Regulation of leaf condensed tannins (CT) and salicylate-derived phenolic glycosides (PG) in fast- and slow-growing cottonwood backcrosses was analyzed by metabolic profiling and cDNA microarray hybridization. Seven hybrid lines of Populus fremontii L. and P. angustifolia James exhibiting growth/CT-PG phenotypes ranging from fast/low (Lines 18 and 1979) to slow/high (Lines 1012 and RL2) and intermediate (Lines NUL, 3200 and RM5) were investigated. Methanol-extractable leaf metabolites were analyzed by gas chromatography-mass spectrometry, and the results evaluated by principal component analysis. The hybrid lines formed separate clusters based on their primary metabolite profiles, with cluster arrangement also reflecting differences in CT-PG phenotype. Nitrogen (N) supply was manipulated to alter CT-PG partitioning and to obtain molecular insights into how primary metabolism interfaces with CT-PG accumulation. Three backcross lines (RM5, 1012, 18) exhibiting differential CT-PG responses to a 10-day hydroponic N-deprivation treatment were chosen for metabolite and gene expression analyses. The fast- growing Line 18 showed a minimal CT-PG response to N deprivation, and a reduction in photosynthetic gene expression. Line 1012 exhibited a strong phenylpropanoid response to N deprivation, including a doubling in phenylalanine ammonia-lyase (PAL) gene expression, and a shift from CT accumulation in the absence of stress toward PG accumulation under N-deprivation conditions. Amino acid concentrations were depressed in Lines 18 and 1012, as was expression of nitrate-sensitive genes coding for transketolase (TK), and malate dehydrogenase (MDH). Genes associated with protein synthesis and fate were down-regulated in Line 1012 but not in Line 18. Line RM5 exhibited a comparatively large increase in CT in response to N deprivation, but did not sustain decreases in amino acid concentrations, or changes in PAL, TK or MDH gene expression. Molecular characterization of the variable CT-PG responses shows promise for the identification and future testing of candidate genes for CT-PG trait selection or manipulation.