Long PCR & Joining 5' of sequence to the rest

[Harry Witchel]

Hello Hello Molecular Explorers --
   I have recently cloned out by PCR the 5' end (300 bases) to a K+ 
channel which I have already found the rest (3.2 kb) of by library 
screening.  I am having difficulty in putting the two halves of the 
sequence together so that I can express it, and I think my problem is 
that my PCR is too long for my methodology.
   The DNA sequence is very GC rich (overall 66% GC).  There is a 26 bp 
overlap of the two halves of the sequence.  There are no convenient 6 
cutter restriction sites, even if I were to change one of the wobble 
bases.  Both sequences are in vectors in which the restriction sites 
would leave one extra unmatching base on the cut out fragment.
   So far I have tried to rePCR the entire gene from my original cDNA 
source without success.  I have recently attempted to PCR the two 
fragments together in a two stage process using the two fragments as 
their own primers, ie:

Set up PCR w/o primers, but with both templates and dNTPs
Run 7-10 cycles, letting each fragment act as a primer for the other 
fragment (they begin at overlapping regions).  This should create a small 
number of full length transcripts.
Dope reaction with proper primers related to 5' and 3' ends of the full 
length gene, and amplify.

So far, this has not worked.  However, I have also been unable to simply 
amplify the longer fragment (3.2 kb) just using primers to that fragment. 
 The short 5' end amplifies well with either Taq or Pfu.  I have been 
wondering about whether I am having problems simply because the 3' end is 
very long for PCR.  I currently am trying various combinations of enzymes 
(Taq, Pfu, Taq+Pfu), longer cycles (72 degrees for 10 minutes), and more 
cycles (50 cycles).  I am wondering if I should up my amount of enzyme (I 
normally use .25-0.5 units Taq or Pfu per 50 ul rxn), up my primers (12 
pM per 50 ul rxn) or lengthen my primers (currently they are 25-30 bp, 
very GC rich).  Currenly when I use cut plasmid as templates, at 30 
cycles I have nothing at all (not even primer dimer) and at 40 cycles I 
start getting nonspecific short bands.
   Any suggestions on either PCR of long fragments or ligating together 
two overlapping fragments will be appreciated.
   Thanks,
   Harry