Molecular Evolution of RAMOSA1 (RA1) in Land Plants

The BLASTp searches recovered SUP sequences along most of the embryophyte species, except from M. polymorpha and S. moellendorffii. To identify the origin of RA1 and understand its evolution, we reconstructed a phylogenetic tree with amino acid sequences obtained from genomes of grasses and other embryophytes. For that, we used SUP sequences from non-seeds plants (P. patens and C. purpureus) as outgroups. The recovered tree topology, and the presence of a single copy sequence of the grass P. latifolia suggests that RA1 originated from two successive duplications of SUP that occurred at the split of the BOP and PACMAD clade (Figure 1). As a result of these duplications, three lineages of sequences were generated; one of them includes RA1 sequences, and the other two are here named arbitrarily RA1-like A and RA1-like B. In this tree, the evolutionary relationships among RA1 and RA1-like remain unclear.

To gain resolution on the relationship between RA1 and RA1-like, we reconstructed a new phylogeny with amino acid sequences obtained from 16 genomes of grass species using as an outgroup the SUP sequence from the monocot J. ascendence (Figure 2a, Figures S2, and S3). The obtained tree topology suggests the possibility that (1) RA1 was originated after a second duplication around the split of the BOP and PACMAD clades, (2) RA1 is sister to the RA1-like B proteins, (3) RA1 was lost in the BOP clade and currently is present in the PACMAD clade, (4) RA1-like A and RA1-like B are found in species of both BOP and PACMAD clades, and (5) species-specific duplications and loss of some copies of RA1 and RA1-like suggest a complex evolution of these molecules during the diversification of the grasses (Figure 2a and Figure S4). Overall, SUP, RA1, and RA1-like are single-copy genes, except for the two RA1 variants found in S. bicolor, as previously described [2] (Figure 2a).

Genome annotations indicate that SUP is on Chr3:8242256…8243372 of the Arabidopsis genome, whereas RA1, RA1-like A, and RA1-like B are on maize chromosomes Chr7:114958642…114959398, Chr5:68312034…68312578, and Chr10:103427449…103428364, respectively. Chromosomal collinearity visualization maps indicate that Arabidopsis chromosome 3 lacks synteny with maize chromosomes 7, 5, and 10 (Figure S5). In contrast, RA1 and RA1-like genes are placed in syntenic chromosome blocks among maize, Setaria, and rice (Figure S6). These results suggest that RA1 and RA1-like are syntenic paralogs.

The secondary structure properties of the SUP/RA1 members are poorly characterized [33,46]. Only the zinc finger C2H2-type of SUP has had its secondary structure experimentally solved by NMR [33]. So, to gain perspective on protein structure among SUP, RA1, and RA1-like, secondary structures predictions were carried out using the PsiPred Workbench (http://bioinf.cs.ucl.ac.uk/psipred) and its PsiPred 4.0 tool [29]. We also used the server’s tool DisoPred 3.1 [30] to predict regions that cannot be assigned a known secondary structure, which are termed as disordered regions. The same algorithm also predicts regions with high probability of binding other proteins. These properties are graphically shown in Figure 4 and Figure S10 and Table S4 as a function of the residue number/name. Overall, the N-terminal region of SUP, RA1, and RA1-like is predicted as disordered with protein binding affinities. Allocated within the N-terminal disordered segment are common motifs shared between lineages, such as Motif 4 and 5 (Figure 2b). The N-terminal disordered region is followed by a region defined as a coil that continues with the C2H2-type domain. The structural predictions suggest the zinc finger to be formed by a β-sheet (two consecutive β-strands) followed by an α-helix (Table S4). In general, the C-terminal of RA1 and RA1-like is presented as a disordered region with protein binding affinities. Allocated within the C-terminal disordered segment is Motif 3 shared by RA1-like A and RA1-like B (Figure 2b). In addition, we found stabilized ordered segments, such as two to three α-helices. These ordered segments are mostly correlated with the allocation of Motif 2 (Figure 4 and Figure S10 and Table S4).

The zinc finger of SUP (1NJQ) is a structure resolved by NMR (20 models), which has 75% sequence identity with the zinc finger of RA1 (Figure S11a). There is no experimental structure of RA1. Given differences in sequence identity between SUP and RA1, we performed 1μs molecular dynamics simulations of each protein zinc finger domain to evaluate differences in the structural stability of these variants. The RMSDs of these simulations as a function of simulated time (Figure S11b) show small departures from their initial values (average RMSD values lower than 0.5Å). This fact is associated with the robustness of the structure of the zinc finger domain of both proteins and the proximity of the initial models to the equilibrated structures. We assumed that the simulated time in the period [0.8μs, 1.0μs] is adequate to calculate equilibrium properties of the systems since the RMSDs of the simulations of both proteins show small fluctuations without any appreciable definite tendency.

Analysis of the time evolution of the secondary structure of SUP and RA1 peptides showed the classical pattern of a zinc finger with a β-sheet composed of two strands (Y47T48-F55E56 and Y46T47-F54E55 in SUP and RA1, respectively) and a turn stabilized by H-bonds (S50F51C52 and G49Y50C51 for SUP and RA1), followed by an α-helix (A59:H69 and A58:R72 for SUP and RA1) (Figure 5a, Table S4). The α-helix structural motif is four residues longer for the case of RA1 (an additional turn). Beyond the α-helix structural motif, the SUP peptide presents a turn of a 310-helix (D72:R75) (Figure 5a, Table S4). We observed an additional structure with fluctuating characteristic 310-helix, α-helix, and turns in RA1 peptide (I76:Y80) (Figure 5a, Table S4).

Previous work suggests that the change of Ala (A) to Gly (G) in the QALGGH motif of the RA1 zinc finger could lead to an enhanced mobility of the residues in the α-helix [2]. However, such a hypothesis has not been tested yet. Here, we performed a comparative RMSF analysis between RA1 and SUP to test whether the G affects the structure dynamic of the zinc finger α-helix (Figure 5b). The RMSF of the SUP and RA1 amino acids were calculated in the equilibrium interval of the simulations for the backbone atoms of each residue. We found that, at this site, the amino acids of RA1 show smaller fluctuations than the respective amino acids of SUP (Figure 5b). One could argue that the effect of the substitution of A by G may not be local; however, we found that the RA1 amino acids show smaller fluctuations along all the extensions of the canonical zinc finger domain. The existence of a longer α-helix motif in the RA1 peptide, followed by the additional helix pattern, causes the region of small fluctuations or relative rigidity of the peptide to extend well beyond the canonical zinc finger motif (Figure 5b). These results indicate that substituting A for G does not impact the three-dimensional structural mobility at that site, nor does it affect the overall structure of the zinc finger.