PCR enrichment - necessary?

[austinso]

Out of curiosity, and with the TruSeq v2 Y-adaptor chemistry:

If you have sufficient amounts of adapted DNA, how important is it to do the PCR enrichment? In principle, my expectation is that cluster generation will work without enrichment since the adaptors should be ligated on both strands. In not a few instances, I've been getting yields upwards to 0.5 nM of adapted DNA, which is more than enough for putting on the Miseq.

Thoughts?

[luc]

You have certainly figured it out by now. For the Truseq style adapters, I know of no reason to carry out a PCR enrichment, if you had enough starting material and the adapter ligation was successful.

[pmiguel]

But see here:

Is there a PCR-free TruSeq Sample Prep protocol available?

The new index adapter design enables PCR-free protocols. (A single cycle of synthesis is required to separate the forked adapter.)

When I wrote to Illumina tech support to clarify, the responded with the following:

We have found that if you simply take the annealed fragment that has the forked adapters on each end and add it to the clustering reaction that even with NaOH treatment before clustering, there is poor initial hybridization and the resultant clustering is poor. By allowing at least one PCR cycle this separates the strands and the forked adapters, allows a single amplification of the entire strands, resulting in a clean, double-stranded species that gives cleaner clustering results.

Which make little or no sense to me. But, in practice, the one time we tried no-amp with a TruSeq library on a high GC% genome, the cluster density was much lower (>10% of expected) than we expected. If I recall correctly, this was based on qPCR quantitation of the no-amp library. In any case, the amplified version of the same library did cluster pretty much as expected.

Lots of caveats here, but the point is: beware. Certainly you want to at least do a qPCR quant of the library. As I have written elsewhere, tests we have run at various points seemed to indicate that if all goes well on the order of 10% of the pre-amp molecules in the right size range to be amplifiable as quantitated by qPCR. Whether this is the result issues with the insert, the ligation or our quantitation, I don't know.

I am just saying don't ignore the 1024 pound PCR gorilla at your back when you do those 10 cycles of enrichment amplification. My take is the world looks a lot friendlier than it actually is with a thousand-fold amplification gorilla pummeling into submission any threats that come your way, eh? Don't assume that you can go traipsing around the same neighborhoods with impunity if you give the gorilla the night off.

--
Phillip

[TonyBrooks]

OK, I'm probably about to show my complete lack of chemistry knowledge, but here goes

NaOH works because the hydroxide ions somehow interfere with the hydrogen bonds between the complementary strands. Maybe the presence of the ssDNA at the fork prevents the ions from accessing the hydrogen bonds. However, PCR'd library has blunt ends meaning the hydrogen bonds would be exposed to the NaOH much more readily.

Obviously 95C does denature forked DNA (otherwise PCR wouldn't work).
I often wonder if a 5 minute 95C denature would actually give better results for a PCR-free library - plus there's no rogue NaOH hanging round in your hyb buffer, thus allowing hybridisation of really tiny amounts of library.

[luc]

Quantification with qPCR is of course required. We have not noticed any detrimental impact of omitting any PCR entirely and I have problems understanding the advice of the Illumina tech person. In my understanding the libraries are (and need to be) denatured (by NaOH) before clustering in any case.

[pmiguel]

Hi luc,

That is interesting. Are you using adapters provided in the TruSeq DNA prep kit?

I agree the Tech support advice makes little sense theoretically.

I just presumed that NaOH left the phosphate back-bone completely de-protonated. The two negatively charged back-bones would repel each other, tearing the strands apart. So that should work fine for forked adapters.

The adapters themselves may have modifications that makes them either poor at hybridizing to the flow cell oligos, or being replicated by the clustering polymerase. They could even be a few bases short at their 3' ends. Just enough to make hybridizing to the flow cell inefficient.

--
Phillip

[TonyBrooks]

Any chemists out there want to explain the nature of Alkaline denaturation of DNA?

[pmiguel]

Not a chemist, per se. But here is the journal article recommended to me by google:

Biophysical Journal, Volume 9, Issue 11, 1281-1311, 1 November 1969
doi:10.1016/S0006-3495(69)86452-0

As is generally the case with papers of this vintage, density centrifugation, rather than gel electrophoresis, is used as the method of choice for size (length) determination. Which seems alien to the modern mind-set.

For what it is worth, the authors found major difference between thermal and alkaline denaturation.

--
Phillip

[luc]

Hi pmiguel,

we are using the barcoded Nextflex adapters from Biooscientific, which have the same sequences as the Illumina Truseq adapters (except some of the barcodes themselves). The pcr-free adapters they sell are identical to the regular Nextflex adapters but are provided at aq higher concentration.
We are using NEB reagents and the Enzymatics ligase for the library preps.

[pmiguel]

Hi luc,
Well then, looks like the effect, if it is real, results from some characteristic of the Illumina TruSeq adapters.

You have established that their sequence is basically the same as the Biooscientific adapters you use. The TruSeq adapters are known to have some modifications. Cytosines are 5-methylated. The last two 3' bases have a phosphorothiate linkage to resist some exonucleases.

Neither of the above seem likely to cause the phenomenon. So we are left with speculation. What could cause this effect? How about a thermolabile interstrand linkage between the strands of the TruSeq kit Y-adapters? I suggest this because last time I read the Applied Biosystem's mate pair library kit protocol, it appeared they had added a new chemistry to their "center adapter", and that sounded just like such a thermolibile inter-strand linkage.

Such a linkage would have the advantage that even if the temperature of the ligation, or the adapters, went above the Tm of the 10(?) bases of the Y-adapters that anneal, there would be little loss of double-strandedness. But it would likely also confer the property of not being denatured by base, and hence would make a poor substrate during clustering.

Should be possible to overcome by heating. Although it is possible this hypothetical linkage might only be labile in the presence of other components of the enrichment PCR reaction.

--
Phillip

[austinso]

Well...I'm glad that everyone agrees that it is a bit odd to say the least...the NaOH denaturation should be good enough to denature everything, and no mods can offset this effect.

So, it seems to me that if you got low loads, there may be enough contaminating adaptor to hyb to the surface to compete with the properly adapted fragments i.e. the other purpose of the PCR is to effectively increase the amount of the bona fide library fragments over the unlighted adaptors (which go through another round of purification).

More to consider I suppose...

[pmiguel]

Well...I'm glad that everyone agrees that it is a bit odd to say the least...the NaOH denaturation should be good enough to denature everything, and no mods can offset this effect.

So, it seems to me that if you got low loads, there may be enough contaminating adaptor to hyb to the surface to compete with the properly adapted fragments i.e. the other purpose of the PCR is to effectively increase the amount of the bona fide library fragments over the unlighted adaptors (which go through another round of purification)

Nope, I can definitively say this is not the cause in our case.

These were bacterial genome libraries, so we clustered them as 10% of the total lane. With the amplified libraries, we got a reasonable fraction of the lane. With the non-amplified libraries, in two different attempts, we got less than 1% of the number of clusters we were trying for.

However, the other libraries in the lane clustered normally. The overall cluster density was more-or-less normal. So the effect was limited to only libraries that had not been enrichment amplified. Hence no global effect on flow cell oligos is a possibility.

I find the idea of having sufficient contaminating adapter oligos present in a library to diminish the loading of the lane far-fetched in any case. Initially the numbers of flow cell oligos must vastly (1000x?) outnumber the library molecules. I can't see how they would be limiting in any normal situation. TruSeq library construction involves Ampure clean-ups as well as a gel cut, post ligation.

Seems more tenable to me to hypothesize that either the adapter strands are covalently linked by some heat-labile moiety or that some genomic DNA contains bases not easily replicated by the Illumina cluster kit polymerase.

--
Phillip

[austinso]

Nope, I can definitively say this is not the cause in our case.

These were bacterial genome libraries, so we clustered them as 10% of the total lane. With the amplified libraries, we got a reasonable fraction of the lane. With the non-amplified libraries, in two different attempts, we got less than 1% of the number of clusters we were trying for.

However, the other libraries in the lane clustered normally. The overall cluster density was more-or-less normal. So the effect was limited to only libraries that had not been enrichment amplified. Hence no global effect on flow cell oligos is a possibility.

Ah...I see...that wasn't all that clear to me. How many libraries were loaded on the flow cell in your situation?

I find the idea of having sufficient contaminating adapter oligos present in a library to diminish the loading of the lane far-fetched in any case. Initially the numbers of flow cell oligos must vastly (1000x?) outnumber the library molecules.

Never heard what that surface density is (I'm not even sure of the flow cell area) but recommended loading gives 600 uL @ 6 pM = 2x10^9 molecules.

I can't see how they would be limiting in any normal situation. TruSeq library construction involves Ampure clean-ups as well as a gel cut, post ligation.

So are you saying that no one (not just you) should see any impurities, and that failed runs or misloaded libraries are primarily due to genomic content? I'm just asking because I do not have the experience that most people here would have...

Seems more tenable to me to hypothesize that either the adapter strands are covalently linked by some heat-labile moiety or that some genomic DNA contains bases not easily replicated by the Illumina cluster kit polymerase.

I suppose that could be a possibility...

Austin

[pmiguel]

Ah...I see...that wasn't all that clear to me. How many libraries were loaded on the flow cell in your situation?

I can look it up, if you like. If memory serves it was 6 or more. We were shooting for 5-10% of the clusters being derived from these 2 bacterial strains.

Never heard what that surface density is (I'm not even sure of the flow cell area) but recommended loading gives 600 uL @ 6 pM = 2x10^9 molecules.

Think about it this way: if the flowcell oligos do not vastly outnumber the number of library molecules, then what is the purpose of the cluster PCR process?

The number you give is more-or-less spec for a MiSeq flowcell. V1 chemistry would give you about 4 million clusters, by the way. Yep, that is right, your yield is 0.2%.

So are you saying that no one (not just you) should see any impurities, and that failed runs or misloaded libraries are primarily due to genomic content? I'm just asking because I do not have the experience that most people here would have...

???

First I was not dealing with edge cases at all -- just what is most likely for a canonical TruSeq DNA library.

Second, "genomic content?" We were discussing TruSeq libraries (or Bioo libraries) and the effect of enrichment PCR on the efficiency of clustering of a library. You opined that the 100x lower clustering density we saw with 2 non-amplified TruSeq libraries might be due to adapters out competing library molecules for binding sites on the flow cell.

I bristled at this interpretation, not just because it seemed an unlikely possibility, but because it is one that seems to be frequently invoked as a possible reason for low cluster densities. Even though it makes no physical sense nor have I seen empirical data to support it as a typical issue.

Anyway, if you would like to cluster a MiSeq flow cell with an unamplified TruSeq library, I would be eager to hear your results. I would suggest at least doing qPCR to quantitate the library, rather than fluorimetry, etc. Because there is the (secondary) reason you would come in low that we all seem to agree about: only a portion of the unamplified molecules will be "true" (fully adaptered) amplicons.

--
Phillip

[ECO]

We run BIOO-ligated amplicon libraries without PCR enrichment all the time with seemingly identical clustering efficiency to amplified libraries.

We do use qPCR to quantitate however.