Iron (Fe) is the sixth most abundant element in the universe and the fourth most abundant in the earth's crust (Morgan and Anders, 1980). For most living organisms on earth, Fe is an essential element. However, its chemical properties in oxygenated environments limit its availability. As a consequence, Fe deficiency is a widespread phenomenon among the animal and plant kingdoms. According to the WHO, anemia (including Fe-deficiency induced anemia) affects more than 2 billion people worldwide (McLean et al., 2009). In crops, and especially in those grown on calcareous soils, Fe deficiency is a major nutritional disorder that causes decreases in vegetative growth and marked yield and quality losses (Abadía et al., 2011). A substantial body of knowledge has arisen during the genomic era with regard to the molecular aspects of plant Fe acquisition mechanisms that are induced in response to limited Fe availability (Palmer and Guerinot, 2009; Schmidt and Buckhout, 2011; Hindt and Guerinot, 2012). In all plants except grasses, the adaptation mechanism to Fe limitation suggests the need for increased reducing power (NADH) for the root plasma membrane (PM) Fe reductase, and increased energy (ATP) for the PM H⁺-ATPase (Römheld and Marschner, 1983). Proteomic approaches are an excellent tool to elucidate the general metabolic reprogramming needed to sustain these requirements. Understanding how plants respond to Fe limitation at the proteomic level provides a new layer of information on Fe homeostasis processes that can ultimately help to reduce the effects of Fe deficiency in crops. This review summarizes, in a comprehensive manner, the major conclusions regarding metabolic adaptations that occur under Fe deficiency conditions in roots and thylakoids of several plant species, and highlights future research directions in the field. Also, the review provides a clear overview on the limitations of the different proteomic techniques available to study a prevalent abiotic stress case in crops.

When Fe is scarce, Strategy I plants develop morphological and biochemical changes leading to an increase in their Fe acquisition capacity. Morphological changes include swelling of root tips and formation of lateral roots, root hairs, and transfer cells that increase the root's surface area. Biochemical changes result in an increased ability to acquire Fe, and include the induction of a plasma-membrane Fe(III)-reductase and an Fe(II) transporter, an enhanced proton extrusion capacity, and the release of low molecular weight compounds such as carboxylates, flavins and phenolic compounds (Abadía et al., 2011).

The most common approach used to study changes in the root proteome upon Fe starvation has been a gel-based approach, using two-dimensional (2-DE) IEF SDS/PAGE electrophoresis. All of the studies reviewed here have been conducted using root extracts of Strategy I plants, whereas proteomic studies in Strategy II plants have only been conducted in root plasma membrane of maize (Hopff et al., 2013). Small gels (7 cm) were used in studies involving Beta vulgaris (Rellán-Álvarez et al., 2010), Medicago truncatula (Rodríguez-Celma et al., 2011a) and Prunus dulcis × Prunus persica (Rodríguez-Celma et al., 2013a), whereas larger gels were used for Cucumis sativus (24 cm, Donnini et al., 2010) and Solanum lycopersicum (24 cm in Li et al., 2008 and 13 cm in Brumbarova et al., 2008). When using small gels the average number of proteins detected in root extracts ranged between 140 for B. vulgaris to ~335 for P. dulcis × P. persica and M. truncatula, whereas in larger gels the number of spots ranged from 1400 in S. lycopersicum to 2200 in C. sativus. Independently of the total number of spots, the number of spots whose relative abundance changed significantly as a result of the different Fe deficiency treatments was within the same range among plant species, with 17, 31, and 61 spots for P. dulcis × P. persica, M. truncatula and B. vulgaris, 57 for C. sativus, and 41 (Li et al., 2008) and 24 (Brumbarova et al., 2008) for S. lycopersicum, respectively. The protein identification rates ranged between 71 and 95%, except for B. vulgaris where only 36% of the proteins changing in abundance upon Fe limitation were identified, underlying the importance of using plant species with sufficient genomic information for proteomic studies. In A. thaliana root extracts, a high-throughput study of changes induced by Fe limitation was performed using HPLC-MS and iTRAQ (Isobaric Tag for Relative and Absolute Quantification). In this study, 4454 proteins were identified in A. thaliana root extracts and 2882 were reliably quantified; from these, a subset of 101 proteins were identified whose abundance changed upon Fe deficiency (Lan et al., 2011). Finally, a phosphoproteomics study with A. thaliana found that among 425 phosphorylated proteins, 45 changed in abundance with Fe deficiency (Lan et al., 2012).

The comparison of proteomes from multiple plant species is not straightforward. This is because of difficulties in handling homology between species, in particular when public genomic data sets are not comprehensive, or in some cases are lacking, as it occurs with B. vulgaris. Furthermore, the reliability of protein identification using database comparisons always has some uncertainty and clearly depends on the stringency of the criteria selected for positive identification, including number of peptides matched and sequence coverage. In addition, because any given proteomic profile is a snapshot of a dynamic process, differences in plant developmental stage and in the manner Fe limitation is imposed can likely lead to final protein profile pictures that may differ significantly. Also, we should keep in mind that changes in spot intensity can be related to post-translational modifications, adding a new layer of complexity to the interpretation of proteomic studies. As a compromise, we used an approach where identified proteins found to change in each plant species were matched to proteins existing in A. thaliana. BLAST searches were run for every protein species found in the different studies and those matches with E-values below e⁻³⁰ were retained as positive Arabidopsis orthologs; if any of the accession numbers overlapped between two or more entries of the same plant species, they were considered redundant and only a single representative entry was retained (Table S1). Also, in an attempt to reduce variability associated with differences in the treatments, the proteomes of B. vulgaris and M. truncatula plants grown in the absence of Fe, but in the presence of CaCO₃, were in principle excluded from the comparison (although data are also provided in Table S1).

Based on this approach, 199 non-redundant Arabidopsis orthologs were found to change in relative abundance with Fe deficiency: 101, 17, 37, 27, and 32 of them were derived from the Arabidopsis (Lan et al., 2011), M. truncatula, S. lycopersicum, P. dulcis × P. persica, and C. sativus studies, respectively. These non-redundant root proteins and the results from the BLAST searches are listed in Table S1. Results showing proteins grouped by function (generally as described by authors, although with some modifications) are summarized in Figure 1A and Table S2 (in the latter case with red and green backgrounds indicating proteins increasing or decreasing in relative abundance, respectively). This comparison did not yield a single specific protein species changing in all plant species as a result of Fe limitation, whereas only 15 proteins did show changes in two or more plant species (Figure 1B; Table S2). Among them, six were related to stress defense, three were related to C metabolism, and three to N metabolism. The remaining three protein species were a β-xylosidase involved in cell wall modification, the translation initiation factor ELF5A-1, and the chaperone HSP70-1. In spite of the low number of common protein overlaps, changes in metabolic pathways do follow a similar trend. It should be noted that changes in Fe homeostasis and transport proteins upon Fe deficiency were only detected in Arabidopsis using HPLC-MS, because most of these proteins are membrane associated and therefore difficult to solubilize and detect with gel-based 2-DE approaches.

Changes induced by Fe deficiency in the shoot proteome have only been reported for S. lycopersicum (Herbik et al., 1996). In that study, although apparent differences were observed in the protein profiles of Fe-sufficient and Fe-deficient leaves, protein identification was not pursued because the aim of the work was to identify differences between a mutant S. lycopersicum genotype (chloronerva) and wild type.

Thylakoid membranes contain the multiprotein photosynthetic complexes photosystems I and II, which include the reaction centers responsible for converting light energy into chemical bond energy, as well as cytochrome b₆/f and ATPase complexes. Fe-deficient plants have a reduced number of granal and stromal lamellae per chloroplast and decreased amounts of many thylakoid membrane components including proteins, electron carriers and lipids and this is consistent with marked decreases in photosynthetic rates (Terry, 1980; Terry and Abadía, 1986; Morales et al., 1990).

Changes induced in the thylakoid proteome by Fe-deficiency have been studied in B. vulgaris and Spinacia oleracea using 2-DE blue native (BN) SDS/PAGE in combination with mass spectrometry (Andaluz et al., 2006; Timperio et al., 2007) and by HPLC and tandem mass spectrometry in A. thaliana (Laganowsky et al., 2009). These studies clearly demonstrate that Fe deficiency causes decreases in the relative amounts of electron transfer proteins, including the core and light harvesting components of the PSI and PSII complexes and in cytochrome b6/f, with the largest decreases seen in PSI. The time course studies performed in S. oleracea also indicate that Fe deficiency affects supercomplex organization, by inducing a transient decrease in trimeric and dimeric organization of the light harvesting complex of PSII. These changes might constitute an adaptive strategy to facilitate energy dissipation in Fe-deficient plants (Timperio et al., 2007).

Contrasting results were obtained for ATP synthase, showing no Fe-deficiency induced changes in B. vulgaris and A. thaliana and decreases in S. oleracea. Also, the relative amounts of proteins involved in leaf C fixation, including Rubisco, increased in B. vulgaris, whereas Rubisco decreased significantly in S. oleracea. These results may be species-specific or more likely are related to plant culture conditions, because S. oleracea was grown without Fe from germination, whereas B. vulgaris and Arabidopsis had a preculture period in the presence of Fe. Other changes observed in proteins present in thylakoid preparations upon Fe limitation included increases in enzymes involved in ROS protection, such as ascorbate peroxidase in A. thaliana and SOD in B. vulgaris.

The analyses of exact masses of thylakoid membrane proteins in Arabidopsis also revealed that several proteins undergo post-translational modifications upon Fe deficiency; for example, the PSII oxygen evolving complex in Fe-deficient Arabidopsis thylakoids was found mostly in its doubly phosphorylated form, while in Fe-sufficient samples the non-phosphorylated form was predominant (Laganowsky et al., 2009).

Results provided in these papers underscore that 2-DE BN SDS/PAGE and HPLC-MS are the techniques more suitable for proteomic studies involving membranes, because they allow the resolution of highly hydrophobic integral membrane proteins. However, the use of other approaches such as 2-DE IEF SDS/PAGE also provides useful complementary information about other sets of proteins tightly linked to membranes.

As proteomic and transcriptomic techniques have evolved toward high-throughput methodologies such as iTRAQ or the new Illumina technology, it has become more evident that a strict correlation does not exist between observed changes in protein and gene expression profiles. The comparison between changes in proteomic and transcriptomic profiles was out of the scope of this review; however, we have compared the datasets obtained from Arabidopsis Fe-deficient root extracts using iTRAQ and RNA sequencing. This comparison is only possible in Arabidopsis, where the genome is available and both studies have been reported; for other plant species, such as Medicago, the genome is still not finished and there are no reports of iTRAQ profiles. The total number of Arabidopsis transcripts identified was ~26 K, from which 2679 (10%) were differentially expressed upon Fe deficiency (Rodríguez-Celma et al., 2013b), whereas the total number of proteins identified by iTRAQ was 4454, from which 101 (2%) were differentially accumulated upon Fe deficiency (Lan et al., 2011). Among these, forty-seven protein species (50% of the differentially accumulated proteins) showed changes in both protein accumulation and gene expression (Table S3). The results of this comparison indicate that both omic approaches are complementary and that the gap in information between them is not only due to methodological limitations, but also reflects the complex regulation of stress responses such as Fe deficiency.

As shown in this review, proteomics has been a powerful tool in the elucidation of general metabolic rearrangements upon Fe deficiency. Different proteomic approaches have demonstrated that Fe deficiency has a profound impact on C metabolism and on the arrangement of the photosynthetic machinery, with many of these changes conserved amongst plant species. However, species- and treatment-specific changes have also been revealed that deserve further attention. The full potential of proteomic approaches is quite far from being fully exploited and detailed studies on specific changes unveiled with this holistic technique may yield new insights into the adaptation of plants to Fe deficiency. Future directions for proteomic studies should focus on the analysis of sub-proteomes crucial for Fe homeostasis, such as those of plant fluids, isolated organelles, and purified membrane preparations. Recent examples have been published for plasma membrane (Meisrimler et al., 2011; Hopff et al., 2013) and phloem sap (Lattanzio et al., 2013). These studies will help to elucidate the specific roles subcellular compartments may play in Fe homeostasis, and also will allow the discovery of proteins with new functions in these processes, which might remain hidden using other, less targeted approaches.