Phylogenetic Framework and Molecular Signatures for the Main Clades of the Phylum Actinobacteria

CSIs That Are Uniquely Present in Most Actinobacteria

We have previously described two CSIs, consisting of a 2-aa deletion in cytochrome c oxidase subunit 1 (Cox1) and a 4-aa insert in CTP synthetase, that were uniquely present in almost all actinobacteria except for the deepest-branching genus, Rubrobacter (97). A 5-aa insert in glutamyl-tRNA synthetase (GluRS) was also identified, but it was lacking in several actinobacterial species (97). Additionally, a large insert in the 23S rRNA is also specific for most actinobacterial species (97, 248). Our recent analyses of protein sequences from actinobacterial genomes identified 6 additional CSIs in various proteins that are uniquely shared by most of the sequenced actinobacterial species. These CSIs include a 4-aa insert in the protein glucosamine-fructose-6-phosphate aminotransferase (Gft) (Fig. 3), which catalyzes the formation of glucosamine 6-phosphate and is the first and rate-limiting enzyme of the hexosamine biosynthetic pathway (290); a 3-aa insert in the enzyme glycyl-tRNA synthetase (GlyRS) that is required for protein synthesis (see File S2 in the supplemental material); a 4- to 6-aa insert in the enzyme tRNA (guanine-1)-methyltransferase (TrmD) that methylates guanosine 37 in various tRNAs (see File S3 in the supplemental material) (230); a 4-aa insert in gyrase A, which plays an essential role in DNA replication and transcription due to its ability to make transient double-strand breaks in DNA to maintain appropriate levels of supercoiling (see File S4 in the supplemental material) (185); a 9-aa insert in the enzyme S-adenosyl-l-homocysteine hydrolase (SAHH) that hydrolyzes S-adenosyl-homocysteine, which is an end product of various methylation reactions (see File S5 in the supplemental material) (143); and, finally, a 5-aa insert in the enzyme serine hydroxymethyltransferase (SHMT), which catalyzes the reversible interconversion of serine and glycine (see File S6 in the supplemental material) (117, 238).

A partial sequence alignment of the protein glucosamine-fructose-6-phosphate aminotransferase showing the Actinobacteria-specific insert is presented in Fig. 3. The absence of this indel in the Archaea as well as other bacterial phyla provides evidence that this indel constitutes an insert in the Actinobacteria rather than a deletion in other groups. The sequence alignments for other newly identified CSIs that are uniquely present in most Actinobacteria are provided in Files S2 to S6 in the supplemental material. In all of these proteins, the identified CSIs are present in highly conserved regions. Table 2 presents information regarding the specificity as well as the presence or absence of these CSIs as well as the CSIs in Cox1, CTP synthetase, and 23S rRNA in different genera of Actinobacteria. As shown in Table 2, most of these CSIs are highly specific characteristics of the phylum Actinobacteria. The CSIs in Cox1, CTP synthase, Gft, TrmD, and 23S rRNA are not found in any other bacteria except Actinobacteria, whereas for the other signatures, CSIs of similar lengths were also present in a small number of distantly related organisms, which could be due to either LGT or the independent occurrence of similar genetic changes in these lineages. From the species distribution profiles of these CSIs, it is clear that while most of these CSIs are commonly shared by virtually all sequenced genera belonging to the class Actinobacteria, they are generally not found in the deeper-branching lineages of Actinobacteria. Of these CSIs, only the Cox1 CSI was present in the genus Acidimicrobium, while the genus Conexibacter contained CSIs in the SAHH and SHMT proteins. However, none of these CSIs were detected in Rubrobacter or the Coriobacteriia. For some of these proteins, their homologs were also not detected in most of the Coriobacteriia (Table 2).

CSPs That Are Specific for the Phylum Actinobacteria

In addition to these CSIs, our Blastp analyses of several actinobacterial genomes (viz., M. leprae TN, Leifsonia xyli subsp. xyli strain CTCB07, Bifidobacterium longum NCC2705, and T. fusca YX) previously identified 29 CSPs that were indicated to be specific for either all or most genome-sequenced Actinobacteria (100). Since the number of sequenced actinobacterial genomes has increased from 25 at that time to >150 at present, the Actinobacteria specificities of these proteins were reexamined. Of the 29 proteins that were reported to be specific for all (or most) of the Actinobacteria, 24 are still specifically present in all of the sequenced actinobacterial genera, except for a few of the deepest-branching lineages (see File S7 in the supplemental material). A summary of the properties of these proteins and information regarding their Actinobacteria specificities are provided in Table 3. Except for Actinobacteria, homologs showing significant similarity to these proteins are not found in any other bacterial phyla. Five proteins that were previously retained despite their presence in some other bacterial groups are now excluded from Table 3. The 24 proteins listed in Table 3 are present in virtually all sequenced genera (total of 57) belonging to the class Actinobacteria (see File S7 in the supplemental material). The homologs of two of them, viz., ML1306 (GenBank accession number NP_301939.1) and ML1009 (accession number NP_301746.1), were also found in Rubrobacter xylanophilus, Conexibacter woesei, and Acidimicrobium ferrooxidans, belonging to the classes Rubrobacteria, Thermoleophilia, and Acidimicrobiia, respectively. Homologs of four additional proteins (viz., ML0642, ML1029, ML0760, and ML0804) were also present in one or two of these three classes (see File S7 in the supplemental material). However, significantly, of all the CSPs identified by comparative genomic analyses, the homologs of none of them were detected in any of the members of the class Coriobacteriia.

Predictive Value and Usefulness of the Identified CSIs and CSPs for Delimiting the Phylum Actinobacteria

The results obtained with various CSIs and CSPs are significant in a number of respects. First, they provide important information for validating the specificity and reliability of these signatures. Many of these CSIs and all of these CSPs were identified when the number of sequenced actinobacterial genomes was very limited (97, 100). However, despite a large increase in the number of sequenced genomes (between 6- and 10-fold) for both Actinobacteria as well as other bacteria, most of these signatures are still specific for Actinobacteria. Additionally, most of these signatures are present in virtually all sequenced genera of Actinobacteria, except those from the deepest-branching lineages. Thus, the presence of these signatures can be used to distinguish member species belonging to the class/phylum Actinobacteria from all other bacteria with a high degree of predictive ability. This inference is further strongly supported by our previous work, where the presence of CSIs in Cox1, CTP synthase, and 23S rRNA was examined in a large number of other Actinobacteria belonging to different families, whose genomes have not been sequenced, by PCR amplification of the corresponding fragments (97). The results of those studies showed that of the 50 gene fragments for these three genes that were sequenced from diverse members of the Actinobacteria, all contained the indicated indels, thereby providing strong evidence that these CSIs are distinctive characteristics of various Actinobacteria, even those for whom sequence information is not available at present (97).

Based upon the presence of these CSIs and CSPs, the class Actinobacteria, which comprises more than 90% of the known actinobacterial genera, can now be delimited and circumscribed in clear molecular terms based upon large numbers of independent molecular markers that are unique characteristics of different members of this class (Fig. 3 and Tables 2 and 3, and see Files S2 to S6 in the supplemental material). Based upon the two CSPs that are uniquely found in the class Actinobacteria and members of the classes Acidimicrobiia, Rubrobacteria, and Thermoleophilia, a case can also be made that these bacterial groups are specifically related to the class Actinobacteria and that they should thus be part of the phylum Actinobacteria. However, detailed analyses of the genomes of Actinobacteria have not identified any CSP or CSI that is commonly shared by the above-mentioned classes Actinobacteria and Coriobacteriia, which is now represented by five sequenced genomes (Table 1) (108, 195, 251). This observation in conjunction with the fact that the Coriobacteriia are separated from all other members of the Actinobacteria by a long branch in the phylogenetic tree (Fig. 2) makes a strong case for the exclusion of the Coriobacteriia from the phylum Actinobacteria. It should be noted in this context that the absence of various CSIs and CSPs in Symbiobacterium thermophilum, which was previously placed into the phylum Actinobacteria, argued against its inclusion within this phylum (97, 100). This inference was later strongly supported by its genome sequence and other lines of evidence (174, 310), and this species is now grouped with the Firmicutes (77). No sequences are available at present for the two genera (viz., Euzebya and Nitriliruptor) that make up the class Nitriliruptoria (77, 176, 191). Hence, the affiliation of Nitriliruptoria with other classes of the phylum Actinobacteria (viz., Actinobacteria, Acidimicrobiia, Rubrobacteria, and Thermoleophilia) cannot be confirmed at present.

CSIs and CSPs That Are Specific for the Order Corynebacteriales

Analyses of protein sequences from Corynebacteriales genomes have identified many CSIs and CSPs that are specific for members of this order. In a macrolide transporter ATP-binding protein, a 2-aa insert in a conserved region is specifically present in all of the Corynebacteriales but no other Actinobacteria (Fig. 5A). Likewise, in the enzyme alpha-ketoglutarate decarboxylase (KGD), which is a part of the tricarboxylic acid cycle (301), a 1-aa deletion in a conserved region is uniquely present in all available Corynebacteriales sequences (see File S8 in the supplemental material). Although sequence information is shown for only a limited number of Corynebacteriales and other Actinobacteria, these indels are highly specific characteristics of all Corynebacteriales and are not found in any other Actinobacteria. Another CSI consisting of a 1-aa insert that is largely specific for the order Corynebacteriales is found in the chromosome segregation DNA-binding protein (ParB) (see File S9 in the supplemental material), which binds to DNA at the origin of replication and is involved in chromosome partitioning (156). The conserved insert in ParB is again present in all of the sequenced genera of Corynebacteriales, and with the sole exception of Leifsonia xyli, it is not found in any other Actinobacteria or in other phyla of bacteria. The presence of this indel in L. xyli could be due to LGT or could result from other possibilities that we cannot distinguish at present. Interestingly, the insert in ParB, although it is present in most of the genome-sequenced Corynebacterium species, is not found in C. aurimucosum and a number of other Corynebacterium species (viz., C. ammoniagenes, C. pseudogenitalium, C. tuberculostearicum, C. accolens, C. striatum, and C. glucuronolyticum), whose genomes are not sequenced and which are not shown in the phylogenetic tree in Fig. 4. In a phylogenetic tree based on ParB protein sequences, the Corynebacterium species lacking this insert formed a distinct clade (see File S10 in the supplemental material). Hence, the most plausible way to explain the species distribution of this indel is that the genetic change leading to this occurred in a common ancestor of the Corynebacteriales, followed by the loss of this CSI from this gene, or LGT of this gene, in this particular subclade of Corynebacterium.

In addition to these CSIs, our Blast analysis of various proteins from the genome of Corynebacterium glutamicum ATCC 13032 identified four CSPs (Table 4), for which homologs showing significant sequence similarity are restricted to all of the sequenced Corynebacteriales species but are not detected in other bacteria. Two of these proteins, viz., arabinosyltransferase (EmbB) and AftA, are involved in the synthesis of cell wall arabinan (6, 24, 259), whereas the other proteins are of unknown functions.

Molecular Signatures of Mycobacteriaceae/Mycobacterium

The family Mycobacteriaceae contains a single genus, Mycobacterium, which harbors some of the most important human pathogens, including those responsible for tuberculosis and leprosy (103, 142, 191, 252, 264). Sequence information for large numbers of species of this genus is now available (viz., M. tuberculosis, M. abscessus, M. avium, M. bovis, M. gilvum, M. leprae, M. marinum, M. ulcerans, and M. vanbaalenii) (Table 1) (1, 32, 55, 56, 88, 101, 186, 204, 242, 260, 288, 342). Multiple strains have been sequenced for a number of species. Mycobacterium species have been divided into two major groups (slow growers and fast growers) depending upon their growth rates (142). The species of these two groups generally branch distinctly in phylogenetic trees (232, 285). Their distinctness is also supported by the presence of a longer helix between positions 451 and 482 in the 16S rRNA gene in the slow growers than in the fast growers (232). Of these, the slow-growing Mycobacterium species/strains are clinically important, whereas the fast growers are ecologically important (142). In the phylogenetic tree shown in Fig. 4, all of the sequenced Mycobacterium species/strains formed a strongly supported clade, and within it, a cluster consisting of the slow-growing Mycobacterium species was also strongly supported. We have identified a number of CSIs and CSPs that are specific for either all sequenced Mycobacterium species or the slow-growing clade. Sequence information for one of the CSIs that is specific for the genus Mycobacterium is presented in Fig. 5B. In the enzyme pantoate-beta-alanine ligase, which is involved in the metabolism of beta-alanine (200), a 1-aa insert in a conserved region is uniquely present in all of the sequenced Mycobacterium species but is not found in any other bacteria (Fig. 5B). Similarly, in the enzyme orotidine-5′-phosphate-decarboxylase (OMP-decarboxylase), which catalyzes the last essential step in the de novo biosynthesis of pyrimidines (199), a 1-aa deletion is specifically present in all Mycobacterium species (see File S11 in the supplemental material). Both these signatures are highly specific for the sequenced Mycobacterium species and provide novel molecular markers for this genus.

In our earlier work, Blastp searches for various proteins from the genome of M. leprae TN led to the identification of 24 CSPs that were indicated to be specific for the genus Mycobacterium (100). A reevaluation of the specificity of these proteins by Blastp searches revealed that all of these proteins are still specific for the genus Mycobacterium. However, of these, the first 18 proteins listed in Table 5 are specifically present in all of the sequenced Mycobacterium genomes (with isolated exceptions as noted), whereas the last 6 proteins are limited to the subclade of slow-growing Mycobacterium species (viz., Mycobacterium bovis, M. tuberculosis, M. ulcerans, M. marinum, M. avium, M. paratuberculosis, M. leprae, and Mycobacterium sp. strain JDM601), which are clinically important members of this genus. Although the exact cellular functions of most of these proteins remain to be determined, some of them are putative virulence factors belonging to the PE/PPE or Lpq family of proteins (158, 289).

Molecular Signatures of Rhodococcus and Nocardia

The species of the genera Rhodococcus and Nocardia form a strongly supported clade in various phylogenetic trees (Fig. 2) (3, 5, 111, 136, 178). The distinctness of species of these two genera from all other genera or families of the order Corynebacteriales is also strongly supported by several CSIs and CSPs that we have identified. A partial sequence alignment of one protein showing a CSI that is specific for Rhodococcus and Nocardia species is presented in Fig. 6. In a protein annotated as an ATP-binding protein, a 3-aa insert in a conserved region is specifically present in species of these two genera. Similarly, another CSI consisting of a 1-aa deletion in a conserved region is found in the alpha-subunit of the enzyme acetyl coenzyme A (acetyl-CoA) carboxylase (ACC), which catalyzes the irreversible carboxylation of acetyl-CoA to produce malonyl-CoA (51) (see File S12 in the supplemental material). Both these indels are not found in Gordonia bronchialis or any other Actinobacteria. Our Blastp searches of the genome of Rhodococcus jostii RHA1 have led to the identification of 14 CSPs whose homologs are specifically found in Rhodococcus and Nocardia species (Table 6), except for isolated exceptions. However, in our analyses, we have not come across any CSI or CSP that is commonly shared by Rhodococcus and Nocardia species as well as by Gordonia bronchialis, whose genome has been sequenced (153). These observations make a strong case that the family Nocardiaceae should be limited to the genera Rhodococcus and Nocardia, as it was in the past (103, 136), and that the genus Gordonia, which was part of the family Gordoniaceae, should not be merged with it, as was recently proposed (191). There is no sequence information available at present for the genera Skermania and Williamsia to determine if they are specifically related to Rhodococcus and Nocardia.

In addition to these CSIs and CSPs that are shared by both Rhodococcus and Nocardia species, we have also identified a 3-aa insert in a hypothetical protein, BlinB_00480, that is specifically shared by all four sequenced Rhodococcus species (viz., R. jostii, R. opacus, R. equi, and R. erythropolis), providing a molecular marker for this genus (see File S13 in the supplemental material).

Molecular Signatures of Corynebacterium and the Corynebacteriaceae

The genus Corynebacterium contains numerous species that are of much interest due to their involvement in human and animal diseases (viz., C. diphtheriae, C. striatum, C. jeikeium, C. urealyticum, C. ulcerans, and C. pseudotuberculosis) and also for large numbers of industrial applications, including the production of amino acids, nucleotides, and other nutritional factors; hydrocarbon degradation; and the bioconversion of steroids, etc. (27, 37, 57, 94, 113, 149, 166, 224, 267, 327). As a result, large numbers of genomes of Corynebacterium species and strains, including C. aurimucosum, C. diphtheriae, C. efficiens, C. glutamicum, C. jeikeium, C. kroppenstedtii, C. pseudotuberculosis, C. ulcerans, and C. urealyticum, have been sequenced, and many are in the process of being sequenced (39, 159, 213, 267, 298, 299, 304, 339). The family Corynebacteriaceae contains two genera, Corynebacterium and Turicella (57, 103, 187, 191). However, currently, no sequences are available for the latter genus, which also lacks mycolic acids, which is a uniquely shared characteristic of most other members of the Corynebacteriales (58, 111, 187, 191). In phylogenetic trees based upon 16S rRNA (187, 227, 343) or concatenated protein sequences (Fig. 2 and 4), Corynebacterium species formed a strongly supported clade, and it was separated from other Corynebacteriales species by a long branch. In the tree shown in Fig. 4, Dietzia was its closest relative, and a clade consisting of these two genera was also strongly supported. Within the clade consisting of Corynebacterium species, a number of distinct clusters or subclades were also well resolved (Fig. 4). The distinctness of Corynebacterium species from all other members of the Corynebacteriales and the existence of a number of discrete clades within this genus are also independently supported by many CSIs and CSPs that we have identified.

The presence of an arabinogalactan polymer in the cell wall is a unique characteristic of members of the orders Corynebacteriales and Pseudonocardiales (111, 182, 187). The enzyme phosphoribose diphosphate:decaprenyl-phosphate phosphoribosyltransferase (UbiA) plays an essential role in this process by catalyzing the transfer of ribose-5-phosphate from phosphoribose diphosphate to decaprenylphosphate to form decaprenylphosphoryl-5-phosphoribose (198). In this enzyme, we have identified a 2-aa insert in a conserved region that is uniquely present in all of the sequenced Corynebacterium species (Fig. 7A) but not in any other Actinobacteria. Similarly, in the enzyme acetate kinase, which carries out the phosphorylation of acetate to produce acetyl phosphate, a 3-aa insert in a conserved region is specifically present in all available sequences of Corynebacterium species (see File S14 in the supplemental material). Another CSI that is specific for Corynebacterium is present in the enzyme protoheme IX farnesyltransferase (CyoE), which is involved in the biosynthesis of heme A (250). Most Corynebacterium species have a 7-aa insert in this protein; however, C. jeikeium and C. urealyticum, which form a distinct clade, contain a longer insert (10 aa) in the same position (see File S15 in the supplemental material). Blast searches for various proteins from the genome of Corynebacterium glutamicum ATCC 13032 have also identified 20 CSPs that are uniquely present in all or most of the sequenced Corynebacterium species (Table 7). While 16 of these 20 CSPs are entirely specific for the genus Corynebacterium, the homologs of three of them are also present in Dietzia cinnamea (belonging to the family Dietziaceae), which forms the outgroup of the Corynebacterium clade in the phylogenetic tree (Fig. 4). The shared presence of these CSPs in Corynebacterium and Dietzia cinnamea supports the inference from the phylogenetic tree (Fig. 4) that species of these two families are distantly but specifically related to each other.

In the phylogenetic tree shown in Fig. 4, C. diphtheriae, C. pseudotuberculosis, C. ulcerans, C. aurimucosum, C. glutamicum, and C. efficiens formed a distinct cluster (marked as cluster I) within the genus Corynebacterium. The existence of this clade is also strongly supported by a number of identified CSIs and CSPs. One example of a CSI that is specific for cluster I Corynebacterium species is shown in Fig. 7B. In this case, in the β′-subunit of RNA polymerase (RpoC), which is highly conserved and universally distributed, a 7- to 8-aa insert in a conserved region is specifically present in all of the cluster I Corynebacterium species, but it is not found in other species, such as C. jeikeium, C. urealyticum, and C. kroppenstedtii, that are not part of this clade. Another 2-aa insert that is specific for cluster I species is present in a conserved region of the GTP-binding protein LepA (see File S16 in the supplemental material), which plays an important role in protein synthesis, particularly under stress conditions (236). For some species that contain the RpoC and LepA inserts (viz., C. matruchotii, C. striatum, C. ammoniagenes, C. accolens, C. lipophiloflavum, C. tuberculostearicum, and C. glucuronolyticum), because their genomes were not sequenced, sequence information is not present in the phylogenetic tree (Fig. 4). However, based upon the shared presence of CSIs in both the RpoC and LepA proteins, it is predicted that these species will also group with cluster I Corynebacterium species. The genetic distinctness of cluster I is also strongly supported by 21 CSPs that are uniquely present in all of the sequenced species of this cluster (see the first 21 entries in File S17 in the supplemental material). Additionally, File S17 in the supplemental material lists 19 other CSPs that are uniquely found in C. jeikeium and C. urealyticum, which form another strongly supported cluster in the phylogenetic tree (Fig. 4) (187). These CSIs and CSPs provide novel molecular markers for the identification and circumscription of the genus Corynebacterium and two of its clades. It should be noted that the genetic distances between these subclades of the genus Corynebacterium are greater than those observed among or between species of the families Mycobacteriaceae and Nocardiaceae. Hence, it can be argued that species of these subclades should be recognized as distinct genera rather than being part of the same genus. It should also be noted that the various CSIs or CSPs that are specific for cluster I Corynebacterium species or for C. jeikeium and C. urealyticum are not found in C. kroppenstedtii, which is separated from both of these clusters by a long branch (Fig. 4). Unlike other Corynebacterium species, C. kroppenstedtii lacks mycolic acid (298), and its phylogenetic position and the absence of signatures for the other two clusters suggest that it forms a distinct subgroup of Corynebacterium species.

Molecular Signatures Supporting the Deeper Branching of Corynebacterium and Dietzia within the Order Corynebacteriales

Within the Order Corynebacteriales, a clade consisting of Corynebacterium species and Dietzia shows the deepest branching, and it is separated from other Corynebacteriales by a long branch (Fig. 4). Within this clade, the species belonging to the families Mycobacteriaceae and Nocardiaceae generally group together, and the other Corynebacteriales species branch in between these two clades. Our analyses of Corynebacteriales genomes have identified a number of CSPs that further strongly support these relationships. Table 8 lists a number of CSPs that are present in most other Corynebacteriales species except Corynebacterium species and Dietzia. The genes for these proteins likely originated from a common ancestor(s) of the other Corynebacteriales following the divergence of Corynebacterium and Dietzia. Of the proteins listed in Table 8, the first four are found mainly in Mycobacteriaceae and Nocardiaceae species, supporting a closer relationship between these two families. The homologs of the remainder of these CSPs are found in Gordonia bronchialis and also, in some cases, in Segniliparus rotundus, Tsukamurella paurometabola, and Amycolicicoccus subflavus, supporting their branching in between the Corynebacterium-Dietzia clade and the Mycobacteriaceae-Nocardiaceae clade.

Molecular Signatures of Micromonosporales and Identification of a Higher Clade Consisting of the Orders Corynebacteriales, Pseudonocardiales, Glycomycetales, and Micromonosporales

The order Micromonosporales contains a single family, Micromonosporaceae, that is made up of 20 genera (77, 103, 110, 283, 309). However, these genera do not form a distinct clade in the 16S rRNA trees, and species from other groups are interspersed within this order. Hence, this order is presently poorly defined in a phylogenetic or taxonomic sense (191). Genome sequences are now available for this order, from Micromonospora aurantiaca, Micromonospora sp. strain L5, Salinispora tropica, and Salinispora arenicola (Table 1) (229, 309). In the phylogenetic tree constructed based upon concatenated sequences for 35 broadly distributed proteins (Fig. 2), the sequenced Micromonosporaceae species formed a strongly supported clade branching in the neighborhood of Pseudonocardiales. Stackebrandtia nassauensis, which is the only species in our data set belonging to the order Glycomycetales, was most closely related to this group, and a clade consisting of the Micromonosporaceae species and S. nassauensis was strongly supported by the bootstrap score. A clade consisting of these species in turn was part of a larger clade that included all species of the orders Corynebacteriales and Pseudonocardiales.

A number of identified CSIs provide useful information regarding the Micromonosporaceae species and their relationships to other orders of the Actinobacteria. First, in a protein of unknown function, Lxx09300, we identified a 1-aa insert in a conserved region that is specifically present in all sequenced Micromonosporaceae species (Fig. 8B). This CSI provides a potential molecular marker for distinguishing species of this order from those of other Actinobacteria. Second, we have also identified 3 CSIs in important proteins that are uniquely shared by all sequenced species of the orders Corynebacteriales, Pseudonocardiales, Micromonosporales, and Glycomycetales (represented by S. nassauensis). The first of these CSIs consists of a 2-aa insert in the delta subunit of DNA polymerase III (HolB), which is involved in replicative DNA synthesis in bacteria (73) (Fig. 9B). This insert is uniquely shared by all species of these orders but is not found in any other Actinobacteria (except Jonesia denitrificans) or bacteria. Another CSI showing a similar species distribution is present in a highly conserved region of the ribosomal protein S3 (see File S19 in the supplemental material). Lastly, in the enzyme alpha-ketoglutarate decarboxylase (KGD), which is involved in the decarboxylation of alpha-ketoglutarate, a 1-aa insert in a conserved region is commonly present in all species of these orders, but except for Acidothermus cellulolyticus, it is not found in any other Actinobacteria (see File S20 in the supplemental material). The shared presence of these CSIs in these important housekeeping proteins by the above-described orders of Actinobacteria, which also cluster together in the phylogenetic tree, strongly indicates that these orders of Actinobacteria shared a common ancestor exclusive of all other Actinobacteria.

Molecular Signatures of Frankia and Identification of a Clade Consisting of the Orders Corynebacteriales, Pseudonocardiales, Glycomycetales, Micromonosporales, and Frankiales

The order Frankiales is presently comprised of six families: Frankiaceae, Acidothermaceae, Nakamurellaceae, Cryptosporangiaceae, Geodermatophilaceae, and Sporichthyaceae (77, 103, 216). Genome sequences are now available for a number of Frankia species (216) as well as a number of other genera (viz., Acidothermus, Nakamurella, and Geodermatophilus) covering three other families (14, 154, 302) (Table 1). As noted above, species of the order Frankiales do not form a coherent phylogenetic lineage, and they branch in a number of independent positions in the 16S rRNA tree and other phylogenetic trees (283, 343). This is clearly seen from the branching positions of G. obscurus, N. multipartite, Acidothermus cellulolyticus, and Frankia species in the tree shown in Fig. 2. As discussed above, G. obscurus and N. multipartite, based upon their branching in the tree and a number of CSIs, are more closely related to the Pseudonocardiales than to the type genus of this order, Frankia, which contains the type species F. alni. Furthermore, although a clade consisting of different sequenced strains of Frankia branches in the proximity of A. cellulolyticus (5), a specific relationship between these species was not supported by our tree. Thus, the order Frankiales, as described currently, cannot be delimited by any means, and its taxonomy needs to be emended. However, we have identified a 7-aa insert in a highly conserved region of the DNA gyrase B protein that is uniquely present in various Frankia species (Fig. 10). In addition to the sequenced genomes, partial information for gyrase B covering this region is available for a large number of other Frankia species and strains (217), and this insert is present in all of them, thus providing a highly specific molecular marker for this genus. In addition to the CSI in gyrase B, sequence information for two additional CSIs that are also specific mainly for Frankia species is provided in File S21 in the supplemental material. These CSIs include a 4- to 5-aa insert in the DNA repair protein RadA and a 3-aa insert in a hypothetical protein, Ncg1. However, besides Frankia, the latter CSIs are also present in a few other Actinobacteria, which could be due to LGTs.

In the phylogenetic tree based on concatenated protein sequences (Fig. 2), although the clade consisting of Frankia spp. branched in the proximity of Micromonosporales, it was not specifically related to this order or to the larger clade consisting of the orders Corynebacteriales, Pseudonocardiales, Micromonosporales, and Glycomycetales (Fig. 2). However, we have identified one CSI consisting of a 1-aa insert in a conserved region of the protein glutamine phosphoribosylpyrophosphate amidotransferase that is uniquely shared by all species of the orders Corynebacteriales, Pseudonocardiales, Micromonosporales, and Glycomycetales and also by various Frankia spp. (see File S22 in the supplemental material). Except for species of these orders, this insert is found only in Propionibacterium acnes and Micrococcus luteus and not in other Propionibacteriales or Micrococcales or other orders of Actinobacteria. This CSI provides suggestive evidence that Frankia spp. may also have shared a common ancestor with the clade consisting of Corynebacteriales, Pseudonocardiales, Micromonosporales, and Glycomycetales.

Based upon the species distribution patterns of various identified CSIs and CSPs that are discussed above, the evolutionary stages in which the genes for these CSPs, or the genetic changes responsible for the observed CSIs, are postulated to have evolved are depicted in Fig. 11. Most of the nodes in this diagram are supported by phylogenetic analysis and independently by many identified molecular markers indicating that these branching patterns are reliable.

CSIs and CSPs That Are Specific for the Order Streptomycetales

The sequence alignments of actinobacterial genomes have led to the identification of 3 CSIs that are of interest. In the enzyme porphobilinogen deaminase (PBGD), which converts porphobilinogen into hydroxymethylbilane and is the third enzyme in the heme biosynthetic pathway (190), a 4-aa insert in a conserved region is specifically present in all sequences of Streptomyces species and also Kitasatospora setae, but it is not found in any other Actinobacteria (Fig. 12A). Similarly, in the enzyme adenylate kinase, which catalyzes the interconversion of adenine nucleotides and plays an important role in cellular energy homeostasis, a 1-aa insert in a conserved region is specifically present in various Streptomyces species and K. setae but not in any other Actinobacteria (see File S24 in the supplemental material). Blastp searches for proteins of the genome of S. coelicolor A3(2) have also identified a number of CSPs, all significant hits of which are present in various sequenced Streptomyces species but not in any other bacteria (Table 10). For the first 5 proteins in Table 10, homologs showing significant similarity were detected in various Streptomyces species but not in K. setae or other bacteria. These proteins could be specific for the genus Streptomyces; however, as the complete genome of K. setae is not yet available, it is possible that homologs of these proteins will also be found in this species. For the next 11 entries of Table 10, homologs were detected in both Streptomyces species and K. setae but not in other Actinobacteria. These CSPs thus could be specific for the entire order Streptomycetales. Due to their specificity for species of the order Streptomycetales, they provide novel molecular markers for distinguishing this group of bacteria from all other Actinobacteria.

CSIs and CSPs That Are Uniquely Shared by the Orders Streptomycetales and Catenulisporales

In the phylogenetic tree shown in Fig. 2, C. acidiphila formed an outgroup of the Streptomyces cluster, indicating that the orders Streptomycetales and Catenulisporales are closely related. This inference is also supported by a 1-aa CSI in the lipid A biosynthesis lauroyl acyltransferase (MsbB) protein, which is uniquely shared by all Streptomycetaceae species, including K. setae, as well as by C. acidiphila but not any other Actinobacteria (Fig. 12B). Further evidence that these two orders of Actinobacteria are closely related is provided by our identification of 3 CSPs listed in Table 10, whose homologs are specifically present in various Streptomycetaceae species as well as C. acidiphila. Based upon phylogenetic evidence as well as the identification of a number of molecular markers that are uniquely shared by species of the orders Streptomycetales and Catenulisporales, these observations make a strong case that the order Catenulisporales, which contains only a limited number of species, should be merged with the order Streptomycetales.

In the phylogenetic tree shown in Fig. 2, a superclade consisting of the orders Corynebacteriales, Pseudonocardiales, Glycomycetales, Micromonosporales, Frankiales, Streptosporangiales, Streptomycetales, and Catenulisporales is strongly supported by its observed bootstrap score (97%). The consensus tree reported previously by Alam et al. (5) also supported a clade consisting of these orders. Although we have not come across any CSI that is specifically shared by species of all of these orders, the placement of the orders Streptosporangiales, Streptomycetales, and Catenulisporales as the outer branches of the large clade shown in Fig. 11 is strongly supported by phylogenetic analyses.

Molecular Signatures of the Bifidobacteriales and Bifidobacteriaceae

Species of the order Bifidobacteriales are generally found in the human gastrointestinal tract, and they are important for establishing and maintaining the homeostasis of the intestinal ecosystem to allow for normal digestion (61, 183, 307, 312, 318). The order Bifidobacteriales is comprised of a single family, the Bifidobacteriaceae, which in turn consists of seven genera, Bifidobacterium, Gardnerella, Scardovia, Parascardovia, Alloscardovia, Metascardovia, and Aeriscardovia (22, 191). Except for the genus Bifidobacterium, which contains 29 species, all other genera are monospecific and contain only a single species (77, 191). Due to the importance of bifidobacteria for human health and also due to their probiotic potential, the genomes of large numbers of Bifidobacterium species and strains as well as Gardnerella vaginalis have been sequenced (15, 93, 102, 141, 164, 183, 254, 262, 292, 306, 313, 314, 318, 326, 337, 344). The genetic, biochemical, and genomic characteristics of Bifidobacterium species were reviewed previously by Ventura and coworkers (28, 61, 307, 308, 313, 314). In addition to Bifidobacterium, sequence information for most of the genes and proteins from the genomes of Scardovia inopinata and Parascardovia denticolens, whose genomes are at assembly stages, is also now available in public databases.

In a phylogenetic tree for the Bifidobacteriales, G. vaginalis was found to branch in between different Bifidobacterium species, making this genus polyphyletic (see File S25 in the supplemental material). In particular, Bifidobacterium animalis was found to branch more deeply than G. vaginalis. Hence, the relationship of G. vaginalis to other Bifidobacterium species and its possible placement in this genus should be considered. Alignments of protein sequences of Actinobacteria species have identified two CSIs that are specific for the Bifidobacteriales. One of these CSIs, consisting of a 1-aa deletion in the ribosomal protein L13, is present in all Bifidobacteriales species, including S. inopinata and P. denticolens, but it is not found in any other Actinobacteria (or other bacteria) (Fig. 13A). Thus, this CSI provides a potential molecular marker for the entire order Bifidobacteriales. Another 1-aa insert in the enzyme glucose-6-phosphate dehydrogenase, which is a part of the pentose phosphate pathway, is uniquely present in various Bifidobacterium species and also in G. vaginalis, but it is not found in Scardovia, Parascardovia, or any other Actinobacteria (Fig. 13B). Thus, this CSI distinguishes the clade consisting of the genera Bifidobacterium and Gardnerella from other genera of this order. Blastp searches for various proteins of the genome of Bifidobacterium dentium Bd1 (318) also identified 16 proteins that are uniquely found in various Bifidobacteriales species as well as 6 CSPs for which all significant Blast hits are from the genera Bifidobacterium and Gardnerella (Table 11). Previously, many CSPs that were specific for B. dentium were also identified (316, 318). These CSPs provide additional markers for distinguishing Bifidobacteriales species from other Actinobacteria.

Molecular Signatures of the Actinomycetales

The order Actinomycetales, which corresponds to the suborder Actinomycineae in the earlier taxonomic scheme (103, 343), contains only one family, the Actinomycetaceae, which is comprised of several medically important genera, such as Actinomyces, Arcanobacterium, Actinobaculum, Mobiluncus, and Varibaculum (103, 191, 253, 343). Of these genera, the genus Actinomyces has been indicated to be quite diverse, and in a phylogenetic tree based upon 16S rRNA, it showed polyphyletic branching into a number of different clusters (253). The genome sequences of Arcanobacterium haemolyticum (336) and Mobiluncus curtisii are now available, and sequence information for most of the proteins from a number of other species (viz., Actinomyces odontolyticus, Actinomyces urogenitalis, Actinomyces coleocanis, and Mobiluncus mulleris) is also available in the NCBI database. Our analyses have identified a number of CSIs that are specific for species of this order. The enzyme deoxy-d-xylulose 5-phosphate reductoisomerase (DXR), which is a part of the nonmevalonate pathway of isoprenoid biosynthesis (245), contains a 12-aa insert in a highly conserved region that is uniquely present in all available sequences of Actinomycetales species, including those of the genera Actinomyces, Arcanobacterium, and Mobiluncus (Fig. 14A). Another CSI consisting of a 6-aa insert that is specific for all sequenced Actinomycetales species is present in the integral membrane protein Lxx09300 (see File S26 in the supplemental material). The high degrees of conservation and specificity of these signatures for species of this order indicate that they provide good and reliable molecular markers for this order of Actinobacteria. Isoleucine tRNA synthetase (IleRS), which is essential for protein synthesis, also contains a 3-aa insert in a conserved region that is specifically present in all available sequences of the genera Actinomyces and Mobiluncus but which is lacking in Arcanobacterium haemolyticum as well as all other Actinobacteria (Fig. 14B). In the phylogenetic tree for Actinobacteria based on protein sequences, A. haemolyticum showed the deepest branching of the three available genera, and a clade consisting of Actinomyces and Mobiluncus species was strongly supported (Fig. 2, and see File S25 in the supplemental material). Thus, it is likely that the genetic change responsible for this CSI took place in a common ancestor of these two genera after the divergence of Arcanobacterium. Lastly, we have also identified a 1-aa deletion in the excision endonuclease UvrC that is specifically present in the two Mobiluncus species (Fig. 14C), providing a molecular marker for this genus.

Molecular Signatures of the Micrococcales and Its Subclades

The order Micrococcales is the most diverse order within the phylum Actinobacteria, containing ecologically, morphologically, and chemotaxonomically divergent species (191, 283). The most updated taxonomic outline of this suborder encompasses 15 families and 86 genera, and information for some of these has been reviewed (38, 78, 157, 169, 191, 258, 282, 343, 343). In view of the importance of species of this order for bioremediation, industrial, and clinical purposes, large numbers of genomes of many species that are part of different genera and families of this order have been sequenced. The sequenced genera include Arthrobacter (202, 203), Beutenbergia, Brachybacterium (180), Cellulomonas (2), Clavibacter (105), Intrasporangium (66), Jonesia (233), Kocuria (296), Leifsonia (205), Renibacterium (328), Rothia, Sanguibacter (155), and Tropheryma (20, 239) (Table 1). In the phylogenetic tree based upon concatenated protein sequences (Fig. 2) or 16S rRNA (191, 343), species of the order Micrococcales are split into a number of clusters, with the orders Actinomycetales and Bifidobacteriales branching between them. Thus, the different families that are presently part of this order do not form a phylogenetically coherent group, and the taxonomy of this order needs to be emended. Hence, novel molecular markers that could serve to define and delimit different subclades of this order are of particular interest.

Our analyses of protein sequences from Actinobacteria have identified some CSIs that are specific for some of the subclades of Micrococcales. In the universally distributed and highly conserved β-subunit of the RNA polymerase (RpoB), a 2-aa insert is present in a conserved region that is specific for clade I Micrococcales (Fig. 15A). This clade, which is comprised of species of the families Micrococcaceae (Arthrobacter, Renibacterium, Micrococcus, Kocuria, and Rothia) and Brevibacteriaceae (Brevibacterium), is also strongly supported in the phylogenetic tree (Fig. 2). Another CSI that is specific for the Micrococcales has been identified in the ribose-5-phosphate isomerase (RPI) protein, which is a key enzyme of the pentose phosphate pathway that catalyzes the conversion of ribose-5-phosphate into ribulose-5-phosphate (340). In this highly conserved protein, a 4-aa insert in a conserved region is specifically present in all of the sequenced Micrococcales that are part of clusters I and III, but this insert is not found in cluster II Micrococcales or any other Actinobacteria (Fig. 15B). It should be noted that although clusters I and II branch in proximity of each other in the tree, they are phylogenetically quite distinct from each other. For cluster III Micrococcales species, although they branch deeply in the tree, their deep branching could be due to a long-branch length effect (82, 97), as Tropheryma, which is a part of this cluster, has a long branch length. Thus, although a clade consisting of cluster I and cluster III Micrococcales is not observed in the phylogenetic tree (Fig. 2), the CSI in the RPI protein suggests that these two subclades of Micrococcales might be more closely related to each other than the cluster II species. We have also identified one additional CSI consisting of a 4-aa insert in the pyruvate carboxylase protein that is uniquely present in various Micrococcales except Tropheryma (see File S27 in the supplemental material). Although homologs of this protein were not detected in all sequenced Micrococcales, this CSI suggests that despite their divergent branching in the phylogenetic trees, all of the Micrococcales might be derived from a common ancestor exclusive of other Actinobacteria.

As noted above, in phylogenetic trees, species of the orders Actinomycetales and Bifidobacteriales branch between the different clusters of Micrococcales, indicating that these orders are closely related. One additional CSI that we have identified supports this inference. In the highly conserved DnaK or Hsp70 family of proteins, a 5-aa insert in a conserved region is present in all of the Bifidobacteriales, Actinomycetales, and Micrococcales (clusters I, II, and III), but with a few exceptions, this insert is not present in most other Actinobacteria (see File S28 in the supplemental material). The presence of this CSI in a few other Actinobacteria could be due to LGTs. The shared presence of this CSI in all species of these actinobacterial orders suggests that they likely shared a common ancestor exclusive of other Actinobacteria.

Molecular Signatures of the Propionibacteriales

The order Propionibacteriales contains the families Propionibacteriaceae and Nocardioidaceae (103, 343). Members of the Propionibacteriaceae thrive in diverse habitats, covering human epidermal surfaces, dairy products, silage, soil, water, Antarctic sandstone, and sewage treatment plants (279). They are either aerobic or facultative anaerobes, have different morphologies, and exhibit different peptidoglycan type variations (279, 281, 343). The genome sequences of several species of this order covering both families are now available. These include sequences of several strains of Propionibacterium acnes as well as those of Propionibacterium freudenreichii, Kribbella flavida, and Nocardioides sp. strain JS614 (34, 79, 235). Additionally, sequence information for large numbers of genes and proteins from Aeromicrobium marinum is also available in the NCBI database. Sequence alignments of actinobacteria have identified two CSIs that are specific for this order. In the helicase DinG, which is involved in DNA repair and replication, a 3-aa insert in a conserved region is specifically present in all available sequences from this order of bacteria but not in any other Actinobacteria or other bacteria (Fig. 16). Another CSI that is specific for this order is present in the cytochrome c oxidase subunit 1 (Cox1) protein (see File S29 in the supplemental material), which also contains a CSI that is specific for most Actinobacteria (97). In this case, all Propionibacteriales homologs contain a 1-aa deletion that is not present in other Actinobacteria (see File S29 in the supplemental material). Both these CSIs provide molecular markers that distinguish species of the order Propionibacteriales from all other Actinobacteria.

Molecular Signatures Identifying Larger Clades Consisting of the Orders Bifidobacteriales, Actinomycetales, Micrococcales, Kineosporiales, and Propionibacteriales

Due to the compact clustering of most actinobacterial orders in the 16S rRNA and protein trees, their branching orders are generally not resolved (Fig. 2). As indicated above, rare genetic changes such as CSIs, due to their rare and highly specific nature, are capable of resolving deep-branching relationships that are not resolved by phylogenetic trees (11, 119, 131, 243, 246). Our analyses have identified a number of CSIs that clarify the evolutionary relationships and branching orders of the actinobacterial orders Bifidobacteriales, Actinomycetales, Micrococcales, Kineosporiales, and Propionibacteriales.

In the phylogenetic tree shown in Fig. 2, the order Kineosporiales branches in the proximity of the order Micrococcales, which is interspersed by the orders Bifidobacteriales and Actinomycetales. Although a clade consisting of these orders is not supported by the 16S rRNA or protein trees (Fig. 2) (5, 343), a number of CSIs provide evidence that species of these orders are specifically related and that they shared a common ancestor exclusive of other Actinobacteria. In the highly conserved and universally distributed ribosomal protein S3, a 5-aa insert in a conserved region is uniquely present in all species of these 4 orders of actinobacteria, but this insert is not found in any other Actinobacteria (Fig. 17). The shared presence of this insert in this important protein involved in the information transfer process strongly suggests that the genetic change leading to this insert was introduced in a common ancestor of these 4 orders of Actinobacteria. Two other CSIs that also support that these orders are specifically related are found in the CgR_2975 and Cox1 proteins. In the CgR_2975 protein, whose cellular function is not known, a 3-aa insert in a conserved region is present in all species of the orders Bifidobacteriales, Micrococcales, and Kineosporiales but not in any other bacteria (see File S30 in the supplemental material). Because homologs of this protein were not detected in Actinomycetales, it is likely that this CSI was introduced in a common ancestor of the orders Bifidobacteriales, Actinomycetales, Micrococcales, and Kineosporiales, followed by the loss of this gene from the order Actinomycetales. Similarly, in the Cox1 protein, which contains two other CSIs, one specific for most Actinobacteria (97) and the other specific for Propionibacteriales (see File S29 in the supplemental material), one additional CSI that is uniquely shared by all species of the orders Micrococcales and Kineosporiales has been identified (see File S31 in the supplemental material). Because Cox1 homologs were not detected in the orders Actinomycetales and Bifidobacteriales, it is likely that this CSI was also introduced in a common ancestor of the above-described 4 orders, followed by the loss of this gene from the orders Bifidobacteriales and Actinomycetales. Recently, the existence of a clade consisting of these 4 actinobacterial orders based upon 16S rRNA trees was also suggested by Ludwig et al. (191). However, in the phylogenetic trees of 16S rRNA genes reported by Zhi et al. (343) and Adekambi et al. (3), this clade was not observed.

Lastly, the triosephosphate isomerase protein, which plays a key role in glycolysis, contains a 2-aa insert in a conserved region that is commonly shared by all species of the orders Bifidobacteriales, Actinomycetales, Micrococcales, Kineosporiales, and Propionibacteriales but which is not found in any other Actinobacteria or other phyla of bacteria (Fig. 18). The genetic changes responsible for this CSI were likely introduced in a common ancestor of these orders, providing evidence that they are specifically related. The evolutionary relationships among various taxa belonging to these actinobacterial orders that emerge based upon various identified CSIs and CSPs are summarized in Fig. 19.

Usefulness of CSIs and CSPs for an Understanding of the Phylogeny and Taxonomy of Actinobacteria

One of the immediate applications of these signatures is that they provide potentially more definitive means for understanding or clarifying actinobacterial phylogeny and taxonomy. Our understanding of the phylogeny and taxonomy of Actinobacteria currently relies solely on phylogenetic trees based on 16S rRNA (103, 191, 283, 343). Although such trees have been and will remain some of the primary means for understanding microbial phylogeny and taxonomy, some of the limitations of these trees should be recognized (50, 171, 225, 278). While phylogenetic trees in general are most effective in resolving evolutionary relationships at intermediate taxonomic levels (viz., genus, family, and order), their resolving power at either higher (among orders, classes, or phyla) or lower (i.e., among species or different strains of a species) phylogenetic depths is quite limited (192, 278, 286, 324, 343). Additionally, the phylogenetic trees are a continuum, with no fixed boundaries. Hence, on the basis of the branching in these trees, it is often difficult to delimit a phylum or any other taxa reliably, unless all members of the proposed clade or taxon share at least some unique and reliable molecular, biochemical, or physiological characteristics. The phylum Actinobacteria represented such a case, where no unique characteristic of any kind was known that was commonly shared by all species that have been assigned to this phylum. In this context, our identification of a number of CSIs and CSPs that are commonly and uniquely shared by most members of all other classes of the phylum Actinobacteria, except Coriobacteriia, argues strongly that the bacteria belonging to the class Coriobacteriia, which branches more deeply than all other Actinobacteria, should at present be excluded from the phylum Actinobacteria. If, in the future, some unique biochemical and/or molecular properties that are specifically shared by Coriobacteriia and Actinobacteria are discovered, the inclusion of Coriobacteriia in the phylum Actinobacteria could be reconsidered.

On the basis of the identified CSIs and CSPs, it is also now possible to identify and delimit most of the main orders within the phylum Actinobacteria in molecular terms (Fig. 11 and 19). For some orders that have been studied in detail (viz., Corynebacteriales, Bifidobacteriales, and Streptomycetales), individual families and genera as well as subclades of some genera (e.g., Corynebacterium and Mycobacterium) can now also be identified in clear molecular terms based upon multiple signatures. Additionally, based upon these molecular signatures, it is also possible to delineate the interrelationships among different orders of Actinobacteria, and several higher levels of clades can be identified (Fig. 11 and 19). These clades include those consisting of (i) the orders Corynebacteriales and Pseudonocardiales; (ii) the orders Corynebacteriales, Pseudonocardiales, Glycomycetales, and Micromonosporales; and (iii) the orders Corynebacteriales, Pseudonocardiales, Glycomycetales, and Micromonosporales and the genus Frankia (Fig. 11). Although the Frankiales species do not form a coherent clade, all sequenced species are part of this larger clade, indicating that they are related to this group of species. Phylogenetic studies also support a larger clade consisting of the orders Corynebacteriales, Pseudonocardiales, Glycomycetales, Micromonosporales, Frankiales, Streptosporangiales, and Streptomycetales, although no molecular signature that is specific for this large clade has thus far been identified. The other higher levels of clades within the phylum Actinobacteria that can be identified on the basis of identified molecular signatures include those consisting of the orders (iv) Bifidobacteriales, Actinomycetales, and Micrococcales; (v) Bifidobacteriales, Actinomycetales, Micrococcales, and Kineosporiales; and (vi) Bifidobacteriales, Actinomycetales, Micrococcales, Kineosporiales, and Propionibacteriales (Fig. 19). Several of these phylogenetic clades were also observed in a consensus phylogenetic tree for a limited number of actinobacteria constructed by using different approaches (5). Additionally, the phylogenetic analysis and molecular signatures reported here provide strong evidence that species of the order Streptomycetales are closely related to Catenulisporales, and a strong case can be made for the merger of Catenulisporales into the order Streptomycetales.

Recently, on the basis of 16S rRNA trees, Ludwig et al. (191) also indicated the identification of two large clades within the phylum Actinobacteria. One of these clades consists of the orders Actinopolysporales, Corynebacteriales, Glycomycetales, Jiangellales, Micromonosporales, Pseudonocardiales, and Propionibacteriales. This clade is similar to one of the large clades identified here, except that our results suggest that the genus Frankia and other species that are currently part of the order Frankiales are also affiliated with this clade, whereas those of the order Propionibacteriales are not part of this clade. There are no genome sequences available at present for the orders Actinopolysporales and Jiangellales. Hence, we are unable to determine the placement of these orders within this clade. The other large clade identified by Ludwig et al. (191), consisting of the orders Bifidobacteriales, Actinomycetales, Micrococcales, and Kineosporiales, is also supported by various identified signatures (Fig. 19). Although the identification of these large clades based upon different CSIs as well as the 16S rRNA trees strongly indicates that these clades are meaningful, it should be recognized that phylogenetic trees are dynamic constructs and that the branching of species within them is dependent upon large numbers of variables and assumptions, including the different species that are part of the data set and the models used to create the sequence alignment and phylogenetic trees (81, 83, 192, 206, 330). This is illustrated by the fact that the two large clades proposed by Ludwig et al. (191) were not observed in the phylogenetic trees for 16S rRNA reported by Zhi et al. (343) and Adekambi et al. (3). In contrast to the highly dynamic (and variable) nature of phylogenetic trees, the inferences derived from CSIs are based upon minimal assumptions, and their interpretation is generally straightforward (119, 126, 132). Based upon these CSIs, all of the identified clades are defined simply based upon the presence or absence of given indels in highly conserved regions of proteins (119, 126, 130, 132). Furthermore, these CSIs provide highly stable molecular markers with strong predictive abilities. This is evidenced by the fact that many of the Actinobacteria-specific CSIs and CSPs, which were identified when the number of sequenced genomes was very limited (97, 100), are still reliable characteristics of this phylum despite the nearly 10-fold increase in the number of sequenced genomes. Additionally, the investigated CSIs are also present in many other actinobacterial species whose genomes have not been sequenced, providing further strong evidence of their reliability and predictive power (97).

The specificity of various identified signatures for actinobacterial species or groups is presently based mainly upon the species and/or strains whose genomes have been sequenced (Table 1). Although these genomes represent only a small fraction of the actinobacterial species (52, 103), they cover most of the major orders and families of Actinobacteria. However, it is of much importance to obtain sequence information for these genes/proteins from other actinobacterial species to further validate and more precisely determine the boundaries of the clades that are defined by these signatures. These signatures are also very appealing for taxonomic studies, as the assignment of various species (or new isolates) to different clades can be readily done based upon the presence or absence of certain diagnostic signatures, without the need for the construction of detailed phylogenetic trees.

Interesting Cases of Lateral Gene Transfers Identified by CSIs and CSPs

Although most of the CSIs and CSPs described in this review are specific for particular clades of Actinobacteria, the shared presence of CSIs or CSPs in unrelated groups of bacteria also provides a novel means for the identification of lateral gene transfers. Two interesting cases of LGTs between Actinobacteria and Chlamydiae that have been identified by these means include those for the genes encoding the enzymes serine hydroxymethyltransferase (SHMT) (or the GlyA protein) and UDP-N-acetylglucosamine enolpyruvyl transferase (MurA) (115, 117). In the enzyme SHMT, which links amino acid and nucleotide metabolisms by generating the key intermediate for one-carbon transfer reactions (238), two CSIs (3 and 31 aa long) are uniquely shared by all Chlamydiae species, Treponema species, as well as a subset of Actinobacteria (117). Interestingly, the actinobacterial species which contain these CSIs have multiple homologs of the glyA gene, and only one of them harbors the indicated CSIs (117). Similarly, in the MurA protein, which plays an important role in the synthesis of cell wall peptidoglycan, a 16-aa CSI was commonly shared by all Chlamydiae and a subset of Actinobacteria (115). In the phylogenetic trees based upon GlyA or MurA protein sequences, the Chlamydiae homologs branched with the various insert-containing Actinobacteria within a clade of other Actinobacteria. These results provide strong evidence that the shared presence of these CSIs in these two groups of bacteria is due to the lateral transfer of genes for these proteins from certain groups of Actinobacteria to a common ancestor of the Chlamydiae (117, 133). It is of much interest to understand the functional significances of the identified CSIs in these proteins and to determine why their genes were laterally transferred from Actinobacteria to the common ancestor of the Chlamydiae. Our work on actinobacterial CSPs has also revealed that homologs of some of them are also found in Magnetospirillum magnetotacticum (100), which is unrelated to the Actinobacteria. The CSI in the glycyl-tRNA synthetase, which is mainly a distinctive characteristic of Actinobacteria, is also found in this bacterium as well as in a few Planctomycetes (Table 2). The shared presence of these CSPs and CSI is again due to LGTs from Actinobacteria to M. magnetotacticum, and it is of much interest to determine what unique properties are shared by these two groups of bacteria.

Application of the Identified Molecular Signatures for Identification of Actinobacteria and Exploring Their Diversity

The phylum Actinobacteria is extremely diverse. In addition to containing many bacteria that are major human, animal, or plant pathogens (e.g., Mycobacterium, Actinomyces, Renibacterium, Atopobium, Gordonia, Gardnerella, Leifsonia, and Clavibacter), other actinobacterial taxa arguably provide the richest source for discovering diverse natural products that have proven to be of seminal importance in clinical and biotechnological applications (12, 21, 36, 45, 86, 87, 220, 249). Thus, it is of much interest and importance to discover novel means by which both known as well as novel species belonging to different actinobacterial groups can be readily and accurately identified in different settings (viz., clinical or environmental). Because some taxa of Actinobacteria (e.g., Streptomyces, Salinispora, Saccharopolyspora, Cellulomonas, Verrucosispora, Pseudonocardia, Micromonospora, Bifidobacterium, and Arthrobacter, etc.) have proven to be particularly important sources for the discovery of novel compounds such as antibiotics and probiotics and compounds useful in bioremediation (42, 110, 162, 197, 220, 221, 308, 325), there is enormous interest in the discovery of novel actinobacterial species belonging to these taxa, which could lead to the discovery of either novel antibiotics or other natural products that can be gainfully employed for various applications (35, 36, 84, 110, 197). As emphasized by Goodfellow and coworkers (110, 323), a sound knowledge of actinobacterial systematics is of particular importance in this regard. A reliable phylogenetic framework for Actinobacteria in conjunction with specific probes for identifying different groups of Actinobacteria can greatly facilitate the discovery of novel actinobacterial species in different environments. In this context, molecular markers (CSIs and CSPs) that are specific for different major clades of Actinobacteria are of particular importance, since probes based on them can serve as novel and specific tools for the identification and discovery of novel actinobacterial species belonging to these taxa. The primary sequences of many of the CSPs and most of the proteins that contain these CSIs are highly conserved. Based upon conserved regions in these genes/proteins, degenerate PCR primers for these genes/proteins can be readily designed, which should specifically amplify gene sequences from these clades (95, 97, 115) and should provide novel means for the identification of new as well as existing actinobacterial species belonging to these clades from different environments. Using these molecular signatures, it should also be possible to readily and more accurately determine the presence or absence of different families and orders of Actinobacteria in metagenomic samples obtained from various environments (35, 109, 110, 147, 201, 241, 276, 277). Likewise, CSIs and CSPs that are specific for the pathogenic Actinobacteria (viz., Mycobacterium, Corynebacterium, Propionibacterium, and Actinomyces) provide novel means for their diagnostics. Some of the CSPs and the proteins containing CSIs that are specific for these groups should also provide potential means for developing vaccines for these bacteria or potential targets for developing drugs that are specific for these bacteria.

Functional Significance of Actinobacterial CSIs and CSPs

An important area for future research is to understand the functional significance of various CSIs and CSPs that are specific for either all Actinobacteria or their various clades. For the phylum Actinobacteria or most of its major clades, no biochemical or physiological characteristics that are unique to them are presently known. Hence, the identified CSIs and CSPs that are specific for different clades of Actinobacteria provide novel means for discovering biochemical and/or other characteristics that are unique to these groups. Most of the identified CSIs are located in widely distributed proteins (e.g., ribosomal proteins, RNA polymerase, gyrase, DNA polymerase, and various enzymes in key metabolic pathways) that are responsible for carrying out essential cellular functions. The primary functions of these proteins are vital for cell survival, and they are expected to remain the same in all organisms. Hence, the question arises, What is the functional significance of these evolutionarily conserved indels that are specific for different actinobacterial lineages?

Recent work on a number of conserved indels in the Hsp60 (GroEL) and Hsp70 (DnaK) proteins showed that the identified CSIs are essential for the groups of species where they are found and that deletions or most changes in them led to a failure of cell growth (269). Based upon this finding, we expect that the CSIs that are specific for Actinobacteria will also be essential for the particular lineages where they are found. An important observation in this regard is that most of these CSIs are generally present in the surface loops of various proteins (4, 118, 269). This is also true for most of the Actinobacteria-specific CSIs described in this work, and it is illustrated by the structures of two of the proteins, viz., S-adenosyl homocysteine hydrolase (240) and serine hydroxymethyltransferase (Fig. 20), which contain 5-aa and 9-aa CSIs, respectively, that are specific for most Actinobacteria (see Files S4 and S6 in the supplemental material for sequence alignments of these proteins). The structures shown in Fig. 20 are from M. tuberculosis, which contains these inserts, and the regions corresponding to the inserts are colored magenta. As shown in Fig. 20, the inserts in both proteins are present in surface loops, and they are seen as patches or knobs on the surfaces of these proteins. The surface loops in protein sequences are known to play an important role in mediating protein-protein interactions, and they can either facilitate or disrupt certain interactions (4, 152). In view of the predicted essential nature of these CSIs and their locations on protein surfaces (generally away from the active sites), we have postulated that these CSIs are involved in conferring novel functional capabilities (i.e., ancillary functions) on these essential proteins through protein-protein or other forms of interactions (269). These ancillary functions are expected to be important for the lineages in which these CSIs are found, and they could include the ability of the protein(s) to interact with other cellular proteins or ligands (with the CSI serving as a docking site) that either modulate the activity of these proteins or confer some new function(s) on them. Recent studies of two large CSIs in the gyrase B and RpoC proteins that are specific for a number of bacterial phyla support this hypothesis (46, 116, 255). Hence, further studies toward an understanding of the cellular functions of these Actinobacteria-specific CSIs should lead to the discovery of novel aspect of many important proteins that contain these CSIs.

Unlike CSIs, which are commonly found in essential proteins of known functions, the cellular functions of most of the CSPs that are limited to particular lineages of Actinobacteria are generally not known. The evolutionary conservation and retention of genes for these proteins by different lineages strongly suggest that they perform important functions (62, 96, 133, 244) that are specific for these lineages and which distinguish them from other Actinobacteria. Hence, an understanding of the cellular functions of these CSPs should provide valuable insights into the biochemical and physiological characteristics that are unique to different taxa of Actinobacteria. The significance of such proteins for particular lineages is illustrated by the examples of the well-studied EmbA, EmbB, EmbC, and AftA proteins, which are CSPs that are limited to either the order Corynebacteriales or the orders Corynebacteriales and Pseudonocardiales (Tables 3 and 9). The species of these two orders have cell wall chemotype IV, defined by the presence of meso-diaminopimelic acid, arabinose, and galactose in their cell walls, and these proteins play key roles in the biosynthesis of arabinan, which is a unique component of their cell walls (7, 24, 106, 259, 263, 300). Thus, the lineage specificity of these proteins correlates with a unique and essential biochemical property of these orders of Actinobacteria.

Of the four CSPs that are distinguishing characteristics of nearly all Actinobacteria, the structures of two of them, viz., SCO1997 and SCO1662 [gene identification from S. coelicolor A3(2), which corresponds to the ML1009 and ML1306 proteins from M. leprae TN], were recently solved (Protein Data Bank [PDB] accession number 3E35) (100). Although the structures of these two related proteins show limited structural similarity to purine nucleoside phosphorylase and the PAC2 family of proteins from the Archaea (PDB accession number 3GAA), at the sequence level, they exhibit no significant similarity to these proteins. Thus, the functions of these Actinobacteria-specific proteins are predicted to be novel. This inference is strongly supported by recent work showing that SCO1662 specifically interacts with the ParA protein, and it likely corresponds to the ParJ protein, which negatively regulates ParA polymerization in vitro, which is important for efficient chromosome segregation in sporulating aerial hyphae (71). However, further studies are needed to understand the roles of the two homologs of this protein (viz., SCO1662 and SCO1997) and why they are uniquely found in Actinobacteria. Similarly, the WhiB protein family, which has several gene copies in all Actinobacteria except the deepest-branching lineages, is indicated to play an essential role in controlling developmental transition in Streptomyces (43, 90). In nonsporulating actinobacterial species such as Mycobacterium species, WhiB proteins are differentially expressed, and they are important in regulating virulence, cell division, antibiotic resistance, and other stress responses (32, 208). These examples indicate that the CSPs that are specific for different actinobacterial lineages likely play important roles in different unique aspects of these bacteria, including their niche adaptation, pathogenic mechanisms, and other genetic, biochemical, and morphological characteristics that are unique to these bacteria. Hence, concerted efforts to understand their cellular functions should provide important insights into the unique biological aspects of these bacteria.