This is a wild guess, so sorry if its unhelpful
Make up a new variable x_const with value 1 everywhere. Then run your SUR as before, but add x_const as a depvar to each equation.
Make sure you run with the noconstant option enabled. The constant in the new equations will be the "slope paramter" of the x_const variable.
This means the slope coefficient on the x_const variable will behave in the same way as the intercept term from your old equation. I think You can use constraints on that slope parameter in the 4 equations to get the output you want.
The constraint you want would be something like:
[E_share]x_const + [K_share]x_const +[L_share]x_const+[M_share]x_const = 1
An example of a sureg with constaints is here
Stata 11 help for estimation options