Quantitative MRI Identifies Heterogeneous Bone Marrow Treatment Responses in a Mouse Model of Myelofibrosis

Abstract

Myeloproliferative neoplasms (MPNs), a class of blood cancers, typically arise from mutations in the thrombopoietin receptor (MPL), JAK2, or calreticulin that constitutively activate JAK/STAT signaling in hematopoietic stem and progenitor cells. Myelofibrosis (MF), the most aggressive of the MPNs, can arise as a primary neoplasm or secondary to other MPNs. MF is characterized by progressive fibrosis of bone marrow (BM), destruction of the hematopoietic niche, and eventual BM failure. Recently, the focus of treatment in MF has shifted toward restoration of healthy BM architecture and a functional hematopoietic niche rather than focusing solely on alleviating symptoms. Accomplishing this paradigm shift in drug development and therapy requires new approaches to monitor spatial and temporal changes in BM in patients during treatment. We present a quantitative magnetic resonance imaging (MRI) approach to analyze key pathologic changes in BM in a mouse model of MPN driven by MPLW515L. To evaluate disease extent, progression, and response to therapy in mice, we use anatomical MRI for spleen volume and three quantitative MRI metrics for BM: 1) apparent diffusion coefficient (ADC), 2) proton density fat fraction (PDFF), and 3) magnetization transfer ratio (MTR). PDFF and ADC assess the fat content and water movement of the BM and are useful in identifying cellularity changes such as hypercellularity in MPNs and MF. MTR probes changes in macromolecular structure, such as those observed with increasing fibrosis in MF. We used quantitative MRI to measure response to treatment in BM and change in spleen volume in response to three drugs: ruxolitinib (JAK1/2 inhibitor), fedratinib (JAK2 inhibitor), and navitoclax (BCL-2/BCL-xL inhibitor). Within 10 days after starting treatment, diseased spleen volumes in the ruxolitinib group (~3.5x healthy volume) had the greatest average reduction in spleen volume measured by MRI (-73.5% ± 11.4%) followed by the fedratinib group (-52.7% ± 16.7%). Interestingly, the spleens of every ruxolitinib-treated mouse displayed consistent and prolonged spleen volume reductions, while only a subset of mice in the other treatment groups (~70% fedratinib mice; ~33% navitoclax mice) demonstrated reductions in spleen volume. Despite the robust spleen response in ruxolitinib-treated mice, only some of the mice (~60%) exhibited sustained changes in BM ADC, MTR, and PDFF signals expected with reversion to healthy BM (decreased ADC, decreased MTR, and increased PDFF). In mice treated with fedratinib or navitoclax, even fewer mice (~40% and ~33% respectively) showed reversion toward healthy BM. Importantly, we observed spatial heterogeneity of BM response within the tibia of single mice, where proximal and distal tibia regions often responded disparately (~49% of cases). Because improving BM cellularity and fibrosis is increasingly the target of therapy in MPNs and in MF specifically, there is a distinct need for better techniques for longitudinally evaluating variations in treatment efficacy in living subjects. In this study, we demonstrate that the combination of three quantitative MRI metrics (ADC, PDFF, and MTR) can detect the impact of therapy on the BM in a preclinical mouse model of MPN/MF. Interestingly, we identified pronounced heterogeneity at multiple levels during treatment: 1) differences in changes in spleen volume versus BM; 2) disparate treatment responses of mice within a treatment group; and 3) contrasting regional treatment responses of the BM within a single mouse. Because these quantitative MRI metrics are directly translatable to clinical medicine and can identify variations in treatment efficacy in BM, both between and within subjects, this study sets the stage for future drug development and analysis of therapeutic interventions on the BM of patients with MF.