On 2021-12-30 15:35:00, user Jacob Cram wrote:
This is a public comment on Szabo et al. “Ecological stochasticity and phage induction diversify bacterioplankton communities at the microscale”, submitted to BioArxiv on Sep 21, 2021.
Understanding the dynamics by which microorganisms attach to and grow on particles is an important and contemporary field in microbial ecology, and in the understanding of the factors that influence the role of particle flux in the global carbon cycle. Szabo et al focus on the randomness of this process. By taking ~1000 identical chiton beads and incubating them in the sea-water from the same sample, and looking at the community structure 100 beads at a time, over the course of seven days, the authors aim to quantify how much variability there is in the microbial take-over of these particles.
The authors applied shotgun metagenomics to each and every particle, focusing on assembling genomes into metagenome assembled genomes (MAGs).
Several key findings stand out to me:
1) There is substantial variability over time in the microbial community structure, and on the number of microorganisms present per particle. <br />
1a) The authors suggest that random variation in which bacteria attach to the particles and when they attach drives much of this variability.
2) There do not appear to be statistical associations between which microorganisms are on a given particle. That is if a given species “A” is common on particle A and not particle B, that has no bearing whatsoever on the abundance of any other microbe on either particle.<br />
2a) Such a finding suggests that there are essentially no meaningful interactions between the microbes on the particles. Cross feeding, predation, symbiosis, chemical warfare, all believed to be important for microbial communities (Fuhrman and Steele 2008; Steele et al. 2011) would each be expected to lead to some sort of statistical association between organisms, but in this scenario at least such patterns are essentially absent.
3) The authors looked for contigs (partial phage genomes) and identified which appeared to “bin into the MAGs of their bacterial hosts”, suggesting that they were lysogenic with and therefore part of the genome of at least some members of that host. The more copies of this contig were present, the more active this phage was said to be. They found associations between the activity of these phages and the apparent growth of their hosts and negative associations between bacterial abundance and the presence of these phages.<br />
3a) The authors suggest that stochastic absence of particular phages can lead to the situations where their hosts can rapidly take over a particle.
I found this to be a very thought provoking manuscript and it raises a number of interesting and testable questions for future research. The sequencing and assembly of so many metagenomes, especially on very low biomass samples is an impressive technological feat (and clearly required diligent work on the part of the authors) which will be of value to the community at large. While some of my comments below are critical, I want to be clear that I was quite impressed with this paper and share these comments because I think the research is important and merits reflection.
I have comments about each of the three main points listed above that I would like to share. I have not, as of yet, been asked to review this manuscript for any journal, but would be happy for any editor to use my comments. After preparing this review, I discussed it with Dr. Mina Bizic and she indicated that she shares my opinions. Dr. Bizic had several additional comments which she plans to make separately.
Comment 1: On Stochasticity
The authors make the case that there is randomness in the attachment and growth dynamics of microbial communities on particles. The authors suggest that because the variability between the communities on the particles is much higher than that of the surrounding water samples. However, I suspect that random variability in which rare taxa end up in each incubation could drive many of the patterns that they see.
As context, in this experiment, chitinous beads (~80 micron diameter) are enclosed, one per well, in 96 well plates and incubated in, 175 ul of sea water. The microbes and particles have been concentrated in this small volume up to ten times by centrifugation. That is, volumes of whole sea-water were filtered, and then centrifuged and the bottom 1/10, presumably containing intact cells and small particles from the environment was retained. This means that each bead is incubated with essentially 1.75 ml of sea-water worth of microbes and microaggregates.
I suspect that microbes that are adapted to degrading chitinous beads are scarce in the water, perhaps near or slightly below a concentration of 1 per 1.75ml. In this case, there could be random variability in whether chitin degrading microbes end up in any given well. Furthermore, a big driver in the randomness between which bacteria are in a well could be the presence of chitinous particles (smaller than the 60 micron filtration cutoff) in the background water. Ambient chitinous particles likely contain communities that would be adapted to break down chitinous beads. If one well happens to have one of these particles that particle is likely to come in contact with the bead near the beginning of the experiment in which case the microbes on the microaggregate can take over the bead. If such particles are absent, then perhaps the takeover of the bead doesn’t happen, or happens more slowly. Thus the stochastic process that drives the variability that the authors see may be in the starting community of the water in which the particle is incubated. If these organisms are rare, they would be likely to be missed by the sequencing, which can only sample the most abundant organisms. As they sequenced the seawater samples to a depth of ~500,000 reads per sample, and maintained about 25% of the samples (Table S5), this means that they essentially considered ~125,000 sequences per sample. Assuming the water had on the order of 1 million bacteria per ml, we might expect that any organism present at lower than ~10 copies per ml would likely be missed by their process. As there is an amplification step in their sequencing (supplementary methods) their method may even be less sensitive to rare organisms.
Indeed, it is clear that the sequencing of the seawater didn’t catch every organism that could colonize the particles because per Table S7, some of the jackpot taxa (taxa that take over some particles) are either never seen or rarely seen in the seawater samples. Since they must have come from the seawater, it is clear that some species are missed by sequencing.
Thus I contend that some of the particle to particle variability is likely from well to well variability in which microbes were stochastically placed in wells with each particle.
On the other hand, it is possible that this stochasticity is environmentally relevant. For instance, an 80 micron bead that sinks through 100 m of the water column only clears a total volume of ~500 μl {π(80 μm / 2) ^2 * 100 m = 503 μl} and so it is possible that microbes beyond this abundance in water would actually be unlikely to encounter a particle as it sinks out of the photic zone, for instance.
Comment 2: On interactions
I’m surprised that there don’t seem to be interactions between organisms, but their graphical lasso based statistics seem reasonable to me.
I’m furthermore surprised the authors did not seem to consider Bižić-Ionescu et al. (2018)’s paper, which has a very complementary design to this paper, but seemed to find the opposite pattern with respect to microbial interactions.
Bižić-Ionescu et al. (2018) presented a very similar project, in which the authors also had replicate particles, though fewer than in the paper by Szabo et al. Key differences were that the authors used a flow-through rolling tank which exposed the particles to more water, and that those authors used (larger) aggregates of algae rather than chitinous beads as their particles. Bižić-Ionescu et al. did not quantify the variability in microbial abundance and so would not have seen the abundance dynamics that Szabo et al. saw, if they had occurred. Like Szabo et al. (this manuscript), they suggested that differences in the timing of microbial colonization of particles drive a lot of the particle-to-particle variability. Bižić-Ionescu et al. also saw statistical patterns that suggested interactions, as well as expression of genes for microbial interactions, including antagonistic processes. I hope the authors will consider the possible differences between the two systems and why those might lead to different dynamics, and what that says about the robustness and environmental realism of the patterns seen in both experiments.
Comment 3: On viral contigs and non-assembled microbes
The authors consider viruses that bin into MAGs which I presume means that they are often or always part of the microbial genome of a particular organism. I am not an expert on this process, but it seems to me a reasonable way of assigning viruses to hosts. I note that other validated tools for metagenomic host assignment are also available (Zielezinski et al. 2021). I presume there are many viral contigs that did not bin to a specific MAG. Why did the authors choose to ignore these?
Similarly the authors focus only on those species that assemble into MAGs, I presume there is a bunch of microbial diversity that doesn’t assemble (since my impression is that in most communities not all sequenced contigs end up as part of a MAG). Could the authors expand on why they chose to ignore this diversity, and what impacts on their analysis only looking at assembled bacteria and not the rest of the microbial diversity might have on the analysis.
I thank the authors for sharing this pre-print in a public forum and encourage them to consider these comments.
Sincerely,<br />
Jacob Cram
References
Bižić-Ionescu M, Ionescu D, Grossart H-P. Organic Particles: Heterogeneous Hubs for Microbial Interactions in Aquatic Ecosystems. Front Microbiol [Internet]. 2018 [cited 2019 Dec 18];9. Available from: https://www.frontiersin.org...
Fuhrman J, Steele J. Community structure of marine bacterioplankton: patterns, networks, and relationships to function. Aquat Microb Ecol. 2008 Sep 18;53:69–81.
Steele JA, Countway PD, Xia L, Vigil PD, Beman JM, Kim DY, et al. Marine bacterial, archaeal and protistan association networks reveal ecological linkages. ISME J. 2011;5(9):1414–25.
Zielezinski A, Deorowicz S, Gudyś A. PHIST: fast and accurate prediction of prokaryotic hosts from metagenomic viral sequences. Bioinformatics. 2021 Dec 14;btab837.