How small is too small for TruSeq RNA?

[eab]

What is the minimum quantity of total RNA below which people would not attempt the standard TruSeq mRNA library prep protocol, but would instead use DSN normalization, or an amplification method like SMARTer or the NuGen kit? The current protocol recommends at least 100 ng total RNA, and I've had success making libraries with even less. Is there reason (either theory or data) to believe that less RNA won't work, and if so, why won't it work? If one generates a library using 10 ng total RNA, sees a nice peak on the BioA that quantifies nicely by qPCR, is there some reason the doubt the sequencing results produced from that library?

Any opinions would be very welcome. Thanks!

[HESmith]

At some point, the representation of your library will decline and the fraction of reads that are PCR replicates will increase. Even with the SMARTer system, the number of genes detected drops by 50% as the input RNA is reduced from 10ng to 0.01ng.

[pmiguel]

10 ng of total RNA is ~100 pg of messenger RNA. Might not seem like much, but that works out to about 200 million 1 kb messenger RNAs. Plenty, right?

So the problem can't be the limited amount of RNA per se. I think one issue is the binding capacity of laboratory plasticware. How much RNA or DNA does a typical microfuge tube or pippette tip bind? Probably not nanograms -- but it may well bind picograms.

There are work-arounds to limit the effects of this issue, but it still means you are in another world than microgram to nanogram world you are used to.

--
Phillip

[eab]

Hi there, thanks very much for your reply. If I'm understanding you right, the loss of representation should occur gradually with declining input amount, with the least abundant species lost first. So there is probably not a single point at which the library becomes "invalid." It just progressively fails to represent the actual complexity of the sample.

Therefore, if one were to use mRNA seq to compare profiles of two cell populations, and one of these was very limited in size, then it would be difficult to compare accurately the expression of lower-abundance species between the samples. Higher-abundance species might be comparable, though, at least until you got down to very very low cell numbers, right? Are unpredictable PCR threshold effects responsible for this, do you think?

If you were to do the experiment described above, what method would you use to make the library from the smaller cell sample, and at what cell number would you be comfortable with using the standard, non-amplified method?

[eab]

Hi Phillip, thanks a lot for your reply. I've had the same question about binding to the plastic. In theory, I guess the best library would be totally PCR-free - who cares about having more clusters, if they dataset they yield is at best what you would have gotten without PCR, just multiplied several fold? It must be that you would lose all your material if you started with a tiny amount and didn't amplify. Do you (or does anyone else) know of any data that addresses the cell numbers where this starts to be noticeable?

Also, can you tell me more about the work-arounds that you mention?

[pmiguel]

Hi eab,

Work-arounds: low bind plastic tips and tubes. Adding Tween (where reasonable--usually for final libraries).

I agree with you about the PCR. Except, PCR just seems to makes everything easier...

One irritating thing is that the TruSeq protocol seems to be designed based on a no-PCR published protocol. PCR is tacked on the end. Which is fine--but no guidance is give for not using PCR!

--
Phillip

[eab]

How did you know the TruSeq protocol was based on a no-PCR protocol? I didn't pick up on that....

Besides the work-arounds, how would you prep mRNA libraries from 10 ng total RNA input?

Thanks again,
Eli

[pmiguel]

How did you know the TruSeq protocol was based on a no-PCR protocol? I didn't pick up on that....

I read:

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2664327/

Amplification-free Illumina sequencing-library prep

and it looks very similar to the TruSeq protocol.

Besides the work-arounds, how would you prep mRNA libraries from 10 ng total RNA input?

Thanks again,
Eli

First, I would try to avoid it if possible. Attempting to construct a library from 10 ng of total RNA is venturing off the straight and narrow protocol-wise. Given that even those following the standard protocol paths are often having a hard time at it, then realistically, you are asking for trouble if you make major changes to the protocol.

Second, if I decided I really must go down this dark path, I would go with one of the commercial kits for constructing libraries from low amounts of RNA.

Third, I would probably spend a year trying to sift useful data out of amplification artifacts and regretting the entire venture.

Here are some artifacts I might run into:

(1) Contamination. It just takes a small amount of material from another source to overtake my limited sample as the subject of my sequencing run. Starting with 1ug of total RNA, minor contaminants are swamped out by sample library. Not necessarily so if I start 2 orders of magnitude less sample.

Any enzymes you are using for library construction become a source of contamination. Why? The companies selling them probably don't QC them under conditions of low sample input. For example, back in the day, we used Invitrogen PCR4TOPO cloning kits to construct shotgun libraries for Sanger sequencing. But it was a PCR product cloning kit--add 100 trillion molecules of your PCR product and there was no issue. Add 1 trillion molecules of sheared BAC DNA -- still good. But add a billion or less molecules of low CoT fractionated genomic DNA and suddenly 10% of your reads derived from the entire catalog of Invitrogen vectors. Vectors you had never heard of, never mind used, in your lab.

(2) Bottlenecks in product amounts in your protocols will result in a limited pool of your sample molecules being PCR amplified into the majority of your sample. You started with 20 billion molecules of RNA (if the molecules have an average length of 1000 nucleotides). Likely no more than 1 billion derive from message. In the process of ribo-depleting your sample, you will lose most of them.

Let's say you are having a good day and you recover 200 million of them. You are using low-bind everything. The stench of silconization hangs heavy in your lab. Now you need to traverse all the basic steps of library construction (reverse transcription, fragmentation, maybe end repair, and adapter ligation), and you may well be working blind. Actually pico RNA and high sensitivity DNA bioanalyzer chips may shed a little light into your benighted world. Or not...

Anyway, each of those steps takes its toll. And the toll is going to be steep. Because the protocol is designed for working with 100 times, or more, the sample molecules you are providing. Worse, because the protocol is scaled for using vastly more starting material, the molar ratios of sample to library construction reagents will be crazily skewed. Yes, the protocol design logic imparted by some beneficent kit designer to gently carry a novice user along a royal path through the mine field is now working against you. Strange edge-effects that you would have been unable to even detect using the kit within normal parameters, become the rule.

(3) Bias. Not all of it derives from PCR. Fire up a new amplification method and you may be the first in your institution to get a tour of a whole new class of bias artifacts. But at least if you go with a kit designed to start with limited amounts of RNA you will be the beneficiary of the kit designers logic, rather than the victim.

Option 2 here is to hang back and wait for this problem to be solved by others. I'm pretty sure it will be eventually. Sure it is the pusillanimous option, but most of the time it will be the valid one.

--
Phillip

[eab]

Hey Phillip, thanks, what you write makes sense. I will probably wind up saying a prayer and heading down the dark path anyway - sometimes there aren't but a few hundred thousand cells (or less!) of a particularly interesting type to be had from any individual - but now at least I know better how to set my expectations. And I know to make one library out of water, just to see what the background is.
Cheers,
Eli