A panel of eight human cell lines were established as previously described from bone marrow or peripheral blood leukocytes of infants with KMT2A-rearranged ALL (5). Following informed consent from a parent/guardian, primary patient specimens were sourced from infants enrolled on legacy Children’s Oncology Group (COG) trials in Perth, Western Australia. Briefly, cells obtained at the time of diagnosis or relapse were collected in preservative-free heparin and separated on Ficoll-Hypaque (Pharmacia Fine Chemicals AB, Uppsala, Sweden) at 841 × g for 25 min. Washed cells at a concentration of 1 × 10⁶/ml were placed in round bottom 96-well culture plates (Falcon, Corning, NY, USA) seeding 100 ul per well. The cells were cultured in RPMI 1640 medium (Life Technologies, CA, USA) supplemented with L-glutamine (2 mM final concentration) (Sigma-Aldrich, MO, USA), 2-mercaptoethanol (50 µM final concentration) (Sigma-Aldrich, MO, USA), sodium pyruvate (1 mM) (MP Biomedicals, CA, USA), 1% non-essential amino acids (MP Biomedicals, CA, USA), and 20% heat-inactivated fetal calf serum (FCS) (Flow Laboratories, Irvine, Scotland). The cells were cultured in a humidified 37°C incubator containing 5% CO₂ in air. Fifty percent of the medium was replaced by fresh medium every third day. Cultures were passaged provided they had reached cell concentrations of 2 × 10⁶/ml. Cells were maintained in RPMI 1640 containing 20 or 10% FCS, 1% non-essential amino acid mix (100×), and 1% 100 mM sodium pyruvate solution. Cell viability was assessed through live cell counting by exclusion of 0.2% Trypan blue (Sigma-Aldrich, MO, USA). For all experiments, cells were seeded at pre-determined densities and incubated at 37°C and 5% CO₂.

The cell lines are presently growing in 24-well cell culture plates (Nunclon, Thermo Fisher Scientific, MA, USA) and are passaged twice a week. The cell lines are repeatedly shown to be free of mycoplasma using the MycoAlert™ PLUS Mycoplasma Detection Kit (Lonza, Basel, Switzerland). DNA Fingerprinting was performed on the primary patient samples and on an annual basis for cell line authentication by the Australian Genome Research Facility (Melbourne, Australia), using the GenePrint 10 kit (Promega, WI, USA).

The AIEOP-BFM consensus antibody panel for pediatric ALL was used to assess the immunophenotype of the cell lines by flow cytometry (6). The cells were stained with manufacturer recommended quantities of fluorochrome-conjugated antigen-specific monoclonal antibodies in 5 ml polystyrene round bottom tubes (Falcon, Corning, NY, USA). The cells were incubated with the antibodies for 20 min at room temperature shielded from light. Phosphate-buffered saline (PBS) containing 2% FCS was added to the tubes and then centrifuged at 473 × g for 5 min. The supernatant was discarded and the cells were resuspended in 300 µl of PBS-2% FCS and used for flow cytometry analysis. For intracytoplasmic antigens, cells were fixed and permeabilized using the Fixation/Permeabilization Solution Kit (BD Cytofix/Cytoperm, BD Biosciences, CA, USA) as per manufacturer instructions. A comprehensive antibody panel with 6–10 color direct immunofluorescent labeling was used, according to manufacturer recommended volumes, for immunophenotyping with a BD LSRFortessa™ X-20 flow cytometer (BD Biosciences, CA, USA) ( Supplementary Table 1 ). Cell populations of interest were gated on the basis of light scattering properties. Where appropriate, unstained cells, isotype controls, or cells known to lack certain cell surface markers were used as appropriate negative controls to set quadrant gates. Conversely, cells known to express particular markers (e.g. Jurkat E6-1) were used as positive controls where required. The percentage of antigen-positive cells was decided using a combination of quadrant statistics and ranged histogram gates.

The cell lines were cultured using unsynchronized cell cultures and G-band conventional analysis was performed on metaphase chromosomes according to standard methods. Fluorescent in situ hybridization (FISH) testing using a commercially available KMT2A (MLL) break apart probe to screen interphase cells was carried out according to the manufacturer’s instructions (MetaSystems, Altlussheim, Germany). Directed metaphase and sequential FISH studies using whole chromosome paints and other FISH probes were performed to further characterize complex rearrangements. Chromosomal abnormalities were defined and reported according to the International System for Human Cytogenetic Nomenclature (7).

In vitro cytotoxicity assays were performed using a modified Alamar Blue (resazurin reduction) assay with cells in logarithmic growth. All in vitro assays were performed in triplicate in 384-well clear tissue-culture treated plates (Corning, NY, USA). Cells were seeded at pre-determined optimal seeding densities ( Supplementary Table 2 ) in 50 µl media using an automated Multidrop Combi Reagent Dispenser (Thermo Fisher Scientific, MA, USA). Cells were treated with drugs dispensed by the automated Tecan D300e Digital Dispenser (Tecan Group Ltd., Männedorf, Switzerland) at concentration ranges listed in Supplementary Table 2 at fixed ratios. Drugs were dissolved in dimethyl sulfoxide (DMSO) or water for those with poor DMSO solubility or stability ( Supplementary Table 3 ). Final DMSO concentration in assay wells receiving drugs dissolved in DMSO was 0.5%. After 72 h drug exposure at 37°C and 5% CO₂, Alamar Blue reagent comprising resazurin, methylene blue, potassium hexacyanoferrate (III), and potassium hexacyanoferrate (II) (Sigma-Aldrich, MO, USA) was added (10% v/v) and cell viability determined by absorbance (570 and 600 nm) 6 h after addition of Alamar Blue using a Synergy™ Mx microplate reader (Biotek, VT, USA). Raw absorbance readings were used to generate dose-response curves. Half maximum inhibitory concentrations (IC₅₀) for each compound were calculated in Microsoft Excel by plotting a log dose-response curve utilizing the slope from the linear trendline to determine the IC₅₀. Cell viability was reported as a percent of the untreated control cells for each cell line.

For the synergy assays, we used a 14 × 7 pairwise checker-board matrix format that assessed each of the nine conventional chemotherapy agents used to treat infants with ALL and carfilzomib, at 13 and 6 concentrations respectively, using a series of twofold dilutions. The fraction of cells affected was calculated for each drug concentration and 72 h percent viability imputed from the growth curves. Excess over Bliss as a measurement for synergy was calculated by the Bliss Independence model using the Chalice Bioinformatics Software (Horizon Discovery, Waterbeach, United Kingdom) (8, 9).

Female NOD/SCID mice were purchased from the Animal Research Centre, Perth. Animals were housed under pathogen-free conditions and all studies were approved by the Animal Ethics Committee, Telethon Kids Institute, Perth (Ethics number #312). Seven-week-old mice were inoculated with 1 × 10⁶ PER-785, 1 × 10⁶ PER-826, 1 × 10⁶ MLL-5, or 2 × 10⁶ MLL-14 cells. PER-785 and PER-826 xenografts were established from our cell line panel, whereas MLL-5 and MLL-14 are well characterized patient-derived xenograft models (10–12). Carfilzomib (Selleckchem, Houston, USA) was formulated in an aqueous solution of sulfobutylether-beta-cyclodextrin (Captisol^®, CyDex Pharmaceuticals, CA, USA) or 2-hydroxypropyl-beta-cyclodextrin (Sigma-Aldrich, MO, USA). The maximum tolerated dose (MTD) of carfilzomib was determined in the PER-785 xenograft. Drug treatment was commenced when the percentage of human CD19⁺ CD45⁺ cells reached 1% in the bone marrow, identified from extensive mapping of leukemia cell kinetics (10). Following a 12-day engraftment period, mice were randomized into seven groups of five mice. The delivery schedule was designed to mimic that of a recently completed phase 1b study of carfilzomib in combination with induction therapy for children with relapsed/refractory ALL (13). Carfilzomib was administered intravenously via tail vein injection on days 1, 2, 8, 9, 15, and 16 with treatment groups comprising of 1.5, 2.0, 2.5, 3.0, 3.5, or 4.0 mg/kg dosing of carfilzomib or vehicle control. Mice were sacrificed 3 weeks from the start of therapy and leukemia burden was determined by measuring the percentage of human CD19⁺ CD45⁺ cells in the bone marrow, spleen, and peripheral blood by flow cytometry with anti-human CD19-APC and CD45-PE antibodies.

All four xenograft models were used to determine the response to drug treatment by EFS and leukemia burden in lymphoid organs post-treatment. The PER-785 and PER-826 xenograft models were used to evaluate the clinical potential of carfilzomib in the setting of minimal disease. For each xenograft model, mice were randomized into two groups of eight mice, comprising of a carfilzomib treatment group and vehicle control. Carfilzomib monotherapy was administered at the established MTD, commencing on day 12 following injection of cells for the PER-785 xenograft and on day 18 for the PER-826 xenograft, corresponding to 1% human CD19⁺ CD45⁺ cells in the bone marrow. MLL-5 and MLL-14 xenografts were used to evaluate the effect of carfilzomib at high leukemic burden and in combination with conventional chemotherapy. For each xenograft model, treatment commenced on day 17 following injection of cells for MLL-5 and on day 21 for MLL-14, corresponding to 1% human CD19⁺ CD45⁺ cells in the peripheral blood. Individual mouse EFS was calculated as the time in days from treatment initiation until mice reached a humane end point with evidence of leukemia-related morbidity.

For in vitro assessment of proteasome activity, the eight cell lines were seeded in 24-well plates and incubated with carfilzomib at their respective IC₅₀ concentrations for 72 h. Untreated cell lines incubated under the same conditions were used as a control. Cells were subsequently harvested and washed and proteasome activity was measured using the Abcam Proteasome Activity Assay Kit (Abcam, Cambridge, United Kingdom) according to manufacturer’s instructions.

For in vivo assessment of proteasome activity, seven-week-old mice were inoculated with 1 × 10⁶ MLL-5 cells. Drug treatment was commenced when the percentage of human CD19⁺ CD45⁺ cells reached 1% in the peripheral blood. Mice were randomized into two groups of five mice, comprising of a carfilzomib treatment group and vehicle control. A 1.5 mg/kg dose of carfilzomib was administered intravenously via tail vein injection on days 1 and 2. Blood was collected from all mice 10 min after the second dose was administered to the carfilzomib treatment group and red cell lysis performed using BD Pharm Lyse™ lysing buffer (BD Biosciences, CA, USA). Proteasome activity was measured using the Abcam Proteasome Activity Assay Kit (Abcam, Cambridge, United Kingdom) according to manufacturer’s instructions.

Statistical analyses and graphics were performed using Prism 8 (GraphPad, La Jolla, CA, USA). Data were analyzed using the two-tailed unpaired Student’s t-test. Survival studies were analyzed using log-rank test. A p value <0.05 was considered statistically significant.

Eight cell lines were established from six infants with KMT2A-rearranged ALL ( Table 1 ). Four of the infants were less than 90 days old at diagnosis, with one infant diagnosed with congenital leukemia on day 9 of life. A cell line was established at both diagnosis (PER-494) and relapse (PER-485) from one infant (P272). Two cell lines were derived at diagnosis (PER-784 and PER-826) from P337 and the MLL-14 patient-derived xenograft was also established from this infant at diagnosis (14). Immunophenotyping with the AIEOP-BFM consensus antibody panel revealed B-lineage for PER-494, PER-784, PER-826, PER-785, and PER-910, corresponding with the immunophenotypic classification from the original patient ( Table 2 ). The dominant lineage assignment for PER-490 was B-cell precursor, however there was co-expression of myeloid markers in keeping with the primary patient, P287, from which it was derived. Co-expression of B and myeloid markers was also seen at relapse for patient P272 and at diagnosis for patient P810. Subsequent to in vitro selection pressure, the corresponding cell lines, PER-485 and PER-703, exhibit a myeloid dominant immunophenotype ( Table 2 ). Cytogenetic analysis identified involvement of the same KMT2A-translocation partner between each cell line and corresponding primary patient sample ( Table 3 ). Sideline clones were identified in several cell lines, with structural and numerical changes that were undetected in patient specimens. This could be explained by less advanced techniques for routine standard of care cytogenetic analysis when the infants were originally diagnosed or may reflect clonal evolution in vitro. The former can be exemplified by the initial diagnostic karyotype of P337 revealing 49,XX,+X,+X,+6[5]/46,XX[16] in 1994, with the complex three-way translocation specified in Table 3 identified in the MLL-14 xenograft and on re-analysis of the corresponding primary patient material using more advanced techniques in 2008 (14). The minor cytogenetic clone identified in the patient was detected in the corresponding cell lines, PER-784 and PER-826. DNA fingerprinting revealed identical short tandem repeat profiles between the primary patient sample and corresponding cell line, except for single markers at the CSF1PO locus for PER-784 and PER-785 and at the TH01 locus for PER-703 ( Supplementary Table 4 ). The high match percentage confirmed authentication of the cell lines.

Carfilzomib demonstrated low IC₅₀ concentrations within the nanomolar range across all cell lines (PER-494: 6.0 nM; PER-485: 15.8 nM; PER-490: 9.1 nM; PER-784: 7.3 nM; PER-826: 10.6 nM; PER-785: 14.8 nM; PER-703: 12.9 nM; PER-910: 12.8 nM) ( Figure 1 ). Carfilzomib was shown to inhibit proteasome activity at the IC₅₀ concentration in all cell lines compared to untreated control ( Figure 2A ). Combination drug testing indicated in vitro synergy between carfilzomib and several conventional chemotherapeutic agents including vincristine, daunorubicin, dexamethasone, L-asparaginase, and 4-hydroperoxycyclophosphamide, with six or more cell lines demonstrating positive sum of synergy scores according to Excess over Bliss ( Table 4 ). Notably, the highest total synergy scores in the two cell lines with a myeloid dominant immunophenotype (PER-485 and PER-703) occurred in combination with cytarabine ( Table 4 ). As a first step towards assessing the potential of such drug combinations, we examined carfilzomib as single agent in vivo.

Leukemia-bearing mice receiving 2.5 mg/kg or more of carfilzomib developed weight loss, diarrhea, shortness of breath, slower movement, and swelling at the injection site. Mice that received 4.0 mg/kg were euthanized due to excessive morbidity prior to completion of therapy, thus establishing the MTD of carfilzomib as 2.0 mg/kg. There was no difference in leukemia burden in the bone marrow, spleen, or peripheral blood between any of the doses tested or vehicle control at the defined endpoint ( Figure 3 ). This finding was extended for assessment of EFS, when carfilzomib monotherapy was used to treat minimal disease burden in both the PER-785 and PER-826 xenograft models. Compared to vehicle control, no significant survival benefit was seen for carfilzomib at the MTD of 2.0 mg/kg ( Figures 4A, B ).

Results from the in vitro synergy screen identified positive Bliss scores in the majority of cell lines for carfilzomib and chemotherapeutic agents used during induction therapy for children with ALL ( Table 4 ). Given this finding, we sought to determine whether carfilzomib could potentiate the effect of multi-agent induction therapy. This was performed using the MLL-5 and MLL-14 xenografts, as they have been extensively characterized for their response to vincristine, dexamethasone, and L-asparaginase (VXL) (12). As induction therapy is administered to children with high leukemic burden and to additionally determine whether carfilzomib could provide an in vivo survival benefit at high leukemic burden, treatment was commenced when the percentage of human CD19⁺ CD45⁺ cells reached 1% in the blood. For each xenograft model, mice were randomized into four groups of 10 mice, with groups comprising of carfilzomib monotherapy, VXL, carfilzomib-VXL combination therapy, or vehicle control. VXL treatment comprised of 0.15 mg/kg of vincristine (Pfizer, NY, USA) once a week for 4 weeks, dexamethasone (Mylan, PA, USA) 5 mg/kg 5 days a week for 4 weeks, and L-asparaginase (Sanofi, Paris, France) 1,000 U/kg 5 days a week for 4 weeks administered by intraperitoneal injection (15). To minimize toxicity when combined with VXL therapy, carfilzomib was administered at one dose level below the established MTD, namely 1.5 mg/kg, with in vivo inhibition of proteasome activity demonstrated at this dose level compared to untreated controls ( Figure 2B ). However, despite the on-target effect, no survival benefit was seen for carfilzomib, either as a single agent when compared to vehicle control, or in combination with VXL therapy compared to VXL alone in MLL-5 or MLL-14 xenografts treated with high leukemic burden ( Figures 4C, D ).