Culturable microbiome NGS

[neokao]

My lab has done 16S and whole genome shotgun on mouse gut/fecal microbiome (wild-type vs knockout). Recently we found that the culturable part of these knockout mouse fecal samples are actually effective in our following assays. So we'd like to further analyze the abundance and composition difference of the culturable microbiome (wild-type vs knockout) at species level.
I could not find much references for this kind of "metagenome" NGS.

Some questions to discuss here:

Assuming there are fewer species of culturable bacteria (compared with fecal samples ), which approach are we supposed to use? 16S or whole genome shotgun sequencing? What kind of NGS depth? How many biological samples are needed to see the difference?

Your input would be appreciated.

[Brian Bushnell]

There are all kinds of different strategies for this kind of study. Partial 16S is generally the cheapest, and probably easiest to analyze, but the most prone to amplification bias. But as long as all you care about are the relative differences before and after culturing, that may not matter. Note that I am not sure one can claim there is a "culturable" and "non-culturable" microbiome... I think the matter is a subject of active research.

Anyway, partial 16S is certainly the easier study. The problem of assembling/clustering/binning WGS metagenomes to get individual whole genomes and determine their abundance in different conditions is currently unsolved in the general case, while it's mostly a solved problem for partial 16S apart from the amplification bias issues. Still, for WGS, determining community gene abundance levels is pretty well solved. So it kind of depends on your goal, how much effort you want to put in, and the complexity of the microbiome.

As for depth, I recommend at least 20x for the least-abundant organism of interest, and ideally >40x, if you want to do WGS and assembly. Complex metagenomes typically follow an exponential abundance distribution (at least, in kmer-space) so doubling your sequencing depth will keep giving you new organisms, seemingly forever. You have to stop at some point, and that point is determined by when you, the researcher, think that organisms at that abundance (say, 1/10000 of the biomass) are still relevant to the function of the community, or when you run out of money. Of course, if you only have the computational capacity to assemble 8 Illumina lanes of data at most, then there's little point in generating more than that.

[neokao]

Thanks for the reply, Brian. As I mentioned, we have done 16S and WGS on our mouse gut/fecal microbiota already. I actually don't care about the differences before and after culturing. What we need to find is the differences between two sources of the fecal content microbiota in several days of liquid culture. I am aware the pros and cons of 16S and WGS since we have done them. My questions are for our particular "culturable" microbiome project, what's a better choice if I need to find the differences at species level? Apparently I am asking because we have budget issue. Otherwise I will just do WGS for 8-10 samples each condition (not practical for my lab).

There are all kinds of different strategies for this kind of study. Partial 16S is generally the cheapest, and probably easiest to analyze, but the most prone to amplification bias. But as long as all you care about are the relative differences before and after culturing, that may not matter. Note that I am not sure one can claim there is a "culturable" and "non-culturable" microbiome... I think the matter is a subject of active research.

Anyway, partial 16S is certainly the easier study. The problem of assembling/clustering/binning WGS metagenomes to get individual whole genomes and determine their abundance in different conditions is currently unsolved in the general case, while it's mostly a solved problem for partial 16S apart from the amplification bias issues. Still, for WGS, determining community gene abundance levels is pretty well solved. So it kind of depends on your goal, how much effort you want to put in, and the complexity of the microbiome.

As for depth, I recommend at least 20x for the least-abundant organism of interest, and ideally >40x, if you want to do WGS and assembly. Complex metagenomes typically follow an exponential abundance distribution (at least, in kmer-space) so doubling your sequencing depth will keep giving you new organisms, seemingly forever. You have to stop at some point, and that point is determined by when you, the researcher, think that organisms at that abundance (say, 1/10000 of the biomass) are still relevant to the function of the community, or when you run out of money. Of course, if you only have the computational capacity to assemble 8 Illumina lanes of data at most, then there's little point in generating more than that.

[Brian Bushnell]

Sorry, I guess I misinterpreted the nature of your experiment. Let me try again -

You will get different answers from the two approaches. If done optimally, WGS will give the more accurate answer - if you had a reference genome for every species involved, so you could measure abundance based on mapping unamplified shotgun reads, that would absolutely be more accurate than 16S. But I assume this is not the case. WGS is much harder to quantify without references, but you could for example quantify based on mapping WGS reads back to the 16S consensuses. This is also tricky because bacteria have different numbers of 16S copies with different levels of internal identity, but it's relatively straightforward to do, even though the answers may not be perfect (but they would be consistent if you used the same reference in all conditions). For the most part, if you have sufficient WGS depth, and you map with ambiguously-mapped reads being assigned randomly to one of their possible mapping locations, and the 16S fragments are assembled and clustered perfectly, then mapping the WGS reads to the 16S fragments should produce a very accurate portrayal of community depth.

I guess you have already done a little sequencing and are trying to decide whether to focus on WGS or 16S for the remainder, is that right? If you are sequencing-budget-constrained, just do 16S; since the amplification bias should be similar in both conditions, the relative levels should be correct. But since you already have some data with both, if you are interested in absolute levels, it might be helpful to determine how well your WGS and 16S data correlate, and whether you are able to accurately quantify your WGS data, before proceeding. Where they differ, the WGS data will be more accurate, if the quantification is done correctly. Again, that's an unsolved problem, but I think mapping WGS to 16S could be useful in cases where you have both. You only need one set of high-quality, high-coverage 16S data to use as a reference for any number of WGS datasets, to get amplification-free quantification.

[thermophile]

You have 16s on both fecal and cultured samples? I'd follow up with qPCR on the dominant species in each type. See how abundant the culturable community is in the original samples.