1. Introduction

Global climate changes subject plants to heightened environmental stresses, which disrupt their physiological processes and cause elevated reactive oxygen species (ROS) accumulation (Castro et al., 2021; Dietz and Vogelsang, 2024; Mittler et al., 2022). ROS serve as dual-edged signals in plant stress responses, at physiological concentrations, transient accumulation of ROS activates defense pathways (Hasanuzzaman et al., 2020), including the activation of MAPK cascades, calcium signaling, hormone signaling especially the ethylene and abscisic acid (ABA), and the transcriptional reprogramming (Wang et al., 2024). A prominent example of such signaling is the localized mitochondrial ROS burst, which provokes plant biotic stress responses and ABA-mediated defense pathways (Gleason et al., 2011). However, uncontrolled ROS accumulation adversely affects components of surrounding biological system, including lipids, proteins, and DNA, and triggers physiological processes that ultimately lead to cellular injury or cell death (Lee and Gould, 2002; Mittler, 2017).

To mitigate oxidative stress, plants employ both enzymatic scavengers and non-enzymatic antioxidants. Among the latter, vacuolar flavonoids such as anthocyanins, the most extensively studied flavonoid group responsible for red to blue plant pigmentation, scavenge ROS in situ and mitigate the ROS-induced damage (Davies et al., 2024; Lee and Gould, 2002). They also protect photosynthetic machinery and enhance tolerance to environmental stress (Chalker-Scott, 1999; Davies et al., 2018; Landi et al., 2015). Anthocyanin biosynthesis is derived from the phenylpropanoid pathway, catalyzed by a series of cytoplasm-localized structural enzymes, including phenylalanine ammonia-lyase (PAL), chalcone synthase (CHS), chalcone isomerase (CHI), flavanone 3-hydroxylase (F3H), flavonoid 3′-hydroxylase (F3′H), dihydroflavonol 4-reductase (DFR), and anthocyanin synthase (ANS) (Jaakola, 2013). Additionally, synthesized anthocyanidins undergo different modifications, such as glycosylation, mediated by uridine diphosphate glycosyltransferase (UGT). Finally, the anthocyanins are transported into the vacuole via glutathione transferase (GST)-mediated transportation to exert their biological functions (Buhrman et al., 2022; Koes et al., 2005; Springob et al., 2003). The biosynthesis of anthocyanins is primarily transcriptionally controlled by a well-defined MYB-bHLH-WD40 (MBW) complex, in which some MYB transcription factors play a determinant role in ROS response besides in regulation of anthocyanin biosynthesis (Mahmoud et al., 2024; Page et al., 2012; Qin et al., 2021; Xu et al., 2017; Zhu et al., 2023). In Medicago truncatula, abiotic stresses like salinity and high light potently induce anthocyanin accumulation like in other plants, implying conserved regulatory mechanisms in legumes (Karami et al., 2015; Wang et al., 2022). Despite this conserved stress-responsive accumulation, the molecular switches that directly coordinate ROS homeostasis with anthocyanin biosynthesis remain largely unexplored in most plants, particularly in legumes.

ROS are generated through intracellular redox reactions, primarily from mitochondrial electron transport chains, plasma membrane NADPH oxidase, peroxisomal oxidative reactions, and other enzymatic and non-enzymatic processes (Corpas et al., 2020; Dietz and Vogelsang, 2024; Noctor et al., 2018; Unal et al., 2020). Within mitochondria, Complex Ⅰ (NADH dehydrogenase) and Complex Ⅲ (ubiquinone-cytochrome c reductase) are the main sites for ROS generation (Lenaz, 2001). Impairment of these complexes, often associated with elevated ROS production or stress, triggers mitochondrial retrograde signaling to reprogram nuclear gene expression (Qiu et al., 2021; Yang et al., 2022). The key components of mitochondrial stress responses, such as mitochondrial chaperone/cochaperone DnaJ proteins which regulate proteostasis under stress (Verma et al., 2019; Zhou et al., 2012), are potential mediators of this communication. However, there is little research on their roles in orchestrating the crosstalk between ROS and anthocyanin biosynthesis.

DnaJ chaperone/cochaperone proteins constitute essential molecular regulators of plant stress responses, functioning in diverse cellular compartments (Chen et al., 2010; Jia et al., 2016; Park and Kim, 2014; Wang et al., 2015; Zhou et al., 2012; Zhu et al., 2015). Classical DnaJ proteins are classified into DnaJA, DnaJB, and DnaJC types based on structural features of the J-domain, G/F-domain, zinc finger domain (DnaJ_CXXCXGXG), and C-terminal domain (Chiu et al., 2013; Kampinga and Craig, 2010; Ohta and Takaiwa, 2014; Verma et al., 2017). Despite their established roles in proteostasis and stress adaptation, the involvement of DnaJs in regulating plant pigment metabolism, particularly anthocyanin biosynthesis, remains largely unexplored. To our knowledge, no DnaJ protein that is involved in the regulation of anthocyanin biosynthesis has been previously reported in any plant species. Current knowledge is primarily limited to the chloroplast-localized DnaJ protein ORANGE (OR), which stabilizes phytoene synthase and magnesium chelatase subunit Ⅰ to modulate photosynthetic pigments accumulation (Lu et al., 2006; Sun et al., 2023; Welsch et al., 2018; Zhou et al., 2015). Function of those organelle-localized (especially mitochondrion) DnaJ proteins in anthocyanin biosynthesis still remained undocumented. Critically, multiple abiotic and biotic stresses can elevate production of ROS (Altangerel et al., 2017; Catalá et al., 2011; Nakabayashi et al., 2014; Steyn et al., 2002), a known inducer of anthocyanin synthesis (Shi et al., 2018; Xu et al., 2017); while the DnaJ proteins are involved in diversified stress response pathways. This correlation might suggest some potential functional links between DnaJ-mediated stress responses and ROS-triggered anthocyanin biosynthesis. However, experimental evidence defining the role of specific DnaJ proteins, particularly in regulating ROS-triggered stress responses that coordinate the anthocyanin biosynthesis, is currently undiscovered.

In this study, we combined forward genetics and histochemical approaches in Medicago truncatula to identify novel regulators of ROS-anthocyanin homeostasis. We isolated a leaves with more ROS and anthocyanin1 (lma1) mutant which exhibits ROS hyperaccumulation and ectopic anthocyanin production, and characterized LMA1 as the causal gene encoding a mitochondria-localized DnaJA protein. The lma1 mutant displays disrupted mitochondrial ultrastructure, reduced mitochondrial complex Ⅰ activity, and elevated ROS levels. Transcriptomic and physiological analyses further revealed that ROS accumulation activates anthocyanin biosynthesis via up-regulating the anthocyanin biosynthesis-related genes in lma1. Collectively, our data first show that loss-of-function of the mitochondria-localized DnaJA protein LMA1 of M. truncatula is sufficient to trigger constitutive anthocyanin accumulation under non-stress growth conditions. This work elucidates that LMA1 serves as a pivotal integrator of mitochondrial function and anthocyanin metabolism, providing new insights into the function of DnaJ-mediated organelle homeostasis in stress adaptation.

2. Materials and methods 2.1. Plant materials and treatments

Ecotype R108 was used as the wild type (WT) for Medicago truncatula, and the NF6046 (lma1-1) mutant was isolated from the transposable element of Nicotiana tabacum (Tnt1)-tagged mutant collections (in R108 ecotype) as previously reported (Tadege et al., 2008). The original NF6046 (lma1-1) mutant was backcrossed with the WT for two generations, and the resulting BC2F2 progenies were used for phenotypic analyses. Seeds were subjected to scarification and low-temperature treatment (4 ℃ for five days) before sowing in sterilized soil. Plants were grown in a greenhouse under 24℃ day/20℃ night temperature, 16 h day/8 h night photoperiod, 50%–60% relative humidity, and about 150 μE m⁻² s⁻¹ light intensity conditions. For H₂O₂ and the DMTU treatment, surface-sterilized germinated seeds of WT and mutant were first grown on the 1/2 MS solid medium for one week, the seedlings were then transferred to another 1/2 MS solid medium supplemented with 4 mM H₂O₂ or 3 mM DMTU for two additional weeks under the aforementioned greenhouse condition.

2.2. Anthocyanin extraction and measurement

Approximately 0.1 g leaves from three-week-old WT, lma1-1, and CR-lma1 plants were ground in liquid nitrogen and soaked in an equivalent volume of extraction solution (methanol containing 0.1% HCl) and kept in the dark overnight at 4℃. After vortex and centrifugation at 12000 g for 5 min, the residues were re-extracted until the supernatants were colorless. The supernatants were pooled as crude extracts for further analysis. For the rapid determination of anthocyanins in leaves, a spectrophotometer (JENWAY 7315 UV-VIS) was used to measure the absorbance of each sample at 530 nm, 620 nm, and 650 nm. The following formula was used to calculate the relative anthocyanin content: (A530–0.25*A657)/g.

For reverse-phase High Performance Liquid Chromatography (HPLC) analysis, the equivalent volume hexane was added to the crude extracts to remove the chlorophyll. After centrifugation, the aqueous supernatant in methanol was then filtered for HPLC separation. The HPLC separation was performed on an Agilent LC1260 infinity Ⅱ HPLC system with Agilent 5 TC-C18 (2) 250 × 4.6 mm column and elution with solution A (0.2% phosphoric acid) and solution B (acetonitrile) at 1 mL/min flow rate. The elution gradient was as follows: 0–2 min, 5%–10% solution B; 2–10 min, 10%–18% solution B; 10–14 min, 18%–20% solution B; 14–18 min, 20%–22% solution B; 18–22 min, 22%–40% solution B; 22–24 min, 40%–100% solution B. The absorbance was detected at 530 nm.

2.3. Gene cloning and gene-editing of LMA1

Equal amounts of leaf tissue were collected from five independent lma1-1 plants (with intense DAB brown or NBT blue precipitates, pink pigmentation on abaxial leaf surface, and reduced leaf size) in the BC2F2 population to prepare a pooled sample. The genomic DNA of this mixture was then extracted by a Plant Genomic DNA Kit (Tiangen, Beijing, China). A whole genome resequencing at 50x coverage of the R108 genome using this DNA mixture was performed to identify all the Tnt1 insertions in lma1-1 as previously reported (Jiang et al., 2015).

Two targets for CRISPR/Cas9-mediated editing of LMA1 were designed via the CRISPR-P 2.0 tool (http://cbi.hzau.edu.cn/CRISPR2/) (Table S1) (Liu et al., 2017) for generating more lma1 alleles. Guide RNAs (gRNAs) were amplified and cloned into the modified pYLCRISPR/Cas9P_35S-B binary vector using the Golden Gate method (Ma et al., 2015). In our modified vector, the promoter of soybean GmUbi3, CaMV35S promoter, and promoters of small nuclear (sn) RNA U6/U3 of M. truncatula are used to drive the expression of hSpCas9 (Feng et al., 2013), the selectable marker gene BAR, and the designed sgRNAs, respectively (Meng et al., 2017). The resulting pYLCRISPR/Cas9P35S-B-LMA1 construct was used for Agrobacterium (EHA105 strain)-mediated transformation of WT (Cosson et al., 2015). Genomic DNA was isolated from young leaves of the T0 and T1 transgenic plants for PCR amplification, a pair of primers was used to amplify the flanking region of the target site in LMA1 gene (Table S1). The resulting PCR products were cloned into pEASY-Blunt cloning vector (TransGen Biotech, CB101) for Sanger sequencing to validate mutations. The homozygous T1 plants of CR-lma1 which showed the same phenotype as lma1-1 were chosen to perform phenotype analyses.

2.4. Phylogenetic and syntenic analysis

A total of 184 newly released high-quality genomes were collected in this study (Table S2). The Hidden Markov Model (HMM) files of DnaJ domain (J-domain, PF00226), DnaJ central domain (zinc finger domain, DnaJ_CXXCXGXG, PF00684), and DnaJ C-terminal domain (PF01556) were acquired from the InterPro database (https://www.ebi.ac.uk/interpro) and used to query the protein files via HMMER (v3.0, http://hmmer.org/). Subject proteins with all these three domains were considered as DnaJA candidates. These candidates were uploaded to the InterPro and NCBI CDD tool (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi) for further confirmation of these domains. Those candidates without the HPD signature in their J-domain were excluded. After the multiple sequence alignment analyses using MAFFT (Katoh and Standley, 2013), the alignment result was trimmed with trimAl (-gt 0.05) (Capella-Gutierrez et al., 2009) and then used for maximum-likelihood tree construction (1000 SH-aLRT replicates/1000 ultrafast bootstrap replicates) using the Q. plant + I + R10 substitution model by the IQ-TREE2 software (Lanfear et al., 2020). Based on the phylogenetic tree, J-domain sequence of all the members from the three main branches were used to generate a seqlogo plot via the TBtools (Chen et al., 2023). The all-versus-all BLASTP results (e-value < 1e-5, score > 100) between Amborella trichopoda, Aristolochia fimbriata, Vitis vinifera, Cercis canadensis, and Medicago truncatula were used for the WGDI toolkit to detect inter-genomic syntenic blocks (Sun et al., 2022; Yang, 2007), and then TBtools were also used to visualize the micro-collinearity of LMA1.

2.5. Subcellular localization

The full-length coding sequence of LMA1 was amplified and inserted into the NcoI restriction site of the pCAMBIA3301-MP vector to generate the 35S::LMA1-GFP construct (Table S1). Two micrograms of 35S::LMA1-GFP plasmid DNA were transformed into Nicotiana benthamiana protoplasts using the previously described method (Yoo et al., 2007). These transformed protoplast cells were gently resuspended with pre-warmed (37℃) Washing and Incubation (WI) solution (4 mM 4-morpholineethanesulfonic acid/MES, pH 5.7, 0.5 M mannitol, and 20 mM KCl) containing 25 nM MitoTracker® Red CM-H₂XRos (40740ES50, YEASEN). After 45 min of incubation, protoplasts were collected by centrifugation and resuspended with 37℃ pre-warmed WI solution, followed by observation under a confocal microscope (Zeiss, LSM900).

For other DnaJA proteins, we adopted nine widely used and high-accuracy methods including DeepLoc 2.1 (Ødum et al., 2024), TargetP 2.0 (Almagro Armenteros et al., 2019), Plant-mSubP (Buckley et al., 2020), LOCALIZER (Sperschneider et al., 2017), TPpred3 (Savojardo et al., 2015), MultiLoc2 (Blum et al., 2009), PredSL (Petsalaki et al., 2006), CELLO v.2.5 (Yu et al., 2006), and Predotar (Small et al., 2004) to computationally predict their subcellular localization. Only the highest-scoring subcellular compartment was considered as the localization for each DnaJA if multiple localizations were predicted.

2.6. RNA extraction and real-time quantitative PCR (qRT-PCR)

Total RNA was isolated from leaves with RNaEx^TM Trizol (GK3005, GENEray), subsequently, 3 μg of total RNA were used for reverse transcription with the Hiscript® Ⅱ 1st Strand cDNA Biosynthesis Kit (R202, Vazyme). qRT-PCR was conducted using the QuantFast SYBR Green qPCR SuperMix (M2211, MAGIC-BIO) on a Light-Cycler 480Ⅱ device (Roche). The MtACTIN gene was used as an internal control for Medicago truncatula (Table S1). For semi-quantitative RT-PCR, 32 cycles were carried out to detect the LMA1 transcript.

2.7. RNA-seq analysis

A total of four mature leaf blades or eight young leaves samples (containing the shoot apex and immature leaves) were collected from three-week-old WT and lma1-1 mutant plants, and each with three biological replicates. Total RNA was extracted using the RNAsimple Total RNA Kit (DP419, TIANGEN) according to the manufacturer's instructions, and RNA quality was assessed using an Agilent 2100 Bioanalyzer. The cDNA libraries were constructed using NEBNext® Ultra^TM Ⅱ RNA Library Prep Kit (E6310, NEB) and then sequenced using Illumina Nova PE150 by the Shanghai Ruixing Bio-Tech (Shanghai, China). Raw sequencing reads were processed with Trimmomatic (Bolger et al., 2014) to remove adapter contaminants and trim low-quality bases, followed by quality verification using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) to ensure read integrity. Clean reads were obtained after filtering and mapped to the reference genome of Medicago truncatula (https://medicago.toulouse.inra.fr/MtrunA17r5.0-ANR/) using HISAT2 software (Kim et al., 2019). The featureCounts software (Liao et al., 2014) was used to assemble and count the mapped reads, and then estimate the transcript abundances for each gene as fragments per kilobase of transcript per million mapped reads (FPKM) using the DESeq2 software (Love et al., 2014). Genes with FDR corrected p-values (adjusted p-value, i.e., p. adjust) < 0.05 and an absolute fold-change value ≥ 2 were considered as differentially expressed genes (DEGs). The TBtools (Chen et al., 2023) was used for Gene Ontology (GO) analysis and enrichment of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways of the DEGs and then plotted with GraphPad Prism 9.4.1. The expression levels of selected DEGs which involves in leaf anthocyanin biosynthesis were validated via qRT-PCR. All primers used in this study were listed in Table S1.

2.8. Transmission electron microscope (TEM)

WT and lma1-1 mutant leaves were cut into 2 mm² and subjected to vacuum infiltration in 2.5% glutaraldehyde solution, and then kept at room temperature overnight. The samples were washed three times for 15 min each with 0.1 M phosphate buffer before fixing them at 4℃ in 1% osmium acid for 4 h. After washing three times with 0.1 M phosphate buffer solution, ethanol gradient dehydration was then carried out and followed by replacing with acetone, soaking in Spurr resin, embedding and curing the embedded blocks. The ultrathin sections (70 nm thick) were prepared using a Leica UC7 ultramicrotome and stained with 2% uranyl acetate-lead citrate for 30 min. These specimens were then examined under a transmission electron microscope (JEM-1400PLUS).

2.9. H₂O₂ and superoxide anion contents and mitochondrial complex Ⅰ activity

The first fully unfolded compound leaf of three-week-old WT and lma1-1 mutant were collected. For visualization of H₂O₂ and superoxide anions (O₂^•-), leaves were stained with 1 × 3, 3-diaminobenzidine (DAB) solution (DA1010, DAB substrate kit, Solarbio) and 1 mg/mL nitroblue tetrazolium (NBT) (in 10 mM potassium phosphate buffer, pH 7.8) at room temperature, respectively. After staining, the leaves were washed and bleached with 95% ethanol in a water bath at 80℃ for decolorization until all chlorophyll completed faded. Quantification of H₂O₂ and O₂^•- in WT and lma1-1 mutant was conducted using Hydrogen Peroxide (H₂O₂) Content Assay Kit (BC3590, Solarbio) and Superoxide Anion Activity Content Assay Kit (BC1290, Solarbio), respectively, on a JENWAY 7315 UV-VIS Spectrophotometer. The mitochondrial complex Ⅰ activity was measured with a Mitochondrial complex Ⅰ/NADH-CoQ reductase Activity Assay Kit (BC0515, Solarbio) following the manufacturer's instructions.

2.10. Statistical analysis

Plant growth and development parameters including plant height, leaf number, petiolule length, rachis length, leaf area, pod size, and seed size were measured. All quantitative data are presented as mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism (v.9.4.1) with unpaired two-sample t-tests for comparisons between groups. Statistically significant differences were indicated by asterisks: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

3. Results 3.1. lma1 mutant of Medicago truncatula exhibited ROS-rich leaves

Plants accumulate ROS to cope with multiple environmental stresses. To investigate the intrinsic factors regulating ROS accumulation in legumes, we screened a tobacco retrotransposon Tnt1-insertion population (approximately 300 lines) of M. truncatula for mutants with elevated ROS levels, via the diaminobenzidine (DAB) or nitroblue tetrazolium (NBT) staining assay. Leaves of certain mutant line stained with brownish deposits (DAB) or dark blue precipitates (NBT) were identified as candidates. One mutant, designated as leaves with more ROS and anthocyanin1 (lma1-1, NF6046) (Fig. 1A and B) displayed a higher level of ROS in leaves (Fig. 1C-H). In a comparison of the hydrogen peroxide (H₂O₂) level by using DAB staining, the lma1-1 leaves displayed speckled dark brown, while the WT were only slightly stained with brown polymerization (Fig. 1C and D). Quantitative analysis confirmed elevated H₂O₂ levels in lma1-1 compared to WT (Fig. 1E). Detection of superoxide (O₂^•-) production via NBT staining also revealed that the lma1-1 leaves accumulated more blue formazan deposits than those in the WT (Fig. 1F and G), in agreement with the quantitative results of O₂^•- (Fig. 1H). The lma1-1 was then backcrossed with WT for two generations to purify its genetic background before subsequent analyses. In the BC2F2 population, the progenies exhibiting the WT-like (with slight DAB brown or NBT blue precipitates, uniform green abaxial leaf, and normal leaf size as WT) and lma1-1 (with intense DAB brown or NBT blue precipitates, pink pigmentation on the abaxial surface of leaf, and reduced leaf size) phenotype presented a segregation ratio of 3:1 (43 WT and 13 mutants; χ² = 0.049 < χ_0.05² = 3.84), indicating mutation of a single recessive gene is responsible for the phenotype. These results demonstrated the successful isolation of a ROS-overaccumulation lma1-1 mutant in M. truncatula.

3.2. Molecular cloning and characterization of LMA1 in Medicago truncatula

To clone the LMA1 gene, we obtained an equal amount of young leaves from backcrossed lma1-1 mutants (from the BC2F2 population) to prepare a pooled genomic DNA sample for a whole genome resequencing (WGS)-based approach to recover the Tnt1 integration sites. Using the Identification of Transposon Insertion Sites (ITIS) algorithm (Jiang et al., 2015), we identified 63 Tnt1 insertions in the lma1-1 mutant sample, including eight putative homozygous insertions (Table S3 and Fig. S1A). Utilizing polymerase chain reaction (PCR)-based genotyping and Sanger sequencing of the mutant individuals from the BC2F2 population, we confirmed that only one homozygous Tnt1 insertion in the third exon of MtrunA17_Chr5g0426951 is co-segregated with the lma1-1 phenotype (Figs. 2A and S1). The remaining seven insertions are also homozygous in some WT-like plants in the BC2F2 population, and therefore they do not contribute to the lma1-1 phenotype. The transcript of MtrunA17_Chr5g0426951 was barely detected in lma1-1 via both semi-quantitative RT-PCR and quantitative real-time PCR (qRT-PCR) assays (Figs. 2B and C). These results unveil that MtrunA17_Chr5g0426951 might be the candidate gene of LMA1. Amino acid sequence analysis indicates that the MtrunA17_Chr5g0426951 encodes a typical DnaJA protein, with full zinc finger motif and C-terminal structure in addition to the core J-domain (DnaJ domain) (Figs. 2D and S2).

To further confirm the corresponding gene in lma1-1, we then generated more loss-of-function alleles of MtrunA17_Chr5g0426951 using a clustered regularly interspaced short palindromic repeats/CRISPR-associate protein9 (CRISPR/Cas9) gene editing system (Ma et al., 2015). We designed two gene specific guide RNAs (sgRNAs) targeting the exons of MtrunA17_Chr5g0426951 for Agrobacterium-mediated transformation of WT (Fig. 2A) (Cosson et al., 2015), and 29 positive edits were acquired among the 44 transgenic lines (T0). Three of the fifteen homozygous T1 lines (all showing lma1-1-like phenotype) were chosen for subsequent analysis. Sanger sequencing had verified that the CR-lma1-1 and CR-lma1-2 have 1 bp insertion and 2 bp deletion in target 2, respectively, while CR-lma1-3 showed a 46 bp and 9 bp deletion in target 1 and target 2, respectively (Fig. 2E). The insertion and/or deletion in these three additional alleles cause premature termination of LMA1 translation and produced the truncated LMA1 protein, which disrupted its zinc finger domain and/or C-terminal domain (Figs. 2D, E and S3). After the DAB and NBT staining, leaves of these three independent CRISPR/Cas9-generated mutant alleles exhibited visible and high-level accumulation of ROS (Fig. 2F and H), and the quantitative analyses also revealed the significantly increased content of H₂O₂ and O₂^•- in these lines (Fig. 2G and I), which are similar to that of the lma1-1. These results confirm that loss of MtrunA17_Chr5g0426951 function causes the lma1 mutant phenotype.

3.3. LMA1 encodes a mitochondria-localized DnaJA protein

Based on the signature of J-domain, 96 typical DnaJ proteins including nine DnaJA members were identified in Medicago truncatula (Fig. S4). The LMA1 is one of the DnaJA proteins with full zinc finger domain and C-terminal domain in addition to the core J-domain. It seems these nine DnaJA proteins could be grouped into three subtypes via the phylogenetic analysis (Fig. S4). We then chose more representative species from major lineage of plants for further study. The phylogenetic analysis, representing 150 families of 89 orders, clearly revealed that the DnaJA is widespread within the plantae, and there are three main well-supported subtypes of DnaJA proteins which were defined as DnaJA Ⅰ, Ⅱ, and Ⅲ (Figs. 3A and S5). Notably, the J-domain of these three subtypes displayed visibly different patterns (Fig. 3B). DnaJA Ⅰ, to which the LMA1 belongs, seems to possess the original J-domain structure since the other two subtypes of DnaJA lacking certain amino acids to varying degrees (Fig. 3B), further supporting the phylogenetic tree. We then performed a syntenic analysis and indeed revealed that the LMA1 is located in a conserved syntenic block with its putative orthologs from Cercis canadensis, Vitis vinifera, Aristolochia fimbriata, and the basalmost angiosperms Amborella trichopoda (Fig. 3C). These results suggested that the LMA1 and its putative orthologs represent a group of evolutionarily conserved DnaJA.

To explore the subcellular localization of LMA1, protoplasts of Nicotiana benthamiana were transiently transformed with a pro35S::LMA1-GFP construct. The green fluorescent puncta near the cell membrane which appear to localize in the mitochondria were clearly observed under a confocal microscopy. We then stained those transformed protoplasts with the mitochondria-specific dye, MitoTracker Red, and it was found the fluorescence of the LMA1-GFP fusion protein co-localized with the MitoTracker Red-stained mitochondria (Fig. 3D), thus confirming the mitochondrial localization of LMA1. Interestingly, the vast majority of members from the DnaJA Ⅰ subtype were computationally predicted to localize in mitochondria like LMA1 (Fig. 3E), which suggests that the subtype DnaJA Ⅰ including the LMA1 is a group of conserved mitochondria-localized DnaJ proteins. To determine the expression profile of LMA1, qRT-PCR analysis was performed using different tissues in both in vegetative and reproductive stages. Our qRT-PCR analysis showed that LMA1 is ubiquitously expressed in all plant tissues checked, with higher expression level in the leaf, shoot, and flower (Fig. 3F). These findings together revealed that the LMA1 is widely expressed and encodes a conserved mitochondria-localized DnaJA protein.

3.4. LMA1 negatively regulates anthocyanin biosynthesis

It is particularly noteworthy that the abaxial surface of lma1-1 leaves exhibit distinct pink pigmentation, in contrast to the uniform green phenotype in WT (Fig. 4A and B), while no obvious color change occurs in other organs of lma1-1. Corroborating with deeper pigmentation in crude leaf extracts, quantification revealed a nine-fold increase in total anthocyanin content in leaves of lma1-1 versus WT (Fig. 4C and D). We then utilized a reverse-phase High Performance Liquid Chromatography (HPLC) to further verify the anthocyanins accumulation in lma1-1. HPLC analysis further confirmed elevated anthocyanin accumulation in lma1-1, manifested by significantly increased peak intensities at anthocyanin-specific wavelengths (530 nm) compared to WT (Fig. 4E). Moreover, the three independent CRISPR/Cas9-generated LMA1 knockout lines (CR-lma1-1, -2, -3) recapitulated the phenotype of lma1-1, all of which displayed similar leaf anthocyanin hyperaccumulation and comparable anthocyanin content elevation (Fig. 4F and G). Collectively, these results indicated that LMA1 functions as a repressor in leaf anthocyanin biosynthesis.

3.5. Transcriptomic profiling reveals LMA1 co-regulates mitochondrial dysfunction and anthocyanin metabolism

To dissect the molecular basis of LMA1-mediated ROS and anthocyanin regulation, we conducted a bulk RNA sequencing (RNA-seq) on young and mature leaves from three-week-old WT and lma1-1 plants. Following rigorous quality assessment (Fig. S6), we identified 354 consistently up-regulated and 231 down-regulated differentially expressed genes (DEGs; adjusted P-value <0.05) shared between young and mature lma1 leaves (Fig. 5A; Tables S4 and S5). The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis uncovered two major metabolic disruptions in lma1-1: the mitochondrial dysfunction, evidenced by DEGs related to mitochondrial biogenesis, chaperones and folding catalysts, and amino acid metabolism; and the secondary metabolic shift, prominently featured by the phenylpropanoid metabolism and flavonoid biosynthesis (Fig. 5B). In consistent with the KEGG results, the Gene Ontology (GO) term analysis also highlighted alterations in ATP-dependent processes (linked to mitochondrial dysfunction) and flavonoid metabolic pathways (Fig. 5C). Notably, terms associated with oxidative stress response and oxidoreductase activity implicated the mitochondrial-derived ROS accumulation; while flavonoid metabolic rewiring, particularly the upregulation of anthocyanin-related genes, might suggest compensatory mechanisms against redox imbalance (Fig. 5C). Strikingly, nearly the entire anthocyanin biosynthesis pathway was transcriptionally activated in lma1-1, including genes encoding biosynthetic enzymes (PAL, C4H, 4CL, CHS, F3H, F3′H, DFR, and ANS), a modification enzyme (UGT78G1), and a transporter (GSTF7) (Fig. 5D). To validate these RNA-seq results, we carried out a qRT-PCR analysis in both lma1-1 and CR-lma1 mutants, which confirmed significant upregulation of anthocyanin-related genes (Figs. 5E and S7). Together, these results demonstrate that LMA1 dysfunction triggers extensive transcriptional changes, involving in the mitochondrial perturbations and anthocyanin hyperaccumulation which represent two interconnected facets of its regulatory role.

3.6. Mitochondrial dysfunction-induced ROS promoted anthocyanin biosynthesis in lma1

The RNA-seq data revealed that numerous DEGs are enriched in mitochondrial biogenesis, ATP-dependent activity, oxidoreductase activity, and responses to reactive oxygen species (ROS) (Fig. 5B and C). Thus, we were wondering if regulation of anthocyanins biosynthesis by mitochondria-localized LMA1 is associated with mitochondrial function and its cellular oxidative status. We first performed an ultrastructural analysis of the mitochondrial morphology in lma1-1 using transmission electron microscopy (TEM). In WT plants, mitochondria displayed well-organized cristae structures, characterized by densely packed and parallel invagination of the inner mitochondrial membrane (Fig. 6A). In contrast, mitochondria of lma1-1 exhibited severely disorganized cristae, with sparse and fragmented membrane infoldings (Fig. 6B), suggesting the integrity of its mitochondria is aberrant. Quantitative analysis confirmed this structural defect that 37% of mitochondria in lma1-1 exhibited abnormal cristae morphology, higher than 11% of that in WT (Fig. 6C). This finding suggests that LMA1 plays a critical role in maintaining mitochondrial architecture. We then evaluated the functional impact of this structure defect on mitochondria of lma1-1, and it showed that the mitochondrial complex Ⅰ activity in lma1-1 was only 37% of that in WT (Fig. 6D), indicating loss of LMA1 function catastrophically disrupts the mitochondrial ultrastructure and bioenergetic capacity. Considering that the mitochondrial complex Ⅰ is the major source of ROS production and its dysfunctions have been associated with ROS accumulation (Lenaz, 2001; Liu et al., 2002), the elevated ROS levels in lma1 likely stem from impaired mitochondria. These findings demonstrate that LMA1 is essential for maintaining mitochondrial structural integrity and redox homeostasis, while loss of LMA1 function leads to ROS overproduction via compromised electron transport chain (ETC) function.

To test whether anthocyanin accumulation in lma1-1 depends on the ROS overproduction, we examined the effects of exogenously applied ROS (H₂O₂) on anthocyanin biosynthesis in lma1-1. After H₂O₂ treatment, the total anthocyanin content was highly increased in both lma1-1 and WT (Fig. 6E and H). In contrast, when adopting the ROS scavenger N, N′-dimethylthiourea (DMTU, a highly permeant molecule for effectively scavenging H₂O₂ and ·OH) to decrease the ROS level in lma1-1 mutant, the total anthocyanin content was dramatically reduced following the DMTU treatment (Fig. 6E–G). Correspondingly, the expression of anthocyanin biosynthetic genes including the MtDFR and MtANS were significantly upregulated under the ROS treatment while inhibited after the application of DMTU (Fig. 6I and J). These findings uncover the mitochondrial dysfunction-induced ROS overproduction causing the anthocyanin accumulation in lma1.

4. Discussions

ROS serve as a pleiotropic factor in plant cells, acting not only as a byproduct of oxidative stress but also as signaling molecules regulating anthocyanin biosynthesis (Zhang et al., 2024). Increased ROS can trigger anthocyanin accumulation by promoting late biosynthetic and regulatory genes, while ROS scavengers like DMTU attenuate this response (Altangerel et al., 2017; Shi et al., 2018; Xu et al., 2017; Zhang et al., 2024). Consistent with these findings, anthocyanin accumulation in lma1-1 was induced by H₂O₂ and alleviated by DMTU (Fig. 6E-H), confirming that ROS signaling links DnaJ dysfunction to anthocyanin biosynthesis. Stress-induced metabolic dysregulation could elevate the ROS levels (Mittler et al., 2022), while anthocyanins act as antioxidants to mitigate this oxidative damage (Bi et al., 2014; Gould et al., 2002; Jiang et al., 2023; Landi et al., 2015; Nakabayashi et al., 2014; Neill et al., 2002; Steyn et al., 2002). Moreover, DnaJ proteins could participate in antioxidant metabolic pathways and adjust their own expression and activity in response to intracellular ROS levels, thereby protecting cells from oxidative damage (Jin et al., 2024; Kampinga and Craig, 2010; Zhang et al., 2023). However, we found that the LMA1 is mildly induced by H₂O₂ treatment while it maintains the normal transcription level after DMTU treatment (Fig. S8). This slight upregulation implies that LMA1 might primarily function as a ROS gatekeeper and be orchestrated through a certain kind of protein-level regulation rather than transcriptional induction, contrasting with the strong transcriptional suppression observed in those anthocyanin-related genes under DMTU treatment (Fig. 6).

In this study, we found lacking functional LMA1 gene had compromised the oxidative homeostasis of Medicago truncatula (Fig. 1C–H), possibly resembling a stress-induced ROS accumulation. In response, anthocyanin synthesis was upregulated in the lma1 mutant (Fig. 4) to mitigate excessive ROS production, thereby balancing the cellular oxidative homeostasis. These findings imply that LMA1 may play a role in both stress response and anthocyanin biosynthesis. It is imperative to elucidate the mechanisms by which LMA1 modulates ROS and anthocyanin homeostasis under different stress conditions. Under stress conditions, ROS act as key signaling molecules that promote anthocyanin accumulation, a process in which abscisic acid (ABA) plays an integral role (Li et al., 2019; Previtali et al., 2021; Song et al., 2024; Xu et al., 2017; Zhang et al., 2024). ABA signaling recruits multiple transcription factors like ABA-INSENSITIVE 5 (ABI5) and ZINC FINGER OF ARABIDOPSIS THALIANA 10 (ZAT10) to coordinately enhance anthocyanin biosynthesis. The ABI5 directly activates genes like CHS and interacts with the MBW complex, while the ZAT10 can also reinforce the activity of MBW complex (Raghavendra et al., 2010; Song et al., 2024; Zhou et al., 2009). Our transcriptomic data also revealed a significant enrichment of those ABA-responsive genes and ABA signaling pathways in the lma1-1 mutant (Tables S4 and S5), suggesting the involvement of this phytohormone in LMA1-mediated anthocyanin accumulation; however, the underlying mechanisms require further investigation.

Phylogenetic analysis identified three DnaJA subtypes, with mitochondrial-localized DnaJA Ⅰ proteins possibly exhibiting the highest evolutionary conservation across plantae (Figs. 3A, B and S5), suggesting their indispensable roles in mitochondria. For instance, DnaJA Ⅰ members Mdj1p of yeast, AtJ1/AtDjB1 of Arabidopsis, and OsDjA9 of rice have been functionally characterized to maintain mitochondrial proteostasis and respiration (Duchniewicz et al., 1999; Jia et al., 2016; Kampinga and Craig, 2010; Rowley et al., 1994; Westermann et al., 1996; Xu et al., 2020; Zhou et al., 2012). As the first identified mitochondria-localized DnaJA protein in M. truncatula, LMA1 affects mitochondrial morphology and respiration (Fig. 6A–D), as supported by RNA-seq data which links it to mitochondrial biogenesis and ATP-dependent processes (Fig. 5B and C). Both mitochondrial DnaJA Ⅰ protein LMA1 and the AtJ1 affect mitochondrial structural integrity and ROS homeostasis, but no published evidence demonstrates that AtJ1 can regulate anthocyanin biosynthesis (Jia et al., 2016; Zhou et al., 2012). LMA1 might have acquired a lineage-specific role in suppression of leaf anthocyanin accumulation, besides its explicit function in controlling growth and development, as evidenced by the reduced fresh weight and organ size in lma1 (Figs. 1A and B, 2F and H, S9). These growth defects in lma1 likely stem from the primary effect of LMA1 in mitochondrial respiration, which directly impacts energy metabolism. In contrast, ROS hyperactivation in lma1 acts as the key inducer of anthocyanin biosynthesis, a stress-response mechanism particularly pronounced in legumes like M. truncatula. This functional divergence between LMA1 and AtJ1 is plausibly due to the constitutively active flavonoid biosynthesis in leaves of legumes, considering that the M. truncatula possess a more complex assortment of various flavonoids than that in Arabidopsis (Dixon and Sumner, 2003; Gholami et al., 2014; May and Dixon, 2004).

Compared to prokaryotes and other eukaryotes, plants possess expanded and diversified DnaJ proteins (Georgopoulos et al., 1980; Hageman and Kampinga, 2009; Miernyk, 2001; Rowley et al., 1994; Sahi and Craig, 2007; Sarkar et al., 2013). Critically, the subcellular compartmentalization of DnaJ proteins might dictate their functional specificity in regulating distinct plant pigment pathways. For instance, plastid-localized OR (DnaJE proteins) regulates chlorophyll and carotenoid levels (Sun et al., 2019, 2023; Welsch et al., 2018; Yuan et al., 2015; Zhou et al., 2015); the tomato chloroplast-targeted DnaJs (DnaJC proteins) stabilize photosystem Ⅰ, Ⅱ, and Rubisco activity to maintain chlorophyll synthesis (Cai et al., 2022; Kong et al., 2014a, 2014b; Wang et al., 2015), and the Arabidopsis chloroplast-localized DnaJA members are essential for chlorophyll synthesis (Zhang et al., 2021). However, the mitochondria-anchored DnaJA protein LMA1 may modulate anthocyanin biosynthesis via a distinct compartment-decoupled mechanism. Its mitochondrial localization (Fig. 3D) spatially segregates it from nuclear/cytosolic anthocyanin-related transcription factors and those ER-localized anthocyanin biosynthetic enzymes. Instead, LMA1 maintains the mitochondrial integrity and ROS balance to influence the biosynthesis of anthocyanins (Fig. 6). We propose that the mitochondrial-derived ROS in lma1 (due to its mitochondrial dysfunction) can serve as the primary trigger to induce the mitochondrial retrograde signaling, which eventually activates the anthocyanin biosynthesis (Khan et al., 2024; Mittler et al., 2022). Previous studies have validated that ROS can activate cytosolic signaling pathways to modulate the activity of transcription factors (Fu and Dong, 2013; Lee et al., 2021; Zhang et al., 2024; Zhu, 2016); and a typical precedent is the ROS-mediated inhibition of PP2A phosphatase, which can enhance the expression of MYB transcription factors for anthocyanin biosynthesis under high light stress (Zhang et al., 2024). In addition, the ROS could directly induce nuclear translocation or DNA-binding activity of redox-sensitive transcription factors such as heat shock factors (HSFs) and NAC for stress response (Albertos et al., 2021; Giesguth et al., 2015; Meng et al., 2019). We hypothesize that mitochondrial-derived ROS in lma1 may adopt the similar mechanisms, potentially targeting at some specific transcription factors that regulate anthocyanin-related structure genes or the core MBW complex. Thus, elucidating the precise components which the LMA1 might affect in the retrograde signaling pathway represents a key objective in future research.

Classical DnaJA proteins could function as co-chaperones by interacting with 70-kD heat shock proteins (HSP70s), stimulating their ATP hydrolysis to maintain protein homeostasis (Kampinga and Craig, 2010). Loss-of-function mutations in Arabidopsis mitochondrial HSP70s (mtHSC70s) cause abnormal mitochondrial morphology, excessive ROS accumulation, and severe growth defects (Li et al., 2021; Wei et al., 2019), which are phenotypes relevant to our study (Figs. 1, 2F–I, 4F, 6A–D and S9). Considering that there are approximately 60 HSP70 members in M. truncatula, with at least two of them predicted to reside in the mitochondria (Table S6). It is notable that our RNA-seq data also unveiled the DEGs between lma1 and WT are enriched in chaperones and folding catalysts (Fig. 5B). Therefore, it deserves intensive investigation whether one or more mtHSC70s recruit LMA1 for their function. In future research, it might be a good option to identify those interacting HSP70s using yeast two-hybrid or bimolecular fluorescence complementation assays. The zinc finger and C-terminal domains of DnaJA proteins are essential for specific interactions with HSP70 partners and substrate delivery (Goffin and Georgopoulos, 1998; Li et al., 2003; Lu and Cyr, 1998). Our CRISPR-Cas9-generated CR-lma1 mutants possess disrupted zinc finger domain and/or C-terminal domain, resulting in identical phenotypes to lma1-1 (Fig. 2D; S3B). These results not only confirmed the loss-of-function phenotype of LMA1, but more importantly demonstrated the necessity of the zinc finger domain and C-terminal domain for the proper function of LMA1. Given the previous reports and this study, it is reasonable to presume that LMA1 needs more specific mitochondrial-localized substrates besides HSP70s for its function. Searching for additional partners interacting with zinc finger domain and/or C-terminal domain of LMA1 might be a good way to comprehend its mechanisms in mitochondrial integrity and anthocyanin metabolism.

5. Conclusions

In summary, this study identified the mitochondrial DnaJA protein LMA1 as a pivotal integrator of mitochondrial function and anthocyanin metabolism in Medicago truncatula. Loss of LMA1 function disrupts mitochondrial ultrastructure and impairs complex Ⅰ activity, leading to ROS overaccumulation that transcriptionally activates anthocyanin biosynthesis (Fig. 7). These results establish that LMA1 maintains mitochondrial integrity to control ROS levels, thereby regulating anthocyanin production, which provides new insight into plant stress adaptation by elucidating how the DnaJA proteins link the organelle integrity to protective pigment biosynthesis.

Acknowledgement

This research was supported by National Natural Science Foundation of China (32170360, 32470334, 32200681, 32570980, and 32200290), Strategic Priority Research Programs of the Chinese Academy of Sciences (XDA26030301), Yunnan Revitalization Talent Support Program (Yunling Scholar Project to J.C., XDYC-QNRC-2022-0179, and XDYC-QNRC-2022-0335), the Youth Innovation Promotion Association CAS (2021395), and the Caiyun Postdoctoral Funding (to S.Z.). We thank Yingqi Guo (Kunming Institute of Zoology, CAS) for assistance with TEM, Prof. Changning Liu’ lab and the Institutional Center for Shared Technologies and Facilities of Xishuangbanna Tropical Botanical Garden, CAS for providing the local Linux server.

CRediT authorship contribution statement

Wu Qing: Data curation, Investigation, Methodology, Validation, Visualization, Writing – original draft. Wang Ruoruo: Data curation, Methodology, Writing – review & editing. Mao Yawen: Formal analysis, Methodology, Software. Yang Liling: Data curation, Funding acquisition, Resources, Writing – review & editing. Zhao Weiyue: Funding acquisition, Resources. He Liangliang: Data curation, Funding acquisition, Project administration, Writing – review & editing. Zhou Shaoli: Funding acquisition, Investigation. Luo Jia: Methodology, Resources. Zhang Hailong: Investigation, Resources. Feng Hanyan: Validation. Fang Yuqi: Investigation. Liu Mingli: Investigation. Liu Yu: Writing – review & editing. Chen Jianghua: Conceptualization, Funding acquisition, Supervision, Writing – review & editing. Zhao Baolin: Conceptualization, Funding acquisition, Data curation, Project administration, Resources, Supervision, Visualization, Writing – review & editing.

Data availability

Information of all Medicago truncatula genes involved in this study can be found at the M. truncatula genome portal (https://medicago.toulouse.inra.fr/MtrunA17r5.0-ANR), the related accession numbers are given in Table S7. The raw data of the WGS and RNA-seq were deposited in the Genome Sequence Archive (http://bigd.big.ac.cn/gsa) under CRA024207 and CRA024208 (under PRJCA037854), respectively.

Declaration of competing interest

The author Jianghua Chen is an Associate Editor In Chief for Plant Diversity and was not involved in the editorial review or the decision to publish this article. All other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2025.10.004.