VAR estimation with variables integrated of different orders

[EktaS]

I have to use macroeconomic model to forecast Exchange rate using VAR. My aim is to capture the residuals/innovations by regressing exchange rate on macroeconomic variables using VAR.
I have six variables. Two variables are I(0), two are I(1), and two are I(2). Exchange rate is also I(1). Now before applying VAR, i have to test for cointegration johnsan test for which all the series should be of same order.
Plz suggest in this case how should i apply the johnson test and how the VAR would be applied. It is necessary to have all the variables of same order to apply VAR.
Also it is necessary to conduct granger causality test before applying VAR.