# Justification for transforming explanatory variables

I am using linear and generalised linear models, and have transformed my explanatory variables using $log10(\bullet)$ and $sqrt(\bullet)$ transformations, and my response variable using an arcsine square root transformation ($\arcsin(\mathrm{sgn}(x)\sqrt{|x|})$ as I had negative values in $x$). For the latter, the justification is to get the data normally distributed.

What is the justification (or point!) of transforming explanatory variables, as they do not need to be normally distributed?