Abstract

Background

   The Flavonoid 3’-hydroxylase gene(F3’H) is an important structural gene
   in the anthocyanin synthesis pathway of plants, which has been proven
   to be involved in the color formation of organs such as leaves,
   flowers, and fruits in many plants. However, the mechanism and function
   in barley are still unclear.

Results

   In order to explore the molecular mechanism of the grain color
   formation of purple qingke, we used the cultivated qingke variety
   Nierumzha (purple grain) and the selected qingke variety Kunlun 10
   (white grain) to conduct transcriptomic sequencing at the early milk,
   late milk and soft dough stage. Weighted Gene Co-expression Network
   Analysis (WGCNA) was used to construct weighted gene co-expression
   network related to grain color formation, and three key modules (brown,
   yellow, and turquoise modules) related to purple grain of qingke were
   selected. F3’H (HORVU1Hr1G094880) was selected from the hub gene of the
   module for the yeast library, yeast two-hybrid (Y2H), subcellular
   localization and other studies. It was found that in purple qingke,
   HvnF3’H mainly distributed in the cytoplasm and cell membrane and
   interacted with several stress proteins such as methyltransferase
   protein and zinc finger protein.

Conclusions

   The results of this study provide reference for the regulation
   mechanism of anthocyanin-related genes in purple grain qingke.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12864-024-10738-9.

   Keywords: Qingke, Anthocyanin synthesis, Weighted gene co-expression
   network, Yeast two-hybrid, Subcellular localization, F3’H

Background

   Qingke (Hordeum vulgare L. var. nudum Hook. F.) being a characteristic
   crop of the Qinghai-Tibet Plateau, plays a crucial role in providing
   food and forage for the Tibetan people [[38]1]. In botanical
   classification, qingke belongs to a variety of barley, whose genome is
   highly similar to that of barley, and the inner lemma of qingke can be
   separated from the caryopsis at maturity, which is also known as naked
   barley [[39]2]. As the main grain crop in the Qinghai-Tibet Plateau,
   with the gradual improvement of people’s living standards, the dietary
   demand has changed from focusing on food and clothing to emphasizing
   health, and its edible value has attracted much attention [[40]3].
   Qingke is rich in protein and contains most of the essential amino
   acids for the human body. Meanwhile, qingke has more dietary fiber and
   less fat than other crops, and the content of β-glucan in qingke is the
   highest among barley crops [[41]4]. In addition, qingke is rich in
   phenols and anthocyanins, which play an important role in anti-aging
   and cancer prevention [[42]5].

   The biosynthesis pathway of anthocyanins involves multiple enzymes and
   intermediate products [[43]6]. Phenylalanine aminotransferase (PAL) is
   the first enzyme in the anthocyanin biosynthesis pathway, converting
   phenylalanine to transaminate coumaric acid, the subsequent steps
   include converting to cinnamic acid, coumarin acid, and ultimately
   synthesizing anthocyanins [[44]7]. In addition, the anthocyanin
   synthesis pathway is affected by various regulatory factors [[45]8].
   Among them, transcription factors are one of the most important
   regulatory factors [[46]9]. The complex of MYB, bHLH and WD40
   transcription factors is called MYB-bHLH-WD40 (MBW) complex, which
   plays a key role in anthocyanin biosynthesis pathway [[47]10, [48]11].
   MYB transcription factors regulate the expression of specific
   anthocyanin synthesis genes by interacting with bHLH and WD40
   transcription factors [[49]12]. In addition, other transcription
   factors and hormone signals are also involved in the regulation of
   anthocyanin biosynthesis, and plant hormones such as ethylene can
   affect anthocyanin biosynthesis through signal transduction pathways
   [[50]13]. These hormones can regulate anthocyanin synthesis pathway by
   altering the expression of related genes [[51]14]. Not only that,
   environmental factors also have a certain impact on anthocyanin
   biosynthesis [[52]15]. Environmental factors such as light, temperature
   and adversity can affect anthocyanin accumulation [[53]16–[54]18]. For
   example, UV exposure can stimulate plants to synthesize more
   anthocyanins as a defense mechanism against UV radiation [[55]19].

   The anthocyanin regulatory pathway is a complex network involving the
   interaction of many factors, including gene expression regulation,
   substrate supply, environmental factors and signal transduction
   pathways [[56]20, [57]21]. These regulatory mechanisms together affect
   anthocyanin synthesis and accumulation in plants. At the same time, the
   anthocyanin regulatory pathway may be different in different plant
   species and environmental conditions. At present, the mechanism of
   regulating anthocyanin synthesis of qingke grain is still unclear.

   Recent high-throughput sequencing analyses, especially in combination
   with metabolome and transcriptome analysis, have characterized
   anthocyanin synthesis pathways in a variety of non-model plants such as
   asparagus, turnip, and grape [[58]22–[59]24]. However, previous studies
   of barley seed pigments have mainly focused on comparisons of color
   differences among seeds; to our knowledge, no studies of the molecular
   mechanisms of seed anthocyanin formation are available and
   understanding this mechanism is important for the improvement of barley
   seedcoat color. Here, genes associated with anthocyanin biosynthesis
   and its regulations were identified using WGCNA based on
   transcriptomics. of these, the hub genes, accumulation-related
   structural genes and transcription factors were further examined. The
   results improve our understanding of the molecular mechanisms
   underlying anthocyanin biosynthesis in qingke and provide valuable
   information for the genetic improvement of purple qingke and other
   purple crops.

Materials and methods

Plant material and data acquisition

   Nierumuzha (purple grain) and Kunlun 10 (white grain) were obtained
   from Academy of Agriculture and Forestry Sciences, Qinghai University.
   According to the growth scale, the seeds of two varieties were
   collected at early milk, late milk and soft dough stage for gene
   expression analysis [[60]25]. Transcriptome sequencing was performed
   using three biological replicates of each sample [[61]26].

Construction and analysis of weighted gene co-expression network

   The WGCNA package of R language was used to construct gene
   co-expression network analysis
   ([62]https://horvath.genetics.ucla.edu/html/CoexpressionNetwork/Rpackag
   es/WGCNA/index.html). Then the weighting coefficient β was calculated
   by the pickSoftThreshold function to make the WGCNA conform to the
   scale-free network distribution. The pickSoftThreshold function in
   WGCNA package is used to calculate the weighting coefficient β to make
   WGCNA conform to the scale-free network distribution. The β value with
   correlation coefficient squared close to 0.8 and certain gene
   connectivity was selected as the soft threshold of WGCNA. The dynamic
   tree cut method and blockwiseModules are used to identify coexpression
   modules and construct the network.

   Principal component analysis was carried out for all genes in each
   module. PC1 value was used as the module feature vector ME to calculate
   the correlation coefficient r and corresponding p value between the
   module feature vector and different traits in each module. Where, r > 0
   represented positive correlation, and r < 0 represents negative
   correlation. Then we selected |r|>0.70, p < 0.001 as correlation
   module.

   SoftConnectivity function was used to calculate gene connectivity, and
   hub genes of each module (top 5 genes of connectivity) were screened.
   The weight values between different genes in each module are calculated
   according to the topological overlap matrix and visualized using
   Cytoscape v3.9.1.

RNA‑seq analysis and functional annotation

   We used the MiniBEST Universal RNA Extraction Kit (TaKaRa, Tokyo,
   Japan) to extract RNA from the seedcoat of Nierumuzha and Kunlun10 at
   three stages, and set three biological replicates for each sample. RNA
   seq and functional annotation by Novogene (Beijing, China). We
   functionally annotated DEGs using KEGG (Kyoto Encyclopedia of Genes and
   Genomes; [63]www.kegg.jp/kegg/kegg1.html) and GO (Gene Ontology;
   [64]http://wego.genomics.org.cn/cgi-bin/wego/index.pl) enrichment
   analyses and associated analyses using the OmicStudio tools
   ([65]https://www.omicstudio.cn/tool/62).

Reverse transcription-polymerase chain reaction (RT-PCR) and
quantitative-real-time polymerase chain reaction (qRT-PCR) analysis

   The reverse transcription of the total RNA of the sample was completely
   carried out according to the steps of the cDNA synthesis kit
   (PrimeScript™ II 1st Strand cDNA Synthesis Kit) and stored at -80℃.
   qRT-PCR were performed using the SYBR Premix Ex Taq II Kit (Takara,
   Tokyo, Japan) on a Roche LightCycler480 II (Roche Molecular
   Biochemicals, Mannheim, Germany). Design primers for 15 hub genes using
   Primer 5.0 (Table [66]S1). The qRT-PCR operation program refers to the
   method of Yao et al. [[67]26]. Using TC139056 as an internal reference
   gene for qRT-PCR analysis [[68]27]. Gene expression levels were
   calculate using the 2^−∆∆Ct method.

Subcellular localization

   We constructed p35S::HvnF3’H: GFP recombinant vector by reconstructing
   HvnF3’H CDS sequence onto a carrier containing green fluorescent
   protein (Table [69]S2). The recombinant vector was transformed into
   Agrobacterium tumefaciens EHA105, and the agrobacterium carrying the
   recombinant vector was transferred into Nicotiana benthamiana by
   injection, cultured in the dark for 2–3 d, and sliced by laser scanning
   confocal microscope (Nikon A1R+, Japan).

Construction of qingke cDNA library

   The total RNA of Nierumuzha seed skin was extracted by Trizol Kit
   (Invitrogen™, Shanghai). The mRNA was isolated and purified using the
   Oligotex mRNA Midi Kit (Qiagen, Shanghai). The cDNA library (Uncut
   type) was constructed using the CloneMiner II cDNA library construction
   kit (ThermoFisher, Shanghai) by reverse transcription of RNA.

   We recombined cDNA onto pDONR222 vector and transformed the recombinant
   vector into Escherichia coli (E. coli) DH10B for primary library
   construction. We reconstructed the plasmid extracted from the primary
   library onto the pGADT7-DEST vector, transformed the recombinant vector
   into Escherichia coli DH10B, then cultured it at 37℃ to construct the
   secondary library. Further, the secondary library plasmid was
   transformed into yeast Y187 to construct a yeast library, and 100 µL
   diluter 1/10, 1/100, 1/1000, 1/10,000 was applied to 100 mm SD/-Leu
   plate to calculate the conversion efficiency. Library capacity
   identification was performed using 10 µL of converted E. coli stock
   solution, which was diluted 100 times with SOC medium, 50 µL of the
   dilute solution was applied to LB plate (containing 50 µg/mL kanamycin)
   and cultured in a constant temperature incubator at 37℃ overnight. The
   clones on the plate were counted on the second day.

Yeast two-hybrid (Y2H) verification

   We used pGBKT7-F3’H as bait protein for yeast library screening,
   sequenced the positive clones, and recombined them into pGADT7 vector
   according to the sequence. Two germplasms containing AD and BK vectors
   were co-transferred into the yeast receptive state, and 600 µL PEG/LiAC
   solution was added with high-speed vortex. Shake at 30ºC at 200 rpm for
   30 min, then add 70 µL DMSO, slowly invert and mix; Heat shock at 42ºC
   for 15 min, place on ice for 2 min (quickly insert into ice bath),
   centrifuge at 14,000 rpm for 5s, remove supernatant, and add 100 µL
   1×TE suspensible bacteria. Apply to SD/-Leu-Trp for screening, grow at
   30ºC for 2–3 d. The monoclones were scraped and gradient diluted with
   SD/-Leu-Trp liquid Dilute the bacterial solution onto SM/-Leu-Trp for
   self-activation validation, and apply it onto SD/-Trp His-Ade deficient
   culture medium. Grow at 30℃ for 2–3 d for protein interaction
   validation.

Bimolecular fluorescence complementation (BiFC) assays

   Cloning the CDS sequence of HvnF3’H and the gene screened by Y2H using
   primers with Pac I and Spe I double enzyme cutting sites. Recombine
   HvnF3’H onto the pBiFC-YC vector, and recombine the genes screened by
   Y2H onto the pBiFC-YN vector (Table [70]S2). The extracted plasmid was
   transformed into Agrobacterium tumefaciens EHA105. Agrobacterium with
   recombinant vector was injected into Nicotiana benthamiana and cultured
   in the dark for 3 d. The sections were observed using laser confocal
   microscopy (Nikon A1R+, Japan).

Results

Data processing and WGCNA

   To better obtain genes associated with anthocyanin synthesis, we
   generated RNA-seq data for the seedcoats of Kunlun10 and Nierumuzha at
   the three different developmental stages (18 samples). In total, we
   generated 112.62 Gb clean data, an average of 6.26 Gb per sample. The
   average Q20 ratio of each sample was 97.41–98.08%, and the Q30 was
   93.17–94.36%. The average GC content was 53.88%. Across the six samples
   (early milk, late milk and soft dough of Kunlun 10; early milk, late
   milk and soft dough of Nierumuzha), 37.34 (91.01%), 37.39(88.89%),
   36.53(87.32%), 37.35(90.32%), 36.35(89.21%), and 36.76 (85.02%)million
   reads, respectively, were mapped to the reference barley (H. vulgare)
   genome in total, while 33.34(81.25%), 33.17(78.87%), 32.44(77.53%),
   33.09 (80.01%), 31.44 (77.14%), and 31.92 (73.83%) million reads,
   respectively, were uniquely mapped.

   After removal of all genes that were unexpressed in 18 samples
   (FPKM = 0), the remaining 25,375 genes were used for WGCNA. The optimal
   soft threshold of 10 is selected to construct an approximate scale-free
   topological overlap matrix (Fig. [71]1A and B). A scale-free network
   was constructed to obtain 25 colored modules, corresponding to clusters
   of co-expressed genes (Fig. [72]1C). The cluster modules ranged in size
   from 49 genes (orange) to 5801 genes (turquoise).

Fig. 1.

   [73]Fig. 1
   [74]Open in a new tab

   Constructions of gene co-expression network. A: scale-free topology fit
   index as a function of the soft-thresholding power, the red line
   represents that R^2 is equal to 0.8. B: mean connectivity as a function
   of the soft-thresholding power. C: Clustering dendrograms of genes and
   module detecting. D: Gene clustering heat map of module

Identification of color-related modules in seedcoat

   Due to the large number of genes in each module, representative genes
   (module characteristic gene, ME) in each module were selected for
   clustering. We conducted cluster analysis on 25 co-expression modules,
   and the results showed that the 25 co-expression modules were divided
   into 3 categories, each of which contained two main branches
   (Fig. [75]1D). At the same time, the heat map of correlation between
   modules shows that there are different degrees of correlation between
   the 25 modules.

   In order to further understand the relationship between modules and
   samples, we used the data of six samples in three grain formation
   periods of two varieties, white grain and purple grain, to conduct
   correlation analysis with 25 modules obtained, and constructed the heat
   map of module phenotypic correlation. When the correlation is red, it
   means that the characteristic gene of the module is positively
   correlated with the phenotype; on the contrary, when the correlation is
   blue, this means that there is a negative correlation, with the
   correlation getting deeper as the color gets darker. Among the 25 gene
   modules, black, red and green modules were positively correlated with
   early milk maturation (r = 0.95, P = 4e-03), late milk maturation
   (r = 0.90, P = 0.02) and soft dough stage (r = 0.95, P = 4e-03) of
   Kunlun10 (white), respectively. brown, yellow and turquoise modules
   were positively correlated with early milk ripening (r = 0.96,
   P = 2e-3), late milk ripening (r = 0.92, P = 9e-3) and soft dough
   ripening (r = 0.95, P = 4e-3) of Nierumuzha (purple), respectively
   (Fig. [76]2).

Fig. 2.

   [77]Fig. 2
   [78]Open in a new tab

   Correlation between modules and traits. Each row represents a module
   and each column represents a sample. The numbers in the rectangular box
   represent the correlation coefficient between the module and the sample
   and the corresponding p value

Analysis of gene expression patterns of key modules

   After obtaining the key modules, we analyzed the expression levels of
   all genes and ME genes in these six modules in all samples
   respectively, and found that the gene expression levels in each module
   were highly correlated, and the ME expression level was also highly
   correlated with the overall expression level of the module, indicating
   that the ME of the target module could fully represent the whole gene
   in its module. Red represents high expression, while green represents
   low expression. The results showed that genes of black, red and green
   modules were highly expressed in early milk, late milk and soft dough
   stage of Kunlun10 (white grain), respectively. Genes of brown, yellow
   and turquoise modules were highly expressed in early milk, late milk
   and soft dough stage of Nierumuzha (purple grain), respectively
   (Fig. [79]3).

Fig. 3.

   [80]Fig. 3
   [81]Open in a new tab

   Expression levels of all genes and corresponding ME in different
   modules of each sample. The above figure (heat map) shows the
   expression level of all genes in modules. The row represents all genes
   in the module, and the column represents each sample. A: brown module;
   B: yellow module; C: turquoise module; D: black module; E: red module;
   F: green module. The below figures show the ME expression level of the
   module in the samples

Functional enrichment analysis of key module genes

   With the progress of the period, the grain color of purple variety
   Nierumuzha gradually deepened. In order to further explore the function
   of gene modules related to anthocyanin synthesis of qingke grains, GO
   enrichment was analyzed using brown, yellow and turquoise modules
   related to purple grains. They were divided into three categories:
   biological process, cell component and molecular function. As you can
   see, in terms of biological processes, brown module was mainly enriched
   in intracellular transport (GO: 0046907), establishment of localization
   in cell (GO: 0051649) and protein transport (GO: 0015031) and other
   regulation and colligation; yellow module was mainly enriched in DNA
   integration (GO: 0015074). Turquoise module was mainly enriched to
   protein ubiquitination (GO: 0016567), protein modification by small
   protein conjugation (GO: 0032446) and nitrogen compound transport (GO:
   0071705). In terms of molecular functions, brown module was mainly
   enriched in microtubule binding (GO: 0008017) and tubulin binding (GO:
   0015631). Yellow module was mainly enriched in histone binding (GO:
   0042393) and transcription factor activity, protein binding (GO:
   0000988) and RNA binding (GO: 0003723); turquoise module was mainly
   enriched in ubiquitin-protein transferase activity (GO: 0004842) and
   ubiquitin-like protein transferase activity (GO: 0019787). In terms of
   cell components, brown module was mainly enriched in three cell
   components, endomembrane system (GO: 0012505), membrane coat (GO:
   0030117) and coated membrane (GO: 0048475). Yellow module mainly was
   enriched in transferase complex (GO: 1990234), nucleoplasm (GO:
   0005654) and nucleoplasm part (GO: 0044451). Turquoise module was
   mainly enriched in peroxisome (GO: 0005777) and microbody (GO:
   0042579). We analyzed the KEGG pathway in order to explore the pathway
   of key module genes. Brown module was mainly enriched in the Glycine,
   serine and threonine metabolism (ath00260) pathway. Yellow module was
   mainly concentrated in phagosome (ath04145), oxidative phosphorylation
   (ath00190) and lysine degradation (ath00310). Turquoise module was
   mainly enriched in biosynthesis of amino acids (ath01230) and
   plant-pathogen interaction (ath04626) pathways (Fig. [82]4).

Fig. 4.

   [83]Fig. 4
   [84]Open in a new tab

   GO functional and KEGG pathway enrichment analysis in the three
   modules. A, B and C: GO enrichment analysis of three modules; D, E and
   F: KEGG enrichment analysis of three modules. The vertical axis
   represents the enriched GO term or KEGG pathway, and the horizontal
   axis represents the ratio of enriched genes in the module. The point
   size represents the gene number enriched in the pathway, and the color
   represents the p-value corrected for multiple testing

Construction of gene interaction network

   In order to obtain the hub genes in the three purple granulose-related
   modules, Cytoscape software was used to visualize the gene regulatory
   network. Five genes with the highest connectivity in each module were
   selected as the hub genes (red dots) (Fig. [85]5). The hub genes in
   brown module were GST (HORVU4Hr1G081110), MFP1 (HORVU4Hr1G069210),
   GR-RBPs (HORVU5Hr1G040380), UBXD5 (HORVU6Hr1G016460) and d-Cdes
   (HORVU6Hr1G079140). The hub genes ID in yellow module were
   IAA(HORVU1Hr1G028170), PMEI(HORVU1Hr1G093370), IMO(HORVU3Hr1G011820),
   ROC8 (HORVU4Hr1G083750), and PRB1-2 (HORVU7Hr1G040730). The hub genes
   ID in turquoise module were CHI (HORVU5Hr1G046480, HORVU5Hr1G112670),
   F3’H (HORVU1Hr1G094880), C4H (HORVU3Hr1G080830) and 4CL
   (HORVU4Hr1G072130) (Table [86]1).

Fig. 5.

   [87]Fig. 5
   [88]Open in a new tab

   Gene co-expression networks within specific modules related to
   anthocyanin synthesis. A, B, and C represent brown, yellow, and
   turquoise modules respectively. The red circles in each network map
   represent the core genes of each module. Genes are connected by line
   segments, indicating that genes are related to each other

Table 1.

   Functional annotations of hub genes in different modules
   Module Hub gene Gene function
   Brown HORVU4Hr1G081110 Encodes Glutathione S-transferase (GST)
   HORVU4Hr1G069210 MFP1 attachment factor 1(MFP1)
   HORVU5Hr1G040380 Glycine-rich RNA-binding protein (GR-RBPs)
   HORVU6Hr1G016460 UBX domain-containing protein 6 (UBXD6)
   HORVU6Hr1G079140 Putative D-cysteine desulfhydrase (d-Cdes)
   Yellow HORVU1Hr1G028170 Auxin-responsive protein IAA1 (IAA)
   HORVU1Hr1G093370 Putative pectin esterase/pectin esterase inhibitor 38
   (PMEI)
   HORVU3Hr1G011820 Indole-2-monooxygenase (IMO)
   HORVU4Hr1G083750 Homeodomain transcription factor ROC8 (ROC)
   HORVU7Hr1G040730 Pathogenesis-related protein PRB1-2 (PRB)
   Turquoise HORVU5Hr1G046480 Probable chalcone-flavonone isomerase 3
   (CHI)
   HORVU1Hr1G094880 Flavonoid 3’-monooxygenase (F3’H)
   HORVU3Hr1G080830 Cinnamic acid 4-hydroxylase (C4H)
   HORVU4Hr1G072130 4-coumarate–CoA ligase-like 1 (4CL)
   HORVU5Hr1G112670 Chalcone-flavonone isomerase (CHI)
   [89]Open in a new tab

The expression of hub genes by qRT-PCR

   15 hub genes related to anthocyanin in qingke with purple grains, and
   15 of them were selected for real-time fluorescence quantitative PCR
   verification. We used the same batch of RNA in transcriptome sequencing
   for reverse transcription, and designed primers to detect the
   expression of these genes in the three stages of the Nierumuzha and
   Kunlun10 by qRT-PCR (Fig. [90]6). The test results showed that the
   expression levels and changing trends of the tested genes in different
   periods were basically consistent with the transcriptome results,
   indicating that the selected core genes had important research value in
   affecting the formation of purple grains. In addition, the qRT-PCR
   results also proved the reliability of the transcriptome sequencing
   results.

Fig. 6.

   [91]Fig. 6
   [92]Open in a new tab

   Expression patterns of the 15 hub genes in Kunlun10 and Nierumuzha. The
   TC139056 were used as internal. Blue represents the relative expression
   level of Kunlun 10, while orange represents the relative expression
   level of Nierumuzha. (A): HORVU4Hr1G081110; (B): HORVU4Hr1G069210; (C):
   HORVU5Hr1G040380; (D): HORVU6Hr1G016460; (E): HORVU6Hr1G079140; (F):
   HORVU1Hr1G028170; (G): HORVU1Hr1G093370; (H): HORVU3Hr1G011820; (I):
   HORVU4Hr1G083750; (J): HORVU7Hr1G040730; (K): HORVU5Hr1G046480; (L):
   HORVU1Hr1G094880; (M): HORVU3Hr1G080830; (N): HORVU4Hr1G072130; (O):
   HORVU5Hr1G112670. Analysis of differences in relative expression levels
   between two varieties at different stages using one-way ANOVA. *:
   p < 0.05; **: p < 0.01

Spatiotemporal expression analysis of HvnF3’H

   Quantitative real time Polymerase Chain Reaction (qRT-PCR) was used to
   detect the relative expression of this gene in the early milk, late
   milk, and soft dough stage of Nierumzha and Kunlun 10 seed coat color
   formation. The results showed that the expression of HvnF3’H of Kunlun
   10 was lower in early milk, late milk and soft dough of grain color
   formation, but there were significant differences. The expression level
   of HvnF3’H in Nierumzha showed a significant increase in the late milk
   and soft dough stage (P < 0.01). The expression of HvnF3’H in Nierumzha
   was significantly higher than that in Kunlun 10 in late milk and soft
   dough stage (P < 0.01). It is speculated that with anthocyanin
   synthesis in seed coat, the expression of HvnF3’H is gradually
   increased, and this gene may positively regulate anthocyanin
   accumulation. In addition, qRT-PCR was performed on the roots, stems,
   leaves and awns of the two varieties of Nierumzha and Kunlun 10. The
   results showed that HvnF3’H was not expressed in the roots, stems,
   leaves and awns of the two varieties, but was significantly increased
   in the seed coats of Nierumzha (Fig. [93]7A and B).

Fig. 7.

   [94]Fig. 7
   [95]Open in a new tab

   Expression and subcellular localization of HvnF3’H. A: Expression of
   HvnF3’H in various tissues. The TC139056 were used as internal gene;
   W1, W2, W3 represent the early milk, late milk and soft dough stage of
   Kunlun10; P1, P2, P3 represent the early milk, late milk and soft dough
   stage of Nierumuzha; *: p < 0.05; **: p < 0.01; B: Observation of grain
   phenotypes of two varieties; C: HvnF3’H was fused with GFP and
   expressed in Nicotiana benthamiana leaves. The fluorescence signal was
   observed by laser scanning confocal microscope. Scale bars = 50 μm

Subcellular localization of HvnF3’H

   In order to investigate the subcellular localization of the HvnF3’H, we
   successfully transformed the recombinant vector into Agrobacterium
   tumefaciens EHA105, and then injected the Agrobacterium tumefaciens
   containing the recombinant vector into tobacco leaves for microscopic
   observation. The green fluorescence signal of the HvnF3’H mainly exists
   in the cytoplasm and cell membrane of Nicotiana benthamiana, and there
   is also a small distribution in the nucleus. The green fluorescence
   signal of the control spread throughout the entire cell, indicating
   that HvnF3’H is a protein located in the cell membrane and cytoplasm
   (Fig. [96]7C).

Y2H screening of cDNA library

   To explore the specific regulatory mechanism of HvnF3’H, we first
   determined the insertion fragment size and recombination rate of the
   cDNA library. The findings revealed that the capacity of the primary
   library is 1.28 × 107 cfu and that of the secondary library is
   1.12 × 107 cfu. The recombination rate of the two libraries reached
   100%, and the average inserted fragment length exceeded 1000 bp,
   indicating that the constructed libraries met the quality requirements
   (Fig.[97]S1). We amplified and constructed the HvnF3’H into a pGBKT7
   vector as a bait protein (pGBKT7-F3’H), and screened it using a yeast
   nuclear library. In the protein toxicity test, the yeast transformed by
   the recombinant bait vector grew normally on SD/-Leu/-Trp medium and
   showed similar colony density as the blank control (pGBKT7 Lam). In the
   self-activation test, we set a negative control (pGBKT7 Lam) and a
   positive control (pGBKT7-53), and through comparison, we found that the
   recombinant vector yeast did not grow on
   SD/-Leu/-Trp/-His/-Ade/X-α-gal/AbA medium. The concentration of AbA is
   125 ng/mL, and the concentration of X-α-gal is 40 mg/L. The above
   results indicate that the HvnF3’H has no significant toxicity to yeast
   and there is no self-activation phenomenon, which can be used for
   subsequent yeast library screening and one-on-one validation. The
   pGBKT7-F3’H was co screened with the yeast cDNA library of qingke
   (DOE20223637), and a total of 34 blue clones were found on the
   SD/-Leu/-Trp/-His/X-α-gal screening plate. The blue clones on the
   screening plate were selected for re-screening, and a total of 27
   clones grew and turned blue.

   We selected genes that were “in-frame” in the library vector for yeast
   two hybrid one-on-one validation, and obtained a total of seven
   proteins that interacted with HvnF3’H (Fig. [98]8A). On this basis, we
   performed functional annotation on seven genes encoding candidate
   proteins. Among them, HORVU4Hr1G003060 encodes Lysine
   N-methyltransferase (KMT), HORVU7Hr1G100810 encodes
   1-aminocyclopane-1-carboxylate oxidase-1-like protein (ACO),
   HORVU6Hr1G093960 belongs to the TCP family transcription factor (TCP),
   HORVU5Hr1G006550 encodes Cation/calcium exchange 1 (CCXs),
   HORVU0Hr1G021850 encodes FCS-Like Zinc finger 6-like (FLZ),
   HORVU7Hr1G026770 encodes F-box protein At5g67140-like isoform X (FBK),
   and HORVU3Hr1G028810 encodes Heme binding protein 2 (HEBP2)
   (Table [99]2).

Fig. 8.

   [100]Fig. 8
   [101]Open in a new tab

   Y2H screening and BiFC analysis. A: The positive clones screened by
   yeast library were verified one-to-one by yeast two-hybridization, and
   the dilute concentrations of 10^− 1, 10^− 2, 10^− 3 and 10^− 4
   bacterial solution were placed on the medium with DDO/X and TDO/X.
   DDO/X: SD/-Leu/-Trp/X-α-gal; TDO/X: SD/-Leu/-Trp/-His/X-α-gal. B: The
   in vivo interaction between HvnF3’H and screened protein in tobacco
   leaves was detected by BiFC method, and the fluorescence signal was
   observed under fluorescence microscope. Scale bars = 50 μm

Table 2.

   Functional annotation of screening library genes
   gene ID gene name
   HORVU4Hr1G003060 Lysine N-methyltransferase (KMT)
   HORVU7Hr1G100810 1-aminocyclopropane-1-carboxylate oxidase-1-like
   protein (ACO)
   HORVU6Hr1G093960 TCP family transcription factor (TCP)
   HORVU5Hr1G006550 Cation/calcium exchanger 1 (CCXs)
   HORVU0Hr1G021850 FCS-Like Zinc finger 6-like (FLZ)
   HORVU7Hr1G026770 F-box protein At5g67140-like isoform X (FBK)
   HORVU3Hr1G028810 Heme-binding protein 2 (HEBP2)
   [102]Open in a new tab

   In order to avoid false positives in Y2H, we detected seven proteins
   interacting with HvnF3’H through biomolecular fluorescence
   complementarity. We fused the HvnF3’H with the C-terminal yellow
   fluorescent protein YFPc, and the candidate protein fused with the
   N-terminal yellow fluorescent protein YFPn, successfully constructing a
   recombinant vector for Agrobacterium tumefaciens transformation. We co
   transformed Agrobacterium tumefaciens containing recombinant vectors
   into tobacco leaves for microscopic observation, and the results showed
   that among the seven candidate proteins, except for HORVU5Hr1G006550
   and HORVU7Hr1G026770, the remaining five candidate proteins all
   interacted with the HvnF3’H (Fig. [103]8B).

Discussion

   Anthocyanins were an important class of plant phenolic pigments, which
   play an important role in the color formation of many plant organs and
   are the key factors for the red, blue or purple appearance of plant
   organs [[104]28, [105]29]. In addition to participating in color
   formation, anthocyanins were also able to improve plant stress
   resistance [[106]30]. Previous studies had shown that anthocyanins play
   an important role in high temperature resistance, cold resistance,
   drought resistance and ultraviolet radiation resistance [[107]31].
   Purple grain qingke is a kind of important germplasm resources, the
   study of the formation of the color and the synthesis pathway of
   anthocyanin of purple grain qingke will play the basic role in the
   improvement of qingke varieties.

   In the study of anthocyanins in qingke, we conducted in-depth analysis
   of the selected hub genes. Functional annotation showed that the hub
   genes of Nierumuzha in soft dough phase encoded key enzymes in
   anthocyanin synthesis pathway, among which C4H encoded by
   HORVU3Hr1G080830 belonged to plant P450 monooxygenase, acting on the
   first stage of anthocyanin synthesis and catalyzing the second step of
   phenylalanine pathway [[108]26]. CHI encoded by HORVU5Hr1G046480 is one
   of the rate-limiting enzymes of flavonoid metabolism pathway and acts
   on the second stage of anthocyanin synthesis [[109]32]. Previous
   studies had shown that CHI is involved in the formation of plant flower
   color. By inhibiting CHI expression in tobacco using RNA interference
   technology, the color of petals and pollen became lighter [[110]33].
   The GST encoded by HORVU4Hr1G081110 is transported to plant vacuoles
   through the combination of glutathione (GSH) and Anthocyanidin to
   promote the accumulation of Anthocyanidin [[111]34]. HORVU6Hr1G016460
   was related to ubiquitination. Promoting ubiquitin degradation also had
   a negative regulation effect on Anthocyanidin accumulation [[112]35].
   HORVU1Hr1G028170 and HORVU3Hr1G011820 were related to photosensitive
   pigments. The transcription factor HY5 encoding photosensitive pigments
   could positively regulate anthocyanidin accumulation by combining with
   anthocyanidin structural gene [[113]36].

   In order to further study these hub genes, we found that the expression
   of HvnF3’H was in line with expectations through fluorescence
   quantification, so we focused on this gene. HORVU1Hr1G094880 encoded
   flavonoid 3’-hydroxylase, which catalyzes the single oxygenation
   reaction involving NADPH and O[2], and is another rate limiting enzyme
   in the flavonoid metabolic pathway [[114]37]. Its metabolites are
   important intermediates in the process of Anthocyanidin synthesis
   [[115]38]. Brugliera et al. detected the expression of F3’H gene in
   petunias petals, ovary, calyx, stem and other organs [[116]39].
   Castellarin et al. found that F3’H was highly expressed in the roots,
   stems, leaves, flowers and seeds of grape [[117]40]. Compared with the
   expression of these plants, the expression of F3’H in purple grain
   qingke was relatively simple, and only showed high expression in the
   seed skin of purple grain qingke. Through subcellular localization, we
   found that the green fluorescence signal was mainly distributed in the
   cell membrane and cytoplasm, indicating that HvnF3’H mainly plays a
   role in the cell membrane and cytoplasm.

   Functional annotation was performed on the coding genes of 5 candidate
   proteins, among which HORVU4Hr1G003060 encoded Lysine
   N-methyltransferase (KMT). Research has shown that colorless
   dihydroflavonols are formed under the catalysis of the F3’H, which are
   then catalyzed by dihydroflavonol reductase (DFR) and anthocyanin
   synthase (ANS) to form colored anthocyanin glycosides. Finally,
   different anthocyanin glycosides are catalyzed by methyltransferase
   (MT), glycosyltransferase (GT), and acyltransferase (AT) to form
   different anthocyanin glycosides [[118]41]. HORVU7Hr1G100810 encodes
   1-aminocyclopropane-1-carboxylate oxidase (ACO). Studies showed that
   the ACO had a promoting effect on the synthesis of ethylene compounds;
   the accumulation of anthocyanins significantly increases after treated
   by ethylene [[119]42]. An et al. found that ethylene responsive factors
   ERFs are involved in regulating plant Anthocyanidin synthesis
   [[120]43]. HORVU6Hr1G093960 belonged to the TCP transcription factor
   family. Xu et al. found TCP might be involved in anthocyanin
   accumulation, and the TCP played an inhibitory role in anthocyanin
   synthesis in camellia [[121]44]. We found HvnF3’H was an important
   structural gene in the early stage of purple qingke in anthocyanin
   synthesis [[122]26]. Based on our study, HvnF3’H is not only a
   structural gene in the anthocyanin synthesis pathway in purple qingke,
   but the specific mechanism of action still needs further exploration.

Conclusions

   Through the construction of a gene co-expression network, this study
   identified hub genes related to anthocyanin, most of which directly or
   indirectly affect the synthesis or accumulation of anthocyanin. HvnF3’H
   is located in the cell membrane and cytoplasm. Through Y2H and BiFC
   experiments, it was found that HvnF3’H interacts with various proteins
   related to anthocyanin synthesis and metabolism. The experimental
   results indicated that HvnF3’H regulated the synthesis of anthocyanin
   in purple-grained qingke through interactions with related proteins,
   but the specific regulatory mechanism requires further exploration.
   These results provide theoretical guidance for the in-depth analysis of
   the molecular mechanism and biological function of anthocyanin-related
   genes in purple qingke.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [123]Supplementary Material 1^ (9.6KB, xlsx)
   [124]Supplementary Material 2^ (11.8KB, xlsx)
   [125]Supplementary Material 3^ (5.2MB, tif)
   [126]Supplementary Material 4^ (34.2KB, xlsx)
   [127]Supplementary Material 5^ (1.1MB, tif)
   [128]Supplementary Material 6^ (1.2MB, tif)
   [129]Supplementary Material 7^ (6.4MB, tif)
   [130]Supplementary Material 8^ (12.2MB, tif)

Acknowledgements