DNA interference: DNA-induced gene silencing in the appendicularian Oikopleura dioica

Abstract

RNA interference is widely employed as a gene-silencing system in eukaryotes for host defence against invading nucleic acids. In response to invading double-stranded RNA (dsRNA), mRNA is degraded in sequence-specific manner. So far, however, DNA interference (DNAi) has been reported only in plants, ciliates and archaea, and has not been explored in Metazoa. Here, we demonstrate that linear double-stranded DNA promotes both sequence-specific transcription blocking and mRNA degradation in developing embryos of the appendicularian Oikopleura dioica. Introduced polymerase chain reaction (PCR) products or linearized plasmids encoding Brachyury induced tail malformation and mRNA degradation. This malformation was also promoted by DNA fragments of the putative 5′-flanking region and intron without the coding region. PCR products encoding Zic-like1 and acetylcholine esterase also induced loss of sensory organ and muscle acetylcholinesterase activity, respectively. Co-injection of mRNA encoding EGFP and mCherry, and PCR products encoding these fluorescent proteins, induced sequence-specific decrease in the green or red fluorescence, respectively. These results suggest that O. dioica possesses a defence system against exogenous DNA and RNA, and that DNA fragment-induced gene silencing would be mediated through transcription blocking as well as mRNA degradation. This is the first report of DNAi in Metazoa.

1. Introduction

Most organisms have gene-silencing systems for protection of cells against invading nucleic acids, such as viruses and transposons. A host defence system was first reported in the petunia flower [1], where overexpression of mRNAs involved in floral pigmentation unexpectedly induced a reduction of such pigmentation. The gene-silencing mechanism known as RNA interference (RNAi) has been clarified in Caenorhabditis elegans [2], where double-stranded RNA (dsRNA) induces sequence-specific mRNA degradation. In RNAi, Argonaute protein binds to dsRNA to form a protein–nucleic acid complex, which recognizes and cleaves the target mRNA [3]. It is now known that RNAi operates in a broad range of organisms from fungi to vertebrates [4,5]. On the other hand, DNA interference (DNAi) has also been reported in a few species of plants, ciliates and archaeans. In tobacco and Adiantum, introduction of DNA fragments without a promoter has been shown to induce sequence-specific gene silencing. These phenomena were termed DNAi [6–8]. In Paramecium, a ciliate, plasmid DNA carrying a protein-encoding region induces post-transcriptional gene silencing of the corresponding endogenous gene [9]. Furthermore, a recent study of the archaean Thermus thermophilus has suggested that it may have a system for DNA-induced DNAi [10]. Argonaute protein in T. thermophilus has shown to cleave exogenously supplied DNA strands in a sequence-specific manner [10]. While RNAi is a highly conserved gene-silencing system in a wide range of eukaryotes, DNAi has not been explored in any multicellular animals studied to date. In addition to DNAi, the phenomenon of co-suppression by transgenic DNA has been reported in many organisms [11,12]. In C. elegans, introduction of transgenic copies of a gene inhibits expression of the transgene as well as endogenous homologous genes, partly using cellular components of RNAi [12]. This co-suppression is distinct from DNAi, because co-suppression depends on the presence of a promoter region of transgene and probably transcription from transgenes [11,12].

In this study, we found a DNAi phenomenon in the chordate Oikopleura dioica, and demonstrate the existence of a new gene-silencing system in Metazoa. An RNAi-mediated gene knockdown method has already been established in this animal [13]. We found that microinjection of double-stranded DNA fragments of the Brachyury gene induced tail malformation reminiscent of the phenotypes induced by RNAi. DNA fragments of the putative 5′-flanking region and intron without the coding region also induced the same phenotype. This malformation was also promoted by linearized plasmid DNAs encoding partial Brachyury cDNA sequences. Quantitative RT-PCR analyses showed that this DNA fragment-induced effect is sequence-specific and reduces the amount of the corresponding transcript. DNA fragments of Zic-like1 and Acetylcholinesterase (AChE) induced the expected loss-of-function phenotypes (i.e. loss of sensory organ and muscle AChE activity, respectively). These results suggest that O. dioica possesses a defence system against exogenous DNA and RNA, and that DNA fragment-induced gene silencing would be mediated through transcription blocking as well as mRNA degradation.

2. Results and discussion

(a). Double-stranded DNA induces sequence-specific gene knockdown phenotypes

To investigate the response of O. dioica to invading DNA, we injected O. dioica with a polymerase chain reaction (PCR) product of 248 bp encoding a partial Brachyury cDNA sequence (PCR-Bra-248 bp) (figure 1a). For introduction of DNA into oocytes, the PCR product and mRNA encoding a fluorescent fusion protein, H2B-EGFP, were co-injected into the ovary of pre-spawning adults [13]. We selected larvae after hatching with EGFP fluorescence in the nuclei (22% ± 10% s.d. of a single cohort on average, n = 11) that were considered to have incorporated PCR product, and these larvae were used for further analyses. More than 90% of larvae that incorporated PCR-Bra-248 bp showed tail malformation (figure 1c,n), and the phenotype closely resembled the result of dsRNA-mediated knockdown of Brachyury mRNA (figure 1m) [13]. This phenotype was categorized into four groups—normal, short tail, shrunken tail and lethal—on the basis of the previous RNAi experiment [13]. Embryos injected with a PCR product encoding EGFP (PCR-EGFP) developed normally (figure 1b,n). The tail malformation is not due to mutation caused by homologous recombination between the O. dioica genome and the PCR products. Sequencing of Brachyury gene in a PCR-Bra-248 bp injected short-tail larva demonstrated that there are no mutations (n = 2, data not shown). These results together suggested that the PCR products had induced sequence-specific gene knockdown phenotypes.

Figure 1.

Tail malformation induced by PCR products encoding the Brachyury gene or cDNA sequence. (a) Regions targeted by each of the PCR products. Upper box is a gene model of Brachyury and the regions targeted by each of the PCR products are shown by blue bars underneath. Yellow, grey, blue and white boxes indicate the ORF, intron, UTR and inter-genic region, respectively. The starting methionine and stop codon are indicated as ATG at positions 1 and *. (b–m) Phenotype of larvae injected with various PCR products of Brachyury and EGFP as a control. Each PCR product (0.3 µg µl⁻¹) was co-injected with H2B-EGFP mRNA (1.3 µg µl⁻¹) and only larvae with EGFP fluorescence localizing into nucleus were scored. exon-20 bp and exon-50 bp DNA were prepared by annealing of 20 and 50 bp of synthesized oligonucleotides, respectively. Total numbers of shrunken tails among those of larvae with EGFP fluorescence are shown at the bottom. Scale bar, 50 µm. (n) Average proportions of hatched larvae that exhibited each phenotype in two to four independent injections (shown in parentheses). S.d. is shown by vertical bars. The results obtained in animals injected with PCR-EGFP, PCR-Bra-248 bp, PCR-Bra-full, PCR-Bra-intron, PCR-Bra-intron-2, PCR-Bra-exon, exon-20 bp, exon-50 bp, PCR-Bra-UTR, PCR-5′Bra-1, PCR-5′Bra-2 and PCR-5′Bra-3 are shown in various colours depicted on the right.

To test whether the DNA-induced malformation was sequence-dependent, we tested 10 PCR products targeting distinct regions of the Brachyury gene: PCR-Bra-full, PCR-Bra-intron, PCR-Bra-intron-2, PCR-Bra-exon, exon-20 bp, exon-50 bp, PCR-Bra-UTR, PCR-5′Bra-1, PCR-5′Bra-2 and PCR-5′Bra-3 (figure 1a). More than 80% of larvae injected with PCR-Bra-full, PCR-Bra-intron, PCR-Bra-exon, PCR-Bra-UTR or PCR-5′-Bra-1 similarly showed tail malformation (figure 1d–f,i,j,n). In addition, PCR-Bra-intron-2 showed a weaker effect (51% shrunken tail), probably because length of the fragment is short (only 74 bp; electronic supplementary material, table S1), and it is not effective enough, as mentioned later. Therefore, PCR products targeting not only Brachyury-encoding regions but also the introns and one of the 5′-flanking regions (PCR-5′-Bra-1, covering −217 to −393 bp upstream from the starting methionine) are able to efficiently exert knockdown effects.

In general, O. dioica mRNAs have tens to a few hundred bases of 5′UTR [14]. In O. dioica, Brachyury cDNA is reported to have 5′UTR of 216 bp in length, and does not have the spliced reader sequence [15]. We confirmed that PCR-5′-Bra-1 does not correspond to 5′UTR of Brachyury mRNA in three ways. First, our 5′-RACE analysis showed that Brachyury cDNA does not have longer UTR in consistent with the previous report. Second, we also carried out RT-PCR to test the possibility that the PCR-5′-Bra-1 corresponds to the 5′UTR of Brachyury mRNA. PCR primers covering the reported 5′UTR and coding region amplified an appropriate band. On the other hand, forward primer covering the 3′region of PCR-5′Bra-1 (figure 1a) could not amplify the Brachyury sequence (data not shown). Third, RNAi targeting the sequence covered by PCR-5′-Bra-1 did not show any knockdown effect (0% shrunken tail, n = 3). These results indicate that PCR-5′Bra-1 indeed corresponds to the 5′-flanking region, but not 5′-UTR, of the Brachyury gene. By contrast, use of PCR-5′Bra-2 (−617 to −793 bp) and PCR-5′Bra-3 (−1268 to −1444 bp) resulted in gradual loss of the capacity to induce tail malformation, which occurred in only 9% (±4% s.d.) and 4% (±2% s.d.) of the larvae, respectively (figure 1k,l,n). Thus, the efficiency of the knockdown effect seems to decline with increasing distance of the target from the Brachyury coding region. Nonetheless, PCR-5′Bra-2 (−617 to −793 bp) and PCR-5′Bra-3 still exert a weak effect. The effect of PCR products targeting the intron and the 5′-upstream region of the Brachyury gene suggests that the introduced DNA might interfere with the transcriptional process.

We examined the possibility that injected DNAs affect mRNA maturation. RT-PCR of Brachyury cDNA covering the PCR-Bra-intron target region was carried out. If PCR-Bra-intron prevents intron excision, the 506 bp band would be detected in addition to the normally spliced 294 bp band. Injection of PCR-Bra-intron did not give any additional band (figure 2a), and reduction of the 294 bp band was just observed, suggesting mRNA maturation was not affected by PCR-Bra-intron.

Figure 2.

Tail malformation induced by linear double-stranded DNA. (a) RT-PCR of Brachyury cDNA covering PCR-Bra-intron target region (red bar). Yellow and grey boxes show exon and intron, respectively. Blue arrows show PCR primers, and the PCR-Bra-intron target region is shown by a red bar. In the bottom picture, cDNAs from PCR-EGFP and PCR-Bra-intron injected larvae were used as PCR templates, respectively. If PCR-Bra-intron prevents intron excision, the 506 bp band would be detected in addition to the normally spliced 294 bp band. (b) Average proportions of hatched larvae that exhibited each phenotype derived from animals injected with different concentration of PCR-Bra-248 bp. ×1/5, ×1/25 and ×1/125 represent the injected concentration, 0.0600 µg µl⁻¹, 0.0120 µg µl⁻¹ and 0.0024 µg µl⁻¹ of PCR-Bra-248 bp, respectively. (c) A partial Brachyury cDNA sequence (blue arrow) was cloned into pBluescript. Direction of arrow indicates that of the inserted gene. Restriction enzyme recognition sites (BamHI and NotI) are shown by red lines. (d) Proportions of each of the phenotypes in hatched larvae. Each construct (0.2 µg µl⁻¹) was injected into the gonad. Blue, red, green and yellow bars show the results for injection of circular pBS-Bra, pBS/NotI, pBS/BamHI and pBS/NotI,BamHI, respectively. (e) Representative phenotype resulting from each injection. Ratios of larvae with shrunken tails are shown at the bottom. Scale bar, 50 µm.

To investigate crucial concentration of the injected DNA, we diluted PCR-Bra-248 bp DNA (initial concentration was 0.3 µg µl⁻¹). As shown in figure 2b, 1/5 concentration (0.06 µg µl⁻¹) was effective. Proportions of the shrunken tail phenotype at 0.3 and 0.06 µg µl⁻¹ were 92% (±1% s.d.) and 80% (±5% s.d.), respectively. By contrast, 1/25 and 1/125 concentrations were not effective any more (figure 2b). Therefore, 0.06 µg µl⁻¹ is the minimum concentration.

Next, we examined whether the knockdown phenotype depends on the length of the target. PCR products that were 100 bp long (PCR-intron and PCR-exon; figure 1e,f,n; electronic supplementary material, table S1) worked. By contrast, the 50 bp product (exon-50; figure 1h,n) exerted moderate effect (22% shrunken tail), and the 20 bp product (exon-20; figure 1g,n) showed no effect. This is consistent with the previous result, in which PCR-Bra-intron-2 (74 bp) also showed weaker effect (51% shrunken tail).

(b). Linear DNA, but not circular DNA, exerts knockdown effects

To investigate the difference in the response to circular and linear DNA, we injected O. dioica with four different configurations of a plasmid harbouring a partial Brachyury cDNA insert of 721 bp (pBS-Bra in figures 1a and 2c). These comprised a circular plasmid pBS-Bra, linearized plasmids pBS-Bra/NotI and pBS/BamHI, and only the insert region pBS-Bra/NotI,BamHI, respectively. Most of the larvae injected with pBS-Bra and pBS-Bra/NotI showed normal development (figure 2d,e), whereas those injected with pBS-Bra/BamHI or pBS-Bra/NotI,BamHI showed tail malformation (figure 2d,e). These results indicate that a linear DNA configuration is essential for induction of tail malformation phenotypes, or that the vector sequence connected to the 5′ end of the Brachyury sequence inhibits the capacity to induce malformation.

(c). Linear double-stranded DNA induces a gene-specific phenotype

Next, to test whether PCR products targeting other genes would induce specific phenotypes, PCR products of the nerve-specific gene Zic-like1 or the muscle marker gene Acetylcholinesterase (AChE) were injected. Zic-like1 is a zinc finger motif protein expressed zygotically only in neural tissue (corresponding to zinc finger protein ap-zic [GSOIDT00008831001] in the OikoBase genome browser at http://oikoarrays.biology.uiowa.edu/Oiko/). In larvae injected with PCR-Zic-like1 (covering 446 bp), loss of the otolith in the brain vesicle was observed at 5.5 h post-fertilization (hpf) (figure 3a). This phenotype was in accord with the result of RNAi-mediated knockdown of Zic-like1 (figure 3a). To investigate the effect of PCR-AChE (covering 819 bp), the amount of AChE protein was monitored by histochemical staining in 7 hpf larvae. In uninjected controls, two row of AChE staining were observed in muscle cells on both sides of the tail (figure 3b). In juveniles injected with PCR-AChE, staining of muscle cells was not detected, although staining was detected in small cells on the left side of the tail (figure 3b). These signals corresponded to the nerve cells in the nerve cord that are present on the left side and express unknown cholinesterase activity, having been hidden by strong muscle staining in the controls. The phenotype induced by each PCR product supported the contention that linear DNAs induced gene-specific knockdown.

Figure 3.

Phenotype induced by PCR-Zic-like1 and PCR-AChE. (a) Phenotype of larvae injected with PCR-Zic-like1 (0.1 µg µl⁻¹) and uninjected control at 5.5 hpf. Ratios of larvae in which an otolith was observed are shown at the bottom. (b) Histochemical staining for acetylcholinesterase was performed in juveniles injected with PCR-AChE (0.2 µg µl⁻¹) and uninjected controls at 7 hpf. Ratios of juveniles that showed AChE staining in muscle cells are shown at the bottom. Scale bar, 50 µm.

(d). Introduced double-stranded DNA reduces the amount of mRNA

Next, we examined which processes, from the gene to the protein level, were involved in the effect of introduced DNAs. In order to investigate the effect of introduced DNAs on the amounts of mRNA, we measured the mRNAs of Brachyury and Thrombospondin, a target gene of Brachyury [16], using quantitative real-time PCR. PCR-Bra-248 bp reduced the amount of Brachyury and Thrombospondin mRNA to 21% and 9% of that in the control group, respectively (figure 4a). On the other hand, PCR-thrombospondin did not reduce the amount of Brachyury mRNA, but reduced that of Thrombospondin mRNA to 25% (figure 4b). This result demonstrated that the introduced DNA specifically reduced the mRNA of the targeted genes and those downstream from them.

Figure 4.

PCR products eliciting mRNA reduction. (a,b) Results of quantitative real-time PCR. Ratios of the amounts of Brachyury and Thrombospondin mRNA in 3 hpf larvae injected with (a) PCR-Bra-248 bp (0.3 µg µl⁻¹) and with (b) PCR-thrombospondin (0.7 µg µl⁻¹) to that in larvae injected with control PCR-EGFP (0.1 µg µl⁻¹) are shown. Three separate experiments per subject were performed, and the average values are shown. (c) Fluorescence in embryos at 2 hpf. Top panels show mCherry and EGFP fluorescence in embryos co-injected with H2B-EGFP and H2B-mCherry mRNAs (1.7 µg µl⁻¹). Middle and bottom panels represent embryos that were co-injected with an mRNA mixture and PCR-EGFP or PCR-mCherry (0.1 µg µl⁻¹), respectively. Scale bar, 50 µm. (d,e) Quantitative representation for (c). *p < 0.05 (Wilcoxon–Mann–Whitney test) compared with the control groups. Bars indicate standard deviations.

(e). Introduced double-stranded DNA reduces the amounts of protein synthesis from exogenous mRNAs

We then examined how introduced DNA reduces the amount of mRNA, either by blocking its transcription or degrading it, as in the case of RNAi. PCR products targeting the intron or 5′-upstream regions of Brachyury induced tail malformation, as mentioned above, thus indicating that introduced DNAs reduce the amount of mRNA through transcriptional inhibition. To investigate whether PCR products also degrade mRNA, O. dioica was co-injected with H2B-mCherry mRNA, H2B-EGFP mRNA, and PCR-EGFP or PCR-mCherry (covering 128 bp of the EGFP sequence and 126 bp of the mCherry sequence). In embryos injected with PCR-EGFP, only the green fluorescence of EGFP was reduced (figure 4c, middle row). Quantitative analysis using the method of Omotezako et al. [13] showed that the ratio of fluorescence obtained by dividing the EGFP green signal by the mCherry red signal (EGFP/mCherry) was reduced to 42% (±17% s.d.) of the control (figure 4d). Likewise, in embryos injected with PCR-mCherry, the red fluorescence of mCherry was reduced to 76% (±14% s.d.) of the control (mCherry/EGFP) (figure 4c, bottom row; figure 4e). These results show that the introduced DNA exerts its effect by reduction of the amount of mRNA through mRNA degradation, at least in part.

Our present results have revealed a new gene-silencing phenomenon induced by linear double-stranded DNA. In O. dioica, exogenously supplied double-stranded DNA induced sequence-specific blocking of transcription and degradation of mRNA, resulting in gene-specific knockdown phenotypes. We term this phenomenon ‘DNAi’ in O. dioica for the following reasons. This gene-silencing phenomenon is different from co-suppression with transgenic DNA in fly and nematodes, which requires promoter region [11,12]. Microinjection of promoterless DNA fragments encoding exon, intron or UTRs were sufficient to induce gene silencing in O. dioica. This characteristic is consistent with the DNAi in plants [7,8]. However, the DNAi in O. dioica differs from DNAi in plants in several aspects. In O. dioica, PCR products targeting the inter-genic region and intron also induced sequence-specific gene silencing. In plants, however, only promoterless-cDNA matching to exons could induce gene silencing. Another difference is conformation of DNA. In O. dioica, DNAi is specifically induced by linear DNA, whereas both linear and circular DNA induced DNAi in plants. Overall, this study represents the first example of DNAi in multicellular animals.

The molecular pathway that mediates DNAi in O. dioica will be the focus of future investigations. As a host defence against invading DNA, it has recently been reported that Argonaute protein of T. thermophilus (TtAgo) mediates small single-stranded DNA-induced DNAi [10]. Similarly to TtAgo, Argonaute of O. dioica may mediate DNAi, inhibiting transcription and degrading mRNA. In a BLAST search of the O. dioica genome database (OikoBase) [17], we found nine Argonaute family protein homologues bearing both the PAZ and PIWI domains. For host defence, some of these homologues might mediate RNAi in response to dsRNA, whereas others might mediate DNAi in response to introduced DNA. Although its primary function would be protection against exogenous nucleic acids, if the invading DNA were to carry a homologous sequence of the O. dioica endogenous gene, this host defence system might inhibit the endogenous gene to induce malformation.

Another intriguing possibility is that the gene silencing in O. dioica is mediated by RNA, which is transcribed from introduced DNA. However, while PCR products targeting the sequences in the 5′-upstream region, PCR-5′Bra-1, induced knockdown phenotype efficiently, dsRNA targeting the same sequence showed no effect, as mentioned above. This shows that transcription suppression by DNAi is unlikely to be mediated by RNA that is transcribed from introduced DNA. This observation, however, does not exclude the possibility that mRNA degradation by DNAi is mediated by transcribed small RNAs.

Another unresolved issue is whether DNAi is a conserved biological event in metazoans. We have investigated this possibility using another chordate species, the ascidian Halocynthia roretzi. However, when we injected H. roretzi eggs with PCR-HrBra, these animals developed normally (data not shown). Further investigations will clarify whether double-stranded DNA-induced gene silencing occurs in other multicellular animals.

Preparation of double-stranded DNA by PCR is much easier, faster and cheaper than preparation of dsRNA. Therefore, DNAi could be a feasible technique for characterization of gene functions in O. dioica. For practical application, PCR products longer than 100 bp at a concentration of over 0.1 µg µl⁻¹ would be recommended.

3. Material and methods

(a). Laboratory culture of Oikopleura dioica

Wild animals were collected at Sakoshi Bay and Tossaki Port in Hyogo prefecture, Japan, to start a culture. The animals have been cultured over generations for more than a year in the laboratory, as described previously [13].

(b). Constructs

To generate Brachyury, Acetylcholinesterase and Zic-like1 constructs, O. dioica genomic DNA for Brachyury and partial cDNAs for Acetylcholinesterase and Zic-like1 (accession no. KM047670 and KM047669) were isolated and cloned into the pCR4-TOPO vector (Invitrogen), except for Acetylcholinesterase cDNA, which was subcloned into the pGEM-T Easy vector (Promega). To amplify the Brachyury genome sequence, PCR primers were designed for the upstream and downstream flanking genes of Brachyury, respectively. Primers used for obtaining genome and cDNA sequences are shown in the primer list (electronic supplementary material, table S1). To generate pBS-Bra, partial Brachyury cDNA (721 bp) was amplified with NotI-BraF and BamHI-BraR primers using Brachyury cDNA inserted into the pCR-TOPO vector, and this was inserted between the NotI and BamHI digestion sites of the pBluescript vector.

(c). Microinjection into the ovary

PCR products and constructs were injected into the gonads of maturing adult animals, as described previously [13]. Because pro-oocytes are connected to a shared cytoplasm through a pore known as the ring canal in the ovary [18], injected constructs were spread into part of the gonad with a gradient and incorporated into oocytes through the ring canal. mRNA encoding H2B-EGFP or H2B-mCherry was co-injected with DNA fragments as an indicator of incorporation. We categorized resulting larvae into two groups (larvae with and without fluorescence in the nuclei), although there was a continuous distribution of different intensities from strongly fluorescent to non-fluorescent eggs. Larvae categorized into ones with fluorescence were scored and used for analysis. A single injected animal spawned approximately 200 eggs, and 22% ± 10% s.d. (n = 11) of hatched larvae showed fluorescence derived from incorporated mRNA on average.

(d). Preparation of PCR products

PCR products to be injected were amplified using KOD plus (TOYOBO) and gene-specific primers or universal primers (see primer list in electronic supplementary material, table S1). We ordinarily get a single band. In one case where non-specific bands were detected, we purified an appropriate band from gel. exon-20 bp and exon-50 bp were generated by annealing of synthesized oligonucleotides. Sense and antisense oligonucleotides were heated at 95°C for 8 min and gradually cooled into 25°C using thermal cycler. They were purified by phenol–chloroform, chloroform and ethanol precipitation, and dissolved in water for microinjection.

(e). Acetylcholinesterase histochemistry

7 hpf larvae were fixed in 5% formaldehyde in artificial seawater at room temperature for 15 min. After washing with PBS-Tween, fixed larvae were stained with reaction solution including 65 mM phosphate buffer (pH 6.0), 10 mM sodium citrate, 3 mM copper sulfate, 0.5 mM potassium ferricyanide and 0.5 mg ml⁻¹ acetylthiocholine iodide at room temperature for 2 h.

Data accessibility

All sequences generated for this study have been deposited under GenBank accession numbers KC253939, KM047669 and KM047670. The datasets supporting this manuscript have been submitted as part of the electronic supplementary material.

Funding statement

This work was supported by grants-in-aid for Scientific Research from the JSPS to H.N. (22370078 and 26650079) and to T.A.O. (24870019), and a grant-in-aid for JSPS Fellows to T.O. (20131402).

Authors' contributions

T.O., T.A.O. and H.N. designed the study, and wrote and revised the manuscript. T.O. conducted the experiment and analysed the data. All authors read and approved the final manuscript.