A metabonomics approach based on high-resolution magic-angle spinning (HRMAS) ¹H NMR spectroscopy was applied to investigate the metabolite composition in intact hepatic tissues and renal cortical tissues from db/db mice of 8 weeks old, an animal model of type 2 diabetes mellitus. Compared to the control group, the hepatic tissues of diabetic mice have elevated levels of triglyceride and bile acid and declined levels of trimethylamine-N-oxide, phosphocholine, glycerophosphocholine, and choline. The biochemical changes are less obvious in renal cortical tissues of diabetic mice. The WET_CPMG pulse sequence was selected for our metabonomics study after the quality and reproducibility of the spectra obtained from the NOEPR, NOEPR_CPMG, and WET_CPMG pulse sequences were analyzed together with principal component analysis. The influence of line-broadening factor of exponential window function for spectral manipulation on class separation was paid attention to for the first time, and an optimal value was obtained under our experimental conditions. These studies show the efficiency of HRMAS ¹H NMR spectroscopy for tissue metabonomics study in combination with multivariate statistical analysis, which may help to explore the etiological factor of diabetes mellitus from a new perspective.