Reconstructing bulk isotope ratios from compound-specific isotope ratios

Abstract

Carbon isotope analysis by bulk elemental analysis coupled with isotope ratio mass spectrometry has been the mainstay of delta(13)C analyses both at natural abundance and in tracer studies. More recently, compound-specific isotope analysis (CSIA) has become established, whereby organic constituents are separated online by gas or liquid chromatography before oxidation and analysis of CO(2) for constituent delta(13)C. Theoretically, there should be concordance between bulk delta(13)C measurements and carbon-weighted delta(13)C measurements of carbon-containing constituents. To test the concordance between the bulk and CSIA, fish oil was chosen because the majority of carbon in fish oil is in the triacylglycerol form and approximately 95% of this carbon is amenable to CSIA in the form of fatty acids. Bulk isotope analysis was carried out on aliquots of oil extracted from 55 fish samples and delta(13)C values were obtained. Free fatty acids (FFAs) were produced from the oil samples by saponification and derivatised to fatty acid methyl esters (FAMEs) for CSIA by gas chromatography/combustion/isotope ratio mass spectrometry. A known amount of an internal standard (C15:0 FAME) was added to allow analyte quantitation. This internal standard was also isotopically calibrated in both its FFA (delta(13)C = -34.30 per thousand) and FAME (delta(13)C = -34.94 per thousand) form. This allowed reporting of FFA delta(13)C from measured FAME delta(13)C values. The bulk delta(13)C was reconstructed from CSIA data based on each FFA delta(13)C and the relative amount of CO(2) produced by each analyte. The measured bulk mean delta(13)C (SD) was -23.75 per thousand (1.57 per thousand) compared with the reconstructed bulk mean delta(13)C of -23.76 (1.44 per thousand) from CSIA and was not significantly different. Further analysis of the data by the Bland-Altman method did not show particular bias in the data relative to the magnitude of the measurement. Good agreement between the methods was observed with the mean difference between methods (range) of 0.01 per thousand (-1.50 to 1.30).