I finally managed to proof this variant of the Dominated Convergence Theorem:
Theorem (Variant of Dominated Convergence Theorem). Let $f, f_k: X \to \overline{\mathbb R}$ be $\mu$-measurable, $g, g_k: X \to \overline{\mathbb R}$ be $\mu$-summable, $f_k \to f$ $\mu$-a.e., $g_k \to g$ $\mu$-a.e., $\vert f_k \vert \leq g_k$ for each $k \in \mathbb N$, $$ \lim_{k \to \infty} \int_X g_k \, \mathrm d\mu = \int_X g \, \mathrm d\mu \; .$$ Then $$ \lim_{k \to \infty} \int_X \vert f_k - f \vert \, \mathrm d\mu = 0 $$ holds.
Here is what I did: I first noted that $f_k, f \in L^1(X, \mu)$. Using Fatou's Lemma, I could show that $$ \limsup_{k \to \infty} \int_X f_k \, \mathrm d\mu \leq \int_X f \, \mathrm d\mu \leq \liminf \int_X f_k \, \mathrm d\mu \; , $$ so $\Vert f_k \Vert_{L^1} \to \Vert f \Vert_{L^1}$. Using this and some continuity properties of the integral, and Egorov's theorem, I could finally conclude that $f_k \to f$ in $L^1(X, \mu)$. I wanted to ask, if somebody knows a more direct way to show this result?