Evaluation of Eribulin Combined with Irinotecan for Treatment of Pediatric Cancer Xenografts

Current therapies for treatment of childhood cancers rely almost exclusively on use of cytotoxic agents, predominantly those that induce DNA damage, and antimitotic agents. Although the use of intensive chemotherapy has resulted in increased 5-year event-free survival (1, 2), and a high cure rate, the effects of chemo-radiotherapy can be devastating (3–5). Development of new therapies for childhood solid tumors is slow and remains challenging. In part, this is because the number of drugs that can be tested is limited by a relatively low number of patients eligible for phase I/II evaluation. We and others have used pediatric patient-derived xenografts (PDX) and cell line-derived xenografts (CDX) to identify novel agents and effective combinations (6–11). These models were valuable in identifying combinations that were highly active in rhabdomyosarcoma, such as irinotecan–vincristine (9), and irinotecan–temozolomide (12) in other sarcomas (13). Several of these studies were confirmed in clinical trials (14, 15).

Camptothecin-based topisomerase I poisons have shown significant antitumor activity in multiple childhood xenograft models (6–8). Of note, both topotecan and irinotecan showed greater than additive activity when combined with vincristine (9, 10), and clinically, the combination of vincristine with irinotecan induces high objective response rates for advanced rhabdomyosarcoma (14), and anaplastic Wilms tumor (15). However, despite interesting preclinical activity and the clear demonstration of clinical utility of this combination, no mechanistic studies were undertaken to elucidate the basis for synergy.

Vincristine, a standard of care drug for many solid tumors and acute leukemias, binds tubulin and destabilizes microtubules, inhibiting mitotic progression. More recently, another microtubule inhibitor, eribulin, has shown impressive antitumor activity against a broad spectrum of cancer xenografts (16). Eribulin is a synthetic derivative of halichondrin B (17), that induces irreversible mitotic arrest. Eribulin prevents microtubule growth and suppresses dynamics by binding to their plus ends (18, 19). Eribulin may have activities in addition to its antimitotic effects; eribulin treatment has been reported to alter gene expression, induce tumor cell differentiation, cause vascular remodeling, reverse epithelial-to-mesenchymal transition (EMT), and inhibit Wnt/β-catenin signaling (16, 20–23). Eribulin demonstrated significant antitumor activity against sarcoma and acute lymphoblastic leukemia models (24), and it had far greater activity against Ewing sarcoma PDX models than vincristine (25). Eribulin has entered pediatric phase I testing (NCT02171260), with at least one partial response in a patient with Ewing sarcoma (26). Eribulin is approved for third-line therapy of breast cancer (27), and it was recently FDA approved for treatment of liposarcoma (28). Eribulin is also reported to have less neurotoxicity in preclinical models and patients (29, 30), and no peripheral neuropathy was observed in the phase I trial in children (26), suggesting eribulin may have advantages over vincristine with respect to long-term toxicity. In this study, we investigated the efficacy and mechanisms of action for eribulin and vincristine alone and in combination with irinotecan against pediatric PDX models.

PDXs (rhabdomyosarcomas Rh10, Rh30, Rh41, Rh65; Wilms tumor KT-11; and kidney rhabdoid tumor KT-14) and CDXs (Ewing sarcoma lines CHLA258, SK-NEP-1, ES-4) have been described previously (31). PDXs were established directly in mice without first being cultured in vitro, whereas CDXs were established from cell lines first grown in culture. RBD1 is an atypical teratoid rhabdoid tumor established from a metastatic lung lesion. All tumors were used at low passage and authenticated by STR analysis against reference profiles developed by this group.

C.B.17SC scid^−/− (C.B-Igh-1b/IcrTac-Prkdcscid) female mice (Taconic Farms) were used to propagate subcutaneously implanted tumors as described previously (31). All mice were maintained under barrier conditions and experiments were carried out using protocols and conditions approved by the institutional animal care and use committee of UTHSA. Mice were randomized into groups of 10, and treatment was started when tumors were 200 to 400 mm³. Methods similar to those developed in the Pediatric Preclinical Testing Program (PPTP) were used, only for these studies tumor volumes were measured for up to 20 weeks, and regrowth of tumors determined. Mice were evaluated weekly for tumor mass and body weight and endpoints were time to event, defined as tumor growing to 400% of its volume at the initiation of treatment, tumor regression [partial response (PR), ≥50% volume decrease; complete regression (CR), tumor volume <40 mm³]. Eribulin was dissolved in sterile water and administered days 1 and 8, and irinotecan was dissolved in saline and administered days 1 to 5, with the cycle repeated at day 21. Eribulin was dosed at 1 mg/kg and irinotecan at 2.5 mg/kg. Both drugs were administered by intraperitoneal injection. Vincristine, as a single agent, was administered at 1 mg/kg every 7 days (intraperitoneally) for 6 consecutive weeks, or at 0.5 mg/kg when combined with irinotecan.

For combination testing, one objective was to determine whether the combination is significantly more effective than either agent utilized at their optimal/standard single-agent dose/schedule. This condition is termed therapeutic enhancement, which represents a therapeutic effect for which a tolerated regimen of a combination treatment exceeds the optimal effect achieved at any tolerated dose of monotherapy associated with the same drugs used in the combination (11, 32).

The cytotoxic activities of eribulin, vincristine, SN-38 (active metabolite of irinotecan), and the combinations were assessed by alamar blue assays (#BUF012; Bio-Rad Laboratories). ES-4 cells were plated in 96-well plates at 3.33 × 10³ cells/well, allowed to adhere overnight, and treated with vehicle (0.5% DMSO) or drug for 72 hours. Alamar blue (10% culture volume) was added to each well and fluorescence was measured at 590 nm (excitation 570 nm). All samples were blank-corrected for the background fluorescence of alamar blue. Concentration–response curves were initially plotted and IC₅₀ values were interpolated from nonlinear regressions using Prism 7 (Graphpad Software). Relevant concentration combinations of eribulin and SN-38 or vincristine and SN-38 were tested and combination indexes (CI) were calculated on the basis of the Bliss model of independence using the formula: CI = (E_A + E_B – E_AE_B)/E_AB, where (E_A, E_B) equal the effects of each individual drug and (E_AB) equals the effect of the drugs in combination. Heatmaps were produced using Prism 7. For the isobologram analysis, cells were treated with both drugs at fixed ratios. Concentration–response curves were generated at these fixed ratios and the IC₅₀ values were determined. From these IC₅₀ values, the concentrations of each individual drug at the IC₅₀ were determined and plotted as an isobologram. CIs for these experiments were calculated on the basis of the Loewe additivity model using the formula CI = a/A + b/B, where (a, b) equal the concentrations of the combined drugs that produce a particular effect level and (A, B) equal the concentrations of each individual drug that produce the same effect level.

Approximately 500 ng of total RNA from xenograft tissue was used for RNA sequencing (RNA-seq) library preparation by following the Illumina TruSeq stranded mRNA sample preparation guide (Illumina). RNA-seq libraries were subjected to quantification, and subsequent 50bp single-read sequencing module with Illumina HiSeq 3000 platform. After the sequencing run, demultiplexing with CASAVA (Illumina) was employed to generate the FastQ file for each sample. An average of ∼35M reads were generated for each sample. All data are publicly available at the Gene Expression Omnibus (GEO) of the NCBI (GSE146687).

All RNA-seq FastQ reads were aligned with the reference genome (UCSC human genome build hg19) using TopHat2 (33) default settings. The BAM files obtained after alignment were processed using HTSeq-count (34) to obtain the counts per gene in all samples. To remove RNA species from host (mouse RNA), a custom-designed algorithm was performed to align reads using TopHat2 to both host genome (mm9) and graft genome (hg19). We kept reads as human-specific or human/mouse-common based on number of mis-match bases, number of gaps, and alignment quality control scores. The human RNA-seq reads were then re-aligned to graft genome hg19 using TopHat2 with default settings, and the BAM files obtained after alignment were processed using HTSeq-count to obtain the counts per gene. The R package DESeq (35) was used to normalize gene expression with the size factor method and perform pairwise comparisons between groups to identify differentially expressed genes (DEG). Genes with an FDR-adjusted P value <0.05 and at least a 2-fold change compared with control were considered to be significantly differentially expressed. Upon obtaining the DEGs from all pair-wise comparisons, we performed k-means clustering on the combined gene set (k = 9) with data ordered by untreated, eribulin (24 and 144 hours), irinotecan (24 and 144 hours), and combination (24 and 144 hours) using MATLAB (MathWorks). Additional quality control statistical analysis of outliers, intergroup variability, distribution levels, PCA, and hierarchical clustering analysis were performed to validate the experimental data. Pathway analysis was performed with ingenuity pathway analysis (IPA; Qiagen). Gene set enrichment analysis was performed with the software package distributed by the Broad Institute (36).

Snap frozen ES-4 and KT-11 tumors were ground under liquid N₂ and total protein extracted with cell lysis buffer (Cell Signaling Technology, #9803) containing Halt Protease Inhibitor Cocktail (Thermo Fisher Scientific) and 1 mmol/L phenylmethylsulfonylfluoride (Sigma-Aldrich). Protein concentrations of each lysate were measured with a Protein Assay Kit (Bio-Rad Laboratories) and 20 μg of total protein was separated on NuPAGE 4% to 12% or 10% Bis-Tris gels (Thermo Fisher Scientific). Proteins were transferred to Immobilon-FL PVDF membranes (MilliporeSigma) and probed with antibodies for cleaved PARP (Cell Signaling Technology, #5625), phospho-S139-H2A.X (#9718), TP53 (#48818), CDKN1A/p21 (#2947), MDM2 (#86934), PUMA/BBC3 (#12450), phospho-S345-Chk1 (#2348), Chk1 (#2360), phospho-T68-Chk2 (#2197), Chk2 (#3440), full length/cleaved caspase 3 (#9668), cleaved caspase 3 (#9664), phospho-S10-histone H3 (#53348), histone H3 (#4499), GAPDH (#2118), or RRM2B/p53R2 (Invitrogen, #PA5-19970). Primary antibodies were diluted in Odyssey Blocking Buffer (LI-COR) with 0.1% Tween 20 (Sigma-Aldrich). Membranes were incubated with IRDye 680 or IRDye 800CW-conjugated secondary antibodies (LI-COR) diluted in Odyssey Blocking Buffer with 0.1% Tween-20 and 0.01% SDS. Fluorescence was detected with an Odyssey CLx Imaging System and analyzed with Image Studio software (LI-COR).

Whole tumors were fixed overnight in PBS-buffered 10% formalin at 4°C, then transferred to 70% ethanol until processing. Tumors were embedded in paraffin, cut into 5-μm-thick sections, and mounted on positively-charged slides. Sections were deparaffinized with high-pH Target Retrieval Solution (Agilent) using a PT Link pretreatment module (Agilent) and stained for cleaved caspase 3 (Cell Signaling Technology, #9579), TP53 (Cell Signaling Technology, #48818), CD34 (Abcam, #ab81289), and alpha smooth muscle actin (α-SMA; Cell Signaling Technology, #19245) using a Dako Autostainer Link 48 (Agilent). Sections were counterstained with hematoxylin and visualized using an EnVision FLEX DAB+ sub-chromo dye system. Images were captured on an EasyScan digital slide scanner (Motic) at 40× magnification.

Tumors were snap frozen, embedded in O.C.T. compound (Scigen Inc., #4583), cut into 5 μm sections, and mounted on positively-charged slides. Immediately before immunostaining the O.C.T. sections were fixed with 4% paraformaldehyde (Santa Cruz Biotechnology, #281692) at room temperature and washed three times with PBS. Sections were blocked with 5% goat serum (Cell Signaling Technology, #5425), 0.3% Triton X-100 in PBS. Slides were incubated with primary antibodies diluted in 1% BSA, 0.3% Triton X-100 in PBS, overnight at 4°C in a humidified chamber. Primary antibodies were 53BP1 (Cell Signaling Technology, #4937), phospho-S10-H3 (#53348), or γ-H2A.X (Abcam, #ab26350). Sections were incubated with Alexa Fluor-488 or Alexa Fluor-647-conjugated secondary antibodies (Abcam, #ab150077 and #ab150115,) diluted in 1% BSA, 0.3% Triton X-100 in PBS. Slides were mounted with Prolong Gold Antifade Reagent with DAPI (Cell Signaling Technology, #8961) and 0.17-mm-thick cover glass. Images were captured with a FLUOVIEW FV3000 laser-scanning confocal microscope and analyzed with Fluoview software (Olympus).

C.B.17SC scid^−/− female mice were implanted subcutaneously with 300 μL VEGF-infused (100 ng/mL) Matrigel, bilaterally into each flank. After 7 days, mice were assigned to four groups and left untreated (control) or treated with one cycle of eribulin, irinotecan, or the combination. One-hundred forty-four hours after the first drug dose (day 13 overall), mice were euthanized and plugs were excised, fixed in PBS-buffered 10% formalin containing 0.25% glutaraldehyde, and processed for IHC. Sections were stained for CD34 (Abcam) and α-SMA (Cell Signaling Technology). Immunostaining was quantified using FIJI software.

Eribulin, vincristine, irinotecan, and the combinations were tested against 12 xenograft models (nine PDX, three CDX). With one exception (Rh10 treated with vincristine, P = 0.322), all treatments resulted in statistically significant increases in event-free survival (EFS) compared with controls (P < 0.001). Results are summarized in Supplementary Table S1, along with estimates of therapeutic enhancement, additivity, and TP53 genotype. Details of the numbers of mice used, numbers that had events, or were excluded are given in Supplementary Table S2. Eribulin was administered at 1 mg/kg, which gives drug exposure in mice similar to that achieved after administration of 1.4 mg/m² to patients (24, 37). Eribulin induced regressions in nine models, with five PR, defined as ≥50% decrease in tumor volume, and four CR, defined as tumor volume <40 mm³. These results are consistent with previous testing by PPTP, in which more intense treatment was administered (6). Irinotecan, administered at a dose and schedule that results in SN-38 (active metabolite of irinotecan) exposure similar to 50 mg/m² administered once daily for 5 days in children (38), induced tumor regressions in seven models, with one PR and six CR (Supplementary Tables S1 and S2).

Combining eribulin with irinotecan resulted in complete tumor regressions in five models, whereas in these models the best response for either single drug was PR. The addition of eribulin to irinotecan significantly prolonged time to event in each model tested. In ES-4 Ewing sarcoma xenografts, eribulin and irinotecan as single agents modestly reduced tumor growth (<50% growth inhibition), with increased time to event compared with control tumors of 8.4 and 9.1 days, respectively. In contrast, the combination induced CR in ES-4 xenografts, with EFS(T-C) of 62.9 days and met criteria for supra-additivity. Eribulin was more active against two other Ewing sarcoma xenografts, CHLA-258 (PR) and SK-NEP-1 (CR), with EFS(T-C) of 49.3 and 67.1 days, respectively (Fig. 1; Supplementary Table S1). Irinotecan induced PR in SK-NEP-1 xenografts but had little activity against CHLA-258 [<50% growth inhibition, EFS(T-C) of 7.1 days], whereas the combination induced CR in both CHLA-258 and SK-NEP-1, with EFS(T-C) of 81.6 and >120 days, respectively. EFS(T-C) for the combination treatment was additive for the individual agents in CHLA-258, but could not be estimated for SK-NEP-1 and KT-11 models as the median time to event exceeded the observation period. Similarly, using the EFS(T-C) metric, the combination effect could not be estimated for Rh10, Rh28, and Rh65 rhabdomyosarcoma models, but was additive in Rh30 and supra-additive in Rh41 models (Supplementary Table S1). However, the combination induced prolonged complete regressions in Rh10, Rh41, KT-11, and KT-14 (Fig. 1). Statistical analysis of therapeutic enhancement and additivity for the eribulin-irinotecan combination are presented in Supplementary Tables S1 and S3.

Vincristine was administered at 1 mg/kg, a dose that achieves plasma systemic exposures consistent with those seen in children administered 1.5 mg/m² (10, 39). In combination with irinotecan, the dose of vincristine had to be reduced to 0.5 mg/kg due to excessive toxicity. However, even with the dose reduction, the combination met criteria for therapeutic enhancement (i.e., significantly more activity for the combination than either single drug) and for supra-additivity in two models, SK-NEP-1 and ES-4 (Supplementary Table S4). The combination had significantly superior antitumor activity compared with vincristine at its MTD (1 mg/kg), against six models, and significantly superior activity compared with irinotecan alone in 10 of 12 models, the exceptions being KT-11 and Rh65. The combination of reduced dose vincristine with irinotecan was less active than vincristine (1 mg/kg dose) against KT-11 Wilms tumor. Of note, in xenograft models for which the combination of eribulin-irinotecan was superior to the combination of vincristine-irinotecan, all four models are TP53 wild type, whereas the models for which the combinations had similar activity are predominantly TP53 mutant (Supplementary Table S1).

Pharmacodynamic studies were conducted on tumors from 3 mice per group and focused on ES-4 (Fig. 2) and KT-11 (Supplementary Fig. S1), models in which we observed more than additive activity for the combination. Tumors were harvested 24 hours after the first drug administration to determine acute drug effects, and at 144 hours (24 hours after the final irinotecan treatment) to determine pharmacodynamic changes that associated with tumor response. In ES-4 xenografts, eribulin induced a robust increase in histone H3 phosphorylation at 24 hours (Fig. 2A), indicative of mitotic accumulation (40). Phospho-H3 was maintained at 144 hours, consistent with the high affinity of eribulin for microtubules and slow rate of release (41). In contrast, phospho-H3 had decreased to control levels at 144 hours after vincristine treatment (Supplementary Fig. S2A). Irinotecan slightly increased S139 phosphorylation of H2A.X (γ-H2A.X) at 24 hours, but induced a robust increase 24 hours after the fifth drug dose (144 hours), consistent with induction of DNA damage. Of note, both the eribulin-induced increase in phospho-H3 and the irinotecan-induced increase in γ-H2A.X were diminished in tumors from combination-treated mice (Fig. 2; Supplementary Figs. S1A and S3). In contrast, irinotecan effects on total and phosphorylated CHK1 and CHK2 proteins were similar in combination treated tumors. Thus, from a mechanism-based marker analysis, the combination would be considered as antagonistic, with apparently less eribulin-induced mitotic blockade and less DNA damage induced by irinotecan. Similar antagonism of phospho-H3 by irinotecan was observed in KT-11 xenografts (Supplementary Fig. S1A), although at 24 hours the induction of γ-H2A.X appeared to be greater in combination-treated tumors than in those from irinotecan-treated mice. Similar to the interaction between eribulin and irinotecan, the combination with vincristine resulted in diminished phospho-H3 at 24 hours and γ-H2A.X at 144 hours in ES-4 xenografts (Supplementary Fig. S2). These data suggest the S-phase arrest induced by irinotecan prevents eribulin- and vincristine-induced mitotic arrest.

We further investigated induction of DNA damage by eribulin, irinotecan, and the combination in ES-4 tumors by measuring 53BP1 and γ-H2A.X foci formation by immunofluorescence microscopy. At 144 hours, robust nuclear γ-H2A.X staining was observed following irinotecan treatment, but less intense staining was observed in the eribulin and combination-treated samples (Supplementary Fig. S4). In contrast, greater 53BP1 staining was observed in combination-treated tumors than in those treated with either single-agent. These data suggest the combination treatment resulted in more DNA double strand break formation than treatment with either single agent.

In ES-4 tumors, both eribulin and irinotecan alone induced modest PARP cleavage at 24 hours, whereas at 144 hours only tumors from irinotecan-treated mice showed detectable levels of cleaved PARP (Fig. 2A). For the eribulin–irinotecan combination, results were variable with detectable levels of cleaved PARP at 24 hours, but no cleaved PARP was detectable at 144 hours. In KT-11, eribulin alone induced modest PARP cleavage in tumors at 24 and 144 hours (Supplementary Fig. S1A). Irinotecan treatment resulted in robust PARP cleavage only at 24 hours in tumors (Supplementary Fig. S1A). For the combination, robust PARP cleavage was detected in two-third tumors at 24 hours and one-third tumors at 144 hours (Supplementary Fig. S1A). Abundance of cleaved caspase 3 was also evaluated by immunoblotting and IHC. Irinotecan induced caspase 3 cleavage at 24 hours and this increased at 144 hours in ES-4 xenografts (Fig. 2A). Cleaved caspase 3 was lower at 144 hours in tumors from combination treated mice (Fig. 2A). In KT-11 xenografts, cleaved caspase 3 was detected by immunoblot at both 24 and 144 hours only after irinotecan treatment (Supplementary Fig. S1A). By IHC staining, there was a significant increase at 24 hours in ES-4 and KT-11 tumors from combination treated mice compared with those receiving single-agent treatments (Supplementary Fig. S5). Collectively, these results indicate the combination treatment resulted in more apoptosis compared with either single-agent treatment.

To further probe the mechanisms underlying the enhanced activity of the combination, gene expression was analyzed by RNA-seq in ES-4 tumors. Total reads and initial mapping rates for individual replicates are summarized in Supplementary Table S5 and the numbers of DEGs are summarized in Supplementary Table S6. At 24 hours, we identified 38, 896, and 984 DEGs (adjusted P value <0.05 and fold-change ≥2) in the eribulin, irinotecan, and combination groups, respectively. At 144 hours, 163, 1,759, and 2,114 genes were differentially expressed in these groups (adjusted P value <0.05 and fold-change ≥2). Some of the most significantly upregulated genes at 24 hours after eribulin treatment were hydroxyprostaglandin dehydrogenase 15-(NAD) (HPGD), chromosome 12 open reading frame 5 (C12orf5), cyclin-dependent kinase inhibitor 1A (CDKN1A/p21), tumor protein p53 inducible nuclear protein 1 (TP53INP1), pleckstrin homology-like domain, family A, member 3 (PHLDA3), BCL2 binding component 3 (BBC3; PUMA), and 15 histone encoding genes. At 144 hours, these genes remained elevated and MDM2 mRNA was found to be significantly increased as well. The upregulation of p21 and MDM2 mRNA after eribulin treatment was consistent with the increased protein levels observed by immunoblotting (Fig. 2A).

Similar to eribulin treatment, irinotecan induced TP53 responsive genes. Some of the most significantly upregulated genes after 24 hours of treatment included SESN1, SESN2, MDM2, TP53AIP1, TPINIP1, RRM2B (p53R2), CDKN1A (p21), and BBC3 (PUMA; Supplementary Table S7), all of which remained upregulated at 144 hours. We collected all DEGs (n = 2,668) and performed k-means clustering analysis (k = 9). Among nine clusters, clusters 1, 3, and 8 (total of 1,615 genes) showed increased response in irinotecan or combination treated tumors (cluster 1, 804 genes; Fig. 3A, left). Overrepresentation enrichment analysis with enrichr (42) indicated “direct p53 effectors” as its main enrichment (q = 1.97 × 10⁻¹¹). IPA further confirmed TP53 signaling functions (P = 1.1 × 10⁻¹¹, Fig. 3B). The effect of combination treatment was additive in terms of increased expression compared with single-agent treatments for TP53INIP1, RRM2B, CDKN1A, and MDM2 (Supplementary Table S7). Two exceptions were BBC3, which at 144 hours was similar to the irinotecan sample, and TP53AIP1, which was greater than the sum of the expression in tumors treated with single agents.

The TP53 status for each of the tumor models, where known, is included in Supplementary Table S1. Of the eight models where eribulin–irinotecan induced therapeutic enhancement, five lines have wild-type TP53, two have mutations, and the status of RBD1 is not determined. All four models where eribulin–irinotecan was dramatically more active than the vincristine–irinotecan combination are TP53 wild type (defined as EFS_ERB/IRN/EFS_VCR/IRN >1.5; range >1.53–4.22). Tumor samples were probed for activation of the TP53 pathway (TP53, MDM2, p21) by immunoblotting. In ES-4 tumors, eribulin treatment induced a robust increase in TP53 that was greatest at 144 hours, paralleled by increases in CDKN1A/p21 and MDM2 (Fig. 2A). The response to irinotecan was less pronounced, with an increase in TP53 at 24 hours which diminished at 144 hours. In tumors treated with the eribulin-irinotecan combination, TP53 and MDM2 levels were slightly lower than in the eribulin-treated tumors at 144 hours. However, the p21 levels were similar to those in the eribulin-treated tumors at both time points (Fig. 2A). Of note, in ES-4 tumors the TP53-regulated subunit of ribonucleotide reductase, RRM2B/p53R2, was upregulated by both eribulin and irinotecan, and the response at 144 hours to the combination was more similar to that induced by eribulin. The responses were more variable in KT-11 xenografts (Supplementary Fig. S1A); TP53 and p21 levels increased with each treatment at 24 hours, but the effects were diminished by 144 hours in all treatment groups.

We further evaluated induction of TP53 protein in ES-4 and KT-11 tumors by IHC (Fig. 2B and C; Supplementary Figs. S1B and S1C). In ES-4 tumors, single-agent treatment with eribulin resulted in a slight increase in nuclear TP53 staining at 24 hours, but a robust increase at 144 hours, compared with control (Fig. 2B and C). With irinotecan treatment, small but statistically insignificant increases in TP53 staining were observed at 24 and 144 hours. In contrast, the combination of eribulin and irinotecan resulted in significant increases in nuclear TP53 staining at both 24 and 144 hours (Fig. 2B and C), although this increase was less than that observed with eribulin alone at 144 hours. TP53 staining in KT-11 xenografts showed rapid accumulation after 24 hours following eribulin, relatively little induction by irinotecan, but a dramatic increase in nuclear TP53 staining in tumors treated with the combination at 24 hours (Supplementary Figs. S1B and S1C). Nuclear TP53 was elevated at 144 hours in irinotecan treated KT-11 tumors, but had decreased in those from eribulin or combination treated tumors (Supplementary Figs. S1B and S1C). These results suggest the combination of eribulin–irinotecan results in a more rapid induction of the TP53 pathway than that occurring following single-agent treatment.

The antitumor activity of the eribulin–irinotecan combination was dramatic in some tumor models, yet the biomarkers reporting individual drug actions were muted in tumors treated with this combination relative to tumors from mice that received single agents. To resolve this discordance and further interrogate potential mechanism(s) by which the combination was synergistic in vivo, the in vitro cytotoxic effects of eribulin or vincristine combined with SN-38 were examined. All three drugs showed potent cytotoxic activity against ES-4 cells as single agents (Fig. 4A). The absolute IC₅₀ values for eribulin, vincristine, and SN-38 were found to be 0.82, 6.9, and 2.3 nmol/L, respectively. CIs were calculated for multiple combination pairs of eribulin + SN-38 and vincristine + SN-38 based on the Bliss model of independence (Fig. 4A). For eribulin + SN-38, CIs were close to 1 at most concentration pairs, indicating additivity. CIs less than 1 were observed with combinations of relatively low concentration of each drug (0.01–0.1 nmol/L), suggesting synergy, but the effect sizes were very small (Fig. 4A). Similar results were observed with combinations of vincristine and SN-38, with most CIs indicating simple additive effects (Fig. 4A). To further assess the combined effects of eribulin or vincristine and SN-38, ES-4 cells were treated with both drugs at fixed ratios of eribulin:SN-38 and vincristine:SN-38. Isobologram analysis showed that combination effects fell above the line of additivity suggesting that eribulin and SN-38 are antagonistic in terms of cytotoxic efficacy (Fig. 4B). CIs were calculated on the basis of the Loewe model of additivity for each fixed ratio and ranged from 1.12 to 4.38 (Supplementary Table S8). To determine if this antagonism was common among microtubule destabilizers, we assessed the in vitro interaction of vincristine and SN-38. In contrast to the antagonism seen with eribulin, most data points for vincristine and SN-38 fell close to or below the “line of additivity” at fixed ratios of vincristine:SN-38, indicating additivity or synergy with these combinations (Fig. 4B). CIs for these combinations ranged from 0.40 to 0.98 (Supplementary Table S8). Collectively, these results indicate that in vitro eribulin and vincristine have different effects, possibly linked to their different mechanisms of action, when combined with SN-38. However, studies in additional cell lines would be required to further establish this.

Eribulin has been shown to increase microvessel density and vascular permeability, which could increase drug uptake (43). To determine if eribulin, irinotecan, or the combination affects tumor vasculature, ES-4 tumor sections were stained for CD34 and α-SMA. There was a trend towards increased CD34⁺ cells in tumors from combination-treated mice compared with control mice, although this was not statistically significant (Figs. 5A and B). Image quantification showed a nearly 6-fold increase in α-SMA-positive cells in combination-treated tumors compared with control tumors at 144 hours (Fig. 5A and B), suggesting an increase in tumor vasculature. We also assessed the effects of vincristine and vincristine combined with irinotecan on ES-4 tumor vasculature. We observed a decrease in CD34 staining after treatment with vincristine for 24 and 144 hours. Decreased CD34 staining was also observed with the combination of 0.5 mg/kg vincristine and irinotecan at 24 and 144 hours. With the combination of 1.0 mg/kg vincristine and irinotecan, we detected a decrease in CD34 staining at 24 hours, but approximately twice the amount of CD34 after 144 hours of treatment (Fig. 5B). In addition, we assessed the effects of vincristine and irinotecan on α-SMA abundance in ES-4 xenografts. After 24 hours of treatment, we did not observe any significant changes in α-SMA abundance compared with control tumors for any of the treatment groups. After 144 hours, the vincristine and vincristine + irinotecan groups showed increased α-SMA abundance compared with control tumors, but these increases were not statistically significant.

To assess the tumor-independent effects of these treatments on vasculature, we implanted Matrigel plugs containing 100 ng/mL VEGF into C.B.17SC scid^−/− female mice, allowed vasculature to form for 1 week and then treated the mice with one cycle of eribulin, vincristine, irinotecan, eribulin + irinotecan, or 1.0 mg/kg vincristine + irinotecan. At 1 week after Matrigel implantation, there was clear vascular formation and CD34 positivity (Fig. 6). At day 14, plugs from control, eribulin, irinotecan, and eribulin + irinotecan-treated mice showed detectable levels of CD34 in the interior of the Matrigel plugs, whereas plugs from vincristine and vincristine + irinotecan-treated mice only showed detectable CD34 around the plug peripheries (Fig. 6), suggesting that vincristine reduced infiltration of vasculature into the Matrigel plugs. The CD34⁺ vasculature from eribulin and eribulin + irinotecan-treated mice appeared to infiltrate the Matrigel plugs to a greater extent and had a smaller diameter than those from the other treatment groups, suggesting that eribulin treatment does not disrupt vasculature to the same extent as vincristine. All plugs showed minimal staining for α-SMA, with plugs from eribulin-treated mice displaying the highest levels (Supplementary Fig. S6).

Vincristine plays an essential role in treatment of pediatric solid tumors and acute leukemias, but is associated with significant peripheral neuropathy when administered over prolonged periods. Potentially, substitution of eribulin for vincristine could reduce peripheral neuropathy in patients, hence improve patient quality of life. At the dose levels and schedules of eribulin and irinotecan tested, there is additive or greater than additive activity for the models evaluable. The combination of eribulin–irinotecan was significantly more active than was vincristine–irinotecan in 6 of the 12 models tested, with the vincristine–irinotecan combination being significantly more active for a single model and with three models not being evaluable for time to event comparisons because of the very high activity for each combination. These results reflect in part the ability to combine eribulin with irinotecan at full-dose, whereas vincristine had to be dose-reduced for the combination. However, clinically, combination of vincristine with irinotecan does not require dose reduction of either agent.

For four models, where the EFS was far greater for the eribulin combination compared with the vincristine combination (EFS_EI:EFS_VI >1.5), each of these models isTP53 wild type, whereas four of five xenograft models where the combinations had similar activity have mutant TP53, the exception being RBD1 in which TP53 genotype has not been determined. Further, the antitumor activity of the combination correlated with activation of the TP53 pathway as determined by RNA-seq. Notably, the predominant changes in expression were determined by irinotecan treatment. Sestrins 1 and 2 (SESN1/2), induced downstream of DNA-damage-induced TP53, suggest that irinotecan may suppress TORC1 activity (44), that could affect DNA damage repair (45). Also, TP53 apoptosis inducing protein 1 (TP53AIP1) was enhanced 40-fold at 24 hours after irinotecan and almost 80-fold at 144 hours following the eribulin–irinotecan combination. In contrast, eribulin did not induce TP53AIP1 and only slightly reduced Sestrins 1/2.

The mechanism of action of the eribulin–irinotecan combination appears complex. In tumors, irinotecan prevents eribulin-induced mitotic accumulation, and eribulin appears to reduce biomarkers of DNA damage induced by irinotecan. Thus, from the perspective of mechanism-related biomarkers, the combination appears antagonistic. Similarly, markers of apoptosis tended to be lower in tumors from mice that received either eribulin or vincristine combined with irinotecan, compared with tumors from mice that received single-agent therapy. The reason for this remains obscure, although more rapid induction of apoptosis, may have occurred prior to the later time point (144 hours) when tumors were sampled. An alternative explanation would be that another mechanism of cell death, other than apoptosis, was induced by combination treatments. In vitro isobologram analysis indicated antagonism between eribulin and SN-38 whereas there was synergy when SN-38 was combined with vincristine. In contrast, analysis using the CI approach showed little interaction between eribulin and SN-38, thus further studies with additional cell lines will be required to determine whether there is an interaction between these agents in vitro.

Giannakakou and colleagues showed that vincristine and paclitaxel enhance TP53 induction in response to doxorubicin (46), and increase TP53 nuclear accumulation. In both ES-4 and KT-11 xenografts, nuclear TP53 was dramatically increased at 24 hours in tumors from combination treated mice, consistent with the onset of caspase 3 cleavage. These results strongly implicate the role of TP53 in mediating the response to eribulin–irinotecan therapy, and support the model proposed whereby inhibition of microtubule dynamics leads to accumulation of nuclear TP53 under conditions of DNA damage. Eribulin would also be anticipated to block trafficking of DNA repair proteins (47), including DNA-PK. As phosphorylation of H2A.X (γ-H2A.X) is partly dependent on the activity of DNA-PK (48), this may account for the decreased γ-H2A.X signal, although in the study by Poruchynsky and colleagues (47), γ-H2A.X was enhanced by microtubule inhibitors.

The antagonism of cytotoxicity in vitro, at least by isobologram analysis, suggests the antitumor effects of the combination may be mediated by drug effects on tumor microenvironment, consistent with reports where eribulin induced vascular remodeling and reversed epithelial-to-mesenchymal transition in breast cancer models (21, 23). In addition, eribulin was shown to increase tumor uptake of radiolabeled liposomes (43), possibly due to its effects on tumor vasculature. The greatest increase in tumor CD34 or αSMA positivity occurred at 144 hours in combination-treated tumors (both eribulin and vincristine at the higher dose level), suggesting that combination treatment increased vascularity. In contrast, irinotecan appeared toxic to vascular cells that had invaded Matrigel plugs, whereas combination of eribulin and irinotecan was not. Of interest was the observation that in mice treated with eribulin–irinotecan, vascular cells had invaded the plugs whereas in vincristine–irinotecan treated mice vascular cells were abundant at the periphery of the Matrigel plug, but had not invaded. Although further studies are necessary, the results could indicate that vasculature may be maintained in combination-treated tumors, whereas it is destroyed in tumors treated with irinotecan, thus preserving drug delivery when the combination treatment was administered. Further studies will be required to resolve the role of microenvironment in the antitumor activity of the eribulin–irinotecan combination, particularly with respect to the TP53 status which associates with enhanced tumor response to this combination.

In summary, the combination of eribulin–irinotecan was highly active in most xenograft models of pediatric solid tumors. Several lines of evidence indicate that the combination activates the TP53 pathway and increases nuclear TP53; further, in tumors for which eribulin–irinotecan had greater activity than vincristine–irinotecan, tumors were wild type for TP53. Thus, TP53 status could affect tumor responsiveness to this combination and should be considered in the design of clinical trials.

No potential conflicts of interest were disclosed.

Conception and design: A.J. Robles, M.A. Smith, P.J. Houghton

Development of methodology: A.J. Robles, A. Bandyopadhyay, S.W. Erickson, M.A. Smith, P.J. Houghton

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.J. Robles, R.T. Kurmasheva, Z. Lai

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.J. Robles, S.W. Erickson, D. Kurmashev, Y. Chen, P.J. Houghton

Writing, review, and/or revision of the manuscript: A.J. Robles, R.T. Kurmasheva, S.W. Erickson, M.A. Smith, P.J. Houghton

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.A. Phelps, Y. Chen, P.J. Houghton

Study supervision: P.J. Houghton

This work was supported by UO1CA199297 and CA165995 from the NCI. A.J. Robles was supported by the Cancer Prevention and Research Institute of Texas-funded UT Health San Antonio Cancer Biology Training Program (CPRIT; RP170345) and T32CA148724 (NCI). The authors acknowledge the contributions of Edward Favours, Vanessa Del Pozo, Samson Ghilu, and Kathryn Bondra for conducting the in vivo studies, and Daniel Robledo for IHC processing.

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