Inconstant Cluster Density on HiSeq 2000

[Miaranne]

Since a couple of months our HiSeq 2000 shows inconstant Cluster Density.
We started having troubles when some library pools (we multiplex our libraries in pools and then run each pool on multiple lanes) showed very low cluster density (generally around 200 K/mm2), and then when we re-run the exact same pools (same concentration, same mix, same tube) on another flowcell we obtain optimal density (up to 800 to 1000 K/mm2). We have troubleshoot this problem extensively with Illumina Tech Support but we were not able to point out what the problem is.

We use the HiSeq 2000 machine since 2 years with good result and we have a robust protocol for library quantification and QC. We do Paired End Sequencing (2 x 200 cycles). The problem is always "pool dependant" and when the libraries are re-run it usually gives good results, so we tend to think the problem might be somewhere between the libraries denaturation and the cBot. We've tried to do a re-hyb on a Flowcell showing poor density at the First Base Report, but the density remains the same afterwards. Only repeating the cBot completely with a new flowcell seems to work. We have consider a flowcell batch problem, but it was ruled out since two flowcells of the same batch did not react the same way (one was perfect and the other showed density problems). The HiSeq dont seem to be part of the problem either, we've run many tests.

For an example, see in attachment the density of 2 runs, where the first was problematic for Pool E (first 4 lanes, 200 K/mm2) and the re-run was good (around 800 K/mm2). The re-run was done using the same tube of pooled libraries.

Any ideas? Anybody experience something similar? Any input would be greatly appreciated, we are running out of explanations...

Many thanks!

[GenoMax]

Has an illumina engineer (not tech support folks) actually diagnosed your instrument? Did it pass validation?

[Miaranne]

Hi GenoMax,
Thanks for your reply. Yes an Illumina engineer looked at the HiSeq and cBot. Nothing suspicious was found, all seems in order according to him.

[GenoMax]

How many flowcells have been run since this started and what fraction of them were affected by the problem?

Best recourse is to keep complaining to your local FAS and illumina tech support. I assume you have a maintenance contract and this would certainly qualify for attention.

[Miaranne]

Since the beginning of the problem we ran 8 flowcells. On those, 5 flowcells had between 4 and 6 lanes that needed to be re-run (clusters generally around 200 K/mm2). Whats weird is that there is always some lanes that have good clustering (800-1000 K/mm2), and often the samples in the "good lanes" are samples from the same projects than the lanes that failed, just not the same pool.

We have a service contract with Illumina of course, but the fact is that they tested the machines and found nothing, furthermore the failed lanes are "pool dependant" and it is an "on and off" problem (it seems to happen randomly on some pools and not others), so they are as clueless than us right now and they are not able to help us further.

[GenoMax]

This almost feels like (not that I would know more than an Illumina engineer) there is a failing manifold/valve/partly-blocked tubing somewhere that can be blamed.

When we had recurring issues with a problem instrument some one finally had to come out from California. The instrument got fixed in the end.

[morellr]

We had similar problems with inconsistent clustering densities on repeats of the same library pools. We determined that it was the denaturation step that was the source of variation. Making sure that the NaOH was fresh, correctly diluted, and the incubation time was exactly 5 minutes for each sample, seems to have fixed most of the problem.

[Miaranne]

Hi morellr,
Thanks for the input. We'll be particularly carefull at the denaturation step. We've been aliquoting the NaOH 2N and keeping it in the freezer until we use it. We dilute it at 0,1N just before using. Are you freezing the NaOH 2N or do you keep it at room temp? What concentration of NaOH do you use before dilution? Do you use the NaOH 2N leftovers of the SBS kits?

Thanks Genomax for your help, it's really appreciated. The manifold or tubing of the cBot have been ruled out since the volumes of reagents and template after the flowcells hybridation run are all equal. We would see a variation if there was a blocage.

[morellr]

Hi Miaranne,

We use the 2 N NaOH supplied with the kits. Just before use we dilute to 0.1 N with MilliQ water and then use 1:1 with the 2 nM DNA sample, as per Illumina protocol. Is your "0,2 N" a typo?

[vlee2]

Have you checked the coolant level in cBot?

[Miaranne]

Hi everyone! It's been a while but I thought I should write to tell you the end of the story! Finally we are using 10N NaOH (purchased separately) to dilute at 0.1N before each cBot denaturation. We didn't have any more clustering problem since we stopped using the 2N NaOH provided in the Illumina kits. It seems the denaturation step is really critical.
Cheers!

[DJ Flowcell]

Hi Miaranne,
I just happened upon this thread today and before reaching the bottom conclusion I was going to recommend the same exact thing. We found very similar results to your inconsistent clustering problem and it took quite a while for us to pinpoint it to the NaOH. We also use 10N bought from an external supplier- Illumina admitted that they've had inconsistencies with their 2N supplied NaOH.

The way we pinpointed it was by analyzing the data from the poor clustering pools. They happened to be Caulobacter (very high GC content) and the data was proving otherwise, which was suspect. We determined that the high GC libraries weren't being denatured properly. Sure enough, this was the cause.