by Christoph Weigel
Another journey to Lilliput
Back in 2008, Merry and Elio introduced a Lilliputian to this blog, Nanoarchaeum equitans, whose minute cells ‒ diameter ~400 nm, cell volume ~0.025 µm3, genome size 0.49 Mb (Mega basepairs) ‒ decorate the normal‑sized cells of its host Ignicoccus hospitalis. Both belong to the Archaea domain of the prokaryotes, . and while I. hospitalis can be cultivated alone, N. equitans depends on its host as its genome lacks nearly all genes for lipid, cofactor, amino acid, and nucleotide biosynthesis. Whether N. equitans is best described as an obligate ectosymbiont or as an ectoparasite of I. hospitalis is still debated. But their intimate relationship ‒ N. equitans only accepts I. hospitalis as host, no other Ignicoccus species ‒ led one researcher, J. Godde, to speculate that a ménage à trois of their ancestors with an Alphaproteobacterium might have started the eukaryote adventure.
But what about small bacteria, smaller than Prochlorococcus marinus (cell volume ~0.1 µm3) or Pelagibacter ubique (cell volume ~0.1 µm3) ? I mean really small ones, small enough for a whole bunch of them to fit into, say, a cell of E. coli with its volume of 0.6 ‒ 0.7 µm3 ? No, these aren't the much hyped nanobacteria again, revealed as artifacts not too long after their discovery. Nor are they the tiny bacteria found in suspended animation in 120,000 year-old ice cores from Greenland (featured here in this blog). What about true bacterial Lilliputians that are happily thriving all around us? Turns out that diverse habitats abound with them but they had simply escaped detection until recently (though tempting, no pun here, sorry). Well, not entirely 'escaped': there have been, over the last few years, sporadic reports of metagenomic analyses pointing to very small genomes of apparently free-living bacteria contained in uncultured environmental samples. And such small genomes should be harbored by... small cells, or at least this was the assumption of Jill Banfield and her co-workers, presented first by Elio in this blog last year. More on that in a moment.
A brief detour to the Metagenomics continent
Metagenomics is based on the ingenious idea of Norm Pace 30 years ago that sequencing nucleic acids directly from environmental samples would allow microbiologists to assess microbial diversity without getting stuck in the ~1% recovery-rate bottleneck for traditional cultivation-dependent methods. Thus, using collections of the universally conserved ribosomal 16S rRNA sequences and applying a 75%-identity cut-off, Yarza et al. recently estimated that approximately 1,500 bacterial phyla exist. A staggering number that, pointing to a wild underestimate of bacterial diversity even after a decade of voracious sequencing. Today, the non-redundant set of the SILVA database contains 152,308 curated bacterial 16S rRNA sequences, i.e., "species signatures". In this wider perspective, the presently accepted 29 phyla and ~60 'candidate phyla' ‒ phyla for which there is presently no cultivated species ‒ will hardly continue to serve as landmarks for navigating the bacterial domain.
Metagenomics took a big leap forward when, more recently, largely un-biased sequencing became standard ‒ achieved by abandoning PCR ‒ and now allows the in silico reconstruction of metabolic properties of entire microbial populations. Metagenomics also allows for the reconstruction of complete bacterial genomes from samples collected on filters, mostly filters with a 0.2 µm cut-off (see here for an example). It is only fair to mention, however, that metagenomics would not have become such a powerful tool without the concomitant development of sophisticated bioinformatic tools, e.g., metagenomic binning.
The Lilliput archipelago
Banfield and her co-workers assumed that bacteria harboring small genomes would be small themselves and thus might be enriched by filtration. They sampled microbial communities from an aquifer adjacent to the Colorado River near Rifle, Colorado, at a depth of ~7 m. They filtered ground water through a 1.2 µm pre-filter and collected cells on serial 0.2 and 0.1 µm filters. The filters were subsequently processed for sequencing.
Some numbers to let you appreciate the dimensions of 'archipelago Lilliput': From 12 samples (6 time points, 0.2 and 0.1 µm filters) the authors obtained 224 Gb (Giga basepairs) of DNA sequence from individual 150 bp reads, which they assembled to 3.9 Gb of contigs, i.e., approx. 780,000 individual ~5 kb pieces of contiguous sequence with 10-fold coverage. These contigs were 'binned' into >1750 genome bins, and >60% of these bins represented genomes from the Lilliputians (i.e., the phyla TM6, Parcubacteria (OD1), Microgenomates (OP11), WWE3, Berkelbacteria (ACD58), Saccharibacteria (TM7), WS6, Peregrinibacteria (PER), Kazan, and the previously unrecognized phyla CPR1, CPR2, and CPR3). From the 'Lilliput bins', Banfield and co-workers obtained 789 'draft quality' genomes (>50% contiguous), and 8 complete genomes by manual curation. All these genomes are small, mostly <1 Mb in length. Their 16S rRNA phylogeny suggests that all the Lilliput phyla except TM6 represent a monophyletic branch in the bacterial tree, a branch that may contain considerably more then 35 phyla and represent ~15% of the bacterial domain.
To their surprise, the authors found insertions in the 16S rRNA genes in ~1/3 of the Lilliputian genomes, and in a lesser percentage of the 23S rRNA genes. Such insertions are rather exceptional and had previously been found only in four members of the Thiotrichaceae sub-branch of the Gammaproteobacteria. These insertions vary in size (10 bp ‒ 2 kb, mean 519 bp), and occur at several distinct sites, both in conserved and variable regions of the rRNA genes. The larger insertions (>500 bp) are self-splicing introns or encode homing endonucleases or small proteins of unknown function. The authors argue that these insertions probably prevented detection of the Lilliputians in earlier metagenomic surveys of environmental samples because the set of PCR primers commonly applied to amplify 16S rRNA genes would miss intron-containing genes (or would result in amplicons with other than the expected lengths).
In addition to the peculiarities of their ribosomal RNA genes, the protein composition of the ribosomes of many members of the Lilliput phyla deviate from conserved norms. All lack large-subunit protein L30, while members of four phyla lack L9 and some individual (sub)phyla lack L25 or L1. Ribosome assembly might follow different routes too, as the gene for a specific ribosomal protein biogenesis factor, GTPase Der, is missing from most of the genomes that lack either L9 or L1.
Having thoroughly explored the dimensions of the 'Lilliput archipelago' and confirmed the small genome sizes of its inhabitants, Banfield and co-workers had not conclusively confirmed that these genomes were actually harbored by tiny cells. They might instead reside in long but slim spirochaete-like bacteria that would also have been collected on their filters (which was in fact the case, as seen in Fig. 2.2).
A close-up of the Lilliputians
From the same ground water aquifer as above, the Banfield team collected samples of filtrate run through 0.2 µm filters and split them for cryo-electronmicroscopic (cryo-TEM) studies of cell morphology and for metagenomic analysis. The majority of cells seen in pictures were small, with a spherical diameter of 253±25 nm (median) and calculated cell volume of 0.009±0.002 µm3 (median). Indeed, since roughly 50 of them would theoretically fit inside a single well-fed E. coli cell, they deserve to be called 'ultra-small'. Most of them have a strong, well-structured cell wall with an associated surface layer (S-layer). Their centrally located nucleoids occupy ~20% of the cytoplasmic volume and comprise large, intertwined spiral structures with a periodicity of ~5.6 nm, indicating tightly packed DNA (Fig. 1). On average, cells contain 42±9.5 putative ribosomes that mostly flock to the cell ends (Fig. 1).
A prominent morphological feature of many of these ultra-small bacteria is their 'hairy' phenotype, their covering by pili (Fig. 2.1 a). Supporting these observations, the authors found a full complement of genes encoding pili components ‒ type-IV pili, including pilT for twitching motility, and several pilins ‒ in genomes from two phyla. They assume that these pili serve in cell-to-cell contacts and in interacting with the environment. Interestingly, a number of pictures revealed ultra-small bacteria in pili‑mediated contact with larger bacterial cells (Fig. 2.1 b).
This reminded me of Gulliver who, having been washed ashore on Lilliput unconscious after a shipwreck, complained when waking up: "I found my arms and legs were strongly fastened on each side to the ground". And indeed, the authors were lucky to actually find a complete 'Gulliver' among their 60+ pictures: a spirochaete strongly fastened to one of the Lilliputians by pili (Fig. 2.2). The dumbbell-shaped Lilliputian portrayed in Figure 2.2 indicates, in addition, active metabolism of the sampled ultra-small bacteria as this one seems to have been caught in the process of dividing/budding.
The authors were dealing with a heterogeneous cell population, so it was necessary to assess ‒ at least to approach ‒ the identity of the bacteria by statistical means. The metagenomic analysis revealed a dominance of sequences from species of three phyla that have genomes of ~1 Mb. The statistical analysis points to an excellent correlation between both data sets: yes, the small genomes are harbored by small cells (Fig. 3).
Just two more questions...
If the ribosomes of the ultra-small bacteria have some unusual features, then one might imagine that their DNA replication is likewise atypical. Jill Banfield and her co-workers did not look at this (at least they don't mention it), but I could not, well, resist peering into seven of the completely assembled genomes from different phyla. All have dnaA genes encoding the initiator protein for chromosome replication, and these DnaA proteins share ~40% homology among each other and also with B. subtilis DnaA. This is a typical result for DnaA proteins of bacteria from different phyla (e.g., 43% for E. coli K-12 (Gammaproteobacteria) vs. B. subtlis 168 (Firmicutes)). Also, they have dnaG genes (primase), dnaB genes (replicative helicase), and dnaN genes encoding the beta-clamp subunit of the replicative DNA polymerase. Lastly, in all but one case their replication origin, oriC, is located between the dnaA and dnaN genes, the 'signature' location of oriC in most Firmicutes, Actinobacteria, and Delta- and Epsilonproteobacteria (but notably not in E. coli). The seven oriC structures seem related but a more thorough analysis ‒ and experiments ‒ would be necessary to confirm this (Fig. 4). What appears somewhat unusual, when compared to other known oriC s, is the strong preference for one orientation of the DnaA-binding sites (consensus 5'-TTWTNCACA) in all of them. But taken together, the Lilliputians are pretty much your garden-variety bacteria, at least with respect to their chromosome replication.
If the ultra-small bacteria are true bacteria they would come with their own set of bacteriophages, right? There are first hints that this is in fact the case. Banfield and co-workers found cells with phages attached among their cryo-TEM pictures (Fig. 5). The phages appear to be of enormous size but this is an illusion: look at the size bar in Figure 5 (left) and compare to the sizes of a 'normal' E.coli cell with P1 phages (right). As is common for bacterial genomes, those of the ultra-small bacteria contain a plethora of prophage remnants ‒ genes encoding integrases, tape-measure proteins and phage-related DNA polymerases ‒ yet complete genomes for phages of bacteria from these phyla are not yet known. Attempts to cultivate consortia of Lilliputians in the presence of their phages will be sort of fun for microbiologists endowed with a special sense of humor: unlike normal-sized bacteria, environmental samples of Lilliputians cannot be made essentially "phage-free" by the customary filtration through 0.2 µm filters. Spoiler: quite a few phages will be lytic ones.
A shift in perspective, for bacteriology
Ever since its first formulation during the mid-19th century, cell theory has been one of the beacons of biology: cells are the fundamental units of life. Period. It was therefore not that difficult for bacteriologists ‒ likewise for archaeologists and protist aficionados in their respective realms ‒ to make a clear distinction between cellular bacteria and their bacteriophages (viruses) as "some form of life" that depends on cells for metabolic activity and reproduction. This "some form of life" blurb accounts for the fact that both, bacteria and their phages, have genomes and participate in the big game of evolution, "descent with modification" (Darwin) and selection. For roughly a century, it was OK to stick to this distinction because bacteria are rather big as compared to phages (Fig. 5 right), and they tend to have large genomes, phages much smaller ones (E. coli 4.6 Mb, bacteriophage λ 48.5 kb). There is the giant genome of the Deltaproteobacterium Sorrangium cellulosum (14.8 Mb; yeast S. cerevisiae 12 Mb!) but the genomes of the smallest known self-sustaining, free-living cells are in the ~1.5 Mb range: 1.75 Mb for P. marinus SS120 (Cyanobacteria), 1.3 Mb for P. ubique HTCC1062 (Alphaproteobacteria), and 1.56 Mb for P. necessarius STIR1 (Betaproteobacteria). And, as already mentioned above, Prochlorococus and Pelagibacter cells are pretty small. On the other side, we now know of huge bacteriophages: the recently discovered Tsamsa phage of Bacillus anthracis with a 169 kb genome and a head diameter of 82 nm (tail length 440 nm), or phage G of Bacillus megaterium with a 0.5 Mb genome. The Lilliputians portrayed here neatly fill the size gap with their ~1 Mb genomes and their physical measures, as well. And since they are not freaky exceptions ‒ think of Carsonella ruddii or other endosymbionts with their extremely reduced genomes ‒ but represent a solid branch with numerous phyla in the bacterial family tree, the crude distinction between bacteria and their phages by size becomes increasingly fuzzy and thus obsolete. But what actually calls for a shift in perspective is the finding by the Banfield team that most known members of the huge group of Lilliputian bacteria have one or multiple incomplete biosynthetic pathways. To thrive ‒ a task they obviously accomplish ‒ they depend on cooperation with others from their tribe, or with foreigners. They are distinctly cellular but with a whiff of a phage-like lifestyle. The evolutionary root of the bacterial family tree is still hidden in haze. Finding out whether this trait is closer to the original bacterial lifestyle or a derived state of later adaptation would require yet another of Gulliver's travels...
Christoph Weigel is lecturer at the Life Science Engineering faculty of HTW, Berlin’s University for Applied Sciences and an Associate Blogger for STC