Insights into physiological roles of flavonoids in plant cold acclimation

SUMMARY

Flavonoids represent a diverse group of plant specialised metabolites which are also discussed in the context of dietary health and inflammatory response. Numerous studies have revealed that flavonoids play a central role in plant acclimation to abiotic factors like low temperature or high light, but their structural and functional diversity frequently prevents a detailed mechanistic understanding. Further complexity in analysing flavonoid metabolism arises from the different subcellular compartments which are involved in biosynthesis and storage. In the present study, non‐aqueous fractionation of Arabidopsis leaf tissue was combined with metabolomics and proteomics analysis to reveal the effects of flavonoid deficiencies on subcellular metabolism during cold acclimation. During the first 3 days of a 2‐week cold acclimation period, flavonoid deficiency was observed to affect pyruvate, citrate and glutamate metabolism which indicated a role in stabilising C/N metabolism and photosynthesis. Also, tetrahydrofolate metabolism was found to be affected, which had significant effects on the proteome of the photorespiratory pathway. In the late stage of cold acclimation, flavonoid deficiency was found to affect protein stability, folding and proteasomal degradation, which resulted in a significant decrease in total protein amounts in both mutants. In summary, these findings suggest that flavonoid metabolism plays different roles in the early and late stages of plant cold acclimation and significantly contributes to establishing a new protein homeostasis in a changing environment.

Significance Statement

Flavonoids play a central role in plant cold acclimation. This study provides evidence for a significant effect of flavonoid deficiency on photorespiration and protein homeostasis under low temperature.

INTRODUCTION

Plant metabolism comprises numerous metabolic compounds, belonging to diverse chemical substance classes, which are synthesised, degraded and interconverted enzymatically within comprehensive and highly interconnected metabolic networks. Carbohydrates are direct products of photosynthetic CO₂ fixation and, thus, represent substrates for all further anabolic metabolic pathways. Hence, the metabolism of carbohydrates is tightly linked to carboxylic and amino acids, which altogether represent the building blocks for protein, lipid and nucleotide biosynthesis. Further, they also represent substrates for the biosynthesis of diverse specialised compounds, for example, alkaloids, glucosinolates or flavonoids, which have been shown to protect plants against environmental stressors both biotic and abiotic, by their specialised functions (Erb & Kliebenstein, 2020).

In many plant species, exposure to low but non‐freezing temperatures induces a process termed cold acclimation which increases freezing tolerance. This process has been studied for decades and has revealed numerous molecular and physiological responses that are involved in its regulation (Garcia‐Molina et al., 2020; Gilmour et al., 1998; Kaplan et al., 2007; Ristic & Ashworth, 1993; Steponkus, 1984). Adjustment of photosynthesis and carbohydrate metabolism is a central cold acclimation response to prevent or mitigate photoinhibition and unbalanced primary and secondary photosynthetic reactions (Stitt & Hurry, 2002). Sucrose metabolism has been found to play a crucial role in photosynthetic acclimation (Strand et al., 2003). It has been discussed that adjustment of sucrose phosphate synthase (SPS) activity to low temperature prevents a limitation of triose phosphate/phosphate translocation between cytosol and chloroplasts, which would result in a limitation of photosynthetic ATP biosynthesis (Nägele et al., 2012). Recently, we found evidence for the role of SPS activity in the regulation of carbon fluxes between carbohydrate and flavonoid metabolism during cold acclimation (Kitashova et al., 2023).

While the metabolism of flavonoids is well known to be involved in many plant–environment interactions (Winkel‐Shirley, 2002), their physiological role in cold acclimation remains elusive. Flavonoids are phenylpropanoids that are synthesised predominantly from phenylalanine through the shikimic acid pathway. For Brassica napus, phenylpropanoid deficiency has been found to result in lowered freezing tolerance and a decreased cold acclimation capacity of photosynthesis (Solecka & Kacperska, 2003). Also, for Arabidopsis, a significant impact of flavonoids and their metabolism on cold acclimation capacity and freezing tolerance has been described (Schulz et al., 2015, 2016). Under high light, recent findings suggested flavonoid metabolism to be tightly interconnected with chloroplast carbohydrate metabolism and the SnRK1‐related signalling network (Zirngibl et al., 2023). Other studies have provided further evidence for the diverse functional roles of flavonoids, comprising, for example, DNA protection, protection against UV radiation or serving as signalling compounds to affect gene expression (Nakabayashi et al., 2014; Naoumkina & Dixon, 2008; Sarma & Sharma, 1999). These findings, together with observations made in many other studies (for an overview, see Tohge et al., 2018), provide strong evidence for a central role of flavonoids in plant stress response and cold acclimation.

Flavonoid accumulation is typically observed in the vacuole (for an overview, see Pucker & Selmar, 2022). Also, flavonoids have been detected in other compartments, for example, the nucleus and chloroplasts (Agati et al., 2007; Peer et al., 2001). Together with the structural diversity of flavonoids, also the diversity of subcellular localization challenges the analysis of their plant biochemical and physiological functions.

The core pathway of flavonoid biosynthesis is located in the cytosol and comprises activities of chalcone synthase (CHS), chalcone isomerase (CHI), flavanone 3‐hydroxylase (F3H), flavonol synthase (FLS), dihydroflavonol 4‐reductase (DFR) and anthocyanidin synthase (ANS). Phenylalanine is a precursor of flavonoid biosynthesis and is synthesised in plastids from erythrose 4‐phosphate and phosphoenolpyruvate (Rippert et al., 2009). Erythrose 4‐phosphate is a product of the Calvin–Benson–Bassham Cycle (CBBC) and, thus, directly links the shikimate pathway and subsequent flavonoid biosynthesis to photosynthetic CO₂ fixation and carbohydrate biosynthesis.

Under changing environmental conditions, for example, low temperature, photosynthetic acclimation prevents unbalanced electron transport, ATP‐biosynthesis and enzymatic CO₂ fixation. Photosynthetic acclimation is a comprehensive process that comprises and affects the regulation of redox and ion homeostasis of the chloroplast, protein amounts of photosystems, metabolism and multicompartmental signalling networks (Gjindali & Johnson, 2023). Under low temperature, enzymes of the CBBC have been found to be upregulated to compensate for the decrease in activity due to thermodynamic constraints (Stitt & Hurry, 2002; Strand et al., 1999). Together with the regulation of the central carbohydrate metabolism, for example, the sucrose biosynthesis pathway, these are immediate cold‐induced processes, which can be observed during the initial hours and days of a cold acclimation period (Nägele et al., 2011; Savitch et al., 2001). Accumulation of flavonoids and other specialised metabolites typically follows these initial acclimation reactions of primary metabolism, resulting in peak values after several days and up to weeks of cold exposure (Doerfler et al., 2013; Kitashova et al., 2023).

The balance between energy absorbed through photochemistry, photosynthetic electron transport and interconverting metabolic processes, also termed photostasis, plays a critical role in plant–environment interactions in general (Huner et al., 2003). Here, regeneration of NADP⁺ as an electron acceptor and ADP as a substrate for the ATP synthase reaction are essential to prevent photoinhibition and generation of reactive oxygen species. The multicompartmental pathway of photorespiration has been discussed to play a central role in balancing ATP and NADPH production and usage (for an overview, see Timm & Hagemann, 2020). Further, photorespiration affects many other pathways due to its role in C1 metabolism, S‐metabolism or C/N balance.

To better understand the role of flavonoids during cold acclimation in a pathway‐specific context, the present study analysed metabolism in flavonoid‐deficient mutants chs (also tt4‐11) and f3h (also tt6‐3) on a compartmental level. While a mutation of the CHS enzyme leads to a general absence of flavonoids, the function of the F3H enzyme can be partially complemented by FLS and ANS (Owens et al., 2008). Here, both mutants were analysed together to reveal traits and effects of plant metabolic cold acclimation, which are due to (i) complete flavonoid deficiency (only in chs) or (ii) partial flavonoid deficiency (in f3h and chs).

RESULTS

Flavonoid deficiency affects citrate and glutamate metabolism during cold acclimation

To investigate cold‐induced dynamics of the subcellular metabolomes in Col‐0, chs and f3h, leaf tissue was analysed by non‐aqueous fractionation coupled with GC–MS measurements. This methodology revealed plastidial, cytosolic and vacuolar metabolite amounts. An overlap of mitochondrial and plastidial fractions could not be excluded due to technical limitations of compartment separation (see Materials and methods section). Hence, all results entitled in the following as ‘plastidial’ might represent an overlap of both compartments. An overview of changes in metabolite abundances on cellular and subcellular levels is provided in the supplements (Figure S6; Table S7). Time series analysis, which compared dynamics of substrate‐product relationships in a metabolic network, revealed that particularly citrate and glutamate metabolism (r23‐r25, see Figure S1) were consistently perturbed in chs and f3h plants during the full cold acclimation period of 14 days (Figure 1; Figures S2 and S3; Table S1). In Col‐0, cytosolic pyruvate, which is the substrate for citrate biosynthesis via the pyruvate dehydrogenase reaction (r24 in Figure S1), was found to drop significantly during the first 3 days at 4°C and to be stabilised again until 14 days of cold acclimation (Figure S3). In both mutants, the cytosolic pyruvate amounts consistently decreased until 14 days at 4°C. In Col‐0 and chs, citrate amounts decreased in all compartments until 3 days of cold exposure before they consistently increased again. In f3h, cold‐induced dynamics of citrate amounts were almost absent. Remarkably, both mutants showed a higher variance of citrate amounts after 14 days at 4°C than Col‐0. Furthermore, in f3h, a high variance of subcellular citrate amounts was observed across all time points of cold acclimation (Figure 1). Glutamate amounts showed an opposite trend in both mutants compared to Col‐0. While cold acclimation resulted in increased glutamate amounts in Col‐0 across all compartments, amounts decreased in both mutants at 14 days of 4°C in comparison to non‐acclimated plants, and the strongest decrease was observed in the plastidial fraction. Plastidial glutamine amounts in Col‐0 decreased after 3 days of cold exposure which was followed by an increase after 14 days. In contrast, chs and f3h mutants demonstrated opposite trends: both mutants showed an increase in plastidial glutamine after 3 days, with no significant change in chs or with a decrease in f3h after 14 days at 4°C (Figure 1).

Figure 1.

Cold‐induced subcellular dynamics of citric acid and glutamic acid.

Dynamics of citric acid and glutamic acid in Col‐0, chs and f3h after 0, 3 and 14 days of cold acclimation. A.U., arbitrary units, that is, peak areas normalised to the internal standard and sample dry weight. Green colour represents 0 day; blue: 3 days; magenta: 14 days of cold acclimation; diamonds represent outliers; n = 5. Asterisks indicate significant difference between time points within genotypes (ANOVA with Tukey HSD, P‐value <0.05). The full list of significance levels is provided in Table S2B.

In addition to a perturbed pyruvate and citrate metabolism, also, LC–MS‐based proteome analysis revealed that the abundances of pyruvate dehydrogenases (PDHs) and isocitrate dehydrogenases (ICDHs) were affected in chs and f3h compared to Col‐0 (Figure S4). Remarkably, cytosolic ICDH (CICDH, At1G65930) significantly decreased in Col‐0 until 3 days at 4°C and stabilised until 14 days. In both mutants, the abundance of CICDH was significantly higher at 3 days and decreased to levels of Col‐0 until 14 days. In contrast, mitochondrially located ICDHs showed similar dynamics across all genotypes. Interestingly, in the full data set, that is, comprising all time points and genotypes, CICDH positively correlated with plastidial pyruvate dehydrogenase (AT1G01090) while it correlated negatively with the mitochondrial IDH2 (AT2G17130) and the mitochondrial PDH component subunit alpha‐1 (AT1G59900; Figure 2).

Figure 2.

Pearson correlation of pyruvate dehydrogenases and isocitrate dehydrogenases.

Numbers show Pearson correlation coefficients; blank fields were not significant (P < 0.05). Blue: positive correlation; red: negative correlation. AT2G17130: mitochondrial isocitrate dehydrogenase 2 (IDH2); AT4G35260: mitochondrial IDH1; AT3G09810: mitochondrial IDH‐VI; AT5G14590: plastidial isocitrate dehydrogenase (pICDH); AT5G03290: mitochondrial IDH‐V; AT1G65930: cytosolic isocitrate dehydrogenase(CICDH); AT1G01090: plastidial pyruvate dehydrogenase component subunit alpha‐3 (PDH‐E1 alpha); AT1G30120: plastidial PDH component subunit beta‐3 (PDH‐E1 beta); AT1G59900: mitochondrial PDH component subunit alpha‐1 (AT‐E1 alpha); AT5G50850: mitochondrial PDH component subunit beta‐1 (MAB1); AT1G24180: mitochondrial PDH component subunit alpha‐2 (IAR4).

Flavonoid deficiency modulates photorespiration, dynamics of amino acid metabolism and tetrahydrofolate‐related C1 metabolism

Primary metabolites and lipids were quantified via chromatography coupled with mass spectrometry to reveal the effects of flavonoid deficiencies on metabolism during cold acclimation. Dimensionality reduction by principal component analysis (PCA) showed a general effect of cold acclimation on metabolomes and lipidomes (Figure 3), separating non‐acclimated from cold‐acclimated samples along PC1, while 3 and 14 days of cold acclimation were separated along PC2. Loadings with the highest contribution to separation along PC1 were carbohydrates (sucrose, maltose, glucose, fructose and raffinose), shikimic acid and glyceric acid. Along PC2, O‐acetylserine, phosphoglyceric acid, malic acid, quinic acid, xylitol and galactonic acid contributed strongest to the separation (see Table S3). Although the low‐temperature exposure had a dominant effect on sample separation (PC1), there was a subtle separation of Col‐0 samples from chs and f3h after 3 days of cold. Thus, the absence of cytosolic flavonoid biosynthesis affected metabolic cold acclimation of primary metabolism, and this effect was most pronounced during the early cold exposure period. An overview of growth phenotypes and PCA loadings, representing the contribution of individual metabolites and lipids to the observed separation, along with the components, is provided in the supplements (Figure S5; Table S3).

Figure 3.

Principal component analysis of the cellular metabolome comprising all considered genotypes and time points of cold acclimation.

Circles: chs; triangles: f3h; crosses: Col‐0. Green colour represents 0 days; blue: 3 days; magenta: 14 days of cold acclimation. Detailed information about loadings and components are provided in the supplements (Table S3).

As subcellular compartments contribute differently to metabolic cold acclimation (Nägele & Heyer, 2013), subcellular metabolic effects of flavonoid deficiency on primary metabolism were analysed by combining NAF of leaf tissue with GC–MS analysis. Comparing subcellular amounts of metabolites before and after cold exposure (0–3 and 3–14 days) revealed that the strongest effects of flavonoid deficiency on cold‐induced metabolic reprogramming, which were conserved across chs and f3h mutants, were located in the plastid and occurred during the early acclimation period, that is, between 0 and 3 days at 4°C (Figure 4a; Figure S6a,b). The metabolites that contributed most to the separation of the chs and f3h cluster from Col‐0 at this time point included proline, glycine, serine, fumaric acid, aspartic acid and malic acid. Due to their central role in photorespiration, the observed accumulation of glycine and a decrease in serine amounts suggested an effect in the regulation of photorespiration in both mutants (Figure 4b). Moreover, glutamine and fumaric acid were found to increase in both mutants while amounts decreased in Col‐0 (Figure 4a). Additionally, chs plants were found to accumulate plastidial carbohydrates more than Col‐0. These were glucose, fructose, raffinose, glucose 6‐phosphate, trehalose and sucrose after 3 days of cold exposure (Figure 4a).

Figure 4.

Plastidial metabolic cold acclimation.

(a) Hierarchical cluster analysis of the subcellular metabolite abundance change rate in plastid between Col‐0 and flavonoid mutants. Rates were calculated between 0 and 3 (0–3) and 3 and 14 (3–14) days of cold acclimation (e.g., rate = [median_{3 day} − median_{0 day}]/3) in Col‐0, chs and f3h. Results are displayed in a dendrogram and a heatmap indicating relationships between metabolites and samples based on Euclidean distance and complete‐linkage clustering. The values were column‐wise standardised using the z‐score method.

(b) Plastidial glycine and serine dynamics in Col‐0, chs and f3h after 0, 3 and 14 days of cold acclimation. A.U., intensities normalised to the internal standard and sample dry weight. Grey: Col‐0; red: chs, blue: f3h; n = 5. The list of significance levels (ANOVA with Tukey HSD) is provided in Table S2B.

(c) Hierarchical cluster analysis of the top 6 amino acid biosynthesis pathway‐associated protein abundances contributing to the separation of chs and f3h mutants and Col‐0 after 3 days of cold acclimation. Complete information about Euclidean distances in the cluster analysis of the amino acid biosynthesis‐related protein dynamics can be found in Table S4. Protein abbreviations: APR3, 5′‐adenylylsulfate reductase 3; DAPB1, 4‐hydroxy‐tetrahydrodipicolinate reductase 1; PAT, bifunctional aspartate aminotransferase and glutamate/aspartate‐prephenate aminotransferase; PSP, phosphoserine phosphatase; SAT3, serine O‐acetyltransferase 3; TSA1, TSK‐associating protein 1.

After 3 days of cold exposure, both mutants showed significantly higher amounts of amino acids than Col‐0 in all three analysed compartments (Figure S7). In the cytosol and vacuole, there were notable differences between the chs and f3h mutants. The f3h mutant formed a distinct cluster and showed a stronger accumulation of amino acids than chs. While serine levels did not change significantly in these compartments, glycine peaked after 3 days, especially in the vacuole of both chs and f3h mutants. Interestingly, both chs and f3h showed stronger accumulation of cytosolic raffinose (Figure S6a).

Given the pronounced effect on amino acids, changes in the proteome were quantified with functional relation to amino acid biosynthesis (Figure 4c; Table S4). Also, in the proteome, a similar hierarchical clustering of chs and f3h mutants became evident after 3 days of cold exposure. The proteins contributing most to this clustering were 5′‐adenylylsulfate reductase 3 (APR3), 4‐hydroxy‐tetrahydrodipicolinate reductase 1 (DAPB1), bifunctional aspartate aminotransferase and glutamate/aspartate‐prephenate aminotransferase (PAT), phosphoserine phosphatase (PSP), serine O‐acetyltransferase 3 (SAT3) and TSK‐associating protein 1 (TSA1; Figure 4c). APR3, PAT, PSP and SAT3 are involved in serine metabolism, which further suggests a differential regulation of the photorespiratory pathway in flavonoid mutants.

Based on the observed differential response of metabolic mutants in amino acids and enzymes related to serine metabolism, protein dynamics in the photorespiratory pathway with its connection to nitrogen, tetrahydrofolate‐related C1, sulphur and methionine metabolism were analysed (Figure 5). The most significant effects in the flavonoid mutants were observed in photorespiration reactions following the serine‐hydroxymethyltransferase 1 (SHM1) reaction. In contrast to Col‐0 and f3h, the chs mutant was not significantly affected in SHM1. Simultaneously, the levels of serine:glyoxylate aminotransferase (SGAT), glycerate kinase (GLYK) and plastid glycolate glycerate transporter 1 (PLGG1) decreased in chs during the initial 3 days at 4°C. Additionally, ferredoxin‐dependent glutamine:oxoglutarate aminotransferase (Fd‐GOGAT) and the exchange of glutamic acid and 2‐oxoglutaric acid across the plastid membrane were downregulated in chs plants. Notably, the chs mutant showed higher abundances of SHM1 and serine O‐acetyltransferase 3 (SAT3), along with slightly elevated abundances of O‐acetylserine (thiol) lyase C (OASC). Consequently, dynamics were observed in tetrahydrofolate‐related (THF‐related) C1 metabolism. 5‐formyl‐THF acts as a (potential) inhibitor of SHM1, and an increased abundance of 5‐formyltetrahydrofolate cycloligase (5‐FCL), that interconverts 5‐formyl‐THF into 5,10‐methenyl‐THF, in chs could possibly indicate a feedback regulation of glycine/serine ratio in mitochondria. The chs mutant showed an upregulated phosphorylated pathway of serine biosynthesis, which is a possible response in the plastid to alternations in serine‐consuming pathways in the mitochondria. THF‐related C1 metabolism connects photorespiration with methionine metabolism in the cytosol, and S‐adenosylmethionine synthase (SAM1) abundance showed a stronger reduction in Col‐0 than in chs and f3h mutants after 3 days of cold exposure (Figure 5). To validate if the deficiency of flavonoids leads to changes in photorespiration, net CO₂ assimilation rates were quantified as functions of light intensities and CO₂ concentrations. From these measurements, CO₂ compensation points, that is, the CO₂ concentrations at which NPS became 0, were calculated (Figure S8). Both flavonoid mutants showed higher CO₂ compensation points than Col‐0 after 3 days of cold acclimation, with the highest difference between chs and Col‐0 of 17.6 μmol mol⁻¹ (Figure S8c). This suggested higher rates of photorespiration in chs and f3h. Since no major differences were observed in glycine decarboxylase (GDC) abundance between the three genotypes, the increase of abundance of formate dehydrogenase (FDH) in chs and f3h indicated higher activity of this enzyme. This increase in FDH activity could contribute to a higher rate of CO₂ release, affecting the overall balance between photosynthesis and photorespiration (Figure 5). In summary, flavonoid metabolism appears to affect photorespiration and, consequently, plastidial amino acid metabolism after 3 days of cold exposure.

Figure 5.

Photorespiration and its interaction with nitrogen, tetrahydrofolate‐related C1 and sulphur metabolism.

Colour boxes below indicated proteins represent a change in average abundance between 0 and 3 days of cold acclimation in chs (first box), f3h (middle box) and Col‐0 (right box). Blue colour represents a decrease, red colour represents an increase in protein abundance, grey colour for PSP and SAT3 in f3h indicates an increase in protein abundance by more than 0.5 A.U. Asterisks and dots indicate significance (t‐test; *P < 0.05; filled dots indicate P < 0.07). In the case of the proteins with several isoforms present, the average median change was calculated. Metabolite abbreviations: 2‐OG, 2‐oxoglutaric acid; 2PG, 2‐phosphoglycolate; 3PGA, 3‐phosphoglycerate; 3PHP, 3‐phosphohydroxypyruvic acid; 3PS, 3‐phosphoserine; Cys, cysteine; FDH, formate dehydrogenase; FOLD1, bifunctional 5,10‐methylene‐THF dehydrogenase/5,10‐methenyl‐THF cyclohydrolase; Gly, glycine; Gln, glutamine; Glu, glutamic acid; OAS, O‐acetylserine; SAM, S‐adenosylmethionine; Ser, serine; THF, tetrahydrofolate. Protein abbreviations: 5‐FCL, 5‐formyltetrahydrofolate cycloligase; CAT2, catalase 2; DiT1 and DiT2.1, dicarboxylate transporters; Fd‐GOGAT, ferredoxin‐dependent glutamine:oxoglutarate aminotransferase; GDC, glycine decarboxylase; GGAT1/2, glutamate:glyoxylate aminotransferase 1/2; GLYK, glycerate kinase; GOX1/2, glycolate oxidase 1/2; GS2, glutamine synthetase 2; HPR1, hydroxypyruvate reductase 1; HPR2, hydroxypyruvate reductase 2; OASB, O‐acetylserine (thiol) lyase B; OASC, O‐acetylserine (thiol) lyase C; PGLP1, phosphoglycolate phosphatase 1; PGDH, 3‐phosphoglycerate dehydrogenase; PLGG1, plastid glycolate glycerate transporter 1; PSAT1, 3‐phosphoserine aminotransferase 1; PSP, phosphoserine phosphatase; SAM1, S‐adenosylmethionine synthase 1; SAT3, serine O‐acetyltransferase 3; SGAT, serine:glyoxylate aminotransferase; SHM1, serine‐hydroxymethyltransferase 1; SHM3, serine‐hydroxymethyltransferase 3. Created with BioRender.com.

Two mitochondrial proteins involved in THF‐related C1 metabolism, 5‐formyltetrahydrofolate cycloligase (5‐FCL) and FDH, were detected and quantified by the LC–MS/MS proteomics approach in the present study, and the deficiency of flavonoids had an impact on both of them (Figure 5). The THF‐related C1 metabolism is connected to the flavonoid biosynthesis pathway through the shikimate pathway, chorismate and the biosynthesis of para‐aminobenzoic acid (Kołton et al., 2022). Although shotgun proteomics applied in the present study could not detect or quantify all proteins of THF metabolism, similar dynamics in aminodeoxychorismate lyase abundance in chs and Col‐0 suggested that flavonoids might interact with THF‐related C1 metabolism not only by influencing precursor availability but also by direct interactions with central enzymes of this pathway. Evidence from a study on mice suggested a possible direct interaction between isoquercetin and cytosolic C‐1‐THF synthase (C1TC), a homologue of the bifunctional 5,10‐methylene‐THF dehydrogenase/5,10‐methenyl‐THF cyclohydrolase (FOLD1, 2, 3 and 4) proteins in Arabidopsis (Manzoor et al., 2022). The mitochondrial FOLD1 protein and its homologues in plants, yeast, mice and humans are evolutionarily related (Figure 6a). The structural alignment between the mitochondrial FOLD1 protein of Arabidopsis and cytosolic C‐1‐THF synthase of mice showed the root mean square deviation (RMSD) of 1.76 Å, despite only 67% similarity in their amino acid sequences. This indicates that the dehydrogenase/cyclohydrolase region maintains a stable structure while allowing diverse amino acid combinations (Figure 6b,c). Following this, isoquercetin was identified as a potential ligand for the highly conserved region of the FOLD1 protein. It formed polar connections with several amino acids: three amino acids interacting with the quercetin backbone and two amino acids interacting with the glucose moiety (Figure 6d; Supplementary File S1). The negative binding affinity score of −8.79 kcal mol⁻¹ indicated an interaction between the protein and the ligand in the identified binding pocket (Guntero et al., 2021; Sefika Feyza et al., 2022). Validation of the molecular docking experiment yielded a binding affinity of −8.94 kcal mol⁻¹, with RMSD of 1.49 Å (Table S5). In addition to the affinity score, negative intermolecular interaction energies, that is electrostatic energy and van der Waals energy, of isoquercetin and FOLD1 further supported a possible intermolecular interaction (Table S5). These findings suggest that flavonoids and their precursors affect THF‐related C1 metabolism, and consequently, photorespiration.

Figure 6.

Bifunctional 5,10‐methylene‐THF dehydrogenase/5,10‐methenyl‐THF cyclohydrolase (FOLD1) protein analysis.

(a) Phylogenetic tree of FOLD1 and its homologues with the Neighbour‐Joining method demonstrating common ancestry of these proteins. The branch length represents the number of substitutions per site. C1TC, cytosolic enzyme of yeast, mouse and human; C1TM, mitochondrial enzyme of yeast, mouse and human; FOLD3/4, plastidial enzymes of Arabidopsis; FOLD2, cytosolic enzyme of Arabidopsis.

(b) Pairwise protein structure alignment of Arabidopsis FOLD1 (blue) and dehydrogenase/cyclohydrolase functional domain of cytosolic mouse C1‐THF‐synthase protein (sand) with root mean square deviation (RMSD) 1.76 Å.

(c) Pairwise amino acid sequence alignment of Arabidopsis FOLD1 and dehydrogenase/cyclohydrolase functional domain of cytosolic mouse C1‐THF‐synthase protein. Turquoise colour indicates highly conservative regions of the protein that potentially interact with isoquercetin. Pink colour indicates interacting with isoquercetin amino acid residues; pink text indicates alignment RMSD of the highlighted regions.

(d) 3D interaction between the FOLD1 and isoquercetin after molecular docking analysis. Cyan colour: carbon backbone of isoquercetin molecule; salmon: carbon backbone of FOLD1; red: oxygen; grey: hydrogen; blue: nitrogen atoms. Yellow dashed lines represent polar contacts within 5 Å of the isoquercetin, numbers represent the distance in Å between the interacting atoms of isoquercetin and amino acid residues. The 3D model is provided in Supplementary File S1 and the molecular docking energy information is provided in Table S5.

Four amino acids involved in the interaction of isoquercetin and FOLD1 (Asp‐106, Gln‐158, Asp‐181 and Thr‐204) were identical between Arabidopsis and mouse (Figure 6c). A molecular docking experiment on the mouse C1TC protein revealed a binding pocket similar to that in Arabidopsis, with an affinity score of −8.075 kcal mol⁻¹ (Supplementary File S2). This interaction in the C1TC protein involved four amino acids, three of which were identical to Asp‐106, Gln‐158 and Asp‐181 in Arabidopsis. Alignment of the individual regions of FOLD1 and C1TC proteins that contain interacting amino acid residues revealed similar 3D structures, with an RMSD varying between 0.145 and 1.083 Å. The evolutionary conservation profile of the FOLD1 protein revealed these interacting amino acids to be highly conserved and exposed functional residues (Figure S9), further supporting the hypothesis of a conserved interaction mechanism between flavonoids and FOLD1 and its homologues.

Flavonoid deficiency affects total protein amounts and proteasomal compounds

Flavonoid deficiencies were found to affect amino acid metabolism (Figures 1, 4, and 5). To determine whether these dynamics in amino acid metabolism might also have an impact on protein homeostasis, the total protein amount was quantified photometrically in a Bradford assay and normalised to dry weight (Figure 7a). During the first 3 days of cold acclimation, total protein amounts decreased across all genotypes, with the chs and f3h mutants showing a more pronounced decrease than Col‐0. Additionally, between 3 and 14 days, protein amounts in both mutants were found to further decline, while Col‐0 stabilised its protein homeostasis. Overall, between 0 and 14 days, the chs mutant showed the most significant reduction in total protein level, with a median decrease of more than 2‐fold. In contrast, Col‐0 showed a stabilised protein content after 14 days of cold acclimation, with only a 1.2‐fold decrease (Figure 7a). After 3 days of cold, both chs and f3h mutants exhibited similar or higher levels of amino acids in the plastid, cytosol and vacuole compared to Col‐0 (Figure S7). After 14 days, both mutants showed reduced amino acid levels in the cytosol, although chs still had more than after 3 days. Only in f3h, amino acid levels in the cytosol significantly decreased after 14 days, indicating that the protein decrease in chs and f3h was not generally due to a general lack of substrates for protein biosynthesis, but rather due to alterations in the regulation of protein biosynthesis and/or degradation.

Figure 7.

Total protein amounts during cold acclimation.

(a) Total protein amount; (b) cytosolic ribosome and (c) proteasome‐associated protein abundance in Col‐0, chs and f3h before (0 day) and after (3 or 14 days) of cold acclimation. A.U., log₂ transformation of LFQ intensities. Points represent biological replicates (n ≥ 4), asterisks represent significant differences (ANOVA with Tukey HSD, P‐value<0.05), and the full list of significance levels is provided in Table S2A.

(d) Pearson correlation analysis between total protein amount (x‐axis) and cytosolic ribosome‐associated protein abundance (primary y‐axis, pink), and between total protein amount (x‐axis) and proteasome‐associated protein abundance (secondary y‐axis, blue) in Col‐0, chs and f3h across 0, 3 and 14 days of cold acclimation. Points represent independent replicates, n = 5; the solid line represents linear regression and the shaded area represents the confidence interval of the linear regression.

During early (0–3 days) and later (3–14 days) stages of cold acclimation, genotypes could be distinguished by dynamics of abundances of (cytosolic) ribosomal and proteasomal proteins (Figure 7). During the first 3 days of cold, compounds of the cytosolic ribosome showed an insignificant differential trend in chs plants, with ribosome abundance slightly decreasing during the first 3 days at low temperature (Figure 7b). In contrast, Col‐0 and f3h medians increased in this period. On the other hand, during the same period, protein abundance of proteasomal compounds decreased in Col‐0 but increased in both flavonoid mutants, potentially contributing to the initial decline in protein content (Figure 7c). During the later cold acclimation period, that is, after 14 days, both mutants displayed a wild‐type‐like increase in the cytosolic ribosome and proteasome abundance (Figure 7b,c). Pearson correlation analysis between total protein amount and either cytosolic ribosome abundance or proteasome abundance revealed a strong significant negative correlation throughout the whole period of cold acclimation in chs and f3h, but not in Col‐0 (Figure 7d). This finding suggests a complex interaction between ribosome and proteasome machinery and flavonoid accumulation during the later stages of cold acclimation.

Differential protein abundance GO enrichment analysis on the whole proteome dynamics in the chs mutant after 14 days of cold revealed an increase of proteins related to cytoplasmic translation, proteasomal protein catabolic process and ribosome assembly (Table S6). Moreover, unlike Col‐0, both chs and f3h did not show an upregulation of ‘protein folding’. In summary, these results suggest an effect (direct or indirect) of flavonoids on protein biosynthesis, degradation and folding.

DISCUSSION

Flavonoid deficiency affects the C/N interface during cold acclimation

Metabolic acclimation of plants during cold exposure has been described in many studies to comprise and affect a large array of enzymatic reactions and transport processes across intracellular membrane systems which ultimately results in adjustment of metabolite concentrations (see Garcia‐Molina et al., 2020; Hannah et al., 2006; Hoermiller et al., 2017; Korn et al., 2008; Koster & Lynch, 1992). Although many of these metabolic adjustments are strong and significant, frequently, their physiological function remains elusive. This might be due to the high number of enzymes involved in biosynthesis, degradation and compartmental sequestration which contribute to the physiological function. For example, the accumulation of soluble carbohydrates, for example, sucrose, hexoses and raffinose oligosaccharides is a well‐known cold response of diverse plant species (Alberdi et al., 1993; Kaplan et al., 2004; Klotke et al., 2004; Koster & Lynch, 1992). Although biophysical studies provided evidence for the protective functions of carbohydrates during cold and freezing in vitro (Hincha et al., 2003), the in vivo function can hardly be predicted without subcellular localization (Knaupp et al., 2011; Zuther et al., 2004).

Based on such observations, in the present study, plant metabolism was resolved to a subcellular and compartment‐specific level to reduce the ambiguity of functional interpretation of metabolite dynamics. Time series analysis of GC–MS derived metabolomics data unravelled a cold‐induced perturbation of citrate and glutamate metabolism in chs and f3h mutants (see Figure 1; Figure S2). Subcellular proteomics suggested a shift of ICDH activity from mitochondria to the cytosol due to flavonoid deficiency (see Figure 2; Figure S4). Such a cytosolic bypass was discussed earlier in the context of C/N balance and interconversion of citrate storage outside the mitochondria (Sweetlove et al., 2010). Even more, for illuminated leaves, such non‐cyclic TCA cycle flux modes were shown to reduce PDH‐catalysed decarboxylation rates by up to 75% (Tcherkez et al., 2009). Carbon skeletons of such fluxes are then substrates for glutamate and glutamine biosynthesis. Due to the finding that both metabolites over‐accumulated in the cytosol and plastids of chs and f3h during the first 3 days of cold acclimation, this suggests that flavonoid deficiency affects, or maybe even co‐regulates, the C/N balance during cold exposure. Also, previous work suggested that a high C/N ratio under low nitrogen availability increases flavonoid content (Li et al., 2023). Further, flavonoid pathway activators PAP1 and PAP2 were found to strongly respond to low temperature and low nitrogen (Olsen et al., 2009). Together with the data of the present study, this suggests that C/N balance and flavonoid metabolism are mutually regulated. This is also supported by the finding that in chs, both GOGAT enzymes and plastidial dicarboxylate transporters were significantly downregulated during the initial cold acclimation period. Hence, during the early phase of cold acclimation, flavonoid biosynthesis may stabilise plastidial C/N metabolism and, indirectly, photosynthesis by consuming carbon skeletons in the cytosol which originate from citrate metabolism. Finally, this might connect mitochondrial, cytosolic and plastidial redox and energy balance under environmental fluctuation.

Flavonoid biosynthesis results from a combination of the phenylpropanoid pathway and the acetate pathway (Perez de Souza et al., 2020). Acetyl‐CoA plays a central role in multiple pathways, including citrate metabolism. Based on the observation that CICDH positively correlated with PDH‐E1 alpha (see Figure 2), this suggests that flavonoid deficiency could affect the metabolic flux through the acetate pathway in a feedback manner. This, in turn, would increase the availability of acetyl‐CoA for other pathways, such as fatty acid and lipid biosynthesis. Indeed, flavonoid deficiency was associated with a significant increase in triacylglycerols under cold (Tables S3 and S7 and S7), further suggesting a potential role of flavonoids in the regulation of associated pathways.

Flavonoid metabolism interacts with photorespiration and amino acid metabolism

The analysis of primary metabolism revealed a strong and significant effect of flavonoid deficiency on amino acid metabolism, and this effect was most pronounced in plastids (see Figure 4). While it may not be surprising that the metabolism of aromatic amino acids, which are precursors for flavonoid biosynthesis, are affected in chs and f3h mutants, it is even more interesting to see emphasised effects in the photorespiratory intermediates glycine and serine. Due to experimental limitations of the NAF protocol, in which separation of plastidial and mitochondrial metabolomes remains difficult (Fürtauer et al., 2019), the presented data cannot be exclusively interpreted as plastidial or mitochondrial metabolite concentrations. However, due to the accumulation of the substrate glycine and the depletion of the product serine of the mitochondrial SHM1‐catalysed reaction, it appears plausible that the observed metabolic phenotype is due to affected photorespiratory regulation. Further, quantified protein dynamics suggested that deficiencies in the flavonoid biosynthesis pathway resulted in upregulated plastidial serine biosynthesis (see Figure 5). This might be a consequence of a depleted mitochondrial serine pool which limits its availability for cellular functions and might be compensated for by this plastidial bypass. Plastidial proteins PGDH, PSP and SHM3 were found to be upregulated in both chs and f3h plants. Simultaneously, cysteine biosynthesis via O‐acetylserine was upregulated during the first 3 days of cold acclimation in these mutants which further indicates that mitochondrial deficiencies in serine biosynthesis may have been compensated for in the plastids. Additionally, the alternations in photorespiration caused by mutations in the flavonoid biosynthesis pathway led to a shift in CO₂ compensation points in chs and f3h after 3 days of cold acclimation, potentially due to increased activity of FDH.

In mitochondria, SHM1 catalyses the interconversion of glycine to serine which comprises a methylene transfer from 5,10‐methylene‐THF (Shi & Bloom, 2021). Alternatively, 5,10‐methylene‐THF might be interconverted into 5,10‐methenly‐THF, catalysed by FOLD1 (AT2G38660), which represents a bifunctional 5,10‐methylene‐THF dehydrogenase/5,10‐methenyl‐THF cyclohydrolase being localised to mitochondria (Collakova et al., 2008; Groth et al., 2016). The competitive inhibitor of SHM1 and other folate‐dependent enzymes, 5‐formyl‐THF, is synthesised by SHM1 in the presence of glycine (Stover & Schirch, 1990). In chs plants, a significant and strong upregulation of 5‐FCL during the first 3 days at 4°C was observed which metabolises 5‐formyl‐THF (Roje et al., 2002). Even though 5‐FCL was also upregulated in Col‐0, the upregulation was stronger in both flavonoid mutants and only significant in chs (Figure 5). Although so far not described in plants, quercetin uptake by mitochondria was reported for human cells (Fiorani et al., 2010). Thus, a potential explanation for the upregulation of 5‐FCL in chs might be that metabolic compounds or pathways that are affected by the chs mutation represent potential inhibitors of 5‐formyl‐THF biosynthesis. Indeed, the FOLD1‐homolgue in mice, C1TC, was observed to be inhibited by isoquercetin, which is produced downstream of CHS (Manzoor et al., 2022). Isoquercetin and quercetin were shown to interact with FOLD1 in silico (see Figure 6). However, isoquercetin, due to its glucoside moiety, formed more connections and had a slightly lower affinity score than quercetin (−8.79 vs. −8.02 kcal mol⁻¹; see Table S5), potentially establishing a more stable complex. Such a possible regulatory role of isoquercetin was further supported by the finding that it significantly accumulated in Col‐0 after 3 days of cold, which was observed neither in f3h nor in chs (Figure S10; Table S2C). Together with the observation that, until 3 days at 4°C, anthocyanin amounts, that is, late products of the flavonoid pathway, did not significantly increase (Kitashova et al., 2023), this further suggests that early intermediates of flavonoid biosynthesis clearly differ in their function from end products of the same pathway. While early intermediates of the flavonoid biosynthesis pathway seem to play a crucial role in metabolic acclimation, late products, such as anthocyanins, might rather have a function as light or UV protectants (Li et al., 1993; Zhang et al., 2017; Zirngibl et al., 2023). In f3h plants, quercetin and isoquercetin were still found to accumulate as already reported before (Owens et al., 2008). However, under low temperature, these amounts were very small compared to Col‐0 (Figure S10). Still, this needs to be considered when the potential conserved effects of flavonoids on metabolism and its regulation are discussed.

Changes in protein homeostasis due to flavonoid deficiency during cold acclimation

Cold acclimation resulted in a decrease of protein amounts normalised to sample dry mass (see Figure 7). While Col‐0 was found to successfully stabilise its protein amount on a slightly lower level than at 22°C, both f3h and chs mutants showed a significant decrease over the full cold acclimation period. Proteomics data revealed that in both flavonoid mutants proteasomal complexes were upregulated during the initial phase of cold acclimation, that is, during 3 days at 4°C, while a negative trend was observed for Col‐0. This resulted in significantly negative correlations of proteasome and protein amounts in f3h and chs while no significant correlation was observed in Col‐0. This points towards a function of flavonoids and/or intermediates of flavonoid biosynthesis in stabilisation of the protein homeostasis at low temperature. Interestingly, in a GO term enrichment analysis, ‘protein folding’ was significantly over‐represented in chs between 0 and 3 days at 4°C while no significance was detected in Col‐0 and f3h (Table S6). The functional category ‘protein refolding’ was decreased in this period in Col‐0 while it remained constant in both mutants. While, to the best of our knowledge, up to today there is no direct experimental proof for flavonoid‐based stabilisation of proteins during plant cold acclimation, such interaction has been shown in other systems. For example, an in vitro study in mice showed that flavonoids, for example, quercetin or myricetin, enhanced the stability and folding of the visual G protein‐coupled receptor rhodopsin, which is expressed in the rod photoreceptors of the bovine eye (Ortega et al., 2019). It was observed that flavonoids modulate the protein's conformation by direct interaction and stabilise it, probably by introducing structural rigidity. Hence, deficiencies of flavonoids might lead to misfolded proteins which subsequently undergo refolding or proteasomal degradation. Another study showed that flavonoids inhibited proteasome 26S activity in pig red blood cells (Chang, 2009), further supporting the role of flavonoids in inhibiting proteasomal compounds and, thereby, regulating protein degradation rates.

Although the effect on ribosomes in chs and f3h was less pronounced than on the proteasome, it still became significant in the correlation with protein amounts (see Figure 7d). Previous work has highlighted the important role of ribosomes and translational regulation during cold acclimation (Garcia‐Molina et al., 2020). Moreover, flavonoids were shown to interact with isolated tRNA (Kanakis et al., 2006). In line with this, data from the present study suggests that flavonoids influence protein homeostasis during cold exposure, maybe through their direct interactions with ribosomes or tRNA. In summary, although it remains speculation from experimental data of the present study, it seems plausible that also in plants, such interactions between flavonoids and proteins as well as the proteasome, ribosomes and/or tRNAs might occur and contribute to a stable protein homeostasis at low temperature.

CONCLUSION

The findings of the present study indicate that flavonoids play different physiological roles during a 2‐week cold acclimation period. During the initial cold response, flavonoids were found to affect, and maybe even regulate, photorespiration and stabilisation of the C/N balance between plastids, cytosol and mitochondria. During the late cold acclimation period, flavonoids affect protein stability and folding which has dramatic effects on a plant's energy homeostasis. While the experimental design of the present study does not allow discrimination between direct and secondary effects of flavonoid deficiency on cold acclimation, we have observed a robust photosynthetic acclimation in f3h and chs in a previous study (Kitashova et al., 2023). This might support the assumption of direct effects on acclimation capacity, but it remains speculation at this point. It remains to be elucidated whether such functional diversity can also be observed under different flavonoid‐stimulating conditions, such as high light or excess UV. Finally, however, the physiological function of flavonoids and involved pathways clearly goes beyond the accumulation of pigments for the dissipation of electromagnetic radiation energy.

MATERIALS AND METHODS

Plant material and growth conditions

Plants of Arabidopsis thaliana accession Col‐0 and homozygous T‐DNA insertion lines chs (chalcone synthase, line SALK_020583C, locus AT5G13930, allele tt4‐11), and f3h (flavanone 3‐hydroxylase, line SALK_113904C, locus AT3G51240, allele tt6‐3) were grown as described before (Kitashova et al., 2023). In summary, plants were grown on a 1:1 mixture of GS90 soil and vermiculite in a climate chamber under controlled short‐day conditions (8 h/16 h light/dark; 100 μmol m⁻² sec⁻¹; 22°C/16°C; 60% relative humidity). After 4 weeks, plants were shifted to a growth cabinet and grown further under long day conditions (16 h/8 h light/dark; 100 μmol m⁻² sec⁻¹; 22°C/16°C; 60% relative humidity). After 2 weeks, plants were either (i) harvested at midday, that is, 8 h of light (0 days of cold acclimation) or (ii) transferred to a cold room for cold acclimation (16 h/8 h light/dark; 90–100 μmol m⁻² sec⁻¹; 4°C/4°C). Cold exposed plants were harvested after 3 or 14 days of acclimation at midday, that is, after 8 h of light. Each sample consisted of nine leaf rosettes which were immediately frozen in liquid nitrogen, ground to a fine powder and lyophilised.

Net CO₂ assimilation rate measurements

Rates of net photosynthetic CO₂ uptake (NPS) were recorded using a WALZ® GFS‐3000FL system with Standard Measuring Head 3010‐S (Heinz Walz GmbH, www.walz.com). NPS was determined within two response curves: (i) light curves where photosynthetically active radiation (PAR) increased stepwise from 0 to 1200 μmol m⁻² sec⁻¹ in 5 min intervals under constant CO₂ concentration of 450 ppm, and (ii) CO₂ curves where CO₂ concentration increased stepwise from 50 to 1200 ppm in 5 min intervals under constant PAR intensity of 1200 μmol m⁻² sec⁻¹. The CO₂ compensation point, that is, the CO₂ concentration at which NPS was 0, was calculated using a linear interpolation of NPS between 50 and 450 ppm CO₂.

Non‐aqueous subcellular fractionation

Non‐aqueous subcellular fractionation (NAF) was performed as described before with slight modifications (Fürtauer et al., 2016). In brief, approximately 10 mg of lyophilised plant material were homogenised in tetrachloroethylene (ρ = 1.60 g cm⁻³) with an ultrasonic homogeniser (Hielscher Ultrasonics UP200St, www.hielscher.com). After centrifugation with 20 000g, the supernatant was transferred to another tube, and its density was reduced with heptane (ρ = 0.68 g cm⁻³). The pellet was then resuspended in tetrachloroethylene. Sonication of the supernatant with the adjusted density, centrifugation, separation of pellet and density adjustment of the supernatant were repeated several times until a gradient of 5–7 densities was obtained. Each of the resulting resuspended pellets was split into 2 sub‐fractions and dried in a desiccator. These fractions were used for (i) marker enzyme quantification and (ii) primary metabolite quantification using GC–MS. Activities of marker enzymes, that is, plastidial pyrophosphatase, cytosolic uridine 5'diphosphoglucose pyrophosphorylase and vacuolar acidic phosphatase, were determined photometrically and correlated with metabolite amounts (Fürtauer et al., 2016). Based on observations about the distribution of marker proteins made in a previous study, it was assumed that the plastidial fraction included both plastidial and mitochondrial metabolites due to limitations in separating both compartments (Fürtauer et al., 2019). Therefore, assumptions regarding the separation of plastidial and mitochondrial metabolomes were made carefully, aligning them with the observed dynamics of the subcellular proteome.

Extraction of lipids and primary metabolites for mass spectrometry

Lipids and non‐polar metabolites were extracted following the method described by (Hummel et al., 2011). In summary, approximately 5 mg of freeze‐dried leaf material or NAF fractionation pellets were subjected to extraction using 1 mL of a pre‐cooled (−20°C) mixture of methanol and methyl tert‐butyl ether in a ratio of 3:1. This extraction solution included 10 μL corticosterone, 5 μL chloramphenicol, 1.25 μL ampicillin, 2.5 μL sorbitol‐¹³C and 5 μL ribitol, each at a concentration of 0.2 mg mL⁻¹, serving as internal standards. The mixture was vigorously mixed until the tissue was fully suspended and then shaken for 10 min at 4°C, followed by 10 min of sonication on ice. Phase separation was initiated by adding 500 μL of water:methanol (3:1) solution, followed by vigorous mixing and centrifugation at maximum speed using a Centrifuge 5417R (Eppendorf, www.eppendorf.com). For lipid analysis, 600–700 μL of the upper phase was collected, while 400–500 μL of the lower phase was divided into equal aliquots for GC–TOF–MS analysis (sample dependent). All samples were dried using a vacuum concentrator (Concentrator 5301; www.eppendorf.com) and subsequently stored at −80°C until further analysis. To prevent oxidation, argon was added during storage.

GC–MS: Primary metabolite analysis

For the derivatization process, the pellet was suspended in 10 μL of methoxyamine hydrochloride solution (20 mg mL⁻¹ in pyridine) and incubated for 90 min at 40°C. Following this, 20 μL of BSTFA (N,O‐Bis[trimethylsilyl]trifluoroacetamide) containing 2.5 μL of a retention time standard mixture comprising linear alkanes (n‐decane, n‐dodecane, n‐pentadecane, n‐nonadecane, n‐docosane, n‐octacosane and n‐dotriacontane) was added, and the mixture was further incubated at 40°C for 45 min. Subsequently, 1 and 2 μL of each sample was injected into a GC–TOF–MS system (Pegasus HT, Leco, www.leco.com) utilising splitless injection as well as split ratios of 10, 30 and 50 (metabolite dependent). Sample injection and processing were automated using an autosampler system (Combi PAL, CTC Analytics AG, www.ctc.ch). Helium was employed as the carrier gas at a constant flow rate of 0.6 mL min⁻¹. Gas chromatographic separation was conducted on an Agilent GC (7890A, Agilent, www.agilent.com) equipped with a 30 m VF‐5 ms column coupled with a 10 m EZ‐Guard column. The temperature of the split/splitless injector, transfer line and ion source was maintained at 250°C. The initial oven temperature was set at 70°C and ramped up to 350°C at a rate of 9°C min⁻¹. Metabolites were fractionated and ionised by a 70 eV ion pulse. Mass spectra were acquired at a rate of 20 scans per second within an m/z range of 50–600. Chromatograms and mass spectra were analysed using ChromaTOF 4.72 and TagFinder 4.1 software (Luedemann et al., 2008). Raw values were normalised by internal standard and dry weight.

LC–MS: Lipid analysis

For LC–MS analysis, the Dionex Ultimate 3000 UHPLC (Thermo Fisher Scientific, www.thermofisher.com) in combination with a timsTOF (Bruker, www.bruker.com) was used. The dry extract was resolved in acetonitrile:isopropanol (7:3) and injected on a C8 reversed‐phase column (Ultra C8 100 × 2.1 mm; Restek, www.restek.com) with 300 μL min⁻¹ flow at 60°C. The solvents used are (A) water and (B) acetonitrile: isopropanol (7:3), both including 1% (v/v) ammonium acetate and 0.1% (v/v) acetic acid. The 26 min gradient started at 55% B, followed by a ramp to 99% B within 15 min. After a 5‐min washing step at 99% B, the gradient was returned to 55% B and kept constant for 5 min equilibration.

MS detection was performed using an electrospray ionisation (ESI) source, operating in positive mode. Nitrogen served as the dry gas, at 8 L min⁻¹, 8 bar and 200°C. The timsTOF mass spectra were recorded in MS and MSMS mode from 50 to 1300 m/z with 40 000 resolution, 1 Hz scan speed and 0.3 ppm mass accuracy. Compounds were annotated in a targeted approach using the specific mass (m/z) at retention time and the isotopic pattern. All data were acquired by Compass HyStar 4.1 and otofControl 6.2. The evaluation was performed by DataAnalysis 5.1 and MetaboScape 2021. All software tools were provided by Bruker. Raw values were normalised by internal standard and gram dry weight.

Extraction and quantification of flavonoids

The extraction of flavonoids was performed as described previously with some modifications (Likić et al., 2014). In brief, 15 mg of dry plant material was treated with 1 mL of a 1:1 (v/v) mixture of methanol and 1.2 M hydrochloric acid containing 2 μg mL⁻¹ of corticosterone as an internal standard. The mixture was thoroughly vortexed and incubated at 80°C with shaking at 1000 rpm for 30 min, supplemented by a 30 sec ultrasonication period (Hielscher UP200St). Following ultrasonication, the samples were centrifuged at 8°C for 15 min at maximum speed to clarify the supernatant, which was transferred into a clean tube. This process was repeated to ensure the complete removal of plant debris. Chromatographic separation was achieved using a Dionex Ultimate 3000 UHPLC system coupled to a timsTOF mass spectrometer (Bruker). An aliquot of 60 μL of each extract was introduced onto a C18 reversed‐phase column (Ultra AQ C18 3 μm 100 × 2.1 mm) maintained at 30°C, employing a mobile phase of (A) water and (B) acetonitrile, each containing 0.1% (v/v) formic acid, at a flow rate of 400 μL min⁻¹. The gradient program started with 95% aqueous phase (A) for 2 min, transitioning to 95% organic phase (B) over 20 min, followed by a 3‐min organic wash and a 5‐min re‐equilibration to initial conditions.

Mass spectrometric detection was performed using an electrospray ionisation source (ESI) operating in positive ion mode. The parameters set included a dry gas flow of 8 L min⁻¹, a nebuliser pressure of 8 bar, and a source temperature of 200°C. Mass spectra were collected from 50 to 1300 m/z with a resolution of 40 000, Scan speed of 1 Hz, and mass accuracy of 0.3 ppm. Identification of compounds based on exact mass, true isotopic patterns and retention time confirmed by commercial standards. Data acquisition and processing were conducted using otofControl 6.2, DataAnalysis 5.3 and MetaboScape 2021, with further analysis in Microsoft Excel.

Protein extraction and analysis by LC–MS/MS

Protein extraction and preparation for the LC–MS/MS quantification was done as previously described with minor adjustments (Marino et al., 2019). In summary, proteins were extracted using 6 M guanidine‐chlorine in 10 mM HEPES pH 7.8, supplemented with 1 tablet of protease inhibitor cocktail per 10 mL buffer (cOmplete™ Proteasehemmer‐Cocktail, ©Roche) with an ultrasonic homogeniser (Hielscher Ultrasonics UP200St, www.hielscher.com). Then, the proteins were precipitated with methanol/chloroform in water, washed with methanol and the supernatant was discarded, leaving protein‐containing pellets. Dried pellets were resuspended in 6 M urea/2 M thiourea in 50 mM HEPES pH 7.8 buffer at 37°C. Total protein content was quantified using the Pierce™ 660 nm Protein Assay Kit (©Thermo Scientific™). Following quantification, 80 μg of protein was aliquoted, reduced with 10 mM DTT and alkylated with 50 mM iodoacetamide. The proteins were subsequently digested overnight with trypsin at 37°C. Acidified with formic acid samples were purified using home‐made C18 stage tips with 80% acetonitrile and 0.5% formic acid solution. Finally, 1 μg of purified protein extract was used for the LC–MS/MS analysis.

LC–MS/MS analysis was conducted as previously described with minor modifications, involving peptide separation over a 90 min linear gradient ranging from 5 to 80% (v/v) acetonitrile (ACN) (Espinoza‐Corral et al., 2023). Raw data files were processed using MaxQuant software version 2.2.0.0 (Cox & Mann, 2008), with peak lists searched against the Arabidopsis reference proteome from Uniprot (www.uniprot.org), employing default settings with ‘match‐between‐runs’ enabled. Protein quantification was achieved utilising the label‐free quantification algorithm (LFQ) (Cox et al., 2014). Subsequent analysis was conducted using Perseus version 2.0.9.0 (Tyanova et al., 2016). To refine the dataset, potential contaminants, proteins identified solely through site modification and reverse hits were excluded. Only protein groups quantifiable by the LFQ algorithm in at least three out of four replicates under at least one condition were retained. LFQ intensities underwent log₂ transformation, and missing values were imputed from a normal distribution within Perseus using standard settings.

Statistics and data analysis

Dynamics of the subcellular metabolomes were analysed in MATLAB® (www.themathworks.com) using a method for time series analysis in the context of biochemical network information (Nägele et al., 2016). In brief, sigmoidal Gompertz functions (Equation 1) were fitted to experimental data (0–3 days at 4°C, 3–14 days at 4°C).

$$h\left(x\right)=d+\left(a-d\right){\mathrm{e}}^{-{\mathrm{e}}^{-b\left(x-c\right)}}$$ (1)

Gompertz functions describe growth rates with the slowest growth at the beginning of the period, that is, reflecting the moment of cold exposure and thermodynamically affected enzymatic reaction rates. Derivatives of Gompertz functions represented metabolic functions, that is, the summed rates of biosynthesis and degradation for each metabolite pool. Dynamics of metabolic functions were then compared to the dynamics of substrate concentrations of each enzymatic reaction described by a metabolic network of subcellular carbohydrate metabolism, carboxylic acids and amino acids (Figure S1). Dynamics were indicated by $\omega$‐functions (Equation 2):

$$\omega \left(M_i\to M_j,t\right)=\frac{\frac{d}{\mathit{dt}}{f}_{M_j}\left(\mathit{M}\mathbf{,}\mathit{p},t\right)}{f_{M_i}\left(\mathit{M}\mathbf{,}\mathit{p},t\right)}$$ (2)

Here, M _i denotes the substrate concentration of an enzymatic reaction and M _j represents the product concentration. Metabolic functions of both substrate and product pools, $f_{M_i}$ and $f_{M_j}$, were derived from Gompertz functions as described before. To identify metabolic deregulation in both mutants, ratios of z‐scaled omega values were built (Figure S2; Table S1).

Statistical analysis and figure preparation was performed with free software R version 4.2.2 (www.r‐project.org) (R Core Team 2019) and RStudio version 2023.9.1.494 (www.rstudio.com), Python version 3.8.8 (www.python.org), Jupyter Notebook version 6.3.0 (https://jupyter.org) and MATLAB® (www.themathworks.com). The protein phylogeny analysis was performed using MUSCLE Multiple Sequence Alignment version 3.8, and pairwise sequence alignment was conducted using the EMBOSS Water tool (https://www.ebi.ac.uk). Pairwise protein structure alignment of proteins was performed using AlphaFold structures (https://www.rcsb.org). Ligand docking simulation was performed using the free software DockThor version 2.0 (https://dockthor.lncc.br/v2). The results from ligand docking were visualised using PyMOL version 2.5.7 (https://pymol.org). The evolutionary conservation profile of the FOLD1 protein was performed using the free software ConSurf (https://consurf.tau.ac.il). Differentially abundant proteins (DAP) were identified according to the ANOVA significance test with Tukey HSD (Table S2A). DAP GO term enrichment analysis (Biological Process) was performed using the R ‘org.At.tair.db’ package version 3.16.0 and ‘clusterProfiler’ package version 4.6.2. GO terms were checked for redundancies with the most significant terms being kept.

AUTHOR CONTRIBUTIONS

AK performed fractionation experiments, statistical analysis and data evaluation, and wrote the paper. ML performed metabolomics analyses, MM performed quantification of flavonoids, SS and DL performed, supervised and conceived proteomics analyses, and TN conceived the study, performed statistics and data evaluation and wrote the paper.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

Supporting information

Figure S1. Graphical representation of the metabolic network of subcellular carbohydrate metabolism, carboxylic acids and amino acids used for the analysis of the dynamics of subcellular metabolomes.

Figure S2. Ratios of omega functions revealed by time series analysis.

Figure S3. Cold‐induced dynamics of the subcellular metabolome.

Figure S4. Dynamics of isocitrate dehydrogenases and pyruvate dehydrogenases in Col‐0, chs and f3h after 0, 3 and 14 days of cold acclimation.

Figure S5. Growth phenotypes of Col‐0, chs and f3h after 0, 3 and 14 days of cold acclimation.

Figure S6. Hierarchical cluster analysis of the subcellular metabolite abundance change rate.

Figure S7. Dynamics of main amino acid pool in Col‐0, chs and f3h in cytosol, plastid and vacuole after 0, 3 and 14 days of cold acclimation.

Figure S8. Net CO₂ assimilation rates.

Figure S9. Evolutionary conservation profile of FOLD1 protein.

Figure S10. Cold‐induced dynamics of flavonoids.

Table S1. Ratios of scaled time series omega functions. Interpolated Gompertz functions were used to calculate omega functions (see Materials and methods section). Functions were scaled (z‐scale, i.e., zero mean—unit variance) before ratios of function values were built for (A) chs/Col‐0 (0–3 days), (B) f3h/Col‐0 (0–3 days), (C) chs/Col‐0 (3–14 days) and (D) f3h/Col‐0 (3–14 days).

Table S2. ANOVA with Tukey HSD results. Significance levels for (A) proteome level; (B) subcellular metabolomics; (C) cellular metabolomics and (D) comparison of CO₂ compensation points; P‐values below 0.05 are highlighted in green.

Table S3. Principal component analysis (PCA) loadings and component scores for metabolites quantified.

Table S4. Cluster analysis results for proteins involved in amino acid biosynthesis. (A) List of proteins involved in amino acid biosynthesis; (B) row‐wise and (C) column‐wise calculated Euclidean distances.

Table S5. Summary of molecular docking results for isoquercetin and quercetin with bifunctional 5,10‐methylene‐THF dehydrogenase/5,10‐methenyl‐THF cyclohydrolase (FOLD1) protein.

Table S6. GO term enrichment analysis. Differentially abundant proteins in chs and f3h were identified based on ANOVA with Tukey HSD results between 0 and 3 days, and 3 and 14 days of cold acclimation (Table S2A). Only GO terms related to biological processes are included in the analysis.

Table S7. Cold‐induced dynamics of metabolome in Col‐0, chs and f3h after 0, 3 and 14 days of cold acclimation. Mean values and standard deviations are provided for (A) subcellular and (B) cellular metabolite amounts.

Supplementary File 1. 3D structure of isoquercetin and bifunctional 5,10‐methylene‐THF dehydrogenase/5,10‐methenyl‐THF cyclohydrolase (FOLD1) protein of Arabidopsis interaction.

Supplementary File 2. 3D structure of isoquercetin and C‐1‐THF synthase protein of mouse interaction.

Supplementary File 3. 3D structure of quercetin and bifunctional 5,10‐methylene‐THF dehydrogenase/5,10‐methenyl‐THF cyclohydrolase (FOLD1) protein of Arabidopsis interaction.

DATA AVAILABILITY STATEMENT

Proteomics and metabolomics data acquired in this study are accessible via the DataPLANT platform, which is part of NFDI: https://git.nfdi4plants.org/thomas.naegele/2024_Kitashova_et_al_subcellular.