Evaluating the response of δ 13 C in Haloxylon ammodendron , a dominant C 4 species in Asian desert ecosystems, to water and nitrogen addition as well as the availability of its δ 13 C as an indicator of water use efﬁciency

. Variations in precipitation and atmospheric N deposition affect water and N availability in desert and thus may have signiﬁcant effects on desert ecosystems. Haloxylon ammodendron is a dominant plant in Asian desert, and addressing its physiological acclimatization to the changes in precipitation and N deposition can provide insight into how desert plants adapt to extreme environments by physiological adjustment. Carbon isotope ratio ( δ 13 C) in plants has been suggested as a sensitive long-term indicator of physiological acclimatization. Therefore, this study evaluated the effect of precipitation change and increasing atmospheric N deposition on δ 13 C of H. ammodendron . Furthermore, H. ammodendron is a C 4 plant; whether its δ 13 C can indicate water use efﬁciency (WUE) has not been addressed. In the present study, we designed a ﬁeld experiment with a completely randomized factorial combination of N and water and measured δ 13 C and gas exchange of H. ammodendron . Then we calculated the degree of bundle-sheath leakiness ( ϕ ) and WUE of the assimilating branches of H. ammodendron . δ 13 C and ϕ remained stable under N and water supply, while N addition, water addition and their interaction affected gas exchange and WUE in H. ammodendron . In addition, δ 13 C had no correlation with WUE. These results were associated with the irrelevance between δ 13 C and the ratio of intercellular to ambient CO 2 concentration ( c i / c a ), which might be caused by a special value (0.37) of the degree of bundle-sheath leakiness ( ϕ ) or a lower activity of carbonic anhydrase (CA) of H. ammodendron . In conclusion, δ 13 C of H. ammodendron is not sensitive to global change in precipitation and atmospheric N deposition and cannot be used for indicating its WUE.


Introduction
Recently, global precipitation pattern has changed significantly (Frank et al., 2015;Knapp et al., 2015), and atmospheric N deposition has continued to rise (Galloway et al., 2004;Liu et al., 2013;Song et al., 2017). Previous researchers have suggested that arid ecosystems are most sensitive to climate change (Reynolds et al., 2007;Huang et al., 2016), while global change in precipitation and atmospheric N deposition has an important impact on water and N availability in desert (Huang et al., 2018). Thus, these changes may have significant effects on desert ecosystems. Haloxylon ammodendron is a dominant species in desert regions, especially in Asia. Studying the physiological responses of H. ammodendron to global change can provide insight into Z. Chen et al.: Evaluating the response of δ 13 C in Haloxylon ammodendron how desert plants adapt to extreme environments by physiological adjustment. Carbon isotope ratio (δ 13 C) in plants depends on the ratio of intercellular to ambient CO 2 concentration (c i /c a ), which reflects the balance between inward CO 2 diffusion rate, regulated by stomatal conductance (g s ), and CO 2 -assimilating rate (A) (Farquhar and Richards, 1984) and has been suggested as a sensitive long-term indicator of physiological acclimatization (Battipaglia et al., 2013;Cernusak et al., 2013;Tranan and Schubertt, 2016;Wang and Feng, 2012). Therefore, investigating the variations in δ 13 C of H. ammodendron under water and nitrogen addition can enhance understanding of physiological responses of desert plants to future changes in precipitation and atmospheric N deposition.
A large quantity of works have been devoted to the relationships between C 3 plant δ 13 C and water availability or precipitation (e.g., Diefendorf et al., 2010;Kohn, 2010;Liu et al., 2005;Ma et al., 2012;Serret et al., 2018;Stewart et al., 1995;Wang et al., 2005Wang et al., , 2008 and nitrogen availability (e.g., Cernusak et al., 2007;Li et al., 2016;Sparks and Ehleringer, 1997;Yao et al., 2011;Zhang et al., 2015). However, a relatively small amount of research has focused on the responses of C 4 plant δ 13 C to water availability or precipitation (Ellsworth et al., 2017;Liu et al., 2005;Rao et al., 2017;Wang et al., 2006) and nitrogen availability (Ma et al., 2016;Schmidt et al., 1993). For C 4 plants, δ 13 C is controlled by both the c i / c a ratio and the degree of bundlesheath leakiness (ϕ), the proportion of CO 2 produced within bundle-sheath cells from C 4 acids that leaks back to mesophyll cells (Ellsworth and Cousins, 2016;Ellsworth et al., 2017;Farquhar, 1983). Thus, the responses of C 4 plant δ 13 C to water and N availability are also affected by ϕ. Genetic factors control ϕ values, which causes the interspecific differences in δ 13 C, even the responses of plant δ 13 C to water and N availability (Gresset et al., 2014). On the other hand, enzymatic activity of carbonic anhydrase (CA) may influence δ 13 C in C 4 plants (Cousins et al., 2006). CA is an enzyme that catalyzes the hydration of CO 2 in mesophyll cells to form bicarbonate . Previous studies showed that CA activity in most C 4 plants is usually low, just sufficient to support photosynthesis (Cousins et al., 2006;Yakir, 2000, 2001;Hatch and Burnell, 1990). H. ammodendron is a typical C 4 plant. How its δ 13 C responds to water and N availability has never been addressed.
Foliar δ 13 C in C 3 plants has been considered as a useful indicator of intrinsic water use efficiency (WUE) (Farquhar, 1983). However, although some studies have suggested that δ 13 C of C 4 plants could also indicate its WUE (Henderson et al., 1992;Wang et al., 2005;Cernusak et al., 2013;Ellsworth and Cousins, 2016), this statement is still controversial. The relationship between δ 13 C and WUE is based on the links between c i / c a ratio and δ 13 C and between c i / c a ratio and WUE (Ehleringer and Cerling, 1995). For C 3 plants, δ 13 C always decreases with an increase in c i / c a ratio; but for C 4 plants, the correlation between δ 13 C and c i / c a ratio de-pends on the ϕ value (Cernusak et al., 2013) and CA activity (Cousins et al., 2006). As mentioned above, ϕ value is under genetic control, and the CA activity changes across species (Cousins et al., 2006;Yakir, 2000, 2001;Hatch and Burnell, 1990); thus, the correlation between δ 13 C and c i / c a ratio, as well as the relationship between WUE and δ 13 C, shows interspecific difference. Whether δ 13 C of H. ammodendron indicates WUE has never been evaluated.
In this study, we designed an experiment with multiple water and nitrogen supplies in the southern Gurbantünggüt Desert in Xinjiang Uygur Autonomous Region, China. We measured the δ 13 C, gas exchange and WUE of the assimilating branches of H. ammodendron. We had two objectives. One objective was to evaluate the response of the dominant plant of Asian desert to future changes in precipitation and atmospheric N deposition by revealing the effects of water and N supply on δ 13 C of H. ammodendron. The other was to explore the availability of δ 13 C as the indicator of water use efficiency in H. ammodendron.

Definitions and basic equations
Stable carbon isotopic ratio (δ 13 C) of natural materials is expressed as where ( 13 C / 12 C) sample and ( 13 C / 12 C) standard are the 13 C / 12 C ratio of the sample and of the Pee Dee Belemnite (PDB) standard, respectively. Farquhar (1983) proposed the pattern of carbon isotopic discrimination ( ) in C 4 plants: = δ 13 C air − δ 13 C plant 1 + δ 13 C plant /1000 ≈ δ 13 C air − δ 13 C plant where δ 13 C plant and δ 13 C air are the δ 13 C values of plants and CO 2 in the ambient air, respectively. The parameter a (= 4.4 ‰; Craig, 1954) is the carbon isotopic fractionation in the diffusion of CO 2 into internal leaves; b 4 (= −5.9 ‰; O'Leary, 1984) is the combined carbon isotopic fractionations occurring in the processes of gaseous CO 2 dissolution, hydration-dehydration reactions of CO 2 and HCO − 3 in mesophyll cells, and HCO − 3 carboxylation by PEP (phosphoenolpyruvate) carboxylase; s (= 1.8 ‰; O'Leary, 1984) is the carbon isotopic fractionation during diffusion of CO 2 out of the bundle-sheath cells; and b (= 27 ‰; Farquhar and Richards, 1984) is the carbon isotopic fractionation of CO 2 carboxylation by RuBP (ribulose-1,5-bisphosphate) carboxylase. The variable ϕ is the proportion of CO 2 production within bundle-sheath cells from C 4 acids that leaks back to mesophyll cells, and c i / c a is the ratio of intercellular to ambient CO 2 concentration. Eq. (2) can be transformed into the following format: is greater than 0, δ 13 C decreases with increasing c i / c a ; if this coefficient is lower than 0, δ 13 C increases with increasing c i / c a . Water use efficiency (WUE) is defined as the amount of assimilated carbon dioxide by plants under the consumption per unit of water. There are two characteristics of WUE, instantaneous WUE (ins-WUE) and intrinsic WUE (int-WUE). Ins-WUE can be calculated by where e i and e a are the water vapor pressure inside and outside the leaves, and p is the atmospheric pressure. The definition of int-WUE is where g s is stomatal conductance.

Study site and species
This experiment was conducted at the Fukang Station of Desert Ecology, Chinese Academy of Sciences, on the southern edge of the Gurbantünggüt Desert (44 • 26 N, 87 • 54 E) in northwestern China. The altitude of the study site is 436.8 m above average sea level (a.s.l.). It is a typical continental arid, temperate climate, with a hot summer and cold winter in the area. The mean annual temperature is 7.1 • C, and the mean annual precipitation is 215.6 mm, with a potential evaporation of about 2000 mm. The mean annual temperature and the annual rainfall amount in the sampling year are 10.23 • C and 122.7 mm (Cui, 2018). The soil type is grey desert soils (Chinese classification) with aeolian sands on the surface (0-100 cm). The percentages of clay (< 0.005 mm), silt (0.005-0.063 mm), fine sand (0.063-0.25 mm) and medium sand (0.25-0.5 mm) range from 1.63 %-1.76 %, 13.79 %-14.15 %, 55.91 %-56.21 % and 20.65 %-23.23 %, respectively . The soil is highly alkaline (pH = 9.55 ± 0.14) with low fertility. The vegetation is dominated by Haloxylon ammodendron and Haloxylon persicum with about 30 % coverage. Herbs include ephemerals, annuals and small perennials, with a cover of ca. 40 % (Fan et al., 2013). Although the coverage of the two Haloxylon species is a little lower than that of herbs, the biomass of the former is much larger than that of the latter, because Haloxylon plants are shrubs with an average height of 1.5 m whereas the latter are very low herbaceous plants. Biological soil crusts are distributed widely on the soil between the herbs and Haloxylon, with almost 40 % coverage (Zhang et al., 2007). The present study focused on Haloxylon ammodendron because it is the dominant species in Asian desert. H. ammodendron is a species of Chenopodiaceae, which is a xerophytic and halophytic woody plant (Cui et al., 2017). The leaves of H. ammodendron have been completely degraded due to the extreme drought, and the assimilation branches, which are the glossy green branches (Fig. S1), perform the same functions as the leaves. Due to its drought tolerance, H. ammodendron is widely distributed in desert areas.

Experimental design
A field experiment with a completely randomized factorial combination of water and nitrogen has been conducted from 2014 to 2017. We designed two water addition levels (0, 60 mm yr −1 ; W0, W1) based on the prediction that precipitation will increase by 30 % in northern China in the next 30 years (Liu et al., 2010), and three levels of N addition (0, 30, 60 kg N ha −1 yr −1 ; N0, N1 and N2), because N deposition has reached 35.4 kg N ha −1 yr −1 in the nearby city Ürümqi (Cui et al., 2017) and will double by 2050 relative to the early 1990s (Galloway et al., 2008). Therefore, there were six treatments (W0N0, W0N1, W0N2, W1N0, W1N1 and W1N2) in this experiment. Four replicates of each treatment were set, making a total of 24 plots with a size of 10 m × 10 m. A small sub-plot with a size of 1.5 m × 1.5 m was set in each plot. A well-grown H. ammodendron was enclosed in the center of the sub-plot. The average height and coverage of an individual H. ammodendron were 1.5 m and 1.9 m 2 , respectively, and did not vary significantly across the plots. The type of nitrogen used in the present study is NH 4 NO 3 . To simulate natural water and N inputs, the treatments were applied in equal amounts, 12 times, once a week in April, July and September, as 5 mm m −2 of water and 2.5 or 5 kg N ha −1 each week (Cui et al., 2017). Usually, water addition was with a sprinkler kettle, irrigating over the canopy of H. ammodendron.

Measurements of gas exchange and WUE
As mentioned above, the main assimilating organ of H. ammodendron is the assimilation branches. Thus, we conducted gas exchange measurements on the assimilation branches of the H. ammodendron grown in the sub-plots by a LI-6400 portable photosynthesis system. The measurements were conducted on 27-29 June 2016, which is the main growing season of H. ammodendron. It may be most appropriate to take measurements during this period, and the results of the measurements are therefore more representative. Previous studies have also usually conducted this measurement during the growing season (Nyongesah and Wang, 2013;Cui, 2018;Gong et al., 2019). The ins-WUE was calculated based on these measured gas exchange traits by Eq. (4) and int-WUE by Eq. (6). At each plot, the top assimilating branches of a mature individual were selected randomly for the measurement of gas exchange, which includes photosynthetic rate (A), stomatal conductance (g s ), transpiration rate (E), the ambient CO 2 concentration (c a ) and the intercellular CO 2 concentration (c i ). Before the measurement of gas exchange, it takes about 5 s to stabilize after the assimilating branches were inserted in the cuvette. We repeated 10 times on the same assimilating branches for each measurement. Our measurements were carried out under the conditions of a standard 450 mmol mol −1 CO 2 concentration at a flow rate of 500 mmol s −1 above saturation in photo flux density of 1600 mmol m −2 s −1 . The temperature of the assimilating branches varied from 29.5 to 30.5 • C during the entire period of gas exchange measurements.

Sample collection
Sample collection was conducted in 20 July, during the addition of water and nitrogen. Considering that there is a considerable difference in δ 13 C between buds and young and matured leaves, we collected the mature assimilating branches of H. ammodendron for the δ 13 C measurements. All H. ammodendron individuals grown in plots (10 m × 10 m) were sampled. Eight pieces of the mature assimilating branches (15-20 cm long) were collected from each individual; two pieces of assimilating branches were collected at each of the four cardinal directions from the positions of full irradiance. All assimilating branches from the same plot were combined into one sample. After the samples were collected, they were immediately divided into two parts randomly and taken back to the laboratory at Fukang Station. The first part was used to determine the chlorophyll content. The second part was immediately inactivated in a 105 • C oven in the laboratory at Fukang Station and then brought back to Beijing in a ziplock bag. The time interval between sample collection and inactivation is very short. After inactivation, the carbon exchange of the assimilating branches stops, so the isotope composition of the samples will not change anymore. All plant samples of the second part were air-dried immediately in the laboratory in Beijing. Then the samples were ground into a fine powder using a steel ball mixer mill MM200 (Retsch GmbH, Haan, Germany) for the measurements of δ 13 C and N contents.
2.6 Measurements of plant δ 13 C, plant N and chlorophyll contents The δ 13 C and N measurements were performed on a Delta Plus XP mass spectrometer (Thermo Scientific, Bremen, Germany) coupled with an automated elemental analyzer (Flash EA1112, CE Instruments, Wigan, UK) in continuous-flow mode, at the Stable Isotope Laboratory of the College of Resources and Environmental Sciences, China Agricultural University. The carbon isotopic ratios were reported in the delta notation relative to the V-PDB standard. For this measurement, we obtained standard deviations lower than 0.15 ‰ for δ 13 C among replicate measurements of the same sample. And standard deviations for the N measurements were 0.1 %.
The chlorophyll contents of all samples were determined immediately when the samples were taken back in the laboratory at Fukang Station. The samples were first extracted by 95 % ethyl alcohol (0.5 g sample to 25 mL ethyl alcohol), and then the absorbency was measured under the wavelengths 665 and 649 mm by the spectrophotometer. The content of chlorophyll a and b was calculated by the following equations: Chlorophyll where OD665 and OD649 are the absorbency under the wave lengths of 665 and 649 mm, respectively.

Calculation of the degree of bundle-sheath leakiness
The degree of bundle-sheath leakiness (ϕ) was calculated by the transformation of Eq. (2): In this equation, parameters a, b 4 , b and s are constant, while δ 13 C plant and c i / c a are the measured values of our samples. We did not measure the δ 13 C air at our study site, so we had to use an approximation of the δ 13 C air to do this φ calculation. The approximated value we used is −9.77 ‰, which was measured at Donglingshan, Beijing, north China, in September 2019. The two sites should have similar δ 13 C air because the two sites are located in countryside with less human activity and have a similar distance from the nearest city. The straight line distances between Donglingshan and the city center of Beijing as well as between our study site and Ürümqi city are about 90 km. In addition, since the δ 13 C air has large diurnal and seasonal variations, we used the published range of δ 13 C air from May to July in Shangdianzi, China (data come from Global Monitoring Laboratory, Earth System Research Laboratories, https://www.esrl. noaa.gov/gmd, last access: 21 March 2021) to calculate the minimum and maximum δ 13 C air of the time period, which is the main growing season for H. ammodendron. Finally, the δ 13 C air used in the calculation ranged from −10.52 ‰ to −9.01 ‰ with an average of −9.77 ‰. Note. ϕ min , ϕ ave and ϕ max represent the ϕ values calculated from the minimum, average and maximum δ 13 C air . * , * * and * * * indicate a significant influence. W · N represents the interaction between water addition and N addition.

Statistical analysis
Statistical analyses were conducted using SPSS software (SPSS for Windows, Version 20.0, Chicago, IL, United States). One-way analysis of variance (ANOVA) and twoway analysis of variance (ANOVA) were used to compare the difference of δ 13 C and other physiological traits between each treatment. Pearson analysis was used to determine the correlation among δ 13 C, WUE and c i / c a in H. ammodendron.

Gas exchange and WUE under water and nitrogen addition
Photosynthetic rate (A), stomatal conductance (g s ), transpiration rate (E) and c i / c a ranged from 12.11 to 39.35 µmol CO 2 m −2 s −1 , from 0.09 to 0.31 mol H 2 O m −2 s −1 , from 2.87 to 8.49 mmol H 2 O m −2 s −1 and from 0.11 to 0.57, respectively. One-way ANOVA showed significant changes in leaf gas exchange across the six treatments (p < 0.01 for A, g s , E and c i / c a , Fig. 2). Two-way ANOVA suggested that water addition had exerted an effect on c i / c a (p < 0.01), that N additions influenced A (p < 0.01) and c i / c a (p = 0.009), and that the interaction between water and N supply played a role in g s (p < 0.001), E (p < 0.001) and c i / c a (p < 0.001, One-way ANOVA showed significant changes in these two indexes across these treatments (both p < 0.001, Fig. 3). Two-way ANOVA suggested that water addition, N addition and their interaction all have a significant effect on these two indexes (all p < 0.05, Table 1).

Correlations among δ 13 C, WUE and c i / c a ratio
In order to test whether δ 13 C in H. ammodendron can indicate WUE, the relationships among δ 13 C, ins-WUE, int-WUE and c i / c a ratio were revealed in this study. Our results showed no correlation between δ 13 C and ins-WUE (p = 0.23, Fig. 4a), between δ 13 C and int-WUE (p = 0.23, Fig. 4c), or between δ 13 C and c i / c a ratio (p = 0.18, Fig. 4e). However, there was a negative correlation between ins-WUE and c i / c a ratio (p < 0.001, Fig. 4b) and between int-WUE and c i / c a ratio (p < 0.001, Fig. 4d).

The degree of bundle-sheath leakiness under water and nitrogen addition
The ϕ value calculated from the minimum δ 13 C air ranged from 0.16 to 0.50 with a mean value of 0.35; the ϕ value calculated from the maximum δ 13 C air ranged from 0.44 to 0.70 with a mean value of 0.55, and the ϕ value calculated from the average δ 13 C air ranged from 0.32 to 0.59 with a mean value of 0.45. One-way ANOVA showed no significant vari-  ation in ϕ calculated from the minimum, average and maximum δ 13 C air across treatments (p = 0.60 for the ϕ calculated from the minimum δ 13 C air , p = 0.77 for the ϕ calculated from the average δ 13 C air and p = 0.90 for the ϕ calculated from the maximum δ 13 C air , Fig. 5). Two-way ANOVA suggested that ϕ was not affected by water addition (p = 0.46 for the ϕ calculated from the minimum δ 13 C air , p = 0.64 for the ϕ calculated from the average δ 13 C air and p = 0.98 for the ϕ calculated from the maximum δ 13 C air ), N addition (p = 0.65 for the ϕ calculated from the minimum δ 13 C air , p = 0.60 for the ϕ calculated from the average δ 13 C air and p = 0.55 for the ϕ calculated from the maximum δ 13 C air ) or their interaction (p = 0.30 for the ϕ calculated from the minimum δ 13 C air , p = 0.52 for the ϕ calculated from the average δ 13 C air and p = 0.87 for the ϕ calculated from the maximum δ 13 C air , Table 1).

Discussion
The δ 13 C of the assimilating branches in H. ammodendron did not change across treatments (Fig. 1, Table 1), suggesting that neither water addition nor nitrogen addition influenced the δ 13 C of H. ammodendron. Previous studies also reported no significant relationship between δ 13 C of C 4 plant and water availability (Swap et al., 2004;Wang et al., 2008) and between δ 13 C of C 4 plant and nitrogen availability (Yao et al., 2011;Yang et al., 2017). In general, the effects of water availability and nitrogen availability on δ 13 C are dependent on c i / c a ratio, which reflects the balance between stomatal conductance (g s ) and photosynthetic rate (A) (Farquhar and Richards, 1984). Stomatal conductance (g s ) usually increases with increasing water availability under water addition. Although two-way ANOVA suggested that water addition had no effect on both A and g s (Table 1), one-way ANOVA showed that g s was higher in W1N0 than that in W0N0 (Fig. 2b), indicating that water addition had a positive effect on g s under ambient N conditions. Increasing g s under water supply will lead to the rise of intercellular CO 2 because of the decrease in diffusional resistance to CO 2 . As a result, c i / c a ratio was observed to increase with increasing moisture (Fig. 2d, Table 1). However, δ 13 C remained stable under water addition (Fig. 1, Table 1). Thus, c i / c a ratio could not explain the observed response of δ 13 C to water supply. For most plants in natural ecosystems, nitrogen is the key factor limiting plant growth (Hall et al., 2011). Thus, nitrogen addition usually causes plants to absorb more N. However, extreme drought could prevent plants from absorbing N even under high N supply. In the present experiment, N supply was found to have an effect on N contents in H. ammodendron. Relative to the control treatment (W0N0), N contents increased with N supply under low N addition but remained unchanged under high addition (Tables S1, S2). Nitrogen is the main constituent of Rubisco (ribulose -1,5 -bisphosphate carboxylase oxygenase) and chlorophyll in plants. Thus, chlorophyll a was found to have a similar pattern as N contents under water and N supply. Chlorophyll a was higher in W0N1 than W0N0, and there was no difference in chlorophyll a between W0N0 and W0N2 (Table S1). Increasing chlorophyll contents in W0N1 should lead to the increase in photosynthetic rate (A). However, different from our prediction, one-way ANOVA suggested that A in W0N1 did not differ from that in W0N0 and that A in W0N2 was lower than that in W0N0 (Fig. 2a). Two-way ANOVA showed that N addition had an influence on A (Table 1). Both the analyses suggested that N supply played a negative role in A. These results might be associated with the extremely high light intensity at the study site. Due to the high light intensity, photosynthetic rate might not be correlated with chlorophyll contents (Gabrielsen, 1948). The negative effect of N supply on A led to the decrease in the consumption of intercellular CO 2 . Consequently, c i / c a ratio increased with N supply (Fig. 2d, Table 1). Therefore, the variations in c i / c a ratio with N addition could not account for the unchanged pattern in δ 13 C under N supply (Fig. 1).
The co-application of water and nitrogen had a negative effect on A but no effect on g s (W0N0 vs. W1N1, W1N2, Fig. 2a, b). The responses of A and g s to the co-application of water and nitrogen resulted in an increase in c i / c a ra-tio (Fig. 2d). Since δ 13 C remained unchanged under the co-application of water and nitrogen (Fig. 1), c i / c a ratio could not also explain the observed δ 13 C response to the coapplication of water and nitrogen.
In summary, the unchanged δ 13 C across treatments was not dependent on the c i / c a ratio in H. ammodendron (Fig. 4e). The observed δ 13 C stability across treatments might be associated with the ϕ value and carbonic anhydrase (CA) in H. ammodendron. For C 4 plants, the relationship between carbon isotope discrimination ( ≈ δ 13 C air − δ 13 C plant ; see Eq. 2) and c i / c a ratio is controlled by ϕ values (Ellsworth and Cousins, 2016;Ellsworth et al., 2017;Farquhar, 1983;Wang et al., 2008). Some studies suggested that ϕ value was stable for a given species under a wide range of environmental conditions (Henderson et al., 1992;Wang et al., 2008;Cernusak et al., 2013). However, other studies had different conclusions that ϕ value was influenced by irradiation (Bellasio and Griffiths, 2014;Kromdijk et al., 2010;Pengelly et al., 2010;Ubierna et al., 2013), temperature (von Caemmerer et al., 2014), water stress (Fravolini et al., 2002;Williams et al., 2001;Yang et al., 2017) and nitrogen supply (Fravolini et al., 2002;Meinzer and Zhu, 1998;Yang et al., 2017). In the current study, the ϕ value of H. ammodendron remained unchanged across six treatments (Fig. 5), and two-way ANOVA suggested that water supply and N supply had no effect on ϕ (Table 1). Therefore, the ϕ value of H. ammodendron was insensitive to water and N addition in this study. Even if the ϕ value remains stable, the relationship between and c i / c a ratio is also associated with the magnitude of the ϕ value. Cernusak et al. (2013) predicted that when the ϕ value is greater than 0.37, the correlation between and c i / c a ratio is positive; conversely, when the ϕ value is less than 0.37, the correlation is negative. In particular, when the ϕ value is equal to 0.37, there will be no correlation between them, because the coefficient ([b 4 +ϕ(b−s)−a] in Eq. 2) of c i / c a ratio equals 0 (Cernusak et al., 2013). The ϕ value calculated from the average δ 13 C air ranged from 0.32 to 0.59 with a mean value of 0.45 in the present study. Thus, the correlation between and c i / c a in H. ammodendron should be positive based on the prediction by Cernusak et al. (2013).
always changes in the opposite direction to δ 13 C plant changes according to Eq. (2); thus, a negative relationship between δ 13 C plant and c i / c a is expected. In fact, this study observed no correlation between δ 13 C and c i / c a in H. ammodendron (Fig. 4e); this indicates that ϕ was not the driver of the observed δ 13 C pattern in H. ammodendron. However, the measured δ 13 C represents the fixed carbon isotope composition throughout the assimilation branch formation period, which usually spans at least several weeks. And the measured c i / c a is an instant indicator. As a result, there were some uncertainties in the calculation of ϕ value using Eq. (2) based on the measured δ 13 C and c i / c a . In addition, the mean ϕ values calculated from the minimum and maximum δ 13 C air were 0.35 and 0.55, respectively, suggesting that the ϕ value of H. ammodendron might be close to 0.37, which led to the observed insensitive response of δ 13 C to water and N addition.
The enzymatic activity of CA may be another mechanism behind the unchanged δ 13 C across treatments. Cousins et al. (2006) suggested that enzymatic activity of CA affects carbon isotope discrimination in most C 4 plants because CA can result in the parameter b 4 changes (see Eq. 2). But in the traditional view, the parameter b 4 was a constant. However, it is only true when the ratio of PEP carboxylation rate to the CO 2 hydration rate (V p /V h ) is equal to zero, which is caused by a high CA activity. If V p /V h is not zero, b 4 will change and be controlled by V p /V h (Cousins et al., 2006). Previous studies reported that CA activity was low in most C 4 plants (Cousins et al., 2006;Yakir, 2000, 2001;Hatch and Burnell, 1990). Thus, CA activity in H. ammodendron might also be low, leading to the change in b 4 with V p /V h , and thus δ 13 C. Cousins et al. (2006) added V p /V h into the discrimination pattern of C 4 plants and predicted that at a given ϕ value, when the V p /V h is 0 or 1, the correlation between and c i / c a ratio is negative or positive, respectively. Since CA activity is low in most C 4 plants, and the V p /V h always ranges from 0 to 1, we speculate that no correlation between and c i / c a ratio may also occur when the V p /V h is a certain value between 0 and 1. The irrelevance of and c i / c a ratio also means that δ 13 C plant is not related to c i / c a ratio due to the negative correlation between and δ 13 C plant according to Eq. (2). Thus, the uncorrelated pattern between δ 13 C and c i / c a ratio in H. ammodendron might be related to this specific V p /V h value due to low CA activity.
In addition, the unchanged δ 13 C across treatments may also be controlled by the water sources of H. ammodendron. A previous study has found that the root of H. ammodendron can be inserted into the soil layer deeper than 3 m (Sheng et al., 2004), which made it easy to uptake groundwater. Therefore, H. ammodendron may be less sensitive to water addition. However, a study conducted in the same region has found that the shallow soil water (0-40 cm) and groundwater are two important water sources for H. ammodendron (Dai et al., 2014), and another study has reported that water addition resulted in an increase in soil water contents in the shallow soil layer (Cui, 2018). Moreover, gas exchange changed across treatments in the present study (Fig. 2). Thus, the utilization of groundwater by H. ammodendron may be one of the reasons why its δ 13 C was not sensitive to water and N addition, but it should not be the main reason.
Whether foliar δ 13 C of C 4 plants can indicate their WUE is still controversial. Henderson et al. (1992) found that δ 13 C of 10 C 4 species has negative correlation with their WUE. Although this result was just opposite to a positive relationship between δ 13 C and WUE for C 3 plants (Farquhar, 1983;Duquesnay et al., 1998;Feng, 1998), it is proof that δ 13 C of C 4 plants can indicate their WUE. In the work of Henderson et al. (1992), they found that the ϕ values in 10 C 4 species were around 0.21 over a range of irradiance and leaf temperature. According to the suggestion by Cernusak et al. (2013) that is negatively related to c i / c a ratio when ϕ value is less than 0.37, the δ 13 C of 10 C 4 species has a positive correlation with c i / c a ratio. In general, under fixed ambient CO 2 concentration, WUE is always negatively correlated with c i / c a ratio (see Eqs. 4 and 6). This is why a negative relationship between δ 13 C and WUE was observed for the 10 C 4 species. The present study showed that ins-WUE and int-WUE both had no correlation with δ 13 C in H. ammodendron (Fig. 4a,  c), which was different from the results published by Henderson et al. (1992). In general, c i / c a ratio is the link between WUE and δ 13 C. As mentioned above, if the ϕ value equals 0.37 and/or the activity of CA is very low, δ 13 C would not correlate to c i / c a ratio and thus leads to the uncorrelation between δ 13 C and WUE. In addition, the different timescales of δ 13 C, ins-WUE and int-WUE may also result in this uncorrelation. As mentioned above, the measured δ 13 C represents the long-term fixed carbon isotope composition (at least several weeks). And the values of ins-WUE and int-WUE were calculated from the gas exchange of a short-term measurement, which lasted only a few minutes. Therefore, this difference in timescale may also drive the uncorrelation between δ 13 C and WUE. Although the defects in measurements could introduce some uncertainty in the observed relationship between δ 13 C and WUE, δ 13 C remained stable under water and nitrogen addition (Fig. 1, Table 1), while the measured ins-WUE and int-WUE were higher in the control treatment (W0N0) than other treatments (Fig. 3), suggesting water and N supply had a significant effect on WUE (Table 1). These results indirectly confirmed that δ 13 C of H. ammodendron could not indicate its WUE.
The present study has found that δ 13 C of H. ammodendron could not be used as an indicator of its WUE. Although this conclusion cannot be analogous to all C 4 plants, the present study has important implications for the understanding of physiological responses of desert plants to future changes in precipitation and atmospheric N deposition. H. ammodendron is a dominant species in Asian desert, which has a great effect on the stabilization of sand dunes, the survival and development of understory plants, and the structure and function of desert ecosystems (Sheng et al., 2005;Su et al., 2007;Cui et al., 2017). Thus, H. ammodendron is widely distributed in desert areas, and the prediction of its drought adaptation is crucial in desert ecosystems.

Conclusion
Global changes including precipitation and atmospheric N deposition have been proven to have an important influence on ecosystems, especially for arid ecosystems. The present study showed that water and N addition had little effect on the δ 13 C values and the degree of bundle-sheath leakiness (ϕ) of H. ammodendron but played an important role in the change of its gas exchange and water use efficiency (WUE). In addition, different patterns of instantaneous WUE (ins-WUE), intrinsic WUE (int-WUE) and δ 13 C across treatment and no correlation between instantaneous WUE (ins-WUE) and δ 13 C and between intrinsic WUE (int-WUE) and δ 13 C have been found in this study, suggesting that δ 13 C of H. ammodendron could not indicate its WUE. This result was caused by the lack of correlation between δ 13 C and the ratio of intercellular to ambient CO 2 concentration (c i / c a ), which might be associated with the degree of bundle-sheath leakiness (ϕ) or the low activity of carbonic anhydrase (CA). Thus, the current experiment implies that the availability of δ 13 C as the indicator of WUE could be not universal for C 4 species.
Data availability. The datasets analyzed in this paper are not publicly available. Requests to access the datasets should be directed to gawang@cau.edu.cn.
Author contributions. GW and JL designed the experiment and modified the manuscript. ZC designed and executed the experiment and wrote the manuscript. XL designed the experiment. XC executed the experiment. YH executed the experiment.
Competing interests. The authors declare that they have no conflict of interest.