Mathematical models of the interactions between root systems and the surrounding soil provide valuable tools for understanding how plant roots influence soil water distribution and availability as well as for designing nature-based solutions to improve soil stability. This presentation introduces a novel optimization-based strategy for simulating the inherently challenging coupled problem of the interaction between a growing root system and the surrounding soil, focusing on the prediction of root water uptake. We address this problem by first rigorously reducing the coupled 3D Richards equation, governing unsaturated flow in the soil, and the 3D Stokes equation, describing water flow within the root xylem, to a well-posed 3D-1D formulation. Then, we cast the problem within a PDE-constrained optimization framework, enabling a direct and efficient quantification of root water uptake. To tackle the non-linearity of the Richards equation and its coupling with the 1D root variables, we propose a tailored iterative solving strategy designed for accuracy and computational efficiency. A key aspect of our work is the pioneering application of the Virtual Element Method (VEM) for the spatial discretization of the three-dimensional soil domain. VEM ability to handle general polyhedral meshes, including elements with aligned edges and faces as well as concave elements, significantly enhances the method capacity to address realistic scenarios, including complex geometries arising from the presence of natural obstacles in soil. Furthermore, our approach incorporates a dynamic model of root system architecture growth, driven by a discrete-hybrid tip-tracking strategy that accounts for key plant tropisms. Notably, we introduce a novel method for modeling negative thigmotropism, i.e. the natural tendency of plants to avoid obstacles, leveraging properties of the VEM basis functions. Several numerical experiments are provided, showcasing the accuracy of the method and demonstrating its applicability to realistic, large-scale simulations.