During the last decades a considerable effort has been made to characterize the diversity of proteins expressed in different tissues under a variety of conditions at the international level. This effort was initiated using 2D-PAGE-based approaches, generating reference maps for various organs in different legume species. Taking M. truncatula as an example, Mathesius et al. (2001) were the first to establish a root reference map, which included ~2,500 protein spots, from which 179 were identified. It was remarkable that close to half of them were present as protein isoforms, including key metabolic enzymes such as S-adenosyl-L-methionine synthase, malate dehydrogenase, or ascorbate peroxidase. Gallardo et al. (2003) reported the characterization of the seed proteome during seed filling, with the identification of 84 proteins including proteins belonging to the main storage protein families as well as proteins involved in carbon and sulfur metabolism, among others. Subsequently, works from the Sumner laboratory published a comprehensive analysis of the M. truncatula proteome at the different organ level, including cell cultures, with the high-confidence identification of close to 2,000 proteins, the largest proteomic identification reported so far (Watson et al., 2003; Lei et al., 2005).

Focusing on the symbiotic perspective, several comparative proteomics works have been devoted to the analysis of the differential root proteome of a number of legumes when inoculated with their corresponding microbial partners. Depending on the focus of the study, works in the field of symbiosis can be divided in two major groups: (i) studies focused on the characterization of the legume plant proteome, and (ii) works focused on the Rhizobium partner (Table 1). In the latter case, a classical comparison is the analysis of the proteome of free-living Rhizobium cells vs. their differentiated nitrogen-fixing forms, named bacteroids. This strategy has been applied to identify symbiosis-specific proteins synthesized in the symbionts of the main legume species. Comparison of the proteomic profiles of cultured cells vs. Sinorhizobium meliloti nodule bacteroids suggested that nodule bacteria do not express sugar transporters or enzymes involved in the early steps of glycolysis, while containing multiple transporters for nitrogen compounds including amino acids and oligopeptides (Djordjevic et al., 2003; Djordjevic, 2004). The high adaptability of symbiotic bacteria depending on the carbon source was also reported in proteomic works carried out in Bradyrhizobium japonicum (Sarma and Emerich, 2005, 2006). Authors reported the unusually low levels in proteins related to fatty acid and nucleic acid metabolism in bacteroids, suggesting that bacteroids and cell cultured bacteria may present differential mechanisms to regulate the levels of ribonucleotides. Subsequent work in B. japonicum using more powerful mass spectrometry techniques, however, did not observe such repression in nucleotide metabolism, identifying almost the full set of enzymes involved in de novo nucleoside and nucleotide biosynthesis expressed at the gen and/or protein level (Delmotte et al., 2010). Regarding the symbiont of L. japonicus, Tatsukami et al. (2013) identified 722 proteins commonly found under the free-living and symbiotic conditions, while 125 proteins were uniquely identified under symbiotic conditions. Interestingly, proteins involved in peptidoglycan biosynthesis and proteins related to the flagellum were uniquely detected under free-living conditions, suggesting that once within the symbiosomes, bacteroids simplify their cell surface by losing their cell wall and motility structures (Tatsukami et al., 2013). Furthermore, the quantitative time-course proteomic analysis of M. loti suggested that bacteroids experience nitrogen-deficiency at early stages of nodule development, while at intermediate stages high levels of nitrogenase protein lead to nitrogen-rich conditions in the symbiosome (Nambu et al., 2015).

In the case of plant-oriented studies, it is common to compare differences in the proteome of uninoculated vs. inoculated plants. One of the pioneer works in this line was carried out by Krause and Broughton (1992), reporting 12 symbiosis-specific proteins potentially involved in root-hair deformation in Vigna unguiculata, although at the time the lack of genomic sequence information limited protein identification. The availability of expressed sequence tags (ESTs) and, subsequently, genomic sequences of several legume plants have led to greatly improve the number of identified proteins. For instance, the model legume M. truncatula has been subjected to detailed proteomic characterization in terms of symbiotic responses. One of the first studies in nodules identified two leghemoglobin isoforms and one enolase protein as some of the proteins that most accumulated in roots upon inoculation with symbiotic bacteria (Bestel-Corre et al., 2002). The protein profiling of the plant fraction of M. truncatula root nodules led to the identification of 377 unique proteins, most of them with roles in amino acid metabolism and protein synthesis and degradation (Larrainzar et al., 2007). Proteomics is also a valuable tool for the analysis of local and systemic responses upon inoculation. Through the analysis of the proteomic changes occurring in shoots and roots of inoculated M. truncatula plants, 18 proteins were found to accumulate in roots including sucrose synthase 1, a fructose-bisphosphate aldolase, and an alcohol dehydrogenase, while in shoots the content of several proteins involved in defense responses or abiotic stress responses was found to increase (Molesini et al., 2014). Proteomics has been also applied to investigate the effects of the addition of the ethylene precursor aminocyclopropane carboxylic acid (ACC) on nodule development using the supernodulating, ethylene-insensitive mutant sickle (Penmetsa and Cook, 1997) during the early stages of the symbiotic interaction (Prayitno et al., 2006). Among other findings, authors observed that Sinorhizobium inoculation increased the abundance of one ACC oxidase isoform in wild-type roots but not in sickle roots, suggesting that a feedback mechanism regulates the expression this gene. Nevertheless, subsequent work using RNA-seq techniques has shown that at least three genes of the ACC oxidase family are induced in sickle upon inoculation (Larrainzar et al., 2015), which highlights the usefulness of combining data at the proteomic and transcriptomic level.

Further characterization of specific metabolic pathways using absolute quantification techniques has been also applied to this model legume, including the detailed analysis of the nitrogen assimilation and ethylene biosynthesis pathways in root nodules (Larrainzar et al., 2009, 2014). Similarly, the symbiotic proteome of soybean, a crop of major economical importance, has been also extensively studied. A time-course proteomic analysis of wild type and the soybean mutant SS2-2, which lacks autoregulation of nodulation, has revealed that there is a protein-mediated suppression of defense-related responses in root cells upon inoculation with Rhizobium bacteria (Lim et al., 2010). A similar observation was done when comparing the proteomic changes associated to inoculation of soybean plants with differential nodulation capacities (Salavati et al., 2012). In this work, a correlation between the levels of a peroxidase isoform and levels of nodulation was found, although the regulation did not occur at the transcript level. The plant fraction of soybean nodules has been also subjected to proteomic analysis, leading to the identification of 69 proteins mainly related to carbon and nitrogen metabolic activities (Oehrle et al., 2008), similarly to previous observations in M. truncatula.

Since root hairs are most frequently the main entry point for Rhizobium bacteria, several works have been devoted to identify the proteomic changes occurring in this specialized root cell upon inoculation. This work has the obvious technical limitation that collecting sufficient amount of plant material is challenging and requires a large number of plants per proteomic sample, with estimations of around 1,500 soybean roots and 4,000 soybean seedlings (Wan et al., 2005; Brechenmacher et al., 2009). A time course of the proteomic changes occurring in root hairs revealed that there is a specific induction of phospholipases and phosphoglucomutases, as well as a lectin and an actin isoform upon inoculation (Wan et al., 2005). Under uninoculated conditions, Brechenmacher et al. (2009) combined traditional 2D-PAGE and shotgun methods for the identification of 1,492 proteins present in root hairs, establishing a reference map for future work.

Several recent studies have provided broad insights into the systemic effects the legume-Rhizobium symbiotic interaction at the root and, especially, leaf metabolic level (Staudinger et al., 2012, 2016; Desalegn et al., 2016; Turetschek et al., 2016). In all these studies, one of the major response found was related to a significantly induction of the plant translational apparatus and an accumulation of plant proteins involved in stress responses.

Proteomic approaches are particularly suited to gain information about protein subcellular localizations. Analyses of enriched fractions or, ideally, purified organelles, or sub-organelle compartments allow the validation of protein compartmentalization and isoform localization data. Furthermore, subcellular fractionation provides valuable information on the specific changes in the proteome of organelles in response to various stresses, allowing for the development of accurate proteomic pathways and networks (Hooper et al., 2014). Legume plants have been also subjected to this type of analysis. Most of work, however, has been done under non-symbiotic conditions. Comprehensive reviews in the field of subcellular proteomics in legumes have been published elsewhere (Lee et al., 2013; Wang and Komatsu, 2016; Yin and Komatsu, 2016). Thus, in the current review we will discuss subcellular proteomic works with a symbiotic focus.

Nodules are complex structures containing a combination of infected and non-infected plant cells. Infected cells are filled with nitrogen-fixing bacteroids arranged in symbiosomes surrounded by a specialized plant membrane named peribacteroid membrane (PBM). The identification of proteins present in this specialized membrane is of key relevance, since it represents the direct interface where nutrient and signal exchange occurs between the legume host plant and Rhizobium bacteroids. In order to identify which proteins are localized at the PBM, extensive proteomic work has been carried out in this membrane fraction. Panter et al. (2000) carried out one of the first studies, with the 2D-PAGE analysis of the PBM of soybean nodules. The proteomic characterization of the pea PBM and peribacteroid space, a much more challenging approach, has also been studied (Saalbach et al., 2002). In this work, proteins of the Coatomer-coated vesicles like V-ATPase, BIP, were found in the PBM fraction, supporting the role of the endomembrane system in PBM biogenesis. These studies were followed by more comprehensive LC-MS/MS-based analyses of the PBM in legumes such as L. japonicus (Wienkoop and Saalbach, 2003), M. truncatula (Catalano et al., 2004), and more recently, soybean (Clarke et al., 2015). Identification of protein components in these membranes has been shown an essential first step for increasing our knowledge on the metabolic exchange processes between plant and bacteroid. For instance, the proteomic analysis of the PBM in L. japonicus led to the identification of a symbiosis-specific sulfate transporter 1 (SST1) and around 80 other abundant membrane or membrane-associated proteins such as the hypersensitive response protein and remorin (SYMREM1; Wienkoop and Saalbach, 2003). Both SST1 and SYMREM1 have subsequently been characterized in detail and shown to be relevant for nodule development and functioning (Krusell et al., 2005; Lefebvre et al., 2010; Toth et al., 2012; Domonkos et al., 2013; Kalloniati et al., 2015).

One of the strengths of mass spectrometry-based proteomic approaches is that it allows the identification of post-translational modification (PTM) sites in proteins, something that cannot be accurately predicted with genomic information alone. PTMs have a huge impact on plant signaling and metabolism, contributing to the regulation of protein activity, stability/degradation, interactions, and ultimately gene expression. In recent years, the number of studies focused on the identification of PTMs in plants has considerable grown (for an extensive review, see Friso and van Wijk, 2015). Particular interest has received the large-scale identification of phosphoproteins at early stages of the legume-Rhizobium symbiotic interaction. The reason behind this interest is that a number of protein kinases have been shown essential for rhizobial infection and/or nodule development. These include LysM-domain-containing receptor kinases, implicated in the binding of Nod factors, as well as other membrane receptor-like kinases and calcium/cadmodulin-dependent protein kinases (Antolín-Llovera et al., 2014). Analysis of the phosphoproteome at early symbiotic stages has been carried out in L. japonicus roots (Serna-Sanz et al., 2011), soybean root hairs (Nguyen et al., 2012), and M. truncatula roots both applying discovery proteomics (Rose et al., 2012) and targeted approaches (van Ness et al., 2016). These studies are important first steps toward understanding how Nod-factor signaling is transmitted from the plasma membrane and decoded to activate the developmental reprogramming root cells undergo to allow nodule formation. Previous studies identified several phosphoproteins present in mature nitrogen-fixing nodules, including several phosphopeptides in sucrose synthase 1 and alkaline invertase, the main sucrose-degrading enzymes in this specialized organ (Wienkoop et al., 2008). Similarly, large-scale analyses of M. truncatula roots (Grimsrud et al., 2010) or during nodulation (Marx et al., 2016) have allowed the identification of a repertoire of in vivo phosphorylated peptides and phosphorylation motifs, which can be queried online (http://www.phospho.medicago.wisc.edu and http://compendium.medicago.wisc.edu, respectively).

Ubiquitination is another PTM that plays important roles for rhizobial infection and nodule organogenesis in legumes. For instance, proteins such as the receptor kinase M. truncatula LYK3 has been found ubiquitinated in vitro by the E3 ubiquitin ligase PUB1 (Mbengue et al., 2010) and a deubiquitinating enzyme named AMSH1 has been shown to be required for the establishment of an effective symbiosis in L. japonicus (Małolepszy et al., 2015). However, to our knowledge, the large-scale identification of ubiquitinated proteins under symbiotic conditions has not been carried out to date.

The role of the signaling molecule nitric oxide and its associated protein modifications, nitrosylation and nitration, has also drawn considerable attention in legume studies. At the level of root nodules, two proteins have been identified as targets of Tyr nitration: the ammonium-assimilating enzyme glutamine synthetase (GS; Melo et al., 2011; Blanquet et al., 2015) and the hemeprotein leghemoglobin (Sainz et al., 2015). In the latter work it was found that leghemoglobin not only plays an important role as an O₂ transporter but may also act as sink of toxic peroxynitrite and thus be part of a protective mechanism in symbiosis.

Other PTM studies in the field include the identification of sulfenylated proteins in the M. truncatula-S. meliloti symbiosis (Oger et al., 2012). Interestingly, both in the plant and bacterial partners a large proportion of the identified proteins were related to the glycolytic pathway, tricarboxylic acid cycle and amino-acid metabolism, including one of the nodule-enhanced sucrose synthase isoforms. Additionally, a cytosolic isoform of glutamine synthetase has been found as a target of sulfoxidation, although this modification did not alter the activity of the enzyme (Matamoros et al., 2013).

Although a number of proteomic studies have been published on the response of legumes to abiotic stresses, in most studies, however, plants under study have not been grown under symbiotic conditions. For instance, analyses of peanut (Arachis hypogaea L.) varieties with contrasting tolerance to water deficit have shown a relation between plant tolerance to the stress and the abundance of proteins involved in stress signaling and wax biosynthesis in leaves (Kottapalli et al., 2009). A similar approach was taken by Katam et al. (2016) who identified proteins related to nitrogen metabolism, defense and cellular biogenesis exclusively in the tolerant peanut cultivar. Abiotic stress responses in a major legume crop like soybean have been reviewed elsewhere (Hossain et al., 2012, 2013; Hossain and Komatsu, 2014; Komatsu et al., 2015). Here we will focus on the proteomic works analyzing the effects of abiotic stress of legumes grown under symbiotic conditions.

Regarding the legume-Rhizobium symbiosis, the effects of drought stress on the M. truncatula nodule proteome have been characterized in most detailed. Proteomic analysis of the plant protein fraction of drought-stressed nodules showed a general decline in enzymes involved in symbiotic nitrogen fixation and N assimilation (Larrainzar et al., 2007). In contrast, the microsymbiont showed an accumulation of chaperonins and heat-shock proteins, along with enzymes involved in energy metabolism. In a later study, the application of an integrative proteomic and metabolic approach to M. truncatula nodules allowed the identification of the main effects of drought followed by a recovery treatment (Larrainzar et al., 2009). Both works highlighted the role of enzymes related to sulfur assimilation in nodules, with the identification of a specific, nodule-enhanced methionine synthase isoform responding to water deficit conditions. Thus, in a subsequent work, Larrainzar et al. (2014) analyzed in detail the involvement of sulfur metabolism in the nodule response to drought. Gene expression and absolute quantification of proteins in the methionine and ethylene biosynthesis pathways were found strongly down-regulated during drought, demonstrating the contribution of nodule sulfur metabolism and the phytohormone ethylene for nodule functioning under water deficit conditions. Using a split-root system and proteomics, local inhibition of nitrogen fixation and nodule metabolism was evidenced in M. truncatula and soybean plants partially exposed to drought stress (Gil-Quintana et al., 2013). More recently, the drought response of two legume species was compared, finding a similar molecular response to drought but a higher drought tolerance in the tropical legume (Gil-Quintana et al., 2015). In this study, protein isoforms that share same function and location could be identified across species.

Induced systemic resistance and the positive effect of microbes in the plant immune system has received considerable in recent years (van Wees et al., 2008). In this line, the effect of Rhizobium in priming plant metabolism in response to salt and drought stress has been analyzed in M. truncatula using a combination of proteomic and metabolomic techniques (Staudinger et al., 2012, 2016). These works highlighted the involvement of another group of nodule proteins in the plant response to drought; the family of lipoxygenases, enzymes related to lipid metabolism and jasmonate biosynthesis. Additionally, authors found an influence of the Rhizobium symbiosis on the abundance levels of stress responsive proteins prior application of the stress treatment. In subsequent works, plants grown under symbiotic conditions were found to present reduced levels of leaf senescence under drought stress, regardless the efficiency of the Rhizobium strain used (Staudinger et al., 2016). Authors concluded that, besides increased jasmonate biosynthesis and potassium ion concentrations, allocation of reserves to osmolytes and a shift in carbon partitioning from starch to sugar are key responses for the observed symbiont-induced stay-green effect.

The plant-AM symbiosis has been shown to have a positive impact on the plant nutritional status, thus, improving growth and crop productivity. In legumes, an overall enhanced tolerance to abiotic stress has been also reported (Ruiz-Lozano et al., 2001). However, specifically in the field of proteomics applied to the legume-AM symbiosis, works describing this positive effect are scarce. One of the few examples are the analyses by Aloui and colleagues focused on the changes in the proteome of M. truncatula roots (Aloui et al., 2009) and shoots (Aloui et al., 2011) when colonized with the AM fungus G. intraradices in cadmium (Cd)-free and Cd-contaminated substrates. Authors identified nine out of 15 proteins changing under Cd treatment in non-mycorrhizal roots that were not observed in Cd-treated, colonized roots. In shoots Cd induced the accumulation of proteins related to photosynthesis in plants under symbiotic conditions. As a conclusion, when exposed to Cd, symbiotically grown plants partially escaped metal toxicity through a concerted increase in shoot biomass and thanks to allocation plasticity strategies compared to non-symbiotic plants.

Collectively, in the last decade proteomics research has provided a significant contribution to the understanding of the legume symbiotic interactions in response to biotic and abiotic stress. A summary of the main conclusions drawn can be found in Figure 1.

Symbiotic interactions and responses to a pathogen attack share some features especially at the early stage of the interaction. Legume hosts initially recognize their symbiotic partners as potential threats, activating plant defense responses that are subsequently down-regulated at later stages (Zamioudis and Pieterse, 2012). Since plants under symbiotic conditions appear to show an improved tolerance to abiotic stresses, it could be hypothesized that a similar pattern would be observed when challenged by pathogens. In terms of proteomic analysis, few are the works that have analyzed this question and most studies have been carried out on legumes grown under non-symbiotic conditions. One of the few works on the topic analyzed the proteomic changes occurring during the tripartite interactions with S. meliloti, mycorrhizal fungi, and the pathogenic oomycota Aphanomyces euteiches in M. truncatula roots (Schenkluhn et al., 2010). Among the proteins detected after inoculation either with G. intraradices and/or S. meliloti the highest increase in relative content was observed for a calmodulin-2, which could be related to the activation of Nod- and Myc-factor signaling cascades. In agreement with previous studies, authors observed the activation of proteins involved in antioxidant defense and/or scavenging of reactive oxygen species. At later stages the pathogen induced the expression of a different set of proteins including pathogen response (PR) proteins, Kunitz-type proteinase inhibitors, a lectin, and proteins related to primary carbohydrate metabolism. Interestingly, the induction of these pathogen responses occurred to a lesser extent in plants under mixed infections (i.e., symbiotic plants infected with A. euteiches), which indeed reinforces the idea of the beneficial effects of symbiosis toward pathogenic infections. A similar observation was made when analyzing the leaf proteome and metabolome of pea plants after the co-inoculation with R. leguminosarum and AM fungi, or either of those combined with the pathogenic fungi Didymella pinodes (Desalegn et al., 2016). The Rhizobium symbiosis considerably increased the levels of proteins involved in the pisatin pathway upon pathogen attack compared to mycorrhized, co-inoculated plants or plants under non-symbiotic conditions. Hence, it was concluded that proteins and pathways involved in symbiotic interactions might indirectly be linked to specific pathogen-response pathways. However, the regulatory link between the different molecular responses induced by symbionts and pathogens remains unknown. In fact, when in a subsequent work the effect of the symbionts and pathogen disease levels were compared on two pea cultivars with differential susceptibility, the influence of the microsymbiont was found superimposed by genotypic resistance traits (Turetschek et al., 2016).

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.