First of all, it exists about 5.000 transcription factors in human. Imagine the number of combination that might makes. Second, they are mostly a platform to recruit modifier enzymes like HDACs, DNMTs, etc.... to open/close chromatin structures. Third, in addition to transcription factors themselves, co-regulators can tunefully regulates gene expressions.

Hope it is helping...

As Florent mentions, the 'limited number' of transcription factors that you mention is actually quite large. Such a diversity is small enough to assure coherence but large enough to permit variation at many levels. For example, gene expression will vary during the cell cycle, among tissues, with different conditions, even with behaviour. This requires tremendous flexibility.

One key thing that I like to keep in mind concerning biological systems is that some of their parts have had 4 billion years to evolve. That is a lot of time to reorganize complex systems in a compact manner and the best way to do that is with interconnected networks of interaction, or regulation in the case we are discussing.

If the idea of networks and their importance is of interest to you, I highly suggest reading 'LINKED', by Albert-Laszlo Barabasi. I think it can be an eye opener for biologists just getting to genomics. Plus, it's fun to read :)

@ Florent, the number is probably a little smaller than that, there probably "only" 1400 TF encoded in the human genome: http://www.nature.com/nrg/journal/v10/n4/full/nrg2538.html

To Raghavendra:

As anybody working with Drosophila will not hesitate to tell you, transcription programs regulating cell fate specification (and probably to a large degree, stimulus responses) in metazoans are largely encoded in enhancers (and has been long before the recent ENCODE paper explosion ;) .

Recent research suggests (see, e.g. Heinz, Mol. Cell 2010 (sorry for the shameless plug ;) ; Ghisletti, Immunity 2010; Mullen, Cell 2011; Yanez-Cuna, Genome Research 2012) that there is a hierarchy of transcription factors, where "master regulators" generate regions of open chromatin (and probably also are responsible for setting up part of the 3D structure of the genome, see e.g. Job Dekker's latest paper with ENCODE in Nature). Thes then get marked by e.g. histone H3K4me1, an epigenetic mark of enhancers, and make binding sites for other transcription factors, both signal-dependent and independent ones accessible. Together, both classes of factors are necessary for the enhancer to "fire" and to activate transcription of nearby (or not so nearby) genes.

Since the master regulators need each other to be able to open chromatin at sites with "weak" sequence motifs, the combination of master regulators defines which sites open up for "business", and the complement of master regulators and non-master regulators defines, which sites become active, and thus which genes become transcribed (e.g. Heinz, Mol Cel. 2010).

The 3D interaction structure of the genome, which is also influenced by master regulator action probably serves as a higher level of control, i.e. which regions (euchromatic/heterochromatic) *can* be hit by master regulators and which can't. It'll be interesting to see whether there is a "folding" program of the genome, which would define the succession of genomic conformations (and thus gene regulatory programs) that a cell can access during development, essentially the 3D part of Waddigton's epigenetic landscape (in analogy to proteins, the primary structure would be the sequence itself, and the folding interactions to cluster enhancers/promoters together the secondary structure, or maybe it should be third, with the protein repertoire currently present to interact with the genome being second?).

OK let's cut the apple in two, the nest paper (from 2400 OK) says 2600 transciption factors that I think have increased since it. But anyway, it was just to say that if you considered just a linear promoter sequence, there is more combination of it than genes in human. And mainly that it is just the first layer in the transcription regulation. Conclusion to answer to Raghavendra, there is enough to precisely, temporally and tissue specifically all human gens in a specific way....

But thanks for your complement....

Babu MM, Luscombe NM, Aravind L, Gerstein M, Teichmann SA (2004). "Structure and evolution of transcriptional regulatory networks". Curr. Opin. Struct. Biol. 14 (3): 283–91. PMID 15193307.

Oops, sorry, the whole enhancer part below the citation was meant for the OP... .

Regarding the 2600 vs. 1400 TF in the human genome, 1400 may be low-balling it a little; On the other hand, the 2600 TF in the 2004 paper came from ENSEMBL back then, which probably wasn't the greatest way to discern DNA binding transcription factors from DNA binding enzymes or RNA binders. Since two of the four authors of the 2009 paper are the same as the one from 2004 that you cite, I would see the number in the latter paper as a refinement of their work from 2004, and probably to be more accurate.

To get back on topic though, I think what Raghavendra is alluding to is the fact that e.g. motifs highly enriched in cell type-specific DNase hypersensitive sites, or otherwise epigenetic marked sites point to limited numbers of transcription factors, and the model of pioneering factors collaborating against histones to to create regions of open chromatin, which in turn serve as beacons for the other (non-pioneering, and often signal-dependent) factors is so far the best explanation for how this phenomenon comes about that we have.