Positive selection analysis highlights key positions in plant PP2A regulatory subunits

ABSTRACT

A versatile hub for cellular control, the eukaryotic protein phosphatase 2A (PP2A) enzyme family is thought to achieve specificity through combinatorial complexity. Phylogenetic analysis has revealed that expansion of PP2A gene families resulted from whole genome duplications followed by non-random gene loss, and selection analysis suggests that retention of B56/PPP2R5 gene family members after genome duplication events was driven by functional diversification. Here we identify the sites at which positive selection is detected in the plant B56 gene family, and discuss the significance of selection at these positions in the context of PP2A holoenzyme structure. The pattern of positive selection observed in the B11 subclade is distinctive, and suggests selective pressure on interactions with substrates and the enzymatic core.

Results and discussion

Canonical heterotrimeric PP2A holoenzymes contain catalytic (C), regulatory (B), and structural/scaffolding (A) subunits; these complexes regulate diverse cellular transactions by dephosphorylating target proteins (reviewed in refs.1-3). We recently reported the results of our phylogenetic analysis of PP2A gene families encoding A, B and C subunits,⁴ with particular focus on the 3 regulatory B subunit types known in plants: B55, B56 and B72 (reviewed in refs.1,2). Although each B subunit type is found in heterotrimeric PP2A complexes, the gene families encoding these proteins are unrelated and the protein products are structurally and biochemically distinct. Members of the B55 family are β-propeller proteins⁵ and B72 family members are calcium-binding proteins,^6-9 while B56 proteins comprise a conserved core of α-helical Huntingtin/Elongation factor 3/A subunit/TOR (HEAT) repeats.¹⁰ The overall pattern of PP2A gene family evolution is one of expansion driven by paleopolyploidization events, followed by non-random gene loss.⁴ Interestingly, each regulatory B subunit family includes one clade that deviates from this overall pattern by showing little or no expansion. We also observed several lineage specific losses among plant PP2A gene families; in the B56 family, for instance, the B5/8 clade has been lost in grasses and the BØ clade is absent in plant species that do not support beneficial microbial associations (Fig. 1).^4,11,12

Figure 1.

Overall Topology of the Seed Plant B56 Gene Family. Bayesian analysis was used to construct a phylogenetic tree for the B56 gene family. The tree at left shows a simplified version of the tree constructed through analysis of B56 cDNA sequences from 34 plant species (Chlamydomonas reinhardtii, Physcomitrella patens, Selaginella moellendorffii, Ceratopteris richardii, Adiantum capillus-veneris, Gnetum gnemon, Podocarpus macrophyllus, Pinus taeda, Picea abies, Amborella trichopoda, Spirodela polyrhiza, Phoenix dactylifera, Musa acuminata, Sorghum bicolor, Brachypodium distachyon, Oryza sativa, Aquilegia coerulea, Nelumbo nucifera, Solanum lycopersicum, Mimulus guttatus, Utricularia gibba, Vitis vinifera, Fragaria vesca, Populus trichocarpa, Citrus sinensis, Medicago truncatula, Theobroma cacao, Carica papaya, Cleome/Tarenaya hassleriana, Eutrema salsugineum, Brassica rapa, Capsella rubella, Arabidopsis lyrata, and Arabidopsis thaliana; for details of analysis see ref.4). The first column shows the taxa in which each clade is found (Present) and the second column indicates lineage-specific losses observed for that B subunit clade (Absent). The final column lists the Arabidopsis genes belonging to each clade. For simplicity, 3 clades present in gymnosperms (2 B5/6/8/9/10-like clades and one B11-like clade) and all clades found in non-seed plant species are not shown. For detailed phylogenetic trees including all B56 sequences analyzed, see Supplemental Figures 2 and 3 in ref. 4.

Several clades within the B56 gene family in plants (Fig. 1) show evolutionary evidence of functional diversification.⁴ We used the branch-site likelihood method¹³ to map the amino acid positions (sites) at which we could detect positive selection within the B56 protein sequence. Among B56 gene families in seed plant species, the B11 clade exhibited the largest number of sites of positive selection (9; Table 1 and Fig. 2). Several other clades showed evidence of positive selection at 2 (B6 and B3/4/7 clades) or 3 (BØ, B6/9/10 and B5/8/6/9/10 clades) sites. Most of these sites of positive selection mapped to positions of lower sequence conservation (< 50% sequence identity); only 5 of the 24 sites mapped to more highly conserved positions (55 – 75% sequence identity; Table 1 and Fig. 2). Three sequence positions showed evidence of positive selection in 2 distinct clades.

Table 1.

Sites of Positive Selection in the B56 Regulatory Subunit Family Selection analysis of the plant B56 gene family was performed using the branch-site likelihood method¹³ as described previously.⁴ Analyses were performed on the amino acid sequence tree (the plant sub-tree shown in Supplemental Fig. S3 of ref.4), and the codon alignment was generated from the amino acid alignment used for the phylogenetic analysis and the nucleotide coding sequence as input into TranslatorX.²² Branch points at the bases of clades containing BØ and A. thaliana isoform lineages were tested. Initial omega values for alternate hypotheses were 1.5 for tests of branch points all clades except the B3 clade (initial omega of 15). All branches were analyzed in independent runs.

Site number	B56 Clade	B11 positiona	BEB valueb	Amino acid identity (%)c
1	B5/8/6/9/10	72	0.977	–
2	BØ	93	0.979	55
3	B3/4/7	123	0.986	–
4	B11	126	0.955	–
5	B5	127	0.981	45
6	B6/9/10		0.964
7	B3/4	133	0.972	–
8	B11	138	0.991	45
9	BØ		0.995
10	B11	148	0.954	45
11	B11	152	0.98	45
12	B11	155	0.987	64
13	B11	157	0.993	64
14	B6/9/10	161	0.977	–
15	B11	302	0.986	–
16	B6/9/10		0.968
17	B11	320	0.97	–
18	B5/8/6/9/10	340	0.985	45
19	B11	364	0.979	73
20	B6	401	0.97	–
21	B5/8/6/9/10	405	0.952	–
22	B3/4/7	413	0.99	64
23	B6	441	0.956	45
24	BØ	447	0.993	–

^aAmino acid residue in the predicted Arabidopsis B11 (AT5G25510) protein sequence

^bBayes Empirical Bayes value; the criterion for reporting positive selection at any position was a BEB value ≥ 0.950 (and a log likelihood ratio test result of p < 0.05 for positive selection on the clade overall, as reported in ref. 4).

^c–, Less than 40% amino acid sequence identity in Fig. 2 alignment.

Figure 2.

Sites of Positive Selection within the Core Domain of Plant B56 Regulatory Subunits. Amino acid sequences of B56 domains for all 9 Arabidopsis B56 isoforms plus the Medicago truncatula BØ (PP2AB’1¹²) and human B’γ-1 were aligned using ClustalX.²⁰ Positions at which positive selection was detected in the B11 clade and in other B56 clades are indicated by red and yellow circles, respectively, above the sequence alignment; boxes in the alignment indicate the clade[s] in which selection was detected. Highlighted amino acids in the human B’γ-1 sequence are points of contact with the A (green) or C (blue) subunit,¹⁰ or with the SLIM motif from RepoMan and BubR1¹⁹ (red). The consensus logo was auto-calculated in Jalview.²¹ Residues in the consensus logo are sized and shaded by percent identity, with the smallest/lightest logo letters indicating 45% identity, and increasing letter size/shading indicating increasing identity (in increments of 9.1%).

We used the structural modeling programs PHYRE2¹⁴ and ProtMod (http://ffas.burnham.org/protmod-cgi/protModHome.pl) to generate predicted structures for A. thaliana B11 and the Medicago truncatula BØ isoforms, and PyMOL (The PyMOL Molecular Graphics System, Version 1.8.6.0 Schrödinger, LLC) to compare the predicted structures to the crystallographically determined structure of human B’γ-1. B56 regulatory subunits carry relatively short and non-conserved amino- and C-terminal regions that flank a highly conserved core, the B56 domain.¹⁵ This core domain comprises 8 HEAT repeat motifs (each consisting of a pair of α helices connected by a short ‘intra-repeat’ loop), followed by an unpaired α helix.¹⁰ The alignments exhibit small gaps in loops with variable lengths, but neither modeling program predicted obvious structural differences between the B11 or BØ protein and the human B’γ-1 protein. Superimposition of the predicted B11 structure with the human B’γ-1 structure yields a very tight alignment through the entire B56 domain (dark vs. light gray in Fig. 3A). Superimposition of the predicted BØ structure yielded nearly identical results (data not shown).

Figure 3.

Mapping Sites of Positive Selection onto the PP2A Complex. Sites at which positive selection was detected in the B11 clade (red) and in other B56 clades (yellow) are highlighted in a structural model for the PP2A complex. A structural model for the B56 domain of the Arabidopsis B11 protein (light gray; residues K81 – K466 in Fig. 2) was generated using FFAS ProtMod and superimposed on the crystal structure of human B’γ-1 (dark gray) in a heterotrimeric PP2A complex (PDB ID 2NPP;¹⁰) using PyMOL. Ribbon diagrams of the subunits (panel A) show the amino termini (and first α-helical HEAT repeats) of the B56 structures at top, with the C-termini at bottom. The scaffolding subunit is shown in green, and the catalytic subunit is shown in blue. The long intra-repeat loop of HEAT repeat 2 in the B56 subunit, which includes several sites of positive selection highlighted in red and yellow, projects toward the enzyme's active site. A surface representation of the complex (panel B, which is rotated slightly relative to panel A) highlights the exposed positions of several sites of positive selection in HEAT repeats 5 – 8.

To determine whether selection affects residues that might play specific roles in the context of the heterotrimeric PP2A holoenzyme, we mapped the sites of positive selection onto this structural model for the B56 subunit in the PP2A complex (Fig. 3). More than half of the amino acid positions identified in our positive selection analysis (13/24) mapped to HEAT repeats 1 and 2, in the amino-terminal portion of the B56 domain. Five of those positions were predicted to fall in the extended intra-repeat loop in HEAT repeat 2 that projects toward the active site in the heterotrimeric enzyme complex; residues in this loop are required for PP2A-B56 activity toward tyrosine hydroxylase.¹⁶ One site mapped upstream of HEAT repeat 1, in the amino-terminal region not included in the structural model. The remaining 10 positions mapped to HEAT repeats 5, 6, 7, 8, and the C-terminal α helix.

None of the 24 sites of positive selection mapped to the highly conserved third and fourth HEAT repeats, which contain the binding site for a short linear interaction motif (SLIM) recently identified in numerous substrate and scaffolding proteins recognized by B56-containing PP2A complexes^17-19 (Fig. 2). However, 5 sites (10, 11, 13, 14 and 18) are predicted to lie at or immediately next to positions in the inter-subunit contact interfaces of the holoenzyme.¹⁰ Strikingly, all but one of these sites (number 18) map to the extended intra-repeat loop sequence in HEAT repeat 2 (Figs 2 and 3A), which has previously been linked to recruitment of specific substrates.¹⁶ Three of these sites were identified in the B11 clade, suggesting that the B11 clade has experienced unique selective pressure in this region.

The 10 sites that map to the region from HEAT repeat 5 to the C-terminus exhibit a wider phylogenetic distribution, affecting the BØ, B6, B11, B3/4/7, B6/9/10 and B5/8/6/9/10 clades. Only one of these sites (number 18) maps close to a predicted point of contact in the core of the heterotrimeric complex, while 5 of the remaining 9 (numbers 17, 20, 21, 23 and 24) are predicted to affect surface-exposed residues distal to the A and C subunits (Fig. 3A and B). Positive selection at these C-terminal positions thus may reflect selective pressure on interactions with regulatory factors or substrates, rather than inter-subunit interactions within the trimeric complex.

Overall, the distribution of sites of positive selection within plant B56 gene families suggests that the B11 gene family has experienced positive selection at the largest number of sites, including a cluster of positions that are predicted to contact the A/C enzyme core and/or mediate substrate recruitment. In contrast, several other B56 subclades exhibit evidence of positive selection at more dispersed positions that appear less likely to affect interactions with the enzyme core. These data are consistent with the hypothesis that members of the B11 clade may recruit a distinct set of substrates. In contrast, selective pressure that may affect interactions with regulatory factors (e.g. those controlling subcellular localization) is observed in several other B56 clades.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

We are indebted to G. Jogl for advice on implementing PyMol modeling. We also thank M. Johnson, J. Bender and other members of the Brown Arabidopsis Molecular Genetics Group for helpful discussions. This material is based upon work supported by the National Science Foundation grant IOS-1145585. M.A.B. was partially supported by the National Institutes of Health predoctoral training program (grant no. GM007601).