Global CO2 flux from leaf photorespiration and day respiration probably represents ≈ 30 Gt of carbon liberated each year in the atmosphere. The cellular mechanisms of CO2 production are thus of paramount importance for our understanding of plant carbon balance and partitioning. However, knowledge of the metabolic control of photorespiration and day respiration is still rather limited. The photorespiratory CO2 flux (Φ) is commonly assumed to be primarily defined by the CO2/O2 specificity of ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) and CO2 and O2 mole fraction at carboxylation sites, as follows: Φ = vo/2 = vcΓ*/cc (see Table 1 for definition of symbols; Von Caemmerer, 2000). By contrast, the day respiratory CO2 flux (Rd) does not follow a similar simple relationship. That is, the metabolic flux associated with glycolysis and the tricarboxylic acid pathway (TCAP) and the value of Rd cannot be predicted (or expressed as a function of [CO2], [O2], vc or vo). Therefore, empirical, correlative or fixed values are used instead. Recently, it has been suggested that photorespiratory and day respiratory CO2 evolution may not purely reflect Rubisco chemistry and the commitment into catabolism, respectively. In other words, it has been assumed that from an optimization perspective, (photo)respiratory CO2 loss is minimized: indeed, in an ideal situation, (photo)respiration may be expected to evolve no CO2 at all, with full recycling of (photo)respiratory products, which may then be incorporated into plant organic matter. For example, glycine molecules synthesized along the photorespiratory pathway have been assumed to feed peptide synthesis under certain circumstances (Dirks et al., 2012) rather than conversion to serine and CO2 production. Here, we will explore the credibility of such an hypothesis. It is now well accepted that CO2 evolved by photorespiration and day respiration is partly refixed by gross photosynthesis. In fact, common equations used to describe gas exchange (Von Caemmerer & Farquhar, 1981) implicitly assume that CO2 liberated by (photo)respiration in the internal CO2 pool (mole fraction ci) can serve as substrate for carboxylation. (It should be noted that the original equations of Von Caemmerer & Farquhar (1981) do take into account refixation, and thus Φ and Rd are not net (photo)respiratory, refixation-substracted rates but are true (photo)respiration fluxes.) This does not mean, however, that all the (photo)respired CO2 is refixed. The common relationship giving net assimilation A = g(ca − ci) = gm(ci − cc) may be interpreted as follows (Fig. 1a): CO2 molecules liberated by (photo)respiration are diluted by the internal CO2 pool, which is permanently renewed by CO2 movement; that is, diffusion of external CO2 into the leaf (flux gca) and retrodiffusion from inside the leaf to the atmosphere (gci). Rather than ci, one may use the intracellular, dissolved CO2 pool (mole fraction cc): similarly, evolved CO2 is assumed to be liberated at the intracellular carboxylation sites and diluted by the diffusion of intercellular CO2 (flux gmci) and the retrodiffusion of intracellular CO2 (flux gmcc; note that this type of arithmetic division (gca, gci, gmci, gmcc) is commonly used to interpret 18O exchange between the leaf and atmosphere (retrodiffusion of 18O-enriched CO2)). At ambient [CO2], photorespired and day-respired CO2 (computed from common gas exchange values and parameters) would be numerically close to 25 and 2.5% of ci, and are thus diluted four and 40 times, respectively. At the same time, the relative exhaust rate of intercellular CO2 (ordinarily, gci/gca ≈ 75%) leads to the loss of most of the (photo)respired CO2 (≈ 95%) and the penetration of ‘new’ CO2 molecules from the atmosphere. With cc-based calculations, one may end up with an exhaust proportion of ≈ 85% (i.e. a potential refixation rate of 15%). Again, these rough calculations assume that (photo)respired CO2 is liberated at the carboxylation site considered (ci or cc). Two recent publications have suggested that such an assumption is incorrect and may lead to erroneous photosynthetic models and thus mitochondrial CO2 production should be divorced from chloroplastic photosynthetic activity (Evans & von Caemmerer, 2012; Tholen & Zhu, 2012). However, it should be emphasized that respiratory CO2 production in the light does not solely involve mitochondria (Fig. 1b). In fact, the general term ‘day respiration’ commonly refers to the nonphotorespiratory CO2 efflux in the light, regardless of the metabolic pathways involved. This is because, in gas exchange experiments, it is impossible to differentiate between CO2 molecules from the ‘pure’ respiratory pathway (TCAP taking place in the mitochondrion) and that from other decarboxylations. In other words, cellular compartmentalization of metabolism is such that photorespired CO2 is produced in mitochondria (glycine decarboxylation) while day respiratory CO2 is simultaneously produced by chloroplasts (chloroplastic pyruvate dehydrogenase, NADP-dependent isocitrate dehydrogenase, NADP-dependent malic enzyme and pentose phosphates), mitochondria (mitochondrial pyruvate dehydrogenase, TCAP, NAD-dependent malic enzyme) and the cytosol (NADP-dependent isocitrate dehydrogenase, pentose phosphates). Such a complex situation associated with day respiration makes modelling quite difficult, because the specific contributions of chloroplastic, mitochondrial and cytosolic decarboxylations are not well known. That said, chloroplastic decarboxylation of pyruvate is believed to be of importance, while mitochondrial TCAP activity is rather low (for a review, see Tcherkez et al., 2012a). Such a division into compartments implies the use of new intracellular environments, in which different values of CO2 mole fraction may be found (chloroplasts (i.e. carboxylation sites, ‘true’ cc) and cytoplasm (cm)), and an additional intracellular conductance (gl) to CO2 diffusion in solution. Photorespiratory CO2 is released by mitochondria into the cytoplasm where it may enter the chloroplast or simply escape. The commitment of photorespiratory CO2 to escaping then depends upon both gm and gl (Fig. 1b) and thus is not easily accessible, unless gm an gl are known. The fate of day-respired CO2 molecules probably depends on their origin: while chloroplastic respired CO2 may be easily refixed, mitochondrial and cytosolic respired CO2 may partly escape from the cell. Further, it is likely that the intracellular position of chloroplasts (adpressed along the plasma membrane, thus covering a significant proportion of cell surface exposed to intercellular airspace) limits escape if CO2. Taken as a whole, there are arguments for and arguments against considering that (photo)respiratory release and carboxylation sites coincide and that such a simplification remains valid for calculations and modelling. Nevertheless, refixation of (photo)respired CO2 should be regarded as inevitable, but rather modest. In fact, the direct analysis of decarboxylated 13CO2 and 13C assimilates upon 13C pyruvate feeding has shown that 13C sucrose and 13C starch represented 0.7 and 0.4% at most of total sucrose and starch, respectively, that is, ≤ 15% of decarboxylated 13CO2 was refixed (Tcherkez et al., 2008). Such estimates are not far from that obtained (17–38%) recently by Busch et al. (2012) using 13CO2 and 12CO2 fluxes, although in that study, gross 13CO2 assimilation was somewhat underestimated (and thus refixation was overestimated) because photorespired CO2 was considered to be in the form of 12CO2 only after 2–3 min in a 13CO2 atmosphere. Other very similar studies taking advantage of 17O–13C (Guillon et al., 2010) or other hetero–nuclear interactions (Cegelski & Schaefer, 2005) and NMR technology have also suggested that under low CO2 or water stress, glycine is not converted to phosphoglycerate and is instead incorporated into proteins. These studies are probably flawed as they have no consideration for 13C enrichment in Ser C-1 or C-2; neglect 17O/16O exchange with solvent water (e.g. exchangeable oxygen of carbonyl groups); show no 17O–13C interaction in Gly C-2 (although expected from consideration of a standard tripeptide); cannot distinguish contributions of individual metabolites from wide peaks (solid-state NMR spectra); and provide no evidence for a causality relationship between 13C-Gly C-2 disappearance and 13C increase in carbonyl signals. Taken as a whole, assuming a complete or predominant consumption of Gly to pathways (such as protein synthesis) other than the photorespiratory ‘cycle’ seems unreasonable, even under particular conditions (low CO2, water deficit). From a metabolic perspective, Gly conversion into Ser is a clear imperative, as Ser is the amino donor of one-half of the glyoxylate molecules produced by photorespiration (the other half being aminated by Glu). Should Gly be mostly consumed by protein synthesis rather than recycling to phosphoglycerate, a huge nitrogen assimilation flux (several μmol m−2 s−1) would be required to compensate for the lack of NH2 recovery from Ser. Such a large flux to nitrogen assimilation (and Glu consumption) is very unlikely: current estimates of N assimilation are, at most, 0.5 μmol m−2 s−1. This does not mean, however, that Gly conversion to Ser is perfectly quantitative or that the leaf Gly content cannot vary. A small amount of Gly and Ser probably escapes and accumulates. From consideration of metabolite build-up, the ‘escaping’ rate is probably near 0.01 μmol m−2 s−1, that is, ≈ 0.4% only of photorespiratory Gly conversion to Ser (Tcherkez, 2011). However, Ser and Gly may also be marginally synthesized by cytoplasmic reactions from glycolysis intermediates (Bourguignon et al., 1999): this contributes to the build-up of whole-leaf Gly and Ser content (and dilutes 13C-Gly and 13C-Ser during labelling experiments). Both photorespiration and day respiration evolve CO2, and at the leaf level, net CO2 fixation is the combination of gross CO2 fixation and CO2 liberation by (photo)respiration. Although some uncertainty remains as to whether CO2 liberation by (photo)respiration occurs at carboxylation sites or should be compartmentalized in photosynthetic models, the proportion of refixed (photo)respired CO2 is probably modest. Quite critically, however, photorespiration and day respiration are complicated processes that interact (or compete) with other metabolic pathways and molecular actors. Such interactions are not well described yet, but they might contribute to adjusting (photo)respiratory recycling fluxes. Day respiration is intrinsically linked to nitrogen assimilation and the provision of 2-oxoglutarate (2OG), as most 2OG molecules neosynthesized in the light are directed to Glu production (Gauthier et al., 2010). Therefore, day respiratory metabolism is accompanied by the abstraction of intermediates, thereby impeding CO2 liberation by the TCAP. In addition, several respiratory enzymes are inhibited in the light (for a review, see Tcherkez et al., 2012a) and this limits CO2 production. Such an inhibition is the result of molecular events (phosphorylation, redox poise effects, feedback exerted by metabolites) that are, in turn, probably influenced by physiological conditions. For example, at high [CO2], CO2 production by the TCAP has been shown to decrease (Tcherkez et al., 2012b) and this further lowers the probability of refixation of day-respired CO2. On the one hand, day respiratory metabolism is required to supply 2OG provision, while on the other, day respiration competes with Suc synthesis for phosphorylated sugars and with Asp (and Asp-derived amino acids) synthesis for oxaloacetate. The compromise found actually seems to satisfy both sides, with a metabolic flux through the TCAP that matches N assimilation. Photorespiratory Gly conversion into Ser involves a complex set of steps catalysed by the glycine decarboxylase complex (GDC) and the serine hydroxymethyltransferase (SHMT). The reaction requires cofactors such as tetrahydrofolate (THF) and flavin adenine dinucleotide (FAD). Recent phosphoproteomics have revealed that SHMT (SHMT1 isoform) may be phosphorylated at a C-terminal Ser residue very close to the THF-binding region in Arabidopsis (Aryal et al., 2012). Similarly, the L-protein of the GDC complex may be phosphorylated at a Ser residue (Nakagami et al., 2010) situated within a nonstructured loop linking two β-sheets that appears to be in the vicinity of the FAD binding site. One may speculate that SHMT and GDC activity or their affinity for cofactors may be finely regulated by phosphorylation under certain photorespiratory circumstances, such as substantial changes in the [Gly]/[Ser] ratio or altered THF and FAD concentration. Still, the use of photorespiratory intermediates by nonphotorespiratory pathways (Gly and Ser ‘leakage’ from the photorespiratory cycle) should be relatively small, so that photorespiratory CO2 evolution is primarily defined by [CO2]/[O2], Rubisco specificity (Sc/o) and GDC stoichiometry (2 moles of Gly consumed for 1 mole of evolved CO2).