Activity of MM-398, Nanoliposomal Irinotecan (nal-IRI), in Ewing's Family Tumor Xenografts Is Associated with High Exposure of Tumor to Drug and High SLFN11 Expression

Irinotecan hydrochloride, a water-soluble analogue of the natural alkaloid camptothecin (CPT; ref. 1), is first-line therapy in combination with 5-fluorouracil (5-FU) and leucovorin for metastatic colorectal cancer (2, 3). Carboxylesterase catalyzes irinotecan into SN-38, which is >100-fold more cytotoxic than irinotecan against tumor cells in vitro (4, 5). However, the complicated pharmacology of irinotecan has likely limited its clinical activity. An active lactone form spontaneously hydrolyzes into the inactive carboxylate form in normal physiologic pH until the carboxylate form reaches equilibrium (∼30% of total irinotecan; refs. 6, 7). Other pharmacologic disadvantages of irinotecan include fast elimination of the drug (8, 9), and diarrhea, due to β-glucuronidases produced by intestinal microflora that mediate hydrolysis of SN-38 glucuronide (inactive phase II metabolite of SN-38) to form the active metabolite SN-38 (10).

Liposomal formulations of irinotecan have been introduced (11, 12), which slow elimination and minimize exposure of the drug to serum (which at neutral to basic pH promotes hydrolysis of irinotecan), and thus stabilize the active lactone form versus the inactive carboxylate form (13). A nanoliposomal formulation of irinotecan, nal-IRI (also known as MM-398 or PEP02), employs a modified gradient loading method with sucrose octasulfate (13). Macrophage-mediated disruption of the nal-IRI, drug release, and conversion to SN-38 for in vivo require >24 hours, and preclinical studies demonstrated activity against colorectal and breast cancer subcutaneous xenografts (13), as well as intracranial glioblastoma tumor xenografts (14–16). A three-arm phase II study of nal-IRI versus irinotecan or docetaxel as a second-line therapy in patients with locally advanced or metastatic gastric or gastro-esophageal junction adenocarcinoma demonstrated an objective response rate comparable with docetaxel (17). A nal-IRI phase II study in patients with metastatic, gemcitabine-refractory pancreatic cancer achieved a one-year survival rate of 25% (18). Nal-IRI combined with 5-FU and leucovorin demonstrated significantly improved overall survival for gemcitabine-refractory pancreatic cancer patients compared with 5-FU/leucovorin alone in a recently completed phase III pivotal trial (NAPOLI-1; ref. 19).

Here, we report the comparative pharmacokinetic profiles and antitumor activity in mouse xenograft models of pediatric solid tumors of nal-IRI and irinotecan (Camptosar). High expression of SLFN11 has been reported for Ewing's family of tumors (20), and SLFN11 expression has been associated with sensitivity to cytotoxic agents (21). Therefore, we also examined the correlation between in vitro activity of SN-38 and SLFN11 expression in panels of pediatric cancer cell lines, the relative expression of SLFN11 in pediatric primary tumors, and the correlation of SLFN11 expression with the Ewing's-specific aberrant transcription factor EWS/FLI (22).

Merrimack Pharmaceuticals provided nal-IRI (nanoliposomal irinotecan·HCl, 5.0 mg/mL). Irinotecan (Camptosar, 20 mg/mL, free drug formulation) was from Baxter Healthcare Corporation. 4-HC was purchased from Niomech (Germany), and cyclophosphamide for in vivo studies was from Baxter. Topotecan, pure irinotecan·HCl, SN-38 (active metabolite of irinotecan), camptothecin (internal standard: IS), and all other chemicals were LC-MS grade and were purchased from Sigma-Aldrich Chemical Company. Methanol and acetonitrile were from Fisher Scientific. Normal mouse plasma was from Lampire Biological Laboratories and filtered by 6 mL charcoal packing Resprep SPE cartridges (Restek). Drug-free human plasma was obtained from the TTUHSC Blood Bank.

For these studies, we employed 20 Ewing's sarcoma (EFT), 20 neuroblastoma (NB), and 12 rhabdomyosarcoma (RMS) cell lines that were established in our laboratory or obtained from the originator of the line (Supplementary Table S1). All cell lines were grown in antibiotic-free RPMI-1640 supplemented with 10% heat-inactivated FBS or Iscoves Modified Dulbeccos Medium (IMDM) supplemented with 3 mmol/L _L-glutamine, 5 μg/mL insulin, and 20% heat-inactivated FBS, were mycoplasma free, and were cultured and treated with drugs in a 37°C incubator with 5% CO₂. Cell line identity was confirmed at the time of experimentation by short tandem repeat (STR) genotyping (23) and compared with the Children's Oncology Group STR database (www.COGcell.org).

All animal studies were approved by the Animal Care and Use Committee at TTUHSC. Female BALB/c mice (6–8 weeks, 16–18 g) and nu/nu mice (6–8 weeks, 24–28 g) were from Charles River Laboratories. Irinotecan or nal-IRI was injected intravenously at 5, 10, or 20 mg/kg into the lateral tail vein of BALB/c mice once. Blood was collected by cardiac puncture at various time points under isoflurane anesthesia and centrifuged immediately at 15,700 × g for 5 minutes at 4°C to isolate the plasma. Organs were harvested, minced, mixed with 3 volumes of PBS, and homogenized on ice using a Tissue Tearor (BioSpec Products). Protein content in tissues was determined as previously described (24). Plasma and tissue homogenates were stored at −80°C until analyses.

For tumor pharmacokinetics, the human CHLA-136 neuroblastoma cell line (15 × 10⁶ cells/100 μL) was injected subcutaneously into nu/nu mice and after 42 days irinotecan or nal-IRI was injected intravenously (20 mg/kg). Blood, organs, and tumors were collected and processed as described above 4, 24, and 48 hours after injection.

The concentrations of irinotecan and SN-38 in mouse plasma and tissues were determined by an Agilent 1200 HPLC system (Agilent Technologies) and a Applied Biosystems 4000QTRAP LC-MS/MS system (Applied Biosystems) with data analysis software Analyst 1.42 (Applied Biosystems). Isocratic separation was achieved using 50% of methanol (0.1% formic acid) and 50% of 2 mmol/L ammonium formate in water (0.1% formic acid) on a 2.1 × 50 mm Waters XBridge C18 Column (3.5 μm; Waters) with a 2.0 × 4 mm Phenomenex Security Guard C18 cartridge (Phenomenex) at a flow rate of 0.3 mL/min. The retention times for irinotecan, SN-38, and internal standard (I.S.) were 1.20, 2.02, and 2.55 minutes, and the total run time was 6 minutes.

The column effluent was connected to an electrospray ionization interface operated in the positive mode (600°C and 3,000 V ionization voltage); nebulizer gas = 40 psi and turbo gas = 50 psi; collision-activated dissociation gas was set at medium and the curtain gas at 15 psi. The collision energies for irinotecan, SN-38, and camptothecin were 59, 37, and 33 EV, and the collision exit potential for the analytes were 12, 20, and 16 EV, respectively. The detection of the analytes was performed using multiple reaction monitoring. The transitions were m/z 588.4 → m/z 167.2 for irinotecan, m/z 393.2 → m/z 349.1 for SN-38, and m/z 348.8 → m/z 305.0 for camptothecin.

Plasma calibration standards of irinotecan were prepared at 1.0 to 500 ng/mL, and for SN-38 at 0.5 to 250 ng/mL in purified mouse plasma. Tissue calibrations of irinotecan and SN-38 were performed in blank mouse liver homogenate (1:3 dilution with PBS). Method validation was per FDA guidelines (http://www.fda.gov/CDER/Guidance/4252fnl.html). The regression parameters of slope, intercept, and correlation coefficient were calculated by least-squares linear-regression analysis using a weight factor of 1/x². For noncompartmental analysis, the area under the concentration versus time curve (AUC_0-inf) was calculated by using the trapezoidal rule extrapolated to infinity. Calculations were as follows: elimination half-life: T_1/2 = 0.693/ke; systemic clearance: CL_tot = dose/AUC); and apparent volume of distribution: Vd = Dose/(AUC_0-inf × ke).

Irinotecan was spiked into blank human and mouse plasma to achieve final concentrations of 0, 10, 20, 50, 100, 250, 500, 750, and 1,000 ng/mL, incubated at 37°C for 24 hours, and reaction terminated by adding 600 μL of ice-cold acetonitrile–methanol (1:1, 0.1% formic acid) and 2 μL of internal standard solution (camptothecin 5 μg/mL). The concentrations of SN-38 in plasma were determined by LC-MS/MS. V_max and K_m values were calculated using GraphPad Prism software.

Subcutaneous xenografts with human cell lines from EFT (CHLA-258, COG-E-352,CHLA-32, SK-N-MC, and CHLA-25; refs. 25), RMS [CB-NJR (26), SMS-CTR (27), and Rh30 (direct xenografts; ref. 28)], and neuroblastoma (CHLA-136, CHLA-79, and CHLA-119; ref. 29) were established in nu/nu mice. All cell lines were grown before xenograft implantation in antibiotic-free medium and verified to be free of mycoplasma (MycoAlert kit; Lonza) and cell line identity confirmed by STR at the time of experimentation (23). For each 6- to 8-week-old female athymic nu/nu mouse, 2.5 to 10 million tumor cells were mixed with RPMI-1640:Matrigel (1:1) and injected subcutaneously between the shoulder blades. Tumor volume was measured twice weekly (28, 30). Drug treatment (irinotecan or nal-IRI, 10 mg/kg/d, Monday and Thursday, a total of 2 doses were given) began when progressively growing tumors reached a mean of 150 mm³. Mice were sacrificed when the tumor volume exceeded 1,500 mm³ or serious morbidity was observed. To assess combination activity of nal-IRI + cyclophosphamide given in multiple doses, mice were treated for 4 weeks with intravenous nal-IRI 10 mg/kg/d, each Monday and Thursday, and intraperitoneal cyclophosphamide 25 mg/kg (every Monday).

Chromogenic immunohistochemistry was conducted on 5-μm-thick tumor frozen sections on fresh cut slides. Slides were baked for 30 minutes at 65°C, soaked in xylene, rehydrated through 100% and 80% graded alcohols, and pretreated with proteinase K (Invitrogen) for 15 minutes at room temperature. All further steps were done at room temperature in a hydrated chamber. Slides were blocked using Peroxidazed1 (Biocare) and serum-free protein block (Dako) to quench endogenous peroxidase activity, and nonspecific binding, respectively, stained overnight with the F4/80 antibody (Serotec, MCA497G, 1.0 mg/mL) incubated at a 1:1,000 dilution, and was developed using the rat on mouse HRP polymer kit (Biocare) as per the manufacturer and counterstained with hematoxylin. Slides were imaged on an Aperio Brightfield microscope at a 20× magnification. The percentage of macrophage-positive area was quantified as a ratio of total viable tumor tissue area was carried out using Definiens Tissue Studio XD (Definiens).

We assessed cytotoxicity of SN-38 using the DIMSCAN cytotoxicity assay as previously described (31, 32) and SLFN11 expression by Taqman real-time reverse transcription-PCR (RT-PCR) as previously described (33) using primers and a probe from Applied Biosystems in 20 EFT, 20 NB, and 12 RMS cell lines (Supplementary Table S1). Relative IC₅₀ (rIC₅₀) was calculated based on NCI Pediatric Preclinical Testing Program (PPTP) guidelines (32). SLFN11 expression was also measured by Human Exon (HuEx) 1.0 ST Array (Affymetrix) in 17 EFT cell lines and expression measured by HuEx arrays correlated with expression by RT-PCR (r = 0.62, P = 0.007). SLFN11 protein expression was assessed by immunoblotting (antibody from Santa Cruz) as previously described (33).

Transcriptional profiling was performed on 150 cases of childhood tumors, 50 each of EFT, NB, and RMS (both embryonal and alveolar, in equal numbers) using Affymetrix Human Exon (HuEx) arrays. Data were imported and normalized using PLIER and quantile normalization. Mean expression values were set to 80 per the manufacturer's recommendation. The expression levels for both Schlafen family member 11 (SLFN11) and Friend leukemia integration site 1 (FLI1) were analyzed for each tumor in each category and compared individually and categorically. Results were expressed as the log to base 10 in each case. For comparison of SLFN11 versus FLI1, a scatter analysis with equal x and y axes (log₁₀) was employed. The R² value was calculated using a simple linear regression model to determine the correlation coefficient.

Mouse event-free survival (EFS) was graphically represented by the Kaplan–Meier analysis, and EFS between groups was analyzed by the log-rank test using GraphPad Prism (GraphPad Software). EFS was defined as the time from initiating treatment until death, tumor size greater than or equal to 1,500 mm³, or treatment-related toxicity. For pharmacokinetic experiments, the statistical differences were evaluated by the Student t test. Correlation between rIC₅₀ and SLFN11 expression was determined by Pearson r correlation. The differences in macrophage content in tumors were assessed by the one-way ANOVA. P < 0.05 was considered significant.

We compared the antitumor activity of nal-IRI (10 mg/kg/dose, total 2 doses: Monday and Thursday) to irinotecan (10 mg/kg/dose, total of 2 doses: Monday and Thursday) in xenograft models (three each) of EFT, RMS, and NB. Treatment with nal-IRI significantly extended EFS compared with control in seven of nine xenograft models and compared with irinotecan in six of nine models (Fig. 1, P values are included in Supplementary Table S2). Maintained complete responses were observed over 110 days in the majority of mice treated with nal-IRI, but not free irinotecan, in all EFT xenograft models (CHLA-258, COG-E-352, and CHLA-32). The number of mice per group, the median EFS, and the P values are listed in Supplementary Table S2. In one of three RMS mouse models (CB-NJRm), nal-IRI–treated mice showed significantly greater EFS (median survival, 38.5 days) compared with irinotecan (22 days, P < 0.0001) or control mice (13 days, P < 0.0001). In the three neuroblastoma xenograft models (CHLA-136, CHLA-79m, and CHLA-119), EFS with nal-IRI was greater than with irinotecan (P < 0.0004) or vehicle control (P < 0.0001), but duration of responses in two of the three NB models was <100 days in the majority of mice treated with nal-IRI. Dosing of 10 mg/kg/d for both nal-IRI and irinotecan was well tolerated in xenografted mice; body weight during the course of therapy is shown for two representative models in Supplementary Fig. S1A. Overall, EFT xenograft mice treated with nal-IRI showed significantly longer EFS relative to RMS or NB (Supplementary Fig. S1B).

We also evaluated nal-IRI alone and when combined with cyclophosphamide when given weekly for 4 weeks in five xenograft models of EFT, including xenografts from cell lines established from progressive disease after myeloablative therapy (CHLA-258 and COG-N-352), compared with three RMS xenograft models. Cyclophosphamide was chosen for the combination studies due to its common use in pediatric oncology, and our data showing that the active metabolite of cyclophosphamide (4-HC) showed the best synergy with topoisomerase I inhibitors among tested agents (34). EFS for all 5 EFT xenograft models treated with nal-IRI + cyclophosphamide was 92% at 150 days, but only 13% at 110 days for the 3 RMS xenografts (Supplementary Fig. S1C). EFS for nal-IRI + cyclophosphamide in RMS xenografts at 80 days was 27% compared with 0% for nal-IRI as a single agent (P = 0.0014). However, due to the high activity against EFT of nal-IRI, the increase in EFS for nal-IRI + cyclophosphamide of 92% compared with 73% for nal-IRI alone was not significant (P = 0.08).

Plasma and tissue pharmacokinetic profiles for irinotecan and SN-38 following the administration of free irinotecan or nal-IRI at 5, 10, and 20 mg/kg in nontumor-bearing BALB/c mice are shown in Fig. 2A. After a 20 mg/kg dose of free irinotecan, the observed maximum plasma concentration of irinotecan was 2.8 ± 0.3 μg/mL at 5 minutes after injection, whereas nal-IRI at the same dose achieved 197.0 ± 10.3 μg/mL at 1 hour (P < 0.01). At 8 hours after injection with free irinotecan, the irinotecan concentration was <0.01 μg/mL, whereas it was 98.0 ± 12.7 μg/mL at 8 hours after nal-IRI injection (P < 0.01). The rapid clearance of SN-38 from mouse blood following free irinotecan administration was consistent with the observed irinotecan clearance. The plasma concentration of SN-38 after injection of nal-IRI showed an increase up to 12 hours after injection followed by a gradual decrease over time. The pharmacokinetic parameters in mice for both formulations are summarized in Supplementary Table S3. The half-life of irinotecan (0.8–1.1 hours) in plasma was significantly shorter than that of nal-IRI (2.6–4.4 hours, P = 0.001), and the clearance was greater with irinotecan (138–300 mL/h) compared with nal-IRI (0.2–0.4 mL/h, P = 0.024).

After the administration of irinotecan we observed a concentration drop to 1/20–1/300 of the maximum concentrations over 8 hours in all tissues (Fig. 2B). Irinotecan and SN-38 levels were higher in liver, spleen, and kidney relative to the lung and brain. Irinotecan and SN-38 concentrations achieved by nal-IRI were comparable with free irinotecan soon after intravenous injection, but elimination of both irinotecan and SN-38 from tissues was substantially slower for nal-IRI than for free irinotecan (Fig. 2C).

In nu/nu mice bearing human NB xenografts (CHLA-136), intravenous injection of 20 mg/kg free irinotecan achieved a irinotecan plasma level of 0.06 ± 0.01 μg/mL at 4 hours after injection, barely detectable above the limit of quantitation at 24 hours (Fig. 3A, left). At the same dose, nal-IRI showed plasma irinotecan concentrations at 4 hours (131.4 ± 35.8 μg/mL) 1,000-fold higher than achieved with free irinotecan. Irinotecan was above the detection limit (0.014 ± 0.005 μg/mL) 48 hours after nal-IRI injection. Plasma SN-38 concentrations were >100-fold higher with nal-IRI relative to irinotecan (Fig. 3A, right).

Elimination of irinotecan and SN-38 for most tissues was more rapid with free irinotecan relative to nal-IRI in tumor-bearing nu/nu mice (Fig. 3B–F), similar to what was observed in nontumor-bearing mice. At 48 hours after injecting free irinotecan, the irinotecan levels in tissues and tumor were decreased by 2 orders of magnitude compared with those at 4 hours, and the concentration in the brain was below the detection limit (Fig. 3B). Elimination of nal-IRI in tissues and tumor was considerably slower than observed for irinotecan (Fig. 3B–F). After 4 hours of nal-IRI injection, the tissue levels of irinotecan and SN-38 in the liver reached 548 ± 396 μg/g protein (irinotecan) and 15.1 ± 8.6 μg/g protein (SN-38), which are the highest in all tissues measured (Fig. 3E). The levels of SN-38 in tumor at 48 hours with nal-IRI were almost 10-fold higher than we observed with irinotecan (Fig. 3F).

As liposomal stabilization of irinotecan should minimize release of the drug before it reaches tumor sites, we did not expect to see >100-fold higher plasma SN-38 levels in nal-IRI–injected mice compared with free irinotecan-injected mice (Fig. 3A). As nal-IRI–injected mice did not experience greater systemic toxicities relative to irinotecan-injected mice, it was speculated that higher SN-38 concentrations might have resulted after samples were collected. Thus, we assessed two methods of blood sample collection, direct extraction without freezing plasma samples versus freezing samples for future analyses. Freezing samples for future analyses resulted in approximately 2-fold higher SN-38 concentrations up to 20 hours after the injection of nal-IRI (Supplementary Fig. S2A and S2B), indicating that irinotecan is further converted to SN-38 during the freeze-thaw process by carboxylesterase in mouse plasma.

A plasma carboxylesterase content is higher in mice than in humans (35), and postcollection conversion of irinotecan to SN-38 may be less of an issue in humans. Indeed, in vitro enzyme kinetic experiments demonstrated more limited conversion of irinotecan to SN-38 for human versus mouse plasma at 24 hours (37°C). The Michaelis–Menten kinetic parameters of V_max (maximum velocity) and K_m (substrate concentration that yield a half-maximal velocity) were 0.3 pmol/hr/mg protein and 12,471 ng/mL (31.8 μmol/L) in human plasma, and 1.3 pmol/hr/mg protein and 393 ng/mL (1.0 μmol/L) in mouse plasma (Supplementary Fig. S2C).

Responses to nal-IRI in xenograft models varied widely, with activity in EFT being consistently greater than what was observed for RMS or NB xenografts (Fig. 1 and Supplementary Fig. S1B). Tumor-associated macrophages could potentially concentrate nal-IRI and facilitate release of the more potent SN-38 into tumors. However, we observed no significant differences in macrophage infiltration among NB, RMS, or EFT tumor xenografts at 0, 48, or 168 hours following nal-IRI treatment (Fig. 4A, P = 0.31, representative pictures in Supplementary Fig. S3). Treatment with nal-IRI significantly increased macrophage infiltration in all three models (P < 0.005), but no greater infiltration was seen in any specific tumor type relative to other models at 168 hours after the treatment (Fig. 4A, P = 0.37).

The tumor concentrations of irinotecan (but not of SN-38) in EFT or RMS xenografts were significantly higher than NB xenografts (Fig. 4B, P < 0.05). Plasma concentrations of irinotecan and SN-38 at 48 hours after injection in RMS xenograft mice were apparently lower compared with NB or EFT xenograft mice, although the differences were not statistically significant (P = 0.07). However, SN-38 plasma concentrations were above the detection limit (2.5 ng/mL) at 168 hours in the RMS model, whereas NB or EFT models were below the limit of detection. Given that no differences in macrophage infiltration were seen for the 3 tumor types, and that the tumor concentrations of the drug and the metabolite were not higher in EFT models (which showed durable complete responses in all five models tested, Fig. 1 and Supplementary Fig. S1B and S1C), macrophage infiltration or tumor concentrations of irinotecan or SN-38 were excluded as the major mechanism for the remarkably high nal-IRI activity observed in EFT relative to NB or RMS xenografts.

High expression of the putative DNA/RNA helicase SLFN11 has been recently reported to correlate with sensitivity to DNA damaging agents (20, 21). We observed that sensitivity to DNA damaging agents (4-HC: active metabolite of cyclophosphamide, topotecan, SN-38), as determined by the DIMSCAN cytotoxicity assay, showed a significant correlation with SLFN11 expression assessed by the Affymetrix Human Exon Array in 17 EFT cell lines [Supplementary Fig. S4, P < 0.005 for all three drugs, r = −0.95 (4-HC), −0.76 (topotecan), −0.84 (SN-38)]. Using quantitative RT-PCR for 20 EFT, 20 NB, and 12 RMS cell lines, we observed that SLFN11 expression was >100-fold higher in EFT relative to NB cell lines, whereas the expression varied widely in RMS (Fig. 5A). SLFN11 protein and mRNA expression were consistent, showing higher expression in the majority of EFT cell lines relative to NB or RMS (Fig. 5B). We observed an inverse correlation between SLFN11 mRNA expression measured by RT-PCR and the relative inhibitory/cytotoxic concentrations (rIC₅₀) concentrations (measured by DIMSCAN) of SN-38 for 20 EFT cell lines (P = 0.004, r = −0.63, Fig. 5C). However, no significant correlation between rIC₅₀ and SLFN11 mRNA expression was observed in RMS or NB cell lines. (P = 0.26 and P = 0.21; Fig. 5D and E).

As SLFN11 mRNA expression assessed by RT-PCR and Affymetrix HuEx microarray correlated well in cell lines (r = 0.62, P = 0.007), we examined SLFN11 mRNA expression data from HuEx microarrays for primary tumors (50 each for EFT, RMS, and NB). Of the 50 EFT samples, 90% had the EWS–FLI1 translocation and 10% had EWS–ERG. Rhadomyosarcomas were 1/3rd alveolar and 2/3rds embyronal and neuroblastomas were 1/3rd favorable biology, 1/3rd MYCN-amplified, and 1/3rd MYCN-nonamplified with unfavorable histopathology. Consistent with the expression in the cell line panels, median SLFN11 mRNA expression was approximately 10-fold higher in EFT primary tumors compared with NB or RMS primary tumors (P < 0.0001; Fig. 6A). FLI1 is not normally expressed, and thus using the 3′end of the gene provides a surrogate for the chimeric EWS–FLI1 gene. Using this approach, coexpression of FLI1 and SLFN11 was evaluated using the same microarray data employed for Fig. 6A, comparing an exon of SLFN11 (3753518) with the 8th exon of FLI1 (3355784), which is expressed at high levels only when translocated, as part of EWS–FLI1 (r = 0.51, P < 0.001; Fig. 6B).

Although diarrhea is the dose-limiting toxicity of irinotecan, it can be managed to some degree by employing antidiarrhea medications at the early stage of treatment (14). Other limitations of irinotecan for its clinical application include (i) poor pharmacokinetic properties only allowing short tumor exposure to the drug and its active metabolite (SN-38) and (ii) hydrolysis of active lactone rings in irinotecan and SN-38 to inactive carboxylate forms in neutral to basic pH (36). Stable liposomal encapsulation of irinotecan may minimize the transformation of the irinotecan into its inactive carboxylate form by reducing the direct exposure of the drug to physiologic pH and prolong the residence time of active drug and metabolites in the system. Because only the lactone form of camptothecin analogues is able to transverse cellular membranes (37), decreasing hydrolysis of irinotecan by encapsulating in liposomes has the potential to improve clinical activity.

Recent advances in intraliposomal encapsulation technology showed that irinotecan formulated as a nanoliposomal formulation (nal-IRI = MM-398) achieved high in vivo chemical stability, liposome retention, and improved residence time of irinotecan in rats (13). We observed that levels of SN-38 in xenografted tumors were significantly higher with nal-IRI than with irinotecan. The increased tumor exposure to irinotecan and SN-38 achieved with nal-IRI relative to irinotecan was consistent with the higher antitumor activity and superior EFS seen in the mouse xenograft models of pediatric solid tumors treated with nal-IRI compared with irinotecan in the majority of xenograft models tested.

The accumulation of liposomal drugs in tumors is suggested to occur via two pathways. One is via the enhanced permeability and retention (EPR) effect (11, 38). Liposomes do not readily diffuse out of the blood vessels in healthy tissues (39), but the leaky vasculature and impaired lymphatic drainage in tumor can facilitate the spontaneous accumulation of liposomes in these areas (40). The other pathway is through the localization of liposomes to mononuclear phagocytes such as macrophages (38). Although the number of macrophages varied greatly among different tumors, the percentage of tumor-associated macrophages is usually maintained at a stable level for a particular tumor type (41). In addition, macrophage carboxylesterases in tumor is a presumed mechanism for the conversion of irinotecan to SN-38 in human tumors (14, 42, 43), and the macrophage levels in nu/nu mice are comparable with levels observed in immunocompetent mice (44). However, we observed that neither macrophage infiltration nor intratumoral concentrations of irinotecan or SN-38 were significantly higher in EFT models in which durable responses to nal-IRI were seen relative to xenografts of NB and RMS that were less responsive to nal-IRI. Our data showed that macrophage infiltration into tumors or tumor pharmacokinetics does not vary among xenografts of the 3 histotypes of pediatric solid tumors that we tested, and thus is unlikely to be the reason for the consistently high responsiveness of EFT xenografts to nal-IRI relative to other histotypes.

The mRNA expression of SLFN11, a putative RNA/DNA helicase, has been reported to correlate with the sensitivity to DNA damaging agents in the NCI60 panel of cell lines (21) or with the sensitivity to irinotecan in three EFT cell lines (20). Consistent with the previous report by Barretina and colleagues showing highest expression of SLFN11 in EFT primary samples among various cancer types (20), we found that SLFN11 mRNA expression was significantly higher (>100-fold) in EFT cell lines relative to NB and RMS cell lines (Fig. 5). Although NB cell lines exhibit much lower SLFN11 expression, the sensitivity to SN-38 in NB cell lines was similar to that in EFT cell lines, suggesting that SLFN11 expression is not the primary driver of SN-38 sensitivity across all histotypes. For example, CHLA-119 (established at time of relapse after intensive consolidation chemotherapy; CCG-3891) that included topoisomerase II but not a topoisomerase I inhibitor is one of four NB cell lines with the lowest SLFN11 expression, but nine of the ten xenograft mice survived >120 days after two doses of nal-IRI (Fig. 1).

In 20 EFT cell lines, we observed a strong correlation between SLFN11 expression and SN-38 sensitivity that was confirmed by both microarray and quantitative RT-PCR experiments. The high sensitivity of EFT cell lines to SN-38 in vitro was consistent with in vivo EFT tumor xenograft responses to nal-IRI. Thus, the high activity seen for nal-IRI in EFT xenografts is likely due to the improved pharmacokinetic properties of nal-IRI enhancing exposure to SN-38 in the tumor combined with the high sensitivity to SN-38 associated with high SLFN11 expression.

We observed significantly higher expression of SLFN11 mRNA in primary tumors from EFT relative to RMS and NB, which is consistent with our cell line data and with the xenograft response data, suggesting that clinically EFT may be particularly sensitive to nal-IRI. EWS/FLI is thought to be an aberrant transcription factor that is specifically expressed in EFT (22). FLI1 is normally not expressed in EFT or other childhood tumors, other than as a partner of EWS (which is ubiquitously expressed in most tissues) to form the EWS–FLI1 chimeric gene found in 85% to 95% of Ewing's sarcomas. Thus, we examined the correlation between expression of FLI1 (which on Affymetrix microarrays provides a surrogate for EWS/FLI expression) and SLFN11. We observed a strong correlation between SLFN11 and FLI1 mRNA expression, and SLFN11 may be a direct target of EWSR1–FLI1 as evidenced by ChipSeq analysis data from Kauer and colleagues (45). The relationship of SLFN11 expression to sensitivity of tumors to nal-IRI will ultimately need to be addressed in clinical trials, and the mechanism by which SLFN11 expression confers sensitivity to SN-38 will need to be explored in further laboratory studies.

Clinical trials of nal-IRI in patients with metastatic colorectal cancer and pancreatic cancer are ongoing (ref. 46; clinicaltrials.gov ID: NCT01494506, NCT01375816), with positive results being recently reported for nal-IRI in a pivotal phase III study of pancreatic cancer (19) Based on published data and data from the present study, evaluations of SLFN11 as a potential biomarker for nal-IRI activity are planned. Our study is the first to show that nal-IRI has substantially improved antitumor activity over irinotecan in xenograft models of pediatric cancers. The striking antitumor activity in EFT xenografts seen with nal-IRI was likely due to the combination of the advantageous pharmacokinetic properties of nal-IRI combined with the high sensitivity to SN-38 of EFT, which was significantly correlated with SLFN11 expression. Based on the preclinical studies reported here, a phase I study of nal-IRI combined with cyclophosphamide in pediatric solid tumors (SPOC 2012-001;NCT 02013336) is under way in the South Plains Oncology Consortium (www.SPONC.org).

J.B. Fitzgerald has ownership interest (including patents) in Merrimack Pharmaceuticals. D.C. Drummond is employed by and has ownership interest (including patents) in Merrimack Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.H. Kang, J.B. Fitzgerald, D.C. Drummond, C.P. Reynolds

Development of methodology: M.H. Kang, J. Wang, M.R. Makena, N. Paz, C.P. Reynolds

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.H. Kang, J. Wang, M.R. Makena, J.-S. Lee, N. Paz, C.P. Hall, M.M. Song, R.I. Calderon, R.E. Cruz, A. Hindle, W. Ko, C.P. Reynolds

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.H. Kang, J. Wang, M.R. Makena, C.P. Hall, W. Ko, J.B. Fitzgerald, T.J. Triche, C.P. Reynolds

Writing, review, and/or revision of the manuscript: M.H. Kang, J. Wang, C.P. Hall, J.B. Fitzgerald, D.C. Drummond, C.P. Reynolds

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.H. Kang, M.R. Makena, R.E. Cruz

Study supervision: M.H. Kang, D.C. Drummond, C.P. Reynolds

The authors thank Annie Chung, Srirupa Cheerla, Dr. Rupeng Zhou, and Dr. Yongku Ryu for their technical support.

This work was supported by Rett's Initiative (www.EFTlab.org) of the Nearburg Foundation via CureSearch National Childhood Cancer Foundation, The Chris Carrey Foundation, and by National Cancer Institute grant CA82830.

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.