First let's express $\neg, \lor$ and $\land$ using $\downarrow$, where $(P\downarrow Q) \iff (\neg P\land \neg Q)$:
$$\begin{align}
\neg P &:= (P\downarrow P) \\
(P\lor Q) &:= \neg(P\downarrow Q) \\
(P\land Q) &:= \neg(\neg P\lor \neg Q). \\
\end{align}$$
Now it's easy to define $\leftrightarrow$:
$$
(P \leftrightarrow Q) := (P\land Q) \lor (\neg P \land \neg Q).\tag{$Equiv_1$}
$$
An equivalent definition is:
$$(P \leftrightarrow Q) := (\neg P\lor Q)\land(P\lor \neg Q).\tag{$Equiv_2$}
$$
Eliminating the defined symbols $\neg, \lor$ and $\land$ in ($Equiv_1$) or ($Equiv_2$) is left as a mechanical exercise. (If you need to do so, it looks like ($Equiv_2$) is the easier definition to work with — it has only one $\land$ not two, so the resulting formula should be shorter than what you'd get starting from ($Equiv_1$).)