An orbital of a closed-shell Hartree-Fock calculation can only contain two elctrons, one with alpha ("up") and one with beta ("down") spin.
You are calculating the Helium-anion. In a minimal STO-3G basis, it comes with one(!) basis function (or, loosly speaking, "orbital"), as it has two electrons in its neutral form. The anion however has three electrons. So, one closed-shell orbital is certainly not enough to host three electrons. The calculation works for Lithium a.) because gaussian automatically assumes that you want to perform an unrestricted-calculation as the number of electrons is uneven (this is also the case for the helium-anion, but I am not sure if you are aware of this). Furthermore, Lithium has three electrons "by nature" (in its neutral state) and therefore, two basis-functions are enough to host four electrons.
You are using the STO-3G minimal basis, which adds one basis-function per two electrons of the system in the closed-shell case and one per electron in the open-shell case. Of importance is not the actual number of electrons in your system but the number of electrons the neutral atoms composing your system have. For a quick fix, substitute "hf/sto-3g" with "uhf/cc-pVDZ" and your calculation will run without issues.
Your system has an uneven number of electrons, so you cannot run a restricted (closed-shell) HF calculation. Either perform an unrestricted Hartree-Fock (UHF) or a restricted open-shell Hartree-Fock (ROHF) calculation. Both are available in gaussian09. As said, gaussian is already automatically doing this as a restricted HF calculation of a system with an uneven number of electrons is not possible.
Furthermore, as the calculation of the Helium-anion is quite fast, you should considering using a larger basis. You can start with "cc-pVDZ", a Dunning correlation-consistent basis-set, and move on from there.
For a comprehensive introduction to electronic-structure theory I suggest you read the book by Szabo and Ostlund. It assumes no previous mathematical or chemical knowledge above high-school-calculus und explains all the conceptual issues I have just mentioned in a detailed and understandible manner. Also, this book is quite affordable.
If this is too much, I suggest you at least look up the Roothan-Hall and the Pople-Nesbet equations for running an electronic structure calculation in a basis, as Gaussian builds on them.