Evolutionary Relationships Among Shell Proteins of Carboxysomes and Metabolosomes

Abstract

Bacterial microcompartments (BMCs) are self-assembling prokaryotic organelles which encapsulate enzymes within a polyhedral protein shell. The shells are comprised of only two structural modules, distinct domains that form pentagonal and hexagonal building blocks, which occupy the vertices and facets, respectively. As all BMC loci encode at least one hexamer- and one pentamer-forming protein, the evolutionary history of BMCs can be interrogated from the perspective of their shells. Here, we discuss how structures of intact shells and detailed phylogenies of their building blocks from a recent phylogenomic survey distinguish families of these domains and reveal clade-specific structural features. These features suggest distinct functional roles that recur across diverse BMCs. For example, it is clear that carboxysomes independently arose twice from metabolosomes, yet the principles of shell the assembly are remarkably conserved.

Compartmentalization is a modular organizing principle fundamental to life. The organelles, cells, tissues and organs of eukaryotes are familiar levels of biological subdivision which demonstrate how compartmentalization confers adaptive and complex functionality. It is only relatively recently that it has become evident that bacteria too have intracellular organization. Metabolic diversity underpins microbial diversity and, as in Eukaryotic organisms, compartmentalization in bacterial organelles expands metabolic flexibility by segregating reactions that are incompatible with cytosolic metabolism. These primitive prokaryotic organelles are known as Bacterial Microcompartments (BMCs). However, unlike eukaryotic organelles, the bounding membrane of BMCs is a protein shell. The BMC shell is composed of two basic types of protein – the hexagon-forming Pfam00936 domain and pentagon-forming Pfam03319 domain – that self-assemble into the facets and vertices of polyhedral, often icosahedral shells (Fig. 1). The potential to form BMCs is found across the Bacterial Kingdom where they are involved in both anabolic and catabolic reactions. The homology among shell proteins is readily recognized, which is remarkable in light of the extensive functional diversity of the encapsulated enzymes.

Figure 1. Overview of BMC shell proteins and structures of empty shells.

Shell structures’ PDB codes: HO: model based on 5V74, model beta-carboxysome shell: 6OWG, GRM2 shell: 6QN1, tandem BMC-H fusion shell: 6NER.

BMCs were first discovered as polyhedral bodies in cyanobacteria [1] and later named carboxysomes for their role in fixing CO₂ [2]. Now, comprehensive phylogenomic surveys of BMC loci reveal a widespread taxonomic distribution, across 45 different bacterial phyla [3,4]. Moreover, classification of BMC loci by clustering their domain profiles distinguishes at least 65 different types or subtypes of catabolic BMC loci (metabolosomes), including a large number of unknown function [3,4]. In contrast, only two types of carboxysomes are found, differentiated by the type of carbonic anhydrase and RuBisCO encapsulated. Crystal structures of many individual shell proteins from carboxysomes and metabolosomes [5] have shown that the cyclic symmetry axis of the hexagonal assembly is typically circumscribed by conserved residues that dictate the size and charge of the pore opening. These pores serve as the conduits for substrates and products of the encapsulated reactions.

In the last decade the study of BMCs has been tremendously advanced by the discovery that empty shells could be formed in the absence of cargo [6-11]. These tend to be smaller than their native (cargo replete) counterparts. Structural studies of empty shells from several functionally distinct BMCs [9,12,13] show that the BMC-H hexamers make up the bulk of the shell (Fig. 1). Interactions of the hexamers with each other and with the pentamers are highly conserved across types, with the sidedness/orientation of the faces of the shell proteins conserved [9,12,13]. In one structure, the BMC-T proteins were also visualized and were found to have slightly different angles of interaction with BMC-H proteins [12,14], indicating that they fulfill a special structural and/or functional role.

Collectively these structures constitute an architectural map of the universal features of BMC shell assembly. Nevertheless, it remains challenging to predict the assembly properties for many of the newly-identified BMC types, which exhibit a large combinatorial diversity of shell protein types encoded by their loci, including many paralogs and type-specific variants.

In the most recent phylogenomic BMC locus survey [4], in which phylogenies were constructed for approximately 9500 BMC-H, 4300 BMC-T, and 4900 BMC-P protein sequences, unexpected relationships among shell proteins from functionally distinct BMC types were found, indicating that a classification system based on sequence features could be devised. By assigning descriptively-named RGB hexcodes to identifiable tree clades, using similar colors for closely-related clades, a function-agnostic rankless classification system was developed for discussing the evolutionary history of BMC shell proteins, complementary to and independent of prior assumptions about function. Here we review these data and situate our findings in the context of conserved architectural principles and the evolution and functional diversity of BMCs.

BMC-P Proteins

Pentamers are presumed to be confined to vertices of fully-assembled shells, and have been shown to be essential for the function of the shell as a diffusive barrier [15]. They generally exhibit strong sequence similarity, including several universally-conserved amino acid positions which correspond to the hexamer-pentamer interface (Fig. 4ab). They resolve into four or five distinct phylogenetic groups: blue/green-type, purple-type, orange-type, and grey-type (Fig. 2a). The majority of BMC loci encode a single BMC-P from the blue/green or purple-type clades, representing a “standard” pentamer. Interestingly, many loci (particularly SPUs, FRAGs, ACI, HO, and PVM, see Fig. 2,3 legends for type nomenclature) encode BMC-P “triplets” containing a PvmM-like orange-type and a PvmK-like grey-type together with a standard pentamer [4]. Given that BMC shells are assumed to require only 12 pentamers, and that the pore formed at the cyclic symmetry axis is small (~4 Å in diameter), they are assumed not to play a significant role in metabolite conductance. However, a recent study of the protein stoichiometry of beta-carboxysomes showed varying occupation of the vertices by CcmL, depending on environmental conditions [16]. Shells without BMC-P proteins at the vertices, referred to as “wiffle ball” shells, have also been obtained in synthetic model systems (Fig. 1) [17,18]. The diversity of roles for BMC-P proteins, and their dynamics, warrants further study.

Figure 4. Clade-specific sequence features of BMC shell proteins.

Sequence conservation logos are shown for sequences of selected BMC-P (a-b) and BMC-H and BMC-T clades (c-d). Boxes indicate regions of residues involved in lateral inter-protein interactions based on assembled shell structures. Logos were generated using WebLogo software upon multiple sequence alignment with MAFFT.

Figure 2. Unrooted phylogenetic trees of BMC shell proteins.

Trees were generated with RAxML for non-redundant sets of shell protein sequences. (a) BMC-P (b) BMC-H (c), BMC-T^s proteins, which are simple trimers that do not dimerize and (d) BMC-T^dp proteins, this family dimerizes across the concave face. Position of characterized proteins are labeled with their gene or model system name. Protein structure PDB codes are indicated in parentheses. PDU: propanediol utilization BMC; EUT: ethanolamine utilization BMC; HO: Haliangium ochraceum BMC; RMM: Rhodococcus and Mycobacterium microcompartment (functional type nomenclature as used in [3,4]).

Figure 3. Relationships of carboxysome shell proteins to nearest metabolosome relatives.

Subtrees are shown for carboxysome and related clades of (a) BMC-P, (b) BMC-H, (c) BMC-T^s and (d) BMC-T^dp phylogenies from Figure 2. Clades with a predominant BMC type are indicated; BMC type abbreviations: SPU: sugar phosphate BMC, PVM: Planctomycete and Verrumicrobia BMC; MUF: Metabolosome of unknown function, MIC: Metabolosome with incomplete core, BUF: BMC of unknown function, GRM: Glycyl radical BMC (functional type nomenclature as used in [3,4]).

Uniquely among all BMC loci, alpha-carboxysomes encode two very similar BMC-P proteins, CsoS4A/CsoS4B [19], from a single purple-type clade that is strongly divergent from other BMC-P clades (Fig. 2a,4a). With the alpha-cyanobacterial CsoS4 sequences virtually indistinguishable from their orthologs in chemoautotrophs, the whole clade maps closest to other Proteobacteria-specific metabolosome purple-type clades, nearest to raspberry GRMguf-type pentamers from Deltaproteobacteria as well as the well-studied EutN/lightViolet clade from Gammaproteobacteria (Fig. 3a), supporting a Proteobacterial origin for the alpha-carboxysome.

In contrast, beta-carboxysome loci encode a single pentamer, CcmL, from a completely different section of the tree (Fig. 2a). CcmL is most closely related to green-type metabolosome pentamers from Firmicutes (Fig. 3a,4b). The mixed nature of the nearest clade (murkyGreen), representing a variety of BMC types (PDU1E, PDU1C, EUT2A, and GRM3A), suggests that beta-carboxysomes co-opted a metabolosome pentamer. Clade-specific features may possibly represent unique interaction surfaces for BMC-type-specific binding partners, or particular cytoplasmic features such as cytoskeletal elements.

BMC-H Proteins

The BMC-H hexamer comprises the main building block of the shell. They have long been assumed to form only homohexamers, but recent evidence suggests that heterohexamer assemblies also can be formed [20,21]. Most BMC shells are composed from multiple BMC-H paralogs; these genes may be encoded distal from the main organelle locus in satellite loci [3]. These are presumed to be differentially regulated from the main locus, and may provide additional flexibility in the permeability properties of the shell.

The comprehensive phylogenomic survey [4] revealed a core set of tightly-conserved clades, centered around the blue cluster, which appear to contain the fundamental, and presumably numerically dominant, building blocks of the metabolosome shell (Fig. 2b, inset). This is consistent with the most abundant BMC-H proteins from biochemical studies of intact metaboloomes such as PduA/J [22,23] and EutM [24]. Other clades, which typically constitute additional paralogs, seem to be more specialized and show greater sequence divergence or sequence extensions (Fig. 2b). Alpha-carboxysome hexamers are all derived from a single evolutionary type, despite many alpha-carboxysomes utilizing up to three (CsoS1A,B,C) paralogs that are nearly identical within a given locus [25](Fig. 3b). They constitute a very simple shell consisting almost exclusively of CsoS1 paralogs [26]. This compact clade again exhibits a long internal stem which separates CsoS1s from their closest metabolosome homologs (mulberry, EutK-like), implying a large sequence divergence and strong conservation upon the establishment of alpha-carboxysome hexamers.

Beta-carboxysomes are also composed of multiple BMC-H paralogs (CcmKs), clustering at the periphery of the hexamer tree, however very far from the CsoS1 clade (Fig. 2b). CcmK1 and CcmK2 paralogs map basally within the CcmK-like subtree and cannot be resolved phylogenetically due to extremely strong sequence constraint (Fig. 3b), in accordance with their essentiality [27-29] and stoichiometric prevalence in intact, natively-purified carboxysomes [16]. CcmK3 and CcmK4, which are encoded together in a satellite locus, appear as sister clades diverging together from the middle of the CcmK subtree (Fig. 3b), were recently recent demonstrated to form heterohexamers capable of dimerizing across their concave faces to form dodecamer in a pH-dependent manner [20], potentially serving as a metabolite channel subject to regulation. With additional genome sequence data now available for ecophysiologically diverse cyanobacteria, additional CcmK paralogs have been discovered [30]. For example, the newly identified CcmK5, from taxonomically-diverse genomes that are distinctive in lacking CcmK3 and CcmK4, structurally resembles CcmK1/2 and CcmK4; while CcmK6, found predominantly in heterocyst-forming cyanobacteria, resembles CcmK3, with which it often co-occurs. These observations suggest the potential for additional plasticity in shell permeability properties. While CcmK paralogs likely arose by a series of gene duplications after the establishment of beta-carboxysomes, three groups of CcmK5-like hexamers are found in heterotrophs (Fig. 3b), including two large sister clades from GRM5 and PVMlike metabolosomes as well as a smaller clade of EUT2x-type metabolosomes (a EUT2 related BMC that lacks the genes for ethanolamine lyase) from the Firmicute genus Sporosarcina, suggesting horizontal gene transfer from beta-cyanobacteria. In contrast to the ocean-dwelling alpha-cyanobacteria, beta-cyanobacteria live in diverse habitats subject to environmental dynamics (e.g., nutrient availability, desiccation, symbioses); suggesting that the multiplicity of CcmK paralogs arose as an adaptation for altering carboxysome permeability in response to fluctuating environments [20] experienced by ecophysiologically diverse cyanobacteria.

BMC-T proteins

BMC-T proteins are a tandem fusion of two copies of the Pfam00936 domain (Fig. 1), having originated from gene duplication and/or fusion events. Crystal structures BMC-T trimers reveal pseudohexameric assemblies that are similar in size and shape as BMC-H hexamers. The three-fold instead of six-fold symmetry allows for twice as many distinct conserved residues converging at the pores, and trimers also have two distinct edges that can interact with (different) surrounding shell proteins in a facet. Each domain in a BMC-T protein can diverge separately and at differing rates, resulting in extensive diversification. One clade from GRM1A/B loci, “T^s_dust” (Fig. 3c) appears to be a recent fusion of two adjacently-encoded BMC-H paralogs (Fig. 4c) [31]. This provides inspiration for the design of shell building blocks that are poised for adaptive evolution for new functions in bioengineering, such as the synthetic BMC-T protein that self-assembles into 20 nm wiffle ball shells even without a BMC-P [18] (Fig. 1). In general, BMC-T proteins to have specialized functions such binding an Fe-S cluster at the pore, as in PduT [32,33](Fig. 2c). Some extreme variants such as PduB, EutL and EtuB exhibit circular sequence permutations, which reorders the secondary structure elements within each domain, and have off-axis pores that possibly regulate opening of the central pore [34-37]. BMC-T^s proteins are widespread, found in 52 of the 68 presently-identified BMC types, although they are conspicuously absent from several BMC types such as GRM2, RMM, PVMlike, EUT2x, BUF1B, MIC1/MIC2, and alpha-carboxysomes [4].

All beta-carboxysome loci encode an enigmatic BMC-T^s, CcmO, which is relatively poorly characterized, even lacking experimental demonstration that it forms trimers/pseudohexamerss. CcmO maps at the periphery of the tree in a tight cluster, indicative of evolutionary constraint, with a long internal stem representing significant sequence divergence, most closely related to three other outlier clades from metabolosomes (Fig. 3c), potentially a result of longbranch attaction. In a homology model, CcmO has a standard CcmK-like (KIGS) motif that would surround the pore in a trimer/pseudohexamer, and typical edge-interaction interfaces in both domains (Fig. 4d). CcmO however is essential for formation of functional carboxsyomes [27,38] so it clearly fulfills an important structural function in beta-carboxysomes. There is a second class of trimers that obligately dimerize across their concave surfaces (BMC-T^dp Proteins, Fig. 1) The pores in these trimer are relatively large (~14 Å) and are gated by absolutely conserved surrounding residues [39-42]. For the beta-carboxysomal protein CcmP there is structural evidence of metabolites bound in a pocket on the inside cavity that was modeled as 3-PGA [39] or ADP [41] and its presence was correlated with pore opening [41]. With more functionally diverse BMC loci are identified, genes encoding BMC-T^dp proteins are found to be widespread, occurring in 21 of the 68 locus types, including HO, RMM, EUT3, and SPU2/3/6, as well as both the alpha- and beta-carboxysomes (Fig. 2d).

Four families of BMC-T^dp can be distinguished phylogenetically (Fig. 2d), with the alpha-carboxysome type, CsoS1D, again very distant from the beta-carboxysome type, CcmP. Unlike the BMC-H and BMC-P sequences of alpha-carboxysomes, the CsoS1D orthologs from cyanobacteria are readily distinguished from their chemoautotroph counterparts, which comprise three distinct clades (Fig. 3d). While each contains a variety of taxa, only Gammaproteobacteria can be found in all three, suggesting this taxonomic class may have been associated with the origins of the alpha-carboxysome. This supports the idea that BMC-T^dp proteins contribute a more specialized function compared to standard hexamers – perhaps in regulation and/or metabolite transport across the shell, or in the context of their different redox dynamics ( chemoautotrophs do not generate O₂). Nevertheless, the entire CsoS1D clade again diverges strongly from metabolosome orthologs, with the long-stemmed melon clade as their closest relatives, encoded mostly by SPU-type loci [3,4] from Proteobacteria.

In contrast, the CcmP clade from beta-carboxysomes is adjacent to several clades from SPU3-type metabolosomes, almost indistinguishably, with the closest derived from Chloroflexi (Fig. 3d). The association of SPU metabolosomes with both carboxysome clades is interesting because the class of substrates that they catabolize, sugar phosphate, is also the immediate product of carboxysomes, 3-phosphoglycerate; thus perhaps these BMC-T^dp proteins conduct this kind of metabolite. The very close connection with these CcmP-like tandem domains is unclear, however, especially as the taxonomy and BMC-type of the nearest metabolosome ancestor is completely different from that of CcmL, CcmK, and CcmO. Because CcmP is always encoded in a satellite locus, it is likely that the protein was horizontally acquired, although such a transfer would have had to predate the establishment of the modern cyanobacteria, as it is rarely missing from any genomes and is found even in the early-branching genera such as Gloeobacter. Nevertheless, again the BMC-T^dp sequences from beta-carboxysomes are more closely affiliated with their metabolosome ancestors and map to a very different region of the tree than the alpha-cyanobacterial BMC-T^dp homologues.

Conclusions

The structure and selective permeability of BMC shells constitutes the interface between the encapsulated reactions and the surrounding metabolism and is intrinsic to BMC function. Bioinformatic analysis suggests that the BMC shell is an ancient innovation and the modularity of the components used in its construction enabled it to evolve for diverse functions, and sometimes more than once for the same function. Multiple lines of evidence suggest that alpha- and beta-type carboxysomes arose independently. Likewise, the shell protein building blocks of most metabolosomes are drawn from multiple lineages that arose from BMC-H, BMC-T and BMC-P evolution, indicating frequent exchange of the modular components. Bioinformatic analyses combined with structural characterization of model shell systems are pointing to some universal principles of shell architectures, such as the role of at least one identifiable paralog from the central Blue BMC-H clade as the standard hexamer, that assembles with other hexamers and trimers oriented with the concave faces outward. The modular assembly from a catalog of functionally distinct but structurally interchangeable building blocks that assemble into a predictable orientation is fundamental to the native functions of shells and their engineering into new contexts.

HIGHLIGHTS:

• Phylogenomic analysis of shell proteins uncovers supra-functional relationships

• Alpha-carboxysome shell proteins arose independently from beta-carboxysome shells

• Beta-carboxysome shell proteins retain similarities with metabolosome shells

• Alpha-carboxysome shells likely diverged from Proteobacterial metabolosomes

• Interactions among building blocks in intact shells are conserved across all types