Differences in the photosynthetic plasticity of ferns and Ginkgo grown in experimentally controlled low [O2]:[CO2] atmospheres may explain their contrasting ecological fate across the Triassic–Jurassic mass extinction boundary

Abstract Background and Aims Fluctuations in [CO2] have been widely studied as a potential driver of plant evolution; however, the role of a fluctuating [O2]:[CO2] ratio is often overlooked. The present study aimed to investigate the inherent physiological plasticity of early diverging, extant species following acclimation to an atmosphere similar to that across the Triassic–Jurassic mass extinction interval (TJB, approx. 200 Mya), a time of major ecological change. Methods Mature plants from two angiosperm (Drimys winteri and Chloranthus oldhamii), two monilophyte (Osmunda claytoniana and Cyathea australis) and one gymnosperm (Ginkgo biloba) species were grown for 2 months in replicated walk-in Conviron BDW40 chambers running at TJB treatment conditions of 16 % [O2]–1900 ppm [CO2] and ambient conditions of 21 % [O2]–400 ppm [CO2], and their physiological plasticity was assessed using gas exchange and chlorophyll fluorescence methods. Key Results TJB acclimation caused significant reductions in the maximum rate of carboxylation (VCmax) and the maximum electron flow supporting ribulose-1,5-bisphosphate regeneration (Jmax) in all species, yet this downregulation had little effect on their light-saturated photosynthetic rate (Asat). Ginkgo was found to photorespire heavily under ambient conditions, while growth in low [O2]:[CO2] resulted in increased heat dissipation per reaction centre (DIo/RC), severe photodamage, as revealed by the species’ decreased maximum efficiency of primary photochemistry (Fv/Fm) and decreased in situ photosynthetic electron flow (Jsitu). Conclusions It is argued that the observed photodamage reflects the inability of Ginkgo to divert excess photosynthetic electron flow to sinks other than the downregulated C3 and the diminished C2 cycles under low [O2]:[CO2]. This finding, coupled with the remarkable physiological plasticity of the ferns, provides insights into the underlying mechanism of Ginkgoales’ near extinction and ferns’ proliferation as atmospheric [CO2] increased to maximum levels across the TJB.


INTRODUCTION
The end of the Triassic marked the beginning of a period of geological and ecological upheaval known as the Triassic-Jurassic mass extinction event (approx. 200 Mya). Although several studies have questioned the high rates (Hallam, 2002;Tanner et al., 2004) and sources (Bambach et al., 2004) of biodiversity loss across the Triassic-Jurassic boundary (TJB), it is widely considered as the third greatest mass extinction in the Phanerozoic (last approx. 450 million years) (Benton, 1995;McElwain and Punyasena, 2007). Plant communities were severely affected in terms of species turnover rates (Harris, 1937;Visscher and Brugman, 1981;Fowel et al., 1994;McElwain et al., 1999McElwain et al., , 2009Olsen et al., 2002) and evenness (i.e. the equality of relative abundances among taxa, McElwain et al., , 2009), yet the available data suggest that higher taxonomic ranks displayed remarkable resilience (Ash, 1986;Willis and McElwain, 2013). In general there is a negligible impact of mass extinctions on plants at the family level (Ash, 1986;McElwain and Punyasena, 2007). Instead, plant communities undergo structural reformation, which includes substantial changes of families' relative abundances and distributions and/or in some cases the total loss of growth habits (McElwain et al., , 2009McElwain and Punyasena, 2007;Bonis and Kuerschner, 2012). In this context, differences in the physiological plasticity between species, families or reproductive groups are expected to play a role in shaping the composition of plant communities under changing environmental conditions.
The causal mechanism of the TJB mass extinction has been actively debated (Hallam, 1997;McElwain et al., 1999McElwain et al., , 2009Palfy et al., 2001;Beerling and Berner, 2002;Hesselbo et al., 2002;Olsen et al., 2002;Marzoli et al., 2004). The organic carbon isotope record revealed a major, and synchronous with the mass extinction event, perturbation of the global carbon cycle in the form of a light carbon excursion across the TJB (McRoberts et al., 1997;McElwain et al., 1999;Palfy et al., 2001;Ward et al., 2001;Hesselbo et al., 2002;Guex et al., 2004;Galli et al., 2005;Kuerschner et al., 2007;Williford et al., 2007;Ruhl et al., 2009;Bachan et al., 2012). In addition, recent data strongly support synchroneity between the disruption of the carbon cycle and the early phases of the Central Atlantic magmatic province (CAMP) eruption, pointing towards a cause-effect relationship between extreme volcanism and biodiversity loss (Hesselbo et al., 2002;Marzoli et al., 2004Marzoli et al., , 2008Blackburn et al., 2013;Dal Corso et al., 2014). Environmental change associated with CAMP volcanism and possible clathrate release has been tracked across the TJB using fossil plant and soil-based proxy approaches. Stomatal density and stomatal index changes in fossil leaves, for example, have allowed atmospheric CO 2 to be tracked at high resolution across the TJB, revealing a 2Á5-to 4-fold increase at the end of the Triassic (McElwain et al., 1999;Bonis et al., 2010;Steinthorsdottir et al., 2011), which is further corroborated by soil isotope-based estimates (Schaller et al., 2011). Furthermore, tracking the carbon and sulphur cycles, which are controls of atmospheric O 2 , using the isotopic composition of carbonates and sulphur has led to the development of massbalance models that allow the estimation of O 2 levels in the atmosphere and reveal that the TJB was also marked by low atmospheric O 2 (Berner, 2001(Berner, , 2006. McElwain et al. (1999) suggested a mechanism by which elevated CO 2 could have functioned as a driver of ecological change by causing a global greenhouse effect with an estimated 4 C increase of mean annual temperature. They argued that this temperature increase combined with an impairment of the leaf cooling capacity of plants due to high CO 2 -induced reductions in the stomatal aperture and/or stomatal density of the leaves resulted in leaf size-dependent thermal damage (Kramer and Boyer, 1995). Consequently, broadleaved Ginkgoales were replaced by related species possessing more dissected leaves, while at the same time monilophytes (ferns), which possess dissected leaves, proliferated across much of the Northern hemisphere including East Greenland (McElwain et al., 2007), Germany, Sweden (van de Schootbrugge et al., 2009 and North America (Fowell and Olsen, 1993;Olsen et al., 2002).
Despite strong evidence that Earth's atmosphere and biota have co-evolved throughout Earth history (Woodward, 1987;Tolbert et al., 1995;Beerling et al., 2001;Ehleringer et al., 2005;Igamberdiev and Lea, 2006;Berner et al., 2007;Franks and Beerling, 2009;Willis and McElwain, 2013), our understanding of the mechanisms by which atmospheric composition can shape patterns in plant evolution, such as those documented across the TJB, remains limited. Although several studies have shown that high CO 2 and sub-ambient O 2 have differential effects upon different species (Fukao and Bailey-Serres, 2004;Wang et al., 2012;Haworth et al., 2013), few studies have attempted to explain past, major shifts in ecological dominance through the prism of these differences in the physiological plasticity under fluctuating atmospheric O 2 and CO 2 .
Due to the dual oxygenation/carboxylation activity of plants' primary carboxylase (i.e. Rubisco), every change in the relative abundance of [O 2 ] and [CO 2 ] in the atmospheric mixture results in corresponding changes in the enzyme's carboxylation and oxygenation rates, which are accompanied by dynamic adjustments of the energy flows that support them with ATP and reducing power in the form of NADPH (Farquhar et al., 1980;von Caemmerer, 2000). Since all the light energy absorbed by a plant needs to be quenched, a fine balance between the energy and the reducing power produced during the 'light reactions' and those consumed at the 'dark reactions' of photosynthesis and photorespiration is essential (Sharkey, 1990;Zhang and Portis, 1999;Andersson, 2008;Parry et al., 2008). However, acclimation to high CO 2 is known to result in a wide range of morphological and physiological adaptations such as the downregulation of Rubisco activity and an increase in the resistances to CO 2 diffusion (Ainsworth and Long, 2005;Kirkham, 2011;Kitao et al., 2015) that further complicate the necessary fine adjustment of the energy flows within the photosynthetic machinery. Accordingly, the aim of the present study was to address this gap in our knowledge by investigating the responses of plants belonging to all three major plant reproductive grades (angiosperms, gymnosperms and ferns) after exposure to O 2 and CO 2 atmospheric concentrations similar to those that prevailed across the TJB. Our hypothesis was that the enhanced resilience of ferns across the TJB low [O 2 ]:[CO 2 ] bottleneck was, at least partly, a result of their inherent increased physiological plasticity under low [O 2 ]:[CO 2 ] conditions compared with Ginkgo. In particular, our study focused on the ability of the different plant groups to readjust the energy flows within their photosynthetic apparatus effectively so that they match the acclimated rates of Rubisco carboxylation and oxygenation. Our approach did not exclude the angiosperms, even though the general consensus is that they first evolved much later than the TJB, during the Cretaceous (Crane et al., 1995;Soltis and Soltis, 2004). A combined consideration of the generic diversity in angiosperms (Niklas et al., 1983) and the atmospheric levels of O 2 and CO 2 (Berner, 2006) reveals that their explosive radiation and final dominance was seemingly unaffected by the rapid increase in the initially low, and similar to that of the

Plant material
Mature plants from five species belonging to some of the most early diverging families within each major reproductive grade (Supplementary Data Table S1), namely the monilophytes Osmunda claytoniana L. and Cyathea australis (R. Brown) Domin, 1929, the gymnosperm, 'living fossil' Gingko biloba L., and the angiosperm species Drimys winteri J.R. Forst. & G. Forst and Chloranthus oldhamii Solms were purchased, repotted and then acclimated to glasshouse conditions (mean temperature ¼ 18 C) at University College Dublin Rosemount Environmental Research Station glasshouses for approx. 2 weeks. Pot size, soil mixture, type and amount of fertilizer used were determined independently based on the size, age and special preferences of each species (Table S1).

Controlled-environment experiments
A trial study to assess potential 'chamber effects' was conducted prior to the initiation of growth chamber experiments (Porter et al., 2015). The study revealed that some of the eight Conviron BDW40 (Winnipeg, Manitoba, Canada) walk-in growth chambers of University College Dublin's P EAC facility (Rosemount, University College Dublin) display significant 'chamber effects', thus their further use in the present study was avoided. Following the 2 week acclimation under glasshouse conditions, 2-3 fully expanded leaves were tagged on each individual and the plants were then transferred into four BDW40 growth chambers that displayed no chamber effects (see Porter et al., 2015). Half of the plants were subjected to a replicated TJB atmospheric treatment in two of the chambers for 2 months, while the rest were grown under ambient atmospheric conditions for the same amount of time, serving as our controls (Supplementary Data Table S2). The atmospheric composition used for the TJB treatment was 16 % O 2 and 1900 ppm CO 2 , which is a good approximation of the corresponding mean values for atmospheric O 2 and CO 2 across the boundary reported by mass balance modelling and palaeo-proxy studies (McElwain et al., 1999;Berner, 2001Berner, , 2006Bergman et al., 2004;Belcher and McElwain, 2008;Bonis et al., 2010;Steinthorsdottir et al., 2011). The O 2 and CO 2 concentrations used for the ambient treatment were 21 % and 400 ppm, respectively (Table S1).
CO 2 in each chamber was monitored by a WMA-4 infrared analyser (PP-Systems, Amesbury, MA, USA), and injection of compressed CO 2 (BOC Gases Ireland Ltd, Bluebell, County Dublin, Ireland) enabled stable within-chamber CO 2 concentrations well above ambient levels. The O 2 concentration in each chamber was monitored by a PP-systems OP-1 oxygen sensor, and injection of compressed N 2 produced by a nitrogen generator (Dalco Engineering, Dunshaughlin, County Meath, Ireland) was used to reduce the O 2 levels below ambient. The rest of the growth conditions were kept constant between the two treatments. All plants were grown under a 16 h/8 h simulated day/night program; 05Á00-06Á00 h, dawn; 06Á00-09Á00 h, light intensity progressively rises from 300 to 600 lmol m -2 s -1 ; 09Á00-17Á00 h, mid-day light intensity of 600 lmol m -2 s -1 ; 17Á00-20Á00 h, light intensity decreases from 600 to 300 lmol m -2 s -1 ; 20Á00-21Á00 h, dusk. Temperature ranged from a nighttime low of 15 C to a mid-day high of 20 C, and relative humidity was kept constant throughout the day at 65 %. Chamber conditions in all chambers were recorded at 5 min intervals and are summarized in Table S2. Plants were watered regularly, receiving amounts of water which depended on the particular needs of each species under the two separate growth regimes. Upon completion of the 2 month treatment period, chlorophyll fluorescence and gas exchange measurements were performed on one of the tagged leaves of each plant. Measuring leaves that were fully expanded before the initiation of the treatments meant that we were able to exclude the effects of the simulated TJB atmosphere on the morphology of developing leaves (e.g. adaptation of stomatal density, stomatal index, leaf expansion, etc.). Consequently, our experimental approach enabled us to focus on differences in the innate plasticity/adaptability of the photosynthetic physiology, and specifically the ability to readjust the energy flows in the photosynthetic apparatus efficiently, among our test species after exposure to a low [O 2 ]:[CO 2 ] air mixture.

Gas exchange measurements
Upon completion of the 2 month acclimation period, the responses of net assimilation rate to incident light (light curves) and intercellular CO 2 partial pressure (A-C i curves) were recorded within the chambers with a CIRAS-2 gas analyser (PP-Systems) attached to a PLC6(U) cuvette fitted with a 4Á5 cm 2 measurement window and a red/white light LED unit. Theoretically, measuring within the TJB chambers at 16 % O 2 could introduce error in our measurements due to bandbroadening effects; however, it has been shown that these effects are negligible even when using O 2 -free air (Loriaux and Welles, 2004). Measurements were performed on intact leaves between 09Á00 and 12Á00 h to avoid potential mid-day stomatal closure. Air flow, leaf temperature and vapour pressure deficit during both the light and A-C i curves were maintained at 300 cm 3 min -1 , 20 C and 1Á0 6 0Á2 kPa, respectively.
For the light response curves, tagged leaves from 3-4 plants per species and treatment were enclosed in the cuvette and illuminated at either 1200 (O. claytoniana, C. australis and C. oldamii) or 1600 (G. biloba and D. winteri) lmol m -2 s -1 until full photosynthetic induction, as judged from three consecutive stable readings of CO 2 assimilation (A) and stomatal conductance (g s ), usually within 30 min. The CO 2 and O 2 concentrations used were identical to the corresponding growth values for each treatment (ambient, 21 % O 2 -400 ppm CO 2 ; TJB, 16 % O 2 -1900 ppm CO 2 ). Light levels were then adjusted from 1200/ 1600 to 20 lmol m -2 s -2 in nine/ten descending steps, each with a 3 min duration, which was always adequate to obtain stable A readings. The light-saturated photosynthetic rate (A sat ) and saturating light intensity were then calculated according to Norman et al. (1992).
For the A-C i curves, the tagged leaves previously used to acquire photosynthetic light response curves were again enclosed in the cuvette and allowed to equilibrate at 400 ppm CO 2 , growth O 2 concentration (ambient, 21 % O 2 ; TJB, 16 % O 2 ) and saturating light intensity (calculated from the light response curves) for 30 min. CO 2 concentration in the cuvette (C a ) was then stepwise decreased from 400 to 50 lmol mol -1 (400, 300, 200, 150, 100 and 50) and then increased from 50 to 2000 lmol mol -1 (50, 400, 500, 600, 800, 1000, 1200, 1600 and 2000). Relative stability of C a and A values at each step typically took 4 min, while a close agreement between the two measurements taken at 400 lmol mol -1 indicated that exposure to low C a had not affected the activation state of Rubisco (von Caemmerer and Edmondson, 1986;Ethier and Livingston, 2004). The resultant response curves were fitted using the model equations of Long and Bernacchi (2003). Implementation of the model allowed the calculation of the maximum Rubisco-limited rate of carboxylation (V Cmax ) and the maximum electron flow rate supporting RuBP regeneration (J max ). Respiration in the light (R d ) was also calculated as the y-axis intercept of the A-C i response curve. We have to note here that the V Cmax and J max values reported in our study are adjusted at 25 C using the temperature functions of Bernacchi et al. (2001Bernacchi et al. ( , 2003. A-C i curves obtained using dried leaves of each species were used to correct all measurements for CO 2 leakages Muir et al., 2014).
Incident growth light intensity (Q) for each species was measured with an MQ-200 quantum sensor (Apogee Instruments, Inc., Logan, UT, USA) and was then used to calculate the in situ electron transport rates from the nonrectangular hyperbola that describes the relationship between photon flux and electron transport (von Caemmerer, 2000) as: where h is an empirical curvature factor with an average value of 0Á7 (Evans, 1989), J max is the maximum electron transport supporting RuBP regeneration and Q 2 is the light utilized by photosystem II (PSII) and is calculated as:  where Q is the incident photosynthetically active radiation, abs is the leaf absorptance and U PSIImax is the maximum efficiency of primary photochemistry. A common absorptance value of 0Á85 was used in our calculations (von Caemmerer, 2000), and U PSIImax , which is equivalent to F v /F m , was measured with a continuous excitation fluorimeter (see below). Average values of Q, F v /F m and J max were used for the calculation of J situ at the species and treatment level.
In situ rates of ribulose 1,5-bisphosphate (RuBP) oxygenation (V O ) and carboxylation (V C ) were calculated from the corresponding light curve recordings (i.e. the step of the light curve with an intensity close to that received by each species in situ) according to Sharkey (1988) as: Chloroplastic CO 2 concentrations (C c ) were calculated from the corresponding C i values as: Mesophyll conductance (g m ) was calculated using the constant J modelling method of Harley et al. (1992). Five measurements of the RuBP regeneration-limited phase (typically at C i values between 50 and 120 Pa) of each A-C i response curve were used to calculate the photosynthetic linear electron flow rate (J) as: Given that J is constant when A is limited by the regeneration rate of RuBP, the g m value that minimizes the variance in J was calculated iteratively using the Solver Microsoft Excel add-in (Warren, 2006). C * is the photorespiratory compensation point (i.e. the chloroplastic CO 2 concentration at which photosynthesis equals photorespiration) and depends on the temperature-sensitive relative affinity of Rubisco for CO 2 and O 2 . C * is considered to be relatively conserved among C 3 species and is linearly correlated with O 2 concentration; thus, we assigned standard values of 3Á16/2Á41 Pa at 20 C and 21 %/16 % O 2 , respectively, for all five species in our study (Bernacchi et al., 2002).

Chlorophyll fluorescence measurements
Chlorophyll fluorescence measurements were performed on all three tagged leaves of each plant for a total of 12-18 measurements per species and treatment. After dark-adapting the leaves for 1 h, a Pocket-Pea continuous excitation fluorimeter (Hansatech Instruments Ltd, Kings' Lynn, Norfolk, UK) was used to capture their fast chlorophyll a fluorescence transients. Saturating light (approx. 3500 lmol m -2 s -1 ) was provided by a single high intensity red LED (peak at 627 nm), and chlorophyll fluorescence values were recorded from 10 ls to 1 s. Data acquisition rates were 10 5 , 10 4 , 10 3 , 10 2 and 10 readings per second in the time intervals of 10-300 ls, 0Á3-3 ms, 3-30 ms, 30-300 ms and 0Á3-1 s, respectively. The cardinal points of the recorded polyphasic fluorescence kinetics [OJIP curves, cardinal points: fluorescence value at 20 ls (F o ), fluorescence value at 300 ls (F 300 ls ), fluorescence value at 2 ms (F J ), fluorescence value at 30 ms (F I ) and maximal fluorescence intensity (F m )] were then used to derive the following parameters according to the JIP-test (Strasser et al., 2004), as extended to include the effect of events related to the final electron acceptors of PSI (Tsimilli-Michael and Strasser, 2008): and is the slope at the origin of the fluorescence rise.
is the relative variable fluorescence at 2 ms.

Statistical analysis
Statistical analysis was performed in R (vÁ3Á1Á1). Data were tested for normality and equal variance, and analysed using mixed effects models. Chamber identity was treated as a random effect to identify a possible chamber effect (no chamber effect was detected). Multiple models were run with random and interaction effects. Models were compared using analysis of variance (ANOVA) comparison, and the best fit model was determined using the Akaike information criterion (AIC).
One-way ANOVA and Tuckey post-hoc analysis were performed to assess the significance level of the differences between all fixed effects (i.e. species and treatments) from the best fit model.

RESULTS
Light response curves indicated that the A sat of the test species did not acclimate uniformly to low [O 2 ]:[CO 2 ] (Fig. 1). Even though the photosynthetic stimulation was statistically significant only in C. oldhamii, the two ferns and the two angiosperms increased their A sat by 13Á6-43Á0 % depending on the species, while the gymnosperm G. biloba displayed a small, non-significant decrease (6Á1 %). We should note here that Ginkgo exhibited a decrease in its light-saturated photosynthetic rate despite the fact that light curves for all species were taken at growth CO 2 concentration, i.e. nearly 5-fold higher CO 2 concentration for the TJB treatment plants compared with controls.
This first sign of photosynthetic downregulation under low [O 2 ]:[CO 2 ] was confirmed by the results of the A-C i curves (Table 1). Acclimation to low [O 2 ]:[CO 2 ] led to significant decreases in V Cmax and to a lesser extent J max values in all test species; however, the decreases were most prominent in G. biloba, reaching 64Á8 and 57Á8 %, respectively, relative to the corresponding values of the control plants (Table 1). As a result of these changes, almost all species exhibited a decreased mean V Cmax /J max ratio under low [O 2 ]:[CO 2 ] (Table 1), which is indicative of an altered balance between RuBP carboxylation and regeneration typically observed in plants growing at high CO 2 (Long et al., 2004;Ainsworth and Rogers, 2007;Osada et al., 2010). Yet, we should note that only in three of the species was this decrease statistically significant.
In situ carboxylation rates (V C ) did not display significant changes, despite the decrease in V Cmax , due to the much higher growth CO 2 concentration of the TJB treatment (Table 1). As expected, in situ oxygenation rates of Rubisco (V O ) diminished due to the very low [O 2 ]:[CO 2 ] ratio of the growth air mixture in combination with the altered absolute concentrations of both gases (Table 1), yet overall G. biloba was the only species to display a significantly decreased combined rate of Rubisco carboxylation/oxygenation under TJB atmospheric conditions Table 1). Furthermore, G. biloba was also the species that displayed the most significant reductions in absolute (V O ) and relative (V O /V C ) rates of oxygenation primarily as a result of its extraordinarily high rates of Rubisco oxygenation under ambient conditions (Table 1). Indeed, calculation of the total conductance (g t ) from the measured values of g s and g m revealed that under ambient atmospheric CO 2 G. biloba poses substantially higher resistances to CO 2 diffusion and its photosynthetic machinery operates under significantly lower C c compared with the rest of the test species ( Fig. 2A, B).
Ribulose-1,5-bisphosphate is the substrate for both carboxylation and oxygenation reactions of Rubisco, thus the reduction of the combined Rubisco carboxylation/oxygenation rate observed in the TJB treatment Ginkgo plants should normally be accompanied by a proportional reduction in the in situ rates of electron transport supporting RuBP regeneration (J situ ). Indeed, J situ values followed a similar pattern, showing small, non-significant changes between control and TJB plants in ferns and C. oldhamii, moderate decrease in D. winteri, a substantial decrease in G. biloba (Table 1) and correlated with the corresponding decreases in V C þ V O (Fig. 3A). It is interesting, however, that when the relatively small changes in V C values are ignored and the changes in J situ are plotted against the corresponding changes in V O , the correlation becomes more robust (Fig. 3B). In addition, there also seems to be good correlation between the relative decreases in the in situ V O [Rel. DV O ¼ (V Oamb -V OTJB )/V Oamb ] and the decreases of Rubisco content-dependent V Cmax when plants are exposed to the TJB atmospheric treatment (Fig. 4).
Exposure to the TJB treatment led to plant group-specific changes in F v /F m (Fig. 5A). Compared with controls, the two fern species maintained high F v /F m values while angiosperms showed moderate, yet non-significant decreases. Ginkgo biloba was the only species to display a substantial drop in its F v /F m , which indicated that low [O 2 ]:[CO 2 ] acclimation resulted in partial photoinhibition. Interestingly, changes in F v /F m were found to correlate linearly with changes in both J situ and V O (Fig. 6A, B). Furthermore, in the case of G. biloba, the substantial decrease in the maximum quantum yield of primary photochemistry was accompanied by a 2Á5-fold increase in the rate of heat dissipation per PSII reaction centre (DI o /RC) (Fig. 5B). This clearly suggested a severely decreased efficiency of photosynthetic light reactions in the low [O 2 ]:[CO 2 ]-acclimated plants.

DISCUSSION
Our results clearly demonstrate that Ginkgo biloba diverts an extraordinarily high percentage of the ATP and NADPH produced during the photosynthetic light reactions to photorespiratory metabolism when grown under ambient CO 2 and O 2 concentrations (Table 1). This is contrary to the pattern observed in the two fern and two angiosperm species studied. Inherent (Yiotis et al., 2010) as well as stress-induced (Flexas et al., 2006) diffusional limitations have been proven to result in a substantial drawdown in C c concentrations and high photorespiration rates. Given that the O 2 concentration in the chloroplast is considered to be about equal to ambient under any circumstance, low stomatal and/or mesophyll conductance to CO 2 diffusion could lead to decreased chloroplastic CO 2 /O 2 ratios and thus increased relative rates of V O . Indeed, our results showed that G. biloba poses significantly higher resistance to CO 2 diffusion compared with the basal angiosperms and ferns of the study ( Fig. 2A, B). This seemingly peculiar and extremely low conductance to CO 2 resembles the decreased conductance observed in plants under stress, but it can be explained by the limited ability of G. biloba to adapt to low [CO 2 ] by altering its conductance to CO 2 diffusion. This is supported by recent evidence that the dynamic range of stomatal conductance of Ginkgo is constrained due to its low anatomical G max (McElwain et al., 2016) and the fact that the mesophyll conductance of gymnosperms is generally low (Flexas et al., 2012). Yet, C c does not depend solely on a species' resistance to CO 2 diffusion. C c is a function of both g t and A (von Caemmerer, 2000); thus, in spite of the anatomical and/or TABLE 1. n ¼ 3-4 depending on species. Different letters for each species denote statistically significant differences between treatments (P 0Á05). . Data are means 6 s.d. from 3-4 measurements. Different letters denote statistically significant differences (P 0Á05) between species. (B) Mesophyll conductance (g m ), stomatal conductance at saturating light intensity (g s ) and total conductance [g t ¼ (g m Â g s )/(g m þ g s )] of the test species under ambient atmospheric conditions (400 ppm. CO 2 , 21 % O 2 ). n ¼ 3-4 depending on species. The error bars denote 1 s.d. Different letters for each parameter denote statistically significant differences (P 0Á05) between species.
biochemical restrictions limiting the g s and g m of G. biloba, the species could still increase its operational C c by photosynthesizing at lower rates. Since the light-saturated photosynthesis of G. biloba is clearly limited by its Rubisco carboxylation capacity, a reduction in the enzyme's content could facilitate an increase in the operational C c values. It is apparent that such a reallocation of nitrogen from Rubisco to other proteins would decrease the relative photorespiratory carbon losses under low [CO 2 ]. Although photorespiration was initially considered a wasteful evolutionary relic resulting from a lack of selective pressure when Rubisco was selected by photosynthetic bacteria as their primary carboxylase, recent evidence has demonstrated the importance of photorespiration in photoprotection (Osmond and Grace, 1995;Kozaki and Takeba, 1996;Niyogi, 2000;Takahashi et al., 2007) and nitrogen assimilation (Somerville and Ogren, 1980;Rachmilevitch et al., 2004;Heldt, 2005;Bloom et al., 2010;Bloom, 2015). Interestingly, our results show that there is good correlation between the decreases in V O and the Rubisco content-dependent V Cmax (Fig. 4). Moreover, due to the sheer amount of leaf nitrogen invested in Rubisco (Evans, 1989), we could argue that V Cmax generally reflects plant nitrogen content and nitrogen assimilation capacity (Walker et al., 2014). Concomitantly, the correlation between the relative decreases in V O and V Cmax seems to support the proposed link between photorespiration and nitrogen assimilation.
Plants have a complex control mechanism that adjusts the rate of the 'light photosynthetic reactions' and the combined rate of photosynthesis and photorespiration so that they match each other (Sharkey, 1990;Zhang and Portis, 1999;Andersson, 2008;Parry et al., 2008); however, stress conditions can slow down the rate of the Calvin-Benson-Bassham cycle, thus generating a potentially harmful imbalance. In these cases, photorespiration is believed to act as a safety valve, quenching the excess absorbed light energy by consuming the produced ATP and NADPH that cannot be used for carbon assimilation. Exposure to a TJB treatment, however, diminished photorespiration (Table 1; Fig. 7) and in conjunction with the stable or decreased in situ V C (Table 1; Fig. 7) led to a drop in the demand for RuBP regeneration, which is reflected in the corresponding drops in operational J situ (Table 1; Figs. 3A,B and 7).
Under these conditions of reduced capacity for photosynthetic and photorespiratory quenching of absorbed light energy, the electron transport chain of the light reactions becomes overreduced and plants are forced to increase the efficiency of energy quenching through heat dissipation in order to avoid permanent photodamage (Demmig-Adams and Adams, 1996;Horton et al., 1996;Müller et al., 2001;Lambrev et al., 2012). The amplitude of this increase depends on the amount of excess absorbed energy or, in our case, the drop in the capacity for photosynthetic and photorespiratory quenching, assuming no significant changes in leaf absorption.  Our results do not allow us to identify the increased rate of photorespiration in G. biloba either as proof of the species' reduced fitness to current low atmospheric CO 2 or as an evolutionary strategy aiming to enhance the species' persistence to stress and/or nitrogen assimilation capacity. Yet, it is apparent that exposure to a low [O 2 ]:[CO 2 ] atmosphere and subsequent diminishment of the photorespiratory sink for photosynthetic electron flow severely affects the species' competitiveness, especially when compared with the fern species, which display a high level of photosynthetic plasticity. In contrast to G. biloba, ferns and angiosperms displayed a remarkable adaptability of their physiology to TJB atmospheric conditions. Ferns' adaptive decreases in V Cmax and J max (Table 1) were counterbalanced by the increased CO 2 concentration in such a way that both the in situ V C and A sat showed an increasing, yet nonstatistically significant, trend (Table 1; Fig. 1). In addition, the absence of changes in the heat dissipation per reaction centre (DI o /RC) and the maximum efficiency of primary photochemistry (F v /F m ) under nearly non-photorespiratory conditions further highlights their ability to acclimate their physiology effectively to low [O 2 ]:[CO 2 ] (Figs 5A, B and 7).
We acknowledge that an increase in global temperature and high atmospheric CO 2 -induced decreased transpiration of the broadleaved Ginkgoales must have played a role in their decline across the TJB (McElwain et al., 1999). However, the present work provides an additional mechanism that may have contributed to the near extinction of Ginkgoales and to the proliferation of ferns evident in the TJB fossil record (Fowell and Olsen, 1993;Olsen et al., 2002;. Our study focused exclusively on innate differences in the photosynthetic plasticity and did not investigate the anatomical adaptations of leaves grown under a TJB atmosphere. Yet, we would like to note that after >100 million years of plant evolution and despite the plummeting of atmospheric CO 2 levels, gymnosperms still display relatively low values of stomatal and mesophyll conductance, even though higher values would have enabled them to maintain higher photosynthetic rates and/or higher photosynthetic nitrogen use efficiency (Flexas et al., 2012;McElwain et al., 2016). Based on this apparent reduced anatomical/physiological adaptability of gymnosperms, we believe that the responses of present-day G. biloba to a TJB treatment are likely to resemble those of Ginkgoales present during the TJB. Overall, our study stresses the importance of differences in the physiological plasticity of different plant groups in shaping evolutionary patterns under fluctuating atmospheric O 2 and CO 2 . Thus, in light of ongoing rapid increases in atmospheric CO 2 levels, we argue that further investigation of the innate physiological characteristics and constraints of different plant groups are of considerable significance as we are likely to be on the verge of major shifts in the composition of plant communities.

SUPPLEMENTARY DATA
Supplementary data are available online at https://academic. oup.com/aob and consist of the following. Table S1: classification, size of pots, compost mix and fertilizer used for each of the studies. Table S2: mean values 6 s.d. of growth conditions in the chambers used in the study.