Most DNA sequences are poly-functional and so must also be poly-constrained. This means that DNA sequences have meaning on several different levels (poly-functional) and each level of meaning limits possible future change (poly-constrained).  For example, imagine a sentence which has a very specific message in its normal form but with an equally coherent message when read backwards. Now let's suppose that it also has a third message when reading every other letter, and a forth message when a simple encryption program is used to translate it. Such a message would be poly-functional and poly-constrained. We know that misspellings in a normal sentence will not normally improve the message, but at least this would be possible. However, a poly-constrained message is fascinating, in that it cannot be improved. It can only degenerate. Any misspellings which might possible improve the normal sentence from will be disruptive to the other levels of information. Any change at all will diminish total information with absolute certainty.

There is abundant evidence that most DNA sequences are poly-functional, and are, therefore, poly-constrained. This fact has been extensively demonstrated by Trifonov (1989). For example, most human coding sequences encode for two different RNAs that read in opposite directions (i.e., both DNA strands are transcribed Yelin et al., 2003). Some sequences encode for different proteins, depending on where translation is initiated and where the reading frame begins (i.e., read-through proteins). Some sequences encode for different proteins based upon alternate mRNA splicing. Some sequences serve multiple functions simultaneously (i.e., as a protein-coding sequence and as an internal transcriptional promoter). Some sequences encode for bot a protein coding region and a protein-binding region. Alu elements and origins of replication can be found within functional promoters and within exons. Basically all DNA  sequences are constrained by isochore requirements (regional GC content), “word” content (species-specific profiles of di-, tri- and tetra-nucleotide frequencies), and nucleosome binding sites (because all DNA must condense). Selective condensation is clearly implicated in gene regulation and selective nucleosome binding is controlled by specific DNA sequence patterns that must permeate the entire genome. Lastly, probably all sequences also affect general spacing and DNA folding/architecture, which is clearly sequence dependent. To explain the incredible amound of information which must somehow be packed into the genome (given the extreme complexity of life), we really have to assume that there are even higher levels of organization and information encrypted within the genome. For example, we know there is another whole level of organization at the epigenetic level (Gibbs, 2003). There also appears to be extensive, sequence-dependent, three-dimensional organization within chromosomes and with the whole nucleus (Manuelidis, 1990; Gardiner, 1995; Flam, 1994). Trifonov (1989) has shown that probably all DNA sequences in the genome encrypt multiple codes (up to 12).    

John Sanford, Genetic Entropy  2008  p.131-2