Given an undirected graph G = (V, E) where each edge e = (i, j) has a length d_ij ≥ 0, the κ-minimum spanning tree problem, κ-MST for short, is to find a tree T in G which spans at least κ vertices and has minimum length l(T) = ∑_(i,j)ε T^d_ij. We investigate the computational complexity of the κ-minimum spanning tree problem in complete graphs when the distance matrix D = (d_ij) is graded, i.e., has increasing, respectively, decreasing rows, or increasing, respectively, decreasing columns, or both. We exactly characterize polynomially solvable and NP-complete variants, and thus, establish a sharp borderline between easy and difficult cases of the κ-MST problem on graded matrices. As a somewhat surprising result, we prove that the problem is polynomially solvable on graded matrices with decreasing rows, but NP-complete on graded matrices with increasing rows.