Increasing genomic or organismal complexity
The case where we increase genomic complexity without a concomitant increase in organismal complexity is fairly elementary. An insertion mutation can add in new "junk" DNA to an intergenic region, or an intron that has no effect on phenotype. Likewise, a duplication mutation (a type of insertion) could duplicate a gene that simply does the exact same job as an existing gene. In both cases, genomic complexity is increased while organismal complexity stays the same.
Increasing organismal complexity without increasing genomic complexity is more interesting. In this case, a point mutation or a deletion affects cis regulatory regions or coding regions in a way that increases the function or expression of a gene product. For example, imagine amylase (an enzyme that digests amylose starches), which is normally expressed in the small intestine. Local activators are present all over the digestive system (including the stomach, mouth, and intestines), but a switch regulated by an signalling pathway is only activated in the small intestine. A series of point mutations in the cis-regulatory regions of amylase now create a novel local activator response element (Figure 2). In a way, this could be seen as an increase in genomic complexity because a new binding site was added, but seen from another point of view, the overall number of nucleotides remains the same. This novel cis regulatory region turns on amylase in the mouth, so now salivary glands produce amylase and starch digestion can begin earlier, diversifying the diet of the animal. In this case, the species has maintained the number of genes and the number of nucleotides in the genome but organismal complexity has increased due to an new cis regulatorysite (figure 2).
Exonic (coding) mutations can also increase complexity, without adding genes or nucleotides to the genome. For example, imagine a mammalian protein with two domains - a functional domain and a repressor domain. The repressor domain can be ubiquitinated to inhibit the protein under particular conditions. In this case, our protein is a leg organizer. The repressor domain is bound by Hox genes of the trunk, limiting leg organizer expression to the region anterior to the thoracic and posterior to the lumbar regions of the body. Along with other limits to expression, this results in an animal with two pairs of legs. A point mutation that converted a protein-coding codon to a stop codon (a nonsense mutation), at the end of the functional domain, would result in a protein missing the repressor domain. This would cause ectopic activation of leg organizer activity and (potentially) additional legs (Figure 2).
Increasing organismal and genomic complexity
Genomic complexity can be added, at the small or large-scale, and it can occur over cis-regulatory or coding regions. It can involve horizontal transfer from another genome or duplication of elements already present in the genome. Here, we will consider different ways to add genetic complexity and how they might affect organismal complexity. I will focus on duplication rather than horizontal transfer for simplicity but similar consequences can occur with horizontal transfer as well. Smaller scale duplications can occur via errors in DNA replication, errors in meiotic crossing over, or transposon insertion. Larger scale duplications can occur via errors in meiotic crossing over, transposon insertion, or whole chromosome duplication via missegregation of chromosomes during cell division.
Small scale duplications can duplicate a single functional domain (like a bHLH domain), a single coding sequence, a single gene, or a single cis-regulatory protein binding site. In the case of the cis-regulatory binding site, the initial mutation might increase or decrease expression in target cells. Additional point mutations in the binding site could optimize for binding of a related protein. This would have the effect of driving ectopic gene expression and potentially new function. If the new site binds to a new local activator, then ectopic gene expression can occur in cells that express these activators when they also have an activating signal transduction pathway turned on. The new site alternatively could be a Signal Pathway Response Element (SPRE) in which case it will turn on expression ectopically only when the initial local activators are present.
Duplicating a gene (or at least a coding region) can have one of three major effects: creation of a pseudogene, DDS (duplication, divergence, subfunctionalization), or DDN (duplication, divergence, neofunctionalization). Pseudogenes do not add organismal complexity, but subfunctionalization and neofunctionalization potentially can. Under both DDS and DDN, we see a release of purifying selection due to the redundant duplicate protein (or RNA). Under DDS, a potentially pleiotropic protein is duplicated and now each duplicate can optimize for a subset of the original functions. This can lead to higher fitness and/or increased organismal complexity. Under DDN, one of the two redundant copies randomly acquires a new function that can then be optimized, as it is not under pleiotropic constraints. This new function increases organismal complexity.
Larger-scale duplications can also occur over either cis-regulatory regions or protein-coding/gene regions. The duplication of an entire enhancer element could potentially affect transcription either by acting on gene expression directly or by binding to and sequestering transcription factors present in the cell. An enhancer element can also be inserted into a new genomic location via transposon insertion or errors in crossing-over. In this case we would expect ectopic gene expression driven by this new element.
Larger-scale duplications over multiple coding regions include the duplication of the Hox-genes. This leads to multiple Hox-clusters in a single animal. An ancient duplication of the Hox-cluster gave rise to the Hox and Parahox clusters. Hox genes act in regionalizing the anteroposterior axis, mostly by acting on the ectoderm and mesoderm, of animals while Parahox genes are generally involved in endoderrmal and central nervous system patterning. Further Hox duplications (as seen in the whole-genome vertebrate duplications) allowed Hox genes to pattern new axes - like the vertebrate limb axis. Likewise, release of pleiotropic constraints on duplicated nervous system patterning genes may have given rise to the neural crest in vertebrates. This "fourth germ layer" plays important roles in patterning craniofacial structures as well as our enteric nervous system, adrenal gland, and other complex structures.