In many cases, people use U6 as normalization control.

You may give it a try.

Hi, you can serch in the literature (http://mcb.asm.org/content/31/1/203.full, http://genesdev.cshlp.org/content/26/19/2180.full for example) what normalization control is used in your condition.

As Sung-Chou Li U6 is the most used, I also try snoRNA (small nucleolar RNA) as these, on life technologies site:

http://www.lifetechnologies.com/order/genome-database/searchResults?searchMode=keyword&CID=&productTypeSelect=microrna&alternateProductTypeSelect=&originalCount=&alternateTargetTypeSelect=&targetTypeSelect=mirna_controls&species=Mus+musculus&otherSpecies=&selectedInputType=&keyword=&sequenceInput=&chromosome=-&chromStart=&chromStop=&vcfUpload=&batchText=&batchUpload=

Every system/cell line/treatment/time frame will have a different answer. Broadly speaking, when selecting a normalising reference, take the following into account:

1. miRNAs are transcribed by RNA Pol II; snoRNAs and some other non-coding RNAs are transcribed by Pol III. When normalising your miRNA to a PolIII transcript, you may simply be expressing the relative activity of the two polymerases. In short, you are assuming that PolIII activity remains unchanged, which is a completely inappropriate assumption under different serum concentrations and cell growth.

2. As your biological question is not fully described, does the half life of your RNAs play a role? Medium-size RNAs (snRNAs, snoRNAs etc) tend to be highly structured and incorporated into protein complexes, and thus very stable. If you are interested in transcription, or using short time points, using different classes of RNA may be biasing your results.

3. If you are using a single RT reaction for all genes and then qPCR for different genes (e.g. in SYBR-based assays), try to find normalising genes with similar abundance to your miRNA of interest. Reverse Transcription can saturate differently for different abundance genes. For example, in my hands, using rRNA as a normaliser for mRNAs is unreliable (!), despite being commonly used.

4. Do not use a single normaliser. Pick at least 3 genes and use their geometric mean (or average Ct, as Ct is a logarithmic concept). Much more stable.

5. If you have the money, do a genome-wide profiling (microarray or sequencing) of all miRNAs and pick the most stable ones for future use. Even just comparing the two extreme conditions will be informative.

Good luck!

Anna Git and others, thanks a tonne for the suggestions. I am a rotation student and I can't really ask for a genome-wide profiling study. My experiment deals with testing the theory that breast cancer can have far-off effects on other tissues. For this, I will be treating C2C12 murine skeletal cells with conditioned media obtained from murine breast cancer cell line and assaying miR486 expression in C2C12 cells. As 486 is essential in skeletal muscle development, I should be seeing a downregulation. This has been published earlier and they used miR30d to normalise. Unfortunately, 486 didn't show the expected result. We, therefore, surmised that it could be the serum concentration in the C2C12 plates. So, currently, my task is to grow C2C12 cells in different serum concentrations and check for levels of 486. I will use the concentration that gives me the highest levels, so for my experiment, I can see quantifiable results. Unfortunately, the normaliser 30d showed wide variations and I couldn't use it. Hence, the search for other nomalisers.

Based on some reading, I realised that it wouldn't be prudent to use snRNA or snoRNA to normalise due to differences in size, concentrations and the way they are synthesised. Thus, I have chosen to test the murine versions of miR 425, 16, 423 and 191 as suitable normalisers. I have attached the references for my choices in a separate file. Thanks. 

Update - After running RT-qPCR on C2C12 cells grown in varying concentrations of FBS, we came to the conclusion that the levels of miR 16 and miR 191 remained steady, thus, can be used as normaliser-controls for assaying other micro RNAs.

We concluded this by looking at the cycle-threshold readings on qPCR. Thanks all for the help.