In this paper, we study the open-loop equilibrium strategy for mean-variance portfolio selection problem under the assumption that the risk tolerance of the investor is a non-negative and nonlinear function of his/her wealth. We derive a sufficient and necessary condition for the existence and uniqueness of an open-loop equilibrium strategy via a coupled forward-backward stochastic differential equation. To the best of our knowledge, such an equation appears for the first time in the literature. The well-posedness of this equation is established by merely imposing Lipschitz condition
on the risk tolerance. We also present two examples with non-monotone risk tolerances, where some interesting findings are revealed and the equilibrium strategies are obtained explicitly and numerically.