Background
Human papillomavirus negative (HPV
neg) head and neck squamous cell carcinoma (HNSCC) is a major subtype of HNSCC, which is associated with poorer clinical outcomes and lower response rate to chemo-radiotherapy as well as treatment targeting EGFR and PD-1 when compared to HPV-positive disease [
1‐
5]. Deregulation of cell-cycle signaling mediated by
CCND1 and
CDKN2A aberrations was observed in more than half of the HPV
neg HNSCCs [
6]; a finding that strongly recommends the investigation of CDK4/6 inhibitors in this disease. Considering the mechanisms of CDK4/6 inhibition, profound effect that causing tumor regression was rarely observed when CDK4/6 inhibitor was used as a monotherapy in HNSCC [
7,
8]. On the other hand, when combined with endocrine therapy, CDK4/6 inhibitor has become a first-line therapy as a combination drug with endocrine therapy for estrogen receptor (ER)-positive and HER2-negative breast cancer [
9], and multiple clinical trials are exploring various combination treatments involving CDK4/6 inhibitors [
10].
Palbociclib is the first FDA-approved CDK4/6 inhibitor which is currently under active clinical investigation as the combinational agent in HNSCC [
11]. Currently, several phase I-II clinical trials are in progress to evaluate palbociclib in combination with cetuximab or platinum-based chemotherapies in HNSCC (NCT02499120, NCT02101034, NCT03024489, NCT03065062, and NCT03088059). Very recently, results from these trials showed conflicting findings regarding the therapeutic efficacy using combination of palbociclib plus cetuximab [
12,
13], while combination of platinum-based chemotherapy plus palbociclib showed insufficient antitumor activity in recurrent/metastatic HNSCC patients [
14]. Thus, rational combinations with palbociclib would fulfill the promise of targeting the cell-cycle pathway in HPV
neg HNSCC and could be more readily integrated to the current treatment regimens during clinical translation.
To circumvent potential drug resistance which is commonly observed using monotherapies and induce a synergistic treatment effect, drug combination strategy is currently the mainstream approach in cancer treatment [
15]. More recently, systematic matrix screening of drug combinations using large-scale compound libraries offers a more preferred pipeline to rapidly identify robust single agent as well as assess the potential of candidate agents as synergistic combinations [
16‐
18]. Additionally, unbiased matrix screening represents a high-throughput means to provide a roadmap of drug-drug pairs that are synergistic, additive, or antagonistic [
19]. By using high-throughput drug screening method, multiple promising drug combinations are currently under clinical investigation, like NCT02756247, in which BKM120 (a PI3K signaling pathway inhibitor) was previously discovered that would cooperate with ibrutinib (an inhibitor of the Bruton’s tyrosine kinase) are being evaluated for the treatment of activated B-cell-like diffuse large B-cell lymphoma. To our knowledge, few such studies have been conducted to define the synergistic potential of other agents when combined with CDK4/6 inhibitors.
Numerous studies showed sensitivity of targeted therapy generated from patient-derived xenograft models that closely recapitulated the genetic and biological features in the original cancer patients [
20]. PDX models that contain certain genetic feature like BRAF-mutated melanoma PDX models [
21] or ER-positive breast cancer PDX models [
22] have been selected as more clinically relevant models for targeted drug evaluation. Previously, we have established and conducted a population-based PDX trial consisting of 24 models established from 24 melanoma patients, which demonstrated robust antitumor effect of palbociclib in CDK4-amplified melanoma [
23]. Thus, molecularly defined PDXs cohort study could serve as a preferred pipeline for targeted drug evaluation and translation, especially for the repurposing of FDA-approved agents.
Here, seeking to identify potential therapeutic agents to boost efficacy of CDK4/6 inhibition, we conducted a high-throughput combination drug screening to evaluate a collection of FDA-approved and investigational drugs against multiple molecularly defined HPVneg HNSCC cell lines. After mapping the drug-drug interaction landscape using the screening data, we identified the most promising candidate combinations, which were further evaluated in vitro across an expanded panel of representative cell lines and in vivo using five molecularly annotated HPVneg PDX models. Very briefly, our study provides preclinical evidence for particular targeted agents would synergize with palbociclib in HPVneg HNSCC patients.
Methods
Cell cultures
HPVneg cell lines used in this study were HN6 (RRID: CVCL_5516), HN30 (RRID: CVCL_5525), CAL27 (RRID: CVCL_1107), CAL-33 (RRID: CVCL_1108), SCC4 (RRID: CVCL_1684), SCC9 (RRID: CVCL_1685), PECA-PJ15 (RRID: CVCL_2678), PECA-PJ41 (RRID: CVCL_2680), FADU (RRID: CVCL_1218), UPCI-SCC-172 (RRID: CVCL_2231), UPCI-SCC-131 (RRID: CVCL_2229), Detroit-562 (RRID: CVCL_1171), and HSC-2 (RRID: CVCL_1287). Human breast cell line MCF-7 (RRID: CVCL_0031) was purchased from ATCC. All cell lines were routinely cultured at 37 °C with 5% CO2 according to the manufacturer recommendations. All cell lines were authenticated using short tandem repeat analysis and confirmed as mycoplasma-free (YEASEN). All cell lines were used for experiments within less than 20 passages.
High-throughput drug screening
Chemical library
We purchased a chemical screening library composed of 162 compounds from Selleckchem for the combination drug screening with palbociclib. The corresponding information, including drug name, MoA (mechanism of action) and resource, is provided in Additional file
2: Table S1. The drugs used in cancer treatment and clinical trials were given priority, especially in HNSCC. To include different MoAs used in cancer treatment, our chemical library was constructed based on several common drug lists published in cancer studies [
24,
25].
Combination high-throughput screening
In the primary drug combination screening, CAL27, HN6, FADU, and SCC9 cells were seeded into 96-well polystyrene tissue culture-treated Corning plates with 3000–6000 cells/well, depending on the specificity of each cell line. During the exponential growth phase, we added 10 μL of palbociclib and 10 μL 162 compounds (10X concentration) to individual wells for a 6×6 matrix screening (a 5-point custom concentration range, with constant 1:5 dilution between each point) (concentration ranging from 20 to 0.032 μM). Bortezomib (final concentration 3 μM) was used as a positive control for cell cytotoxicity. After 72 h of treatment, 10 μL of CCK8 (Beyotime) was added to each well. The plates were transferred to a standard incubator with a stainless steel lid for 2 h. OD values were taken using Synergy H1 (Biotek). All plates passed Quality Control conditions, with a Z factor greater than 0.6. In the secondary drug combination screening, cells were seeded into 384-well polystyrene tissue culture-treated plates at a concentration of 800–2000 cells/well. A total of 10 μL of compounds were dispensed for a 10×10 matrix screening (a 9-point custom concentration range, with constant 1:3 dilution between each point) (concentration ranging from 20 to 0.003 μM). The plates were incubated for 72 h in a standard incubator covered by a stainless steel gasketed lid to prevent evaporation. We added CCK8 and measured OD values. The efficacy of different drug combinations was estimated using the ExcessHSA score and a Bliss independence model.
Cell viability assay
Cells were placed in 96-well plates to perform drug combination in three independent replicates. Cell viability was measured using Cell Counting Kit-8 assay according to the manufacturer’s instructions (Beyotime). To evaluate the effect of drug combination, cells were treated with each drug individually or in combination for 72 h. Data were processed in GraphPad Prism 5.0 (GraphPad Software, Inc.). Combination index (CI) was determined using CalcuSyn software (Biosoft), which allowed us to measure drug synergy using the median effect model (T.-C. Chou) [
26]. A CI lower than 1.0 was considered to be synergistic.
Cell-cycle analysis
After 48 h of palbociclib or combination treatment, HNSCC cells were harvested, washed, fixed in 75% ethanol overnight, and centrifuged at 300g for 5 min. The pellets were then washed and resuspended in 500 μL PI/RNase Staining Buffer (BD Pharmingen™). Cell-cycle analysis was performed on a BD Facscalibur. Data were processed using FlowJo v.10 (FlowJo LLC).
EdU assay
Cell proliferation ability was measured using a Click-it EdU imaging kit (Beyotime Biotechnology, Alexa Flour 555). Briefly, post-treatment cells were incubated in culture medium with 10 μM EdU for 2 h. After fixation with 4% PFA and permeabilization with 0.3% Triton X-100, cells were incubated in Click-iT Plus reaction cocktail at room temperature for 30 min. Cell nuclei were stained with 10 μg/mL Hoechst 33342 for 2 h. Images were captured with Axio Vert.A1 and analyzed using ZEN software.
Whole exome sequencing and data analysis
All HPVneg cell lines, PDXs, and paired patients’ tumors were subjected to whole exome sequencing. DNA concentration of the enriched sequencing libraries was measured with the Qubit 2.0 fluorometer dsDNA HS Assay (Thermo Fisher Scientific). Size distribution of the resulting sequencing libraries was analyzed using Agilent BioAnalyzer 2100 (Agilent). DNA sequencing was performed with paired-end 2 × 150 base reads on the Illumina NovaSeq6000 platform at Mingma Technologies Co., Ltd. Raw FASTQ files were initially processed by a proprietary algorithm to filter out contaminated mouse sequencing reads.
Somatic mutations detection
For each paired sample, somatic SNVs and InDels were detected with Sentieon TNseq. We excluded mutations in low-complexity regions, including tandem repeats and highly homologous regions. Low-confidence variants were removed in case any of the following criteria was not satisfied: total depth > 10, alternative allele depth > 3 and mutation frequency > 0.01. All high-confidence mutations were then annotated with ANNOVA (Version 2016-02-01).
Copy number variant detection
We used CNVkit to analyze somatic copy number variations (CNVs). CNVs were called by comparing normalized tumor and normal data. Regions with absolute log2 copy number ratios at least 0.58 (= log2 (1.5)) were viewed as losses (deletions) or gains (amplifications) [
27].
RNA sequencing and data analysis
RNA sequencing was performed according to our previous research [
28]. Briefly, cells were treated with either a drug or vehicle for 48 h, and then lysed in Trizol and stored at – 80 °C. Total RNA was extracted using the RNeasy Mini Kit (Qiagen) following the manufacturer’s instructions. We checked RIN number to evaluate RNA integrity using Agilent Bioanalyzer 2100 (Agilent technologies, Santa Clara, CA, US). Qualified total RNA was further purified using the RNAClean XP Kit (Beckman Coulter, Inc. Kraemer Boulevard Brea, CA, USA) and RNase-Free DNase Set (QIAGEN, GmBH, Germany). Library preparation was completed with the Illumina TruSeq RNA Sample Preparation Kit (Illumina). RNA sequencing was performed on a Illumina Hiseq 2000 platform. Raw reads were trimmed with Skewer (v0.2.2) to remove adapter sequences, and then aligned to the reference human genome GRCh37/hg19 with STAR (v2.4.2a). Gene expression abundance was estimated using RSEM (1.2.29) based on reads mapping uniquely to specific parts of the human genome. The presence of differentially expressed genes (DEGs) was determined using a negative binomial model with EdgeR, by comparing DMSO- and drug-treated tumors.
P-values were calculated using the Wald test after controlling for multiple testing using B-H procedure. A |fold change| > 2 and an adjusted
P-value < 0.05 were used as cut-off criteria.
Gene Set Enrichment Analysis (GSEA) was performed on normalized RNA-Seq expression data using the Desktop Application [
29] and using the genes whose average CPM value was greater than 100 in at least one of the tested conditions (either DMSO or any of the 6 drug-treated sets). For each individual treatment vs. DMSO or palbociclib, a pre-ranked GSEA was performed based on the log2FC values, measured against “Hallmark” gene sets and epithelial or mesenchymal gene sets (Additional file
2: Table S2). Weighted enrichment statistics were based on 1000 gene-set permutations. Only the gene sets with an FDR adjusted
q-value < 0.1 were selected for further analysis.
Quantitative real-time PCR
Total RNA was extracted using TRIzol reagent (Invitrogen) following the manufacturer’s instructions. RNA was reverse transcribed using a Prime-Script RT Reagent Kit (Takara). Quantitative real-time PCR (qRT PCR) was conducted using a Roche LightCycler system with SYBR Green Reagent (Takara). Gene expression levels were normalized based on the GAPDH level. The primers used for qRT PCR are listed in Additional file
2: Table S3. Data were analyzed using the 2
−ΔΔCT method.
Western blot analysis
Cells were lysed in SDS Lysis Buffer supplemented with protease and phosphatase inhibitor cocktail. Protein concentration was determined using BCA assay. Gel electrophoresis was performed on a stacking gel (90 V) and a separation gel (120 V). The protein samples were then transferred to PVDF membranes with 300 mA for 1.5 h. The membranes were incubated with primary antibodies at 4 °C overnight (Additional file
2: Table S4). Horseradish peroxidase (HRP) and secondary antibodies were used at room temperature for 1 h in the following day. Ultimate blots were visualized with enhanced chemiluminescence (ECL; Thermo Fisher Scientific).
Generation of stable RRM2-overexpressing cell lines
pHBLV-CMV-MCS-3FLAG-EF1-ZsGreen-T2A-PURO and vector (Hanbio) were constructed and transfected into FADU and CAL27 cells using Fugene HD [
30]. RRM2 overexpression was validated by Western blot analysis. Stable cells were routinely cultured and authenticated according to the ATCC guidelines.
Histology and immunohistochemistry
Patient tissues and PDXs were embedded in paraffin. After this, we generated 4-μm-thick paraffin sections. Tissue morphology was determined by hematoxylin and eosin (H&E) staining. For immunohistochemistry (IHC) assays, paraffin sections were dewaxed and rehydrated through a graded ethanol series. After being treated with heat-mediated antigen retrieval using Citrate Unmasking Solution (Cell Signaling Technology), the sections were blocked with goat serum at room temperature for 1 h. Primary antibodies (Additional file
2: Table S4) diluted in 3% BSA were incubated at 4 °C overnight. We used DAB (Cell Signaling Technology) to perform the color reaction. All H&E and IHC images were obtained using an OLYMPUS microscope BX51 and a DP 71 camera (OLYMPUS). The positive grade was determined by the immuno-reactive score (IRS), which was according to the staining intensity (negative = 0, weak = 1, moderate = 2, strong = 3) and the percentage of positive tumor cells (0%=0, 1–10%=1, 11–50%=2, 51–80%=3, 81–100%=4). IRS ranging from 0 to 12 was calculated from the values of the staining intensity multiplied by the percentage of positive cells [
31].
Cell migration and invasion assays
Cell migration and invasion assays were performed using Transwell assay with uncoated polycarbonate transwell inserts (Millipore) for migration, or BioCoatTM transwell champers (Corning) for invasion. 10 ~ 12 × 104 transfected cells were seeded in the upper chamber. After staining with crystal violet, positive cells were counted and analyzed under the microscope.
In vivo experiments
To measure the antitumor efficacy of palbociclib monotherapy, we suspended CAL27, FADU, and HN6 cells in the mixture of PBS and Matrigel (1:1). Immuno-deficient athymic BALB/c-nu/nu female mice (6-week-old) were used in this experiment. A total of one million cells were injected subcutaneously into the left flank region of each mouse. When tumor size reached at least 100 mm3, the mice were randomly divided into palbociclib and vehicle groups (6 mice in each group). Palbociclib (60 mg/kg) and vehicle (sodium lactate buffer, 50 mmol/L, pH 4.0) were orally administrated once a day. HN6 and FADU xenografts were treated with a continuous application of palbociclib for a total of 14 days, while CAL27 xenografts were treated with a phased schedule (days 1–14: palbociclib treatment; day 15–31: treatment suspension; day 32–69: palbociclib treatment).
In the drug combination experiment on FADU xenografts, the mice were randomly assigned into four groups receiving vehicle (sodium lactate buffer, 50 mmol/L, pH 4.0, orally, daily), palbociclib (60 mg/kg, orally, daily), alpelisib (20 mg/kg, orally, daily), or both for a total of 21 days. To further evaluate the efficacy of the clinical treatment, PDX models were used for different drug combinations. Xenograft-bearing mice were randomly assigned into seven groups (7 mice per group) to receive vehicle, palbociclib, alpelisib, cetuximab (1 mg/kg, intraperitoneal, twice a week), palbociclib plus cetuximab, palbociclib plus alpelisib, and palbociclib plus cisplatin (3 mg/kg, intraperitoneal, once a week). Treatment began when tumors reached at least 100 mm
3. The treatment scheme for each model varied from 8 to 29 days (mice were sacrificed either when tumor volume exceeded 1500 mm
3 or at the experiment endpoint). Tumor size and bodyweight were measured twice per week. The percentage of tumor growth inhibition [
32] was defined as 100 × [1 − (TVf_treated − TVi_treated)/ (TVf_control − TVi_control)]. TVi and TVf represent the mean tumor volume at the start and end of treatment, respectively [
23].
Statistical analysis
Statistical analysis was performed using GraphPad Prism 9.0 software. Data are presented as the mean ± S.D. or S.E.M., as per indicated in the figure legends. Pairwise comparisons between experimental and control groups were performed using unpaired two-tailed Student’s t tests, one-way ANOVA, or two-way ANOVA where appropriate. P < 0.05 was considered to be statistically significant. P-value symbols are denoted as follows: * (P < 0.05), ** (P < 0.01), *** (P < 0.001), and **** (P < 0.0001).
Discussion
Our study aims to identify different ways to improve the efficacy of CDK4/6 inhibitors in HPVneg HNSCC. By using palbociclib-based combination drug screening, we found that alpelisib, a PI3K inhibitor, exerted potent synergistic effects when combined with palbociclib. Notably, this combination strategy was illustrated to have a higher synergism than palbociclib combined with cetuximab or cisplatin, both of which have been evaluated in phase II clinical trials. Additionally, this increased efficacy was particularly evident in cases with PIK3CA alterations. Thus, our study adds to the growing number of rational combinations with CDK4/6 inhibitors in molecularly defined patients that could fulfill the promise of targeting cell-cycle pathway in molecularly defined patients in the clinic.
High-throughput drug screening in a matrix is a pivotal strategy exploring drug-drug pairs for potential synergy use. This approach led to the discovery of novel, highly effective combination options for osimertinib-based and ibrutinib-based therapeutics [
18,
57]. However, few studies have systematically evaluated the synergistic potential of CDK4/6 inhibitors with other therapeutic agents. Here, we conducted a palbociclib-based combination matrix screening and identified PI3K pathway, EGFR, and MEK inhibitors containing 75% of the top 20 palbociclib-based synergistic drugs. Notably, the therapeutic efficacy of palbociclib plus selective PI3K pathway inhibitors was also investigated preclinically or in clinical trials among other solid tumors (e.g., breast cancer [
58,
59], oral squamous cell carcinoma [
60], and hepatocellular carcinomas [
61]), supporting our conclusion that PI3K inhibitors were promising synergistic agents for palbociclib-based treatment. In addition to PI3K inhibitors, we also observed that mTOR and AKT inhibitors showed relatively well average ExcessHSA in the palbociclib combinations, which has also been reported by other researchers in HNSCC [
60].
Activation of PI3K pathway is a frequent observed in hallmark of cancer, which is highlighted by the prevalence of somatic mutations or amplification of
PIK3CA in HNSCC. Our results demonstrated that PI3K inhibitors showed prominently therapeutic efficacy in
PIK3CA-altered HPV
neg HNSCC while CDK4/6 inhibition has also been proven to improve the response to PI3K inhibitors in
PIK3CA-mutated breast cancers [
62]. Interestingly, a PI3K inhibitor-based combination drug screening conducted on
PIK3CA mutant breast cancer revealed that combined CDK4/6-PI3K inhibition could overcome intrinsic and acquired resistance of PI3K inhibitors and lead to tumor regressions [
63]. Recently, alpelisib was approved by the FDA as the first PI3Kα inhibitor for the treatment of HR-positive, HER2-negative,
PIK3CA-mutated, advanced or metastatic breast cancer patients [
64]. Triple PI3K:CDK4/6:ER combination therapy was also proposed to be evaluated to delay and/or prevent the acquired resistance of CDK4/6-based therapy [
65‐
67]. Additionally, preclinical studies and clinical trials demonstrated the safety and combinatory potential of alpelisib with cisplatin, cetuximab, paclitaxel, or docetaxel in HNSCC (NCT02145312, NCT01602315, and NCT02051751). To sum up, alpelisib, as a representative PI3K inhibitor, was demonstrated to be a promising synergistic drug for the treatment of HPV
neg HNSCC.
We also observed other drugs with high average ExcessHSA in HNSCC cell lines that represented potentially valuable combination strategies. For example, appreciable synergistic effects were induced in 3 out of 4 cell lines after combining palbociclib with the CDK2 inhibitor, milciclib. Importantly, Cyclin E-CDK2 complexes drive cell-cycle progression (S-phase entry) and the activation of CDK2 represents one of the mechanisms of CDK4/6i resistance in breast cancer [
68]. In addition, we also found two inhibitors (JQ1 and I-BET-762) targeting BRD4, one of the bromo- and extra-terminal domain (BBD) proteins, which ranked 2nd and 18th in the 6 × 6 screening that represents another promising combined therapeutic strategy. BBD proteins are emerging therapeutic targets in TNBC [
69] and able to elicit synthetic lethal with several genes including CDK4 [
70]. Nevertheless, it is absolutely necessary to refine the potential biomarkers used in combinational treatment to ensure highly efficient responses.
EMT is a well-studied complex biological process in embryonic development that is recapitulated during tumor progression, metastasis, and drug resistance [
68]. Previous studies have demonstrated that palbociclib treatment promoted or inhibited EMT in different types of cancer [
46,
71]. Here, we found that EMT gene set was upregulated by palbociclib while two targeted treatment combinations exerted significant synergistic effects through suppression of the EMT-related impacts induced by palbociclib. The synergistic effect of these two combinations was attenuated by RRM2 overexpression and restored after subsequent treatment with osalmid, an RRM2 inhibitor. Recently, the oncogenic role of RRM2 has been linked to the EMT process in esophageal adenocarcinoma, glioma, and prostate cancers [
53,
54,
72] and multiple RRM2 inhibitors have been developed and are being investigated in clinical trials as monotherapy or combinational treatment option. In particular, RRM2 inhibitors was found to be highly effective when combined with cell-cycle checkpoint inhibitors in Ewing sarcoma [
73]. Thus, further studies are warranted to fully explore the potential of such promising therapeutic combinations.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.