Let $X=(X_t, t\geq 0)$ be a superprocess in a random environment described by a Gaussian noise $W^g=\{W^g(t,x), t\geq 0, x\in \mathbb{R}^d\}$ white in time and colored in space with correlation kernel $g(x,y)$. We show that when $d=1$, $X_t$ admits a jointly continuous density function $X_t(x)$ that is a unique in law solution to a stochastic partial differential equation \begin{align*} \frac{\partial }{\partial t}X_t(x)=\fracΔ{2} X_t(x)+\sqrt{X_t(x)} \dot{V}(t,x)+X_t(x)\dot{W}^g(t, x) , \quad X_t(x)\geq 0, \end{align*} where $V=\{V(t,x), t\geq 0, x\in \mathbb{R}\}$ is a space-time white noise and is orthogonal with $W^g$. When $d\geq 2$, we prove that $X_t$ is singular and hence density does not exist.