The idea is to define connected sum along the coordinate balls, i.e. the balls that are mapped via coorinate chats to standard balls in $\mathbb R^n$. For such balls the argument is easy, and no annulus theorem is needed. For notational convinience let's write a coordinate ball as $B_x$ where $x$ is the point in $M$ that cooresponds to the center of the corresponding ball in $\mathbb R^n$. Fix $y\in M$ and consider the subset $Y$ of $M$ consising consisting of points $x$ such that there is a homeomorphism of $M$ taking $B_x$ to $B_y$. It is easy to show that $Y$ is open and closed (the point is that given two metric balls in $\mathbb R^n$ there is a homeomorphism that maps one ball into the other one, and is the identity outside a compact set; such a homeomorphism can be constructed with bare hands, and this is where the work is). Thus if $M$ is connected, then $Y=M$. The same argument shows that any two coordinate balls are ambiently isotopic.
The idea is to define connected sum along the coordinate balls, i.e. the balls that are mapped via coorinate chats to standard balls in $\mathbb R^n$. For such balls the argument is easy, and no annulus theorem is needed. For notational convinience let's write a coordinate ball as $B_x$ where $x$ is the point in $M$ that cooresponds to the center of the corresponding ball in $\mathbb R^n$. Fix $y\in M$ and consider the subset $Y$ of $M$ consising of points $x$ such that there is a homeomorphism of $M$ taking $B_x$ to $B_y$. It is easy to show that $Y$ is open and closed (the point is that given two metric balls in $\mathbb R^n$ there is a homeomorphism that maps one ball into the other one, and is the identity outside a compact set; such a homeomorphism can be constructed with bare hands, and this is where the work is). Thus if $M$ is connected, then $Y=M$. The same argument shows that any two coordinate balls are ambiently isotopic.