How to normalize/pool a 16S library when samples are highly variable concentrations??

[samd]

Hi all,

I've been having this issue that has been plaguing my PhD. Some of my sequencing runs come back with terrible uneven read coverage. I am guessing it is a pooling problem but still not 100% sure.

For some background on my pipeline: after my final bead clean step I quantify all my samples and usually get a huge range of concentrations from 5ng/ul to 100+ ng/ul. This isn't surprising as my gut microbiome samples come from a variety of fish species. I then proceed to pool the samples by adding different volumes to create a 20ng/ul final pooled library. Theoretically, I am adding the same amount of moles of each sample to the pool. Obviously it isn't perfect but I do not understand why my read coverage is so terrible. I usually get back many samples with <1000 reads and some with over 300k reads on an Ilumina Miseq (16S V4 amplicon).

My question is: is there a different way to normalize the concentrations before pooling?

Is diluting samples with water to normalize concentrations (adding different amount of H20 to get them all the same concentration) helpful before pooling?

Are highly variable sample concentrations prior to pooling a culprit for uneven read coverage?

For indexing (2nd PCR), I usually add 1ul of PCR product or 2ul depending on the brightness from the gel. This has worked fine before but would it help if I quantified my PCR products and then added a more precise volume for indexing to keep it more normalized?

I pool in a way so I am adding a minimum of 2ul so as to avoid pipette error with small volumes.

Some additional info:
I do a 2-step PCR process with 515-806R primers
Qiagen multiplex kit for PCR 1 and Kappa Hi-Fi for PCR 2
Nextera Unique Dual indexes for indexing
Omega RXN Pure Beads for bead cleans
Qubit for quantification

Any help would much be appreciated!!
Cheers,
Sam

[nucacidhunter]

High variability of concentration could be one reason.

To reduce variability:
1- Quantify 1st PCR amplicons and use the same mass for indexing PCR. This will reduce gel invisible artefacts.

2- Normalize 2nd PCR in two steps, first to 10 ng/ul and then to 5 ng/ul. I think normalizing to 20 ng/ul is too high.

3- If PCR yield is not limiting use 3 ul for Qubit quantification

You still will see variability but it should be limited max 2x. This will increase hands on time but depending on your workload might worth.

[samd]

@nucacidhunter Thank you for this detailed answer. All these make sense especially the 2 time normalization. Why is lower concentration better? More accurate?
I will employ these modifications once things return back to normal.
Thanks again.

[MU Core]

Do you normalize your template input for the 1st PCR? We have found this very useful and critical in getting an even representation of reads in the pool. As long as there are no PCR inhibitors in the sample (i.e. similar PCR efficiency in each reaction), we are able to pool all wells across a 96-well plate by equal volume and perform a single bead clean-up. The variation in read count across samples is minimal.

[samd]

@MU Core I actually also have gotten similar advice from someone at a sequencing core. The one thing is that I do have huge variability in PCR efficiency and output. However I feel like this pre-normalization step combined with what @nucacidhunter said would greatly help my problem.

As for the beadclean step, are you saying that you simply do a single beadclean on the pooled library? That would save me a great amount of time.
Cheers,
Sam

[MU Core]

That's correct, a single bead clean-up on the pool. As long as you have good amplification across the plate and normalize template input amount, there is no need to quantitate and normalize the PCR products.

[cement_head]

SequalPrep Normalisation Kit (96)

Load 25, or 50 microL of PCR and wash away the rest...

[itstrieu]

When we prep 16s libraries, the input is normalize for 1st PCR. We normalize based on molarity after the clean-up from 2nd PCR using Quant-iT/Qubit and a Bioanalyzer trace. This usually gives us uniform coverage for all samples.

[cement_head]

Originally Posted by itstrieu

When we prep 16s libraries, the input is normalize for 1st PCR. We normalize based on molarity after the clean-up from 2nd PCR using Quant-iT/Qubit and a Bioanalyzer trace. This usually gives us uniform coverage for all samples.

Interesting. I use one-step construction (EMP Protocol), but I could see how normalising the first step in a two step would be a good idea.

[ATϟGC]

I have used the SequalPrep Normalisation 96-well plates for about 4-5 years now after the second and final round of PCR and they seem to work OK when you have strong and even amplification of your target amplicon (I think their manual even says something like this). However, in my hands there is large variation in per-sample read number when I get slightly uneven amplification across my 96 samples. An added problem I have encountered with these plates is that after I elute out of the 96 wells, pool them all together, bead-clean to concentrate for the Pippin prep, and then run on the Pippin prep to eliminate off-target bands, I barely have enough product for a Miseq run with 2 other amplicons based on my sequencing providers minimum input limits.

If you have a nice concise amplicon product I would recommend trying the Qubit method itstrieu mentioned above as this should work well so long as you do not have too much off-target amplification (off-target amplification makes normalization based on product quantity more difficult).

I have done a few hundred fish gut extractions over the years and since your samples are coming from various species I should also add that these types of samples can contain moderate to high levels of PCR inhibitors and the amount of inhibitors can vary by species. Some have more mucous and some have more plant-based inhibitors like polyphenols and tannins from their diet and/or the prey of their diet. This is usually only a problem in round 2 PCR. If inhibitors are causing uneven amplification, dilution of template DNA even in the low-yielding samples can sometimes help even this out. Using a better extraction method can help in most cases as well if inhibitors are co-purifying with your DNA.

[ATϟGC]

correction....inhibitors are typically only a problem in round 1 of PCR. By the second round they are usually diluted enough that they do not cause problems.

[cement_head]

Originally Posted by ATϟGC

I have used the SequalPrep Normalisation 96-well plates for about 4-5 years now after the second and final round of PCR and they seem to work OK when you have strong and even amplification of your target amplicon (I think their manual even says something like this). However, in my hands there is large variation in per-sample read number when I get slightly uneven amplification across my 96 samples. An added problem I have encountered with these plates is that after I elute out of the 96 wells, pool them all together, bead-clean to concentrate for the Pippin prep, and then run on the Pippin prep to eliminate off-target bands, I barely have enough product for a Miseq run with 2 other amplicons based on my sequencing providers minimum input limits.

If you have a nice concise amplicon product I would recommend trying the Qubit method itstrieu mentioned above as this should work well so long as you do not have too much off-target amplification (off-target amplification makes normalization based on product quantity more difficult).

I have done a few hundred fish gut extractions over the years and since your samples are coming from various species I should also add that these types of samples can contain moderate to high levels of PCR inhibitors and the amount of inhibitors can vary by species. Some have more mucous and some have more plant-based inhibitors like polyphenols and tannins from their diet and/or the prey of their diet. This is usually only a problem in round 2 PCR. If inhibitors are causing uneven amplification, dilution of template DNA even in the low-yielding samples can sometimes help even this out. Using a better extraction method can help in most cases as well if inhibitors are co-purifying with your DNA.

We do a lot of environmental (soil, lake-water) and we do sometimes get off-target bands, but we don't bother cleaning-up, just pool, reduce volume using SpeedVac, KAPA and run them on the MiSeq. We sort everything bioinformatically on the backend. I tried the gel purification approach and ended up losing too much sample, like you've found.

[samd]

Hi Grace,

This is interesting. What do you guys use to elute after the last clean up stage? We use water anyway (based off the manufacturers instructions) so I figured adding more water wouldn't hurt.