Comprehensive phylogenomic analysis of ERF genes in sorghum provides clues to the evolution of gene functions and redundancy among gene family members

Abstract

APETALA2/Ethylene-Responsive transcription factors (AP2/ERF), with their multifunctional roles in plant development, hormone signaling and stress tolerance, are important candidates for engineering crop plants. Here, we report identification and analysis of gene structure, phylogenetic distribution, expression, chromosomal localization and cis-acting promoter analysis of AP2/ERF genes in the C4 crop plant sorghum. We identified 158 ERF genes in sorghum with 52 of them encoding dehydration-responsive binding elements (DREB) while 106 code for ERF subfamily proteins. Phylogenetic analysis organized sorghum ERF proteins into 11 distinct groups exhibiting clade-specific expansion. About 68% ERF genes have paralogs indicating gene duplications as major cause of expansion of ERF family in sorghum. Analysis of spatiotemporal expression patterns using publicly available data revealed their tissue/genotype-preferential accumulation. In addition, 40 ERF genes exhibited differential accumulation in response to heat and/or drought stress. About 25% of the segmental gene pairs and eleven tandem duplicated genes exhibited high correlation (> 0.7) in their expression patterns indicating genetic redundancy. Comparative phylogenomic analysis of sorghum ERFs with 74 genetically characterized ERF genes from other plant species provided significant clues to sorghum ERF functions. Overall data generated here provides an overview of evolutionary relationship among ERF gene family members in sorghum and with respect to previously characterized ERF genes from other plant species. This information will be instrumental in initiating functional genomic studies of ERF candidates in sorghum.

Electronic supplementary material

The online version of this article (10.1007/s13205-020-2120-y) contains supplementary material, which is available to authorized users.

Introduction

Sorghum, a valuable model for C4 grass research, is grown for food, forage, and feedstock, worldwide (Mathur et al. 2017; Ahmad Dar et al. 2018). Its extraordinary ability to thrive under hot and dry conditions makes it suitable for planting on marginal lands (Lukas et al. 2017). However, with the continuously changing climatic patterns and limited freshwater resources, ensuring high quality and quantity of the biomass and/or grain yield would require targeted molecular engineering of the available sorghum cultivars.

Transcription factors have been the most important targets for crop engineering due to their role as master regulators in diverse biological processes (Century et al. 2008; Kaufmann and Airoldi 2018). APETALA2-Ethylene Responsive Factors (AP2-ERFs) are plant-specific transcriptional regulators characterized by one or more DNA binding AP2/ERF domains of approximately 60–70 amino acids. AP2-ERF superfamily has been classified into three families namely, AP2 (APETALA2), ERFs and RAVs (Nakano et al. 2006; Gu et al. 2017). AP2 family proteins contain two tandemly repeated AP2/ERF domains, whereas, RAV family members contain a B3 domain in addition to a single AP2/ERF domain. ERFs, on the other hand, possess a single AP2/ERF domain and have been further divided into CBF/DREB (C-repeat binding factor/ dehydration-responsive element binding) and ERF (Ethylene responsive transcription factor) subfamilies (Nakano et al. 2006; Licausi et al. 2013). Sequence alignment of AP2/ERF domains of DREB and ERF proteins revealed different degrees of amino acid conservation in the two sub-families. However, amino acid residues, reported as indispensable for DNA binding in DREB and ERF families, were highly conserved.

ERF family genes have received a lot of attention over the past two decades as overexpression of several ERF/DREB genes across different plant species led to broad-spectrum resistance to pathogens and/or improved abiotic stress tolerance in transgenic plants (Phukan et al. 2017; Xu et al. 2011). Genes of CBF/DREB subfamily have been shown to regulate abiotic stress response by recognizing A/GCCGAC (DRE/CRT; Dehydration-Responsive or C-Repeat element) in the regulatory regions of target genes, whereas, ERF genes mainly regulate biotic stress response by binding to the GCC box (Hao et al. 1998; Thomashow 1999; Tang et al. 2017; Dey and Corina Vlot 2015; Bihani et al. 2011). However recently, several ERF and DREB proteins have been shown to regulate both abiotic and biotic stress response (Phukan et al. 2017; Jisha et al. 2015; Tian et al. 2015). With their ability to coordinate multiple signaling and hormonal pathways, ERFs are excellent candidates for engineering biotic and abiotic stress tolerance in crop plants (Xu et al. 2011; Gu et al. 2017).

Considering their importance in crop improvement, genome-wide investigation of ERF family has been undertaken in several crop species with the number of ERF family genes ranging from 96 in strawberry to 292 in maize (Srivastava and Kumar 2019). With the expansion of gene families due to large number of gene duplications in plant genomes, identification of gene family members at whole genome scale followed by structural and expression analysis is crucial for prioritizing candidates and choosing appropriate strategies for functional characterization (Qiao et al. 2018; Jung et al. 2015). Earlier, Yan and coworkers (2013) reported 105 ERF genes in sorghum. However, with the availability of more sequencing data and recent update in the sorghum genome annotation (McCormick et al. 2018), a detailed investigation of the complete ERF gene family in sorghum was lacking. To fill this gap, we performed genome-wide identification and in-depth analysis of ERF genes in sorghum. With the addition of 53 newly identified ERF members, a total of 158 unique ERF genes were annotated. Further, to gain insights into their modus operandi, genome-scale phylogenomic analysis of sorghum ERFs was performed. The insights gained from structural, subcellular localization, phylogenetic, duplication, expression and promoter analysis combined with the knowledge available about the functions of the genetically characterized ERF genes in other plant species provided clues to ERF gene functions in sorghum.

Materials and methods

Identification of AP2/ERF genes in sorghum

AP2/ERF transcription factor sequences, extracted from Plant Transcription Factor Database v3.0 (PTFDB, https://planttfdb.cbi.pku.edu.cn/), were used to generate HMM profile for AP2 and ERF families through HMMER 3.0 tool using default parameters (Finn et al. 2011). Sorghum bicolor proteome v3.1, available at Phytozome database (https://phytozome.jgi.doe.gov/pz/portal.html), was searched using the in-house generated HMM profiles to identify sorghum AP2/ERF genes. After consolidating the resulting sequences with AP2/ERF transcription factor genes available in PTFDB, redundant sequences were removed. Unique sequences were evaluated for the presence of the AP2-ERF domain using a web-based NCBI Conserved Domain Database (NCBI-CDD; https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi) and SMART domain analysis (https://smart.embl-heidelberg.de/) tools. Based on the domain distribution, these were categorized into AP2, ERF and RAV proteins. Further, unique sequences were also compared with those reported by Yan and colleagues (2013) and the nomenclature of ERFs, proposed by them, was extended to name the newly identified members. A list of ERF proteins with known functions in diverse plant species was compiled using manual literature search. The sequence information for characterized genes was downloaded from NCBI.

Subcellular localization

To predict the subcellular localization of SbERF proteins, we used four independent online prediction tools including WoLF-PSORT (https://www.genscript.com/wolf-psort.html), DISTILL (https://distill.ucd.ie/distill/), MultiLoc2 (https://abi.inf.uni-tuebingen.de/Services/MultiLoc2) and CELLO v2.5 (https://cello.life.nctu.edu.tw/). Only high confidence predictions were considered for prediction of localization using the following thresholds: WoLFPSORT: confidence score > 12, MultiLoc2: confidence score > 0.7, DISTILL: “High” confidence predictions and CELLO v 2.5: score > 3.0.

Phylogenetic analysis

Protein sequences of rice ERF family, as reported by Nakano and workers (Nakano et al. 2006), were downloaded from Phytozome. AP2/ERF domain sequences were retrieved from all the rice and sorghum ERF proteins using the NCBI CDD tool (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi). Multiple sequence alignment of AP2/ERF domain sequences was performed using the MAFFT version 7 program (Katoh and Standley 2013) using local alignment parameters. We used MEGA 6 (Tamura et al. 2013) to generate an unrooted phylogenetic tree using Neighbor-joining method. A separate tree was constructed for AP2/ERF domains of only sorghum ERF proteins using the same parameters. A conjoined phylogenetic tree comprising of AP2/ERF domain sequences from sorghum and 74 genetically characterized ERF proteins from diverse plant species were also generated as described above. The phylogenetic groups were named as earlier (Nakano et al. 2006).

Gene structure analysis, identification of conserved motifs and cis-acting regulatory elements prediction

The coding and genomic sequences of SbERF genes and information about the number of introns as well as alternative transcripts were obtained from Phytozome. The gene structures were illustrated using Gene Structure Display Server v2 (Hu et al. 2015). Further, to identify additional conserved motifs in AP2/ERF proteins, we used online MEME (Multiple Expectation–Maximization for Motif Elicitation) motif identification tool version 4.11.1 (https://meme-suite.org/tools/meme) using following parameters: motif length: 6–50 amino acids with any number of repetitions and the maximum number of motifs set to 15. The functional significance of MEME motifs was analyzed using the TOMTOM tool (Gupta et al. 2007). We used the MAST tool to examine the presence of these motifs in the ERF proteins with known functions from other plant species. Weblogo 3.6.0 was used to create an ERF domain logo for ERF and DREB proteins. Promoter sequences (1.5 Kb) of sorghum ERF genes were extracted from Phytozome v12.1 (https://phytozome.jgi.doe.gov/pz/portal.html) using Biomart tool. Various cis-acting regulatory element sequences were collected from the literature. ERF promoter sequences were investigated for the presence of these cis-regulatory elements with Plant Promoter Analysis Navigator (PlantPAN 2.0; Chow et al. 2015). Data were recorded in table format and analysed.

Chromosomal localization, duplication analysis and identification of orthologs of SbERF genes

Gene coordinates for SbERF genes were obtained from the Phytozome and localized on respective chromosomes using Mapchart 3.0 (Voorrips 2002). Segmental duplication data for Sorghum bicolor was obtained from the Plant Genome Duplication Database (https://chibba.agtec.uga.edu/duplication/) and visualized using Circos (Krzywinski et al. 2009). Tandem duplications were identified using Plant Tandem Duplicated Genes Database [(PTGBase; https://ocri-genomics.org/PTGBase/; (Yu et al. 2015)] supplemented with manual analysis where ERF genes lying in tandem on sorghum chromosomes with > 40% protein identity were considered as potential duplicates. Ortholog identification was performed using the Inparanoid standalone tool version 4.1 (Sonnhammer and Ostlund 2015).

Expression profiling

Publicly available microarray and RNA seq data were used for expression profiling of sorghum ERF genes in diverse developmental tissues and in response to abiotic stress treatments. Pre-normalized Affymetrix-based microarray data for six diverse genotypes representing grain, sweet, forage and bioenergy sorghum available at NCBI-GEO (GSE49879) was used for expression profiling during vegetative stages of development in grain sorghum and analyzes the genotype-specific expression of SbERF genes. Similarly, data generated in response to heat and/or drought stress, using Agilent 28 K array (GSE48205) was used for identifying abiotic stress-responsive ERFs in sorghum by extracting information about ERF genes from the supplementary data provided (Johnson et al. 2014). RNA seq data (TPJ5005), generated by Davidson and coworkers (2012), was used for comparative transcriptomic profiling of ERF family genes in rice and sorghum by fetching data on RPKM values for ERF genes.

qRT-PCR analysis

Mature leaves (fourth from top) and middle stem internodes were sampled from booting stage of Sorghum bicolor genotype M35 plants. Total RNA was isolated from three independent biological replicates using TRIzol reagent (Invitrogen). After genomic DNA removal using Turbo DNA free kit (Invitrogen), RNA was converted to cDNA using iScript cDNA synthesis kit (BioRad). qRT-PCR was conducted on a CFX96 Real-Time System (Biorad) with Brilliant III UltraFast SYBR green qPCR mix (Agilent Technologies). Reaction conditions were as follows: Initial denaturation at 95 °C for 2 min followed by 40 cycles of denaturation at 95 °C for 15 s, annealing at 59–60 °C for 30 s and extension at 72 °C for 20 s, followed by melt curve analysis to ensure primer specificity. Three technical replicates for each biological replicate were analyzed. Relative expression of ten ERF transcription factor encoding genes was evaluated in stem and leaf tissues using the double delta CT (ΔΔCT) method. Eukaryotic Initiation Factor 4α (eIF4α) was used as an internal control and expression values were expressed as mean ± SD. while ERF100 that exhibited lowest expression in both the tissues was used as a reference for calculating relative expression of other genes in stem and leaf tissues.

Results and discussion

Identification and structural analysis of ERF proteins

We have identified a total of 184 unique AP2-ERF sequences from sorghum. Among these, 22 proteins containing two AP2-ERF domains each were annotated as AP2 family transcription factors while four genes containing single AP2-ERF domain along with a B3 domain were grouped into RAV subfamily. Remaining 158 sequences containing a single AP2-ERF domain were annotated as ERF family proteins (Supplementary Table S1). A comparison of manually annotated ERF sequences with the recent version of Phytozome revealed a few discrepancies in gene annotations. The sequence corresponding to Sobic.001G320600 (SbERF108) does not contain the ERF domain and Sobic.005G077532 (SbERF059) is no more annotated as a protein-coding gene in the recent version of Phytozome (v3.1). However, manual curation confirmed the presence of an ERF domain in both the sequences. Since expression evidence is also available in the form of ESTs or RNA seq reads for both the sequences, these were included in subsequent analysis. Yan and co-workers (2013) had earlier reported 105 ERF sequences from sorghum and named them from SbERF001 to SbERF105. Supplementing their nomenclature, newly identified 53 genes were named from SbER106 to SbERF159 (Supplementary Table S1). SbERF099, annotated as ERF protein by Yan and colleagues (2013), had two AP2/ERF domains and, therefore, was excluded from further analysis.

More than 60% ERF genes (100 in number) were intron-less while 38 had a single intron (Supplementary Fig. S1). Remaining 20 genes contained two or more introns. Presence of long, multiple introns and splicing influence transcriptional output exhibited in the form of delayed expression, whereas, fewer or smaller introns lead to efficient expression (Heyn et al. 2015; Jeffares et al. 2008). Hence, a large number of intron-less ERF genes could be attributed to their role in swift response to the environment. The presence of multiple introns, on the other hand, may cater to multiple functions through different protein isoforms resulting due to alternative splicing. For instance, SbERF091 a DREB subfamily member has four alternative transcripts. Rice ortholog (OsDREB2B) of SbERF091 is temperature and drought responsive. It exhibits alternative splicing wherein on exposure to high temperature an alternative isoform predominates, which leads to the formation of a functional protein (Staiger and Brown 2013). SbERF091 may also employ similar mechanism to cater to its diverse functions.

SbERF048 comprising 639 amino acids is the longest protein in the family while the shortest ERF gene, SbERF106, codes for 103 amino acids. The molecular weights of ERF proteins range from 11 to 68 kDa. The subcellular localization of all ERFs was predicted using four different databases (Supplementary Table S2). Based on the high confidence collated predictions, the majority of ERF proteins (128 in number) were predicted to localize in the nucleus. Three proteins including SbERF030, 158 and 159 were predicted to localize in chloroplasts while SbERF150 was predicted to localize in mitochondria. The localization results for the rest of the ERF proteins indicate that they likely shuttle between two or more locations in the cell. Subcellular localization also provides important clues regarding gene functions. Previously, ERF genes have been shown to be involved in plastid-nucleus retrograde signaling (Wu et al. 2018; Leon et al. 2012), therefore, it would be interesting to determine the functions of ERF genes predicted to localize in chloroplasts.

Phylogenetic and motif analysis of SbERF proteins

Further, to gain insights into phylogenetic relationship and domain evolution in sorghum ERF proteins, ERF domain sequences were extracted from 133 rice and 158 sorghum ERF proteins. ERF domains of three of the sorghum ERF proteins (SbERF019, 107 and 136), however, could not be aligned with other domains due to low sequence similarity and were, therefore, not used for phylogenetic analysis. A combined phylogenetic tree using both rice and sorghum ERF domains was constructed (Fig. 1). Out of the eleven phylogenetic groups, identified earlier by Nakano and coworkers (2006) in rice, ten could be clearly demarcated in the combined phylogenetic tree. Though rice group XI merged with sorghum group V, some of the rice proteins from this clade formed a separate clade with sorghum proteins. This new clade was named group XI^* to differentiate it from previously identified group XI in rice (Nakano et al. 2006).

Fig. 1.

Phylogenetic analysis of ERF family proteins in sorghum and rice. The phylogenetic groups as defined by Nakano and coworkers (2006) are marked using the Roman numerals. Groups I to IV comprise DREB subfamily and groups V to XI* are ERF subfamily proteins

ERF family comprises of two subfamilies named, ERFs and DREBs. Both groups could be clearly demarcated in the tree with groups I to IV corresponding to DREB subfamily while groups V to XI comprised ERF subfamily of ERFs (Fig. 1 and Table 1). Most of the clades contained members from both the species indicating a common origin with a few exceptions. For example, Group XI^* was split into two separate clades in the combined tree. Rice and sorghum proteins in one of these clades formed distinct subclades corresponding to each species indicating possible structural and functional divergence. We also observed that several clades, especially in groups II, III, VIII, IX, and XI contained an uneven number of proteins in rice and sorghum indicating clade-specific expansion post rice-sorghum divergence. Nine proteins, SbERF003, 5, 17, 30, 66, 67, 68, 90 and 104, that were assigned to the DREB subfamily by Yan et al. (2013) grouped with ERF subfamily in our study. This difference may have occurred due to updated gene annotations. Nevertheless, the sequence alignment revealed that these have conserved residues characteristic to the ERF subfamily.

Table 1.

Summary of phylogenetic placement of SbERF genes

Subfamily	Phylogenetic group	No of SbERFs
DREB	I	8
	II	11
	III	26
	IV	7
ERF	V	10
	VI	17
	VII	13
	VIII	18
	IX	26
	X	13
	XI	6

Since DREBs and ERFs differ in the amino acid composition of respective AP2/ERF domains and their binding specificities (Licausi et al. 2013), a separate sequence logo for the domains extracted from all ERFs and DREBs was generated (Fig. 2). The height of each letter in the logo is proportional to the observed frequency of each amino acid in all the sequences analyzed corresponding to the level of conservation across subfamily members. The region corresponding to a three-stranded beta-sheet and an alpha helix was marked on each logo (Fig. 2). Arginine and aspartic acid residues in the second strand of the beta-sheet of ERFs are conserved in most of the ERF sequences (marked by a red arrow). Similarly, among DREBs, valine at the 14th position is largely conserved while position 19 has a high frequency of glutamic acid. Most of the amino acids, previously implicated in DNA binding (marked by an asterisk) in the ERF domain, were conserved in both the subfamilies. A “WLG” element in the third beta-strand was also conserved in both DREB and ERF subfamily proteins.

Fig. 2.

Sequence logo of AP2/ERF domains extracted from a ERF and b DREB proteins. The beta-strands and alpha-helix is marked using yellow bars. Red asterisks represent conserved amino acids required for DNA binding. The conserved “WLG” motif is highlighted by the rectangular red box

The reliability of the phylogenetic grouping of ERF proteins was further supported by motif analysis of the ERF proteins using the MEME motif analysis tool (Fig. 3). A total of 15 motifs were identified from sorghum ERF proteins and for each motif, a separate sequence logo was generated (Supplementary Fig. S2). Motifs 1–4 correspond to the AP2-ERF domain and, were conserved across the family except for group XI^* members indicating divergence of AP2-ERF domain in group XI^* proteins. Overall, most of the members belonging to a particular clade share a similar motif organization indicating the coevolution of the ERF domain with the rest of the protein sequence. The demarcation between DREB and ERF proteins was also obvious from the motif analysis with motifs 5, 8, 9 and 12 exclusively present in DREBs and motifs 6, 7, 11, 14 and 15 being specific to ERF subfamily proteins. Several motifs were exclusive to specific phylogenetic groups indicating their role in the functional specialization. For instance, motif 11 was exclusive to group VII, whereas, motifs 9 and 10 were specifically present in genes governing abscisic acid-mediated drought responses. Motif 15 was exclusively detected from tomato stress-responsive factor TSRF1, which regulates pathogen resistance to Ralstonia solanacearum. These results indicate that both the ERF domain as well as domains outside the ERF domain have undergone distinct evolutionary changes in both the subfamilies.

Fig. 3.

MEME motif analysis of ERF family proteins in sorghum. Colored boxes designate the meme motifs and illustrations of meme motifs present in ERFs proteins are arranged group-wise for easy identification of group-specific motifs. The color legend for motifs is provided at the base. In total, 15 MEME Motifs were identified in SbERF proteins, where Motif 1–4 represent the AP2/ERF domain. Detailed structure of each motif is given in supplementary Fig. 2

Among DREBs, motif 9 was exclusively present in group I proteins. Its taxonomic distribution showed that it is specific to monocots, whereas, motifs 5 and 12 were exclusive to members of group II and III. Motif 5 was specifically rich in proline and alanine, and these residues were conserved at several positions in the motif (Supplementary Fig. 2). Motif 8 was exclusive to group III proteins and its taxonomic distribution in the NCBI protein database revealed it to be monocot exclusive. A conserved “LWSY” motif was identified at the C-terminal region of some of group III members. This motif has earlier been reported in OsDREB1A/B/C and in AtCBF3/DREB1A as well (Dubouzet et al. 2003). LWSY motif has been earlier implicated in transcriptional activation and cold stress response (Jin et, Liu 2008; Yan et al. 2013). Among ERFs, group VI members contain a 22-residue motif (motif 14) with conserved aspartic acids at positions 9, 11 and 14. Motif 11, comprising 12 residues, is exclusively present in Group VII members. N-terminal region of the motif 14 “MCGGAI/L” is highly conserved and has been characterized as an MC motif. It is required for protein stability under hypoxia and is essential for oxygen-mediated targeted proteolysis via the N-end rule pathway (Gibbs et al. 2011; Licausi et al. 2011; Wei et al. 2019). Overexpression of ERF genes containing MC motif has been shown to provide tolerance to oxygen-deprivation stress (Xu et al. 2006; Licausi et al. 2011; Wei et al. 2019). Motif 10, identified from several DREB and ERF proteins, is specifically rich in glutamine (Q), whereas, motif 5 contains several alanine residues (Supplementary Fig. 2). The presence of poly (Q) motif stabilizes the protein–protein interactions and has been associated with protein aggregation in plants as well as animals (Schaefer et al. 2012; Liu et al. 2011). Conversely, both poly(Q) and poly (A) motifs have been associated with molecular pathogenesis responsible for several diseases in humans (Pelassa et al. 2014). Motif 13 is specifically rich in serine residues, which is the third most common disorder-promoting amino acid residue (Uversky 2015). Previously it has been shown that poly serine repeats serve as flexible linkers and the presence of disordered regions enhances their functionality by expanding their potential to interact with multiple partners. Therefore, SbERF proteins having motif 13, may be involved in multiple pathways. Motifs 6, 7 and 15 do not show any resemblance to characterized motifs and are exclusive to group IX proteins.

Expression profiling of SbERF genes in diverse tissue types, genotypes and abiotic stress treatments

We analyzed the expression of SbERF genes in diverse tissue types, genotypes and in response to abiotic stress treatments using publicly available microarray and RNA sequencing data. The expression of SbERFs was examined in seedling roots and shoots, shoot tips, leaves and internodes using microarray-based expression data available for grain sorghum genotype R159 (Fig. 4, Supplementary Table S3). The DREB genes from groups I and II, in general, exhibited higher expression in seedling shoots, leaves, and basal internodes. Whereas, those belonging to group III and IV had predominant expression in seedling shoots and basal internodes. Though many ERF genes exhibited high expression in all the analyzed tissues, several genes showed tissue-specific accumulation. For example, SbERF011, 31 and 62 were predominantly expressed in leaves. Similarly, SbERF005 and SbERF082 had the highest expression in shoot tips while SbERF023 exhibited the highest expression in basal internodes. SbERF019, 20, 81, 94 and 119 were expressed at very low levels across all the tissues analyzed here (Fig. 4, Supplementary Table S3). To study the genotype-specific expression of ERF genes, we compared microarray-based expression data from various tissues in diverse sorghum genotypes. Data indicated tissue-specific enrichment of some of the ERF genes in a genotype-dependent manner. For example, SbERF10 was specifically expressed in shoot tips of high biomass/bioenergy lines, PI152611 and AR2400 and leaf tissue of PI455230 (Supplementary Fig. S3, Supplementary Table S4).

Fig. 4.

Expression profiling of sorghum ERF genes in developmental tissues. The tissues are marked on the top of heatmap. Internode T Top Internode and Internode B Bottom internode. The color legend is provided at the base where green represents low expression, red high expression and black medium level expression. The duplicated genes exhibiting expression correlation > 0.7 are marked with the blue rectangular boxes

Further, to study the effect of abiotic stress conditions on ERF gene expression, we analyzed microarray-based expression data from grain sorghum in response to heat, drought and, combined heat & drought stress. Out of 121 genes that were represented on the array, 40 SbERF genes were upregulated and 14 were downregulated in response to at least one of the stress conditions (Fig. 5; Supplementary Table S5). Very little overlap was observed in the genes induced or downregulated by these stress treatments (Fig. 5a) thereby, suggesting that SbERF genes were specifically regulated in response to the respective stress treatments. For example, drought stress exclusively stimulated higher expression from SbERF097 and 98. Likewise, SbERF015, 28, 37 and 96 were induced specifically in response to heat stress, whereas, combined stress led to exclusive upregulation of nine SbERF genes namely, SbERF005, 13, 16, 50, 51, 64, 75, 79 and 89. SbERF100 was the only ERF member that was upregulated in response to all three stress conditions, whereas, two genes SbERF010 and 11 were downregulated in response to all three stress treatments. SbERF147 was induced by more than 300 folds in response to drought stress and > 100 folds in response to combined heat and drought stress (Fig. 5b). SbERF40 exhibited maximum induction (> 1000 folds upregulation) in response to heat stress (Fig. 5c) with 30 folds induction in response to combined heat and drought stress treatments (Fig. 5d). Promoters of SbERF040 and 147 contain cis-regulatory elements related to dehydration responses. SbERF001, 63 and 83 exhibited the highest level of induction in response to combined heat and drought stress (Fig. 5d). While SbERF001 promoter is rich in cis-elements related to dehydration response, the SbERF063 promoter is rich in cis-elements related to hormonal response.

Fig. 5.

Expression profiling of sorghum ERF genes in response to abiotic stress treatments. a Venn diagram showing overlap in the number of up and down-regulated genes in response to drought, heat or simultaneous heat and drought treatments. b Bar diagram showing log₂ fold changes of ERF genes in response to drought stress. c Bar diagram showing log₂ fold changes of ERF genes in response to heat stress. d Bar diagram showing log₂ fold changes of ERF genes in response to simultaneous drought and heat stress

Identification and analysis of expression divergence among orthologs and paralogs

The expression profiles of all SbERF genes and their rice orthologs were extracted from transcriptomics data (Davidson et al. 2012) available for diverse developmental tissues including stigma, anther, pistil, embryo, endosperm, leaf and seedlings (Supplementary Fig. 4). SbERF genes, as well as their rice orthologs placed in groups III, IX and X, expressed at very low levels. Whereas, most of the genes in the phylogenetic groups IV, VI, VII, and VIII exhibited high-level ubiquitous expression across all the tissues analyzed. Similar expression patterns of most of the rice and sorghum ERF orthologs suggest conservation in ERF gene functions in both the species (Supplementary Fig.4).

Based on the information available in Plant Tandem Duplicated Genes Database, positional coordinates on sorghum chromosomes and percent identity, 48 ERF genes comprising clusters of two to six genes were identified as potential paralogs resulted due to tandem duplications. In addition, 57 pairs of paralogous genes lying on segmentally duplicated regions of sorghum chromosomes were identified based on the information available in the Plant Genome Duplication Database (PGDD). These 57 gene pairs corresponded to 69 ERF genes as many of these had multiple paralogs. Overall, 106 ERF genes out of 158 had at least one paralog due to segmental or tandem duplication, indicating that gene duplications are the major factor underlying expansion of the ERF family in sorghum.

Chromosomal localization of SbERF genes on respective chromosomes, based on the coordinates available in Phytozome, clearly revealed uneven distribution across sorghum chromosomes with the highest number of genes located on chromosomes 4 and 2 (Fig. 6a). Notably, majority of the SbERFs were localized on the lower arms of the chromosomes (Fig. 6a). The maximum number of segmentally duplicated ERF genes were shared between chromosomes 4, 6 and 10 (Fig. 6b).

Fig. 6.

Chromosomal localization and duplication analysis of ERF family genes in Sorghum. a ERF genes are marked on sorghum chromosomes and clusters of tandemly duplicated genes are highlighted using the rectangular box. b The circos diagram shows segmental duplications of ERF family genes in sorghum genome

The phylogenetic groups, III, VIII, and IX mainly expanded due to tandem duplications, whereas, groups III, VII and X contain the maximum number of segmentally duplicated gene pairs. A correlation greater than 0.7 was observed in expression patterns of about 25% of the segmentally duplicated gene pairs and eleven of the tandem duplicated genes, indicating a high level of genetic redundancy among them (Blue rectangles, Fig. 4). The Ka/Ks ratios of most of these gene pairs were below 1 indicating purifying selection (Supplementary Table S6).

Further, there was a significant overlap between segmental and tandem duplicates indicating dispersion of duplicated segments as leading cause of the expansion of the ERF family in sorghum. This is especially true for groups III, IV, VII and X. For example, SbERF014, 50, 51, 52, 69, 71, 72 and 108 of group III are paralogous and, therefore, seem to have originated from the same gene. In group IV, SbERF091, 96 and 109 are paralogous to each other. Similarly, SbERF009, 22, 56, 31 and 102 of group VII resulted due to segmental duplications. In group IX, SbERF005, 65 and 128 are paralogous. Conversely, paralogous genes, SbERF45, 58, 63, 89, 95 and 120 of group X are placed on different chromosomes indicating dispersion of segmentally duplicated regions. In addition, SbERF023, 83 and 129, also seem to have originated from the same gene. However, groups III, VIII and IX seem to have mainly expanded due to tandem duplications. The largest cluster of tandemly duplicated genes consisting of SbERF025, 26, 27, 28, 29 and 30 belonged to group III. To estimate the redundancy among paralogous ERF genes, we compared the expression profiles of paralogous genes. Out of 57 pairs of ERF genes lying on segmentally duplicated regions, 10 gene pairs exhibit expression correlation of ≥ 0.7, whereas, another set of eleven gene pairs had an expression correlation between 0.5 and 0.7. Among rest, eighteen gene pairs had a correlation of less than 0.5 indicating divergence in their expression patterns. For another eighteen gene pairs, the expression correlation could not be calculated due to missing data for one of the paralogs.

Similarly, among paralogous genes arisen due to tandem duplications, SbERF015 and 16 as well as SbERF026, 27, 28 and 29 on chromosome 2; SbERF050 and 51 on chromosome 4 and SbERF075, 76 and 77 on chromosome 7 exhibited a correlation greater than 0.7 in their expression patterns indicating redundancy in their gene functions. Also, it is noteworthy that tandemly duplicated genes SbERF009, 10 and 11 on chromosome 1 were downregulated in response to drought, heat as well as combined heat and drought stress indicating their conserved role in the stress response. Similarly, paralogous genes, SbERF026 and 27, were upregulated in response to both heat as well as combined stress treatments.

Analysis of regulatory regions

Genome-wide duplication events are the major force behind the origin of multigene families (Nei and Rooney 2005). Evolution of regulatory elements in their promoter sequences has played an important role in neofunctionalization of the duplicated genes (Arsovski et al. 2015). A large number of cis-acting elements related to diverse metabolic pathways were identified in the SbERF gene promoters by PlantPan 2.0 (Supplementary Table S7). A clear difference in the distribution of cis-regulatory elements was observed in promoters of DREB and ERF genes. About 50% of the SbERF genes had DRE/CRT element and ~ 25% had GCC-box in their own promoters, which indicates their regulation by other AP2/ERF proteins. All cis-elements identified from ERF regulatory regions were categorized into four groups i.e. Stress-associated (Biotic and abiotic), hormone-responsive (ABA, salicylic acid, jasmonic acid, methyl jasmonate etc.), developmental tissue-specific (embryo, endosperm, seed, meristem, etc.) and transcription factor binding sites (Supplementary Table S7). ABA-responsive elements, categorized as abscisic acid-responsive elements (ABRE), were detected in DREB gene promoters at a higher frequency with some of the ABRE elements exclusively detected in promoters of DREB genes. Several hormone-responsive elements (auxin, ethylene, methyl jasmonate, salicylic acids, and jasmonic acid), biotic stress (W-box, YTGTCWC, AG motif) and light-responsive elements were also detected in the regulatory regions of ERF genes (Supplementary Table S7). Wider distribution of cis-regulatory elements related to abiotic/biotic stress, hormonal responses or other transcription factors, suggests that these proteins are controlled by diverse signaling pathways involved in various developmental processes and responses of the plant towards stress conditions. In addition, binding sites for specific transcription factors such as WRKY, bZIP, MYB, NAC and bHLH, could also be delineated. bZIP TF can bind to TACGTA and CACGTG, and these elements were detected in both ERF and DREB gene promoters. However, the TACGTA element was more common in DREB gene promoters while CACGTG was more common in ERF gene promoters. Phylogenetic clade-specific bias was also obvious in the cis-regulatory element distribution. Though GCC-box was common in ERF gene promoters, none of the gene promoters from the group I and II (except one) contained GCC-box. Whereas, > 35% of the DREB genes from group III and IV contained GCC-box. Similarly, DRE/CRT element was present with higher frequency in group IV, V, VIII and X genes. Frequency of multiple stress-related binding elements like TCATCTTCTT, AGATCCAA or GTTAGTT, was higher in the ERF gene promoters. The presence of these elements and their biased distribution in different phylogenetic clades and DREB or ERF groups indicate their importance in gene regulation, however, empirical validation is required to confirm these predictions.

Phylogenomic analysis of sorghum ERFs with genetically characterized ERF proteins from diverse plant species

To facilitate prediction of gene functions using a phylogenomic approach, we compiled a list of genetically characterized ERF genes from various plant species. In total, gene function information for more than 100 genes was collected (Supplementary Table S8). The ERF domains from these sequences were extracted and aligned with those of sorghum ERF proteins and a conjoined phylogenetic tree was generated (Fig. 7). To facilitate the mapping of functions from model plants, a total of 42 and 106 SbERF orthologs were identified in Arabidopsis and rice, respectively (Supplementary Table S9).

Fig. 7.

Conjoined phylogenetic tree of sorghum ERF genes with genetically characterized ERF proteins from other plant species. The phylogenetic groups are marked using roman numerals. The functions of the genes assigned to each group are highlighted

Among characterized genes from the group I, MtWXP1 and MtWXP2 of Medicago enhance drought resistance by regulating cuticular wax accumulation (Zhang et al. 2007). ZmDBF1 and OsERF48 also play a role in combating water deficit stress and enhancing root growth (Saleh et al. 2006; Jung et al. 2017). Sorghum ortholog (SbERF080) of rice OsERF48, is upregulated by drought as well as combined heat and drought stress in sorghum. Its regulatory region contains ABRE and dehydration response-related cis-regulating elements. SbERF080 is, therefore, an important candidate for evaluation in drought response. Group II did not contain any of the mapped characterized genes. However, based on the ortholog analysis, we identified SbERF094 as an ortholog of SERF1, which has been shown to regulate root to shoot signaling during salinity stress in rice (Schmidt et al. 2014). SbERF094 was induced by heat stress as well as combined heat and drought stress indicating its involvement in abiotic stress response.

Most of the DREB proteins characterized by involvement in abiotic stress response clubbed with group III proteins. The tree highlights sorghum homologs of rice DREB genes implicated in abiotic stress tolerance in rice. Arabidopsis CBF proteins, AtCBF1, AtCBF2 and AtCBF3, which are involved in drought and low-temperature stress tolerance (Pino et al. 2008; Novillo et al. 2004; Kasuga et al. 2004), also clubbed with group III but formed a separate clade. The sorghum orthologs of rice DREB genes including SbERF027 (OsDREB1A), SbERF025 (OsDREB1B), SbERF100 (OsDREB1C), SbERF085 (OsDREB1D), SbERF072 (OsDREB1E and G) and SbERF042 (OsDREB1F) are recommended candidates for engineering abiotic stress tolerance in sorghum. All of these genes were affected by abiotic stress treatments. Promoters of these genes contained cis-elements related to hormonal responses, dehydration and TF binding sites. Group IV also features several known genes from rice and Arabidopsis including OsDREB2A, OsDREB2B, OsABI4, AtDREB2A, AtDREB2C and AtABI4, mainly involved in regulating salt and drought stress tolerance, through ABA-mediated pathways (Matsukura et al. 2010; Sakuma et al. 2006; Zhang et al. 2013; Chen et al. 2010; Penfield et al. 2006). Promoter regions of sorghum orthologs of these genes, including SbERF092 (OsABI4 and AtABI4), SbERF84 (AtDREB2C), SbERF109 (AtDREB2C), SbERF091 (OsDREB2B) and SbERF034 (OsDREB2A), were specifically enriched in dehydration response elements and are, therefore, worth examining further. Rice OsERF117/OsABI4, involved in drought tolerance, and its sorghum ortholog SbERF92 exhibited embryo-specific expression in the data analyzed here. It would be interesting to evaluate its role in drought tolerance during seed development.

Groups V to XI comprised ERF subfamily proteins, which have been shown to play more diverse roles in plant development as well as biotic and abiotic stress tolerance. Group V contained several genes implicated in lipid biosynthetic pathways including AtWIN1, AtSHN1 and HvNUD (Broun et al. 2004; Aharoni et al. 2004; Taketa et al. 2008), suggesting sorghum group V proteins as candidates for engineering lipid biosynthesis. The closest homologs of these genes, SbERF122 and 153, are orthologous to rice OsWR1 and OsWR2, respectively, implicated in wax biosynthesis and drought tolerance (Wang et al. 2012). Both these genes exhibited high expression in top internodes and are good candidates for in-depth characterization. SlPti6 of group VI regulates both biotic and abiotic stress tolerance, whereas SiCRF2 regulates abiotic stress response (Shi et al. 2014). OsSnorkel1 and 2 regulate internode elongation in response to submergence stress (Hattori et al. 2009). SbERF159 is the closest ortholog of these genes in sorghum and its promoter contains TF binding and dehydration response binding elements. It would be worth examining further.

Group VII genes play a role in diverse abiotic stresses as well as pathogen response. SbERF009 is orthologous to rice OsBIERF4, which regulates both biotic and abiotic stress tolerance (Cao et al. 2006). SbERF009 is down-regulated by heat as well as combined heat and drought stress. Whereas, SbERF011, orthologous to rice root abundant factor OsRAF/OsLG3, which is involved in drought tolerance (Xiong et al. 2018), is down-regulated by all three abiotic stress treatments. SbERF012 is orthologous to rice OsERF060/OsEBP89 implicated in shoot development and seed maturation in rice (Yang et al. 2002). Both the genes shared similar expression patterns with peak expression in seedling stage and significant transcript levels in endosperm indicating conservation in function (Supplementary Fig. S4).

Several of the group VIII proteins including TmERF1, ESR1, AtERF4 have been shown to act as negative regulators of plant growth and stress response (Yang et al. 2005; Banno et al. 2001). SbERF047, orthologous to Arabidopsis AtBOLITA and ESR1, was induced by drought stress in sorghum but exhibited very low-level expression in developmental tissues examined. Its promoter had hormonal response elements but lacked most of the abiotic stress-responsive binding elements. It would be interesting to examine if it is involved in regulating the interplay of plant growth and stress response.

Group IX genes are also implicated in diverse stress responses. However, it is interesting to note that three of the Arabidopsis proteins, AtERF04, AtERF05, and AtERF6 that form a separate clade are involved in high light acclimation in Arabidopsis. The sorghum proteins in the same clade, however, form a separate subclade indicating a possible divergence in function. Group X includes AtABR1 that regulates ABA response during seed germination (Pandey et al. 2005) and OsEATB of rice implicated in internode elongation in response to gibberellins (Qi et al. 2011), indicating a role of this group in mainly regulating plant development in response to phytohormones. Also, SbERF063 and SbERF095 gene promoters contained several hormonal response related binding elements. SbERF023, orthologous to rice OsEATB that regulates internode elongation (Qi et al. 2011), exhibited the highest expression in seedling shoots and basal internode indicating conservation in function. Similarly, OsCRl5 of group XI regulates crown root initiation pointing to a role in plant growth and development (Kitomi et al. 2011). SbERF146 orthologous to this gene was downregulated in response to combined drought and heat stress. Downregulation in response to stress might infer growth compensation due to stress response.

Since ERF genes play important roles in regulating interplay between plant growth and diverse stress conditions, these are of prime importance for studying stress responses and developmental networks in plants (Phukan et al. 2017) and engineer them to grow on marginal lands. Based on the phylogenomic analysis, expression profiles across tissue types and in response to stress conditions, and the presence of cis-regulatory elements present in their promoters, a list of 19 ERF genes was compiled as high priority candidates for engineering abiotic stress tolerance in sorghum (Supplementary Table S10). Manipulating expression of some of these genes can lead to developmental phenotypes due to their multifunctional roles in other aspects of plant growth and development as well. For example, SbERF023 is orthologous to rice OsEATB that directly regulates stem internode elongation and, therefore, plant height. We checked the expression of 10 of these genes in mature leaf and middle stem internodes, harvested at booting stage, using qPCR. As expected, SbERF023 exhibited maximum expression in stem internodes and mature leaves followed by SbERF032, SbERF073 and SbERF042 (Supplementary Fig. S5). These patterns are also in agreement with the trend observed in RNAsq data presented in Fig. 4. Manipulation of these four genes would lead likely lead to plant height or leaf senescence-related phenotypes. In conclusion, our work provides a platform for initiate and design suitable strategies for in-depth characterization of sorghum ERF proteins.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

SM, VS, IV and SSP performed the analysis and prepared the figures. SM, RS and MKS participated in writing the manuscript. MKS conceived, conceptualized and supervised the complete study. K-H.J. participated in finalizing the manuscript. All authors read and approved the final manuscript.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.