Vegetatively propagated A. kolomikta vines were used for experiments at Changchun, China (43°48′45″ N, 125°24′15″ E). The study site receives annual precipitation of 617 mm, and the average annual maximum and minimum temperatures are 37°C and −35°C, respectively. The plants were exposed to full sunlight (the maximum photosynthetic photon flux density [PPFD] on a clear day is ~1,200 ± 128 μmol m⁻² s⁻¹). The maximum light intensity showed no significant change during leaf development. All field trials were conducted from May to July from 2019 to 2021. The mean humidity during the experimental period was 85% at night and 60% in the daytime. Each field plot was divided into three subplots, and the seedlings were planted in each plot (3 m × 4 m). The nitrogen (30.65 ± 2.06 g/Kg), phosphorous (3.44 ± 0.26 g/Kg) and potassium (3.48 ± 0.26 g/Kg) concentrations were sufficient. Seedlings in all experiments were of uniform size. White and green leaves were sampled during late May and early June. Investigation trials were conducted at Changchun and in natural habitats.

Net photosynthetic rate (P_n), stomatal conductance (G_s), and intercellular CO₂ concentration (C_i) were measured under ambient conditions with a portable photosynthesis system (CIRAS-2; PP-Systems, Hitchin, UK). The photosynthesis parameters were recorded at 2-hour intervals between 05:00 and 17:00 on sunny days.

Data for the photosynthetic light response curve were recorded from three white leaves between 08:00 and 11:00 (to avoid midday depression of photosynthesis). Measurements were recorded separately on the adaxial and abaxial leaf surfaces. The PPFD was gradually decreased stepwise using an integrated light-emitting diode (LED) light source and ranged from 1,200 to 0 μmol m⁻² s⁻¹. The net photosynthetic rate (P_n) at each PPFD was recorded when it was stable (usually 2–3 min). Quantum use efficiency (apparent quantum yield, AQY) was calculated according to the initial slope of the linear (light-limited) portion of the light-response curve between PPFD of 0 and 200 μmol m⁻² s⁻¹. Initial carboxylation efficiency (CE) was calculated according to the initial slope of the linear (light-limited) portion of the CO₂-response curve between CO₂ of 0 and 200 mmol mol⁻¹.

To understand the type of white leaf and green leaf, leaf structure and ultrastructure were observed using light microscopy and transmission electron microscopy (TEM). Sample preparation for semi-thin sections and TEM was performed in accordance with previously described methods (Konoplyova et al., 2008; Sheue et al., 2012). Semi-thin sections were stained with 1% toluidine blue and observed under a light microscope (ECLIPSE 80i, Nikon Corporation, Tokyo, Japan; Sheue et al., 2012; Wang et al., 2015). Photomicrographs were captured using a Zeiss Axiolab with a digital camera (DXM1200, Nikon). Ultrathin sections were observed using a TEM (JEM-1200EX; JEOL Ltd., Tokyo, Japan) at 80 kV. Leaf thickness parameters were measured using an ocular micrometer and measurements were rounded to the nearest 0.1 μm.

During leaf development, leaves were sampled at pre-dawn (03:00–05:00). Chlorophyll a fluorescence transience (OJIP) was measured with a Plant Efficiency Analyzer (Pocket-PEA, Hansatech Instruments, King’s Lynn, United Kingdom). Saturating red light of 3,000 μmol m⁻² s⁻¹ was produced by an array of four LEDs (peak wavelength 650 nm). Fluorescent signals were recorded within a time scan from 10 μs to 1 s with a data acquisition rate of 100 readings ms⁻¹ for the first 2 ms, and 1 reading ms⁻¹ after 2 ms. The maximum quantum yield of PSII (F_v/F_m) was calculated in accordance with the method of Strasser (1997). The F_v/F_m was determined using a 1 s pulse of red radiation (3,500 μmol m⁻² s⁻¹) on fully dark-adapted (for 1 h) samples. The following parameters from the raw fluorescence measurements were calculated: minimal fluorescence of the dark-adapted state (F₀); the maximal fluorescence of the dark-adapted state (F_m); and variable total chlorophyll fluorescence yield (F_v), defined as F_m − F₀. Quantum yield for electron transport, ET₀/ABS = [1–(F₀/Fm)] × (ET₀/TR₀). The efficiency of electron flow beyond QA^–, Ψ₀ = 1–V_J.

Autofluorescence imaging and determination of F_v/F_m were performed in accordance with previously described methods (Jacobs et al., 2016). Cross sections were prepared from dark-adapted leaves with a Plant Microtome (NK SYSTEM, Japan). Fluorescence images were captured with a fluorescence kinetic microscope equipped with FluorCam7 software (Photon Systems Instruments, Czech Republic). Initial fluorescence (F₀) was recorded in leaves adapted to the dark for 1 h. A 0.8 s pulse of saturating white light (>4,000 μmol m⁻² s⁻¹) was applied to determine the maximum fluorescence (F_m). The F_v/F_m was calculated following the procedure of Jacobs et al. (2016).

Total RNA was extracted in triplicate from white leaves and green leaves (the control). Illumina paired-end read libraries were constructed using the NEBNext^® Ultra RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, United Sates), and sequenced using an Illumina HiSeq 2000 platform (GENEWIZ, Inc.). Trimming was done to obtain ≥50 million paired reads per sample. The remaining clean reads were assembled using Trinity, which is a novel method for efficient and robust de novo reconstruction of the transcriptome from RNA sequencing (RNA-seq) data. Gene expression levels were calculated using the reads by kilobase per million (RPKM) method. Differentially expressed genes (DEGs) were identified using the false discovery rate (FDR) procedure (Robinson et al., 2010). Selection of DEGs was based on two criteria: (i) FDR ≤ 0.05, and (ii) fold change ≥ 2.0 for green leaves vs white leaves. Significantly enriched gene ontology (GO) terms were determined using GOseq using a hypergeometric test. We applied p < 0.05 as the cutoff for a significantly enriched GO term. The Kyoto Encyclopedia of Genes and Genomes (KEGG) is a collection of databases for genomes, biological pathways, diseases, drugs, and chemical substances¹. We used in-house scripts to identify significantly enriched DEGs in KEGG pathways. The criterion Q ≤ 0.05 was applied for significantly enriched KEGG pathways.

To validate the reliability of RNA-Seq, qRT-PCR for transcripts was carried out as described by Sun et al. (2019). qRT-PCR was performed using the SYBR Premix Ex Taq^TMII (TaKaRa, Bio., Dalian, China) according to the manufacturer’s protocol. The relative expression levels of the selected genes were calculated using the 2^−ΔΔct method (Zhang et al., 2022). The sequences of the primers used for qRT-PCR are listed in Supplementary Table S1.

Statistical analysis was performed using JMP 6.0 software (SAS Institute, Cary, NC, United States). The variables analyzed were as follows. Gas exchange data reported are the mean of at least individual measurements of five leaves. The photosynthetic light response curve data represent the means of individual measurements of three leaves. Chl a fluorescent parameters represent means individual measurements of eight leaves. Autofluorescence imaging were calculated and are presented as the average of five independent measurements. qRT-PCR is presented as the average of three independent measurements.

Diurnal variation of P_n in green leaves exhibited a double-peak pattern with peaks observed at 09:15 and 16:45 (Figure 2A). The diurnal trend of P_n in white leaves was similar to that observed in green leaves (Figure 2A). Although P_n in green leaves was higher than that of white leaves, the value of P_n in white leaves was 82% of that in green leaves at 09:15 (Figure 2A). The diurnal trend of G_S in green leaves and white leaves also exhibited a double-peak pattern, but the G_S of green leaves was higher than that of white leaves (Figure 2B). Diurnal variation of C_i in green leaves and white leaves showed almost no significant difference (Figure 2C).

To exclude the impact of light intensity on photosynthesis, P_n under different PPFDs was measured on the adaxial and abaxial leaf surfaces. The light-saturated P_n of green leaves was higher than that of white leaves measured on the adaxial surface (Figure 3A), but little difference in light-saturated P_n values between green and white leaves was observed when measured on the abaxial surface (Figure 3B). The AQY and CE of photosynthesis of white leaves was 51.8 and 41.2% lower than that of green leaves on the adaxial surface (Figures 3C,D). The P_n and AQY of green and white leaves showed no significant differences when measured on the abaxial surface (Figure 3C), but CE of white leaves was 72% higher than that of green leaves on the abaxial surface (Figure 3D).

The anatomy of green leaves and white ones differed distinctly (Figure 4). The palisade mesophyll of green leaves consisted of a well-established layer of packed and elongated cylindrical cells (Figure 4B) that were in tight contact with the adaxial epidermis. However, in white leaves, only rounded or irregularly shaped cells were observed (Figure 4A), rather than the typical palisade-type cells. Intercellular spaces were observed between the epidermal and mesophyll cells or within the mesophyll cell layer in white leaves (Figure 4A). In cross section, the palisade and spongy mesophyll tissues of white leaves were thicker than those of green leaves (Figure 4G). Compared with green leaves, the thickness of the palisade and spongy mesophyll of white leaves increased by approximately 73% and 91%, respectively.

Observation of the chloroplast ultrastructure by TEM revealed that palisade and spongy parenchyma cells of green leaves contained normal chloroplasts with abundant thylakoid membranes and starch grains (Figures 4E,F). The number of starch grains in palisade parenchymal cells was less than that in spongy parenchymal cells (Figures 4E,F). In contrast, chloroplast development of palisade parenchyma cells in white leaves was abnormal and impaired, and some chloroplast membranes were absent or appeared faint (Figure 4C). Compared with palisade parenchyma cells of white leaves, spongy parenchymal cells of white leaves contained regular and functional chloroplasts with abundant grana and obvious starch grains (Figure 4D).

The values of F₀ and F_m of green leaves were about 4.4-fold and 4.8-fold higher, respectively, compared with those of white leaves measured on the adaxial surface (Figures 5A,B). However, F₀ and F_m in green leaves measured on the abaxial surface were about 81% and 72%, respectively, of those observed in white leaves (Figures 5A,B). The F_v/F_m of green leaves was higher than that of white leaves when measured on the adaxial surface, whereas F_v/F_m in green leaves was lower than that of white leaves when measured on the abaxial surface (Figure 5C). Efficiency of electron flow beyond QA^–(Ψ₀) of white leaves was significantly higher than in green leaves (Figure 5D). Quantum yield for electron transport (ET₀/ABS) of white leaves on the adaxial surface was similar to green leaves, but on the abaxial surface, ET₀/ABS of white leaves was significantly higher than in green leaves (Figure 5E).

Direct insight into the photosynthetic efficiency of different mesophyll tissues in leaf cross section can be gained by chlorophyll fluorescence imaging (Figures 6A,B). The fluorescence intensity of the palisade mesophyll in green leaves was significantly higher than that of the spongy mesophyll (Figure 6B), whereas the fluorescence intensity of the palisade mesophyll in white leaves was significantly lower than that of the spongy mesophyll (Figure 6A). The values of F₀ and F_m of the palisade mesophyll in green leaves were approximately 5.3-fold and 7.4-fold higher, respectively, than those of white leaves (Figures 6C,D). However, F₀ and F_m of the spongy mesophyll in green leaves was 79% and 64%, respectively, of that of white leaves (Figures 6C,D).

The F_v/F_m of the different mesophyll tissues in green leaves and white leaves differed significantly. The F_v/F_m of the palisade mesophyll in green leaves was 136% higher (p < 0.05) than that of white leaves (Figure 6E), whereas the F_v/F_m of the spongy mesophyll of green leaves was 68% of that of white leaves (Figure 6E).

To gain insight into changes in expression of photosynthesis-associated genes, a comprehensive transcriptome analysis comparing white leaves and green leaves at flowering was conducted. The RNA-seq analysis identified 501 DEGs (q-value <0.001). Compared with green leaves, 215 and 286 genes were exclusively upregulated and downregulated, respectively, in white leaves. To further investigate the biochemical pathways in which these DEGs participate, we analyzed all DEGs according to terms in the KEGG database. 105 genes had KEGG annotations and were categorized into 28 pathways (Figure 7). Significant enrichment of DEGs associated with organismal systems, metabolism, genetic information processing, environmental information processing, and cellular processes was observed. In particular, three pathways were significantly enriched: genes involved in carbon metabolism, starch and sucrose metabolism, and carbon fixation in photosynthetic organisms were significantly upregulated (Figure 7).

To detect genes associated with photosynthesis, we identified seven DEGs that had been previously associated with photosynthetic electron transport and carbon assimilation (Table 1). These genes encode a photosystem II cp43 protein (CP43), Chloroplast Nucleoids DNA-binding Protease (CND41), NAD(P)H-quinone oxidoreductase subunit 5 (NDHF), glyeraldehyde-3-phosphate dehydrogenase (GAPDH), RuBisCO activase (RCA), and sucrose-phosphate synthase (SPS). Some important genes involving photosynthesis were upregulated in white leaves, for example, GAPDH, RCA, and SPS.

To confirm the RNA-Seq data, we quantified 3 transcripts by qRT-PCR between white leaf and green leaf, which mainly include sucrose metabolism, photosynthesis, and antioxidant-related genes. The expression of GAPDH, RCA, and SPS of white leaf was 14-, 1.3-, and 3.9-fold higher than green leaf respectively (Figures 8A,B,D), the expression of CND41 of green leaf was 1.8-fold higher than white leaf (Figure 8C). These results from the real-time PCR matched the RNA Seq data very well, thus confirming the expression pattern of those genes in white leaf and green leaf (Figure 8).

Plants with non-green leaves (such as white or red leaves) show a significant reduction of photosynthetic capacity compared with that of plants with green leaves (Burger and Edwards, 1996; Gould et al., 2002; Hughes et al., 2005; Zeliou et al., 2009). In some cases, photosynthetic function was almost completely lost in non-green leaves of Arabidopsis (e.g., the Var variegation; Yu et al., 2011; Lundquist et al., 2014) and Ficus pumila “Sonny” (Sheue et al., 2012). It is proved that some specific ecological functions of non-green leaves are usually at the expense of photosynthesis. For example, white bracts of Davidia involucrata and Saururus chinensis maintained an extremely low P_n, although the plants attracted significantly higher numbers of pollinators (Karageorgou et al., 2008; Sun et al., 2008; Zhao et al., 2013; Song et al., 2018). It is reported that approximately 2,500–3,000 proteins and enzymes are involved in the photosynthetic process (Motohashi et al., 2003; Eberhard et al., 2008; Ouyang et al., 2011), and expression of the genes encoding these proteins and enzymes is correlated with photosynthetic capacity (Müller et al., 2010). In the present study, we detected only seven DEGs associated with photosynthetic electron transport and carbon assimilation (Table 1). This finding implied that the majority of photosynthesis-associated genes showed no significant difference in expression between white leaves and green ones.

More importantly, the present study demonstrated that the white leaves of A. kolomikta still maintained a relatively high P_n (Figures 2, 3), and the highest P_n ratio of white leaves to green ones was 82% over the course of the day (Figure 2A). In addition, the total P_n of white leaves over the course of the day was 70% of that of green ones. Notably, white leaves comprise almost 42–57% of the total leaf number during blossoming. Therefore, the photoassimilates from white leaves on a reproductive branch might effectively support flower development on that branch.

The present results also demonstrated that the P_n of white leaves and green leaves under different PPFDs differed on the adaxial leaf surface (Figure 3). Although the P_n of white leaves and green ones under saturated light showed little difference, the P_n of green leaves was significantly higher than that of white leaves under low light intensity. Thus, the AQY of green leaves was significantly higher than that of white ones on the adaxial leaf surface (Figure 3A). Significant decreases of G_s in white leaves were associated with changes in C_i (Figures 3B,C), which suggested that the reduction of P_n in white leaves was caused by non-stomatal limitation.

Leaf structure is an important determinant of leaf photosynthetic characteristics (Green and Kruger, 2001). However, contrary to previous studies, the present results showed that functional chloroplasts were not developed in palisade parenchyma cells (Figure 4C), although the palisade mesophyll of white leaves was significantly thicker than that of green ones (Figure 4G). In contrast, functional chloroplasts and higher abundance of starch grains were observed in the spongy mesophyll cells of white leaves (Figure 4D). These structural traits suggested that photosynthetic activity of the spongy mesophyll in white leaves might be greater than that of the palisade mesophyll. Compared with green leaves, although the fluorescence intensity and F_v/F_m on the adaxial surface of white leaves were significant lower (Figure 5C), Ψ₀ and ET₀/ABS were significant higher or similarity. Furthermore, these parameters measured on the abaxial surface of white leaves were significantly higher than those of green ones (Figure 5C). Photochemical activity on the adaxial and abaxial sides of leaves mainly reflect photosynthetic activity of the palisade and spongy mesophyll tissues (Peguero-Pina et al., 2009; Wu et al., 2020). Therefore, higher fluorescence intensity, F_v/F_m and quantum yield for electron transport on the abaxial side of white leaves (Figure 5) demonstrated indirectly that the photosynthetic apparatus and photochemical activity of the spongy mesophyll in white leaves were enhanced. The data from fluorescence kinetic microscope with leaf cross section further confirm that the fluorescence intensity and F_v/F_m of the spongy mesophyll in white leaves were significantly higher than those of green ones (Figure 6). These results reflected that the spongy mesophyll of white leaves may compensate the constraints to the photosynthetic apparatus and photochemical activity in the palisade mesophyll. The structural traits and photochemical activity in white leaves may explain that the photosynthetic capacity of white leaves and green ones showed little difference when the P_n under different PPFDs was measured on the abaxial leaf surface. These results showed that the photosynthetic apparatus is located predominantly in the palisade mesophyll in green leaves, whereas in white leaves the spongy mesophyll accumulated more components of the photosynthetic apparatus. In addition, the lower AQY and CE on the adaxial side of white leaves might be due to higher light reflection and abnormal chloroplast development in palisade mesophyll cells of white leaves, which resulted in a lower light energy utilization ratio. However, AQY and CE on the abaxial side of white leaves was slightly higher than that in green leaves (Figure 3C). These data further suggested that the higher light energy utilization ratio of the spongy mesophyll facilitated maintenance of the photosynthetic capacity of white leaves. Therefore, changes of the leaf structure in white leaves may greatly influence the photochemical activity and carbon assimilation capacity of different cell layers.

Previous studies suggest that white areas of Begonia formosana leaves can maintain higher photosynthetic capacity (Sheue et al., 2012), which is the result of palisade mesophyll cells in white areas containing functional chloroplasts. In addition, in some plants, for example, Ranunculus ficaria and Arum italicum (Konoplyova et al., 2008; Rocca et al., 2011), white leaves possess two or three palisade cell layers, which effectively maintain the photosynthetic capacity of white leaves. However, for A. kolomikta, the specialized structure of white leaves suggested that the spongy mesophyll compensated for lower photosynthetic capacity of the palisade mesophyll and that the functioning of the spongy mesophyll as the main photosynthetic tissue plays a vital role in maintenance of the photosynthetic capacity of white leaves. Thus, the photosynthetic mechanisms of white leaves in A. kolomikta differ significantly from those of white leaves in other plant.

Based on the analysis of DEGs between white leaves and green ones, seven of the eight photosynthesis-related DEGs were downregulated (Table 1). These genes regulate expression of proteins or enzymes involved in photosynthetic electron transport and carbon assimilation. Therefore, these downregulated genes may be associated with a slight decrease in photosynthetic capacity of white leaves. GAPDH and SPS are two key enzymes involving the Calvin–Benson cycle and sucrose biosynthesis (Sun et al., 2022). Rubisco activase (RCA) can keep catalytic active state of Rubisco (Zhang et al., 2022). Chloroplast Nucleoids DNA-binding Protease (CND41) is involved in Rubisco degradation (Kato et al., 2005). In this study, RCA and GAPDH gene was upregulated (Table 1 and Figures 8A,B), which suggested higher Rubisco activity and CO₂ fixation efficiency. Higher expression of RCA and SPS may match higher CE and obvious starch grains on the abaxial side of white leaves. Upregulation of SPS promoted accumulation of starch and sucrose (Figure 8D). Meanwhile, downregulation of CND41 can avoid Rubisco degradation (Figure 8C). This regulation of transcriptional level may compensate for the lowered photosynthetic capacity of the palisade mesophyll, which could be important in the maintenance of photosynthetic capacity in white leaf.