State-of-the-art LES models have made significant progress in the prediction of heat release and flame propagation for a wide range of combustion regimes. Predicting pollutants remains however a difficult issue, especially in situations where sub-grid scale wrinkling is high. A novel strategy, named COPLES (Chemistry OPtimized for LES), is presented in this paper to account for the effect of sub-grid scale wrinkling on pollutants formation in a transported chemistry context and under flamelet regime assumption. New chemical mechanisms are built through an optimization process to recover the properties of filtered flames. Two alternative formulations are proposed. The first one consists in optimizing the COPLES parameters to retrieve the structure of Filtered Planar Flame (FPF) and to include separately the impact of sub-grid scale wrinkling. The second formulation targets directly a collection of Filtered Wrinkled Flamelets (FWF), so that the influence of turbulence on the chemical flame structure is intrinsically accounted for. The two approaches are successfully validated on 1-D freely propagating premixed flames and then challenged against the standard Thickened Flame for LES model on a turbulent premixed swirled combustor, experimented at Cambridge University. Whereas all formulations retrieve the temperature field, the COPLES-FWF model improves significantly the prediction of CO formation.