Labelling with stable isotopes has under-exploited potential for studies of polysaccharide endotransglycosylation in vivo. Ideally, the labelled polysaccharides should have the highest possible buoyant density. Although [13C6]glucose has previously been used as a precursor, it was unclear whether 2H would be efficiently incorporated from [2H]glucose or lost as D2O. Rose (Rosa sp.) cell-suspension cultures efficiently incorporated 13C from D-[13C6,2H7]glucose into wall polysaccharides with negligible dilution from atmospheric 12CO2. Also, approximately 70% of the 2H atoms in D-[13C6,2H7]glucose were retained during polysaccharide biosynthesis. This shows that relatively few cycles of intermediary metabolism leading to the release of D2O occurred before sugar residues were incorporated into wall polysaccharides. In agreement with these observations, isopycnic centrifugation in caesium trifluoroacetate gradients showed that the hydrated buoyant density of xyloglucan synthesised by rose cells growing on [13C6,2H7]glucose and [13C6]glucose was 3.7 and 2.6% higher, respectively, than in isotopically non-labelled cultures. Thus, [13C,2H]glucose-feeding enabled a 42% better resolution of 'heavy' from 'light' xyloglucan than [13C]glucose-feeding.