It's NOT so much "how" as "why". I presume that just before the section you quoted they showed that the derivative of [itex]e^x[/itex], for e any number, "[itex]M(e)e^x[/itex]" where M(e) is a number depending on e only, not on x.
It is not too difficult to show that M(2) is less than one and that M(3) is larger than 1 (by numerically approximating the limits). "Choosing" M(e) to be 1 is really choosing a specific value of e such that M(e)= 1. My point about M(2) and M(3) is that there is such a value of e, between 2 and 3. You could, then, by a succesion of numerical approximations, show that M(2.7) is less than 1 but that M(2.8) is greater than 1 so "e" is between 2.7 and 2.8. Or that M(2.71) is less than 1 but that M(2.72) is greater than one so that "e" is between 2.71 and 2.72, etc.
Choosing "e" to be the number such that M(e)= 1 means that the derivative of the function [itex]y= e^x[/itex] is just [itex]e^x[/itex] again.