New Miseq learning curve

[creeves]

We just started using a new MiSeq and are discovering that the devil can be in the details. We have done two runs, the first with four indexed yeast genomes and the second with three indexed yeast genomes. The libraries were prepared using the Nextera kit and quantified by picogreen. The reads in these runs contained a high number of Ns (no base call) that increased as read length got longer. Illumina says this should not happen. The cluster densities for these runs were 437 and 312 K/mm2, respectively. The distribution of reads between the pooled genomes was skewed away from the desired 25/25/25/25 and 33/33/33 for the two runs as well.

I have very carefully quantified the yeast genomic DNA preps with picogreen and then run 200 ng next to 200 ng of lambda DNA on a gel. There is certainly some variability in prep quality from the gel image. In particular, yeast DNA preps always have a significant fraction of the DNA that is trapped at the well. I don't know if this is a problem or not.

Is there anyone out there with tips for maximizing quantity and quality of sequence data from a MiSeq run with three indexed yeast genomes?

[frozenlyse]

Can you quantitate your libraries using qPCR with a standard curve made from a library of known cluster density?

[ECO]

Have you successfully made libraries from this DNA with other (traditional shear-ER-AT-Ligate) methods?

I would be wary of combining your first foray into MiSeq with Nextera on what sounds like questionable quality (or not clean!) DNA.

You definitely need to get qPCR quantitation...using bioanalyzer or picogreen is a recipe for improperly loaded flowcells. N's are (in my limited experience) a sign of reading off the end of fragments...meaning you probably have a high % of short fragments.

I'd love to see a FASTQC on your run(s) to help troubleshoot.

[docbio]

Accidentally posted twice, sorry about that.

[docbio]

Adapter trimming?

Incidentally, are the N's all appearing at the end of the reads? That is, do you have strings of NNNNNNN? If so, did you leave "Trim Adapter Sequences?" turned on when you made your sample sheet in the IEM software? In my experience, it is turned on by default, and it will replace the adapter sequence with N's in your reads if you have small inserts and have read into the adapter on the other end of the insert.

I had this exact experience the first time I tried the Nextera kit. The transposon is very active, and if you aren't quick about purifying it after the tagmentation step (or if your PCR thermal cycle hasn't been calibrated or is running a little "warm") you will wind up with a lot of really short inserts.

Since then I've switched to NexteraXT which seems to work better as they include a "stop solution" you add after tagmentation, and I imagine the enzyme concentration is dialed down. I also leave adapter trimming turned off so I can better assess the actual quality of my reads, then trim the adapter sequences out later.

[creeves]

New MiSeq user

OK, so we now know that the Ns are at the ends of fragments where the adapter sequence was masked out and that we have a high percentage of short fragments in our library. I thought the Ampure beads would remove the short fragments. Perhaps there are so many short fragments going into the Ampure clean up step that they still make it through. If this is the case, the tagmentation step may be suspect. Can anyone tell me what is the expected DNA concentration (by picogreen) for a good Nextera library made from 50 ng genomic DNA? I have been getting about 4 ng/ul. If that is low, do I perhaps need to increase the amount of gDNA at the tagmentation step?

[docbio]

I can't tell you the concentration you should expect but I feel like the libraries coming out of the Nextera are fairly dilute, so your concentration may not be off.

When I talked to tech support about the small insert issue they dissuaded me from troubleshooting the Nextera and pushed me towards the XT for my purposes. They did suggest several possibilities:

1) Allowing the tagmentation to sit at 10degC after running. They say the enzyme is still active at this temperature, just "less active", and they recommended immeadiately removing it from the cycler and continuing the protocol. This is how the XT protocol is written, and XT includes a stop solution. I have much much larger inserts in my XT library according to Bioanalyzer, very consistent sizing too.

2) Thermal cycler calibration being "off" - hotter cycler = more tagmentation was the implication.

Those were the only two explanations they really had. I asked about decreasing the tagmentation incubation time but they swore that the reaction chemistry was very complex and simply decreasing the time might cause unexpected results. YMMV. I did watch an Illumina webinar on microbial geomes by some folks from Australia, and they had the same problem with small inserts (150-200bp, I think) on one particular organism, which made me wonder if there is some influence of overall GC content on the performance of the reaction.

I only tried it once, the XT kit was released the same week so I moved to that. Surely some other people have done some troubleshooting.

[creeves]

Thanks for feedback

Thanks docbio - those were very helpful comments. The Nextera XT kits should help a lot. It would seem best to limit the amount of transposase relative to DNA so that the DNA would not be shredded even after the reaction went to completion. Seems fundamental unless I don't understand the reaction.

As to quality of the genomic preps, we suspect this is not the problem, because we have sent many yeast gDNA samples of similar and sometimes worse quality to Prognosys and still received good sequence data.

[pmiguel]

By the way, if you can run a bioanalyzer high sensitivity chip prior to the XT normalization step, it is much easier to troubleshoot issues of this sort.

--
Phillip

[creeves]

Bioanalyzer data

We finally got our software upgrade for the 2100 working and could analyze our MiSeq libraries on the high sensitivity chips. The results are puzzling because the distribution is skewed to larger fragments compared to the distribution shown in the Nextera protocol book. Instead of an average of approximately 450 bp as shown in the protocol book, the average size of our libraries ranged between about 700 to over 1000 bp. This could explain the lower cluster densities we got, but why are there so many reads that go through the insert and into the adapter sequence? Such reads imply a length of about 250 bp, but the Bioanalyzer shows baseline fluorescence below about 300 bp. Any thoughts?

[koadman]

Hi creeves,
In our lab's experience using the original Epicentre kits (e.g. not Nextera 2.0, not Nextera XT), there is still a large population of <200nt fragments remaining after an Ampure 0.7x cleanup. Our solution has been to use a gel cut at 300-600nt for hiseq, bigger for the miseq. The smaller fragments are somewhat enriched for favored Tn5 target sites, eliminating them reduces this bias and evens out the coverage to some extent in our hands. YMMV! If you are doing de novo genome assembly I would be wary of increasing DNA input without also increasing enzyme unless Epillumina has done some work to reduce the tagmentation site bias in the newer kits (does anyone know?). On the other hand if you just want a handful of tagmentation-site associated SNPs (think RAD-seq) then increasing DNA input should be just fine.

[Keritiger]

Having used both Epicentre and Illumina Nextera kits I can definitely say that the library concentrations output with the Illumina Nextera kit is much lower than the earlier kits. Speaking with their tech support (Illumina and Epicentre together) we learned poor library concentrations are common to the Illumina Nextera preps and to boost your DNA library concentrations by running more PCR cycles. The original Epicentre kit called for 9 cycles of PCR while Illumina changed to protocol to 5 cycles. Subsequently, low concentrations are apparently an issue.

If you do run some of your libraries to check them first on a high-sensitivity bioanalyzer chip and see that the small fragments are indeed still an issue I would also recommend doing an AmpureXP purification twice (but only if you added those PCR cycles). That's something that has helped me clean up small fragments from other libraries before and is recommended by Roche for cleaning up small fragments in library preparation (works the same for Illumina preps too though!)

I also just wanted to add that we also get different insert sizes for different bacterial genomes even if they're all prepped at the same time, so your insert sizes could vary just based on the genome your using. All our libraries had larger inserts than those featured in the Illumina/Australia seminar with most metagenomic DNA being ~450bp insert and we had 700bp inserts for some pure bacterial genomic preps.

[ScottC]

We've also seen this type of thing happen with different bacterial gDNA preps. We have had sizes ranging from 2kb down to 200bp. The larger peaks can be reduced by using lower quantities of starting material with the XT kits (we've tested the same prep at 1.0 - 0.2ng in 0.2ng increments).

[ScottC]

Instead of an average of approximately 450 bp as shown in the protocol book, the average size of our libraries ranged between about 700 to over 1000 bp. This could explain the lower cluster densities we got, but why are there so many reads that go through the insert and into the adapter sequence?

We've also seen this type of thing happen with different bacterial gDNA preps. We have had libraries with peaks at 2kb, 1kb, etc, down to 200bp, when all preps were done at the same time in the same tray.

In our experience the broad 700-1kb type 'peak' on the bioanalyzer is indicative of too high a starting concentration. This is especially true with the XT kits, which are very sensitive to starting DNA quantity. The larger peaks can be reduced by using lower quantities of starting material with the XT kits - we've tested the same prep at 1.0ng to 0.2ng in 0.2ng increments. You need to dilute your samples very carefully, measure very carefully and pipette very carefully. It all sounds like very basic information, but it makes a lot of difference for these preps (with TruSeq, you had a huge range in which to work).

Scott.

[koadman]

@ScottC, did you sequence anything from the lower input preps, e.g. 0.2-0.5ng or just bioanalyze it? If sequenced how did the library complexity look? Did you have many PCR duplicates? How many PCR cycles did you have to run to get a sequencable library from 0.2ng?

We are thinking to try Nextera XT for ultralow inputs. We had some luck with the Epicentre Nextera at 500pg input but library complexity was too low for comfort at 50pg. I'm wondering if XT might give improved results...