Size Variation Of The 16S Gene

Question

[xapple]

I have a bunch of sequences obtained from a metagenomic experiment targeting the 16S gene of bacteria. The primers are designed to span from V3 to V4. The 16 rDNA gene is relatively well conserved, nonetheless I expect some variation in the size of this region. However, I'm seeing pretty large variations in the size of these fragments as you can see in the graph below. I would have thought that the distribution was going to be more compact. Where could I check for the sanity of this data ? I can't seem to find any sources on the amplitude of the naturally occurring length variation of the bacterial 16S gene.

EDIT - A few clarification to answer question from @Mabeuf:

• The source sample is a lake sediment core (highly phosphorus-saturated sediments) from which DNA was extracted.
• The primers used are 341F 5' -CCTACGGGNGGCWGCAG-3' and 805R 5' -GACTACHVGGGTATCTAATCC-3' and was run with no other sample on an Illumina MiSeq.
• Subtracting those two numbers 805 - 341 = 464 base pairs would fall right between the two right peaks on the graph.

Answer 1

[Daniel]

My initial thought is that you are picking up multiple taxonomic kingdoms i.e Bacteria, Archaea, Eukaryotes.

A few questions which might point in the right direction:

• What was your source samples? Could they include some eukaryotic 18S (I've seen this happen in my own data)
• Which primers did you use and which peak is the one which you would be expecting? Do they have a history cross amplification? 357f-518r do (although that's too small for this data)
• Is there any chance that the sequencers that you used (if an external company) ran your samples on different machines/runs. I had a dataset return with two peaks because one was ran on 454 and one on 454+. Unlikely, but just mentioning it.

[xapple]

I edited my question to provide more clarifications ! So you do think that the spread is too large to be of natural cause ?

Answer 2

[fanli.gcb]

An old question, but recently came up in some of our work using the V1V2 (27F-338R) primer set. There seems to be a lot of length variation in our amplicons:

This paper also reports substantial length variation in the V3V4 amplicon (341F-534R): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4643231/

[5heikki]

I think such variation is completely normal. In e.g. current RDP (large file) the longest bacterial 16S rRNA is 2,487 bp. I imagine the majority of the length difference between that and the 1,541 bp reference E. coli 16S is due to hyper variable regions. If it's important, you could align a random subset of RDP to the E. coli reference to see for yourself..