A golf hole cutter was utilized to eliminate the effect of errors in initial aboveground and belowground biomass (10 cm diameter × 20 cm long). Annual ryegrass was transplanted in the experimental farm at Hebei University of Engineering, Handan City, Hebei Province, China. Then, the collected grasses were transplanted into pots (10 cm diameter × 100 cm long) filled with fritted clay and moved to artificial climate chambers (Model BDP-2000, Ningbo Prandt Instrument Co., Ltd, China). We trimmed grasses every 30 days to a 5-cm canopy height during the 90 days experimental treatments to keep grass plants in good growth condition (Yu et al., 2012b).

Eight artificial climate chambers were utilized to automatically monitor and control CO₂, four of which were set as modern CO₂ (a[CO₂]; 400 μmol mol^-1) and the remaining four were set as elevated CO₂ (e[CO₂]; 800 μmol mol^-1). The environmental settings for all eight environmental growth chambers were at 25/20°Cday/night temperature, 800 μmol m^-2 s^-1 PAR canopy light intensity, 65% relative humidity, and a 12-h photoperiod of 7:00-19:00. To minimize confounding effects of environmental variation between two chambers, we changed the [CO₂] of each growth chamber every 7 days, and then relocated the CO₂ treated annual ryegrass plants to the growth chambers with corresponding [CO₂] during the whole experiment. In each artificial climate chamber, six randomly selected pots of annual ryegrass plants were watered to through-flow twice a week with half-strength Hoagland’s solution modified to generate six P concentration treatments of 0.004, 0.012, 0.02, 0.06, 0.1-, and 0.5-mM P as KH₂PO₄, respectively. To ensure that all grasses have the same amount of potassium kalium (K) in the nutrient solution at each watering, we add an additional moderate amount of KCL to supplement the K in the half-strength Hoagland’s solution. Four artificial climate chambers with a[CO₂] or e[CO₂] are biological replications (n = 4).

To characterize the maximum stomatal pore size of annual ryegrass, we selected recently expanded leaves for sampling stomatal imprints from the middle section on the abaxial surface using colorless nail varnish in artificial climate chambers on the 30th, 60th, and 90th days after CO₂ treatment and P treatments (Zheng et al., 2013; Xu, 2015). We observed and photographed the collected imprints using the method of Zheng et al. (2013) and measured the stomatal aperture length (SAL), stomatal aperture width (SAW), stomatal aperture circumference (SAC) and stomatal aperture area (SAA) using the Image J quantification software (NH, Bethesda, MD). Stomata on each surface were counted and combined for calculating stomatal density (SD) (Ceulemans et al., 1995) and the stomatal aperture shape index (SASI) was also calculated as SAA /SAC × 100% . The morphological traits of stomata were visualized and photographed with a scanning electron microscopy (FEI Corp, USA). We randomly selected four images (a magnification of 100) from each treatment to estimate the stomatal spatial distribution pattern. The selected images were digitized with a GIS software (ArcGIS 10.0; ESRI Inc., Redlands, CA). In this study, the center of each stoma was treated as a single point. Then the point pattern analysis was conducted with the Ripley’s K-function (Ripley, 1976). Comprehensive guidelines for the analysis of stomatal spatial distribution pattern can be found in Xu (2015) and Zheng et al. (2020).

A portable LI-6400 photosynthesis system (Li-Cor Inc., Lincoln, NE, USA) was utilized to determine the net photosynthetic rate (P _n), stomatal conductance (G _s) and transpiration rate (T _r) on recently expanded leaves on the 30th, 60th, 90th days after CO₂ treatment and P treatments. All measurements were performed in the standard cuvette chamber (2 cm × 3 cm) with the CO₂ concentration of 400 μmol mol^-1, the saturating light at 1000 μmol photons m^-2 s^-1, the leaf-to-air vapor pressure deficit (VPD) of 1.5 KPa and the temperature of 20°C. The intrinsic water use efficiency (WUE) was calculated as P _n/T _r.

The aboveground and belowground biomass were harvested using the physical cutting at the end of the 90-day experiment and oven-dried the separated tissues at 80°C to a constant weight. Finally, the data of biomass were weighed using an electronic scale. The aboveground and belowground portions were grinded to fine powder using a ball mill (MM2, Fa. Retsch, Haan, Germany). The tissue phosphorus (P), carbon (C), and nitrogen (N) contents of shoots and roots were determined using an elemental analyzer (Vario Max CN; Elemnetar Corp., Germany). All the biochemical analyses were repeated four times (n = 4).

Two-way analysis of variance (ANOVA) was utilized to test the interactive effects of P concentration and [CO₂] on plant biomass, stomatal traits, and leaf gas exchange as well as the contents of P, C, and N among different treatments (p < 0.05). Additionally, a three-way analysis of variance (ANOVA) was also used to estimate the interactive effects of [CO₂] × P × plant tissues on P, C, N contents (p < 0.05). Furthermore, we used linear and non-linear regressions to analyze the relationship between biomass and other variables (p < 0.05). All statistical analyses were conducted using the SPSS 20.0 software (Chicago, IL, USA).

Our results showed that the total plant biomass, aboveground biomass, belowground biomass and below/above biomass ratio of annual ryegrass were substantially changed by P supply (all p < 0.001), while e[CO₂] only remarkably affected the aboveground biomass (p = 0.014) and belowground biomass (p = 0.024; Figure 1 ). Specifically, e[CO₂] marginally decreased the total plant biomass by 6.6% at P_0.004 (p < 0.05), but the total plant biomass under the P concentrations of 0.02 Mm (P_0.02) and 0.06 mM (P_0.06) was obviously increased by 7.1% (p < 0.05) and 10.4% (p < 0.01) ( Figure 1A ). Moreover, elevated [CO₂] dramatically increased the aboveground biomass by 14.5% (p < 0.001) under the highest P concentration of 0.5 mM (P_0.5) ( Figure 1B ). By contrast, rising [CO₂] decreased the belowground biomass by 13.7% (p < 0.01), 9.2% (p < 0.05), and 11.1% (p < 0.05) under the P concentrations of 0.004 mM (P_0.004), 0.012 mM (P_0.012), and 0.5 mM (P_0.5) ( Figure 1C ). Similar to the changes in belowground biomass, elevated [CO₂] also changed the allocation of plant biomass between belowground and aboveground (below/above biomass ratio) with decreasing the below/above biomass ratio by 11.9%, 11.4%, and 18.5% at P_0.004 (p < 0.05), P_0.012 (p < 0.01), and P_0.5 (p < 0.05), while the below/above biomass ratio of annual ryegrass at P_0.006 was substantially enhanced by 31.5% under elevated [CO₂] (p < 0.001; Figure 1D ). Moreover, our one-way ANOWA results revealed that the total biomass (p < 0.001), aboveground biomass (p < 0.001), belowground biomass (p < 0.001), and the ratio of below/above biomass (p < 0.001) were significantly changed by the P deficiency, while e[CO₂] only marginally effected the aboveground biomass (p = 0.014) and belowground biomass (p = 0.024) of annual ryegrass ( Figure 1 ). Additionally, the remarkably interactive effects of P and [CO₂] were found on the total plant biomass, aboveground biomass, belowground biomass, and below/above biomass ratio of annual ryegrass ( Figure 1 ).

Sufficient P supply (P_0.5) had a strong CO₂ fertilization effect on the net photosynthetic rates of annual ryegrass, as evidenced by the 25.3% increase of net photosynthetic rates (p < 0.001) ( Figure 2A ). By contrast, the net photosynthesis rates of annual ryegrass at lower P supply of P_0.004, P_0.02, and P_0.06 were significantly decreased by 13.6% (p < 0.01), 19.8% (p < 0.001) and 16.9% (p < 0.001) under elevated [CO₂]. Meanwhile, elevated [CO₂] substantially reduced the stomatal conductance by 54.2% (p < 0.001), 56.0% (p < 0.001), 39.6% (p < 0.01), 47.3% (p < 0.01), and 36.7% (p < 0.05) at the P treatments of P_0.012, P_0.02, P_0.06, P_0.1, and P_0.5 except for the stomatal conductance under the P supply of P_0.004 ( Figure 2B ). However, elevated [CO₂] only increased the leaf transpiration rate at P_0.06 by 14.1% (p < 0.01), and barely affected the leaf transpiration rates under other P treatments ( Figure 2C ). Consequently, e[CO₂] dramatically reduced the water use efficiency of annual ryegrass by 13.1% (p < 0.05), 13.5% (p < 0.01), and 27.1% (p < 0.001) under lower P treatments of P_0.004, P_0.02 and P_0.06 ( Figure 2D ), whereas the water use efficiency at higher P treatments of P_0.1 and P_0.5 was significantly enhanced by 22.3% (p < 0.001) and 24.8% (p < 0.001; Figure 2D ). Moreover, the significantly interactive effects of P supply and e[CO₂] were also found on the net photosynthetic rates (p < 0.001), stomatal conductance (p = 0.043), transpiration rates (p = 0.032) and water use efficiency (p < 0.001) of annual ryegrass ( Figure 2 ).

Elevated [CO₂] substantially affected the stomatal density (SD) of annual ryegrass regardless of P supply ( Table 1 ). Specifically, e[CO₂] dramatically increased the SD by 72.3%, 34.8% and 25.6% under P_0.012, P_0.1, and P_0.5, whereas obviously decreased the SD by 17.1%, 27.1%, and 35.5%, respectively, at the P supply of P_0.004, P_0.02, and P_0.06 (all p < 0.05; Table 1 ; Figure 3 ). Moreover, elevated [CO₂] dramatically decreased the stomatal area (SA) by 13.7%, 12.5% and 11.5% at P_0.004, P_0.1, and P_0.5 (all p < 0.05), which may be due to the smaller stomatal length and width ( Table 1 ; Figure 4 ), and the minimum and maximum values of the stomatal area were occurred at P_0.012 and P_0.1, respectively ( Table 1 ; Figure 4 ). Our two-way ANOVA results showed that the SD of annual ryegrass was substantial changed by [CO₂] or P supply ( Table 2 ). Additionally, [CO₂] × P supply also significantly affected the SD, SAL, SAW, SAA, and SAA (all p < 0.05), but barely changed the SAC ( Table 2 ).

The spatial distribution pattern of annual ryegrass was also changed by P supply and e[CO₂] ( Figure 5 ). In general, the spatial pattern of stomata distributed on leaves of annual ryegrass followed a regular pattern at small scales (<150 µm) and a random distribution at larger scales (>200 µm) regardless of [CO₂] and P supply ( Figure 5 ). Interestingly, the most regular pattern both at the scale of c. 110 μm regardless of the [CO₂] concentration was observed in the current study, as evidenced by the average minimum Lhat(d) values of -9.24 under a[CO₂] and -8.00 under e[CO₂] ( Figure 5 ). Moreover, e[CO₂] produced more regular spatial patterns of stomata at small scales when annual ryegrass was subjected to three higher P supply of P_0.06, P_0.1, and P_0.5, due to the lower Lhat(d) values at the same spatial scales ( Figure 5 ). Meanwhile, elevated [CO₂] also increased the range scale of regular pattern of stomata ( Figure 5 ).

Elevated [CO₂] generally reduced phosphorus contents in both shoots and roots of annual ryegrass ( Table 3 ). Specifically, elevated [CO₂] significantly decreased the phosphorus in shoots by 33.9%, 15.2%, 18.9% and 12.7% under P_0.012, P_0.06, P_0.1 and P_0.5 (all p < 0.05; Table 3 ). However, e[CO₂] only obviously reduced the phosphorus content in roots by 19.9% under P_0.06 (p < 0.05; Table 3 ). In addition, e[CO₂] significantly decreased shoots N by 25.8% and 20.6% under P_0.004 and P_0.06 (both p < 0.05; Table 3 ), but increased shoots N by 19.1% under P_0.5 (p < 0.05; Table 3 ). Consequently, the C/N ratio in shoot was enhanced by 29.2% and 25.0% under P_0.004 (p < 0.05) and P_0.06 (p < 0.05; Table 3 ) but lowered by 17.9% under P_0.5 (p<0.05; Table 3 ). Additionally, the shoot C (p > 0.05), root C (p > 0.05), N (p > 0.05) and C/N ratio in roots (p > 0.05) were barely changed by e[CO₂] regardless of P supply ( Table 3 ).

The tissue P contents were substantially affected by [CO₂] (p < 0.05) or P supply (p < 0.001) from the results of three-way ANOVA ( Table 4 ). However, we found statistical differences in the contents of phosphorus (p < 0.001), C (p < 0.001) and N (p < 0.001) as well as the C/N ratio (p < 0.001) between tissues (shoot and root) of annual ryegrass ( Table 4 ). Moreover, there were obviously interactive effects of [CO₂] × P supply on the tissue N (p < 0.001) and P contents (p < 0.05) as well as the C/N ratio (p < 0.001; Table 4 ), whereas P supply × tissue only changed the phosphorus content of annual ryegrass (p < 0.001; Table 4 ). In addition, our results also showed that [CO₂] × P supply × tissue significantly changed the contents of N, P and the C/N ratio of annual ryegrass (all p < 0.05; Table 4 ).

The aboveground (R² = 0.80, p = 0.017; Figure 6D ) and total biomass (R² = 0.86, p = 0.008; Figure 6F ) demonstrated a linear increase with the elevation of leaf photosynthesis at e[CO₂]. However, no linear or parabolic relationships were found between leaf photosynthesis and aboveground biomass (R² = 0.13, p = 0.815; Figure 6A ), as well as total biomass at a[CO₂] (R² = 0.31, p = 0.252; Figure 6C ). Similarly, we found linear relationships of shoot phosphorus content between aboveground (R² = 0.73, p = 0.031; Figure 7A ) and total biomass (R² = 0.59, p = 0.076; Figure 7C ) at a[CO₂], as well as aboveground (R² = 0.64, p = 0.055; Figure 7D ) and total biomass (R² = 0.89, p = 0.005; Figure 7F ) at e[CO₂]. Nevertheless, there is no obvious correlation between belowground biomass and leaf photosynthesis ( Figure 6 ), shoot phosphorus content ( Figure 7 ), or shoot nitrogen content ( Figure 8 ) regardless of CO₂ concentration. Moreover, we also found parabolic relationship of shoot nitrogen content with leaf photosynthesis, irrespective of CO₂ concentration ( Figure 9 ).

Previous studies have established that since the current atmospheric [CO₂] is suboptimal for the Rubisco enzyme involved in leaf photosynthesis (Ainsworth and Rogers, 2007; Pandey et al., 2015). As a result, the “CO₂ fertilization effect” would benefit crops, given the enriched atmospheric [CO₂], leading to plant growth and crop yield (Sakurai et al., 2014; Xu, 2015). For example, a study on winter wheat indicated that increasing [CO₂] had a maximum boost of more than 50% on its biomass (Butterly et al., 2015). However, it should be noted that most of these studies focused on the effects of e[CO₂] on plant growth and physiological processes were conducted under sufficient nutrition supplies (Li et al., 2007). Therefore, this CO₂ fertilization effect is likely to be required more essential nutrients for sustaining plant growth (Plénet et al., 2000; Lee et al., 2011; Menge et al., 2012; Pandey et al., 2015; Zheng et al., 2017). Increasing CO₂ from 400 to 800 μmol mol^-1 only substantially enhanced the aboveground biomass of annual ryegrass by 14.5% at P_0.5, indicating that e[CO₂] indeed boosts the plant growth of annual ryegrass with sufficient P supply. However, the aboveground biomass of annual ryegrass subjected to P deficiency was barely affected by elevated CO₂ and total biomass even obviously decreased by 6.6% at P_0.004, which suggested that soil P deficiency down regulated the favorable impacts of CO₂ fertilization effect on annual ryegrass. Moreover, we also found that the belowground biomass was substantially increased under P deficiency, indicating that plants may preferentially distribute more biomass to the roots for nutrient uptake when subjected to P limitation, which is consistent with the conclusions from previous studies that P deficiency may favor root growth more than shoot growth, and thus result in a higher below/above biomass ratio (Péret et al., 2011; Pandey et al., 2015). Additionally, the pronounced interactive effect of CO₂ concentration and P supply on the aboveground, belowground and the total biomass of annual ryegrass was evident in two-way ANOVA results. The total biomass and aboveground biomass also increased with the increase of P supply, suggesting that the positive CO₂ fertilization effect on the growth of annual ryegrass was triggered by P supply and the stimulation of biomass accumulation by CO₂ depends on the P status.

It is well known that CO₂ is one of the key reactants needed by plants to engage in the biochemical process of photosynthesis, and elevated CO₂ stimulates leaf photosynthesis (Kimball et al., 2002). However, several lines of evidence suggest that the increased photosynthesis associated with e[CO₂] may be diminished during prolonged exposure, particularly in plants limited by nutrient availability (Lauer et al., 1989; Campbell and Sage, 2006; Reich et al., 2006; Pandey et al., 2015). In this study, e[CO₂] substantially enhanced the net photosynthetic rate at P_0.1 and P_0.5 was observed, whereas the net photosynthetic rate of annual ryegrass subjected to P deficit was barely affected or even obviously decreased by elevated CO₂ concentration, indicating that the stimulating effect of e[CO₂] on the photosynthetic response of plants is weakened as the P concentration decreased. This result may be explained by the fact that high CO₂ improves uptake efficiency in the presence of adequate P supply (Nilsson et al., 2010; Jin et al., 2011; Pandey et al., 2015). Moreover, Duchein et al. (1993) found in a research on clover (Trifolium subterraneum L.) that e[CO₂] obviously increased its net photosynthetic rate under a P concentration supply of 2 mM, while it was inhibited when subjected to P limitation, which was similar to our study. In addition, we found that the highest P supply (0.5 mM) further enhanced the photosynthetic rate under e[CO₂], as evidenced by the higher increase of net photosynthetic rate at P_0.5 than that of plants treated with P_0.1, indicating that plants will probably demand ultra-optimal levels of P supply if they want to benefit from the general trend of increasing atmospheric CO₂ concentration in the future. Moreover, the aboveground biomass of annual ryegrass was barely affected by e[CO₂] when the P supply was lower, although the net photosynthetic rates were higher under P deficit. This suggested that the CO₂ fertilization effect is more pronounced in stimulating the growth of annual ryegrass under sufficient P supply. Additionally, the obvious decrease in the net photosynthetic rate at elevated CO2, compared to ambient CO_2, under the lowest P supply (0.004 mM) likely indicates the crucial role of stomatal limitation on photosynthesis. Furthermore, one of the most consistent responses of plants to elevated atmospheric CO₂ is a decrease in stomatal conductance (Fleisher et al., 2012; Singh and Reddy, 2014; Zheng et al., 2019; Zheng et al., 2020). This aligned with our results, as we observed a remarkable decline in stomatal conductance with increasing CO₂ regardless of P supply. However, other nutrient studies have reported that reduced stomatal conductance under lower P supply did not seem to be the major reason for the limitation of photosynthesis (Jin et al., 2011; Singh et al., 2013a). From the above discussion, reduced photosynthesis may be a mechanism by which crops cope with soil phosphorus limitation, which may largely contribute to lower biomass.

It has been reported that plant stomata normally exhibit a variety of short-term behavioral and long-term morphological reactions to CO₂ (Zheng et al., 2013; Xu, 2015), soil moisture (Sun et al., 2014), thus stomatal regulation is a potential mechanism for adaptation to the external environment during plant growth. In addition to CO₂ concentrations and soil moisture conditions, Sekiya and Yano (2008) also pointed out that phosphorus supply level can modulate the rate of stomatal production in plant epidermal cells, subsequentially influenced the SD. Our results have further confirmed that stomatal traits of annual ryegrass varied with the P supply and atmospheric CO₂ concentration. Our results showed that lower P supply reduced the stomatal width and stomatal area, suggesting that annual ryegrass improved their adaptability to different P situations through regulating their stomatal opening. We found that the response of annual ryegrass stomatal density under e[CO₂] depended on P concentration, i.e., stomatal density increased at higher P concentrations and decreased at lower P supply. This CO₂-induced decrease of stomatal density under lower P supply may explain the downregulation of leaf photosynthesis, since the SD partially determines the efficiency of CO₂ diffusion from the atmosphere to the mesophyll tissues (Jin et al., 2011). However, observations from previous studies regarding the influence of elevated CO₂ concentration on SD differed (Ryle and Stanley, 1992; Woodward et al., 2002; Marchi et al., 2004). Regarding this inconsistency, Gray et al. (2000) pointed out that interspecific differences may play a role. However, it should be noted that, as revealed in this study, we cannot deny the strongly interactive effects between environmental variables. Furthermore, it has been well demonstrated that stomatal distribution patterns can affect the net photosynthetic rate and transpiration rate (Soares et al., 2008; Zheng et al., 2013). In the current research, we found that e[CO₂] made the spatial distribution pattern of stomata more regular under sufficient P supply, while the opposite was true when annual ryegrass was subjected to P deficiency. This may partially explain why increasing CO₂ in this study did not enhance or even lowered the net photosynthetic rate under a lower P supply. This is because the spatial distribution of the stomata on the blade surface affects the diffusion distance of carbon dioxide between the stomata (Zheng et al., 2013; Xu, 2015), which means that the more regular the spatial distribution of the stomata, the more efficient the blade is in terms of gas exchange. Overall, these results implied that P supply partially decided the response of the stomatal distribution pattern to e[CO₂].

It is now well established from a variety of studies that P limitation and e[CO₂] are likely to alter the distribution patterns of tissue constituents as well (Taub and Wang, 2008; Singh et al., 2013a). However, previous studies on leaf P concentration have had contradictory results (Fangmeier et al., 1999; Lewis et al., 2010). For instance, Lewis et al. (2010) reported that leaf P content of Populus deltoides was reduced by 22.2-48.6% with increasing the CO₂ concentration from 350 to 700 μmol mol^-1. While Fangmeier et al. (1999) found that e[CO₂] did not significantly lower the P concentration in the leaves of wheat. These inconsistencies suggested that the potential complexity of the effects of e[CO₂] on P nutrition. In the current study, we found that the biomass - net photosynthetic rates relationship followed a similar linear or bell-shaped curve like the biomass-shoot P content relationship at the e[CO₂] ( Figures 6 and 7 ). It shows that e[CO₂] did not enhance the net photosynthetic rate at a lower P supply, possibly owing to the decrease in leaf P content. Additionally, other studies also suggest that the decrease in nitrogen content in leaves may also be attributed to the decrease in photosynthesis, as the leaf N content is tightly correlated with the content of Rubisco enzymes (Zheng et al., 2019). Moreover, the P content in shoots dramatically decreased under P limitation, but the differences in roots under 0.001 mM - 0.1 mM P supply were mostly insignificant, indicating that shoots may be more sensitive to P limitation than roots.