Consider $P = \{p_1, \ldots,p_l\}$ the set of numbers coprime to $n$, and set $P^{-1} = \{p_j^{-1}\}$ of inverses in $\Bbb Z_n$, that means: $$ p_j^{-1}p_j \equiv 1 \pmod n $$ I have proved that $P = P^{-1}$ as sets. Suppose that exists $q \in P^{-1}$ which is inverse to some $p_j \in P$: $$qp_j \equiv 1$$ such that $q$ isn't coprime with $n$, then exists non-zero $k \in \Bbb Z_n$ for which we have: $$ qk \equiv 0$$ that implies $$k \equiv (p_jq)k = p_jqk = p_j(qk) \equiv 0 $$ Contradiction.
I am interested in more interesting proof of this fact, may be something theoretical-set, combinatorial, etc...
If $q p \equiv 1 \pmod n$ then $qp =1+nk$, so $qp-nk=1$. But $\gcd(q,n)$ divides every lineal combination of $q$ and $n$, so $\gcd(q,n) | qp-nk=1 $.