Here is a different way to reach the same transform, by first writing the solution in real space as a convolution and then using the convolution theorem, instead of transforming the PDE directly.
The general solution of the heat equation $u_t=ku_{xx}$ on the whole real line, with initial data $u(x,0)=g(x)$, is the convolution of $g$ with the heat kernel:
\[ u(x,t) = \int_{-\infty}^{\infty} g(y)\, K(x-y,t)\, dy, \qquad K(x,t)=\dfrac{1}{\sqrt{4\pi kt}}\,e^{-x^2/(4kt)} \]Taking the Fourier transform of a convolution turns it into a product of the two individual transforms, so
\[ \hat u(w,t) = \hat g(w)\cdot \hat K(w,t) \]The Fourier transform of the heat kernel $K(x,t)$, with the convention $\hat f(w)=\int f(x)e^{iwx}dx$, works out to $\hat K(w,t)=e^{-kw^2t}$; this is the standard fact that a Gaussian in $x$ transforms into a decaying exponential in $w$.
Now find $\hat g(w)$, the transform of the initial data $g(x)=e^{-a|x|}$. Split the integral at $x=0$:
\[ \hat g(w) = \int_0^\infty e^{-ax}e^{iwx}dx + \int_{-\infty}^0 e^{ax}e^{iwx}dx = \dfrac{1}{a-iw}+\dfrac{1}{a+iw} = \dfrac{2a}{a^2+w^2} \]Multiplying the two pieces together:
\[ \hat u(w,t) = \dfrac{2a}{a^2+w^2}\cdot e^{-kw^2t} \]Let's check this against the other options. Option B is missing the factor of $2$ that comes from adding the two halves of the integral for $g$. Option C puts $a^2$ in the exponent instead of $w^2$, but the exponential decay in time must depend on $w$ (the Fourier variable dual to $x$), not on $a$ (which only measures how fast the initial condition itself decays). Option D is just $\hat K(w,t)$ alone, the transform for a point source (delta function) initial condition; it has no $a$ in it at all, so it cannot be the right answer once the actual initial data $e^{-a|x|}$ is folded in.
Let's summarize:
So the correct expression is $\hat u(w,t) = \dfrac{2a}{a^2+w^2}e^{-kw^2t}$.
\[ \boxed{\hat u(w,t)=\dfrac{2a}{a^2+w^2}\,e^{-kw^2t}} \]