Genome-wide identification and comparative evolutionary analysis of the Dof transcription factor family in physic nut and castor bean

Datasets and sequence retrieval

As shown in Table S1, Arabidopsis and rice Dof genes described before were retrieved from The Arabidopsis Information Resource (TAIR10, http://www.arabidopsis.org/) and Rice Genome Annotation Project (RGAP7, http://rice.plantbiology.msu.edu/), respectively. Castor genomic sequences were accessed from Phytozome v12 (https://phytozome.jgi.doe.gov/pz/portal.html). The physic nut genome and other data such as nucleotides, Sanger expressed sequence tags (ESTs) and RNA-seq reads were obtained from NCBI (http://www.ncbi.nlm.nih.gov/, last accessed January 2018).

Identification and manual curation of Dof family genes

The hidden Markov model (HMM)-based profile of the Dof domain (PF02701) was obtained from the Pfam database (http://pfam.sanger.ac.uk/), and was used as a query to perform the HMMER search (Version 3.0, http://hmmer.janelia.org/). The resulted protein sequences were further adopted for the TBLASTN search (Altschul et al., 1997), and the positive genomic sequences with an E-value of less than 1e−5 were predicted as described before (Zou et al., 2015a). Predicted gene models were further validated with available cDNAs, ESTs, and RNA-seq reads. The presence of the Dof domain was confirmed using SMART (http://smart.embl-heidelberg.de/). Homology search for nucleotides or ESTs and expression annotation using RNA-seq data were performed as described before (Zou et al., 2015b).

Sequence alignment and phylogenetic analysis

Multiple sequence alignments were performed using MUSCLE (Edgar, 2004) and phylogenetic trees were constructed using MEGA (version 6.0) (Tamura et al., 2013) with the maximum likelihood method (bootstrap: 1,000 replicates). The alignment of Dof domains was displayed using Boxshade (http://www.ch.embnet.org/software/BOX_form.html), and gene structures were displayed using GSDS (Hu et al., 2015) via aligning the cDNA to the gene sequence. For synteny analysis, the all-to-all BLASTP method was used to identify duplicate pairs as previously described (Zou et al., 2018), and syntenic blocks were inferred using MCScanX (Wang et al., 2012). Considering the relatively fragmented status of the current castor genome assembly, MicroSyn (Cai et al., 2011) was also used to detect microsynteny. Best reciprocal hit (BRH)-based BLAST analysis described before (Zou et al., 2017a; Zou, Xie & Yang, 2017b) was also used to define orthologs across different species, where the E-value was set to 1e–10. Systematic names were assigned based on the phylogenetic analysis for castor, or chromosome distribution for physic nut.

Protein properties and conserved motif analysis

Theoretical molecular weight (MW), isoelectric point (pI), and grand average of hydropathicity (GRAVY) were determined using ProtParam (http://web.expasy.org/protparam/). Protein subcellular localization was predicted using CELLO (http://cello.life.nctu.edu.tw/). Conserved motifs in different Dof proteins were analyzed using MEME (http://meme-suite.org/tools/meme), and the optimized parameters were: any number of repetitions; maximum number of motifs, 15; and, the optimum width of each motif, between six and 100 residues. The MAST program (http://meme-suite.org/tools/mast) was also used to search detected motifs in protein databases.

Gene expression analysis

Global gene expression profiles over various tissues were investigated based on Illumina pair-ended RNA-seq samples, that is, expanding leaf (SAMEA1034168), developing male flower (SAMEA1034167), developing endosperm II/III (SAMEA1034165), developing endosperm V/VI (SAMEA1034166), developing seed (SAMN02688010), and germinating seed (SAMEA1034164) in castor; root (SAMN03160709), leafage (half expanded leaf, SAMN03160710), expanded leaf (SAMN03486846), flower bud (SAMN03733282), and seed (SAMN03160711) in physic nut. Expression profiles of JcDof genes during flower development were examined over five typical stages, that is, IND (undifferentiated inflorescence of 0.5 cm diameter, SAMN07525083), PID1 (carpel primordia beginning to differentiate, SAMN07525085), PID2 (three distinct carpels formed, SAMN07525084), STD1 (stamen primordia beginning to differentiate, SAMN07525082), and STD2 (10 complete stamens formed, SAMN07525086). JcDof genes involved in Colletotrichum gloeosporioides response were identified using SAMN02905749, SAMN02905750, SAMN02905751, and SAMN02905752. The filtered reads were mapped to the coding sequences (CDS) of Dof and other protein-coding genes using Bowtie 2 (Langmead & Salzberg, 2012), and the fragments per kilobase of exon per million fragments mapped (FPKM) method (Mortazavi et al., 2008) was used for the determination of transcript levels. To analyze differentially expressed genes (DEGs), the parameters “FDR< 0.001” and “log₂Ratio ≥1” were used. Unless stated otherwise, the tools used in this study were performed with default parameters.

Characterization of 25 Dof family genes in physic nut

As shown in Table 1, the search of the physic nut genome resulted in 25 Dof-coding loci from 16 out of the 6,023 scaffolds. Compared with the original genome annotation (Wu et al., 2015), one more locus was identified from scaffold241 (denoted JcDof6.1, see File S1). Interestingly, this gene was also identified by Wang et al. (2018), however, JcDof7.4, another member encoded by scaffold341, was not reported in their study. Although most scaffolds harbor a single Dof gene, four of them were found to contain more than one, that is, scaffold341 (4), scaffold684 (4), scaffold328 (3), and scaffold18 (2) (Table 1). Based on available genetic markers (Wu et al., 2015), these scaffolds can be further anchored to nine out of the 11 chromosomes (Chrs), only excluding Chromosomes 1 and 11. The gene distribution looks uneven: Chromosomes 4 and 10 contain a single Dof gene, whereas Chromosome 7 harbors the most of five (Fig. 1).

Homology search showed that 12 JcDof genes have EST hits in GenBank, that is, JcDof3.2, JcDof3.3, JcDof3.4, JcDof5.1, JcDof5.2, JcDof5.3, JcDof6.3, JcDof7.1, JcDof9.1, JcDof9.2, JcDof9.3, and JcDof10.1, where JcDof5.3 harbors maximum of five hits (Table 1). The full-length cDNA sequences of JcDof3.2 (i.e., FJ605173), JcDof5.1 (i.e., KF022019), JcDof5.2 (i.e., GQ256647; JcDof1), JcDof7.2 (i.e., KF022024; JcDof2), and JcDof9.1 (i.e., GQ256648 and KF022017; JcDof3) have also been reported in GenBank. Moreover, the expression of other family members was supported by RNA-seq reads derived from various tissue transcriptomes, including callus, root, leaf, flower, inflorescence meristem, seed, and embryo. Based on read alignment, the transcription regions of all JcDof genes were extended and four predicted gene models were also optimized (Table 1; Files S2–S5). The CDS of three genes (i.e., JcDof7.2, JcDof7.3, and JcDof7.4), which are located on scaffold341, were shown to exhibit relatively high identity, that is, 87.7% between JcDof7.2 and JcDof7.3, 87.4% between JcDof7.3 and JcDof7.4, and 97.4% between JcDof7.2 and JcDof7.4. They were defined as tandem duplications for their adjacent organization (Lee et al., 2013).

Characterization of 24 Dof family genes in castor

A total of 24 Dof-coding loci were identified from 21 castor scaffolds (Table 1). As observed in physic nut, most of these castor scaffolds contain a single Dof gene, whereas scaffold30138, scaffold30190, and scaffold29676 harbor two instead. The expression of all these genes was supported by available ESTs and/or RNA-seq reads. Compared with the automatic genome annotation (Chan et al., 2010), nine predicted gene models were manually optimized based on read alignment (see Files S6–S14). Conserved microsynteny was observed between castor and physic nut Dof-coding scaffolds. Although the current castor genome is fragmented in 25,763 scaffolds, 23 out of the 24 RcDof genes could be anchored to nine physic nut chromosomes based on synteny analysis (Fig. 1).

Phylogenetic analysis and definition of orthologous groups

To reveal the evolutionary relationships of physic nut and castor Dof genes, an unrooted phylogenetic tree was constructed using full-length Dof proteins. As shown in Fig. 2, 49 Dofs are split into nine groups representing four previously defined subfamilies (i.e., A, B, C, and D): Groups 1, 2, 4, 5, 7, 8, and 9 correspond to D1, D2, A, C2.2, C1, C2.1, or B1, whereas Groups 3 and 6 correspond to B2 as described in Arabidopsis (Lijavetzky, Carbonero & Vicente-Carbajosa, 2003). The classification is supported by analysis of exon-intron structures and conserved motifs as shown below.

To facilitate the transfer of functional information obtained in model plants, orthologs of Jc/RcDofs were further identified in Arabidopsis and rice, two well-studied species of dicotyledons or monocotyledons respectively. The detailed orthologous relationships are shown in Table 2, whereas 24 orthologous groups (OGs) were also proposed based on the definition of a certain OG with at least one member found in at least two out of four examined species. Genes whose functions have been characterized were also summarized in each OG that they belong. Nine phylogenetic groups identified above were shown to contain one to five OGs. Consistent with the phylogenetic analysis, species-specific gene loss was observed and several OGs were only found in physic nut and castor (i.e., OG-1b and OG-2c) (Table 2).

Exon-intron structures, sequence features, and conserved motifs

The exon-intron structure was analyzed based on optimized gene models and results showed that these genes contain no or a single intron (Table 1). Approximately 48.00% of JcDof genes and 43.48% of RcDof genes are intronless. Similar exon-intron structure was also observed in rice and Arabidopsis, though two-intron-containing genes (i.e., AtDof3.6 and AtDof3.7, accounting for 5.6%) were also found in Arabidopsis (Table S1). Generally, the exon-intron structure is highly conserved within a certain phylogenetic group especially within a certain OG: Groups 2–5 usually contain no intron, whereas Groups 1 and 6–9 feature one intron with the exception of OG-7c.

Deduced JcDof and RcDof proteins consist of 160–518 or 162–506 amino acid residues, respectively. The MW ranges from 18.22 to 55.72 kDa for JcDofs, or from 18.46 to 54.98 kDa for RcDofs. The predicted pI value ranges from 4.65 to 9.44 for JcDofs, or from 4.48 to 9.51 for RcDofs. Nevertheless, the average protein length, MW and pI value are similar between two species, that is, 323 residues, 34.92 kDa or 8.00 in physic nut; 334 residues, 36.23 kDa or 7.78 in castor. The GRAVY value of all Dof proteins is less than 0, indicating their hydrophilic feature. All these proteins were shown to harbor a single Dof domain with the Cx₂Cx₂₁Cx₂C zinc finger pattern, and were predicted to localize to the nucleus (Fig. 3).

Conserved motifs were also analyzed using MEME and results were shown in Fig. 2 and Fig. S1. Among 15 motifs identified, the broadly distributed Motif 1 is characterized as the Dof domain, in contrast, little is known about other motifs: Motif 8 is present in all members of Group 1, and several members of Groups 5 and 8; Motif 10 is widely present in Group 2, and several members of Groups 3 and 4; Motif 11 is present in all members of Group 1, and several members of Group 2; Motif 15 is present in all members of Group 9, and several members of Group 7; Motifs 2, 4, 5, 7, 9, 12 is limited to Group 1; Motifs 3 and 13 are limited to Group 2; Motif 6 is limited to Group 9; and, Motif 14 is limited to OG-1d of Group 1 (Fig. 2).

Gain or loss of certain motifs was observed within a certain group even between different OGs in the same group. Group 1 usually contains Motifs 8, 11, 1, 9, 4, 5, 12, 2, and 7 in orders; however, OG-1a only harbors Motifs 8 and 1, suggesting their early divergence; moreover, OG-1d harbors Motif 14 instead of Motifs 5 and 12 found in other OGs. OG-2b of Group 2 usually contains Motifs 11, 1, 13, 3, 10 in orders, in contrast, OG-2c only harbors Motif 1 and OG-2a has Motifs 1, 3, and 10, supporting this group can be further divided into three subgroups. Interestingly, JcDof7.3 has also lost Motif 10. OG-4b of Group 4 contains Motifs 1 and 10, whereas OG-4a only harbors Motif 1, supporting this group can be subdivided into two subgroups. OG-5b of Group 5 contains Motifs 1 and 8, whereas OG-5a only harbors Motif 1, supporting this group can also be subdivided into two subgroups. Group 8 usually features a single motif; however, OG-8c gained Motif 8 (Fig. 2).

Expression profiles of JcDof genes

Despite the expression of all JcDof genes, transcriptional profiling showed that JcDof6.1 was lowly expressed in all tissues examined. Based on the FPKM value, the total transcripts of JcDof genes were shown to be most abundant in flower bud, moderate in expanded leaf and leafage, and relatively low in seed and root. Most genes were expressed in all five examined tissues, though their transcript levels are variable. According to the expression pattern, these genes can be divided into five main clusters: Cluster I is preferentially expressed in root, where JcDof2.2 and JcDof6.3 were also abundant in leafage and flower bud; Cluster II is predominantly expressed in seed as well as high abundance in leafage; Cluster III is mostly expressed in leafage, where JcDof8.2 and JcDof9.1 were also abundant in expanded leaf; Cluster IV is typically expressed in expanded leaf, where JcDof3.4 and JcDof5.2 were also abundant in root or leafage respectively; Clusters V and VI are mostly expressed in flower bud, where Cluster V members were also highly expressed in seed (Fig. 4). Several key family members were also identified in a certain tissue: JcDof7.2 represents the most expressed gene in both flower bud and root; JcDof5.3 represents the most expressed gene in both expanded leaf and seed; and, JcDof9.1 represents the most expressed gene in leafage.

Low seed yield, which was caused by the lower number of female flowers, is a bottleneck for commercial-scale exploration of physic nut. Transcriptome profile analysis of five typical stages during floral sex differentiation revealed thousands of DEGs involved in ovule and stamen development (Hui et al., 2017). With a focus on Dof family genes, we found that the total transcripts were most abundant in IND, followed by PID2, STD1, STD2, and PID1 (Fig. 5). Compared with IND, eight genes were significantly downregulated in PID1; three genes were significantly regulated in PID2, including one upregulated and two downregulated; 10 genes were significantly downregulated in STD1; 11 genes were significantly regulated in STD2, including one upregulated and 10 downregulated. Compared with PID1, nine genes were significantly regulated in PID2, including one downregulated and eight upregulated. Compared with STD1, six genes were significantly regulated in STD2, including two upregulated and four downregulated (Fig. 5; Table S2). More remarkable, the expression pattern of JcDof9.3 suggests its key role in the later stage of both ovule and stamen development.

Anthracnose, caused by C. gloeosporioides, is an important disease causing damages on physic nut leaves, stems, and fruits (Ferreira-Pinto, Silva & Santos, 2011). To analyze the response of JcDof genes to this pathogen, two different genotypes (i.e., the susceptible RJ127 and the resistant 9-1) were artificially infected with pure isolate of C. gloeosporioides, and the induced leaves were collected at 2, 24, 96, and 144 h after infection and pooled for Illumina RNA-seq. Differential expression analysis showed that three JcDof genes were significantly regulated, that is, JcDof3.3, JcDof7.2, and JcDof9.3: JcDof3.3 was upregulated in both cultivars, whereas JcDof7.2 and JcDof9.3 exhibit an opposite trend (Fig. 6), implying something different regulation mechanism of these two cultivars.

Tissue-specific expression profiles of RcDof genes

Similar to its ortholog in physic nut (i.e., JcDof6.1), the transcript level of RcDof14 was extremely low in all examined tissues, that is, leaf, male flower, endosperm I/III, endosperm V/VI, developing seed, and germinating seed. The total transcripts were most abundant in leaf (Class I); followed by male flower and developing seed (Class II, accounting for 79.3–79.4% of Class I); moderate in endosperm I/III and germinating seed (Class III, accounting for 38.2–56.0% of Class I); and, relatively low in endosperm V/VI (Class IV, accounting for 13.7% of Class I). Based on their expression patterns, these genes were grouped into seven main clusters: Cluster I is mostly expressed in germinating seed, including RcDof04 and RcDof12, which were also abundant in male flower or endosperm I/III respectively; Cluster II is preferentially expressed in developing seed; Cluster III is predominantly expressed in endosperm I/III, where RcDof19, RcDof06, and RcDof18 were also abundant in developing seed; Cluster IV is typically expressed in male flower; Cluster V is preferentially expressed in leaf and male flower; Cluster VI is preferentially expressed in leaf and developing seed; and, Cluster VII is preferentially expressed in leaf. As for a specific tissue, several key isoforms were also identified: RcDof09 represents the most and second most expressed gene in leaf or male flower, respectively; RcDof13 represents the most expressed gene in male flower; RcDof19 represents the most expressed gene in both developing seed and endosperm I/III; RcDof12 represents the most expressed gene in both germinating seed and endosperm V/VI; RcDof18 represents the second most expressed genes in germinating seed, endosperm I/III and endosperm V/VI; RcDof18 represents the second most expressed genes in developing seed (Fig. 7).