Best kits for gene expression analysis in fungi (and other gene expression questions)?

[hypervalent_iodine]

As per the title, I'm looking at doing some gene expression analysis in S. cerevisiae as well as some pathogenic fungi (C. neoformans, for example). I am fairly okay with extracting the total RNA and assessing integrity of the samples, having done this pretty extensively in my last job, however I am unsure of where I should go for isolating the mRNA. My plan currently is to use the Qiagen RNeasy kit followed by mRNA purification with their poly A minikit (I think it's called Oligotex). It is a little expensive, however. If there are kits that are better or the same in quality for less, I'd be more than willing to try them. Biology is an expensive discipline to be involved in.

 

My second question is more generally about methods for looking at gene expression. Would it be worth going for a next-gen sequencing approach over a microarray? I have never done microarrays before and do not really know what the output is like. I'm not sure that the extra data is really necessary, which pushes me towards a microarray.

 

Last question. I have read some papers where they use RT-PCR to confirm their microarray results. Should this also be done with next-gen data?

[BabcockHall]

I am not an expert, but I would like to ask for some clarification. When you say RT-PCR, do you mean real-time PCR or do you mean reverse transcriptase PCR?

[hypervalent_iodine]

I am not an expert, but I would like to ask for some clarification. When you say RT-PCR, do you mean real-time PCR or do you mean reverse transcriptase PCR?

Yes, sorry, I should have specified. I mean real-time.

[CharonY]

A couple of things. In general, if you go the kit route, there is not a huge difference between them (on average). Typically, familiarity and practice with a given protocol is a stronger determinant than the kit itself. If you got only one system, such as yeast and under mostly standard condition, RNeasy is fine. If you have different cells and/or may run into situation of low cell count and increase of interfering substances a phenol-type extraction (e.g. trizol is more flexible and does a better job at cleanup).

Microarrays by now are cheaper, but more restricted in many ways, I would probably go for NGS for an easier sell. At least for now, considering there are still statistical issues that have not been ironed out completely. Though that does not mean that microarays are better in that regard, it is more that their issues are better known. Crucial, as usual, are biological replicates, which still are going to be expensive.

Validation: that is somewhat of a biggie and people often do get it wrong (but still get it published).

In the end, it depends on the conclusions you intend to draw with your data set and in some cases it is sufficient to run some qPCRs with sufficient biological replicates. Unfortunately some use the original RNA extraction, instead of growing new batches which ends up just to be a comparison of RNAseq to qPCR data, with little power to infer biological effects (as an example). Also there is the huge fight over how to properly normalize and calibrate qPCR data in the first place, especially in the low abundance regime.'

[hypervalent_iodine]

Thanks, CharonY. I have used trizol in combination with the spin column mini kits from Qiagen in the past, though more as a learning exercise than for anything else. It is also highly possible that we have trizol in one of the labs I work in.

 

 

 

Validation: that is somewhat of a biggie and people often do get it wrong (but still get it published).

In the end, it depends on the conclusions you intend to draw with your data set and in some cases it is sufficient to run some qPCRs with sufficient biological replicates. Unfortunately some use the original RNA extraction, instead of growing new batches which ends up just to be a comparison of RNAseq to qPCR data, with little power to infer biological effects (as an example). Also there is the huge fight over how to properly normalize and calibrate qPCR data in the first place, especially in the low abundance regime.'

 

 

I will keep this in mind when it comes time for validation. Is there a particular method you would suggest for normalising and calibrating data or is it more of a project-specific thing?

[CharonY]

It depends largely on how much you want to infer from your runs.

There are several levels of normalization but I will limit the discussion to the use of reference runs for now (other aspects are well covered in literature). Theoretically, the best way, which no one does, is to use artificial mRNA of your target gene in known quantities to create a calibration curve. This would take care of the RT reaction (which is somewhat underexplored as a source of error) as well as the actual qPCR run. A second method, which is cheaper is to use cDNA of your gene in known quantities. In some cases people utilize genomic or plasmid DNA to do that. While the RT bias would not be included and there is going to be some deviation from the true results, in many cases one assumes a constant error, which still allows for comparisons.

What most people are doing, however, are simply the use of reference genes, which are assumed to be expressed independent on the variables you are testing (i.e. housekeeping genes). Many only use one, which is rather bad practice. But even if using a panel there is the general issue that there are no true housekeeping genes. In fact, the more omics data you generate and look at, the more you will note that almost everything can change under given conditions. This can be due to external sources (e.g. nutrients, stressors) or interior signaling (e.g. cell aging, cell cycle).

So one should validate whether this reference genes are truly stable in your test conditions. Even then, there is a lot of uncertainty and everyone has to decide what level of accuracy one is aiming for (or can afford).

[hypervalent_iodine]

It depends largely on how much you want to infer from your runs.

There are several levels of normalization but I will limit the discussion to the use of reference runs for now (other aspects are well covered in literature). Theoretically, the best way, which no one does, is to use artificial mRNA of your target gene in known quantities to create a calibration curve. This would take care of the RT reaction (which is somewhat underexplored as a source of error) as well as the actual qPCR run. A second method, which is cheaper is to use cDNA of your gene in known quantities. In some cases people utilize genomic or plasmid DNA to do that. While the RT bias would not be included and there is going to be some deviation from the true results, in many cases one assumes a constant error, which still allows for comparisons.

What most people are doing, however, are simply the use of reference genes, which are assumed to be expressed independent on the variables you are testing (i.e. housekeeping genes). Many only use one, which is rather bad practice. But even if using a panel there is the general issue that there are no true housekeeping genes. In fact, the more omics data you generate and look at, the more you will note that almost everything can change under given conditions. This can be due to external sources (e.g. nutrients, stressors) or interior signaling (e.g. cell aging, cell cycle).

So one should validate whether this reference genes are truly stable in your test conditions. Even then, there is a lot of uncertainty and everyone has to decide what level of accuracy one is aiming for (or can afford).

 

I was speaking to a old professor of mine about this earlier today, who told me that they usually use 4 housekeeping genes for confirmation for the reasons you have described. I like the idea of using cDNA, though.

 

Thank you for your help, as always. Now I just have convince my supervisor(s) that it's worth spending money on.

[CharonY]

On the note of housekeeping genes, it is important also to check whether the ratio of the proposed housekeeping genes are constant. Also, as a rule of thumb if your growth diverges massively from your "reference" condition, there is a good chance that your housekeeping genes are also going to be altered. The problem here is that it is typically unknown how the change looks like in reference to your genes of interest. If you go to tissues it is even worse.