Experiments were conducted in an experiment field in Beijing (115.7°E–117.4°E, 39.4°N–41.6°N). The sorghum hybrid Liaoza 13 was selected for investigation. The seeds (provided by National Center for Sorghum Improvement, Liaoning province, PR China) were imbibed on wet paper for 1 d, and germinated seeds were sown in containers (30 cm ×20 cm × 10 cm) filled with vermiculite. Plants were watered every 2 days (d). Two weeks later, 4-leaf seedlings were transplanted into plastic pots (15 cm in diameter, 20 cm in height) containing Hoagland’s nutrient solution and grown in a site having simple rain shelters in the field exposing to natural light with a maximum intensity at 1800 ± 23 μmol photons m^-2s^-1. The nutrient solution was renewed every 3 d. The seedlings were used for measurements after 20 d.

Leaf reflectance and transmittance on each surface of the sorghum leaves were measured using a bifurcated fiber optic cable and a leaf clip (PP Systems, USA) across the spectrum of 310 - 1100 nm at 1-nm intervals. To calculate reflectance, leaf spectral radiance was divided by the radiance of a 99% reflective white reference standard (Spectralon, Labsphere, North Dutton, NH, USA) (Gamon and Surfus, 1999). Leaf transmittance was measured using two straight-fiber optics and a custom-made device. One straight-fiber optic irradiated the leaf from the adaxial side, and the other was used as a detector on the abaxial side. Absorptance = 1- reflectance - transmittance. The reflectance, transmittance, and absorptance could be obtained over 300 - 1100 nm wavelength. The average photosynthetically active radiation (PAR) (reflectance, transmittance, and absorptance of leaves was calculated from a trapezoidal integration over the 400 - 800 nm wavelength.

Photosynthetic gas exchange rates were measured using an infrared gas analyzer (Li-6400, Li-Cor, Lincoln, NE, USA) equipped with a combined LED light source which contained multispectral hybrid white light. By setting the voltage, the light source could provide direct light with an intensity range from 0 μmol m^-2 s^-1 to 1800 μmol m^-2s^-1. All the light near the earth’s surface is diffuse, which atmospheric aerosols have scattered. In many ecological studies, the DI mostly remained less than 0.27 during sunny days and more than 0.65 during cloudy days, respectively (Roderick et al., 2001; Li et al., 2004; Urban et al., 2007; Urban et al., 2012; Wang et al., 2016). This study measured the DI over the 400 to 800 nm wavelength range using a fiber optic spectrometer equipped with a cosine corrector (AvaSpec-ULS 2048XL, Avantes, Netherland). The light with DI > 0.55 was recognized as direct light from a LED, and the light with DI < 0.27 was recognized as diffuse light from using a scattered plate (Fotodiox Inc., USA) to convert the direct light to diffuse light ( Figure 1 ).

A transparent film (TF) that passed our optical testing was used to block the flux between the leaf surface and chamber in a gas-exchange analyzer system (LI-6400), allowing the gas exchange rates of the opposite surfaces to be measured. Thus, the integrated overall and specific gas exchange rates of a leaf in direct, diffuse, and self-transmitted light were measured as follows:

In total, 24 plants were selected for the studies. The plants were measured in batches for either integrated or specific measurements. All measurements were performed on the fully-expanded leaves. Measurements were conducted from 9:00 to 12:00 in a leaf chamber set at 30°C, with 45% - 50% relative humidity, an ambient CO₂ concentration of 380 µmol mol^-1, and an irradiance of 1,000 μmol m^-2 s^-1. The light-response curves for gas exchange were investigated under direct, diffuse, and self-transmitted light, with the light intensity gradient of 1,200, 1,000, 800, 600, 400, 300, 250, 200, 150, 100, 50, 20, 0, 0 and 0 μmol m^-2 s^-1, three logging in 0 μmol m^-2 s^-1 was to obtain the steady gas-exchange in dark condition. Our preliminary study of leaf optical properties showed that the reflectance rates of either surface were 10%, and the absorbance rates of two surfaces were 85.4% under the LED light. We assumed that the absorbance rate of the two surfaces was the same, so the residual light reflected and intercepted by illuminating side was 47.7% of the incident light. Thus, the photosynthetic light-response curves under incident and self-transmitted light were obtained respectively by using the full and 47.5% of incident light intensity as the actual PPFD (PPFD _act). The intercellular CO₂ concentrations (C_i ) corresponded with the values of P_N and G_s in adaxial illumination and abaxial illumination, respectively. The ratios between the intercellular CO₂ concentration (C _i) and atmospheric (C _a) CO₂ concentration (C _i /C _a ratios) were presented in separate experiments. When the leaf was acclimated in 1,200 μmol m^-2 s^-1 for 30 min, the CO₂ supply in the chamber was controlled by a CO₂ supply system equipped with a CO₂ supplysystem equipped with LI-6400 to maintain the C _a near 310 µL L^-1. The values were logged when P _N and stomatal conductance (G_s ) reached a steady state, and C _a reached 310 µL L^-1. This study measured the boundary layer conductance with wet filter paper before the gas-exchange measurement. In the experiments, the boundary layer conductance at the flow rate of 500 μmol s^-1 was 0.87 mol m^-2 s^-1 and 0.96 mol m^-2 s^-1 for the adaxial and abaxial half-chambers, respectively. The apparent quantum yield (AQY) was the initial slope of the regression curve of PPFD _act versus P _N.

The SD was determined using the “leaf imprints” (Coupe et al., 2006). A fast-drying nitrocellulose and ethyl acetate were smeared evenly on the leaf surfaces to obtain a replica of the leaf surfaces. The replicas were observed under a light microscope (Olympus BH-2; Olympus Optical Co. Ltd, Tokyo, Japan), and a digital camera was used to photograph the replicas. Samples were taken from the same area of the leaves used for the gas-exchange measurements. The SDs (the number of stomata mm^-2) were determined by the numbers of stomata with the constant area. We measured the long axis of each stoma to determine the size because the short axes of stomata change depending on the degree of opening. In total, 100 marked leaves from 10 individual plants per treatment were selected for measurement.

The light-response curves of stomatal conductance were used for calculating stomatal sensitivities. Stomatal conductance increased as the light intensity increased. Stomatal conductance had a strong linear correlation with light intensity at low-light intensities (< 200 μmol m^-2 s^-1). The regression line of stomatal conductance (y) versus light intensity (x) was:

y= β x + b,

where β represents the initial slope of the regression line at low-light intensities, and the value of b was the steady stomatal conductance in the dark conditions. The value of β could reflect the stomatal insensitivity to light intensity (Wang et al., 2008).

Leaf sections (2 × 2 mm) were immersed in a fixative consisting of 1% (v/v) par-formaldehyde and 3% (v/v) glutaraldehyde in a 0.1 mM sodium phosphate buffer. The sections were subsequently washed, post-fixated, and dehydrated (Wang et al., 2019). Samples were embedded in Spur resin. Transverse semi-thin leaf sections (1 μm) were observed using a light microscope (Nikon-E800, Scientific Imaging Inc, USA).

The mesophyll thickness of two surfaces, the area of the sub-stomatal cavity per unit transection (S _a /S), and the chloroplast surface areas exposed to intercellular airspace per unit leaf area (S _c /S) were determined from the pictures. Moreover, chloroplast surface areas exposed to intercellular airspace (S _c) and the contact area between the bundle sheath and mesophyll cells (S _b) in transverse semi-thin leaf sections were determined by (Soares-Cordeiro et al., 2009; Wu et al., 2014) methods. S_b was calculated according to the following specific steps: (i) assuming that mesophyll cells were spheroids, following Thain (Thain, 1983), the total cell surface area per unit volume of tissue is equal to the total length of the cell profile perimeter in a unit area of a tissue section; (ii) the lengths and the widths of cells can be easily measured, and the curvature correction factors (F) can be determined by using length/diameter ratios, and (iii) the contact area of the chloroplast facing the outer mesophyll cells was considered as

S _b = L _c/W · F,

where L _c represents the lengths of chloroplasts facing the mesophyll cells, and W represents the width of the leaf cross-section (Wu et al., 2014). The value of F in mesophyll cells of sorghum leaves was 1.45, and the values of W were calculated using the pictures of transverse semi-thin leaf sections with Image J software.

The asymmetry index (ASI) was used in our study to reflect the degree of dorsoventral asymmetry. The ASIs for morphological and anatomic structures, and gas-exchange parameters, were calculated as the ratio of the value of the adaxial side to that of the abaxial side. By definition, if a given parameter shows the same value for both sides, its ASI is 1.00. Thus, the greater deviation of ASI from 1.00 denotes a greater degree of ASI.

Data were subjected to separate statistical analyses using parametric tests. Differences between means were analyzed using Student’s t - tests. Regression lines were obtained by the least squares method. Differences in the regression coefficient and in intercept were detected using an analysis of covariance. The multiple comparisons were assessed using Tukey’s test. Linear regression models were fitted to pairwise scatter plots to analyze the relationship between stomatal sensitivity β and SD for each individual surface. Data were analyzed using the Statistical Package for Social Sciences (SPSS, Version 18.0, for Windows). Correlations of linear regressions were calculated using Sigmaplot (version 14.0).

Our testing showed that the TF-covered treatment only changed the absorbance rate over 530 - 550 nm and 750 -1100 nm in adaxial illumination and 530 -550 nm and 590 – 620 nm in abaxial illumination ( Figure 2A ); the absorbance rate changed slightly (<5%) in rather a narrow band ( Figure 2B ). For this reason, it was considered that the leaf absorbance/reflectance rate in PAR (400 – 800 nm) was unchanged when covering the illuminated leaf surface with the TF.

The absorbance could be up to approximately 80% of incident light, whether adaxial or abaxial illuminating over the photosynthetically active region of the spectrum (400 - 800 nm) ( Figure 3A ). Further, more light illuminated into the leaf as the result of lower reflectance in case of adaxial illuminating (P<0.05) ( Figure 3B ), but more light transmitted from the leaf ( Figure 3C ), which caused the no-significance of absorbance between two surfaces. Consequently, there was no difference in the absorbance levels between the two surfaces of sorghum leaves over the photosynthetically active wavelengths.

Transverse sections observed by light microscopy showed that the sorghum leaves were not differentiated into palisade and spongy tissues. However, the long axes of the adaxial mesophyll cells were longer than those of the abaxial mesophyll cells. There were many motor cells in the adaxial epidermis but none in the abaxial epidermis. The vascular bundles were surrounded by parenchymal cells, and the paths from bundle sheath cells to both surfaces were nearly identical. The sub-stomatal cavities of the adaxial and abaxial mesophyll cells were separated by compact mesophyll tissues ( Figure S1 ). Both sides of leaves had stomata, and the SD of the abaxial surfaces were 1.5 times greater than those of the adaxial surfaces (P < 0.001), while the adaxial stomata lengths were approximately the same as the abaxial values ( Table 1 ). Further analyses showed that the S_a/S, S_c/S, and S_b of the abaxial surfaces were significantly greater than those of the adaxial surfaces (P < 0.01) ( Table 1 ).

The integrated P _N and G _s of the adaxial illumination were significantly higher than that of abaxial illumination both in direct and diffuse light, but the ASI of the integrated P _Nwas 2.83 in direct light, while significantly dropped to 1.69 ( Figure 4A ). In addition, the G_s changed as the P _N varied, and the ASI of the integrated G_s dropped from 2.23 in direct light to 1.67 in diffuse light ( Figure 4B ). The C _i /C _a remained nearly constant under direct and diffuse light conditions, and the G_s was not affected by the C _i ( Figure 4C ).

The specific G_s increased rapidly as the light intensities (PPFD _act) increased in direct light, including the direct incident light and self-transmitted in direct light ( Figures 5A, B ). The response of specific G_s to PPFD _act was a little lower in diffuse light than that in direct light ( Figures 5C, D ) . The adaxial specific G_s were greater than the abaxial specific G_s at most PPFD _act in direct incident light ( Figures 5A ), while there was no obvious variation of the specific G_s between adaxial and abaxial side in self-transmitted direct light ( Figures 5A, B ). The adaxial specific G_s showed lower than the abaxial specific G_s in diffuse light ( Figures 5B, D ).

The initial slope of the regression (β) reflected that the stomatal sensitivity. The β of adaxial surfaces was slightly greater (but significantly) than that of abaxial surfaces in direct incident light, while extremely lower than that of abaxial surfaces in self-transmitted in the direct light ( Table 2 ). These results suggested that stomatal sensitivity varied significantly between the adaxial and abaxial surfaces when exposed to direct light, and these differences depended on which surface was illuminated. However, there was no significant variation of β value between the two surfaces whether in incident light or self-transmitted light when the leaves were exposed in diffuse light ( Table 2 ), indicated that variation of the stomatal sensitivity between two leaf surfaces might be constant.

The adaxial specific P _N was greater than the abaxial specific P _N at the light below 600 μmol m^-2 s^-1, but lower than the abaxial specific P _N at the higher light intensity in direct light ( Figure 6A ), which was not consistent with the change of G_s ( Figure 5A ). Changes of adaxial specific P _N as the PPFD _act increased were consistent with G_s in diffuse and self-transmitted light ( Figures 6B–D ). The adaxial-specific AQY was significantly greater than the abaxial-specific AQY in direct light ( Figure 6A ) but was significantly lower in diffuse and two types of self-transmitted light ( Figures 6B–D ).

The number, size, and distribution of stomata are quite different in amphistomatic leaves, and amphistomaty has been associated with a greater capacity for gas diffusion because of the reduction in the pathway of the CO₂ diffusion from the atmosphere to sites of carboxylation. Significant morphological differences between the two surfaces might cause photosynthetic asymmetry in the amphistomatic leaves (Wang et al., 2020). Field environments change rapidly, and photosynthetic regulation is required to adapt to diverse light conditions because leaf photosynthesis is optimized through evolutionary fine-tuning. In our investigation, in spite of the same intensity, the diffuse light decreased the integrated adaxial P _N and G _s compared with direct light, but in abaxial ones, the opposite happened ( Figures 4A, B ). However, this phenomenon could not reverse the dorsoventral asymmetry as the ASI value greater than 1 owing to >3.5 billion years of evolution in leaf photosynthesis.

Here, we examined the great dorsoventral asymmetry of both direct and diffuse light, and the ASI was greatly higher in direct light (2.83) than in diffuse light (1.69). Leaf morphology and structure affect the photosynthetic capacity through SD, stomatal size, and movement (Soares-Cordeiro et al., 2009), as well as mesophyll tissue (Oguchi et al., 2005; Earles et al., 2017) and vascular bundle (Oguchi et al., 2005; Soares et al., 2008) structures. Photosynthesis in C₄ plants requires the collaboration of mesophyll cells and vascular bundle sheath cells. The gas exchange rate between these two types of cells plays a crucial role in photosynthesis (Sowinski et al., 2008). Here, we measured the S_a/S, S_c/S, and S_b values between adaxial and abaxial mesophyll sides, which are important in determining P _N. The greater SD, S_a/S, S_c/S, and S_b values in the abaxial leaf side compared with the adaxial side demonstrated that the gas-supply potential was greater in the former than the latter ( Table 1 ). The photosynthetic capacities of C₄ isobilateral leaves exposed to direct light tend to be greater in abaxial surfaces, owing to the higher SD and higher G _s (Long et al., 1989; Driscoll et al., 2006; Soares et al., 2008; Soares-Cordeiro et al., 2011). Our results showed that the integrated adaxial P _N was higher than the abaxial one ( Figure 4A ), and the abaxial specific P _N was higher than or the same with adaxial one, respectively in direct (in high light) and diffuse light ( Figures 6A, B ). These findings could imply that self-transmitted light was of sufficient intensity to be effective for driving photosynthesis.

In addition to CO₂-supply capacity, light-use capacity, which involves light absorptance and photochemical activity, is also a key trait in determining the P _N. Leaf optical properties reflect the proportion of the light incident into leaves, which may be influenced by surface properties, pigment content, and leaf morphology (Gorton et al., 2010). However, our data demonstrated that leaf absorptance and reflectance profiles were almost the same on the adaxial and abaxial surfaces in sorghum over most of the 400-800 nm wavelength range ( Figure 3A ). This result was consistent with other studies on isobilateral leaves (Soares et al., 2008; Brodersen and Vogelmann, 2010; Gorton et al., 2010). In fact, the light environments of the two surfaces were more complex, whether illuminated on the adaxial or abaxial side. For example, the light that illuminated on the abaxial surface was rarely direct light but usually diffuse and self-transmitted light. In our experiments, the integrated P _N of illuminated the adaxial surfaces was greater than those of illuminated on abaxial leaf surfaces ( Figure 4A ), which was inconsistent with previous reports (Long et al., 1989; Driscoll et al., 2006; Soares et al., 2008; Soares-Cordeiro et al., 2011). Actually, the integrated P _N and G _s, shown in Figure 4 , was the sum of two surfaces, whether the incident light illuminated on the adaxial surfaces or the abaxial ones. For example, in direct light, the integrated P _N illuminating on the adaxial surface (the black bar in Figure 4A ) was the integration of the adaxial specific P _N in direct light ( Figure 6A , closed circle) and the abaxial specific P _N in self-transmitted light ( Figure 6C , open circle), while the integrated P _N illuminating on the abaxial surface (the open bar in Figure 4A ) was the integration of the abaxial specific P _N in direct light ( Figure 6A , open circle), and adaxial specific P _N in self-transmitted light ( Figure 6C , closed circle). Thus, we speculated that the greater photosynthesis on the adaxial illuminating ( Figure 4A ) was due to the higher adaxial specific P _N and AQY ( Figure 6A ) in direct incident light and slightly changed adaxial specific P _N and AQY in self-transmitted direct light ( Figure 6C ). Although the greater photosynthesis on the adaxial illuminating was observed than that on the abaxial illuminating ( Figure 4A ) in diffuse light, the ASI dropped from 2.83 to 1.69 compared with that in direct light. We can apply the same reasoning that the higher photosynthesis but reduced ASI (from 2.83 to 1.69) on the adaxial illuminating was due to significantly higher abaxial specific P _N and AQY in self-transmitted diffuse light and lower adaxial specific P _N and AQY in diffuse incident light. Thus, the apparent conflicts with photosynthetic asymmetry in previous conclusions (Long et al., 1989; Wang et al., 2008; Soares-Cordeiro et al., 2009; Urban et al., 2012; Cheng et al., 2015; Earles et al., 2017) might be caused undistinguished light properties. The results in Figure 6 also showed that the adaxial specific P _N could be higher than abaxial one, indicated that symmetric photosynthetic characteristics of sorghum leaves might be changed by light intensity, which was consistent with the results in tobacco (Wang et al., 2020).

The CO₂ diffusion was nearly unaffected by the opposite side, even though the stoma was blocked from the outside, owing to restriction between two sides by the compact mesophyll cells. This was further supported by the ratio of the integrated G _s to the sum of specific G _s responses shown in Figure 7 , the calculated specific G _s in incident light + the G _s of the opposite side in self-transmitted light was closely corresponded with the integrated G _s, respectively in adaxial illuminating (R² = 0.9415, P < 0.0001, slope = 0.9657) and abaxial illuminating (R² = 0.9640, P < 0.0001, slope = 1.0086), which was not significantly different from 1:1.

As a major pathway of the gas exchange between the atmosphere and the plant, stomata plays an important role in the variation of adaxial and abaxial photosynthesis (Yu et al., 2004; Xu and Zhou, 2008). Light intensity arrived at the leaf surface is the most important ecological factor in stomata opening, despite the fact that stomata opening has been observed in the darkness in some species (Shimazaki et al., 2007). Although the SD was significantly greater in abaxial surfaces than in adaxial surfaces ( Table 1 ), the abaxial specific G _s were lower than the adaxial specific G _s ( Figure 5A ) in the direct light. Stomatal sensitivity, as indicated by β, in leaf surfaces has an important role in G _s(Goh et al., 2002; Wang et al., 2008), and the differences in the specific β between adaxial and abaxial surfaces were consistent with variations of G _s versus incident light ( Table 2 ). The results in self-transmitted light were consistent with previous conclusions (Wang et al., 2008), but the specific β was greater in adaxial surfaces in direct light and nearly the same in the two surfaces in diffuse light ( Table 2 ). Numerous in vitro investigations indicated that the stomatal sensitivity to light is greater on the abaxial surface than adaxial surface (Goh et al., 2002; Wang et al., 2008). The different conclusion might be due to the distribution of the stomatal apparatus (stomatal size and SD) on the two surfaces and the incident light, which might have considerable impacts on β. In our investigation, the stomata in both leaf surfaces were similar in size ( Table 1 ), and β had strong correlations with SDs of adaxial ( Figure 8A ) and abaxial ( Figure 8B ) surfaces from different sorghum leaves. Hence, even though there was no direct effect of SD on specific G _s, the specific β could be affected by the SD.

When the leaf was illuminated with polychromatic light, mesophyll cells of the illuminated side selectively absorbed red and blue light, while the remaining light that arrived at the opposite side consisted mostly of green light. Green light might be more efficient for stomata opening (Wang et al., 2008; Terashima et al., 2009). Wang et al. (2008) showed that the self-transmitted and leaf-transmitted light were more efficient in opening the abaxial stomata of sunflower leaves than the direct light, which was explained by the mostly green light in transmitted light (Wang et al., 2008; Falcioni et al., 2017). Consequently, discrepancies in the specific β in self-transmitted light might be cased variations in the spectral composition of incident light. Estimations of leaf stomatal sensitivities should take leaf surface traits and the light spectrum into consideration.