**3. Discussion**

The structural propensities and association features of IDPs and disordered protein domains are still poorly understood compared to those of well-folded protein regions. The difference is especially important for proteins, such as LrtA, which is formed by two distinct domains (i.e., N-LrtA and C-LrtA) with roughly the same sequence length, but with completely distinct conformational features. In a previous work [12], we have hypothesized that N-LrtA has a distinct folding topology that is in common with other members of the HPF protein family [5,11], whereas C-LrtA was predicted to be unfolded. In the present work, we have tested that hypothesis and we have found evidence that isolated C-LrtA is an IDP. The intrinsically disordered nature of C-LrtA was suggested by several pieces of evidence: (i) the lack of dispersion in NMR spectra (Figure 1); (ii) the shape (Figure 2A) and the deconvolution of far-UV CD spectrum; (iii) the absence of all-or-none co-operative transitions in the thermal and chemical denaturations (Figure 2B,C); and (iv) a model of the structure without the His-tag obtained with MD simulation (Figure 3).

It could be thought that the observed C-LrtA self-association is non-specific, that is, the protein has solvent-exposed hydrophobic patches which induce highly unspecific, self-association. However, there are two pieces of evidence that suggest that association was not the result of random solvent-exposed hydrophobic residues. First, the fact that Tyr residues were implicated in the oligomerization indicates that only regions containing those residues were involved. Second, if self-association was to be unspecific very large high-molecular species should be observed in some of the techniques used; in contrast, the highest molecular-weight species is observed to be an octamer (in SEC experiments) and glutaraldehyde cross-linking showed the presence of higher molecular-weight species only at very long incubation times (Figure S3C). Finally, it is interesting to note that we have shown that the

intact protein, which also self-associates, did not have any large amount of close solvent-exposed hydrophobic patch at physiological pH [12].

Although it is devoid of secondary and tertiary structures, C-LrtA is involved in quaternary contacts, involving different orders of self-associated species, as indicated by several of the techniques used. In general, IDPs intervene in protein-protein contacts [13], but only a few are reported as self-associated species in solution, while keeping their disordered nature. In C-LrtA, some of the four tyrosine residues were involved in its self-association, as judged by the changes in the quenching parameters as the concentration of protein was increased. In the parental, whole protein, tyrosine residues were already found to participate in the oligomerization interface [12]. With the new results in hand, we suggest that some of the four tyrosine residues in C-LrtB were responsible for the self-oligomerization of the whole protein. Then, whereas tyrosine residues in N-LrtA (residues 1–101) seem to contribute to the rigidity of the β-sheet scaffold, tyrosine amino acids in C-LrtA (residues 102–191) seem to be responsible for the quaternary structure of the intact protein. It is important to note that the detected order of the self-associated species in C-LrtA varied among the different biochemical and hydrodynamic techniques used (as it happens for the whole LrtA [12]) indicating the presence of different oligomerization equilibria. That is, the self-association did not involve the simple dimer-monomer equilibrium, but rather a sequence of different order equilibria.

The exact biological significance of self-association in LrtA remains unknown, and it is still a matter of debate. In fact, in spite of similar structural organization and high sequence homology among the members of the HPF family, their functions during stress responses are very different in the organisms to which they belong to [4,7–10,30]. There are some examples of oligomeric HPFs reported in the literature: for instance, the short *Vibrio choleras* HPF is a dimer, whose dimerization occurs through Zn ions at one of the β-strands of the β-α-β-β-β-α fold; however, it is not known if such dimerization is due to the crystallization process [11]. On the other hand, the HPF of *Staphylococcus aureus* is also a member of the long HPF family, but its C-terminal region is shorter (60-residue long) and it is folded [5]. The protein is a dimer, and its C-terminal region is responsible for this dimerization; furthermore, interactions of the dimeric C-terminal region with ribosomes are responsible for ribosome dimerization. Thus, we hypothesize that in LrtA the oligomeric, disordered C-LrtA domain might be responsible for the dimerization of 70S particles (whose abundance in the cell has been associated with the presence of LrtA [3]) during stress conditions. The fact that C-LrtA is disordered (in contrast to that of *Staphylococcus aureus* HPF [5]) could provide the advantage that LrtA may be bound not only to the 70S particles, but also to other macromolecules, regulating several cyanobacterial processes triggered by stress conditions.

Evidence of the possibility of self-association and supra-molecular order in IDP are starting to be mounting. One of the first reported examples of oligomeric IDPs was the "fuzzy" dimer formed by the cytoplasmic domain of the T-cell receptor zeta subunit [31,32], although this putative dimer was later shown to be a monomer under a wide range of conditions, and its oligomerization was attributed to non-ideal protein-column-resin interactions [33]. However, other studies have used a plethora of biophysical and biochemical techniques to unambiguously show the existence of intact self-associated IDPs. For instance, there are reports describing oligomeric plant IDPs [34]; dimeric proteins in the disordered *umu*D gene products [35]; oligomeric mitochondrial IDPs [36]; oligomeric IDPs which bind to the Polycomb complex [37]; disordered, oligomeric acid-rich proteins of rod photoreceptors [38]; and disordered oligomeric oncogen products [39]. Many of these homoligomeric interactions are deposited in the MFIB database (http://mfib.enzim.ttk.mta.hu) [40], together with others involving hetero-oligomer assemblies, where in all cases mutual folding of the IDP chains occurs. The presence of such a quaternary (homo or hetero-oligomeric) structural organization for an IDP has been explained as due to multiple transient interactions or long-range contacts, which yield a fuzzy self-associated species [41–43], involving different degrees of organization in the association process [44].
