Well to well variation in thermal cycler fluorescence?

Question

We have an old BioRad ICycler Thermal Cycler with MyIQ single color fluorescent detector. While it's meant for RT-PCR, we've been using it for melt curves and binding assays for different types of RNA using thiazole orange as the intercalating fluorophore. However, when we tried to get serious about this assay and do things in triplicate we noticed that the well to well variation was terrible, making statistical significance difficult to get.

After reading through the manual I tried to get "persistent well factors" for the instrument to correct for this variation. However, the instrument's protocol for this requires heating to 95⁰C, so intercalating fluorophores won't work, the RNA melts and fluorescence is lost, so I tried with fluorescein. This didn't solve the variation.

So I tried to make my own well factors by making 10mL of Poly(rA) Poly(rU) RNA solution with thiazole orange and filling 96 PCR tubes with 100µL each and measuring fluorescence. Then I calculate the average fluorescence and divided the average by the value for each well to determine a factor, F, for each well. Then I tried to normalize subsequent experiments by multiplying each well's fluorescence by its F to get a corrected value. However, while this reduces the variation a little, it's still pretty bad.

Do you have any suggestions to reduce the well to well variation in these experiments or otherwise correct the data?

I have no prior experience with RT-PCR instruments, everything I know about it comes from playing around with the instrument. I don't know if I'm missing something really basic.

Step by step protocol

1. Prepared 1xSSC buffer, 150mM NaCl, 15mM Sodium Citrate, pH 7.5
2. Weighed 5mg Thiazole Orange, dissolved in 1mL Methanol
3. Diluted Thiazole Orange 1000 fold in 1xSSC
4. Took 71.4µL (100µg) Poly(rU) Poly(rA) solution, 1.4µg/uL
5. Diluted RNA with 8.93mL 1xSSC
6. Added 100µL diluted Thiazole Orange to diluted RNA, final volume 10mL
7. Set up 96 PCR tubes, used 8 channel pipette and trough to transfer 100µL RNA solution to each
8. Placed PCR tubes in all 96 positions of thermal cycler
9. Ran melt curve protocol, start at 25⁰C, increase 1⁰C every cycle, 60 cycles, so final temp is 85⁰C. Each cycle is 10 seconds. Measure fluorescence every cycle.
10. Repeated melt curve 3 times, total of 4 curves.
11. Put data from each curve into spreadsheet. 96 data points at each temperature.
12. Averaged fluorescence across the whole plate at each temperature. Used this average, standard deviation, maximum, and minimum values to make the plots below.
13. Calculated correction factors for each well at each temperature by dividing the whole plate average fluorescence by each well's fluorescence for melt curves 2, 3, and 4. Averaged correction factors from each curve.
14. Calculated corrected fluorescence for curves 1, 2, 3, and 4 by multiplying the averaged correction factor for each well by the observed fluorescence for that well.
15. Replotted the corrected values to show reduced variation.

Data is shown below. I repeated the melt curve 4 times because the first melt usually has a lower intensity than subsequent melts. Y axis is fluorescence, X axis is temperature in celsius, and I plotted the average fluorescence at each temperature with the standard deviation along with the minimum and maximum to show how far outside the standard deviation some wells are falling. I calculated the correction factors for melt curves 2, 3, and 4 and averaged them. When I used these averaged factors on melt curves 2, 3, and 4 it did a good job at eliminating variation. When I applied the correction factors to melt curve 1, it reduced the variation, but not by much.

Answer 1

There can be several reasons for variability and since you said that the machine has not been serviced for a long time, I guess that should be one main source of errors. Multichannel pipettes can also sometimes cause problems (based on personal experience) and it is mostly because the tips don't fit properly in all channels.

Try varying some parameter (for e.g. concentration of RNA) and see the effect on melting temperature; if the inter-replicate variability is not high enough to make the inter-sample difference insignificant then you can continue with your experiment. Take more number of replicates to minimize variance.

If the variability is very high then I guess you would need to get your machine repaired or use more sensitive equipments designed specifically for these kind of studies.