Multiple render targets: useful for deferred 3D lighting rendering, not sure what it could do in 2D. It just means being able to draw the same set of objects to multiple surfaces at once, which our engine doesn't need to do.
Antialiasing/multisampling: this is generally a massive performance hit for a sometimes imperceptible improvement in rendering quality. Since most 2D games use alpha-channelled textures instead of geometry, I don't think there's anything to gain from it.
Query objects: I guess they could return some performance stats or debug info, but I'm not clear in what way that would be useful.
Culling: already done in the engine before it issues draw calls. I don't think there's any point moving it to the GPU, since it means submitting many more draw calls, which would probably actually degrade performance.
In other words, WebGL 1 has almost everything a 2D engine needs already.