Quantitative MRI reveals heterogeneous impacts of treatment on diseased bone marrow in a mouse model of myelofibrosis

Abstract

Purpose: Analyzing bone marrow in the hematologic cancer myelofibrosis requires endpoint histology in mouse models and bone marrow biopsies in patients. These methods hinder the ability to monitor therapy over time. Preclinical studies typically begin treatment before mice develop myelofibrosis, unlike patients who begin therapy only after onset of disease. Using clinically relevant, quantitative MRI metrics allowed us to evaluate treatment in mice with established myelofibrosis.

Methods: We used chemical shift-encoded fat imaging, DWI, and magnetization transfer sequences to quantify bone marrow fat, cellularity, and macromolecular components in a mouse model of myelofibrosis. We monitored spleen volume, the established imaging marker for treatment, with anatomic MRI. After confirming bone marrow disease by MRI, we randomized mice to treatment with an approved drug (ruxolitinib or fedratinib) or an investigational agent, navitoclax, for 33 days. We measured the effects of therapy over time with bone marrow and spleen MRI.

Results: All treatments produced heterogeneous responses with improvements in bone marrow evident in subsets of individual mice in all treatment groups. Reductions in spleen volume commonly occurred without corresponding improvement in bone marrow. MRI revealed patterns associated with effective and ineffective responses to treatment in bone marrow and identified regional variations in efficacy within a bone.

Conclusions: Quantitative MRI revealed modest, heterogeneous improvements in bone marrow disease when treating mice with established myelofibrosis. These results emphasize the value of bone marrow MRI to assess treatment in preclinical models and the potential to advance clinical trials for patients.

Keywords: MRI; bone marrow; myelofibrosis; treatment response.

Figure 1.. Experimental Setup.

Experimental plan (A) outlining the major imaging steps of the study and expected changes in the corresponding BM MRI metrics. Increases (up arrows) and decreases (down arrows) are considered relative to healthy with the magnitude of the increase or decrease corresponding to the number of arrows. We used pre-treatment (pre-tx) time point spleen volumes to randomize mice into treatment groups (B). Unfilled symbols represent mice that survived seven or fewer days of treatment. Survival of each treatment group (C) provides the number of mice imaged at each time point.

Figure 2.. Spleen volume reduction occurs more quickly and to a greater extent than BM improvement.

Representative spleen MRIs from the pre-treatment time point and after 14 days of treatment (A) with quantification (B) of spleen volumes of mice in each treatment group. BM quantitative MRI trajectories for proximal and distal tibia BM for (C) ADC, (D) PDFF, and (E) MTR. Spleen volume P values calculated against the vehicle control group using the Benjamini, Krieger, and Uekutieli Unpaired t test adjusting for the false discovery rate. * P<0.05. Number of mice for each treatment group given with notation: Days post-BMT (Days post-Tx): N= n vehicle, n ruxolitinib, n fedratinib, n navitoclax. Day 44 post-BMT(pre-tx): N=7,7,7,7; Day 49(3): N=5,7,7,7; Day 51(5): N=5,6,7,6; Day 53(7): N=5,6,7,5; Day 56(10): N=5,5,7,3; Days 58–78(12–32): N=5,5,7,3.

Figure 3.. Treatment improves BM in subset of mice with spleen volume reduction (SVR).

When compared to the vehicle control condition for spleen volume and BM (A), we observed three distinct situations: (B) SVR and BM improvement, (C) SVR without BM improvement, and (D) neither SVR nor BM improvement. BM changes in panels A-D observed in each MRI metric and displayed using representative MTR images. (E and F) Quantification of SVR (E) and BM improvement (F) by treatment group. (G) Two-way table data identify association between SVR and BM improvement by Fisher’s exact test *p=0.0128 (two-sided). Only mice surviving at least 14 days of treatment included in analyses. For spleen volume analysis (E), N = 5 vehicle, 5 ruxolitinib, 6 fedratinib, 3 navitoclax (1 fedratinib excluded due to no spleen volume change from healthy at the pre-treatment time point). For BM improvement analyses (F, G), N = 5 vehicle, 5 ruxolitinib, 5 fedratinib, and 3 navitoclax (one additional mouse excluded from fedratinib group because BM imaging did not confirm BM disease at the pre-treatment time point).

Figure 4.. Treatment impacts BM heterogeneously even for mice in the same treatment group.

Representative MRI images for mice separated into effective (A), moderately effective (B), and ineffective (C) treatment response groups with longitudinal distal quantification for ADC (D), PDFF (E), and MTR (F). Data shown normalized to the pre-treatment time point and presented as the mean change ± s.e.m. Differences among groups at each time point evaluated using Tukey’s multiple comparisons test (* denotes p<0.05 compared to ineffective, † denotes p<0.05 compared to moderately effective, and ‡ denotes p<0.05 compared to vehicle control). Only mice with imaging-confirmed disease at pre-tx time point included in analysis. For days 44–77 post-BMT (pre-tx –days post-tx) N = 5 vehicle, N = 5 ineffective tx, N = 3 moderately effective tx, N = 3 effective tx. For day 77 post-BMT (31 days post-tx) N = 5 vehicle, N = 3 ineffective tx, N = 3 moderately effective tx, N = 2 effective tx. The change in mouse numbers from day 70 to day 77 post-BMT reflects removal of navitoclax treated mice from the study prior to endpoint as reflected in Figure 1C. (G) Count of mice in each treatment condition that had effective, moderately effective, or ineffective BM treatment responses. Only mice surviving 14 days of treatment and demonstrating significant BM disease at the pre-tx time point are included in analyses. Treatment = Tx

Figure 5.. MRI reveals two BM treatment response profiles.

BM treatment responses can be separated into two groups: (A) Regionally heterogeneous responses where the distal and proximal BM have different magnitudes of response and (B) Regionally homogeneous responses where the treatment response occurs similarly throughout the BM. Quantification (C) of the representative examples from panels A and B with proximal-distal heterogeneity (left) or homogeneity (right). Box plot and whiskers show range containing 95% of the data. Pre-Tx is considered Day 0 of treatment. X-axis is not to scale. (D) Counts of mice in each treatment group with heterogeneous or homogeneous BM responses. Analysis for mice at any point of treatment and excluding mice with limited bone marrow disease progression at the pre-tx time point. N = 7 mice for each treatment group at study commencement with Figure 1C denoting number of mice imaged at subsequent time points.