About the topology of manifolds we have
paracompactness $\iff$ partitions of unity $\iff$ every component has a countable basis $\iff$ ...
Not hard to prove.
About differentiable structures, it is more convenient to talk about diffeomorphisms.
Here are the facts (classic results from the $50$'s) -- (assume paracompact to be safe)
Let $M$ a $C^k$ manifold ($k\ge 1$). Then there exists a $C^{\omega}$ manifold ( real analytic) $\tilde M$ and a $C^k$ diffeomorphism $M \simeq \tilde M$ ( this is equivalent to: there exists a $C^{\omega}$ atlas on $M$ which is $C^k$-compatible with the given $C^k$ atlas.
Let $1\le k
\le l$ $M$, $M'$ two $C^l$ manifolds that are isomorphic as $C^k$ manifolds. Then they are isomorphic as $C^l$ manifolds. From the previous statement only the case $k=1$, $l=\omega$ has to be considered.
Whitney showed that every differentiable manifold is a submanifold of some $\mathbb{R}^N$ with image real analytic. This implies the existence of a real analytic structures.
John Nash ( of a beautiful mind) proved that every Riemannian manifold can be imbedded in some $\mathbb{R}^n$ with image a real analytic submanifold.
The previous results do not work for $k=0$:
There exists a topological ($C^0$ ) manifold that is not homeomorphic to a ($C^1$) differentiable manifold ( Kervaire).
There exist non-diffeomorphic differetiable manifolds that are homeomorphic (Milnor).
See also exotic $\mathbb{R}^4$.