Palbociclib monotherapy shows limited drug efficacy in HPVneg HNSCC
To evaluate the therapeutic efficacy of palbociclib monotherapy in HPVneg HNSCC, we investigated the cellular response of CDK4/6 inhibition in different HPVneg cell lines. An ER-positive MCF-7 human breast cancer cell, which has been demonstrated to be highly sensitive to palbociclib [33], was used as positive control. All of the 13 selected HPVneg HNSCC cell lines showed varied degree of sensitivity to palbociclib, with a mean IC50 ranging from 0.908 to 47.88 μM, and we defined 5 cell lines with relative lower IC50 (< 5 μM) as “palbociclib sensitive” (PS) and the remaining 8 cell lines as “palbociclib resistant” (PR; Fig. 1a and Additional file 1: Fig. S1a, b). Well-recognized genetic lesions (CDKN2A, CCND1 and PIK3CA) of all the 13 cell lines were annotated by whole exome sequencing (WES; Fig. 1b). Notably, the mutations identified in 7 of these 13 cell lines were similar to that which have also been characterized by Cancer Cell Line Encyclopedia [34] and Genomics of Drug Sensitivity in Cancer [25] database (Additional file 1: Fig. S1b and Additional file 2: Table S5). Additionally, the mRNA and protein expression levels of representative cell-cycle pathway-related genes were determined by qRT PCR and Western blot (Fig. 1b and Additional file 1: Fig. S1e).
Rb/pRb, whose expression serves as an inclusion eligibility criterion in several clinical trials of palbociclib [35, 36], was universally expressed among all the PS cell lines (Fig. 1b and Additional file 1: Fig. S1e). As expected, we found that significantly lower expressions of Rb1 mRNA, total Rb, and phosphorylated Rb at Ser807/811 protein were detected in PR group when compared with PS group (P = 0.0246, 0.0036, 0.0404, respectively, Additional file 1: Fig. S1c, e). On the other hand, the mRNA and protein expression levels of CDK6, which have been reported to be correlated with intrinsic resistance to palbociclib [37], were significantly higher in PR group when compared with PS group (P = 0.0347 and 0.0454, respectively, Additional file 1: Fig. S1c, e). However, CDKN2A deletion/mutation, CCND1 or CDK6 amplification, which contribute to Rb pathway activation and have been reported as predictive biomarkers of CDK4/6 inhibition [7, 38], were not found to be correlated with palbociclib sensitivity (Additional file 1: Fig. S1d), possibly due to insufficient sample size.
Two PS cell lines (PS-CAL27, PS-FADU), one PR cell lines (PR-HN6) and MCF7 were selected for evaluation of the cellular response to palbociclib treatment. When treated with different doses of palbociclib, PS-CAL27, PS-FADU and PR-HN6 showed significantly less reduction in DNA replication and G1 phase arrest when compared with MCF7 cells which displayed prominently decreased proliferation and G1-arrest as previously reported [39] (Fig. 1c, d and Additional file 1: Fig. S1f). Western blot analysis also showed palbociclib (ranging from 100 to 500 nM) exhibited limited inhibition of total and phosphorylated Rb (Rb and pRb), as well as the key regulators of G1/S-phase transition (CDK2 and Cyclin A2) (Fig. 1e).
Additionally, palbociclib monotherapy exerted significant inhibitory effects on the growth of PS-FADU xenografts in a 2-week treatment schedule, while it had no effects on PR-HN6 xenografts (Fig. 1f, g, Additional file 1: Fig. S1g). To evaluate the long-term therapeutic efficacy of palbociclib [40], an “on-off-on” phased treatment schedule was conducted on CAL27 xenografts (Additional file 1: Fig. S1h). As illustrated in Fig. 1h, palbociclib monotherapy exerted significantly inhibitory effects on the growth of PS-CAL27 xenografts after the first treatment phase (days 1–14, P = 0.0084). During the treatment suspension phase, tumor progressed rapidly upon the removal of palbociclib (days 15-31, treatment off), reflecting the cytostatic rather than cytotoxic effect of CDK4/6 inhibition. When the mean tumor volumes in the treatment group reached comparable to that of the control group, palbociclib was resumed and PS-CAL27 xenografts were first responsive to palbociclib (from days 31–52) while started to regrow and progressed under drug pressure (from day 53 to day 69), indicating the emergence of acquired resistance to palbociclib in PS-CAL27 (Fig. 1h and Additional file 1: Fig. S1g). Taken together, HPVneg HNSCCs tend to be less responsive or develop drug resistance to treatment with palbociclib monotherapy, highlighting the need to investigate combinational therapeutic strategies.
Quantitative high-throughput 6 × 6 matrix combination screening identifies synergistic agents
To identify alternative combinational approaches for palbociclib in HPVneg HNSCC, we performed high-throughput combination drug screening in four HNNCC cell lines with varied sensitivity to palbociclib (PS-CAL27, PS-FADU, PR-SCC9, PR-HN6) using a customized library within 6 × 6 checkerboard matrix (Fig. 2a). The library includes 162 agents covering multiple inhibitors for well-explored oncogenic targets [for instance, phosphoinositide 3-kinase (PI3K), anaplastic lymphoma kinase (ALK), and mitogen-activated protein kinase (MEK)] and targeting 54 distinct mechanisms of action (Additional file 2: Table S1). Notably, most of the FDA-approved agents that are currently used in HNSCC treatment including chemotherapeutics (cisplatin, docetaxel, 5-fluorocrail) and targeted therapy (cetuximab) were also incorporated in the library [41].
The 6 × 6 discovery checkerboard matrix generated a total of 16,200 drug-drug interactions. Of the 648 discrete screened plates, the average Z-prime score as a control for robustness was 0.83, and Z-prime score of all plates was greater than 0.5, indicating that the screening assay was experimentally robust (Additional file 1: Fig. S2a, b). Each 6×6 matrix was scored by the sum of ExcessHSA (Excess over Highest Single Agent) for evidence of synergistic (ExcessHSA score < − 20), additive (− 20 ≤ ExcessHSA score ≤ 20), or antagonistic effects (ExcessHSA score > 20, Additional file 2: Table S6), and the average ExcessHSA score of each compound in four cell lines was ranked accordingly (Fig. 2b).
Overall, we found that 77.16% of these compounds (125/162) showed distinct patterns of synergistic effect with palbociclib and ranked all compounds based on average ExcessHSA scores (Fig. 2b). The most obvious trend was that 7 PI3K pathway inhibitors (e.g., pictilisib, everolimus, alpelisib) were among the top 20 hits for synergism with palbociclib. It was also notable that receptor tyrosine kinase (RTK) related agents including 5 EGFR inhibitors or antibody (e.g., osimertinib, dacomitinib, and cetuximab) and 3 MEK inhibitors (e.g., binimetinib, cobimetinib) were ranked among the top 20 hits in the combination drug screening. Bromodomain and extra-terminal protein (BRD4) has been increasingly appreciated as a key oncogene during the tumorigenesis and development of HNSCC [42]. Two BRD4 inhibitors, JQ1 and GSK-525762A, demonstrated significant synergy with palbociclib (ranking the 2nd and 18th). On the other hand, we found that most of the evaluated conventional chemotherapies displayed additive or antagonistic effects when combined with palbociclib (e.g., cisplatin, nedaplatin, ExcessHSA -7.63, -13.85; paclitaxel, docetaxel; ExcessHSA 173.86, 62.76, Fig. 2b). Taken together, the quantitative high-throughput matrix combination screening characterized the synergistic potential of a list of well-recognized agents when combined with palbociclib.
Quantitative 10 × 10 matrix screening validates PI3K inhibitors as top synergistic option
Considering the significant synergistic effect observed between palbociclib and the PI3K pathway inhibitors, we conducted a more detailed analysis of 13 agents targeting different components of PI3K pathway, including 6 PI3K inhibitors, 3 AKT inhibitors, and 4 mTOR inhibitors. Four of the top 7 PI3K pathway inhibitors belongs to isoform-selective PI3K inhibitors, including PI3Kα/δ inhibitor pictilisib (ExcessHSA: − 300.41), FDA-approved PI3Kα selective inhibitor alpelisib (ExcessHSA: − 227.83), dual PI3K/ mammalian target of rapamycin (mTOR) inhibitors apitolisib (ExcessHSA: − 202.16), and PI3Kα/β/γ inhibitor taselisib (GDC-0032) (ExcessHSA: − 185.12, Fig. 2c). Three AKT inhibitors (e.g., AZD5363, perifosine, miltefosine) and 4 mTOR inhibitors (e.g., temsirolimus, everolimus, MLN0128, AZD2014) also showed strong synergistic effects with palbociclib (average ExcessHSA: − 194.86, − 138.83, − 94.83, and − 188.08, − 187.27, − 171.81, − 147.45, respectively, Fig. 2c). Notably, PI3K inhibitors showed significantly higher average ExcessHSA when compared with AKT inhibitors and mTOR inhibitors (− 230.47 versus − 127.04 and − 185.5, P = 0.0002 and 0.039, respectively, Fig. 2d).
To capture more comprehensive information about drug-drug interactions, we conducted a 10 × 10 matrix study to evaluate the synergistic scores by incorporating up to 81 different concentration combinations. As defined by multiple metrics including the ExcessHSA score and Bliss independence model, this experiment further confirmed the synergistic effects between palbociclib and the four PI3Ki (ranging from − 476.86 to − 690.54), and alpelisib demonstrated the most synergistic effect when combined with palbociclib (Fig. 2e, f; Additional file 1: Fig. S2c). Moreover, we found that PIK3CA-amplified cell line FADU outperformed other cell lines with the highest average ExcessHSA score of these combinations (− 748.28, Additional file 1: Fig. S2d). Thus, these screening results indicated that PI3K pathway contributed the most to sustain the viability of HNSCC under palbociclib treatment, with PI3K inhibitors being identified as the most promising combination option.
PI3K inhibitors show enhanced synergistic effect in PIK3CA-altered HPVneg HNSCC
PIK3CA is the most commonly altered oncogene in HNSCC, with mutations or amplifications detected in 34% of all the HPV-negative HNSCC tumors and previous studies have reported that PIK3CA mutation or amplification was predictive for the response of PI3K/AKT/mTOR inhibitors [43]. To assess whether PI3K inhibitors plus palbociclib would lead to more prominent combination potency in PIK3CA-altered cell lines, we then evaluated these inhibitors (alpelisib, pictilisib, GDC-0032, and apitolisib) in combination with palbociclib in eight selected cell lines (4 PS and 4 PR), which represent two major genetic subtypes (PIK3CA mut/amp: Detroit-562, FADU and HN30; PIK3CA wild type (WT): HN6, PECA-PJ41, SCC9, CAL27, and UPCI-SCC-131). Drug dose-response curves showed that all four PI3K inhibitors resulted in obvious shifts of drug response curve and corresponding decreased IC50 when combined with palbociclib, especially for three PIK3CA-altered cell lines (Fig. 3a; Additional file 1: Fig. S3a).
We next measured cell viability across different drug doses or combinations in each cell line. The result showed that combinational treatment with PI3K inhibitors resulted in stronger anti-proliferation activities in PIK3CA mut/amp cell lines when compared with others (PIK3CA WT) at low concentrations (Fig. 3b; Additional file 1: Fig. S3b-e). The resulting combination index (CI) theorem of Chou-Talalay offers quantitative definition for additive effects (CI = 1), synergism (CI < 1), and antagonism (CI > 1) in drug combinations [26]. Analysis with this algorithm indicated strong synergy between palbociclib plus alpelisib, and to a lesser degree, palbociclib plus GDC-0032, palbociclib plus apitolisib, and palbociclib plus pictilisib (Fig. 3c). Importantly, the highest extent of synergism was observed in three PIK3CA mut/amp cell lines, suggesting that disruption of PI3K pathway substantially sensitized PIK3CA mut/amp HNSCC cells to palbociclib-mediated cytostatic effect (Fig. 3c). Notably, significant synergetic effect of these combinations was also observed using another CDK4/6 inhibitor, abemaciclib, with four PI3K inhibitors, especially in PIK3CA mut/amp cell lines (Additional file 1: Fig. S4). Additionally, combinations of palbociclib and PI3K inhibitors showed consistently synergistic effects on cell proliferation and G0/G1 cell cycle arrest across all 3 PIK3CA mut/amp cell lines while divergent synergistic effects were observed in 5 PIK3CA WT (Fig. 3d and Additional file 1: Fig. S3f-h). Alpelisib, which showed the most consistently synergistic effect in almost all cell lines, was further evaluated as a combinational agent with palbociclib in vivo using FADU-derived xenograft (annotated with PIK3CA amplification). As expected, palbociclib or alpelisib monotherapy slightly delayed tumor progression, while palbociclib plus alpelisib showed significantly synergistic therapeutic efficacy (Fig. 3e and Additional file 1: Fig. S3i-k). Thus, we found that alpelisib had a broad synergistic effect in HPVneg HNSCC cell lines, with particularly higher therapeutic potential in cases harboring PIK3CA alterations.
Alpelisib exerts synergistic effect via suppression of EMT induced by palbociclib
To explore potentially synergistic mechanisms between palbociclib and alpelisib, we performed RNA-seq analysis on PS-FADU, PS-HN30, and PS-CAL27 cells treated with DMSO, palbociclib, alpelisib, or two combination treatments using low or high doses of alpelisib plus palbociclib. Significantly upregulated and downregulated genes were then identified in palbociclib, alpelisib, and combination treatment groups when by comparison with the DMSO group (Fig. 4a). As expected, gene set enrichment analysis (GSEA) analysis showed cell-cycle-related genes (including E2F targets, G2M, and MYC target-related genes) were consistently downregulated in all treatment groups, while PI3K signaling pathway genes were only downregulated by alpelisib and the combination treatments (Fig. 4b). Leading-edge analysis from GSEA gene sets indicated that E2F targets and PI3K signaling pathway genes were both strongly downregulated in the treatment groups of all three cell lines (Additional file 1: Fig. S5a, b). Western blot analysis of the combination treatment group in all three cell lines showed decreased levels of E2F targets (e.g., Cyclin A2 and Aurora B) and PI3K pathway markers (e.g., pAKT and pS6 phosphorylation, Additional file 1: Fig. S5d).
Epithelial mesenchymal transition (EMT) has been reported to be associated with resistance to targeted therapy through various mechanisms of action [44, 45]. More recently, a study demonstrated that HPVneg HNSCC cancer cells harboring a mesenchymal phenotype were less sensitive to CDK4/6 inhibition than those with an epithelial phenotype [8]. In agreement with these reports, we found that genes related to EMT were consistently upregulated under palbociclib treatment in all three cell lines (Fig. 4b). Previous study has demonstrated that CDK4/6 knockdown could upregulate TGF-β/Smad3 signaling and induce EMT via increasing the expression of downstream targets including p15 and p21 [46,47,48]. To investigate whether palbociclib treatment elicited EMT through TGF-β/Smad3 signaling in HNSCC, CAL27, FADU, and HN30 cells were next incubated with TGF-β1, SB-431542 (a TGF-β type I receptor kinase inhibitor) and palbociclib in the absence or presence of SB-431542. We first observed that TGF-β1 activated TGF-β/Smad3 signaling as evidenced by increased phosphorylated Smad3 and Smad3 downstream targets p15/p21, which were similarly upregulated under palbociclib treatment (Fig. 4c). Notably, SB-431542 completely inhibited the expression of palbociclib- or TGF-β1 (a well-known EMT inducer)-mediated induction of p15/p21 and EMT-associated genes, including N-cadherin and Slug (Fig. 4c). Thus, these results indicated that the EMT elicited by CDK4/6 inhibition was largely caused by TGF-β/Smad3 signaling-mediated induction of p15 and p21.
As for the combination treatment groups, we observed that low-dose combination resulted in a lower degree of EMT gene-set upregulation compared to that under palbociclib monotherapy, while no upregulation of the EMT gene set was detected in high-dose combination treatment group of any of the three cell lines (Fig. 4b). Leading-edge analysis from the GSEA EMT gene set highlighted genes that were strongly up- and downregulated only in combination-treated cells (Additional file 1: Fig. S5c). Western blot analysis showed that the combination treatments resulted in decreased levels of representative mesenchymal proteins including N-cadherin, Vimentin, and Snail whereas epithelial protein E-cadherin appeared to be unaffected in all three cell lines under the combination treatments (Fig. 4d). Moreover, cell viability assay showed that TGF-β1 treatment effectively increased the palbociclib resistance (a 1~4-fold increase in IC50 values of palbociclib) in all three HNSCC cell lines, and PI3K inhibitor alpelisib combined with palbociclib could at least partially restore palbociclib sensitivity (Fig. 4e). Collectively, these results show palbociclib plus alpelisib does not induce the EMT and alpelisib might exert synergistic effect through suppressing EMT-related impacts induced by palbociclib.
Evaluation of selected palbociclib-based treatment combinations in vitro
Cisplatin is a standard therapeutic regimen for HNSCC [49], while the EGFR-targeted antibody cetuximab, except for anti-PD-1 antibody, remains the only molecular therapy approved for the treatment of HNSCC [50]. Thus, various treatment strategies using these two agents as combinational agents are currently under active clinical investigation in HNSCC [51]. To this end, the therapeutic efficacy of these three combinations: palbociclib plus alpelisib, along with palbociclib plus cetuximab and palbociclib plus cisplatin, were further examined in a total of 13 HPVneg cell lines. We found that palbociclib plus alpelisib led to the highest degree of synergism in inhibiting tumor cell growth of the three combinations (average ExcessHSA score: − 213.24; Fig. 5a), with significantly higher ExcessHSA scores observed in PIK3CA mut/amp cell lines than that of PIK3CA WT lines (P = 0.0253, Fig. 5b). Palbociclib plus cetuximab also exhibited synergistic effects, but to a lesser extent (average ExcessHSA score: − 99.33). By contrast, palbociclib plus cisplatin combination resulted in antagonistic effects in 8 of the 13 examined cell lines (average ExcessHSA score: 96.12; Fig. 5a and Additional file 1: Fig. S6a).
To further evaluate whether EMT was involved in these synergistic or antagonistic effects, RNA-seq was conducted in PS-FADU, PS-HN30, and PS-CAL27 cells treated with DMSO, palbociclib, or the three combination treatments. GSEA analysis indicated that the EMT gene set was consistently upregulated by palbociclib rather than palbociclib plus alpelisib treatment, whereas treatment with palbociclib plus cetuximab or palbociclib plus cisplatin led to inconsistent results across the three cell lines (Additional file 1: Fig. S6b). Considering that the EMT gene set includes two distinct groups of genes representing the epithelial phenotype or mesenchymal phenotype of cancer cells [52], we then separately examined the expression of epithelial-related (n = 169) and mesenchymal-related genes (n = 48) in combination treatment groups compared with that under palbociclib monotherapy to identify differences in the effects of these combinations on the EMT process in detail.
Notably, palbociclib plus alpelisib induced significant downregulation of mesenchymal-related genes in CAL27 and FADU cells (NES = − 1.85 and − 1.64, P = 0.008 and 0.015, respectively, Fig. 5c), thus indicating EMT inhibition. Palbociclib plus cetuximab induced significant upregulation of epithelial-related genes in FADU and HN30 cells (NES = 1.51 and 1.29, P = 0.002 and 0.049, respectively) and significant downregulation of mesenchymal-related genes in HN30 (NES = − 1.69, P = 0.012, Fig. 5c) also demonstrated similar EMT inhibition. On the contrary, palbociclib plus cisplatin did not result in any obvious inhibitory effects on EMT in all three cell lines and indeed appeared to induce a significant decrease in transcription of epithelial-related genes in CAL27 cells (NES = − 1.53, P = 0.004) and upregulation of mesenchymal-related genes in FADU cells (NES = 1.58, P = 0.014), suggesting the activation of EMT process (Fig. 5c and Additional file 1: Fig. S6c).
Western blot analysis revealed that palbociclib monotherapy induced an increase of mesenchymal markers N-cadherin, Vimentin, and Slug, while the palbociclib plus alpelisib combination attenuated the levels of corresponding mesenchymal markers in all these three cell lines (Fig. 5d and Additional file 1: Fig. S6d). However, the palbociclib plus cetuximab induced a relatively minor reduction in the levels of these markers in FADU and CAL27 cells, which was consistent with the limited effects of this combination on the Mes signature (Fig. 5c). In addition, neither palbociclib plus alpelisib nor palbociclib plus cetuximab elicited any discernible effects on E-cadherin in each cell line, indicating that E-cadherin was not likely required for the upregulation of epithelial gene set (Fig. 5c, d and Additional file 1: Fig. S6d). Notably, palbociclib plus cisplatin did not elicit any detectable effects in any of these mesenchymal markers. Taken together, our results demonstrate that palbociclib plus alpelisib and palbociclib plus cetuximab might enhance the therapeutic effects of palbociclib monotherapy through the blockade of EMT.
RRM2 can induce EMT to attenuate the synergistic effects of palbociclib combined with alpelisib or cetuximab
Ribonucleotide Reductase Regulatory Subunit M2 (RRM2) has been reported as a therapeutic target via activating EMT in HNSCC and other cancers [53, 54]. Our transcriptomic and Western blot analysis also confirmed that RRM2 was downregulated in HNSCC cells treated with palbociclib plus alpelisib or palbociclib plus cetuximab (Fig. 5d and Additional file 1: Fig. S6d, e). To evaluate the potential contribution of RRM2 in these combination treatments, we generated RRM2 stable overexpression (OE) and empty vector (EV) control lines in the FADU and CAL27 backgrounds (Fig. 5e and Additional file 1: Fig. S6f). The synergistic effects of palbociclib combined with alpelisib or cetuximab were significantly attenuated by RRM2 OE, which were then reversed by RRM2 inhibitor osalmid [55] (Fig. 5f and Additional file 1: Fig. S6g). However, neither RRM2 OE nor osalmid treatment exerted similar effect in cells treated with palbociclib and cisplatin combination (Additional file 1: Fig. S6h). Based on these findings, it was reasonable to speculate that RRM2 functioned specifically in response to the palbociclib plus alpelisib and palbociclib plus cetuximab combinations.
To further understand whether RRM2 OE counteracted the synergistic effects of these two combinations via inducing EMT, we first conducted transwell assays. The results showed that RRM2 overexpression led to increased cell migration and invasion, which could be reversed by osalmid treatment (Additional file 1: Fig. S6i, j). Moreover, the expressions of three mesenchymal markers were upregulated in RRM2-OE cells (Fig. 5e and Additional file 1: Fig. S6f). These findings indicated that RRM2 OE could induce EMT in HNSCC cells, which then led us to further investigate the genetic factors responsible of EMT regulation. We found that the decreases in N-cadherin, Vimentin, and Slug induced by these palbociclib combination treatments in RRM2 EV cells were partially attenuated by RRM2 OE (Fig. 5g and Additional file 1: Fig. S6k). However, we observed a similar downregulation of these genes following co-administration of osalmid with either of the two combinations (Fig. 5g and Additional file 1: Fig. S6k). To sum up, these data indicate that RRM2 overexpression could induce EMT to attenuate the synergistic effects of palbociclib combined with alpelisib or cetuximab.
Evaluation of palbociclib-based treatment combinations in molecularly defined PDXs
To validate the synergistic effects of palbociclib-based treatment combinations from a more clinically relevant perspective, palbociclib/alpelisib/cetuximab monotherapy and three combinations mentioned above were further evaluated in five molecularly defined PDX models (Fig. 6a and Additional file 2: Table S7). The CNV landscape and percentage of tumor variants defined by variant allele frequency (VAF) were generally maintained between PDX models and the matched tumors (Fig. 6b, c). The treatment efficacy was evaluated by the percentage of tumor growth inhibition [32], and TGI > 60% was considered meaningful as a significant responder as previously reported [56].
Our results demonstrated that palbociclib monotherapy caused significant tumor inhibition in three of five selected PDX models (SHHN001, SHHN002, and SHHN004) with TGI = 68.47%, 61.45%, and 63.51%, respectively (Fig. 6d, e, g, i). In particular, SHHN005 showed no response to palbociclib monotherapy (TGI = 1.60%, Fig. 6h, i), in which CCND1 amplification and CDKN2A mutation were found to be concurrent with RB1 deletion, and the lack of Rb expression was further validated in the xenograft that might explain its resistance (Fig. 6a and Additional file 1: Fig. S7a). Cetuximab monotherapy showed a lack of tumor suppressing effect in four of five PDX models (TGI ranging from 3.60 to 58.81%, SHHN001 TGI = 80.08%, Fig 6d–i). Alpelisib monotherapy exerted significant antitumor efficacy in the PIK3CA amplified SHHN001 and SHHN003, with TGI = 65.78%, 65.65% (Fig. 6d, f, i). Though not defined as significant responders, partial antitumor effect was observed in the other three models (TGI ranging from 47.41 to 57.20%, Fig. 6e, g–i). Thus, palbociclib generally has a better therapeutic effect as monotherapy in Rb-proficient HNSCCs when compared with cetuximab or alpelisib monotherapy.
Palbociclib plus cetuximab showed superior therapeutic effect in three of five PDX models (SHHN002 TGI = 90.39%, SHHN003 TGI = 92.48%, SHHN004 TGI = 94.90%) when compared with cetuximab monotherapy (P = 0.003, P < 0.0001, P = 0.0061, Fig. 6e–g), and in three PDX models (SHHN001 TGI = 91.66%, SHHN002 TGI = 90.39%, and SHHN003 TGI = 92.48%) when compared with palbociclib monotherapy (P = 0.0459, P = 0.0066, P < 0.0001, Fig. 6d–f). More importantly, palbociclib plus alpelisib had superior therapeutic effect in four of five evaluated PDXs (except for the Rb-deficient SHHN005) when compared with alpelisib monotherapy (Fig. 6e–g), and in all five PDX models when compared with palbociclib monotherapy (Fig. 6d–h). Notably, tumor regression (TGI > 100%) was only observed in palbociclib plus alpelisib-treated groups in two PIK3CA-amplified models, SHHN001 and SHHN003 with a TGI = 107.45% and 103.66%, respectively (Fig. 6d, f).
No significant synergistic effect was observed in palbociclib plus cisplatin combination in all five evaluated PDXs (Fig. 6d–h). Of note, palbociclib plus cisplatin was found with a reduced therapeutic effect when compared with palbociclib monotherapy in SHHN001 (TGI = 48.40% vs TGI = 68.47%), SHHN002 (TGI = 46.16% vs TGI = 61.45%), and SHHN004 (TGI = 24.50% vs TGI = 63.51%, Fig. 6d, e, g). Consistent with the drug screening results from the cell line experiments, palbociclib plus cetuximab or alpelisib targeted treatment combination showed superior therapeutic effects when compared with palbociclib plus cisplatin treatment in all 5 PDX models (Fig. 6d–h). Notably, all treatment combinations were well-tolerated in mice with no significant reductions in bodyweight in all models (Additional file 1: Fig. S7b), as well as no obvious hematologic aberrations or organ lesions (Additional file 1: Fig. 7c, d). Taken together, our in vivo studies using five PDX models with heterogeneous genetic backgrounds provide further evidence that palbociclib can synergize with PI3K and EGFR inhibition and suggest that patients harboring PIK3CA alterations are more likely to benefit from these treatment combinations.