Expression profiling reveals functionally redundant multiple-copy genes related to zinc, iron and cadmium responses in Brassica rapa



  • Genes underlying environmental adaptability tend to be over-retained in polyploid plant species. Zinc deficiency (ZnD) and iron deficiency (FeD), excess Zn (ZnE) and cadmium exposure (CdE) are major environmental problems for crop cultivation, but little is known about the differential expression of duplicated genes upon these stress conditions.
  • Applying Tag-Seq technology to leaves of Brassica rapa grown under FeD, ZnD, ZnE or CdE conditions, with normal conditions as a control, we examined global gene expression changes and compared the expression patterns of multiple paralogs.
  • We identified 812, 543, 331 and 447 differentially expressed genes under FeD, ZnD, ZnE and CdE conditions, respectively, in B. rapa leaves. Genes involved in regulatory networks centered on the transcription factors bHLH038 or bHLH100 were differentially expressed under (ZnE-induced) FeD. Further analysis revealed that genes associated with Zn, Fe and Cd responses tended to be over-retained in the B. rapa genome. Most of these multiple-copy genes showed the same direction of expression change under stress conditions.
  • We conclude that the duplicated genes involved in trace element responses in B. rapa are functionally redundant, making the regulatory network more complex in B. rapa than in Arabidopsis thaliana.


Among the 49 nutrients required by humans in low amounts to meet their metabolic needs, 16 trace element minerals are essential, including iron (Fe) and zinc (Zn) (Welch & Graham, 2004). Zn deficiency is a prevalent nutritional problem for the human population because agricultural systems in many countries fail to provide products containing adequate quantities of Zn (Graham et al., 2001; Welch & Graham, 2004). The main reasons for this are the low bioavailability of Zn in many agricultural systems and the inability of many plant species to store substantial quantities of Zn in edible plant parts. In general, Zn deficiency is widespread among plants grown in highly weathered acidic soils and in calcareous soils, where it is often associated with deficiency of Fe and Mn (Graham et al., 1992; Marschner, 1995; Cakmak et al., 1997, 1999; Wu et al., 2007). Although Zn is usually considered a nontoxic micronutrient for plants, toxicity symptoms may occur upon extremely high Zn intake (Mitchell & Fretz, 1977; Fosmire, 1990; Ebbs & Kochian, 1997; Paschke et al., 2000). The effects of Zn toxicity are even more detrimental to plant growth than Zn shortage (Marschner, 1995; Wu et al., 2007). A previous study on three Brassica species indicated that high Zn exposure could reduce Fe and Mn concentrations in shoots, resulting in Fe and Mn deficiencies (Ebbs & Kochian, 1997). Although soils are generally rich in Fe, most of it is in a biologically unavailable form, causing severe problems for agricultural production and consequently for human nutrition (Guerinot, 2001). Cadmium (Cd) is a toxic element to plants, animals and humans. Cd shares many physical and chemical properties with Zn and Fe, adding to the toxicity of Cd (Das et al., 1997). Uptake by crop plants is the main way Cd enters into the human food chain (Clemens et al., 2013a). Because of rapid urbanization in many countries, soils used for agriculture are becoming easily and increasingly polluted by Cd (Das et al., 1997).

Microarray studies have reported dramatic differences in gene transcription in response to Zn and Fe deficiency and Zn and Cd toxicity in the model species Arabidopsis thaliana (Wintz et al., 2003; Becher et al., 2004; Chiang et al., 2006; van de Mortel et al., 2006, 2008). Most of the differentially expressed genes encode metal transporters and transcription factors that play important roles in trace metal element responsive processes. Cd enters into plant cells through Fe, Zn and Ca transporters/channels of low specificity (Clemens, 2006). Because Cd competes with Fe and Zn during uptake and translocation from roots to shoots, some overlap is expected in the genes responding to Cd exposure and Fe or Zn deficiencies (Clemens, 2006; Wu et al., 2012).

Previous research has focused mainly on the expression profiles of trace element mineral-related genes in roots rather than in shoots, with the idea that minerals enter the plant through the root. However, shoots are as important as roots for keeping mineral homeostasis in plants. For example, Zn accumulators store Zn primarily in the vacuoles of leaf cells (Becher et al., 2004). In addition, the Fe concentration in shoots functions as a signal to regulate the expression of genes involved in Fe homeostasis and to prevent plants suffering from Cd toxicity (Wu et al., 2012). Therefore, it is important to know how genes are expressed in shoots in response to metal-stress conditions. Here, we focused on the differentially expressed genes in leaves under conditions of Fe deficiency, Zn deficiency, Zn excess and Cd exposure.

We studied this in the diploid species Brassica rapa (containing the Brassica A genome). Screening of 188 accessions of B. rapa showed large genetic variation in leaf Zn and Fe accumulation (Wu et al., 2007). To further research the genetic improvement of nutritional Zn and Fe concentrations and reduction of toxic Cd concentrations in this widely grown vegetable and oilseed crop, it is necessary to generate an overview of the Zn, Fe and Cd responsive genes expressed in B. rapa leaves. Recently, the whole genome of B. rapa was sequenced and annotated (Wang et al., 2011), providing a solid foundation upon which to study the expression of metal responsive genes. B. rapa is derived from the same ancestor as A. thaliana and has undergone genomic triplication since its divergence from the A. thaliana lineage (Wang et al., 2011; Cheng et al., 2013). Therefore, many of the genes in A. thaliana have two or more orthologs in the B. rapa genome. Gene annotation of B. rapa genome predicted that gene families related to environmental adaptability had tended to expand after genome triplication, including genes responsive to Cd and Zn ions (Wang et al., 2011). Although it was proposed that the duplicated genes enhanced environmental plasticity (Brichler & Veitia, 2007; Van de Peer et al., 2009), the functional conservation of duplicated genes related to Zn, Fe and Cd is not very well understood.

In this study, we present the first genome-wide analysis of gene expression in leaves of B. rapa grown under Fe deficiency (FeD), Zn deficiency (ZnD), Zn excess (ZnE) and Cd exposure (CdE) conditions, using Tag-Seq based on the Illumina GA platform. The retention ratio of trace element responsive genes and the expression levels of paralogs were further analyzed.

Materials and Methods

Plant materials and growth conditions

Plants of Brassica rapa L. accession R-o-18 were grown in a hydroponic system with 0.5 × Hoagland's nutrient solution in a climate-controlled growth chamber set at 70% humidity with a 20 : 15°C (12 h : 12 h) day : night temperature regime. The plants were grown in normal nutrient solution (2 μM Zn, 3 μM Fe, without Cd) for 14 d. Then, the plants were subjected to Fe deficiency (FeD; 0.05 μM), Zn deficiency (ZnD; 0.005 μM), Zn excess (ZnE; 50 μM), or Cd exposure (CdE; 1 μM) treatments. Plants kept under normal nutrient conditions were used as a control. Three replicates were used for each treatment, with three plants for each replicate. Zn was added into nutrient solutions as ZnSO4·7H2O, Fe was supplied as Fe(Na) EDTA and Cd was applied as 3CdSO4·8H2O. After 7 d exposure to the treatments, the first pair of expanded leaves of each plant was harvested for RNA extraction. The nutrient solutions were refreshed every 4 d.

RNA extraction and data generation

Tag-Seq was applied in this research. Tag-Seq is a tag-based transcriptome sequencing approach where short raw tags are generated by an endonuclease (Morrissy et al., 2009). Total RNA was extracted from the harvested leaves of B. rapa using the RNeasy Plant Mini Kit (Qiagen) and DNase treated with RNase-free DNase set (Qiagen) according to the manufacturer's instructions. The RNA concentrations were determined spectroscopically and the quality of the RNA was checked on a 1.0% agarose gel. At least 20 μg of total RNA (≥ 1000 ng μl−1) from each sample was sent to the Beijing Genomics Institute (BGI, for Illumina sequencing (a commercial service). Digital gene expression libraries were prepared using the Illumina gene expression sample prep kit and sequencing was performed as described by Xue et al. (2010). In total, there were 15 digital gene expression libraries, one for each of the three replicates under the four trace metal element treatments and normal nutrient supply conditions as a control. The raw data has been deposited in NCBI's Gene Expression Omnibus (Barrett et al., 2009) and are accessible through GEO series accession number GSE55264,

Analysis and mapping of digital gene expression tags in B. rapa genome

The genome sequence of B. rapa retrieved from Brassica Database (BRAD) (, together with a complete set of predicted B. rapa genes, was used as the reference database (Wang et al., 2011). Before mapping of the reads to the reference database, all sequences were filtered to remove the adaptor sequences, low quality sequences (tags with unknown nucleotides), empty tags (no tag sequence between the adaptors) and singletons (tags occurring only once that might result from sequencing errors). A pre-processed database of all possible CATG (recognition site of NmeI, which was used for digital gene expression library construction) plus 17 nucleotide tag sequences was created to analyze the expression of B. rapa genes (Xue et al., 2010). For gene annotation, the following steps were performed: (1) all clean tags were mapped to the complete set of predicted B. rapa genes using the Short Oligonucleotide Analysis Package v2.20 (SOAP; Li et al., 2009), allowing no more than one nucleotide mismatch; (2) the unmapped tags were then mapped to the B. rapa genome sequence by SOAP allowing no more than one nucleotide mismatch on the sense and antisense strands. The tags mapped to the reference database were defined as unambiguous tags. For expression analysis, the number of unambiguous tags for each gene was normalized to TPM (number of transcripts per million clean tags). A statistical analysis of the frequencies of tags in each library was performed to compare the expression levels of genes under different trace metal element supply conditions according to the method described by Audic & Claverie (1997). The false discovery rate (FDR) was used to determine the threshold P-value in multiple tests. We used an absolute value of the log2(TPMtreatment/TPMcontrol) ≥ 1 and FDR ≤ 0.001 as the threshold to judge the significance of differentially expressed genes.

Annotation of differentially expressed genes (DEGs) and identification of DEG paralogs in B. rapa

The annotation of DEGs was performed by analyzing syntenic relationships between B. rapa and A. thaliana, and a genome-wide BLASTP search. We first retrieved syntenic orthologs of DEGs in A. thaliana from BRAD (, which were determined by both sequence similarity (cutoff: E ≤ 10−20) and the colinearity of flanking genes (Cheng et al., 2012). As not all B. rapa genes have a syntenic relationship with their orthologs in A. thaliana, we used BLASTP to identify A. thaliana orthologs of DEGs with an E-value ≤ 10−20 and ≥ 70% sequence coverage (Altschul et al., 1997). Complete lists of DEGs, annotations and orthologs in A. thaliana are shown in Supporting Information Table S1. B. rapa genes sharing the same A. thaliana ortholog were defined as a paralogous set.

Phylogenetic analysis of the ZIP and HMA gene families in A. thaliana and B. rapa

B. rapa genes of the ZIP and HMA families were identified using the same approach as for identification of DEG paralogs. The amino acid sequences of all HMA and ZIP members in A. thaliana and B. rapa were aligned using MUSCLE v3.8.31 with default parameters (Edgar, 2004). PhyML v3.0 was used to construct maximum likelihood phylogenetic trees with default parameters (Guindon & Gascuel, 2003).

Quantitative RT-PCR analysis

In order to confirm the DEGs identified by Tag-Seq, 15 genes were randomly selected to perform quantitative reverse transcriptase PCR (qRT-PCR). In addition, certain DEGs for which orthologs in A. thaliana had been reported to play important roles in the regulation and transportation of trace metal elements were also examined by qRT-PCR. cDNA was synthesized from DNA-free RNA using the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA, USA) following the manufacturer's instructions. The cDNA was then used as a template in a 20-μl reaction using the F-416L DyNAmo Color Flash SYBR Green qPCR kit (Thermo Scientific, Waltham, MA, USA) to carry out subsequent reactions on a Mastercycler realplex real-time PCR machine with programs recommended by the manufacturer (Eppendorf, Hauppauge, NY, USA). All gene expression analyses were performed with three independent biological replicates. Two independent technical replicates were performed for each sample. The samples were normalized first to a selected internal control gene (GAPDH, glyceraldehyde-3-phosphate dehydrogenase) and then the relative gene expression levels were determined using the inline image method (Livak & Schmittgen, 2001). The primers used in the qRT-PCR are listed in Table S2.


Digital gene expression library sequencing

In total, 15 digital gene expression libraries of B. rapa leaves were created, from plants exposed to normal (2 μM Zn, 3 μM Fe, without Cd), FeD (0.05 μM Fe), ZnD (0.005 μM Zn), ZnE (50 μM Zn) and CdE (1 μM Cd) conditions. All raw sequence data generated from 15 libraries are available from gene expression omnibus (GEO) accession number GSE55264. The number of tags generated for each library was similar (Table 1). In total, 18.14 × 106, 18.26 × 106, 17.87 × 106, 18.05 × 106 and 17.86 × 106 raw tags were generated from three biological replicate libraries, respectively, for normal, FeD, ZnD, ZnE and CdE conditions (Table 1). After filtering the tags with low quality, the proportion of clean tags was all above 96% for each library (Table 1). All clean tags were aligned to the reference B. rapa genome and a complete set of predicted B. rapa genes (Wang et al., 2011). The proportion of clean tags that could be mapped to the reference genes of B. rapa in each dataset was between 51.45% and 60.03% (Table 1). Around 70% of the B. rapa transcriptome was represented in the complete dataset (29 183 of 41 019 reference genes), with between 53.06% and 56.48% in each individual library. Finally, the tags mapped to the reference genes generated c. 26 000 tag-mapped transcripts for each dataset, with the proportion of genes mapped with > 100 clean tags varying from 9.53% to 12.44% (Table 1).

Table 1. Overview of raw and clean tags under normal (N), iron (Fe) deficiency (FeD), zinc (Zn) deficiency (ZnD), Zn excess (ZnE) and cadmium (Cd) exposure (CdE) conditions in Brassica rapa
Librar-iesTotal tagsClean tags% clean tags in raw dataAligned clean tags% aligned clean tagsNo. unique genes hit% genes hit% genes with > 100 tag hits
N16108 8235900 93196.603542 48160.0322 87155.7511.48
N26178 0905947 52496.273304 52055.5622 35854.5010.78
N35848 5515637 73896.403146 99455.8222 54054.9510.46
FeD16198 7335965 08096.233541 44759.3722 96955.9912.44
FeD25842 3905640 66196.553349 13059.3723 03956.1612.40
FeD36223 2315997 20196.373408 12656.8322 83155.6612.26
ZnD16139 6155908 30196.233256 72555.1223 16956.4811.03
ZnD25971 5535748 24796.263365 31458.5522 89855.8211.11
ZnD35759 9025575 67596.803303 53559.2522 46754.7710.51
ZnE15984 1725776 51496.533200 42855.4022 21254.1510.77
ZnE26045 0435825 03496.363321 84857.0322 98456.0311.86
ZnE36018 9465803 41196.423228 95055.6422 72555.4011.29
CdE15872 1265691 06896.922927 86851.4521 76553.069.53
CdE26032 8165833 68496.703180 01854.5122 32854.4310.30
CdE35954 6045764 02896.803227 44955.9922 49654.8410.87

DEGs under FeD, ZnD, ZnE and CdE conditions

In order to identify the differentially expressed genes (DGEs) under FeD, ZnD, ZnE and CdE conditions, the expression abundance of tag-mapped genes was analyzed by calculating the number of TPM clean tags. High correlations of TPM value were observed between replicates under the same conditions, indicating a high degree of reproducibility of biological replicates, with an average correlation coefficient of = 0.970, 0.980, 0.939, 0.967 and 0.980, for the replicates of normal, FeD, ZnD, ZnE and CdE conditions, respectively (Fig. S1). We therefore calculated the mean TPM of three biological replicates under the same conditions for subsequent analysis.

The TPM values showed ranges of 0–21287.6 (mean 132), 0–14518.6 (mean 126), 0–15931.0 (mean 125), 0–21266.0 (mean 122) and 0–18972.8 (mean 129) in the FeD, ZnD, ZnE, CdE and control datasets, respectively. The DEGs under the four metal treatment conditions were identified using an algorithm developed by Audic & Claverie (1997). The results revealed 812, 543, 331 and 447 DEGs under FeD, ZnD, ZnE and CdE conditions, respectively (Table S1). Upregulated DEGs accounted for 76.7% of FeD, 81.0% of ZnD and 53.8% of ZnE DEGs, whereas only 13.6% of DEGs were upregulated under CdE. The 15 most upregulated DEGs under FeD, ZnD and ZnE conditions and the 15 most downregulated DEGs under CdE conditions are shown in Table 2.

Table 2. The15 most up-regulated differentially expressed genes (DEGs) under iron (Fe) deficiency (FeD), zinc (Zn) deficiency (ZnD) and Zn excess (ZnE) conditions and the15 most down-regulated DEGs under cadmium (Cd) exposure (CdE) conditions in Brassica rapa
TreatmentsGene IDGene annotationGene expression level log2(TPMtreatment/TPMcontrol)False discovery rates (FDR)
  1. a

    There was no orthologous gene identified in Arabidopsis thaliana.

FeDBra004609bHLH10013.27091.56 × 10−170
Bra001472MYB1010.07252.73 × 10−18
Bra040957a9.73951.39 × 10−14
Bra0403029.56777.73 × 10−13
Bra016946bHLH1009.43691.12 × 10−11
Bra000363Phagocytosis and cell motility protein ELMO1-related9.05144.39 × 10−9
Bra003055ORG18.74392.26 × 10−7
Bra000194EXPA88.24952.11 × 10−5
Bra0322348.17544.77 × 10−5
Bra015328AWPM-19-like membrane family protein7.85290.40 × 10−3
Bra014657bHLH0387.80915.37 × 10−43
Bra009099Unknown protein7.24322.24 × 10−219
Bra000362WRKY family transcription factor6.58433.17 × 10−68
Bra014659bHLH0386.24002.16 × 10−14
ZnDBra000247Unknown protein9.00724.82 × 10−9
Bra006681F-box family protein8.87732.83 × 10−8
Bra039194HMA18.04722.04 × 10−192
Bra019777DDF18.00600.18 × 10−3
Bra0330017.88840.23 × 10−3
Bra021948ZIP87.74860.50 × 10−3
Bra015415ZIP55.07505.08 × 10−7
Bra027108IRT34.56023.62 × 10−31
Bra031470IRT34.53021.62 × 10−32
Bra040152NAC domain containing protein 474.27550.34 × 10−3
Bra012263Zinc-binding family protein4.10621.37 × 10−5
Bra0327673.18495.24 × 10−11
Bra021965CYP707A33.16581.54 × 10−18
Bra003936AP2 domain containing transcription factor3.14104.99 × 10−13
Bra019248Short-chain dehydrogenase/reductase (SDR) family protein3.04080.95 × 10−13
ZnEBra004609bHLH10013.47741.17 × 10−198
Bra04095710.59231.89 × 10−26
Bra016946bHLH10010.43181.12 × 10−23
Bra001472MYB1010.14332.47 × 10−19
Bra000363Phagocytosis and cell motility protein ELMO1-related9.77785.26 × 10−15
Bra003055ORG19.59783.03 × 10−13
Bra0403029.54599.24 × 10−13
Bra011193TET99.04764.39 × 10−9
Bra0322348.94721.65 × 10−8
Bra004086Expressed protein8.13495.22 × 10−5
Bra015328AWPM-19-like membrane family protein7.83920.35 × 10−3
Bra022041ACD117.80950.44 × 10−3
Bra014657bHLH0387.45265.43 × 10−43
Bra014659bHLH0386.75491.04 × 10−20
CdEBra012938ERF104−8.22853.17 × 10−5
Bra012513Terpene synthase−4.25684.94 × 10−4
Bra007149Overexpression leads to PEL (Pseudo-Etiolation in Light) phenotype−3.74218.95 × 10−10
Bra021110Unknown protein−3.64711.57 × 10−31
Bra009464ZAT6−3.34575.30 × 10−7
Bra019002Unknown protein−3.34574.56 × 10−44
Bra038316RPK1−3.19450.58 × 10−3
Bra030673Glycine-rich protein−3.09270.13 × 10−3
Bra005447CIPK13−3.09011.90 × 10−4
 Bra039273Pyridoxine 5′-phosphate oxidase-related−3.01417.99 × 10−14
Bra010404Unknown protein−2.90082.82 × 10−5
Bra012254RNA recognition motif (RRM)-containing protein−2.60546.68 × 10−4
Bra027243Unknown protein−2.59894.44 × 10−30
Bra010237Unknown protein−2.56890.94 × 10−3
Bra011636SEN1−2.53915.18 × 10−48

Considering the overlap of DEGs under the four conditions, we identified 1647 nonredundant DEGs in total (Table S3). We used Gene Ontology (GO) annotations to classify the functions of the identified B. rapa DEGs. Among the 1647 DEGs, 987 could be assigned to GO categories (Table S3). According to the GO classification, 531 of the 987 DEGs were classified into the nine major functional categories, including transportation, stress response, signal transduction pathway, regulation, plant growth, photosynthesis, metal ion response, defense mechanism and cell related. The effects of trace element treatments on plants have been found to be related to these nine functional categories, particularly the transportation, metal ion response, regulation and signal transduction pathway categories in previous research (Wintz et al., 2003; Colangelo & Guerinot, 2004; Lanquar et al., 2004; Arrivault et al., 2006; Séguéla et al., 2008; Lingam et al., 2011; Sivitz et al., 2012; Aksoy et al., 2013). The information for the remaining 456 GO annotated DEGs was not clear enough to define their functional categories, which may have resulted in bias of the functional classification.

Then, we examined the number of DEGs in each GO functional category under FeD, ZnD, ZnE and CdE conditions (Fig. 1, Table S4). Amongst the 812, 543, 331 and 447 DEGs identified in the respective conditions, there were 285, 162, 119 and 133 DEGs, respectively, belonging to the nine major functional categories, and of these, 73.7%, 82.1%, 58.0% and 12.8%, respectively, were upregulated. The majority of the upregulated DEGs were involved in transportation (FeD, 75; ZnD, 63; ZnE, 30; CdE, 2), regulation (FeD, 127; ZnD, 61; ZnE, 38; CdE, 12) and metal ion response (FeD, 72; ZnD, 58; ZnE, 16; CdE, 3) (Fig. 1a–c, Table S4). Here, the number of DEGs in different GO functional categories was redundant because one gene with two or more GO IDs could be assigned to more than one category. The DEGs of CdE covered all nine functional categories, indicating that toxicity caused by Cd does not affect one particular biological process (Fig. 1d, Table S4).

Figure 1.

Number of differentially expressed genes (DEGs) in each Gene Ontology (GO) category under iron (Fe) deficiency (a), zinc (Zn) deficiency (b), Zn excess (c) and cadmium (Cd) exposure (d) conditions in Brassica rapa. Red, upregulated; green, downregulated.

Among the most differentially regulated FeD genes, six genes encoding transcription factors were highly induced, including bHLH proteins (bHLH100 and bHLH038), MYB10 and a WRKY family protein (Table 2). Twelve of the 15 most upregulated DEGs under FeD were also among the most upregulated DEGs under ZnE. For ZnD, the most upregulated DEGs included genes encoding metal transporters (HMA1, ZIP5, ZIP8 and IRT3) and genes involved in abiotic stress responses (CYP707A3, DDF1 and a gene encoding a NAC domain containing protein). Under CdE, genes related to stress responses and defense mechanisms (ERF104, ZAT6, RPK1, SEN1, a gene encoding terpene synthase and a gene encoding a DNAJ heat shock protein) were the most downregulated.

Expression patterns of DEGs under FeD, ZnD, ZnE and CdE conditions

The 1647 nonredundant DEGs were classified into 15 clusters based on their expression patterns under the four treatments (Fig. 2, Table S3). There were 28.5%, 20.6%, 8.5% and 18.1% DEGs specifically found under FeD (cluster XIII), ZnD (cluster XIV), ZnE (cluster XV) or CdE (cluster XII) conditions, respectively. The remaining 24.3% of the DEGs were grouped into the other 11 clusters as they were shared between or among the four conditions.

Figure 2.

Venn diagram showing unique and shared differentially expressed genes (DEGs) in Brassica rapa between or among the four trace metal element treatments: iron (Fe) deficiency (FeD), zinc (Zn) deficiency (ZnD), Zn excess (ZnE) and cadmium (Cd) exposure (CdE). The total number of DEGs under each condition is indicated below the treatment labels. The number and percentage of DEGs in each cluster is shown beside the Roman numeral group code.

Nearly 50% of the ZnE responsive DEGs were also found under FeD (Fig. 2), the vast majority (almost 97%) were regulated in the same direction by ZnE and FeD (Table S3), including all 117 genes in cluster X (DEGs shared between ZnE and FeD) (Table 3). Nearly 30% of the ZnD responsive genes were differentially expressed and regulated in the same direction under FeD (Fig. 1, Table 3). Most of them belonged to cluster IX (DEGs shared between FeD and ZnD), including genes encoding transcription factors (such as NAC domain containing proteins, MYB family proteins, BTB and TAZ domain proteins), protein kinase, zinc-finger family proteins and hormone associated proteins (Table S3).

Table 3. Directional comparisons of gene expression changes under iron (Fe) deficiency (FeD), zinc (Zn) deficiency (ZnD), Zn excess (ZnE) and cadmium (Cd) exposure (CdE) conditions in Brassica rapa
ClusteraTreatments caused the changes of gene expressionTotal no. of genesNo. of fitted genesCoverage of fitted genes (%)
  1. a

    Genes only responsive to one treatment were not included in this table.

  2. b

    ●● or ○○ indicate changes of gene expression in the same direction; ●○ indicates changes of gene expression in the opposite directions; ● image_n/nph12803-gra-0002.png or ● image_n/nph12803-gra-0001.png indicate changes of gene expression partly in the same direction and partly in the opposite direction.

II 383797.4
III  image_n/nph12803-gra-0001.png 141178.6
IV 33100
V image_n/nph12803-gra-0002.png  191684.2
VI  543870.4
VII  2222100
VIII  121083.3
IX  959095
X  117117100
XI  201785

Furthermore, we found that 25.1% (112/447) and 15.4% (69/447) of the CdE responsive DEGs were differentially expressed under FeD and ZnD conditions, respectively (Fig. 2). Because excess Cd may compete with Fe and Zn uptake and thus cause plants to experience Fe and Zn deficiency (Clemens, 2006; Wu et al., 2012), the overlapping DEGs were expected to be regulated in the same direction under these three conditions. However, in clusters I, II, IV, VI and VII the majority of the FeD or ZnD responsive genes were regulated in the opposite direction under CdE conditions (Table 3).

Confirmation of tag-mapped genes by quantitative RT-PCR

In order to confirm the differential expression of the DEGs under the four trace metal element treatment conditions, one DEG from each expression profile cluster was randomly selected for qRT-PCR analysis. Expression of the selected DEGs as determined by qRT-PCR fitted well with the expression as determined by Tag-Seq analysis (Fig. 3, Table S3). In addition, we checked the qRT-PCR expression profiles of several important genes known to be involved in trace metal element homeostasis (BrbHLH100, BrbHLH038, BrMYB10, BrHMA1, BrZIP8 and BrIRT3). These results were also consistent with the expression as determined by Tag-Seq analysis (Fig. S2).

Figure 3.

Expression analysis of selected differentially expressed genes (DEGs) in Brassica rapa by quantitative RT-PCR under control, iron (Fe) deficiency (FeD), zinc (Zn) deficiency (ZnD), Zn excess (ZnE) and cadmium (Cd) exposure (CdE) conditions. Control means normal culture conditions. = 3, and the values are means ± SD. The DEGs were selected randomly from the 15 groups classified according to their expression patterns. The gene IDs in order from group I to XV: unknown protein (Bra011686; I), ZAT6 (Bra009464; II), CAT2 (Bra034674; III), APRR5 (Bra009768; IV), FER3 (Bra007215; V), XTH33 (Bra018433; VI), Unknown protein (Bra006727; VII), Pyridoxine 5′-phosphate oxidase (Bra039273; VIII), CYP707A3 (Bra021965; IX), ZIF1 (Bra023432; X), APT2 (Bra003547; XI), unknown protein (Bra001421; XII), YSL1 (Bra013764; XIII), IRT3 (Bra027108; XIV) and TET9 (Bra011193; XV).

bHLH038 and bHLH100 centered gene network involved in metal homeostasis in B. rapa leaves under (ZnE-induced) FeD conditions

We examined the expression levels of the genes in B. rapa belonging to gene families known to play a role in Fe homeostasis in A. thaliana. As shown in Table 2, BrbHLH038 and BrbHLH100 were strongly expressed under FeD and ZnE conditions. In A. thaliana, bHLH038 and bHLH100 were both upregulated under FeD conditions and the transcription factors they encode were shown to participate in regulating several genes involved in Fe homeostasis such as AtFRO2, AtIRT1, AtNAS1, AtNAS2, AtNRAMP1, AtZIP9, as well as genes involved in heavy metal detoxification including HMA3, MTP3, IRT2 and IREG2 (Sivitz et al., 2012; Wu et al., 2012). bHLH038 interacts with FIT, another bHLH transcription factor (Colangelo & Guerinot, 2004), to directly or indirectly activate the expression of several FeD responsive genes in A. thaliana roots (Yuan et al., 2008; Wu et al., 2012), while bHLH100 functions independent of master regulator FIT in both shoots and roots (Sivitz et al., 2012). Consistent with previous results from A. thaliana (Colangelo & Guerinot, 2004), B. rapa homologs of FIT gene were not expressed (Bra011972) or expressed at a very low levels (Bra000495 and Bra034392, 0.1–0.4 TPM) in leaves under either ZnE or FeD conditions. There is no report on how bHLH038 regulates expression of the target genes in shoots. The upregulation of bHLH038 and bHLH100 in B. rapa leaves under FeD led us to consider whether there was a similar FeD responsive gene network as in A. thaliana roots, and if so, how it responds to ZnE conditions. Because research on the bhlh100/101 mutant indicated that the target genes of bHLH038 and bHLH100 were different (Sivitz et al., 2012), we clustered the B. rapa genes into two groups, namely the bHLH038-centered group and the bHLH100-centered group. The bHLH038-centered group included BrNAS3, BrNRAMP4 and BrFRO1 (Robinson et al., 1999; Waters et al., 2002; Lanquar et al., 2004; Oomen et al., 2009; Haydon et al., 2012). The bHLH100-centered group included BrZIF1 and BrMYB10 (Haydon & Cobbett, 2007; Sivitz et al., 2012). Our findings showed that genes from both groups were differentially expressed under (ZnE-induced) FeD conditions, among which BrNRAMP4, BrFRO1, BrZIF1 and BrMYB10 were highly expressed, while BrNAS3s showed low expression (Fig. 4).

Figure 4.

Relative expression of the genes hypothetically involved in BrbHLH038 or BrbHLH100 centered gene networks and the genes playing a role in zinc (Zn) or iron (Fe) homeostasis in Brassica rapa. Gene expression was determined by quantitative RT-PCR. ‘-1’, ‘-2’, ‘-3’,’-4'or ‘-5’ indicates the copy number of each gene. The metal ions related with the genes of interest are indicated above each gene name. The relative expression levels (Log2(TPMtreatment/TPMcontrol)) of each gene are color coded. A solid oval indicates a transcription factor, represented by multiple copies of bHLH038 or bHLH100. A dashed oval indicates a possible dimerization partner of bHLH038.

Expression of DEG paralogs in response to FeD, ZnD, ZnE and CdE exposure conditions

B. rapa has undergone triplication and subsequent fractionation of the genome, during which gene retention was biased (Cheng et al., 2012). To test whether there was biased gene retention of trace metal element responsive genes, we analyzed the retention rate of A. thaliana orthologs in B. rapa. Based on synteny analysis between A. thaliana and B. rapa (Wang et al., 2011), 1488 B. rapa DEGs were syntenic to 1317 A. thaliana genes. Amongst these 1317 A. thaliana genes, 388 (29.5%) had one syntenic ortholog and 929 (70.5%) had two or three syntenic orthologs in B. rapa. The ratio of multiple-copy to single-copy genes was significantly higher (= 1.0 × 10−84) for the detected trace element responsive genes (2.39; 929/388) than when considering the whole genome (0.745; 6836/9175; Wang et al., 2011), indicating that the trace element responsive genes were over-retained in the B. rapa genome.

We then examined the expression of B. rapa paralogous genes. To minimize the chance of missing orthologs that were not in synteny with A. thaliana genes, we used BLAST searching in addition to synteny analysis to identify the A. thaliana orthologs. We found that 1592 (96.7%) of the 1647 nonredundant B. rapa DEGs had orthologs in A. thaliana (Table S3). B. rapa genes with same A. thaliana ortholog were considered paralogous genes. We defined each DEG and its paralogs as a paralogous set. We identified 478, 332, 190 and 291 paralogous sets under FeD (including 578 DEGs), ZnD (including 367 DEGs), ZnE (including 212 DEGs) and CdE (including 314 DEGs) conditions, respectively (Table S5). Most of the paralogous sets contained only one DEG (Table 4), while the remaining 91 (19%), 32 (10%), 20 (11%) and 21 (7%) paralogous sets under FeD, ZnD, ZnE and CdE conditions, respectively, included two or three DEGs. Multiple DEGs from the same paralogous set were almost all expressed in the same direction, except for paralogous genes encoding alpha dioxygenase (AT1G73680; Bra015975, Bra003833) (Table 4). Further analysis on the expression of the paralogs showed that > 70% were expressed in the same direction as the detected DEGs, even if they were not significantly different compared with the normal conditions (Fig. 5). Co-regulation of the paralogous genes by trace metal elements indicates that they are functionally redundant at certain levels in B. rapa.

Table 4. Analysis of paralogous sets in which the expression trends of the differentially expressed genes (DEGs) were in the same direction in Brassica rapa
TreatmentsNo. of paralogous setsOne DEGTwo DEGsaThree DEGsb
  1. a

    The paralogous set included two DEGs.

  2. b

    The paralogous set included three DEGs.

  3. c

    The number of paralogous sets in which the expression trends of the DEGs were in the same direction. Outside the parenthesis is the number of paralogous sets that included two DEGs.

  4. Treatments: FeD, iron deficiency; ZnD, zinc deficiency; ZnE, zinc excess; CdE, cadmium exposure.

FeD47838780 (79)c11 (11)
ZnD33230029 (29)3 (3)
ZnE19017018 (18)2 (2)
CdE29127019 (19)2 (2)
Figure 5.

Percentage of paralogs whose expression trends were in the same direction (closed bars) or opposite direction (open bars) to the corresponding differentially expressed genes (DEGs) under iron (Fe) deficiency (FeD), zinc (Zn) deficiency (ZnD), Zn excess (ZnE) and cadmium (Cd) exposure (CdE) conditions in Brassica rapa.

Biased distribution of DEGs in the three B. rapa sub-genomes (LF, MF1 and MF2)

Whole genome analysis of B. rapa distinguished three sub-genomes by their degrees of gene loss, namely LF, MF1 and MF2 (Wang et al., 2011; Cheng et al., 2012). Amongst the 1647 nonredundant DEGs, 1636 were assigned to these sub-genomes (Table S3); 712 DEGs were assigned to the LF genome, 491 to MF1 and 433 to MF2. As expected, LF contained more DEGs than the other two sub-genomes (< 0.001). To determine whether genes from LF were also more responsive to trace element stress conditions, we used the ratio of the number of DEGs to the total gene number in each sub-genome (LF: 17 504; MF1: 12 543; MF2: 10 354; Wang et al., 2011) to evaluate whether there was a biased distribution of DEGs among the three sub-genomes. This showed that the ratio of LF DEGs was not significantly higher than that of the MF DEGs (= 0.610), which indicated that the higher proportion of DEGs in LF comes from biased gene retention in the LF sub-genome, rather than from the LF sub-genome playing a more important role in trace element homeostasis.


In this study, we have provided the first description of genome-wide expression profiles of genes in the leaves of Brassica rapa which are responsive to Zn and Fe deficiency, as well as Zn excess and Cd exposure.

FeD-induced genes are also involved in ZnE and ZnD responses

We found that many of the DEGs were shared among FeD, ZnD and ZnE, and often they were regulated in the same direction by these treatments (Fig. 2, Table 3). Before these experiments, we expected to find a large overlap of DEGs between FeD and ZnE because high Zn exposure had been found to induce FeD in Arabidopsis thaliana (van de Mortel et al., 2006). However, because previous research on wheat showed that Zn-deficient plants in general had a significantly higher shoot Fe concentration than controls (Imtiaz et al., 2003), we did not expect to find that 30% of the ZnD responsive genes were differentially expressed and regulated in the same direction under FeD. In nature, ZnD is often associated with FeD when plants are grown in calcareous soil. The HCO3 present in calcareous soil inhibits both Zn and Fe uptake and translocation to shoots (Marschner, 2012). According to our results, most of the genes responsive to both low Zn and low Fe encode transcription factors, protein kinases, zinc-finger proteins and hormone-related proteins (Table S3). It will be worthwhile to study plants grown in calcareous soil to determine whether the same genes are induced and unravel their roles in coping with the stress caused by this growing environment. Furthermore, the majority of the FeD or ZnD responsive genes were regulated in the opposite direction under CdE conditions. This is inconsistent with a previous report that excess Cd may compete with Fe and Zn uptake and thus cause plants to experience Fe and Zn deficiency. This might be because the Cd concentration (1 μM) we used caused physiological disruption in the treated plants but was not high enough to induce Cd toxicity compared with previous research on other Brassica species (Belimov et al., 2007; Van Engelen et al., 2011; Wan et al., 2011).

Enhanced expression of BrbHLH038 and BrbHLH100 may contribute to gene regulation in B. rapa leaves under (ZnE-induced) FeD conditions

In A. thaliana, many of the genes involved in Fe homeostasis are also highly expressed under Zn excess conditions, indicating competition of Zn at high concentrations with the Fe uptake mechanism, effectively causing Fe deficiency (van de Mortel et al., 2006, 2008). As shown in Table 2, BrbHLH038 and BrbHLH100 were strongly expressed in B. rapa leaves under FeD and ZnE conditions. This was in accordance with the similarly high expression of AtbHLH038 and AtbHLH100 in A. thaliana leaves upon low Fe or high Zn treatments (Wang et al., 2007). In A. thaliana, it has been suggested that there is a strict requirement for AtbHLH100/AtbHLH101 for genome-wide reprogramming in response to Fe deficiency via a FIT-independent pathway both in roots and shoots (Sivitz et al., 2012). In contrast to AtbHLH100, AtbHLH038 needs to dimerize with FIT in regulating Fe uptake in A. thaliana roots (Yuan et al., 2008; Wu et al., 2012). However, FIT is root-specifically expressed (Colangelo & Guerinot, 2004), and it is not known whether bHLH038 interacts with a protein similar to FIT to carry its regulation of gene network in shoots under Fe deficiency. Further investigation will be required to dissect the regulation pathway of bHLH038 and bHLH100 in B. rapa leaves.

Nicotianamine synthase (NAS) catalyzes the trimerization of S-adenosylmethionine (SAM) to form nicotianamine (NA), which is thought to play an essential role in the internal transportation of Fe (Higuchi et al., 2001) and Zn (Clemens et al., 2013b). Previous research indicated that AtNAS1 and AtNAS2 expression was induced by the co-expression of FIT with bHLH038 (Wu et al., 2012). In the present study, the expression of NAS1 and NAS2 was not enhanced under ZnE or FeD despite the induced expression of bHLH038, although one of the two copies of NAS3 was found to be repressed under ZnE and FeD. In A. thaliana, NAS3 is normally not expressed in leaves or is expressed at a very low level, but is highly induced by low Fe in roots (Wintz et al., 2003). Recently Palmer et al. (2013) reported that the transcript accumulation of NAS4 was upregulated by induced expression of transcription factors MYB10 and MYB72 under Fe deficiency. In line with this observation, MYB10 was also upregulated in B. rapa both under FeD and ZnE. However, transcription of NAS4 was not found to be induced in this study. These results together suggest that members of NAS gene family might be functionally differentiated between A. thaliana and B. rapa.

Plant signaling molecules modulate the (ZnE-induced) FeD and CdE response in B. rapa leaves

The biosynthesis of ethylene, auxin and nitric oxide is related to Fe acquisition in plants (Graziano et al., 2002; Chen et al., 2010; Lingam et al., 2011). In line with previous research, we observed that the majority of DEGs involving in the signal transduction pathway of these compounds were upregulated by Fe deficiency and Zn excess (Fig. 1a,c). For instance, genes functioning in the ethylene signaling pathway, such as ERFs, EBFs, EIN3 and EIL3, were highly expressed under FeD and ZnE (Table S3), indicating that the low Fe signal in leaves activated the ethylene responsive pathway to increase Fe content in the plants. EIN3, a transcription factor that regulates a series of ethylene responses from the vegetative to reproductive stage (Solano et al., 1998), is able to interact physically with FIT, and probably also bHLH038-type proteins, to trigger FeD responses by preventing these proteins from proteasomal degradation (Lingam et al., 2011). Hence, we deduced that the EIN3 encoded molecule in leaves probably helped to stabilize yet unknown regulators of FeD responses in leaves.

Cd is a toxic element in plants that negatively affects various physiological processes such as nutrient uptake, photosynthesis, development and cellular homeostasis, as we also observed in this study (Fig. 1). The plant response to Cd involves the synthesis of signal molecules including ethylene, abscisic acid and salicylic acid (DalCorso et al., 2008). ERF genes were induced after 2 h exposure to Cd in A. thaliana roots and shoots (Herbette et al., 2006; Weber et al., 2006). By contrast, in this study all DEGs related to signal transduction pathways were downregulated under CdE in B. rapa leaves (Fig. 1d). The main reason for this apparent inconsistency between our results and previous research is most likely the low concentration of Cd we applied in this study (1 μM). While these are concentrations comparable to what plants may experience at Cd contaminated sites, it was much lower than the concentrations used in other research (varying from 5 to 125 μM) (Herbette et al., 2006; Weber et al., 2006). Also differing from those experiments, we examined the gene expression profile of plants exposed to Cd treatment for 7 d – that is, a long-term response – whereas the work of Herbette et al. (2006) and Weber et al. (2006) used plant materials exposed to Cd for only a few hours. This assumption is further supported by the result recently reported by Cabot et al. (2013) that in A. thaliana plants exposed to 1 μM Cd the expression of BGL2, a marker of the salicylate signaling pathway, was either unaffected (20 h exposure) or decreased (7 d exposure), while treatment with 10 μM Cd caused enhanced expression after 7 d.

The ZIP and HMA gene families are important for the ZnD response in B. rapa leaves

Previously, several genes involved in Zn homeostasis were identified in A. thaliana (Becher et al., 2004; van de Mortel et al., 2006). Many metal transporters belonging to the ZIP and HMA gene families were highly induced by low Zn (van de Mortel et al., 2006). In particular, ZIP genes including AtZIP4, AtZIP5, AtZIP9 and AtIRT3 were strongly and specifically induced by Zn deficiency in roots and leaves (Grotz et al., 1998; Wintz et al., 2003; Talukdar, 2007; TAIR In B. rapa, we found that ZIP and HMA genes were highly induced by ZnD (Table S1), amongst which BrZIP5, BrZIP8, BrIRT3 and BrHMA1 were the most upregulated (Table 2). It is notable that both AtHMA1 and AtZIP8 were reported to function under excessive Zn conditions (Williams & Mills, 2005; van de Mortel et al., 2006; Kim et al., 2009). As there are two copies of HMA1 in B. rapa, and the one differentially expressed was a nonsyntenic ortholog (Bra039194), we performed phylogenetic analysis of HMA genes in B. rapa and A. thaliana to make sure that the gene was not functionally misclassified. The results showed that Bra039194 was more diverged compared with the syntenic ortholog (Bra011750) (Fig. S3a), but it was closer to AtHMA1 than the other members of the HMA family. In the case of ZIP8, which has three copies in B. rapa genome, the differentially expressed copy was a syntenic ortholog (Bra021948) and grouped most closely to AtZIP8 (Fig. S3b). These results suggest that HMA1 and ZIP8 in B. rapa might have been functionally diversified from their orthologs in A. thaliana during their evolution.

Genes associated with Zn, Fe and Cd responses are inclined to be over-retained in the B. rapa genome

Previous research showed that genes associated with important environmental factors were apt to be over-retained in the B. rapa genome during the gene loss process following the triplication event (Wang et al., 2011). Our results were in accordance with this finding. In theory, the high retention rate of duplicated genes allows the species to adapt to environmental stresses (Ha et al., 2009). In this study, even though certain paralogs were not significantly expressed under the trace element treatments, > 70% of them were expressed in the same direction as the paralogous DEGs. This included genes playing important roles in heavy metal responses such as BrbHLH038, BrbHLH100 and BrZIF1 (Haydon & Cobbett, 2007; Table S5). This suggests that the trace element responsive genes are not only over-retained at the gene dosage level, but also redundant at the transcriptional level. Nowak et al. (1997) classified genetic redundancy into three types, including true redundancy, generic redundancy and almost redundancy, according to the different levels of fitness when loss one of the duplicated genes. Our result that > 70% paralogs regulated by the Zn, Fe and Cd in the same direction but on different levels indicates they might be partially functional redundancy. A widespread view is that if the genes were fully redundant, then they would not be protected against accumulated deleterious mutations and most of them are not evolutionary stable. Recently, it was proposed that diversified expression patterns between duplicated genes are a major evolutionary mechanism to meet the enhanced demand for a wide spectrum of functionality across different developmental stages and differentiated cell types (Padawer et al., 2012). Further experiments in screening phenotypes of B. rapa plants carrying functional/nonfunctional alleles and examining gene expression levels in different tissues/cell types at different developmental stages are needed to clarify the metal responsive gene network.


The work was funded by the National High Technology R&D Program of China (2012AA100101) and the National Program on Key Basic Research Projects of China (The 973 Program: 2012CB113900 and 2013CB127000), as well as the Program Strategic Scientific Alliances between China and the Netherlands from the Royal Dutch Academy of Sciences (KNAW-project 08-PSA-BD-02). Research was carried out in the Key Laboratory of Biology and Genetic Improvement of Horticultural Crops, Ministry of Agriculture, China and the Sino-Dutch Joint Lab of Horticultural Genomics Technology in Beijing.