D gene segments with irregular spacers (DIR) are D gene segments that are specific to higher primates. Their use is controversial because of their G؉C-rich long sequences. In the human, it has always been tempting to assume that a complementarity-determining region 3 sequence has been added by terminal deoxynucleotidyltransferase (TdT) activity and is not derived from DIR recombination. Herein, we examine the use of human DIR gene segments by cross-breeding the human Ig heavy chain minilocus pHC1 transgenic mice and TdT-deficient mice. In the absence of TdT and with a defined set of human D gene segments, it is relatively easy to demonstrate that DIR2 is used to form human Ig heavy chains, contributing to 7% of the human heavy chain rearrangements. V H DJ H rearrangements (where H is heavy chain) in the minilocus TdT ؊/؊ mice use small portions of DIR2 located throughout the coding sequence. These results constitute the strongest evidence to date that DIR gene segments are used to form human antibodies. Additionally, we show that direct and inverted DIR2J H and V H DIR2 rearrangements occur in the minilocus transgenic mice. During these rearrangements, D M2 3 signal sequence and a new DIR2 5 signal sequence are used. These rearrangements generally follow the 12͞23 recombination rule. Our results at the V H DJ H , DJ H , and V H D levels indicate that DIR2 is used to form human heavy chains in transgenic mice. The rearrangement of this gene segment likely involves, however, other mechanisms in addition to the classical V H DJ H recombination.The variable region of Ig heavy chains (IgH) and T cell receptor  and ␦ chains derive from the combination of three separate DNA elements termed the variability (V), diversity (D), and junctional (J) gene segments through a cell-specific process termed V H DJ H recombination (where H is heavy chain). Typically, the recombination process is performed in two steps with D to J H rearrangement preceding V H to DJ H rearrangement (1-3). The rearrangement is directed by recombination signal sequences (RSSs) flanking each coding gene segment. RSSs are formed of a consensus palindromic heptamer related to the sequence CACAGTG and a nonamer related to the sequence ACAAAAACC separated by 12 Ϯ 1 or 23 Ϯ 1 nucleotide spacers. D gene segments are flanked on both sides by 12 Ϯ 1 spacer RSSs, whereas V H and J H gene segments are flanked by 23 Ϯ 1 spacer RSSs at the 3Ј and 5Ј sides, respectively.