**5. Metabolic Demand for ATP and NADPH**

While plant metabolism employs ATP and NADPH in a myriad of biochemical reactions, the vast majority of ATP and NADPH flux in an illuminated leaf enters metabolic networks at relatively few nodes of central metabolism, most notably CO2 assimilation and related processes, making it possible to reasonably estimate total ATP/NADPH demand [66]. Some reactions require reductive energy from alternative redox carriers (i.e., Fd or NADH) but for convenience in calculation and discussion, we will refer to reductive demand in terms of NADPH (2 e−) equivalents. The fixation of each CO2, and subsequent regeneration of the C3 cycle intermediates requires 3 ATP and 2 NADPH for a total demand of 1.5 ATP/NADPH [67]. In C3 plants growing under ambient conditions, the next largest demand for ATP and NADPH is photorespiration, which results from the molecular fixation of O2 by the first enzyme of the C3 cycle (rubisco, [66,68]). Photorespiration requires 3.5 ATP and 2 NADPH for

complete operation, meaning that as photorespiration increases relative to CO2 fixation, ATP/NADPH demand increases as well.

Altering the relative rates of photorespiration and carbon fixation will alter the relative demands for ATP and NADPH. Rates of rubisco carboxylation (*Vc*) and oxygenation (*Vo*) determine downstream rates and energy requirements for carbon fixation and photorespiration respectively. Since *Vc* and *Vo* in C3 plants are constrained by rubisco kinetics, rates of each can be estimated for a given rate of net CO2 exchange (A) and CO2 and O2 concentration to calculate subsequent ATP and NADPH demand [68–72]. While these calculations have been presented in part across many publications, we compile them all herein to make their use more convenient for the non-specialist to use measured gas exchange data to calculate *Vc*, *Vo*, ATP and NADPH demand, and extra ATP needed above that provided from LEF (see Appendix A). This quantitative framework requires several simplifying assumptions, but these estimates are close enough to show the magnitude of fluxes and relative impact between conditions.

While carbon fixation and photorespiration comprise the largest portion of central metabolic demand, other metabolic processes such as nitrate assimilation requires a significant contribution. Nitrogen assimilation in leaves involves nitrate reduction into nitrite by nitrate reductase (NR) in the cytosol, translocation of nitrite to chloroplast where it is reduced to ammonium by nitrite reductase (NiR), followed by ammonium assimilation into amino nitrogen via the glutamine synthetase (GS)-glutamine-2-oxoglutarate aminotransferase (GOGAT) pathway in the chloroplasts [73]. Nitrate assimilation to glutamine requires 5 NAD(P)H and 1 ATP. Specifically, reduction of one molecule of nitrate (oxidation state +5) to ammonium (oxidation state −3) requires eight electrons (equivalent to four NADPH), whereas the production of a glutamate via the GS-GOGAT pathway requires an additional two electrons (equivalent to 1 NADPH) and 1 ATP [66]. The reducing power required by the plastidic NiR and GS-GOGAT is supplied from photosynthetic electron transport via the reduced Fd. Higher rates of nitrate assimilation in the light than in the dark [74] reflects the tight connection between photosynthetic metabolism and nitrate assimilation. The reducing power needed for nitrate reduction via the cytosolic NR could be provided by the plastidic NAD-driven malate valve [75]. The NADPH demand for nitrate assimilation is estimated to range from ~ 0.35 to 3 μmol m−<sup>2</sup> s−<sup>1</sup> on an area basis, based on the nitrate assimilation rate measured by prior studies [76,77]. These rates of nitrate reduction would require ~2.5%–23% of total LEF in the sample dataset examined in Table 1, making nitrate assimilation a significant electron sink in terms of total electron flux.

Lipid biosynthesis represents another sink for NADPH and ATP consumption in plants, but quantitative estimates of its magnitude have not been reported. Lipids, being an important structural component of membranes, constitute approximately 5% to 10% of the dry weight of vegetative cells of plants [78]. The major constituents of lipids are fatty acids, which can represent up to 10% of the chemical energy of leaves on a biomass basis [79]. The synthesis and breakdown of fatty acids occur constitutively during leaf development. As much as 4% of total fatty acid content in leaves is degraded per day [80]. The turnover of fatty acids is exceeded by the rate of de novo fatty acid synthesis in non-senescent leaves. The net fatty acid accumulation generally increases during leaf expansion, with a rate ranging from 0.16 to 8 μmol carbon atoms mg−<sup>1</sup> chlorophyll h−<sup>1</sup> [80–83]. Plant de novo fatty acid synthesis is an energy-demanding process occurred in plastids. ATP drives the first committed step of fatty acid synthesis, the formation of malonyl-CoA from acetyl-CoA catalyzed by acetyl-CoA carboxylase. Reducing power in the form of NADPH and NADH is required for the two reductases involved in each round of fatty acid synthesis [78]. The predominant carbon source of plastidic acetyl-CoA is pyruvate, which is generated from photosynthetically fixed 3-phospho-D-glycerate (3-PGA) via the intermediate phosphoenolpyruvate. For every molecule of palmitic acid (16:0) produced, eight molecules of acetyl-CoA (generation of each acetyl-CoA from 3-PGA regenerates one ATP and one NADH), seven molecules of ATP, and 14 molecules of NAD(P)H are needed, resulting in the consumption of six molecules of NAD(P)H and surplus of one ATP collectively. Based on the total fatty acid content measured in Arabidopsis and *Brachypodium distachyon* (40 μg cm−<sup>2</sup> leaf area, [83]), we estimate that the NADPH demand to maintain the 4% turnover rate of fatty acids is ~0.5

μmol m−<sup>2</sup> s<sup>−</sup>1, which represents approximately 2% and 0.5% of the total NADPH demand under low light and high light, respectively (Table 1). Due to the small pool size of fatty acids in young leaves, the NADPH demand for fatty acid synthesis would be even smaller in the developing leaves. Although the NADPH demand for fatty acid synthesis is relatively small, this process is highly dependent on light and subject to redox regulation [84]. Nevertheless, while up to 2% of total NADPH demand has potential implications to some situations, this is insufficient to significantly affect calculations for total leaf energy balancing.

**Table 1.** Requirements for energy production for the supply and demand of the energy balancing network under low and high light in *Nicotiana tabacum*. For metabolic demand, shown are rates of CO2 assimilation (A), intercellular and chloroplastic CO2 concentration, rates of rubisco carboxylation (*vc*) and oxygenation (*vo*), rates of nitrate reduction (*Vn*), rates of lipid production (Vl) and total ATP and NADPH demand. For energy supply shown are photosynthetically active radiation (PAR), measured rates of electron transport through PSII (LEF) and PSI (JPSI), rates of linear electron flux needed to provide sufficient NADPH for metabolic demand (LEFpred), ATP produced from LEFpred (ATPLEF) and the ATP deficit. For energy balancing, shown are the electron and photon demands for the ATP deficit to be provided by CEF via the NDH, FQR or *b6f* pathways or the malate valve. Details for these calculations found in the text. Values taken from Miyake et al. 2005 [1] indicated with a star (\*), with remaining values calculated or assumed herein. For these calculations Rl, Γ\* and gm were assumed to be 1.5 μmol CO2 m−<sup>2</sup> s<sup>−</sup>1, 4.7 Pa and 6 μmol CO2 Pa CO2 <sup>−</sup><sup>1</sup> m−<sup>2</sup> s<sup>−</sup>1.


#### **6. Determining the E**ffi**ciency of Energy Balancing Mechanisms**

As an autotrophic organism, the energy that fuels metabolism in plants is derived ultimately from absorbed photons, providing a metric by which to gauge the efficiency of an energy balancing mechanism. Photon use efficiency has thus provided a logical objective function for approaches that assume photoautotrophs use light energy optimally (i.e., [85]), but given the massive amount of absorbed energy that is dissipated as NPQ under high light, it is not clear that light energy is always limiting to growth. Additionally, given the dynamic fluctuations in energy demand and light supply many plants face under growing conditions, it is likely that the capacity for a given energy balancing mechanism may become more important than its efficiency when light energy supply is adequate. In this section we outline the photon costs of various energy balancing mechanism and incorporate them into a quantitative framework. We then use this framework to examine past work and hypothesize that the energy balancing network operates in a high or low-efficiency mode based on light availability. To examine the energy requirements for energy balancing under various light conditions, the ATP and NADPH demand for the C3 cycle and photorespiration has been determined from past work which paired concurrent gas exchange with measurements of electron flux through PSII and PSI (Figure 1, Table 1 and [86]).

Different pathways of CEF have different costs for energy balancing, depending on how many H<sup>+</sup> are pumped per electron excited by an absorbed photon. As outlined above, the highest efficiency CEF pathway proceeds through NDH, which pumps 4 H+/2 e<sup>−</sup> (Table 2). The FQR and *b6f* pathways have identical yields of 2 H+/2 e−. Since 14 H<sup>+</sup> are required to generate 3 ATP in the chloroplast, CEF has an ATP/photon or e<sup>−</sup> stoichiometry of 0.43 via NDH and 0.21 via FQR or *b6f*. Additionally, the e- and photon demand for energy balancing can be calculated for the low and high-light conditions presented in Table 1 and data from Miyake et al., 2005. Under low light, 11 or 22 μmol photons m−<sup>2</sup> s−<sup>1</sup> are needed to produce the ATP needed to balance energy supply via NDH or FQR/*b6f* pathways, respectively, or between 7% and 15% of the total incident light (Table 1). Under high light, this requirement drops to 4%–8%. To gain a more complete picture of the relative energy cost of these mechanisms under their respective light conditions, the photon demand can be expressed in terms of actual absorbed photon energy that enters photochemistry by adding the rates of flux through PSII and PSI. Interestingly, this recalculation reveals that as light level increases, a greater percentage of photon energy absorbed and passed through the photosystems would need to be partitioned to CEF processes for energy balancing, specifically 11%–22% under low light and 13%–27% under high light. Under high light, however, energy from more photons is dissipated as NPQ compared to low light. If energy to NPQ is considered as excess, this means that there is more excess energy under high light that can be used to drive CEF. Specifically, based on the data from Miyake et al. 2005 [86], energy from only 48 μmol photons m−<sup>2</sup> s−<sup>1</sup> was dissipated via NPQ under low-light conditions, but energy from 776 μmol photons m−<sup>2</sup> s−<sup>1</sup> was dissipated via NPQ under high-light conditions. These numbers reveal that while photon energy could be limiting to drive CEF under low-light conditions, there appears to be enough surplus photon energy available under high light to drive CEF. This is expected because if NPQ limits excitation of PSII, it should also limit flux to PSI through LEF, but it will not necessarily limit PSI electron flow of electrons through CEF.


**Table 2.** Energy requirements and efficiencies of CEF pathways and the malate valve to produce supplemental ATP. Shown are the number of absorbed photons used for the calculation of each pathway. Further details and assumptions for calculations are found in the text.

The energetics of the malate valve are more difficult to assess given the added complexity of transport and flexibility of mitochondrial electron transport. The initial energetics and efficiency of the malate valve are tied to LEF; eight photons produce two NADPH and 12 H+, resulting in 2.57 ATP. The energetics following the transport of the reducing power of 2 NADPH into mitochondrial electron transport and ATP generation depend on the e−/H<sup>+</sup> and, more generally, the e−/ATP efficiency of the mitochondria. For our theoretical evaluation of malate valve energetics, we will first assume mitochondrial electron transport operates optimally and each electron contributes maximally to the proton gradient, passing through Complex I, III and IV to pump 10H+/2e−. To produce ATP, these protons pass through a ring of c-subunits of ATPase, with each full rotation producing three ATP and the number of H<sup>+</sup> per rotation depending on the number of c-subunits present in the ring [87–91]. We assume the number of c-subunits is the same as found in animal cells since there is no available data on plant mitochondrial c-subunit number, requiring eight H+/3 ATP, although in yeast there are 10 c-subunits [92]. Since each molecule of ATP synthesized requires the (electroneutral) transport of one Pi with the associated (electrogenic) ADP3−/ATP4<sup>−</sup> exchange activity of the mitochondrial adenine nucleotide translocase (equivalent to the uptake of an additional proton per molecule of ATP synthesized) [93,94], the final stoichiometry is 11 H+/3 ATP, making a theoretically maximum ATP/oxygen ratio of 2.7 [95]. This stoichiometry is closely matched in experimental measurements of the ratio of 2.6 ADP/oxygen consumed in intact mitochondria in potato [96], suggesting that these stoichiometries reasonably approximate mitochondrial respiration in plants despite the highly-branched potential of mitochondrial electron transport. Therefore, for every two NADPH (4 e−) that are processed via the malate valve, 5.45 ATP are produced in the mitochondria. The above discussion focuses specifically on the ATP produced via mitochondrial respiration fueled by electrons provided from the light reactions, we recognize that some ATP may be produced in the light from pyruvate produced during "dark-type" glycolysis. Exact rates of glycolysis-supplied mitochondrial respiration in the light are not available, but estimates from CO2 gas exchange indicate these rates are rather small compared to net assimilation and lower than rates measured in the dark [97–99], suggesting that the bulk of ATP generated in the mitochondria may come from other sources (such as the Mehler valve), but more information is needed to explore this in more detail.

To integrate the production stoichiometries into a complete malate valve cycle, the costs of transporting ATP from the mitochondria back into the chloroplast where it is primarily needed for the phosphorylation of C3 and photorespiratory cycle intermediates must also be considered. Transport of ATP from the mitochondria proceeds via the ADP/ATP translocase [100,101] and into the chloroplast via the plastidic ADP/ATP transporter [102,103].

The above energetics determine that the malate valve is a highly efficient ATP producer on a photon basis. For every eight photons of light energy, 2.57 ATP are produced in the chloroplast and 5.45 are produced in the mitochondria for a total ATP/photon ratio of 1, much higher than the 0.21–0.43 determined for CEF (Table 2). This high ATP/photon ratio means that much less absorbed light energy is needed for energy balancing assuming low and high-light conditions (Table 1). Specifically, only 4.7 and 18.5 μmol photons m−<sup>2</sup> s−<sup>1</sup> were needed under low and high light, respectively (Table 1). This comprises only 3.2\$ and 1.7% of incident irradiance for the low and high light intensities.

## **7. With an E**ffi**cient Malate Valve, Why is CEF Important?**

Given the theoretically greater efficiency of the malate valve, how can we explain the commonly observed participation of CEF in energy balancing? We propose that the real energy cost of CEF will depend on the light intensity and other factors. At low light, when energy capture is strongly limited by the number of photons captured by the photosystems, activating CEF will require diverting energy from LEF, limiting the overall efficiency of energy capture. However, as light input nears saturation imposed by downstream reactions, PSII efficiency drops substantially, either by decreased efficiency of antenna (increased NPQ), or by increases in the fraction of closed PSII centers. In this case, activating CEF may have little effect on the efficiency of LEF (because it is already light saturated) but will increase

total energy capture, albeit with a higher fraction stored in ATP/NADPH. Indeed, experimentally, CEF appears to play a larger role in energy balancing under high, but not low irradiances when there is limiting energy available from absorbed photons [72,86].

For example, when high light intensities drive light-saturated photosynthetic rates, CEF shifts proportionally in response to changing CO2 to cover ATP deficient predicted from gas exchange [72]. These measurements were made across CO2 concentrations and reveal the capacity for CEF to respond to changes in leaf energy demand (Figure 3). We refer to the measured change of CEF in response to changing energy balancing requirements as the "dynamic range of CEF". By comparing the measured dynamic range of CEF to the change predicted from gas exchange modeling, the degree by which CEF participates in energy balancing can be determined. The dynamic range of CEF can run into a self-regulating upper limit when the high rates of CEF increase the ΔpH and initiate qE-dependent NPQ, as occurs in shade leaves exposed to increasing light [104]. This could serve as a protective mechanism, when CEF increases ΔpH to sufficient levels, light harvesting is down-regulated and further energy mismatch is avoided when the capacity for energy balancing by CEF is reached. Interestingly, the dynamic range of CEF was minimal in response to changing energy demanding conditions when measured under low light, indicating that processes other than CEF (e.g., the malate valve) accomplish energy balancing under these low flux conditions [72].

**Figure 3.** Comparison of the measured relative rate of cyclic electron flux (CEF: circle symbols) to the predicted change in CEF required to match ATP/NADPH supply with demand across CO2 concentration (line). Shown are n = 3–4 ± SE. This data is a replotted subset of measurements from Walker et al. (2014) [72].

A role for "low-light, high-efficiency" and "high-light, low-efficiency" energy balancing networks is further supported by flux balance analysis of photosynthetic systems and in mutant lines deficient in CEF. Flux balance analysis of photosynthetic systems that are optimized for energy production per photon of absorbed light predict the malate valve to be the optimal mechanism of energy balancing unless the additional costs of enzymatic interconversions are introduced into the model [105]. This is also supported in work using a modified flux balance analysis approach, which weights flux solutions based on pathway complexity [106]. As light intensity increases, absorbed light energy is actively released as NPQ, indicating that under high light, the system is no longer light limited and the energy balancing network could trade the more light-optimal malate valve for CEF. This position is supported by work showing that CEF is critical for plant growth under high, but not low-light conditions [107].

It is not clear what the advantages of CEF might be over the malate valve, but the speed and flexibility of CEF may provide an explanation. The operation of the complete malate valve requires tight coordination of enzymatic and transport activity between the chloroplast and mitochondria, limiting the dynamic range of its energy balancing capacity over short time scales. The malate valve also requires the careful coordination of two electron transport chains in separate organelles, further complicating the upregulation of this pathway under greater energy balancing demand. CEF occurs only in the chloroplast, simplifying the signaling network required to up- or down-regulate ATP production. By contrast, CEF is likely regulated by stromal ATP levels as well as stromal redox state [108] and thus may also be more rapidly responsive to alterations in energy demands, e.g., during induction or rapid changes in light, CO2, etc. whereas the malate valve appears to require (potentially slower) redox activation (see below). Thus, having at two routes of ATP/NADPH balancing, provides photosynthetic systems with greater flexibility to balance diverse metabolic imbalances as well as providing optimal efficiency under low light (malate valve) or more rapid/responsive responses (CEF) pathways. We further hypothesize that the baseline requirements for energy balance are achieved by the more light-optimal malate valve. This baseline activity satisfies the needs for energy balancing until greater capacity is needed, such as occurs under higher light regimes. Under high light, CEF acts as a highly flexible stop-gap to allow energy balancing to occur under dynamic conditions.

#### **8. Demand-Side Energy Balancing Processes**

While there is much focus on how supply-side reactions mediate energy balancing, there is less focus on how metabolic demand itself changes. For example, under increased ATP/NADPH demand, metabolism could either increase the supply of ATP via supply-side mechanisms like CEF or the malate valve or reprogram metabolism itself to use the supply more optimally. A simple example of this is in the redox regulation of the C3 cycle, where multiple redox post-translational modifications may tune activity to available reducing power availability [109].

There is additional evidence for this demand-side reprogramming to achieve energy balancing in the unique link between nitrate assimilation and photorespiration. C3 plants have lower nitrate assimilation rates when photorespiration is reduced through altered atmospheres [76,77,110–112]. This link could be explained if under ambient conditions, the increased demand for ATP/NADPH imposed by photorespiration is offset by increased rates of nitrate assimilation, which has a much lower ATP/NADPH demand. This would achieve ATP/NADPH balance not exclusively by increased ATP supply, but by repartitioning demand-side processes themselves. Interestingly, expression of nitrate assimilation genes increase in NADP-MDH mutants, suggesting that nitrate assimilation could be a compensatory response to achieve energy balancing when the malate valve is disrupted [75]. NADP-MDH mutants also show improved growth on nitrate-rich media [75,113]. These experiments demonstrate that nitrate assimilation and the malate valve may cooperate to maintain a baseline level of energy balancing, increasing the complexity of the energy balancing network.
