The MtDMI2-MtPUB2 Negative Feedback Loop Plays a Role in Nodulation Homeostasis^,

MtPUB2 and MtDMI2 form a negative feedback loop that regulates nodulation homeostasis in the Medicago truncatula/Sinorhizobium meliloti symbiosis.

Abstract

DOES NOT MAKE INFECTION 2 (MtDMI2) is a Leu rich repeat-type receptor kinase required for signal transduction in the Medicago truncatula/Sinorhizobium meliloti symbiosis pathway. However, the mechanisms through which MtDMI2 participates in nodulation homeostasis are poorly understood. In this study, we identified MtPUB2—a novel plant U-box (PUB)–type E3 ligase—and showed that it interacts with MtDMI2. MtDMI2 and MtPUB2 accumulation were shown to be similar in various tissues. Roots of plants in which MtPUB2 was silenced by RNAi (MtPUB2-RNAi plants) exhibited impaired infection threads, fewer nodules, and shorter primary root lengths compared to those of control plants transformed with empty vector. Using liquid chromatography-tandem mass spectrometry, we showed that MtDMI2 phosphorylates MtPUB2 at Ser-316, Ser-421, and Thr-488 residues. When MtPUB2-RNAi plants were transformed with MtPUB2^S421D, which mimics the phosphorylated state, MtDMI2 was persistently ubiquitinated and degraded by MtPUB2^S421D, resulting in fewer nodules than observed in MtPUB2/MtPUB2-RNAi-complemented plants. However, MtPUB2^S421A/MtPUB2-RNAi-complemented plants showed no MtPUB2 ubiquitination activity, and their nodulation phenotype was similar to that of MtPUB2-RNAi plants transformed with empty vector. Further studies demonstrated that these proteins form a negative feedback loop of the prey (MtDMI2)-predator (MtPUB2) type. Our results suggest that the MtDMI2-MtPUB2 negative feedback loop, which displays crosstalk with the long-distance autoregulation of nodulation via MtNIN, plays an important role in nodulation homeostasis.

Legumes possess the capacity for symbiotic nitrogen fixation, in contrast to nonleguminous plants such as rice (Oryza sativa), maize (Zea mays), and wheat (Triticum aestivum; Vijn et al., 1993). Biological nitrogen fixation by legumes is estimated to have contributed approximately 90% of the 100 to 140 teragrams (Tg) of annual nitrogen fixation that occurred on earth prior to agricultural activity (Gage, 2004). Leguminous plants use specific organs, namely nodules, to accommodate symbiotic rhizobia and provide a compatible environment for nitrogen fixation (Desbrosses and Stougaard, 2011). The host plant provides rhizobia with photosynthetic products and controls nodulation during the symbiotic process (Magori et al., 2009; Desbrosses and Stougaard, 2011), and in conjunction with this, rhizobia supply the host with ammonia (Magori et al., 2009). Nodulation must be regulated, because nitrogen fixation consumes a large amount of energy (Ferguson et al., 2010; Mortier et al., 2012). Nitrogen and carbon metabolism and photosynthesis are also in homeostasis in plants (Huppe and Turpin, 1994). As noted above, nodulation homeostasis might be maintained through many mechanisms.

Previous studies have proposed a model for the autoregulation of nodulation (AON) involving root-to-shoot communication to maintain an optimal number of nodules by systemically reducing nodulation (Searle et al., 2003; Soyano et al., 2014). In another study, nod factor (NF)-induced ethylene production was shown to inhibit nodulation, suggesting a local equilibrium mechanism that controls nodule numbers (Oldroyd et al., 2001). It is very likely that more than one mechanism plays a critical role in maintaining this equilibrium. If such a mechanism exists, which molecule is the regulatory target?

The DOES NOT MAKE INFECTION 2 (MtDMI2)-MtDMI1-MtDMI3 signaling pathway plays a vital role in the Medicago truncatula/Sinorhizobium meliloti symbiotic process (Oláh et al., 2005). MtDMI1 (the ortholog of Lotus japonicus CASTOR and POLLUX) is an ion channel that controls calcium oscillations, which are required for symbiosis in legumes (Ané et al., 2004; Riely et al., 2007; Charpentier et al., 2008). MtDMI3 (the ortholog of L. japonicus CCaMK) is a Ca²⁺/calmodulin-dependent protein kinase that acts downstream of calcium spiking and is required for nodulation signal transduction (Lévy et al., 2004; Mitra et al., 2004). MtIPD3 (the ortholog of L. japonicus CYCLOPS) interacts with MtDMI3 and is also required for the symbiotic signaling pathway (Horváth et al., 2011). Recent studies have indicated that LjCYCLOPS is a phosphorylation substrate of LjCCaMK and that it can bind the promoter of the nodule inception gene (LjNIN) to initiate symbiotic root nodule development (Singh et al., 2014). MtDMI2 (AJ418369.1, also known as SYMBIOSIS RECEPTOR KINASE [SymRK] in L. japonicus) is a Leu-rich repeat-type receptor kinase with an extracellular domain containing three Leu-rich motifs, a malectin-like domain in the extracellular domain, and a Ser/Thr-type kinase domain in the intracellular domain (Endre et al., 2002; Antolín-Llovera et al., 2014). MtDMI2 acts downstream of NOD FACTOR PERCEPTION (MtNFP) and LYS MOTIF-CONTAINING RECEPTOR-LIKE KINASE 3 (MtLYK3; Endre et al., 2002; Antolín-Llovera et al., 2014).

Because of the importance of MtDMI2 in the rhizhobia and Arbuscular mycorrhiza symbiosis signaling pathway, previous studies have identified a number of its interacting proteins. MtPUB1, a plant U-box (PUB)-type E3 ligase, interacts with MtLYK3 and negatively regulates nodulation (Mbengue et al., 2010). A recent study suggests that MtDMI2 can interact with and phosphorylate MtPUB1 in the symbiosis pathway (Vernié et al., 2016). However, in these two cases, the targets of MtPUB1 and the molecular mechanisms remain unknown. MtHMGR1, a 3-hydroxy-3-methylglutaryl CoA reductase 1 in the mevalonate pathway, is involved in nodulation, interacts with MtDMI2, and is necessary for nodulation in M. truncatula (Kevei et al., 2007). Furthermore, research indicates that mevalonate is the direct product of MtHMGR1 and that it might be a novel second messenger in the nodulation pathway (Venkateshwaran et al., 2015). Similarly, LjSymRK-INTERACTING PROTEIN 2, a type of MAPKK from L. japonicus, interacts with LjSymRK to regulate nodule organogenesis (Chen et al., 2012). LjSymRK-interacting E3 ubiquitin ligase (LjSIE3) is a REALLY INTERESTING NEW GENE finger-type E3 ligase that can use LjSymRK as a substrate for ubiquitination. LjSIE3 is a positive regulator involved in nodulation. Moreover, SEVEN IN ABSENTIA 4 (LjSINA4) has been reported to be involved in the turnover of LjSymRK, and their interaction is spatiotemporally controlled in L. japonicus during nodulation (Den Herder et al., 2012); however, the exact mechanisms through which LjSIE3 and LjSINA4 regulate LjSymRK remain unclear (Den Herder et al., 2012; Yuan et al., 2012).

As mentioned above, although several proteins interact with MtDMI2 and LjSymRK, the molecular mechanisms contributing to nodulation homeostasis via phosphorylation and ubiquitination are poorly understood. To determine the direct regulator of MtDMI2, we used a yeast two-hybrid (Y2H) approach and identified a novel PUB-type E3 ubiquitin ligase that interacts with MtDMI2, which we named MtPUB2. Biochemical functional assays of MtPUB2 showed that it has E3 ubiquitin ligase activity, and the conserved amino acid Val-274 in the U-box domain is required for E3 activity. MtDMI2 activates MtPUB2 via phosphorylation at Ser-421, and activated MtPUB2 directly targets MtDMI2 for ubiquitination-mediated degradation. MtPUB2-RNAi roots demonstrate impaired infection threads, fewer nodules, shorter root length, and lower nodule densities. Biological and biochemical assays show that MtPUB2 and MtDMI2 exhibit similar tissue expression, and they each show a wave-shaped curve during nodulation. We adopted the numerical solutions of Lotka-Volterra equations (LV equations; Lotka, 1910) to describe the protein levels of MtDMI2 and MtPUB2 during nodulation from 24 h after inoculation to 21 d after inoculation (DAI). Thus, the purpose of the negative feedback loop between MtDMI2 and MtPUB2 is to achieve a stable root nodulation system.

RESULTS

MtPUB2 Is a New PUB-Type E3 Ligase That Is Homologous to AtPUB13 and OsSPL11

To identify proteins that interact with MtDMI2, we performed Y2H screening using a cDNA library. The roots of M. truncatula genotype cv Jemalog A17 plants were treated with S. meliloti strain 1021 at different time points. MtDMI2IR (intracellular region, amino acids 543–919) was cloned into pGBKT7 and used as the bait. The screening data also included previously identified MtHMGR1 (Kevei et al., 2007; Supplemental Fig. S1). We then selected MtPUB2 (KU285617.1) for further testing. Full-length MtPUB2 DNA contains 4,531 bp and includes four exons (Supplemental Fig. S2A). Full-length MtPUB2 cDNA contains an open reading frame of 1,989 nucleotides that encodes a polypeptide of 662 amino acids. MtPUB2 has the structure of a UND-PUB-ARM protein and contains one U-box domain followed by a region containing six ARMADILLO (ARM) repeats (Supplemental Fig. S2B). Protein domains were identified with the SMART (http://smart.embl-heidelberg.de/) database. A phylogenetic analysis (MEGA4.0) showed that the closest homologs of MtPUB2 are AtPUB13 (AT3G46510; Lu et al., 2011) and OsSPL11 (AAT94161; Zeng et al., 2004), which are distantly related to MtPUB1 (DAA33939) (Supplemental Fig. S3). The protein sequence of MtPUB2 shares 66.77% identity with AtPUB13 and 62% identity with AtPUB12 and OsSPL11.

We selected two additional PUB-type proteins, namely Medtr1g094021.1 and Medtr4g028960.1, to determine whether they would interact with MtDMI2IR. They did not interact with MtDMI2IR in yeast cells (Supplemental Fig. S4).

MtPUB2 is associated with the plasma membrane in onion (Allium cepa) epidermal peels (Fig. 1A). This result was further confirmed with a cell fractionation assay using specific antibodies against H⁺-ATPase and cFBPase (cytosolic Fru-1,6-bisphosphatase) to validate the membrane and cytosolic fractions, respectively (Fig. 1C). MtDMI2 was also localized to the cell plasma membrane in M. truncatula (Riely et al., 2013). This finding suggested that MtPUB2 and MtDMI2 localize to the same subcellular compartment.

Figure 1.

MtPUB2 is a membrane-associated protein and a novel PUB-type E3 ubiquitin ligase. A, Subcellular localization analysis of GFP in onion epidermal peels following particle gun-mediated transformation with Pro-35S:MtPUB2-cGFP and the vector alone (control). The peels were imaged by epifluorescence using identical exposure settings. Plasmolysis occurred 5 min after treatment with 4% NaCl solution. The scale bars indicate 60 μm. B (right), In vitro ubiquitination assays with a MBP-MtPUB2 and MBP-MtPUB2^V274R in the presence of E1 (UBE1), E2 (UbcH5c), and Arabidopsis ubiquitin. Multiple HMW bands indicate the poly-ubiquitination of MBP-MtPUB2 in the presence of E1 and E2 enzymes, ubiquitin, and ATP. Reactions with various components omitted (-) and MBP-MtPUB2^V274R were used as controls. The MBP-MtPUB2 protein is approximately 112 kD. The MBP tag is approximately 40 kD. Both anti-MBP and anti-Ubi antibodies were used on individual immunoblots (top and bottom, respectively). M, represents protein marker. C (bottom left), Cellular fractionation assays of transient transgenic N. benthamiana leaves expressing MtPUB2-HA. Anti-cFBPase and anti-H⁺ATPase (pm) were used as cytoplasm and membrane markers, respectively.

To analyze the function of MtPUB2, an in vitro ubiquitination assay was performed (Fig. 1B). Val-272 of OsSPL11 and Val-273 of AtPUB13 are the key amino acids for maintaining the ubiquitination ability of these E3 ligases. This residue is highly conserved in different U-box proteins and was demonstrated to be vital for the biological and biochemical functions of U-box proteins (Zeng et al., 2004; Liu et al., 2012). Therefore, Val-274 in MtPUB2 was substituted with Arg to further study its role in ubiquitination. MtPUB2 and MtPUB2^V274R were both tagged with a maltose-binding protein (MBP-MtPUB2 and MBP-MtPUB2^V274R), expressed in Escherichia coli strain BL21, and purified using amylose resin. Mutant MBP-MtPUB2^V274R did not self-ubiquitinate (Fig. 1B, lane 1). MtPUB2 was converted to a mixture of high-M_r ubiquitinated protein products in the presence of E1 (Ub-activating), E2 (Ub-conjugating), ubiquitin, and ATP, suggesting that MtPUB2 is capable of self-ubiquitination (Fig. 1B, lane 2). The high-molecular-weight (HMW) bands were not observed in reactions lacking ubiquitin, E1, E2, or MtPUB2 (Fig. 1B, lanes 3–6). Antibodies recognizing MBP and ubiquitin were both used in immunoblot analysis (Fig. 1B). However, there were some unknown proteins, recognized by the anti-Ubi antibody in lanes 1, 3, 4, 5, and 6. These bands may be E1 or E2 with ubiquitin (Fig. 1B, anti-Ubi; Lu et al., 2011).

In summary, these results indicate that MtPUB2, a homolog of AtPUB13/PUB12 and OsSPL11, is a novel PUB-type E3 ligase in M. truncatula.

MtPUB2 Affects the Number of Infection Threads and Nodules

Mtpub2 mutants were not available in the fast-neutron (Rogers et al., 2009) or transposable element of tobacco (Nicotiana tabacum) cell type 1 insertion collection (Noble Foundation; Cheng et al., 2011). Therefore, to functionally characterize MtPUB2, we created Agrobacterium rhizogenes-mediated RNAi hairy root cultures and stable Agrobacterium tumefaciens-mediated RNAi lines.

The A. rhizogenes-mediated RNAi plants were planted on square petri dishes (130 × 130 mm) with Fåhraeus medium lacking nitrogen (pH 6.5) and inoculated with S. meliloti strain 1021 after 7 d of growth. The RNAi efficiency in A. rhizogenes-mediated hairy root cultures was approximately 72.8% in the test transgenic samples (Fig. 2B). Infection threads (ITs) were imaged after staining of hairy root cultures with 5-bromo-4-chloro-3-indolyl β-d-galactopyranoside acid (X-gal) and counted in the control (empty vector [EV], GUS-RNAi) and MtPUB2-RNAi lines at 5 DAI. In both GUS-RNAi and MtPUB2-RNAi cultures, two types of ITs exist: ITs in the outer cortical cells and ITs in the inner cortex cells (Fig. 2C; Supplemental Fig. S5). However, the average total number of ITs was decreased approximately 2- to 4-fold in MtPUB2-RNAi root cultures compared to that in control root cultures (Fig. 2D).

Figure 2.

Inhibition of nodulation and infection threads by knock-down of MtPUB2 in A. rhizogenes-mediated MtPUB2-RNAi hairy root cultures. A, Nodulation in GUS-RNAi and MtPUB2-RNAi hairy root cultures generated by A. rhizogenes-mediated transformation at 21 DAI. The black arrows indicate enlarged photos of the roots and nodules. B, The MtPUB2-RNAi construct efficiently down-regulates MtPUB2 gene expression. MtPUB2 mRNA quantification was normalized against two internal reference genes (Mtactin and MtEF1α) in each sample. The SDs between samples are shown as error bars. C, ITs in GUS-RNAi roots (top) and MtPUB2-RNAi roots (bottom) at 5 DAI imaged by staining with X-gal. ITs are labeled by black arrows. The scale bars indicate 100 μm. D, Quantification of ITs depicted in C. Shown are mean values ± se calculated from the three independent experiments (** indicates significant differences between GUS-RNAi and MtPUB2-RNAi, P ≤ 0.01, Student’s t test, n = 20, three repeats). E, Quantification of nodulation depicted in A. Nodule numbers were counted in GUS-RNAi and MtPUB2-RNAi hairy root cultures at 21 DAI. Depicted are mean values ± se calculated from the three independent experiments (** indicates significant differences between GUS-RNAi and MtPUB2-RNAi, P ≤ 0.01, Student’s t test, n = 20, three repeats).

We scored the average nodule numbers in GUS-RNAi and MtPUB2-RNAi hairy root cultures at 21 DAI. The average number of nodules in the MtPUB2-RNAi lines was decreased compared with the control GUS-RNAi lines. Approximately 83% fewer nodules formed on average in the MtPUB2-RNAi root cultures than in control root cultures (Fig. 2, A and E).

Taken together, our data indicate that MtPUB2 participates in the development of ITs and nodules. Previous studies revealed that MtDMI2 knock-down blocks the release of rhizobia in the nodule, and the ITs of dmi2-1 plants are aberrant (Limpens et al., 2005). Therefore, MtPUB2 and MtDMI2 both participate in the development of ITs and nodulation formation.

MtPUB2 Affects Primary Root Length, Nodule Densities, and Nodule Development in M. truncatula

We analyzed the nodulation phenotypes of stable-transgenic MtPUB2-RNAi lines to further verify the biological function of MtPUB2. The RNAi efficiencies in stable transgenic MtPUB2-RNAi lines (T₃ generation) were 0.60, 0.65, and 0.70 in lines 32, 115, and 119, respectively (Fig. 3E). We also tested the relative expression of the other PUB gene, Medtr1g094025.1, in stable-transgenic MtPUB2-RNAi lines; the relative expression of this gene was not reduced in the roots of MtPUB2-RNAi plants compared to that in control plants (Supplemental Fig. S6). This result indicates the efficient and specific down-regulation of MtPUB2.

Figure 3.

Inhibition of nodulation, root length, and development by knock-down of MtPUB2 in A. tumefaciens-mediated MtPUB2-RNAi and EV stable-transformed lines. A, Whole plants of EV-transformed lines and MtPUB2-RNAi lines (lines 115 and 119) at 21 DAI. Left, EV-transformed lines; right, MtPUB2-RNAi lines 115 and 119. B, Close-up stereomicroscope view of the roots and nodules at 21 and 42 DAI. Bar = 5 mm. C, The development of 49-d-old plants of MtPUB2-RNAi lines and EV-transformed lines in soil. D, Lugol staining of the nodules of EV-transformed lines and MtPUB2-RNAi line 119 at 21 and 42 DAI. Bar = 200 μm. E, RT-qPCR analysis of relative MtPUB2 expression in the stable MtPUB2-RNAi lines 32, 115, and 119 (T₃ generation). Shown are mean values ± se calculated from three independent experiments (** indicates significant difference between EV-transformed lines and MtPUB2-RNAi lines, P ≤ 0.01, Student’s t test). F, Nodule density (No. cm⁻¹ root length) assay in EV-transformed lines and MtPUB2-RNAi lines (lines 32, 115, and 119). Shown are mean values ± se calculated from three independent experiments (** indicates significant difference between EV-transformed lines and MtPUB2-RNAi lines, P ≤ 0.01, Student’s t test, n = 10, three repeats). G, Root length (cm) of 4-week-old plants of EV-transformed lines and MtPUB2-RNAi lines (lines 32, 115, and 119). Shown are mean values ± se calculated from three independent experiments (** indicates significant difference between EV-transformed lines and MtPUB2-RNAi lines, P ≤ 0.01, Student’s t test, n = 10, three repeats). H, The number of nodules on each of the transgenic root systems of each plant transformed with the EV or pANDA-MtPUB2 (lines 32, 115, and 119, n = 15, three biological repeats/line). The x axis represents the nodule numbers, while the y axis represents the number of plants.

An RNAi efficiency of 0.95 caused severe reactions, such as delayed plant growth, sterility, and death. For the nodulation phenotype assays, the seeds of EV-transformed plants and stable-transgenic MtPUB2-RNAi plants (32, 115, and 119) were planted on Fåhraeus medium without nitrogen (pH 6.5) and inoculated with S. meliloti strain 1021 after 7 d of growth. The nodule number, primary root length (cm), and nodule density (No. cm⁻¹) were measured in the EV-transformed plants and stable-transgenic MtPUB2-RNAi plants at 21 DAI.

An analysis of the distribution of nodule numbers per plant indicated that approximately 63.3% of MtPUB2-RNAi (line 32, line 115, and line 119) roots produced zero or one nodule, whereas 20% of roots showed two nodules and 13.3% of roots had three nodules (n = 15, three biological repeats/line; Fig. 3, A and H; Qiu et al., 2015). All three stable-transgenic MtPUB2-RNAi plants exhibited retarded growth (Fig. 3, A and C). Conversely, 100% of the EV-transformed plants developed 5 to 10 nodules (Fig. 3, A and H). Moreover, the MtPUB2-RNAi plants exhibited a significant reduction in primary root lengths (5.8–7 cm, Fig. 3, A and G) and nodule densities (0.5–0.8 cm⁻¹ No. root length; Fig. 3, A and F) compared with EV-transformed plants (7.8–8 cm root length and 1.5 No. cm⁻¹ root length; Fig. 3, A, G, and F). These results demonstrated that the reduced nodule numbers were not caused by the short root length. As shown in Figure 3, B and D, the nodules on EV-transformed plants and stable-transgenic MtPUB2-RNAi plants were both pink at the time of harvest (21 DAI). At 21 DAI, although all appeared pink in color, the maturation of nodules was different (Fig. 3B). We further investigated the role of MtPUB2 in nodule senescence. Lugol staining is used to indicate the presence of amyloplasts, the accumulation of which is linked to premature nodule senescence as shown in several studies using bacteria defective in nitrogen fixation (Lodwig et al., 2003; Harrison et al., 2005). Lugol staining was performed in 21-DAI and 42-DAI nodules. At 21 DAI, the nodules of EV-transformed plants were mature (Fig. 3D); however, the nodules of MtPUB2-RNAi plants were immature (Fig. 3D). At 42 DAI, more senescent cells were observed in the nodules of EV-transformed plants than in those of MtPUB2-RNAi plants (Fig. 3D). In summary, the down-regulation of MtPUB2 can induce a significant delay in nodule development.

Stable-overexpression transgenic plants were also created using a pCAMBIA1302-MtPUB2 vector alongside control plants transformed with pCAMBIA1302 as the EV. The plants were also inoculated with S. meliloti strain 1021 after 7 d of growth on square petri dishes (130 × 130 mm). The nodule numbers were counted 21 DAI. However, there was no difference in nodule number or primary root length (cm) between the MtPUB2-overexpressing plants and the EV-transformed plants (Supplemental Fig. S7, A and C). Reverse transcription quantitative-PCR (RT-qPCR) was used to measure the relative expression of MtPUB2 in the overexpressing lines (Supplemental Fig. S7B).

Thus, MtPUB2 not only is involved in the development of plants, but also participates in the formation and development of nodules.

MtPUB2 Shows Similar Tissue Expression Patterns as MtDMI2

To assess the tissue expression pattern of MtPUB2, a 2.3-kb promoter region of MtPUB2 was cloned and inserted into the pkGWFS7 vector (Huault et al., 2014) carrying a downstream GUS fusion construct. Stable M. truncatula transformants were prepared. In noninoculated plants, proMtPUB2-GUS expression was observed in the leaves, stems, and roots (Fig. 4, A–C) and particularly in the cortex, endodermis, pericycle, and phloem of the roots (Fig. 4, C and G). proMtPUB2-GUS was expressed in the primordial of nodules at 5 DAI with S. meliloti strain 1021 (Fig. 4F). At 21 DAI with S. meliloti strain 1021, proMtPUB2-GUS was expressed in different parts of the nodules, such as the distal, apical, and persistent meristem (zone I) and in zones of increasing cell age, comprising an infection zone (zone II), an interzone (zone II-III), and a nitrogen-fixation zone (zone III; Fig. 4, D and H). RT-qPCR assays confirmed the expression of MtPUB2 in the roots, stems, leaves, and nodules (Fig. 4E). The results showed that MtPUB2 is primarily expressed in the roots and nodules (Fig. 4, A–E).

Figure 4.

Histochemical localization of proMtPUB2-GUS in M. truncatula. The expression pattern of MtPUB2 was tested using its 2.3-kb promoter region. M. truncatula plants transformed by A. tumefaciens were used for proMtPUB2-GUS assays during nodulation using X-gluc (blue in A—H). A to C, Expression of proMtPUB2-GUS in the leaves, stems, and roots, respectively, of 4-week-old M. truncatula plants. The scale bars indicate 200 μm in A and B, and 100 μm in C. D, Expression of proMtPUB2-GUS in nodules of M. truncatula roots at 21 DAI. The scale bar indicates 200 μm. E, RT-qPCR analysis of relative MtPUB2 gene expression in the root, stem, leaves, and nodules of 4-week-old wild-type R108 plants (internal reference genes were MtEF1α and Mtactin). Error bars in this panel represent the SDs calculated from 15 samples. F, Expression of proMtPUB2-GUS in the primordia of nodules at 5 DAI (black arrows indicate the primordia of nodules). The scale bar indicates 200 μm. G, Expression of proMtPUB2-GUS in the cortex, inner cortex, pericycle, and the phloem of the roots. The scale bar indicates 100 μm. H, Expression of proMtPUB2-GUS in the longitudinal section of 21-d-old nodules (section, 80 μm thick). I (Zone I, a distal, apical, and persistence meristem followed by zones of increasing cell age), II (Zone II, comprising an infection zone), II to III (Zone II–III, an interzone), III (Zone III, a nitrogen-fixation zone). The scale bar indicates 200 μm. All images were taken under a light microscope.

The expression pattern of MtDMI2 has been studied previously (Bersoult et al., 2005), and is similar to that of MtPUB2 in the roots and nodules of M. truncatula.

MtPUB2 and MtDMI2 Interact in Vitro and in Planta

To confirm that MtPUB2 and MtDMI2 interact, a Y2H assay, a glutathione S-transferase (GST) pull-down assay, and a bimolecular fluorescence complementation (BiFC) assay were performed. The results of the Y2H assay showed that in yeast, MtPUB2 specifically interacts with MtDMI2. COMPACT ROOT ARCHITECTURE 2 (MtCRA2) and MtLYK3, which do not interact with MtPUB2 in yeast (Fig. 5A), were used as controls. The truncated domain of MtPUB2 was cloned into pGBKT7, and the sequence encoding the protein kinase domain of MtDMI2 was cloned into pGADT7. In addition, this interaction occurred in the ARM repeats domain but not in the U-box domain of MtPUB2 (Supplemental Fig. S2C). The interaction was also tested through a GST pull-down assay. Purified His-MtPUB2 and GST-MtDMI2IR proteins were employed in GST pull-down assays (Fig. 5B). The 76-kD His-MtPUB2 protein was found to associate with the 71-kD GST-MtDMI2IR protein, whereas the control GST protein (27 kD) did not display a positive interaction.

Figure 5.

Interaction assays between MtDMI2 and MtPUB2. A, Y2H assays of MtPUB2, MtDMI2IR, MtLYK3, and MtCRA2 (intracellular region) in transformed Saccharomyces cerevisiae AH109 cells grown on SD/-Leu/-Trp and SD/-Ade/-His/-Leu/-Trp medium. pGBKT7-p53/pGADT7-RecT and pGBKT7/pGADT7 were used as positive and negative controls, respectively. An α-Gal assay was performed on SD/-Ade/-His/-Leu/-Trp medium. B, GST-pull down assay of MtPUB2 and MtDMI2IR. His-MtPUB2 was pulled down by GST-MtDMI2IR protein but not by the GST protein alone. The GST-MtDMI2IR fusion protein is approximately 71 kD. The GST tag is 27 kD. The His-MtPUB2 fusion protein is approximately 76 kD. The molecular weights of these proteins are indicated on the left of the blot. C, A BiFC analysis was performed by coexpressing split YFP pair combinations in onion epidermal peels. The C-terminal domain of YFP was fused to the C terminus of MtDMI2 (MtDMI2-Yc), and the N-terminal domain of YFP was fused to the N terminus of MtPUB2 (Yn-MtPUB2). YFP fluorescence (green) was imaged by epi-fluorescence using identical exposure settings. The scale bar indicates 60 μm.

Furthermore, a BiFC assay was performed by coexpressing the split yellow fluorescent protein (YFP) pair combinations in onion epidermal peels. The C-terminal domain of YFP was fused to the C terminus of MtDMI2 (MtDMI2-Yc), whereas the N-terminal domain of YFP was fused to the N terminus of MtPUB2 (Yn-MtPUB2). YFP fluorescence (green) was detected in onion epidermal cells 16 h after excitation in the transient BiFC assay. The results showed interaction at the plasma membrane of the cells (Fig. 5C).

In summary, these results show that MtDMI2 interacts with MtPUB2 in vitro and in planta.

MtPUB2 Mediates MtDMI2 Ubiquitination and Its Associated 26S Proteasome-Dependent Degradation in Vitro and in Planta

MtPUB2 has E3 ubiquitin ligase activity and interacts with MtDMI2. Thus, we performed in vitro ubiquitination assays and in vivo degradation assays to test whether MtDMI2 is the ubiquitination substrate of MtPUB2. In vitro ubiquitination assays revealed that MtPUB2 ubiquitinates MtDMI2 (Fig. 6A, lane 4). MtPUB2^V274R did not ubiquitinate MtDMI2 in vitro (Fig. 6A, lane 2). This result demonstrated that Val-274 is a key amino acid for E3 ubiquitin ligase activity, as observed in the self-ubiquitination assays. The HMW bands were not observed in reactions lacking ubiquitin, E1, E2, or MtPUB2. GST protein was also used as a negative control. Next, we examined these samples by immunoblot analysis using antibodies that recognize GST and ubiquitin. However, there were some unknown proteins recognized by the anti-Ubi antibody in lanes 1, 2, 6, 7, 8, and 9. This result might be caused by E1 or E2 with associated ubiquitin (Fig. 6A).

Figure 6.

MtPUB2 targets MtDMI2 for ubiquitination and degradation via the 26S proteasome pathway. A, In vitro ubiquitination of MtDMI2 by MtPUB2. The ubiquitination of GST-MtDMI2IR (MW = 71 kD) by MBP-MtPUB2 (MW = 112 kD) was detected by anti-GST and anti-Ubi antibodies in the presence of E1 (UBE1), E2 (UbcH5c), and Arabidopsis ubiquitin. MBP-MtPUB2^V274R, MBP, and GST proteins were used as negative controls. B, Degradation of MtDMI2 promoted by MtPUB2 in N. benthamiana leaves. Either MtPUB2-HA or Myc-MtDMI2 was coexpressed with GFP-HA by agro-infiltration in N. benthamiana. Tissues were harvested 3 d after infiltration. MG132 (50 μm) was added to the corresponding protein mixture samples to prevent protein degradation by the 26S proteasome. Estimates of MtDMI2 protein levels relative to the GFP-HA protein are shown at the bottom. GFP-HA was expressed as an internal control. Expressed target genes and NbEF1α mRNA expression levels were analyzed by RT-PCR (bottom). The Myc-MtDMI2 protein is 111.8 kD. The MtPUB2-HA protein is 84 kD. The 6× Myc tag is approximately 7.9 kD. The 3× HA tag is 12 kD. C and D, MG132 was either added (C) or not added (D) to the corresponding protein mixture samples. Ponceau S staining (bottom) of the Rubisco protein is shown as a loading control. Reactions with various components omitted (−) were used as controls.

We then co-infiltrated agrobacterium host constructs expressing MtPUB2-HA and Myc-MtDMI2 with controls into the same leaf areas of N. benthamiana to test in vivo degradation. Samples were collected to detect the protein and RNA levels of the constructs. In this experiment, GFP-HA was used as an internal control to determine whether equal amounts of MtPUB2 were expressed in the different co-infiltrations. Quantitative immunoblot analysis showed that MtDMI2 was degraded by the addition of MtPUB2 (Fig. 6B, lane 1), but this degradation was prevented by the addition of (N-[(phenylmethoxy)carbonyl]-l-leucyl-N-[(1S)-1-formyl-3-methylbutyl]-l-leucinamide (MG132; 26S proteasome inhibitor; Fig. 6B, lane 2). The mutant MtPUB2^V274R-HA did not mediate the degradation of MtDMI2 in the presence or absence of MG132 (Fig. 6B, lanes 3 and 4). RT-qPCR was used to further verify that this degradation occurred at the protein level. NbEF1α was used as an internal reference gene. There were no obvious changes in MtDMI2 or NbEF1α transcript levels. These results showed that MtPUB2 mediates the degradation of MtDMI2.

To determine whether the degradation of MtDMI2 involved ubiquitination, we performed semi-in vivo degradation assays. Total proteins were extracted from N. benthamiana expressing Myc-MtDMI2 or MtPUB2-HA. The protein extracts were then mixed at 4°C for the indicated times (0–6 h). These samples were probed with anti-HA and anti-Myc antibodies. In this experiment, GFP-HA was used as an internal control. To test whether this degradation was mediated by the 26S proteasome instead of another pathway, dimethyl sulfoxide and MG132 were added to the samples. The results show that the degradation of MtDMI2 was enhanced over time (0–6 h; Fig. 6C), this degradation was prevented by the addition of MG132 (0–6 h, Fig. 6D).

In summary, MtPUB2 mediates the ATP-dependent degradation of MtDMI2 through ubiquitination and the 26S proteasome pathway.

MtDMI2 Phosphorylates MtPUB2 in Vitro and in Vivo

Because MtDMI2 has a protein kinase domain (amino acids 595 to 867), we investigated whether MtDMI2 kinase could phosphorylate MtPUB2. As expected, strong MtDMI2IR auto-phosphorylation was observed (Fig. 7A, lane 4). In the presence of GST-MtDMI2IR, MBP-MtPUB2 was phosphorylated (Fig. 7A, lane 3), whereas MBP protein alone showed weak phosphorylation (Fig. 7A, lane 8). Alignment of the MtDMI2 intracellular domain and LjSYMRK revealed that Thr-762 in MtDMI2 is equivalent to Thr-760 in LjSYMRK (Yoshida, 2005). In vitro phosphorylation assays indicated that GST-MtDMI2IR^T762A did not undergo auto-phosphorylation unlike GST-MtDMI2IR (Fig. 7B).

Figure 7.

Phosphorylation assays with MtDMI2 and MtPUB2 in vitro and in M. truncatula. A, Phosphorylation of MtPUB2 by activated MtDMI2 in vitro. Purified GST-MtDMI2IR was incubated with MtPUB2-MBP in the presence of [^γ-32P]ATP and kinase reaction products were resolved on SDS-PAGE gel. The star indicates the phosphorylation of MtPUB2 by MtDMI2. Reactions with various components omitted (−) were used as controls. The CBB showed the SDS-PAGE of GST-MtDMI2IR and MBP-MtPUB2 proteins (8% SDS-PAGE gel). The red arrows represent MBP-MtPUB2, GST-MtDMI2IR, and GST proteins. M, protein marker. B, Auto-phosphorylation of MtDMI2 in vitro. Purified GST-MtDMI2IR and GST-MtDMI2IR^T762A were incubated in the presence of [^γ-32P]ATP, and kinase reaction products were resolved on SDS-PAGE gel (12%). C, Phos-tag assays with MtPUB2. Proteins were harvested from the roots of wild-type A17 roots inoculated with S. meliloti strain 1021 at 21 DAI. Samples were separated in a phos-tag SDS-PAGE gel. β-Actin was used as a loading control. Reactions with various components omitted (−) were used as controls. For CIAP assays, the proteins were treated with CIAP at 37°C for 30 min and then detected through immunoblot analysis with anti-MtPUB2 polyclonal antibodies (BPI Co) and the antiactin antibodies (CWBio). D, Phos-tag mobility shift detection of phosphorylated MtPUB2 in wild-type A17 roots and dmi2-1 (TR25) roots after inoculation with S. meliloti strain 1021 at 7 DAI, 14 DAI, and 21 DAI. Protein extracts from the roots treated with S. meliloti strain 1021 for different times were separated in an SDS-PAGE gel with phos-tag reagent. β-Actin was used as a loading control. P in B, C, and D represents the phosphorylated bands, while np represents the nonphosphorylated bands.

A phos-tag reagent (BOPPARD) was used in a phospho-protein mobility shift assay to detect the phosphorylation of MtPUB2 in vivo. Proteins were acquired from the roots and nodules of wild-type roots (M. truncatula cv Jemalong A17) treated with S. meliloti strain 1021 at 7, 14, and 21 DAI. Immunoblot analysis showed that MtPUB2 is phosphorylated by MtDMI2 at the three time points in vivo (Fig. 7, C and D). At 21 DAI, the root proteins were treated with calf intestinal alkaline phosphatise (CIAP), and the results showed that the phosphorylation bands of MtPUB2 disappeared (Fig. 7C).

To explore whether MtPUB2 can be specifically phosphorylated by MtDMI2 in vivo, we subsequently performed phos-tag assays using nodulated 7-DAI, 14-DAI, and 21-DAI wild-type A17 and dmi2-1 (TR25) roots. The results showed that the smear of MtPUB2 phosphorylated bands was greatly diminished in samples prepared from dmi2-1 (TR25) roots compared with that from wild-type A17 roots, and the MtPUB2 bands were also diminished (Fig. 7D). Moreover, immunoblot analysis revealed more phosphorylated forms in the nodulated wild-type A17 roots at 7 DAI than at 14 DAI or 21 DAI (Fig. 7D). These results clearly suggest that MtDMI2 has strong auto-phosphorylation activity and that it can phosphorylate MtPUB2 in vitro and in vivo.

In summary, MtPUB2 can be phosphorylated by MtDMI2 in vivo, and the stability of MtPUB2 depends on its phosphorylation status.

MtPUB2 Can Be Phosphorylated by MtDMI2 at Ser-421, Which Appears to Enhance the E3 Ubiquitin Ligase Activity of MtPUB2

We performed a liquid chromatography-tandem mass spectrometry (LC-MS/MS, SCIEX) analysis to identify which amino acid of MtPUB2 is phosphorylated by MtDMI2. Purified GST-MtDMI2IR and His-MtPUB2 proteins were used in the in vitro phosphorylation reaction (three biological repeats), and the phosphorylated MtPUB2 was used for LC-MS/MS. The results suggest that Ser-316, Ser-421, and Thr-488 are the potential MtPUB2 phosphorylation sites (Supplemental Fig. S8, A–C). These sites were identified in the three biological replicates, and we found that the prediction of phosphorylation sites passed a 99% confidence level. Furthermore, Ser-421 and Thr-488 were located in the ARM-repeats domain, and Ser-316 was in the U-box domain of MtPUB2. We subsequently only focused on these three amino acids.

We then performed the in vitro ubiquitination assays to investigate which phosphorylation site is most important. Unphosphorylated state protein mimics (MtPUB2^S421A, MtPUB2^T488A, and MtPUB2^S316A) and phosphorylated state protein mimics (MtPUB2^S421D, MtPUB2^T488D, and MtPUB2^S316D) were purified in E. coli for in vitro ubiquitination analysis. An immunoblot analysis indicated that the self-ubiquitination activity of MtPUB2^S421A was significantly reduced, whereas MtPUB2^S421D exhibited enhanced activity compared with MtPUB2 (Fig. 8A). The MBP-MtPUB2^S421D protein was also tested for ubiquitination in the presence or absence of E1, E2, and ubiquitin (Supplemental Fig. S9).

Figure 8.

Identification of the phosphorylation sites of MtPUB2 and in vitro ubiquitin assays. A, In vitro ubiquitination assays with MBP-tagged MtPUB2 (MBP-MtPUB2), MBP-MtPUB2^S421A, MBP-MtPUB2^S421D, MBP-MtPUB2^S316A, MBP-MtPUB2^S316D, MBP-MtPUB2^T488A, and MBP-MtPUB2^T488D in the presence of E1 (UBE1), E2 (UbcH5c), and Arabidopsis ubiquitin. Multiple HMW bands indicated the poly-ubiquitination of MBP-MtPUB2. The MBP-MtPUB2 protein is approximately 112 kD. The MBP tag is approximately 40 kD. Immunoblots were incubated with anti-MBP (top) and anti-Ubi (bottom). B, In vitro ubiquitination of MtDMI2 by MtPUB2. The assays were performed with MBP-MtPUB2, MBP-MtPUB2^S421A, MBP-MtPUB2^S421D, MBP-MtPUB2^T488A, and MBP-MtPUB2^T488D. Poly-ubiquitination was detected with anti-Ubi antibody; anti-GST and anti-MBP antibodies were also used in immunoblot analysis. The red boxes labels in A and B represent the in vitro ubiquitination assays with MBP-MtPUB2^S421A and MBP-MtPUB2^S421D. C, In vitro assays for the ubiquitination of MtDMI2 by MtPUB2^S316A and MtPUB2^S316D. The ubiquitination of GST-MtDMI2IR (MW = 71 kD) by MBP-MtPUB2, MBP-MtPUB2^S316A, and MBP-MtPUB2^S316D (MW = 112 kD) were detected using anti-MBP, anti-GST, and anti-Ubi antibodies. The red arrows represent MBP-MtPUB2 and GST-MtDMI2IR, while the red boxes labels represent the in vitro ubiquitination assays with MBP-MtPUB2^S316A and MBP-MtPUB2^S316D.

In contrast, the self-ubiquitination activity of MtPUB2^S316A was comparable to that of MtPUB2, and MtPUB2^S316D exhibited no self-ubiquitination activity (Fig. 8A). However, MtPUB2^T488A and MtPUB2^T488D displayed activity comparable to that of MtPUB2. We then performed the in vitro ubiquitination assays using MtDMI2 as the substrate. MtDMI2IR ubiquitination by MtPUB2^S421A was greatly reduced compared with ubiquitination by MtPUB2^S421D or MtPUB2. These results reveal that Ser-421, and not the Thr-488 site in the ARM-repeat domain, is the key phosphorylation site of MtPUB2 required for its activation (Fig. 8, A and B). We also performed in vitro ubiquitination assays with MtDMI2IR, MtPUB2^S316A, and MtPUB2^S316D. The results showed that MtPUB2^S316D could not poly-ubiquitinate MtDMI2, whereas MtPUB2^S316A could use MtDMI2 as a poly-ubiquitinated substrate (Fig. 8C). The results indicate that the phosphorylation of Ser-316 inactivates the capability of E3 ubiquitin ligase; however, it is activated by the phosphorylation of Ser-421.

Taking into account the involvement of MtPUB2 and MtDMI2 proteins in nodulation, we decided to generate the complemented plants and analyze their nodulation phenotypes. This study centers upon the biological function of Ser-421 during nodulation.

Ser-421 of MtPUB2 Plays an Important Role in Nodulation

We tested the biological functions of MtPUB2^S421A and MtPUB2^S421D during nodulation. MtPUB2^S421A and MtPUB2^S421D were overexpressed in stable-transgenic T₃ generation MtPUB2-RNAi lines 32 and 119, respectively, via the hairy-root transformation method. pCAMBIA2301-modified overexpression transformants were used as EV controls.

Analysis of the distribution of nodule numbers per line demonstrated that 98.6% to 100% of MtPUB2^S421D/MtPUB2-RNAi (line 119 and line 32) roots developed zero to two nodules; by contrast, for EV/MtPUB2-RNAi lines (line 119 and line 32), 90% to 94.3% of roots developed zero to two nodules and 5.7% to 10% of roots developed three nodules (n = 20–25 per repeat, three biological repeats per line; Qiu et al., 2015). Similarly, for MtPUB2^S421A/MtPUB2-RNAi lines (line 119 and line 32), 91.4% to 94.3% of roots developed zero to two nodules and 8.6% to 20% of roots developed three to five nodules (n = 20–25 per repeat, three biological repeats per line; Fig. 9, A–E; Supplemental Fig. S10; Qiu et al., 2015). These data indicate that the roots of MtPUB2^S421D/MtPUB2-RNAi lines (line 119 and line 32) form fewer nodules than those of MtPUB2/MtPUB2-RNAi lines, demonstrating that the stability of MtDMI2 is necessary to produce an appropriate nodule number (Fig. 9, B and D; Supplemental Fig. S10). The nodule number in roots of MtPUB2^S421A/MtPUB2-RNAi lines (line 119 and line 32) was comparable to that of EV/MtPUB2-RNAi lines; they do not recover to the same level observed in MtPUB2/MtPUB2-RNAi lines (Fig. 9, B and D; Supplemental Fig. S10). MtPUB2/MtPUB2-RNAi lines were used as the control; 35.7% to 37.1% of roots of MtPUB2/MtPUB2-RNAi lines (line 119 and line 32) developed zero to three nodules, whereas 62.9% to 64.3% of roots developed six to eight nodules (n = 20–25 per repeat, three biological repeats per line; Fig. 9, B and D; Supplemental Fig. S10). The relative expression of MtPUB2 in the rescued roots was 1.0 in EV/MtPUB2-RNAi roots, approximately 75-fold up-regulated in MtPUB2/MtPUB2-RNAi roots, approximately 70-fold up-regulated in MtPUB2^S421D/MtPUB2-RNAi roots, and approximately 72-fold up-regulated in MtPUB2^S421A/MtPUB2-RNAi roots (Fig. 9, C and E; Supplemental Fig. S10). Together, these findings clearly demonstrate that the phosphorylated form of MtPUB2 is essential for determining the number of nodules. In summary, Ser-421 is not only necessary for MtPUB2 ubiquitin ligase activity but also critical for nodulation in M. truncatula.

Figure 9.

Nodulation phenotypes of A. rhizogenes-transformed roots in rescued MtPUB2-RNAi lines. A, pCAMBIA2301-MtPUB2, pCAMBIA2301 (EV), pCAMBIA2301-MtPUB2^S421A, and pCAMBIA2301-MtPUB2^S421D were introduced into the stable MtPUB2-RNAi lines (lines 32 and 119), respectively. The nodulation phenotypes were counted at 21 DAI. B and D, Quantification of nodule numbers at 21 DAI in EV/MtPUB2-RNAi lines, MtPUB2/MtPUB2-RNAi lines, MtPUB2^S421A/MtPUB2-RNAi lines, and MtPUB2^S421D/MtPUB2-RNAi lines based on lines 119 (B) and 32 (D). The nodules were counted by the grading statistic method (n = 20–25 per repeat, three biological repeats per line). C and E, RT-qPCR analysis of relative MtPUB2 expression in the rescued lines EV/MtPUB2-RNAi, MtPUB2/MtPUB2-RNAi, MtPUB2^S421A/MtPUB2-RNAi, and MtPUB2^S421D/MtPUB2-RNAi based on lines 119 (C) and 32 (E). MtPUB2 mRNA quantification was normalized against two reference genes (Mtactin and MtEF1α) in each sample. Shown are mean values ± se calculated from three independent experiments (***P ≤ 0.001, Student’s t test).

Therefore, the results lead to the question of whether a negative feedback loop between MtPUB2 and MtDMI2 exists during nodulation to maintain homeostasis.

MtDMI2 and MtPUB2 Function in a Prey-Predator Relationship during Nodulation

To directly test the above hypothesis, we performed quantitative immunoblot assays (with two biological repeats, using actin protein as the loading control) in M. truncatula and analyzed the data using numerical solutions in detail (Lev Bar-Or et al., 2000; Marciano et al., 2016). Wild-type A17, dmi2-1 (TR25), and MtPUB2-RNAi plants were inoculated with S. meliloti 1021 and sampled at different time points (0 h, 5 h, 12 h, 24 h, 3 DAI, 5 DAI, 7 DAI, 14 DAI, 21 DAI, and 28 DAI) for immunoblot analysis. The results showed that both MtDMI2 and MtPUB2 levels increased and then decreased during the nodulation process in wild-type plants, whereas this trend was not observed in dmi2-1 (TR25) or MtPUB2-RNAi plants (Fig. 10A; Supplemental Fig. S11). We speculated that the interaction between MtDMI2 and MtPUB2 follows a prey-predator relationship during nodulation in wild-type plants.

Figure 10.

Quantitative immunoblot assays and the prey-predator relationship between MtDMI2 and MtPUB2. A, Quantitative immunoblot analysis of wild-type (A17) plants inoculated with S. meliloti strain 1021. The abundance of MtDMI2 (top) and MtPUB2 (bottom) was determined by quantitative immunoblot analysis using the whole root systems. The roots were collected at different time points after inoculation (0 h, 5 h, 12 h, 24 h, 3 DAI, 5 DAI, 7 DAI, 14 DAI, 21 DAI, and 28 DAI). Actin protein was used as the loading control. The abundance of MtPUB2 and MtDMI2 was determined using anti-MtPUB2 and anti-MtDMI2 polyclonal antibodies (BPI, Co. Ltd.) in experiments. Data from quantitative immunoblot bands were analyzed using ImageJ software. These experiments were repeated twice. B, The changes in MtPUB2 and MtDMI2 present as a wave-shaped curve during the nodulation process. The red box shows the time points selected for solving the prey-predator model (24 h–21 DAI). C, The curves derived from LV equations. The equations were solved numerically using MATLAB. The blue curve shows the changes in MtDMI2, and the gray shows the changes in MtPUB2 (repeat 1). The initial x and y were set to the MtDMI2 and MtPUB2 protein levels measured at 24 h. D, Similar to C, the orange curve shows the changes in MtDMI2, and the yellow shows the changes in MtPUB2 (repeat 2).

To substantiate this hypothesis, we adopted LV equations, which are pairs of first-order, nonlinear, differential equations. This model was first proposed by Lotka in 1910 (Lotka, 1910),

where X represents the number of prey (or MtDMI2 in our study); Y represents the number of predators (or MtPUB2); αX and δxy represent the increasing rates of X and Y, respectively; and βXY and γY represent the decreasing rates of X and Y, respectively. We applied the aforementioned method to designate repeat 1 and repeat 2 and determined two sets of best-fit parameters (α = 0.7, β = 1.4, γ = 1.2, δ = 0.1, and α = 1.7, β = 2.6, γ = 1.0, δ = 0.1).

These results verified our conjecture of the negative feedback between MtDMI2 and MtPUB2 from 24 h after infection to 21 DAI (Fig. 10B). As shown in Figure 10, C and D, the curves derived from LV equations perfectly depict the MtDMI2 and MtPUB2 protein levels measured at 24 h, 3 DAI, 5 DAI, 7 DAI, 14 DAI, and 21 DAI in both repeat 1 and repeat 2. This negative feedback loop was nonexistent in the inoculated dmi2-1 plants and MtPUB2-RNAi lines (Supplemental Fig. S11).

The MtDMI2-MtPUB2 Negative Feedback Loop Displays Crosstalk with the AON via MtNIN

To investigate the relationship between the MtDMI2-MtPUB2 negative feedback loop and the AON, the nodulation marker genes MtNSP1, MtNIN, MtDMI1, and MtDMI3 were detected by RT-qPCR in stable-transformed MtPUB2-RNAi and wild-type roots inoculated with S. meliloti strain 1021.

The relative expression of MtDMI1, MtDMI3, MtNSP1, and MtNIN was induced approximately 1.8-, 1.6-, 10-, and 9-fold by inoculation with S. meliloti strain 1021 in MtPUB2-RNAi roots compared to that in wild-type roots at 7 DAI (Fig. 11, A–D). The relative expression of MtNIN was up-regulated throughout the experiment; for instance, it was up-regulated approximately 9- to 16-fold in MtPUB2-RNAi roots compared to wild-type roots at 5 and 7 DAI with S. meliloti strain 1021 (Fig. 11B). MtNIN not only is required for nodule organogenesis (Marsh et al., 2007), but also plays a key role in the AON pathway (Soyano et al., 2014). The up-regulation of MtNIN will cause the up-regulation of CLAVATA3/ENDOSPERM SURROUNDING REGION-related small peptides (CLEs; Soyano et al., 2014). We inferred that potential crosstalk exists between the MtDMI2-MtPUB2 negative feedback loop and AON in M. truncatula.

Figure 11.

RT-qPCR analysis of nodulation marker gene expressions in the nodulated MtPUB2-RNAi roots and wild-type roots. A and D, RT-qPCR analysis of relative MtNSP1, MtNIN, MtDMI1, and MtDMI3 expression in nodulated MtPUB2-RNAi and wild-type roots (0 h, 5 h, 12 h, 24 h, 3 DAI, 5 DAI, 7 DAI, 14 DAI, 21 DAI, and 28 DAI). mRNA quantification was normalized against two internal reference genes (Mtactin and MtEF1α) in each sample. Shown are mean values ± se calculated from three independent experiments. The gray box shows the MtPUB2-RNAi roots, and the black box shows the wild-type roots.

Taken together, nodulation signals first triggered an increase in MtDMI2 through MtLYK3 and MtNFP. MtDMI2 phosphorylates MtPUB2 on Ser-316 and Ser-421. Before 7 DAI, because the phosphorylation on Ser-316 inactivated MtPUB2, MtDMI2 was accumulated and some unknown downstream genes were activated. From 7 DAI to 21 DAI, MtDMI2 was degraded because of the activation of MtPUB2 caused by the phosphorylation on Ser-421. As a result, nodulation homeostasis was achieved. After nodulation homeostasis was disrupted, we found that MtDMI2 was no longer degraded, MtNIN expression was up-regulated excessively, and the number of nodules decreased. In the future, we will study in detail how MtPUB2^Ser-316 and MtPUB2^Ser-421 cooperate during nodulation (Fig. 12).

Figure 12.

Proposed MtDMI2-MtPUB2 negative feedback loop working model. M. truncatula perceives NFs signals via MtNFP and MtLYK3. The signals are then transferred to MtDMI2. MtDMI2 activates MtPUB2 by phosphorylation (potential Ser-316 and Ser-421 sites of MtPUB2). Before 7 DAI, the phosphorylation on Ser-316 inactivated MtPUB2, which induced the accumulation of MtDMI2 and the activation of some unknown downstream genes. From 7 DAI to 21 DAI, the phosphorylation on Ser-421 caused the activation of MtPUB2, which induced the degradation on MtDMI2 and maintained the nodulation homeostasis. Once this balance was broken, MtDMI2 was no longer degraded, MtNIN was up-regulated excessively, and the nodule numbers were decreased. The graph shows the LV equations from the second repeat. The blue curve shows the changes in MtDMI2, and the gray shows the changes in MtPUB2. However, the details of how MtPUB2^Ser-316 (in red italic font, Ser-316) and MtPUB2^Ser-421 (in red italic font, Ser-421) cooperate during nodulation require further study.

DISCUSSION

The MtDMI2-MtDMI1-MtDMI3 symbiotic signaling pathway transfers nodulation signals, and the roots of dmi2, dmi1, and dmi3 mutants display no nodules (Wais et al., 2000; Endre et al., 2002; Lévy et al., 2004; Mitra et al., 2004; Riely et al., 2007). The dmi2-1 mutant (TR25) shows swollen root hair tips, but does not exhibit branching and has a nonnodulation phenotype (Esseling et al., 2004). The down-regulation of MtDMI2 is associated with various phenotypes, including numerous bulbous infection threads in the central tissue of the nodule but without the release of bacteria (Limpens et al., 2005). This study also showed that the majority of M. truncatula lines with down-regulated MtDMI2 did not develop nodules (Limpens et al., 2005). Here, we identified a new PUB-type E3 ubiquitin ligase that can degrade MtDMI2 via the 26S proteasome pathway. We showed that MtPUB2-RNAi roots exhibit impaired infection threads, fewer nodules, a shorter primary root length, and decreased nodule densities compared with EV-transformed roots (Fig. 3). Our results clearly identify a novel MtDMI2-MtPUB2 negative feedback loop that depends on phosphorylation and ubiquitination during nodulation. We speculate that this mechanism locally and tightly calibrates nodulation homeostasis.

Regarding the negative feedback loop working model, we analyzed the quantitative immunoblot analysis data via numerical solutions (Lev Bar-Or et al., 2000; Marciano et al., 2016). As shown in Figure 10, the curves derived from LV equations perfectly depicted the MtDMI2 and MtPUB2 protein levels measured at 24 h, 3 DAI, 5 DAI, 7 DAI, 14 DAI, and 21 DAI in both repeat 1 and repeat 2. The curves in the two repeats were not completely identical due to discrepancies in plant development. The curves showed time points from 24 h to 21 DAI but not before 24 h, because the ITs did not show any development from 0 to 24 h (Lohar et al., 2006). We speculate that the negative feedback between MtDMI2 and MtPUB2 is not yet established during this period (Fig. 12). At 3 DAI, the number of plants responding to rhizobia inoculation was significantly different compared to those subjected to mock treatment (Bersoult et al., 2005). At 5 DAI, most of the plants exhibited developing nodules emerging from the roots (Bersoult et al., 2005). In addition, most nodule development occurred from 7 to 21 DAI, and nodule ageing occurred after 28 DAI (Puppo et al., 2005). Our data showed that the dynamics occurred within the period from 24 h to 21 DAI, and once nodulation signals triggered an increase in MtDMI2, the activated MtPUB2 degrades MtDMI2 and itself (Fig. 10).

This negative feedback loop between MtPUB2 and MtDMI2 was not established in the inoculated dmi2-1 plants and MtPUB2-RNAi plants (Supplemental Fig. S11). In the inoculated MtPUB2-RNAi roots (0 h, 5 h, 12 h, 24 h, 3 DAI, 5 DAI, 7 DAI, 14 DAI, 21 DAI, and 28 DAI), MtDMI2 was always present at a relatively higher level compared with that in inoculated wild-type roots (Supplemental Fig. S11, A and B). Additionally, in the inoculated dmi2-1 (TR25) roots (0 h, 5 h, 12 h, 24 h, 3 DAI, 5 DAI, 7 DAI, 14 DAI, 21 DAI, and 28 DAI), MtPUB2 was always found at a relatively lower level compared with that in inoculated wild-type roots (Supplemental Fig. S11, C and D). The results of phos-tag assays further confirmed that MtPUB2 phosphorylation was decreased in dmi2-1 roots compared with that in wild-type roots (Fig. 7). Based on the results of the quantitative immunoblot and phos-tag analyses, we hypothesized that the stability of MtPUB2 was dependent on phosphorylation by MtDMI2.

To further test the working model of feedback between MtDMI2 and MtPUB2, we analyzed the rhizobial induction of crucial gene expression of the CSP (MtDMI1, MtDMI3, MtNIN, and MtNSP1) pathway in wild-type and MtPUB2-RNAi roots. MtDMI1, MtDMI3, MtNIN, and MtNSP1 expression was required for a NF-activated signal transduction pathway (Catoira et al., 2000; Gonzalez-Rizzo et al., 2006; Messinese et al., 2007; Lopez-Gomez et al., 2012). MtNSP1, MtNIN, MtDMI1, and MtDMI3 expression increased simultaneously through the nodulation time course in MtPUB2-RNAi roots, suggesting that MtDMI2 was partly beyond the control of MtPUB2 in MtPUB2-RNAi roots (Fig. 11, A–D). In particular, MtNIN expression was up-regulated in nodulated MtPUB2-RNAi roots compared to that in wild-type roots during the entire experiment (Fig. 11, A and B); however, the nodule numbers of MtPUB2-RNAi roots were decreased (Figs. 2 and 3). As we mentioned regarding AON, MtNIN expression can systemically suppress nodulation via AON to control the number of nodules (Soyano et al., 2014). We assumed that the up-regulation of MtNIN expression, which could further lead to the up-regulation of CLEs, inhibited the nodulation by AON in our study. Thus, we hypothesized that the MtDMI2-MtPUB2 negative feedback loop, which displays crosstalk with the long-distance AON via MtNIN, plays a vital role in nodulation homeostasis. Recently, Saha and DasGupta claimed that crosstalk may exist between supernodulating null and SYMRK receptors for the activation as well as restriction of nodule development in M. truncatula (Saha and DasGupta, 2015). These data suggest that regulation of nodulation is complex.

After its phosphorylation, MtPUB2 gains the ability to ubiquitinate MtDMI2; identifying the key amino acid that MtDMI2 phosphorylates MtPUB2 via LC-MS/MS is very important (Supplemental Fig. S8). We selected three potential amino acids, namely Ser-421, Ser-316, and Thr-488, for further study. The results demonstrated that Ser-421 in the ARM-repeats domain is the key phosphorylation site of MtPUB2 that affects its ubiquitination activity. Further study revealed that MtPUB2^S421D (mimicking the phosphorylated form) could ubiquitinate MtDMI2 in vitro, but MtPUB2^S421A had minimal activity (Fig. 8A). Furthermore, we performed rescue assays with MtPUB2^S421D and MtPUB2^S421A in stable MtPUB2-RNAi lines to test their functions during nodulation. MtPUB2/MtPUB2-RNAi lines and EV/MtPUB2-RNAi lines were used as controls in this experiment (Fig. 9). The results showed that MtPUB2^S421D/MtPUB2-RNAi lines formed the fewest nodules. MtPUB2^S421A/MtPUB2-RNAi lines had almost the same number of nodules as EV/MtPUB2-RNAi lines, and MtPUB2/MtPUB2-RNAi lines formed similar numbers of nodules as wild-type plants (Fig. 9, B and C). The reasons for the different phenotypes were as follows: (1) In the MtPUB2^S421D/MtPUB2-RNAi lines, because of the mimicked sustained phosphorylation of MtPUB2, MtDMI2 was ubiquitinated constantly, leading to the fewest nodules during nodulation. (2) In the MtPUB2^S421A/MtPUB2-RNAi lines, MtPUB2 lacked ubiquitination activity, and thus the nodulation phenotype was similar to that of EV/MtPUB2-RNAi lines (Fig. 9). We also assessed the function of Ser-316 by performing in vitro ubiquitination assays with MtDMI2 and in vivo rescue assays with MtPUB2^S316D in stable MtPUB2-RNAi lines to test its function during nodulation (Fig. 8; Supplemental Figs. S12 and S13). Based on alignment of the U-box domain with AtPUB13, AtPUB12, OsSPL11, Glyma20g32340.1, Lj5g3v1796810.2, AtPUB14, MtPUB1, and AtPUB22, Ser-316 conservation is 100% (Supplemental Fig. S11). MtPUB2^S316D cannot perform self-ubiquitination; however, MtPUB2^S316A retains self-ubiquitination capacity and transfers ubiquitin to MtDMI2 (Fig. 8C). In addition, the results of rescue assays with MtPUB2^S316D showed that MtPUB2^S316D/MtPUB2-RNAi lines had similar nodule numbers as EV/MtPUB2-RNAi lines (lines 115 and 119; Supplemental Fig. S13).

The quantitative immunoblot analysis results demonstrate that before 7 DAI, MtDMI2 and MtPUB2 levels both show a steady increasing trend in the two repeats (Figs. 10, C and D and 12). Thus, it is likely that MtDMI2 phosphorylated MtPUB2 at Ser-316; however, because MtPUB2^S316D lost the capacity for self-ubiquitination and the transfer of ubiquitin to MtDMI2 (Fig. 8C), MtPUB2 might accumulate. At 7 DAI, the accumulation of MtDMI2 and MtPUB2^S316D may ensure the activation to the downstream genes of MtDMI2. After 7 DAI, the phosphorylation of Ser-421 was greater than that of Ser-316, which enhanced self-ubiquitination and MtDMI2 ubiquitination. From 7 DAI to 21 DAI, MtPUB2 caused the degradation of MtDMI2 via ubiquitination, and MtPUB2 caused its own degradation via self-ubiquitination (Figs. 10, C and D, and 12). These data also substantiated the prey (MtDMI2)-predator (MtPUB2) working model during nodulation. Further studies are still needed to fully understand the phosphorylation changes of Ser-316 and Ser-421 and their contributions to the nodulation phenotype.

In conclusion, this study describes a novel dynamic model involving an E3 ubiquitin ligase and a protein receptor kinase during nodulation in M. truncatula. Moreover, this negative feedback model plays a critical role in controlling nodulation homeostasis in legumes. Therefore, we propose that there is crosstalk between the AON and the MtPUB2-MtDMI2 negative feedback loop during nodulation in M. truncatula.

MATERIALS AND METHODS

cDNA Library Construction

To generate a high-quality Medicago truncatula Y2H cDNA library, we used the roots of M. truncatula genotype cv Jemalog A17 plants inoculated with S. meliloti strain 1021 or mock-inoculated for different durations (0 h, 5 h, 12 h, 24 h, 3 d, 5 d, 7 d, 14 d, 21 d, 28 d, and 35 d). The library was constructed according to the protocols supplied with the Mate & Plate kit (Clontech, cat. no. 630490). The complexity of the cDNA library was 2.8 × 10⁷, the concentration was 1,150 ng/μL, the average insertion size was 750 bp, and the recombinant efficiency was 90% (nonamplified transformant random check).

Plant Materials

M. truncatula cv Jemalog ecotype A17 and ecotype R108 were used for phenotypic and genotypic analyses. The homozygous mutant dmi2-1 (TR25, frame shift due to a 1-bp deletion, premature translational termination in the extracellular domain, Nod^-Myc^-) was acquired from Jean-Michel Ané (Endre et al., 2002), and the seeds were treated for germination as previously described (Pennycooke et al., 2008). The seeds were planted in a chamber under the following environmental conditions: 14 h/10 h light/dark cycle, dim light (60 μmol photons s^-1m⁻², Philips; TL5 Essential Super 80), 22°C/18°C day/night regime, and 70% relative humidity (RH).

Phylogenetic Tree Analysis

The PUB domains for each subfamily were used as an input for the phylogenetic tree analysis. For the phylogenetic analysis of MtPUB2 and homologous PUBs in M. truncatula, Lotus japonicas, Arabidopsis (Arabidopsis thaliana), Glycine max, and Oryza sativa, sequences were obtained from Phytozome v11.0 (http://www.phytozome.net; Supplemental File S2). We analyzed the multiple sequence alignment of PUB amino acids using ClustalW2 software and constructed the phylogenetic tree using the neighbor-joining method. The numbers on the branches indicate bootstrap values based on 1,000 replicates.

Spatiotemporal Expression of MtPUB2

The 2.3-kb promoter of MtPUB2 was cloned from the DNA of M. truncatula cv Jemalog A17. The promoter was initially cloned into the pTOPO-Entry vector (Invitrogen) using Gateway technology and then into the pkGWFS7 vector that carried a downstream GUS fusion construct (Huault et al., 2014). The final proMtPUB2-GUS construct was transformed into M. truncatula via Agrobacterium tumefaciens. The plants were then used for proMtPUB2-GUS assays during nodulation. The roots, nodules, leaves, and stems were immersed in a X-gluc solution at 37°C for approximately 18 h in a vacuum for 2 h. The color was removed using 75% ethanol. The samples were sliced with a Leica VT1000S and photographed using an Olympus BX50 microscope. The primers related to this experiment are listed in Supplemental Table S2.

M. truncatula Transformation

An MtPUB2-specific RNAi construct was prepared to target a 450-bp fragment containing the first 450 bp of the MtPUB2 coding sequence (1–450 bp starting from ATG). The primers used for PCR are listed in Supplemental Table S1. This 450-bp fragment was cloned into the pANDA vector (Clontech), which is driven by the ubiquitin promoter, to create pANDA-MtPUB2. The stable overexpression transgenic lines were prepared using the pCAMBIA1302 vector (pCAMBIA1302-MtPUB2) driven by the cauliflower mosaic virus 35S promoter (CaMV35S). Hygromycin (10 mg/L, Roche) was used to select transformed plants. The regenerated identified T₀ transformed plantlets were further cultivated in soil. After 8 to 10 months, the seeds of the transgenic plants were harvested for further analysis (www.noble.org/MedicagoHandbook/).

For the Agrobacterium rhizogenes-mediated hairy root transformation of M. truncatula genotype R108 plants, germinated seeds were transformed with the A. rhizogenes strain ARqual1, as described by Boisson-Dernier et al. (2001). For RNAi experiments, the PCR product was cloned into the pFRN-RNAi vector (Gonzalez-Rizzo et al., 2006). For overexpression assays, the CDS of MtPUB2 was cloned from A17 cDNA via PCR and inserted into the modified pCAMBIA2301 vector under the control of the dual CaMV35S to obtain the recombinant pCAMBIA2301-MtPUB2 vector. The transgenic seedlings were then transferred to square petri dishes (130 × 130 mm) on Fåhraeus medium without nitrogen (pH 6.5) after 4 weeks of growth. Five to six plants were placed on each plate. All plates were incubated in the growth chamber for 2 weeks under the following conditions: 16-h/8-h-light/-dark cycle, dim light (60 μmol photons s⁻¹ m⁻²), 24°C day/night regime, and 70% RH. The inoculation of S. meliloti strain 1021 grown to an OD₆₀₀ of 0.5 was performed 7 d after transferring the seedlings to the square petri dishes.

For the nodulation assay, roots of 1-week-old seedlings were inoculated with rhizobial bacteria; plants were flood-inoculated with a total of 20 mL of a resuspended cell suspension (OD₆₀₀ = 0.5) per plate. Nodulation phenotypes were recorded at 21 DAI. Photographs of nodules were taken under a stereomicroscope (MZFLIII, Leica). All primers used in this experiment are listed in Supplemental Table S2.

RT-qPCR Analysis

Total RNA was extracted with TRIzol (Invitrogen) from the roots, leaves, stems, and nodules of M. truncatula plants (different treatments), and 1 mg of total RNA was used for reverse transcription. RT-qPCR reactions were performed using SYBR Premix Ex Taq (TaKaRa) on a Bio-Rad CFX-96 PCR system (Bio-Rad) according to the manufacturer’s instructions. The internal reference genes were MtEF1a and Mtactin. Three independent biological replicates and three independent technical replicates were performed for each condition tested. We sampled the roots, leaves, stems, and nodules taken from three different plants at the same time grown in different pots. Total RNA was extracted from pooled roots, leaves, stems, and nodule tissues; there are three plants/seedlings pooled per sample. Gene expression was calculated using the 2-^△△CT method. The primers used in this experiment are listed in Supplemental Table S2.

Infection Threads Assay and Lugol Staining of Nodules

Infection threads assays were performed 5 DAI. Rhizobia used in this assay are S. meliloti 1021 pXLD4 expressing HemA-lacZ. The roots and nodules were in a vacuum for three cycles of 30 s with 2.5% glutaraldehyde in 0.1 m PIPES (pH 7.2), fixed for 2 h, and rinsed in 0.1 m PIPES (pH 7.2) three times. The whole roots were then incubated in staining solution containing 50 mm potassium ferrocyanide, 50 mm potassium ferricyanide, 0.1 m PIPES (pH 7.2), and 0.08% X-gal for 16 h at 28°C. Finally, the stained roots were rinsed with 0.1 m PIPES (pH 7.2) with 2.4% sodium hypochlorite for 15 min before being photographed with an Olympus BX50 microscope.

For the amyloplast detection of nodules, 21-DAI and 42-DAI nodules were collected and embedded in 6% agarose. The samples were then sliced into 80 μm with Leica VT1000S microtome. The amyloplasts were stained in the lugol solution (Sigma-Aldrich) for 5 min and then washed with water. The sections were photographed using an Olympus BX50 microscope.

Y2H Screen and Pairwise Interactions

The sequence encoding the intracellular region of MtDMI2 was amplified by PCR and cloned into the pGBKT7 (binding domain) vector (Clontech), which was transformed into the yeast AH109 strain using the LiAc-mediated yeast transformation method (Gietz and Schiestl, 2007). The cDNA inserts of positive clones were amplified by PCR from yeast cells and were then sequenced. The protein domains of MtPUB2 were cloned using primers specific to the U-box (258 to 321 aa), ARM (382 to 628 aa), an unnamed domain (1–257 aa), and the full-length (1–662 aa). MtCRA2, MtLYK3, Medtr1g094021.1, and Medtr4g028960.1 were cloned using specific primers. The truncated domains of MtPUB2 were cloned into pGBKT7, and the sequence encoding the protein kinase domain of MtDMI2 was cloned into pGADT7 (activation domain). Yeast transformants were screened on 4D selective medium (SD/-Ade/-His /-Leu/-Trp) supplied with 8 mm 3-amino-1,2,4-trazole to suppress self-activation and incubated at 28°C for 4 to 5 d. The primers used in this experiment are listed in Supplemental Table S2.

Subcellular Localization and BiFC Experiments in Onion Epidermal Peels

The open reading frame of MtPUB2 without the termination codon was inserted into the modified pE3025 plasmid containing yellow fluorescent protein (YFP, green) instead of red fluorescent protein at the EcoRI and SalI sites to generate the Pro-35S:MtPUB2-cGFP vector. YFP fluorescence was visualized using a confocal laser-scanning microscope. BiFC experiments were performed in a structural complementation assay using pSY735- and pSY736-derived plasmids driven by the CaMV35S (MtDMI2-Yc, Yn-MtPUB2, Yn-Medtr1g094021.1) for transient expression of the fusion proteins in onion (Allium cepa) epidermal peels. Relevant negative controls and positive controls were produced at the same time. YFP fluorescence (green) was monitored by excitation at 488 nm with an argon laser combined with a 505- to 520-nm bandpass filter using a Nikon ECLIPSE TE2000-E inverted fluorescence microscope equipped with a Nikon d-ECLIPSE C1 (Nikon) spectral confocal laser-scanning system. All primers used in this experiment are listed in Supplemental Table S2 (pE3025-MtPUB2-F, pE3025-MtPUB2-R, pSY736-MtPUB2-F, pSY736-MtPUB2-R, pSY735-MtDMI2-F, pSY735-MtDMI2-R, pSY736-Medtr1g094021.1-F, pSY736-Medtr1g094021.1-R).

GST Pull-Down Assay

MtDMI2IR was cloned into pGEX-4T-1 as a GST fusion and then transformed into the Escherichia coli BL21 strain. The recombinant protein was purified by GST-agarose affinity 610 chromatography, as described previously (Yang et al., 2011). MtPUB2 was cloned into pET-30a (+) as a His-fusion and then transformed into the E. coli BL21 strain. The His-MtPUB2 protein was purified by chromatography using Ni-NTA agarose (Invitrogen) following the manufacturer’s instructions. Immobilized GST-MtDMI2IR was obtained from the lysate by incubation with Glutathione Sepharose 4B agarose beads for 1 h at 4°C (GE Healthcare Life Science). Unbound proteins were removed with lysis buffer at 4°C for 5 min. Finally, the interaction complex was analyzed by immunoblot analysis using anti-GST and anti-His antibodies (mouse monoclonal, Abcam). The primers used in this experiment are listed in Supplemental Table S2 (pET30a-MtPUB2-F, pET30a-MtPUB2-R, pGEX-4T-1-MtDMI2IR-F, pGEX-4T-1-MtDMI2IR-R).

In Vitro Ubiquitination Assay

The reactions contained 500 ng of substrate protein, 20 ng of E1 (UBE1, rabbit, Boston Biochem), 40 ng of E2 (UbcH5c, human recombinant, Boston Biochem), 5 μg of Arabidopsis ubiquitin (Boston Biochem), and 3 μg of purified MBP-MtPUB2 (MtPUB2 was cloned into pMAL-c2x, expressed in E. coli strain TB1 with an MBP-tag at the N terminus, and purified using amylose resin) in ubiquitination buffer (0.1 m Tris-HCl pH 7.5, 25 mm MgCl₂, 2.5 mm dithiothreitol, and 10 mm ATP) in a final volume of 30 μL. The reactions were incubated at 30°C for 2 h and were stopped by the addition of 4× SDS sample buffer. The samples were detected by immunoblot analysis with anti-MBP, anti-GST, and anti-Ubi (mouse monoclonal, Abcam) antibodies.

The primers used in this experiment are listed in Supplemental Table S2.

In Vitro Phosphorylation Assay

For the protein kinase assays, approximately 5 μg of kinase and 50 μg of the substrate protein were incubated in 45 μL of 0.5 m HEPES, pH 7.4, 15 mm MnCl₂, 1 m dithiothreitol, 50 μm ATP, and 1 μCi of [^γ-32P]-ATP per reaction at 28°C for 30 min. The reaction was stopped by the addition of 4× SDS loading buffer and boiling at 100°C for 5 min. The samples were then used directly and separated through 10% SDS-PAGE. After electrophoresis, the gel was dried, and kinase autophosphorylation activities were detected via autoradiography. The primers used in this experiment are listed in Supplemental Table S2 (pMAL-C2x-MtPUB2-F, pMAL-C2x-MtPUB2-R, pGEX-4T-1-MtDMI2IR-F, pGEX-4T-1-MtDMI2IR-R, pGEX-4T-1-MtDMI2IR^T762A-F, pGEX-4T-1-MtDMI2IR^T762A-R).

CIAP and Phos-Tag Assays in M. truncatula

Proteins were extracted with native extraction buffer 1 (NB1; 10 mm HEPES, pH 7.5, 100 mm NaCl, 1 mm EDTA, 10% glycerol, 0.5% Triton X-100 and protease inhibitor cocktail (Roche), and PMSF [AMRESCO]). For in vivo calf intestinal (CIAP, New England Biology) assays, the proteins were extracted from the roots of wild-type Jemalong A17 and dmi2-1 (TR25) plants at 21 DAI. The proteins were treated with CIAP at 37°C for 30 min and then detected with an anti-MtPUB2 (BGI, Co. Ltd.) polyclonal antibody and an antiactin (CWBio) antibody on immunoblots. For the phos-tag (BOPPARD) assay, proteins were extracted from the roots of wild-type A17 plants at 7 DAI, 14 DAI, and 21 DAI, and phos-tag was added to SDS-PAGE gels at a concentration of 5.0 mm. Anti-MtPUB2 and antiactin antibodies were used in the immunoblot analysis.

Cellular Fractionation Assay

For the cellular fractionation assay, wild-type N. benthamiana plants were used as the host plants for the assays. Plants were grown in a growth chamber at 22°C/18°C day/night regime, 70% RH under a 16-h/8-h-light/-dark photoperiod for approximately 4 to 5 weeks before infiltration. Agrobacterium EHA105 strains carrying MtPB2-HA and p19 expression constructs were co-infiltrated at different ratios. Three days after infiltration, the samples were collected for analysis. The fusion proteins were extracted according to the protocols of Lei et al. (2015) and Duan et al. (2017). Then the fractions (1 mg) were separated by 10% SDS-PAGE and subjected to immunoblot analysis where a 1:5,000 dilution of H⁺ATPase (Agrisera; approximately 95 kD, plasma membrane marker antibody) and a 1:5,000 dilution of cFBPase (Agrisera; approximately 37 kD, cytosolic fraction marker antibody) were used as the marker antibodies.

In Vivo Degradation

The methods were modified from those described by Liu et al. (2010). Wild-type Nicotiana benthamiana plants were used as the host plants for the assays. For the in vivo degradation experiments, Agrobacterium EHA105 strains carrying E3 ligase, substrate, p19 expression constructs, and internal control plasmids were co-infiltrated at different ratios. Three days after infiltration, the samples were collected for analysis. MG132 (N-[(phenylmethoxy)carbonyl]-l-leucyl-N-[(1S)-1-formyl-3-methylbutyl]-l-leucinamide, Sigma-Aldrich) dissolved in 10 mm MgCl₂ at a final concentration of 50 μm was infiltrated into the previous infiltrated region 12 h before sample collection. The proteins were extracted with NB1 (10 mm HEPES, pH 7.5, 100 mm NaCl, 1 mm EDTA, 10% glycerol, 0.5% Triton X-100, protease inhibitor cocktail [Roche], and PMSF [AMRESCO]). The samples were used in immunoblot analysis. The expression of MtDMI2 and NbEF1α were detected by RT-qPCR assays to confirm that the degradation of MtDMI2 was mediated by MtPUB2 at the protein level. The primers used in this experiment are listed in Supplemental Table S2 (pCAM1300-Myc-MtDMI2-F, pCAM1300-Myc-MtDMI2-R, pCAM1300-HA-MtPUB2-F, pCAM1300-HA-MtPUB2-R).

Semi-in Vivo Degradation Assay

The degradation assay and semi-in vivo protein degradation analysis were conducted by modifying a previously reported method (Liu et al., 2010). For the semi-in vivo protein degradation analysis, 3 d after infiltration, Myc-MtDMI2, MtPUB2-HA, MtPUB2^V274R-HA, GFP-HA, and wild-type samples were harvested. The samples were separately extracted in NB1. To preserve the function of the 26S proteasome, ATP (Sigma-Aldrich) was added to the cell lysates at a final concentration of 10 μm. The MtPUB2-HA/MtPUB2^V274R-HA extract was then mixed with Myc-MtDMI2 or wild-type extract in a volume ratio of 1:1. MG132 was added to the corresponding mixtures to a final concentration of 50 μm. The mixtures were incubated at 4°C with gentle shaking. The samples were used in immunoblot analysis and detected with anti-HA and anti-Myc antibodies. All primers used in this experiment are listed in Supplemental Table S2 (pCAM1300-Myc-MtDMI2-F, pCAM1300-Myc-MtDMI2-R, pCAM1300-HA-MtPUB2-F, pCAM1300-HA-MtPUB2-R).

LC-MS/MS Assay

To detect the sites of MtPUB2 that were phosphorylated by MtDMI2, we performed an LC-MS/MS assay using in vitro-expressed proteins (MBP-MtPUB2 and GST-MtDMI2IR) in the E. coli BL21 strain. The samples were separated using 8% SDS-PAGE gels and digested with trypsin. The digested peptides were analyzed in a SCIEX TripleTOF 5600+ high-resolution mass spectrometer (AB SCIEX). The ion spray voltage was 2.4 kV. The time of flight masses of time of flight MS and the product ion were 3,500 to 1,500 D and 100 to 1,500 D, respectively. The time-of-flight masses (TOF MS) and the production ion were 350 to 1,500 Da and 100 to 1,500 Da, respectively. In the LC-MS/MS assay, three biological replicates were performed for control and phosphorylated samples. We sampled each specimen of cultures of unique E. coli BL21 strain transformants in three different biological replicates.

The data were analyzed using ProteinPilot 5.0 software and Paragon algorithm search parameters (AB SCIEX).

Quantitative Immunoblot Analysis and Dynamic Analysis in M. truncatula

In the quantitative immunoblot assay, the M. truncatula ecotype A17 was the wild-type material, and MtPUB2-RNAi roots and dmi2-1 (TR25) roots were the mutant materials. The roots of 7-d-old plants without a nitrogen source (0 h) were treated with S. meliloti strain 1021, and the entire root systems were analyzed at different time points (0 h, 5 h, 12 h, 24 h, 3 DAI, 5 DAI, 7 DAI, 14 DAI, 21 DAI, and 28 DAI). The abundance of actin protein was used as a loading control. The abundance of MtPUB2 and MtDMI2 in the total root systems was assessed using anti-MtPUB2 and anti-MtDMI2 polyclonal antibodies (BPI).

MtPUB2 (full length) and a truncated MtDMI2 (amino acids 543–919) were expressed in the E. coli BL21 strain using the pET-30a (+) vector and pGEX-4T-1 vector, respectively, purified, and used to produce anti-MtPUB2 and anti-MtDMI2 antibodies, respectively. The antibodies were produced by Beijing Protein Innovation (BPI, http://www.proteomics.org.cn). The arterial blood serum was collected, and the antibodies were purified by Protein A Agarose (GE Healthcare; 17-0402-01).

Roots for quantitative immunoblot analysis were collected at 0 h, 5 h, 12 h, 24 h, 3 DAI, 5 DAI, 7 DAI, 14 DAI, 21 DAI, and 28 DAI. The quantitative immunoblot band data were analyzed with ImageJ software. Finally, the data were assessed as the MtPUB2/Mtactin and MtDMI2/Mtactin ratios.

To model the negative feedback loop between MtDMI2 and MtPUB2 across various time points, we adopted LV equations (better known as prey-predator equations), which are a pair of first-order, nonlinear, differential equations (Lotka, 1910). These equations are widely applied to many natural systems, such as chemical reactions and aerosol-cloud-precipitation interactions (Lotka, 1910; Koren and Feingold, 2011). The equations are expressed as follows:

Therefore, a set of four positive parameters (α, β, δ, and γ) dictates the concentration of MtDMI2 and MtPUB2 as a function of time. In this study, we let the parameters α/β vary from 0.1 to 10 and the parameters δ/γ vary from 0.1 to 1 with an interval of 0.1. The initial values of x and y were set to the MtDMI2 and MtPUB2 protein levels measured on day 1 (24 h). The equations were solved numerically using the MATLAB program. The best set of the α, β, δ, and γ parameters was determined using the least square fit for the modeled x and y with the MtDMI2 and MtPUB2 protein levels measured at 24 h, 3 DAI, 5 DAI, 7 DAI, 14 DAI, and 21 DAI. The program is shown in Supplemental File S1.

Rescue of the MtPUB2-RNAi and Nodulation Phenotype Assays

For the A. rhizogenes-mediated hairy root transformation of M. truncatula genotype R108 plants, germinated seeds were transformed with the A. rhizogenes strain ARqual1, as described by Boisson-Dernier et al. (2001). For the rescue tests, the CDSs of MtPUB2, MtPUB2^S421A, MtPUB2^S421D, and MtPUB2^S316D were inserted into the modified pCAMBIA2301 vector under the control of dual CaMV35S to obtain the recombinant pCAMBIA2301-MtPUB2 vector, pCAMBIA-MtPUB2^S421A vector, pCAMBIA-MtPUB2^S421D vector, and pCAMBIA-MtPUB2^S316D vector. pCAMBIA2301 was used as the empty vector control. The primers used in this experiment are listed in Supplemental Table S2.

Accession Numbers

Sequence data from this article can be found in the M. truncatula genome database and the GenBank data library under accession numbers Medtr5g030920 (MtDMI2), Medtr8g043970 (MtDMI3), Medtr2g005870 (MtDMI1), Medtr5g099060 (MtNIN), XM_003627246 (MtNSP1), AY372402 (MtLYK3), XM_013606619 (MtCRA2), and KU285617.1 (MtPUB2).

SUPPLEMENTAL DATA

The following supplemental materials are available.

• Supplemental Figure S1. Selection of MtDMI2-interacting proteins by Y2H.

• Supplemental Figure S2. Structure analysis and phylogenetic tree analysis of MtPUB2.

• Supplemental Figure S3. Phylogenetic tree of the homologous PUBs in Medicago truncatula (Medtr), Arabidopsis thaliana (At), Glycine max (Glyma), Oryza sativa (Os) and L. japonicus (Lj).

• Supplemental Figure S4. Interaction assays between MtDMI2IR, Medtr1g094021.1, and Medtr4g028960.1 in yeast cells.

• Supplemental Figure S5. Representative infection threads in GUS-RNAi and MtPUB2-RNAi roots.

• Figure Figure S6. Relative expressions of MtPUB2 and Medtr1g094025.1 in the GUS-RNAi lines and MtPUB2-RNAi lines.

• Supplemental Figure S7. Nodulation phenotypes of the MtPUB2-overexpressing lines.

• Supplemental Fig. S8. MtPUB2 sites with high possibility of phosphorylation based on LC-MS/MS assay.

• Supplemental Figure S9. In vitro ubiquitination assays with MBP-MtPUB2^S421D and GST-MtDMI2IR.

• Supplemental Figure S10. Nodule quantification in transgenic lines.

• Supplemental Figure S11. Quantitative immunoblot assays of the nodulated MtPUB2-RNAi roots and dmi2-1 (TR25) roots.

• Supplemental Figure S12. Amino acid sequence and domain structure of PUB members in M. truncatula (Medtr), Arabidopsis (At), soybean (Glyma), rice (Os), and L. japonicus (Lj).

• Supplemental Figure S13. Nodulation phenotypes of A. rhizogenes-transformed roots in rescued MtPUB2-RNAi lines.

• Supplemental Table S1. LC-MS/MS data of phosphorylation sites in MtPUB2.

• Supplemental Table S2. Primers used in this study.

• Supplemental File S1. MATLAB program.

• Supplemental File S2. Text file of the amino acid sequence of PUB-type proteins in M. truncatula (Medtr), Arabidopsis (At), soybean (Glyma), rice (Os), and L. japonicus (Lj) that were used as an input for the phylogenetic tree analysis.