A matrix $A$ is reducible if there is a permutation matrix $P$ such that
$$\tag{1}
P^TAP=\begin{bmatrix}A_{11} & A_{12} \\ 0 & A_{22}\end{bmatrix}.
$$
If $A$ would have a zero row, then it could be permuted to
$$
P^TAP=\begin{bmatrix}A_{11} & A_{12} \\ 0 & 0_{1\times 1}\end{bmatrix},
$$
by some $P$, which is (1) with $A_{22}=0$. Similarly, if $A$ has a zero column, it can be permuted to
$$
P^TAP=\begin{bmatrix}0_{1\times 1} & A_{12} \\ 0 & A_{22}\end{bmatrix},
$$
which is (1) with $A_{11}=0$.
Hence, if $A$ has a zero row/column, it is reducible. Reversing this implication shows that an irreducible matrix cannot have a zero row and/or column.