Conservation and Evolution of the Sporulation Gene Set in Diverse Members of the Firmicutes

ABSTRACT

The current classification of the phylum Firmicutes (new name, Bacillota) features eight distinct classes, six of which include known spore-forming bacteria. In Bacillus subtilis, sporulation involves up to 500 genes, many of which do not have orthologs in other bacilli and/or clostridia. Previous studies identified about 60 sporulation genes of B. subtilis that were shared by all spore-forming members of the Firmicutes. These genes are referred to as the sporulation core or signature, although many of these are also found in genomes of nonsporeformers. Using an expanded set of 180 firmicute genomes from 160 genera, including 76 spore-forming species, we investigated the conservation of the sporulation genes, in particular seeking to identify lineages that lack some of the genes from the conserved sporulation core. The results of this analysis confirmed that many small acid-soluble spore proteins (SASPs), spore coat proteins, and germination proteins, which were previously characterized in bacilli, are missing in spore-forming members of Clostridia and other classes of Firmicutes. A particularly dramatic loss of sporulation genes was observed in the spore-forming members of the families Planococcaceae and Erysipelotrichaceae. Fifteen species from diverse lineages were found to carry skin (sigK-interrupting) elements of different sizes that all encoded SpoIVCA-like recombinases but did not share any other genes. Phylogenetic trees built from concatenated alignments of sporulation proteins and ribosomal proteins showed similar topology, indicating an early origin and subsequent vertical inheritance of the sporulation genes.

IMPORTANCE Many members of the phylum Firmicutes (Bacillota) are capable of producing endospores, which enhance the survival of important Gram-positive pathogens that cause such diseases as anthrax, botulism, colitis, gas gangrene, and tetanus. We show that the core set of sporulation genes, defined previously through genome comparisons of several bacilli and clostridia, is conserved in a wide variety of sporeformers from several distinct lineages of Firmicutes. We also detected widespread loss of sporulation genes in many organisms, particularly within the families Planococcaceae and Erysipelotrichaceae. Members of these families, such as Lysinibacillus sphaericus and Clostridium innocuum, could be excellent model organisms for studying sporulation mechanisms, such as engulfment, formation of the spore coat, and spore germination.

INTRODUCTION

A variety of bacteria are capable of producing resting forms, commonly referred to as spores. However, the ability to form heat-, solvent- and UV-resistant endospores has been observed only in members of the phylum Firmicutes (low-G+C Gram-positive bacteria, recently renamed Bacillota) (1–8). Occasional reports of endosporeformers from other phyla have not been validated so far (9, 10). Sporulation enables bacteria to survive adverse environmental conditions. Thus, when this ability is found in human pathogens that cause severe diseases, such as anthrax, botulism, colitis, infectious diarrhea, gas gangrene, sepsis, tetanus, and food poisoning, it makes them particularly dangerous and difficult to eradicate (3, 4). Even sporeformers that used to be considered benign could turn out to be opportunistic pathogens, as exemplified by the recently described involvement of Paenibacillus thiaminolyticus in postinfectious hydrocephalus (11).

Formation of endospores is a complex process that starts with the asymmetric division of a vegetative cell, producing a mother cell and a prespore, and proceeds through several stages of spore maturation. In the best-studied model organism, Bacillus subtilis, the sporulation process affects expression of more than 500 genes, some of which are essential for sporulation, whereas others appear to be involved in various regulatory circuits (12–20). While some genes are exclusively involved in sporulation and are expressed only at certain stages of the process, spores also contain many housekeeping proteins that function during spore germination and subsequent vegetative growth (21–23). Studies on sporulation mechanisms in clostridia identified a somewhat smaller set of sporulation genes than in bacilli, leading to the conclusion that clostridia encode a streamlined version of the sporulation machinery (22, 24–27).

The ability to form spores is widespread in the two major classes of Firmicutes, Bacilli and Clostridia, and also has been observed in the other, more recently described, firmicute lineages (3, 28–30). However, some well-studied taxa within the Bacilli, such as lactobacilli, listeria, staphylococci, and streptococci, do not include any spore-forming members (3). Similarly, no sporeformers have been identified in the clostridial family Halanaerobiaceae (order Halanaerobiales) and in several families of the order Clostridiales (recently renamed Eubacteriales) (3, 31). Many other firmicute lineages include both spore-forming and non-spore-forming members (3).

These observations, coupled with the attempts to identify potential drug targets for extermination of spores, call for identification of the core set of essential sporulation genes that are necessary and, possibly, sufficient for a bacterium to be a sporeformer. The most obvious candidate, the sporulation master regulator Spo0A, indeed, appeared to be essential, as no Spo0A^– organism had been shown to form spores (3, 32). However, Spo0A is encoded in the genomes of many nonsporeformers and, therefore, could not be used as a reliable predictor of the sporulation ability (3, 32). Three more genes, sspA, dpaA (spoVFA), and dpaB (spoVFB), initially proposed as sporulation signatures (33), presented the same problem, being present in a variety of nonsporeformers (32).

The availability of complete genome sequences of many diverse firmicutes offers an opportunity to identify sporulation genes through comparative genomics. Following the approach pioneered by Stragier (34), several studies took advantage of the constantly expanding genome list to define the conserved core of sporulation genes that are present in all (or at least most) known sporeformers (2, 32, 35–37). The resulting list included about 60 genes, many of which had been previously shown to be essential for sporulation because mutations in these genes caused sporulation arrest and/or production of immature spores. However, deletion of some other genes from the core set appeared to have only a minor effect on the sporulation efficiency, indicative of a substantial redundancy of the sporulation machinery. Indeed, in B. subtilis, many sporulation genes are found in two or more paralogous forms (37, 38).

Here, we used the latest update of the Clusters of Orthologous Genes (COG) database (39) to analyze the sporulation gene sets in the current genome collection of diverse firmicutes, including members of the recently defined classes Negativicutes, Tissierellia, Erysipelotrichia, and Limnochordia. The COG database covers a limited number of selected, completely sequenced microbial genomes (typically, a single representative per bacterial genus) and features COG-specific patterns of presence-absence of genes in the respective organisms (40, 41). Thus, COG profiles offer an easy way to identify genomes in which a given gene, e.g., one involved in sporulation, is missing (39, 41, 42). Analysis of COG profiles of 180 genomes representing 76 spore-forming and 102 asporogenous species allowed us to detect numerous events of lineage-specific gene loss in various components of the sporulation machinery. Of particular interest was the widespread loss of sporulation genes in the families Planococcaceae and Erysipelotrichaceae, which resulted in a greatly streamlined machinery for engulfment, formation of the spore coat, and spore germination. These observations suggest that certain members of these families could be excellent model organisms for studying the fundamental mechanisms of sporulation in greater molecular detail.

RESULTS

Sporeformer genome collection.

The analyzed set consisted of 180 genomes from 160 genera, representing every named genus of Firmicutes that included at least one completely sequenced genome by 1 April 2019 and several genomes released after that date (39). These organisms belong to six currently recognized classes of Firmicutes, including Bacilli, Clostridia, Erysipelotrichia, Limnochordia, Negativicutes, and Tissierellia, and represent 76 species whose original descriptions mentioned their ability to sporulate, 102 nonsporeformers (based on the descriptions of the respective strains), and two species with an unclear sporulation status (see Table S1 in the supplemental material). At the time of this study, there were no completely sequenced genomes from any representatives of the classes Culicoidibacteria and Thermolithobacteria; members of these classes have been described as non-spore forming (43, 44).

To ensure reliability of the gene presence-absence patterns, this work relied on high-quality genomes included in the recent release of the COG database (39), most of which had been vetted by the NCBI’s RefSeq team and selected as RefSeq reference genomes (45). For two organisms, Caproiciproducens sp. strain NJN-50 and Thermincola potens JR, the sporulation status had not been described, and they had both spore-forming and asporogenous relatives. To ensure proper coverage of the class Erysipelotrichia, its two members featured in the COG database were supplemented with five additional genomes, including two from known sporeformers, [Clostridium] innocuum (the brackets indicate that this organism has been misnamed, which remains to be rectified [28]) and Erysipelatoclostridium (formerly Clostridium) ramosum (28, 46, 47).

Sporulation and genome size.

As reported previously, sporeformers generally have larger genomes than nonsporeformers (32). In our set, the mean genome size of the former was ~3.9 Mb, compared to ~2.7 Mb for the latter (Fig. S1 in the supplemental material). Although the genome size distributions overlapped, the previously reported 2.3-Mb boundary for cultivated sporeformers (32) held for the expanded genome set analyzed here. Only two sporeformers in the set, both still uncultured, “Candidatus Arthromitus” SFB-mouse and “Candidatus Desulforudis audaxviator” MP104C, had genome sizes of less than 2.4 Mb (see Table S1; the recently cultivated spore-forming strain of “Ca. Desulforudis audaxviator” with a 2.2-Mb genome [48] was not part of the COG set). In agreement with the previous reports (3, 49, 50), the ability to sporulate often differed even between closely related organisms: the families Bacillaceae and Planococcaceae in the class Bacilli, most families in the class Clostridia, and the families Tissierellaceae and Erysipelotrichaceae all included both spore-forming and asporogenous members (Table S1). A comparison of the genome sizes of sporeformers and nonsporeformers from different classes is presented in Fig. S1B.

Selection of sporulation genes.

For the purposes of this work, sporulation genes were defined as genes that participate in spore formation but have no known housekeeping roles in vegetative cells. This approach excluded most metabolic enzymes, as well as proteins involved in DNA replication and repair, transcription, translation, motility, secretion, and other processes. However, we included the cell division genes that are (also) involved in asymmetric cell division, as well as Spo0A-regulated genes that function at the onset of sporulation, prespore and forespore genes expressed under the control of SigF and/or SigG, mother cell genes that are expressed under the control of SigE and/or SigK, genes involved in the formation of the spore cortex and spore coat, and genes involved in spore germination. The list of the genes belonging to these groups was compiled based on the previous studies (2, 15, 16, 28, 32, 36, 37) and the data from the SubtiWiki database (http://subtiwiki.uni-goettingen.de/) (38). In the COG database, products of these genes formed 237 clusters of orthologous groups (COGs). We extracted COG phyletic profiles (the patterns of gene presence-absence in the respective genomes) for these sporulation genes and sorted them by their functions and representation among the 76 spore-forming organisms. The list of the most common sporulation genes identified in this manner is presented in Table 1. The patterns of distribution of these genes among all 180 analyzed species are listed in Table S2 in the supplemental material and are shown graphically in Fig. S2.

TABLE 1.

Conservation of sporulation genes in spore-forming Firmicutes

Sporulation stage or regulon	Sporulation gene(s)a
	Universally conserved in Firmicutes	Conserved mostly in Bacilli
Sporulation onset and checkpoints; asymmetric cell division	spo0A, spo0H (sigH), spoIIE, spoIIIE, spoIIIJ, spoVC (pth), spoVG,b spoVS,b divIB,c divIC,c divIVA,b ftsA,b ftsE,c ftsH, ftsX,c ftsY, ftsZ, jag, minC,b minD, minJ, obgE, sweCc	spo0B, spo0F, spo0E, spoVN (ald), ftsL, ricA (ymcA), ricF (ylbF), ricT (yaaT), sda
Spo0A regulon	sigE, sigF, sigG, spoIIAA, spoIIAB, spoIIGA, spo0JA (parA),c spo0JB (parB)	yisK, yusE
Engulfment	spoIID,c spoIIM,b spoIIP,c spoIIQ, spoIIIAA,b spoIIIAB,b spoIIIAC,b spoIIIAD,b spoIIIAE,b spoIIIAF,b spoIIIAG,c spoIIIAH	spoIIB, spoIIIL, fisB (yunB)
Forespore-expressed genes
SigF regulon	spoIIR,c spoIVB, spoVT,c dacB/dacF, gerW (ytfJ), ydfS/yetF,c yhcV/ylbB,b ylbC,b yloC,b yuiC,c yyaCb	bofC, rsfA, fin (yabK), yjbA, ymfJ, yqhG, ywzB
SigG regulon	spoVAC, spoVAD, spoVAEB, nfo (yqfS),b sspA/sspB/sspC/sspD	spoVAA, spoVAB, spoVAF, sspE, sspF, sspH, sspI, tlp, yqfX
Mother cell-expressed genes
SigE regulon	sigK, spoIIID, spoIVFB,c spoVB (spoIIIF)/ykvU, spoVE, spoVK,b ctpB, dacB/dacF, alr (yncD), spmA,b spmB,b yisY/yfhM, yitE/yqfU,c ylmC/ymxH,c ytaF,c ytvI, yyaD/ykvI	spoIVFA, spoVM, bofA, ydcA, ydcC, yhbH
SigK regulon	dpaA (spoVFA),b dpaB (spoVFB),b cgeD,b ykuD/yciB,b ytdA,b ytlDb
Spore cortex	spoVD, spoVV (ylbJ), cwlC/cwlD, lytH,b stoA (spoIVH), yabP, yabQ,c yqfC,b yqfD	cotD
Spore coat	spoIVA, cotJC/yjqC,b cotSA, gerM,b lipC (ycsK),b safA,b sleL/ydhD,b yhaX	spoVID, spoVIF, cotE, cotJA, cotJB, cotM/cotP, yhjR
Germination	gerA,b gerB,b gerC,b gerF(lgt), gpr, cwlJ/sleB,b cspA,c gdh	gerD, gerE, gerQ, ypeB

^aDistribution of the COGs that contain the indicated Bacillus subtilis sporulation genes. No distinction is made for members of the same COG. Paralogs are separated with a slash; parentheses indicate alternative names of the same gene (see Table S2 for details). Genes whose products were used to build the consensus tree (Fig. 3) are shown in bold typeface.

^bGene conserved in diverse sporeformers but missing in three or more sporeformer genomes.

^cGene missing in one or two genomes of sporeformers (not counting the frameshifts); see Table S3 for a detailed list.

Distribution of spo0A, dpaAB, and sspA genes.

In 2004, Onyenwoke and colleagues identified four conserved sporulation genes, namely, spo0A, dpaA, dpaB, and sspA, and proposed using these genes as markers of the ability of a bacterium to form endospores (33). We have previously reported the presence of these genes in several clostridial nonsporeformers and argued that the presence of these genes was insufficient—and, in the case of dpaAB, unnecessary—for sporulation (32). It was instructive to explore the distribution and evolution of these four genes in the current, much more diverse collection of firmicute genomes.

(i) spo0A.

Analysis of the current genome set confirmed that the master cell regulator Spo0A is an essential component of the sporulation machinery. Indeed, the genome of every experimentally characterized spore-forming member of the Firmicutes encodes Spo0A, and its absence is an excellent predictor of the organism’s inability to sporulate. Among the 180 species analyzed in this work, the spo0A gene was present in 118, including all 76 sporeformers and 40 nonsporeformers (Table 1; Table S1). These observations confirm that the presence of spo0A is insufficient to conclude that a bacterium is a sporeformer, although its absence unequivocally indicates that it is not. In agreement with the previous reports (32, 33, 51), Spo0A-encoding nonsporeformers were found in the classes Bacilli, Clostridia, and Erysipelotrichia. Among Bacilli, our set included six Spo0A-encoding nonsporeformers, all in the order Bacillales: Lentibacillus amyloliquefaciens in the family Bacillaceae, Kurthia zopfii and Planococcus antarcticus in Planococcaceae; Macrococcus caseolyticus in Staphylococcaceae, Novibacillus thermophilus in Thermoactinomycetaceae, and Exiguobacterium sp. from Bacillales family XII, incertae sedis. No Spo0A-encoding genomes and accordingly no sporeformers were represented among the 23 members of the order Lactobacillales. The current version of GenBank includes several spo0A genes from several clinical isolates of Streptococcus pneumoniae, but it remains unclear whether these isolates are correctly classified.

Spo0A is encoded by 32 of the 42 non-spore-forming members of Clostridia in our set and the two members with an unclear sporulation status. Among the 10 members of Negativicutes and nine members of Tissierellia, all four Spo0A-encoding organisms were sporeformers. Among the seven representatives of Erysipelotrichia, Spo0A was encoded in two sporeformers, C. innocuum and E. ramosum, and two nonsporeformers, Amedibacterium intestinale and Turicibacter sp. (Table S1). Most Spo0A⁺ genomes, 107 of 118, carried a single spo0A gene. The only exceptions were observed in the order Clostridiales, where 10 genomes encompassed two paralogous spo0A genes each, and one, the asporogen Anaerostipes hadrus, had three spo0A paralogs.

The phylogenetic tree of Spo0A proteins (Fig. S3 in the supplemental material) showed only minor deviations from the 16S rRNA-based phylogeny of the Firmicutes: Spo0As from the members of the same bacterial family typically formed distinct clades. Among the Bacilli, monophyly was observed for the families Alicyclobacillaceae and Planococcaceae, whereas Spo0As from members of Bacillaceae and Paenibacillaceae each formed two separate clades (Fig. S3). In Clostridia, Spo0As from most members of the families Clostridiaceae, Lachnospiraceae (formerly known as Ruminococcaceae or Hungateiclostridiaceae), and Peptostreptococcaceae (Clostridioides difficile and Paeniclostridium sordellii), as well as members of the order Halanaerobiales, appeared to be monophyletic. At the class level, Spo0As from spore-forming members of the Erysipelotrichia formed a strongly supported clade that mapped within the Bacilli, whereas members of the Negativicutes (family Sporomusaceae) and Tissierellia mapped within the clostridial clade (Fig. S3). Altogether, these groupings were consistent with vertical inheritance of the spo0A genes across the entire Firmicutes phylum.

Notably, the monophyly of Spo0As from the members of the same family did not depend on whether they came from sporeformers or nonsporeformers. Thus, Spo0As from all members of Planococcaceae formed a single clade, with Spo0As from nonsporeformers Kurthia zopfii and Planococcus antarcticus grouping together with Spo0As from planococcal sporeformers (Fig. S3). Likewise, Spo0A from the sporeformer Lachnoclostridium phytofermentans fit into a clade with non-spore-forming members of the family Lachnospiraceae, whereas Spo0A from the nonsporeformer Dehalobacter sp. strain CF mapped together with spore-forming members of the family Peptococcaceae (Fig. S3).

Irrespective of their ability to form spores, Spo0A-encoding organisms had larger genomes than Spo0A⁻ ones (Fig. S1B) and possessed many more sporulation genes than Spo0A⁻ genomes (Fig. 1). This trend was most pronounced for widespread sporulation genes (Fig. 1A) but also held for narrowly conserved sporulation genes (Fig. 1B) and for those sporulation genes that were conserved mostly in Bacilli (not shown).Thus, Spo0A-encoding bacteria, whether spore forming or not, differ from their Spo0A⁻ relatives in terms of their sporulation gene content. In other words, non-spore-forming Spo0A⁺ organisms still carry a fair number of sporulation genes. Because stepwise loss of sporulation genes in formerly spore-forming lineages appears far more likely than acquisition of such genes by nonsporulating organisms, these observations serve as additional evidence of the ancient origin of sporulation (see below).

FIG 1.

Distribution of sporulation genes in Spo0A⁺ and Spo0A⁻ firmicutes. The graphs show the numbers of widespread sporulation genes (out of 112) (A) and more narrowly conserved sporulation genes (out of 74) (B) encoded in sporeformers (which are all Spo0A⁺; blue columns), Spo0A⁺ nonsporeformers (orange columns), and Spo0A⁻ nonsporeformers (gray columns). “Other classes” includes Negativicutes, Tissierellia, Erysipelotrichia, and Limnochordia.

(ii) dpaAB.

Dipicolinate, a key spore component, is produced by oxidation of dihydrodipicolinate, which is catalyzed by dipicolinate synthase, whose two subunits are encoded by dpaA (spoVFA) and dpaB (spoVFB), genes that are expressed in the mother cell under σ^K control. Among the members of the class Bacilli in our genome set, the dpaAB gene pair was found in all sporeformers and was absent in nearly all nonsporeformers. Therefore, within Bacilli, these genes indeed could serve as markers of sporulation. The dpaAB gene pair was also found in Limnochorda pilosa and in both spore-forming members of Erysipelotrichia. In clostridia, however, the picture was more complicated. The dpaA and dpaB genes have been previously shown to be missing in Clostridium perfringens, C. botulinum, and C. tetani, in which dihydrodipicolinate oxidation is catalyzed by the electron transfer flavoprotein, encoded by the etfA-etfB gene pair (32, 52). Nonorthologous displacement of dpaAB by etfAB was also observed in several other clostridial sporeformers, including Clostridium acetobutylicum, Alkaliphilus metalliredigens, Paeniclostridium sordellii, and Tepidanaerobacter acetatoxydans, as well as two spore-forming representatives of Negativicutes, Methylomusa anaerophila and Pelosinus fermentans, and in Tissierella sp. strain JN-28 (also called Sporanaerobacter sp. strain NJN-17), a member of the class Tissierellia. Finally, Gottschalkia acidurici does not encode either DpaAB or EtfAB, suggesting that in this organism, dipicolinate production is catalyzed by yet another, currently unidentified oxidoreductase.

(iii) sspA.

The distribution of the sspA gene, which encodes an α/β-type small acid-soluble sporulation protein (SASP), is also dramatically different between the classes Bacilli and Clostridia. Among bacilli, sspA is found in all sporeformers and is missing in nearly all nonsporeformers (Table S2), making it another good marker of the ability of bacilli to sporulate. Essentially the same picture was observed in the classes Negativicutes, Tissierellia, and Erysipelotrichia. However, in Clostridia, the sspA gene is represented both in sporeformers and in many nonsporeformers, albeit only in Spo0A-encoding ones. The only sporeformer in our set that did not carry the sspA gene was C. innocuum. The ability of this organism to form spores in the absence of SspA (or any other known SASPs; see below) merits further study.

Identification of the core sporulation gene set.

The COGs that included the B. subtilis sporulation genes were sorted by their representation in the 76 sporeformer genomes and, specifically, in the genomes of spore-forming bacilli and clostridia. We classified these COGs into four groups (Table 1): (i) genes (COGs) that were represented in every sporeformer genome, with the possible exception of one or two; (ii) genes that were widespread in sporeformers but were missing in three or more genomes, often those from a specific lineage; (iii) genes that were widespread in bacilli but poorly represented in other classes; and (iv) genes that were present in only a limited number of sporeformers. Representation of these genes (COGs) in all analyzed firmicute genomes was recorded as well (Table S2).

As mentioned above, the analyzed genomes were extracted from the COG collection (39), most of which were selected from the NCBI’s RefSeq database (45). Nevertheless, it has been shown that even highly conserved ribosomal proteins are occasionally missed during genome assembly, left untranslated because of sequencing errors, or simply overlooked in the course of genome annotation (53). Therefore, cases of widespread sporulation genes missing in one or more sporeformer genomes were individually verified by checking for the potential presence of highly diverged or partial open reading frames (ORFs) using a DELTA-BLAST search (54) against the NCBI protein database and/or a TBLASTn search (55) against the nucleotide sequence database, each time selecting the respective species or strain as the target database and a representative protein from a closely related organism as a query (56). This analysis identified frameshifts and nonsense mutations in several widely conserved sporulation genes in several genomes (Table S3 in the supplemental material). However, the occurrence of such suspicious frameshifts was typically limited to one or two per genome, even in Sulfobacillus acidophilus strain TPY (GenBank accession number CP003179.1), which had several unannotated ribosomal proteins (53) and was excluded from RefSeq because of the missing rRNA genes. This genome also lacked an unusually high number of widespread sporulation genes (Table S4 in the supplemental material). Another exception was the genome of Jeotgalibacillus malaysiensis D5 (GenBank number CP009416.1), which also had several frameshifted and missing sporulation genes (Tables S3 and S4), some of which were found in other Jeotgalibacillus spp. (Table S2). The peculiar phylogenetic position of the genus Jeotgalibacillus between Bacillaceae and Planococcaceae (57, 58) warranted keeping this genome in the analyzed set, but its sporulation genes patterns were considered unreliable.

Comparison of Table 1 with the previously identified sets of core sporulation genes (2, 32, 36, 37) showed that the list remained largely stable, despite the substantially expanded coverage of Bacilli and Clostridia and the inclusion of representatives of three more classes of Firmicutes, Negativicutes, Tissierellia, and Limnochordia (Table S5 in the supplemental material). Most genes previously assigned to the conserved core were found in all or nearly all of the 76 sporeformers in the current genome set. Further, several genes, such as spoIVFB, spoVE, cotS, yfhM, yhaX, yisY, and yyaA, turned out to be more widespread than previously thought (Table S5). The only deviation from this pattern was observed in the class Erysipelotrichia, in which the spore-forming members, C. innocuum and E. ramosum, were missing more than a dozen genes from the conserved sporulation gene set (see below). This comparison also showed that, despite the overall conservation of the core sporulation genes in all classes of Firmicutes, many of these genes were found to be completely missing (as opposed to being disrupted by frameshifting) in one, two, or more genomes (Table S6 in the supplemental material). Although some of these gaps in phyletic patterns could be caused by errors in genome sequencing and/or assembly, the consistency of such patterns in certain phylogenetic groups appeared to reflect an actual evolutionary trend. In the following sections, we describe the streamlining of the sporulation gene complement in sporeformers from the families Planococcaceae and Erysipelotrichaceae, followed by discussion of the gene loss within specific sporulation systems.

Sporulation genes in Planococcaceae.

The analyzed genome set included six spore-forming members of the family Planococcaceae (also known as Caryophanaceae), order Bacillales: Paenisporosarcina sp. strain K2R23-3, Rummeliibacillus stabekisii, Solibacillus silvestris, Sporosarcina psychrophila, and Ureibacillus thermosphaericus. In addition, we included Jeotgalibacillus malaysiensis, which is currently assigned to the Planococcaceae, and Lysinibacillus sphaericus, which is usually assigned to Bacillaceae but in phylogenetic analyses robustly falls within the Planococcaceae (58). These seven organisms showed a consistent pattern of sporulation gene loss, with most of them lacking four of the nine components of the SpoIIQ–SpoIIIA complex (spoIIIAA, spoIIIAB, spoIIIAD, and spoIIIAF genes), as well as such conserved sporulation genes as spmA, spmB, spoIIM, gerM, gerW (ytfJ), yqfC, and ytxC, which are found in most sporeformers (Table 1; Tables S2 and S4). Furthermore, these seven organisms also lacked some genes that are generally conserved among the spore-forming members of the Bacillaceae, including bofC, csfB, spoIIIL, spoIVFA, spoVAA, spoVAB, spoVAEA, spoVID, yabG, yhbB, yrrD, ysxE, and ytxC (Table S2), although they encoded SpoVAC, SpoVAD, and SpoVAEB, proteins that collectively are sufficient for the calcium dipicolinate uptake by the developing spore (59). They also lack the genes that encode such SASPs as SspK, SspL, SspN, SspO, SspP, and CsgA, spore coat proteins CotM, CotI, CotO, CotP, and CotW, and spore maturation proteins SpsJ and CgeB, which are found in most spore-forming bacilli (Table S2). The most dramatic deviations from the common phyletic pattern were detected in J. malaysiensis, in accordance with its uncertain phylogenetic position (58). The consistent absence of all these genes suggests that members of the Planococcaceae possess substantially streamlined sporulation machinery. The analyzed genome set also included two asporogenous representatives of Planococcaceae, K. zopfii and P. antarcticus, which have spo0A but lack almost all other sporulation genes (Table S2).

Sporulation genes in Erysipelotrichaceae.

The evolution of the Erysipelotrichia lineage likely involved a massive loss of genes, including many related to sporulation (60). Indeed, the genomes of C. innocuum and E. ramosum, despite their relatively large sizes (4.7 and 3.25 Mb, respectively), show patterns of extensive loss of sporulation genes. Consistent with the previous observations (60, 61), they both lack seven of the nine genes of the SpoIIQ-SpoIIIA complex, retaining only SpoIIQ and SpoIIIAH components. Further, they both lack such widely conserved sporulation genes as spoIIID, bofA, cotQ, safA, yabQ, ypeB, yyaC, yyaD, sleL (yyaH), and gerW (ytfJ) (Tables S2 and S4). Despite the evolutionary proximity of Erysipelotrichia and Bacilli (28, 60), many widely conserved bacillar sporulation genes are missing as well, including bofC, csfB, spoIIIL, spoIVFA, spoVAA, spoVAB, spoVAEA, spoVID, yabG, yhbB, yrrD, ysxE, and ytxC. Some sporulation genes appear to have been lost in only one of these genomes. In particular, C. innocuum lost spoIIR and spoVS, whereas E. ramosum retained these genes but lost spoVG and spoVT (Table S4). Another system with an unusual phyletic pattern is the spore germination receptor GerABC. While non-spore-forming members of Erysipelotrichia encode either all three subunits of GerABC (Turicibacter sp. strain H121) or none at all (three other genomes), the sporeformers C. innocuum and E. ramosum encode the membrane subunit GerA (GerAA) but not the membrane transporter GerB (GerAB) or the lipoprotein GerC (GerAC) subunits. Although the GerABC receptors are not strictly required for germination (they are missing in Clostridioides difficile [62, 63] and could be dispensable in C. botulinum [64]), they are nearly ubiquitous among sporeformers (Table S2). The presence of GerAA but not GerAB or GerAC subunits is so far unique among known sporeformers.

Sporulation onset and asymmetric division.

Along with spo0A, genes involved in sporulation onset and asymmetric division are highly conserved in sporeformers of the class Bacilli, and many, although not all, of these genes are also conserved in spore-forming members of Clostridia and four other classes of Firmicutes (Table 1; Table S2). As discussed above, many clostridia lack the Spo0B-Spo0F-Spo0A phosphorelay system, and their Spo0As are phosphorylated by orphan histidine kinases that are not orthologous to any of the five sporulation histidine kinases (KinA to KinE) of B. subtilis. However, many of these orphan kinases contain at least one PAS sensor domain (25, 51, 65). Nevertheless, the spo0B-spo0F gene pair is present in several clostridia, primarily members of the families Heliobacteriaceae, Peptococcaceae, and Thermoanaerobacteraceae (51), as well as in Negativicutes and Tissierellia (Table S2).

The genes involved in asymmetric cell division, such as divIB, divIC, divIVA, spoIIE, ftsE, ftsH, ftsX, ftsY, ftsZ, minC, minD, and minJ, are widespread in bacteria, including most firmicutes and nearly all sporeformers (Table S2). As expected, genes encoding the four sporulation-specific sigma subunits, sigF, sigG, sigE, and sigK, are found in every sporeformer, but the Spo0A-regulated genes spoIIAA, spoIIAB, spoIIGA, spo0JA (parA), and spo0JB (parB) are also highly conserved in sporeformers of every firmicute lineage (Table 1). However, such Spo0A-dependent genes of B. subtilis as ykuJ, ykuK, and yneF, all three with unknown functions, are not conserved even within Bacillales (Table S2). It should be noted that sporulation of Sporosarcina ureae has been observed to involve symmetric division into two daughter cells (66). We do not know how widespread this phenomenon is, but it could reflect the above-mentioned genome streamlining in the Planococcaceae.

Formation and activation of σ^K.

While σ^K is universally encoded in the genomes of all spore-forming firmicutes, its activation is tightly controlled, preventing premature expression of late-stage sporulation genes in the mother cell (67–69). In B. subtilis strain 168, the σ^K-encoding gene is interrupted by a 48-kb prophage-like skin (sigK-intervening) element, which results in two separate coding regions designated spoIVCB and spoIIIC, neither of which is expressed. These fragments of sigK are joined as a result of SpoIVCA-dependent excision of skin to form the complete, intact sigK gene that encodes pro-σ^K (70, 71). Pro-σ^K is then activated in the mother cell through cleavage by the intramembrane metalloprotease SpoIVFB (67, 69). In C. difficile 630, the skin element is only 14.6 kb long, and its excision yields a sigK gene encoding a mature σ^K, which does not require proteolytic activation (72, 73). SigK-interrupting skin elements have also been identified in C. tetani and C. perfringens (74, 75). We confirmed the presence of split sigK genes in B. subtilis, C. difficile, and C. tetani and identified split sigK genes in 12 additional organisms from Bacilli, Clostridia, Negativicutes, and Tissierellia (Fig. 2). These skin elements range in size from 2.6 to 48 kb and encode from 2 to 67 proteins (see Table S7 in the supplemental material).

FIG 2.

Organization of skin elements in 15 firmicute genomes. spoIVCB genes are shown in light green, spoIIIC in dark green, and spoIVCA in orange. The organism names are followed by the GenBank genome entries and genomic coordinates of the regions between spoIVCB and spoIIIC. The numbers above the gaps indicate the distances between spoIVCB or spoIIIC and spoIVCA (~1.5 kb less than the total length of the skin element). R. cellulolyticum Ccel_0434 is annotated as a single 7,394-nucleotide pseudogene.

The orientation of skin elements with respect to the sigK component genes is variable, but the spoIVCA gene is always located at the edge of the element, usually immediately upstream of the spoIIIC gene, which encodes the C-terminal fragment of σ^K (Fig. 2). The SpoIVCA proteins from different skin elements (see Fig. S4A in the supplemental material for a sequence alignment) are members of a single family of site-specific DNA recombinases (COG1961) that is represented by multiple copies in most firmicute genomes, including up to 26 genes in (non-spore-forming) Flavonifractor plautii and Oscillibacter valericigenes. In the phylogenetic tree of SpoIVCA-like recombinases/integrases (Fig. S4B), SpoIVCAs from skin elements are interspersed with those from tailed phages (Caudovirales), suggesting a (pro)phage origin of the skin elements. Remarkably, aside from spoIIIC, spoIVCA, and spoIVCB, no genes are shared by all skin elements (Fig. S4C). The presence of skin elements that split sigK genes in two nonadjacent fragments often causes confusion in genome annotation. Thus, in the current genomic entries for B. subtilis strain 168 (GenBank accession numbers AL009126.3, CP051860.2, CP053102.1, and CP052842.1), sigK is erroneously marked as a pseudogene, whereas in genomic entries for C. tetani E88 (GenBank number AE015927.1) and Pelosinus fermentans (GenBank number CP010978.1), the C-terminal spoIIIC-like fragment is left untranslated (Tables S5 and S7). Remarkably, sigK is not unique as the skin insertion site: skin-like elements were also found interrupting the cotJC gene of Laceyella sacchari and spoIIID gene of Ruminiclostridium cellulolyticum (Table S3).

In B. subtilis, activation of pro-σ^K by SpoIVFB is regulated by BofA and also SpoIVFA, which is itself regulated by proteolysis by the forespore-expressed protease SpoIVB (76). σ^K is encoded in every sporeformer genome, as is, with a single exception, SpoIVFB, which is a member of a widespread metalloprotease family (COG1994). However, SpoIVFB-interacting BofA and SpoIVFA are far less widespread. BofA is missing in 10 of the 76 sporeformers, including six clostridia, C. innocuum, E. ramosum, and Limnochorda pilosa. The distribution of SpoIVFA is generally limited to the members of the class Bacilli. It is missing in the Planococcaceae, in nearly all clostridia, and, again, in C. innocuum, E. ramosum, and L. pilosa. In B. subtilis, the SpoIVFA-cleaving serine protease SpoIVB also cleaves another signaling serine protease, CtpB (77), and plays an additional, essential role in sporulation that appears to involve cleavage of SpoIIQ (78, 79). This cleavage of SpoIIQ has been proposed to be the function of SpoIVB in C. difficile, which has a pro-less sigK gene and does not need SigK processing (80). Accordingly, SpoIVB is nearly universal among sporeformers, being represented in numerous genomes that do not encode SpoIVFA. The details of SpoIVFB-BofA interaction in the absence of SpoIVFA and the functions of SpoIVB homologs in these organisms remain to be elucidated. This system also includes BofC, which has been proposed to interact with SpoIVB (81). BofC consists of two distinct domains, and its full-length version is found almost exclusively in Bacillaceae and Paenibacillaceae but not in other families of Bacillales, including Planococcaceae (Table S2). The C-terminal domain of BofC, PF08955, is more widespread but is still missing in most spore-forming clostridia. The functions of BofC and its C-terminal domain remain obscure.

Conservation of the engulfment complex.

In B. subtilis, engulfment involves two key protein complexes, SpoIID-SpoIIM-SpoIIP (the DMP complex) and SpoIIIA-SpoIIQ, the first of which is targeted to the growing septum by either SpoIIB or the SpoIVFA-SpoIVFB pair (82–84), with additional involvement of GerM (85). Both protein complexes are found in C. difficile (86) and are conserved in almost all sporeformers (Table 1), whereas the targeting proteins are less widespread: as noted previously (32), both spoIIB and spoIVFA are missing in clostridia (Table S2).

Among the three genes encoding the peptidoglycan-degrading DMP complex, spoIID was missing in a single genome in our 76-sporeformer genome set, and spoIIP was missing in two (Table S6); these three cases could be due to the sequencing problems. In contrast, spoIIM, although widespread as well, was missing in nine genomes (Table S2). This pattern is consistent with the reports that SpoIIM is not essential for engulfment in C. difficile (87, 88), whereas SpoIID and SpoIIP appear to be indispensable.

The transmembrane SpoIIIA–SpoIIQ complex consists of nine proteins, eight of which are encoded in the mother cell by the genes of the spoIIIAA-AB-AC-AD-AE-AF-AG-AH operon, which is under SigE control. The ninth protein, SpoIIQ, is produced in the forespore under SigF control and contains the M23-type metallopeptidase LytM domain (84, 89). The core of the SpoIIIA-SpoIIQ complex consists of SpoIIIAH and SpoIIQ proteins. Both these proteins form dodecameric (or even larger) rings that interact to form a 60-Å (or even a 140-Å) transmembrane channel, which connects the mother cell and the forespore and could potentially enable trafficking of small molecules (or even proteins) between the two compartments (90, 91). The structure of this complex from B. subtilis (Fig. S5A) revealed a key role of the LytM domain of SpoIIQ in its interaction with SpoIIIAH. However, despite the presence of the common LytM domain, SpoIIQ proteins from bacilli and clostridia are substantially distinct: the M23 peptidase is inactivated in the former but appears to be intact in the latter (92). Accordingly, these proteins belong to two different COGs, COG5820 and COG5821, in the COG database (39). Both varieties of SpoIIQ are often encoded in similar spoIID-spoIIQ-spoIIID operons, although the genomic neighborhoods of spoIIQ are highly variable even within the Bacillaceae (Fig. S6). Although we previously reported not being able to find orthologs of clostridial spoIIQ (CD0125 in C. difficile) in the genomes of several members of the family Peptococcaceae (32), in this work, using synteny and DELTA-BLAST (54) searches, we identified these in the genomes of all clostridial sporeformers but very few nonsporeformers (Table S2). Two spore-forming members of Erysipelotrichia, C. innocuum and E. ramosum, encode the bacillar form of SpoIIQ, whereas spore-forming members of Negativicutes, Tissierellia, and Limnochordia encode the clostridial form. Thus, all 76 sporeformers in our set encode both SpoIIIAH and one of the two forms of SpoIIQ, indicating strict conservation of the SpoIIQ-SpoIIIAH pore among spore-forming bacteria. Despite its conservation, the functions of the SpoIIQ-SpoIIIAH complex are not yet fully resolved. There is growing evidence that this complex plays a key role in keeping the two forespore membranes together, functioning as a zipper (93, 94) and/or a noncanonical piliation system (95).

As noted above, two groups of Firmicutes, Erysipelotrichaceae and Planococcaceae, have apparently lost multiple components of the SpoIIQ-SpoIIIA complex. The Erysipelotrichia members C. innocuum and E. ramosum retain only the spoIIIAH and spoIIQ genes. In addition, they possess SpoIVFB and GerM but lack SpoIIB and SpoIVFA (Table S2; Fig. S5B and C) (61). Six members of the family Planococcaceae show stronger conservation of the SpoIIQ-SpoIIIA complex: in addition to SpoIIIAH and SpoIIQ, they retain SpoIIIAC, SpoIIIAE, and SpoIIIAG components (Fig. S3). However, all these organisms lack SpoIIB, SpoIVFA, SpoIVFB, and GerM proteins, which participate in the assembly of the SpoIIQ-SpoIIIA complex in B. subtilis. Members of Planococcaceae, such as Lysinibacillus sphaericus, could be useful model organisms for studying the core mechanisms of engulfment in the Firmicutes.

SASPs.

SASPs are short proteins that constitute up to 20% of the total spore protein content and are the major source of amino acids during spore germination (96–100). In B. subtilis, the majority of the SASPs belong to two types, α/β (COG5852, which unifies the SspA, SspB, SspC, and SspD families, and COG5854 [SspF]) and γ (COG5853), whereas several minor SASPs (96, 99, 101) make up COGs from COG5855 through COG5864. As observed previously, the α/β-type SASPs, which bind double-stranded DNA and protect spores from heat, UV radiation, and other adverse conditions, are nearly universal in spore-forming firmicutes, whereas the distribution of the other SASP families varies (32, 102). This observation is supported by the data from our new set. Indeed, COG5852 is represented in 75 of the 76 sporeformers and in 31 of the 40 Spo0A-encoding nonsporeformers (Table S2). The only sporeformer missing SspA was C. innocuum; the other spore-forming member of Erysipelotrichaceae, E. ramosum, has two sspA-like paralogous genes and a single sspI gene for a minor SASP. Homologs of sspA are also found in other members of Erysipelotrichaceae, such as [Clostridium] spiroforme and [Clostridium] cocleatum (Fig. S7). However, C. innocuum lacks known ssp genes, so the nature of its SASPs, if any, remains enigmatic. Most of the minor SASPs described in B. subtilis are also encoded in at least some members of Bacillales (102). In contrast, members of other classes of Firmicutes encode only α/β-type SASPs of the SspF and Ssp4 families (103, 104); some clostridia also encode Tlp-like SASPs. Thus, the diversity of SASPs in B. subtilis does not appear to be shared by other groups of Firmicutes, which is consistent with the reports of nonessentiality of minor SASPs (97, 104).

Spore cortex.

Genes involved in the biosynthesis of the spore cortex, the peptidoglycan layer that surrounds the forespore membrane, have been previously found to be conserved throughout Bacilli and Clostridia (32). Examination of the current expanded genome set showed that this conclusion still holds for spoVB, spoVD (ftsI), spoVE (ftsW), spoVV (ylbJ), yabP, and yqfD genes (Table 1). One more gene, yabQ, which is essential in B. subtilis (105, 106), was missing only in the genomes of C. innocuum and E. ramosum, whereas the yqfC gene was missing in the genomes of six Planococcaceae members and E. ramosum (Table S6).

Spore coat.

As noted previously (5, 32, 107), bacilli and clostridia have dramatically different spore coats: many of the ~70 proteins that form the coat of B. subtilis (7) are conserved only within Bacillaceae or not even in all members of this family (32, 108). Among the 10 morphogenetic spore coat proteins of B. subtilis (SpoIVA, SpoVM, SpoVID, SafA, CotE, CotH, CotO, CotX, CotY, and CotZ), only SpoIVA is universally conserved in all spore-forming firmicutes (Table S2). In bacilli, SpoIVA interacts with SpoVID (109, 110), whereas in C. difficile and some other clostridia, its interaction partner is SipL (CD3567), which shares a C-terminal LysM domain with SpoVID (111, 112). Of the other SpoIVA- and/or SpoVID-interacting proteins, SpoVM is found in some clostridia but is not essential for sporulation (113), CotE is found in only a few clostridia, and the full-length SafA (as opposed to its LytM domain) is not represented outside bacilli. CotH, CotO, CotX, CotY, and CotZ are missing even in some members of Bacillaceae, and, except for CotH, none of these genes is found in clostridia or members of other classes of Firmicutes (Table S2).

Among the proteins of the B. subtilis inner coat, the most widespread are proteins of the CotJC/YjqC family, which belong to COG3546, “Mn-containing catalase.” Three members of this family (CotC, CotD, and CotE) were found in the spore coat of C. difficile, and the latter two have been shown to retain enzymatic activity (107, 114). Two more spore coat proteins of C. difficile, the peroxiredoxin CotE and the superoxide dismutase SodA, could be involved in protection from oxidative stress (107). Other widely conserved proteins of the B. subtilis basement layer and inner coat are also (not necessarily active) enzymes: the acetyltransferase CgeE, the peptidoglycan hydrolases CwlJ and SleL (YaaH), the phospholipase LipC, the HAD family phosphatase YhaX, the ATP-grasp family enzymes YheC and YheD, the MenH-related esterase YisY, and the glycosyltransferases CotSA and YdhD (Table S2). A similar pattern is apparent in the outer coat, which contains the alanine racemase YncD, the nucleotidyl transferase YtdA, the predicted FAD-dependent dehydrogenase CotQ, and the LysM domain-containing protein YkzQ. Both inner and outer coats of B. subtilis contain members of the small heat shock protein (HSP20) family, CotP and CotM, respectively. These examples demonstrate the widespread recruitment (exaptation) of common housekeeping enzymes as structural components of the spore coat. Their broad representation among bacillar sporeformers, as opposed to the scarcity in nonsporeformers (Table S2), suggests that such recruitment is a common mechanism of spore coat assembly in Bacilli. There are still insufficient data on spore coat proteins in Clostridia and none on those in other classes of Firmicutes, but studies in C. difficile (107, 114) suggest that such recruitment occurs in clostridia as well.

Germination proteins.

As mentioned above, the common germination receptors consisting of GerA, GerB, and GerC components are encoded in all classes of Firmicutes (Table S2). In our set of 76 sporeformers, the only exceptions were two members of the family Peptostreptococcaceae, C. difficile and Paeniclostridium sordellii, which lacked all three components of the receptor, and two members of Erysipelotrichaceae, C. innocuum and E. ramosum, which lacked GerB and GerC. In C. difficile, the bile acid germinant receptor has been identified as the pseudoprotease CspC (5, 115, 116), a member of the vast family of subtilisin-like serine proteases (COG1404). The presence of multiple members of this protein family in most spore-forming firmicutes suggests that the CspC-dependent germination mechanism is not limited to C. difficile and could potentially function in other organisms, as an addition to the better-studied GerABC-mediated signaling.

The germination process involves degradation of the spore protective layers, the cortex, and the inner and outer coats, as well as hydrolysis of DNA-shielding SASPs. In B. subtilis, hydrolysis of the spore cortex is catalyzed by peptidoglycan hydrolases CwlJ and SleB, members of the same COG3773, which is widespread in spore-forming bacilli and clostridia and is also found in spore-forming members of Tissierellia and Limnochordia, albeit not in the Negativicutes or C. innocuum (Table S2). The regulators of the activity of these enzymes, however, are less widespread. YpeB, which regulates SleB activity in bacilli and many clostridia (117–119), is missing in C. difficile and several other clostridia, as well as in all members of Negativicutes and Erysipelotrichia in the current set (Table S2). In C. difficile, the principal spore cortex-hydrolyzing enzyme is SleC (CD630_05510), a multidomain protein that consists of a structurally, but not functionally, characterized DUF3869 domain (Pfam [120] domain PF12985) (121), a SpoIID-like amidase domain, and a C-terminal peptidoglycan-binding (PF01471) domain. SleC homologs are found in many clostridia, as well as C. innocuum and E. ramosum.

Degradation of SASPs is catalyzed by the germination protease, Gpr, an aspartate protease (104, 122) that is universally conserved in spore-forming firmicutes (Table 1) but is absent in non-spore-forming members of Bacilli, Negativicutes, Tissierellia, and Erysipelotrichia. Among the clostridial nonsporeformers, the distribution of Gpr is generally similar to that of Spo0A: Spo0A-encoding sporeformers usually have the gpr gene, whereas genomes that lack spo0A also lack gpr (Table S2).

Among other genes involved in spore germination in B. subtilis, only a few, such as spoVAD, gdh, gerF (lgt), ykvU, and ypeB, are widespread, whereas others, such as csgA, gerD, gerE, gerPA/gerPF, gerPB, gerPC, gerPD, gerPE, and spoVAEA, are found mostly (or exclusively) in spore-forming bacilli (Table S2).

Uncharacterized sporulation proteins.

Despite numerous studies on sporulation in B. subtilis and other model organisms, functions of several widespread sporulation genes have been clarified only recently, whereas functions of several others remain unknown (Table 1). Altogether, the SubtiWiki database (38) lists >250 y genes that appear to be specifically expressed during sporulation (15, 16, 20, 123, 124), but have not been characterized experimentally. Assignment of such genes to the COGs allowed for making at least tentative functional predictions for several y genes (Table S8 in the supplemental material). Examples include YdgB and YteA proteins, both annotated as “uncharacterized” in UniProt, which belong to COG1734, “RNA polymerase-binding transcription factor DksA,” and the uncharacterized proteins YqcK and YqjC that fall into COG0346, “Catechol 2,3-dioxygenase-related enzyme, vicinal oxygen chelate (VOC) family.” Experimental characterization of these genes and other widespread genes of unknown function could provide further insight into the details of sporulation mechanisms in diverse firmicutes.

Evolution of sporulation.

We employed the updated list of core sporulation genes (Table 1) to investigate the evolution of sporulation and compare it with the overall phylogeny of Firmicutes. To this end, we constructed a ribosomal protein-based phylogenetic tree for all 180 analyzed organisms as well as sporulation protein-based and ribosomal protein-based trees for the 76 sporeformers. The maximum-likelihood phylogenetic tree of 180 firmicutes (Fig. S8) was built using IQ-TREE 2 (125) from a concatenated alignment of 54 ribosomal proteins (6,951 total positions, 6,366 unique) essentially as described previously (28, 126). This tree had excellent bootstrap support and largely agreed with the currently accepted 16S rRNA gene-based phylogeny of the Firmicutes. Members of most previously defined Firmicutes families clustered together, typically as clades. The notable deviations from the current taxonomy (56) included members of the class Erysipelotrichia (family Erysipelotrichaceae), mapped within Bacilli as a sister group to Listeriaceae and Lactobacillales, and members of three other classes mapping within Clostridia as follows: (i) all three families of class Tissierellia as a sister group to Peptostreptococcaceae; (ii) members of the class Negativicutes as a sister group to families Peptococcaceae and Syntrophomonadaceae, and (iii) Limnochorda pilosa, the sole member of the class Limnochordia, as a sister group to Sulfobacillus acidophilus and Thermaerobacter marianensis (Eubacteriales family XVII, incertae sedis). This tree also suggested the placement of some organisms with previously obscure taxonomy: Intestinimonas butyriciproducens mapped into Oscillospiraceae, whereas Ndongobacter massiliensis and Ezakiella massiliensis fell within the radiation of the family Peptoniphilaceae (class Tissierellia).

For a bird’s-eye view of the phylogeny of sporulation, we constructed a maximum-likelihood tree based on a concatenated alignment of 41 core sporulation proteins (as listed in Table 1) from all 76 sporeformers in the current set (Fig. 3, left) and compared it with the ribosomal protein-based tree for the same 76 species (Fig. 3, right). This comparison showed a strong, clear trend: with a few exceptions, sporulation protein sets from members of the same family formed well-supported clades. As an example, members of the Planococcaceae in both trees formed a sister group to the Bacillaceae, the only differences being the branching order of Sporolactobacillus terrae, a member of the family Sporolactobacillaceae, and Alkalihalobacillus (formerly Bacillus) halodurans, and two members of the Erysipelotrichaceae, C. innocuum and E. ramosum. The clostridial portions of both trees had closely similar topologies with essentially the same branching order for most groups. Members of the recently recognized classes Negativicutes (family Sporomusaceae), Tissierellia, and Limnochordia mapped within the clostridial lineage but, again, in the same positions on both trees (Fig. 3). The congruence of both trees indicated that the sporulation genes largely evolved by vertical inheritance, without substantial horizontal gene transfer.

FIG 3.

Phylogenic trees for sporulation and ribosomal proteins from Firmicutes. Maximum-likelihood phylogenetic trees were built using IQ-TREE2 (125) with the LG+F+R9 substitution model from concatenated alignments of 41 core sporulation proteins (left panel) and 54 ribosomal proteins (right panel) encoded in the genomes of 76 spore-forming members of the Firmicutes.

The evolutionary history of gains and losses of these sporulation genes (the same 41 except for spoIIQ) was reconstructed using GLOOME, by mapping the patterns of gene presence and absence in each lineage onto the phylogeny of 180 representative Firmicutes, to infer the posterior probabilities of gene presence in the ancestral tree nodes (127). This reconstruction placed all 40 core sporulation genes at the root of the Firmicutes phylogenetic tree and suggested only a few gene gains (jag in three Lactobacillus spp, the dipicolinate transporter-encoding spoVV/yjiH/ylbJ genes in several nonsporeformers, and ycaP in Veillonella and Megasphaera). An interesting example of a likely gain was the spo0A gene in Macrococcus caseolyticus, one of the few instances of this gene in non-spore-forming bacilli and the only such gene in the Staphylococcaceae members in our set (in GenBank, spo0A is also found in the members of the newly recognized genus Mammaliicoccus). The maximum-likelihood approach implemented in GLOOME considered the gain of spo0A by M. caseolyticus more likely than the loss of this gene in five other staphylococcal genomes, although the highly degraded sequence of this gene and its unusual placement in the phylogenetic tree of Spo0A (Fig. S3) suggested the latter alternative as more realistic.

With all 40 core genes mapped to the root of the Firmicutes, GLOOME predicted extensive loss of sporulation genes in non-spore-forming lineages, such as lactobacilli, staphylococci, and listeria. A major loss of sporulation genes was also inferred for the asporogenous members of the order Bacillales, such as Salimicrobium jeotgali, Gemella haemolysans, and Exiguobacterium sp. (see Table S1). A similar picture was observed in the classes Negativicutes and Tissierellia. Many sporulation genes are conserved in the sporeformers Methylomusa and Pelosinus in the former class and Gottschalkia and Tissierella in the latter, whereas asporogenous representatives of these classes retained very few sporulation genes (Table S1). Accordingly, GLOOME interpreted the absence of these genes in the nonsporeformers as lineage-specific gene losses. The two lineages mentioned above, Planococcaceae and Erysipelotrichaceae, showed loss of some sporulation genes in their spore-forming representatives and Spo0A-encoding nonsporeformers and a far more extensive loss in the Spo0A⁻ bacteria and nonsporeformers.

Among Clostridia, massive, family-wide gene loss was seen only in Eubacteriaceae, as most other families included both spore-forming and asporogenic genera. Here, dramatic differences in sporulation gene content were often observed within the same family. Thus, in the family Peptostreptococcaceae, the analyzed genome set included five members, two sporeformers (C. difficile and Paeniclostridium sordellii) and three nonsporeformers (Acetoanaerobium sticklandii, Filifactor alocis, and Peptoclostridium acidaminophilum) (49). Accordingly, the first two encompassed nearly complete sets of core sporulation genes (with the exception of those coding for the germination receptor GerABC and several others; see above), whereas the three nonsporeformers showed the loss of at least 29 of the 40 analyzed genes (see Table S2). Further, many clostridial nonsporeformers still carried spo0A and retained a fair number of sporulation genes (Fig. 1; Table S1). As an example, all six representatives of the order Halanaerobiales in the current set were nonsporeformers, but they all carried spo0A, and a major loss of sporulation genes was observed only in Halanaerobium hydrogenoformans (Table S1). This case, as well as the examples of lineage-specific gene loss in Butyrivibrio proteoclasticus (family Lachnospiraceae) or Fastidiosipila sanguinis and Mageeibacillus indolicus (family Oscillospiraceae) (Tables S1 and S2), shows that dramatic loss of sporulation genes can occur over relatively short evolutionary distances.

DISCUSSION

Sporulation is a complex process that involves dozens of genes that function in a tightly regulated fashion to ensure the survival of the bacterial cell under adverse conditions. Spores are resistant to a variety of damaging factors, including extreme conditions, and can persist for thousands—and potentially millions—of years (128). A recent study suggested that spores of B. subtilis could even survive on a simulated Martian surface (100). A better understanding of the sporulation mechanisms could help in finding new approaches to eradicate spore-forming human and animal pathogens. Furthermore, recent studies of the human microbiome have revealed a variety of sporeformers that remain to be characterized in detail (129), suggesting that the impact of Firmicutes sporeformers on human health could be even greater than currently appreciated.

An essential step toward understanding the sporulation process is delineation of the set of genes that are necessary and, possibly, sufficient for sporulation across the entire diversity of bacterial sporeformers. Given that the functions of many genes involved in sporulation remain unknown, those genes that belong to the core set would become priority targets for structural and functional characterization. A major stumbling block for the identification of the core set of sporulation genes is the uncertainty of the sporulation status of many bacteria with sequenced genomes. While many genomes come from well-characterized type species, many others do not, and it is usually not known whether the sequenced strain exactly corresponds to the one that was described previously, in some cases many years ago. A case in point is Acetohalobium arabaticum, which was initially described as a sporeformer (130) but apparently lost the ability to sporulate during the laboratory passages (31, 131); we listed it as a nonsporeformer (Table S1). Conversely, Turicibacter sanguinis, a Spo0A-encoding member of the class Erysipelotrichia, was originally described as asporogenous (132) but subsequently shown to form typical subterminal spores (see Extended Data Figure 4 in reference 129). The genome of Turicibacter sp. strain H121 included in this work has almost the same set of sporulation genes as C. innocuum and E. ramosum but lacks, among others, divIVA, dpaB, spoIIGA, spoVK, and spoVN (Table S2), making this strain unlikely to sporulate. The spore-forming representatives of Erysipelotrichia analyzed in this work, C. innocuum and E. ramosum, were occasionally described as forming very low numbers of spores (133, 134). Nevertheless, there was no doubt that these organisms are sporeformers, which made defining their set of sporulation genes a worthwhile exercise.

Previous studies by several independent groups (2, 32, 34–37) identified essentially the same set of ~60 sporulation genes of B. subtilis that were shared by spore-forming members of Bacilli and Clostridia. However, in the genomes of some sporeformers, certain genes from this conserved core were found to be missing (32, 36, 37). The goal of this work was to reassess the distribution of the key sporulation genes using the widest possible selection of complete genomes from all branches of the Firmicutes. To this end, we employed the latest version of the COG database (39) and supplemented it with five recently sequenced genomes from various members of the Erysipelotrichia. This resulted in a set of 180 species from 160 genera, representing 45 different families from six classes of the Firmicutes. At this scale, most genera were represented only by a single genome of a single species, which provided a different perspective from most previous studies (2, 32, 34–37). Therefore, this work complements, rather than supersedes, those studies.

As discussed elsewhere, the COG database displays the patterns of presence-absence of orthologous genes (COG members) across all covered genomes and therefore allows straightforward identification of those genes that are missing in a given genome (39, 42). However, there is an important caveat: a single COG often includes several paralogous genes, which can give the impression that each of these paralogs has a wider phylogenetic distribution than it actually does. For example, cwlJ and sleB, members of COG3773, “Cell wall hydrolase CwlJ,” are both shown as universal (Table 1), although some genomes, such as that of C. perfringens or C. difficile, may encode only a single member of that family. Likewise, dacB and dacF are paralogous genes that encode closely related d-alanyl-d-alanine carboxypeptidases (penicillin-binding proteins 5* and 1, respectively, members of COG1686) that are involved in peptidoglycan metabolism. However, the former is expressed in the mother cell under SigE control, whereas the latter is expressed in the prespore under the control of SigF and/or SigG. Members of this COG, which also contains a sporulation-independent paralog, dacA, are found in most firmicutes, both spore forming and not.

Further, certain gaps in phyletic patterns can arise from errors in genome sequencing and annotation. In some cases, genes that were frameshifted in the examined genomes (Table S3) were found in full-size versions in the genomes of other strains or closely related species. Although we made an effort to address the potential annotation problems, fixing possible sequencing errors was not part of this project. As a result, we recorded the instances of rarely missing genes in Table S6 but were mostly interested in cases where a certain gene was missing in more than one or two genomes, particularly when these genomes came from the same bacterial lineage.

In our previous work (32), we noted the absence of many key sporulation genes in the relatively large (4.8 Mb) genome of Lysinibacillus sphaericus C3-41 (135) but blamed it on the poor quality of the genome sequence, calling it “the worst offender” among all bacilli and excluding it from most analyses. Here, many of those genes were also found to be missing in the genome of the type strain of this species, L. sphaericus DSM 28. Further, such genes as spoIIB, spoIIM, spoIIIAA, spoIIIAB, spoIIIAD, spoIIIAF, spoVAA, spoVAB, spoVID, bofC, gerM, spmA, spmB, and yqfC were missing in the genomes of all members of the family Planococcaceae (Table S4), to which L. sphaericus was recently reassigned (58). These observations showed that, rather than being “the worst offender” in terms of genome content, L. sphaericus could be a valuable model organism to study the basics of the engulfment process.

The expansion of the initial genome set to include two spore-forming members of Erysipelotrichia, C. innocuum and E. ramosum, provided a window into an even greater loss of core sporulation genes (Table S4). In these organisms, sporulation appears to be less frequent than in members of other lineages, making them somewhat similar to oligo-sporulating mutants of B. subtilis. Nevertheless, the ability of C. innocuum and E. ramosum to form spores in the absence of numerous core genes (Table S4) makes them useful model organisms for studying minimal requirements for sporulation among the Firmicutes.

The conservation of core sporulation genes in Bacilli and Clostridia suggested that the ability to form spores was a common ancient feature of the Firmicutes (2, 3). Although the scenarios of the origin of Gram-negative (diderm) bacteria from spore-forming Gram-positive (monoderm) ones (136, 137) do not seem to be supported by phylogenetic trees of universal genes (126, 138, 139), understanding the fundamental mechanisms of sporulation and clarifying the origin of sporeformers remains a major goal in microbiology. Here, we used an expanded collection of the firmicute genomes to trace the phylogeny of the core sporulation genes within this phylum. Despite a certain variability in the branching order, proteins coming from the members of the same phylum typically formed well-supported clades. This was particularly clearly demonstrated by a comparison of the trees built from concatenated alignments of sporulation proteins and ribosomal proteins (Fig. 3). We and others have previously demonstrated that such a ribosomal protein-based tree is closely similar to the 16S rRNA gene-based tree and faithfully represents the organismal phylogeny of the Firmicutes (28, 49, 140). The remarkable congruence of these trees strongly suggested vertical inheritance of the sporulation genes from a common ancestor of all Firmicutes. Further, in the Spo0A tree (Fig. S3), proteins from nonsporeformers often grouped with Spo0As from sporeformers of the same bacillar or clostridial family, suggesting that Spo0A-encoding nonsporeformers lost the ability to form spores, rather than acquired their spo0A genes. Accordingly, non-spore-forming Spo0A⁺ bacteria typically retain many more sporulation genes than Spo0A⁻ nonsporeformers (Fig. 1). These observations lend further support to the hypothesis that the ability to form (endo)spores was an ancestral feature of the Firmicutes that emerged in this phylum at the early stages of bacterial evolution (3, 32). This ability remains a unique feature of some Firmicutes, distinguishing them from all other bacterial lineages. The evolution of Firmicutes from their spore-forming last common ancestor involved extensive, but in most cases incomplete, loss of sporulation genes in lineages that lost the sporulation capacity. Subsets of the sporulation genes were also lost in several lineages of sporeformers. Experimental study of sporulation in these organisms with reduced complements of sporulation genes can be expected to yield further insight into the minimal biochemical requirements for sporulation.

Despite the substantial progress made in the past several years, a large fraction of sporulation genes remain uncharacterized. For some genes, this is reflected in their y names, but many genes with specific names have been characterized only with respect to the timing of their expression, as being controlled by sporulation-specific sigma factors, or as encoding proteins of the spore coat, whereas their biochemical activities remain obscure. For example, spore maturation proteins SpmA and SpmB, which are conserved in all sporeformers except for the members of Planococcaceae (Table S6), participate in spore core dehydration and are required for heat resistance of spores (141, 142). However, the biochemical mechanisms of spore dehydration by SpmA and SpmB remain unknown. We hope that phylogenetic profiles of the key sporulation proteins derived in this work (Table S2) will help in identifying priority targets for experimental research and allow the selection of suitable model organisms for such studies.

MATERIALS AND METHODS

Genome coverage.

The list of Firmicute genomes used in this work was taken from the recent release of the COG database (39). The COG genome set includes complete genomes of 175 members of the Firmicutes: 73 members of the class Bacilli, representing 59 different genera; 79 members of the class Clostridia, representing 76 genera; 10 members of the class Negativicutes; 9 members of the class Tissierellia; 2 members of the class Erysipelotrichia; 1 member of the class Limnochordia; and an additional genome of “Ndongobacter massiliensis,” which remained unclassified at the time of this work (143). Most genera were represented by a single genome, except for five members each of Bacillus spp. and Streptococcus spp., four members each of Clostridium spp. and Lactobacillus spp., three Streptococcus spp., and two Listeria spp. The list of these organisms is available in Table S1 in the supplemental material and also on the NCBI website https://ftp.ncbi.nih.gov/pub/COG/COG2020/data/cog-20.org.csv. Examination of the available publications allowed extraction of sporulation characteristics for 173 of these 175 organisms. For 74 of these, the ability to form spores was documented in the descriptions of the respective species (Table S1). Ninety-nine organisms were non-spore forming, based on the explicit descriptions of the respective strains and/or their lineages (such as the order Lactobacillales and families Acidaminococcaceae, Halanaerobiaceae, Listeriaceae, Peptoniphilaceae, Selenomonadaceae, Staphylococcaceae, and Veillonellaceae). For two organisms, the sporulation status remained unknown, as their (in)ability to form spores had not been properly documented and the respective lineages included both sporeformers and asporogens (Table S1).

Since both members of the Erysipelotrichia included in the COGs were asporogenic, the genomic list for this class was supplemented with five recently sequenced complete genomes: two from spore-forming members of the genus Erysipelatoclostridium, [Clostridium] innocuum ATCC 14501 (GenBank accession number CP048838.1) and Erysipelatoclostridium ramosum DSM 1402 (GenBank accession number CP036346.1), and three genomes of nonsporeformers, Amedibacterium intestinale (GenBank accession number AP019711.1), Faecalibaculum rodentium (GenBank accession number CP011391.1), and Intestinibaculum porci (GenBank accession number AP019309.1). This brought the total set to 180 genomes, 76 of which came from proven sporeformers (Table S1).

Protein selection.

The list of sporulation proteins analyzed in this work was based on the lists of genes that were previously demonstrated to play a role in the sporulation of B. subtilis and/or were regulated either by Spo0A or by the sporulation-specific sigma subunit SigE, SigF, SigG, or SigK (2, 7, 12, 13, 15–18, 33, 34, 123, 144), as documented in the SubtiWiki database (38).

The selected genes were matched against the COG database (39), and the patterns of presence or absence of the respective COGs in the selected organisms were recorded in Table S2 in the supplemental material (in some cases, a single COG included several paralogous sporulation genes). The resulting set consisted of 237 protein families (COGs), which included 112 COGs for widely conserved genes from the previous COG releases, such as ald (spoVN), ftsH (spoVK), ftsI (spoVE), ftsK (spoIIIE), spmA, spmB, spoVG, spoVS, and so on (Table S2). That list was expanded by the addition of 125 sporulation COGs that were made available in the recent release of the COG database (39). While this list mostly contained genes (proteins) characterized in B. subtilis, it also included clostridial forms of SpoIIQ (COG5833) and small acid-soluble spore protein (SASP) (COG5864), as well as the Amidase_6 domain (COG5877). The widely conserved sporulation genes (proteins) were assigned to one of the following eight functional groups: (i) sporulation onset and checkpoints (22 genes); (ii) Spo0A-regulated genes (9 genes); (iii) engulfment (12 genes); (iv) spore maturation, SigF- or SigG-regulated forespore expression (19 genes); (v) spore maturation, SigE- or SigK-regulated mother cell expression (25 genes); (vi) spore cortex synthesis (9 genes); (vii) spore coat and crust proteins (8 genes); and (viii) germination proteins (8 genes) (see Table S2). Two more groups included sporulation genes conserved in Bacilli (51 genes) and narrowly represented genes (74 genes) (Table S2). This selection was based on the previous studies (2, 15, 16, 28, 32, 36, 37) and the data from the SubtiWiki database (http://subtiwiki.uni-goettingen.de/) (38).

Verification of the COG profiles.

The patterns of presence-absence of the sporulation genes in the 180 analyzed genomes were extracted from the COG master file cog-20.cog.csv (available at https://ftp.ncbi.nih.gov/pub/COG/COG2020/data/ folder) and matched against the assembly numbers listed in Table S1, and the counts of each gene in the 76 sporeformer genomes were calculated, ignoring the paralogs. The cases of apparently essential sporulation proteins missing in certain sporeformer genomes were verified using a recent version of the TBLASTn program (55) that allows selection of specific target organisms (or lineages) based on their entries in the NCBI Taxonomy database (56). Proteins from either the B. subtilis reference set or closely related organisms (same genus or family) were used as queries. The resulting BLAST hits (cutoff E value, 0.1) were verified using CD-search (145) and compared against the protein sets in GenBank and RefSeq databases. Confirmed sporulation genes were used to correct the COG profiles. Newly translated sporulation ORFs will be reported to GenBank and/or RefSeq.

Phylogenetic analysis.

Unless indicated otherwise, protein sequences for phylogenetic analyses were aligned with MUSCLE v5 (146), and the maximum-likelihood trees were inferred using a locally installed version of IQ-TREE 2 (125) with the best-fit model automatically selected by ModelFinder (147) and branch support calculated using a Bayesian-like approximate-likelihood test (148). For each COG that included paralogs, a consensus was produced as described in reference 149; a pairwise similarity score between each sequence and the consensus sequence was calculated using the BLOSUM62 matrix (with the value of −4 assigned to a gap in the sequence, matching a nongap character of the consensus sequence); the paralog with the maximum score was used in the alignment. Individual phylogenetic trees, constructed using IQ-TREE 2 (125), are available from the authors.

For the phylogenetic analysis of Spo0A, the alignment included sequences from all 118 Spo0A-encoding organisms in the analyzed set: 76 sporeformers, 40 nonsporeformers, and 2 organisms with uncertain sporulation status (Table S1). The MUSCLE alignment was edited to remove variable-length linkers between the REC and DNA-binding domains (corresponding to residues 131 to 148 in the Spo0A of B. subtilis), which left 225 informative sites. The maximum-likelihood tree of Spo0A (Fig. S3) was constructed with IQ-TREE 2 (125) using the LG+R6 substitution model.

The SpoIVCA tree (Fig. S4B) was constructed from an alignment of 1,220 SpoIVCA-like site-specific DNA recombinases/integrases from bacteria and phages that were identified in a PSI-BLAST (55) search. The alignment contained 476 informative positions. The approximately maximum-likelihood tree was built with FastTree 2 (150) using the “–gamma” option to rescale the branch lengths and the WAG (151) model of amino acid evolution.

For phylogenetic analysis of sporulation, alignments of 41 core sporulation proteins from 76 sporeformers were constructed (the 40 genes marked in Table 1 with the addition of YmfB [COG0740], a SASP-degrading paralog of ClpP [152]). The few missing proteins (Table S2) were replaced with the respective sequences from closely related organisms (the same genus, where available). For the overall phylogenetic tree of sporulation core, these 41 alignments were concatenated, resulting in an alignment with a total of 10,129 informative sites, and the maximum-likelihood tree (Fig. 3, left) was built using IQ-TREE 2 (125) with the LG+F+R9 substitution model. The ribosomal protein-based tree (Fig. 3, right) was built from a concatenated sequence alignment of 54 ribosomal proteins from 76 sporeformers with IQ-TREE 2 and the LG+R8 substitution model.