PCR Recombination

In my naive mental model of how PCR works I thought once the polymerase starts extending the 3’ end of the primer, it just keeps going until it disconnects from the template and then that strand is done. Especially if you use a $$ high fidelity, high processivity polymerase like Q5 and tightly control the extension time. But I was wrong.

We were trying to amplify a library of hairpin sequences. The library was a single stranded oligo pool from Twist. Each oligo once cloned into a plasmid and transcribed in a plant cell would create an ssRNA hairpin to target a different endogenous gene. The structure of each oligo is shown below.

enter image description here

I will not get into the bigger picture of why we were doing this. The project on the whole did not work and is irrelevant to this topic. But we were trying to add the second strand using PCR and primers that landed on either end of these oligos.

Each oligo had a region labeled stem which is 65nt and homologous to a targeted endogenous gene. The region labeled stem’ is the reverse complement to stem. The loop region and the cloning and primer sites were the same on all oligos. I had 2 primers - one that would bind on one end to initiate second strand synthesis, and the other would bind to the other end one a second strand was created. After the template oligos had second strands created, the primers should act like normal PCR primers.

We had to do a lot of optimization to the PCR conditions to get amplification to happen. Adding betaine to the Q5 master mix allowed the reaction to proceed and generate a correct sized band on an agarose gel. We cloned this amplified library into a plasmid backbone and sequenced the plasmid library. Thank you to Sayeh Gorjifard at UW for advice on how to do this library cloning. Everything was going great up to this point.

Then we sequenced the plasmid library and noticed a common phenomenon. The two stems cloned into a single plasmid did not match up. Each oligo in the library was designed to only target one gene. Oligo #1 would contain A-loop-A’ oligo #2 would contain B-loop-B’ etc. The plasmids should not contain A-loop-B’. But the Plasmidsaurus nanopore sequencing said that they did.

Three categories of explanations:

  1. Twist somehow botched the oligo library synthesis in a very specific
    and coincidental way.
  2. Our cloning process was recombining the oligos.
  3. Plasmidsaurus was recombining portions of the plasmid, presumably during the library prep.

You can probably guess which one it was based on the title of this essay. (1) seemed incredibly unlikely. I know the oligo synthesis process is error prone but only in a way that produces truncations and point mutations. Not randomization of oligo fragments. Testing this would also be very difficult so we ignored it and would come back to this explanation as a last resort.

To test option (3) we made a couple different non-library plasmid minipreps that each contained only one specific hairpin. Miniprep 1 only contained a plasmid with A-loop-A’. Miniprep 2 only contained a plasmid with B-loop-B. We made 5 different versions of this and we sequenced these individually to make sure this was true. Then we blended them together and sent the mixture to Plasmidsaurus. The results showed no recombination. So option (2) was looking more likely.

At that time Plasmidsaurus could do amplicon sequencing but (A) the standard service did not support library sequencing1 and (B) the minimum amplicon size was like 500bp if I remember correctly. Our amplicon was somewhere around 170bp. So we set up a custom service with them to sequence our super short amplicon library. The results were very illuminating.

[image]

  1. Not officially at least. We sent plenty of libraries to them using their standard $15 service and just analyzed the raw reads ourselves. Works great. ↩︎