For the 5' bias in base content: check this answer and the blog post linked to there.

Additionally to the 5' bias, you can see 3' bias in base content. If this bias is gradual I'd say adapter contamination and you should trim your data. In this case the bias is abrupt and pertains to the last base. This could be observed if your data are already trimmed? When you trim adapters with high stringency, all the reads ending in the first base of the adapter (e.g. "A") will have that base removed, therefore none will end in "A".