Improving the yield of carbohydrate to lipid conversion and lipid productivity are two critical goals to develop an economically feasible process to commercialize microbial oils. Lignocellulosic sugars are potential low-cost carbon sources for this process but their use is limited by the toxic compounds produced during biomass pretreatment at high solids loading, and by the pentose sugars (mainly xylose) which are not efficiently metabolized by many microorganisms. Adaptive laboratory evolution was used to select a Rhodosporidium toruloides strain with robust growth in non-detoxified wheat straw hydrolysates, produced at 20% solids loading, and better xylose consumption rate. An arabinose-inducible cre-lox recombination system was developed in this evolved strain that was further engineered to express a second copy of the native DGAT1 and SCD1 genes under control of the native xylose reductase (XYL1) promoter. Fed-batch cultivation of the engineered strain in 7-L bioreactors produced 39.5 g lipid/L at a rate of 0.334 g/Lh(-1) and 0.179 g/g yield, the best results reported in R. toruloides with non-detoxified lignocellulosic hydrolysates to date.