The NF-κB family of transcription factors is very modular, with different combinations having different effects. The active (nuclear, DNA-bound) TF is a dimer, composed variously of RelA/p65, RelB, c-rel, NFKB1/p50, and/or NFKB2/p52 subunits. For example, the "canonical" p65/p50 dimer is activated in response to stimulants like TNF-α (tumor necrosis factor alpha, released in response to inflammatory signals like the presence of pathogens) and LPS (lipopolysaccharide from the cell walls of Gram-negative bacteria), while the RelB/p52 dimer plays an important role in the development of B cells (the immune system component that produces antibodies). The AP-1 transcription factor is also heterodimeric, containing proteins from the Jun, Fos, JDP, and ATF families, and there are numerous other examples of multimeric transcription factor complexes. This recent article in Nature Immunology (disclaimer: I was not involved in that research, but my old lab did similar work) shows DNA sequence-specific binding by different NF-κB dimers.
Polymerism in general, and dimerism in particular, are quite common modes of transcriptional activation and regulation. The large number of ways in which a relatively small number of transcription factors can be combined allows for the exquisite control of genes, responding to a huge variety of cellular situations. Unfortunately, it also means that mutations in key, common components can result in transformation, unrestricted growth, and the generation of tumors.