"Proof by contradiction" is, I take it, another label for the Reductio rule which can helpfully be displayed as
$$\quad\quad | \quad A$$
$$\quad\quad | \quad \vdots$$
$$\quad\quad | \quad C$$
$$\quad\quad | \quad \vdots$$
$$\quad\quad | \quad \neg C$$
$$\neg A$$
or (in another formulation)
$$\quad\quad | \quad A$$
$$\quad\quad | \quad \vdots$$
$$\quad\quad | \quad \bot$$
$$\neg A$$
When $A$ is a temporary assumption, which (via some subproof) leads to an explicit contradition or an absurdity $\bot$, we are allowed to discharge that temporary assumption and conclude (from whatever other premisses are in play) that it must be false, $\neg A$.
This is a valid rule of inference in systems of logic which lack a conditional (and even where a conditional can't be defined). So it is unhelpful -- to say the least -- to say that is "based on" a tautology involving a conditional (or on modus tolens).
If you do have reductio and modus ponens that modus tollens will be a derived rule:
$$A \to C$$
$$\neg C$$
$$\quad\quad | \quad A$$
$$\quad\quad | \quad C$$
$$\quad\quad | \quad \bot$$
$$\neg A$$
And conversely, if you have a conditional proof rule for introducing conditionals, modus tollens, and the assumption $\neg\bot$ then you could get reductio as a derived rule.
But, to repeat, it would be wrong to say that the result that (with a bit of help) you can get reductio from modus tollens is what "really" underlies reductio. Reductio is a warranted inferential rule because of the meaning of negation, not (even in part) because of the meaning of the conditional.
That, as they say, is the take-home message!