...what are you actually trying to measure?

Quantitation by "band intensity following PCR" is not a terribly good method, as PCR is exponential and also frequently performed to completion, so by 35 cycles you'll get exactly the same band intensity from one starting transcript or ten thousand starting transcripts. If you're trying to do this quantitatively, you'll need to be very, very careful in your cycle number, and you might save yourself a lot of time by just using qPCR. It's more expensive, but much more accurate.

Add to this, GAPDH is generally very highly expressed (especially in largely glycolytic cell lines) so shouldn't be difficult to detect anyway (your cDNA could be diluted 100-200-fold and would still give you bands, I suspect).

...what are you actually trying to measure?

Quantitation by "band intensity following PCR" is not a terribly good method, as PCR is exponential and also frequently performed to completion, so by 35 cycles you'll get exactly the same band intensity from one starting transcript or ten thousand starting transcripts. If you're trying to do this quantitatively, you'll need to be very, very careful in your cycle number, and you might save yourself a lot of time by just using qPCR. It's more expensive, but much more accurate.

Add to this, GAPDH is generally very highly expressed (especially in largely glycolytic cell lines) so shouldn't be difficult to detect anyway (your cDNA could be diluted 100-200-fold and would still give you bands, I suspect).

In agreement with John above:  you may still be inhibiting your PCR by using too much template.  Once you have eliminated your propensity for setting up the reactions incorrectly, try a dilution series of your cDNA:  1:10, 1:20, 1:50, 1:80, 1:100, 1:200, 1:500.  Then look at your band intensities -- you may find that the 1:500 dilution gives you the brightest band.  qPCR is definitely another option.  Always be aware of sample-related inhibition of the PCR process - especially with abundant targets such as reference genes like GA3PDH and 18S rRNA.  Also, if your tissues are from mouse - you will be dealing with a pseudogene for GA3PDH as well...

Hi everyone,

Thank you so much for your advice, I'll try this again. Actually, I tried 30 cycles, but in the protocol I received from my supervisor, the annealing temperature is 62oC and the extension temperature is 72oC (I'm using Amplitaq Gold), so I wonder if these temperatures are not suitable as well.

@John: As I mentioned above, since I started with 1:20 diluted samples and the bands were not very clear, so I just wanted to see how different the intensity of bands is if higher concentrations of DNA would be used. Because the results were not very good and there was not much different between non-diluted samples and diluted samples, I mistakenly chose the wrong concentration for PCR of other samples; thus I got no band, and there was also a problem with reproducibility.

@Raghavendra: My lab is using Trizol RNA isolation protocol. When I used such protocol, the OD 260/280 of all samples ranged between 1.6~1.7, so I wonder if this value is acceptable.

260/280 is far less important than 260/230. 260/280 is primarily for protein contamination, and you rarely get enough protein carryover to significantly affect downstream synthesis (and if you DO, it's usually pretty obvious even without speccing). 

260/230 is an indicator of (primarily) guanidium salt contamination, and this IS potentially bad. As chaotropic agents, guandium salts can really screw up cDNA synthesis reactions. TRIzol is a guanidium isothiocyanate based method, so this might be a concern.

I try to get 260/230 ratios above 2.00, but will generally accept anything above about 1.7 (or 1.6 if I'm feeling lazy).

As for dilution, remember that PCR is exponential, so diluting a sample 2 fold is the same as simply using one less cycle. A 20 fold dilution is only a difference of about 4 cycles, and 200 fold dilution only 7 or 8. 

Given your cDNA isn't very dilute, and you're looking at GAPDH, I would guess that you have already achieved saturation by 30 cycles (i.e. the reaction has exhausted all the reactants and further cycles will make no real difference to intensity). If you dilute your cDNA much more strenuously (so 200-fold or so), or do a dilution series as suggested by Jack (above), you might be able to detect a range over which band intensity is (semi) quantitative.