WM and CSF masks after denoising

I’ve performed nonaggressive denoising on the *space-T1w_desc-preproc_bold.nii.gz and now I’d like to re-calculate the WM and CSF signals from the *_space-T1w_desc-AROMAnonaggr_bold.nii.gz. Do the WM and CSF ROIs need to be in MNI space, or can I calculate mean signal from native bold time series (and use native anatomical space masks)?
Also, in fmriprep is ICA-AROMA run after smoothing?