Differences on photosynthetic limitations between leaf margins and leaf centers under potassium deficiency for Brassica napus L.

Analyzing the proportions of stomatal (SL), mesophyll conductance (MCL) and biochemical limitations (BL) imposed by potassium (K) deficit, and evaluating their relationships to leaf K status will be helpful to understand the mechanism underlying the inhibition of K deficiency on photosynthesis (A). A quantitative limitation analysis of K deficiency on photosynthesis was performed on leaf margins and centers under K deficiency and sufficient K supply treatments of Brassica napus L. Potassium deficiency decreased A, stomatal (gs) and mesophyll conductance (gm) of margins, SL, MCL and BL accounted for 23.9%, 33.0% and 43.1% of the total limitations. While for leaf centers, relatively low limitations occurred. Nonlinear curve fitting analysis indicated that each limiting factor generated at same leaf K status (1.07%). Although MCL was the main component of limitations when A began to fall, BL replaced it at a leaf K concentration below 0.78%. Up-regulated MCL was related to lower surface area of chloroplasts exposed to intercellular airspaces (Sc/S) and larger cytosol diffusion resistance but not the cell wall thickness. Our results highlighted that photosynthetic limitations appear simultaneously under K deficiency and vary with increasing K deficiency intensity.

activity was decreased under K deficiency, becoming a major limiting factor for photosynthesis in Oryza sativa leaves 3 . Chlorophyll synthesis was observed to be significantly impaired under K deficiency in Eucalyptus grandis leaves 1 . Moreover, K starvation up-regulated the fraction of electron transport to O 2 , resulting in an increased reactive oxygen species (ROS) 19 . Carbohydrate accumulation which may feedback regulation of leaf photosynthesis is more easily observed in K starved leaves 20,21 . Indeed, the relative contributions of these three limiting processes to photosynthesis under K deficiency and the underlying mechanisms have not been fully explored, due to the complicated physiological processes and variation of dominant limiting factors under differ K deficiencies 2,5 . No matter what the primary cause of decrease A, the discrepancy between researches was believed to be derived from differ physiological K deficiency severities. For this reason, a comprehensive consideration of whole limiting factors and their relationships with leaf K status seems to be important.
In 2005, Grassi and Magnani proposed a method to accurately quantify photosynthetic limitations by separating the relative controls on A resulting from S L , mesophyll conductance (MC L ) and biochemical limitations (B L ) 22 . This method has been successfully applied for evaluating the relative control of leaf A under water stress and during their recovery processes, among inter-and intra-species 13,[23][24][25] . It showed not only great potential for elucidating the magnitude changes of limitations and their dominance in photosynthetic restraints with increasing severity of K deficiency, but also revealing the corresponding critical K concentrations for their transformation.
Winter oilseed rape (Brassica napus L.), a model-plant of winter cover crops who needs substantial amount of potassium to growth was used for a deeply aggregate analysis of K deficiency on photosynthetic limitations 26 . However, malfunction of physiological processes like photosynthesis is hard to be affected when K concentration above the threshold value (1.5% in dry matter, or less) 10 . On consideration of the fact that potassium deficiency symptoms, characterized by a chlorosis and even scorch around the periphery can be obviously observed when leaf K concentration below 1.0% in most species 27 . And the withdrawal K initially occurred at the edge of leaf tip, as tip cells are initially proliferated and oldest 28 , resulting in different K levels as well as visible distinctions between centers and margins. These natural K gradients are therefore precious for photosynthetic limitation analysis, from which we may seek out the main limiting factors under variable leaf K status and the corresponding threshold values. This phenomenon occurred more frequently under a complex biological and abiological environment system during a long-time and low-temperature wintertide, which may conducive to generate a physiology K deficiency in a K-deficient soil, i.e., it may bring K function into full play 29,30 . Accordingly, the objectives of the present study were to: (1) estimate the differences of contributions for three limiting factors to photosynthesis between leaf margins and leaf centers, (2) uncover the relationships between photosynthetic limitations and diminishing leaf K status, and therefore the critical K concentration for the predominate restraint transformation, (3) reveal the mechanism underlying the K-induced variation of limiting factors. It is hoped that this research will facilitate a better understanding of the photosynthetic physiological mechanism by which potassium deficiency leads to growth retardation in oilseed rape.

Results
Plant performance, leaf K concentration and net photosynthesis. The total dry matter of the -K treatment decreased significantly by 29.9% on average versus the + K treatment ( Table 1). The leaf expansion was also restrained, with a 22.1% and 18.0% decline in the individual leaf dry matter and leaf area, respectively. Leaf K concentration was dramatically influenced by potassium supply and leaf position, which was significantly lower in the -K treatment than in the + K treatment. Meanwhile, within an individual leaf, K concentration was remarkably lower in margins than in centers. The mean net photosynthesis (A) in the leaf margins of the -K treatment was 56.9% that of the + K treatment. However, there was no significant difference between leaf margins and centers under the -K treatment, as well as the two positions under the + K treatment.
Stomatal conductance. Potassium deficiency led to a significant decline of the mean stomatal conductance (g s ) in leaf margins, which was 63.6% that of the + K treatment. However, the mean g s value of the leaf centers was not influenced by K nutrient (Table 2). There was a significantly lower g s in leaf margins than in leaf centers under the -K treatment, whilst the g s values of these two positions were the same under the + K treatment. Despite a  Table 1. Effects of K deficiency on plant dry matter, leaf dry matter, leaf area, leaf K concentration, and net CO 2 assimilation rate (A) for the two positions in the fifth fully expanded leaves. Values are mean ± SE of six replications for total dry matter, individual leaf dry matter and leaf area, and of four replications for K concentration and A. 1 Different letters in the same column of total dry matter, individual leaf dry matter and leaf area indicate significant differences between treatments (P ≤ 0.05). 2 Different letters in the same column of K concentration and A indicate significant differences between positions (P ≤ 0.05). 3 *shows significant differences between the two K treatment in same position (P ≤ 0.05).
Scientific RepoRts | 6:21725 | DOI: 10.1038/srep21725 decrease of the g s value in leaf margins of the -K treatment, the intercellular CO 2 concentrations (C i ) value was raised, and the mean C i were similar to those of other groups. Potassium supply and leaf position had no effects on stomatal frequency and stomatal length (Table 2). However, stomatal width was significantly decreased in the -K treatment, especially in the leaf margins where the width decreased by 20.9% as compared with the + K treatment. Stomatal pore area was therefore considerably decreased due to K deficiency, particularly in the leaf margins with a 28.0% decline in the single stomatal pore area. Nevertheless, the stomatal length and width as well as the stomatal pore area showed no significant difference in these two positions under the + K treatment.
Mesophyll conductance. Despite a dramatic decrease in the mean mesophyll conductance (g m ) in the leaf margins of the -K treatment, the mean chloroplastic CO 2 concentrations (C c ) was 9.4% higher than that of the + K treatment ( Table 3). The mean g m and C c values were similar in the leaf centers of different K treatments, as well as between the two positions in the + K treatment. Potassium niutrient and leaf position did not affect mean intercellular CO 2 compensation point (C i * ) and mitochondrial respiration rate in the light (R d ), however, the mean chloroplastic CO 2 compensation point (Γ *) was significantly increased in leaf margins of the -K treatment, but showed no statistical differences among the other three groups.
Biochemical characteristics. The mean maximum rate of electron transport (J max ) and maximum rate of carboxylation (V c,max ) in the leaf margins of the -K treatment were the lowest, and minor changes were observed among the other three treatments (Table 4). However, the mean J max /V c,max in leaf margins of the -K treatment was dramatically increased compared with the mean values of the other three groups in the range from 1.42 to 1.46. The variation of photosynthetic parameters was verified by chemical analyses (Table 4). A significant decline of leaf chlorophyll concentration was found in the -K treated leaves, especially in the leaf margins, with a 31.1% decrease. Furthermore, Rubisco activity was dramatically decreased in leaf margins of the -K treatment, but it was the same in the leaf centers of the -K treatment and the two positions of the + K treatment. Potassium deficiency caused severe ROS production in leaf margins where O 2 .− generation rate increased by 22.8%, and meanwhile, POD activity increased by 25.5%.
The relationship between relative A, g s and g m with leaf K concentration. A significant curvilinear relationship between relative A, g s or g m and leaf K concentrations is shown in Fig. 1. The relative values increased with increasing leaf K concentration, and remained stable when the leaf K concentration was beyond a certain concentration. Here a photosynthesis-based concentration threshold with the relative values reaching 95.0% of  Table 2. Effects of K deficiency on stomatal conductance (g s ), intercellular CO 2 concentrations (C i ), stomatal frequency, length, and width, single stomatal pore area and total stomatal pore area of the two positions in the lower epidermis of the fifth fully expanded leaves. Images were taken at a magnification of × 500 with a scanning electron microscope. Values are mean ± SE of four replications for g s , C i , of 20 replications for stomatal frequencies, and of 300 replications for stomatal lengths, stomatal widths, stomatal pore areas and total stomatal pore areas. 1 Different letters in the same column at a given treatment indicate significant differences between positions (P ≤ 0.05). 2 *shows significant differences between the two K treatment in same position (P ≤ 0.05). Table 3. Effects of K deficiency on mesophyll conductance (g m ), chloroplastic CO 2 concentrations (C c ), intercellular CO 2 compensation point (C i * ), mitochondrial respiration rate in the light (R d ), and chloroplastic CO 2 compensation point (Γ*) in the two positions of the fifth fully expanded leaves. C i * and R d were measured by Laisk method, Γ * was calculated according to the equation Values are mean ± SE of four replications for g m , C c , and C i -C c , of three replications for C i * , R d and Γ *. 1 Different letters in the same column at a given treatment indicate significant differences between positions (P ≤ 0.05). 2 *shows significant differences between the two K treatment in same position (P ≤ 0.05).
the maxima was defined. The relative A values increased rapidly with increasing leaf K concentration when it was less than 1.07% (Fig. 1a), and varied little when the leaf K concentration was above 1.07%. Therefore, the K concentration (1.07%) was used to evaluate the relative g s and g m (Fig. 1b,c), and the calculated result (93.7% and 94.3%) was close to 95.0%, indicating that this threshold value was acceptable for g s and g m .
Quantitative limitation analysis. The restrictions of A max in the -K leaves posed by stomatal (S L ), mesophyll conductance (MC L ) and biochemical limitations (B L ) are presented in Fig. 2a. In symptomatic margins, total limitations reached a value of 46.9%, and the contribution of S L , MC L and B L represented 23.9%, 33.0% and 43.1% of total limitations, respectively. By contrast, despite the relatively low limitation (4.8%) in the leaf center, MC L contributed a primary limitation to A max . Accordingly, the dominant limitations changed from symptomatic leaf margins to centers. The relationship between relative limitations and leaf K concentrations were further analyzed (Fig. 2b). All the limitations declined precipitously with the leaf K concentration increased from 0.6 to 1.07% (according to the K-based concentration threshold), particularly the B L with the maximum slope of the fitted curve, but they gradually decrease as leaf K concentration continues to increase. Their relative contribution also varied with the change of the leaf K status. While MC L largely predominated at the leaf K concentration of less than 1.07%, B L replaced it when the K concentration was below 0.78% (leaf K concentration of the intersection point between B L and MC L fitted curves).

Discussion
Limitations imposed by K deficiency occur at the same time. In the present study, A in leaf margins were weakened by K deficiency. Generally, the declining A is considered to be limited by stomatal and mesophyll resistances to CO 2 diffusion, and biochemical obstacles 13,22 . Here we demonstrated that g s , g m and biochemical activities were profoundly restricted as A down-regulated. Stomatal conductance which determine the vital step of CO 2 diffuse from the atmosphere to the interior of leaf was markedly decreased in -K leaf margins, as reported for Eucalyptus grandis 1 , Gossypium hirsutum 17 , and Oryza sativa 3 . This is mainly because the lack of vacuole K to keep stomatal aperture by providing driving force to promote water inpour into the guard cell vacuole 31 . The declined A, to a certain extent, revealed that the K in cytoplasm identified as biochemical functional component was below the critical value 10 . Therefore, malfunction of physiological process could come with limited A.
Likewise, g m was decreased in parallel with A. Indeed, g m might be down-regulated by increasing leaf dry mass per area (M A ) 7,23 , however, in the present study, there was no remarkable difference in M A between the -K and + K leaves (Table 5; Supplementary Fig. S1). Cell wall thickness (T cell-wall ) and surface area of chloroplasts exposed to intercellular airspaces (S c /S) are reported to be the most substantial anatomical traits in determining g m 23,32 . However, significant differences in mesophyll cell wall surface area exposed to intercellular airspace per leaf area (S m /S) and S c /S, but not T cell-wall between leaf margins of two K treatments were observed (Table 5). Besides, chloroplast size 33 has also been proved to influence g m . In the present study, though the chloroplast length (L chl ) decreased under lowest K status, the thickness (T chl ), surface area (S chl ) and volume (V chl ) of chloroplast were largely increased, however, the S chl /V chl was smaller (Fig. 3). The chloroplast enlarging under lowest K concentration was not completely same to that discovered under low nitrogen conditions 33,34 . The increase of T chl was more likely to be based on the sacrifice of length owing to roughly circular envelope (Fig. 3a,b; Supplementary Fig. S2). Mathematically, ellipsoidal chloroplasts, combining with an increscent chloroplast number (see Supplementary  Fig. S3) were more probably to have longer length of chloroplasts facing the cell wall than swollen even sphere envelopes. Furthermore, the resistance along diffusion pathway length in cytoplasm (distance of chloroplast from cell wall, D chl-cw ) and stroma (taken as half of the chloroplast thickness) account for 10-50% of g m limitation 23 , which however, reported only up to 22% of liquid phase resistance (r liq ) by Evans et al. in 1994 35 . Low K status strongly increased T chl and D chl-cw (Fig. 3b,f), accordingly, the corresponding resistance would be increased. It is therefore proved that the decreased g m is primary due to the reduced S c /S and larger cytosol diffusion resistance but not T cell-wall . More evidences may seek from the influence of K on plasma membrane and chloroplast envelope conductance 32 , carbonic anhydrase and aquaporins that participated in determination of g m 14,15,23 .
It should be noted that A, g s or g m started to decline almost at the same time with an extremely similar leaf K status. By another way, the quantitative analysis of limitations indicated that three limiting factors coexist when K concentration below 1.07%. This was similar to the results reported by Grassi Table 4. Effects of K deficiency on the maximum rate of electron transport (J max ), maximum rate of carboxylation (V c,max ), ratio between J max and V c,max (J max /V c,max ) estimated from A-C c curves, chlorophyll concentration, Rubisco activity, O 2 .− generation rate, and POD activity in the two positions of the fifth fully expanded leaves. Values are mean ± SE of four replications. 1 Different letters in the same column at a given treatment indicate significant differences between positions (P ≤ 0.05). 2 *shows significant differences between the two K treatment in same position (P ≤ 0.05).
in plants suffering from water stress. However, the investigation carried out by Galmés et al. 13 revealed that B L of Hypericum balearicum and Phlomis italica still remained zero under mild water stress even if the total limitation reached 20-30%. The present finding highlights that all photosynthetic limitations simultaneously occur when leaf is in a physiological K-deficiency state.    Table 5. Effects of K deficiency on leaf thickness (T leaf ), mesophyll cell wall thickness (T cell-wall ), mesophyll cell wall surface area exposed to intercellular airspace per leaf area (S m /S), and surface area of chloroplasts exposed to intercellular airspaces (S c /S) in the two positions of the fifth fully expanded leaves. Data are mean ± SE of eight replications for S m /S and S c /S, at least thirty replications for T leaf and T cell-wall . 1 Different letters in the same column at a given treatment indicate significant differences between positions (p ≤ 0.05). 2 * shows significant differences between the two K treatment in same position (p ≤ 0.05). centers, respectively. This is mainly ascribed to the discrepancy of relative severity of K deficiency 5, 6 . As has been stated in the previous studies that some irreversible damages, such as impaired ATP synthesis, depressed Rubisco activity, and cell damage occurred when the limiting A, for the most part, is attributed to B L 3,38 . Some of which were verified in the present study, such as degraded chloroplast, limited photoassimilate transportation (see Supplementary Fig. S2), and increased O 2 .− generation rate under severe K deficiency. The obstacle of these physiological processes alleviated as K deficient stress mitigating, however, the role of MC L on A began to stand out.
The relationship between relative limitations and leaf K concentration verified that, at a leaf K concentration of less than 1.07%, MC L represented the main component of limitations, but B L replaced it when leaf K concentration below 0.78%. This pattern, to a lesser extent, could be found in plants suffering from water stress which suggested that the variation of limitations depends on the stress intensity and duration 22 . Regrettably, the present study failed to reveal whether or not there is a critical concentration in the shifting process from S L predominance to MC L predominance. Further studies focusing on the photosynthetic limitations of rapeseed leaves subjected to a serial K gradient may help to elucidate this issue. Methods Study site and growth conditions. A field experiment was conducted in Wuxue county, Hubei province, central China (30° 06′ 46″ N, 115° 36′ 9″ E) during the 2013-2014 oilseed rape growing season. The mean temperature of the season was 13.8 °C, and the average temperature during winter (from December 2013 to February 2014) was 5.9 °C. The total precipitation during oilseed rape cropping season was 660.7 mm, with wintertide accounting for 26.1% of the total. The soil was a sandy loam with pH 5.3, organic matter 30.5 g kg −1 , total N 1.7 g kg −1 , NH 4 OAc-K 42.5 mg kg −1 , Olsen-P 15.7 mg kg −1 and hot-water soluble B 0.78 mg kg −1 in the topsoil layer (0-20 cm). As stated by Zou, the soil belongs to a K-deficient type, which would cause yield reduction without K fertilizer addition 4 .

Experimental design. A complete randomized block design was set up with two treatments and four rep-
licates. The treatments were: (1) Sufficient K supply treatment ( + K), with a K fertilizer recommendation rate of 120 kg K 2 O ha -1 which was tested and well-proved to ensure the optimal growth and yield formation of oilseed rape based on field experiments in this region 39 . (2) K deficiency treatment (-K), with no K fertilizer applied throughout the growing season.
To ensure that nutrients other than K did not limit plant K uptake, 180 kg N ha −1 , 90 kg P 2 O 5 ha −1 , and 1.6 kg B ha −1 were applied for these two treatments. The N, P, K, B fertilizers used in the experiment consisted of urea (46% N), superphosphate (12% P 2 O 5 ), potassium chloride (60% K 2 O), and borax (10.8% B). The N fertilizer was applied in three splits: 60% prior to transplanting, i.e., BBCH (Biologische Bundesantalt, Bundessortenamt and Chemische Industrie) 15-16 40 , 20% at the over-wintering stage (i.e., BBCH 29), and 20% at the initiation of stem elongation (i.e., BBCH 30). Besides, all the P, K, B fertilizers were applied as basal fertilizers. The experimental field was plowed and leveled with a rotary tiller, and basal fertilizers were incorporated during the process. The plot measured 20 m 2 with a length of 10 m and a width of 2 m.
The oilseed rape cultivar was Zhongshuang 11, supplied by Oil Crops Research Institute of the Chinese Academy of Agricultural Sciences. Rapeseeds were sown in prepared seedbeds on 16 September 2013, and then, on 22 October, about 36 d after sowing, oilseed-rape seedlings with five to six leaves (i.e., BBCH 15-16, 3-4 g dry weight plant −1 ) were uniformly selected and transplanted by hand in double rows spaced approximately 0.3 m apart with 0.2-0.3 m between plants, corresponding to 112 500 plants ha -1 . The oilseed rape was grown under rain-fed conditions. Meanwhile, weeds, pests and disease stresses were controlled by spray herbicides, insecticide and fungicide so that no obvious weeds, insect pests, and diseases infestation occurred during cropping season.
Plant and leaf tagging. There was an obvious phenotypic difference in plants between the -K and + K treatments 60 d after transplanting. The discrepancy was highlighted in the fifth to ninth fully expanded leaves (with a total average of 9 fully expanded leaves (i.e., BBCH 19) in both treatments) from apex downwards, specifically, obvious etiolation symptoms around the periphery in K-deficient leaves and asymptomatic leaves in the + K treatment. For each treatment, 24 fifth fully expanded leaves and six uniform plants were tagged in each of the four replicate plots for destructive and non-destructive analysis described later in the methods.

Leaf gas exchange and fluorescence measurements. Leaf gas exchange and chlorophyll fluorescence
were measured simultaneously at either leaf margins (Fig. 4, about 2 cm of leaf surface from the margin) or leaf centers (the rest part between half-elliptic and vertical dashed lines), using a portable, open circuit, infrared gas analysis system (Li-6400, Li-Cor Inc., Lincoln, NE, USA) equipped with an integrated leaf chamber fluorometer . Measurements were performed on at least four randomly selected leaves of both treatments in the late morning (11:00-12:30) under a light-saturating photosynthetic photon flux density (PPFD) of 1200 μ mol m −2 s −1 (with 90% red light and 10% blue light). CO 2 concentration in the leaf chamber (C a ) was set at 400 μ mol mol −1 air, leaf temperature was controlled at 25 ± 0.2 °C, relative humidity was between 50 and 60%, and the flow rate was 500 μ mol s −1 . In addition to net photosynthesis (A), stomatal conductance to water vapour (g s ) and intercellular Scientific RepoRts | 6:21725 | DOI: 10.1038/srep21725 CO 2 concentration (C i ), the incorporated fluorometer allowed determination the steady-state fluorescence yield (F s ) under actinic light and maximum fluorescence ( ′ F m ) during light-saturating pulse (0.8 s) of approx. 8000 μ mol m -2 s -1 . The relative A, g s and g m values were the relative proportion of measured values over the mean values of K-sufficient leaf centers.
A/C i curves were measured on the two positions that had been previously acclimated to saturating light conditions for 20 min. The CO 2 concentration (C a ) in the gas exchange chamber was reduced stepwise from 400 to 300, 250, 200, 150, 100, 50 μ molCO 2 mol −1 , and then increased from 50 to 400, 600, 800, 1000, 1200, 1500, 1800 μ mol CO 2 mol −1 at a constant PPFD of 1200 μ mol m −2 s −1 at 25 ± 0.2 °C, and 50-60% relative humidity. In all cases, the parameters were recorded after the gas exchange rate stabilized at the given C a . At least four leaves were performed in each treatment.
The actual photochemical efficiency of photosystem II (Φ PSII ) was then determined as follows 41 : The electron transport rate (J) can be calculated as: Where α is the leaf absorptance, and β is the fraction of light distributed to PSII. As routinely assumed, α was taken as 0.85 42,43 and β was taken as 0.5 44,45 . A sensitivity analysis of J biases resulting from rough assumption of α and β on g m variations was also conducted (See Supplementary Table S5).
Mesophyll conductance was estimated according to Harley et al. from combined gas exchange and chlorophyll fluorescence measurements 46 where A, C i and J were determined as previously described for each treatment, mitochondrial respiration rate in the light (R d ) and the intercellular CO 2 compensation point (C i * ) were measured by Laisk method, as described by Brooks and Farquhar 47 . Briefly, the A/C i curves generated with PPFD values of 75, 150, 500 μ mol m −2 s −1 , respectively, with each having five different C a in chamber (i.e. 50, 80, 100, 120 and 150 μ mol CO 2 mol −1 ). A linear regression was then fitted to each A/C i curve. The x-axis and y-axis of intersection point of three A/C i curves were defined as C i * and R d
The Γ * is the chloroplastic CO 2 photocompensation point calculated from C i * and R d as: For each data point generated, we checked whether it met the range of < / < C dA 10 d 50 c 46 . The CO 2 concentration in the chloroplast stroma (C c ) was calculated as: Therefore, A-C i curves were converted into A-C c curves. On the basis of C c , the maximum rate of Rubisco-catalysed carboxylation (V c,max ), and the maximum rate of electron transport (J max ) as defined by Farquhar et al. 49 , were calculated 11,50 . Since variable J method is sensitive to many sources of errors, e.g. (1) Γ * and R d biases; (2) a wrong assumption of p 1 and p 2 ; (3) biases in the measurements of C i , A, and J, a sensitivity analysis would be great values to improve the confidence in g m estimates and following limitation calculations 51 . Following the method of Harley et al. 46 , we used actual Γ *, R d and J values calculated in this study and a deviation from the measured values to analyze the effects of Γ * , R d and J on g m estimates (see Supplementary Table S1, S3, S5). RuBP regeneration can be limited by either insufficient NADPH or ATP, according to Farquhar model, A and J can be linked as follows: For insufficient NADPH, p 1 = 4 and p 2 = 8; for insufficient ATP, p 1 = 4.5 and p 2 = 10.5 or p 1 = 4 and p 2 = 9.33. Finally, the sensitivity analysis for photosynthetic limitations was conducted basing on these calculated g m values (see Supplementary Table S2, S4, S6). The analysis showed that the g m was significantly affected by varying Γ * and R d (see Supplementary Table S1), p 1 and p 2 inputs (see Supplementary Table S3) and J biases (see Supplementary  Table S5). However, the g m variation derived from Γ * , R d , J, p 1 and p 2 biases, did not cause profound effects on photosynthetic limitations (see Supplementary Table S2, S4, S6). In addition, g m appears to be strikingly affected by C i 12,14,51 , nevertheless, the similar C i in different treatments and positions here seems to have no impact on g m (Table 3). Therefore, the results obtained was unlikely to be altered by these methodological artifacts.
Plant dry matter, leaf area and dry matter. Six tagged leaves and six tagged plants in each plot were used to determine the individual leaf area, dry matter, and total dry matter. Each leaf was digitally scanned using an Epson ES-1200C scanner (Epson, Long Beach, CA, USA), and the area determined using ImageJ software (National Institutes of Health, Bethesda, Maryland, USA) 1 . Individual leaf dry matter and total dry matter were weighed after oven drying at 65 °C for 48 h. Biochemical analysis. Twelve tagged leaves per plot were picked immediately after the determination of photosynthesis. They were divided into two parts along vertical dashed lines (Fig. 4), followed by dissecting the leaf apexes into leaf margins and leaf centers, and removing all the veins. A portion of segments were immersed in liquid N and then stored at − 78 °C, and the rest were used for leaf K concentration determination. There were four replications for biochemical determinations.
Leaf segments (2 g) were oven dried at 65 °C for 48 h. After that, about 0.15 g dried leaves were milled and digested with H 2 SO 4 -H 2 O 2 as described by Thomas et al. 52 , and K concentration in digestion solution was measured by a flame photometer (M-410, Cole-Parmer, Chicago, IL, USA).
The Rubisco extracts were prepared according to Weng et al. with minor modifications 3 . Briefly, leave segments (0.2 g) were ground to a powder using a chilled mortar and pestle with liquid N 2 and a small amount of quarzsand, followed by homogenization with 4 mL pre-cooled extraction buffer containing 50 mM Tris-HCl (pH 7.5), 1 mM EDTA,10 mM MgCl 2 , 12.5% (v/v) glycerol, 10 mM (v/v) β -mercaptoethanol and 1% (w/v) PVP-40 (soluble PVP) at 0-4 °C. The homogenate was centrifuged for 15 min at 15 000 g at 4 °C, and then the supernatant was immediately used to determine the activity of ribulose-1, 5-bisphosphate carboxylase/oxygenase (Rubisco, EC 4.1.1.39) by an enzyme-linked immunosorbent assay method with a RuBPcase ELISA kit (CK-E91697P, Shanghai jijin Chemistry and Technology Co., Ltd, China) according to the manufacturer's instructions. The chlorophyll concentration was determined according to the method of Huang et al. 53 .
Superoxide radical O 2 .− production rate was measured by monitoring the nitrite formation from hydroxylamine in the presence of O 2 .− according to Elstner and Heupel 54 . A 0.5 g aliquot of leaf margins and centers was ground and homogenized in 5 mL of 65 mM pre-cooled phosphate buffer (pH 7.8), followed by centrifuging the homogenate at 10,000 g for 15 min at 4 °C and mixing 0.5 mL of the supernatant with phosphate buffer (0.5 mL) and 0.1 mL of 10 mM hydroxylamine hydrochloride. This mixture was incubated at 25 °C for 20 min, followed by the addition of 1 mL of 58 mM sulfanilic acid and 1 mL of α -naphthylamine, and then another 20 min incubation at 25 °C. The as-prepared solution was shaken with equal volume of ether, followed by centrifuging the mixture at 10,000 g for 3 min and measuring the absorbance of the pink water phase at 530 nm. The activity of POD (EC 1.11.1.7) was determined using the guaiacol oxidation method 55 .
Anatomical analysis. Another six tagged leaves per treatment were collected, and removed all the veins for anatomical analysis. The stomatal size and frequency were measured in six sub-samples either for leaf margin or center. The materials were prepared as described by Meng et al. 56 . Briefly, leaf samples (about 1cm in length and 1 cm in width) were fixed in 2.5% glutaraldehyde (v/v) at 4 °C for 2 h, and washed twice in 0.1 M phosphate buffer (pH 6.8). Next, they were sequentially dehydrated in ethanol (30%, 50%, 70%, 80%, 90%, 95%, and 100%) for 10 min at each gradient concentration, with 100% ethanol repeated twice. After further drying and spraying with gold, the as-treated leaf samples were observed and photographed with a scanning electron microscope (JSM-5310LV, Jeol Co, Tokyo, Japan). Images were taken of the lower leaf surface for five microscope fields per sub-sample at a magnification of × 500. The number of stomata was counted in each field (a total of 20 measurements of stomatal frequency for each position) as described by Battie-Laclau et al. 1 , and the stomatal frequency was calculated by dividing the stomatal count by the area of the field of view 57 . Moreover, the length and width of ten stomata selected at random were measured in each field. Assuming the stomatal pore as an ellipse, the total stomatal pore area was calculated (stomatal frequency × π × 0.25 × stomatal length × stomatal width).
Leaf segments (1-2 mm 2 ) were cut from each part and fixed with 2.5% glutaraldehyde (v/v) in 0.1 M phosphate buffer (pH 7.4) for 4 h, followed by washing twice in the same buffer for 30 min and postfixing with 2% osmium tetroxide for 4 h at 4 °C. Next, the samples were dehydrated with an ethanol series (10-100%) and in propylene oxide, followed by embedding them in Epon 812 resin.
For the light microscope observation, they were cut into 1 μ m transverse sections by LKB-5 ultramicrotome 359 (LKB Co., Ltd., Uppsala, Sweden), and stained with 0.5% toluidine blue. Micrographs were captured at a magnification of × 400 with a Nikon Eclipse E600 microscope equipped with a Nikon 5 MP digital microscope camera DS-Fi1 (Nikon Corporation, Kyoto, Japan). There were four samples per treatment. For each samples, three cross-sections were chosen to measure their thickness (T leaf ), mesophyll cell wall surface area exposed to intercellular airspace per leaf area (S m /S), and surface area of chloroplasts exposed to intercellular airspaces (S c /S) according to Tosens et al. (2012) 32 . Where L mes and L c are the length of mesophyll cell wall exposing to intercellular air space and chloroplast surface area touching the intercellular air space. W is the width of measured cross-section. F is the curvature correction factor which was obtained as the weight average of palisade and spongy mesophyll.
For the ultrastructural observations, ultrathin sections (90 nm) were examined with a transmission electron 360 microscope (H-7650, Hitachi, Japan) after staining with 2.0% uranyl acetate (w/v) and lead citrate. Cell wall thickness (T cell-wall ), chloroplast length (L chl ) and thickness (T chl ) were measured from at least 30 chloroplasts. Chloroplasts were assumed as ellipsoids, and chloroplast surface area (S chl ) and volume (V chl ) were calculated according Cesaro formula 58  where = . × a L 0 5 chl ; = . × b T 0 5 chl . Distance of chloroplast from cell wall (D chl-cw ) was determined according to Tomás et al. 23 .
Quantitative limitation analysis. The limitations (stomatal limitations, S L ; mesophyll conductance limitations, MC L ; biochemical limitations, B L ) imposed by K deficiency on photosynthesis were investigated by analyzing the leaf margins and centers under two treatments using the quantitative limitation analysis method proposed by Grassi and Magnai 22 . Relative changes in light-saturated assimilation is expressed in terms of relative changes in stomatal, mesophyll conductance, and biochemical capacity as Equation (11) where l s , l mc , and l b are the corresponding relative limitations calculated as Eqns from (12) to (14), g sc is stomatal conductance to CO 2 (g s /1.6), and V c,max is maximum rate of carboxylation estimated from A-C i curve.