CDK4/6 Inhibitors Sensitize Rb-positive Sarcoma Cells to Wee1 Kinase Inhibition through Reversible Cell-Cycle Arrest

Sarcomas are a heterogeneous group of malignancies derived from mesenchymal cells that affect both children and adults of all ages and can arise anywhere in the body (1). More than 80 subtypes of sarcomas exist, broadly categorized into soft tissue sarcomas (STS) and primary bone sarcomas, each with significant differences in underlying biology, natural history, and treatment paradigms (2–4). Overall, newer treatment strategies for sarcoma patients lag behind many other solid tumors, partly due to the difficulty in performing translational research and clinical trials in selected sarcoma subtypes due to each entity being so rare. Sarcomas account for only 1% of all adult solid tumors and up to 20% of childhood solid tumors. Standard chemotherapy regimens usually consist of doxorubicin-based combinations, and many tumors are aggressive and refractory to, or eventually progress on standard chemotherapy (1, 5). Surgery plays a critical role in local control of primary tumors, along with radiation and chemotherapy in certain histologies that are known to be more sensitive to radiation and/or chemotherapy (6). However, rates of distant metastasis are still high and are the major cause of death in sarcoma patients (3) and newer, more targeted treatments are needed. However, many of the genetic drivers of sarcomas such as the fusion-genes used to define diagnosis are currently undruggable (7, 8). While there are examples of sarcoma subtypes with known active targeted therapies (e.g., tyrosine kinase inhibitors in gastrointestinal stromal tumors, and pazopanib in several STS) there is a paucity of described pathway alterations that can be used to guide treatment (9–12).

Deregulation of the cell cycle is one of the hallmarks of cancer, and may be a reasonable target for the development of new therapies in sarcoma (13). In particular, the G₁ checkpoint may be targeted due to the frequent alterations involving cyclin-dependent kinases (CDK) CDK4 and CDK6, p16, and the retinoblastoma (Rb) protein across multiple subtypes of sarcoma (14). CDKs are kinases that regulate the cell cycle, as well as transcription/RNA processing through phosphorylation of substrates such as Rb. CDKs that are involved in the cell cycle are regulated through binding to cyclins, which are transcriptionally regulated by the cell cycle. However, cancer cyclins, such as cyclin D and E, can be amplified or overexpressed leading to cell-cycle–independent constitutive activation of CDKs that drive cell-cycle progression. For this reason, targeting CDKs with small-molecule inhibitors has been an area of interest in many malignancies, including sarcoma (13). Although there are currently no approved cell-cycle–targeted therapies in sarcoma, there is an emerging body of promising preclinical and early clinical trial data that show signs of efficacy. CDK4/6–specific inhibitors are the most clinically advanced type of CDK inhibitor, with palbociclib being the first drug in this class to receive FDA approval in the setting of metastatic estrogen receptor (ER)-positive breast cancers given in combination with the aromatase inhibitor letrozole or the selective estrogen receptor downregulator fulvestrant (15, 16). Two other CDK4/6 inhibitors, abemaciclib and ribociclib, are undergoing FDA review for approval in a similar setting (17). More recently, these drugs are also being investigated in other solid tumors ranging from melanoma to non–small cell lung cancer, both as single agents and in combination with other signal transduction inhibitors (18, 19).

CDK4/6 inhibitors have already been examined clinically in well-differentiated and dedifferentiated liposarcomas, which almost universally harbor high-level amplifications of 12q14, the region containing CDK4. High expression of CDK4 has also been associated with a poor prognosis in these patients (14, 20). In a clinical trial with palbociclib in adults with liposarcoma, there was only 1 partial response noted, but the 12-week progression free survival (PFS) rate was 66%. This PFS exceeded the primary endpoint that would be considered a positive signal in sarcoma (21). A follow-up study using a different dose and schedule also resulted in a favorable efficacy signal in this disease, including 1 complete response out of 30 patients (22). The goal of our work was to explore the use of CDK4/6 inhibitors in other sarcoma subtypes and to determine whether they may be used to synchronize tumor cells in the G₁ phase of the cell cycle, positioning them to enter S and G₂–M phase synchronously where they would be more responsive to G₂–M kinase inhibitors.

One such G₂–M target, Wee1 kinase, is a key regulator of the G₂–M checkpoint, through inhibitory phosphorylation of CDK1, preventing progression of cells with DNA damage into mitosis (23). Wee1 kinase inhibitors have been found to synergize with DNA-damaging chemotherapies (such as platinum agents or gemcitabine) or radiation in many cancer cell lines due to promoting premature mitosis in cells with lethal DNA damage, leading to cell death via mitotic catastrophe. In this report, we tested the hypothesis that palbociclib-induced G₁ arrest and subsequent release primes the cells to respond to agents that are active maximally in S–G₂ phase such as Wee1 kinase inhibitors. To this end, we interrogated the utility of the sequential combination treatment of palbociclib followed by AZD1775 (24) in sarcomas using in vitro and in vivo (cell line and patient-derived xenografts) model systems.

Sarcoma cell lines were acquired from ATCC, and cultured as indicated in Supplementary Table S1. All cell lines were authenticated by short tandem repeat profiling upon receipt. Large batches of cells were frozen down so that cells could be maintained in culture for no more than 6 weeks for experiments. Palbociclib was a generous gift from Pfizer Oncology and AZD1775 was purchased from Proactive Molecular Research.

Western blot analyses were performed as previously described with the following modifications. The cell pellet was sonicated in PBS containing a cocktail of protease/phosphatase inhibitors (25 g/mL leupeptin, 25 g/mL aprotinin, 10 g/mL pepstatin, 1 mmol/L benzamidine, 10 g/mL soybean trypsin inhibitor, 0.5 mmol/L phenylmethylsulfonyl fluoride, 50 mmol/L sodium fluoride, and 0.5 mmol/L sodium orthovanadate). The primary antibodies used were phospho-Rb Ser-780 (Cell Signaling Technology #9307), Rb (Cell Signaling Technology #9309), CDK4 (Santa Cruz Biotechnology sc-260), CDK6 (Santa Cruz Biotechnology sc-177), p16 (Santa Cruz Biotechnology sc-1207), E2F1 (Santa Cruz Biotechnology sc-251), β-actin (Millipore MAB1501), PARP (Cell Signaling Technology #9542), caspase-3 (Cell Signaling Technology #9662), and Vinculin (Sigma V9131).

The cell cycle was analyzed using propidium iodide (PI) staining and flow cytometry analysis using standard methods. Briefly, cells were plated and treated as indicated in the text describing each figure. At the end of treatment, cells were harvested by trypsinization and fixed in 70% ethanol (in PBS). Following fixation and rinsing with PBS, cells were stained with 1 μg/mL propidium iodide (PI) in buffer overnight. The staining buffer consisted of PBS + 0.5% Tween-20 and 0.5% BSA with 20 μg/mL RNase A. A FACSCalibur flow cytometer was used with data generated using CellQuest Pro software, version 6.0.2 (BD Biosciences).

To generate the SK-LMS1 shRNA cells, we used lentiviral transduction of shRNA plasmids. HEK293T cells were transfected using PEI at a 3:1 ratio with the shRNA, and packaging vectors. Media was replaced on day 2, and supernatant containing virus collected on day 3 and 4. Viral infection was allowed to proceed for 24 hours. Puromycin was used to select for transduced cells; however, after selection, cells were maintained in regular DMEM:F12 media. The Rb shRNAs used were V3LHS_340825 and V3LHS_340827 (Dharmacon) and the nontargeting control was catalog no. RHS 4346; sense sequence, CTCGCTTGGGCGAGAGTAA (Open Biosystems).

This 96-well format high-throughput screening assay was performed with minor changes to previously published protocols from our laboratory. The density for each cell line was optimized using growth curves prior to beginning single drug treatments (listed in Supplementary Table S1). The changes to the timeline for this assay are shown as a schematic in Supplementary Fig. S3A, due to the slow mechanism of action of palbociclib necessitating 6 days of treatment. After palbociclib treatment, fresh media containing no drugs were used for “recovery” to allow cells to re-enter the cell cycle. The concentrations of each drug used for combinations are listed in Supplementary Table S2.

Senescence was measured by the senescence-associated galactosidase (SA-ß gal) staining kit (Millipore) according to the manufacturer's standard protocol. Briefly, cells were plated at a low density of 2,000 to 4,000 cells (depending on the plating efficiency of the cell line) in each well of 12-well plates and treated as described for high-throughput screening assay (HTSA; see Supplementary Fig. S3A). Following drug treatments and recovery, cells were then washed with PBS, fixed, and stained with SA-ß-gal solution overnight. The cells were then photographed using the Evos XL Core cell imaging system (Thermo Fisher) and senescent cells were quantified by counting 100 cells in three different fields for each replicate. A minimum of three technical and three biological replicates were performed for each condition.

All mouse experiments were performed under Institutional Animal Care and Use Committee–approved protocols at the University of Texas MD Anderson Cancer Center. Briefly, 1 × 10⁶ SK-LMS1 cells were injected subcutaneously into 4- to 5-week-old male nude mice, and randomized into treatment groups once the tumors reached 200 mm³. Palbociclib was dissolved in sterile 50 mmol/L sodium lactate buffer, and AZD1775 was dissolved in 0.5% methylcellulose. Palbociclib was administered by oral gavage daily (70 mg/kg for combination treatment experiments), and AZD1775 was administered twice daily (70 mg/kg) also by oral gavage. PDX lines were generated from surgical samples under an Institutional review board–approved protocol (LAB07-0659). PDX lines were generated using established methods, and the patient details can be found in Supplementary Table S3. PDX model 22 was used for assessment of the efficacy of palbociclib and AZD1775 as single agents and in combination. To propagate enough tumor to allow for the 4 treatment arm experiment (Fig. 6C–E), PDX model 22, was passaged into the flanks of 10 nude mice. Once the tumors reached 1.5 cm³, they were excised and cut into 30 mm³ segments and transplanted into the flanks of 20 nude mice. Mice were randomized (when tumors reached 150 mm³), into 4 treatment arms (5 mice/arm) using the same treatment strategy as in the SK-LMS1 experiments. Animals were weighed weekly to document toxicity, and tumors were measured in two dimensions using calipers. Tumor volumes were calculated using the formula π/6 × L × W². Tissues were fixed in 10% neutral buffered formalin for 24 hours and then transferred into 70% ethanol prior to processing for paraffin blocks.

To assess the recovery of SK-LMS1 xenograft cells into active cell-cycle posttreatment with palbociclib, mice were injected intraperitoneally with 0.25 mg/g body weight of bromodeoxyuridine (BrdUrd) 2 hours before euthanasia. Paraffin-embedded sections of the tumors were deparaffinized and rehydrated, and antigen retrieval performed using 10 mmol/L sodium citrate (pH = 6). Rat anti-BrdUrd mAb (GeneTex clone BU1/75, GTX26326) was used at a dilution of 1:500 for staining, and slides counterstained with Hematoxylin, Mayer's (Lillies's modification, from Dako). Two representative images per slide were taken under a 40× objective lens, and BrdUrd cells counted manually, including an average of 400–800 cells per animal.

All experiments were performed with a minimum of three technical and three biological replicates, and values reported are the mean of the three biological replicates, unless otherwise indicated. Error bars represent the SD from the mean, unless otherwise indicated. Pairwise comparisons were analyzed using multiple t tests (one unpaired t test per row). For all tests, differences were considered statistically significant at a P value of 0.05 or less. For all figures, ns: P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. All statistical analyses were performed using the GraphPad Prism software.

Breast cancer studies have demonstrated that cell lines that are most sensitive to CDK4/6 inhibitors are the ER-positive cell lines, most of which retain wild-type Rb (25). However, in sarcoma, the utility of Rb as a predictive biomarker has not been established. Therefore, to select both Rb-positive and Rb-negative cell lines for analysis of drug sensitivity and cell-cycle profiling, we subjected a panel of soft tissue and primary bone sarcoma cell lines to Western blot analysis for Rb and related G₁ checkpoint proteins (Fig. 1A). From this analysis, we found that SaOS2 (osteosarcoma) completely lacks Rb expression, and Rb is expressed at a low level in RD cells (rhabdomyosarcoma). All other cell lines expressed abundant Rb protein, which was also phosphorylated as we harvested actively cycling cells. CDK4 and CDK6 are highly expressed in all of the cell lines we tested, as is cyclin D1. We also found that 4 of the 6 cell lines do not express significant levels of p16, whereas SaOS2 and RD express p16. From these cell lines, we selected HT-1080 (fibrosarcoma) and SK-LMS1 (leiomyosarcoma) as representative Rb-positive, p16-negative cell lines and RD and SaOS2 as Rb-negative/low cell lines. We also used U2OS (osteosarcoma) as another Rb-positive cell line in many in vitro experiments to confirm our results.

CDK4/6 inhibitors approved for patients with hormone-receptor positive breast cancers have been shown to work more efficiently in cell lines that have intact G₁ checkpoint function (i.e., Rb present). We confirmed this relationship holds true in sarcoma cell lines as well using a modification of our previously described high-throughput survival assay (26), treating cells with palbociclib (PD) for different time intervals (Fig. 1B; Supplementary Fig. S1A). Both Rb-positive cell lines (HT-1080, SK-LMS1, and U2OS) show a time-dependent increase in sensitivity to palbociclib, whereas Rb-negative cell lines (RD and SaOS2) require significantly higher doses (i.e., >5 μmol/L) that are not pharmacologically achievable in vivo to significantly inhibit growth (Fig. 1B; Supplementary Fig. S1B).

As CDK4/6 inhibitors have been shown to be largely cytostatic due to causing G₁ cell-cycle arrest, in aggressive malignancies such as sarcomas, combination treatments with other agents are likely needed for significant tumor response. We hypothesized that inducing a reversible G₁ arrest with CDK4/6 inhibitors, then allowing recovery and reentry into the cell cycle, may prime cells for sensitivity to agents that preferentially act during S or G₂ phase, such as DNA-damaging chemotherapy or Wee1 kinase inhibitors. As a first step prior to testing the effect of combination treatment, we examined the sensitivity of cells to doxorubicin, as many standard first-line chemotherapy regimens are doxorubicin-based, and a Wee1 kinase inhibitor (AZD1775). In contrast to palbociclib, Rb status did not predict sensitivity to AZD1775 or doxorubicin, and 1 and 2 days of treatments with either agent resulted in similar dose–response curves, independent of Rb status (Fig. 1C and D; Supplementary Fig. S1C–S1F).

It is well-established that palbociclib induces a G₁ arrest with concomitant S-phase reduction in many cancer cell lines by continuous inhibition of CDK4/6 over several population doubling times (25, 27). We confirmed this in our sarcoma model systems using 6 days of treatment, as this was the period of treatment necessary to see consistent strong antiproliferative effects with an IC₅₀ within the pharmacologically achievable range (28). Rb-proficient cell lines treated with physiologically relevant doses of palbociclib arrested in G₁, whereas Rb-deficient cells did not (Fig. 2A; Supplementary Fig. S2A). Furthermore, increasing the dose of palbociclib in SaOS2, Rb-negative cells, not only failed to increase the proportion of cells in G₁ phase, but also resulted in a modest increase in G₂ phase, suggesting off-target effects of palbociclib at high doses (Fig. 2B). We next addressed the reversibility of the G₁ arrest, by treating all 5 cell lines with palbociclib, and then removing the drug and harvesting cells for flow cytometry analysis at several time intervals to determine the time needed to progress to S and G₂–M phases of the cell cycle. These analyses revealed that in SK-LMS1 and HT-1080 cells, the cells entered S phase within 6–9 hours of drug removal (Fig. 2C, top). Similar transitions to S and G₂–M were observed in U20S cells (Supplementary Fig. S2B). However, in the Rb-low/Rb-negative cells that did not arrest in G₁, drug removal did not induce significant changes in S–G₂ phase proportions at any concentration of palbociclib used during the course of the study.

To further investigate Rb as a biomarker of palbociclib response, we generated SK-LMS1 cells stably expressing Rb shRNAs, and examined whether Rb is necessary and sufficient for palbociclib-mediated G₁ arrest. Knockdown of Rb did not significantly affect the expression of CDK4, CDK6, or Wee1 kinase as compared with the nontargeting shRNA control cells (Fig. 3A). To rule out proliferation differences as an explanation for phenotypic changes, we conducted growth curves with the cell lines. The doubling time for the Rb knockdown cells (KD1 and KD2) were not different from nontargeting control (NTC; NTC -12.55 hours, KD1 -13.95 hours, and KD2 - 14.71 hours; Fig. 3B). We next subjected these cells to palbociclib treatment for 4, 6, and 8 days similar to previous experiments, and examined survival on day 13, as well as counted surviving cells at the end of 6 days of treatment. Rb knockdown cells were 1.5- to 4-fold more resistant to palbociclib depending on treatment duration and dose (Fig. 3C and D), compared with the NTC cells. To determine the requirement for Rb in mediating G₁ arrest, we also examined the cell cycle profiles following 6 days of treatment, and found that Rb knockdown cells had moderately less G₁ arrest than the control cells (Fig. 3E). We also examined the sensitivity of the cell line panel to AZD1775, and as expected found no significant difference indicating that Rb expression does not affect the activity of Wee1 kinase inhibitors (Fig. 3F). These results suggest that expression of Rb is necessary for inducing G₁ arrest by palbociclib at physiologic concentrations.

Combination treatment efficacy was evaluated using our HTSA, which allows comprehensive evaluation of 2 or more drugs either in sequence or given concomitantly in any adherent cell line in a 96-well format (schematically depicted in Supplementary Fig. S3A). For palbociclib, we selected 6 days of treatment, followed by 6 or 9 hours of recovery in drug-free medium, followed by treatment with the second drug (doxorubicin or AZD1775) for 48 hours, and quantification of cell survival at the end of the experiment on day 13. Doses and conditions for each cell line and each drug are listed in Supplementary Tables S1 and S2. Combination indices were calculated using CalcuSyn software using the median effect models described by Chou and Talalay (29). The absolute value of the combination index (CI) determines the degree of synergy or antagonism of drug combinations, CIs below 0.9 represent synergism and above 1.1 represent antagonistic interaction of two agents. Combination treatment with palbociclib followed by doxorubicin (Fig. 4A) and palbociclib followed by AZD1775 (Fig. 4B) resulted in synergistic activity of the combination in SK-LMS1 and HT-1080 cells, but not in Rb-deficient RD and SaOS2 cells (compare blue to red bars in Fig. 4A and B). We also compared the effectiveness of concomitant treatment with palbociclib and AZD1775 in SK-LMS1 cells, and observed antagonism at all doses of palbociclib when treating with both drugs on day 1 or day 3 (Supplementary Fig. S3D). To address the role of the fresh media recovery period, we performed side-by-side experiments comparing 6 hours recovery to no recovery in HT-1080 cells. At all doses of palbociclib, the average CI values were lower in the recovery experiment compared with no recovery, indicating that this recovery/cell-cycle reentry is important for maximal synergy between these agents (Supplementary Fig. S3E). We also asked whether expression of Rb is a prerequisite for synergistic activity of palbociclib and AZD1775. We treated the Rb shRNA SK-LMS1 cells with the combination of palbociclib and AZD1775 and results showed that once Rb is knocked down, the combination treatment did not result in synergistic activity (Supplementary Fig. S4).

To examine the mechanism of growth inhibition, we examined cleaved caspase-3 and PARP levels in SK-LMS1 cells treated similarly to the HTSA protocol to determine whether apoptosis occurred following combination treatment. The results revealed that apoptosis is not the main mechanism of synergism between these two agents (Supplementary Fig. S5). Next, we asked whether palbociclib induces senescence in sarcoma cells and if so, would the combination treatment with palbociclib and AZD1775 significantly increase senescence. The rationale for this hypothesis is that treatment of cells (in vitro and in vivo) with palbociclib alone induces senescence in breast and ovarian cancer cell lines (30, 31). To test this hypothesis, we treated two soft tissue sarcoma cell lines with either single agent or combination therapy with palbociclib and AZD1775 and subjected them to SA-ß galactosidase activity measurement (established assay for senescence detection). The data clearly show that while treatment of cells with palbociclib, but not AZD1775, as a single agent causes senescence, the sequential combination treatment of palbocilcib followed by AZD1775 causes significantly more senescence (Fig. 4C). Next, we asked whether there was DNA damage and/or mitotic catastrophe associated with increased senescence. We measured γ-H2AX (biomarker of DNA double strand breaks) at the end of the treatment and found that while the untreated and single drug–treated cells contained discernable γ-H2AX foci, the cells treated with the combination did not have any foci. Upon further examination of the nuclei of those cells, we found that they were undergoing mitotic catastrophe as measured by extent of abnormal nuclei detected by DAPI staining of the remaining cells (Fig. 4D). We next measured the percentage of cells with abnormal nuclei and found that in the combination treated cells >30% of cells are undergoing mitotic catastrophe as a result of micronucleation (Fig. 4D).

Given the synergy observed in vitro in Rb-proficient cell lines treated with palbociclib followed by either doxorubicin or AZD1775 (Fig. 4), we next interrogated whether this sequential combination treatment strategy also results in synergistic activity in vivo, using both cell line and patient-derived xenograft (PDX) models. As doxorubicin is a standard chemotherapy agent used in the treatment of sarcoma with overlapping toxicity (neutropenia) with palbociclib (28, 32), we focused on AZD1775 as the second drug to provide rationale for a novel dual targeted therapy trial if preclinical studies proved to be successful. The in vitro studies also suggested that the most synergistic schedule of treatment was sequential treatment with palbociclib first, followed by a recovery period so that cells could traverse through S-phase, followed by treatment with AZD1775. To recapitulate this in vitro scheduling (dose and time) in the in vivo design, we first optimized the schedule and dose of palbociclib in vivo using SK-LMS1 xenografts (Fig. 5B and C). For this experiment, 51 nude (nu/nu strain) mice were injected with 1 × 10⁶ viable SK-LMS1 cells subcutaneously. Tumor burden was measured daily and when the volumes reached 200 mm³, mice were randomized to 17 arms (3 mice/arm) and were treated with vehicle, 50 mg/kg, 100 mg/kg, or 150 mg/kg of palbociclib daily for 3 or 5 days and then allowed to recover for 0, 3, or 6 days. Animals were sacrificed and BrdUrd staining was performed following the last treatment/recovery period. Treatment of mice with doses higher than 50 mg/kg of palbociclib resulted in inhibition of tumor growth (Fig. 5A). BrdUrd labeling of the tumor tissue revealed that treatment of mice with either 100 mg/kg or 150 mg/kg every day for 5 days was sufficient to significantly inhibit S-phase fraction of the cells, and upon recovery for 3 and 6 days, the S-phase fraction returned to normal (Fig. 5B and C). On the basis of this dose and schedule, we next compared single agent to sequential treatment in vivo in SK-LMS1 cell line xenografts, and found that under similar dosing schedules, the strategy that was most effective was the sequential combination strategy (Fig. 5E, purple line). For this in vivo study, we treated cohorts (n = 10/cohort) of SK-LMS1 xenografts with the following 4 treatment arms: (i) vehicle; (ii) palbociclib alone; (iii) AZD1775 alone; (iv) sequential treatment with palbociclib followed by recovery and treatment with AZD1775. We used a dose of 70 mg/kg/day of palbociclib to minimize toxicity because we treated the mice for 3 cycles of 10 days/cycle (Fig. 5D). Results showed that continued administration of palbociclib alone for 3 cycles (5 days on-5 days off) or AZD1775 alone for 3 cycles (2 days on, 8 days off) are not as effective as 3 cycles of sequential combination treatment (5 days palbociclib, recovery for 3 days, followed by 2 days of AZD1775 treatment, Fig. 5E). Analysis of tumor samples at the end of the 3 cycles of treatment showed that combination treatment resulted in lower phosphorylation of Rb compared with control or palbociclib-treated animals, and also decreased cyclin D1 and E2F1 expression (Fig. 5F). Levels of CDK4 and CDK6 were similar in each of the study groups. As we showed in vitro (Supplementary Fig. S5), markers of apoptosis (caspase-3 and PARP cleavage) were not altered by any of the treatments suggesting that downregulation of Rb is likely the key pathway leading to decreased tumor volume by the combination therapy. To assess the toxicity of the treatments, we recorded the weights of the mice every other day, and observed no significant decreases in weight over time suggesting that the treatment was well-tolerated (Supplementary Fig. S6A). These in vivo studies with cell line xenografts suggest that sequential treatment with palbociclib, followed by a period of recovery and then treatment with AZD1775 leads to synergistic inhibition of tumor growth.

We next examined the combination treatment strategy in vivo using leiomyosarcoma PDX models. We have recently generated PDX models from surgical samples obtained from leiomyosarcoma patients and have established 13 stable PDX lines (Fig. 6). To determine the Rb status of these PDX lines and identify those that would be more likely to respond to sequential treatment with palbociclib → AZD1775 combination therapy, we subjected tumor tissues from the first passage (P1) of each established PDX line to Western blot analysis with Rb pathway regulators (Fig. 6A). The results revealed that 9 of the 13 PDX models express phospho-Rb protein and that CDK4 and CDK6 were uniformly expressed in the majority of the Rb-positive lines. We also noted that p16 was highly expressed in PDX #6 (no Rb expression) and was not expressed in the PDX lines with the highest phospho-Rb expression (#28 and #18). As most of the P1 lines can be passaged in vivo to multiple passages, we next examined the stability of the Rb pathway in 4 of these PDX models (#18, #20, #21, and #22, boxed in Fig. 6A) across several passages (Fig. 6B). The results revealed that the expression of p16, CDK4 and CDK6 were similar across all passages examined in all 4 PDX models. The one exception was the expression of phospho-Rb, which was diminished at P3 and P4 in PDX model #21. These results suggest that PDX models #22, #18, and #20 have maintained a stable Rb/CDK axis. The rate of tumor growth of these 4 PDX models over multiple passages are similar to each other and to the SK-LMS1 xenograft model, allowing us to translate our dosing schedule from the SK-LMS1 xenograft model to any of these 4 PDX models. We used PDX model #22 for the in vivo treatment study and as with the SK-LMS1 cell line xenografts, we treated cohorts (n = 5/cohort) of PDX #20 with the following 4 treatment arms: (i) vehicle; (ii) palbociclib alone; (iii) AZD1775 alone agent; (iv) sequential palbociclib followed by recovery and treatment with AZD1775. The PDX model was similarly sensitive to the combination treatment, showing both a significant decrease in tumor growth and final tumor weight at day 25 (Fig. 6C and D). Similar to the SK-LMS1 xenograft tumors, phosphorylation of Rb and levels of cyclin D1 were decreased in the combination-treated tumors, while levels of CDK4 and CDK6 remained unchanged by the treatment (Fig. 6E). To assess the toxicity of the treatments, we recorded the weights of the mice every other day, and observed no consistent decreases in weight over time suggesting that the treatment was well tolerated (Supplementary Fig. S6B). Taken together, these results indicate that sequential treatment with palbociclib and AZD1775 is effective in inhibiting leiomyosarcoma growth in vivo without significant toxicity, and that Rb is an important mediator of this synergistic activity.

Here we identify a novel combination treatment strategy for Rb-positive sarcomas that uses a currently FDA-approved drug (palbociclib) to potentiate a newer drug (AZD1775) that is in clinical trials for several solid tumor types (33). This sequential regimen synchronizes cells in vitro and in vivo to accumulate in G₁ phase by treatment with a selective CDK4/6 inhibitor, followed by a period of recovery that allows cells to re-enter the cell cycle where they are more sensitive to subsequent treatment with agents that are active in S -phase such as doxorubicin or AZD1775. Our in vitro work showed that only Rb-positive cell lines were synergistically inhibited by these treatments, and we directly showed in vivo that the palbociclib and AZD1775 combination treatment significantly inhibited tumor growth. We also show that the mechanism by which combination treatment with palbociclib and AZD1775 leads to growth inhibition is through sustained induction of senescence and mitotic catastrophe, but not apoptosis.

The cyclin D1/CDK4/6 pathway is relevant to sarcoma pathogenesis, and hence makes for an ideal therapeutic target. Beyond direct hyperactivation of this growth-promoting pathway through amplification of kinases such as CDK4 in liposarcoma and rhabdomyosarcoma, cyclin D1 may be overexpressed in various subtypes due to oncogenic activation by several transcription factors. For example, in synovial sarcoma the histology-defining fusion chromosome (X;18) creating fusions such as SS18-SSX, results in nuclear expression of β-catenin leading to upregulation of cyclin D1 (34). In alveolar rhabdomyosarcoma, activation of cyclin D1/CDK4 leads to phosphorylation and activation of the signature PAX3-FOXO1 fusion oncogene, known to be important in the aggressiveness of this tumor type (35). Suggesting these sarcoma histologies may be responsive to this novel combinatorial treatment strategy.

Successful clinical translation of the combination regimen of palbociclib and AZD1775 will require careful clinical trial design, both in terms of drug delivery schedule as well as enrollment criteria. Recent trials in STS combining agents with doxorubicin have had mixed results in randomized phase II–III trials when all histologic subtypes are included for enrollment, even for agents that have a solid preclinical rationale (36). One agent that has recently been shown to potentiate doxorubicin efficacy is olaratumab (a human PDGFRα mAb), which has received FDA approval for use in anthracycline-naïve STS patients in combination with doxorubicin. The approval is based on a phase II study showing an 11.8 month improvement in overall survival; however, the mechanism underlying this benefit is not entirely clear (37). One explanation for the mixed clinical results is that there are insufficient validated biomarkers available to select patients for these trials. Our work strongly implicates Rb as a biomarker for CDK4/6 inhibitor combinations, similar to data from other cancer types such as breast cancer and glioblastoma. Selection for enrolment based on biomarkers is more likely to be successful versus the standard approach of enrollment by histologic criteria alone (38). The frequency of Rb positivity within sarcoma appears to differ among histologic subtypes, and not be universally expressed in any subtype. Among liposarcoma and leiomyosarcoma, the two most common adult soft tissue sarcoma subtypes, approximately 40% are Rb-positive, with an enrichment in the well-differentiated liposarcoma histology for Rb expression compared with the dedifferentiated or myxoid subtypes (90%–100% IHC positive in well-differentiated, vs. 33% in dedifferentiated and 0% in myxoid; refs. 14, 39–41). This heterogeneity in the pattern of Rb expression by histologic subtype suggests that screening for expression of Rb will be necessary for trial enrollment to enrich for potential responders. Osteosarcoma, the most common childhood sarcoma, also may be a good candidate for this therapy, as 50% are Rb positive, and p16 is frequently hypermethylated (42). Our in vitro data (Fig. 1) showing that RD cells with very low levels of Rb levels are not responsive to palbociclib, raises the possibility that there may be a threshold of Rb activity that is needed for palbociclib to mediate its growth inhibitory (i.e., G₁ arrest) activity. Unlike Rb, where there is consensus among different tumor types regarding the predictive value of this protein, it remains controversial as to whether CDK4 or CDK6 levels may be useful in selecting patients for CDK4/6 inhibitor–based therapy. In one preclinical study using a different CDK4/6 inhibitor (LEE011), CDK4 amplification in rhabdomyosarcoma was shown to be a biomarker of resistance due to inability to activate the G₁ checkpoint and induce G₁ arrest (43). However, another recent study showed that in other soft tissue sarcomas (cell lines and PDXs), those with high levels of CDK4 were more sensitive compared with low-expressing tumors (44).

A concern with using drugs such as palbociclib to induce G₁ arrest and then allowing for a period of recovery is that normal cycling tissues may also be affected, making them more sensitive to subsequent DNA damage and cell death. However, this should not be the case due to the presence of intact G₁ and G₂ checkpoints and wild-type p53 in normal tissues. In mice, it has recently been shown that CDK4/6 inhibition protected normal intestinal epithelium from radiation-induced damage, through the inhibition of radiation-induced p53 activation (45). In normal hematopoietic cells, nonselective cell-cycle inhibition with cyclohexamide was able to protect them from Ara-C toxicity, while leukemic cells were not protected, again suggesting that the basal differences in cell cycle between normal and tumor cells provides a therapeutic window for the use of smart combination treatment strategies (46).

Combination therapies are often necessary to treat aggressive tumors such as sarcomas, and palbociclib may be an ideal drug to consider for other combinations. Indeed, in a large screen of PDX models, LEE011 (Novartis CDK4/6 inhibitor) synergized with several different signal transduction pathway–targeting agents, based on the premise that targeting a single pathway more potently leads to more resistance (47). Recent work has identified that the PI3K pathway may be a mechanism of resistance to palbociclib in breast cancer, making the argument that cotargeting PI3K may be beneficial (48). In osteosarcoma particularly, it is known that this pathway is a therapeutic vulnerability, making a triple combination strategy (CDK4/6, Wee1, and mTOR) possible as mTOR inhibitors do not have overlapping toxicities with CDK4/6 and Wee1 kinase inhibitors (49).

Our study adds to the growing literature that sequencing matters to the effectiveness of combination therapies in cancer, both in terms of tumor cell cytotoxicity as well as minimizing normal tissue damage. Our group has previously identified that CDK1/2 inhibition prior to doxorubicin selectively targets triple-negative breast cancer cells for DNA-damage-induced cell death in a mutant p53-dependent manner, through CDK1-mediated inhibition of DNA repair (50). Previous work on palbociclib in breast cancer cells expressing Rb has demonstrated that concomitant treatment with anthracyclines is antagonistic, as would be expected due to G₁ arrest (51). In ovarian cancer models, sequential palbociclib and chemotherapy (carboplatin or paclitaxel) without recovery was also shown to be antagonistic (52). Our work shows, however, that in Rb-positive sarcomas, palbociclib and doxorubicin can be combined successfully by allowing a period of recovery in between, indicating that sequence, drug stability/half-life, mechanism of action and timing of drug delivery must be considered when designing combination treatments.

No potential conflicts of interest were disclosed.

Conception and design: A. Alexander, V. Ravi, K. Keyomarsi, K.K. Hunt

Development of methodology: A. Alexander, K.H. Low, K. Keyomarsi, K.K. Hunt

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.M. Francis, A. Alexander, Y. Liu, S. Vijayaraghavan, N. Somaiah, K. Keyomarsi, K.K. Hunt

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Alexander, S. Vijayaraghavan, D. Yang, T. Bui, K. Keyomarsi, K.K. Hunt

Writing, review, and/or revision of the manuscript: A.M. Francis, A. Alexander, S. Vijayaraghavan, N. Somaiah, V. Ravi, K. Keyomarsi, K.K. Hunt

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.M. Francis, Y. Liu, K.K. Hunt

Study supervision: K. Keyomarsi, K.K. Hunt

This work is dedicated to the memory of Arshia Tabrizi who lost his courageous battle with metastatic leiomyosarcoma.

This work was supported by a Department of Defense Breast Cancer Research Program Postdoctoral Fellowship (BC112270 to A. Alexander), and T32 Fellowship (CA009599 to A.M. Frqancis), NIH grants CA87458 and CA1522228, a Sarcoma Foundation of America grant, and a joint grant from Alan B. Slifka and National Leiomyosarcoma Research Foundations (to K. Keyomarsi) and John Wayne Cancer Foundation (to K.K. Hunt). Institutional resources utilized included the Characterized Cell Line Core Facility and the Flow Cytometry and Cellular Imaging Facility, which are supported in part by the NIH through MD Anderson's Cancer Center Support Grant CA016672 (to K. Keyomarsi and K.K. Hunt, co-leaders of the Breast Program).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.