The presence of unknown mutual coupling between array elements is known to significantly degrade the performance of most high-resolution direction of arrival (DOA) estimation algorithms. In this paper, a robust subspace-based DOA estimation and array auto-calibration algorithm is proposed for uniformly linear array (ULA), when the array mutual coupling is present. Based on a banded symmetric Toeplitz matrix model for the mutual coupling of ULA, the algorithm provides an accurate and high-resolution DOA estimate without any knowledge of the array mutual couplings. Moreover, a favorable estimate of mutual coupling matrix can also be achieved simultaneously for array auto-calibration. The algorithm is realized just via one-dimensional search or polynomial rooting, with no multidimensional nonlinear search or convergence burden involved. The problem of parameter ambiguity, statistically consistence and efficiency of the new estimator are also analyzed. Monte-Carlo simulation results are also provided to demonstrate the effectiveness and behavior of the proposed algorithm.