Kinome Reprogramming of G2/M Kinases and Repression of MYCN Contribute to Superior Efficacy of Lorlatinib in ALK-Driven Neuroblastoma

Abstract

Mutations in the tyrosine kinase domain of the anaplastic lymphoma kinase (ALK) oncogene in neuroblastoma occur most frequently at one of three hotspot amino acid residues, with the F1174* and F1245* variants conferring de novo resistance to first- and second-generation ALK inhibitors, including crizotinib and ceritinib. Lorlatinib, a third-generation ALK/ROS1 inhibitor, overcomes de novo resistance and induces complete and sustained tumor regressions in patient-derived xenograft models unresponsive to crizotinib. Lorlatinib has now completed phase 1 testing in children and adults with relapsed/refractory ALK-driven neuroblastoma and entered pivotal phase 3 testing within the Children's Oncology Group. To define mechanisms underlying the superior activity of lorlatinib, we utilized a chemical proteomics approach to quantitatively measure functional kinome dynamics in response to lorlatinib and crizotinib in clinically relevant ALK-driven neuroblastoma patient-derived xenograft models. Lorlatinib was a markedly more potent inhibitor of ALK and preferentially downregulated several kinases implicated in G2/M cell-cycle transition compared with crizotinib. Lorlatinib treatment also led to the repression of MYCN expression and its occupancy at promoters of the same G2/M kinases. These data provide mechanistic insight into the superior efficacy of lorlatinib over crizotinib for the treatment of ALK-driven neuroblastoma.

Figure 1.

Lorlatinib induces significantly more kinome reprogramming compared with crizotinib in ALK-driven PDXs and CDXs. A and B, Volcano plots of aggregate MIB/MS data for COG-N-453x, FELIX-PDX, NB-1, and NB-1643 xenografts combined. Data for kinases from both mouse (gray) and human (A, blue; B, red) species were plotted following 6 days of treatment with 100 mg/kg crizotinib (A) or 10 mg/kg lorlatinib (B) daily (n = 3). The horizontal dotted line represents a P value threshold of 0.05. C, Statistically significant changes in MIB binding of the kinases upon treatment with crizotinib (top) or lorlatinib (bottom) for 6 days in the indicated neuroblastoma models. Blue circles indicate downregulation, and red circles upregulation. Circle size correlates with the magnitude of change up to a 256-fold change. Kinome trees were generated using CORAL (PMID 30172842).

Figure 2.

MIB/MS profiling in neuroblastoma models reveals direct and indirect kinase targets of lorlatinib. A, ALK inhibitor treatment decreased ALK MIB binding in individual neuroblastoma models. B, Quantification of MIB binding by other kinases previously reported as direct targets of lorlatinib and/or crizotinib. Data shown represent the aggregate data from COG-N-453x, FELIX-PDX, NB-1, and NB-1643 tumor models. Error bars indicate the mean ± SD. C-E. COG-N-453x lysates were incubated with 0, 10, 100, or 1000 nmol/L lorlatinib for 30 minutes prior to MIB/MS to identify direct kinase targets (n = 3). Shown is the relative abundance for the indicated kinases compared with vehicle-treated (0 nmol/L) lysates. Error bars indicate the mean ± SEM. C and D, Loss of MIB binding for known targets of lorlatinib from human tumor cells (C) or mouse stroma (D). E, Novel putative lorlatinib targets from human tumor models. LFQ, Label Free Quantification. Statistical significance is indicated by an asterisk. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 derived via unpaired t test; ns, not significant.

Figure 3.

Lorlatinib downregulates G2/M kinases. A, Gene ontology analysis of kinases downregulated by lorlatinib treatment in xenograft tumors shows strong enrichment of mitotic and G2/M-related biological processes. B, Relative abundance of G2/M kinase captured by MIBs. Data shown are aggregate data from COG-N-453x, FELIX-PDX, NB-1, and NB-1643 tumor models normalized to vehicle control. Each point represents a single sample in which the indicated kinase was quantified. C, MIB/MS data from COG-N-453x tumors shows significant downregulation of G2/M kinases in response to loratinib. LFQ intensities were normalized to vehicle control. D, NB-1643 xenograft or (E) COG-N-453x were treated with vehicle, 100 mg/kg crizotinib, or 10 mg/kg lorlatinib for 2 or 6 days (three animals per treatment). RT-PCR was performed in triplicate using probes for G2/M kinases as indicated in the figure. Sustained downregulation of G2/M kinases at the transcript level by lorlatinib was observed in both models. Error bars indicate the mean ± SD. Statistical significance is indicated by an asterisk *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 derived via unpaired t test; ns, not significant.

Figure 4.

Lorlatinib treatment results in the loss of PLK1 kinase. A, NB1643 and (B) COG-N-453 tumor-bearing mice were treated with vehicle, 100 mg/kg crizotinib, or 10 mg/kg lorlatinib for 6 days and immunoblotted to validate MIB/MS data, revealing loss of activating phosphorylation of ALK and total PLK1 in response to lorlatinib treatment. Dose-escalation studies of crizotinib or lorlatinib in (C) Felix, (D) NB-1, and (E) SH-SY5Y neuroblastoma cell lines show loss of activating phosphorylation of ALK and reduction in PLK1 levels. F, ALK knockdown using CRISPRi in SH-SY5Y results in decreased levels of PLK1. Dose–response curves show that the combination of crizotinib and the PLK1 inhibitor BI6727 phenocopies lorlatinib. G, SH-SY5Y and (H) FELIX cell lines were treated with crizotinib with or without BI6727 (IC₅₀ concentration) or lorlatinib for 5 days, and viable cells were assayed using CellTiter-Glo. RLU, Relative Light Units. Plotted are the relative luminescence values normalized to DMSO (crizotinib and lorlatinib) or to BI6727 monotherapy (crizotinib + BI6727). Error bars indicate the mean ± SD.

Figure 5.

Transcriptomic analysis after ALK inhibition in neuroblastoma CDX and PDX models. A, Bar graphs showing differentially expressed kinases involved in G2/M cell-cycle transition downregulated at least 2-fold upon treatment with crizotinib (100 mg/kg) or lorlatinib (10 mg/kg) daily for 2 days in NB-1643 xenograft and COG-N-453x (three animals per treatment) show differential pattern, log₂ FC >1. B, Log₂ FC of kinase mRNA, that showed significant reduction in MIB binding upon treatment with lorlatinib in NB-1643 xenograft and COG-N-453x, do not show significant downregulation at transcript level, indicating that the regulation is posttranscriptional. MYCN mRNA is downregulated upon treatment with crizotinib and lorlatinib in (C) NB-1643 xenograft and (D) COG-N-453x showing ALK mutation–specific effect. Error bars indicate the mean ± SD. *, P < 0.05; **, P < 0.01; ****, P < 0.0001, derived via unpaired t test.

Figure 6.

Lorlatinib induces displacement of MYCN and Pol lI from promoters of several G2/M kinases. A, COG-N-453x was treated for 6 days with DMSO, 100 mg/kg crizotinib, or 10 mg/kg lorlatinib, and Western blot was performed with indicated antibodies. B, Quantitation of Western blot for MYCN normalized to β-actin expression. To examine the effect of ALK inhibitor treatment on MYCN binding to G2/M kinase promoters, COG-N-453x was treated with vehicle, crizotinib, or lorlatinib for 6 days, tumors were collected (n = 3), and ChIP was performed using MYCN and RNA Pol II antibodies. Species-matched IgG was used as a control. C, Relative occupancy of MYCN at E-box was examined by RT-PCR using probes for genes indicated in the figure and (D) for Pol II enrichment at the transcription start site. Error bars indicate the mean ± SD. Statistical significance is indicated by an asterisk **, P < 0.01; *, P < 0.05, derived via unpaired t test; ns, not significant.