Pooling RNA samples for differential display (DD)?

Dom Spinella dspinella at chugaibio.com
Tue Jun 2 17:03:52 EST 1998

> We are trying to display transcripts differentially expressed between
> tumor and normal tissue. Rather than try perform a DD reaction with
> each primer set (240 combinations) for each tumor sample, I thought it
> would be better to perform DD on pools of tumor RNA (with the thought
> that individual differences would be reduced, thus enhancing the display
> of transcripts differentially expressed in the tumor population and
> reducing the number of DD reactions that would have to be performed). I
> asked a company that specializes is DD and they did not like the idea of
> pooling the RNAs (but, without an explanation).
> So, my questions is . . . what do you think? Is pooling the tumor (and
> "normal") RNAs a bad idea or not? Any input would be appreciated.
> Randy

I think it's a bad idea too, for the following reason: your idea makes
the tacit assumption that the relevant genes are going to be the same in
every tumor. That's not necessarily the case.  If, for example, tumor
sample 1  shows up-regulation of oncogene A, while tumor 2 shows
down-regulation of tumor supresor B, you'll miss that fact by pooling
the samples.  While I do think it important to use a variety of tumors
and not restrict DD analysis to pairwise comparisons, I much prefer
running reactions derived from individual tumor samples in adjacent
lanes rather than pooling the RNA. (Are you not interested in a gene
that is up-regulated in 4/5 tumors -- which you would likely miss by
pooling RNA from all 5?) 

Now to a bit of a diatribe.  IMHO, derived from looking at global gene
expression proifiles of a lot of tissues and tunmors, DD is frought with
danger.  First of all, what constitutes the "tumor" from which you
extract RNA?  Is it a mass of flesh removed by some surgeon?  Think
about all the kinds of tissue in suce a mass (epithelia, vascular,
endothelia, parenchymal, blood cells, etc.) to say nothing about the
relative mix of normal and transformed cells in the mass itself. 
Second, there's a bit of a Heisenberg principle at work here.  After
all, the first thing a surgeon does when removing the tumor is clamp off
the relevant blood vessel.  That results in a decrease in temperature
and oxygen -- both of which turn on many genes and turn off others (heat
shock genes for example).  And of course tumors undergo many genetic
changes that may be only indirectly related to the transformation -- for
example, turning on of cyclin genes in a mitotically active cells that
was hitherto terminally differentiated. Then of course there's allelic
polymorphism between tissues from unrelated patients.  The end result of
all this in my experience is that, a fair proportion of ALL the genes
(as much as 10% or more) will be apparently differentially expressed
when such tissues are compared.  You can discover such genes by any
method you like -- including fishing out cDNAs at random (where 10% or
so will differ on Northern analysis on the parent tissues).  The point
is, are they really relevant to the oncogenic event or are they just
secondary changes or artifacts?  It isn't for nothing that DD is
sometimes referred to as "Differential Dismay" and the real key is
choosing the samples to be analyzed very wisely and carefully.  Anyway, 
enough ranting -- good luck!
--Dom Spinella

More information about the Methods mailing list