Cytokinins (CKs) are involved in different aspects of nodule development, playing a dual role in nodulation, depending on the time and place of their action. The exogenous application of cytokinin to legume roots induces responses similar to those of rhizobial Nod-factor (NF), including cortical cell division and expression of early nodulin genes (Cooper and Long, 1994; Heckmann et al., 2011). Moreover, the gain-of-function mutation in Lotus histidine kinase1 (Lhk1) cytokinin receptor gene causes the formation of spontaneous nodules in Lotus japonicus (Tirichine et al., 2007). The loss-of-function mutations in Lhk1 gene or RNAi-mediated downregulation of its ortholog in M. truncatula leads to a dramatic reduction in the nodule formation (Gonzalez-Rizzo et al., 2006; Murray et al., 2007, respectively). All these data suggest that cytokinins have a positive influence positively on the nodule development. However, elevated cytokinin levels can also locally contribute to the negative regulation of nodulation. ckx3 mutants that are defective in cytokinin oxidase/dehydrogenase gene, involved in cytokinin degradation, exhibit reduced nodulation and infection thread formation together with elevated root levels of tZ (trans-zeatin) and DHZ (dihydrozeatin) cytokinins (Reid et al., 2016). In addition, the lhk1 mutant that is defective in a cytokinin receptor gene in L. japonicus displays epidermal hyperinfection, suggesting that CKs negatively regulate epidermal infection (Held et al., 2014). However, in the lhk1-1 lhk1a-1 lhk3 triple mutant, the infection thread is unable to enter the root cortex, which indicates that CKs positively regulate the infection thread entry and growth in the cortex (Miri et al., 2016). Moreover, the ectopic expression of Arabidopsis AtCKX (pEPI:AtCKX3) gene in the root epidermis of Medicago truncatula results in increased nodule number, while the ectopic expression of AtCKX3 (pCO:AtCKX3) in the root cortex is associated with decreased nodule number (Jardinaud et al., 2016). Thus CKs act both positively and negatively on the nodulation depending on the tissue: while CKs are essential for the nodule primordium initiation and the infection process in the root cortex, they negatively regulate the infection in the epidermis (for review see Gamas et al., 2017).

CKs biosynthesis is a multistep process. The first and rate-limiting step of CKs biosynthesis is catalyzed by the adenosine phosphate isopentenyl transferase (IPT). Nine IPT genes (AtIPT1–AtIPT9) have been identified in Arabidopsis thaliana, exhibiting different expression patterns both in the shoots and the roots (Takei et al., 2001; Miyawaki et al., 2004). Among them, seven IPT genes (AtIPT1 and AtIPT3–AtIPT8) encode ATP/ADP-isopentenyltransferases that have been shown to isopentenylate ATP and ADP (Kakimoto, 2001). AtIPT2 and AtIPT9 are tRNA-isopentenyltransferases which are supposed to supply cis-zeatin-type cytokinins in A. thaliana (Miyawaki et al., 2004). The production of trans-zeatin (tZ) nucleotide is catalyzed by the cytochrome P450 monooxygenase (CYP735A family) via an hydroxylation of the isopentenyl adenine nucleotide (Takei et al., 2004). Eventually, the formation of active cytokinin nucleobases from nucleotides is mediated by the product of LONELY GUYs (LOGs) genes (Kurakawa et al., 2007). In legumes, several cytokinin biosynthesis genes are activated in response to rhizobium inoculation or NF treatment. For example, in L. japonicus, the expression of IPT1/2/3/4 together with CYP735A and LOG1/4 is increased in response to rhizobial inoculation (Chen et al., 2014; Reid et al., 2017). Furthermore, the overexpression of cytokinin biosynthesis pathway genes (IPT3, LOG4, and CYP735A) is sufficient to induce spontaneous nodule formation (Reid et al., 2017). In M. truncatula, the expression of some cytokinin biosynthesis genes occurs as early as 3 h after NF treatment (van Zeijl et al., 2015). Two LOG genes are activated in response to bacterial inoculation (Mortier et al., 2014). Moreover, RNA-seq analysis shows the expression of some LOG genes along with that of CYP735A1 gene (Medtr6g017325) in the epidermis of inoculated roots (Jardinaud et al., 2016). Activation of IPT and LOG genes has also been demonstrated in nodules of Pisum sativum (Azarakhsh et al., 2015; Dolgikh et al., 2017).

Furthermore, LjIPT3 is activated in the shoot phloem via the components of the AON (autoregulation of nodulation) system which negatively affects nodulation (Sasaki et al., 2014). AON represents a systemic regulation controlling the number of root nodules; it involves long distance signaling and communication between the roots and the shoot (for review see Reid et al., 2011; Mortier et al., 2012; Soyano et al., 2014). AON was shown to be activated when the first nodule primordia are formed (Caetano-Anolles and Gresshoff, 1991; Li et al., 2009). One of its key components is the product of a CLV1-like (CLAVATA1-like) receptor kinase gene [SUPER NUMERIC NODULES (SUNN) in M. truncatula and HYPER NODULATION ABERRANT ROOT FORMATION1 (HAR1) in L. japonicus], which acts in the shoot. Mutations in the CLV1-like kinase gene lead to a shoot-controlled supernodulating phenotype in different legume species (Krusell et al., 2002; Nishimura et al., 2002; Searle et al., 2003; Schnabel et al., 2005). According to the present model of AON (for review see Soyano et al., 2014), CLE-peptides produced in nodules are transported to the shoot, where they bind to the CLV receptor complex including CLV1-like kinase; consequently, the response to suppress excessive nodule formation on roots is triggered. A set of studies suggests that the CLV receptor complex may include several proteins. In M. truncatula, the proteins CORYNE (CRN) and CLAVATA 2 (CLV2) interact with SUNN (Crook et al., 2016), and a mutation in the CLV2 gene also leads to a supernodulating phenotype in legume plants (Krusell et al., 2011). In L. japonicus, the KLAVIER gene, the defect of which is also characterized by a shoot-controlled supernodulating phenotype, encodes a receptor-like kinase, structurally unrelated to CLV1-like kinase, that was shown to interact with HAR1, suggesting that these proteins may form a receptor complex which would perceive CLE-peptides (Miyazawa et al., 2010). Among the factors acting downstream of CLV1-like kinase in the root, the product of the TOO MUCH LOVE (TML) gene, an F-box protein, was revealed; its mutation results in a root-controlled supernodulating phenotype (Takahara et al., 2013).

The exact molecular nature of the shoot-derived signal which inhibits nodulation is still not well understood. It was shown that the Mtsunn mutant had increased amount of auxin transported from the shoot to the root. This indicates that the activation of CLV1-like kinase in the shoot leads to a reduction of auxin transport to developing nodules, thereby to a reduction in nodulation (van Noorden et al., 2006). The activation of LjIPT3 in the shoot phloem, downstream of CLV1-like kinase, suggests that the shoot-to-root transport of both auxin and cytokinin is targeted by CLV1-like kinase, so that in roots the auxin amount is reduced and the cytokinin level is increased to restrict subsequent nodule formation. Other hormones such as jasmonic acid (JA) have also been implicated in the shoot-to-root communication during AON (Kinkema and Gresshoff, 2008; Reid et al., 2012).

In our previous study we had showed that the orthologs of LjIPT1 and LjIPT3 genes in M. truncatula, Medtr1g110590 (MtIPT1) and Medtr1g072540 (MtIPT3) respectively, were also upregulated in developing nodules at 7–9 days after rhizobial inoculation (Azarakhsh et al., 2015). However, the detailed analysis of MtIPTs expression, both in the roots and the shoots, in response to nodulation has not been performed and the link between IPTs and AON has not been studied in M. truncatula so far. In this study, we estimated, both in the shoots and the roots, the expression levels of all MtIPTs identified in databases at different time points after rhizobial inoculation. Moreover, we analyzed the expression of MtIPTs in the sunn-3 mutant to identify MtIPTs potentially involved in the systemic control of nodulation in a SUNN-dependent manner.

In the entire Medicago genome Mt4.0v1, 23 sequences were annotated as isopentenyl transferases (IPT). Among them, according to the phylogenetic analysis, 15 sequences were clustered together as a separate group of sequences with a high degree of similarity, and two pairs of sequences from this group appeared to be exactly the same. This suggests that this group of IPTs sequences may represent new genes which appeared during evolution as a result of recent duplication events in the Medicago genome. However, according to the expression analysis, IPT genes from this group exhibited either no or very low level of expression in the Medicago tissues tested. Thus, a functional and more comprehensive phylogenetic analysis including IPT sequences from other legumes is required to elucidate the possible role of these evolutionary new IPT genes in legume development.

Cytokinins have been previously shown to regulate different aspects of nodulation, including rhizobial infection during which CKs act as negative regulators and nodule primordia development and nitrogen fixation for which CKs act as positive regulators (Tirichine et al., 2007; Held et al., 2014; Boivin et al., 2016; Jardinaud et al., 2016). In M. truncatula, accumulation of cytokinins (iP: isopentenyladenine, iPR: isopentenyladenosine, tZ: trans-zeatin) was first observed at 3 h after NF exposure and this occurs in MtCCaMK-dependent manner. (van Zeijl et al., 2015). In our work, we did not study IPTs expression at very early stages nodule development, i.e., few hours after rhizobial infection. However, van Zeijl et al. (2015) reported that Medtr2g022140 (MtIPT4 according to our nomenclature) induced at 3 h after Nod factor treatment, therefore this IPT gene may contribute to cytokinin accumulation at early stages after rhizobial inoculation. Moreover, according to LCM-RNA-seq data, MtIPT2 is activated in root epidermis 24 h after NF treatment (Jardinaud et al., 2016). Moreover, the expression of the KNOX3 gene encoding homeodomain-containing TF that activates IPT3 expression in developing nodules (Azarakhsh et al., 2015) was also induced in root epidermis according to LCM-RNA-seq data (Jardinaud et al., 2016). The early induction of cytokinin biosynthesis in root epidermis may contribute to its negative effect on rhizobial infection reported previously (Held et al., 2014, see Figure 7).

In pea, a two-stage increase in the content of CKs during nodule development was observed (Dolgikh et al., 2017). The accumulation of tZ and IP at 3 dpi can be associated with cortical cell division where CKs are known to be positive regulators, while the increase of tZ content at later stages suggests that CK also plays a role in subsequent stages of the nodule development and functioning (Dolgikh et al., 2017). The temporal expression dynamics of MtIPTs we observed are in line with this observation, suggesting that cytokinin biosynthesis also occurs at later stages of nodule development. According to our previously reported data, the expression of MtIPT1 and MtIPT3 increased in the inoculated roots reaching a maximum at 9 dpi (Azarakhsh et al., 2015). Here, we found that the expression of MtIPT2, MtIPT9, and MtIPT5 genes were also induced during the nodule development, and the expression maximum of these genes is observed at 9–12 dpi, when the nodule primordia are completely developed.

The nodulation phenotype of the Mtcre1 mutant defective in a cytokinin receptor suggested that besides having a crucial role in the early steps of nodule initiation, CKs regulate the subsequent stages of nodule development (Plet et al., 2011). The rare nodules which formed on Mtcre1-inoculated roots exhibit incomplete differentiation and have frequently multiple lobes, suggesting that a functional MtCRE1 may regulate the transition between meristematic and cell differentiation/elongation zones in the indeterminate nodules (Plet et al., 2011). Taking into account this finding, we suggest that IPTs expression that we observed after 7 dpi (at the stages when nodule differentiation occurs and the nodule meristem is being formed) may contribute to cytokinin involvement in the nodule differentiation and the meristem formation (see Figure 7).

Moreover, recently, it was shown that functional cytokinin receptor in M. truncatula is required for nitrogen fixation (Boivin et al., 2016). In cre1 mutant showing reduced and delayed nodule formation, nitrogen fixation was decreased. Notably, the AHK4/CRE1 gene from Arabidopsis was able to complement nodule initiation, but not nitrogen fixation in the cre1 mutant, indicating that legume-specific determinants encoded by MtCRE1 are likely required for nitrogen fixation activity (Boivin et al., 2016). This finding represents one more evidence that CKs act at later stages of nodule development and functioning.

In L. japonicus, it was shown that LjIPT3 gene is activated in the shoot in response to rhizobial inoculation, and its activation is dependent on the CLV1-like kinase LjHAR1 (Sasaki et al., 2014). This increase in LjIPT3 expression was observed at 3–5 dpi, which is consistent with the timing of AON induction determined previously for this legume in a split-root system (Suzuki et al., 2008). In our experiment we found that the expression of MtIPT3, the ortholog of LjIPT3, was activated in shoots at 7 dpi. The activation of MtIPT3 in the leaves of inoculated plants is consistent with the data from phytozome database⁴. MtIPT4 and MtIPT5 were also slightly induced in shoots after rhizobial inoculation. Interestingly, the shoot MtIPT3 activation in response to inoculation was absent in the sunn-3 mutant defective in the CLV1-like kinase MtSUNN. These findings indicate that the ortholog of LjIPT3 gene in M. truncatula is also activated in shoots in response to inoculation in a CLV1-like kinase-dependent manner, whereas the expression of MtIPT3 together with other MtIPT genes in roots occurs independently of CLV1-like kinase. The activation of MtIPT4 and MtIPT5, which was still observed in the sunn-3 mutant in response to inoculation is unlikely to be regulated by MtSUNN, but instead it must be controlled by other regulatory pathways. The timing of AON induction in Medicago was determined to be 3 days after inoculation (Kassaw et al., 2015). In our experiments, the statistically significant activation of MtIPTs in leaves was observed at 7 dpi, although in one of three biological repeats the increased expression of MtIPTs in leaves was found at 5 dpi after rhizobial inoculation (Supplementary Figure S5). We suppose that temporal dynamics of nodule development may vary depending on experimental plant growth conditions, and the timing of AON activation may vary accordingly. Moreover, the expression of MtCLE13, a CLE-peptide that triggers AON, was shown to be activated as early as 3 h after NF treatment at a time-point preceding nodule primordium initiation (van Zeijl et al., 2015). According to Mortier et al. (2010), MtCLE13 was significantly increased in inoculated roots at 4 dpi, and its expression level remained relatively high until 10 dpi, with a slightly pronounced maximum at 8 dpi. These findings allow us to propose that AON may comprise different levels of regulation, involving the perception of different signals. The shoot-derived signal (SDI) in AON also appeared to have a complex molecular nature. AON involves different hormones such as auxins, JA and cytokinins (van Noorden et al., 2006; Kinkema and Gresshoff, 2008; Sasaki et al., 2014), that may also inhibit nodulation at different levels via multiple targets. Among the later ones, the NIN transcription factor was shown to be targeted by AON in inoculated roots in a negative feed-back manner, where NIN also directly induces CLEs expression to trigger AON (Soyano et al., 2014). Together with CKs, NIN represents a crucial regulator of different aspects of nodulation that plays both positive and negative roles in nodulation. Along with the fact that NIN transcription factor is important both for rhizobial infection and nodule primodia development (Schauser et al., 1999; Marsh et al., 2007), NIN expression in the cortex was shown to inhibit rhizobial infection (Yoro et al., 2014). Moreover, NIN expression in the root cortex is activated by CK (Gonzalez-Rizzo et al., 2006; Plet et al., 2011), and NIN can promote CRE1 expression (Vernié et al., 2015) (see Figure 7).

A large body of evidence indicates that CKs play a complex role in nodulation, mediating negative feedback regulatory mechanisms (Mortier et al., 2014; Sasaki et al., 2014). The exact molecular mechanisms underlying the dual effects of CKs on nodulation are still far from being understood. It is likely that CKs acting in a tight crosstalk with other hormones, such as auxins, may exert different effects on nodulation depending on the auxin concentration that is known to fluctuate through different stages of nodulation. Future studies should reveal the molecular pathways regulated by CKs and other hormones to elucidate their complex action during nodulation.

MA performed all the experiments on gene expression analysis. MA and ML analyzed the data and drafted the manuscript. ML and LL planned the experiments and supervised the research. All authors read and approved the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.