Fractional integrals and Ap-weights: a sharp estimate