Motivated by large domain problems, there has been significant progress in estimating various properties of discrete distributions from samples. For a number of emph{symmetric properties}, we show that a natural MLE approach, called Profile Maximum Likelihood achieves the optimal sample complexity.