16s v4 region library preparation with negative control

[trong.wised]

Hi.
I am about to run 16s v4 region for bacterial population comparison between groups.
I used Illumina nextera index and kit earlier.
But now I would like to save cost, so I will turn to custom primers and custom preparation.
So, the pipeline of labwork is.
1. PCR the 16s v4 region with primer from this reference.

http://aem.asm.org/content/suppl/2013/08/06/AEM.01043-13.DCSupplemental/zam999104626so1.pdf

2. Check the band in gel electrophoresis.
3. Clean the PCR product with Ampure xp beads.
4. Measuring concentration and size with Qubit broad range and agilent tapestation.
5. Normalise all samples to 4nM concentration and pool them together into one tube.

Is this the correct workflow? or Should I change or add something else?

then my bigger question is:
for negative PCR control, if there is no band, many people said that I have to sequence it and delete it from other samples.
Then I have to make the negative control at 4nM and pool it to the library.
So, if I have a low negative control then amp it up to 4nM, isn't is a bias?
And when I delete it from other samples, if my sample has some of that OTU in it as a normal, so I delete it, isn't it a bias too?

I would like to make sure that I do not have a contaminant in my sequence results.
But I cannot answer myself about this question too.
Hope I can find some answers here.
thank you very much in advance.
Trong

[thermophile]

I don't know that there is a good way currently to deal with the sequences from negative controls. Here's a discussion that some of us have had on the mothur forum about neg controls https://www.mothur.org/forum/viewtopic.php?f=5&t=4145
What I will say is you need them. You need negatives from the extractions too, not just the PCR. Currently, I'm processing the negatives with samples including through clustering but then not doing anything with those. I check that I'm getting an order of magnitude fewer seqs from the negatives than from real samples (this is easy to achieve with the PCR controls, more difficult with the extraction controls).

On your pooling question, I run samples on a gel (actually qiaxcel) then pool a given #ng per sample up to some volume. So we'll pool our PCRs to 10ng of the ~400bp band, for most samples that's 1-10ul. We'll then pool 10ul of the negative controls. Our SOP is still in progress but if you'd like to see it offline, message me.

[trong.wised]

Thank you thermophile.
I was thinking about this when I was on the bed last night.
The same thing as you said.
Pool the samples first, make sure that all of them are equal.
Then just put x amount of negative on top of it.
Then sequence maybe 100,000 reads per sample and maybe 1,000 reads from negative.
So if all 1,000 reads are present in every samples. Then I should delete them out.
In this way, I will feel much more comfortable to do that.

Anyway what I am about to do in my study is, I will extract DNA from samples, every 7 samples will have one batch negative control.
So I will have lots of negative control, 12-16 batch controls.
That's why I am worry that if I normalise all of them my 90 samples may be diluted by these neg. control.

If you have any suggestions on batch control, I would love to hear them.
Thanks again.
Trong

[trong.wised]

I would like to send private message for the SOP but I cannot, maybe my junior status do not allow me to do it?
Is there any other way I can message you?

[thermophile]

sure email me at mars at uconn.edu

[fanli]

I'll just chip in my (rather long) 2c, as we've spent a great deal of time recently dealing with these issues.

First off, like @thermophile said, you absolutely need to include negative controls throughout your experiment. See the placental microbiome papers for motivation. For example, we include DNA extraction blanks pre- and post-sample processing, as well as PCR controls for each plate.

For pooling, we will include negative controls at the same equimolarity if it was successfully quantified by Tapestation, otherwise we add a larger volume (60uL currently). I don't think there is a clear consensus on whether you should add a similar volume, e.g. 10uL or less, if you don't get a real quantification on a blank, but our intent is to get as much sequencing as possible on those as possible to determine their composition more robustly.

After sequencing, we look at the negative controls both in the context of the run as well as over time wrt other runs. We examine the read depth even though in theory this is more a reflection of your pooling strategy than anything else. We look at the composition of the OTUs detected in the negative controls - do they look like any of the samples they were sequenced with? Do they look like the positive controls? Is there any pattern to their compositions? The expectation generally is that you should see fairly divergent compositions if the negatives represent stochastic amplification from low level amounts of DNA. For example, if all the blanks from one plate showed the same composition, we would be suspicious that something happened there. I should note that this is done both automatically by flagging certain characteristics and manually by eye.

In terms of what to do with OTUs/sequences detected in the blanks, we typically don't do anything. As long as we are happy with the composition and quality of the run, we simply proceed with analysis of our samples. Like Pat and others have said, it is difficult to say that simply deleting all of them is a good idea. "Subtracting" them out somehow may be better, but that runs the risk of introducing other biases that are a function of your pooling strategy.

[trong.wised]

Thank you so much @fanli.
I got your point now.
As long as they don't look like real or obvious contamination.
We should be happy with that results.