Without writing code (don't have the time right now), I can write out the steps for you. However, be warned that although you are accounting for correlation between variables by using principle components, you are sacrificing interpretability of the results, as it's often difficult to interpret what each component represents.
1. Extract predictor variable values for each point (raster package function "extract")
2. Run PCA on the matrix from above ("prcomp", etc.).
3. Use the rotated data values in the output ($x) as the predictor variable inputs to Maxent, or any other SDM.
Jamie Kass
PhD Candidate
City College, NYC