[en] It is suggested that the usual path integral representation of Euclidean vacuum amplitude (tunneling amplitude) in QCD must be supplemented by the explicit boundary condition corresponding to the spontaneous breaking of chiral SU(N) x SU(N). Adopting the trial wave function introduced by Nambu and Jona-Lasinio, one sees that such a path integral automatically breaks also the additional chiral U(1) symmetry of massless quarks. The catastrophe of semi-classical approach to QCD and 'U(1) problem' would be avoided in this way and one has, in principle, a better starting point for the self-consistent calculation