3700 sequencing problems

Phillip San Miguel pmiguel at purdue.edu
Tue Dec 23 16:54:34 EST 2003

Hall, L.S. wrote:

>	[...]
>	I am currently faced with what should be an 'easy fix' problem but in reality has been dragging on for months: Specifically, we possess a 3700 and of late, samples that sequenced very well have started to fail or produce top heavy sequence with alot of noise; What is even more curious is the fact that all samples produce better sequence in terms of balanced signal and superior signal to noise if the same samples on the same plate are rerun. Nevertheless, in nearly all cases even those sequences that second time round produce raw data demonstrating an optimal signal strength and a good spread in terms of 5' -3' signal demonstrate background contamination, invariably amorphous red 'noise' under sharp blue specific peaks. On face value and to my mind this might suggest the following:
>	1. The consistent/persisitent presence of one color under another may be a matrix artifact, necessitating a new spectral assay.
>	2. The top heavy peaks and overall weak data might be due to some sort of salt contamination in the sequencing preparations; Further, augmented signal from reruns of the same sample could be due to any salt present being depleted or eradicated during the initial injection. Consequently, when the same 'desalted' sample is reinjected by rerunning the same plate, extension products in general and 3' products in particular are more efficiently injected because salt is absent or considerable reduced.
    While I have no real answer to your problem, note that a sample that
has been run once by a 3700 is in no way 'desalted'. I repeat: its salt
composition should not change. This is because a volume of your sample
(2.5 ul by default) is taken from your plate on the plate deck and
transfered to loading dimples for electrokinetic injection. Hence the
sample which is left behind would not be altered. This is contrary to
the situation with all other machines which use electrokinetic injection
of which I'm aware (3100, 3730, megabace).
    But this is not to say that there would be no change to your sample
at all--three spring to mind:

1)   You sample will be exposed to air--hence subject to oxidation. With
version 1.0 Big Dye chemistry I saw loss of "G" peaks over time--I think
because of oxidation. Probably not important in your case.

2)   You sample will be exposed to air and may evaporate. If it contains
volatiles, such as ethanol or chloroform, these would be substantially
reduced in concentration in your sample. Were your samples contaminated
with ethanol, perhaps the time they are exposed to the air reduces said
ethanol contamination and ameliorates the effect you see.

3)   Diffusion and dissolution will occur (whether the samples are open
to air or not.) I have only just realized what an important effect this
is for us. That is, if your samples are not mixed, then over time they
will mix via diffusion. If your samples are to some extent pellets, they
will dissolve.

    Recently we have realized the importance of this last effect. We do
reactions in 384-well plates and clean-up the terminators by doing an
ethanol precipitation. After washing a couple of times with 70% ethanol
and removing any remaining ethanol with an inverted spin, we add 15 ul
of ddH2O, resuspend by shaking for 5 minutes or so. Spin down and load
onto the plate deck of the 3700. To test side-by-side reactions on our
3700 and our new 3730XL, we resuspended in 25 ul instead and transfered
10 ul new plates to be run on our 3730XL. To our surprise the signal was
very weak--yet the 3730XL should be more sensitive to than the 3700.
    It was either a post or an email from Paul Shinn I saw long ago
defending the "denaturation" of sequencing reaction products after
cycle-sequencing/ ethanol precipitation. I thought it not necessary
because the product concentration would vastly outweigh the template
concentration--hence loss of product to double-stranded forms seemed
unlikely to be important. Paul noted that he heated his samples to
promote there resuspending. So years later I saw demonstrated the
importance of this step for rapidly disolving reaction products after
precipitation. During the intervening time I'd completely ignored Paul's
advice on this issue. How? Because, I guess, the samples have about an
hour or more to dissolve in the plate deck of the 3700 before they are
    Anyway I doubt I would have had that insight with your comment long
ago, Paul (if you still read this newsgroup.)
    Good luck with your problem, Laurence Hall. Perhaps some of this
will be of help to you...

Phillip SanMiguel
Purdue Genomics Core Facility


More information about the Autoseq mailing list