Sequential Targeting of Retinoblastoma and DNA Synthesis Pathways Is a Therapeutic Strategy for Sarcomas That Can Be Monitored in Real Time

Soft-tissue sarcomas (STS) are a rare and highly heterogeneous group of tumors that account for 1% of adult malignancies and 20% of pediatric cancers (1). There are more than 50 distinct subtypes of STS; leiomyosarcoma (LMS), liposarcoma (LPS), and rhabdomyosarcoma (RMS) represent the most common types of adult and pediatric cases (2, 3). Surgical resection, with or without radiotherapy and chemotherapy, has been the cornerstone of therapeutic intervention for STS patients (4). In the advanced setting, first-line chemotherapy consists of doxorubicin (5), ifosfamide (6), or gemcitabine (7), given as single agents or in combination, with response rates in the range of 10% to 25% (8, 9). Prognosis is poor with a median overall survival (OS) of 12.8 to 14.3 months (10). Therefore, new treatment strategies that exploit specific targets/pathways are required to improve outcomes.

One such target is the retinoblastoma (Rb) protein and pathway, which are commonly altered in many adult and pediatric STSs. Rb is a tumor-suppressor protein; in its hypophosphorylated state, it sequesters the E2F transcription factor (11), inhibiting its ability to transcribe the genes required for G₁ to S phase transition (12). Rb is sequentially phosphorylated by cyclin-dependent-kinase (CDK) 4/cyclin D or CDK6/cyclin D and CDK2/cyclin E complexes at early and late G₁ phases, leading to transactivation of key cell-cycle genes and progression of cells into S phase by releasing E2F from the Rb/E2F complex. Multiple STS subtypes, including some LPS and RMS, exhibit amplification of chromosome 12q13–15, a region where the CDK4 gene is located (13, 14). Overexpression or amplification of CDK4 results in inactivation of Rb as a tumor suppressor. Hence, the suppression of cell-cycle progression through reactivation of the Rb pathway poses a potential avenue for novel targeted therapies in STS that harbor intact Rb signaling. Clinical efficacy data with single-agent abemaciclib (CDK4/6 inhibitor) in adult dedifferentiated LPS patients (NCT02846987) showed median progression-free survival (PFS) was prolonged to 30.4 weeks but did not meet the criterion for partial response, underscoring the need for biomarkers and rational combination treatments for STS patients who receive CDK4/6 inhibitors (CDK4/6i; ref. 15).

Concurrent combinations of CDK4/6is with chemotherapy have been examined in several preclinical models, but antagonistic responses have been observed (16). One reason these concurrent combination therapies are not effective is that CDK4/6is induce G₁ arrest and prevent S phase–specific chemotherapeutic agents from mediating their therapeutic effects (17). To address this challenge, a two-step sequential combination treatment strategy was designed in which CDK4/6is and chemotherapy dosing were separated (i.e., sequential) to induce a synergistic cell killing in STS models in vitro and in vivo. The impact of each agent on the cell cycle was evaluated in real time through functional imaging and noninvasive blood-based biomarkers, providing a potential treatment strategy for patients with STS harboring an intact Rb pathway.

Formalin-fixed, paraffin-embedded (FFPE) sarcoma surgical specimens (n = 1,043) were retrieved from the University of Texas MD Anderson Cancer Center (MDACC, Houston, TX) pathology archives under an Insititutional Review Board (IRB)–approved protocol, and all patients provided written, informed consent. Clinical and pathologic characteristics are summarized in Supplementary Table S1. Whole-slide H&E-stained sections were reviewed by a pathologist, and areas of viable tumor were selected and marked for sampling. IHC was performed on FFPE tissue sections in the clinical IHC laboratory, which is certified under the provisions of the United States Clinical Laboratory Improvement Act and accredited by the College of American Pathologists. Purified mouse anti-human Rb protein (clone G3-245, BD Pharmagen, RRID:AB_385259) antibodies were used for IHC staining (Supplementary Materials and Methods). All sections were scored independently by two pathologists to estimate the percentage of Rb-positive tumor cells. A cutoff point of 5% immunoreactive tumor cells was used to define a positive phenotype for Rb.

Sarcoma cell lines were purchased from ATCC and maintained as described previously (18). Rhabdomyosarcoma cells were gifted by Dr. Peter J. Houghton (Greehey Children's Cancer Research Institute, The University of Texas Health Science Center at San Antonio). Cell lines were regularly tested for Mycoplasma and authenticated regularly (every 6 months) by karyotype and short tandem repeat analysis at MDACC Characterized Cell Line Core Facility (Houston, TX). The detailed procedures for all in vitro assays (Western blot analysis, high-throughput survival assay, cell-cycle analysis, apoptosis assay, immunofluorescence staining, and IHC staining) were described previously (18–24) and are included in the Supplementary Materials and Methods.

To generate RB1 knockout HT-1080 cell lines, sgRNA-targeting exon 1 (GGTGGCGGCCGTTTTTCGGG) or exon 20 (GGATCTTCCTCATGCTGTTC) of human RB1 gene was cloned into pX330 plasmid (RRID:Addgene_101733). For human TP53 knockout, sgRNA-targeting exon 9 (GGAGAGGAGCTGGTGTTGT) was used. FUCCI-expressing cells were generated by transducing HT-1080 cells with VSV-G pseudotyped lentiviral particles containing pBOB-EF1-FastFUCCI-Puro (Addgene; Plasmid #86849; RRID:Addgene_86849).

The combination index and synergy scores were analyzed using cell viability data obtained from high-throughput screening assay on a 96-well plate format following previously published protocols from our laboratory (20, 21, 25). To better predict synergism of drug combination treatment, cell viability data collected from combination treatment were processed by three different mathematical models: (i) Bliss independence (26, 27), (ii), Bayesian dose–response framework (28), and (iii) Chou–Talalay (29). The combination indices or summary synergy scores for each drug combination is the average of individual synergy scores obtained from each of the dose combination measurements. The detailed procedures for combination treatments and combination index and synergy analysis are included in the Supplementary Materials and Methods.

For the in vitro radio-tracer uptake test, cells seeded in 12-well plates were treated with IC25 concentrations of abemaciclib for 6 days before the indicated recovery period in drug-free media (cell density and drug concentration are presented in table S5). In vitro radio-biotracer uptake was performed following the methods of previous publications (30, 31) and is presented in Supplementary Materials and Methods.

Preclinical [¹⁸F]FLT–positron emission tomography (PET) and CT imaging were performed using a Bruker Albira PET/SPECT/CT scanner. Tumor-bearing mice were anesthetized and injected with 100 to 150 μCi of [¹⁸F]FLT tracer intravenously. After 45 minutes, PET imaging was initiated, followed by a subsequent CT scan (10 minutes) for anatomic reference. The imaging process was completed within 2 hours of the tracer injection. The PET analysis was performed with PMOD software (version 3.612, PMOD Technologies, Ltd., RRID:SCR_016547). [¹⁸F]FLT uptake in tumors was reported as a ratio of standardized uptake value (SUV) activity, obtained using the SUV of the most intense pixel in the tumor regions and normalized with the corresponding SUVmax obtained from the right/left leg muscle value from the same mouse. A series of PET/CT images were captured for each mouse over 4 different time points (baseline–before treatment initiated, following 6 days of abemaciclib treatment, 3 and 6 days following abemaciclib removal).

Blood samples were collected from tumor-bearing mice by retro-orbital bleeding or cardiac puncture at baseline, day 7 (6 days on), day 9 (3 days off), and day 12 (6 days off) of the treatment cycle. Serum was isolated from each blood draw collection, aliquoted, and stored at −80°C for future studies. Serum enzymatic activity of TK1 was determined by a DiviTum assay (Biovica International) using a refined ELISA-based method, according to the manufacturer's instructions (32).

All mouse experiments were performed under Institutional Animal Care and Use Committee–approved protocols at the University of Texas MD Anderson Cancer Center. For cell line xenograft models, 1×10⁶ HT-1080 (RRID:CVCL_0317) or HT-RB1-KO cells suspended in 150 μL of matrigel were injected subcutaneously into the right flank of 4- to 6-week-old female nude mice, which were randomized into treatment groups once the tumors reached 200 mm³. Leiomyosarcoma patient-derived xenograft (PDX) lines were generated from surgical samples under an IRB-approved protocol (LAB07-0659). Rhabdomyosarcoma PDX lines (Rh30, RRID:CVCL_0041 and Rh36, RRID:CVCL_M599) were gifted by Dr. Peter J. Houghton (Greehey Children's Cancer Research Institute, University of Texas Health Science Center at San Antonio). IHC conditions, including the source for all antibodies used, are included in Supplementary Table S6. The patient details can be found in Supplementary Table S3 and the in vivo therapeutic efficacy details can be found in Supplementary Materials and Methods.

Differences in the cohorts were evaluated using Fisher exact tests for categorical variables and ANOVA for age. The endpoint was OS, calculated from the time of diagnosis to death or last follow-up date. Patients who did not experience the endpoint were censored at the last follow-up. Median follow-up times were computed using the reverse Kaplan–Meier estimator. The Kaplan–Meier method was used to calculate the 5-year OS rate for each factor. The differences in survival curves were evaluated using the log-rank test. Univariable regression analyses were performed using the Cox proportional hazards model. Stata statistical software (SE 16, StataCorp LP; RRID:SCR_012763) was used for statistical analyses. All P values were two-tailed, and P ≤ 0.05 was considered statistically significant.

Linear regression and Pearson correlation analyses between protein expression and abemaciclib sensitivity were performed using GraphPad Prism 9 (RRID:SCR_000306). All P values were two-tailed with a 95% confidence interval.

All experiments were conducted with at least 3 biological replicates; each biological replicate had at least two technical replicates. The statistical analysis was carried out using Prism 9 (GraphPad Software Inc.). All statistical tests of comparative data were performed using a two-sided, unpaired Student t test or one-way or two-way ANOVA for a differential comparison between two or more groups, respectively. Statistical significance was denoted as P value > 0.05 (ns); *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; and ****, P ≤ 0.0001.

The data generated in the study are available within the article and its Supplementary Data files, and the raw data are available from the corresponding author upon reasonable request.

To determine whether Rb protein can be used as a stratifying biomarker for targeted therapy in STS patients, a Clinical Laboratory Improvement Amendments (CLIA)-approved Rb IHC assay was used to examine the expression of Rb in 1,043 sarcoma tumor specimens including RMS (n = 46), angiosarcoma (AS; n = 113), undifferentiated pleomorphic sarcoma (UPS; n = 102), radiation-associated sarcoma (RAS; n = 62), LPS (n = 100), uterine leiomyosarcoma (ULMS; n = 211), leiomyosarcoma (ST-LMS; n = 195), and osteosarcoma (Osteo, n = 214). Approximately 50% (range, 37% for OS to 91.3% for RMS) of all tumor specimens stained positive for Rb (cutoff at >5% positivity; Fig. 1A–C). OS was examined as a function of Rb expression, in three patient cohorts (AS, ST-LMS, and ULMS) that had a complete annotated follow-up with clinical and pathologic characteristics (Supplementary Table S1). OS analysis revealed that Rb status was not significantly associated with outcome (Fig. 1D). The median OS times for AS, ULMS, and ST-LMS patients were similar between the Rb-positive (Rb⁺) and Rb-negative (Rb⁻) cohorts within each subtype (Supplementary Table S1). Assessment of Rb staining in matched primary and recurrent tumor specimens from 24 patients (5 AS, 4 UPS, and 15 ULMS) showed that 80% of tumors retained Rb positivity in matched samples (Fig. 1E). Because Rb is the substrate for CDK4/6, we hypothesized Rb⁺ STS would be good candidates for CDK4/6i-based therapies, which are currently FDA approved in advanced hormone receptor–positive, HER2-negative (HR⁺, HER2⁻) breast cancer patients (33–35).

The CDK4/6is induce transient cell-cycle inhibition, which can be reversed when the drug is removed. To sustain the therapeutic efficacy of CDK4/6is, a two-step sequential combination strategy was developed in which the CDK4/6i abemaciclib acts first to prime Rb⁺ sarcoma cells to reversibly arrest in G1 phase of the cell-cycle (first step) and upon drug removal induce synchronized traverse to S phase. The second drug, an S phase–targeted agent (second step) is then used to mediate a synergistic cell-killing response (Supplementary Fig. S1A).

To identify Rb⁺ and Rb⁻ cell lines for analyzing drug sensitivity, four different sarcoma cell lines (Osteo, ST-LMS, UPS, and RMS; Supplementary Table S2) were subjected to Western blot analysis examining key cell-cycle proteins governing the Rb pathway (Fig. 2A). Total Rb and its phosphorylated forms (pRb) were expressed at high levels in HT-1080, SK-LMS1, and most RMS cell lines, whereas RD and SaOS2 exhibited low to undetectable protein levels (Fig. 2A). Further analysis of the proteins that regulate the G₁ to S transition (CDK4, CDK6, cyclin D1, p16, and E2F1) revealed no detectable differences among the cell lines. The sensitivity of the cell lines to abemaciclib was examined and results revealed that whereas the doubling times (Td) of the cell lines at baseline were similar (47.3–54.8 hours; Supplementary Fig. S1B) that there was a 10-fold difference in response to abemaciclib based on Rb status [IC₅₀ in Rb⁺ cell lines (0.1 –0.5 μmol/L) and in Rb⁻ cell lines (1–5 μmol/L); Fig. 2B]. Pearson correlation analysis of the cell-cycle protein levels (Fig. 2A) and IC₅₀ (Fig. 2B) of all eight cell lines revealed that sensitivity to abemaciclib was significantly (P < 0.05) associated with Rb and pRb status but not with any of the other cell-cycle proteins examined (Fig. 2C; Supplementary Fig. S1C). Knockout of Rb via CRIPSR/Cas9 in the Rb⁺ HT-1080 cell line resulted in undetectable levels of Rb in HT-RB1-KO, resembling Rb⁻ SaOS2 cells (Fig. 2D; Supplementary Fig. S1D). The deletion of Rb did not modulate the cell-cycle distribution or cell proliferation rates (Td = 39.3–48.5 hours; Fig. 2E and F; Supplementary Fig. S1D and S1E). However, with Rb deleted, HT-RB1-KO cells were resistant to abemaciclib with a 5-fold increase in IC₅₀ (Fig. 2G; P < 0.01). Compared with Rb knockout, p53 knockout had no effect on response to abemaciclib (Fig. 2H–K). CRISPR/CAS9 knockout of p53 in the HT-1080 cell line (harboring wild-type p53) did not alter the Rb pathway protein expression (Fig. 2H; Supplementary Fig. S1F), cell-cycle profile (Fig. 2I; Supplementary Fig. S1G), rate of proliferation (Fig. 2J, Td = 45–48.5 hours), or response to abemaciclib (Fig. 2K).

Rb status and transition of cells through each phase of the cell cycle were examined during and after treatment with abemaciclib. Treatment with increasing concentrations of abemaciclib for 6 days induced significant (P < 0.05) G₁ arrest in all Rb⁺ cell lines (80%–90% G1 cells) but not in Rb-knockout (HT-RB1-KO E20 sgRNA) or Rb-deficient cells (SaOS2 and RD; Fig. 3A and B; Supplementary Fig. S2A).

The next step was to determine whether the antiproliferative effect of abemaciclib in Rb⁺ cells is reversible and if this effect allows cells to synchronously reenter the cell cycle upon drug removal. Results showed that Rb⁺ cells (Rh30, SK-LMS1, and HT-1080) resumed cycling and synchronously entered the S phase (Fig. 3C–E; Supplementary Fig. S2B and S2C). Within 9 to 12 hours after abemaciclib removal, the fraction of cells in the S phase increased from 4% to 35% (i.e., 8.75-fold) between pre- and post-recovery time intervals in HT-1080 cells and increased 8.1- and 2.7-fold in SK-LMS1 and Rh30 cells, respectively. In contrast, the proportion of cells in the S phase remained unchanged in all three Rb⁻ cell lines (20%, 17.2%, and 19.8% in HT-RB1-KO, SaOS2, and RD cells, respectively) after 24 hours of drug removal (Fig. 3D–F; Supplementary Fig. S2B and S2C).

The reversible G₁ arrest following abemaciclib removal led to sequential changes in the expression of cell-cycle regulators in Rb⁺, but not Rb⁻ cell lines, which can be rationally targeted using sequential combination treatment with S (gemcitabine) or G₂–M (Wee1 kinase inhibitor) targeting agents (Fig. 3G; Supplementary Figs. S2D–S2F, S3, S4, and S5). These changes included a reduction in Rb phosphorylation on serine 807/811 (CDK4/6 phosphorylated site), E2F1, and thymidine kinase 1 (TK1, an E2F1 target), followed by temporal inductions without any detectable changes in CDK4, CDK6, and cyclin D1 expression, only in Rb⁺ cells (Supplementary Figs. S2E, S2F, S3A, and S3B). Examination of late S and G₂–M regulatory proteins revealed an increase in cyclin B, CDK1, phospho-CDK1 (pCDK1), and Wee1 kinase levels (Supplementary Fig. S5A and S5B). The expression of E2F1 and its downstream targets in Rb⁻ cells remained unchanged during the recovery period (Fig. 3G; Supplementary Figs. S2D–S2F, S3A, S3B, and S5A–S5C).

Sequential combination treatment with abemaciclib, followed by S phase targeting agents, was examined with gemcitabine, which is often utilized in STS patients, and a Wee1 kinase inhibitor (AZD1775) as a late S and G₂–M targeting agent for these studies. All sarcoma cell lines exhibited similar sensitivities to single-agent gemcitabine (IC₅₀ 1.9–2.2 nmol/L; Fig. 3H) or AZD1775 (IC₅₀ 0.2–0.4 μmol/L; Supplementary Fig. S5D), independent of Rb status.

Sequential and concomitant combination treatment with abemaciclib, followed by either gemcitabine or AZD1775, was examined in Rb⁺ and Rb⁻ cell lines as depicted by the treatment schedules shown in Fig. 3I, using the high-throughput survival assay (HTSA) methodology. HTSA was used to increase the capacity of screening drug combination in many cell lines versus classic clonogenic assays, which is not conducive to high-throughput combination treatment experiments. However, in side-by-side comparisons, the HTSA and clonogenic assays are highly concordant (24, 36–38), and HTSA has been routinely used to determine combination drug synergy in a variety of cell lines (22, 23, 39–42). The sensitivity of drug combination was initially assessed by comparing the viability of cells treated with single agents versus combination treatment (43). The results showed that only Rb⁺, but not Rb⁻ cell lines, were significantly (P < 0.05) sensitive to sequential combination treatment of abemaciclib and gemcitabine (Supplementary Fig. S3C and S3D) or Wee1 kinase inhibitor (Supplementary Fig. S3E and S3F).

To further infer the degree of drug interactions that contributes to the drug combination sensitivity, independent of the single drug effects, synergy scores of drug combinations were calculated using three different mathematical models: (i) Bliss independence (by SynergyFinder; refs. 26, 27), (ii) Bayesian dose–response framework (using “d-chain” program; ref. 28) for concomitant and sequential treatment strategies, and (iii) Chou–Talalay models for concomitant treatment strategies (using Calcusyn software; refs. 29, 44; Supplementary Figs. S3J, S3K, S4A, S4B, and S5E–S5H). The summary synergy scores for each drug combination are averaged over all the dose combination measurements for each method. Using Bliss independence model, our analysis showed that the sequential combination treatment with abemaciclib, followed by drug wash out, recovery, and addition of gemcitabine or AZD1775, resulted in synergistic cell killing in Rb⁺ with synergy scores greater than 10, indicating synergism of the sequential combination treatment in these cell lines (Fig. 3J and K; Supplementary Figs. S4A, S5E, and S5F). In contrast, Rb⁻ cell lines treated sequentially have synergy scores ranging from −10 to 7, underscoring the additive and antagonistic nature of the sequential combination treatments in these cell lines. Similarly, concomitant combination treatment failed to induce drug synergy in either Rb⁺ or Rb⁻ cell lines (Fig. 3J and K; Supplementary Figs. S4A, S5E, and S5F).

Consistent with the synergy scores calculated by the Bliss independence model, the synergy scores obtained from the Bayesian dose–response framework for sequential drug treatment in Rb⁺ cell lines were at least 4-fold higher than those obtained in Rb⁻ cell lines (Supplementary Figs. S4C and S5H). Using the Chou–Talalay approach, the concomitant treatment with abemaciclib + gemcitabine or abemaciclib + AZD1775 showed antagonism in Rb⁺ cells, with synergy score values greater than 1 for both HT-1080 and SK-LMS1 (Supplementary Figs. S4B and S5G). The sequential and concomitant combination treatment with abemaciclib and gemcitabine or AZD1775 was also antagonistic in Rb⁻ cells (average synergy score = 1.2 and 1.1 for sequential and concomitant treatment, respectively) and RB1-KO cells (average synergy score, 10–100; Supplementary Figs. S4B and S5G). The most synergistic cell killing was observed in Rb⁺ cell lines treated sequentially with abemaciclib and gemcitabine or AZD1775, with average synergy scores of 0.5 and 0.3 for SK-LMS1 and HT-1080 cells, respectively (Supplementary Figs. S4B and S5G). Collectively, all three mathematical models used in synergy score calculations confirmed that only when Rb⁺ cells are treated sequentially with abemaciclib followed by gemcitabine or Wee1 kinase inhibitor, the combination treatment is synergistic.

The mode of cell death to the sequential combination of abemaciclib followed by gemcitabine was examined using the Annexin V/PI assay (Supplementary Fig. S6A and S6B). Single-agent abemaciclib or gemcitabine induced apoptotic cell death (12%–20%) in the Rb⁺ cell lines HT-1080 and Rh30; however, sequential treatment with abemaciclib followed by gemcitabine resulted in significantly higher rates of apoptosis (>80% for HT-1080 and >30% for Rh30, P < 0.0001). Concomitant treatment with abemaciclib + gemcitabine resulted in similar levels of apoptosis as treatment with single agents (i.e., 8%–22%; Supplementary Fig. S6A and S6B). Cell-cycle analysis revealed an increase in the sub-G₁ population of Rb⁺ cell lines treated with sequential combination treatment but not in those cells treated with single-agent or concomitant therapy (Supplementary Fig. S6C and S6D).

To monitor cells in each phase of the cell cycle in response to the different treatment strategies (monotherapy, sequential, or concomitant combination treatment), fluorescent ubiquitination-based cell-cycle indicator (FUCCI)–expressing HT-1080 Rb⁺ cell lines (Fig. 4A) were generated, where cells emit red fluorescence during G₁, yellow during early S, and green during late S/G₂–M (Fig. 4B). Treatment with abemaciclib induced G₁ arrest in FUCCI-expressing cells, as demonstrated by 80% of cells emitting red fluorescence (Fig. 4C–E). Eighteen hours after the removal of abemaciclib, 60% of cells emitted yellow and green fluorescence, indicating enrichment of cells in S/G₂–M phases of the cell cycle (Fig. 4C). Treatment of cells with gemcitabine did not alter cell-cycle distribution (Fig. 4D and E; Supplementary Table S3). However, treatment with abemaciclib followed by gemcitabine resulted in approximately 50% of cells enriched in early S (yellow) and G₂–M (green) phases (Fig. 4E). Concomitant combination treatment captured significantly (P < 0.05) fewer cells in S phase (34%) with a smaller fraction of cells in G₀–G₁ (29%; Fig. 4E; Supplementary Table S3).

DNA damage was assessed by quantitation of γH2AX (marker of DNA double-stranded break) in FUCCI-expressing cells, after each treatment strategy (Fig. 4D). In cells treated sequentially with abemaciclib followed by gemcitabine, the average number of γH2AX foci per cell was 5.1, significantly higher (P < 0.001) than the concomitant (i.e., 2.7 foci/cell) or single-agent (i.e., < 1.6 and 1.8 foci/cell for abemaciclib and gemcitabine, respectively) treatment strategies (Fig. 4F; Supplementary Table S4). In the sequential treatment group, 40% of cells had more than five γH2AX foci, 2- and 3-fold higher than that of cells in the concomitant and single-agent treatment groups, respectively (Fig. 4G). There was a significantly higher (P < 0.0001) number of γH2AX foci per cell induced by sequential treatment in both G₀–G₁ nonproliferative cells and S/G₂–M proliferative cells (Fig. 4H and I). These results imply that sequential targeting of the CDK4/6 and DNA synthesis pathways potentiates DNA damage in all phases of the cell cycle in which cells from the previous cycle may carry DNA damage upon reentry into the next cell cycle. Consistently, the significant induction (P < 0.001) of γH2AX foci was observed only in abemaciclib-treated cells but not in cells treated with other CDK4/6is, such as palbociclib or ribociclib (Supplementary Fig. S7A–S7C).

Cell line (CDX) and PDX models were used to examine the treatment strategies in vivo. The time required for S phase reentry in tumors cells in vivo was optimized using the schedule and dose of abemaciclib shown in Supplementary Fig. S8A. Treatment of mouse models with abemaciclib at 75 mg/kg for 6 days resulted in up to 25% regression of tumor growth compared with day 0; Fig. 5A).

Rb, pRb, and TK1 protein levels were downregulated after 6 days of treatment with abemaciclib, with levels returning to baseline 3 to 6 days after drug removal (Fig. 5B–D). BrdUrd labeling (as a measure of tumor cell proliferation) revealed that treatment with 75 mg/kg per day for 6 days was sufficient to inhibit S-phase fraction of the tumor cells (i.e., <5% for the 75 mg/kg treatment group compared with 20% BrdUrd-positive cells in the vehicle arm). Levels returned to baseline at 3 (>25%) and 6 (30%) days of recovery, following abemaciclib removal (Fig. 5C and D). Similar trends were detected when examining Ki67 and Rb status across all treatment groups (Fig. 5C and D).

LMS and RMS PDX models comprising Rb⁺ and Rb⁻ subtypes were used to examine the experimental approach (Fig. 5E; Supplementary Fig. S7B). Of 10 PDX models examined, 44M, Rh30, and Rh36 showed positive Rb and pRb expression accompanied by intact CDK4, CDK6, and cyclin D1, indicating that these models harbor a wild-type Rb pathway phenotype. Rh30 RMS tumor growth regressed and failed to increase in volume following 6 days of abemaciclib treatment (Fig. 5F; Supplementary Fig. S8B). The expression of proliferative biomarkers such as TK1, BrdUrd, and Ki67 was consistent with tumor volume status, with enrichment in the S phase detected after 3 days of recovery (Fig. 5G and H; Supplementary Fig. S8B).

The sequential combination treatment strategy of abemaciclib followed by gemcitabine treatment was evaluated in vivo, using Rb⁺ cell lines and PDX models (Fig. 6; Supplementary Fig. S9). Based on the ex vivo biomarker analysis (Fig. 5), mice were treated with the sequential combination treatment for 12 days/cycle; each cycle comprised 6 consecutive days of treatment with abemaciclib (orally, q.d., 75 mg/kg), recovery for 3 days and 1 day of gemcitabine treatment (i.p., q.d., 50 mg/kg; Supplementary Fig. S9A).

In the HT-1080 model, there was a regression of tumors in mice treated with sequential combination therapy for three cycles (Fig. 6A–C). Concomitant treatment with abemaciclib and gemcitabine was similar to treatment with either single-agent therapy (Fig. 6A). There was 80% tumor growth inhibition (TGI) at the end of the first cycle with sequential combination treatment and only 25% to 50% TGI with either a single-agent or concomitant treatment (Fig. 6A and B). Similar results were observed with the treatment-naïve Rb⁺ RMS Rh30 PDX model, where sequential combination treatment resulted in 80% TGI (first cycle), which was more effective than the single-agent treatments (25% and 47% TGI for gemcitabine and abemaciclib, respectively) or concomitant treatment (26% TGI; Fig. 6D–F). Rh36, an RMS Rb⁺ PDX model collected from a patient who experienced disease progression on chemotherapy (DITC/DOX), was resistant to gemcitabine (−32% TGI) but was sensitive to abemaciclib (22% TGI; Supplementary Fig. S9B–S9D). This was enhanced when gemcitabine was administered sequentially following recovery from abemaciclib treatment (46% TGI).

Survival analysis performed after 3 cycles of treatment revealed that the sequential combination of abemaciclib and gemcitabine was more effective than individual drugs or concomitant treatment, resulting in longer survival times (Fig. 6C and F; Supplementary Fig. S9D). In the HT-1080 model, 100% of the mice in the sequential treatment group remained alive until day 40, whereas all of the mice in the single agent and concomitant treatment groups had to be euthanized by day 25, because of large tumor burden (Fig. 6C). In the sequentially treated Rh30 PDX model, there was an increase in survival duration (38 days) compared with the concomitant (27 days) or single-agent–treated mice (23 and 19.5 days for abemaciclib and gemcitabine, respectively; Fig. 6F). The Rh36 PDX model also benefited from sequential combination treatment. The median survival time with sequential combination therapy was 27 days, compared with 17 days for concomitant, 13 days for gemcitabine, and 15 days for abemaciclib treatment alone (Supplementary Fig. S9D).

Levels of γH2AX, cleaved PARP, cleaved caspase-3, and pATM all increased in mouse tumors treated sequentially with abemaciclib and gemcitabine compared with the single agent or concomitant treatment (Fig. 6G–I; Supplementary Fig. S9E). There was a decrease in proliferation, as indicated by the percentage of phospho-Histone H3 (pHH3) positive cells, only evident in the sequential treatment cohort (Fig. 6G–I).

To translate this sequential treatment strategy clinically, in vivo real-time monitoring modalities were evaluated. TK1 is a key enzyme in the pyrimidine salvage pathway that takes part in the initial thymidine phosphorylation process required for DNA synthesis. TK1 expression is strongly associated with Rb⁺ cell-cycle progression both in vitro (Fig. 3G; Supplementary Fig. S2D–S2F) and in vivo (Fig. 7A and B, P < 0.001). The changes in tumor TK1 can be captured by functional PET imaging using ¹⁸F-fluoro-3′-deoxy-3′-L-fluorothymidine (¹⁸F-FLT) as a TK1-regulated proliferative tracer and by circulating thymidine kinase activity (TKa) in the in vivo models (Fig. 7C). Although recovery from abemaciclib treatment did not alter ¹⁸F-FLT uptake in Rb⁻ and RB1-KO cells, a 10-fold increase in biotracer uptake after 12 to 18 hours of recovery was detected in Rb⁺ cell lines that were pretreated with abemaciclib (Fig. 7D). Unlike ¹⁸F-FLT, which directly measures thymidine uptake, neither ¹⁸F-FDG, a glucose analogue, nor its regulator hexokinases II (HKII) were able to capture the recovery of Rb⁺ (Supplementary Fig. S10A–S10D).

Finally, the utility of [¹⁸F]FLT–PET/CT imaging in monitoring the reversible antiproliferative activity of abemaciclib in vivo was examined. To demonstrate the kinetic changes in ¹⁸F-FLT tumor uptake from individual abemaciclib-treated mice of a cohort of n = 3/model, a series of images were taken for each mouse over 4 different time points (baseline, 6 days on, 3 days off, and 6 days off). Mice bearing Rb⁺ tumors displayed a 40% reduction in tumor ¹⁸F-FLT uptake compared with the vehicle-treated or Rb⁻ cohorts (Fig. 7E–H). Three days following drug removal, the intratumoral ¹⁸F-FLT accumulations across all Rb⁺ mice examined were elevated by 10-fold compared with mice on-treatment (6 days on) and 4-fold compared with no treatment (Fig. 7E and G). Following 6 days of recovery, the ¹⁸F-FLT signal returned to basal levels, indicating the return of tumor cells to an asynchronous proliferative state (Fig. 7E and G). In HT-RB1-KO tumors, intratumoral ¹⁸F-FLT accumulation was not responsive to abemaciclib treatment at any time interval during or after treatment (Fig. 7F–H), consistent with the lack of change in TK1 expression after abemaciclib treatment in vitro (Supplementary Fig. S2E and S2F).

TKa in the serum of treated mice was evaluated and correlated with ¹⁸F-FLT uptake. TKa was measured using the DiviTum assay (45) in sera collected from Rb⁺ HT-1080 and Rb⁻ HT-RB1-KO tumor-bearing mice (Fig. 7I and J). This revealed that TKa mirrors ¹⁸F-FLT uptake (Fig. 7E and G) and is significantly associated with TK1 protein expression in tumors (Fig. 7A and B; P < 0.001). TKa changes were observed in Rb⁺ models as a measure of response to abemaciclib, but no significant changes in TKa were observed in the Rb⁻ tumor model (Supplementary Fig. S10E).

The clinical management of sarcomas is challenging due to the rarity, heterogeneity, and aggressive nature of these tumors. Conventional chemotherapy is the mainstay of therapy, but survival benefit is limited, and almost all patients progress on the few standard treatment options available. Although molecularly targeted approaches to STS treatment have gained traction in recent years, the reported OS for targeted therapies, such as pazopanib, was only modestly prolonged compared with that of placebo (12.5 vs. 10.7 months; ref. 46). There is a clear need for new rational biomarker-driven therapeutics to improve patient outcomes.

CDK4/6is have shown significant clinical benefit in breast cancer patients with advanced or metastatic hormone receptor–positive disease (33–35, 47). These promising results have prompted the use of CDK4/6is in other cancers harboring deregulation of the cyclin D-CDK4/6-INK4-Rb pathway, including sarcomas (15, 48–51). However, reports from phase II trials in advanced LPS sarcoma patients with CDK4 overexpression treated with palbociclib or abemaciclib showed only modest efficacy, with only 1 out of 60 palbociclib-treated patients showing response (15, 51). Given the lack of effective therapeutic options for this LPS subtype, CDK4/6is are included as an option in the National Comprehensive Cancer Network guidelines even though not formally approved by any regulatory agencies. The use of CDK4 overexpression as a stratifying biomarker may have been the main limitation for patient selection in these studies, because upregulation of CDK4 or CDK6 contributes to the CDK4/6i resistance observed in various cancer types (52). Other biomarkers, such as CDKN2A loss/mutation or CCND1, 2, or 3 amplification, also failed to predict response to CDK4/6is (53–55). On the basis of these negative clinical findings, we hypothesized that a more functional marker for response to CDK4/6i in STS may be the Rb protein. Indeed, Rb was the only marker that persisted across all STS models we examined for response to abemaciclib. This finding provides a compelling rationale to evaluate abemaciclib activity in subtypes of sarcoma beyond those with CDK4 overexpression. None of the other G₁/S regulators, including p16, cyclin D1, CDK4/6, and cyclin E1, showed a correlation with response to abemaciclib across multiple sarcoma models (Fig. 2).

There is a precedent for using Rb as a biomarker for patient selection for CDK4/6is. Results from the PALOMA 1 and 2 trials in breast cancer patients revealed that PFS in palbociclib-treated patients whose tumors harbored RB1 loss was 3 times shorter than that in patients with intact RB1 (3.6 months vs. 10.1 months; refs. 53, 55–58). These results suggest that RB1 loss is associated with resistance or lack of response to CDK4/6is in breast cancer patients. In sarcomas, only 20% of all STS cases examined exhibited RB1 loss/mutation at the genome level (59). The expression of Rb protein in over 1,000 tumor specimens with different sarcoma histologies showed that over 50% retained Rb expression. There was no apparent survival benefit from standard-of-care management (surgery, chemotherapy, and radiotherapy) when patients were stratified based on Rb status, suggesting that other treatment strategies, such as CDK4/6is, which are likely to show response in the patients with Rb-positive tumors, could be offered to these patients. Notably, more than 90% of rhabdomyosarcoma and angiosarcoma tumors are Rb-positive, making these patient cohorts ideal candidates for clinical trials of CDK4/6is sequential combination therapy.

Because CDK4/6is are cytostatic agents that induce cell-cycle arrest (i.e., G₁) rather than cell death, combination strategies are needed to maximize therapeutic efficacy. However, the optimal treatment approach for combining CDK4/6is with chemotherapy, given sequentially or concurrently, has remained controversial. A recent study that examined 108,259 pairs of combination drug treatments across breast, colon, and pancreatic cancer cell lines revealed that the frequency of synergy observed in concurrent combination treatment is a rare event, with only 5.2% of all combinations examined showing synergistic response (60). Specifically, data from the concomitant combination of palbociclib with gemcitabine in 52 breast cancer cell lines were 18-fold less potent (median log₂[∆IC₅₀] = 4.2) compared with the expected combination response for synergy threshold (60). One reason for the lack of synergism of concurrent combination treatments with palbociclib may be that inhibition of CDK4/6 results in G₁ arrest in cells, not allowing the chemotherapeutic agents that target the S phase to mediate their cytotoxic activity (17).

A combination treatment strategy with CDK4/6is (followed by chemotherapeutic agents) that is biomarker-driven, mechanism-based, and limits the overlapping toxicities of the two agents through sequential treatment is a novel approach, presented in this report. Specifically, a two-step sequential combination strategy in which abemaciclib acts to prime Rb-positive tumor cells to reversibly arrest at G₁ phase, followed by a short recovery period to allow cells to synchronously enter S phase, at which point, they are treated with S phase-targeting drugs, such as gemcitabine. The synergistic cell-killing effects mediated by sequential combination treatment with abemaciclib and gemcitabine are effective in several in vitro and in vivo models of sarcoma, including a chemotherapy-resistant PDX model. The mechanism of this treatment strategy is through sustained induction of DNA damage in both proliferative and nonproliferative cells.

A challenge in the translation of the proposed sequential combination treatment is the identification of real-time pharmacodynamic biomarkers to ascertain when CDK4/6is induce G₁ arrest and how long it takes for the tumor cells to enrich in the S phase in patients undergoing such treatment strategies. One such biomarker is TK1, which can be measured in tumor tissues (Fig. 7A and B). Its activity can be quantitated in serum (Fig. 7I and J), and its effector uptake, as a biotracer, can be monitored through functional imaging (PET/CT; Fig. 7E-H). The link between tumor cell proliferation and serum TK1 was reported with the neoadjuvant NeoPalAna phase II “To Reverse ENDocrine Resistance” (TREnd) trial (NCT02549430), in which changes in plasma TK1 were concordant to Ki67 and can be used to predict response to palbociclib (32, 61). These clinical studies suggest that TK1 is a valid pharmacodynamic marker for tumor proliferation. TK1 levels and activity in Rb⁺ models were highly concordant with tumor proliferation status pre- and post-abemaciclib treatment. As a result, real-time monitoring through functional PET imaging and a blood-based biomarker, in concert with the ex vivo biomarkers presented herein, can comprehensively monitor the early treatment response of abemaciclib-based sequential combination treatment, and be translated into future clinical trials.

One limitation of our study is the reliance IHC to identify Rb⁺ tumors. It is important to also perform a genomic analysis of the RB1 gene and correlate its genetic loss or mutations to Rb protein expression. Our data show that therapeutic benefits can be achieved in multiple PDX models as long as the Rb pathway is intact, although the mechanism of cell death induction will be different as each cancer type and CDK4/6i have distinct biological and pharmacodynamic profiles. In addition, the mechanism by which tumor cells secrete TK1 into the circulation and how CDK4/6is treatment affects such mechanisms remain unclear. These analyses will be instrumental in determining the utility of these biomarkers in patients with a functional Rb pathway who can be identified upfront as potential responders to CDK4/6i-based sequential combination therapy.

In summary, a novel, sequential treatment strategy with abemaciclib followed by gemcitabine (or Wee1 kinase inhibitor) is effective in inhibiting tumor growth in vitro and in vivo. Rb is an important mediator of this synergistic activity and Rb expression can readily be assessed in tumors using IHC. Finally, the cellular response to this sequential treatment can be monitored noninvasively by ¹⁸F-FLT uptake or circulating TK1 activity in the serum, and through IHC of TK1 in tumor tissues.

No disclosures were reported.

T.D. Nguyen: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. Y. Wang: Resources, data curation, validation, investigation, methodology, writing–review and editing. T.N. Bui: Resources, data curation, validation, investigation, methodology, writing–review and editing. R. Lazcano: Data curation, formal analysis, writing–review and editing. D.R. Ingram: Resources, data curation, methodology, writing–review and editing. M. Yi: Data curation, software, formal analysis, methodology, writing–review and editing. V. Vakulabharanam: Data curation, software, formal analysis, methodology, writing–review and editing. L. Luo: Data curation, software, formal analysis, methodology, writing–review and editing. M.A. Pina: Data curation, formal analysis, investigation, methodology, writing–review and editing. C. Karakas: Formal analysis, investigation, visualization, writing–review and editing. M. Li: Data curation, investigation, methodology, writing–review and editing. N.M. Kettner: Investigation, methodology, writing–review and editing. N. Somaiah: Conceptualization, resources, data curation, methodology, writing–review and editing. P.J. Houghton: Resources, writing–review and editing. O. Mawlawi: Resources, formal analysis, supervision, investigation, methodology, writing–review and editing. A.J. Lazar: Conceptualization, resources, data curation, formal analysis, supervision, investigation, writing–review and editing. K.K. Hunt: Conceptualization, resources, supervision, funding acquisition, investigation, visualization, methodology, writing–review and editing. K. Keyomarsi: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

The research reported in this article was supported by the NIH under award P30CA016672 to The University of Texas MD Anderson Cancer Center, the Department of Defense Breakthrough Post-Doctoral Fellowship BC170615 (to N.M. Kettner), NIH, NCI R01 grants CA255960 and CA223772 to K. Keyomarsi, CPRIT multi-investigator grant RP180712 (to K.K. Hunt and K. Keyomarsi), CPRIT Research Training Program grant RP170067 and RP210028 (to N.M. Kettner and M. Li), the National Leiomyosarcoma Foundation (to K. Keyomarsi and K.K. Hunt), and the John Wayne Cancer Foundation (K.K. Hunt). The authors thank Dr. Natividata (Robert) Fuentes from MD Anderson Cancer Center, Department of Radiation Oncology for his advice on FUCCI system, cell confocal image analysis. They also thank Dr. Lucia Martini from MD Anderson Cancer Center, Department of Experimental Therapeutics for her advice on the PET/CT analysis on the mouse models. The authors are also grateful to Dr. Amy Williams (Biovica Inc.) who advised them on the use of and provided them with the DiviTumTKa kits, free of charge, that they used for both in the in vitro and in vivo studies presented in this report.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.