Not going to be that efficient, but does the job (just to_csv) after)
In [29]: df = DataFrame(dict(A = [1.5,2.5], B = [2.5,3.0]))
In [30]: df
Out[30]:
A B
0 1.5 2.5
1 2.5 3.0
In [31]: df.applymap(lambda x: str(x).replace('.',','))
Out[31]:
A B
0 1,5 2,5
1 2,5 3,0
This should be a bit faster (needs at least 0.12)
In [37]: df.applymap(str).replace(r'\.',',',regex=True)
Out[37]:
A B
0 1,5 2,5
1 2,5 3,0