The Effect of Increased Ozone Levels on the Stable Carbon and Nitrogen Isotopic Signature of Wheat Cultivars and Landraces

: Several studies have highlighted the negative effects of ozone (O 3 ) on wheat development and productivity. The negative effects of O 3 are mediated by changes in photosynthetic carbon and nitrogen metabolism, which are difﬁcult and time-consuming to assess and are thus only measured sporadically throughout the plant cycle. Stable isotope measurements in grains can help integrate the effects of chronic O 3 exposure over the lifespan of the plant. This particular study focuses on the extent to which the stomatal conductance and productivity of Mediterranean wheat are related to carbon and nitrogen isotopic signatures under chronic O 3 exposure. An open top chamber experiment was designed to analyse the effects of the pollutant on 12 Spanish wheat genotypes, which included modern cultivars, old cultivars and landraces. Four O 3 treatments were considered. Stomatal conductance (g s ) measurements were carried out during anthesis, and yield and nitrogen content parameters were taken at maturity, along with the carbon ( δ 13 C) and nitrogen ( δ 15 N) isotopic composition measured in grains. Modern and old cultivars responded similarly to O 3 and were sensitive to the pollutant regarding yield parameters and g s , while landraces were more O 3 -tolerant. Grain δ 13 C had a strong negative correlation with grain yield and stomatal conductance across genotypes and O 3 conditions, and increased under higher O 3 concentrations, showing its capacity to integrate O 3 stress throughout the wheat cycle. Meanwhile, a higher nitrogen concentration in grains, coupled with smaller grains, led to an overall decreased grain nitrogen yield under higher O 3 concentrations. This nitrogen concentration effect within the grain differed among genotypes bred at different ages, following their respective O 3 -sensitivity. δ 15 N showed a possible indirect effect of O 3 on nitrogen redistribution, particularly under the highest O 3 concentration. The correlations of δ 15 N and δ 13 C to the usual effects of ozone on the plant suggest their potential as indicators of chronic ozone exposure.


Introduction
Even though tropospheric ozone (O 3 ) is both a naturally occurring and anthropogenic gas, current global levels are mainly caused by photochemical reactions, due to precursors from industrial and urban emissions [1,2]. Local and regional air masses can transport precursors throughout long distances before leading to ozone production. When high ozone precursors and O 3 -favorable meteorological conditions for atmospheric chemistry coincide, pollutant formation increases, causing chronic and extensive O 3 -pollution problems in rural and forested areas, far away from the source of emissions [3]. This, along with the complexity of atmospheric O 3 photochemistry, coupled with high NO x levels, explains the difficulty in maintaining O 3 levels under safe values for crops and ecosystems [2,4]. In Europe, several EU Directives have prompted the reduction of O 3 -precursor emissions (2008/50/EU and 2016/2284/EU), which caused the high O 3 peaks that were of maximum concern during the eighties. Nonetheless, background O 3 levels still do not present a clear decreasing trend [5,6], and safe levels are not expected to be reached by the year 2030 [2].
Regarding chronically-high O 3 levels, the Mediterranean basin is the most affected area in Europe [5,7]. Its stable summer conditions, high temperatures and high solar radiation favour photochemical O 3 formation, which causes surface levels during spring and summer that surpass the threshold targets of the EU Directives (2008/50/EU and 2016/2284/EU) and the critical levels of the United Nations Air Convention for plant protection [5, 8,9].
Along with the pollution problem, the expected impact of Climate Change is also recognized as a major concern in the Mediterranean area, a transition zone located between the arid North African climate and the temperate central European climate, which is prone to suffer substantial climatic changes [10,11]. Temperature is expected to increase by up to 5 • C by 2100 and, coupled with 25% less precipitation in the Mediterranean basin, these changes are expected to greatly harm crop yields [11] and worsen the O 3 pollution problem [6,12].
O 3 affects the plant by entering through the leaf stomata and producing reactive oxygen species (ROS), which react with the cell membranes and walls [13]. This early O 3 -induced oxidative burst drives cellular chain effects that alter physiological processes such as carbon assimilation by affecting stomatal conductance (g s ), inducing partial closure or less efficient stomatal control [14], reducing foliar pigment content [15], affecting the electron transport chain [16], reducing the efficiency of Calvin cycle enzymes such as Rubisco [17], reducing photosynthates made available for redistribution [18] and increasing respiration rates [19]. As a final result, yield and grain quality are reduced [20,21]. Over 20% of all crop production in Europe is currently considered at risk of damage by O 3 pollution, as O 3 -sensitive and moderately sensitive crops are grown in around 16% and 7%, respectively, of the agricultural land in Europe [21,22].
Wheat, one of the most important staple crops on a global scale, is also one of the most O 3 -sensitive crops [21,23,24]. Studies from the last decades have shown that this pollutant can induce important wheat yield losses and reduce grain quality (i.e., starch content, seed size), while also affecting N use efficiency and protein yield [25]. O 3 -induced wheat yield reduction is important both for its economic impact and its effect on global food security. In terms of wheat yield, modelling exercises have shown that O 3 pollution can account for an average global reduction of 7.1% [26], ranging between 4.1-12.1% for Europe [19]. Moreover, data worsens when coupled with other constraints such as climate change and pressures to maintain natural resources [27].
Breeding for higher grain yield (GY) in wheat and other herbaceous crops has been coupled frequently with an enhancement of stomatal conductance (g s ), which favours higher photosynthetic rates and cools down plants [28][29][30][31]. However, the higher g s of modern wheat cultivars can also potentially increase their O 3 -sensitivity, as they absorb the pollutant more readily. This has led to a clear trend in the genotypic variation regarding how wheat cultivars respond to O 3 . In general, modern cultivars are more O 3 -sensitive than older ones [28,32]. This also applies to wheat bred for Mediterranean conditions, although their adaptation to water-limited environments might have also made them more O 3 -tolerant (Bermejo-Bermejo et al., in press).
Ozone-induced effects in plants are usually studied through discrete gas exchange measurements during the crop season [28,33,34]. However, as these measures are very time consuming, they are usually only performed a small number of times throughout the growth cycle [26,35,36]. Usually, what is observed in these instantaneous measurements is extrapolated to the entire cycle, even though O 3 responses can differ depending on the plant growth stage, with grain filling appearing to be the most susceptible stage [37,38]. A better-integrated response of plant gas exchange and assimilation can be analysed through the study of the stable carbon (δ 13 C) and nitrogen (δ 15 N) isotope compositions [39][40][41][42], particularly when analysed in grain after harvest. Carbon fractionation is driven by the discrimination of stable carbon isotopes, favouring the uptake and further assimilation from the surrounding air of the lighter isotope, 12 C, against 13 C [39,43]. It should, however, be noted that frequently, rather than expressing the isotope signature as a composition (denoted as δ) which is the deviation of the 13 C/ 12 C molar abundance ratio between the analysed substance and a standard, values are transformed to discrimination (denoted as ∆ 13 C) of the 13 C from the atmosphere surrounding the plants [43]. From a physiological point of view, this discrimination depends on the ratio of intercellular to atmospheric CO 2 concentration (Ci/Ca) [39,43]. If this ratio is integrated over time, it shows the balance between the g s and the net photosynthetic assimilation of the plant during the considered period [43][44][45]. Thus, δ 13 C discrimination occurs due to the diffusion of CO 2 through the stomata and by Rubisco carboxylation [43,46]; therefore, if the stomata close, for example as a response to water stress, g s decreases, restricting the CO 2 supply to the carboxylation sites, and overall decreasing the Ci/Ca and consequently ∆ 13 C, while increasing δ 13 C [39,40,43].
Meanwhile, δ 15 N values are driven by discrimination of 14 N against 15 N, which depends largely on the capacity of the plant to assimilate the nitrogen available in the soil. Thus, δ 15 N in the plant may reflect δ 15 N of the nitrogen available in the soil, as well as metabolism, including recycling, transport through the plant, including the grains and even volatilization during senescence [38,47,48].
In that sense, changes in g s may affect the transpiration stream and, subsequently, nitrogen uptake from the soil, its transport to the leaves and volatilization, thus affecting δ 15 N [49]. Similarly, retranslocation of nitrogen through the plant via phloem, as well as photorespiration, can cause δ 15 N to increase in the upper organs, whereas roots tend to have lower δ 15 N than the stems and leaves, and these have less δ 15 N than the ears [47]. Nitrogen uptake after anthesis is quite low [50], therefore, as O 3 -effect is more predominant post-anthesis [37], N redistribution could be expected to be affected more than uptake by the pollutant.
In the present study, the effects of O 3 exposure on the stable carbon and nitrogen isotope composition and its relation to yield parameters, as well as g s , were studied, considering a set of Mediterranean bread wheat genotypes that included landraces, along with old and modern commercial varieties, of which all except for one were developed after the so-called Green Revolution, which took place in the middle of the last century. The objective of the study was to assess the performance of grain δ 13 C and δ 15 N as an integrated tool to monitor chronic O 3 effects on C assimilation and N metabolism in Mediterranean wheat genotypes. Both δ 13 C and δ 15 N have been shown to perform as useful indicators in wheat under other types of stresses, such as water-stress and salinity [39,42,51].
The starting hypotheses were that higher O 3 -stress will decrease g s , increasing δ 13 C (or lowering carbon isotope discrimination, ∆ 13 C) particularly on the more O 3 -sensitive genotypes, in comparison to the tolerant genotypes. We postulate that this would cause a negative correlation between δ 13 C and yield parameters or g s due to the effect of O 3 , differing among wheat genotypes developed at different times. Even though the mechanisms explaining δ 15 N are less established than those of δ 13 C, since N absorption, metabolism and redistribution affect the nitrogen isotope signature, continuous O 3 -exposure is also expected to affect δ 15 N of mature grains.

Field Site
The experiment was developed in the CIEMAT Open Top Chambers (OTCs) facility located at "La Higueruela" MNCN/CSIC Research Farm (Santa Olalla, Toledo) in central Spain (450 m.a.s.l.; 40 • 3 N, 4 • 26 W) (Figure 1), which was specifically designed to study the effects of increasing O 3 levels on crop plants. According to the Köppen-Geiger climate classification, this area experiences a hot-summer Mediterranean Climate (Csa) [52] and is used for rainfed cereals, characteristic of the agriculture of the central Iberian Peninsula.

Open Top Chamber Facility
The experiment employed fifteen NCLAN-type OTCs, which are circular-like greenhouses (3 m high × 3 m diameter) that are open at the top through a conic structure (frustrum) to prevent wind from entering the chambers [53].
The experimental design of the facility included a complete randomized 3-block design with four O 3 treatments replicated three times: thus, 3 OTCs per O 3 treatment: Charcoal Filtered Air (FA), Non-filtered air (NFA) to reproduce ambient levels, Non-filtered air + 20 nL L −1 of O 3 (NFA+) and Non-filtered air + 40 nL L −1 of O 3 (NFA++); where the last two reproduce O 3 levels that are expected along this decade, as well as three chamberless plots (ambient air-AA) to check for chamber effect. More detailed information on the OTC facility can be found in Calvete-Sogo et al., 2016 [54].

Instrumentation OTC Facility
An O 3 generator (A2Z Ozone Systems Inc., Louisville, KY, USA) system, located in a control cabin and fed with pure O 2 , was used to apply O 3 to the NFA+ and NFA++ OTCs for 8 h day −1 (6:00 to 14:00 GTM), seven days a week. Air pollution was monitored continuously above the canopy inside each of the OTCs and AA plots by employing a timesharing system, with a sampling period of 10 min per plot. The air pollutants measured were: Ozone (ML ® 9810B, Teledyne Monitor Labs, Englewood, CO, USA) and nitrogen oxides (NO 2 and NO; ML ® 9841, Teledyne Monitor Labs, Englewood, CO, USA). Monitors were calibrated at the beginning of the experiment following a standard protocol.
Meteorological parameters were also monitored within six of the OTCs and the three AA plots: air relative humidity (RH), temperature (HOBO ® Pro v2, Onset, Bourne, MA, USA) and photosynthetic active radiation (PAR; OSO-SUN HOBO ® , Onset, Bourne, MA, USA).

Ozone Exposure Indices
Several ozone exposure indices were calculated to describe ozone exposure during the fumigation period: the accumulated AOT40 index was calculated as the sum of exceedances of O 3 -hourly mean values over 40 nL L −1 during daylight hours through the O 3 -exposure period [55]. In addition, 24 h-mean, 7 h mean and maximum hourly O 3 values were also calculated.

Plant Material
Twelve Spanish wheat genotypes were selected for the assay: 4 modern commercial varieties (CVs), 4 old CVs and 4 landraces ( Table 1). Pane 247 and the landraces were released pre-green revolution. The Plant Breeding Unit of the Agronomic Engineer School of Madrid (UPM) provided the seeds which were germinated in a plant nursery greenhouse. Seedlings were transported to the OTC field for transplanting into pots (18 cm 3 ) using a mixture of turf (60%), perlite (20%) and vermiculite (20%) on January 26-28 of the year 2015. Four plants per pot were considered as well as four pots per cultivar and OTC (totalling 12 pots per CVs and O 3 treatment). Pots were fertilized following the commercial management of wheat fields in the area, with 120 kg N ha −1 applied in two doses, on April 14th and 30th. Regarding irrigation, plants were watered manually based on the plant needs at the time (e.g., climatic conditions) to maintain water saturation at field capacity, approximately every two days, until the beginning of grain hardening towards the end of May, at which point watering stopped to allow for grain maturity. Plant O 3 exposure began on April 14th, 77 days after transplant to pot (DaT), when plants were close to the anthesis stage. Fumigation lasted for 51-71 days, depending on the CV, until plants reached full grain maturity and were harvested.

Physiological and Yield Parameters
Stomatal conductance (g s ) measurements were performed using a portable gas exchange system (Li-Cor 6400; Li-Cor, Lincoln, NE, USA) between 9:00-13:00 h GMT from May 4-14 (depending on the cultivar) after 20-30 days of exposure to the different O 3treatments and, therefore, around 100 DaT. Two plants per OTC for two blocks were randomly selected for the measurements, with one flag leaf per plant. The standardized conditions during gas exchange measurements for all the cultivars were: PAR 1000 µmol m 2 s −1 , RH 55% and a temperature range of 20-22 • C.
Yield, agronomical yield components and nitrogen content parameters were taken at grain maturity from June 3 to June 23 depending on each genotype (Table 1). After harvesting plants from the base, they were divided into straw and spikes, dried at 60 • C for 48 h and weighted to obtain the dry weight of the different plant parts (DW pot −1 ), totalling 48 pots per CV and 12 pots per CV and O 3 treatment. Spikes were manually shelled to calculate the grain yield. Harvest Index (HI) was calculated as the ratio between grain and the total weight of the aerial plant (straw plus grain husks). Individual grain weight was then calculated and further grain nitrogen content analysed as described below. The number of grains per ear was calculated by dividing the total number of grains over the total number of ears for each pot.

Nitrogen Content and Stable Carbon and Nitrogen Composition
After measuring harvest parameters, the whole grain production from each genotype and OTC was pooled for isotopic analysis (15 samples per genotype and 3 samples per O 3 treatment and genotype). Pooled grain per OTC and genotype was ground and analysed for total nitrogen content and 13 C/ 12 C and 15 N/ 14 N signature using an isotope ratio mass spectrometer (Delta C IRMS, ThermoFinnigan, Bremen, Germany) paired with an elemental analyser (Flash 1112 EA, ThermoFinnigan, Bremen, Germany). One-mg samples of the ground wheat, and reference materials, were weighed, placed in tin capsules, sealed and loaded onto an automatic sampler (ThermoFinnigan, Bremen, Germany) for EA-IRMS analysis. These measurements were carried out at the CCiT (Centres Científics i Tecnológics) at the University of Barcelona. Total nitrogen content was expressed as percentage (%) of dry matter, while the isotope 13 C/ 12 C ratio was expressed in δ notation [46]: where "sample" refers to plant material and "standard" to the international secondary standards of known 13 C/ 12 C ratios (IAEA CH7 polyethylene foil, IAEA CH6 sucrose, and USGS 40 L-glutamic acid) calibrated against Vienna Pee Dee Belemnite calcium carbonate (VPDB). Analyses were performed with an analytical precision (SD) of 0.10‰. The 15 N/ 14 N ratio is also expressed in δ notation (δ 15 N); however, using international secondary standards of known 15 N/ 14 N ratios (IAEA N1 and IAEA N2 ammonium sulphate and IAEA NO3 potassium nitrate) calibrated against N 2 in the air, with an analytical precision of 0.18‰. Nitrogen as a concentration (N%) was converted into Nitrogen Grain Yield (NGY) following the function: NGY = (N% × Grain Yield)/100.

Statistical Analysis
Effects of O 3 and genotype age-type factors on the different variables were analysed via two-way analysis of variance (ANOVA) with IBM ® SPSS ® Statistics 20 (Chicago, IL, USA) using block as a random factor. Significant differences among levels of the factors (p < 0.05) were assessed with the Tukey Honestly Significant Difference test (HSD). The two-way ANOVA data were previously checked to fulfil the ANOVA requirements of normality and homoscedasticity through Shapiro-Wilk and Levene's tests, respectively, using the SPSS Explore Descriptive Statistics function, with Plots. The same procedure was then followed with a one-way ANOVA analysis of each genotype, using the post-hoc Tukey's HSD test. The old CV Pane 247 was excluded when the grouped analysis of the modern and old CVs was done, due to its proximity to the landrace group and its behaviour as an outlier within these age-groups.
Pearson correlations between δ 13 C and δ 15 N and stomatal conductance and yield parameters were analysed. The significance of the regressions was assessed by p-value (p < 0.05) and the goodness of fit was judged by the determination coefficient (R 2 ).
Finally, a Principal Component Analysis, without rotation and with Keiser Normalization, was performed. The standardized data, excluding Pane 247, were checked for Kaiser-Meyer-Olkin Measure of Sampling Adequacy and Bartlett's Test of Sphericity for the usefulness of the factor analysis. Then, the components with an Eigenvalue higher than 1 were selected (leading to two components) and verified with the scree plot to check for a clean drop in the Percentage of Variance after PC2. Variables selected for each component presented a Component Matrix value higher than 0.5. This was repeated for the data in general, and then the data were divided by genotype group. The PCA images were acquired with R, version 3.6.3 (29 February 2020) to include the ellipses. Table 2 summarizes the methodology and instrumentation for each experimental stage.   Mean values for the meteorological parameters inside the chambers during the exposure period are represented in Figure 2b. All the parameters followed the seasonal trend of the climatic area at the OTC site. Air and soil temperature increased from April to June, with maximum mean daily values ranging from 22-36 • C for air, as the growth cycle progressed, and 22-34 • C for soil. On the contrary, RH values dropped from 48% to 24% between April and June, reaching the minimum mean daily values at midday (minimum hourly mean values around 24% in June). Photosynthetic active radiation (PAR) also increased throughout the cycle, ranging between 850-1200 µmol m −2 s −1 inside the OTCs, and peaking in June at the end of the experiment.

Genotype Age-Type Effects
The age-type genotype factor caused clear significant differences in the means of all the parameters assayed: GY and GNY, g s and stable isotopic composition ( Table 3, Age-type factor). In general, no significant differences existed between the modern and old CVs, which, in turn, differed from the landraces significantly. More specifically, breeding caused an improvement in the GY of both categories of CVs, which increased by 34% (mean across O 3 treatments and cultivars) when compared to the landraces. Regarding HI, it increased by 37% in CVs compared to the landraces. Cultivars exhibited a higher number of grains per ear (by 54%) than the landraces, but at the expense of lower grain size. Thus, single seed size was 14% lower in CVs than landraces; also, breeding caused a decreased straw yield (SY) (38%) and total aboveground biomass (3%) (Appendix A: Table A1). Table 3. Ozone and age type effects on yield parameters, g s and isotopic signals: GY, grain yield (g pot −1 ); HI, Harvest Index; GNY, grain nitrogen yield (g pot −1 ); g s , stomatal conductance (mmol H 2 O s −1 ); δ 13 C, stable carbon isotope composition (‰) of grains; N, nitrogen concentration (% DW); δ 15 N, stable nitrogen isotope composition (‰) of grains. FA = charcoal filtered air, NFA = non filtered air, NFA+ = non filtered air supplemented with 20 nl L −1 of O 3 , NFA++ = non filtered air supplemented with 40 nL L −1 of O 3 ; Mean values ± SE. Different letters indicate statistically significant differences in the mean (p < 0.05). These improvements on yield parameters, caused by the varietal selection process, have been accompanied by an increment in the g s of the CVs. Post-anthesis measurements of g s were 27% higher in CVs than in landraces.
This same pattern was observed for δ 13 C, where landraces showed 2.13 ‰ higher (i.e., less negative) values compared with the modern and old CVs. The breeding process decreased grain N concentration by 6%, while enhancing GNY by 32% for the CVs (modern and old) vs. landraces. δ 15 N signal decreased by 0.96‰ for the cultivars compared with landraces.

Ozone Effects
Considering the general response across genotypes, O 3 -exposure affected GY, grain nitrogen yield (GNY) and both isotopic signals (Table 3, Ozone factor). When O 3 increased from FA to NFA++, yield parameters decreased: GY by 13% and GNY by 10% (mean values across the three age-type genotypes), with the effect being highly significant for GY (p = 0.003) and a trend near significant (p = 0.062) for GNY. Grain size (p < 0.0001), biomass (p < 0.0001) and Straw Yield SY (also a trend near significant-p = 0.057) also decreased by 11%, 9% and 7%, respectively, when FA increased to NFA++ (Appendix A: Table A1). Neither HI nor number of grains per ear were affected by the pollutant.
The observed δ 13 C enrichment induced by the pollutant was small; 0.61‰ in NFA++ compared with FA; however, this response was stable and highly significant (p < 0.0001). The significant response of grain δ 15 N to O 3 exposure was not linear: increasing from FA to NFA+ 0.27‰ to then decrease by 0.54‰ from NFA+ to NFA++. Meanwhile, the O 3 -induced reduction observed for g s was only significant for the CVs.

Age-Type Genotypes × Ozone Interactions
Modern and old CVs together responded similarly to O 3 and contrasted with landraces ( Table 3, Age × Ozone). Cultivars subjected to the breeding process, no matter their age, were O 3 -sensitive. Considering yield parameters, the general response was an O 3 -induced yield decrease. The percentage of yield loss for the NFA++ treatment compared with FA across CVs were 17% for GY and 11% for GNY. Meanwhile, grain δ 13 C increased by 0.79‰ (again with a stable and significant response, <0.0001). When only CVs were considered, the O 3 effect on g s was also significant (p < 0.05), inducing a strong (37%) reduction in NFA++ treatment when compared to the FA control. Other parameters affected were grain size and biomass, which decreased by 15% and 11%, respectively. The non-linear O 3response pattern observed in the general O 3 -effect on the grain δ 15 N signal was softened but maintained within the CVs group: the pollutant first caused an increase from FA to NFA+ by 0.10‰, to then decrease by 0.23‰ from NFA+ to NFA++. CVs also presented a general N% increasing trend of around 7%. Meanwhile, when landraces were analysed alone, they presented no significant response to the O 3 exposure to any of the parameters assayed.
Nonetheless, despite their O 3 -sensitivity, CVs subjected to genetic selection still presented an overall 34% higher GY when compared to Landraces (mean value across cultivars and O 3 treatments) and maintained this yield advantage under the highest O 3 treatment (26% higher). This breeding improvement under NFA++ exposure also persists when other parameters are considered, such as GNY (30% higher), HI (34% higher) and grain δ 13 C (1.48‰ lower; 6% more negative); or other parameters, such as seed size (25% lower), number of grains per ear (54% higher), SY (41% lower) and total biomass (8% lower) (Table A1). On the contrary, for g s and nitrogen, the pollutant outweighed the benefits of varietal selection and values between CVs and landraces under NFA++ were similar (without significant differences). For grain δ 15 N, the increased discrimination by breeding also remained even when exposed to the highest O 3 treatment, with CVs values 0.52‰ lower than Landraces.
Intra-variability in the O 3 -responses within each of the three age categories of genotypes was lower for the δ 13 C measurements (Figure 3a) than for the g s measurements (Figure 3b). However, some genotypes differed from the average response of their own group. All of the δ 13 C values for the modern and old CVs, except for Marius, showed a significant response to increasing O 3 levels; however, some exhibited a more sensitive response: Berdún, Califa Sur, Arthur Nick and Ablaca suffered the most intense effect under the NFA++ treatment, with increases ranging between 0.87-1.04‰, while Califa Sur, Yecora and Pane 247 were also significantly affected by the lower O 3 -exposure of the NFA+ treatment (increments from 0.43-0.81‰). This was followed by Pane 247, Yecora and Nogal, which, although less O 3 -sensitive, had a corresponding increased δ 13 C signal ranging from 0.56-0.69‰ under NFA++. In contrast, none of the landraces showed significant changes. When g s was considered, Califa Sur, Arthur Nick and Ablaca can be considered as the most sensitive; O 3 induced a g s decrease in the range 34-64% for these CVs (Figure 2). The other δ 13 C-affected CVs presented a non-significant gs-reduction pattern. A more surprising behaviour was observed in the landrace group for Candeal de Vellisca, whose g s strongly increased (by 85%) from FA until NFA++.
The same individual analysis per cultivar was done for grain δ 15 N ( Figure 4a) and nitrogen concentration (Figure 4b). Individual patterns of the modern and old CVs showed the overall average trend of the CVs group: grain δ 15 N presented an increase in NFA+ and a subsequent drop by NFA++. Marius had the only significant differences in O 3 , decreasing from NFA to NFA++ by 0.56‰. Landraces showed a higher variability, and clear amplification of the quadratic response, especially for Chamorro and Candeal de Vellisca; meanwhile, Mocho Rojo presented a decreasing trend. Regarding grain N values (Figure 4b), increases were significant for Arthur Nick and Yecora; Aragon showed an increase until NFA+ but then decreased by NFA++.

Modern & Old CVs
Landraces Meanwhile, landraces did not present a significant correlation for δ 13 C against total biomass or GNY, but a significant (p < 0.05) negative correlation with GY (R 2 = 0.27) and highly significant correlation (p < 0.01) with g s (R 2 = 0.44). All δ 15 N and g s correlations for Modern & Old CVs and Landraces were non-significant.
Regarding other traits, the correlation between grain δ 13 C and HI, though still negative, was stronger for landraces (R 2 = 0.44) than for modern and old CVs (p < 0.01; R 2 = 0.18); SY, single seed weight (p < 0.05) and grain nitrogen (p < 0.01) only showed correlations under the landraces but not for the CVs, with R 2 values of 0.29, 0.24 and 0.52, respectively.
Regression functions for δ 13 C with total biomass, GY, GNY and g s are shown in Figure 5 specifying the different O 3 treatments with different symbols. It can be observed that the higher O 3 treatments also presented the lowest δ 13 C, with some exceptions such as for the tolerant CV Pane 247, which behaves closer to landraces and has not been included in the analysis. No significant correlations were found between grain δ 13 C and δ 15 N, or between grain δ 15 N or g s with any of the assayed parameters for both the CVs and landrace groups.

Principal Component Analysis
The Kaiser-Meyer-Olkin Measure of Sampling Adequacy and Bartlett's Test of Sphericity deemed the data for each of the three groups of genotypes (general), the modern and old CVs together and the landraces alone as useful for factor analysis ( Figure 6a). As harvest index and biomass already correlated strongly with GY and SY, respectively, they were removed to avoid redundancies. Nitrogen concentration was also removed as it had a component matrix under 0.5.
On the general biplot (Figure 6a), both components explained 77.8% of the total variance. Strong differences can be seen between the different age-type genotypes as shown by the separate ellipses. GY, GNY and number of grains per ear were associated positively with g s , and negatively with grain δ 13 C, grain δ 15 N, single seed weight and SY. The ellipse embracing the CVs is associated with higher yields, and, to a lower extent, a higher g s , while the ellipse embracing the landraces is associated with a higher SY and higher grain δ 13 C values. Seed weight does not seem to be generally associated with the other variables, only slightly positively with δ 13 C and SY.  When the response to O 3 was considered in the PCA analysis (Figure 6b), modern and old CVs differed from landraces. Landraces showed a great overlap of the ellipses of the different O 3 treatments, while modern and old CVs together show a deviation under the NFA++ treatment. For the modern and old CVs together, both components explained 61.8% of the total variance. PC1 showed negative associations between grain δ 13 C, and GY, GNY, SY and number of grains per ear, whereas grain δ 13 C was positively associated with increasing ozone concentrations. In PC2, grain δ 15 N and single seed weight do not seem to be associated with the latter, but correlate positively to each other, though with a short grain δ 15 N vector.
In the landraces, the two components explained 70.5% of the variance. The PC distribution here mirrors the distribution seen in the general analysis, though SY is less correlated to all the other variables, and seed weight is negatively correlated to the isotopic signals.
For both groups of age-type genotypes, the grain δ 15 N component matrix values were extremely low, lying almost directly on the component 2 axis and following the lack of correlations seen before.

Discussion
Spain has the largest surface dedicated to cereal production in the Mediterranean basin [56]. This mainly rainfed, drought-adapted wheat is located in the middle of the Iberian peninsula, where the recorded O 3 values chronically remain high above the protection thresholds for crops and vegetation defined by the European Directive of Air Quality (2008/50/CE) [57]. As hourly O 3 frequently reaches and exceeds 60 nL L −1 during early spring [58], which coincides with the non-water limitation period in the area, it allows for maximum physiological activity and gas exchange in the crop and thus, for maximum O 3 absorption [17,26], increasing the O 3 -risk for the wheat despite the characteristic water-deficit later in the season.
During the experiment, the FA O 3 values, in the range 20-30 nL L −1 , well reproduced the natural background levels currently found in the area during the winter and early fall seasons, which are considered as natural O 3 background levels for pre-industrial times [59]; thus, this treatment can be referred to as an adequate control treatment. The NFA treatment represented the daily and seasonal pattern and levels currently registered in the rural areas of the central Iberian Peninsula during spring and summer. It increases from early April until the end of the wheat season, and spans no longer than June, reaching values around 55 nL L −1 at the central hours of the day. Similarly, the O 3 levels in the fumigated treatments, NFA and NFA++, reached maximum hourly values of around 90 nL L −1 in May, reproducing the maximum hourly values sporadically achieved during the high O 3 episodes in the area [57]. Therefore, the experiment could be considered to adequately represent the poor air quality conditions under which rainfed cereals are growing in the area, and to represent the situation in the Mediterranean basin [60]. This chronic O 3pollution is enough to experimentally induce negative effects on crops such as tomato [61], leafy crops [62], watermelon [63] or durum wheat [36]. The latter is more O 3 -tolerant than the spring bread wheat assayed in this study [64]; in fact, bread wheat is currently considered the most economically important sensitive crop and is used as a reference crop for O 3 risk assessment [9]. Air meteorological conditions during the experiment also resembled the characteristic spring-summer conditions of the Mediterranean wheat fields, with temperatures increasing during the season and reaching maximum monthly mean values of around 35 • C at the central hours of the day in June, towards the end of the assay, along with the lowest HR value of around 25%.
Results showed strong differences between age-type genotypes, where breeding improved both yield and yield components; for CVs, these effects were strongly related to a decreased carbon isotope composition and an increase in g s . Landraces and CVs differed significantly; however, old and modern CVs behaved similarly. These results follow previous works performed under field conditions [30], which compared 12 Spanish varieties of soft and durum wheat under different tillage, also considering the age-type factor. Here, landraces had the lowest GY and behaved rather steadily throughout the trial seasons, while CVs presented a higher yield but depended more on the environmental conditions of the year and their independent growth habit. Winter genotypes, such as Marius, presented a higher GY in the humid years, and spring genotypes such as Califa Sur and Yecora performed comparatively better in the dry years. In the present study, no differences were seen due to growth habit since water availability was not a limiting condition in the experimental design and the date of sowing and transplantation was homogenized for all genotypes; however, CVs presented a 34% higher GY than landraces (across all O 3 treatments), which is similar to the 32% observed in Ruiz et al., 2019 [30] across seasons. Nonetheless, differences were seen between genotypes released pre-and post-green revolution (Table 1), particularly displaying the different behaviour Pane-247 had from the other CVs due to its pre-green revolution release. In this experiment, its very high δ 13 C and low g s fit much better within the landraces group, which was also released pre-green revolution, causing its removal when all genotypes were joined to analyse the general O 3 response. Actually, strong increases in GY and HI characterized the green revolution; however, further increases in these traits during the following six decades have been comparatively lower, explaining the lack of clear differences between old and modern CVs.
Achieving a higher GY in modern and old CVs was possibly not only due to a higher HI, but also a higher g s , which favoured photosynthetic activity and caused the more negative δ 13 C values in the newer genotypes [31,42,65]. Therefore, when compared to the landraces, the higher g s of the post-green revolution cultivars (Table 1) allowed for increased discrimination, favouring the lighter 12C isotope and diminishing the δ 13 C by 2.13‰. A decrease in the δ 13 C (or increase in ∆ 13 C) has been reported before for Spanish and Italian modern durum wheat cultivars when compared with old cultivars and landraces [42,65], where the difference in δ 13 C between some modern genotypes and landraces from the Iberian Peninsula was around 1.2‰.
However, it should be noted that the higher δ 13 C values seen in the landraces could also be influenced by their longer phenology when compared with the CVs, which exposes them to a more severe terminal (i.e., during grain filling) stress, even under irrigated conditions. The higher g s observed in the more modern genotypes might seem contradictory for the adaptation process required for the characteristic water limited conditions of Mediterranean rainfed wheat; however, an increase in transpiration would not only aid photosynthesis, but would also limit leaf temperature, which is another important stress that affects plant physiological activity in hot and dry areas [66].
The effect of breeding on increasing δ 15 N and GNY would be associated with the higher assimilation of the more modern varieties, while the observed reduction in grain N% would be related to a grain dilution effect, providing N availability is constant for the plant [28]. In essence, as GY increases with breeding, N% is diluted in the grain when paired with a constant N availability.
The PCA for the combined modern plus old CVs group and the landraces group (Figure 6a) reinforced the aforementioned two very different behaviours. The PCA pattern of the analysed parameters, except for seed weight, which did not correlate well with the others, showed the expected behaviours for the two age-groups, where the yield increase (GY and GNY) linked to breeding selection was positively associated with g s , and negatively associated with the isotopic δ 13 C and δ 15 N signals and SY. This behaviour is opposite to that of the landraces.
In terms of O 3 -response, the effect was age-type dependent; while modern and old CVs presented an O 3 -induced reduction in yield parameters and g s , as well as an alteration of both isotopic signals (less negative δ 13 C and lower δ 15 N), landraces were not affected by O 3 exposure, even under the highest levels of the NFA++ treatment. The similarity in the O 3 response of the post-green revolution CVs (modern and old CVs without Pane 247 as shown in Table 1) was striking and allowed them to be analysed jointly. Exposure to O 3 induced high yield losses, 17% and 11 % for GY and GNY, respectively, as an average value across CVs in the NFA++ treatment. A meta-analysis derived by Feng et al., 2008 [67], considering more than 50 studies, obtained mean GY losses of 29% in modern wheat varieties for O 3 -exposures based on Accumulated Ozone Exposure (AOT) functions that were slightly higher than the present NFA++ treatment. Meanwhile, in a review for Europelevel O 3 effects, similar O 3 exposures between 80-100 ppb produced yield reductions that rounded 30% [68]. Compared with these results, the Spanish CVs seem to lie within the O 3 -sensitivity range observed for modern wheat CVs from other agronomic regions and for Central and North European levels, which do not have the characteristic water limitation of the Mediterranean region.
The O 3 effects on yield parameters of the sensitive CVs paralleled those of the g s reduction. The greater sensitivity of modern CVs, compared to traditional genotypes not subjected to a programmed selection process, has been observed in previous works on wheat [28,32]. All of them showed smaller O 3 -effects on the lower yielding old cultivars, and associated breeding selection for higher yield with an increased gas exchange via higher g s , which led to an unintentional greater O 3 absorption and damage. Other studies have defined some of the O 3 -damage mechanisms that affect modern varieties, which relate the O 3 induced lower photosynthetic rates to non-stomatal factors, such as a decreased carboxylation or electronic transport and lower capability of the antioxidative system [69].
Previous studies have concluded that, for wheat, the most O 3 -sensitive period is anthesis, as O 3 -effects on previous stages were not statistically significant [37]. Therefore, the present experiment was designed to start fumigation after spike emergence, close to anthesis. Accordingly, the overall negative O 3 -effects were detected more in grain parameters than in biomass parameters, which follow the observed decrease in HI induced by the pollutant. The GY reduction was caused mainly by smaller grains while maintaining the number of grains per ear, also following results found in previous studies [68].
While this experiment followed many of the yield related O 3 -effects seen in previous studies, few studies have delved into the effects O 3 has on the isotopic signature of wheat [25,40], and none have sought this for Mediterranean wheat. In experimental assays, gas exchange and carbon fixation factors affected by O 3 are often measured sporadically, and the results are extrapolated to the entire exposure period; even though these parameters usually present daily and seasonal patterns depending on the plant development stage [28], which might lead to a misinterpretation of the O 3 response. As δ 13 C in plant dry matter depends on both stomata and Rubisco carboxylation [46], if the stomata close or their conductance decreases due to any stress, such as high O 3 , stomatal discrimination decreases, restricting the CO 2 supply to the carboxylation sites and overall increasing δ 13 C (or decreasing ∆ 13 C) [39,40]. The integrated effect on the physiological activity of the plant throughout its lifespan, for whichever metabolic route the pollutant has affected, can be assessed by considering the isotopic parameters [43].
Results of the present work confirm the starting hypothesis that the δ 13 C of mature grains becomes less negative (increases) under O 3 stress for O 3 -sensitive CVs, due to the g s reduction induced by the pollutant, which both decreases the capacity of the plant for carbon discrimination and negatively affects photosynthesis, decreasing GY. On the contrary, the landraces O 3 -tolerance correlates with the non-effect seen in δ 13 C of mature grains, along with a decrease in g s and yield parameters. This also follows the postulate that, under stress conditions, cultivars use their current photosynthetic activities to produce carbohydrates for grain filling, while landraces rely more on storing carbohydrates in the stem to then redistribute them to the grain [28,29]. This could suggest that if the landraces had been exposed to ozone earlier, when they were storing carbohydrates in the stem, they could have shown a more sensitive behavior in their GY, which could be analyzed with an experiment with chronic ozone exposure beginning before anthesis.
Both the regression functions and the PCA analysis pointed to these results. It should be noted that a negative correlation between δ 13 C of mature grains and g s was observed within CVs and landraces; however, for the latter there was no direct relationship with the pollutant. The observed results of the PCA for each of the O 3 treatments (Figure 6b) reiterate the commented O 3 -tolerance of the landraces, and the sensitivity of the modern and old CVs, particularly under the NFA++ treatment related to the decreased g s , and hence, in 13 C discrimination (and thus higher δ 13 C), and lower GY, GNY, SY and number of grains per ear. Meanwhile, for both CVs and landraces, this effect is not shown in single seed weight, nor grain δ 15 N, most possibly due to its more complex isotopic fractionation processes [47].
The general response pattern of the grain N concentration, which increases under increasing O 3 levels, was more evident under the sensitive CVs, presenting a 7% increase due to the pollutant. This aligned with previous studies that analysed wheat grown under O 3 stress [28,70], and has also been seen with drought [71]. As senescence is accelerated by O 3 -stress, it would favour N translocation to the reproductive parts, enriching the grain more with protein than starch [72]; moreover, as the individual grain becomes smaller, it further increases the concentration [19]; however, when coupled with a decreasing GY, it actually leads to an overall decrease in GNY.
In general, previous studies on wheat indicate that when there is a positive O 3 -effect on N grain concentration, there is an approximately equally negative effect on the yield [73]. Therefore, even though grain quality improves, the amount of protein accumulated per unit area, related to GNY, decreases, which can lead to food security and economic problems. In the present study, the effect on GNY was softer than the one observed in yield loss, with a GNY reduction of 11% compared to the 17% found for GY. This could be because O 3 exposure began close to anthesis, and even though this period is the most O 3 -sensitive for wheat [37], it does not match the main period for N absorption, which usually decreases greatly after anthesis and during grain filling [50]. At anthesis, wheat focuses more on N redistribution from the vegetative tissues (e.g., flag leaf) for grain filling [47]. Therefore, a reason for a lack of stronger differences regarding the N-related parameters could be because N uptake from the soil had greatly decreased before the O 3 treatment began. Nonetheless, the landraces are characterized by a later phenology, in the sense that heading and anthesis occur later than in CVs; however, in the present study, all the genotypes were homogenized for growing so differences caused by their growth habit would not apply here.
As landraces rely more on ears, which are generally less affected by stress, and on protein degradation from other plant organs during grain filling stages [49,74], the O 3effect on assimilation and redistribution could be particularly detrimental to the sensitive CVs when under stress; since they rely more on the flag leaf for N translocation for grain filling which senesces faster under O 3 exposure. At the same time, as active nutrient uptake by the roots occurs at early spring, future experiments designed to begin O 3 fumigation at the pre-anthesis stage are needed to take O 3 effects on N into consideration, especially for rainfed wheat.
Concerning the δ 15 N of mature grains, as the plants were grown in pots with artificial substrate and had the same N fertilizer source (with a given δ 15 N), the starting pattern of assimilation and further fractionation is expected to be very similar for all the plants. δ 15 N values are largely driven by nitrogen metabolism, including nitrogen assimilation, recycling and redistribution [38,[47][48][49]. When N moves from one organ to another, the δ 15 N signal in the organ increases as it builds up moving upwards towards the grain [47]. Therefore, alterations in plant metabolism and growth can modify δ 15 N. However, in our study, the assimilated N was accumulated in the flag leaf before the O 3 fumigation treatment began [49]. Due to the low levels of N absorption after anthesis [73], this suggests that only processes related to the N translocation from the leaves to the grains during grain filling were responsible for the effect of O 3 on δ 15 N, relating the O 3 response of the N parameters, such as δ 15 N, more with N redistribution.
The general response of δ 15 N of the grain to the pollutant followed a non-linear pattern, upon which it increased from the control to the NFA+ treatment, to then being decreased by NFA++. Here, the effect of O 3 on δ 15 N of grains was significantly driven by the CVs as opposed to the landraces and could again be related to the higher O 3 -sensitivity of the CVs as well as on their dependence on N redistribution from organs that are highly affected by O 3 , such as the flag leaf. Meanwhile, as mentioned, N redistribution from the ears, which is more associated to landraces, works well both with and without stress [49], giving landraces an advantage in redistribution under O 3 -stress. The negative correlation between δ 15 N and GY has been observed before [42], although it was not significant in the present assay. An alternative explanation for the decrease in δ 15 N of the CVs as a response to the more severe O 3 treatment might be related to the decrease in g s , which would limit volatilization of the lighter 14 N isotope with relation to the heavier 15 N [47,49,75], in turn leading to a lower δ 15 N of mature grains.
It should also be noted that the landraces displayed a large heterogeneity in the δ 15 N signal compared to the CVs, which could be due to their higher genetic variability [76] coupled with the aforementioned N redistribution from the ears and protein degradation of other plant parts [49]. The importance of the N redistribution mechanisms in the grain N should be considered for future plant breeding programs.

Conclusions
Modern and old CVs behaved very similarly under O 3 exposure, showing a high O 3 -sensitivity based on the significant impacts seen on yield parameters, such as GY or GNY, and biomass, and, accordingly, on g s and δ 13 C and δ 15 N values. On the contrary, landraces showed a stable and tolerant behaviour under the increasing O 3 exposures.
A different behavior was found between genotypes depending on their release date, where pre-green revolution genotypes presented very high δ 13 C and low g s. δ 13 C grain correlated strongly with both stomatal conductance and yield factors such as GY, GNY or total biomass; therefore, δ 13 C grain , which can congregate all the intrinsic mechanisms by which the plant reacts to chronic O 3 exposure, and which hinder its yield, could be used as a good indicator of the cumulative effects of O 3 at the end of the wheat plant lifespan. Moreover, δ 13 C values presented a more homogeneous response within each age-type group than other physiological response parameters such as g s , backing its potential use as an indicator for accumulated O 3 effects.
Since the O 3 exposure period started post-anthesis, δ 15 N grain response was more indicative of alterations caused by the pollutant on N redistribution and N volatilization, and requires a more integrated future analysis due to its dependence on other N-related factors, such as absorption, metabolism and redistribution.
In the present study, the benefits of modern breeding through varietal selection mostly counterbalanced the negative O 3 effects on the modern and old CVs; however, the strong differences found in O 3 -sensitivity between CVs and landraces indicate that there is still a broad genetic base for improving O 3 -tolerance to enhance productivity and profits in Mediterranean agriculture, and to better adapt Mediterranean wheat to global change factors such as chronic high O 3 levels.  Table 1 can be found at https://genvce.org/productos/fichas/ (accessed on 4 July 2021).

Conflicts of Interest:
The authors declare no conflict of interest.
Appendix A Table A1. Ozone and age type effects on yield parameters: SY, straw yield (g pot −1 ); Single sw., single seed weight (mg); # grains ear −1 , number of grains per ear per pot; Biomass, added value of the straw yield and the ears with grains (g pot −1 ). FA= charcoal filtered air, NFA = non filtered air, NFA+ = non filtered air supplemented with 20 nL L −1 of O 3 , NFA++ = non filtered air supplemented with 40 nl L −1 of O 3 ; Mean values ± SE. Different letters indicate statistically significant differences (p < 0.05).