Electing a candidate: a speculative history of the bacterial phylum OP10

Authors


E-mail pfdunfie@ucalgary.ca; Tel. (+1) 403 220 2469; Fax (+1) 403 289 9311.

Summary

In 1998, a cultivation-independent survey of the microbial community in Obsidian Pool, Yellowstone National Park, detected 12 new phyla within the Domain Bacteria. These were dubbed ‘candidate divisions’ OP1 to OP12. Since that time the OP10 candidate division has been commonly detected in various environments, usually as part of the rare biosphere, but occasionally as a predominant community component. Based on 16S rRNA gene phylogeny, OP10 comprises at least 12 class-level subdivisions. However, despite this broad ecological and evolutionary diversity, all OP10 bacteria have eluded cultivation until recently. In 2011, two reference species of OP10 were taxonomically validated, removing the phylum from its ‘candidate’ status. Construction of a highly resolved phylogeny based on 29 universally conserved genes verifies its standing as a unique bacterial phylum. In the following paper we summarize what is known and what is suspected about the newest described bacterial phylum, the Armatimonadetes.

Discovery of OP10

Obsidian Pool is a shallow geothermal spring in Yellowstone National Park, with pH-neutral water at c. 79°C near the source. Its name comes from the black sand deposited at the bottom of the pool, probably volcanic glass coated with pyrite (Shock et al., 2005). In the 1990s, the group of Norman Pace described the microbial community of Obsidian Pool by constructing clone libraries of 16S rRNA genes amplified from DNA extracts of water and sediment (Barns et al., 1994; Hugenholtz et al., 1998). The most remarkable finding of these studies was the detection of 12 new phylum-level phylogenetic clusters within the Domain Bacteria. Bacterial phyla (sometimes referred to as divisions or Kingdoms) are not recognized as valid taxonomic levels by the Bacteriological Code, but have nevertheless been adopted into common usage by microbiologists. They represent the primary branches in the radiation of the Domain Bacteria, a radiation that may have occurred as a kind of evolutionary ‘big bang’ (Pace, 2009). The new phyla found in Obsidian Pool were unaffiliated to any cultured bacteria, and were therefore dubbed ‘candidate divisions’ OP1 to OP12. Hugenholtz and colleagues (1998) defined a candidate division as ‘… an unaffiliated lineage in multiple analyses of data sets with varying types and number of taxa and having < 85% identity to reported sequences, indicating its potential to represent a new bacterial division’. This paper is now considered a landmark work in molecular microbial ecology, and started a rush to discover more unknown lineages on the tree of life. At present there are over 70 proposed bacterial candidate divisions, more than twice the number of phyla containing cultured members (Achtman and Wagner, 2008).

Time, and the growth of the public-domain 16S rRNA gene database, has been kinder to some of the OP candidate divisions than to others. Few sequences closely related (> 90%) to the original OP4, 6 and 7 clones have been found, so reliable phylogenetic placement of these is still difficult. Others tend to cluster within known bacterial phyla during phylogenetic construction, and taxonomic databases list them as either ‘unclassified bacteria’ or as lower-level taxa than phyla (Table 1). For example, OP12 is now usually assigned to the Firmicutes. Hugenholtz and colleagues (1998) recognized these possibilities when originally defining their divisions, stating that ‘Since the depth of a lineage cannot be estimated based on only oneor a few examples, analysis of further sequences will be necessary to confirm or refute the status of a candidate division’.

Table 1. Present taxonomic assignment of the original OP candidate divisions based on the ARB-SILVA reference database (http://arb-silva.de), the RDP Classifier (http://rdp.cme.msu.edu) and the Greengenes database (http://greengenes.lbl.gov).
 Phylum based on ARB-Silva taxonomy (with class or subphylum cluster where defined)RDP taxonomyGreengenes taxonomyARB Reference sequencesaARB Ref. sequences per 10 000Maximum 16S rRNA gene difference within group (%)b
  • The number of 16S rRNA gene sequences belonging to the clusters defined by the ARB reference database (SSUREf_108_SILVA_09_09_11) and the ratios of these per 10 000 database entries (530 197 total Bacteria) are given. The maximum pairwise distance between any two species in groups defined in the ARB-Silva Reference database was calculated using the ‘Calculate Matrix’ function in ARB.

  • a. 

    Refers to the smallest cluster listed in column 1 of this table.

  • b. 

    In order to limit erroneously high values, calculation utilized only high-quality sequences (alignment quality value > 75) that were > 1450 nt long.

  • c. 

    The original OP6 clone (AF027079) is a short sequence (647 bp) that does not appear in the ARB-SILVA reference database, and based on a BLAST search still has no close relatives (> 90% identity).

  • d. 

    The assignment to Elusimicrobia (Geissinger et al., 2009) is very tentative. Distances of the original OP7 sequences to the type strain Elusimicrobium minutum are > 25%.

OP1Candidate division EM19Unclassified BacteriaCandidate division OP1 or unclassified Bacteria721.3630.9
OP2 Thermotogae (EM3)Unclassified BacteriaUnclassified Bacteria470.8928.4
OP3Candidate division OP3Unclassified BacteriaCandidate division OP3 or unclassified Bacteria2274.2829.5
OP4 Spirochaetes (Brevinema)Unclassified BacteriaUnclassified Bacteria160.0920.8
OP5 Caldiserica Caldiserica Caldiserica 981.8527.4
OP6cNoneUnclassified BacteriaNone000
OP7 Elusimicrobia (12–31)dUnclassified BacteriaUnclassified Bacteria110.2125.3
OP8 Nitrospirae (OPB95)Unclassified BacteriaCandidate division OP8 or unclassified Bacteria3145.9224.3
OP9Candidate division OP9Unclassified BacteriaCandidate division OP92735.1522.8
OP10 Armatimonadetes Armatimonadetes Armatimonadetes 56810.7128.8
OP11Candidate division OP11Candidate division OP11Candidate division OP114939.3033.4
OP12 Firmicutes, Clostridia (OPB54) Firmicutes, Clostridia Unclassified Bacteria2915.4924.2

However, a few OP divisions, particularly OP1, 3, 9, 10 and 11, remained distinct phylum-level clusters and expanded dramatically as more sequences from uncultured bacteria were found. The latest version of the ARB-Silva reference database (SSUREf_108_SILVA_09_09_11) contains 530 197 nearly complete (> 1200 bp) 16S rRNA gene sequences from the Domain Bacteria (Pruesse et al., 2007). Although this is an incomplete survey of bacteria detected in nature, it gives a reasonable estimate of the relative detection frequency of different phyla. In the database, 568 sequences are classified as OP10, or about one OP10 sequence per 1000 entries in the Domain Bacteria, making it the most frequently detected of the OP divisions (Table 1).

Environmental distribution

Hugenholtz and colleagues (1998) reported the temperature of Obsidian Pool as 75–95°C, but did not specify the temperature of the individual samples taken. Shock and colleagues (2005) noted that the source was usually 79°C, so likely the lower limit reported by Hugenholtz et al. is realistic for the detection of OP10. Along a thermal gradient (48–76°C) in a Japanese hot spring, OP10 occurred only below 60–66°C (Nakagawa and Fukui, 2002). A similar trend was observed in a hot spring system of the Alvord Desert, Oregon where OP10 sequences (comprising 5% of a 16S rRNA gene clone library) were found in a thick downstream biofilm at 72°C but not at an upspring biofilm at 78°C (Connon et al., 2008). The distribution of OP10 species in thermal gradients may be an autecological effect or a consequence of their lifestyle. They may grow heterotrophically on exudates of other bacteria and therefore prefer thick mat communities (see below). The hottest sites where OP10 sequences have been detected are between 75°C and 80°C, including Obsidian Pool (c. 79°C), Great Boiling Hot Spring 04b in the Great Basin of the western USA (76.8°C; Costa et al., 2009), and a hot spring in the Rupi Basin of Bulgaria (79°C; Tomova et al., 2010). The maximum growth temperature for OP10 species appears to fall just short of the 80°C borderline used to define hyperthermophiles.

Although OP10 are commonly detected in thermal springs, only about 6% of the 16S rRNA gene sequences classified as OP10 in the ARB-Silva Reference database originate from geothermal habitats. Most are from temperate soils (37%), skin swabs (17%), aerobic and anaerobic bioreactors and waste treatment facilities (12%), or freshwaters and freshwater sediments (7%). The remainder include sequences associated with bacterial mats or biofilms (6%), plants or animals (6%), marine sediments (3%), ice and snow (2%), mine tailings (1%), fossil fuel deposits (1%) and dust/atmospheric samples (1%). The group as a whole appears to be primarily associated with temperate habitats rather than geothermal sites, although to some extent these numbers reflect a bias of scientific effort. The large number of sequences associated with skin swabs is probably misleading, as these are from massive (> 300 000 sequences) skin microbiome sequencing projects based on cloning and Sanger sequencing, in which OP10 makes up < 0.05% of all 16S rRNA sequences obtained (Grice et al., 2009). Although intriguing, this study may simply reflect the presence of OP10 as part of the rare biosphere in soils and dust attached to skin.

Individual studies point to an ecologically broad and adaptive group. Although all three isolated strains are aerobes (see below), it is likely that many members of the phylum are anaerobes or facultative anaerobes, based on their presence in highly anoxic environments like methanogenic bioreactors (Wang et al., 2010) and methanogenic aquifers (Dojka et al., 1998). Besides hot springs, sequences have been found in other extreme habitats, including hypersaline waters (Burns et al., 2004; Ley et al., 2006), Antarctic cryptoendolithic communities (de la Torre et al., 2003), and acidic soils at pH 4.5 (Faoro et al., 2010). OP10 species seem unlikely to form strong parasitic or symbiotic relationships with plants or animals, although sequences have been detected in animal intestinal contents and faeces (Ley et al., 2008; Wu et al., 2010), and human skin swabs (Grice et al., 2009), as well as in the plant phylosphere (Delmotte et al., 2009) and rhizoplane (Chelius and Triplett, 2001). One oddity is the detection of OP10 in the synovial fluid of some arthritis patients, where it is likely to be an opportunistic colonizer (Siala et al., 2009).

Preferred habitats

Only rarely have OP10 bacteria been detected in 16S rRNA-based bacterial community surveys. When they are detected, they are usually a minor community component, comprising a relative detection abundance below 1%. Although the rate of detection in different environments will undoubtedly increase with the growing popularity of next-generation sequencing approaches, OP10 should remain members of the rare biosphere in most environments. The most intriguing community surveys involving OP10 are therefore those few where it comprises a larger proportion of the community. Such studies may provide clues to preferred habitats of members of this phylum, although it should be stressed that the OP10 group is diverse environmentally and phylogenetically, and probably includes species with a wide variety of metabolic potentials.

The first reported community with a predominance of OP10 was in Horestooth Reservoir, Colorado, a metal-rich lake that becomes fully stratified in late summer (Stein et al., 2002). Metal oxides from the underlying sandstone are reduced by anaerobic respiration in the anoxic hypolimnion, and reprecipitate as metal oxides when they diffuse into oxic surface waters. OP10 comprised 50% of the bacteria detected in a 16S rRNA gene clone library constructed from these metal precipitates. This sheer predominance makes certain speculation reasonable: perhaps these bacteria have a role in redox metal cycling, either as lithotrophs, anaerobic metal respirers, or anoxygenic phototrophs. OP10 strains have been detected in other metal-rich sites as well, particularly mine tailings (Rastogi et al., 2009; Navarro-Noya et al., 2010).

OP10 sequences also comprised over 50% of the 16S rRNA genes detected in a green bacterial mat in a 50–57°C hot spring in Thailand (Portillo et al., 2009). The proportion decreased in layers of the mat with fewer Cyanobacteria. This suggests that some OP10 strains may themselves be phototrophic, or more likely that they grow chemotrophically on exudates or decay products of photosynthetic bacteria. In fact, OP10 are commonly detected, often at several percent of total 16S rRNA genes, in areas predominated by photosynthetic bacteria or eukarya, including freshwater cyanobacterial blooms (Pope and Patel, 2008); stromatolites (Burns et al., 2004); hypersaline photosynthetic mats dominated by Chloroflexi (Ley et al., 2006); photosynthetic hypolithic communities (Wong et al., 2010); and macrophyte-dominated eutrophic sediment (Shao et al., 2010). Some role in scavenging products of photosynthetic organisms therefore seems likely. Their prevalence in plant-fed anaerobic bioreactors further indicates a role in degradation of plant material (Wang et al., 2010).

OP10 are also typical inhabitants of carotenoid-pigmented biomats commonly found in geothermal springs. They comprised 5–9% of the 16S rRNA genes detected in orange thermal biomats in the USA, New Zealand and Thailand (Connon et al., 2008; Stott et al., 2008; Portillo et al., 2009). Many thermophilic bacteria produce carotenoids as UV protectants, and the few known OP10 isolates are themselves pigmented pink-to-orange (see below).

Possible detection biases

The low representation of OP10 in most 16S rRNA-based diversity surveys indicates that they are usually members of the rare biosphere. However, low detection rates may also be a function of methodological biases. Liles and colleagues (2003) assayed the bacterial diversity of soil via a standard method in which 16S rRNA genes are PCR-amplified from soil DNA extracts using the widely employed ‘universal’ PCR primers 27f and 1492r. They then compared this with a metagenomic approach in which soil DNA was first ligated into BAC vectors and cloned into Escherichia coli. This BAC library was then used as a template for PCR-amplification of 16S rRNA genes. Intriguingly, no OP10 clones were detected in the direct PCR approach, but they comprised 5% of the 16S rRNA clones obtained from the BAC library approach. Only 28 individual BAC clones were found to contain 16S rRNA genes, so the results should probably be considered suggestive rather than statistically conclusive. However, the authors ascribed the differences to different lysis protocols: a gentle protocol was necessary to extract high-quality DNA for BAC library construction, while a bead-beating protocol was used in the direct PCR approach. Spore-forming Bacilli, for example, were under-represented when using gentler extraction. Presumably then, OP10 are easily lysed and not prevalent as recalcitrant resting forms. In agreement with this, Nogales and colleagues (2001) detected OP10 at a higher frequency in a clone library constructed from soil-extracted RNA (16S rRNA) than in one constructed from DNA (16S rRNA genes). This suggests that OP10 bacteria make up a larger portion of total soil rRNA than soil DNA, and therefore that they exist primarily in active rather than dormant states. This would be consistent with a role as K-selected oligotrophs rather than feast-or-famine copiotrophs (Janssen, 2009). The importance of OP10 within the metabolically active community may therefore be higher than indicated by 16S rRNA gene surveys.

It is also possible that OP10 species are under-represented in community surveys due to template mismatches with commonly used ‘universal’ 16S rRNA gene primer sets. We therefore examined the complete 16S rRNA gene of Chthonomonas caldirosea, assembled from a draft genome sequence, for complements to universal primer sequences. However, there were no mismatches to the universal primer regions 27f (AGAGTTTGATCMTGGCTCAG) or 1492r (ACGGYTACCTTGTTACGACTT). At least for this member of the OP10 group, there should not be poor amplification due to primer biases.

Isolation of OP10 strains

In the decade immediately following the 1998 Hugenholtz et al. study, none of the original OP groups were obtained into pure culture. However, in 2008 two publications reported the initial cultivation of members of OP10 (Stott et al., 2008) and OP5 (Mori et al., 2008). As expected, both isolates were obtained from geothermal springs and both were thermophilic. Perhaps unexpectedly, both were also organoheterotrophs. The OP5 division was the first to have its candidate status upgraded, with the validation of Caldisericum exile as the first member of the phylum Caldiserica (Mori et al., 2009). The OP10 phylum was renamed the Armatimonadetes in 2011 (Lee et al., 2011; Tamaki et al., 2011).

Stott and colleagues (2008) first reported cultivation of an OP10 bacterium, isolate P488, from the Te Manaroa Spring in Waikite, New Zealand. This boiling (98°C) spring is situated in a depression several metres deep in a densely vegetated area. Plant material growing along the inside rim of the depression contributes to a dense mat of decaying organic material on the edges of the spring. Based on a 16S rRNA gene clone library, the microbial community in this decaying organic material was abundant in OP10 (comprising 9% of the clones). Strain P488 was obtained in pure culture using media that carefully reproduced in situ conditions of pH and temperature in this mat, and incorporated modifications of standard enrichment procedures suggested particularly by the group of Peter Janssen (Janssen et al., 2002; Joseph et al., 2003; Janssen, 2008). After optimizing culture conditions, phylogenetically identical bacteria were obtained from other locations as well. One of these, strain T49 from Hell's Gate, NZ, was later described as Chthonomonas calidirosea (Lee et al., 2011).

Keys to the cultivation of P488 and T49 were (Stott et al., 2008; Lee et al., 2011):

  • (i) High abundance in the original site (9% of the clone library). Given that OP10 is usually only a minor community component, the Waikite community was a natural enrichment, and therefore an ideal inoculum for isolation.
  • (ii) Use of dilute, low-nutrient media. Bacteria in nature can be conceptually divided into K-selected oligotrophs and r-selected copiotrophs (Langer et al., 2004; Fierer et al., 2007; Janssen, 2008; 2009; Lauro et al., 2009). The latter group probably includes most cultured species, which thrive on the nutrient-rich conditions presented by standard microbiological media. However, most bacteria in nature are expected to be oligotrophs adapted to low substrate concentrations. These are either inhibited by nutrient-rich media, or their slow growth is obscured by the rapid proliferation of copiotrophs. Strains T49 and P488 cannot grow on many standard media like nutrient broth, tryptic soy broth, or Luria–Bertani broth, even when these are adjusted to environmental pH and temperature. Even R2A, which was developed as a low-nutrient medium for oligotrophs (Reasoner and Geldreich, 1985), only supports growth of these bacteria when diluted 10 ×.
  • (iii) Extended incubation times. Many oligotrophic bacteria have slow growth rates and require extended incubations before colonies can be seen on plates (Davis et al., 2005). The OP10 isolates grow slowly (k = 0.012 h−1, generation time of 58 h). Three to seven days of incubation were reported to be necessary to visualize colonies on plates. This seems much faster than suggested by the 58 h generation time, because the original colony-forming units were already large cell aggregates. Colonies simply did not grow from cfus of well-distributed cells (K.Y.C. Lee and M.B. Stott, unpubl. data). This observation suggests that growth was not only slow, but probably also required quorum signalling or the protection of a biofilm growth state.
  • (iv) Gellan as a medium solidification agent. Using gellan instead of agar as a solidifying agent increases the likelihood of cultivating some bacteria (Janssen et al., 2002; Janssen, 2008). Tamaki and colleagues (2009) noted that 20% of bacterial isolates obtained from a sediment sample grew on gellan-solidified medium but not on agar-solidified medium. Agar can inhibit growth of some bacteria (Johnson, 1995), and conversely gellan can stimulate growth (Janssen et al., 2002). In the case of P488 and T49 both effects probably played a part. On the one hand, the complex polysaccharide gellan actually served as an energy source for these bacteria. Colonies could be seen to bore holes into the gellan substrate as they consumed it. Polymeric growth substrates like gellan can favour oligotrophic bacteria because slow hydrolysis into simple sugars may prevent substrate-induced cell death and overgrowth by copiotrophs (Janssen, 2008). On the other hand, strain T49 could no longer grow when the gellan in a suitable mineral salts medium was replaced by agar plus glucose (K.Y.C. Lee and M.B. Stott, unpubl. data), indicating that agar inhibition is also a factor.
  • (v) Low-pH medium replicating in situ conditions. Analysis of isolate T49 revealed that it had an extremely narrow pH range for growth: 4.7–5.8. The organism probably has little genetic capacity for pH homeostasis and deviation of the medium away from this 1-unit range would have resulted in no growth.

A second member of the OP10 phylum was isolated from the rhizoplane of Phragmites australis, an aquatic plant from a mesophilic freshwater lake in Japan (Tamaki et al., 2011). Strain YO-36 was successfully cultivated on a low nutrient medium and characterized as Armatimonas rosea. This report appeared a few months before the report on Chthonomonas, so the name Armatimonadetes has been given to the OP10 phylum. The authors do not describe the isolation procedure in detail, nor the growth rate, so it is difficult to determine the key factors that allowed this bacterium to be cultivated. Interestingly, Armatimonas did grow on agar-solidified plates although like Chthonomonas it is capable of metabolizing gellan and may therefore have preferred gellan plates. Like Chthonomonas it was isolated on a low-nutrient medium, and could not grow on many standard complex media unless they were diluted. The relatively low nutrient medium R2A did allow growth of Armatimonas, although it was too nutrient-rich for Chthonomonas.

Comparison of OP10 isolates

Chthonomonas and Armatimonas share a few morphological and physiological parameters: both are aerobic Gram-negative cells, produce pink pigment, grow on oligotrophic media and degrade polysaccharides. However, Armatimonas is a mesophile (optimum 30–35°C) and a neutrophile (optimum pH 6.5) while Chthonomonas is a thermophile (optimum 68°C) and acidophile (optimum pH 5.3). Their 16S rRNA gene sequence identity is less than 80%, and not surprisingly they have dramatically different phenotypes (Table 2). An interesting feature of Chthonomonas is its unique lipid composition, which includes several rare Δ5 cis-monounsaturated fatty acids, and a novel cyclopropanoic acid never before detected in bacteria: 5,6-methylene hexadecanoic acid (Vyssotski et al., 2011). This new fatty acid comprised 5% of the total fatty acids by weight. These compounds were not reported in Armatimonas, although the methods used were not suitable for their identification.

Table 2. Phenotypic distinctions between the three known isolates of candidate division OP10.
  Armatimonas rosea (YO-36)a Chthonomonas calidirosea (T49)b‘Fimbriimonas ginsengisoli’ (Gsoil348)c
MorphologyRod to ovoid (1.4–1.8 × 2.4–3.2 µm)Rod [0.5–0.7 × 2.5–3.0 µm (l)]Rod (0.5–0.7 × 2.5–5.0 µm)
MotilityNon-motileMotiledNon-motile
Temperature range (optima)20–40°C (30–35°C)50–73°C (68 6°C)18–30°C (30°C)
pH range (optima)pH 5.5–8.5 (6.5)pH 4.7–5.8 (5.3)6.0–8.5 (7.0)
Maximum salt tolerance0.5% (w/v)2.0% (w/v)n.r.
Pairwise 16S rRNA gene sequence similarity C. calidirosea 79.4%‘F. ginsengisoli’ 77.1% A. rosea 80.2%
‘F. ginsengisoli’ 80.2% A. rosea 79.4% C. calidirosea 77.1%
G + C content62.4 mol%54.6 mol%66.7 mol%
Cell envelopeGram-, smoothGram-, irregular and invaginatedGram-
Major fatty acids16:0 (39.2%), 16:1 (28.0%), 14:0 (24.5%), 15:0 (8.3%)16:0 (25.8 %), i17:0 (19.3 %), ai17:0 (13.5 %), cis-16:1Δ5 (8.8 %), cis-i17:1Δ5 (6.8 %), 5,6-methylene 16:0 (5.2 %)i15:0 (30.9%), i17:0 (19.5%), 16:0 (17.1%), cis-16:1Δ5 (11.3%), i13:0 3-OH (5.8%)
Primary quinonesMK-12 (%?)MK-8 (70 %), MK-9 (17 %), MK-7 (10 %) and MK-6 (3 %)MK-11 (65%) and MK-10 (32%)
CarbohydratesLimited (arabinose, raffinose, maltose, sucrose and gentiobiose)Most mono-, di- and tri-saccharides testedn.r.
PolysaccharidesLimited (gellan, xanthan and pectin)Most amorphous polysaccharidesn.r.
Yeast extractYesNon.r.
R2AYesNoYes

In addition to these peer-reviewed publications, another OP10 isolate dubbed ‘Fimbriimonas ginsengisoli’ strain Gsoil348 was reported in a meeting abstract in 2006 (Im et al., 2006). It was described as a strictly aerobic, non-motile rod (0.5–0.7 mm × 2.5–5.0 mm) growing organotrophically on R2A medium. The isolate was neutrophilic (pH 7 optimum) and mesophilic (30°C optimum), and therefore, at least superficially, seems similar to A. rosea. A more detailed description of this bacterium is hopefully forthcoming.

The genome analyses of these organisms, which are presently underway, should further elucidate characteristics that distinguish them from each other and from other bacteria. Although they are the only available representatives of a new phylum, one should probably not expect too much similarity given the deep evolutionary divergence among them. One might just as well compare Rhizobium with Escherichia.

Phylogenetic structure of OP10

We constructed a rigorous phylogenetic tree using 490 high-quality and non-redundant OP10 sequences, an abridged version of which is presented in Fig. 1 (see Supporting information Fig. S1 for the complete maximum likelihood phylogenetic tree). Phylogenetic reconstructions were performed using ARB (Ludwig et al., 2004), and Tree-Puzzle (Schmidt et al., 2002). Twelve distinct groups, including candidate phylum WS1 (Dojka et al., 1998) were delineated, of which six (Groups 1, 2, 3, 5, 9 and 11) showed > 90% support values using four distinct phylogenetic construction methods. Groups 6, 7 and 8 also showed strong support (> 90%) using three of the four methodologies. Only Groups 4 and 12 (WS1) failed to demonstrate good support.

Figure 1.

Maximum-likelihood tree based on 16S rRNA gene sequences (> 1200 bp) of isolates and environmental clones belonging to the phylum Armatimonadetes (n = 106). Known isolates as well as the original Obsidian Pool clones are highlighted. Bifurcations supported at bootstrap levels of > 90% are indicated by circles, and values > 80% are indicated by triangles (based on 1000 data resamplings). Phylogenetic trees were also generated using Neighbour-Joining (NJ), and Maximum Parsimony (MP) and Tree-Puzzle (TP) methodologies (trees not presented). Where bifurcations in the ML tree are supported by high bootstrap values in each of the ML, NJ and MP methods, and well as by equivalent support values in a TP construction (TP, based on 10 000 puzzling steps), the node symbols are filled. Open symbols represent node confidence values supported by ML, MP and NJ methodologies only. The dendrogram was rooted to an out-group containing multiple representatives from multiple bacterial phyla (not shown). The scale bar represents 0.10 substitutions per nucleotide site.

The three known OP10 isolates are only distantly related to each other, and begin to represent the wide phylogenetic breadth of the OP10 phylum. Armatimonas clusters in Group 1, Chthonomonas in Group 3, and ‘Fibriimonas’ in Group 9. They share < 80% 16S rRNA gene sequence identity with each other. According to Konstantinidis and Tiedje (2005), bacteria that share only the class as their lowest common taxonomic level typically show 80–90% identity in their 16S rRNA sequences. The groups delineated in Fig. 1 should therefore be considered to represent separate taxonomic classes at the very least, and given that the phylum boundary is 78.4–84.7% identity (Yarza et al., 2010) an argument could be made that some of these groups would actually be better described as separate phyla.

Some phylogenetic groupings also loosely correspond to ecological niches (see Supporting information Fig. S1 for colour-coded niche characterization). For example, Group 1 (class Armatimonadia) contains primarily phylotypes detected in soil and water, Group 10 contains geothermal phylotypes, Groups 11 and 12 phylotypes detected in activated sludge and anaerobic sediments, and Group 3 (class Chthonomonadetes) primarily phylotypes detected in soil and biofilms.

Highly resolved phylogenetic placement of OP10

Phylogenies based on 16S rRNA genes, and for that matter any single-gene phylogenies, typically do not resolve deeply branching phyla due to insufficient information. To create a highly resolved phylogenetic placement of OP10 within the Domain Bacteria, we therefore constructed a database containing 29 universally present, vertically inherited proteins (Ciccarelli et al., 2006). The genes for C. caldirosea T49 were obtained from its draft genome sequence.

An initial database of 23 ribosomal proteins (L1, L3, L5, L6, L11, L13, L14, L15, L16, L18, L22, S2, S3, S4, S5, S7, S8, S9, S11, S12, S13, S15 and S17) from 136 bacterial species was assembled using the Ribalign program and database (Teeling and Gloeckner, 2006). Phylogenetic analysis (results not shown) indicated that C. calidirosea appeared to be a deep-branching relative of both Cyanobacteria and Chloroflexi. Therefore, the Ribalign-derived database was manually curated and six non-ribosomal universal proteins (GTP-binding protein YchF, phenylalanyl-tRNA synthetase alpha subunit, seryl-tRNA synthetase, preprotein translocase SecY subunit, DNA-directed RNA polymerase alpha subunit, and leucyl-tRNA synthetase) collected from the public-domain databases were added to selected species. A total of 29 concatenated proteins from 60 bacterial species across 15 phyla were aligned using clustalW (Thompson et al., 1994). The 60 bacterial species included at least five members each of the Clostridia, Bacilli, Chloroflexi and Cyanobacteria. Trees were calculated using NJ (with a Kimura distance correction, 1000 bootstraps), MP (using the PHYLIP algorithm in ARB, 1000 bootstraps), and TP quartet puzzling (10000 data resamplings) methods while excluding gaps in the alignment. Phylogenetic reconstructions where performed using ARB (Ludwig et al., 2004), SeaView (Gouy et al., 2010) and Tree-Puzzle (Schmidt et al., 2002).

This phylogenetic reconstruction, based on 29 concatenated proteins with an average length of 6399 amino acids, firmly verified that C. calidirosea represents a deeply branching lineage of sufficient divergence to be considered a novel phylum, but also strongly suggests that this is a sister phylum to the Chloroflexi (Fig. 2). The support for a common ancestor of Armatimonadetes and Chloroflexi was 99% for NJ, 82% for MP and 76% for TP constructions.

Figure 2.

Phylogenetic tree constructed using Neighbour-Joining and a Kimura distance correction, based on a concatenation of 29 universal core proteins averaging 6399 amino acids in length. Sequences from 60 species across 15 bacterial phyla were included. A consensus tree based on 1000 bootstrapped reconstructions is shown. The support values for each phylum were high (> 80%), but the resolution of the relationships among phyla were generally very low (< 50%). An exception was the association (99% support) of the two phyla Chloroflexi and Armatimonadetes (Chthonomonas calidirosea). The scale bar represents 0.1 change per amino acid position.

Conclusion: bringing culture to the uncultured masses

It has become axiomatic to claim that 99% of all bacterial species are still uncultured. The origin of this figure is obscure. However, given that there are less than 10 000 described bacterial species (Euzéby, 1997; accessed Dec., 2011) and that reasonable estimates of bacterial pan-diversity are in the millions or tens of millions (Curtis et al., 2002; Gans et al., 2005), then even 99% seems a conservative estimate. Remarkably, our knowledge gap is not limited to the taxonomic level of species. We apparently have not cultivated a single member of most higher-level bacterial taxa either. There are now, with the addition of Armatimonadetes, about 30 phyla of bacteria containing cultured members (an exact number is difficult to pinpoint because of the unofficial status of phyla). Many candidate divisions remain, including some first detected more than a decade ago, like OP1 and OP11. There may be 70 or more candidate divisions still (Achtman and Wagner, 2008). There is great excitement over the use of single cell techniques to obtain genomes from uncultured groups. Members of the candidate divisions TM7 and OP11 have been partially characterized in this way (Marcy et al., 2007; Youssef et al., 2011). However, while genomic data are suggestive of functional and evolutionary relationships, there is still no substitute for a growing culture.

The failure to cultivate most bacteria is due to a variety of factors. Available cultivation techniques, for whatever reasons, have not facilitated their growth. However, improved cultivation protocols are starting to open up the uncultured biosphere. Many of these improvements have been covered in recent reviews (Giovannoni et al., 2007; Cardenas and Tiedje, 2008; Janssen, 2008; Overmann, 2010; Dedysh, 2011). New isolation methods include microfluidic cell sorting to start with single cell inocula, in situ incubation systems, and simple modifications of traditional batch- and plate-cultivation methods. The latter usually involve dilute nutrient-poor media, long incubations, replication of in situ conditions, addition of signalling molecules to stimulate growth, complex polymeric substrates, and gellan as a gelling agent. These have been successfully applied to cultivate groups that were traditionally considered to be recalcitrant to growth in the laboratory, like Acidobacteria (Janssen, 2008). Another major improvement in cultivation efforts has been the ability to rapidly screen many colonies via 16S rRNA gene sequencing for only those few that are taxonomically the most interesting and deserve further study. This saves large amounts of effort in isolation and characterization.

The recent successful cultivation of several members of OP10 represents a victory for improvements in bacterial cultivation, particularly for those improvements based on simple alterations in traditional techniques. High-throughput, high-tech methods were not necessary to describe this new phylum, only some common sense and some elbow grease. The vast majority of this ecologically diverse phylum still awaits successful cultivation, but at last inroads are being made. We can now look forward to the genomic and physiological characterization of these isolates.

Acknowledgements

This work was supported in part by the Natural Sciences and Engineering Research Council of Canada Discovery Grant Program, Alberta Innovates-Technology Futures, and the New Zealand Ministry for Science and Innovation Geothermal Research Core funding. K.L. is supported by the Sarah Beanland Scholarship. The authors wish to thank the ongoing support of the Tikitere Trust, Guardians of Hell's Gate.

Ancillary