PCR-involved 454 pyrosequencing does not meet standards of qPCR

[linkingchen]

Hello fellow researchers,

We have recently submit a manuscript, used barcoded 454 pyrosequencing based on 16S rRNA genes to reveal the microbial community composition of our samples. One of the reviewer point out that, we cannot state the abundances of taxa in the samples, for PCR conditions do not meet standards of qPCR (we know the 454 pyrosequencing of 16S rRNA genes is PCR involved, when we used primer set (e.g. F515/R806) to get amplicons), and also we cannot use the wordings like "relative abundance" and "dominated" throughout the manuscript.
e.g.,
"The relative abundance of Proteobacteria in sample 1 is 10%".

What do think? Would you please give some suggestions for responding this comment.

Thanks.

Linking

[ajthomas]

I would have to agree with the reviewer. I do amplicon sequencing on the 454 as well, but my amplicons are MHC genes rather than bacterial 16S genes. A little background: In my species, there are ~6-10 class I MHC genes per haplotype that all amplify with the same primer set. Since there is only one diploid DNA genome per sample, I already know that all of the sequences I get out should have a copy number of either 1X or 2X, depending on if that particular allele is present in one or both haplotypes. However, I find that some sequences amplify better than others. Some sequences, when present, are always a large percentage of the number of reads and others are always a very low percentage. For example, a typical sample might have 600 sequence reads for a given amplicon. Of those, I might have 130 reads of one allele, 95 reads of another, 10 reads of another, 8 reads of another, and 30-40 of the rest. If these were 16S sequences I might conclude that these numbers correlated to abundance, but since these are MHC sequences coming from alleles present on one chromosome, I know I have the original copy number is the same in most cases, and occasionally 2X.

In short, unless you can verify that all of your sequences amplify with the same efficiency and you use some PCR strategy to stop amplification while in the linear phase, you can only say that a species is present or absent but not how much of the initial population was made up of that species.

[linkingchen]

I would have to agree with the reviewer. I do amplicon sequencing on the 454 as well, but my amplicons are MHC genes rather than bacterial 16S genes. A little background: In my species, there are ~6-10 class I MHC genes per haplotype that all amplify with the same primer set. Since there is only one diploid DNA genome per sample, I already know that all of the sequences I get out should have a copy number of either 1X or 2X, depending on if that particular allele is present in one or both haplotypes. However, I find that some sequences amplify better than others. Some sequences, when present, are always a large percentage of the number of reads and others are always a very low percentage. For example, a typical sample might have 600 sequence reads for a given amplicon. Of those, I might have 130 reads of one allele, 95 reads of another, 10 reads of another, 8 reads of another, and 30-40 of the rest. If these were 16S sequences I might conclude that these numbers correlated to abundance, but since these are MHC sequences coming from alleles present on one chromosome, I know I have the original copy number is the same in most cases, and occasionally 2X.

In short, unless you can verify that all of your sequences amplify with the same efficiency and you use some PCR strategy to stop amplification while in the linear phase, you can only say that a species is present or absent but not how much of the initial population was made up of that species.

Hi ajthomas,

Thanks. In the response, i agree with the reviewer's comment and change relevant description in my manuscript, and the manuscript is accepted by the journal now.

I wanna know if there are any solutions to stop PCR amplification while in the linear phase?

Linking

[ajthomas]

I've never tried to do that before, and I don't know of any protocols designed to do that. The only thing I can think of off the top of my head would be to do the amplification in a real-time instrument with Sybr Green in the reaction so you can monitor the amplification in real time and then stop the reaction at the appropriate time. That should work, but it would be rather low-thoughput and labor-intensive.

[Palecomic]

Hello fellow researchers,

We have recently submit a manuscript, used barcoded 454 pyrosequencing based on 16S rRNA genes to reveal the microbial community composition of our samples. One of the reviewer point out that, we cannot state the abundances of taxa in the samples, for PCR conditions do not meet standards of qPCR (we know the 454 pyrosequencing of 16S rRNA genes is PCR involved, when we used primer set (e.g. F515/R806) to get amplicons), and also we cannot use the wordings like "relative abundance" and "dominated" throughout the manuscript.
e.g.,
"The relative abundance of Proteobacteria in sample 1 is 10%".

What do think? Would you please give some suggestions for responding this comment.

Thanks.

Linking

Hi Linking,

I'm going to take the opposite view and disagree with the reviewer. 16S sequencing and other PCR microbial ecology techniques that have been established for years are considered semi-quantitative and do a good job of giving you the relative abundance of the most common organisms in the sample. I would say it's also completely fine to use the term dominated (measures of community evenness and diversity look at whether one organism is relatively more abundant than the others that are present).

He is right in saying that this isn't as accurate (or sensitive) as qPCR, but the point of the study, I would guess, is to look at all bacteria (or archaea) and not just those that you can design qPCR primers for.

Not every organism will display the same kinetics in the PCR reaction, and there are biases, but if you use the same set of primers and conditions for each sample, then you can compare beta-diversity perfectly well between the samples. Don't consider it to be an absolute abundance, but relative abundance within a sample and between sample differences are fine. I think your reviewer is not familiar with the field, it's very standard, there are lots of examples out there.

The important step is to rarefy (randomly resample) the number of reads per sample, to remove sequencing coverage bias and to ensure that every sample is handled and analysed in an identical way. MOTHUR, QIIME and a number of other pipelines for handling 16S data will do the rarefaction for you.

If your question concerns a particular organism, then it's better to use a specific qPCR, or to combine this with 16S on the same sample and validate abundance of this organism. In the past where I've validated the abundance of reads across a set of samples with qPCR in this way, I usually have good correlation. This will not be the case for all organisms, so be cautious.

With a mixed template, it would be impossible to stop the PCR in the exponential phase for every template as it would be different for every sample and community composition. Avoiding PCR amplification altogether and using a metagenomic approach, just sequencing all the DNA in a sample directly and sorting out the mess afterwards, bioinformatically is the alternative. That comes with a whole new set of challenges, not least - cost!