The question is what transformation to use in order to make the distributions normal?
Just looking at a pairplot isn't enough to suggest changes, in my opinion. Moreover, what are you trying to test? What is your null hypothesis, what is the alternative hypothesis? What kind of data is it, how neat was your collection process? Are there any co-variables?
With the info provided, all I can suggest is to perform parametric and non-parametric tests. May try Box-Cox transformation!