If both of your antibodies are raised in rabbit, you will not be able to do double labelling.

Niels is correct. The only option you have is to change the animal of your primaries. Most of the alexas do not have too much bleed through in other channels. Lammel et al. use a mouse anti-TH from Millipore in their paper. I would not use a TH antibody meant for chicken without first checking the sequence homology for TH between rat and chicken, particularly at the site that the antibody recognizes.

Article Diversity of Transgenic Mouse Models for Selective Targeting...

I have not done it myself, so I do not know any more details. But when you first  stain one primary ab and use a fluorescent substrate that precipitates instead of a fluorophore coupled ab, you can boil the slides in citrate buffer (like antigen retrieval) and the precipitate stays stable while the antibodies will be removed. Now just stain your other primary antibody.

You can directly conjugate the fluorophores to primary antibodies if you have to use the antibodies from same species.  With appropriate filters, Cy3 or Alexa 555 and Cy5 should not have any bleed-through.

In our group we've done double labeling IHC (with DAB and DAB-nickel) to co-localize c-Fos and TH in free-floating rat brain sections. The primary antibodies we used were both from rabbit, so when we wanted to label with Alexa-conjugated secondary antibodies, we used a different TH antibody originated in goat.

Niels is correct of course.If you have primary TH antibodies raised in rabbit, the specific labeling of the rat brain using these antibodies is to be validated by the preabsorption procedures  using antibodies rabbit anti-TH and rat anti-TH.Posiive control for TH may be performed using the adrenal glands. 

Niels Haans is correct: if both your primary antibodies are raised in rabbit, then the secondary antibodies will both be anti-rabbit and both will bind to the primary antibodies. The easier thing to do would be to choose a primary antibody in a species that is not rabbit. Both anti-Fos and anti-TH antibodies are available from other species.

However, there is a way to do colocalization with primary antibodies raised in the same species. It involves using tyramide signal amplification (TSA) to amplify the signal of one of the primary antibodies used a very high dilution. The procedure is explained in detail in this paper from my lab:

Marvizon, J. C., et al. (2007). "Calcitonin Receptor-Like Receptor and Receptor Activity Modifying Protein 1 in the rat dorsal horn: localization in glutamatergic presynaptic terminals containing opioids and adrenergic α2C receptors." Neuroscience 148: 250-265.

Juan Carlos is correct.But appropriate controls are crucial in the result evaluation.In fact,one problem with this technique is the high probability of a false positive for colocalization.Unless the exact concentration  of the nonimmune serum required to completely quench any uncoupled F(ab')2 has been determined unbound F(ab')2 can bind to any free primary antibody from the indirectly labeled antigen.

Giacomo, that problem can be avoided by using a movalent Fab fragment as the secondary antibody. That completely avoids the F(ab')2 secondary antibody biding the second primary. Details are in my paper.

I agree with Juan Carlos. Use the Fab secondary antibody (labeled with a fluophores or  biotin) at first  and then the secondary anti- rabbit  (labeled with fluophores or biotin).  But before I start double labeling I woud run a single IHC to optimise the concentration an to get impression of the expression of your proteins of interest. TH shows a fiber and injtensiv golgi like stain of neurons. C-fos should show only a nuclear stain.  If you like to work with biotinylated secondary antibodies you can detect them with fluophore or enzyme labelled Streptavidin. This could enhance your results.

I am greatly thankful to all of you guys for your valuable suggestions.