Relationships between leaf color changes，pigment levels，enzyme activity，photosynthetic fluorescence characteristics and chloroplast ultrastructure of Liquidambar formosana Hance

Guoping Yin·Yong Wang·Yufei Xiao·Jisheng Yang·Renjie Wang·Ying Jiang·Ronglin Huang·Xiongsheng Liu·Yi Jiang

Abstract Liquidambar formosana Hance is an attractive landscape tree species because its leaves gradually change from green to red，purple or orange in autumn.In this study，the red variety of L.formosana was used to establish a quantitative model of leaf color.Physiological changes in leaf color，pigment levels，enzyme activity，photosynthetic fluorescence characteristics and chloroplast ultrastructure were monitored.The relationship between leaf color and physiological structure indices was quantitatively analyzed to systematically explore the mechanisms behind leaf color.Our data showed that with a decrease in external temperatures，chloroplast numbers and sizes gradually decreased，thylakoid membranes became distorted，and chlorophyll synthesis was blocked and gradually decreased.As a result，chloroplast membranes could not be biosynthesized normally;net photosynthesis，maximum and actual photochemical efficiency，and rate of electron transfer decreased rapidly.Excess light energy caused leaf photoinhibition.With intensification of photoinhibition，leaves protected themselves using two mechanisms.In the first，anthocyanin synthesis was promoted by increasing chalcone isomerase and flavonoid glycosyltransferase activities and soluble sugar content so as to increase anthocyanin to filter light and eliminate reactive oxygen species to reduce photoinhibition.In the second，excessive light energy was consumed in the form of heat energy by increasing the non-photochemical quenching coefficient.These processes tuned the leaves red.

Keywords Anthocyanins·Leaf color parameter·Enzymatic activities·Chlorophyll fluorescence characteristics·Liquidambar formosana

Introduction

With increasing demands for ornamental plants，color and brightness are prominent，with multi-color plant forms favored by consumers.Colored trees are attractive because of their bright，rich colors and high ornamental value，and contribute considerably to modern urban landscapes.Mechanisms behind leaf color changes have become a major research focus (Hughes 2011;Li et al.2019;Zhao et al.2020).In recent years，new advances have been made in the study of leaf color changes in colored-leaf species，including theoretical research related to ecological significance，physiological pigment levels，leaf structure and molecular genetic research，micro RNA and transcription factors (Li et al.2016;Ma et al.2018).Methods have incorporated traditional physiological and biochemical approaches to high-throughput sequencing，transcriptome，small RNA analysis，identif ciation and genetic transformation of related functional genes in molecular biology (Noda et al.2017;Xu et al.2019).

TheLiquidambar formosanaspecies of the genusLiquidambaris a widely distributed deciduous species which exhibits strong adaptability，fire resistance and regenerates naturally.In autumn，its leaves gradually change from green to red，purple，orange and other colors，but primarily red.The species is of considerable ornamental value and is an excellent landscape species (Sun et al.2016).A previous study showed that anthocyanin is the main pigment involved inL.formosanaleaves turning red (Luo et al.2 017).Several studies have shown that this pigment is the main basis of leaf color (Feild et al.2001;Junker and Ensminger 2016;Moustaka et al.2020).Anthocyanins are water-soluble flavonoids that vary greatly due to the different hydroxyl or methoxy groups in their core structures (Moustaka et al.2018).Different biotic and abiotic stresses can induce different anthocyanin combinations (Kovinich et al.2014).Research has indicated that anthocyanins absorb blue-green light and reduce the quality and quantity of light captured by chlorophyll and carotenoids during leaf senescence (Hoch et al.2001).This reduces the degree of photoinhibition and the damage induced by oxygen free radicals and superoxides produced by photosynthesis during photoinhibition or light damage，which benefits the reflux of nutrients in old leaves (Hughes 2011;Menzies et al.2016).However，some researchers believe that anthocyanin accumulation regulates osmotic substances and improves plant resistance to low temperatures，drought，nutrient deficiencies and other stresses (Sherwin and Farrant 1998;Stiles et al.2007).Anthocyanins may accumulate in plant leaves，making them red to conduct chemical defenses (anthocyanins have a pungent taste and can protect against predators) (Sinkkonen et al.2012)，concealment (most herbivorous insects are insensitive to red color，and the red leaves are difficult for predators to find) (Zeliou et al.2009)，and warning (red colour of leafis a visual warning to approaching insects，insects perceive this signal and thus avoid consuming red leaves) (Esteban et al.2008;Cooney et al.2012).Controversies related to the function of anthocyanin in leaves have focused on light damage and ecological defenses，and stress signal transduction.Wang et al.(2019) described pigment changes inL.formosanaleaves during color changes under different light quality treatments.Liu et al.(2017)also described the relationship between leaf color changes and pigments.Similarly，Luo et al.(2017) indicated that the photosynthetic capacity ofL.formosanaleaves decreased gradually during senescence and becoming red.These studies not only reported that anthocyanin was the main pigment responsible for coloration inL.formosanaleaves，but they also speculated that the main function of anthocyanin was to defend against light damage.However，this inference cannot be confirmed due to the lack of systematic research on the physiological mechanisms of leaf color changes inL.formosana.

In view of these issues，to demonstrate the physiological function of anthocyanin in the leaves ofL.formosana，theL.formosanared plant was used as the research focus to continuously monitor physiological and structural changes in leaf color，pigment，enzyme activities，photosynthetic fluorescence characteristics and chloroplast ultrastructure during the process of turning red under natural conditions.This quantitative model was established using the International Commission on Illumination LAB color space to quantify the original fuzzy leaf color description and to analyze the relationship between leaf color and changes of physiology and structure.The mechanisms and physiological functions of anthocyanins inL.formosanaleaves were explored systematically.The results will provide quantitative data analysis and theoretical guidance for in-depth research on the color mechanisms ofL.formosanais，and provide a basis for its utility in landscaping.

Materials and methods

Experimental design

Experimental materials were obtained from anL.formosananatural forest (23°21′19"N，106°39′5"E) in Hongye Forest Park，Debao County，Baise City，Guangxi Province，China.Hongye Forest Park is located at south of the Tropic of Cancer.It has a subtropical humid climate，with mainly brown soil covering 729.4 ha，including 243.8 ha ofL.formosanaforest.Changes in daily average temperatures，humidity and light intensity during theL.formosanacoloration period are shown (Fig.1).

Fig.1 a Daily mean temperatures，relative air humidity; b light intensity during the L.formosana leaf coloration period

In autumn 2018，with reference to Junker and Ensminger(2016)，we selected a sunny slope in Hongye Forest Park and built three 20 m×50 m temporary sample plots，measured the DBH and tree height of all adultL.formosanain the sample plot，and then calculated the average value.TheL.formosanaplants with DBH and tree height close to the average value were selected as the objects.Five matureL.formosanatrees with red leaves were selected，and at the end of September，we selected one branch in each direction of east，west，south and north of each selected tree，and make sure that all four branches were at the same height.Then 30 healthy and intact leaves were selected randomly from each selected branch and marked by hanging a tag.A total of 120 leaves were marked per tree.Every 15-20 days，the leaf color of these 120 leaves was measured，after which 5 of the 30 leaves from each marked branch were selected to determine photosynthesis and chlorophyll fluorescence.In addition，10 healthy，intact leaves from each branch in the four directions were randomly collected，in total，40 leaves were collected per tree.These leaves were mixed and evenly divided into two groups;one group was wrapped in tin foil and temporarily preserved in liquid nitrogen.They were analyzed for pigment content，enzyme activities，total soluble sugar (TSS)，and total soluble protein (TSP).Leaves in the other group were cut along the main vein into 2 mm long and 1 mm wide pieces and fixed in 2.5% glutaraldehyde for ultrastructural determination.This was carried out five times from the start of the study to the end (Fig.2).

Fig.2 Liquidambar formosana leaves at different coloration stages.S1，green (29th September 2018);S2，minor reddish areas＜1/3 (13th October 2018);S3，some reddish areas ＞ 1/3，＜2/3 (2th November 2018);S4，reddish areas ＞ 2/3 (22th November 2018);and S5，all red (12th December 2018);Yellow circle in S1 stages represents the measuring point of leaf color parameters

Leaf color determination

Upper surface leaf color was measured using a spectrophotometer CS-650 (Hangzhou CHNSpec Technology Company Limited，China) under a D65 light source using an 8-mm diameter window and an angle of 10°.Each leaf was measured 10 times (Fig.1 S1).The following parameters were measured:brightness，L*(the larger theL*，the brighter the color)，red and green attributes，a*(the larger thea*，the more red color)，yellow and blue attributes，b*(the larger theb*，the more yellow color)，saturation，C*(the larger theC*，the brighter the leaf color)，and the hue angleh°(the smaller theh°，the redder the leaf) (Sarker and Oba 2018).

Determination of pigment，enzyme activity，and soluble sugar and protein levels

Chlorophyll and carotenoid (Car) determination according to Lichtenthaler and Wellburn (1983) was used as a reference.A 0.2 g fresh leaf powder was weighed and 5 mL 80% acetone added，and extracted in a refrigerator for 24 h at 4°C.After supernatant filtration，absorption values at 445 nm，645 nm，and 663 nm were recorded on a UV-4802 double beam spectrophotometer (Unoco (Shanghai) Instrument Company Limited，China)，after which chlorophylla(Chla)，chlorophyllb(Chlb)，total chlorophyll content (Chla+b)and Car levels were calculated.

The Marquis of Carabas did what the Cat advised him to, without knowing why or wherefore. While he was washing the King passed by, and the Cat began to cry out:

For anthocyanin determination，1.0 g fresh leaf powder was mixed with 10 mL 1% hydrochloric acid methanol solution，and the mixture was extracted for 5 h at 32°C.It was then filtered and diluted fivefold.Absorbance at 530 nm and 657 nm was determined as previously described，and anthocyanin levels calculated (Kytridis and Manetas 2006).

For phenylalanine ammonia-lyase (PAL) and chalcone isomerase (CHI) activity determination，fresh 0.5 g leaves were added to 5mL0.05molL-1Na2H PO4/KH2P O4(pH 7.0)，0.05 mol L-1ascorbi cacid，0.018mol L-1mercaptoethanol and mixed in an ice bath.The mixture was centrifuged at 15，000×gfor 20 min at 4°C，after which the supernatant containing crude enzyme was assayed for PAL and CHI activity (Shaked-Sachray et al.2002).

For peroxidase (POD) activity，0.1 g of fresh leaf powder was weighed and placed in a 4°C precooled mortar to which PVP，quartz sand and 5.0 mL 0.05 mol L-1phosphate buffer (pH 7.0) were added.The solution was ground for 30 s at a uniform speed and then poured into a test tube and centrifuged at 11，100×gfor 20 min at 4°C.The resulting supernatant contained active enzyme which was assayed for POD activity (Jiang et al.2016).

To determine UDP-glycose flavonoid glycosyltransferase (UFGT) activity，1.0 g of fresh leaves were ground，mixed with 5.0 mL propanol and centrifuged at 11，100×gfor 20 min at-20°C.The supernatant was discarded and the extraction repeated in 0.1 mol L-1boric acid buffer and 5.0 mmol L-1ascorbic acid solution.The final supernatant was UFGT crude extract.Its activity was assayed according to Lister and Lancaster (1996).

Total soluble sugar (TSS) was determined by weighing 0.5 g fresh leaf powder to which 5.0 mL deionized water was added.The solution was extracted in boiling water for 30 min，filtered，after which 5.0 mL anthrone reagent was added to a 1.0 mL extract.This was placed in boiling water for 3 min and stored for 15 min at room temperature.The absorbance at 630 nm was measured.A curve of different glucose standards was made (1.0 mg mL-1) from which TSS levels were calculated (Eris et al.2007).

The total soluble protein (TSP) was determined by weighing 0.2 g fresh leaves ground into a homogenate in 5 mL PBS at pH 7.8.The mixture was centrifuged at 3000×gfor 10 min at 4°C，after which 1.0 mL of the supernatant was placed in a test tube to which 5.0 mL Coomassie Brilliant Blue GMel250 solution was added.The solution was mixed，incubated for 2 min，and the absorbance measured at 595 nm and compared to a protein standard curve (Eris et al.2007).

Determination of photosynthetic characteristics and chlorophyll fluorescence

With reference to Zhu et al.(2017)，measurements occurred between 9:00 and 12:00 a.m.;net photosynthetic rate (P n)，stomatal conductance (G s)，transpiration rate (Tr)and intercellular CO2concentration (C i)were measured on a Li-6400 portable photosynthesis meter (Beijing Ecotek Technology Company Ltd.，China).The light intensity was 1000 μmol m-2s-1，the CO2concentration 380 μmol mol-1，the chamber temperature of saturated light intensity 37°C，and the relative humidity 75%.With reference to Liu et al.(2018)，chlorophyll fluorescence was measured using a pulse modulated fluorimeter FMS 2.02 (Hansatech，United Kingdom).Leaves were dark adapted for 20 min before measurements.Initial fluorescence (F0) was measured under dark adaptation and maximum fluorescence (Fm)determined by providingas a turated pulse light(6000 μmol m-2s-1，pulse time 0.7 s)，and then leaves were exposed to continuous photochemical active light (1000 μmol m-2s-1) for 5 min to measure steady-state fluorescence under light adaptation(Fs).Following this，a second saturation pulse was applied for 0.7 s to obtain light-adapted maximum fluorescence(Fm′).Finally，the leaves were shaded，the far red light was turned on after dark adaptation for 3 s and the minimum fluorescence (F0′) measured.The maximum photochemical efficiency (Fv/Fm=(Fm-F0)/Fm)，actual photochemical effi-ciency (ΦPSII=(Fm′-Fs)/Fm′)，non-photochemical quenching coefficient (NPQ=(Fm-Fm′)/Fm′)，photochemical quenching coefficient (qP=(Fm′-Fs)/(Fm′-F0′) and electron transfer rate (ETR=ΦPSII·PFD·0.84·0.5 (PFD=luminous flux density) were calculated.

Chloroplast ultrastructural observations

Samples were fixed in 2.5% glutaraldehyde solution for ＞ 24 h，washed in phosphate buffer，fixed in 1% osmic acid for 2 h and rinsed in phosphate buffer.Samples then underwent gradient dehydration in different ethanol concentrations and then infiltrated and embedded in epoxy resin.They were cut into 80 nm slices with an ultra-thin slicer，observed and photographed with a JEOL-1230 transmission electron microscope after staining (Liu et al.2018).

Data analysis

Changes in leaf color parameters and physiological factors at different periods of coloration were statistically analyzed and plotted in Excel (2016).SPSS19.0 was used to analyze data by single factor analysis of variance，multiple comparisons(using Duncan’s new complex difference method) and correlation analysis.Stepwise multiple regression analyses were performed with leaf color parameters as dependent variables and pigment，enzyme activity，photosynthetic fluorescence characteristics and chloroplast size as independent variables.The path analysis of the multiple regression equation which passed the significance test was performed.Correlation coefficients between respective variable factors and dependent variables were decomposed into direct and indirect action coefficients，from which decision-making coefficients were calculated (Kumar et al.2015).Significance levels were set atP＜0.05.

Results

Changes in meteorological factors and leaf color parameters

Daily mean relative humidity and light intensity also fluctuated.Relative humidity was approximately 70% in S1 and S5，and 97.8%，92.8%，and 87.5% in S2，S3，and S4，respectively (Fig.1 a).Average daily light intensity in S1 was 47，600 lx，and in S2 and S3，it was 4600 lx and 8500 lx，respectively.It was higher in S4 and S5，i.e.，21，600 lx and 31，800 lx，respectively (Fig.1 b).

Liquidambar formosanaleaf color parameters varied significantly at different coloration periods (P＜0.01，F(xiàn)ig.3).L*(brightness) andb*(yellow and blue attributes) showed an initially increased and then decreased with a small range of change.a*(red and green attributes) andC*(saturation) gradually increased whileh°(hue angle) gradually decreased.In S1 when the leaves were green，L*，b*andC*values were small，a*was negative andh°was the highest.During S2，there was a gradual increase in leaf redness，withL*，a*，b*andC*all increasing.L*andb*were the highest，whileh°decreased.During S3 and S4，L*，b*andh°gradually decreased whilea*andC*continued to increase.During S5，L*，b*andh°values were lowest，whilea*andC*increased to their maximum levels.The leaves were dark red at this time.

Fig.3 Color parameters of L.formosana leaves at different coloration stages.L*:brightness; a*:red and green attributes; b*:yellow and blue attributes; C*:saturation; h°:hue angle.Data show the mean of five leaves sampled from five trees (± standard error，SE).Significant differences (P＜0.01 using Duncan’s new complex difference method) indicated by different letters

Changes in pigment，enzyme activity，soluble sugar and protein levels

Chla，Chlb，Car，and anthocyanin in leaves varied significantly (P＜0.01) at different coloration stages (Fig.4).With increased redness in leaf color，Chlaand Chlblevels decreased (Fig.4 a)，while anthocyanin levels increased，and Car levels fluctuated (Fig.4 b，c).Chla，Chlb，anthocyanin and Car levels at S1 were 0.469 mg g-1，0.319 mg g-1，0.973 mg g-1and 0.155 mg g-1，respectively，while levels at S5 were 0.189 mg g-1，0.086 mg g-1，2.348 mg g-1and 0.113 mg g-1，respectively.When compared with S1，Chla，Chlband Car decreased by 59.6%，73.0% and 14.5%，respectively，while anthocyanin levels increased by 41.3%.As leaves increased in redness，pigment levels changed，i.e.Chlaand Chlblevels decreased gradually，accounting for 9.8% and 4.4% of total pigment at S5，respectively.Car levels first increased and then decreased，accounting for 9.8%of total pigment content at S2 and 5.8% at S5.However，anthocyanin levels gradually increased and accounted for 50.8% of the total pigment content at S5.Chl/anthocyanin and Car/anthocyanin levels decreased gradually from 0.813 and 0.160 at S1 to 0.117 and 0.048 at S5 by 85.6% and 70.0%，respectively (Fig.4 d).

Fig.4 Pigment levels in L.formosana leaves at different coloration periods.a Chlorophyll a (Chl a) and chlorophyll b (Chl b) levels.b Carotenoid(Car) levels.c Anthocyanin(Ant) levels.d Pigment percentages.Data show the mean of five leaves sampled from five trees (± standard error SE).Significant differences (P＜0.01 using Duncan’s new complex difference method) are indicated by different letters

PAL，CHI，POD，and UFGT activities were significantly different (P＜0.01) at different coloration stages (Fig.5).With increasing leaf redness，PAL activities showed a continuous downward trend;the highest activity was 7.964 U g-1at S1 but decreased to 6.509 U g-1at S5 (Fig.5 a).CHI activity decreased from S1 to S2，continuously increased from S2 to S4，but decreased from S4 to S5.Overall it increased by 14.9% when compared with S1 (Fig.5 b).POD levels fluctuated but reached a maximum at S2 (11.256 U g-1).POD activity at S5 was 7.847 U g-1，which increased by 67.4%when compared with S1 (Fig.5 c).UFGT activity increased at first and then decreased;increasing continuously from S1 to S4 and plateauing at S4 (45.278 U g-1).At S5，UFGT activity decreased to 43.114 U g-1but overall increased by 164.3% when compared with S1 (Fig.5 d).

Fig.5 Enzymatic activity in L.formosana leaves at different coloration stages; a phenylalanine ammonia-lyase (PAL) activity，b chalcone isomerase (CHI) activity，c peroxidase (POD) activity，d UDPglycose flavonoid glycosyltransferase (UFGT) activity.Data show the means of five leaves sampled from five trees (± standard error，SE).Significant differences (P＜0.01 using Duncan’s new complex difference method) are indicated by different letters

TSS and TSP levels in leaves varied significantly(P＜0.01) at different coloration stages (Fig.6).With increasing leaf redness，TSS levels increased from 15.17 at S1 to 29.93 mg g-1at S5 (Fig.6 a).TSP levels increased from S1 to S4，plateauing at S4 (2.88 mg g-1)，and decreasing to 2.06 mg g-1at S5.However，TSP levels at S5 increased by 74.6% when compared with S1 (Fig.6 b).

Fig.6 a Total soluble sugar and b total soluble protein levels in L.formosana leaves at different coloration stages.Data show the means of five leaves sampled from five trees (± standard error，SE).Significant differences (P＜0.01 using Duncan’s new complex difference method) are indicated by different letters

Changes in photosynthetic characteristics and chlorophyll fluorescence

Significant differences (P＜0.01) inP n(Fig.7 a)，G s(Fig.7 b)，C i(Fig.7 c)，andT r(Fig.7 d) were observed in leaves at different coloration stages.With increasing leaf redness，P n，G sandT rdecreased，whileC igradually increased.P n，G s，Ciand Tr at S1 were 5.54μmol m-2s-1，0.19 mol m-2s-1，244.96μmol mol-1and2.35mmol m-2s-1，respectively.At S5，they were 1.17 μmol m-2s-1，0.068 mol m-2s-1，335.65 μmol mol-1and 1.06 mmol m-2s-1，respectively.When compared with S1，P n，G s，andT rat S5 decreased by 78.9%，64.2%，and 54.9% respectively，whileC iincreased by 36.2%.

Fig.7 Photosynthetic parameters of L.formosana leaves at different coloration stages; a net photosynthetic rate (P n) ; b stomatal conductance (G s) ; c intercellular CO2 concentration (C i) ; d transpiration rate (T r) .Data show the means of five leaves sampled from five trees(± standard error SE).Significant differences (P＜0.01 using Duncan’s new complex difference method) are indicated by different letters

Significant differences inF0(Fig.8 a)，F(xiàn)v/Fm(Fig.8 b)，ΦPSII(Fig.8 c)，qP(Fig.8 d)，NPQ (Fig.8 e)，and ETR(Fig.8 f) were observed in leaves at different coloration stages.With increasing redness，F(xiàn)0，F(xiàn)v/Fm，ΦPSII，and ETR gradually decreased，qPfirst decreased and then increased，whereas NPQ levels fluctuated.F0，F(xiàn)v/Fm，ΦPSIIand ETR values at S1 were 220.00，0.84，0.30 and 121.97 respectively，but decreased to 42.60，0.81，0.20，80.21 at S5，reflecting a decrease of 80.6%，3.6%，33.3% and 34.2%，respectively.From S1 to S4，qPdecreased rapidly，increased slightly at S5 but overall decreased by 51.8% when compared with S1.NPQ at S1 was 2.36，decreased slightly at S2，but then increased to 3.16 at S4，and decreased to 3.04 at S5.Overall，NPQ at S5 increased by 29.8% when compared with S1.

Fig.8 Chlorophyll fluorescence parameters for L.formosana leaves at different coloration stages; a initial fluorescence (F0) ; b maximum photochemical efficiency (Fv/Fm) ; c actual photochemical efficiency(ΦPSII) ; d photochemical quenching coefficient (qP); e non-photochemical quenching coefficient (NPQ); felectron transfer rate (ETR).Data show the means of five leaves sampled from five trees (± standard error，SE).Significant differences (P＜0.05 and P＜0.01 using Duncan’s new complex difference method) are indicated by different letters

Ultrastructural changes in chloroplasts

There were significant differences in chloroplast numbers(CN) per cell，and chloroplast lengths (CL) and widths (CW)at different coloration periods (Table 1).Chloroplast numbers and length were largest at S1，i.e.10.2 and 5.2 μm，respectively.With increased redness，chloroplast numbers and chloroplast lengths gradually decreased but chloroplast width increased，reaching a maximum at S4，i.e.，3.2 μm.

Leaf chloroplasts in S1 were irregular and oval，internal solutes deeply colored and arranged along the cell wall.The thylakoid lamellae were clear and tightly stacked，starch granules were large，and there were several smaller osmiophilic granules distributed in the thylakoid lamellae(Fig.9 S1).During S2，chloroplasts became oval，the color of inner solutes darker，the size of starch granules smaller than S1，and osmiophilic granules larger (Fig.9 S2).At S3，chloroplasts became oblate，the color of the inner solution lighter，and the lamellar arrangement of the thylakoid was scattered but still clear，and osmiophilic granules continued to increase，whereas the size of starch granules continued to decrease (Fig.9 S3).At S4，chloroplasts were irregular ovals，the internal solute lightly stained，the thylakoid lamellar blurred，only part of the thylakoid lamellar was observed and osmiophilic granules were condensed into lipoids，whereas starch granules had disappeared (Fig.9 S4).At S5，chloroplasts were oblate and elongated，and a large number of thylakoid lamellae had disintegrated.A small number of disordered grana lamellae were observed，with lipoids further enlarged (Fig.9 S5).

Fig.9 Chloroplast structures of L.formosana leaves at different coloration stages.S1:green;S2:minor reddish areas＜1/3;S3:some reddish areas ＞ 1/3，＜2/3;S4:reddish areas ＞ 2/3;and S5:all red.CH:Chloroplast;TL:thylakoid lamella;SG:starch grain;OG:osmiophilic granule;L:lipoid

Relationship between leaf color and physiological factors

L*andb*were significantly negatively correlated with anthocyanin levels.a*andC*were significantly negatively correlated with Chla，Chlb，Car，Chl/anthocyanin，carotenoid/anthocyanin ratios，and positively correlated with anthocyanin levels.h°was positively correlated with Chla，Chlb，carotenoids，Chl/anthocyanin，carotenoid/anthocyanin ratios，and negatively correlated with anthocyanin (Table 2).

There were significant correlations between leaf color parameters，pigment levels and physiological factors(Table 3)，e.g.，a*was positively correlated with TSS，TSP，CHI，UFGT andC i，and negatively correlated with PAL，P n，G s，T r，F(xiàn) 0，ΦPSII，qP，ETR and CN.Anthocyanin was positively correlated with TSS，CHI，UFGT andC i，and negatively correlated withP n，G s，T r，F(xiàn) 0，ΦPSII，qP，ETR，CN and CL (Table 3).

Table 1 Chloroplast numbers (CN)，length (CL) and width (CW) in different coloration periods

Table 2 Correlation between leaf pigment levels and color parameters

The direct and indirect effects of related independent variables on dependent variables were reflected by the path coefficient，which showed the relative importance of specific variables.As observed (Fig.10 b)，the direct effect coefficient of PAL ona*was negative (-0.453) and the indirect effect coefficient of PAL ona*via anthocyanin，UFGT，CHI，Chla，andF0were-0.639，-0.045，-0.562，0.562 and 0.184，respectively.The superposition of direct and indirect effects resulted in a large negative effect of PAL ona*(-0.954).The direct effects of anthocyanin and UFGT ona*were positive and the indirect effects via other factors were also positive.Thus，anthocyanin and UFGT had large positive effects ona*.The direct effects of CHI，Chla，andF0ona*were all positive，while the indirect effects via other factors were negative.After counteracting each other，the comprehensive effects of CHI ona*were positive and the comprehensive effects of Chla，andF0ona*were all negative.

The decision-making coefficient was calculated using direct and indirect path coefficients.The largest and positive decision-making coefficient was the main decision-making factor，and the smallest and negative was the main limiting factor (Kumar et al.2015).As observed (Fig.10)，the decision coefficients of each factor affecting changes inL*were:TSP (-0.363) ＞ CHI(-1.604) ＞ Ant (-2.548) ＞ Chla(-3.979) (Fig.10 a).The decision coefficient of each factor affecting changes ina*was:Ant (0.803) ＞ PAL (0.659) ＞ CHI (0.254) ＞ UFGT(0.085)＞F0(-0.413) ＞Chla(-2.341)(Fig.10b).The decision coefficient of each factor affecting changes inb*was:Ant (0.187) ＞ Chlb(-0.391) (Fig.10 c).The decision coefficient of each factor affecting changes inC*was:(0.784) ＞ UFGT (0.517) (Fig.10 d).The decision coefficients of the factors affecting h changes were:Chl/anthocyanin (0.679) ＞C i(0.523) ＞ TSS (0.348) ＞ UFGT(0.247) ＞ NPQ (-0.029) (Fig.10 e).

Stepwise regression analyses showed that multiple regression equations for the five leaf color parameters all passed the significance test (R≥ 0.741，P≤0.05) (Table 4).Path analysis (Fig.10) showed that anthocyanin was the main decision factor of leaf color parametersa*andC*，Chl/Ant was the main decision factor ofh°，Chlawas the main limiting factor ofL*anda*，Chlbwas the main limiting factor ofb*，and NPQ was the main limiting factor ofh°.

Table 4 Multiple stepwise regression equations between leaf color parameters and physiological factors

Fig.10 Path analysis results for leaf color parameters and physiological factors.Black dotted arrows indicate the direct effects，i.e.，direct path coefficients.Different color solid arrows indicate the indirect effect of one physiological factor on leaf color parameters via another physiological factor，i.e.，indirect path coefficients.Numbers in brackets indicate the decision coefficient and R the correlation coefficient.L *:brightness; a *:red and green attributes; b *:yellow and blue attributes; C *:saturation; h ° :hue angle;CHI:chalcone isomerase;TSP:total soluble protein;Ant:anthocyanin;Chl a:chlorophyll a;PAL:phenylalanine ammonia-lyase;UFGT:UDP-glycose flavonoid glycosyltransferase; F0 :initial fluorescence;Chl b:chlorophyll b

Table 3 Correlation analysis of leaf color parameters，pigment levels and physiological factors

Discussion

Relationship between pigment，enzyme activity，soluble inclusion and leaf color

Earlier studies have shown that pigment type，proportion and distribution underpin leaf color changes (Archetti 2000;Lee et al.2003).Leaf redness is due to the production of red pigments，e.g.，anthocyanin，the levels and proportions of which are higher than other pigments (Hughes 2011;Novak and Short 2011;Becker et al.2014).With the gradual decrease of external temperatures (Fig.1)，the leaves ofL.formosanabecame red，chlorophyll levels decreased，anthocyanin increased gradually，and carotenoids remained stable.Our stepwise regression and path analysis showed that anthocyanin was positively correlated witha*andC*，and it was a decision factor ofa*andC*，while Chlawas negatively correlated witha*，and was the limiting factor that affected leafL*anda*.Therefore，decreases in chlorophyll levels and anthocyanin accumulation contributed to redness levels inL.formosanaleaves.

Other studies have shown that anthocyanin biosynthesis was related to enzyme activity，soluble sugars and proteins，where PAL was the first enzyme to regulate and catalyze anthocyanin biosynthesis (Osamu 1988;Gu et al.2015).CHI is an important enzyme in anthocyanin biosynthesis and plays a key role in flavonoid levels (Lister and Lancaster 1996;Nishihara et al.2005).UFGT is the last enzyme in the anthocyanin synthesis pathway and catalyzes anthocyanin stability (Kobayashi et al.2001).POD affects polyphenol levels in leaves，promotes anthocyanin synthesis and participates in chlorophyll degradation (Zhu et al.2017).However，some studies have shown that PAL (Lister et al.1996)，CHI (Ju et al.1995)，UFGT (Ju et al.1999) and POD (Boo et al.2011) were not closely related to anthocyanin synthesis，and are not key enzymes.Anthocyanin synthesis begins with phenylalanine metabolism.PAL is the first enzyme in the phenylalanine metabolic pathway (Shi et al.2020)，but the final product includes a variety of components such as flavonoids，anthocyanin，lignin，tannin，and cutin.Therefore，PAL catalytic reactions not only provide precursors for anthocyanin synthesis but also provide precursors for the synthesis of other substances (Ju et al.1999;Wang et al.2004).In this study，there was a significant negative correlation between PAL activity and anthocyanin content.In the early stage of coloration (S1)，PAL activity was higher but anthocyanin content was lower.With the gradual reddening of leaves and increased anthocyanin content，PAL activity gradually decreased.This may be because the direct precursors of anthocyanin formation are dihydroquercetin and dihydromyricetin flavonoids.Part of the PAL is used to catalyze sufficient dihydroquercetin and dihydromyricetin flavonoids to provide precursors for anthocyanin synthesis，while the other part is used to catalyze the synthesis of other products (Wang et al.2004).WhenL.formosanaleaves turn red，the rate of synthesis of anthocyanins increases gradually，while the rate of synthesis of other products may gradually decrease，resulting in an overall decrease in PAL enzyme activity (Wang et al.2004).Therefore，the relationship between PAL activity and anthocyanin levels is affected by the synthesis of other products.This particular mechanism requires further study.

CHI activity fluctuated and was positively correlated witha*and anthocyanins with a positive effect ona*andL*and promoted redness inL.formosanaleaves.There was no significant correlation between POD，leaf color parameters，and anthocyanin.CHI activity was not introduced into the regression equation and had no effect on the change of leaf color.Change in colour is a reaction to low temperatures (Jiang et al.2016)，and therefore the specific reasons for this further study.UFGT was positively correlated with anthocyanins，a*andC*，and negatively correlated with Chl/anthocyanin，carotenoid/anthocyanin andh°，indicating that the higher the UFTG activity，the higher the anthocyanin levels and the redder the leaves.Therefore，UFGT was a key enzyme in anthocyanin synthesis，promoting redness inL.formosanaleaves.

Sugar is not only an energy source for anthocyanin synthesis but can also be used as metabolic precursors or signal molecule to promote anthocyanin formation (Murakami et al.2008).Soluble proteins are components of most enzymes.These enzymes are involved in the synthesis and transformation of pigments，enzymes and sugars，thereby promoting anthocyanin accumulation in leaves (Feng et al.2011).Studies have shown that soluble sugars (Sperdouli and Moustakas 2012) and soluble proteins (Osamu 1988)exert no significant effects on anthocyanin synthesis.Our data showed that，as the leaf turned red，TSS gradually increased and was positively correlated with anthocyanin，a*andC*，and negatively correlated withh°，i.e.，the smaller theh°，the redder the leaf.These data indicate that TSSs were beneficial to anthocyanin biosynthesis，and promoted leaf reddening inL.formosana.Under normal circumstances，plant sugars are derived from photosynthesis (Paul et al.2001).In this study，with the gradual reddening ofL.formosanaleaves，the chlorophyll content and the photosynthetic capacity of the leaves gradually decreased，but TSS levels gradually increased.This may be due to the fact that during leaf senescence and coloration，most nutrients such as soluble sugars，nitrogen and phosphorus are recovered and transferred to the mitochondria.After the leaves become red and lose their photosynthetic capacity，these nutrients provide energy for mitochondrial respiration.As a result，the physiological metabolism of the plant in winter is maintained (Keskitalo et al.2005;Murakami et al.2008;Zhang et al.2013).Therefore，increased TSS content in leaves may come from nutrients recovered and transferred to the mitochondria.

Total soluble protein (TSP) increased initially，then decreased，and were negatively correlated withL*，indicating that TSP exerted a particular effect on leaf brightness.TSP had a positive correlation with anthocyanin and a significant positive correlation witha*，but was not introduced into the regression equations ofa*，C*andh°，indicating TSP had no direct effect on leaf redness but indirectly promoted leaf redness.This may be because anthocyanin synthesis requires enzymes such as CHI and UFGT，and TSP are the components of these enzymes.TSP indirectly promotes the synthesis of anthocyanin by promoting the synthesis of CHI，UFGT，and other enzymes，which indirectly turn leaves red(Osamu 1988).

Relationship between photosynthetic fluorescence characteristics and leaf color

The photosynthetic function of pigmented colored leaf plants is often due to decreased chlorophyll levels and an increase in other pigments (e.g.，anthocyanin).This leads to a decrease in available light energy absorption in leaves，ultimately affecting photosynthesis in colored leaves (Hanke and Mulo 2013).The difference in chlorophyll content in the coloration process ofL.formosanaleaves lead to differences in the structure of the light system in the chloroplasts.By comparing differences in chloroplast structure in mesophyll cells of leaves at different coloration stages，there were dysplastic chloroplast structures in red leaves observed，especially losses and distortions of thylakoid membranes which led to the abnormal biosynthesis of chloroplast membrane structures (Zhang et al.2006;Moustaka et al.2018).Several proteins related to photosynthesis are distributed in the thylakoid membrane structure，where the PS II photoreaction center and the corresponding antenna chlorophyll and electron transport proteins are the main structures in grana lamellae.When plant photosynthesis is inhibited，the chloroplast photosystem PS II center is initially affected (Bertamini and Nedunchezhian 2003;Snider et al.2015).Fv/Fmreflects the plants ability to utilize light energy.Decreases in this value are the most obvious features of photoinhibition，and an increase in NPQ reflects increased heat dissipation (Muller 2001).During the coloration ofL.formosanaleaves，P ndecreased rapidly，as didFv/Fm，ΦPSIIand ETR，indicating that as leaves reddened，excess light energy in the leaves caused photoinhibition (Hiiner et al.1998;Zhang et al.2011).When plants are subjected to photoinhibition，they usually protect themselves by reducing the photosynthetic active radiation incident on leaves and chloroplasts，or by dissipating absorbed energy before transferring to the chlorophyll reaction center (Hoch et al.2001;Hughes et al.2007).Under strong light conditions，leaf anthocyanins filter and attenuate light，the photosynthetic active radiation incident on leaves or chloroplasts is reduced by absorbing blue-green light in incident strong light，and dissipates the absorbed quantum energy in the form of heat (Zhang et al.2011;Junker and Ensminger 2016;Moustaka et al.2020).Therefore，in the middle and later stages of coloration ofL.formosanaleaves (S4 and S5)，F(xiàn)v/Fmdecreased continuously，photoinhibition intensified and anthocyanin and NPQ increased rapidly.On the one hand，leaves filtered light by increasing anthocyanin levels to reduce the photosynthetic active radiation incident on leaves，whereas on the other hand，they protected themselves by increasing the nonphotochemical quenching coefficient by consuming excess light energy.Thus，it is believed that the red leaves ofL.formosanaprimarily relied on anthocyanin for light damage defenses (Hughes et al.2007).

Conclusions

During leaf coloration，leaf color parameters，pigment levels，enzyme activity，soluble sugar and protein content，photosynthetic fluorescence characteristics and chloroplast structures were significantly altered.Of these，red and green attributes and saturation，anthocyanin levels，total soluble sugar，and non-photochemical quenching coefficient are gradually increased.Chlorophyll and carotenoid levels，phenylalanine ammonia-lyase activity，net photosynthetic rate，maximum photochemical efficiency，actual photochemical efficiency，electron transfer rate and the number and length of chloroplasts were gradually decreased.Chalcone isomerase and UDP-glycose falvonoid glycosyltransferase activities appeared to fluctuate.Correlation analyses，stepwise regression analyses and path analyses showed that anthocyanin was significantly positively correlated with red and green attributes and saturation，and was the main decision factor affecting red and green attributes and saturation.Chlorophyllawas negatively correlated with red and green attributes and was the main limiting factor affecting brightness and red and green attributes.Therefore，decreases in chlorophyll levels and anthocyanin accumulation were the direct reasons ofL.formosanaleaves reddening.Chalcone isomerase and UDP-glycose flavonoid glycosyltransferase were the key enzymes of anthocyanin synthesis，and total soluble sugar was an important substance of anthocyanin synthesis，all of which were beneficial to anthocyanin biosynthesis and promoted the reddening ofL.formosanaleaves.

The following mechanism is proposed for the color changes inL.formosanaleaves:with decreased external temperatures，chloroplast numbers and sizes gradually decrease and thylakoid membranes become distorted.Chlorophyll synthesis is blocked，and levels gradually decrease.As a result，chloroplast membrane structures are abnormally biosynthesized and net photosynthetic rate decreases rapidly.Maximum photochemical efficiency，actual photochemical efficiency and electron transfer rate also decrease rapidly，and excess light energy results in photoinhibition.With such intensification of photoinhibition，leaves promote anthocyanin synthesis by increasing chalcone isomerase and UDP-glycose flavonoid glycosyltransferase activities and total soluble sugar levels，increased anthocyanin levels to filtered light and eliminated reactive oxygen species，and reduced photoinhibition.Alternatively，self-protection is implemented by increasing the non-photochemical quenching coefficient by consuming excess light energy by heat energy，which gradually turns the leaves red.