Molecular Mechanisms of Lymphocyte Differentiation 
pendent on IL-2 for growth. One of these genes, 
which we call itk (for IL-2-inducible T cell ki- 
nase), is expressed specifically in T cells. The itk 
gene encodes a tyrosine kinase that resembles src 
kinases, but from its sequence we predict that it 
will differ from those kinases with respect to its 
location in the cell and the regulation of its activ- 
ity. IL-2 is known to regulate expression of a num- 
ber of genes, including a gene for its own recep- 
tor. In cells that have been withdrawn from IL-2, 
itk is expressed at low levels. Soon after addition 
of IL-2, itk is strongly turned on in parallel with 
the IL-2 receptor gene, suggesting that itk func- 
tions in the IL-2 response. 
Antibody Gene Rearrangement 
The total number of immunoglobulin or T cell 
receptor gene segments is large, but when any 
particular immunoglobulin or T cell receptor 
gene is assembled, only a handful of segments are 
selected and joined. As a result, many different 
combinations of segments are possible. It is this 
shuffling of small bits of DNA that generates 
much of the diversity of the immune response. 
By using artificial DNA molecules that rear- 
range after they are introduced into immature B 
cells, we have been able to outline the general 
features of antibody gene assembly, but the mech- 
anism at the level of interactions between mole- 
cules is not known. The products of at least three 
genes are known to play an intimate role in rear- 
rangement. Two of these genes, RAG-1 and RAG- 
2, have been shown by others to activate some- 
how antibody gene rearrangement. A third gene 
encodes a protein that all cells use to help repair 
DNA damage and that also functions at a late step 
in antibody gene assembly. In addition to the 
products of these genes, our laboratory has found 
a protein — NBP — that binds specifically to a 
DNA sequence element required for rearrange- 
ment, suggesting that it may be a part of the re- 
combinational machinery. Members of our labora- 
tory are studying the products of the RAG genes 
and NBP, with the aim of understanding their 
functions and interrelationships. 
Since NBP was isolated from calf thymus, we 
cloned the RAG-1 and RAG-2 genes of the cow 
and identified regions of identity among the 
genes from cow, mouse, and human. Based on the 
predicted protein sequences of these conserved 
regions, antibodies were raised that specifically 
bind to the products of the RAG-1 and RAG-2 
genes. Using these antibodies, RAG-1 and RAG-2 
proteins were identified in extracts of calf thy- 
mus and isolated. We found these proteins are 
distinct from NBP. Interestingly, the sizes of the 
RAG-1 or RAG-2 proteins are slightly different 
from those predicted by their gene sequences, 
suggesting that they are modified after they are 
made. We have gone on to find that both RAG-1 
and RAG-2 are modified by addition of phosphate 
groups. As was discussed earlier, this type of mod- 
ification often serves to regulate protein func- 
tion. We are intrigued by the possibility that such 
is the case for RAG-1 and RAG-2. 
Molecular Mimicry by Antibody Molecules 
As a result of antibody gene rearrangement, the 
structures of potential antigen-binding sites vary 
almost endlessly. It has been a long-standing no- 
tion in immunology that the structure of an anti- 
gen could be mimicked by a special type of anti- 
body — an antibody raised against another 
antibody that was originally elicited by the anti- 
gen of interest. We can draw an analogy for this 
idea from the art of casting. A sculptor creates a 
positive image, which is used to produce a nega- 
tive image, the mold, which in turn is used to 
reproduce the final work. An essential difference 
between this metaphor and reality is that in cast- 
ing, one object is shaped by another, while in the 
antibody response, the binding of antigen selects 
a complementary shape from a universe of preex- 
isting ones. 
We have tested this idea in collaboration with 
research groups in Baltimore and Paris (led by 
Mario Amzel of the Johns Hopkins School of Medi- 
cine and Pierre Ronco, INSERM, Hopital Tenon). 
Antibodies (Abl) were produced against a hor- 
mone, angiotensin II. Then a second antibody 
(Ab2) was raised against the anti-angiotensin II 
antibody. A third antibody (Ab3) was in turn 
raised against Ab2. The Ab3 antibodies bound to 
angiotensin II just as well as the original Abl anti- 
bodies, suggesting that Ab2 could mimic the hor- 
mone's structure. 
This was proved by analyzing the structures of 
Abl and Ab3. Remarkably, we found that the se- 
quences of their antigen-binding sites were 
nearly identical. We then determined the three- 
dimensional structure of the antigen-binding site 
of Ab3 in complex with angiotensin II. We discov- 
ered that the atoms in critical contact with angio- 
tensin II in this complex are also found in the 
sequence of Abl. Thus both Abl and Ab3 were 
elicited by similar structures, one represented by 
the original antigen, angiotensin II, and the other 
by a surface feature of the Ab2 antibody that mim- 
ics angiotensin II. 
110 
