Abstract
Bone marrow adipocytes (BMAds) are not just passive fillers inside the bone marrow compartment but respond to various metabolic changes. Assessment of those responses requires evaluation of the number of BMAds and their morphology for which laborious and error-prone manual histological analysis remains the most widely used method. Here, we report an alternative image analysis strategy to semi-automatically quantitate and analyse the morphology of BMAds in histological bone sections. Decalcified, formalin-fixed paraffin-embedded histological sections of long bones of Sprague-Dawley rats were stained with either haematoxylin and eosin (HE) or by immunofluorescent staining for adipocyte-specific protein perilipin-1 (PLIN1). ImageJ-based commands were constructed to detect BMAds sized 200 µm2 or larger from standardized 1 mm2 analysis regions by either classifying the background colour (HE) or the positive and circular PLIN1 fluorescent signal. Semi-automated quantitation strongly correlated with independent, single-blinded manual counts regardless of the staining method (HE-based: r=0.85, p<0.001; PLIN1 based: r=0.95, p<0.001). The detection error was higher in HE-stained sections than in PLIN1-stained sections (14% versus 5%, respectively; p<0.001), which was due to false-positive detections of unstained adipocyte-like circular structures. In our dataset, the total adiposity area from standardised ROIs in PLIN-1-stained sections correlated with that in whole-bone sections (r=0.60, p=0.02).