Barcoding 16S amplicons for Illumina Seq

[alfredman1]

Hi Folks,

I'd like to sequence a few 16S rRNA amplicons in one lane of an Illumina flow cell. I need to barcode my samples and would like a bit of advice. I have already read the threads in this forum discussing barcoding, but I guess I'm in need of more specific advice and would like to know if there is anyone out there with recent experience. A recent PNAS paper used a barcoding protocol which seems suitable (see Fig 1 of attachment). As I understand it, these researchers prepared their libraries as follows:

1) Made composite PCR primers consisting of (5`--> 3´) the Illumina adapter+the barcode+linker+16S reverse primer. Same for the forward primer but without the barcode.

2) Amplification products thus had Illumina adapters and barcodes

3) Sequenced in the GAIIx using custom sequencing primers (Read1, Read2 and Index Sequencing Primer). The use of custom sequencing primers, as I understand it, allowed them to get 100 bp reads that covered the insert (product of interest) and thus read length was not sacrificed to sequencing the 16S primer.

I have thought of hopefully replicating this, perhaps designing the barcodes with barcrawl and the linkers with PrimerProspector. However, I believe there might be some pitfalls of which I'm not aware. Besides, I am not 100% confident if I understand how this library was prepared.

I'd enormously appreciate help/comments from the community! Thanks!

All the best,

A

[TonyBrooks]

Sorry for the delay in answering. We're looking to do something similar and I found the post after a quick search.
As I understand it, the approach mentioned would sequence the 16s primers too. The Read 1 & 2 sequencing primers would sit outside the 16s amplicon (on the linker sequences), sequencing the whole lot (including the 16s rev and fwd primers).

------------------->>R1>>---------------------------------->>In>>---------------------------
ILLUMINA ADAPTER + LINKER + 16S FWD + NNNNNNNN + 16S REV + LINKER + INDEX + ILLUMINA ADAPTER
---------------------------------------------------------- <<R2<<---------------------------



Our intended approach is slightly different. Our reverse primer would use the Illumina adapter+index+16s rev. Our forward primer would be Illumina adapter + 16s fwd. We then intend to use custom R1, R1 and indexing primers. These are essentially untailed versions of the above primers (the indexing primer is the reverse compliment of the 16s rev sequence). This is a method similar to that suggested in the Quail 2008 paper (Supplimentary note 10)


------------------ >>>R1>>-------------->>In>>>---------------------------
ILLUMINA ADAPTER - 16S FWD - NNNNNNNN - 16S REV - INDEX - ILLUMINA ADAPTER
----------------------------------------<<R2<<<---------------------------


One thing me might do though is to look at a few 16s variable regions to increase the diversity of the library as much as possible. Some of the data we get from the 454 would suggest that the 16s variable regions aren't diverse enough for the GAII's to correctly identify clusters.

[kmcarr]

Tony,

Just curious, are the PCR primers (16S fwd & rev) degenerate? This is typically the case, is it not, for 16S amplicon sequencing. If so then do you intend to cary the degeneracy through to the sequencing primers?

[pmiguel]

Originally Posted by TonyBrooks

One thing me might do though is to look at a few 16s variable regions to increase the diversity of the library as much as possible. Some of the data we get from the 454 would suggest that the 16s variable regions aren't diverse enough for the GAII's to correctly identify clusters.

One generic work-around would be to mix in other, unrelated, bar coded libraries. This reduces the likelihood of any two adjacent clusters having identical sequence during the first 4 cycles.

It worked for some SMART adapter flanked cDNAs we wanted to sequence. Just put them in a lane with a genomic DNA library.

Of course if you are using custom sequencing primers that might be trickier...

--
Phillip

[TonyBrooks]

Originally Posted by kmcarr

Just curious, are the PCR primers (16S fwd & rev) degenerate? This is typically the case, is it not, for 16S amplicon sequencing. If so then do you intend to cary the degeneracy through to the sequencing primers?

Our current 454 design contains a couple of degenerate bases. We'll probably maintain these throughout the run, so that our sequencing primers also contain the same. We're still thrashing out the details of how to run it.

Originally Posted by pmiguel

One generic work-around would be to mix in other, unrelated, bar coded libraries. This reduces the likelihood of any two adjacent clusters having identical sequence during the first 4 cycles.

It worked for some SMART adapter flanked cDNAs we wanted to sequence. Just put them in a lane with a genomic DNA library.

Of course if you are using custom sequencing primers that might be trickier...

We could use that approach and mix the Illumina and custom primers. However, I'm not sure we'd be happy losing reads to something we don't want to sequence.

[DBL]

You can find protocol along with primer constructs in paper below:

Caporaso, J.G., C.L. Lauber, W.A. Walters, D. Berg-Lyons, C.A. Lozupone, P. Turnbaugh, N. Fierer, R. Knight. 2010.
Global patterns of 16S rRNA diversity at a depth of millions of sequences per sample. Proc. Natl. Acad. Sci. USA.
doi:10.1073/pnas.1000080107

[TonyBrooks]

Just to complete this thread, we finally got our first run on last week (yes, it really takes 2 years to get things done around here) and we're happy with the results.

Dual indexing worked beatifully but there were a few negatives. We had a slightly low PF rate of 68% (CD=793k/mm2, PhiX @ 5% gave 6.4% aligned) and Q-scores tanked at 200bp (although we've seen than on genomic libraries too). MiSeq analysis seems fine, although we had ~25% of samples that were unassigned to any phylum.

We quality trimmed at Q20 and stitched the reads together but lost about 75% of reads that either no longer overlapped, or had a mismatch in the overlap region and thus discarded, We still had about 24k reads per sample (about 2.5X more than we had on 454, so we're still happy). Those are about to go through offline analysis.

[pmiguel]

Originally Posted by TonyBrooks

We quality trimmed at Q20 and stitched the reads together but lost about 75% of reads that either no longer overlapped, or had a mismatch in the overlap region and thus discarded,

Might want to try merging the reads before quality trimming.

--
Phillip

[kmcarr]

Originally Posted by TonyBrooks

Just to complete this thread, we finally got our first run on last week (yes, it really takes 2 years to get things done around here) and we're happy with the results.

Dual indexing worked beatifully but there were a few negatives. We had a slightly low PF rate of 68% (CD=793k/mm2, PhiX @ 5% gave 6.4% aligned) and Q-scores tanked at 200bp (although we've seen than on genomic libraries too). MiSeq analysis seems fine, although we had ~25% of samples that were unassigned to any phylum.

We quality trimmed at Q20 and stitched the reads together but lost about 75% of reads that either no longer overlapped, or had a mismatch in the overlap region and thus discarded, We still had about 24k reads per sample (about 2.5X more than we had on 454, so we're still happy). Those are about to go through offline analysis.

Tony,

What tool/pipeline are you using for offline analysis; have you looked at Mothur? In mothur's standard pipeline for MiSeq data the read pairs are first overlapped, then screened.