Introduction

Auxin regulates a host of plant developmental and physiological processes, including embryogenesis, organogenesis, tropic growth, and root and shoot architecture (Quint and Gray 2006). Two types of transcription factor families are required for controlling the expression of auxin response genes, auxin response factors (ARFs), and Aux/IAA repressors (Guilfoyle and Hagen 2007). Members of the Aux/IAA family are generally regarded as repressors of auxin-induced gene expression (Ulmasov et al. 1997). Meanwhile, ARFs activate or repress the expression of auxin response genes by binding to auxin response elements (AuxREs) on promoters of auxin response genes (Tiwari et al. 2003). A number of putative AuxREs have been defined within the upstream promoter regions of primary/early auxin responsive genes, including one or more copies of the conserved motif TGTCTC (Ulmasov et al. 1999b). A typical ARF protein contains a conserved N-terminal B3-like DNA-binding domain (DBD) that regulates expression of auxin response genes, a conserved C-terminal dimerization domain (CTD) that resembles domains III and IV in Aux/IAA proteins, and a variable middle region (MR) (Ulmasov et al. 1997; Guilfoyle and Hagen 2007), located between the DBD and CTD, that determines whether the ARF functions as a transcriptional activator or repressor (Ulmasov et al. 1999a; Tiwari et al. 2003).

Recent advances have provided information on regulation of ARF gene expression, ARF roles in growth and developmental processes, and target genes regulated by ARFs (Liscum and Reed, 2002; Guilfoyle and Hagen 2007). It has been demonstrated that the ARF proteins participate in the transcriptional regulation of a variety of biological processes related to growth and development such as embryogenesis (Hamann et al. 2002; Weijers et al. 2006) leaf expansion (Wilmoth et al. 2005) leaf senescence (Lim et al. 2010), lateral root growth (Tatematsu et al. 2004; Okushima et al. 2007; Marin et al. 2010), and fruit development (Goetz et al. 2006, 2007; Guillon et al. 2008; Jong et al. 2009), as well as various responses to environmental stimuli. Recently, the involvement of ARF family members was reported in ethylene (Li et al. 2006), brassinosteroid (Vert et al. 2008), and ABA responses (Yoon et al. 2010).

Twenty-three ARF genes have been identified in the Arabidopsis genome, distributed over all five chromosomes (Wei and Cui 2006). Sequencing of the rice (Oryza sativa) genome (Rice Genome Initiative 2000), revealed 25 genes, distributed over 10 of the 12 rice chromosomes, that were postulated to encode proteins belonging to the ARF family (Wang et al. 2007; Shen et al. 2010). Phylogenetic analyses revealed that individual members of transcription factor families are clustered into subgroups of genes that are most closely related to other members of that same subgroup in Arabidopsis and rice. Recently, a total of 39 PoptrARFs and 24 SvARFs genes were also identified in Populus trichocarpa (Kalluri et al. 2007) and sorghum (Sorghum vulgare) genome (Andrew et al. 2009), respectively. In addition, the complete cDNA sequences of all 31 maize ZmARFs genes were also submitted to GenBank (Alper et al. 2009). However, in tomato, only 6 SlARF genes, including SlARF2, SlARF3, SlARF4, SlARF6, SlARF7, and SlARF8, have been identified and shown to be homologous to AtARFs (Alvarez et al. 2006; Goetz et al. 2007; Feng et al. 2009; Jong et al. 2009). To date, no systematic investigations of ARF family proteins have been reported in tomato until recently (Kumar et al. 2011). Moreover, Functional analysis of each transcription factor of the ARF family has not been performed, despite the importance of ARF proteins in multiple aspects of plant physiology.

The Genome Sequencing Project for tomato has been completed lately, and the ARFs of Arabidopsis, rice, and maize have also been published, so it is now feasible to carry out a genome-wide search for tomato homologues and to conduct a comparative analysis of ARFs for these four species. To elucidate the structure of SlARF and characterize expression during reproduction in tomato, 21 putative genes with ARF domains were identified through genomic data mining. The full cDNA sequences of 15 novel tomato ARFs were isolated by PCR-based approaches. The genomic structure, chromosomal location, and sequence homology of all SlARFs were then investigated, followed by comparative phylogenetic analysis, exon/intron mapping, and structural analysis of conserved protein motifs of the ARF family genes. Subsequently, the different temporal and spatial expression patterns during flowering and fruiting in tomato plants were determined for each SlARF gene by quantitative real-time PCR (qRT-PCR). The resulting classification of groups, identification of putative functional motifs, and characterization of the expression patterns will be useful for future analysis of the biological functions of ARF family genes in tomato.

Materials and methods

Searching for the ARF genes

Multiple database searches were performed to find all members of the ARF family in Arabidopsis (Arabidopsis thaliana), rice (Oryza sativa L. subsp. japonica), and maize (Zea mays L.). To find ARF genes in Arabidopsis (AtARFs) and rice (OsARFs),“auxin responsive factor” was used as a query to search the protein and nucleotide databases of The National Center for Biotechnology Information (NCBI), and the matching genes were confirmed by previous reports (Wang et al. 2007). Similarly, all 31 ZmARFs genes in maize were identified from the MaizeGDB Database (http://www.maizegdb.org).

To find previously identified and potential ARF family genes in tomato, multiple database searches were performed. First, “auxin responsive factor” was used as a query to search the SGN database (The tomato Information Resource, http://solgenomics.net). Six known SlARF family genes were identified, including SlARF2, SlARF3, SlARF4, SlARF6, SlARF7, and SlARF8. To find other potential ARFs, we initially surveyed the SGN database using the amino acid sequences of the conserved ARF domains from all the known ARF families (including AtARFs and OsARFs) as queries. To increase the number of potential ARF proteins, we also performed the database searches using amino acid sequences of the ARF domains in some ARF family members in Cucumis sativus(3 members) and Solanum lycopersicum (6 members). Based on the combined results from all searches, we finally identified all members of tomato ARF family from the currently available genomic databases. After searching for ARF genes, bioinformatics tools, such as DNASTAR and FGENESH (http://linux1.softberry.com/berry) were used to analyze and predict those unknown SlARFs. NCBI ORF finder was used to find putative open reading frames and functional domains were determined by BLASTP of NCBI.

Isolation of the full-length cDNA sequence using RT-PCR

Total RNA was extracted from tomato ovaries using TRIZOL reagent (Invitrogen, Germany) according to the manufacturer’s instructions. The first cDNA strand was generated using the Improm-TM Reverse Transcription system (Promega, Madison, WI, USA) following the manufacture’s protocol. The full-length cDNA sequences of 15 novel SlARFs were amplified by PCR using primers designed based on the predicted results by FGENESH (listed in Supplementary Table 1). Since the predicted cDNA sequences were quite long (some even longer than 3.0 kb), it was difficult to design an adequate single pair of primers for the entire segment. Therefore, we designed two or more primer pairs for each ARF to amplify and clone the fragments. Then we assembled them into the whole target fragment, and verified the full-length cDNA by PCR with gene-specific primer sets and by BLASTN against SGN database. After optimization, the PCR conditions included denaturation at 94°C for 4 min, followed by 35 cycles of 30 s at 94°C, 30 s at 50–55°C (depending on the specific primers) and 70 s at 72°C, and a final 7-min elongation at 72°C. The amplifications were carried out using TGRADIENT Thermal Cycler machines (Biometra, Germany). The amplified cDNA fragments were cloned and sequenced using the ABI Prism 3730 sequencer (Invitrogen, Bioasia Biotech Co. Ltd). The DNA sequences were amplified utilizing gene-specific primer sets designed from the full length cDNA. Finally, the ORFs of the 15 unknown SlARFs were determined by ORF Finder (http://www.ncbi.nlm.nih.gov/gorf/gorf.html) and homologous alignment.

Mapping SlARF and ZmARF genes on chromosomes

To determine the location of SlARF and ZmARF genes on chromosomes, the SlARF and ZmARF sequences were further used as query sequences for the BLASTN search against SGN Tomato Whole genome Scaffolds data (2.30) (http://www.sgn.cornell.edu/tools/blast/) and Maize GDB B73 RefGen_v1 databases (http://www.maizegdb.org), respectively. Finally, the locations of all 21 SlARFs and 31 ZmARFs were detected.

Multiple-sequence alignments and phylogenetic analysis

Gene sequences were analyzed by DNAStar software and the net service ExPASy Proteomics Server (http://ca.expasy.org). Multiple-sequence alignments employed ClustalX v1.81 (Thompson et al. 1997). Phylogenetic analysis was performed using MEGA 4.1 program by the neighbor-joining (NJ) method (Saitou and Nei 1987). Conserved motifs were investigated by multiple alignment analyses using Clustal W.

Expression analysis of SlARFs

The plant materials used for expression analysis were sampled from tomato (S. lycopersicum L.) cv. Micro-Tom plants (Tomato Genetics Resource Center, University of California, Davis, USA) were grown until flowering in a temperature-controlled greenhouse at the experimental farm at Zhejiang University.

The leaves, stems, roots, and buds were collected from flowering tomato plants, and the various floral organs (sepal, petal, stamen, and ovary) were isolated from the flower buds (about 3 days before opening). To analyze the expression pattern of auxin response genes at different flower developmental stages, flower buds were collected at three stages of early floral development, which was roughly defined by the length of flower buds as follows: stage I: 3–4 mm, stage II: 5–6 mm, and stage III at 7–8 mm (Brukhin et al. 2003). In addition, the ovaries were sampled at 0, 3, 6, and 9 days after the flower fully opened. All the samples were frozen in liquid nitrogen immediately and stored at −75°C until RNA isolation.

Total RNA and the first cDNA strand were prepared as described earlier. QRT-PCR techniques were employed to determine characterize the gene expression profiles of new SlARFs. The quantitative real-time PCR (qRT-PCR) was carried out using the primer pairs listed in Supplementary Table S2. Because the four sister pairs, including SlARF6/SlARF6-1, SlARF8/SlARF8-1, SlARF13/SlARF13-1, and SlARF19/SlARF19-1 are so similar in nucleotide sequences, we only design appropriate primers for one member of each pair. A sample of cDNA (1 μg) was subjected to real-time PCR in a final volume of 20 μl containing 12.5 μl SYBR Green Master Mix Reagent (Takara, Japan) and specific primers (3 pmol). Two biological and three technical replicates for each sample were performed in the real-time PCR machine (STRATAGENE, MX3500), programmed to heat for 30 s at 95°C, followed by 40 cycles of 5 s at 95°C and 45 s at 50°C, and at the end, one cycle of 1 min at 95°C, 30 s at 50°C, and 30 s at 95°C. To normalize the total amount of cDNA present in each reaction, the Ubi3 gene was co-amplified as an endogenous control for calibration of relative expression. The ΔΔC t method of relative gene quantification recommended by Applied Biosystems was used to calculate the expression level of different treatments.

Results

Identification and isolation of SlARF family genes in tomato

To identify the ARF family genes in tomato, BLAST searches of the SGN database were performed using the ARF domain of the Arabidopsis and rice protein as a query sequence. A total of 30 ARF-domain genome DNA sequences and two unigenes were obtained by TBLASTN at an e value of 1e−3 that were similar to ARF genes. All sequences were predicted by FGENESH (http://www.softberry.com/berry.phtml?topic=fgenesh). These predicted amino acid sequences were analyzed by blastp of NCBI to find their conserved domains, followed by homologous alignment with known SlARF genes. For these analyses, the tomato genome appeared to contain, in addition to the six previously known SlARF genes (Alvarez et al. 2006; Jong et al. 2009; Feng et al. 2009), 15 other putative novel ARF genes.

The PCR primers (Supplementary Table S1) were designed based on the predicted results of FGENESH. The potential full-length cDNA sequences of all 15 putative SlARFs were isolated through PCR-based approaches. The open reading frame (ORF) length of SlARF genes varied from 1,218 bp (SlARF12) to 3,371 bp (SlARF7), encoding polypeptides of 375–1,123 aa, with a predicted molecular mass range of 42.4–126.4 kD. The theoretical pI ranged from 5.48 to 8.58 (Table 1) and the calculated molecular masses of the deduced ORFs were almost identical with the sizes of ARF polypeptides previously determined in other plants (Wang et al. 2007).

Table 1 ARF gene in tomato
Full size table

It is noteworthy that the nomenclature system for SlARFs used in the present study, a generic name from SlARF1 to SlARF19-1, was provisionally used to distinguish each of the ARF genes according to the homology between AtARFs and SlARFs. However, homologous genes for SlARF11, SlARF12, SlARF13, SlARF13-1, and SlARF14 were not found in Arabidopsis, rice and maize, so they were named according to the order of submitted to the GenBank database.

Chromosomal locations of SlARFs and ZmARFs

The chromosomal locations and transcription directions of the 21 SlARF genes were determined and demonstrated using BLASTN analysis on Tomato WGS Chromosomes (Fig. 1). Similar to Arabidopsis and rice, SlARF family genes in tomato appeared to be distributed among all the linkage groups, except chromosome 9. The number of SlARF genes per chromosome ranged from one to three. Three SlARFs were present on chromosome 5, 6, and 11, two each were localized to chromosomes 2, 3, 8, and 12, and only one each to chromosomes 1, 4, 6, and 10 (Table 1; Fig. 1).

Fig. 1
figure 1

a Genomic distribution of ARF genes on tomato chromosomes. b Genomic distribution of ARF genes on maize chromosomes. The arrows next to gene names show the direction of transcription. White ovals on the maize chromosomes (vertical bar) indicate the position of centromeres. The chromosome numbers are indicated at the top of each bar

Full size image

Interestingly, SlARF8 and SlARF8-1, SlARF19 and SlARF19-1, SlARF6, and SlARF6-1 were present in different chromosomes although they shared more than 90% amino acid sequence identity. The fact that the two genes on different strands were nearly identical suggested that they might be derived from recent gene duplication events. This finding is consistent with a previous report demonstrating that duplicated genes involved in signal transduction and transcription are preferentially retained compared with other functional gene categories (Blanc and Wolfe 2004).

Similarly, 31 ZmARFs distributed on 9 of 11 maize chromosomes. No ZmARFs was detected on chromosomes 7 or 11. Five ZmARFs were present on chromosome 5, four each on chromosomes 2, 4, 5, 6, and 10, two each on chromosomes 1 and 8, and only one SlARF on chromosomes 5. The location of ZmARF8 by BLASTN was not found in present database (Fig. 1; Table 2).

Table 2 ARF gene in maize
Full size table

Sequence analysis of the SlARF and ZmARF proteins

All the tomato SlARF protein sequences were found to contain DNA-binding domains (DBDs) and MR (middle region) domains (Table 1). All SlARF proteins contained a highly conserved region of about 390 amino acid residues in the N-terminal portion that corresponded to the DBD of the Arabidopsis ARF family (Fig. 2a). Fourteen deduced SlARF proteins contained a carboxyl-terminal domain (CTD) related to domains III and IV found in Aux/IAA proteins (Fig. 2b). Other seven of the SlARF proteins including SlARF3, SlARF6-1, SlARF13, SlARF13-1, SlARF14, and SlARF17 lacked a CTD.

Fig. 2
figure 2

a Alignment profile of tomato ARF proteins obtained with the ClustalX program. The height of the bars indicates the number of identical residues per position. The shaded regions indicate the high sequence similarity among DBDs regions. Motifs III and IV are consensus sequences shared by Aux/IAA proteins. b Multiple alignments of Motifs III and IV in tomato ARF proteins obtained with ClustalX. Black and light gray shading indicate identical and conversed amino acid residues, respectively. Conserved domains are also underlined and correspond to part a

Full size image

Similarly, all ZmARF proteins contained a highly conserved N-terminal region of about 380 amino acid residues that corresponded to the DBD of the Arabidopsis ARF family (Fig. 3a). Twenty-two ZmARFs protein sequences contained three ARF domains, while 9 out of 31 ZmARFs only contained two domains and lacked a CTD domain (Table 2). In maize, the molecular mass of ZmARF protein sequences generally ranged from 50.56 kDa (ZmARF31) to127.49 kDa (ZmARF20) (Fig. 3b).

Fig. 3
figure 3

a Alignment profile of maize ARF proteins obtained with the ClustalX program. The height of the bars indicates the number of identical residues per position. The shaded regions indicate the high sequence similarity among DBDs regions. Motifs III and IV are consensus sequences shared by Aux/IAA proteins. b Multiple alignments of Motifs III and IV of in maize ARF proteins obtained with ClustalX. Conserved residues are highlighted in gray boxes. Conserved domains are also underlined and correspond to part a

Full size image

Gene structure and phylogenetic analysis of ARFs

A comparison of the full-length cDNA sequences with the corresponding genomic DNA sequences revealed the numbers and positions of exons and introns for each individual SlARF gene. The coding sequences of all the SlARFs except SlARF14 were disrupted by introns. The number of introns varied from 1 (SlARF15) to 13 (SlARF1, 2, 5, 8-1) (Fig. 4a). It was suggested that SlARF14 was the product of an mRNA inserted into the tomato genome (Babenko et al. 2004). Based on the presence of triplets containing SlARF10, 14, and 16 in the phylogenetic tree (Fig. 4a), we surmise that this mRNA might come from SlARF10 mRNA, SlARF16 mRNA, or both.

Fig. 4
figure 4

a Left part illustrates the phylogenetic relationships among the tomato ARF proteins. The unrooted tree was generated using MEGA4.1 program by the neighbor-joining method. Bootstrap values (above 50%) from 1,000 replicates are indicated at each branch. Right part illustrates the exon–intron organization of corresponding ARF genes. The exons and introns are represented by black boxes and lines, respectively. b The same information for maize ARF proteins as shown in part a

Full size image

An unrooted phylogenetic tree was generated from the alignment of the full-length protein sequences of all SlARFs. The 21 SlARFs could be divided into three major classes (I–III, Fig. 4a) similar to those in rice (Wang et al. 2007). Class I was further divided into two sub-classes, Ia-1 with seven members and Ib with two members. Class II was also further divided into two sub-classes IIa and IIb, each containing four members. Class III contained four members that were the most divergent compared to those grouped into the other two classes. The SlARFs in class III contained fewer introns in their ORF regions than those in other groups, with one, SlARF14, even possessing no introns. The 21 SlARFs formed seven sister pairs (Fig. 4a), all with very strong bootstrap support (>99%).

The 31 ZmARFs were divided into four major classes, I–IV (Fig. 4b). Class I and class II were further subdivided into two subgroups, Ia-1 with seven members and Ib with five members, and IIa with six members, and IIb with five members. Classes III and IV contained six and two members, respectively. The 31 ZmARFs formed 13 sister pairs (Fig. 4b), while the ARF genes from Arabidopsis and rice formed 6 and 9 sister pairs, respectively (Wang et al. 2007).

To investigate the relationships of ARF proteins, the full-length protein sequences of the 23 AtARFs, 25 OsARFs, 31 ZmARFs, and 21 SlARFs were used to build the phylogenetic tree. All 100 ARF proteins could be classified into four major classes: class I contained 46 members, class II contained 33 members, class III contained 17, and class IV contained 4 gene members. Class I was divided into three subgroups, Ia-1 (24 members), Ia-2(8 members) and Ib (14 members). Class II was further divided into two subgroups, IIa-1 with 16 members and IIb with 17 members. This classification is very similar to that of AtARFs except for Class IV (Supplementary Fig. 1).

In the joint phylogenetic tree, a total of 51 sister pairs were formed, including 7 SlARF–SlARF pairs, 13 ZmARF–ZmARF pairs, 6 AtARF–AtARF pairs, 9 OsARF–OsARF pairs, 10 OsARF–ZmARF pairs, 5 SlARF–AtARF pairs, and one OsARF–AtARF pair. Interestingly, subgroups Ia-1, Ib, IIa, IIb, and class III contained ARF genes from all the four species, but only Arabidopsis AtARFs proteins were present in subgroup Ia-2, while in class IV, only ARFs from rice and maize (monocotyledon) were present (Fig. 5).

Fig. 5
figure 5

Phylogenetic relationships among tomato, rice, maize, and Arabidopsis ARF proteins. The unrooted tree was generated using MEGA4.1 program by the neighbor-joining method. Bootstrap values (above 50%) from 1,000 replicates are indicated at each branch

Full size image

Expression characterization of SlARF genes

A previous report demonstrated that ARF genes were constitutive expressed (Wang et al. 2007). In our study, we also found that most of the SlARFs could be detected in root, stem, buds, and ovary using qRT-PCR (Fig. 6). The SlARF5 mRNA was more highly expressed in stem and leaf than in root, flower, and ovary. Stem exhibited higher expression of SlARF6, SlARF13, and SlARF19-1 than other organs, while SlARF1, SlARF2, and SlARF3 were mainly expressed in leaf. Meanwhile, higher mRNA level of SlARF9, SlARF16, and SlARF17 was detected in root.

Fig. 6
figure 6

qRT-PCR analyses of 17 SlARF genes in different organs (root, stem, leaf, buds, and ovary) of the tomato plant

Full size image

qRT-PCR analysis demonstrated that expression of most SlARF genes could be detected in all parts of the flower (Supplementary Fig. 4). SlARF6, SlARF7, and SlARF8 were expressed at a higher level in sepal and petal than in stamen and ovary, while SlARF7, SlARF14, SlARF16, SlARF17, and SlARF19-1 were expressed mainly in petal. The SlARF9, SlARF10, and SlARF11 mRNAs were detected in ovary at a higher level than in other tissues.

During the different developmental periods of tomato flowers, most of the SlARF genes detected exhibited a similar expression pattern (Supplementary Fig. 5). Most SlARF mRNAs increased during tomato flower development, while only SlARF4, SlARF9, and SlARF17 mRNA levels significantly decreased during flower development. Similar expression patterns were also detected during the different periods of ovary development (Supplementary Fig. 6). With the development of ovary and young fruit, the expression level increased to a peak on the third day after flower opening and then markedly decreased on the nineth day. Only SlARF16 mRNA reached its highest level on sixth day after flower opening, and then decreased significantly on the ninth day. Noticeably, the expression level of SlARF4 was quite different with the other ARF genes, constantly increasing even after pollination during young fruit development.

Discussion

In this study, 15 novel tomato ARF genes were identified and the full-length cDNA sequences of these SlARFs were isolated. The total number of 21 SlARFs detected in present study is more than the 17 SlARFs identified by Kumar et al. (2011) using publically available tomato EST databases. Our gene isolation and sequencing strategy was based on TBLASTN and PCR-based methods which allowed us to find more ARF genes and get exact and comprehensive data. Comparing the deducing amino acid sequences, 14 corresponding SlARF genes can be found in the report of Kumar et al. (2011). The number of SlARF members from tomato is comparable to that of Arabidopsis (23) and rice (25) (Okushima et al. 2005; Wang et al. 2007), although <39 in populus (Kalluri et al. 2007). Meanwhile, a total of 31 putative maize ARF genes were also predicted and analyzed from the MaizeGDB Database. The genome sizes of tomato, maize, Arabidopsis, and rice are quite different (tomato ~950 Mb, maize ~2,300 Mb, Arabidopsis ~125 Mb, rice ~450 Mb), as are the estimated total number of genes (tomato ~35,000, maize ~53,760, Arabidopsis ~25,000, rice ~37,000), so it is interesting to find roughly a similar number of ARF genes in these four different species.

Large-scale duplication of the tomato genome has been reported (Ku et al. 2000). It was suggested that tomato was likely a paleopolyploid (Hoeven et al. 2002) in which large-scale genome duplication occurred approximately 50–52 million years ago (Schlueter et al. 2004). The maize genome is replete with chromosomal duplications and repetitive sequences, the result of an ancient polyploid event that occurred over 11 million years ago (Gaut et al. 2000).

In this study, 7 sister pairs of SlARFs and 13 sister pairs of ZmARFs were identified by phylogenetic analysis. All sister pairs were compared with their corresponding chromosomal locations. Except for SlARF13 and SlARF13-1, which were likely the products of alternative splicing of mRNA, none of these sister pairs were genetically linked, as was also observed in the OsARFs (Wang et al. 2007). Conversely, all closely linked SlARF and ZmARF loci, such as SlARF7 and SlARF6-1 on chromosome 7, SlARF4 and SlARF10 on chromosome 11, ZmARF3 and ZmARF4 on chromosome 2, ZmARF2 and ZmARF22 chromosome 6, ZmARF23 and ZmARF24 on chromosome 6, and finally ZmARF29, ZmARF30, and ZmARF31 on chromosome 10, were not grouped in sister pairs. Similarly, none of the six sister pairs in Arabidopsis were genetically linked (Okushima et al. 2005). Sister pairs were located on chromosomes 12 and 7 (SlARF6 and 6-1), chromosomes 2 and 3 (SlARF8 and 8-1), chromosomes 7 and 5 (SlARF19 and 19-1), chromosomes 11 and 6 (SlARF10 and 16), chromosomes 2 and 11 (SlARF3 and 4), and chromosomes 3 and 12 (SlARF2 and 11). Only SlARF13 and SlARF13-1 were found on the same chromosome. Among the 12 ZmARF sister pairs, three were found on chromosomes 2 and 10 (ZmARF3 and 30, ZmARF4 and 29, ZmARF5 and 31), two on chromosomes 1 and 5 (SlARF1 and 20, SlARF2 and 17), and two on chromosomes 3 and 8 (SlARF10 and 25, SlARF11 and 26). The chromosomal locations of these SlARFs and ZmARFs sister pairs may represent duplicated chromosomal blocks.

Based on the above results, we conclude that whole genome and chromosomal segment duplications are the main factors responsible for the expansion of SlARFs, and especially ZmARFs. In Arabidopsis and rice, tandem duplications played a more important role in AtARF duplication (Wang et al. 2007), as evidenced by the fact that seven very closely related AtARFs (12, 13, 14, 15, 20, 21, and 23) in a single cluster are physically located near each other in a region of chromosome 1 in Arabidopsis (Remington et al. 2004; Okushima et al. 2005).

Phylogenetic analysis revealed that the organization of Arabidopsis, maize, tomato, and rice ARF proteins was very similar to each other in classes I, II, and III, implying that ARFs within these classes derived from a common ancestor. The 15 interspecies sister pairs, including OsARF25 and ZmARF9, SlARF1 and AtARF1, and OsARF14 and AtARF14, indicate these gene groups were descended from a common ancestor and possess well-conserved functions. The ten sister pairs between OsARFs and ZmARFs detected by phylogenetic analysis indicate that OsARFs and ZmARFs have a close evolutionary relationship, as may SlARFs and AtARFs with four interspecies sister pairs. In contrast, only one sister pair was found between monocots and dicots (OsARF14–AtARF14), indicating that the common ancestor of this sister pair appeared before the divergence of monocots and dicots. Remington et al. (2004) also suggested that ARF lineages originated before the monoco–eudicot divergence. The separation between dicot and monocot ARF sequences within each of the clades arose from this gene duplication, indicating that the ARF family expanded and diversified after the divergence between the two lineages. Most of the duplications in the Arabidopsis genome occurred shortly after the divergence between asterids (tomato) and rosids (Arabidopsis) 112–156 million years ago (Baumberger et al. 2003).

Class Ia-2 is a special subclass that only contained AtARFs, suggesting that these AtARFs were generated over the long-term evolution of Arabidopsis, but after the divergence of monocots and dicots. Moreover, segregation to a separate subclass suggests that these proteins have species-specific functions. The ARFs in class IV were all from maize and rice, the two representative monocots, suggesting that class IV proteins were either lost in dicots after divergence of monocots and dicots or evolved solely in monocots after the divergence.

Groups containing multiple ARFs from all four species, such as class IIa, were also found in the phylogenetic tree, which may indicate that a diversification of functions has occurred in all four species. Furthermore, six triplets containing one OsARF and multiple ZmARFs were found, while only one triplet containing multiple OsARFs and one ZmARF was found (OsARF5/OsARF19/ZmARF20). Groups containing one AtARF and multiple SlARFs were found in AtARF6/SlARF6/SlARF6-1, and groups containing one SlARF and multiple AtARFs were found in SlARF7/AtARF7/AtARF19. These classes are presumed to represent conserved functions in rice, tomato, maize, and Arabidopsis, but these functions might have begun to diversify in corresponding species as a result of gene duplication. Compared with OsARFs, the diversification of ZmARFs occurred more frequently, leading to a larger ARF family (31 members) than in rice (25 members).

The features and number of domains present in the protein sequences is very useful information for predicting the function of a new gene. Indeed, careful analysis of protein sequence is the first step when postulating the functions of novel ARF genes. The middle regions of ARFs function as activation domains (ADs) or repression domains (RDs) (Ulmasov et al. 1999a). Protoplast transfection assays indicated that AtARF1, AtARF2, AtARF4, and AtARF9, which contain middle regions rich in proline (P), serine (S) and threonine (T), are repressors, while AtARF5, AtARF6, AtARF7, and AtARF8, which contain middle regions rich in glutamine (Q), are activators (Tiwari et al. 2003; Ulmasov et al. 1999a). Interestingly, all four species encode some CTD-truncated ARFs, which is consistent with previous report that flowering plant tend to encode more CTD-truncated ARFs (Paponov et al. 2009). Compared with CTD-truncated ARFs of AtARFs (4 out of 23 ARFs) in Arabidopsis, more numbers of ARFs lacking a CTD were found in rice (6 out of 25 ARFs) and tomato (6 out of 21 ARFs), especially in maize (9 out of 31 ARFs). Kumar et al. (2011) also proved that tomato SlARF 2, 3, 6, 7, and 13 (5 out of 17 ARFs) showed absence of C-terminal Aux/IAA domains. So tomato, as a relatively advanced dicotyledon, has a higher percentage of CTD-truncated ARFs than Arabidopsis (28.6 and 17.4%, respectively). This situation is also been found in maize and rice as monocot crops (29.0 and 24%, respectively). The CTD is required for ARF-IAA dimerization, it seems that more CTD-truncated ARFs appeared during evolution, which may regulate gene expression in an auxin-independent manner (Shen et al. 2010). In previous studies, CTD-truncated ARFs are all putative repressors (Shen et al. 2010; Guilfoyle and Hagen 2007). ARF6-1, as a putative activator, also was found to be lack a CTD. So this gene might function in a different way, a more in-depth study is needed to further explain this phenomenon.

The protein sequences of all 21 SlARFs were analyzed, and the proline (P), serine (S), and threonine (T) rich regions were found in the MR domain sequences of SlARF1, SlARF2, SlARF3, SlARF4, SlARF9, SlARF10, SlARF11, SlARF12, SlARF13, SlARF13-1, SlARF16, SlARF14, and SlARF17, indicating these genes are more likely acting as repressors. In contrast, glutamine (Q)-rich regions, which are also somewhat rich in leucine (L) and serine (S), were found in the MR domain sequences of SlARF5, SlARF6, SlARF6-1, SlARF7, SlARF8, SlARF8-1, SlARF19, and SlARF19-1, implying that these genes are likely to be transcriptional activators (Supplementary Fig. 2, Supplementary Fig. 3). Although the MR of SlARF5 was enriched in Q, it differed from other Q-rich ARFs such as ARF6, ARF7, and ARF8 in having no homopolymeric Q stretches. Nevertheless, its MR has activation potency nearly equivalent to that of ARFs with homopolymeric Q stretches (Ulmasov et al. 1999a). All SlARFs and ZmARFs proteins with Q-rich MRs belong to class I, while PST-rich SlARFs and ZmARFs belong to other classes (Fig. 5).

Among 21 SlARFs, 17 SlARF genes were detected in all sampling tissues and organs in tomato, implying that most tomato SlARF genes exhibited constitutive expression. The mRNA levels of SlARF1 and SlARF2 were significantly higher in leaf than other organs (Fig. 6), implying that they, regarded as a repressor, might play an important role in the development of leaf, and this conjecture has been confirmed in Arabidopsis (Ellis et al. 2005; Lim et al. 2010). AtARF2 functions in the auxin-mediated control of Arabidopsis leaf longevity by acting as a repressor of auxin signaling (Lim et al. 2010). The loss of expression of AtARF7 genes is directly responsible for the reduced gene expression observed in mesophyll cells (Wang et al. 2005).

In Arabidopsis, arf6/arf8 double-null mutant flowers were arrested as infertile closed buds with short petals, short stamen filaments, undehisced anthers that did not release pollen, and immature gynoecia (Nagpal et al. 2005). The AtARF2 gene regulated floral organ abscission independently of the ethylene and cytokinin response pathways, and AtARF1 was partially redundant with ARF2 (Ellis et al. 2005). Similar expression patterns for several SlARF genes also indicate about their possible overlapping functions during various developmental processes in plants (Fig. 6, Supplementary Fig. 4, Supplementary Fig. 5, Supplementary Fig. 6).

In tomato, SlARF7 acts as a negative regulator of fruit set before pollination and fertilization and moderates the auxin response during fruit growth (Jong et al. 2009). The AtARF8 gene acts as an inhibitor to stop further carpel development in the absence of fertilization and the generation of signals required to initiate fruit and seed development (Goetz et al. 2006). DR12 (auxin response factor 4), can finely modify tomato fruit texture; when this gene was down-regulated, the pericarp tissue of tomato fruit became thicker (Guillon et al. 2008). In contrast to other genes, SlARF4 mRNA levels increased during fruit development, consistent with a previous study (Jones et al. 2002), implying that SlARF4 might be essential for fruit development. Other SlARFs may function in early fruit development, despite the difference in expression level. Kumar et al. (2011) proved that SlARF1 and 9 exhibited maximum expression level at open flower stage, and the expression level of SlARF16 was high at flower bud stage, indicating that these genes might be involved in flower development in tomato, while the mRNA levels of SlARF3, 5, 6, 13, 15, and 17 exhibited low expression during floral development and an increase at either 30 DAP (days after pollination) or mature green fruit stage, indicating that these ARF genes could be involved in the regulation of aspects of plant development. In sum, these studies indicate that some ARFs are indispensable for tomato flower and fruit development (Jones et al. 2002; de Jong et al. 2009).

In conclusion, the full cDNA sequences of 15 novel SlARFs were identified using PCR-based method. A comprehensive genome-wide analysis of SlARF gene family is presented, including the gene structures, chromosome locations, phylogeny, and conserved motifs. The expression characteristics of all 17 SlARFs were also analyzed. The major challenge for the future is to define the specific functions of each individual ARF gene during plant growth and development.

Accession numbers: Sequence data from this article can be found in the GenBank/EMBL data libraries under the following Accession numbers: HM061154(SlARF1), HM19248.1(SlARF5), HM187579.1(SlARF6-1), HM560979.1(SlARF8-1), HM037250.1(SlARF9), HM143941.1(SlARF10), HM143940.1(SlARF11), HM565127.1(SlARF12), HM565128.1(SlARF13), HM565129.1(SlARF13-1), HM565131.1(SlARF14), HM195247.1(SlARF16), HM456923(SlARF17), HM130544.1(SlARF19), HM565130.1 (SlARF19-1).