Protein disorder in plants: a view from the chloroplast

Background The intrinsically unstructured state of some proteins, observed in all living organisms, is essential for basic cellular functions. In this field the available information from plants is limited but it has been reached a point where these proteins can be comprehensively classified on the basis of disorder, function and evolution. Results Our analysis of plant genomes confirms that nuclear-encoded proteins follow the same trend than other multi-cellular eukaryotes; however, chloroplast- and mitochondria- encoded proteins conserve the patterns of Archaea and Bacteria, in agreement with their phylogenetic origin. Based on current knowledge about gene transference from the chloroplast to the nucleus, we report a strong correlation between the rate of disorder of transferred and nuclear-encoded proteins, even for polypeptides that play functional roles back in the chloroplast. We further investigate this trend by reviewing the set of chloroplast ribosomal proteins, one of the most representative transferred gene clusters, finding that the ribosomal large subunit, assembled from a majority of nuclear-encoded proteins, is clearly more unstructured than the small one, which integrates mostly plastid-encoded proteins. Conclusions Our observations suggest that the evolutionary dynamics of the plant nucleus adds disordered segments to genes alike, regardless of their origin, with the notable exception of proteins currently encoded in both genomes, probably due to functional constraints.


Background
A relevant fraction of genomes encode for proteins with structural disordered regions. Intrinsically protein disorder refers to segments or to whole proteins that do not fold into well-defined regular three-dimensional structures in isolation (i.e. not bound to other molecules) [1,2]. This disorder covers local flexible loops, extended domains, molten globule domains and folded domains with flexible linkers [3]. Thus, proteins might be either entirely disordered or partially disordered, characterised by regions spanning just a few (<10) consecutive disordered residues (loops in otherwise wellstructured proteins) or long stretches (>30) of contiguously disordered residues. The presence of protein disorder is thought to confer dynamic flexibility to proteins, allowing transitions between different structural states [4]. This increased flexibility is advantageous to proteins that recognise multiple target molecules such as DNA, RNA, other proteins or small ligands [3,5]. It is predicted that between 30% and 60% of proteins contain stretches of 30 or more disordered residues, with multi-cellular eukaryotes having much more predicted disorder than unicellular eukaryotes [6]. There is evidence that the unstructured state, common to all living organisms, is essential for basic cellular functions [5,7]. Whole-cell NMR experiments demonstrate that intrinsic disorder can exist in vivo [3,8] and therefore this state does not result merely from the failure to find the correct conditions for folding or ligand binding. Despite their lack of a well-defined three dimensional (3D) structure, these proteins carry out basic functions, mostly associated with regulatory processes in the cell, including transcription, translation, cellular signal transduction, protein phosphorylation, the storage of small molecules, and the regulation of the self-assembly of large multi-protein complexes such as the ribosome, in which interactions with multiple partners and highspecificity / low-affinity interactions are often required. The functional diversity provided by disordered regions complements that of ordered protein regions [9][10][11]. It has been also reported the importance of disordered interfaces in the modulation of cellular regulatory response, which participate in subtle regulation by switching its specificity for different binding partners [12].
In plants, the available information about intrinsic disorder in proteins is rather limited compared to other eukaryotic organisms and concerns basically to Arabidopsis thaliana, which was the first complete genome sequenced. Particularly, it has been pointed out that late embryogenesis abundant (LEA) proteins, with chaperone activity, and dehydrin proteins, lack a stable threedimensional structure being probably fully disordered [13][14][15]. These proteins are associated with abiotic stress tolerance, particularly with cold stress and dehydration. The computational prediction of disorder by Dunker et al. [1] did not reveal notable disorder differences among the proteome of A. thaliana and those of other eukaryotes. However, currently it is not known whether this scenario is general for all plant proteomes. Additionally, another overlooked aspect is the comparison of the degree of disorder in organelle and nuclear proteomes. Evolutionary analysis of A. thaliana, cyanobacterial and chloroplast genomes have revealed that many genes were transferred from plastids to the nucleus during plant evolution [16]. In particular, it has been estimated that in A. thaliana approximately 18% of the total protein-coding genes were acquired from the cyanobacterial ancestor of plastids.
At present computational analysis are considered crucial and indispensable for the identification and characterization of unstructured proteins [2,17]. Several methods have been developed to predict intrinsic disorder from amino acid sequences, such as DisEMBL [18]; GLOBPROT2 [19]; DISOPRED2 [20,21]; IUPred [22]; PONDR VL-XT [23][24][25], among others. Among these we decided to use the DISOPRED2 software, which has achieved specificities of 0.95 at the residue level in four successive Critical Assessment of Techniques for Protein Structure Prediction experiments (CASP6-9), and has been shown to be the best predictor of long disordered regions in CASP9 [26,27].
Here we report the disorder analysis of proteins from 8 vascular plants, 1 bryophyta and 3 chlorophyta encoded in either plastid, mitochondrial or nuclear genomes by using the DISOPRED2 method. In order to gain biological and evolutionary insights, we focus on the subset of chloroplast genes which moved to the nucleus during plant evolution. It is observed that originally chloroplastencoded proteins acquired disorder after their genes moved to the nucleus. In contrast, proteins still encoded in the chloroplast chromosome barely become disordered. Finally, in order to further evaluate these findings, we review the incorporation of disorder to chloroplast ribosomal subunits, one of the most representative transferred gene clusters, in comparison to their bacterial counterparts.

Analysis of disorder and occurrence of amino acids in protein sequences
We have analyzed the occurrence of protein disorder in 12 complete plant proteomes (see Materials and Methods). Chloroplast (ca. 85 proteins in average), mitochondrial (ca. 64 proteins in average) and nuclear (ca. 25,000 proteins in average) proteomes were separately analyzed and the occurrence of disordered regions of different length (L) was calculated. In plant nuclear proteomes the percentages of predicted disordered segments with L ≥ 30, L ≥ 40, and L ≥ 50 were determined (full detail in Additional file 1: Table S1). The data showed in average a range of disorder ranging from 40 to 56%, 26 to 44% and 19 to 33%, respectively. Figure 1 summarizes the data corresponding to predicted to-be-disordered segments with L ≥ 30. The highest percentages of disorder were found in Zea mays (56.2%), Glycine max (53.3%), Physcomitrella thaliana (52.6%), Micromonas sp. RCC299 (52.9%) and Ostreococcus tauri (52.5%). In general, no statistically significant differences between vascular plants (8) and bryophyta (1) and chlorophyta (3) species were found (Χ 2 values of 2.367 for bryophyta and 0.060 for chlorophyta, see Additional file 2: Table  S2). Nonetheless Physcomitrella patens had the lowest percentage, 38.2%, a value close to those found in Archaea and bacteria. It is also worth mentioning that no obvious differences were observed between monocots and eudicots.
In an attempt to validate our disorder predictions, we searched in the Protein Data Bank (PDB) for homologous proteins to those of A. thaliana identified as intrinsically disordered proteins in our analysis, as explained in Materials and Methods. This was a very limited validation effort, since it was only possible to recover data for 70 sequences. Nevertheless, we found that 49/70 (61/70 if we consider terminal sequences partially aligned to predicted disordered regions) contained segments with unresolved 3D-structure.
The distribution of disordered segments of L ≥ 30 along complete protein sequences was calculated, splitting proteins in N-terminal (40 aa), C-terminal (40 aa) and internal regions. The results in Table 1 indicate that in nuclear proteomes the disordered regions are slightly more abundant in the internal regions of proteins (50 -65%) compared with the extremes of the protein sequence (14 -30%), being the N-terminal part (20 -31%) more disordered than the C-terminal one (14 -20%). This distribution differs to that calculated for chloroplasts and mitochondria; in organelles the results indicate a more similar occurrence of disorder in the internal regions (21 -41% in chloroplasts, and 28 -46% in mitochondria) compared with the terminal regions (15 -44% in chloroplasts and 24 -41% in mitochondria). This scenario was common for all the plant proteomes studied with the exception of the chloroplast from C. reindhartii, where the disorder distribution was similar to that observed in the nuclear proteome (i.e., the internal part was more disordered than the terminal regions).
Amino acid frequencies in disordered proteins were also analyzed. The amino acid residues Ser, Pro, Gln, Lys and Glu are over-represented in intrinsically disordered regions from nuclear proteomes. In contrast, the amino acid residues with lowest frequencies were Trp, Cys, Tyr, Phe, Ile, Leu and Val (Additional file 3: Figure S1A). In chloroplasts and mitochondria some differences were observed: Lys and Met showed higher frequencies, being Ser and Pro less abundant (Additional file 3: Figures S1B and S1C).

Disorder in proteins encoded by plastidic genes in the nucleus
Intrinsic disorder was investigated in proteins believed to be originally encoded in chloroplast genomes, which were subsequently transferred to the nuclear genome in the course of evolution. With this aim we retrieved from the PLAZA database (for details see Materials and Methods) all Arabidopsis thaliana protein-coding genes within the nuclear genome with a plastid origin as reported in Martin et al. [16]. The analysis revealed that in A. thaliana 147 of 298 total proteins (49.3%) contain L ≥ 30 segments disordered. The analysis for the rest of plant proteomes was done with the transferred nuclear genes identified by homology (see Materials and methods). We found that disordered proteins were 84 of 253 (  Arabidopsis thaliana (49.3%). As illustrated in Figure 2A, the acquisition of disorder by transferred proteins is not uniform across plant species. In 125 out of 226 orthologous groups of transferred genes there are instances where a protein contains long disordered segment in some species but not in others. The percentages of disorder in transferred proteins seem to follow the same trend observed for overall disorder in the corresponding proteomes. In order to further validate this observation we plotted the disorder frequencies of nuclear proteins for L ≥ 30 versus the frequencies of disorder in proteins originally encoded by chloroplast genes and currently placed in nuclear genomes ( Figure 2B). The Pearson correlation obtained was r = 0.826. However, when we plotted the frequencies of protein disorder in the chloroplast for L ≥ 30 versus the disorder frequencies of transferred chloroplast genes ( Figure 2C), the obtained correlation coefficient was insignificant (r = 0.0154).
Martin et al. [16] reported that some genes encoding for cyanobacterial proteins and identified in the plant nuclear genome still conserve a copy in the chloroplast genome. We have found that this group of proteins has a much lower percentage of disorder (ca. 7%) than those that have lost their original chloroplast sequences (20 -52%). In the case of A. thaliana our results revealed that a group of 47 nuclear-encoded proteins maintain putative orthologous copies in the chromosome of the chloroplast. In particular we found that these nuclear proteins correspond to 27 chloroplastic non-disordered proteins, indicating that some of them might be paralogues. For instance, this is the case of the chloroplast NAD(P)H-quinone oxidoreductase subunit 2B (AtCg01250), the NAD(P)H dehydrogenase (AtCg01090), the RNA polymerase beta' subunit (AtCg00180) or the second-largest subunit of DNA-dependent RNA polymerase (AtCg00190). In addition, ribosomal proteins L14 (AtCg00780), L22 (AtCg00810), S8 (AtCg00770) and S19 (AtCg00820), which are among the most conserved ribosomal proteins and bind directly to 23S and 16S rRNAs, respectively, are included in this group [28][29][30] (Additional file 4: Table  S3). As mentioned above, these conserved proteins barely acquire disorder. The scheme in Figure 3 summarizes the protein transfer scenario from chloroplast to nucleus in A. thaliana.
We have further grouped transferred intrinsically disordered proteins in gene clusters (Figure 4), reminiscent of the ancestral bacterial operons, finding that the fts, inf, acc, psa, rpl and ycf gene clusters encode more frequently for disordered proteins (40 -58% of disorder). These genes are involved in cell division, translational initiation and acetyl-CoA carboxylase pathways, or photosystem I, large ribosomal subunits. In contrast, the atp, chl, ndh, men, pet, psb and rps gene clusters, which encode for ATP synthase subunits, protochlorophyllide reductase, NADH-plastoquinone oxidoreductase subunits, succinyl or naphtoate synthase, cytochrome b 6 /f, photosystem II subunits and ribosomal small proteins, contain less disordered proteins (8 -25% of disorder). These observed differences do not appear to be related to protein length, as the average length of intrinsically disordered proteins was found to be 390 aa, a similar value to that of nondisordered proteins (391 aa).

Gene ontology annotations of disordered proteins of plastidic origin
In order to put in perspective the previous observations we investigated the annotated function of disordered proteins in the 12 plant species studied by using the Gene Ontology (GO). In the course of this examination a protein was considered disordered if it contained a contiguous stretch of predicted disordered residues of L ≥ 30 amino acids. The analysis revealed that disordered proteins encoded in nuclear genes assumed to be of plastidic origin were enriched in 29 biological processes (P), 39 cellular components (C) and 13 molecular functions (F) GO categories with corrected p-values < 10E-5 (see Additional . Colour scale goes from blue (order) to red (disorder); magenta is shown in cases were not all members within a group of paralogous genes encode disordered proteins. Absent orthologues are painted in grey. RUBISCO small subunit is marked with an asterisk. B) Scatter plot of disorder frequencies in complete plant nuclear proteomes (X-axis) versus disorder frequencies in the nuclear proteome fraction with chloroplast origin (Y-axis). C) Linear plot of disorder frequencies in complete plant chloroplast proteomes (X-axis) versus disorder frequencies in the nuclear proteome fraction with chloroplast origin (Y-axis). A protein is considered disordered if it contains a contiguous stretch of predicted disordered residues of L ≥ 30 amino acids. file 5: Table S4). As to the cellular component, we found that these proteins were mainly associated to "plastid" (4.60E-43) and "chloroplast" classes, which supports our homology-based selection of chloroplast-transferred genes. The most significant association among specific biological processes was with "cellular nitrogen compound biosynthetic process" (1.10E-13), including cofactor, heterocycle and tetrapyrrole biosynthetic processes. Finally, a few molecular functions were found to be associated to these disordered proteins, such as "structural constituent of ribosome" (8.01E-09) and "ATPase activity" (4.35E-06). These reported corrected p-values are relative to A. thaliana, which is probably the best-annotated plant genome for its role as a model organism. Altogether, these results suggest that disordered transferred proteins as a whole are not strongly linked to any one function. Moreover, nuclear-encoded genes still maintaining a copy in the plastid chromosome were mainly associated to GO cellular components "ribosome" (5.43E-30) and "ribonucleoprotein complex" (2.24E-26). Among biological processes, they were mainly associated to "gene expression" (5.35e-36) including "translation" (2.61E-25), "transcription" (5.97E-14) or "RNA biosynthesis" (9.8E-11). Finally, at the level of molecular function, these proteins were found to be annotated as "structural constituent of ribosome" (2.95E-32), "structural molecule activity" (1.56E-28), "DNA-directed RNA polymerase activity" (2.18E-15) or "NADH dehydrogenase activity" (2.45E-7) (Figure 3).
We also reviewed the annotated function of nondisordered proteins of chloroplast origin and the results were more compelling, as this set of proteins is more homogeneous (see Additional file 6: Table S5). Among biological processes, several translation-related annotations were considerably associated, such as "ribosome biogenesis" (1.28E-31). These agree well with the most significant cellular component found, which "cytosolic large ribosomal subunit" is (1.05E-46). In addition, the strongest association found at the level of molecular function was "structural constituent of ribosome" (4.59E-45).   Additionally, the functions of intrinsically disordered nuclear-encoded proteins were also analyzed (data not shown). Among biological processes the most notable annotations were related to "regulation", including "regulation of nucleobase" (1.96E-267), "regulation of nitrogen compound" (2.48E-266), "regulation of macromolecule biosynthetic process" (5.94E-265) or "regulation of RNA metabolic process" (9.61E-265). At the level of cellular component, significant associations were found with "nucleus" (7.63E-162), "membrane-bound organelle" (5.78E-144) and "organelle" (8.79E-129). These annotations correspond well with those of molecular function categories, such as "nucleic acid binding transcription factor activity" (1.19E-260), "nucleic acid binding" (1.38E-250) or "DNA binding" (2.23E-209). Overall, these functional classes match those reported for eukaryotes in general [5].

Disorder in ribosomal proteins
An in-depth analysis of chloroplast ribosomal proteins was performed with the aim of better understanding the evolution of protein disorder in plants. These proteins were selected for three reasons: i) they are the largest gene cluster transferred to the nuclear genome; ii) they are part of a highly conserved and essential cellular system, and iii) they were highlighted in the GO annotation study described above. The idea was to compare A. thaliana (eudicot) and O. sativa (monocot) proteins with their orthologues in prokaryotic ribosomes (4 Archaea, 3 Gram +, 4 cyanobacteria, 7 eubacteria and 4 proteobacteria). For details see Materials and Methods and Additional file 7: Tables S6A and S6B. We have calculated that 30% and 65% of these proteins are intrinsically disordered in chloroplast 30S and 50S subunits, respectively. The data show that protein disorder is not uniform across bacteria species. There are instances where a protein contains long disordered segment in some species but not in others. It is worth mentioning that no differences were found between the two plant species.
Figures 5A/C and 5B/D colour ribosomal proteins that were predicted to be disordered in our analysis, (and observed experimentally in some cases as described in [3], in at least one prokaryote (top) and one plant chloroplast (bottom) genome, respectively. It can be observed that the disorder degree of the small (30S) subunit does not increase in chloroplast ribosomes ( Figure 5B). On the L30 L5 L31 S13 S11 S6 S10 S21 S9 S20 S5 B L4 Figure 5 Distribution of disordered proteins on the bacterial (A,C) and chloroplast (B,D) ribosome (mapped over PDB entries 1JOO, 1VQ8, 3BBN and 3BBO, respectively). Panels A and B correspond to the 30S subunit, C and D to the 50S. Disordered proteins in bacterial ribosomal subunits are highlighted in pale yellow, pale blue, light blue, dark blue, orange, green, magenta and pink. Their chloroplast orthologues in Arabidopsis thaliana are marked in the same colour. Additional proteins found to be intrinsically disordered in chloroplast 30S and 50S subunits are highlighted in red. Numbers following S and L identify small and large subunit proteins, respectively. The disordered protein L7/L12 in the chloroplast 50S subunit is not marked because of it is absent in the structural data retrieved from the PDB. The average number of intrinsically disordered proteins (IDP) calculated for each ribosomal subunit is written below (for details see in Additional file 7: Table S6). The threedimensional cartoons were drawn using PyMol 1.4.1 (Schrodinger LLC).
contrary, the disorder increases notably in the chloroplast large (50S) subunit ( Figure 5D). An interesting feature that might explain this finding is that the majority of L-proteins are nuclear encoded (33/42) being this ratio lower (12/22) in the case of S-proteins. Interestingly, in certain plant genomes (i.e., O. sativa, S. bicolor, Z. mays, P. trichocarpa, V. vinifera, G. max, P. patens) it was found that some ribosomal proteins are encoded by both nuclear and plastid genes, and in the majority of cases the resulting protein products are identical.
In the small subunit, we found that chloroplast proteins S10, S11, S13 and S20 have acquired disorder with respect to their prokaryotic orthologues, but have also lost disordered segments observed in bacteria (for instance in S2, S3 and S18). Note that plant S10, S13 and S20 protein sequences are much longer than their prokaryotic counterparts (see in Additional file 8: Table S7), and this might explain the gain of disordered segments. Overall, there is not a clear net gain of disorder in this subunit (see in Additional file 7: Table S6A). Within the large subunit, L1, L6, L7/L12p, L9, L11, L13, L17, L18, L24, L27, L28, L34, L35 and L36 proteins gain disorder in the chloroplast. With the exception of L36, all these are nuclear-encoded.

Discussion
The analysis of 12 plant proteomes reveals a similar occurrence of disordered proteins to that found in other eukaryotic organisms [1]. Therefore, there is no clear separation among animals, yeast and plants in terms of the total amount of predicted disordered segments. Nor clear differences were observed among different plant species belonging to bryophyta, chlorophyta and vascular plant, or among eudicots and monocots.
The amino acid composition of disordered segments in plants corresponds well with that reported for other eukaryotes [3,5,11], which can be defined by a low frequency of bulky hydrophobic residues, which normally form the core of a folded protein, and high frequency of polar residues contributing to net charge. The minor presence of cysteine residues within disordered regions was also a characteristic feature observed in either chloroplast, mitochondrial or nuclear proteins, which fits well with other predicted disordered protein profiles [5]. This finding supports that these features in disordered protein regions are stable during evolution. On the other hand, the distribution of disordered regions along the complete protein sequence was slightly higher in the internal parts than in the terminal parts of proteins. This feature was common for all the plant proteomes investigated and no differences were found among different species. This observation differs from the data obtained from protein 3D structures from the Protein Data Bank [31]. These authors reported that the fraction of disordered residues is more abundant in the terminal parts (72%), constituted by 40 residues near to the N-terminal and the C-terminal compared with the middle part (all other residues).
Interestingly, a survey of chloroplasts and mitochondria revealed significant differences concerning the occurrence of disordered regions when compared with the nuclear genome. The percentages calculated in these organelles are in the order of magnitude of those determined in Archaea and bacteria [1]. These data are in agreement with the bacterial origin of genes coding for these proteins. We also observed differences concerning the distribution of disordered regions in the protein chain.
It has been suggested that between 800 and 2,000 genes in the Arabidopsis thaliana genome might come from cyanobacteria, with a majority of proteins included in the functional category of biosynthesis and metabolism [32][33][34][35]. Furthermore, the analysis of 15 sequenced chloroplast genomes revealed 117 nuclear-encoded proteins that are also still present in at least one chloroplast genome [16]. Based on these reports we evaluated the degree of disorder in both nuclear-encoded proteins, which were transferred from the plastid to the nuclear genome, and those transferred to the nucleus that also still conserve a copy in the chloroplast genome. Our results indicate that transferred proteins acquired disorder with a frequency similar to that of nucleusencoded proteins. During evolution, organelles export their genes to the nucleus, but many of these proteins are imported to the chloroplast, with the help of transient peptides and protein-import machinery, to carry out their function. This gain of disorder can be hypothesized to be an advantage during the import-pathway across a double-membrane barrier. However, these disordered segments are not preferentially associated to transient peptides localized in the N-terminal region. Indeed, they were found to be slightly more abundant in the internal region of the protein chain. Moreover, those transferred protein coding-genes that maintain a copy in the chloroplast genome exhibit much lower disorder than those that have lost the plastid copy, similar to proteins encoded by chloroplast or bacterial genes. This fact might be revealing a selection pressure during evolution. These proteins are mainly involved in translation, transcription or RNA biosynthesis, being structural constituents of the ribosome and the ribonucleoprotein complex. The disorder in proteins encoded by ancient chloroplast genes but currently in the nucleus follows the order bryophyta < vascular plants < chlorophyta. In this context, the data suggest that the level of disorder introduced into plastid proteins that have moved to the nuclear genome has increased during evolutionary time, but further investigations will be necessary to clarify this issue.
The gain or loss of disorder in transferred proteins might be to some extent a stochastic process, since orthologous copies found in different plant species do not necessarily conserve disordered segments, despite presumably carrying out similar functions. This observation is in agreement with the finding that gene transfer events from the chloroplast to the nuclear genome occur much more frequently than generally believed, contributing significantly to genetic variations [35]. In this respect it is also noted that disorder distribution in ribosomal proteins among bacterial species appears rather at random (Additional file 7: Table S6).
Non-folding unstructured proteins and regions might be expected to change more rapidly during evolution than structured proteins because buried amino acid residues are highly constrained while disordered regions are not constrained by the structure [11]. It is believed that disordered proteins do not exist as a single structure but rather as a conformational equilibrium of states, which interconvert into each other over a range of time scales. This feature can be an evolutionary advantage for adaptation, for instances, under stress conditions. Additionally, intrinsically disordered proteins could be more susceptible to proteolytic degradation in vitro. The classical PEST hypothesis states that the presence of segments rich in Pro, Glu(Asp) and Ser/Thr flanked by Arg/Lys residues in proteins correlates with a short lifetime in the cell [36,37]. Accordingly, the fact that a group of proteins related to the ribosome biogenesis preserved its ordered character when transferred to the nucleus could be explained by this critical role within the protein synthesis machinery which should be maintained.
On the other hand, around 25% of chloroplast ribosomal proteins transferred to the nucleus are predicted to be intrinsically disordered in our analysis. In this respect it has been argued that flexibility favours the structural assembly of components of large complexes such as those involved in ribosome and therefore such characteristic should be prevalent in certain ribosomal proteins [38]. Moreover, RNA-binding proteins usually contain unstructured regions as is the case of the ribosomal protein L5, which is reported to be associated with 5S rRNA [39]. Our results also indicate that intrinsic disorder is a well-conserved character in some ribosomal proteins. This is the case of L4 and L15, predicted to contain unstructured segments in all the bacterial and plant proteomes analysed. Ribosomal protein L4 is localized near the peptidyl transferase center of the bacterial ribosome [40] and displays significant RNA chaperone activity [41]. The L15 protein is involved at later stages during assembly [41].
The comparison of disorder between bacterial and chloroplast ribosomal proteins unveiled a disorder increase in the chloroplast large 50S subunit, where proteins are in average 55 residues longer, as previously reported by Yamaguchi and Subramanian [42], and the majority are produced by nuclear genes. This finding contrasts with the data obtained with the whole proteome, which show no differences in length between disordered and non-disordered proteins. In the case of the small 30S subunit such differences were not so clear, probably due to the higher content of chloroplastencoded proteins, which most of them are predicted to be non-disordered. These results support our hypothesis that proteins encoded in the nuclear genome are more likely to stochastically acquire disorder. On the other hand, however, we cannot preclude that differences in rRNA composition between chloroplast (23S, 5S and 4.5S) and bacterial (23S and 5S) large 50S ribosomal subunit could also explain the gain of disorder observed in this subunit [43,44].
Differences in the genetic machinery between plastids (prokaryotic) and nucleus (eukaryotic) could also help to explain our observations. When plastid genes reach the nucleus they move from a genetic apparatus that is compact, operon-harbouring and intron-poor, to one that is more complex, operon-splitting and intron-rich [45]. While the gain of disorder is thought to be advantageous or neutral in many cases, there must be selective pressures that put restrictions to this apparently random process, as is the case of the chloroplast RUBISCO small subunit protein, a nuclear-encoded protein with a plastid origin, which was found to be ordered in most of the plant proteomes investigated (see Figure 2).
The comparison of 3D structures of bacterial and chloroplast ribosomal subunits revealed the localization of the extra disordered proteins. For instance, S11 is localized in the mRNA path, next to the intrinsically disordered S21, which directly interacts with the 5' untranslated region of the mRNA [46]. In the ribosomal 50S subunit, L24 and L29 are localized surrounding the polypeptide tunnel exit site. It is worth noting that some of these chloroplastic disordered proteins are normally found in cyanobacteria (see in Additional file 7: Table S6), but in some cases are unstructured in gram-positive bacteria and not in cyanobacteria (i.e. S9, L29 and L31). This might be related with the fact that more Arabidopsis proteins branched with their homologues from gram-positive bacteria (Mycobacterium) than did with cyanobacteria (Prochlorococcus, Synechocystis). This has been interpreted as if the Arabidopsis lineage acquired genes specifically from grampositive bacteria subsequent to its divergence from the yeast lineage [16].

Conclusions
Taken together, our chloroplast-based analyses demonstrate that disordered segments are acquired by proteins most probably due to the process of nuclear integration during plant evolution. However, we observed that some parts of the ancestral chloroplast and mitochondria organelles present in eukayotic cells are being preserved from acquiring disordered segments, probably due to functional constraints and evolutionary pressure.
genes and transferred to nuclear genome. A) biological process (P) GO categories; B) cellular components (C) GO categories; C) molecular function (F) GO categories.