I am optimizing conditions for an RNA-IP Sequencing (RIP-seq) experiment. The goal is to identify associated RNAs that co-IP with a particular transcript.
IPs will be performed with antibodies against an RNA-binding protein. The goal is to identify transcripts that co-IP with this RBP. We know a handful of RNAs that bind this RBP and can use those RNAs as positive controls.
For my experimental samples, I am getting about 10-fold more RNA pulled down than for the negative controls (as measured by Qubit). I'm testing a few negatives, including IgG and another unrelated protein.
An initial screen with RT-qPCR shows that a positive control transcript is much more enriched in the samples than the negative control (great news).
My concern is the difference in total RNA pulled down in the IP. Will the discrepancy between RNA concentrations affect downstream steps in library preparation for RNA-seq? Is it ok to use a negative control for RIP-seq that pulls down significantly less RNA than the experimental samples? Do I need to find a negative that pulls down the same total concentration of RNA but fails to pull down my transcript of interest?
Any advice would be appreciated. Thanks!
Im not sure what your negative control is but in my experience you will always get more RNA recovered (5-10-fold) in the negative control than the experimental condition (this in itself is a good indicator that your IP is specific). For my experiments I was using a FLAG-tagged RNA-binding protein and my control was the untagged protein (Ellis et al, 2015, Mol Microbiol).
The RNA concentration in the negative control would then dictate how much IP RNA you can analyze. I used RT-PCR or Northern Blot and Primer extension to analyze my transcripts and because the IP RNA was much more dilute then the input I was typically analyzing about 30-fold more input RNA than IP. You can take this difference in amount of RNA analyzed into account when calculating the enrichment of your target transcripts.
Since you are analyzing your RIPs by RNA-seq you don't need a lot of RNA. I would recommend using the control IP sample to set the upper limit of how much RNA you can use to make your libraries. For example if your control IP RNA is at 0.02 ug/uL then you might be better off diluting all samples to 0.02 ug/uL.
The bigger challenge will be normalizing read-counts from sample to sample since the IP will be inherently biased based on the number of transcripts that bind your RNA-binding protein.
Hope this helps.
The library preparation should not be affected as long as you still have the minimal number of RNA transcripts for your sequencer available.
How much it will matter for the RNA-Seq data depends on how you want to normalize it. If you measure all the genes, globalization does the trick. But if I understand it correctly, you will only analyze a small fraction of your genes? If so, you cannot use globalization (like averaging your gene expression by TPM, RPKM and so on). By globalization you would assume that most genes are not differential expressed and therefore oversee true changes easily.
So it definitely would matter in that case. That means you need a different source to get the scaling of your samples. If your determined pulldown would be constant for all the genes, that could be used. Any gene that is known to be constantly expressed independent of the condition can also help scale library size differences.
If non of that works, I would not be sure how to estimate the bias caused bei the IP if it truly does affect different genes differently.
Im not sure what your negative control is but in my experience you will always get more RNA recovered (5-10-fold) in the negative control than the experimental condition (this in itself is a good indicator that your IP is specific). For my experiments I was using a FLAG-tagged RNA-binding protein and my control was the untagged protein (Ellis et al, 2015, Mol Microbiol).
The RNA concentration in the negative control would then dictate how much IP RNA you can analyze. I used RT-PCR or Northern Blot and Primer extension to analyze my transcripts and because the IP RNA was much more dilute then the input I was typically analyzing about 30-fold more input RNA than IP. You can take this difference in amount of RNA analyzed into account when calculating the enrichment of your target transcripts.
Since you are analyzing your RIPs by RNA-seq you don't need a lot of RNA. I would recommend using the control IP sample to set the upper limit of how much RNA you can use to make your libraries. For example if your control IP RNA is at 0.02 ug/uL then you might be better off diluting all samples to 0.02 ug/uL.
The bigger challenge will be normalizing read-counts from sample to sample since the IP will be inherently biased based on the number of transcripts that bind your RNA-binding protein.
Hope this helps.
I agree with Michael - I wouldn't worry about getting more total RNA from your negative control, it is probably a good sign to suggest your IP is specific. I also find this to be the case most times with my RNA-IPs. What actually matters is what proportion of that total RNA pool is relevant ie. you might have a lot more different RNAs in your negative control as opposed to a few of them in your IP that are present in bigger quantities. Hope that makes sense!
I definitely wouldn't change the negative control so as to force it to have less total RNA. We recently performed a RIP-Seq experiment that started off with more total RNA present in the negative control and we have found it had no impact on the library preparation.
Thank you all for your input! A few questions based on your replies.
Bastian and Michael - Indeed, I will only analyze a small fraction of transcripts, as only a subset will be enriched in the IP. Couldn't I still do differential expression analysis with a program like edgeR or deseq? I believe this uses counts of raw sequencing reads (via htseq), instead of direct comparison of RPKMs.
Bastian - Finding a gene that is constantly expressed independent of condition is very difficult. We've encountered this problem with RT-qPCR normalization, and have since started using a spike in control. Could this also be done for RIP-seq? Adding in a known concentration of a single exogenous transcript and normalizing libraries that way? I know ERCC spike-in controls exist, but these are very expensive, and may provide more information than I actually need.
Thank you again for your help!
- Badges
- Science method