I need to see if a gene has higher expression in one tissue type compared to another. I would prefer not to use housekeeping genes for normalization.
May I ask why you dont want to use a HKG as a reference? You can design a specific primer for your GOI and check their abundance/expression by doing a qPCR assay. Using a HKG makes your data reliable and error free!
Good luck.
HKG cannot be assumed to have constant expression across tissue types. So a lower relative expression of the GOI in one tissue compared to another could just be variability in the HKG expression rather than in the GOI.
Unfortunately, comparing two types of different tissues of selected GOIs is not accurate without a validated reference gene (RG). As a a RG can display altered levels of expression across different tissues, it is advisable to validate first in your tisues of interest, then normalise your GOI and only after verify the GOI gene expression between the different tissues. It is advisable that you confirm that your RG does not change across the data under analysis or else your data may be misleading.
Are you really only interested in mRNA expression? I ask because this is not neccessarily a good indicator of a protein's expression
You could use an RNA DNA prep Kit to simultaneously purify mRNA and genomic DNA from the same sample, and then perform qPCR for each primer pair against genomic DNA and cDNA in parallel. From this you can calculate the ratio of mRNA copies per gene copy for each tissue. However, depending on the tissue there can be some variation in the efficiency of genomic DNA preparation, which may become more problematic when you want to compare relatively small differences in expression levels..
I agree with Ammon in that HKG cannot be assumed to have constant expression across tissue types (Our in-house RNA-Seq data also supported this). so I think, Normalization using HKG only apply to expression level comparison between different states (for example, disease vs. normal) of the same tissue. I'm not sure whether I'm right because i am not expert at experiment.
I would choose an absolute quantification using same input amounts of RNA. In addition, I would recommend to spike your RNA with RNA from another species to have a control for RT efficiency since this can vary between different tissues. But in line with Filipas answer, in my eyes, your data won't be reliable without a reference gene.
In the MIQE guidelines, it is strongly suggested that normalisation is done using a number (minimum 3, in most cases) of validated reference genes. There are several ways to do this, but my personal preference is using a panel of potential reference genes and a software called geNorm to pick the appropriate ones.
You can amplify your interested genes by PCR and quantify the product. Then you can generate your own standard curve using your PCR product as your reference. then running your PCR samples with each run adding one specific amount of PCR product (usually 1E5 molecules ) as reference. Then you can quantify the exact molecules in your samples.
If you don't want to use reference genes you need to make the samples comparable in the amounts of RNA subjected to RT and the amount of cDNA subjected to PCR. You can use spike-in controls to check for RNA recovery, RT efficiency, and qPCR bias.
For reference genes it would be best to select the genes that are most stable across a multitude of samples. In a paper by de Jonge et al., PLoS ONE 2007 we have identified candidate housekeeping genes with enhanced stability among a multitude of different cell types (13,629 human gene array samples). It might be usefull to use these genes for normalization purposes.
Jochen is on the right track in terms of carefully monitoring the yields at each step of the process. Most answers about the MIQE guidelines are telling only part of the story. The MIQE guidelines suggest using HKGs, but it also says, and this is important, that you carefully justify the HKGs. If the HKGs are not expressed at the same levels in different tissues then this will create a bias that you will need to deal with. You haven't provided many details about what you are trying to measure and why. It matters because you may be able to deal with biases in how you set up the experiment. A detailed answer may also depend on what measurement you care about.
This also depends on how many genes you are analyzing, you may be able to apply global normalization if you have enough data.
Do you want to do absolute quantification? You will need an mRNA standard and ensure that you have quantified your starting material. You need to generate a standard curve then you can make comparisons to the curve.
Can I ask you why you don’t want to use house keeping genes for your qRT-PCR normalization?
Some kind of normalization is required since you are comparing between samples. Even if you were to absolutely quantify down to the number of RNA molecules of your target, if you load more template into one sample over the other, your cDNA rxn is more efficient in one over the other, pipetting errors, etc. your data is meaningless.
Using MIQE quidelines and the HKG mentioned above, you can be reasonably sure that your normalization is not skewing any data.
The suggestion to do a standard curve for your specific primer set is a good idea if you are doing some serious comparisons since that way you will know the efficiency of that qPCR reaction, but also will not help if the amount of template is different, it simply means that normalized data can be quantified and compared.
If you find a way to normalize without using HKG though, that would be very nice to know!
good luck!
I think it is a risky business, the selection and use of housekeeping genes. Several papers report that genes of that idealized type actually aren't.
Instead, nonparametric analysis can be employed. This is probability, not so much statistics, and the bible is by Hollander and Wolfe. For example, suppose you assay 100 markers in two samples. Suppose the top 10 (lowest Cq values) of the first assay are all in the bottom 20 of the second assay. This is highly unlikely by chance. Suppose you assay 10 samples of one type and 10 of another type. Suppose there are 10 markers always in the top 20 of the first type but never in the top 20 of the second type. This, too, is highly unlikely by chance. In both examples the 10 markers would be worthy of further experimentation (including bioinformatics).
Of these things we can be confident. Of housekeeping gene comparisons we should be wary.
I would use housekeeping genes. You can use quite a number of HKGs (>5) and compare the outcome using Bestkeeper software (http://www.gene-quantification.de/bestkeeper.html). Using Bestkeeper you will see HKGs that are coherent and those that are not and you can exclude the latter from the analysis. For the selection of HKGs to start with I would look at general expression of the HKGs using for example in silico transcriptomics at genesapiens.org.
Finding a reference gene(s) that is stably expressed within several tissues and several biological replicates is quite a challenging task.
Besides, even such genes are being identified, the copy number of RG in tussue X could be different to the copy number of RG in tissue Y, and that would lead to wrong conclusions in case of relative quantification.
Tissue X RG: AAAA molecules GOI : B molecules ratio GOI/RG - 0.25
Tissue Y RG: A molecules GOI : B molecules ratio GOI/RG - 1.
If using this layout, even is someone proves that RG is stable in Tissue X and Tissue Y, the conclusion would be that GOI is 4 times higly expressed in Tissue Y - but it will not be true.
As other suggested, in your case I would go for absolute quantification, after normalization for RNA quantity that you retrotranscribe and with spiking with an alien RNA to check if there is a difference in RT efficiency.
The other option would be directly to do RNAseq and quantify the positive calls for your target amongst various tissues.
Thank you all for your advice. I am thinking of using total RNA estimated by nano drop to normalize my data. In 32 different RNA preps I've noticed very little variation in Ct value for three different genes I've looked at. I think this means total RNA concentration is good enough and potentially superior to normalizing to the geometric mean of multiple "validated" reference genes. I know rRNA makes up a large percentage of total RNA but I dont think its large enough that mRNA concentration becomes noisy and lost in the level of precision of the nano drop measurements. Do any of you have experience that would be informative? What do you think? I kind of think level of expression relative to total RNA is more biologically informative than level of expression relative to a few commonly expressed genes. Maybe this is wishful thinking.
I should mention that these RNA samples are biological replicates where I controled cDNA concentration in the qPCR experiment. The samples were fish skeletal muscle.
If the amount of RNA is a good reference depends if the states or treatments you want to compare do not (much) influence the overall RNA content of the cell. Comparing different cell types is thus always very problematic. Also, if the cell proliferation of differentiation is affected, RNA content may well be systematically different.
If you use nanodrop measurements for normalization make sure that the purity of your samples is comparable (260/280nm and 260/230nm ratios) since differences in protein content or phenol carryover contribute to absorption at 260nm and affect efficiency of reverse transcription and PCR
Dear Ammon, in that case it's about the same tissue type, only the samples are different (replicates in your case). Replicates are replicates, biological replicates are the best in order to check the consistency of your results from statistical point of view. Therefore the normalization here could be a relative constant parameter, such as mass of muscle, starting RNA amount or cDNA input. However, using the total RNA amount should exclude mRNA/other species RNA ratio variations (if any) in you tissue. I will use cDNA quantification, for the reverse transcription is probably the most variable reaction in terms of efficiency, due to the reverse transcriptase intrinsic specificity which ranges between 15-30% (the lowest if one ranks enzymes using this parameter). What I don't understand, why are you comparing gene expression in replicates of the same tissue?
- Badges
- Science topic