Hi, I prefer mRNA. For designing primers, for you need an RT primer which can bind to the 3' end. Normally this is an universal primer having oligo(T) followed by 40-mer long nucleotide of your choice. For example: TTT TTT TTT TTT TTT TTT AGC CGT ATG GAG ATC GCA TAC CGT CAG TAC CGA TGC TAG GAC. Use this primer for RT as well first PCR along with an gene specific primer which is designed based on the gene of interest.
Use first PCR product for nested PCR using second set of primers which are designed based on the sequence of RT primer as well as gene of interest.
If you know where the 3' end of the gene is, sequencing genomic DNA would be a much simpler task. You need only design genomic primers spanning the location and PCR up the product followed by sequencing. If what you are actually interested in is whether there are multiple different 3' UTRs of the gene then obviously, you need an mRNA based approach.