Researchers in Texas (1) looked at the cross-training effects of running, biking and swimming, and how they affected the VO2 max (the standard test of cardio fitness).
While they found that excessive cross-training negatively effects performance (i.e. sport specific training is still the best way to go for the bulk of your training), they also discovered that cycling can improve, or at least maintain, non-cycling sports performance for both speed and endurance events.
When comparing running, cycling and swimming, they found that there is a transfer of training effects (VO2max) from one to another, and that running had the greatest positive transference and swimming the least, with cycling in the middle.
Overall, there was evidence that there is evidence that under certain conditions, cycling can help to improve (or at least maintain) non-cycling sports performance for both endurance and speed activities.
Similarly, researchers in California (2) found that when competitive runners replaced 50% of their off-season running with cycling, their race times declined only 2% more than those who continued to run 100% of their training time. Furthermore, there was no decline in VO2 max.
Researchers concluded that this would help athletes in the long term to recover from a tough racing season, and would also give them a break mentally from their usual training regime.
Because cycling can produce the same metabolic stress on the body as running, but without the loading stresses on the muscles, it offers a long term way for runners to protect their bodies.
My own experience with numerous athletes is that cycling also provides a greater measure of protection from knee damage, by exercising the knee joint without weight bearing loads upon it, and also by strengthening the muscles around the knee.
In fact many runners (myself included) have recovered from long-term knee injuries after taking up cycling, and are able to stay relatively injury free so long as they continue to include cycling as a regular part of their training program.