Hadn't someone already mentioned Descartes' Rule of Signs? Maybe I imagined it. Anyway ...
Take $k$ and $c$ non-zero (the zero cases are trivial), and name the polynomial $p(t)$.
* The coefficient sequence of $p(t)$ is $(c, k, −c)$, which has one sign change ($k$ matches sign with either $c$ or $−c$). Thus, by the Rule of Signs, $p$ has exactly one positive real root.
* The coefficient sequence for $p(−t)$ is $(c, −k, −c)$, which also has one sign change, so that $p$ has exactly one negative root.
Since zero isn't a root, and since we've run out of real candidates, the remaining roots must be non-real.