I think I would explain it this way. Due to Nyquist theory, the highest frequency you can detect is half your sample rate (must have at least 2 points in one period). The lowest is obviously 0. The assumption here is that the detectable frequencies increase linearly from 0 to srate/2.
If you inspect the peaks in your plot, you'll notice this approach is an approximation, as the peaks are about 1 point off.