Difficulties with HiSeq for Hi-C libraries

[Marianna Zazhytska]

Dear all,

I'm creating Hi-C libraries according to the protocol from Rao et al. paper and all the time facing problems with sequencing. I'm able to generate libraries with high enrichment of hybrid inserts, but when I was going for HiSeq I've got really very low clusters density and a few reads of the library in the end. I should mention that for MiSeq run, I've done as a control, the situation was also not perfect - I was able to reach ~1Mb. I'm wondering whether anybody has facing the problem like this and can give an advice how to overcome it? 

Many thanks in advance,

Marianna

[Suhas Rao]

Hi Marianna,

I would recommend quantifying your libraries via qPCR before running them on the sequencer to check the concentration of actually clusterable molecules. It could be possible that some step in the library prep isn't working for you and thus you may have lots of DNA but without Illumina adapters attached. Also, you should check the size distribution of the library carefully, if your library has too many long fragments, those won't cluster efficiently either and that could lead to low cluster density. 

Hope that helps,

Suhas

[Marianna Zazhytska]

Dear Suhas,

Thanks a lot for quick reply. Now I have a bit naive question: since I'm checking the quality of library with MisSeq - isn't it a better quantification then with qPCR? Or you mean that Illumina adapters are of the same type on the both sides of insert and they are not able to produce clusters, meaning that I have to check and quantify adapters themselves? If it's the case, would you be so kind to give an idea which steps of library preparation it would be worth to check to get correct type of indexes on both ends of insert? I'm stuck with this problem for 4 months already and have no a single idea how to overcome this problem and believe me I tried to change a lot of things and still nothing is helping.

Many thanks,
Marianna

[Suhas Rao]

Hi Marianna,

How are you determining how much to load on the MiSeq? Presumably you quantify the concentration somehow? I'm suggesting that you do a qPCR then with the Illumina primers (i.e. one of the Kapa qpcr kits). This will tell you the concentration of clusterable molecules (i.e. adapters properly ligated on both ends of the fragment). You mentioned that you're having issues with the cluster density on the MiSeq run as well, which is why I suspected that your concentration calculations are off. Are you adjusting the amount that you load on the HiSeq based on the cluster density you get on the MiSeq and still getting low cluster density? How many cycles of PCR have you been doing to amplify your library? 

Are the problems you're having only for Hi-C libraries? The library prep in the in situ Hi-C protocol is very standard, so I'm not sure why you would be having sequencing problems only for Hi-C libraries. Have you made sure that your insert size distribution is fairly tight within the 300-500bp range? i.e. no very short or very long stuff? 

Best,

Suhas

[Marianna Zazhytska]

Hi Suhas,

Regarding your questions: I'm quantifying library concentration with Qubit and doing fragment analyzer. Taking measurements from Qubit and peak from FA as input, I'm calculating the nano molarity of library. As I'm doing it for a set of libraries I take the lowest concentration and then normalize all the rest of libraries according to it (the lowest concentration is 4.2 nM). Then dilutions and  loading on flow cell. Since I had issues with cluster densities on MiSeq, for HiSeq the concentration of libraries was adjusted according cluster formation on MiSeq. And yes, after adjusting I still have low cluster density (in the end I got 5% of what I was expected). 

To amplify my libraries I'm doing 8 cycles.

I completely agree that the library preparation protocol is very standard and even Illumina tech support has no clue what might be wrong. Before Hi-C experiments I have been doing ChIP-Seq libraries and everything was fine. The only thing that is different - library preparation in running on Dynabeads MyOne Streptavidin beads (I'm using C1, Life technologies), but I don't think that beads should influence somehow library preparation.

And regarding insert size: the average peak for my libraries is ~350bp.

Many thanks for all your comments,

Marianna 

[Suhas Rao]

Hi Marianna,

I would recommend measuring via qPCR in addition to measuring via Qubit. If there is a large discrepancy between the measurements, that would give you some indication as to whether you have many DNA molecules without proper adapters in your library.

I have to say though, I don't have any great ideas about why you are having clustering issues; seeing 5% of what you expect on the HiSeq after adjusting the concentrations based on a MiSeq run is particularly strange. Perhaps you have something in your sample inhibiting clustering? Are you doing the library PCR on the C1 beads? At one point, we had issues with a particular lot of C1 beads where the beads were breaking down during PCR and inhibiting the PCR reaction. We switched to T1 beads and haven't had problems since (and we started detaching DNA from the beads prior to PCR), but I think we've also occasionally used newer lots of C1 beads and not had issues. It's a bit of a long shot, but if you are doing the PCR off the beads, perhaps you could try detaching the DNA from the beads before PCR (98C for 10 minutes is what we use). Sorry I couldn't be of more help.

Best,

Suhas

[Marianna Zazhytska]

Hi Suhas,

I will definitely do qPCR to check whether I have proper adapters. 

I was thinking about what can inhibit clustering for ages, but I'm washing libraries after each step (i.e. end repair, adenylation, adapter ligation) and resuspend in RSB buffer from Illumina. And yes I', doing PCR of library on the C1 beads (maybe I should try with T1). Anyway, you have given me ideas for troubleshooting. I'll inform asap the results here.  

Many thanks,

Marianna

[Marianna Zazhytska]

Hi Suhas and all here!

I'm glad to inform that I've managed to get affordable number of reads from my Hi-C libraries. Indeed, final amplification without beads improved much cluster densities. Nevertheless  I didn't do qPCR instead I have done several spike-in tests with all the libraries. After them we have started to run HiSeq and I've just got the results from the first lane - 124M! Not perfect (you've got 150M from lane in average if I'm not mistaking) but it works!!!

I'll be happy to help if anybody has facing the same problems,

Marianna