Discovery of a multigene family of aquaporin silicon transporters in the primitive plant Equisetum arvense

Authors

  • Caroline Grégoire,

    1. Département de Phytologie – Faculté des Sciences de l’Agriculture et de l’Alimentation, Centre de Recherche en Horticulture, Université Laval, Pavillon Paul-Comtois, Québec city, QC, G1V 0A6, Canada
    Search for more papers by this author
  • Wilfried Rémus-Borel,

    1. Département de Phytologie – Faculté des Sciences de l’Agriculture et de l’Alimentation, Centre de Recherche en Horticulture, Université Laval, Pavillon Paul-Comtois, Québec city, QC, G1V 0A6, Canada
    Search for more papers by this author
    • Present address: Institut National de la Recherche Agronomique, UMR Santé et Agroécologie du Vignoble, 71 Av Edouard Bourleaux, 33140 Villenave d'Ornon, France.

  • Julien Vivancos,

    1. Département de Phytologie – Faculté des Sciences de l’Agriculture et de l’Alimentation, Centre de Recherche en Horticulture, Université Laval, Pavillon Paul-Comtois, Québec city, QC, G1V 0A6, Canada
    Search for more papers by this author
  • Caroline Labbé,

    1. Département de Phytologie – Faculté des Sciences de l’Agriculture et de l’Alimentation, Centre de Recherche en Horticulture, Université Laval, Pavillon Paul-Comtois, Québec city, QC, G1V 0A6, Canada
    Search for more papers by this author
  • François Belzile,

    1. Département de Phytologie – Faculté des Sciences de l’Agriculture et de l’Alimentation, Université Laval, Pavillon Charles-Eugène-Marchand, Québec city, QC, G1V 0A6, Canada
    Search for more papers by this author
  • Richard R. Bélanger

    Corresponding author
    1. Département de Phytologie – Faculté des Sciences de l’Agriculture et de l’Alimentation, Centre de Recherche en Horticulture, Université Laval, Pavillon Paul-Comtois, Québec city, QC, G1V 0A6, Canada
    Search for more papers by this author

(e-mail richard.belanger@fsaa.ulaval.ca).

Summary

Plants benefit greatly from silicon (Si) absorption provided that they contain Si transporters. The latter have recently been identified in the roots of some higher plants known to accumulate high concentrations of Si, and all share a high level of sequence identity. In this study, we searched for transporters in the primitive vascular plant Equisetum arvense (horsetail), which is a valuable but neglected model plant for the study of Si absorption, as it has one of the highest Si concentrations in the plant kingdom. Our initial attempts to identify Si transporters based on sequence homology with transporters from higher plants proved unsuccessful, suggesting a divergent structure or property in horsetail transporters. Subsequently, through sequencing of the horsetail root transcriptome and a search using amino acid sequences conserved in plant aquaporins, we were able to identify a multigene family of aquaporin Si transporters. Comparison of known functional domains and phylogenetic analysis of sequences revealed that the horsetail proteins belong to a different group than higher-plant Si transporters. In particular, the newly identified proteins contain a STAR pore as opposed to the GSGR pore common to all previously identified Si transporters. In order to determine its functionality, the proteins were heterologously expressed in both Xenopus oocytes and Arabidopsis, and the results showed that the horsetail proteins are extremely efficient a transporting Si. These findings offer new insights into the elusive properties of Si and its absorption by plants.

Introduction

Silicon (Si) holds a unique position in plant biology. Although it is not recognized as an essential nutrient, its beneficial effects are widely acknowledged in protecting plants against a diverse range of biotic and abiotic stresses (Epstein, 1999; Fauteux et al., 2005). Plants absorb Si from the soil solution in the form of silicic acid [Si(OH)4], but vary greatly in their ability to accumulate Si, accumulating between 0.1 to 10% Si in top dry weight. There appears to be a direct correlation between Si absorption and benefits (Epstein, 1999). This ability to absorb Si is presumably linked to the presence of specific transporters (Ma et al., 2006).

Si transporters are elusive in all forms of life, and have been identified only in diatoms (Hildebrand et al., 1997) and more recently in the roots of some higher plants, including rice (Oryza sativa L.), barley (Hordeum vulgare L.), wheat (Triticum aestivum L.) and pumpkin (Cucurbita moschata Duch.), which are known to accumulate relatively high concentrations of Si (Ma et al., 2006; Mitani et al., 2011; Montpetit et al., 2012). Plant Si influx transporters belong to the ancient family of aquaporins, small membrane proteins that are part of the major intrinsic protein (MIP) family present in all kingdoms of life (Gomes et al., 2009). Aquaporins have six transmembrane domains and assemble into tetramers in the cell membrane (Maurel et al., 2008). Plant Si transporters belong to the NIP subfamily (nodulin 26-like proteins) of aquaporins, membrane channels that facilitate passive transport of water and/or small uncharged solutes (Gomes et al., 2009), including glycerol, ammonia, boric acid and silicic acid (Bienert and Chaumont, 2011). The selectivity of NIPs for a particular solute appears to be linked to two regions forming the pore in the central channel: two highly conserved NPA motifs and four amino acid residues forming the aromatic/arginine (ar/R) selectivity filter (Wu and Beitz, 2007). In the specific case of Si (silicic acid), only the residues GSGR appear to have the correct specificity for transport, as they are unique and common to the currently known plant Si influx transporters (Mitani-Ueno et al., 2011). Other plant aquaporins include the water channel PIPs (plasma membrane intrinsic proteins) and TIPs (tonoplast intrinsic proteins) (Gomes et al., 2009). Additionally, Si efflux transporters known as Lsi2 have been identified in the roots of some higher plants, including rice (Ma et al., 2007). As aquaporins are passive channels, the presence of active Lsi2 transporters facilitates loading of Si into the xylem.

The primitive vascular plant Equisetum arvense L. (horsetail) has one of the highest Si concentrations in the plant kingdom (Hodson et al., 2005), and requires Si to complete its lifecycle (Chen and Lewin, 1969). Despite its high affinity for Si, the mechanisms by which horsetail absorbs this element are unknown. Although it represents a valuable model to study and understand Si absorption in plants, it has not been exploited for this purpose to date. Many reasons may explain this apparent discrepancy, an important one being its large genome size of 14 Gbp (Obermayer et al., 2002; Bainard et al., 2011), which has greatly restricted the availability of genetic resources for this species. In addition, attempts to identify homologous horsetail influx Si transporters on the basis of conserved sequences from previously described Si transporters in higher plants may have proven difficult because of sequence divergence.

On the basis of the rapidly evolving field of high-throughput sequencing, we used an alternative approach to identify Si transporters in horsetail. As this species requires Si for survival, we hypothesized that transporters involved in Si absorption in horsetail roots may be identified through a transcriptomic approach (RNA-Seq), with subsequent screening based on common amino acid sequences of aquaporins. In this work, we describe the discovery of a multigene family of Si transporters in horsetail. These transporters have properties and activity distinct from the previously identified plant Si transporters, and may explain the remarkable Si absorption by horsetail.

Results

In order to identify horsetail Si transporters, we first relied on a conventional approach of homology cloning using sequences from higher plants. This approach yielded negative results, which suggested that horsetail Si transporters had divergent sequences. Sequencing and assembling the horsetail root transcriptome led to identification of in silico NIP contigs. Cloning full-length cDNAs and genes allowed confirmation of the horsetail NIP sequences. Horsetail proteins shared common pore residues, and modeling showed that silicic acid was a potential substrate. The function of three horsetail NIPs was then assessed using a Xenopus oocyte assay, confirming Si transport capacity. A horsetail Si transporter was also transferred into Arabidopsis, increasing Si accumulation in shoots. During analysis of the horsetail root transcriptome, four other non-NIP aquaporins were also identified and described.

Homology cloning attempts

The search for a Si transporter in horsetail based on sequence homology with transporters from higher plants proved unsuccessful. A PCR approach using primers derived from monocot and dicot Lsi1 sequences, including conserved loops B and E, did not generate any amplicons from horsetail root cDNA. Similarly, Southern blot analysis using three probes from rice Lsi1 did not yield any signal for horsetail genomic DNA (Figure S1).

Identification and analysis of horsetail NIPs in the root transcriptome

De novo assembly of the root transcriptome generated 7254 contigs with a mean length of 354 bp. Twelve contigs encoding putative NIPs were identified in the root horsetail transcriptome. These contigs were used as a starting point to design primer pairs for cloning NIP coding sequences (CDS) from the same root cDNA used in the root transcriptome experiment (see Table S1 for primer sequences). In the case of EaNIP3;1, a single 1248 bp horsetail candidate contig was identified that contained a 747 bp CDS coding for a putative NIP aquaporin (see Figure 5, and text below for description of horsetail aquaporin nomenclature ). Using the primer pair EcoRI-EaNIP3;1F/XbaI-EaNIP3;1R, a CDS was amplified and perfectly matched the assembled contig. For EaNIP3;2 and 3;3, five closely related contigs shared common sequences at the beginning and end of the CDS, which led to the design of a single primer pair (EcoRI-EaNIP3;2-3;3F /XbaI-EaNIP3;2-3;3R) that allowed amplification of two distinct PCR products. One of the in silico sequences perfectly matched EaNIP3;3, while another shared 96% identity with the closest contig. The EaNIP3;43;9 subgroup was more complex to resolve, as a single primer pair (Eco-RI-EaNIP3;4-3;9F/XbaI-EaNIP3;4-3;9R) derived from four closely related contigs led to cloning of six distinct CDS, none of which were identical to the in silico sequences. In all cases, CDS were cloned using a high-fidelity Pfu polymerase to minimize PCR errors, several clones were sequenced, and all RNA-Seq reads were mapped back using strict parameters to confirm sequencing results. The two remaining in silico contigs appeared to be chimeric, comprising the beginning of one protein and the end of another. Mixes of primer pairs (for example, EcoRI-EaNIP3;4-3;9F and XbaI-EaNIP3;1R) were used experimentally to rule out the existence of the chimeric contigs (no PCR amplification). Chimeric contigs were identified in the assembled transcriptome, but they do not correspond to the biological reality in planta. It was necessary to perform RACE PCR to obtain complete untranslated region (UTR) sequences, as assembly of the contigs in these regions was often incomplete. For the EaNIP3;43;9 subgroup, the cDNA sequences were extremely similar in some cases, and it was difficult to design distinct primers, which explains why some RACE PCRs in this subgroup were not completed (Table 1). RACE PCRs were performed using Taq DNA polymerase, which has a higher error rate than Pfu, but RNA-Seq reads were again used to confirm sequences and correct any enzyme errors. For EaNIP3;1, two polyadenylation sites, separated by 51 bp, were discovered at the 3′ end of the transcript.

Table 1.   Sequence statistics for horsetail MIPs identified by root RNA-Seq and/or cloning methods
Protein nameMean read coverage in roots (cDNA)CDS (bp)Gene (bp)5′ UTR3′ UTRExpression in shoots
  1. Using CLC Genomic Workbench, RNA-Seq reads were mapped back to cloned cDNA sequences (for NIPs) or in silico contigs (for PIPs and TIPs) to give a measure of expression by mean read coverage. Full-length genes were cloned and sequenced from horsetail genomic DNA. 5′ and 3′ UTR sequences were obtained by RACE PCR. Expression in shoots or roots was assessed by RT-PCR.

  2. aCloning not completed.

Sequences obtained by in silico analyses and cloning
 EaNIP3;14397477026202330Yes
 EaNIP3;2587893866100207No
 EaNIP3;34397806025234402Yes
 EaNIP3;4a127783 a a 216No
 EaNIP3;4b128783 a a 215No
 EaNIP3;51507834446 a a No
 EaNIP3;61217834282 a 215No
 EaNIP3;7a2047834423 a a No
 EaNIP3;7b189783 a 226 a No
 EaNIP3;81047834059212215No
 EaNIP3;9177783 a a a No
Sequences obtained and analysed in silico
 EaPIP1;1247894    
 EaPIP2;1263843    
 EaTIP1;1229765    
 EaTIP2;1191750    

For most of the described horsetail NIPs, corresponding genes were cloned and fully sequenced. They range in size from 3.9 to 7 kb (Figure 1a and Table 1). The intron–exon structure of the cloned genes was identified (Figure 1a), and all horsetail NIPs appear to share a common position and length of exons 2, 3 and 4. The EaNIP3;1 gene is 7 kb long. The sequence of the exons perfectly matched the sequenced cDNA. Two distinct genomic sequences were documented among the PCR amplification products. These showed no differences within the exons, and hence encode the same protein. Minor insertions and deletions were observed in introns 1 and 2. This pattern of variation suggests that these are allelic variants of a single gene. In the case of EaNIP3;2, two full-length 3.9 kb genes were identified. They encode the same protein, but have minor differences in all four introns. EaNIP3;3 is a 6 kb gene with no identified variants. For the EaNIP 3;43;9 subgroup, the very closely related sequences made it difficult to design specific primers. Using a single primer pair, several full-length genes of 4–4.5 kb were cloned, and four were matched to root cDNAs (Figure 1a and Table 1). Two putative pseudogenes were also identified, showing premature interruption of the reading frame. In some cases, allelic differences are present only in the non-coding regions, while in other cases, very closely related proteins have more than 95% sequence identity and are suspected to come from closely related, perhaps recently duplicated, genes. For further analysis, only the genes that were shown to be expressed in the roots were included, as it is not known whether the numerous other cloned NIP genes are functional.

Figure 1.

 Gene structure and expression pattern of horsetail NIPs.
(a) Gene structure of selected fully sequenced horsetail NIPs. Exons are boxed in gray.
(b) Expression pattern of horsetail NIPs in roots and shoots by RT-PCR.
(c) Digestion of EaNIP3;2 and EaNIP3;3 RT-PCR products from shoots using two restriction enzymes: SacI cleaves only the EaNIP3;2 PCR product, and ScaI cleaves only the EaNIP3;3 PCR product. Such digestion shows that only EaNIP3;3 is expressed in the shoots.

Horsetail NIPs have differential expression patterns and levels in roots and shoots

Mapping back RNA-Seq reads to cloned cDNAs allows comparison of expression levels between horsetail NIPs in roots. EaNIP3;1 and EaNIP3;3 show the highest mean read coverage of all NIPs, while EaNIP3;2 and the EaNIP3;43;9 subgroup show significantly lower expression (Table 1). Additionally, using RT-PCR and specific restriction enzymes, it was shown that only EaNIP3;1 and EaNIP3;3 are expressed in horsetail shoots (Figure 1b,c). To obtain an estimate of the abundance of these transcripts, reads were mapped back to the cloned cDNAs using strict alignment parameters. The mean read coverage varied from 58 for EaNIP3;2 to 439 for EaNIP3;1 and EaNIP3;3, the two most strongly expressed MIPs in horsetail roots (Table 1).

Homology modeling of EaNIP3;1 and its putative substrate

The four pore-forming residues of all horsetail NIPs are S (serine), T (threonine), A (alanine) and the conserved R (arginine) after the second NPA motif (Figure 2). In order to determine whether horsetail NIPs possessed the characteristics deemed essential for silicic acid transport, 3D modeling of one of the horsetail proteins, EaNIP3;1, was performed. The results showed that it contained the six transmembrane α-helices common to all MIP proteins (Figure 3). Amino acid positions 1–30 and 247–248 of EaNIP3;1 could not be modeled due to lack of homology with the crystal structure model. The substrate with the strongest total ligand–receptor interaction was silicic acid (−8.76 kcal mol−1). On the basis of this result, together with the conserved sequences of loops B and E, we tested the silicic acid transport activity of three horsetail NIPs in a heterologous system.

Figure 2.

 Alignment of NIP silicon transporter proteins from horsetail and two higher plants.
The six putative transmembrane regions are indicated by solid lines above the sequences. Loops A–E are indicated by dashed lines above the sequences. The two signature NPA motifs are boxed. ar/R filter residues (STAR for horsetail, GSGR for rice and pumpkin) are shaded in dark gray. The primary selectivity-related motifs P1–P5 are shaded in pale gray. The species acronym is included before the gene name: Cm, Cucurbita moschata; Ea, Equisetum arvense; Os, Oryza sativa.

Figure 3.

 Structural elements of EaNIP3;1 and its interaction with substrate.
(a) 3D model of EaNIP3;1 showing six transmembrane domains (H1–H6) connected by loops A–E (LA–LE) surrounding the central channel.
(b) Interaction between the four ar/R filter residues of EaNIP3;1 and the silicic acid substrate.

Si transport activity in Xenopus oocytes

We tested one representative member from each of the three horsetail NIP subgroups for Si transport in oocytes: EaNIP3;1, EaNIP3;3 and EaNIP3;4. Expression of rice or horsetail Si transporters in oocytes revealed an efficient transport activity in the presence of the substrate (Figure 4). Compared to control oocytes injected with water and incubated in medium supplemented with Si, oocytes injected with complementary RNA (cRNA) of OsLsi1, EaNIP3;1, EaNIP3;3 and EaNIP3;4 showed a significant increase in intracellular Si concentration after 60 min incubation in a 1.7 mm Si solution (Figure 4). The horsetail Si transporters EaNIP3;1 and EaNIP3;4 showed significantly higher import of Si than the rice transporter, reaching a level 10 times above the basal level. No differences in Si transport activity were detected between water-injected oocytes and oocytes incubated in solution that was not supplemented with Si.

Figure 4.

 Influx activity of Si transporters in Xenopus oocytes.
Si uptake was measured over time in Xenopus oocytes injected either with water (control) or cRNAs from rice (OsLsi1) or horsetail (EaNIP3;1, EaNIP3;3 or EaNIP3;4). Before sampling, oocytes were incubated for 0 or 60 min in Barth medium containing 1.7 mM Si. Values are means ± standard error (n = 4).

Horsetail NIPs form a family of closely related proteins

By comparing the amino acid sequence of the horsetail aquaporins described in this work with previously identified plant NIPs, the candidate proteins were all classified as NIP3s (Figure 5 and Figures S2 and S3). They appear to form three distinct subgroups, the first containing only EaNIP3;1, the second containing the closely related proteins EaNIP3;2 and 3;3, and the last comprising numerous sequences classified as EaNIP3;4–3;9. Within each subgroup, proteins share high percentage identities (>90%; Table S2).

Figure 5.

 Phylogenetic tree of plant NIP proteins for the nomenclature of horsetail NIPs, including silicon transporters from monocots and dicots (NIP III).
The species acronym is included before the gene name: At, Arabidopsis thaliana; Cm, Cucurbita moschata; Ea, Equisetum arvense; Gm, Glycine max; Hv, Hordeum vulgare; Os, Oryza sativa; Pp, Physcomitrella patens; Pt, Populus trichocarpa; Zm, Zea mays. Horsetail proteins with more than one isoform are marked ‘a’ and ‘b’ (putative allele at the gene level).

Expression analysis of the EaNIP3;1 transgene in Arabidopsis and dosage of Si in shoots

We selected EaNIP3;1 for further study of heterologous expression in Arabidopsis because it had a higher activity in oocytes and also a higher expression level than most other NIPs in horsetail roots. Under the control of the constitutive CaMV 35S promoter, the horsetail EaNIP3;1 gene is expressed in both roots and leaves of Arabidopsis (Figure 6a), and the expression appears to be relatively equal in both the presence and absence of Si. The Si content of Col-0 and p35S-EaNIP3;1 Arabidopsis plants was measured after 2 weeks of growth in substrate fertilized with regular or Si-supplemented modified Hoagland’s solution (Figure 6b). In the presence of Si, the transgenic line expressing the horsetail transporter showed a significant rise in Si content compared to the Col-0 line.

Figure 6.

 Heterologous expression of horsetail Si transporter EaNIP3;1 in Arabidopsis.
(a) Expression of EaNIP3;1 in roots and leaves of Arabidopsis in the presence or absence of Si using RT-PCR, with ubiquitin as an internal control.
(b) Si content in Arabidopsis Col-0 and a transgenic line expressing the horsetail Si transporter EaNIP3;1 under the control of constitutive CaMV 35S promoter, in the presence or absence of Si (n = 3).

Identification and phylogenetic analysis of horsetail TIPs and PIPs

All identified horsetail NIPs belong to the NIP3 group. However, four horsetail contigs coding for non-NIP aquaporins were also identified in the root transcriptome. Alignment of these horsetail proteins with classified MIPs from Physcomitrella patens, rice and Arabidopsis (Figure S4) and phylogenetic analysis (Figure S3) revealed that two of the horsetail MIPs belong to the PIP subfamily, and the other two to the TIP subfamily. The two TIPs share 58.4% sequence identity and 70.6% sequence similarity at the protein level, while the two PIPs show 65.0% sequence identity and 73.8% sequence similarity. The mean read coverage for these proteins varies from 191 to 247 (Table 1). No matches were found in the horsetail root transcriptome for other subfamilies of MIPs, including SIPs (small basic intrinsic proteins), or the more recently identified subfamilies HIP (hybrid intrinsic proteins), XIP (X intrinsic proteins) and GIP (GlpF-like intrinsic proteins).

Discussion

Despite the ubiquity of Si in nature, its role in biological systems remains poorly defined. Part of the problem can be ascribed to the scarcity of data linking Si to biochemical functions or activities (Fauteux et al., 2005; Chain et al., 2009). In this context, horsetail represents a unique system to study the role of Si in plant biology, as it is one of the strongest Si accumulators in the plant kingdom (Hodson et al., 2005), and one of the rare multicellular organisms that require this element to complete its lifecycle (Chen and Lewin, 1969). In this paper, we have identified a multigene family of aquaporin Si transporters through sequencing of the horsetail root transcriptome. Using heterologous expression in Xenopus oocytes, we have shown that some of the horsetail aquaporins exhibit higher Si transport activity than previously characterized transporters in higher plants. When expressed in Arabidopsis thaliana, the horsetail Si transporter EaNIP3;1 significantly increased Si absorption.

Conventional approaches based on sequence homology with known plant Si transporters failed to yield potential candidates in horsetail. This result was rather unexpected given that all previously identified transporters, from both monocots and dicots (cucurbits), have similar conserved sequences that are apparently key to Si transport. Diatoms are the only other organisms known to have Si transporters, but these transporters have no known homologs in plants (Hildebrand et al., 1997). Consequently, we surmised that, if Si transporters exist in horsetail, they may have some characteristics common to plant aquaporins. Alignment of aquaporins, with particular emphasis on sequences of known Si transporters, revealed two amino acid sequences that were then compared against the root transcriptome of horsetail. This strategy led to identification of a multigene family of NIPs in horsetail.

Horsetail NIPs have distinct characteristics that set them apart from all other known plant Si transporters to date, including a different ar/R filter (STAR) and overall sequence divergence (Figure 5). They do not belong to the NIP III group, which comprises all known Si transporters to date, but instead show stronger sequence homology to the NIP II group (Figure 5), although, to date, no Si transporters have been associated with the latter group. NIP II proteins are instead recognized as transporters of small uncharged substrates, although the function of only a few proteins from this group has ever been assessed in vitro. These include the boric acid transporters AtNIP5;1 (Takano et al., 2006) and AtNIP6;1 (Tanaka et al., 2008). Neither AtNIP5;1 (Mitani-Ueno et al., 2011) nor OsNIP3;1 (Mitani et al., 2008) transports silicic acid. The other main difference between the horsetail and higher-plant transporters is the ar/R filter residues. So far, all characterized higher-plant Si transporters have a GSGR ar/R filter. Plant aquaporins exhibit a great diversity of pore-forming residues (Liu et al., 2009), but this paper describes a previously unknown STAR ar/R filter in vascular plant NIPs. Very recently, an algal aquaporin with a STAR ar/R filter was discovered (Anderberg et al., 2011); however, this algal protein does not belong to the NIP subfamily. In rice, OsTIP4;1 has a closely related TTAR ar/R filter, and was shown by Li et al. (2008) to be permeable to water and glycerol. Both the STAR and GSGR ar/R filters comprise small amino acids, forming pores with a diameter large enough to allow for the passage of silicic acid (Rougé and Barre, 2008). The first residues (S for horsetail, G for higher plants) are both neutral, but differ in polarity, as S is polar but G is not. The second residues (T for horsetail, S for higher plants) are very similar in size, and are both neutral and polar. The third residues (A for horsetail, G for higher plants) are also small, neutral and non-polar. The last residue, arginine, is conserved in almost all plant aquaporins, and is involved in exclusion of positively charged ions from the channel (Beitz et al., 2006).

Despite these differences, the horsetail proteins also share common elements with higher-plant Si transporters. The intracellular loop B and the extracellular loop E were recently recognized to have consensus sequences among higher-plant Si transporters (Hove and Bhave, 2011). These consensus sequences are also found in the horsetail protein sequences (Figure 2). However, nucleotide sequence divergence may explain the lack of PCR amplicons using degenerate primers targeting these zones, as the reverse primer targeted a zone in loop E that is not conserved in horsetail. In addition to the four ar/R residues, five more residues, designated P1–P5, appear to be important in aquaporin substrate selectivity, and are conserved among monocot and dicot Si transporters (Hove and Bhave, 2011). Residues P3–P5 are conserved between the horsetail and higher-plant transporters (Figure 2). In addition, all the horsetail NIPs sequenced in this study have five exons and four introns (Figure 1a), a gene structure that is common to the majority of plant NIPs (Liu et al., 2009). The position of putative transmembrane helices relative to intron position is also conserved with respect to Physcomitrella (Danielson and Johanson, 2008) and higher plants (Liu et al., 2009).

Horsetail NIP3s are classified within the NIP II group (Figure 5), which has no known Si transporters to date. In contrast, higher-plant Si transporters belong to the NIP III group. On the basis of their divergent structure from NIP III proteins and different biological properties to NIP II proteins, horsetail NIP3s may belong to a new group, which may be confirmed if future research expands the number of aquaporins bearing similar characteristics. In a phylogenetic analysis, Danielson and Johanson (2010) noted that NIP3s were present in a common ancestor to bryophytes and higher plants. The authors speculated that NIP3s may have retained the original function of NIPs in early terrestrial plants. In accordance with this analysis, it is interesting to note that NIP3s had never previously been reported in any pteridophyte. A surprising characteristic of the horsetail NIP family is its abundance of very closely related sequences, especially in the NIP3;4–3;9 subgroup. Using complementary techniques to confirm DNA sequences, we have found that horsetail has an abundance of NIP3s, and no other subfamilies of NIPs were identified in this study. As horsetail has a unique requirement for Si to complete its lifecycle, it is possible that the multiplication of NIP3 genes is linked to the affinity for this element.

In our search for Si transporters, we have identified four other MIPs in horsetail that show fairly conventional features. Both PIP1 and PIP2 proteins are also found in higher plants and the moss Physcomitrella, indicating that the split between these two groups must have occurred early in terrestrial plant evolution (Danielson and Johanson, 2010). Vascular plants typically show a greater diversity of TIPs than identified here in horsetail (Danielson and Johanson, 2010), but as RNA-Seq was performed only on root tissues, it is possible that other TIPs or MIPs may be present in the horsetail genome, but not expressed in this organ or at the time of sampling. The expression levels of horsetail MIPs showed that the two most strongly expressed genes in horsetail roots were NIP3;1 and 3;3, which were expressed at rates twice as high as PIPs and TIPs. This is surprising as NIPs are usually expressed in planta at much lower levels than other MIPs (Sakurai et al., 2005; Park et al., 2010; Bienert and Chaumont, 2011). This unusually high expression of Si transporters in roots, the organ responsible for Si absorption, may account for the exceptionally high concentrations of Si found in horsetail tissues. Two of the horsetail transporters, EaNIP3;1 and EaNIP3;3, also showed a different in planta expression pattern than higher-plant root Si transporters. While expression of OsLsi1 (Ma et al., 2006) and HvLsi1 (Chiba et al., 2009) is restricted to the roots, these two horsetail NIPs were expressed in both the roots and the shoots (Figure 1b,c).

Accurate determination of the Si-transporting ability of horsetail NIPs was a key challenge in this study. For this purpose, we relied on oocyte assays, a well-recognized technique to study transmembrane flux (Miller and Zhou, 2000) that is commonly used for the study of aquaporins (Gomes et al., 2009). In our experiments, we tested one member from each of the three horsetail subgroups: EaNIP3;1, EaNIP3;3 and EaNIP3;4. We found that all three aquaporins with a STAR pore were very efficient Si transporters, a property that was previously thought to be exclusive to a GSGR pore (Mitani-Ueno et al., 2011). Although we did not measure the amount of injected cRNA translated into protein, previous studies have shown that the expression of proteins in oocytes is dependent on the quantity of injected cRNA (Goldin, 2006). Therefore, at equal concentration, this method allows comparison of the relative efficiency of various transporters by bypassing the natural genetic regulation observed in planta. Based on this premise, it is noteworthy that two of the horsetail transporters exhibited higher activity than the rice one (Figure 4). This would thus make EaNIP3;1 and EaNIP3;4 the most efficient plant Si transporters known to date. In their comparison of Si transporter activity, Chiba et al. (2009) found equal Si absorption among plant transporters with a GSGR pore, even though they expected Lsi1 from rice to be more efficient given the superior ability of rice to accumulate silicon. Additionally, the presence of Lsi2 efflux transporters may also be required for xylem loading of Si. Work is in progress to characterize homologs of Lsi2-type transporters in horsetail.

To complete the functional characterization of horsetail NIPs, we selected EaNIP3;1 for heterologous expression in the higher plant Arabidopsis. This protein is a strong Si transporter in oocytes and shows high expression in horsetail roots. Under the control of a constitutive promoter, EaNIP3;1 was expressed in both Arabidopsis roots and shoots, and also significantly increased the accumulation of Si in Arabidopsis shoots. This result confirmed previous findings by Montpetit et al. (2012) who showed that it was possible to increase the natural ability of a plant to absorb Si through heterologous expression of a wheat Si transporter.

In conclusion, this study has uncovered the existence and diversity of a multigene family of aquaporin Si transporters in horsetail. Based on their unique properties and features, their high efficiency at transporting Si, and their unusually high expression levels in roots, they offer unprecedented insight into the biology of the little-studied horsetail plant model and its unrivalled ability to accumulate Si. These results may contribute to better exploitation of Si in agriculture as a beneficial element when absorbed by plants.

Experimental procedures

Plant material

Horsetail material was collected from a natural colony of asexually reproducing plants growing at the Jardin Van den Hende botanical garden on the Université Laval campus, Québec, Canada. For RNA extraction, roots were collected from a single plant in late June, during a period of active growth.

Rice (cv. M201) seeds were sown in 9 cm pots in a bed consisting of nylon stockings cut into small ribbons. Pots were placed into a hydroponic system that immersed roots for 15 min every 30 min. Plants were grown in a greenhouse under the following conditions: 16 h/8 h photoperiod at 22°C day/18°C night, 80% humidity. Plants were immersed in distilled water during the first week after sowing, then the water was replaced by Hoagland’s solution that was refreshed each week and adjusted to pH 6.5 until harvest of the roots 3 weeks later.

Arabidopsis seeds (cv. Columbia ecotype; Col-0) were obtained from the Arabidopsis Biological Resource Center (https://abrc.osu.edu/). Col-0 and transformed plant seeds were sown in 9 cm pots containing commercial soil (Connaisseur® Premium Potting Soil, Fafard, http://www.fafard.com) at 4°C for 4 days in the dark. Plants were allowed to grow in a growth chamber under a 16 h/8 h photoperiod and a temperature regime of 22°C day/20°C night, 70% humidity. Plants were watered with distilled water for 2 weeks after sowing, then the water was replaced by a modified Hoagland’s solution (Tocquin et al., 2003) without soluble Si (control) or containing 1.7 mm Si (H4SiO4) for 2 weeks.

RNA and DNA extractions, and cDNA synthesis

Horsetail total RNA was extracted from roots and shoots using TRIzol reagent (Invitrogen, http://www.invitrogen.com) according to the manufacturer’s instructions, checked for integrity on a denaturing agarose gel, and stored at −80°C until use. Rice and Arabidopsis total RNA were extracted from roots using an RNA purification kit (Qiagen, http://www.qiagen.com) and stored at −80°C until use. Horsetail genomic DNA was extracted from leaves and stems using a DNeasy plant mini kit (Qiagen), and stored at −80°C until use. For rice, horsetail and Arabidopsis, first-strand cDNAs were prepared from 1 μg total RNA treated with RQ1 RNase-free DNase (Promega, http://www.promega.com), then reverse-transcribed using SuperScript III reverse transcriptase (Invitrogen) and oligo(dT)18 primers.

Identification of horsetail NIPs and other MIPs

A total of 16 913 711 72-base single-end reads were generated by sequencing horsetail root cDNA on an Illumina Genome Analyzer II sequencer (Illumina, http://www.illumina.com). De novo assembly of the contigs was performed using CLC Genomic Workbench (CLC bio, http://www.clcbio.com/). To identify possible horsetail NIPs, tBLASTn was performed on the assembled horsetail contigs using two peptides that were recently recognized to be conserved among higher-plant Si transporters (Hove and Bhave, 2011): intracellular loop B (GHISGAHMNPA, OsLsi1 positions 100–111) and extracellular loop E (GGSMNPART, OsLsi1 positions 215–223). To identify possible horsetail MIPs in addition to NIPs, tBLASTn was also performed on the assembled horsetail contigs using the 33 MIP protein sequences from rice (Sakurai et al., 2005), 23 MIP sequences from P. patens (Danielson and Johanson, 2008) and 35 MIP sequences from Arabidopsis (Johanson et al., 2001) as queries. The P. patens sequences included members from the recently described GIP, XIP and HIP subfamilies.

Cloning of OsLsi1 and horsetail NIPs

The coding sequence of the rice Lsi1 transporter (OsLsi1) was amplified from root cDNA using extended PCR primers (EcoRI-OsLsi1F and XbaI-OsLsi1R) based on the sequence of the OsLsi1 gene (Genbank accession number AK069842). All primers were synthetized by Invitrogen. The resulting amplicon was inserted into pGEM-T Easy (Promega) to produce pGEM.OsLsi1.

The candidate EaNIP3;13;9 coding sequences were amplified using the high-fidelity Phusion DNA polymerase (New England Biolabs, http://www.neb.com) from horsetail root cDNA using specific primers listed in Table S1. For EaNIP3;1, a single 747 bp product was amplified and subcloned in pUC18 (Fermentas, http://www.fermentas.com) to produce pUC18.EaNIP3;1. For EaNIP3;2 and EaNIP3;3, two amplification products (789 and 780 bp, respectively) were subcloned in pUC18 to produce pUC18.EaNIP3;2 and pUC18.EaNIP3;3. For the EaNIP3;43;9 subgroup, several 783 bp PCR products were subcloned into pUC18, sequenced, and resolved into different sequences. For each cDNA, three to nine clones were sequenced on a 3730xl DNA Analyzer (Applied Biosystems, http://www.appliedbiosystems.com). After sequencing, RNA-Seq reads were mapped back to cloned cDNA sequences using CLC Genomic Workbench to correct any sequencing or enzymatic errors.

To obtain full-length cDNA sequences, both 5′ and 3′ RACE PCRs (Frohman et al., 1988) were performed using HotMaster Taq DNA (5 PRIME, http://www.5prime.com) (Table 1). The amplification products were subcloned into pGEM-T Easy. For each sequence, two to nine clones were sequenced.

To sequence horsetail NIP genes, PCR was performed on genomic DNA using Phusion DNA polymerase and the same primers used for cDNA cloning (Table S1). The amplification products were subcloned into pGEM-T Easy. For each gene, 12–20 clones were sequenced using internal primers that were designed as needed. To identify the intron–exon gene structure, full gene sequences were compared using Spidey (http://www.ncbi.nlm.nih.gov/spidey/) for perfect sequence match with the corresponding cloned cDNA.

Horsetail NIP protein sequences were subjected to pairwise comparisons after alignment with Clustal X (http://www.clustal.org). Identity and similarity percentages are shown in Table S2.

Expression analysis of horsetail NIPs in roots and shoots

To assess expression in roots and shoots, RT-PCR was performed on respective cDNAs using the following primer pairs: EaNIP3;1 5′F/EaNIP3;1Rn1, EaNIP3;2-3;3 F519/EaNIP3;2-3;3R, and EaNIP3;4+ F418/EaNIP3;4+R583 (Table S1). The resulting PCR products were separated on an agarose gel (Figure 1b). As one of the primer pairs targeted the two closely related genes EaNIP3;2 and EaNIP3;3, the resulting PCR product from shoot cDNA was digested using the restriction enzyme SacI, which cleaved only the EaNIP3;2 PCR product, or the restriction enzyme ScaI, which cleaved only the EaNIP3;3 PCR product (Figure 1c).

Alignments and phylogenetic analysis

Amino acid sequences of OsLsi1, CmNIP2;1 and all root-expressed horsetail NIPs were aligned using Clustal X. The transmembrane domains were predicted using the TMHMM server (http://www.cbs.dtu.dk/services/TMHMM/) and the 3D model for EaNIP3;1. For the phylogenetic analysis, amino acid sequences for all rice, Arabidopsis and P. patens NIPs were obtained from the literature (Johanson et al., 2001; Danielson and Johanson, 2008; Liu et al., 2009). Additional NIP sequences from higher plants with putative or proven Si transport activity were also included in the analysis (Ma et al., 2006; Chiba et al., 2009; Liu et al., 2009; Mitani et al., 2009; Yamaji and Ma, 2009; Mitani-Ueno et al., 2011). These proteins and the horsetail NIPs were aligned using Clustal X. For the phylogenetic analysis of all horsetail MIPs, amino acid sequences for all Arabidopsis, rice and P. patens MIPs and the horsetail proteins were aligned using Clustal X. The phylogenetic trees were constructed using MEGA5 version 5.05 (Tamura et al., 2011) using the bootstrap neighbor-joining method with 1000 replicates and the Jones-Taylor-Thornton (JTT) model.

Homology modeling of EaNIP3;1 and its putative substrate

The EaNIP3;1 model was built from the crystal structure of AQP0 (PDB code 2B6O) using MODELLER (Eswar et al., 2008). The model quality was assessed by analysis of the Ramachandran plot using PROCHECK (Laskowski et al., 1993). The 3D structure of silicic acid [Si (OH)4] was obtained from http://molecules.gnu-darwin.org/. The docking of silicic acid in EaNIP3;1 protein was performed using PatchDock (Schneidman-Duhovny et al., 2005) with the following parameters: clustering RMSD = 1.5 Å; complex and type = protein–small ligand. PEARLS (Duhovny et al., 2002) was used to estimate the ligand–receptor interaction energy. The images were generated using Jmol (http://jmol.sourceforge.net/).

Plasmid constructions for heterologous expression in Xenopus oocytes

Fragments containing OsLsi1, EaNIP3;1, EaNIP3;3 or EaNIP3;4 coding sequences were excised from pGEM.OsLsi1, pUC18.EaNIP3;1, pUC18.EaNIP3;3 or pUC18.EaNIP3;4 by EcoRI/XbaI digestion and inserted into EcoRI/XbaI pre-digested Pol1 vector, a Xenopus laevis oocyte expression vector derived from pGEM and comprising the T7 promoter, the X. laevis globin untranslated regions and a poly(A) tract (Caron et al., 2000). All vectors were transformed into Escherichia coli TOP10 strain and kept frozen at −80°C.

Si transport assays using heterologous expression in Xenopus oocytes

Plasmids containing either the OsLsi1, EaNIP3;1, EaNIP3;3 or EaNIP3;4 coding sequence were recovered from a fresh bacterial culture using a QIAprep Spin Miniprep kit (Qiagen). Aliquots (5 μg) of each plasmid were linearized using NheI (Roche, http://www.roche.com). Digestions were purified using a PCR purification kit (Qiagen), and 1 μg of DNA was transcribed in vitro using the mMessage mMachine T7 Ultra kit (Ambion, http://www.invitrogen.com/site/us/en/home/brands/ambion.html). cRNAs were precipitated using phenol/chloroform, and solubilized in water treated with 0.1% DEPC (Sigma-Aldrich, http://www.sigmaaldrich.com). Defolliculated stage V–VI oocytes were injected with 25 nl H2O or 25 nl of a 500 ng μl−1 cRNA solution, and eggs were maintained at 18°C in modified Barth's medium (MBS) (88 mM NaCl, 1 mM KCI, 2.4 mM NaHC03, 0.82 mM MgSO4, 0.33 mM Ca(N03)2•4H20, 0.41 mM CaC12, 15 mM Hepes, pH 7.6) supplemented with 100 M of penicillin/streptomycin. Three days after injection, pools of 10 oocytes for each condition were incubated 0 or 60 min in Barth medium containing 1.7 mm Si or no Si. After incubation, oocytes were rinsed in sucrose/HEPES solution and frozen until intracellular Si measurement. The experiment was repeated four times.

Dosage of Si in oocytes

Concentrated nitric acid (25 μl) was added to each pool of 10 oocytes, which were then dried for 2 h at 82°C. Plasma-grade water (100 μl) was added, and samples were incubated for 1 h at room temperature. Samples were vortexed, then centrifuged for 5 min at 13 000 g. The intracellular Si concentration was measured in 10 μl of supernatant by Zeeman atomic absorption using a Zeeman atomic spectrometer AA240Z (Varian; http://www.varian.com) equipped with a GTA120 Zeeman graphite tube atomizer. The standard curve was obtained using a 1000 ppm ammonium hexafluorosilicate solution (Fisher Scientific, http://www.fishersci.com). Data were analyzed with SpectrA software (Varian).

Plasmid constructions for heterologous expression of EaNIP3;1 in Arabidopsis

For stable expression of EaNIP3;1, Col-0 Arabidopsis plants were transformed using the floral-dip method (Clough and Bent, 1998). The p35S-EaNIP3;1 construct was obtained as follows. The EaNIP3;1 coding region was amplified from horsetail root cDNA using gene-specific primers (Bsp-EaNIP3;1 and Bgl-EaNIP3;1bis; see Table S1). Amplicons were cloned into pGEM T Easy (Promega). A BspHI–BglII fragment carrying EaNIP3;1 was inserted into the corresponding site in pCAMBIA1302 (CAMBIA, http://www.cambia.org/), thereby producing an EaNIP3;1 coding sequence regulated by the CaMV 35S promoter (construct p35S-EaNIP3;1). Transformants (T1) were selected on Murashige and Skoog basal medium containing Gamborg’s Vitamins (MS) (Sigma-Aldrich) containing hygromycin (15 mg l−1), and the presence of the EaNIP3;1 transgene was verified by PCR using EaLsi1-Fow and EaLsi1-Rev primers (Table S1). T2 seeds were harvested and sown on MS medium containing hygromycin (15 mg l−1) to select single-locus lines. Expression of EaNIP3;1 was determined in homozygous T3 transgenic plants.

Expression analysis of the EaNIP3;1 transgene in Arabidopsis grown with or without Si

The expression of EaNIP3;1 in Arabidopsis was assessed by RT-PCR. Total RNA from roots or leaves of homozygous T3p35S-EaNIP3;1 plants grown with or without Si was extracted, and cDNA was synthesized as described above. EaNIP3;1 (plus the AtUBQ1 gene as an internal control) were amplified using the following primers: Atactin2-Fow and Atactin2-Rev for the reference gene, and EaNIP3;1-Fow and EaNIP3;1-Rev for the EaNIP3;1 gene (see Table S1).

Dosage of Si in Arabidopsis shoots

Col-0 plants and the transgenic line carrying p35S-EaNIP3;1 were used to assess Si uptake. Si concentration was measured by inductively coupled plasma optical emission spectrometry (JY2000-2, Horiba Scientific, http://www.horiba.com). Aerial parts of four plants from each treatment (three replicates) were collected and freeze-dried 2 weeks after the start of Si supplementation. Samples were ground to a powder and total Si analysis was performed at 251.611 nm by inductively coupled plasma optical emission spectrometry as described by Côté-Beaulieu et al. (2009).

Acknowledgements

The project was funded by a grant from the Natural Sciences and Engineering Research Council of Canada (NSERC) in collaboration with Syngenta Biotechnology and the Canada Research Chairs Program to R.R.B. C.G. is supported by a Canada Graduate Scholarship from NSERC. The authors would like to thank Dr P. Isenring’s Nephrology Research Group of the Centre Hospitalier Universitaire de Québec/L’Hôtel-Dieu de Québec Institution for help with the oocyte assays. The authors also thank the anonymous reviewers for constructive feedback. The authors would like to thank Halim Maaroufi from the Institut de biologie intégrative et des systèmes (IBIS) at Laval University for excellent assistance with 3D modeling of the horsetail protein. [Correction added on 13 May 2013 after original online publication: An acknowledgement thanking Halim Maaroufi was added.]

Ancillary