Literacy has strong similarities to synesthesia, and indeed is really a form of learned synesthesia. If you can read, you look at visual marks and automatically, involuntarily experience sound and meaning. Yes, these sense- and meaning-associations are initially learned, but they then become automatic. Even when you encounter marks that resemble characters in your language that are the chance result of natural processes - for example, a rock-arch that looks like the letter A - you can't look at it, and not think "A". (To understand what I mean by "automatic", go learn 10 characters from a writing system you don't currently know, and then find a website using that writing system. Understanding the characters is effortful, and by not concentrating, you can look at it, and not think of the sound or meaning.)
Since the serious study of synesthesia began with Francis Galton, it has also been noted that synesthesia runs in families, and that these families are enriched for artists and poets. This has led to the idea that the basis of synesthesia is some genetic influence resulting in insufficient cortical "pruning" in early life; extra fibers are left in areas like the fusiform or superior temporal gyrus, and this leads to color-grapheme or color-sound synesthesia. (An interesting implication is that infants and toddlers may actually all be synesthetes, prior to pruning.)
It stands to reason that if synesthetes are able to more highly associate sensory and meaning experiences, rates of dyslexia (if writing is a form of learned synesthesia) should be lower than in the general population. Doing a web search for this, I inadvertently found a synesthesia discussion forum where participants reported a higher-than-average rate of dyslexia. (Note, you won't have to rely on this dangerous foundation of anecdotal internet discussions for long; but in any event it was interesting that the possible correlation was the opposite of my expectations.)
Now along comes a new Ramachandran paper with David Brang (previously at UCSD, now at Northwestern) using made-up characters in varying colors. Grapheme-color synesthetes have a harder time learning new color-character associations than the rest of us. Extending to dyslexia, it's as if synesthetes' neuronal connections are richer but less trainable. Color-grapheme synesthetes report that it's unpleasant when real characters are printed in colors other than their "normal" synesthetic ones, much like they're constantly taking a Stroop test.