Wasteful, essential, evolutionary stepping stone? The multiple personalities of the photorespiratory pathway.

The photorespiratory pathway, in short photorespiration, is a metabolic repair system that enables the CO2 fixation enzyme Rubisco to sustainably operate in the presence of oxygen, that is, during oxygenic photosynthesis of plants and cyanobacteria. Photorespiration is necessary because an auto-inhibitory metabolite, 2-phosphoglycolate (2PG), is produced when Rubisco binds oxygen instead of CO2 as a substrate and must be removed, to avoid collapse of metabolism, and recycled as efficiently as possible. The basic principle of recycling 2PG very likely evolved several billion years ago in connection with the evolution oxyphotobacteria. It comprises the multi-step combination of two molecules of 2PG to form 3-phosphoglycerate. The biochemistry of this process dictates that one out of four 2PG carbons is lost as CO2 , which is a long-standing plant breeders' concern because it represents by far the largest fraction of respiratory processes that reduce gross-photosynthesis of major crops down to about 50% and less, lowering potential yields. In addition to the ATP needed for recycling of the 2PG carbon, extra energy is needed for the refixation of liberated equal amounts of ammonia. It is thought that the energy costs of photorespiration have an additional negative impact on crop yields in at least some environments. This paper discusses recent advances concerning the origin and evolution of photorespiration and gives an overview of contemporary and envisioned strategies to engineer the biochemistry of, or even avoid, photorespiration.


INTRODUCTION
Photorespiration is the process by which the alternate oxygenation product of the enzyme ribulose 1,5-bisphosphate carboxylase (Rubisco) -2-phosphoglycolate (2PG) À is recycled back into central carbon metabolism by what it is the highest flux-bearing metabolite repair cycle on Earth (Bauwe et al., 2010). Photorespiration becomes more prevalent at high temperatures and under drought conditions, which favour the above-mentioned oxygenation reaction relative to CO 2 fixation. Notably, cumulative evidence suggests the photorespiratory pathway remains fully operational, albeit at a lower rate (Carmo-Silva et al., 2008), even in those photosynthetic bacteria, algae and plants that have evolved carbon-concentrating mechanisms (CCM; Eisenhut et al., 2008;Zelitch et al., 2009;Levey et al., 2019). CCMs, including C 4 photosynthesis, serve to facilitate the uptake of CO 2 at low concentrations and in aqueous environments of bicarbonate, which additionally reduces the oxygenation reaction of Rubisco. Before addressing the conundrum set by the title of this article, we will provide a brief overview concerning the elucidation of the photorespiratory pathways themselves, and discuss how they are embedded within plant primary metabolism.
Photorespiration was identified shortly after the elucidation of the CalvinÀBenson cycle as the loss of freshly assimilated CO 2 to the atmosphere via a process unrelated to the light-independent decarboxylations catalysed by enzymes of the tricarboxylic acid (TCA) cycle (Decker, 1959). The biochemistry of photorespiration was first adumbrated by Rabson et al. (1962) and, as we will return to later, is highly likely to have co-evolved with oxygenic photosynthesis some 3 billion years ago. As mentioned above, at the biochemical level photorespiration starts with the oxygenation side-reaction catalysed by Rubisco. This reaction yields one molecule of 3-phosphoglycerate (3PGA) and one of 2PG, the latter of which inhibits the CalvinÀBenson cycle enzymes triose phosphate isomerase and sedoheptulose 1,7-bisphosphate phosphatase, subsequently RuBP regeneration and starch synthesis (Anderson, 1971;Fl€ ugel et al., 2017;Li et al., 2019). In higher plants, the pathway leading to the recycling of 2PG to 3PGA, which can re-enter the CalvinÀBenson cycle, consists of at least eight core enzymes plus a number of auxiliary enzymes and shuttle proteins, and is split across four subcellular locationschloroplast, cytosol, peroxisome and mitochondrion, as shown in Figure 1. That being so, considerable research effort has been put into identifying the metabolite transporters that are needed to support photorespiration. Recently, some advances have been made in this area with the functions of not only the transporters involved in photorespiratory ammonium assimilation (Kinoshita et al., 2011), but also of the mitochondrial uncoupling protein (Sweetlove et al., 2006;Monn e et al., 2018), the mitochondrial glutamate transporter BOU (Porcelli et al., 2018), and the plastidal glycolate/glycerate (Pick et al., 2013) and glycolate (South et al., 2017) transporters being resolved. As such, the photorespiratory pathway in eukaryotes is much more complex than the CalvinÀBenson cycle itself.
Recycling of 2PG begins in the chloroplast via the reaction catalysed by 2PG phosphatase (PGLP), whereby it is hydrolysed to glycolate that moves to the peroxisome, likely through the protein pore PM22 (Reumann, 2011), where it is oxidized to glyoxylate by glycolate oxidase (GOX) and the by-product H 2 O 2 decomposed by catalase (CAT). Also in the peroxisome, transamination reactions catalysed by the glutamate:glyoxylate and serine:glyoxylate aminotransferases (GGAT and SGAT, respectively) are active. GGAT and particularly SGAT are promiscuous enzymes that transaminate a range of different substrates (Zhang et al., 2013;Modde et al., 2017). In the course of photorespiration, GGAT uses glutamate as amino donor to glyoxylate, generating glycine and the 2-oxoglutarate needed for the refixation of photorespiratory ammonium in the glutamine synthetase-glutamate synthase (GS-GOGAT) cycle, whilst SGAT simultaneously acts on the 2-C and 3-C branches of the pathway transferring the serine amino group to glyoxylate and thereby producing glycine and hydroxypyruvate. Glycine is transferred to the mitochondria whilst hydroxypyruvate is reduced to glycerate by hydroxypyruvate reductase 1 (HPR1) in the peroxisome and, in some conditions, by hydroxypyruvate reductase 2 (HPR2) in the cytosol (Timm et al., 2008).
Within the mitochondria, the central function of the photorespiratory pathway, namely the conversion of the 2-C compound glycine to the 3-C compound serine (which is ultimately recycled to 3PGA), proceeds. This process is catalysed by the glycine cleavage system (GCS) in concert with serine hydroxymethyltransferase (SHMT; Kikuchi et al., 2008), with the former in photorespiring mitochondria being an association of four polypeptides in an as yet poorly characterized fragile multiprotein glycine decarboxylase complex (GDC) of approximately 1300 kDa (Neuburger et al., 1986;Oliver et al., 1990). The GCS catalyses three reactions, namely the pyridoxal 5-phosphate-dependent enzyme P-protein (glycine decarboxylase), the polyglutamyltetrahydrofolate (THF)-dependent T-protein (aminomethyltransferase) and the NAD + -dependent enzyme L-protein (dihydrolipoamide dehydrogenase), whilst the lipoic amide-containing H-protein (hydrogen carrier protein) successively interacts as a shared substrate with the P-, T-and L-proteins to transfer reaction intermediates and reducing equivalents from one enzyme to another and ultimately to NAD + . The process converts one molecule of glycine, tetrahydrofolate and NAD + to produce a molecule of methylenetetrahydrofolate, liberating CO 2 , ammonia and NADH. In order to maintain a sufficient supply of NAD + , any NADH generated is rapidly oxidized in a process that is partially coupled to ATP synthesis and subsequently sucrose synthesis in the cytosol (Gardestr€ om and Igamberdiev, 2016). As such, photorespiration has strong effects on the cellular NADH/NAD + balance, and influences other cellular processes such as the TCA cycle and nitrate assimilation (Bauwe et al., 2010). Subsequently, the pyridoxal 5-phosphate-dependent enzyme SHMT combines methylene-THF with a second molecule of glycine to make serine and regenerate THF. SHMT additionally produces considerable amounts of 5-formyl-THF, which auto-inhibits the activity of the enzyme thus necessitating continual detoxification, adding a second-level metabolic repair system on top of that catalysed by photorespiration (Collakova et al., 2008). The serine produced by SHMT is transported back to the peroxisome where it delivers its amino group to glyoxylate to produce glycine and hydroxypyruvate, the latter being reduced to glycerate by HPR1 (Givan and Kleczkowski, 1992). HPR1 as well as the peroxisomal malate dehydrogenase, which normally generates the NADH needed to fuel the reaction, are not essential due to the fact that, as mentioned above, there is a cytosolic bypass of this reaction (Murray et al., 1989;Timm et al., 2008;Cousins et al., 2011). The glycerate enters the chloroplast in exchange for glycolate through the PLGG1 transporter (Pick et al., 2013). In the chloroplast, in low light alternatively in the cytosol (Ushijima et al., 2017), D-glycerate 3-kinase (GLYK) completes the photorespiratory pathway by returning three out of four 2PG C atoms back to the CalvinÀBenson cycle.
Having described the core pathway of photorespiration above, it is also important to understand how photorespiration is embedded within the broader metabolic network of central metabolism. Several studies cover regulatory interactions that photorespiration has with the CalvinÀBenson cycle (Timm et al., 2013(Timm et al., , 2015Fl€ ugel et al., 2017;Jiang et al., 2018;Zhang et al., 2018), photoinhibition (Takahashi et al., 2007), metabolite shuttles (Sweetlove et al., 2006;Eisenhut et al., 2013;Monn e et al., 2018;Porcelli et al., 2018) and shunts with the most recent addition Figure 1. The plant photorespiration core pathway (light blue) spans three organelles: the chloroplast, the peroxisome and the mitochondrion. The enzymes are PGLP, GOX, GGAT, SGAT, GCS (comprising three enzymes plus their substrate H-protein), SHMT, HPR1 and GLYK. CAT detoxifies the generated hydrogen peroxide. The GS-GOGAT cycle (green) re-fixates the ammonia released by the GCS. Complex I, which links photorespiration with oxidative phosphorylation, and several shuttle systems (light grey underlay) keep the redox balance. This also involves uncoupling proteins (UCP), which actually are mitochondrial transporters of aspartate, glutamate and dicarboxylates, collaborating with mitochondrial and cytosolic aspartate aminotransferase (ASAT). At least two enzymes of the core pathway can be circumvented in specific conditions. First, the cytosolic HPR2 bypass supports HPR1, likely when low NADH limits peroxisomal hydroxypyruvate reduction. Second, ATP consumption by cGLYK in the cytosol of shade-grown plants helps alleviate photoinhibition of chloroplasts. Known plastidal transporters are PLGG1 (glycolate/glycerate antiporter) and BASS6 (glycolate exporter). The mitochondrial glutamate importer BOU is necessary to produce the poly-glutamylated THF cofactor.
© 2020 The Authors. of a glucose 6-phosphate shunt that alleviates the inhibition of RuBP regeneration caused by 2PG accumulation (Li et al., 2019). Additional interactions include one-carbon metabolism (Engel et al., 2007;Collakova et al., 2008), N metabolism (Keys et al., 1978;Bloom, 2015;Abadie et al., 2016;Modde et al., 2017;Bloom and Lancaster, 2018;Huma et al., 2018;Zhang et al., 2018) and respiration including the TCA cycle (Strodtk€ otter et al., 2009;Geisler et al., 2012;Daloso et al., 2015;Obata et al., 2016;Tcherkez et al., 2017). These manifold interactions indicate that plant metabolism has evolved in a manner that renders photorespiration highly central to plant primary metabolism. Given that the metabolic interactions of photorespiration have been much reviewed recently (Obata et al., 2016;Timm et al., 2016), we will not detail them here; suffice to stress that photorespiration has developed a considerable number of roles that are not directly related to the detoxification of 2PG.

THE PAST AND FUTURE OF PHOTORESPIRATION
Plant photorespiration, very much like the plant CalvinÀBenson cycle, is a mosaic of archaeal, proteobacterial and cyanobacterial enzymes . Its initial evolution was very likely coupled to the appearance of the dual-photosystem bearing, oxygen-evolving cyanobacteria À the 'oxyphotobacteria'and their divergence from non-phototrophic cyanobacteria sometime between 3.2 and 2.5 billion years ago, as illustrated in Figure 2 (for review, see Schirrmeister et al., 2016;Soo et al., 2017). These ancestral phototrophic cyanobacteria were the first organisms to be challenged by high to super-saturating daytime oxygen tensions in locally oxygen-rich environments, with such conditions being particularly prominent in microbial mats (for review, see Dick et al., 2018;Hamilton, 2019). These conditions triggered new metabolic processes and innovations, including massively increased rates of 2PG synthesis and its subsequent degradation by an ancient photorespiratory pathway. Reconstruction of this pathway revealed that cyanobacterial 2PG recycling typically comprises two partially redundant pathways, a photorespiratory pathway very similar to that of plants plus the bacterial glycerate pathway (Eisenhut et al., 2008). Moreover, certain cyanobacteria can completely decompose glyoxylate to CO 2 via the concerted action of oxalate decarboxylase and formate dehydrogenase. Following these rather localized environmental changes, the radiation of cyanobacterial photosynthesis massively enriched the atmosphere with oxygen and led to the Great Oxidation Event about 2.3-2.4 billion years ago, when the oxygen content of the atmosphere rose from essentially anoxic 10 À5 to about 10 À2 of its present atmospheric level. It is thought that, at this time, a toxic rise in oxygen levels in ancient Archaean microenvironments could have driven the establishment of an oxygen-tolerant archaeal lineage and its transformation into the first stem eukaryotes (Gross and Bhattacharya, 2010;Gold et al., 2017), followed by the evolution of aerobic heterotrophic eukaryotes (Knoll and Nowak, 2017), including the acquisition of mitochondria approximately 1.5 billion years ago (Roger et al., 2017). A widely advocated hypothesis posits that the closest relative to mitochondria-containing eukaryotes on the host side was a hydrogen-consuming anaerobic phagotrophic Archaeon of the Asgard supergroup that engulfed a proteobacterium that eventually became the mitochondrion (Eme et al., 2017;Martijn et al., 2018). By contrast, the single-membrane-bounded peroxisome did not evolve by endosymbiosis but was rather generated de novo by the fusion of endoplasmic reticulum (ER)-derived buds (van der Zand and Tabak, 2013), or alternatively formed as hybrids between the ER-and mitochondria-derived preperoxisomes (Sugiura et al., 2017;Kao et al., 2018).
Plants, as all Archaeplastida, trace their origin to the socalled primary endosymbiosis 1.25-1.4 billion years ago in which a biflagellate phagotrophic eukaryote engulfed a photosynthetic cyanobacterium similar to the extant freshwater cyanobacterium Gloeomargarita lithophora to become the primary chloroplast (Ponce-Toledo et al., 2017;Nowack and Weber, 2018). During this first major split in eukaryote evolution, photosynthesis and also the basic framework of the photorespiratory pathway were conveyed from cyanobacteria to algae and, subsequently, land plants (Eisenhut et al., 2008). Recent data identified the charophytic alga Chara braunii at the root of all modern land plants (Nishiyama et al., 2018). When land plants appeared on Earth about 500 million years ago (Morris et al., 2018), oxygen levels are thought to have increased from 0.2 to 0.3% at the time of the primary endosymbiosis to approximately the levels we have today. However, CO 2 remained 15-fold higher than in our present atmosphere (Hetherington and Raven, 2005). Massive photosynthetic activity in the Carboniferous period (360-300 million years ago) subsequently resulted in an intermittent drop in CO 2 levels coupled to a similarly transient rise in oxygen levels. During the past 25 million years, CO 2 levels were generally lower than those observed today.
Given that the specificity of Rubisco for CO 2 over oxygen is worse at high temperatures and the solubility of oxygen declines more slowly than that of CO 2 (Jordan and Ogren, 1984), temperature was another important factor in some locations. Approximately 30 million years ago (as an adaptation to lower CO 2 levels and higher temperatures), a number of land plants independently evolved a type of CCM that has commonly become known as C 4 photosynthesis (Christin et al., 2008). This pathway is based on a pre-fixation of CO 2 in the mesophyll by phosphoenol pyruvate carboxylase to initially generate a four-carbon (C 4 ) compound. This enzyme efficiently captures CO 2 even at a very low concentration, creating a large gradient in CO 2 concentration between the inside of the leaf and the outside environment. The C 4 compound then moves to the Rubisco-containing bundle sheath cells, where the CO 2 is released by a C 4 acid decarboxylase. Through this mechanism, Rubisco operates at greatly elevated CO 2 levels and 2PG synthesis is low albeit not entirely absent (Zelitch et al., 2009;Arrivault et al., 2017). Despite these advantages, C 4 photosynthesis is not competitive in all climates typically being particularly inefficient in cool climates (Sage et al., 2012) and most land plants, by far, use the C 3 photosynthetic pathway (Sage et al., 2018).
Paradoxically, photorespiration not only provided selective pressure towards the evolution of C 4 photosynthesis but also triggered this process by inventing the first, albeit not overly efficient, plant CCM (Bauwe, 2011). This was demonstrated by the study of C 3 -C 4 intermediate plants (plants with C 2 photosynthesis) such as Moricandia arvensis, Panicum milioides, Flaveria anomala, Mollugo verticillata, Neurachne minor and Steinchisma hians (for review, see Sage et al., 2018). These plants produce glycine in all photosynthetic cells but, due to a restriction of the GCS to the bundle sheath, the photorespiratory glycine is decarboxylated only in the vein-surrounding bundle sheath cells (Rawsthorne, 1992;Schulze et al., 2013;Keerberg et al., 2014). By this means, the resulting higher concentration of photorespired CO 2 in the bundle sheath and the need to return the surplus photorespiratory ammonia nitrogen back to the mesophyll were crucial in preparing the path for C 4 plant evolution (Mallmann et al., 2014).
During the past 60 years, global CO 2 concentration has increased from~0.032% to a present level of~0.041%. This change has led to a decrease in the worldwide photorespiration/photosynthesis ratio by a quarter over the 20th century (Ehlers et al., 2015). Given that the CO 2 concentration will almost certainly increase in the future, photorespiration can consequently be anticipated to continue to fall. The higher efficiency of C 3 plants under these conditions could be levelled to some extent by global warming (Dusenge et al., 2019), which by and large promotes C 4 more than C 3 plants, but might affect competition between C 3 and C 4 plants and alter the species composition and diversity in many ecosystems (Ehleringer et al., 1997). Whilst some researchers forecast that higher atmospheric CO 2 levels will increase crop yields, notwithstanding the continued impact of photorespiration in future climates (Walker et al., 2016), global climate is highly difficult to predict, and local changes to temperature, rainfall and other determinant factors of agricultural productivity could more than offset any productivity gains at a given location.

TAILORED PHOTORESPIRATION FOR BETTER CROPS?
Given that RuBP oxygenation directly impairs CO 2 fixation, decreases the energy efficiency of photosynthesis as a whole and therefore results in an estimated loss of C 3 net-photosynthesis and crop yields between 20 and 50% (Sharkey, 1988), photorespiration has long been a biotechnological target for attempts at enhancing crop productivity (Ort et al., 2015;Weber and Bar-Even, 2019). One focus, which will not be discussed in this review, is on the improvement of Rubisco's catalytic efficiency, including a better ratio of carboxylation versus oxidation of RuBP, by directed evolution and introduction of such an enzyme into the chloroplast of C 3 crops (Keys, 1986;Carmo-Silva et al., 2015;Sharwood, 2017;Flamholz et al., 2019;Zhou and Whitney, 2019). Here, we will detail four different strategies that have been taken to date to lower the photorespiration-related costs and losses from a plant breeder's perspective by avoiding the mitochondrial decarboxylation of glycine and its concomitant ammonia release before presenting the conundrum presented by the confounding data detailing how plant growth can be improved by enhancing the expression of the H-and L-proteins of the GCS. We will start by detailing synthetic bypasses to sections of the photorespiratory pathway chronologically. Apart from specific features, they all intend to reduce or even avoid photorespiratory ammonia release and the resulting need for energy-expensive ammonia refixation in the presence of the fully operational native photorespiratory pathway. As seen in Figure 3, the first of these approaches (Kebeish et al., 2007) was inspired by the glycolate oxidizing bacterial glycerate pathway and constituted the chloroplastidal overexpression of up to five bacterial genes Figure 3. Strategies aiming to improve plant growth by manipulating photorespiration. The schemes at the top and at the very right side illustrate the findings that an artificially enhanced photorespiratory capacity due to increased PGLP or GCS activity improved photosynthesis and plant growth by intra-organellar or inter-organellar, respectively, regulatory feedback on Calvin-Benson cycle activity. The five schemes at the bottom show artificial bypasses aimed at decreasing photorespiratory ammonia release and/or to increase CO 2 concentrations in the chloroplasts. These attempts can be roughly grouped according to the intended degree of glycolate carbon oxidation (25% -Kebeish and Carvalho versus 100% -Maier, South and Shen). The shown peroxisomal bypass was not fully established, as indicated by failed EcHYI expression. Overexpressed enzymes are shown in red. Ac-CoA, acetyl-CoA; CmMS, malate synthase; CrGLDH, single-protein glycolate dehydrogenase; EcGDH(DEF), multi-subunit glycolate dehydrogenase; EcHYI, hydroxypyruvate isomerase; EcTSR, tartronate-semialdehyde reductase; EcTSS, tartronate-semialdehyde synthase; GCS, glycine cleavage system; ME, endogenous malate enzyme; OsGOX3, bifunctional glycolate oxidase; OsOxOx, oxalate oxidase; PDH, endogenous pyruvate dehydrogenase; SBPase, sedoheptulose 1,7-bisphosphate phosphatase; TPI, triosephosphate isomerase. encoding multi-subunit or single-protein glycolate dehydrogenase, tartronate-semialdehyde synthase and tartronate-semialdehyde reductase in Arabidopsis. If these three changes would come into effect, they ought to directly link the reactions catalysed by PGLP and GLYK, bridging all other reactions of the photorespiratory pathway. This strategy was adopted by Dalal et al. (2015) for Camelina sativa, The second approach attempted to stably introduce a related pathway into the peroxisome of tobacco in order to convert glyoxylate to hydroxypyruvate, avoiding the release of CO 2 and ammonia by the GCS, but failed to overexpress one of the required enzymes, hydroxypyruvate isomerase (Carvalho et al., 2011). The third strategy finally is not in fact a bypass as there is no reconnection to the major pathway. Instead, it attempts to fully oxidize glycolate, avoiding ammonia release and the need for ammonia refixation, but one should also note that this approach sacrifices the normally occurring recovery of three out of four glycolate carbons. This is achieved by a combination of endogenous and introduced enzymes, including overexpression of the plant peroxisomal enzymes glycolate oxidase and malate synthase in the chloroplast with a concomitant expression of Escherichia coli catalase in order to detoxify the H 2 O 2 produced (Maier et al., 2012). Malate synthase, in concert with endogenous PGLP, converts 2PG to malate, which subsequently is degraded to acetyl-CoA and two molecules of CO 2 by the chloroplast enzymes malic enzyme and pyruvate dehydrogenase, parallel to the production of reducing equivalents. Thus, the net reaction balance of this pathway is the oxidation of glycolate to CO 2 and H 2 O in the chloroplast, which is thought to improve CO 2 refixation and photosynthetic energy efficiency by avoiding photorespiratory ammonia release and the need for costly refixation. It is further argued that the CO 2 concentration inside the chloroplast would be increased, fostering CO 2 assimilation (Peterhansel et al., 2013), which however is not supported by metabolic modelling (Xin et al., 2015). Concerning the performance of the transformed plants, all papers above reported elevated photosynthesis and higher biomass accumulation.
This year, two variants based on these early strategies have been published. South et al. (2019) modified the Maier et al. (2012) approach for tobacco by replacing glycolate oxidase with Chlamydomonas reinhardtii glycolate dehydrogenase (a membrane-associated single-polypeptide enzyme of which the physiological electron acceptor is unknown), avoiding the need for catalase, along with the repression of the plastidal glycolate exporter PLGG1 in order to further enhance glycolate oxidation. Field-grown tobacco-transformed lines showed a 41% (25% without PLGG1 repression) increase in total dry biomass, whilst lower but still significant biomass increases were observed with tobacco transformed according to the Kebeish et al. (2007) and Maier et al. (2012) schemes. Shen et al. (2019) altered the Maier approach for rice in a different way, by overexpression of a rice glycolate oxidase isoform that is known to efficiently convert glycolate not only to glyoxylate but further to oxalate. The oxalate is then further oxidized to CO 2 and formate by overexpressed rice oxalate oxidase, using rice catalase to protect the chloroplast from the generated H 2 O 2 . Despite this strategy not restricting glycolate efflux from the plastid, it was capable of resulting in enhanced grain yield albeit in a manner that is highly dependent on the seeding season: in the spring or fall, respectively, moderately or massively decreased grain setting rates lead to higher or lower single plant yields of the bypass plants in comparison with wild-type rice.
In light of the reported growth improvements, the above bypass approaches have been much praised despite the fact that a large number of open questions remain unanswered (Box 1). For example, a kinetic modelling approach suggests a negligible advantage from a photorespiratory bypass and confirms what one intuitively would expect: full oxidation of glycolate during photorespiration would decrease photosynthesis by the large amount of lost CO 2 (Xin et al., 2015). Similarly, irrespective of which of the two chloroplastidal metabolic reconstructions or which ratio of carboxylation/oxygenation of Rubisco is used, the increases in biomass obtained on modelling either the Kebeish or Maier pathways were considerably lower than those reported experimentally (Basler et al., 2016). Potential reasons for the lack of consistency between the modelling and experimental results are that the used models do not include all relevant diffusion pathways, such as those concerning the species-specific strategies of how (and how much) photorespiratory CO 2 is recaptured (Busch et al., 2013), biochemical reactions and regulatory interactions by which photorespiration is embedded into central metabolism. Moreover, the artificial pathways could simply not operate at a significant level. This might be even more likely in light of the fact that growth improvements similar to those noted above were achieved by the plastidal overexpression of E. coli glycolate dehydrogenase alone, omitting the two glycerate pathway enzymes, in the chloroplast of Arabidopsis (Kebeish et al., 2007;Bilal et al., 2019), potato (N€ olke et al., 2014;Ahmad et al., 2016) and camelina (Dalal et al., 2015). Similarly, the strong impairments seen after overexpression of glycolate oxidase only in the chloroplast (Fahnenstich et al., 2008) and which can be cured by concomitant catalase overexpression (Maier et al., 2012) strongly suggest that the first step of the introduced glycolate oxidation pathway functions as intended.
More ambitious approaches have been proposednamely CO 2 -neutral and CO 2 -positive photorespiration shunts that overcome the stoichiometric feature of native photorespiration, which is the release of one CO 2 per two glycolate molecules recycled (Shih et al., 2014;Schwander et al., 2016;Bar-Even, 2018). A synthetic photorespiratory bypass based on the 3-hydroxypropionate bi-cycle into the cyanobacterium Synechococcus elongatus sp. PCC 7942 is also worth mentioning (Shih et al., 2014). A very recent development involves the engineering of a glycolyl-CoA synthetase from the E. coli acetyl-CoA synthetase, and a glycolyl-CoA reductase from the Rhodopseudomonas palustris propanediol utilization protein. These two enzymes were used for the reduction of glycolate to glycolaldehyde, followed by the condensation of glycolaldehyde and phosphoglyceraldehyde (via aldolase) to form arabinose 5-phosphate and then RuBP via additional enzymes (Trudeau et al., 2018). That said, proof of the feasibility of these fascinating strategies within plants and specifically crop plants is currently lacking. In order for such approaches to reach the exalted claim that they will usher in the next green revolution (Eisenhut et al., 2019), they will need to be transferred into food crop plants and tested under a range of environments including ones non-optimal to the crop in question.

Box 2: Open questions concerning plant photorespiration
Plants, in contrast to animals, cannot excrete autotoxic metabolic side-products and therefore depend on highly efficient repair pathways, such as the photorespiratory pathway, which has evolved over several billion years and is increasingly well understood. In addition to the need for a better understanding of chloroplastidal glycolate/glyoxylate metabolism mentioned in Box 1, several other questions remain to be answered. These concern the little understood regulation of photorespiration on the levels of transcription, post-translational modification and metabolites involved in intra-and inter-organellar enzyme regulation as well as the identification of shuttle proteins that manage the flow of photorespiratory metabolites into and out of the peroxisome and the mitochondrion. For example, a plant mitochondrial serine transporter was not yet identified, but likely exists in analogy to the need of a mitochondrial serine importer for human one-carbon metabolism (Kory et al., 2018). Another poorly understood aspect is the identity, role and extent of alternative decarboxylation pathways identified at the gas-exchange level (Giuliani et al., 2019). One of these could be the glucose 6-phosphate shunt that was recently identified in a non-lethal photorespiratory knockout mutant (Li et al., 2019); however, its significance in native photorespiration is currently unknown. Related questions concern the actual magnitude of the cytosolic hydroxypyruvate reduction (Timm et al., 2008) and likewise cytosolic glycerate phosphorylation (Ushijima et al., 2017) in response to the prevailing environments encountered during the life cycle of a plant. To mention just one more enigma, the overexpression of the L-protein and particularly the H-protein of the GCS facilitates photosynthesis in the chloroplast. The exact mechanism of this inter-organellar regulation is yet unknown . Notably, the observed changes are unaccompanied by noticeable increases in the respective amounts of the other three GCS proteins. Moreover, in catalytic terms, the T-protein is present in considerable excess over the other GCS proteins (Timm et al., 2018). Both features are difficult to reconcile with the current belief that all or at least most of the GCS proteins in leaf mesophyll mitochondria associate in a GCS multiprotein complex, considerably increasing the enzymatic activity compared with the kinetic interaction of the non-complexed GCS proteins.

Box 1: Open questions surrounding the photorespiratory bypass approaches
In the approaches discussed in the text, carbon fluxes through the introduced bypasses were, without exception, not determined. However, such measurements would be essential to assess the degree of functionality of the intended artificial pathways. The same applies to missing data on changes in the release of photorespiratory ammonia. Such experiments are technically possible and should be performed. Until experimental evidence proves otherwise, all of these metabolic reconstructions should be regarded as dysfunctional or functioning only at a negligible level compared with native photorespiration. In other words, the mechanism (s) by which the genetic manipulations exert their enhancing effect on photosynthesis is currently unknown. We have repeatedly stated this before (for example in , and it was likewise recently stressed by others: "The actual flux distribution between native photorespiration and the various synthetic bypasses remains elusive" (Eisenhut et al., 2019). The Kebeish glycerate pathway approach and derivatives thereof maintain the natural stoichiometry of photorespiration by releasing only one out of four glycolate carbons as CO 2 , which is sensible. By contrast, the Maier approach and derivatives thereof aim at full decarboxylation of glycolate at the same site at which it is produced from the CO 2 assimilation product RuBP. Intuitively, at least in our view, but supported by the modelling of effects on photosynthesis (Xin et al., 2015) and on biomass production (Basler et al., 2016), the latter strategy hence may turn out to represent a cul-de-sac. Irrespective of the critical points we raise above, both approaches undoubtedly demonstrate that a native or artificially introduced/modified chloroplastidal metabolism of glycolate/glyoxylate interacts with CO 2 assimilation in the CalvinÀBenson cycle. Particularly supportive for this view is the fact that significant growth improvements were achieved by overexpressing the E. coli glycolate dehydrogenase alone, omitting the two glycerate pathway enzymes. Previous reports on chloroplastidal glyoxylate reductase 2, which catalyses the essentially irreversible NADPH-dependent reduction of glyoxylate to glycolate [and concomitantly of succinic semialdehyde to c-hydroxybutyrate (Simpson et al., 2008), potential roles of glyoxylate in the regulation of photosynthesis discussed in Allan et al., 2009], and on an as yet unidentified chloroplastidal glycolate:quinone oxidoreductase system that is associated with photosynthetic electron transport (Goyal and Tolbert, 1996) might be promising starting points for future research in this direction.
© 2020 The Authors. Given the above data, it is perhaps surprising that overexpression in Arabidopsis and tobacco of individual GCS proteins (the GCS/GDC is believed to have high control over photorespiratory carbon flux) also resulted in improved plant performance. Specifically increasing the levels of the H-protein Simkin et al., 2017;L opez-Calcagno et al., 2019), which is the shared substrate of the three GCS enzymes, or the GCS L-protein (Timm et al., 2015) resulted in decreased CO 2 compensation points and enhanced growth. T-protein is present in large excess in leaf mitochondria and hence its overexpression has limited consequences (Timm et al., 2018), whilst the overexpression of peroxisomal SGAT caused reduced CO 2 assimilation, elevated serine and asparagine consumption culminating in an altered C/N balance and diminished growth (Modde et al., 2017).
The complexity of the relationship is further highlighted by two papers published this year. The first of these revealed that efficient 2PG degradation is also required to maintain carbon assimilation and allocation in the dicot C 4 plant Flaveria bidentis (Levey et al., 2019), thereby underlining the importance of the photorespiratory pathway in C 4 plants (Zelitch et al., 2009). The second likewise confirmed that knockdown of the GCS in rice resulted in a decreased carbon assimilation combined with more GCS-independent decarboxylation of photorespiratory metabolites (Giuliani et al., 2019), possibly due to increased CO 2 release in the proposed glucose 6-phosphate shunt (Li et al., 2019). These results are easy to reconcile with those following the enhanced expression of the GDC but, as discussed before, it is less easy to do the same with those of the bypass strategies. This statement is reinforced by the number of questions alluded to above that are left unanswered by the work published to date (Box 2). Several of these questions, such as the real-life carbon fluxes though the artificial bypasses or full decarboxylation pathways and their consequences for the magnitude of photorespiratory ammonia production, are tractable by currently available research methods and therefore seem obvious priorities for future research.

CONCLUSIONS
Returning to the question addressed (and partially answered) in the title, photorespiration can be regarded as wasteful, essential and an evolutionary stepping stone. Whilst at first glance the former two of these seem to be irreconcilable, this depends on the standpoint. In purely energetic terms, which is too much focussed on energy efficiency, photorespiration is inarguably wasteful in that it is costly in terms of CO 2 and ATP losses. The mighty counter argument is that any reduction of photorespiratory capacity has dramatic consequences as it has evolved to be a central repair pathway embedded within the core of cyanobacterial and later plant photosynthetic metabolism (as well as having poorly defined roles in non-heterotrophic tissues; Nunes-Nesi et al., 2010). This is not to imply that photorespiration merely hitch-hiked during the evolution of plant metabolism À far from it. Indeed, as discussed above, strong evidence suggests that the opposite is true and that photorespiration was a stepping stone towards the evolution of oxygenic photosynthesis as we know it and much later of C 4 plants, which fix CO 2 considerably more efficiently in warm climes (Sage et al., 2012). Thus, we are left with the conclusion that photorespiration does indeed have multiple personalities, yet it is clearly a highly important component of plant primary metabolism. The fact that attempts to improve photosynthesis by manipulating chloroplastidal glycolate metabolism have apparently proven successful remains somewhat of a conundrum. Similarly, with the exception of todays and still limited understanding of how altered 2PG levels affect the operation of the CalvinÀBenson cycle, the mechanism by which the catalytic capacity of the photorespiratory pathway feeds back on photosynthetic carbon assimilation is not exactly known. However, a comprehensive analysis of the existing transgenic plants including broad range metabolite and flux profiling would allow the resolution of this issue and should be carried out with high priority. Whilst we have made great advances in our understanding as to how the photorespiratory pathway evolved and how it is currently fully embedded within the primary metabolism of plants and algae, new research avenues in understanding the native pathway still exist (Box 2). Prominent recent examples include the interaction of photorespiration with S metabolism maybe by providing serine for O-acetylserine and hence cysteine formation (Samuilov et al., 2018), and the intriguing suggestion that metal ion availability could regulate photorespiration (Bloom and Lancaster, 2018). Thus, considerable research is needed in order to complete our understanding of this fascinatingly complex pathway.