As this question is a first post it is probably just a basic question about protein synthesis, which @Sean Johnson has answered adequately. However I’m not quite sure. And as I used to work in protein synthesis (but am a rather out of touch now) I decided to look at the recent literature a little to address some more esoteric or obscure questions it raised in my mind. I’ve set out the results below, and divided my answer into mature mRNA (which Sean covered primarily) but have also considered pre-mRNA. If the answer is not very useful to the questioner it may still be of interest to others.
Start and Stop Codons in Mature mRNA
The general picture of initiation and termination of protein biosynthesis in eukaryotes (bacteria and archea are different) has been described by Sean and can be found in standard texts1. In most cases the first AUG from the 5'-end of the mRNA is the one recognized as the ‘start’ signal for protein synthesis by the ribosomal-subunit initiation complex that scans along from the cap, according to the Kozak model2; and any one of the three termination codons encountered in the reading frame is recognized as ‘stop’ and (almost) invariably lead to termination of the polypeptide chain.
START EXCEPTION 1
As mentioned, in some 5% of mRNAs certain aspects of the environment of the first AUG from the 5'-end results in it being ignored by the complex, and the insertion of methionine and start protein synthesis occurs at the second (or perhaps even a subsequent) AUG.2
START EXCEPTION 2
Certain eukaryotic viruses were found to initiate the translation of the mRNAs internally, by a mechanism distinct from the 5'-scanning method. This involves an ‘internal ribosomal entry site’ — IRES. It subsequently transpired that a small subset of host mRNAs posses an IRES and can initiate translation internally. These include some growth factors, transcription factors and translation factors.3 Such (rare) mRNAs can therefore have several AUGs in their 5'-untranslated region.
Although it is possible to envisage an mRNA with several IRESs, and thus encoding several different proteins (analogous to a bacterial polycistronic mRNA), this does not appear to happen. It is striking, for example, that poliovirus RNA has just a single IRES for the start of the synthesis of single a large polyprotein, from which it generates multiple proteins by proteolysis. It would seem that the main purpose of this alternative initiation is to allow protein synthesis to proceed in the absence of a 5'-cap.
STOP EXCEPTION 1
There are some AGU codons that are recognized by a specific tRNA that inserts selenocysteine into the polypeptide chain. Some aspect of the environment of these codons causes them to be ignored by the protein termination system, and the ribosome continues translating, eventually terminating at a subsequent stop codon4.
STOP EXCEPTION 2
It has long been known that certain bacterial and eukaryotic viral mRNAs can increase their genetic repertoire by what is known as ‘read-through’ of ‘leaky’ stop codons. This is a ‘percentage’ effect, rather than all or nothing, and results in a mixture of ‘normal-length’ and extended products. It is now clear that the same phenomenon can occur in some eukaryotic mRNAs: quite frequently in Drosophila5, but also in mammals6.
STOP CODONS & THE 3'-UTR
I’m not clear what the questioner means by “is the stop codon always the final sequence before the poly-A tail?”. If the question is whether there can be additional stop codons in the 3'-UTR between the functional termination codon and the AATAAA polyadenylation signal (purple in Sean’s diagram) the answer is yes. Looking at the sequences of some muscle protein cDNAs done years ago, I found several stop codons in this region in each that I looked at. Of course they are without any function as the ribosome has already departed, but there is no reason for them not to be there on a random basis (and at apparently random positions).
Start and Stop Codons in pre-mRNA
By pre-mRNA I assume is meant the initial RNA transcript before splicing. There is no reason why an intron that is spliced out so that it does not appear in the final mRNA should not contain start or stop codons, which if not removed would have altered the reading frame. Many do.
A more interesting situation is where there is alternative splicing to produce different mRNA transcripts that encode different proteins. Pertinent to the question is the situation with start codons where the proteins differ in their N-termini, and stop codons where the proteins differ in their C-termini. The latter are quite common, but the former also occur, particularly were there are alternative forms of a protein, with and without a signal peptide7. In these cases the start or stop codons will be present in both of two exons, one of which is spliced out in each case. Examples of this from a couple of Drosophila genes are illustrated:
In both cases the coding regions are in orange, and the untranslated regions in grey. The thin lines represent the position of introns that have been spliced out. The arrowhead represents the direction of transcription and is at the C-terminal side of the protein. (i) Different N-termini for alternative products of gene vha14-2; (ii) Different C-termini for alternative products of gene vhaSFD.9