Sustained, complete response to pexidartinib in a patient with CSF1R-mutated Erdheim-Chester disease

Abstract

Erdheim-Chester disease (ECD) is a histiocytic neoplasm that predominantly harbors mitogen-activated protein kinase (MAPK) pathway variants. MAPK inhibitors typically are effective treatments, but mutations outside the MAPK pathway, such as CSF1R variants, may cause refractory ECD. We describe a patient with a novel somatic mutation in CSF1R (CSF1R^{R549_E554delinsQ} ) that resulted in refractory ECD affecting the central nervous system. Cell model studies, RNA sequencing analysis, and in silico protein modeling suggested that she had a gain-of-function mutation occurring in a region critical for autoinhibition. The patient was treated with pexidartinib, a CSF1R inhibitor, and has had a complete clinical and metabolic response lasting more than 1.5 years to date. To our knowledge, this is the first report to describe successful treatment of a patient with ECD by using an agent that specifically targets CSF1R. This case also highlights the critical role of individualized molecular profiling to identify novel therapeutic targets in ECD.

Figure 1:. Location, validation, and structural consequences of the juxtamembrane CSF1R^{R549_E554delinsQ} variant.

a, Structure of the CSF1R protein. The sequencing trace shows a mixed allelic sequence, indicating a heterozygous alteration in the primary tumor tissue (sample obtained before treatment). Subcloning of this material identified a 15-nucleotide deletion (light pink) that resulted in an in-frame, 6−amino acid deletion (dark pink) and insertion of a glutamine (cyan blue). The germline sequence from peripheral blood (sample obtained after treatment) showed only the wild-type CSF1R genotype, thus confirming the somatic status of the CSF1R^{R549_E554delinsQ} variant. b, CSF1R wild-type structure, including the N-terminus (aa 453–543). The model was created with the I-Tasser server and based on Protein Data Bank 4r7h. The pexidartinib molecule was stabilized by the residues indicated, particularly Trp550. In the patient’s variant structure (Arg549_Glu554>Gln), the juxtamembrane region (orange) and adjacent portion (green) were deleted. The activation loop is depicted in magenta. c, CSF1R wild-type (yellow, orange, green) and variant Arg549_Glu554>Gln (blue) structures are superimposed. The deleted region of the variant structure was modeled with Coot software (after applying all regulation and refinement restraints, as described in the Methods). The activation loop is depicted in magenta. Structural differences were clearly visible between the wild-type and variant proteins, with the variant showing a shortened and differently positioned JM region. d, Conformational changes between wild-type and variant protein structures. Although the pexidartinib-binding location was unchanged, the Tyr546 interaction with Asp633 was altered by moving the loop and forming an interaction with Asp796 (variant Asp732). In the CSF1R^{R549_E554delinsQ} protein structure, the Trp550 residue was lost, thereby eliminating a hydrophobic interaction with pexidartinib; however, a new hydrophobic interaction with pexidartinib formed with a different rotamer of Phe733 (wild-type Phe797) and Tyr546 of the variant structure. Additionally, a minimal conformation change of the autoinhibitory loop (wild-type, magenta; variant, blue) that suggested constitutive activation. e, Main residues affected in the variant structure. Residues Cys666, Glu664, and Glu633 showed minimal conformational change. Residue Tyr546 moved and formed a hydrophobic interaction with Phe733 (wild-type Phe797, see part d) and with pexidartinib. Asp732 (wild-type Asp796, see part d) moved slightly to facilitate hydrogen bonding with the hydroxyl group of Tyr546. (Abbreviation: aa indicates amino acid; Ig, immunoglobulin; JM, juxtamembrane; PCR, polymerase chain reaction; TKD, tyrosine kinase domain; WT, wild-type).

Figure 2:. Tumor gene expression profiles and pathway analysis.

a, Whole-transcriptome profiling of the tumor and normal human microglial samples (controls) were compared. The heat map shows that genes commonly expressed in microglial cells of origin were expressed in the tumor sample (column indicated by the red arrow). Further, expression of canonical microglial genes did not differ between the tumor and controls. For details regarding microglial genes, please see the supplementary file; Microglial-Genes.xlsx. A small variation (median coefficient of variation, 8.1%) in expression of microglial genes in control samples was observed, which was mainly due to technical and cohort differences among the studies used to derive the microglial gene set^, and the study containing control samples. The patient’s tumor sample showed robust expression of key microglial genes (see the supplementary file Microglial-Genes.xlsx for the full list of genes), and its gene expression pattern closely paralleled that of microglial cells. b, Key genes in NFкB, PI3K-AKT, cell cycle, and MAPK pathways were upregulated in the tumor sample (column indicated by the red arrow). We did not detect significant upregulation of key genes in the Janus kinase−signal transducer and activator of transcription (JAK-STAT) signaling pathway, which is consistent with CSF1R activation^,. c, Graphical representation of the effect of CSF1R on the MAPK and PI3K-AKT pathways. d, The summary of gene set enrichment results shows upregulation of key pathways. We did not detect any downregulated pathways in the tumor sample compared with controls. The PI3K-AKT pathway was relatively more upregulated when compared with the MAPK pathway (red arrows). **** denotes FDR<0.0001, **denotes FDR<0.01, and * denotes FDR<0.05. e, Patient’s tumor tissue stained for p-ERK 1/2 and p-AKT before initiating treatment with pexidartinib (original magnification, ×400). (Abbreviations: FDR, false-discovery rate; JM, juxtamembrane; KEGG, Kyoto Encyclopedia of Genes and Genomes; MET, MET receptor tyrosine kinase; p-, phosphorylated; WT, wild-type.)

Figure 3:. Functional validation of the patient-specific CSF1R^{R549_E554delinsQ} alteration.

a, Schematic depiction of the retroviral constructs of CSF1R-wt and CSF1R-mut. b, OCI-Ly19 cells were infected with the CSF1R-wt or CSF1R-mut provirus, and fluorescence-activated cell sorting was used to select GFP-positive cells (thus selecting cells expressing CSF1R-wt or CSF1R-mut). Histogram in green is the profile of CSF1R-wt expressing cells while the histogram in red is the profile of CSF1R-mut expressing cells. c, Western blot images of total AKT, total ERK, p-AKT, and p-ERK 1/2 proteins from uninfected controls or cells expressing CSF1R-wt or CSF1R-mut. d, Relative quantification of AKT and ERK activation. The graph shows the ratios of densitometric measures of protein bands shown in Figurer 3c (p-AKT, total AKT, p-ERK 1/2, and total ERK). (Abbreviations: c, coding DNA; del, deletion; GFP, green fluorescent protein; IRES, internal ribosomal entry site; LTR, long terminal repeat; MCSV, murine stem cell virus; mut, CSF1R^{R549_E554delinsQ} alteration; p-, phosphorylated; wt, wild-type.)

Figure 4:. Radiographic response to pexidartinib treatment.

Sagittal FDG F 18 positron emission tomography before (a) and 3 months after (b) initiation of pexidartinib. The patient had a complete metabolic response in the bulky FDG-avid disease (7.9 maximum standard uptake value) encasing the lumbar nerve roots (black arrows). Axial T2-weighted brain MRI before (c) and 2 months after (d) initiation of pexidartinib showed near-complete resolution of a T2-isointense nodule involving the left oculomotor nerve (white circles). Coronal, T1-weighted, contrast-enhanced brain MRI before (e) and 2 months after (f) initiation of pexidartinib showed resolution of abnormal enhancement of the leptomeninges (white arrows) and the oculomotor nerve (red arrows). Axial, T2-weighted brain MRI before (g) and 2 months after (h) initiation of pexidartinib showed resolution of bulky nodular lesions involving the lumbar nerve roots (brackets). (Abbreviations: FDG, fluorodeoxyglucose; MRI, magnetic resonance image.)