Distinct bone metabolic networks identified in Phospho 1-/- mice vs. wild type mice using [18F]FDG total-body PET

Abstract

Introduction: Total-body PET is a recent development in clinical imaging that produces large datasets involving multiple tissues, enabling the use of new analytical methods for multi-organ assessments, such as network analysis-a well-developed method in neuroimaging. The skeletal system provides a good model for applying network analysis to total-body PET, as bone serves many classical whole-body functions as well as being an endocrine regulator of metabolism. Previous reports have suggested an association between the expression of bone-specific phosphatase, orphan 1 and disorders of altered energy metabolism such as obesity and diabetes. Here, we explore how lacking phosphatase, orphan 1 affects the skeletal metabolic networks of mice as a test approach for deploying network analysis in total-body PET.

Methods: We retrospectively analysed [¹⁸F]fluorodeoxyglucose total-body PET/CT images from six 13-week-old wild type mice, three 22-week-old wild type mice, and three 22-week-old Phospho1 ^-/- mice. Pearson correlation networks were created using the dynamic data from seven bone regions, with a Pearson threshold of r>0.6 (significant at p < 0.005).

Results: The bone metabolic networks of 13-week-old wild type mice were found to robustly resist changes to the data from different PET measurements, increased noise, and shortened scan length. Key features were repeatedly observed, namely that all bones except the spine are highly inter-correlated, while the spine has minimal correlation to other bones. When network analysis was used to compare the three cohorts, the older wild type network had similar features to the young mouse, whereas the Phospho1 ^-/- network had increased correlations across all bones. An all-cohort network separated the data into one part including only bones from the wild type mice (13 nodes) and one part only bones from the Phospho1 ^-/- mice (8 nodes, 95% separation purity). Within the wild type section, the same bone from each young and old mouse were correlated.

Discussion: We demonstrated network analysis is a promising method for studying whole-body PET, sensitive to dynamic details in the data without relying on assumptions or modelling. The proposed method could be applied to other total-body PET data-of healthy and diseased subjects, with different radiotracers, and more-to further elucidate tissue interactions at a systems level.

Keywords: PHOSPHO1; bone; network analysis; positron emission tomography; systems biology.

Figure 1

Traditional static SUV analysis yields no statistically significant differences between cohorts. (A) A maximum intensity projection of total-body PET data (120 min) for a 13-week-old wild type mouse. (B) Group SUV curves of the whole spine region for each of the trial groups: Phospho1^−/− (blue), old WT (red), and young WT (green). (C) Group SUV curves of the whole humerus region for each of the three trial groups. (D) Comparison of the mean SUV at equilibrium across the three trial groups with two-way ANOVA returned no statistically significant differences in any of the seven bones. Data are presented as mean ± SEM, n = 3 for Phospho1^−/−, n = 3 for old WT, and n = 6 for young WT.

Figure 2

(A) Whole-bone dynamic SUV network for a single mouse, including all 120 min of scan time. (B) Whole-bone dynamic SUV network for a single mouse, including only data from the first 60 min of scan time. (C) Cuboid region dynamic kBq/mL network for a single mouse (60 min). (D) Cuboid region averaged dynamic kBq/mL network of six 13-week WT mice (60 min). (E) Whole-bone averaged dynamic SUV network of six 13-week WT mice (60 min).

Figure 3

Network analysis is sensitive to the differing metabolic behaviours in the three trial groups. (A) Whole-bone averaged dynamic SUV network for the three old WT mice. (B) Whole-bone averaged dynamic SUV network for the three Phospho1^−/− mice. (C) Whole-bone averaged dynamic SUV network, including all three cohorts: Phospho1^−/− (orange), old WT (light blue), and young WT (white).