A single non-zero coefficient is not sufficient for any n>1 as the only simple polynomial functions with a single non-zero coefficient are P(x)=xn and P(x)=−xn but in both cases n∤P(1). Let us show that the values of the polynomial function Pn(x)=xn−xn−φ(n) at all integral arguments are divisible by n. (Here φ is the Euler's totient function.) This shows that having 2 non-zero coefficients is sufficient.
Let k be an integer. Let the canonical form of n be p1α1⋯pmαm and let us assume without loss of generality that k is divisible by primes p1,…,pl and is not divisible by primes pl+1,…,pm. Define u=p1α1⋯plαl and v=pl+1αl+1⋯pmαm. Let us now show that u∣kn−φ(n) and v∣kφ(n)−1. Having uv=n, we can conclude that n∣Pn(k) as Pn(k)=kn−kn−φ(n)=kn−φ(n)(kφ(n)−1).
To prove that u∣kn−φ(n), it is sufficient to prove for all i=1,…,l that piαi∣kn−φ(n). It is sufficient to prove that αi≤n−φ(n), as by the assumption pi∣k. Inequality αi≤n−φ(n) holds as pi, pi2,…,piαi are αi positive integers which are not greater than n and not coprime with n.
To prove the statement v∣kφ(n)−1, we derive from Euler's theorem that v∣kφ(v)−1 as k and v are coprime. Also, u and v are coprime, therefore, φ(n)=φ(uv)=φ(u)φ(v) from which kφ(v)−1∣kφ(n)−1. Consequently, v∣kφ(n)−1.