mean centering and scaling when doing partial least squares regression

Hi,

I am tryinmg to do partial least squares regression (PLSR) in Matlab. My question regards whether it's needed to mean-center and rescale the data before doing PLSR. I have rescaled the explanatory variables, X, by using the zscore function, and I get a very different result (and maybe wrong?) when don't rescale and mean-center. Does X need to be rescaled and if so why?

Furthermore, I am not sure whether Y needs to be mean centered and rescaled too. However, it doesn't seem to make a difference. Should Y be mean centered and rescaled?

Thanks!

Mads