Spiky high basepair noise in RNAseq libraries

[Rammohan]

Hello
I picked 100, 150, 200 neuronal cells from mouse brain tissue using arcturus lcm.
extracted RNA using picopure kit and then using SMARTer Stranded Total RNA-Seq Kit - Pico Input Mammalia kit I made libraries.
The bioanalyser traces (appended along) show spiky peaks ranging from 300bp to 23000bp. The spikes are beyond the 10380bp markers (shown in purple). I have following questions:
1.Are these spikes PCR artifacts (I took 16 cycles of PCR).
2.Can I do gel extraction from these libraries from size ranging from 300 to 1000bp, if yes, then what is the best method?

For reference:
sample 1,4,7 and 10 are from RNA extracted out of 100 cells
sample 2,5,8 and 11 are from the RNA extracted out of 150 cells
and samples 3, 6 and 9 are from 200 cells.

Thanks in advance,
Ram

[nucacidhunter]

1- It does not seem like a usual PCR artefact.
a .I wonder how long the fragmentation time was.
b. Did RNA was treated with DNase
c. 16 PCR cycle is for lowest recommended input. Did your workflow included QC of input?

2- You can gel extract but it will affect the outcome if you are doing differential expression analysis.

[Rammohan]

many thanks for the help.

1.How does the usual PCR artifact look like in bioanalyser traces?
As few of the spikes are more than 2000bp I thought that to be a PCR artifact.
1a. The fragmentation time was 3min at 72 degrees. The fragmentation time was same for all the samples, but it seems to work well with library made out of 200 cells, It might be possible that the low input from 100 cells might have given PCR artifacts?

1b.Yes The RNA was treated with DNAse. (I feel that the DNAse treatment also reduces the RNA amount, do you have any suggestion for a good DNAse treatment method for small amount of RNA input)

1c.I am not sure what are the QCs for the input RNA? I think the RNA was degraded after LCM and the RIN value of the scrap (tissue from which the cells were picked) was about 3.

2. At this point I am not sure if the gel extraction will be useful or not for differential analysis but I want to make sure that the input for the library should be in the range of 300 to 1000bp, which I assume will give a good and even cluster!

[HESmith]

We typically observe spiky peaks from low-input libraries (RNA and DNA).

The profiles look like the chip was overloaded, which produces a high-molecular-weight tail that is an artifact. Also, lanes 1-4 migrated aberrantly; the HMW marker is actually the band above the purple indicator.

Bottom line: the libraries should be fine for sequencing without gel size selection.

[jdk787]

I agree with HESmith.

Also, to get a better idea of what the actual size of these first four libraries is, you just need to correct the marker calling .

To do this you should go to each individual sample in the BA software, click the button on the bottom labeled "peak table", move your cursor over the correct upper marker, then right click and select "manually set upper marker".

This should then give you correct sizes for these samples.

[nucacidhunter]

Examples of overamplification are shown in following document copied from KAPA “High-Throughput NGS Library Preparation Technical Guide”.

Overamplification examples.pdf

Thanks HESmith for spotting HMW shift that made me to think possibility of DNA in RNA prep used.

[Rammohan]

Thanks for the support.
I reran the bioanalyzer experiment, this time, loading exactly 1ng of each library (after measuring the concentration using Qubit) and the traces look fine this time (with an average fragment size of 300 to 400bp, see the attached pdf file of the traces).

Thanks nucacidhunter for sending the example of PCR overamplification.

But one thing I am still not able to figure out is the noisy spikes. Why the curves for 150 and 200 cells are relatively smooth but that for 100 cell is not?

For reference:
sample 1,4 and 7 are from 100 cells
sample 2,5 and 8 are from 150 cells
samples 3, 6 and 9 are from 200 cells
Sample 10 and 11 are positive control from 250pg starting material

Thanks

[Rammohan]

Hello again,
I thought of sharing the data of my recent run.
This time, I got higher total yield (in terms of ng/ul as measured using Qubit) as compared to my previous run.
With same amount of starting material, I got ~3ng/ul last time but this time, I got around 12ng/ul. I probably use to lose a lot during bead purification.

I was going with 16 PCR cycle before, I was wondering if I should go for 14 or 15 cycles this time?

Some curious observation with this run were:
Spike in sample 4 and 8 (which are replicates) near 300-350bps.
The control Library using 250pg starting material (sample 11) is showing multiple humps. nucacidhunter, is this what you call daisy chain (based on the over amplification example you posted)?

Would like to receive valuable comments from you all!

Thanks

[nucacidhunter]

For PCR cycle number it is better to follow recommended cycles based on input quantity. Expected yield is 3-10 ng/ul. You can do less cycles and check the yield (without clean up) and if it is not enough follow with more cycles assuming doubling yield for each cycle.

The spikes are indicator of a particular fragment that is over-represented in
the library. Their source could be incomplete removal of rRNA or environmental contaminant especially if 250 pg input was not done in a clean space. Sometimes spikes are result of BA vibration during run but in this case it can be dismissed because they are from the same input.

Sample 11 shows signature of overcycling and I am not sure if they are daisy chain or not as I have not seen any evidence proving the assumption. They could be chimera as well. Convincing evidence for daisy chain theory would be re-amplification of a fraction of that library in a few PCR cycles which should give a peak corresponding to smaller one.

[Rammohan]

Thanks for the support Nucacidhunter!