Statistical Inference — sufficient statistic for a gamma shape parameter

Let $x_1, x_2, x_3, …, x_n$ be a random sample from the gamma distribution having pdf $f(x; \theta, \mu) = \frac{1}{\Gamma(\theta)\mu^\theta}e^{-\frac{x}{\mu}}x^{\theta - 1}$; $x > 0$, $\theta > 0$ and $\mu > 0$ is known constant. What is the sufficient statistic for $\theta$?