Genome-wide analysis of ATP-binding cassette transporters sheds light on the genes related to bioactive metabolite transportation in Salvia miltiorrhiza

Background: ATP-binding cassette (ABC) transporters have been found in plants and play important roles in metabolic transport in cells, which affect the subcellular compartmentalisation and tissue distribution of these compounds in particular. The dry roots and rhizomes of Salvia miltiorrhiza Bunge, a widely used medicinal plant, known as Danshen in traditional Chinese medicine, contain biologically active secondary metabolites of tanshinone and salvianolic acid. Given an assembled and annotated genome and a set of transcriptome data of S. miltiorrhiza, we analysed and identied the candidate genes that likely involved in the bioactive metabolite transportation of this medicinal plant, starting with the members of the ABC transporter family. Results: A total of 114 genes encoding ABC transporters were identied in the genome of S. miltiorrhiza. All the ABC genes were divided into eight subfamilies, including 3ABCA, 31ABCB, 14ABCC, 2ABCD, 1ABCE, 7ABCF, 46ABCG and 10 ABCI. Gene expressional proling analysis revealed the tissue-specic expression proles of these ABC transporters, particularly the highly expressed transporters in the roots of S. miltiorrhiza, which might be involved in transporting bioactive compounds of this medicinal plant. Notably, 18 genes that were highly expressed in the roots of S. miltiorrhiza were selected and veried by quantitative reverse transcription polymerase chain reaction to determine the co-expression proling of these candidate genes with the key enzyme genes involved in the tanshinone and salvianolic acid biosynthetic pathways. Finally, three ABC genes (SMil_00020022, SMil_00010949, SMil_00027268) and one gene (SMil_00016361) might be involved in tanshinone and salvianolic acid transport in cells, respectively. In addition, the biological functions of S. miltiorrhiza ABC transporters were predicted on the basis of the phylogenetic analysis and expressional prole exploration, including the transportation of lipids, hormones, ions and primary metabolites. Conclusions: Our systematic analyses on the ABC transporters in S. miltiorrhiza provided information on ABC

S. miltiorrhiza is a commonly used medicinal plant for the treatment of cardiovascular diseases and in ammations because of its high accumulation of biologically active hydrophilic salvianolic acid (SA) and lipophilic diterpenoids (tanshinones) in its roots or rhizomes [22]. S. miltiorrhiza is an ideal model medicinal plant for studying secondary metabolic biosynthesis. GGPP is the biosynthetic precursor of tanshinone, which is catalysed by copalyl diphosphate synthase (CPS) to form copalyl diphosphate, and then, a series of cytochrome P450 monooxygenases (CYP450s) catalyses the downstream oxidation reactions. Ferruginol, the catalytic product of CYP76AH1, is an important intermediate product in the biosynthetic pathway of tanshinone [23]. CYP76AH3 and CYP76AK1 are responsible for the conversion of ferruginol into 11,20-dihydroxy ferruginol and 11,20-dihydroxy sugiol en route to tanshinones [24]. SA biosynthesis is derived from 4-coumaroyl-3′,4′-dihydroxyphenyllactic acid (4C-DHPL), which is a combination of 3,4-dihydroxyphenyllactic acid (DHPL) and 4-coumaroyl-CoA. These two compounds are coupled by the rosmarinic acid synthase (SmRAS) [25]. The 3-hydroxyl group is introduced by a cytochrome P450-dependent monooxygenase (SmCYP98A14) to form rosmarinic acid [25]. Signi cant progress has been made in the biosynthetic pathways of these active ingredients in S. miltiorrhiza, but the transport and storage mechanisms of these compounds in cells have not yet been elucidated.
Here, we described the rst complete analysis of the ABC transporter superfamily from the S. miltiorrhiza genome. A total of 114 genes, divided into seven subfamilies, were annotated to encode the ABC transporter proteins in S. miltiorrhiza. We characterised all the ABC proteins of S. miltiorrhiza and performed a phylogenetic analysis in comparison with the members of the Arabidopsis ABC superfamily. On the basis of the co-expression analysis of key enzyme genes involved in the biosynthetic pathway of the active ingredients of S. miltiorrhiza, we predicted that three ABCG and one ABCC subfamily ABC transporter genes were involved in the transport of the bioactive metabolites of tanshinone and SA, respectively. In addition, the ABC proteins involved in the transport of plant hormones, secondary metabolites, ions and other substances were predicted on the basis of phylogenetic analysis and gene expression pro le characteristics of S. miltiorrhiza.

Results
Identi cation of ABC transporters in the S. miltiorrhiza genome A total of 204 homologous ABC transporters were annotated in the S. miltiorrhiza genome on the basis of the sequence alignment with all the ABC transporters in the Arabidopsis TAIR11 database. The ABC transporters of S. miltiorrhiza were veri ed by con rming the integrity of the conserved domains and motifs of ABC proteins. Finally, 114 genes encoding ABC transporters were identi ed in the S. miltiorrhiza genome ( Table 1). Considering that a typical full-sized ABC protein contained at least 1,200 amino acid residues [2], these ABC transporters ranged in length from 186 to 1,978 amino acid residues in S. miltiorrhiza (Table 1), indicating that some predicted ABC proteins with short sequences might be pseudogene or not full-length ABC transporters. Amongst the 114 ABC transporters, 86 were intrinsic membrane proteins with TMDs. Of these intrinsic membrane proteins, 50 were putative full-sized ABC transporters, containing at least two TMD and two NBD domains, which were distributed in ABCB, ABCG and ABCC subfamilies (Table 1). Of the other 36 intrinsic membrane proteins, 31 proteins were half-sized ABC transporters, with one TMD and one NBD domain, primarily distributed in the ABCF, ABCG and ABCI subfamilies ( Table 1). The remaining ve proteins were non-integrated proteins harbouring two TMD domains and one NBD domain or two NBD domains and one TMD domain, most of which belonged to ABCB and ABCG subfamilies (Table 1). In addition, the remaining 28 genes were identi ed as nonintrinsic proteins, which encoded proteins lacking TMD. Eighteen of these non-intrinsic proteins were covered in ve subfamilies (ABCB, ABCD, ABCE, ABCF and ABCG), and 10 proteins encoded a group of non-intrinsic ABC proteins (NAPs) and were divided into the ABCI subfamily (Table 1).
Fifteen motifs of candidate ABC transporters in S. miltiorrhiza were predicted and identi ed by the MEME Web server to characterise the diversity of ABC proteins. The results showed that the conserved motifs amongst the ABC proteins were similar. For example, the motifs of ABC signatures, Walker A and Walker B were present in these proteins (Additional le 1: Figure S1). Moreover, the motifs of the ABC proteins belonging to the same subfamily were distributed in the same position (Additional le 1: Figure S1). The integrity of the full-sized transporter could be checked by analysing the arrangement of these three motifs in the ABC transporters. The ABC proteins with high similarity had the same motif and gene structure, whereas ABC proteins containing different motifs indicated the diversity of gene functions. Moreover, the subcellular localisation of these ABC transporters has been predicted, together with the gene names (Additional le 2: Table S1), indicating that most of these transporters were located in the plasma membrane.
Phylogenetic analysis and subfamily classi cation Phylogenetic analysis was used to con rm the subfamily classi cation of S. miltiorrhiza ABC transporters. The 114 ABC transporters were divided into eight subfamilies, including 3 ABCA, 31 ABCB, 14 ABCC, 2 ABCD, 1 ABCE, 7 ABCF, 46 ABCG and 10 ABCI transporter genes (Fig. 1). The distribution of the ABC subfamily of S. miltiorrhiza was similar to that of other plants, and the number of genes in the ABCG subfamily was signi cantly higher than that of other subfamily. A phylogenetic tree was constructed using the ABC transporter of S. miltiorrhiza and the ABC proteins identi ed in other plants to infer the function of the transporter of S. miltiorrhiza. Such functional ABC proteins identi ed in other plants were listed in Additional le 3: Table S2.
Subfamily analysis of ABC transporters in S.miltiorrhiza

ABCA subfamily
The plant ABCA subfamily included one full-sized ABC one homolog (AOH) and several half-sized ABC two homologs (ATH). In Arabidopsis, AtABCA1, also known as AtAOH1, was the only full-sized ABCA transporter and was the largest ABC protein encoding 1,882 amino acid residues with domains arranged in a forward direction (TMD1-NBD1-TMD2-NBD2) [2,8]. AtAOH1 was probably involved in pollen germination and seed lipid accumulation and mobilisation [6]. ATHs were half-sized transporters with domains ranging in forward direction (TMD1-NBD1). These transporters were only found in plants and prokaryotes [26,27]. The expression of AtATH14 and AtATH15 was down-regulated when the seedlings were grown in hydroponic culture with long-term NaCl treatment [28].
Three genes (SMil_00000810, SMil_00004803, SMil_00004804) were annotated to be ABCAs in the S. miltiorrhiza genome (Fig. 2). SMil_00000810 was a full-sized ABCA transporter with high sequence homology to AtABCA1, namely, SmABCA1 (Table 1). SmABCA1 was also a large ABC transporter of S. miltiorrhiza, encoded by 1,978 amino acid residues. AtABCA1 might play a role in lipid accumulation and transport during seed maturation and germination in Arabidopsis, respectively [6]. Compared with other tissues, SMil_00000810 was highly expressed in the roots of S. miltiorrhiza (Table 1), implying that SmABCA1 might have a similar function to AtABCA1 in Arabidopsis. On the contrary, SMil_00004803 and SMil_00004804 were half-sized transporters and might be ATH homologues in the S. miltiorrhiza genome.

ABCB subfamily
The ABCB subfamily, the second largest subfamily, consisted of full-sized members, known as MDR or Pglycoprotein (PGP), and half-sized transporters TAP and ATM [7]. The domains of ABCB transporters were arranged in a forward direction: TMD1-NBD1-TMD2-NBD2. A full-sized MDR encoded approximately 1,200 amino acid residues [2]. Most MDRs from animals were plasma membrane e ux pumps, which were capable of transporting amphiphilic cations and cationic phospholipids [29]. AtMDR1 (AtPGP1) was the rst cloned and identi ed ABC transporter with multiple herbicide tolerances in plants [30]. MDR played an important role in bidirectional auxin transport [31], stomatal regulation [32] and metal tolerance in Arabidopsis [33], most of which were located in the plasma membrane of plants [34]. ATM was involved in the biogenesis of Fe-S clusters in the mitochondria [35].
Thirty-one genes were assigned to the ABCB subfamily in S. miltiorrhiza, 17 of which were full-sized transporters ( Table 1; Fig. 3). SMil_00017134 (SmABCB10), SMil_00002175 (SmABCB11) and SMil_00005832 (SmABCB13), encoding full-sized transporters, had sequence homology with Arabidopsis AtABCB1 [36] and AtABCB19 [37] (Fig. 3) and OsABCB14 [38] and tomato SlABCB4 [39], all of which were involved in auxin transport. The expression pro les of these three transporter genes had no tissue speci city in S. miltiorrhiza (Table 1). SMil_00020066 (SmABCB30) was highly expressed in the roots of S. miltiorrhiza, particularly in the periderm ( Table 1). The tissue-speci c expression of SmABCB30 was similar to that of the berberine transporter CjABCB2 in Coptis chinensis [40], indicating that SmABCB30 might be involved in the transport of secondary metabolites of S. miltiorrhiza. We also found that SMil_00018186 (SmABCB29), SmABCB30 and SMil_000 15238 (SmABCB31) showed sequence homology with AtABCB4/AtPGP4 and AtABC21 (Fig. 3), whereas the latter two transporters were responsible for the auxin transport in Arabidopsis [41,42]. The full-sized transporters of SMil_00016160 (SmABCB14) were highly expressed in the owers, whereas SMil_00001053 (SmABCB28) and SMil_00015518 (SmABCB18), likely to SmABCB10, were actively expressed in the roots (Table 1). SMil_00009650 was clustered nearly to AtABCB15, which was implicated in auxin transport of Arabidopsis [43]. We hypothesised that these SmABCB genes might participate in auxin transport of S. miltiorrhiza. The half-sized transporter SMil_00021941 (SmABCB9) was the only ATM homologue in S. miltiorrhiza, which was particularly similar to the ATM of AtABCB23, AtABCB24 and AtABCB25 in Arabidopsis (Fig. 3). These three transporters in Arabidopsis were involved in the biogenesis of Fe/S clusters [35], and their expression was up-regulated after methyl jasmonate (MeJA) treatment, which was similar to the MeJA-induced expression pro le of SmABCB9. The half-sized transporter SMil_00007974 (SmABCB4) was highly expressed in all organs (Table 1). It clustered closer to AtTAP1 (Fig. 3), which was involved in aluminium sequestration [33]. We hypothesised that SmABCB9 and SmABCB4 might be related to the metal tolerance of S. miltiorrhiza. ABCC family ABCC, also known as MRP, was encoded by at least 1,500 amino acid residues and comprised only the full-sized ABC transporter in Arabidopsis [6]. Compared with MDR, MRP harboured the ABCC-speci c additional hydrophobic N-terminal transmembrane domain (TMD0) [44]. The domains of the ABCCs were arranged in a forward direction (TMD0-TMD1-NBD1-TMD2-NBD2) [6]. Most MRPs in plants were located in the vacuole membrane, and few have been reported to reside on the plasma membrane [10,45,46]. MRPs have been identi ed in Arabidopsis and maize [44,47,48]. MRP involved heavy metal tolerance [49,50], stomatal regulation, glutathione S-conjugate transport [51] and phytate storage in plants [46]. In addition, MRP was responsible for the transport of secondary metabolites in several plants. For example, ZmMRP3 transported anthocyanins in corn [52], and CsABCC4 of saffron mediated crocin accumulation in the vacuoles [53].
The transporter genes of the ABCC subfamily were expressed in all organs and tissues of S. miltiorrhiza (Table 1). SMil_00016360 (SmABCC2) and SMil_00016361 (SmABCC1) were preferentially expressed in the roots of S. miltiorrhiza (Table 1), and they were homologous to AtABCC11, AtABCC12, AtABCC1 and AtABCC2 in A.thaliana (Fig. 4). Given that AtABCC1 and AtABCC2 were involved in the transport of folic acid, glutathione S-conjugates and chlorophyll [51], and the function of AtABCC11 and AtABCC12 was unknown, SmABCC1 was predicted to be related with the transport of primary metabolites in S. miltiorrhiza. SMil_00015789 (SmABCC5) was constitutively expressed in all organs (Table 1) and clustered with CsABCC4a and AtABCC4 (Fig. 4). Considering that CsABCC4a acted as a crocin transporter located on the stigma vacuole membrane of Crocus sativus (saffron) [53], whereas AtABCC4 participated in the transport of folic acid and glutathione S-conjugate to regulate stomatal movement [54], we presumed that SmABCC5 might be similar to its homologous protein in S. miltiorrhiza. SMil_00002827 (SmABCC4) was highly homologous to ZmMRP3 in maize [52] and VvABCC1 in grape [55], and the latter two transporters were related to anthocyanin transport (Fig. 4). Compared with other organs, the expression of SmABCC4 in the leaves was higher under MeJA induction (Table 1), and this ABC transporter might be involved in the transport of anthocyanin in S. miltiorrhiza leaves. SMil_00028509 (SmABCC8) was located on another branch near SMil_00002827 and was highly expressed in the leaves (Table 1), implying that SmABCC8 might also participate in the transportation of substances in the leaves (Fig. 4). SMil_00004040 (SmABCC11) was highly expressed in the owers and roots, and its homologue AtABCC5 in Arabidopsis was related to the storage of phytate and loading of InsP6 in the seeds [46], indicating that SmABCC11 might contribute to the accumulation of phytic acid in S. miltiorrhiza. SMil_00020247 (SmABCC13) was highly expressed in the leaves and roots (Table 1) and clustered with AtABCC3 and AtABCC6 (Fig. 4). AtABCC3 and AtABCC6 were related to heavy metal tolerance [56,57], inferring that SmABCC13 might be involved in the heavy metal tolerance of S. miltiorrhiza.

ABCD subfamily
The ABCD subfamily, also known as PMP, was located in the peroxisome membrane. In plants, this subfamily contained full-sized and half-sized transporters. The full-sized transporter AtABCD1 was related to the import of long-chain fatty acyl-CoA into peroxisomes in Arabidopsis [58] and transport of 12-oxophytodienoic acid [59] and jasmonic acids [60]. Two ABCD members, SMil_00013326 (SmABCD1) and SMil_00009714 (SmABCD2), were found in the S. miltiorrhiza genome (Table 1; Fig. 5). SmABCD1 was constitutively expressed in all organs and was homologous to AtABCD1 (Table 1; Fig. 5). We hypothesized that SmABCD1 had a similar function to AtABCD1 in S. miltiorrhiza. ABCE and ABCF subfamilies ABCE, also known as RNase L inhibitor (RLI), was a soluble protein with only two conserved NBDs and without any detectable TMD, which was conserved in eukaryotes and archaea. In Arabidopsis, AtABCE1 and AtABCE2 (AtRLI2) were involved in RNA interference (RNAi) regulation other than transport [61,62]. AtRLI2 catalysed the conversion of mRNA and participated in the biogenesis of the ribosome and initiation of translation in Arabidopsis [62]. ABCF, also known as a GCN homologue, similar to ABCE, was a soluble protein containing only two fused NBDs. Yeast GCN20, as the translational regulator of GCN4, elongated the ribosome during the activation of eIF2a kinase GCN2 [63].
Only SMil_00000396 (SmABCE1) was assigned to the ABCE subfamily of the S. miltiorrhiza genome, and it was constitutively expressed in all organs (Table 1; Fig. 5). Based on the functions of their homologues AtABCE1 and AtABCE2 in Arabidopsis, SmABCE1 might play roles in the regulation of gene silencing. The ABCF subfamily of S. miltiorrhiza contained seven members, four of which (SMil_00025510, SMil_00013582, SMil_00023741 and SMil_00004895) were highly expressed in all organs (Table 1; Fig.  5). Amongst the members, SMil_00013582 (SmABCF6) was signi cantly expressed in high abundance in the leaves and was down-regulated after treatment with MeJA ( Table 1). Considering that the homologues of SmABCF6 in yeast and humans were involved in the regulation of gene expression [63], SmABCF6 might negatively regulate the expression of leaf tissue-speci c genes under MeJA-induced conditions.

ABCG subfamily
The ABCG subfamily was the largest subfamily in plants, which was represented by full-sized PDR and half-sized WBC transporters. The NBD-TMD domains of this subfamily were arranged in opposite directions. Most of the characterised ABCGs were located in the plasma membrane [64,65]. To date, SpTUR2 was an early identi ed transporter protein of the ABCG subfamily, which involved in the transport of sclareol and resistance of herbicides [66]. In addition, the transporters of the ABCG subfamily were related to the transport of paraquat, thereby changing the tolerance of plants to herbicides [67]. ABCG transporters were widely involved in the transport of various compounds in plants [68,69]. Lr34 was involved in the resistance of wheat to various fungal pathogens [70], and CrTPT2 was responsible for the transport of vinblastine in Catharanthus roseus [71]. CsPDR8 and CsPDR12 were related to the hormone response of cucumber [72]. StPDR2 conferred resistance to the biotic and abiotic stresses in tomato [73], and PhPDR2 was identi ed as a petuniasterone transporter in leaves and trichomes of Petunia hybrida [74]. Four PDR genes have been identi ed in tobacco, amongst which, NpABC1 and NtPDR1 were involved in the secretion of antifungal terpenoids [75,76]. NbABCG1/2 was involved in the export of antimicrobial diterpenes and capsidiol for defence against Phytophthora infestans [77], and NtPDR3 was induced to express iron de ciency in the culture medium [78].
ABCG was also the largest subfamily of ABC transporters in S. miltiorrhiza, including 20 PDRs and 26 WBCs (Table 1; Fig. 6). Four genes (SMil_00010949, SMil_00020022, SMil_00027268 and SMil_00004712) had tissue-speci c expression pro les in this subfamily, all of which were highly expressed in the roots of S. miltiorrhiza (Table 1). Notably, SMil_00027268 (SmABCG4) was the most highly expressed gene in the periderm of S. miltiorrhiza roots ( Table 1). Given that tanshinone was synthesized and accumulated in large amounts in the roots of S. miltiorrhiza, particularly in the periderm tissues [24], these four transporters might be related to the transport of tanshinone in S. miltiorrhiza. Phylogenetic analysis showed that SmABCG4 and SMil_00010949 (SmABCG40) were clustered relatively near to the ginsenoside transporter PgPDR3 [86] and the antifungal terpenoid transporter NpABC1 and NtPDR1 [75,76] (Fig. 6). SMil_00020022 (SmABCG46) and SMil_00004712 (SmABCG44) were closely related to AtABCG39/AtPDR11 [67] and AtABCG34 [68] in Arabidopsis, which participated in the stress response of Arabidopsis. MeJA induced the expression of SmABCG46 and SmABCG44 at different levels, which was identical to the MeJA induction of AtABCG34 in Arabidopsis (Table 1). Another full-sized transporter, SMil_00016963 (SmABCG45), possessing the same gene structure and abundance as SmABCG46, was highly expressed in the roots of S. miltiorrhiza (Additional le 1: Figure S1; Table 1). These ve genes of the SmABCG subfamily might be involved in terpenoid transport in S. miltiorrhiza, which might mediate the stress responses of this medicinal plant. SMil_00023314 (SmABCG35) was only expressed in the owers, although it has the same gene structure as SmABCG46, suggesting that this gene might be involved in the transport of substances in the owers of S. miltiorrhiza (Table 1; Additional le 1: Figure S1).
SMil_00004104 (SmABCG32) was a full-sized transporter and highly expressed in the leaves. Its homologous protein CrTPT2 was responsible for the transport of catharanthine [71], suggesting that SmABCG32 might be involved in the transport of secondary metabolites in the leaves of S. miltiorrhiza. In addition, six half-sized WBCs that were expressed in various organs, including SMil_00027466 (SmABCG12), showed high expression levels in the owers, and its homologue AtABCG25 participated in the export of abscisic acid [65], indicating that SmABCG12 might be involved in the transport of plant hormones in the owers of S. miltiorrhiza. SMil_00015148 (SmABCG19) was highly expressed in the owers and was homologous to AtABCG11 [79] and GhWBC1 [85], suggesting that SmABCG19 likely played roles in the transport substances that were related to the growth and development of S. miltiorrhiza. SMil_00000399 (SmABCG27) and SMil_00005271 (SmABCG28) showed the same expression patterns, both of which were half-sized proteins and expressed in all organs other than the leaves (Table 1). Their homologue AtABCG14 mediated the root-to-shoot translocation of trans-Zeatin in Arabidopsis [87]. Thus, SmABCG27 and SmABCG28 likely involved in the hormone transport of S. miltiorrhiza. SMil_00010332 (SmABCG15) was highly expressed in the leaves and induced by MeJA ( Table 1), indicating that SmABCG15 might participate in the MeJA signal transduction pathway.

ABCI subfamily
No ABCH subfamily was found in plants; thus, an additional subfamily (ABCI) containing prokaryotic-type ABC proteins was used to instead of ABCH. ABCI was designated as non-intrinsic ABC protein (NAP) with only one NBD. The Arabidopsis genome contained 15 ABCIs, whereas the rice genome contained 10 members of this subfamily [2,8]. The ABCI subfamily of S. miltiorrhiza consisted of 10 genes (Fig. 7), all of which contained only one soluble NBD. These ABCI transporters were expressed in all organs of S. miltiorrhiza (Table 1). SMil_00009816 (SmABCI4) might be involved in the biosynthesis of Fe/S clusters in the leaves because its expression pro le was similar to the homologous gene AtABCI6 [88]. SMil_00003192 (SmABCI5) had a high degree of homology with ABCI13, which involved in the formation of plastid lipids [3]. SMil_00001364 (SmABCI2) showed high similarity to AtABCI1, which was related to the maturation of cytochrome c [89]. The results showed that SmABCI2 and SmABCI5 might have similar functions to their homologous transporters in Arabidopsis.

Veri cation of gene expression of candidate transporters in the transport of tanshinone and SA
The tissue-speci c expression of some transporter genes might be related to their function in speci c tissues or organs. By contrast, some genes showed indistinguishable expression pro les in all tissues, suggesting that they might play a role in the transport of basic substances in cells, such as transporter genes that were involved in the transport of primary metabolites. Considering that tanshinone and SA were primarily synthesised and accumulated in the roots of S. miltiorrhiza [22,24], we hypothesised that the highly abundant transporter genes expressed in the roots of S. miltiorrhiza might be related to the transportation of tanshinone and SA. Based on the characteristics of gene expression pro les (Table 1), we screened out 18 candidate genes that were highly expressed in the roots for qRT-PCR veri cation (Fig.  8), including 1 ABCA (SmABCA1), 5 ABCBs (SmABCB10, SmABCB13, SmABCB18, SmABCB28 and SmABCB30), 4 ABCCs (SmABCC1, SmABCC2, SmABCC11 and SmABCC13) and 8 ABCGs (SmABCG8, SmABCG27, SmABCG28, SmABCG40, SmABCG44, SmABCG45 and SmABCG46). Amongst these candidate ABC genes, we found that the expression patterns of SMil_00020022 (SmABCG46), SMil_00010949 (SmABCG40) and SMil_00027268 (SmABCG4) were nearly identical to that of CYP76AH1 and SmCPS1, which were key enzyme genes involved in the biosynthetic pathway of tanshinone (Fig. 9). Moreover, SMil_00016361 (SmABCC1) was co-expressed with CYP98A14 and SmRAS, which encoded the key enzymes in the biosynthetic pathway of SA in S. miltiorrhiza (Fig. 9). The subcellular localisation of these four candidate transporter genes related with tanshinone and SA transport in S. miltiorrhiza was predicted to be located in the plasma membrane (Additional le 2: Table S1). Therefore, these four candidate ABC transporters co-expressed with key enzyme genes in the biosynthesis of tanshinone and SA likely participated in the intracellular transport of these two active compounds in S. miltiorrhiza.
In addition, the 3D model of these four candidate ABC transporters was created by Swiss-model using 6vxi.1.A (ABCG2) and 6pza.1.A (ABCC8) as templates, which had a broad substrate speci city, in the Swiss database. The amino acid sequence homology between SMil_00027268, SMil_00020022, SMil_00010949 and SMil_00016361 with the template protein was 25.68%, 24.79%, 25.98% and 30.46%, respectively (Additional le 4: Figure S2).

Discussion
In plants, the ABC protein was initially identi ed as a transporter involved in the nal detoxi cation process [90]. Since this discovery, many reports have shown that the function of this type of transporter extends far beyond detoxi cation. In recent years, ABC transporters and have become a major focus for research in plants. This is not only due to their overall roles in a variety of processes, such as pathogen reaction, surface lipid deposition, accumulation of phytic acid in seeds, and the transport of plant hormones, but more speci cally, due to their essential roles in plant growth and development, response to abiotic stress and the interaction between plants and the environment.
In this study, a total of 114 ABC proteins were identi ed in the genome of S. miltiorrhiza, amongst which, 86 members encoded for ABC transporters with TMDs, including 50 full-sized ABC transporters. The number of the ABC proteins and the full-sized ABC transporters of S. miltiorrhiza was similar to that of Arabidopsis because the Arabidopsis genome contained 120 gene ABC proteins, including 51 full-sized transporters [2,8]. The total number of genes encoding ABC proteins was nearly identical in the two species, despite of the large differences in genome size (615 Mb versus 125 Mb) and gene content (30,478 versus 25,498) [8,17]. The identi cation of S. miltiorrhiza ABC proteins and their comparative analysis with the Arabidopsis ABC transporters revealed an obvious conservation of ABC transporters between the two species. A single plant species can synthesize thousands of different molecules, and these molecules can be transported across the plasma membrane of one or more organelles, which probably explains the large size of the ABC transporter gene family in plants [91].
On the basis of the phylogenetic analysis, the S. miltiorrhiza ABC proteins were grouped into their respective subfamilies, from ABCA to ABCI, except ABCH. Members of ABCG (46 genes), ABCB (31 genes) and ABCC (14 genes) subfamilies were the most abundant, whiles ABCA, ABCD and ABCE subfamilies were less abundant (Table 1), which was similar to the subfamily distribution of tomato [13] and grape [9]. The members of most subfamilies, except of the ABCI (NAP) subfamily, grouped more tightly with each other than with the members of other subfamilies (Fig. 1). Similarly, some members of NAPs also were not clustered to a group with high homology in Arabidopsis [2].
In Arabidopsis, various subfamilies of ABC transporters contain different conserved domains and perform various biological functions. Similar to Arabidopsis [2] and grape [9], only one full-sized ABC transporter (SMil_00000810) has the longest gene sequence in the S. miltiorrhiza genome, belonging to the ABCA subfamily, and homologous to AtABCA1 (Table 1; Fig. 2), which implied the function of SMil_00000810 maybe similar to AtABCA1. There were two members (SMil_00004803 and SMil_00004804) belonging to the ATH homologs in the S. miltiorrhiza genome, was the half-size transporter (Table 1; Fig. 2), whereas the Arabidopsis genome contains 11 ATH members [2]. The expression of AtATH14 (AtABCA10) and AtATH15 (AtABCA11) in Arabidopsis was regulated in response to salt stress [28]. Phylogenetic analyses showed that SMil_00004803 clustered closest to AtABCA11in Arabidopsis, while SMil_00004804 clustered with the group containing AtABCA10 (Fig. 2), indicating that these two ATH homologues may be involved in the stress response of S. miltiorrhiza. In Arabidopsis, the AtABCB1, AtABCB4, AtABCB14, AtABCB15, AtABCB19 and AtABCB21 have been reported to be involved in polar auxin transport as their mutants showed reduced auxin transport [41-43, 92, 93]. It contributed us to speculate that the functions of the homologous to these Arabidopsis ABCB in S. miltiorrhiza involved in the polar auxin transport (Fig. 3). These S. miltiorrhiza ABCB transporters included SMil_00000837 and SMil_00005832 homologous to AtABCB1, SMil_00018186, SMil_00020066 and SMil_00015238 homologous to AtABCB4 and AtABCB421, SMil_00009650 homologous to AtABCB15, SMil_00017134 and SMil_00002175 homologous to AtABCB19 (Fig. 3). AtABCC1 and AtABCC2 contributed to the tolerance of Arabidopsis to Cd (II) and Hg (II), while in the absence of AtABCC2, AtABCC1 conferred great tolerance to divalent heavy metals [94]. Overexpression of AtABCC1 in Arabidopsis enhanced the tolerance and accumulation of Cd(II) [94]. Similarly, AtABCC3 [95] and AtABCC6 [56] might also contribute to improve plant heavy metal tolerance. Otherwise, the vacuolar membrane-localized ABC transporters, such as AtABCC1 [47] and AtABCC4 [54], regulated the concentration of folate in the cytoplasm by transporting excess folic acid to the vacuole, indicating that ABCC transporters, such as AtABCC1, are important for folic acid storage. The knockout mutants of Arabidopsis AtABCC5/MRP5 exhibited a low phytate phenotype [46]. The subsequent transport experiments using vesicles isolated from yeast heterologous expressing AtABCC5 con rmed that AtABCC5 acts as a phytate transporter [46]. These results indicated that ABCC transporters played important roles in transporting of primary products and improving the heave metal tolerance in plants. These results provide valuable clues for studying the ABCC genes in S. miltiorrhiza, such as SMil_00016361, SMil_00019856, SMil_00015789, SMil_00004040 and SMil_00000135 (Fig. 4). Like other plants, ABCG was the largest subfamily in the ABC transporter family of S. miltiorrhiza (Table 1; Fig. 6). Several members of ABCG subfamily in Arabidopsis, such as AtABCG11, 12, 13, and 32, were involved in the transport of cuticular lipid precursors for cuticle layer formation from epidermal cells to the surface [96]. Whereas the AtABCG25 [65] and AtABCG40 [64] were high a nity ABA transporters. ABCG14 participated in transport of cytokinin [87]. AtABCG36 regulated the sensitivity of plants to the auxin precursor indole-3-butyric acid [97]. Whereas the homologous of AtABCG36, AtABCG37 participated in the secretion of scopoletin and derivatives by Arabidopsis roots in response to iron de ciency [98]. These results are su cient to show the diversity of gene functions in the ABCG subfamily. It was worth noting that several members of the ABCG subfamily participated in pathogen defense and/or the crosstalk between plants and microorganisms, with secondary metabolitedependent. The tanshinone and SA in are also secondary metabolites with diverse pharmacological activities in S. miltiorrhiza. Some members of the ABCG subfamily may participate in the transport of these active compounds in this medicinal plant.
Gene expression pro les are complex phenotypic features that can re ect the biological processes of target genes involved in metabolism, tissue and organ development and differentiation, and response to environmental changes in plants. In this study, we analysed a subset of the gene expression pro les in several organs/tissues of S. miltiorrhiza. Since the genes in the same biosynthetic pathway are generally coexpressed, we compared the expression patterns of all the 18 candidate ABC transporter genes with the upstream genes encoding SmCPS1, CYP76AH1, RAS and CYP98A14, those were key enzymes involved in tanshinone and SA biosynthesis, respectively (Fig. 9). The coexpression analysis suggested that three ABCG (SMil_00020022, SMil_00010949 and SMil_00027268) and one ABCC (SMil_00016361) members might be involved in transport of tanshinone and SA in S. miltiorrhiza, respectively (Fig. 9). The coexpression of this transporter gene and key enzyme genes in the secondary metabolic pathway indicated that these transporters might be involved in the transport of secondary metabolites. For example, CsABCC4a and CsABCC2 were highly expressed in the stigmas of C. sativus and enabled to transport crocin in yeast microsomes and were highly co-expressed with total crocin levels and/or CsCCD2, which was the rst dedicated enzyme in the crocin biosynthetic pathway [53]. ABCG14 was highly co-expressed with cytokinin biosynthesis and was the major root-to-shoot cytokinin transporter [87]. We anticipate that a functional study in the near future will elucidate the molecular and physiological functions of the lead candidate ABC transporter involved in tanshinone and SA transport in this medicinal plant.
In addition, we found and con rmed the existence of tissue-speci c gene expression pro les, which represent the predictive tissue-dependent functions of these ABC transporters in S. miltiorrhiza (Table 1; Fig. 8). These results provided not only a valuable information for investigating the functions of the ABC transporter gene in S. miltiorrhiza but also a practical procedure for screening candidate genes involved in bioactive secondary metabolite transport in medicinal plants based on genome and transcriptome dataset.

Conclusion
In this study, we identi ed and analysed ABC transporters in S. miltiorrhiza for the rst time and provided the fundamental and detailed information about S. miltiorrhiza ABC proteins. The information included all the ABC proteins in S. miltiorrhiza with the gene IDs, protein topology, gene expression pro les and phylogenetic trees of subfamily members and orthologues in other plants, showing the reported physiological functions. Based on the previous studies on the functions of ABC genes, the functions of some ABC transporters with domain or expression characteristics were hypothesised in S. miltiorrhiza. Combined phylogenetic and co-expression analyses identi ed three genes (SMil_00020022, SMil_00010949, SMil_00027268) and one gene (SMil_00016361) to be the lead candidates involved in tanshinone and SA transport, respectively. The transporters identi ed in the ABCB and ABCC subfamilies might be involved in the transport of secondary metabolites of S. miltiorrhiza. In addition, the transporters might be involved in the transport of anthocyanins, auxin and metal resistance have been identi ed in several ABC subfamilies of S. miltiorrhiza. Our study outlined the ABC proteins in the S. miltiorrhiza genome and explained their possible transporting pathways for some compounds, laying an important foundation for further research on the metabolic regulation, synthetic biology and utilisation of these compounds in S. miltiorrhiza. Our analysis provides new insight into the diversity and the predicted function of the entire ABC transporters in S. miltiorrhiza compared with Arabidopsis. These results will provide new insights into the function of ABC transporters in S. miltiorrhiza.

Methods
Identi cation of ABC transporter genes in the S. miltiorrhiza genome BLAST was used to align all the proteins in the S. miltiorrhiza genome with the ABC proteins in the Arabidopsis TAIR11 database (E value is less than 1e-5) and identify ABC homologues in S. miltiorrhiza. HMMER (https://www.ebi.ac.uk/Tools/hmmer/) was used to characterise the topology of ABC proteins containing TMD and NBD domains. The TMD domain in these ABC proteins was identi ed by PF12698, PF06472 and PF00664, and the NBD domain was identi ed by PF00005 in Pfam database (https://pfam.xfam.org/). A protein with at least one NBD domain was predicted to encode an ABC transporter. The domain analysis also provided evidence for the subfamily classi cation of the ABC family in S. miltiorrhiza. The obtained S. miltiorrhiza ABC protein sequences were submitted to the MEME Web server (http://meme-suite.org/) to con rm the relationship between the conserved motifs and genes. The subcellular location of S. miltiorrhiza ABC proteins was predicted by the WOLF PSORT online tool (https://wolfpsort.hgc.jp/). A 3D model of the selected protein was created using Swiss-model (https://swissmodel.expasy.org/interactive/).

Phylogenetic analyses
Clustal X (http://www.clustal.org/) was used to perform sequence alignment on the deduced amino acid sequence of S.miltiorrhiza ABC proteins, and then, MEGA 6 was used to construct a phylogenetic tree with 1,000 bootstrap repeats through the neighbor-joining method. The maximum likelihood method was used to establish phylogenetic trees of S.miltiorrhiza ABC transporter subfamilies with the ABC transporters that have been functionally identi ed in Arabidopsis and other plants to predict the function of these transporters in S.miltiorrhiza. Phylogenetic trees were embellished using the interactive Tree Of Life Platform (https://itol.embl.de/).  [23][24]. The expression pro les of 112 putative S. miltiorrhiza ABC genes were analysed with these transcriptome data. The candidate genes that were highly expressed in the roots of S. miltiorrhiza were selected for quantitative reverse transcription polymerase chain reaction (qRT-PCR) veri cation.
qRT-PCR veri cation of gene expression pro les The relative expression levels of 18 candidate ABC genes in different organs/tissues of S. miltiorrhiza were analysed by qRT-PCR. Total RNA was extracted from the ower, stem, leaf, root and tissues of the periderm, phloem and xylem isolated from the roots of 2 year-old S. miltiorrhiza, according to the manufacturer's instructions of the RNAprep Pure Plant Kit (TIANGEN, China), and reversed into cDNA using the Promega GoScript Reverse Transcription System (Promega, Beijing, China). qRT-PCR was performed on an ABI PRISM 7500 real-time PCR system (Applied Biosys) using SYBR Premix Ex Taq™ (Takara, Beijing, China) with the following program: 95 °C for 30 s, 1 cycle; 95 °C for 5 s and 60 °C for 34 s, 40 cycles. The relative gene expression level was calculated using the 2 −ΔΔCT method [99] with the SmActin gene (Genbank number HM231319.1) as an internal reference. The experiments were performed in three independent biological experiments with three technical replicates. qRT-PCR was performed to determine the relative expression levels of the 18 candidate ABC genes along with CYP98A14, SmRAS1, CYP76AH1 and SmCPS1, which were used as positive genes to calculate the correlation coe cient of coexpression for these ABC genes with the key enzyme-encoding genes involved in the biosynthesis of tanshinones and SAs. The primers used in qRT-PCR were listed in Additional le 5: Table S3.

Declarations
Ethics approval and consent to participate Not applicable.

Competing interests
The authors declare that they have no competing interests.

Consent for publication
Not applicable.

Availability of data and materials
The datasets supporting the conclusions of this article are included with in the article and its additional les. Authors contributions HL conceived and designed the work. LY drafted the manuscript and was responsible for the data analysis, collected the sample and performed RT-qPCR. JZ and HC assisted to collected the sample and manuscript revision. All authors read and approved the nal version of the manuscript.