocation of RPL5 and RPL11 from the cytoplasm to the nucleus. We also detected activation of p53 as indicated by elevated p53 S9 phosphorylation, and upregulation of the p53 functional target p21 ( Figure 4A ). Translocation of RPL5 and RPL11 was also confirmed by immunofluorescence staining ( Figure 4B and Supplementary Figure 5A ). The drug-induced relocalizations of RPL5 and RPL11 were specific to oxaliplatin – not cisplatin ( Figure 4C and Supplementary Figure 5B ).

We demonstrated by co-immunoprecipitation an increase in MDM2-RPL5/11 complex in the resistant cells upon oxaliplatin treatment ( Figure 4D and Supplementary Figure 5C ). We also observed the departure of p53 from the MDM2 complex upon oxaliplatin treatment ( Figure 4D ). We, therefore, concluded that under oxaliplatin treatment, RPL29 was suppressed, and RPL5/11 were relocalized to the nucleus to sequestrate MDM2, which caused p53-mediated growth inhibition in the resistant cells ( Figure 4E ).

We set forth to evaluate the efficacy of CDK4/6 inhibitor and oxaliplatin combination therapy for preventing cancer drug resistance. We performed in vitro cancer inhibition assays of different oxaliplatin+palbociclib combinations, and we calculated the combination index (CI) for each combination. We found that most of the evaluated combinations generated synergistic effect (CI <1), and that a higher dose of oxaliplatin produced stronger synergism ( Figure 5A ). Similar results were observed in the phenanthipatin+palbociclib combination ( Supplementary Figure 6A ). Results from isobologram confirmed the synergistic effect between oxaliplatin and palbociclib with 90%, 75%, and 50% growth inhibition (Fa 90, 75, 50, respectively) ( Figure 5B ). These results were also validated in 3D spheroid models ( Supplementary Figure 6B ).

Mechanistically, palbociclib exerts a cytostatic effect in vitro ( Figure 5C ). The addition of low-dose oxaliplatin (GR50 of parental KKU-055wt cells) effectuated small upregulation of the dying annexin V-positive KKU-055wt cells. In contrast, it caused significant increases in annexin V-positive cells in all resistant clones, as well as in the pooled cells. Similar results were observed in the oxaliplatin + plabociclib combination ( Figure 5C ).

Drug resistance, whether preexisting or acquired, is largely thought to be a stable and heritable process. However, over the past few decades, clinical evidence has suggested the role of unstable (reversible) non-heritable mechanisms of acquired drug resistance that affect chemotherapy and targeted agents. We, therefore, examined whether oxaliplatin can overcome both stable and reversible types of CDK4/6 inhibitor resistance. After a 6-week drug holiday, several CDK4/6 inhibitor-resistant clones (i.e., 22, 25, and 26) retained their CDK4/6 inhibitor resistance, which indicated stable resistance in these clones. In contrast, the resistance to palbociclib was reversed in some clones, such as 9 and 21 ( Figure 5D ), which indicates reversible resistance.

We found oxaliplatin to be effective against both stable and reversible resistance when given to the cells without drug holiday, or when given in combination with palbociclib ( Figure 5D , left panel). After a 6-week drug holiday, oxaliplatin was still effective against the clones with stable resistance (clones 22, 25, 26). However, oxaliplatin monotherapy was not effective against the cells with reversible resistance. Interestingly, only the oxaliplatin+palbociclib combination was relatively effective against the reversed clones ( Figure 5D , right panel). These findings indicate that continuous pressure from CDK4/6 inhibition is required to maintain acquired vulnerability to oxaliplatin. We then investigated the application of CDK4/6 inhibitor and oxaliplatin combination for preventing the emergence of drug-resistant cells. We performed long-term cultures of KKU-055wt cells under high doses of palbociclib, oxaliplatin, and the combination of the two for 2 months in 2 of 96-well plates (192 wells) ( Figure 5E ). Within 2 months, we found that 52 (27.1%) and 74 (38.5%) colonies of drug-resistant cells emerged in palbociclib and oxaliplatin monotherapy, respectively; however, there was no emergence (0%) of resistant cells from the combination treatment, which indicates a complete block of drug resistance by this drug combination. In contrast, complete suppression of resistant cells was not achieved by the palbociclib+cisplatin combination ( Supplementary Figure 6C ). Lastly, we found that the combination regimen was more effective compared to the sequential regimen in suppression of the pool drug resistant KKU-055R, and KKU-214BR cells ( Figure 5F ).

To assess the physiological effect and relevance of the oxaliplatin+palbociclib combination, we developed a mouse xenograft model for CDK4/6 inhibitor-resistant CCA. CDK4/6 inhibitor-sensitive KKU-055wt cells were implanted into the flanks of NOS/SCID mice, and the tumor was allowed to grow to 0.5 cm in diameter. To obtain a CDK4/6 inhibitor-resistant tumor, we treated the mice with 75 mg/kg/day palbociclib until the tumor regrew under the treatment. The CDK4/6 inhibitor-resistant tumor was then engrafted into a cohort of NOD/SCID mice to generate the CDK4/6 inhibitor-resistant model ( Figure 6A ). Although 13/16 (81.3%) and 16/16 (100%) of the drug-resistant tumors developed under palbociclib monotherapy and oxaliplatin monotherapy, respectively, only half (10/20, 50%) of the palbociclib-resistant tumors grew under the oxaliplatin+palbociclib combination treatment, which suggests that the combination therapy was more effective than monotherapy in the in vivo setting ( Figure 6B ). This combination treatment was also effective in suppressing tumor invasion and metastasis. We did not detect any tumor metastasis or invasion in the mice under combination therapy (0/20 tumors injected). In contrast, we found 12/16 and 3/16 metastasis sites/tumor injected in mice treated with palbociclib and oxaliplatin monotherapy, respectively ( Figure 6C ). In agreement with the in vitro cell death results, tumors extracted from combination treated mice and oxaliplatin monotherapy treated mice contained a large proportion of cancer cell death and necrosis ( Figures 6D, E ). To evaluate the clinical relevance of the combination, we tested it against 2 models of drug-naïve patient-derived intrahepatic CCA organoid, and we found that the combination was effective against the CCA organoids and lowered of the effective dose of each drug ( Figures 6F, G ). Taken together, these results indicate the comparative effectiveness of oxaliplatin+palbociclib combination therapy against drug-naïve CCA and CDK4/6 inhibitor-resistant CCA cells in the in vivo settings.