The question in your title is different than the question you ask at the end of your post. So I will answer them both.
Firstly, it's not true that polynomials with rational coefficients have integer roots. For instance, $x^2 - 2$ has roots $\pm \sqrt 2$.
However, it is true that polynomials of the form $p(x) = a_0 + a_1 {X \choose 1} + \ldots + a_d {X \choose d}$, where $a_i \in \mathbb{Z}$, only take integer values. More generally, the monomial $a_k {X \choose k}$ only takes integer values when $X$ is an integer. This is equivalent to the denominator, which in this case is $k!$, dividing $X(X-1)(X-2)\dots(X-k+1)$ for any $X \in \mathbb{Z}$.
This last claim admits number theoretic and combinatorial proofs. Number theoretically, you might look at each prime in $k!$ and show that its multiplicity in the numerator is at least the multiplicity in the denominator. This is doable, but not particularly easy. Bill showed a completely arithmetic approach in his answer here.
Combinatorially, you simply notice that for integral $X$, the coefficient ${X \choose k}$ is the classical combination, and counts the number of ways of choosing $k$ things from $X$ choices, and is therefore an integer.
Regardless, as each $a_k {X \choose k}$ is an integer for each $X \in \mathbb{Z}$, you must have that polynomials formed from these monomomials also take integral values on the integers.