Mechanistic and Genetic Biases in Human Immunoglobulin Heavy Chain Development

Broadly neutralizing antibodies against HIV are rare; most patients never develop them at detectable levels. The discovery of four such antibodies therefore warrants research into their origins and their presumed unique characteristics. Such studies, however, require baseline knowledge about commonalities and biases affecting human immunoglobulin development. Obtaining that knowledge requires large sets of gene sequence data and the appropriate statistical techniques and tools.

The Genbank repository provides a free and easily accessible source for such data. Several large datasets cumulatively comprising over 10,000 human Ig heavy chain genes were identified, downloaded, and carefully filtered. We then developed a special software tool called SoDA, which employs a unique dynamic programming algorithm to provide a statistical reconstruction of the events that led to a given antigen receptor gene. Once developed, tested, and peer-reviewed, we used SoDA to provide initial data about each downloaded gene with respect to gene segment usage, n-nucleotide addition, CDR3 length, and mutation frequency, thereby establishing the most precise estimates currently available for human Ig heavy chain gene segment usage frequencies.

We compared data from productive non-autoreactive Ig to non-productive Ig and found evidence for gene segment usage biases, D/J segment pairing preferences resulting from multiple sequential D-to-J recombination events, and biases in TdT action between the V-D and D-J. Further analysis of autoreactive Ig genes yielded evidence that n-nucleotide addition comes at a cost: the higher the ratio of n-nucleotides to germline-encoded nucleotides for a given CDR3 length, the greater the probability of autoreactivity. These results suggest that the germline gene segments have been selected for lack of autoreactivity.

It has previously been shown that human Ig gene segments have evolved efficient evolvability under somatic hypermutation. We have now extended these results, showing that Ig gene sequences are "tuned" to preferentially produce consequential mutations in the antigen-binding domains, and synonymous mutations in the framework regions.

Together, these analyses provide new insights into the genetic and mechanistic biases shaping the human Ig repertoire.