Three allelic mutants of Arabidopsis thalma which lack mitochondrial serine transhydroxymethylase activity due to a recessive nuclear mutation have been characterized. The mutants were shown to be deficient both in glycine decarboxylation and in the conversion of glycine to serine. Glycine accumulated as an end product of photosynthesis in the mutants, largely at the expense of serine, starch, and sucrose formation. The mutants photorespired CO2 at low rates in the light, but this evolution of photorespiratory CO2 was abolished by provision of exogenous NH3. Exogenous NH3 was required by the mutants for continued synthesis of glycine under photorespiratory conditions. These and related results with wild-type Arabidopsis suggested that glycine decarboxylation is the sole site of photorespiratory CO2 release in wild-type plants but that depletion of the amino donors required for glyoxylate amination may lead to CO2 release from direct decarboxylation of glyoxylate. Photosynthetic CO2 fixation was inhibited in the mutants under atmospheric conditions which promote photorespiration but could be partialiy restored by exogenous NH3. The magnitude of the NH3 stimulation of photosynthesis indicated that the increase was due to the suppression of glyoxylate decarboxylation. The normal growth of the mutants under nonphotorespiratory atmospheric conditions indicates that mitochondrial serine transhydroxymethylase is not required in C3 plants for any function unrelated to photorespiration. (3,12,18,31), by comparing the specific activity of photorespiratory CO2 to that of photorespiratory cycle intermediates (16), and by demonstrating, in the presence of chemical inhibitors, that the glycine to serine conversion is tightly coupled to photorespiratory CO2 evolution (11,14,15). These experiments cannot unequivocally exclude the possibility that there are other sites of photorespiratory CO2 release (4). Inhibitor studies are also plagued by nonspecific effects which preclude an appraisal of the long-term physiological effects of blocking the decarboxylation reaction.It has been suggested (4, 7, 8, 15, 34) that, under certain circumstances, glyoxylate, the immediate precursor of photorespiratory glycine, may undergo decarboxylation to produce CO2 and formate. This reaction has been demonstrated in vitro with isolated peroxisomes (7), but evidence substantiating the occurrence of this reaction in vivo is totally lacking. Here, we provide evidence that, under normal physiological conditions, glycine decarboxylation is the sole site of photorespiratory CO2 release but that, under conditions of severe amino depletion, photorespiratory CO2 may arise from direct decarboxylation of glyoxylate. This conclusion is based upon an analysis of mutants of Arabidopsis that lack mitochondrial serine transhydroxymethylase activity and are, therefore, defective in both glycine decarboxylation and the glycine to serine conversion. Photorespiratory CO2 is the product ofa complex mitochondrial reaction in which glycine is converted to stoi...