PDEs & numerical analysis (the latter of which depends heavily on linear algebra) for physics & engineering.
Algebra (including algebraic number theory & all things combinatorics-related) for computer science.
They're kind of all coupled to physics in the end. I mean you also need statistics & probability theory for physics, too. And engineering. And just like...stuff.
So I guess the answer is, "all of them except topology," because topology's just gay, who cares about shapes.
Topology crops up more than one would think. At an elementary level, the intermediate and extreme value theorems in calculus are just special cases of the fact that continuous maps preserve connectedness and compactness. Functional analysis, which is the theoretical underpinning for PDEs, employs multiple topologies (e.g. norm, weak, weak*) to address the different flavors of limit and convergence.
However, when people use the term "topology" colloquially, I think they're probably referring to deforming one shape into another (e.g. saying that a coffee cup and a donut are somehow equivalent). This branch, which could be broadly described as algebraic topology or homotopy theory, is quite far removed from any kind of "applicable" mathematics. Even so, you can see shades of it when you're doing vector calculus. For example, since the curl of the gradient is zero, you know that conservative 3d vector fields necessarily have zero curl. The question of whether a zero curl vector field is conservative is a much more subtle issue that depends on the topology of the nonsingular locus of the vector field.
For example, if a vector field is infinitely differentiable on the whole of real three space, then it will be conservative if and only its curl is zero. This is because real three space can be deformed to a single point, which is homotopically trivial.
On the other hand, consider F(x,y,z) = [-y/(x^2+y^2), x/(x^2+y^2), 0]. If you mechanically compute its curl, you can verify that it is indeed zero. On the other hand, if you integrate F along one counterclockwise turn of the unit circle in the xy-plane, you'll get 2*pi, so F is not conservative. The nonsingular locus of F is all of 3-space with the z-axis removed, which is homotopic to the unit circle. This space has a nontrivial first cohomology group, which can be thought of as an "obstruction" to F being conservative.