First of all, be aware that Granger causality is not the same as the common concept of causality used in Econometrics or policy evaluation.

Basically, in the equation of interest, with the appropiate number lags of the left-hand-side variable (y), one should include leads of the right-hand-side (x; that is, xt+1, xt+2...). One would expect that those leads are not relevant if there is not Granger causality from x to y.

Although the context is different, given that the explanation is clear, you might want to have a look at Mostly Harmless Econometrics (Angrist & Pischke, 2008, chapter 5).

Thank you so much for your kind help