The first-order accurate Discontinuous Galerkin Material Point Method (DGMPM), initially introduced by Renaud et al. [55], considers a solid body discretized by a collection of material points carrying the history of the matter, embedded in an arbitrary grid on which a nodal Discontinuous Galerkin approximation is defined, and that serves to solve balance equations. This method has been shown to be promising, especially for solving hyperbolic problems in finite deforming solids [57, 56]. The main goal of this research is to extend the first-order DGMPM to arbitrary highorder accurate approximations. This is performed by adapting the ADER (Arbitrary high order DErivative Riemann problem) approach [15] to the particular spatial discretization of the DGMPM. Firstly, the predictor step permits to design a particle-to-grid projection of arbitrary high order of accuracy, consistent with that of the nodal discontinuous Galerkin approximation defined on the arbitrary grid. This is performed using a Moving Least Square approximation for the ADER predictor field. Secondly, since the degrees of freedom of the predictor field are now defined at material points, the computation of the constitutive response of the material is ensured to be always performed at these material points. This is of crucial importance for history-dependent constitutive models because it avoids any diffusive transfer of internal variables on a new computational grid. Finally, a total Lagrangian formulation of equations is kept, which allows to precompute once and for all both the nodal discontinuous Galerkin approximation and that of the ADER predictor field, until the arbitrary grid is discarded if required. The method is illustrated on a few two-dimensional numerical examples, on which comparisons are shown with the ADER-DGFEM and Runge-Kutta-DGFEM.