How clean is your RNA? If you're getting inhibition from adding more RNA it would tend to suggest contaminants, since I assume (as it's a one-step) you're using transcript-specific primers (thus ruling out 'saturation effects from too much RNA' that one might expect from oligodT/random priming).

Also, further to the points Graeme makes, above: larger volumes can actually be beneficial when dealing with low transcript number, since we're dealing with DNA molecules in the countable range. You can keep all reagents at the same concentration (as noted) but effectively double the number of transcripts. If you only have 10 transcripts per 10ul volume, performing exactly the same reaction but with a final vol of 25ul puts this up to roughly 25 transcripts, which should lead to more reliable data (lower incidence of stochastic partitioning etc).

Having said all that, I'd be tempted to suggest that a one-step method is perhaps not the most appropriate approach here. Have you considered preparing total cDNA first (via oligodT/random priming) and then using this in subsequent qPCR reactions? It gives you a lot more flexibility over the conditions.

Thanks for the answers. I originally tried using two-step with oligo dT, and random hexamers, and had a similar experience. We switched to one-step because I had read it can give lower Ct values... But now I'm still stuck in the same place. What kind of "flexibility" would I be afforded by using total cDNA?

I also forgot to mention that when I run my qPCR reactions on the Bio-Rad iQ cycler, the minimum amount you can input as the reaction volume is 15ul. So I have been selecting that, even though my volume is only 10. Could this be an indication that my machine prefers larger volumes due to reduced sensitivity?

(I'm grasping at straws here...)

Thanks again.

Ok, so: more questions.

Is your RNA of good quality?

Ideally bioanalyzer-verified, but at the very least, specced (nanodrop or similar) to confirm 260/280 and 260/230 ratios are good.

260/230 ratio is key here (high 230 usually indicates guandium contamination, so I would be cautious of ratios lower than 1.8 or so). Should be 2.0 or higher ideally.

Is your RNA reasonably concentrated?

Most cDNA synthesis kits seem to have a cap around the 2ug mark, so in an ideal scenario you should have sufficiently concentrated RNA to be able to use 1-2ug for cDNA synthesis without encountering volume constraints.

If all the above is fine, and you're still seeing very high CT values, then you probably are simply in the stochastic "very low transcript number" range. If you're trying to estimate reaction efficiency and transcript number, you don't need to rely purely on dilution of your cDNA. You can simply use your primers in a conventional PCR with your cDNA (use as many cycles as necessary to amplify) to generate a reasonably large quantity of your PCR product, then gel extract it, clean it up and spec it, and use

concentration of DNA (in ug.ul-1) divided by (PCR product length x 660)

to determine your molar concentration (mol.l-1), then multiply by avogadro's to determine molecules per litre. Working from there you can prepare stocks at 10^6, 10^5, 10^4, 10^3 (and so on) molecules.ul-1, and use those to generate a standard curve, generally including some non-specific DNA for stability (add as much non-spec DNA as you usually add cDNA. I use sheared herring sperm DNA: it's relatively cheap and cheerful).

Hopefully if all goes well, the samples prepared from 10^1 and 10^0 molecules.ul-1 should give quite a lot of variation in CT value (since these are well into "uneven partitioning per well" territory), and the whole curve should allow you to estimate what CT value is equivalent to 1 transcript for your reaction (usually around 35).

First, thank you for taking your time to answer.

For your first question, I am fairly certain that my RNA is of decent quality. I nanodropped it, and 268/280 was about 2.08, and 260/230 was 1.82. That is for the RNA that I have been using exclusively for generating standard curves. However, in my actual samples, the 260/230 is much lower (unfortunately). However, if I can't even get a decent standard curve to work with good RNA, it won't much matter how the quality of my samples are.

The nanodrop showed concentration of my RNA to be 1220ng/ul, so I think that's plenty concentrated.

All of this sound pretty bad, but I just discovered a different problem: When using 1-step vs 2 step, my melt curves are at different temps... back to the drawing board. I plan on posting a new question regarding this including snap-shots of the curves.

I will reference your protocol for stochastic determination, if I can work the rest of the kinks out.

Thanks again

This diffrence in melt temperature may be a result of the different cation types and concentrations in the two mastermixes. Cation concentration has a pretty big effect on melt temperature.

The one step mix will probably contain a mix of Magnesium and Manganese (IIRC), whereas the normal PCR mastermix will only contain Mg.

Yup! I just called Bio-Rad, and that was the reason they gave as well. I feel better about it, but now I need to figure out what to do with these ultra low transcript numbers.

Thanks everyone! This website is so helpful.

John, I think I am going to have to use the method you suggested. Is there any publications I can use as a reference? I'd like to get a little more reading done before diving in, so I was hoping you could point me in the correct direction. Really appreciate it. Thanks.

OK, posting this here just for general reference (and an attempt at shameless upvote farming -also check your inbox, Austin), but for the record...no, this isn't really based on any particular publication, more of a sort of aggregated base of knowledge accumulated via a few years of working with low transcript numbers.

Generally speaking, it's not technically necessary to determine actual transcript numbers, not least because very few journals will really believe them, and also because once you factor in normalising it all gets a little confusing as to what the numbers even really represent (transcripts per....uh...dunno).

For most studies, you're working in comparative terms (expression of gene X in condition Y vs expression of gene X in condition Z), so you can avoid any transcript number issues by simply determining that "gene X is increased 40 fold in condition Y vs condition Z". It doesn't matter whether there are 10, 100 or 1000 transcripts, all that matters is there's 40 times as many.

Determining transcript numbers can be useful for determining biological relevance, however, and can help plan experiments (also helps "researcher peace of mind", I suppose), but because of the exponential nature of PCR, ascribing numerical values to transcripts is...always going to be a bit tentative. Especially in the low transcript numbers. Saying "6 transcripts per cell" is risky, whereas "ca. 10 per cell" is usually acceptable.

Anyway. A rough protocol would be as follows.

Perform a PCR of your GOI, using the qPCR primers. Run this on a gel. Determine that there is just the one band (primer dimers not ideal, but acceptable. Actual mispriming is, however, BAD).

Cut this band out, gel extract it, clean it up if necessary (I find gel extraction leaves a lot of salt, you may be luckier).

Spec it. From a cleaned PCR redissolved in 10-20ul of H2O I usually get somewhere between 30-80ng.ul-1. It's not a lot, but in terms of molecules it's boatloads. As long as it's fairly clean (decent 260/230 and 260/280) it should be fine.

Now calculate how many molecules you have per ul.

The molecular weight of a doublestranded DNA molecule is essentially 660 times the number of basepairs, so if your product is 150bp, then it has a mol wt of 660x150 or 99000 daltons.

Thus the concentration of your DNA (in Molar) is [conc in g.l-1]/mol wt

So assuming our 150bp fragment, above, is in a solution at 30ng.ul-1 (i.e. 30mg per litre), that's

0.03/99000 = 3x10^-7 M or 300nM

Now multiply by avogadro's number

3x10^-7 x 6x10^23 = 1.8x10^17 molecules per liter, so 1.8x10^8 molecules per microliter.

See, told you it was a lot of molecules.

So we have a stock at 180 million molecules per microliter, and I generally find that 100,000 to 1 million starting molecules is the absolute upper range worth plotting on a standard curve (especially in your case, where you're primarily concerned with low transcript levels). So dilute your stock down to give 1x10^6 molecules per ul (i.e. a 1 in 180 dilution in this case, or 5.6ul in 1ml -make this dilution carefully: pipetting errors etc).

Now make a series of ten-fold dilutions of this (I usually use 100ul into 900ul, because larger volumes generally lower pipetting error), so you have stock solutions ranging from 1x10^6 molecules per ul all the way down to 1x10^0 or even 1x10^-1 molecules per microliter (i.e. 1 or even 1/10th of a molecule per ul), and use these in a qPCR with your primers, 1ul per well. Because you're adding quite small amounts of DNA (albeit all the SAME bit of DNA), it's usually wise to add non-specific DNA to make the final well [DNA] the same as you would have with your actual cDNA.

I use sheared herring sperm DNA as my non-spec DNA (because we had a load in the freezer that wasn't being used) but any generic non-specific DNA should work. Since I usually qPCR for low expression transcripts using 30ng of cDNA.well-1, I add 30ng of herring sperm DNA to each well.

I usually do this in duplicate wells (triplicate samples is overkill here, really).

If you do this, what you should see something like this (qPCR curvesmall):

A fairly predictable increase in CT as number of molecules decreases (should increase by 3.32 CT values per 10-fold dilution if 100% efficient), and as you near the Y-axis (i.e. as number of molecules approaches 1 or fewer per ul) the trace should become more variable (because small numbers of molecules often partition unequally, so you might have 8 and 14 molecules in your duplicate wells, rather than 10 and 10). Beyond the Y-axis we're into "molecules are here or not" territory, so if you have primer dimers, this is where you'll see them. For this example, primer dimers clearly appear with a CT of about 48, while single transcripts have a CT of 35 (yes, I ran the machine for a lot of extra cycles just for illustrative purposes).

From this, you can gather a lot of information: you can draw a trendline through the values you trust (should have an R-square value of 0.99 or greater, usually corresponding to values from 100 molecules.ul-1 and above) and get the reaction efficiency, and you can extrapolate to the Y-axis to estimate your "CT value equivalent to 1 molecule".

What you can also do is include your standards WITH YOUR TEST SAMPLES. In other words, have replicate wells containing your test sample, all with the same amount of cDNA, and add your standard curve samples to these. At a certain point the dilution curve should flatten off (when amount of standard curve templates falls below the number of transcripts in your test sample) and from this it's relatively straightforward to extrapolate across to determine the number of transcripts this is (roughly) equivalent to.

So as shown here (qPCR curve2small): my 'standards only' dilution curve (black squares) flattens off as it nears the 1-10 transcript range (stochastic range), but the three test samples all flatten off much earlier, and extrapolating back allows us to determine that those samples have about 1000, 10000 and 60000 transcripts per cDNA added, respectively.

Of course, this is less useful if all your samples are in stochastic territory (and of course it uses a lot of sample cDNA), but it's worth noting.

For samples that are present in very low amounts, the best advice I can give to obtain meaningful values is to use a lot of replicates. If you have 5 transcripts per ul of cDNA, then you're going to get a lot of variation in your CT values, but if you perform 6 or more replicates (rather than the usual 3) the average value obtained should be fairly representative. Also, always always include a negative control, because with CT values higher than 30, primer dimers become a serious issue. If your negative control shows primer dimers, but they appear at a CT of 40 or similar, then you're probably ok. If they're appearing at the same sort of cycle as your transcripts, then you probably need to redesign your primers.