Mutations aren't passed on as an evolutionary progression would be.
It's a random event that happened to happen, and the chances of it being passed on are exactly the same as it happening in the first place.
And as a random event, the mutation may be just as likely a step backwards, from an evolutionary standpoint.
And e. coli, is still just e. coli. If your theory was fact, e. coli'd be walkin and talkin, by now.
It works like dog breeding.
Every wolf pup is just ever so slightly different along various dimensions from his litter mates. If you are in the practice of keeping wolves (dogs, per se, not yet existing), you would likely disadvantage the reproduction of those wolves that tended to bite you or your family members by culling them early. You might also cull the wolves who prove not interested in defending your family, and run off at the first sign of danger. Conversely, those wolves who seem particularly interested in helping you on the hunt would be prized, and you'd likely make sure they mated often, were particularly well cared for, and produced pups with your other wolves that are good at that. The result is that you will tend, in time, to have wolf pups who don't bite you or your family members, who consider themselves members of your pack, and will thus stand and defend you and your family against danger. You will also tend to have wolves who are good at assisting in the hunt. Keep doing that sort of thing long enough, and you will end up with hundreds of very distinct breeds of wolf (as we have today, although we call them dogs), of very different sizes, physical characteristics, and behavioral traits.
That's just man acting on the natural, DNA-driven, variability of the wolf over a relatively short period of time. Imagine what nature can do, acting over immensely longer periods of time.
But they are all wolves and a Chihuahua could breed with a Dire Wolf, there were any left, and produce viable offspring.
