MiSeq cluster generation problems

[skosuri]

Hi all,

I'm running a library of synthetic amplicons (~175-250bp), and reading it out with custom primers on the MiSeq. I'm having a lot of problems getting enough cluster density. My last MiSeq run, I combined the custom sequencing primer with the Illumina Read 1 mix, and ran the mixed 20% 8pM phiX with with 80% 20pM library. I got ~1.5 million phiX reads and ~500K of the custom reads, with at total cluster density on ~225K/mm2. The custom reads looked fine and were what I expected of my library. I reran the Kapa QPCR quantitation again on my dilutions against phiX and it looked pretty accurate.

I was wondering what the concentration range of getting good cluster density is. Would it make sense to run 10X (200pM) of the library to get 10X the number of clusters? This seems ludicrously high. Does it something to do with the length of my library? Has anyone seen such a drastic difference between Kapa Library quantification and cluster density? I'm a little lost and want to figure stuff out without burning too many flowcells.

Thanks,
Sri

[ECO]

I would guess it's your molecule lengths, if there is nothing else wrong with the library (see other discussions in the 454 forum about the presence of primer-dimer throwing off qPCR calcs), and your adapter design is sound. In my very limited experience, clustering efficiency drops off dramatically with molecules <250 and >800. Clustering was much less efficient for 100-120bp amplicons (which end up at ~220 after ligation) than it was for 150-170bp amplicons (~290 after ligation). I'd check for a correlation between #reads and length of amplicon...

To answer your second point, I have never had to go above 11pM with a library pool to get very high cluster densities (>1100-1300). This is with libraries of amplicons (amplicons being 150-170bp before adapters), shotgun libraries, and enriched libraries (both with insert sizes >200bp before adapters).

Did you denature in a way that would result in higher NaOH concs than recommended? This kills cluster density as well.

[skosuri]

I would guess it's your molecule lengths, if there is nothing else wrong with the library (see other discussions in the 454 forum about the presence of primer-dimer throwing off qPCR calcs), and your adapter design is sound.

The reads that I do get look good. I don't see many primer dimer reads; everything maps to my library or phiX. The index reads look fine as well. The sequencing results also make me think that the adaptor design works.

Did you denature in a way that would result in higher NaOH concs than recommended? This kills cluster density as well.

Interesting, but I use the recommended NaOH (0.1M) concentrations.

As for the lengths, I sent a similarly sized library for HiSeq sequencing, and the data looked great (300 million good reads in a lane).

[ECO]

Originally Posted by skosuri

Interesting, but I use the recommended NaOH (0.1M) concentrations.

Does more than 1mM end up hitting the flow cell?

[skosuri]

Does more than 1mM end up hitting the flow cell?

Well it's about 1mM. 10uL in 1mL of 0.1M; this gets diluted slightly, but not much. Though those are the recommendations based on the MiSeq sample prep guide. Plus, the cluster density of the phiX is not affected.

[ECO]

Hmmm...reload that sucker at a higher concentration and see!

I would really love ILMN to facilitate a cheaper/faster way of checking cluster density...

[ITseq]

Is there heterogeneity for the first 6 bases of your amplicon after your custom sequence primer?

[skosuri]

After several failed runs in a row (even loading at 50pM), I figured out my problem (with the help of very patient illumina tech support representatives ); so I thought I would update.

I had been using the old P5 and P7 sequences (single end versions, Nature 2008 paper; bentley et al; supp info) for the adaptors I PCR'd on. Apparently these changed for the paired end versions (same sequence with a few extra bases on the 3' end; see the nature 2008 paper).

So while my libraries worked for past GAIIx runs (because both primers were on the flow cells), at some point this changed and the MiSeq flow cells no longer contain these shorter P5/P7 sequences. Thus any clusters I was getting were due to very low levels of mispriming off the forward primer (reverse primer was fine since I am using standard multiplexing primers). Though those misprimed clusters would sequence fine.

Since all the quantification kits have the old P5/P7 sequences, all my RT-PCR titrations looked fine; it was just that the extra bases inhibited most cluster formation of my library.

I reordered the primers now, so I'll re-run everything again, but this explained all my previous problems.

Thanks everyone,
Sri

[LSAD]

Originally Posted by skosuri

After several failed runs in a row (even loading at 50pM), I figured out my problem (with the help of very patient illumina tech support representatives ); so I thought I would update.

I had been using the old P5 and P7 sequences (single end versions, Nature 2008 paper; bentley et al; supp info) for the adaptors I PCR'd on. Apparently these changed for the paired end versions (same sequence with a few extra bases on the 3' end; see the nature 2008 paper).

So while my libraries worked for past GAIIx runs (because both primers were on the flow cells), at some point this changed and the MiSeq flow cells no longer contain these shorter P5/P7 sequences. Thus any clusters I was getting were due to very low levels of mispriming off the forward primer (reverse primer was fine since I am using standard multiplexing primers). Though those misprimed clusters would sequence fine.

Since all the quantification kits have the old P5/P7 sequences, all my RT-PCR titrations looked fine; it was just that the extra bases inhibited most cluster formation of my library.

I reordered the primers now, so I'll re-run everything again, but this explained all my previous problems.

Thanks everyone,
Sri

I am interested to know whether the new P5/P7 solved your previous problem, because I have been experiencing the same problem. I added 6nt on old P5 and 2nt on old P7 based on Nature 2008 paper. I re-run everything again, but I still don't have enough clusters. It might be because that my adaptor part is different from TruSeq library and I used a different indexing strategy.

[skosuri]

Originally Posted by LSAD

I am interested to know whether the new P5/P7 solved your previous problem, because I have been experiencing the same problem. I added 6nt on old P5 and 2nt on old P7 based on Nature 2008 paper. I re-run everything again, but I still don't have enough clusters. It might be because that my adaptor part is different from TruSeq library and I used a different indexing strategy.

They did solve my problems. I got 6 million or so reads on my last run. I am also using custom primers (on the forward read), my own indices, and not using the Truseq sequences.

More specifics might be helpful to understanding what might be going on. How many reads are you seeing and how much are you loading? Are you using custom sequencing primers? What exactly is the indexing strategy you are using?

[LSAD]

Originally Posted by skosuri

They did solve my problems. I got 6 million or so reads on my last run. I am also using custom primers (on the forward read), my own indices, and not using the Truseq sequences.

More specifics might be helpful to understanding what might be going on. How many reads are you seeing and how much are you loading? Are you using custom sequencing primers? What exactly is the indexing strategy you are using?

We thought that was the only problem for us but now it seems not. It is great to know that you solved your problem, so that I know I am on the right track. It seems that after adding the 6 nt onto old P5 and P7, we do have cluster formation on MiSeq flowcell. However the fluorescence intensity of base calling is low and all sequences we’ve got are A due to this low intensity. I guess something might be wrong for my sequencing primer.
What sequencing primer do you use?
From 5’ to 3’, my design are P5, 50bp amplicon, 20bp adapter, 6bp index and P7. I am using P5 sequences for Read1 sequencing primer, and a sequence complementary to adapter as index sequencing primer.
Basically what I did was diluting my sequencing primers in HT1 buffer to a final concentration of 0.5uM and loading Read1 sequencing primer to the customized position 18 and index sequencing primer to the position on the reagent cartridge. Do you have any clue about my problem?

[skosuri]

Originally Posted by LSAD

We thought that was the only problem for us but now it seems not. It is great to know that you solved your problem, so that I know I am on the right track. It seems that after adding the 6 nt onto old P5 and P7, we do have cluster formation on MiSeq flowcell. However the fluorescence intensity of base calling is low and all sequences we’ve got are A due to this low intensity. I guess something might be wrong for my sequencing primer.
What sequencing primer do you use?
From 5’ to 3’, my design are P5, 50bp amplicon, 20bp adapter, 6bp index and P7. I am using P5 sequences for Read1 sequencing primer, and a sequence complementary to adapter as index sequencing primer.
Basically what I did was diluting my sequencing primers in HT1 buffer to a final concentration of 0.5uM and loading Read1 sequencing primer to the customized position 18 and index sequencing primer to the position on the reagent cartridge. Do you have any clue about my problem?

A few things.
1. I use a custom sequencing primer for read 1. I use the following sequence: GAAGCACAGCAGCTCTTCGCCTTTACGCATATG for my sequencing primer. It's closely matched in Tm to the standard Illumina primer; your sequences seem like they would have pretty low Tm comparitively. Also, I'm not sure if you can actually use the P5 sequence for sequencing.
2. I use a final concentration of 0.5µM. One thing that did help a lot is I actually spike in my sequencing primer into the Read 1 Primer Mix (I think it's position 12 off the top of my head). I do this so that I can spike in PhiX into the library; it really helps diagnose problems.
3. Another thing that really helped me was checking out the thumbnail images of the flow cell output on every run; it really helped me understand that I was having a serious clustergen problem.

Sri

[LSAD]

Originally Posted by skosuri

A few things.
1. I use a custom sequencing primer for read 1. I use the following sequence: GAAGCACAGCAGCTCTTCGCCTTTACGCATATG for my sequencing primer. It's closely matched in Tm to the standard Illumina primer; your sequences seem like they would have pretty low Tm comparitively. Also, I'm not sure if you can actually use the P5 sequence for sequencing.
2. I use a final concentration of 0.5µM. One thing that did help a lot is I actually spike in my sequencing primer into the Read 1 Primer Mix (I think it's position 12 off the top of my head). I do this so that I can spike in PhiX into the library; it really helps diagnose problems.
3. Another thing that really helped me was checking out the thumbnail images of the flow cell output on every run; it really helped me understand that I was having a serious clustergen problem.

Sri

Thank you a lot for your reply. The Tm problem is highly probable. We have been using this home-brewed library and P5 sequence for sequencing on HiSeq single run kit with no problem. Do you think there is any reason I should not use P5 on MiSeq except the Tm issue?

I used PhiX before. When I use only 10% PhiX spike in, I only got 10% total cluster formation. When I don't use PhiX, at least I got cluster formation although low intensity. I thought somehow PhiX might not be compatible with my design. But now maybe I should reconsider using PhiX and use sequencing primer with higher Tm.

I also noticed that the Tm of Illumina's Index Read Sequencing primer is different from illumina's Read 1 primer. Do you also use the primer match the Tm of Index Read primer for index sequencing?

Based on your experience, can I assume that the new P5/P7 is the only elements need to be identically the same as Illumina's design for cluster formation on MiSeq?

[Yepler]

Thanks, skosuri, for posting your solution. I have been having the same problem...(homebrew) library looked fine on previous HiSeq runs, but wouldn't form clusters on a MiSeq. Looks like I have the same problem with my P5 sequence being the shorter version. At least it's only one oligo to order, but it's been a pretty frustrating week trying to figure out why something suddenly wouldn't sequence!

Deb/Tucson,AZ

[LSAD]

Originally Posted by Yepler

Thanks, skosuri, for posting your solution. I have been having the same problem...(homebrew) library looked fine on previous HiSeq runs, but wouldn't form clusters on a MiSeq. Looks like I have the same problem with my P5 sequence being the shorter version. At least it's only one oligo to order, but it's been a pretty frustrating week trying to figure out why something suddenly wouldn't sequence!

Deb/Tucson,AZ

Yepler,

Except the few nucleotides mentioned earlier by skosuri, another reason of failed run for us is that we used sequencing primer targeting complementary to P5. We tried a different sequencing primer with higher Tm covering P5 and a few more nucleotide but it still doesn't work. It seems that it is not only Tm issue but based on our experience, you should not use primer that is complimentary to P5. Now we solved the problem by using a different sequencing primer that is complimentary to other region of the constructed library but not P5 region. Good luck.

[Yepler]

Just wanted to update (in case this helps someone else) and say that changing to the longer P5 sequence solved the problem. MiSeq returned an avalanche of good reads!

[jacksor1]

Could you post the Illumina adpater sequences you had success with? I am having issues with cluster generation on our MiSeq usng PCR primers with adapters. I have found a number of references and they all seem to have different adapter sequences.
Thank you so much,
jacksor1

Originally Posted by skosuri

After several failed runs in a row (even loading at 50pM), I figured out my problem (with the help of very patient illumina tech support representatives ); so I thought I would update.

I had been using the old P5 and P7 sequences (single end versions, Nature 2008 paper; bentley et al; supp info) for the adaptors I PCR'd on. Apparently these changed for the paired end versions (same sequence with a few extra bases on the 3' end; see the nature 2008 paper).

So while my libraries worked for past GAIIx runs (because both primers were on the flow cells), at some point this changed and the MiSeq flow cells no longer contain these shorter P5/P7 sequences. Thus any clusters I was getting were due to very low levels of mispriming off the forward primer (reverse primer was fine since I am using standard multiplexing primers). Though those misprimed clusters would sequence fine.

Since all the quantification kits have the old P5/P7 sequences, all my RT-PCR titrations looked fine; it was just that the extra bases inhibited most cluster formation of my library.

I reordered the primers now, so I'll re-run everything again, but this explained all my previous problems.

Thanks everyone,
Sri

[Yepler]

Take a look at the supplemental info here:

David Bentley in Nature (2008) 456, 53-59 Accurate whole human genome sequencing using reversible terminator chemistry (supplementary material: http://www.nature.com/nature/journal...re07517-s1.pdf)

the adapter sequences I used were the ones described for the the paired-end flow cell. The ones used on the single-end flow cell were the ones I was previously using (and had failures with).

Hope that helps.

[nbogard]

Originally Posted by skosuri

Thus any clusters I was getting were due to very low levels of mispriming off the forward primer (reverse primer was fine since I am using standard multiplexing primers). Though those misprimed clusters would sequence fine.

Can you please elaborate on this? I am having similar problems with cluster generation on MiSeq but expect that my new libraries with PE adapters will work fine. My concern is whether or not the target for the sequencing primer is also abnormal. I'm worried that two low density runs I have completed (using single-read adapters) consist of useless and misleading data because, in my analysis, I only use the first base immediately downstream from the Read 1 sequencing primer (i.e. where the insert meets the adapter). In my case reads generated beyond that first base would be artifacts.

Do you think the SR adapters only cause a problem with flowcell binding? Would you expect sequencing priming to be normal?

[TonyBrooks]

We're getting similar problems, compounded by low diversity.
Our library is an pool of three different amplicons with three different R1 sequencing primers. We added some N's in our amplicon generation primers to immediately follow R1 primer annealing site in order to increase diversity during cluster registration. Apparently, it wasn't enough as our run was terrible (160k/mm2 density - Q-scores along the floor). We're going to repeat with 5% PhiX and the Josh Quick/Nick Loman fix (http://pathogenomics.bham.ac.uk/blog...illumina-miseq)

The issue I have is that several things worry me when looking at the thumbnails and general SAV analysis

1) Density is clearly higher than 160k/mm2 when eyeballing the thumbnails (poor cluster registration due to low diversity?)
2) There are a mix of low and high intensity clusters that can't be explained by laser cross-talk. These low diversity clusters are appearing on all channels (ATCG) not just the alternate base from the respective laser which is a bit strange, no? You can clearly see the same general cluster pattern for all bases - as if there's cross talk between all bases with some clusters lighting up fairly bright.
3) Only one amplicon seemed to generate any data post indexing

I'm wondering if two of the three read primers have too low Tm and hence generating lower signal. These are being filtered out leading to the lower cluster density estimate.
Tm's of each of the R1 primers are ~66 according to OligoCalc basic. Maybe I need to add some bases on to the other oligos? How do people check their Tm's? MiSeq runs at 65, if I'm not mistaken.

Comments and theories welcome!