Given that
 - kD(k-1) = -[D(k-1) - (k-1)D(k-2)] )
for some

.
Set

to get
That red expression on the RHS looks very similar to the LHS.
Set

to get
Substitute (2) back into (1) to get:
Set

to get
Substitute (4) back into (3) to get:
If you look at (1), (3) and (5):
It's pretty clear what is happening on the RHS.
If you keep repeating this process, then for any finite

you will eventually get
So how many minus signs are on the RHS? Well let's look at those equations again:
That red number happens to be the same as the number of minus signs on the RHS.
Now
 = D(n-(n-2)))
so when we eventually reduce that term on the RHS to
 )
then there will be

minus signs on the RHS.
That is,
Note that the LHS contains
 )
and
 )
is not defined. Therefore we need

.