Bone Collagen Fibers Fall Into Disarray Within Weeks of Diabetes Onset in Rats

Within just two weeks of type 2 diabetes taking hold, the collagen fiber network inside rat femurs had already lost its orderly alignment, even though the bones’ mineral density and visible microstructure looked essentially unchanged. That is the central finding of a new animal study published in European Radiology Experimental, which tracked skeletal changes from the earliest moments of the disease and linked the collagen disruption to fat accumulating in the liver. The work suggests that in early type 2 diabetes, the organic scaffold of bone may deteriorate before the mineral framework shows measurable damage.

Type 2 diabetes has long been known to raise fracture risk through mechanisms that standard bone scans fail to capture. People with the disease can show normal or even elevated areal bone mineral density on dual-energy x-ray absorptiometry, yet still break bones more easily, a paradox that has pushed researchers toward the broader concept of bone quality: the microarchitecture, turnover rate, mineralization, and above all the integrity of the type I collagen network that gives bone its toughness. In a high-glucose environment, advanced glycation end products form abnormal cross-links between collagen fibers, stiffening them and undermining their capacity to absorb energy and repair themselves.

A team at Hebei Medical University Third Hospital in Shijiazhuang, China, designed a prospective animal experiment to examine exactly that window. Twelve seven-week-old male Sprague-Dawley rats, each weighing roughly 270 grams at purchase, were randomly assigned to either a control group or a diabetes group, six animals per side. After a two-week acclimatization, the diabetes group spent four weeks on a high-fat diet containing 45 percent fat, designed to induce insulin resistance, and then received low-dose intraperitoneal streptozotocin injections over three days. Streptozotocin selectively poisons pancreatic beta cells via the glucose transporter type 2, and in the context of a high-fat diet it pushes the animals into a decompensated hyperglycemic state that recapitulates the two core features of human type 2 diabetes: insulin resistance and impaired insulin secretion. The model was considered established once fasting blood glucose consistently exceeded 7.5 millimoles per liter.

The timing of the assessments was deliberate. Within one week of diabetes confirmation, every rat underwent dual-energy x-ray absorptiometry for whole-body composition and magnetic resonance imaging of the liver and femur on a 3.0-T system using a three-dimensional multi-echo chemical shift-encoded sequence to quantify proton density fat fraction, a precise magnetic resonance measure of fat content in tissue. Blood was drawn for a full biochemical and lipid panel. During the following week, the animals were euthanized and the right femurs were harvested for micro-computed tomography at a voxel size of 16.62 micrometers, three-point bending biomechanical testing at the midshaft, and Masson’s staining to visualize collagen fibers. Tibia and fibula samples were processed for enzyme-linked immunosorbent assays of bone turnover markers and advanced glycation end products, and for quantitative polymerase chain reaction of osteoblast-related genes. All measurements were therefore completed within two weeks of diabetes onset.

The metabolic phenotype emerged exactly as expected. Rats in the diabetes group weighed more than controls from just over ten weeks of age onward, and dual-energy x-ray absorptiometry revealed significantly greater fat mass and body fat percentage. Serum total cholesterol, alanine aminotransferase, and the alanine-to-aspartate aminotransferase ratio were all elevated, and hepatic proton density fat fraction was significantly higher, confirming early fatty liver. Femoral fat fraction also trended upward in the diabetic animals, though that difference did not reach statistical significance. Together the data portray a systemic shift of adipose tissue, from peripheral stores into the liver, occurring hand in hand with rising liver enzymes.

The skeletal results were more surprising. On micro-computed tomography, the diabetic femurs showed lower cancellous bone volume fraction and trabecular number, higher trabecular separation and structure model index, and greater total cortical bone area and cortical porosity, but none of these differences were statistically significant. Biomechanical three-point bending likewise showed no significant differences in ultimate load, elastic stiffness, or total work to failure. Serum calcium, magnesium, and phosphorus were unchanged. In other words, by every conventional measure of bone quantity and mineral structure, the early diabetic bones looked broadly normal.

The collagen network told a different story. Using Masson’s staining and image analysis, the researchers quantified both the area occupied by collagen fibers and a measure the authors call collagen fiber goodness, which reflects fiber orientation, with values closer to 1 indicating highly ordered alignment and lower values indicating disorganization. Fiber area was slightly reduced in the diabetic group, but not significantly. Fiber organization, however, was significantly worse in the diabetic femurs, with a p value of 0.022. This was the only bone-related parameter that differed significantly between the groups, pointing to matrix-level deterioration as the earliest detectable skeletal change in this model. Consistent with an organic-first pattern of damage, the bone formation marker procollagen type I N-terminal propeptide was significantly elevated in the diabetic tibia and fibula, while osteocalcin relative messenger RNA expression was significantly reduced. Carboxy-terminal collagen cross-links, osteopontin expression, and advanced glycation end product levels showed no significant differences, the last of which suggests that measurable accumulation of glycation cross-links in bone had not yet occurred at this early stage.

The most striking connection emerged when the skeletal and metabolic data were combined. Femoral collagen fiber goodness correlated strongly and negatively with hepatic proton density fat fraction, with a Spearman correlation coefficient of −0.709 and a p value of 0.012. The fatter the liver, the more disordered the collagen architecture in the femur. Because bone advanced glycation end products were not elevated, the authors speculate that hepatic fat accumulation may disrupt bone collagen organization through molecular pathways independent of glycation, a mechanism they note requires further investigation. The finding aligns with a growing literature linking ectopic lipid deposition to systemic metabolic injury: a 2025 study in Radiology reported that liver fat fraction is associated with the odds of type 2 diabetes in a dose-dependent manner, partially mediated by liver enzymes, and that severe hepatic steatosis was associated with markedly increased odds of prediabetes and diabetes.

The authors interpret their results as evidence for a temporal and spatial sequence in early diabetic bone disease. Damage appears to begin in the organic matrix, with reduced and disorganized collagen fibers preceding any measurable inorganic or mineral changes, and to affect cancellous bone, with its higher metabolic activity, more prominently than cortical bone. The slight, non-significant increase in cortical porosity and the modest shift in biomechanical parameters, including reduced work to failure, are consistent with microstructural deterioration beginning at the margins, though the authors are careful to note these trends need confirmation. They also frame the dissociation between elevated procollagen type I N-terminal propeptide and reduced osteocalcin expression as a possible compensatory osteoblastic response unique to the earliest phase of the disease, one that diverges from patterns reported in longer-established human diabetes.

The study’s limitations are considerable and the authors state them plainly. With only six rats per group, statistical power is limited and generalizability is constrained. The work did not obtain dual-energy x-ray absorptiometry-derived indices such as trabecular bone score, hip structural analysis, or the bone strain index, which could bridge these preclinical observations to clinical bone quality assessment. Imaging used a single 3.0-T magnetic resonance system, and vitamin D levels were not measured. The two-week observation window captures only the earliest phase, leaving longer-term progression unexamined. Most importantly, this is a rodent model, and findings in streptozotocin-treated rats cannot be assumed to translate directly to human patients. What the study does establish is a testable hypothesis: that in early type 2 diabetes, the collagen scaffold of bone is the first skeletal structure to falter, that its disorganization tracks with fat redistribution into the liver, and that magnetic resonance measures of liver fat might one day serve as a flag for clinicians to look beyond bone mineral density when evaluating skeletal health in metabolic disease.

Subject of Research: Early-stage type 2 diabetes disrupts bone collagen fiber organization in rats in association with adipose tissue redistribution and hepatic fat accumulation

Article Title: Disorganized bone collagen fibers in early-stage type 2 diabetes mellitus and its association with adipose tissue redistribution

Article References: Liu, Y., Meng, Q., Chen, S., Zhang, X., Gao, L., Wang, M., Bai, L., Zhang, X., Zhang, P., Zheng, N., & Zhao, J. (2026). Disorganized bone collagen fibers in early-stage type 2 diabetes mellitus and its association with adipose tissue redistribution. European Radiology Experimental, 10(1), Article 147. https://doi.org/10.1186/s41747-026-00820-0

Image Credits: AI Generated

DOI: 10.1186/s41747-026-00820-0

Keywords: type 2 diabetes, bone collagen, bone quality, hepatic fat fraction, MRI, Sprague-Dawley rats, advanced glycation end products, bone turnover markers, micro-computed tomography, adipose tissue redistribution, osteocalcin, proton density fat fraction