Introduction

Root nodule symbioses between plants and nitrogen-fixing soil bacteria are restricted to species from four orders of dicots: Fabales, Fagales, Cucurbitales and Rosales. Unicellular Gram-negative soil bacteria, collectively called rhizobia, induce the formation of nodules on roots of legumes (Fabales) and, exceptionally, on those of a non-legume, Parasponia sp. (Cannabaceae, Rosales). Filamentous Gram-positive soil actinobacteria from the genus Frankia induce nodules on roots of 25 genera from eight families of the Fagales, Rosales, and Cucurbitales, collectively called actinorhizal plants [1]. Both symbioses possess a common ancestor and are restricted to a single clade, the Fabids [2]. Phylogenetic studies indicate that the ancestor of Fabids was symbiotic [2–4], however proof of whether this ancestor could already form root nodules remains to be shown [5]. While legume symbioses are essential in agriculture insofar as they yield protein-rich seeds while rendering their host plants with an independent source of nitrogen fertilizer, thereby contributing to increases of nitrogen pools in surrounding ecosystems, actinorhizal species mostly represent pioneer plants and are often used in reforestation or soil recovery [6].

In all these interactions, the microsymbionts are not vertically transmitted; although the microsymbionts can exceptionally be carried by host-produced seeds [7], nodule induction always starts with bacteria in the soil, not with bacteria in the plant. Thus, each plant must be colonized ab initio in a process that requires signal exchange. Rhizobia produce lipochitooligosaccharide (LCO) nodulation factors (Nod factors; [8]) to induce the response of early nodulins genes. In comparison with model legumes, whose response to Nod factors have been analysed in detail, and enabled the characterization of many components of the Nod factor signal transduction pathway [9], in actinorhizal plants the key molecular components remain less understood. In fact, in actinorhizal plants, only a few genes encoding enzymes involved in nodule induction have been identified and characterized and the foundation of these studies was based on direct homology with the corresponding legume proteins [10]. On the microsymbiont side, there is indication that most Frankia strains do not produce LCO Nod factors [11]. Notwithstanding, homologs of the rhizobial canonical nod genes nodABC, which code for enzymes committed with steps for the assembly of LCO backbone, are present in two strains of the earliest branching Frankia clade and are expressed in symbiosis [12, 13]. Frankia strains infecting Datisca glomerata and Datisca cannabina (Datiscaceae, Cucurbitales) contain nodABC genes, however experimentation with these strains remains hampered by the impossibility of culturing them, making it challenging to address questions concerning the symbiotic signalling associated with these ancient symbioses.

The involvement of phytohormones in legume nodule induction has been examined in detail. Cytokinin signaling is central to nodule organogenesis and it has been shown that the external application of cytokinin or an autoactivated form of a cytokinin receptor can be sufficient to induce the formation of legume nodules in the absence of rhizobia; however, cytokinin has a negative effect on infection [14–18]. It has to be pointed out that in the only non-legumes nodulated by rhizobia, Parasponia spp., the involvement of cytokinin signalling in nodule organogenesis is not yet clear since knockout mutants of one cytokinin receptor are not affected in nodulation; this may be a result of receptor redundancy–or not [19]. The role played by cytokinin in nodulation has also changed when the intercellular infection pathway of the model legume Lotus japonicus was examined, in that knockout mutants of the cytokinin receptor showed a much stronger phenotype than those used for intracellular infection studies [20]. Multiple studies have shown the involvement of auxin signaling not only in the induction of legume nodules but also on their actinorhizal counterparts [21–24]. Altogether, it is widely accepted that the interplay of cytokinin and auxin signalling is required to form a local and polarized auxin maximum, essential to trigger cell division and nodule primordium formation [25–27]. Gibberellin signalling mediated by DELLA proteins interacting with different transcription factors underlines a positive and a negative role: gibberellins promote nodule organogenesis [28], but inhibit nodule infection [29–32]. Controversial roles were also discussed for ethylene: while negative effects were reported on induction of legume nodule primordium and infection thread formation (reviewed by [33]), positive effects were shown for ethylene governing the transport of auxin from the shoot to the nodulation zone [34]. Furthermore, a positive effect for ethylene was also observed on the nodulation of semi-aquatic legumes under flooded conditions [35].

Phytohormone signalling is involved in several steps of legume nodule induction [36]. Expression of the first transcription factor specifically involved in nodulation, Nodule Inception (NIN), is induced by Nod factor signalling via Ca²⁺ spiking sensed by the calmodulin-dependent protein kinase CCaMK which interacts with the transcription factor CYCLOPS [37]; NIN expression can also be induced by exogenous supply of cytokinin in L. japonicus [17, 18]; NIN acts as a central regulator and is crucial to induce the expression of downstream nodule-specific transcription factors, such as the CCAAT box-binding NF-Y transcription factor NF-YA1 [38–40]. Previous studies indicate that NIN is involved in nodule organogenesis and infection in the actinorhizal species Casuarina glauca (Casuarinaceae, Fagales) [11, 41]. D. glomerata is one of the few symbiotic species known so far whose genome contains two copies of NIN [3]. Promoter sequence analysis of D. glomerata NIN1 led to the identification of multiple cytokinin- and auxin-responsive elements [42–45], suggesting that DgNIN1 signaling might depend on auxin and/or cytokinin.

Nodules of D. glomerata differ from those of L. japonicus in anatomy, physiology, and ontology [46]. To understand whether the symbiotic programs of these two species share conserved mechanisms for nodule induction, which might unravel conserved traits across far related lineages, we set out to investigate the effects exerted by auxin and cytokinin on expression of D. glomerata NIN1, while substantial information already existed for L. japonicus NIN [17, 47, 48]. The analysis was extended to the orthologous genes DgCYCLOPS and DgNF-YA1, which, based on direct inference from research carried on model legumes, act upstream and downstream of DgNIN1, respectively; DgNF-YA1 codes for the CCAAT box-binding transcription factor NF-YA1 and its promoter is transactivated by NIN in model legumes [38–40].

Using an axenic hydroponic system that facilitates the transfer of plantlets from one medium to another, gene expression studies (RT-qPCR) on L. japonicus and D. glomerata were performed on three genes which are induced consecutively during nodule induction in legumes: CYCLOPS, NIN1, and NF-YA1. Their expression in response to treatment with cytokinin and auxins was examined alone and in combination with inhibitors of gibberellin and ethylene signalling. The synthetic cytokinin 6-Benzylaminopurine (BAP) was chosen for this study because it had been used in several legume nodule induction studies [17, 49]. The decision for the type of auxin was onerous because the most common endogenous plant auxin, Indoleacetic Acid (IAA), is light sensitive and thus may lead to false negative results when applied in glass vials. The most commonly used synthetic auxin, 1-Naphthaleneacetic Acid (NAA), in contrast to natural auxins, diffuses through membranes in the absence of specific transporters, which, again, might distort the results [50]. Since the dominant auxin in roots of D. glomerata is not Indoleacetic Acid, but Phenylacetic Acid (PAA) [51], and significant amounts of PAA were also found in other actinorhizal plants [22, 23], special emphasis was given to PAA in this study. Since gibberellin and ethylene are involved in nodule induction, in particular ethylene whose effects are known to often overlap with those of cytokinin [18, 19, 33, 52, 53], the effects of the ethylene biosynthesis inhibitor aminoethoxyvinyl glycine (AVG) and those of the gibberellin biosynthesis inhibitor paclobutrazol (PBZ) were examined in this study.

With this work, we strove to extend to actinorhizal nodules the abundant body of knowledge that transpired from studies carried on model legumes to what concerns the role played by auxin, cytokinin, ethylene, and gibberellin signalling in legumes root nodules development.

Results

Establishment of a hydroponic system and identification of reporter genes for phytohormone responses in roots of D. glomerata

To analyse the effects of phytohormones on the expression of genes involved in nodule development, we established an axenic hydroponic system that served to exogenously challenge roots of Lotus japonicus and Datisca glomerata (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Establishment of an axenic hydroponic system.

Seeds were surface-sterilized and imbibed with sterile MilliQ water at 4ºC for 5 days in the dark. After thorough washing, seeds were placed on 20x20 (mm) metal grids and allowed to germinate in a thin layer (ca. 2 mm) of ¼ strength Hoagland’s medium [87] supplied with 10 mM nitrogen (Hoaglands N⁺) and 0.8% plant agar (Duchefa). Panel A shows a detail of a recently germinated Lotus japonicus root. One week after germination, metal grids containing germinated seeds (like the one depicted in panel A) were placed over vials filled with 18 ml of Hoagland’s N⁺. This way, recently emerged root tips were able to develop within a fully controlled environment. Growing proceeded in Magenta boxes under white light [200 μEm^{−1 s−1} m⁻² at 16h light/23°C and 8h dark/18°C]. Panels B and C show hydroponically treated roots of Datisca glomerata.

https://doi.org/10.1371/journal.pone.0315798.g001

D. glomerata contains two copies of NIN [3] and although both NIN1 and NIN2 are induced in nodules compared to roots, only NIN2 is expressed nodule-specifically with no detectable levels of expression on roots of either seedlings or greenhouse-grown plants (Fig 2A; for NIN1 see also [55]). Cathebras et al. [54] have shown the presence of the cis-regulatory element PACE within the promoter of DgNIN2 (termed DgNIN1 in [54]) as a requirement for induction by the CCaMK/CYCLOPS complex, whilst the promoter of DgNIN1 contains an uncommon and extended version of PACE, which compared to that of DgNIN2, or those of other root nodule-forming plants, contains an insertion of eight nucleotides at the core of PACE and displays high nucleotide conservation across PACE-Y and PACE-X, and a fair conservation (ca. 67%) at the core of the PACE element (Fig 2B). DgNIN1 combined with its PACE element could complement the nin15 mutant of L. japonicus, i.e., this non-standard PACE element conferred induction by the CCaMK/CYCLOPS complex of L. japonicus [54]. To examine whether DgNIN1 PACE holds physical properties for distortion and to adopt a structural conformation to bend around a protein, inference about the DNA rigidity of the helices formed by PACE was conducted by propeller-twisting (ProT) of base-pairs. The angle predicted for DgNIN1 PACE by ProT analysis was always lower than 11,06º across the landscape of 37 nucleotides, a measure that correlates with lower rigidity of the helices spanning the PACE sequences DgNIN2 and those of the model legumes L. japonicus and Medicago truncatula (Fig 2C). To gain a better understanding of the PACE-CYCLOPS interactions in D. glomerata, we compared the amino acid sequence of DgCYCLOPS with those of CYCLOPS proteins from a range of root nodule-forming plants and one non-nodulator, Trema tomentosa. Alignment of CYCLOPS orthologs showed that the primary structure of CYCLOPS, along with its ortholog in Ceanothus thyrsiflorus, displays a ~40 amino acid sequence extension at its N-terminus that is neither present in Alnus glutinosa, Parasponia andersonni, nor in the legumes included in the analysis (S2A Fig); importantly, the first 20 amino acids represent a disordered domain whose presence at the N-terminus seems to be a feature of actinorhizal plants as it was only found in Datisca glomerata (position 1–20), Alnus glutinosa (position: 1–33) and Ceanothus thyrsiflorus (position 16–36) (S2A Fig). Furthermore, 17 amino acids upstream of the largest and well-conserved disordered domain, placed at the C-terminus, the primary structure of DgCYCLOPS shows high dissimilarity with that of other orthologs (Fig 2D; full alignment in S2A Fig). To investigate whether this dissimilarity could lead to alterations in protein secondary structure, AlphaFold structure predictions were carried out and the comparison of the predicted models showed the presence of an α-helix in Datisca glomerata that is absent in Ceanothus thyrsiflorus, L. japonicus, and in Medicago truncatula (Fig 2E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The PACE elements of DgNIN1 and DgNIN2 put in context.

(A) Expression profile of genes encoding orthologs of nuclear transcription factors associated with nodule development in Datisca glomerata; transcript abundance was analysed by RT-qPCR in roots (R) and nodules (N) of greenhouse-grown plants and is given relative to that of the housekeeping gene EF1-α (n = 3 for both technical and biological replicates); differences between R and N are shown at p<0.001 (student’s t test). (B) In silico analysis of the cis-regulatory element PACE in DgNIN1 and DgNIN2 promoters; representative logo profile of PACE depicted as position weight matrix out of 14 FaFaCuRo species and 2 non-nodulators (Prunus persica and Ziziphus jujuba) along with a comparison of NIN PACE sequences in D. glomerata showing an 8-nucleotide insertion at the core of DgNIN1 PACE. (C) Evaluation of secondary order effects by propeller-twist (ProT) of base-pairs across the PACE landscape of D. glomerata and model legumes. (D) Implications of primary sequence dissimilarity in secondary structure of CYCLOPS shown as 3D models. Partial alignment of CYCLOPS from 8 nodulating species and 1 non-nodulator (Trema tomentosa) shows dissimilarities in a region encompassing 24 residues (see dashed box) (full alignment in S2A Fig). (E) AlphaFold 3D models showing the implications in secondary structure of DgCYCLOPS at the site of the region covered by the dashed box, highlighting the presence of an α-helix in DgCYCLOPS, from 402Q to 412D, which is absent in Ceanothus thyrsiflorus and model legumes.

https://doi.org/10.1371/journal.pone.0315798.g002

Like for M. truncatula NIN, promoter sequence analysis of D. glomerata NIN1 [3] showed the presence of multiple cytokinin- and auxin-responsive elements [42–45] (see https://doi.org/10.17045/sthlmuni.9275285.v1). To link transcriptional responses in roots to exogenous application of phytohormones, the genetic dependencies of the effects exerted by phytohormones had to be confirmed independently. Although marker genes for phytohormone responses are well established in several plant species, so far none were characterized in D. glomerata. To assess potential candidates that could serve as molecular readouts for phytohormone responses in D. glomerata, we conducted homology analysis by mining a previously published transcriptome [55, 56]. For cytokinin, three potential marker genes were identified from the orthologous group ENOG411BR8V and named after the two-component response regulator orthologs ARR4, ARR5, and ARR9 of Arabidopsis thaliana. Preliminary assays confirmed ARR9 as a suitable marker gene for cytokinin in D. glomerata, owing to an increase of ARR9 mRNA levels in the presence of BAP (see below).

To identify reporter genes for auxin, we searched for candidate members of the Gretchen Hagen3 (GH3), Small Auxin Upregulated RNA (SAUR), and GATA families. Although some representatives of the GATA family of transcription factors could be identified in the nodule transcriptome of D. glomerata (GATA15-, 16-, 24-, and 26-like), none of these candidates were orthologous to auxin-responsive GATA23 of Arabidopsis [57]. However, a member of the Small Auxin Upregulated RNA (SAUR) gene family was identified based on the orthologous group ENOG410JVA8 and the encoded protein shares high sequence similarity with other relatives in Viridiplantae (SAUR1; S1A Fig). Furthermore, the hidden Markov model returned by the profile of AtGH3.5 enabled the identification of two proteins in D. glomerata, which we named GH3.1 and GH3.2. These proteins were analysed in silico for the presence of motifs previously linked with auxin responsiveness [58] and this analysis indicated that GH3.1 was very likely orthologous to the auxin responsive AtGH3.5/WES1, while GH3.2 was probably responsive to jasmonic acid [59] (S1B Fig). As a result, SAUR1 and GH3.1 were selected for downstream analysis as putative auxin markers in D. glomerata.

To validate the responsiveness of GH3.1 and SAUR1 to auxins, their transcript abundance was analysed by real-time quantitative PCR (RT-qPCR) on cDNA from roots treated with NAA and PAA across a wide range of concentrations. Under the conditions tested in this experiment, only GH3.1 showed a convincing transcriptional response to NAA (p = 0.09, 10 nM; p = 0.0021, 100 nM; p = 0.0014, NAA 500 nM; S2 Table; S1C Fig). Given that no differences were observed between 100 nM and 500 nM, this observation suggested that 100 nM may be already above the optimum for GH3.1 response to NAA. Expression levels of SAUR1 were neither affected by NAA nor by PAA, but the range of concentration tested for PAA in this assay (5nM—25nM) was very likely below the optimum (S1C Fig). Because PAA is the dominant auxin in roots of D. glomerata [51], we reasoned that the response of SAUR1 should be examined in greater detail for a higher concentration of PAA (see below). Taken together, results showed GH3.1 as a suitable candidate to address effects exerted by NAA, whereas SAUR1 did not respond to NAA (10nM—500nM) nor to PAA (<25 nM) under the conditions applied in this assay.

Use of the hydroponic system to address responses in gene regulation in roots exogenously treated with phytohormones: Proof of concept using the model legume Lotus japonicus

To assess the possibility of using the hydroponic system coupled with RT-qPCR to assess the effects of phytohormones on nodulation-related gene expression in D. glomerata, the effects exerted by phytohormones were first examined in the well-studied symbiotic model L. japonicus. The expression of L. japonicus cytokinin oxidase/dehydrogenase3 (LjCkx3; [60]) was induced by BAP when applied for a period of 8 h (p<0.01) and its effect persisted after 24 h at 100 nM (p<0.01; Fig 3). Auxin responsiveness was analysed based on the expression of the well-known L. japonicus auxin-responsive promoter LjGH3 [61, 62], which in this assay was induced exclusively by NAA at 10 nM and 100 nM when applied for 8 h or 24 h (p<0.01). Yet, LjGH3 suppression by 100 nM of BAP applied for 8 h (p<0.01) indicates that auxin and cytokinin act antagonistically on LjGH3 expression in L. japonicus roots.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Lotus japonicus transcriptional response to cytokinin (BAP) and auxins (PAA and NAA).

Transcript abundance was analysed by RT-qPCR after 30-day-old Gifu roots treated during 8h and 24h (technical replicates, n = 3). Y-axis shows the mRNA quantity relative to that of PUQ, ATPs, and PP2A. Phytohormones and molarities are given on the X-axis. With the exception of Ctr, 24h (n = 5), Ctr, 8h; PAA100, 8h; NAA100, 8h/24h (n = 4), the number of biological replicates was n = 3. Gene names are given. Significant differences to control are highlighted at p<0.05 (*) and p<0.01 (**).

https://doi.org/10.1371/journal.pone.0315798.g003

As for differential regulation of LjCYCLOPS and LjNIN expression in roots of L. japonicus, the expression level of LjCYCLOPS did not change under any of the conditions examined, in contrast with that of LjNIN which did change in response to cytokinin and auxins. The transcription of LjNIN was induced after 8 h of exposure to 100 nM of BAP (p<0.01) and this effect persisted after 24 h at both 10 nM and 100 nM (p<0.05). The auxin NAA induced changes of LjNIN expression in roots treated with 10 nM or 100 nM over a period of 8 h or 24 h, respectively (p<0.01). Likewise, NIN was induced by the natural auxin PAA, but only in roots exposed for 24 h to a concentration of 10 nM (p<0.05).

These results show that the cytokinin BAP is able to modulate the expression of LjNIN in a dosage- and time-dependent manner when exogenously applied to roots of 30-day-old seedlings grown hydroponically, which is in line with a previous study carried with roots grown in Petri dishes on agar containing BAP [15]. Thus, proof of concept was obtained.

Nanomolar concentrations of the synthetic cytokinin BAP repressed the transcription of CYCLOPS, NIN1, and NF-YA1 in 30-day-old roots of D. glomerata seedlings

The D. glomerata orthologs of CYCLOPS and NF-YA1 were identified phylogenetically (S2 and S3 Figs). To address the question whether the phytohormone background required for nodule development in D. glomerata shares commonalities with that of L. japonicus, 30-day-old roots of D. glomerata were challenged with concentrations of BAP ranging from 10 nM to 500 nM, in 24 h assays. Unlike the transcriptional response observed in roots of L. japonicus (see Fig 3), increasing amounts of cytokinin led instead to repression in expression of both CYCLOPS (p = 0.09, 50 nM BAP; p<0.001 for both 100 nM BAP and 500 nM BAP) and NIN1 (p = 0.032, 500 nM BAP) in a dosage-dependent manner (S2 Table; Fig 4). The transcriptional profile exhibited by CYCLOPS in this assay indicates that its transcription is tightly regulated by BAP levels within the range of 10 nM to 100 nM. Since the promoters of legume Nuclear Factor-Y (NF-Y) subunit genes are expected to be targeted and activated by NIN1 in analogy with legumes [38–40], we then set to investigate whether increases in BAP levels would be accompanied by changes in expression of NF-YA1. Notably, the transcriptional profile exhibited by NF-YA1 was similar to that of CYCLOPS and NIN1 insofar as the transcript abundance of NF-YA1 decreased inversely with BAP levels (p<0.001, 100 nM and 500 nM; S2 Table; Fig 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Datisca glomerata transcriptional response to cytokinin (BAP).

Transcript abundance was analysed by RT-qPCR after 30-day-old roots treated during 24h (technical replicates, n = 3). Y-axis shows the mRNA quantity relative to that of EF1-α, PUQ and TIP41. BAP molarities are given on the X-axis (number of biological replicates, n = 6). Significant differences to the control are highlighted at p<0.05 (*) and p<0.01 (**). Gene names are given.

https://doi.org/10.1371/journal.pone.0315798.g004

Transcription of NIN1 and NF-YA1 was induced by nanomolar concentrations of PAA in 52-day-old roots of D. glomerata

Since in D. glomerata, BAP acted antagonistically towards the expression of genes whose protein products act upstream of NF-YA1 induction in model legumes, we hypothesized that either D. glomerata roots exhibit a different temporal window in which nodulation-related gene expression can be activated when compared with roots of L. japonicus, or that instead of cytokinin, another phytohormone was involved in regulating the expression of these nodulation-related transcription factors. The different temporal window could be explained by the fact that while legume seeds are large and nutrient-rich [63] which means that legume seedlings can afford the carbon expense required for nodule formation, seeds of D. glomerata are similar in size to those of Arabidopsis, and greenhouse plantlets do not nodulate before they have reached a height of at least 7 cm (K. Pawlowski, unpublished observations).

In a trial experiment using roots of 30-day-old D. glomerata seedlings, the exogenous application of 50 nM PAA did not lead to any changes in expression of either CYCLOPS or NIN1 in 24 h treatments (results not shown). To explore the possibility that root development could have a pivotal influence on the transcription of CYCLOPS and NIN1 in assays carried with PAA, a time course experiment was conducted to test the effects of exposure for 2 h, 8 h, and 24 h but, this time, in 54-day-old roots. This assay indicated that the temporal dynamics of NIN1 expression in response to PAA may be linked to and depend on the age of the plant, meaning that ideally the treatments should be performed on roots of ~54-day-old seedlings over a 24 h period (p = 0.013; S2 Table; S4 Fig). This result was then reproduced in two independent series of experiments showing induction of NIN1 in PAA-treated roots at concentrations as low as 10 nM (p = 0.008) and 50 nM (p = 0.006) (S2 Table; Fig 5). Consistently, 50 nM PAA exerted a positive effect on NF-YA1 transcription (p = 0.037) (S2 Table; Fig 5). Although neither BAP nor NAA exerted an effect on the expression of CYCLOPS, NIN1, or NF-YA1, the changes observed on the expression of their respective marker genes ARR9 (p = 0.036, 10 nM BAP; p = 0.008, 50 nM BAP) and GH3.1 (p = 0.2, 10 nM NAA; p = 0.002, 50 nM NAA) convincingly reflected the effects of these phytohormones (S2 Table; Fig 5). Collectively, these findings reinforce the idea that, in D. glomerata, GH3.1 represents a marker gene to address responses to NAA, while SAUR1 represents a marker gene for responses to PAA (p = 0.061, 10 nM PAA; p<0.001, 50 nM PAA; S2 Table). More important, these findings suggest that DgNIN1 mRNAs were translated, and their products were able to induce the expression of their presumable target gene, DgNF-YA1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Datisca glomerata transcriptional response to cytokinin (BAP) and auxins (PAA and NAA).

The data was collected from two independent experiments (empty and full circles). Transcript abundance was analysed by RT-qPCR after 52-day-old roots treated during 24h (technical replicates, n = 3). Y-axis shows the mRNA quantity relative to that of PUQ and TIP41. Phytohormones and molarities are given on the X-axis (number of biological replicates: n = 8 for Ctr, P10, and N50; n = 7 for P50 and N10; n = 6 for B10 and B50). Significant differences to the control are highlighted at p<0.05 (*) and p<0.01 (**). Gene names are given.

https://doi.org/10.1371/journal.pone.0315798.g005

Addition of BAP or inhibition of ethylene or gibberellic acid biosynthesis abolished the induction of NIN1 by 50 nM PAA

In an attempt to understand the effects caused by the removal of ethylene and gibberellic acid (GA), inhibitors of their biosynthesis were introduced into the experiments, namely L-alpha-(2-aminoethoxy vinyl) glycine (AVG) and paclobutrazol (PBZ), which inhibit, respectively, key enzymes for ethylene and GA biosynthesis. Experiments without added inhibitor (control group) were included to interpret the crosstalk between ethylene or GA signalling and auxin and cytokinin signalling. The cytokinin BAP was used at 10 nM, the natural auxin PAA was used at 50 nM, and, in the combined treatment, 10 nM of BAP were mixed with 50 nM of PAA.

The effects exerted by BAP in this experiment were not as prominent as in previous assays, however BAP effects were still moderately at play based on the expression of ARR9 in the control group (p = 0.057) (S2 Table; Fig 6). SAUR1 expression was significantly induced by PAA in the control group (p<0.01), but this induction was abolished in the presence of BAP (p = 0.081), AVG (p = 0.18) or PBZ (p = 0.25) (S2 Table; Fig 6). Consistent with these results, expression of NIN1 (p = 0.015) and NF-YA1 (p = 0.006) was also induced by PAA in the control group, but not when BAP was added as well (NIN1, p = 1.0; NF-YA1, p = 1.0), or when ethylene (NIN1, p = 0.77; NF-YA1, p = 0.089) or GA (NIN1, p = 0.31; NF-YA1, p = 1.0) were depleted (Fig 6; S2 Table). It is worthwhile to note that roots treated with PBZ for 11 days before analysis showed a significant decrease in mRNA abundance of CYCLOPS (p = 0.031) when compared with their control and AVG-treated counterparts (Fig 6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Datisca glomerata transcriptional response to BAP, PAA, and the co-treatment BAP+PAA under the presence or absence of inhibitors for ethylene (AVG) or gibberellin biosynthesis (PBZ).

Transcript abundance was analysed by RT-qPCR after 52-day-old roots treated during 24h (technical replicates, n = 3). With the exception of BAP10+AVG, BAP10+PAA50+AVG, Ctr+PBZ, and PAA50+PBZ (n = 4), the number of biological replicates was n = 3). Y-axis shows the mRNA quantity relative to that of TIP41. Phytohormones and molarities are given on the X-axis (+, present; -, absent). Significant differences to the control are highlighted at p<0.05 (*) and p<0.01 (**) after Tukey HSD test. Gene names are given.

https://doi.org/10.1371/journal.pone.0315798.g006

In summary, ethylene and gibberellin were required for the induction of SAUR1, NIN1 and NF-YA1 expression by PAA, while BAP abolished the induction of the expression of these three genes by PAA (Fig 7). In combination with the negative effect of BAP on the expression of CYCLOPS, NIN1 and NF-YA1 in 30-day-old plantlets (Fig 4), these results suggest that in contrast with legumes, cytokinin acts negatively on the nodulation of D. glomerata.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Summary of the main findings reported in this manuscript.

Positive and negative effects of exogenously added phytohormones auxin (PAA and NAA), cytokinin (BAP), or inhibition of endogenous biosynthesis of ethylene or gibberellins are shown for Datisca glomerata (black) and Lotus japonicus (blue) on the expression of key genes encoding transcription factors involved in nodule development (CYCLOPS, NIN, and NF-YA1) in 30- and 52-old-day seedlings (see red dashed line). Information about the Figure in the manuscript from which the evidence was drawn is provided on top. CYCLOPS and NIN proteins are respectively highlighted in maroon and red.

https://doi.org/10.1371/journal.pone.0315798.g007

Discussion

With this study we sought to (i) gain qualitative insight in the involvement of phytohormones in the developmental program of actinorhizal nodules of Datisca glomerata, (ii) quantify the impact of phytohormones on transcription of early transcription factors in nodule development and (iii) build up knowledge on the prospect that the D. glomerata NF-YA1 promoter is, as is the case in model legumes, targeted by the transcription factor NODULE INCEPTION (DgNIN1; [55]). From here, we set out to understand the regulation of DgNIN1 and DgNIN2; even though no functional studies have been carried out for either NIN gene, their relevance for root nodulation has been established by comparative phylogenomics and phylotranscriptomics studies [3, 64]. In silico analysis of PACE sequences from DgNIN1 and DgNIN2 promoters support the possibility of protein-DNA interactions (Fig 2B and 2C), a prediction that is line with the findings of Cathebras et al. [54] that have shown that the PACE elements of both DgNIN genes suffice for induction by the Lotus japonicus CCaMK/CYCLOPS complex; indeed, DgNIN1 combined with its PACE element could complement the nin15 mutant of L. japonicus, meaning that this non-standard PACE element conferred induction by the CCaMK/CYCLOPS complex of L. japonicus [54]. Thus far no functional studies were performed on CYCLOPS of actinorhizal plants, however protein 3D models analysis indicate that the distinct features of DgCYCLOPS might correlate with a larger plasticity of the protein [65], in comparison with CYCLOPS of other actinorhizal plants or those of legumes (Fig 2D and 2E). Altogether, there is no reason to exclude the possibility that DgCYCLOPS might play a pivotal role by transactivating both DgNIN1 and DgNIN2, but evidence about these events has yet to be provided.

The induction of NIN by exogenously applied cytokinin had been demonstrated for the model legumes Lotus japonicus [17] and Medicago truncatula [18] using seedlings growing on agar in Petri dishes. As proof of concept for our hydroponic system, we took advantage of the knowledge transpired from these studies to address the response of NIN transcription to exogenously applied synthetic cytokinin 6-Benzylaminopurine (BAP) and synthetic auxin 1-Naphthaleneacetic acid (NAA) in roots of L. japonicus, using L. japonicus Ckx3 cytokinin oxidase/dehydrogenase3 as a marker gene for the cytokinin response [61], and Gretchen Hagen3 (GH3; [62]) as marker gene for the auxin response. In line with the findings of Heckmann et al. [17], we showed that LjNIN expression could be induced using BAP and NAA in the hydroponic system (Fig 3). Additionally, we showed that the expression of LjNIN could also be induced in a time- and dosage-dependent manner by the natural auxin Phenylacetic Acid (PAA), although PAA could not be detected in roots of L. japonicus [66]. Interestingly, expression of the auxin response marker gene GH3.1 was upregulated by NAA, but not by PAA. This divergence suggests that some auxin effects are likely to have been overlooked in L. japonicus.

The LjNIN promoter is activated by the CYCLOPS/CCaMK complex [37]. Since the expression of LjCYCLOPS was neither induced by BAP, NAA nor by PAA in this study, we conclude that the induction of LjNIN expression did not require an increase in CYCLOPS mRNA. The induction of LjNIN by cytokinin had been linked to the well-established fact that cytokinin signaling occurs in the root cortex during legume nodule organogenesis [15]. The induction of LjNIN by auxin was not reported previously and is likely to be linked to the accumulation of auxin at the onset of nodule induction as well as to the fact that NIN recruits the lateral root development program for nodule organogenesis [25, 67–69].

When establishing marker genes to track responses to NAA, PAA, and BAP in D. glomerata, the results were in line with those observed previously for L. japonicus: the expression of the ortholog of the cytokinin-induced type A response regulator gene ARR9 from Arabidopsis [70] was upregulated by BAP in D. glomerata, while the expression of the D. glomerata ortholog of LjGH3, DgGH3.1, was upregulated by NAA. DgSAUR1, a member of another group of genes known for auxin responsiveness [71], was identified and served as marker gene for PAA effects. PAA had been discovered as auxin-type phytohormone in 1935 [72–74]; however, because experiments on pea and oat showed that the activity of PAA was much lower than that of IAA, PAA was not examined in detail. The synthesis of PAA does not follow the same pathway as that of IAA [75], and it does not show basipetal transport like IAA [76, 77]. In spite of this, IAA and PAA effects on gene expression showed some overlap when examined for Arabidopsis, e.g., four members of the Arabidopsis GH3 family were induced by both IAA and PAA (GH3.2, GH3.3, GH3.4 and GH3.5), while all Arabidopsis SAUR genes reacting to auxin were only induced by IAA, not by PAA [77]. It should be pointed out, however, that Sugawara et al. [77] examined gene expression already after one hour of exposure to auxin and used much higher auxins concentrations, namely 1 μM IAA and 10 μM PAA– 20 or 200 times higher, respectively, than the concentration established as optimal for the induction of LjNIN/DgNIN1 expression in this study, 50 nM for both NAA and PAA. At any rate, these results showed that the differential effects of NAA and PAA on gene expression during nodule development require further examination.

As discussed above, the induction of legume NIN expression by exogenously applied cytokinin had been ascribed to the fact that cytokinin signalling takes place in the root cortex in the course of legume nodule organogenesis, leading to the formation of the legume nodule primordium [17, 18]. On the other hand, shoot-derived cytokinin transported in the phloem has been implicated in the systemic repression of nodulation during the autoregulation of nodulation [78]. Legume nodule primordia are induced in the root cortex and the root pericycle, while actinorhizal nodule primordia are induced in the root pericycle [46]. Therefore, we do not necessarily expect the same involvement of phytohormones in both processes. For induction of an organ primordium in the root pericycle, close to the auxin maximum at the protoxylem pole [79], auxin would be expected as inducer. Indeed, the results of this study showed unambiguously that the expression of D. glomerata NIN1 is induced by the natural auxin PAA, though not by the synthetic auxin NAA, and that the induction by PAA is abolished in the presence of the synthetic cytokinin BAP. The difference between PAA and NAA effects on DgNIN1 expression is consistent with the facts that (i) PAA is the dominant auxin in D. glomerata roots [51] and (ii) NAA is unlikely to replace PAA in every context since both auxins were shown to be transported differently in Arabidopsis [77]. Intriguingly, when 50 nM PAA and 10 nM BAP were applied simultaneously, the induction of expression of DgNIN1 as well as that of DgSAUR1 by PAA was abolished (Fig 6). This was a striking finding that, for DgNIN1, can be discussed from two angles: first, it reinforces the idea that BAP is indeed acting as a negative regulator of DgNIN1 expression, but treatment with 10 nM BAP alone does not seem to suffice to reduce DgNIN1 expression in 30-day-old plants. At this age, as shown in Fig 4, BAP has the strongest negative effect on the expression of DgCYCLOPS, in concentrations equal to or above 50 nM, while it only affected DgNIN1 expression at 500 nM. Combined, the data suggest that in the presence of exogenous BAP, two synergistic effects for negative regulation of DgNIN1 expression occur: one, direct, via downregulation of DgNIN1 transcription, the other, indirect, abolishing the induction by PAA. In this context, it is interesting to note that the response of DgCYCLOPS expression to 50 nM PAA was inconsistent throughout our experiments, ranging from non-induced (Fig 5, p = 0.19; Fig 6, p = 0.12) to induced (S4 Fig; p = 0.013). This raises the possibility that the discrepancies observed were linked to variable amounts of endogenous auxins or cytokinins (S5A Fig).

The response difference between 30-day-old D. glomerata seedlings to phytohormones compared to 52- or 54-day-old seedlings can be explained by the fact that nodulation requires a considerable carbon investment which has to be supplied from seeds or from photosynthesis. Legume seedlings are routinely nodulated in Petri dishes, i.e., very young, given that they obtain most of their nutrients from their large seeds. This is also possible for seedlings of the actinorhizal shrub Coriaria myrtifolia which forms seeds similar in size to those of L. japonicus [80]. However, in 30-days-old D. glomerata seedlings grown from seeds of 0.8–1 mm diameter, nodulation has to be downregulated before they have produced enough leaves to provide photosynthates for additional strong carbon sinks (nodules), a process that would likely abolish or at least weaken the effects of exogenous application of phytohormones on DgNIN1 expression, as was observed in this study–no effect of PAA was found, and BAP only acted at 500 nM.

Altogether, our results support the observations of Gauthier-Coles et al. [81] for the actinorhizal tree Alnus glutinosa who showed that here, cytokinin (cis-zeatin, trans-zeatin, dihydro-zeatin, and kinetin were used) does not act as positive regulator of pseudonodule development as it does in legumes. Rodríguez-Barrueco et al. [82], on the other hand, found pseudonodule development in response to treatment with two different cytokinins, 2-isopentenyl adenin (2IP) and kinetin; however, the latter study was not performed under axenic conditions, which means that other phytohormones produced by bacteria or fungi may have been involved.

Regarding DgSAUR1 repression, the available evidence suggests that the induction of DgSAUR1 expression by exogenous PAA is abolished by the addition of exogenous BAP, which might be connected to the presence of a dozen of cis-regulatory motifs for cytokinin response in its promoter and which may bind repressor(s) (S5B Fig). Interestingly, the hypothesis that cytokinin is downregulating DgSAUR1 expression is similar to the situation observed for L. japonicus GH3 in presence of 100 nM BAP (Fig 3), further emphasizing that the interplay between auxins and cytokinins must be considered when using these promoters as markers.

To test the effects of ethylene, D. glomerata plantlets were grown on media supplied with an inhibitor of ethylene biosynthesis, 10 μM of L-alpha-(2-Aminoethoxyvinyl) glycine (AVG). Inhibition of ethylene biosynthesis had a negative effect on DgNIN1 expression (Fig 6). Similarly, DgSAUR1 was not induced by exogenous PAA In the presence of AVG (Fig 6). In short, the effect of AVG in combination with PAA resembled the effect of the combination of BAP and PAA. This effect could be ascribed to increased amounts of cytokinins as in legumes, ethylene negatively regulates cytokinin accumulation during nodule induction [19]. However, it has to be mentioned that in the actinorhizal species Casuarina glauca, whose nodule primordia form in the root pericycle like in D. glomerata, ethylene has a negative effect on nodulation like in legumes [83]. Another possible explanation could involve the role of ethylene in the transport of auxin from the shoot to the nodulation site, shown to result in high auxin accumulation [34] and implying that endogenous auxin levels in AVG-treated roots could be below the level required for the induction of DgNIN1. At any rate, the fact that the inhibition of ethylene biosynthesis abolished the induction of DgNIN1 expression by PAA in D. glomerata roots (p = 0.77) was the second relevant difference found between D. glomerata and legumes with regard to phytohormone effects on NIN expression [33]. It is relevant to point out that studies of the effects of ethylene on legume nodulation were normally not performed in liquid culture, but on seedlings growing on solid agar. While ethylene has been found as a requirement for nodulation of Sesbania rostrata under waterlogged conditions [84], it inhibited nodulation under non-flooded conditions in the same species [85]. This suggests that the positive role of ethylene in the regulation of DgNIN1 expression might be a feature of nodulation under waterlogged conditions, or for nodulation that does not involve the formation of infection threads in root hairs, as is the case for D. glomerata [46]. However, later studies have shown that nodulation via root hairs in L. japonicus involves the production of ethylene by the plant [53].

For the inhibition of gibberellin biosynthesis, the growth conditions had to be modified to enable root development before exposure to the inhibitor paclobutrazol (PBZ); PBZ was added after 41 days of seedling growth. This pre-treatment with PBZ was necessary because the inhibition of gibberellins biosynthesis during the 24 h exposure to auxin/cytokinin would not affect existing gibberellins [86]. Thus, the possibility that a growth-retardant effect caused by PBZ interfered with the experiment cannot be ruled out, given the significant differences in the effect on DgNIN1 expression observed between roots assays on day 30 vs. day 52. At any rate, in the presence of PBZ, PAA did not cause significant changes in the expression levels of any gene examined, including that of SAUR1 (p = 0.25; Fig 6). Yet, in the presence of PBZ, the expression of DgCYCLOPS was significantly reduced (Fig 6), suggesting that the induction of DgCYCLOPS expression involves a DELLA protein. Combined, these results are consistent with the positive involvement of gibberellin signalling in legume nodule organogenesis mediated by the GRAS transcription factor DELLA1 as shown in Medicago truncatula [24].

Overall, these findings indicate that the symbiotic program for nodule induction in D. glomerata significantly differs from that of L. japonicus (Fig 7). LjNIN could be induced directly by application of Nod factors [17] as well as directly by exogenous supply of BAP and NAA, and later also by PAA, in a time- and dosage-dependent manner (Fig 3). The data obtained with the model legume L. japonicus confirm the well-known role played by cytokinin signaling towards nodule organogenesis in the root cortex of legumes, a divergent trait that has probably emerged ~60 Mya ago in nodulating species of legumes but is absent in non-nodulating legumes as well as in non-legumes, including actinorhizal plants [81, 87]. In contrast, in D. glomerata, the expression of DgNIN1 was induced by PAA, an effect that was abolished in the presence of cytokinin (Fig 6), arguing for a pivotal role of auxin in the nodulation of D. glomerata regarding the regulation of DgNIN1, and in turn DgNF-YA1. In analogy to legumes, local auxin accumulation–equivalent to cytokinin accumulation in legumes as a result of rhizobial Nod factor signaling–might be induced via Frankia signalling, maybe involving ethylene on the plant side (Figs 5 and 6), inducing DgNIN1 expression and, consequently, nodule organogenesis.

The possibility that the expression of DgNIN1 is linked to a role in the regulation of lateral root formation, as was found for lipochitooligosaccharide Nod factor-responsive NIN genes of legumes and poplar [88], should be considered, however the fact that the expression of DgNIN1 cannot be induced in plantlets too young to nodulate (Fig 4) and it shows high levels of expression in nodules (Fig 2) make it clear that DgNIN1 has to play a relevant role in nodulation. Since the standard cis-regulatory element PACE of DgNIN2 is involved in infection thread progression [54], a role for DgNIN1 and its uncommon PACE in nodule organogenesis can be hypothesized, however functional characterization studies are required to unravel the role played by DgCYCLOPS in this context. As it stands, it is tempting to speculate that DgCYCLOPS plays a pivotal role in the nodule development program of Datisca glomerata by driving nodule organogenesis via DgNIN1 and Frankia accommodation via DgNIN2. At this point, however, it is unclear whether the regulation of CYCLOPS in roots by phytohormones observed in this study is relevant for nodulation, or only for arbuscular mycorrhization. If the former, it would have to involve the induction of the expression of other transcription factor genes as DgNIN2 expression was not detected in roots under the conditions described in this study. Altogether, the data obtained in this study suggest that the ancient program for symbiosis established between Frankia and D. glomerata, whose nodules originate from the root pericycle, requires auxin to induce the expression of DgNIN1 and DgNF-YA1.

Conclusions

An axenic hydroponic system was established to examine the effects of exogenously applied phytohormones on the expression of key genes required for nodule organogenesis on roots of the actinorhizal plant Datisca glomerata and on those of the model legume Lotus japonicus, which we used as a control. Marker genes for phytohormone perception were established: DgARR9 served as marker for the synthetic cytokinin BAP, whilst DgGH3.1 and DgSAUR1 served as markers for the synthetic and natural auxins NAA and PAA, respectively. The collected data showed that DgNIN2, whose promoter contains a standard PACE cis-regulatory element, is not expressed in roots, whereas DgNIN1, whose promoter harbors a non-standard PACE, showed induction by the auxin PAA, but not by the cytokinin BAP. In contrast, LjNIN expression was induced by BAP, NAA, and, when measured 24 h after application of 10 nM, by PAA. The induction of DgCYCLOPS, DgNIN1 and DgNF-YA1 transcription by PAA was abolished when either BAP, the ethylene biosynthesis inhibitor AVG or the gibberellin biosynthesis inhibitor PZB were added together with PAA, suggesting that the induction by PAA required certain levels of ethylene and gibberellin, but could be abolished by low levels of exogenously applied cytokinin BAP. Altogether, the phytohormone involvement in the induction of nodules in the actinorhizal species D. glomerata differs from that of legumes concerning the roles of auxin, cytokinin and ethylene, while the role of gibberellin has been likely conserved across lineages.

Material and methods

Supporting information

Acknowledgments

The authors are very grateful to Max Griesmann and Martin Parniske (Department of Genetics, Ludwig Maximilians University of Munich, Germany) for multiple helpful discussions about genes and phytohormones involved in the nodulation of legumes. The support from Science for Life Laboratory, the National Genomics Infrastructure, NGI, and UPPMAX (UPPNEX project ID b2013247) computational infrastructure is gratefully acknowledged.