View metadata, citation and similar papers at core.ac.uk brought to you by CORE
provided by Elsevier - Publisher Connector Cell, Vol. 118, 539–544, September 3, 2004, Copyright 2004 by Cell Press The Lingering Enigma Review of the Allelic Exclusion Mechanism
Raul Mostoslavsky,1,2 Frederick W. Alt,1,2,4,* Subsequently, ubiquitous DNA repair proteins seal RAG- and Klaus Rajewsky2,3,4,* generated DSBs to form V(D)J coding and RS joins (re- 1Howard Hughes Medical Institute viewed by Bassing et al., 2003) The Children’s Hospital 2 The Center for Blood Research Chromatin Structure and Recombination and Department of Genetics Accessibility 3 Department of Pathology While all Ig and TCR loci share similar RSs, rearrange- Harvard Medical School ment of the different loci is differentially regulated (Gor- Boston, Massachusetts 02115 man and Alt, 1998). In B cell development, IgH gene assembly generally precedes that of IgL genes. For IgH
genes, D to JH rearrangements occur on both allelic B lymphocytes produce diverse antibody specificities chromosomes before appendage of a VH to DJH (Figure by “randomly” assembling antibody genes from germ- 1). In addition, there are two different IgL loci, Ig and line segments. Yet, though each B lymphocyte has Ig, and rearrangement of Ig occurs mostly in cells that multiple allelic loci for the different antibody chains, have already rearranged or deleted both Ig loci (Fig- each clonally derived mature B lymphocyte expresses ure 2). Finally, Ig loci rearrange specifically in B cells a single species of antibody with a unique specificity and TCR loci specifically in T cells. Given the common via a process termed allelic exclusion. Despite some RAG recombinase and target RSs, lymphocytes need progress, the precise mechanism of allelic exclusion a separate mechanism for achieving specific receptor remains an enigma. locus assembly. In this regard, early studies demon- strated that differential accessibility to the common V(D)J More than 50 years ago, Burnet proposed his clonal recombinase determines which gene segments rearrange selection theory of acquired immunity, in which an anti- (Yancopoulos and Alt, 1986). Over the years, many studies body serves as a receptor on cells which produce it. have addressed mechanistic correlates of accessibility. Unspecified genetic mechanisms were proposed to In particular, in vivo studies demonstrated trans-acting generate a large repertoire of cells, each with a different proteins, enhancers, and other cis-regulatory sequences mediate local events that may reflect changes in acces- receptor specificity. Contact between an antigen and sibility, including increased germline transcription and receptor stimulates cells to proliferate and secrete more chromatin changes (reviewed by Krangel, 2003). antibody of the same specificity (Burnet, 1959). Early in development, contact with self-antigens through the The Problem of Allelic Exclusion receptor kills the cells, establishing self-tolerance. This In Burnet’s theory, each Ig-producing B cell expresses prophetic theory was verified by a wealth of experimen- a unique antibody specificity. Yet, each B cell has two tal data. Indeed, B lymphocytes mediate humoral immu- HC loci and mulitple LC loci in its genome. As each nity by producing immunoglobulin (Ig) first as membrane locus can potentially randomly assemble different V, (D), bound B cell antigen receptors (BCRs) and, subse- or J segments and encode multiple, distinct IgH or IgL quently, as secreted antibodies. Ig molecules are a com- chains, a given cell might form multiple BCRs with differ- plex of two identical heavy chains (HC) and two identical ent specificities. Given this, how do B cells achieve light chains (LC). HC and LC variable regions are juxta- production of a single type of antigen receptor, a mecha- posed to form the antigen binding site. During B cell nistic requirement of the clonal selection mechanism? differentiation, HC variable region exons are assembled Seminal studies using rabbit antisera demonstrated that somatically from variable (V), diversity (D), and joining more than 99% of individual B lymphocytes express (J) gene segments and LC variable region exons from only one HC and one LC allele as surface Ig receptors V and J segments via V(D)J recombination (Tonegawa, (Cebra et al., 1966; Pernis et al., 1965), a process termed 1983). Each mammalian B cell expresses a unique anti- allelic exclusion. Two prevailing theories to explain al- body specificity with the enormous diversity of specifici- lelic exclusion are the regulated and the stochastic mod- ties generated largely by V(D)J recombination. els (Gorman and Alt, 1998). The relevant phenomenon T lymphocytes mediate cellular immunity through sim- of IgL isotype exclusion limits B cells to production of ilar receptors, consisting of one of two distinct hetero- either Ig or Ig LCs, but not both (Figure 2). Mechanistic dimers (TCR␣/ or TCR␥/␦). As in B cells, TCR genes issues associated with IgL isotype exclusion, still largely are assembled somatically in differentiating T cells by unsolved, have been reviewed (Gorman and Alt, 1998). V(D)J recombination. In both B and T cells, V(D)J recom- To simplify this review, Ig rearrangement, which gener- bination is initiated by the RAG endonuclease. RAG in- ally does not influence Ig allelic exclusion, will be con- troduces DNA double-strand breaks (DSBs) between sidered at the end in the context of its relevance to variable region gene segments and short, conserved major unresolved issues. recombination signal (RS) sequences that flank them. Regulated Versus Stochastic Models *Correspondence: [email protected] (F.W.A.), rajewsky@ of Allelic Exclusion cbr.med.harvard.edu (K.R.) The regulated model of allelic exclusion proposed that 4 Senior authors listed in alphabetical order. IgL chain V gene assembly must proceed on one chro- Cell 540
Figure 1. Heavy Chain Gene Rearrangement Partially adapted from Yancopoulos and Alt, 1986; see text for details. mosome at a time and that protein products from a tence of nonfunctional or deleterious gene segments functional IgL rearrangement (one that encodes an IgL (Rajewsky, 1996) and potential successive rearrange- chain that associates with the preexisting IgH chain to ments (Melchers et al., 1999). Thus, about 50%–60% of make surface Ig) mediates allelic exclusion through mature B cells have a productive VJ/germline configu- feedback inhibition of further IgL rearrangement (Alt et ration and approximately 40%–50% of cells have a two al., 1980; Figures 1 and 2). Aspects of this model required VJ rearrangements in a productive/nonproductive modification with the subsequent discovery of receptor configuration. Similar ratios also are observed for VDJ/ editing, a process that allows continued rearrangement DJ and VDJ/VDJ rearrangements at the IgH and TCR of an allele that generates an autoreactive antibody loci. Based on past usage of the term, we refer to this (Nemazee and Weigert, 2000). This model also was ex- approximate distribution as the “60/40” ratio (Yanco- tended to cover IgH and TCR allelic exclusion, where poulos and Alt, 1986). The 60/40 ratio generally has been variable regions are generated from two separate rear- interpreted as indicating a level of productive rearrange- rangements (Yancopoulos and Alt, 1986; von Boehmer, ment of 33% or less and a form of feedback to inhibit 2004; see below). In these loci, D to JH rearrangements further rearrangement following a productive rearrange- occur first and on both alleles (i.e., are not allelically ment (Coleclough et al., 1981; Alt et al., 1984; Figures excluded), with V to DJ being the regulated step. The 1 and 2), although other possibilities are conceivable regulated model argues that a certain percentage of (see below). cells make productive rearrangements on their first at- Support for regulated allelic exclusion came from tempt, leading to freezing of a germline Ig (or DJH or transgenic mouse studies that showed expression of DJ) on the second allele, generating a population of transgenic or light chains inhibits rearrangement of mature lymphocytes with a productive rearrangement endogenous LC loci or that expression of IgH or TCR and a germline Ig (or DJ) allele. The model also pre- transgenes inhibited V to DJ rearrangement at those loci dicted that cells that make a nonfunctional VJ rear- (reviewed in Gorman and Alt, 1998). Thus, these studies rangement (or VHDJH or VDJ) on the first allele go on supported the notion that feedback from expressed en- to rearrange the second allele with the same probability dogenous IgL (or IgH or TCR) chains plays a critical of achieving a productive rearrangement, leading to the role in mediating allelic exclusion. An exception to the population of mature lymphocytes with VJ (or VHDJH or feedback rule was noted for B cells that produce multiple VDJ) rearrangements, on both alleles (one productive LC proteins, but express only one on the surface, which and one nonproductive) (Figures 1 and 2). Given junc- led to the suggestion that a LC protein must also associ- tional diversification mechanisms that randomly add or ate with the HC in a “functional” manner to mediate delete nucleotides, only 33% of rearrangements would feedback (Alt et al., 1980). With the discovery of surro- be predicted to be in frame (productive) and, thereby, gate LCs and preTCR␣, this general type of “checking” generate a protein; although, for individual loci, this model was extended to IgH and TCR allelic exclusion number may be somewhat lower or higher due to exis- (Rajewsky, 1996). Review 541
Figure 2. Light Chain Gene Rearrangement Partially adapted from Yancopoulos and Alt, 1986; see text for details.
Stochastic models, in their most simple form, propose (Alt et al., 1980): (1) Rearrangement should be initiated that inefficient V(D)J rearrangement mediates allelic ex- at only one of the two homologous loci (initiation of clusion (Coleclough, 1983). One early model proposed allelic exclusion); and (2) rearrangement on the second that V(D)J rearrangement proceeds simultaneously on allele should be prevented by expression of an appro- both HC or LC alleles, but both would rarely be ex- priate Ig or TCR chain (maintenance of allelic exclusion). pressed in an individual B cell because the generation While feedback mechanisms can explain maintenance, of functional joins occurs infrequently. However, such they do not explain the perplexing problem of initiation. a model, in the absence of feedback, predicts no regula- If most developing B lymphocytes initiated V to J (or tion; so both alleles should ultimately be rearranged in VH to JDH) rearrangement simultaneously at both allelic mature B cells. This is not the case, as mature B and loci, a substantial percentage would exhibit allelic inclu- T cells conform roughly to the “60/40” ratio for all alleli- sion. Therefore, chromosomal accessibility must be reg- cally excluded loci including Ig, IgH, and TCR alleles. ulated such that only one allele at a given locus becomes Another form of a stochastic model argues that B cells accessible over a time period in which feedback could are limited to one productive rearrangement because be achieved. Many potential mechanisms have been rearrangement efficiency per se is so low that the proba- proposed as to how this might occur (Gorman and Alt, bility of two complete rearrangements in a cell would 1998). be very low (Liang et al., 2004; see below). This scenario A recent study from Schlissel and coworkers has mon- would predict that most lymphocytes would only assem- itored potential mechanisms of Ig allelic exclusion via ble a complete variable region exon at one IgL, IgH, or a gene targeting approach that allowed measurement of TCR allele; however, this is not the case (Gorman and Ig transcriptional activation preceding rearrangement Alt, 1998). Thus, purely stochastic models cannot readily (Liang et al., 2004). By inserting a GFP cDNA into a J, explain allelic exclusion with respect to rearrangement they followed activation of germline transcription on an patterns; although hybrids of stochastic and regulated allelic basis. Strikingly, only a minor portion of pre-B activated the Ig locus as determined by (%5ف) models could explain the observed 60/40 ratios (Cole- cells clough et al., 1981; Liang et al., 2004; see below). this reporter assay and such activation was monoallelic. Furthermore, when the GFPϩ pre-B cells were sorted Initiation Versus Maintenance of Allelic Exclusion and cultured in vitro, the vast majority preferentially re- In the context of the original regulated model, allelic arranged the GFP allele, indicating that measured open- exclusion was proposed to involve two separate steps ing of the allele for GFP transcription is an index of Cell 542
rearrangement potential. They interpreted these find- differentiation to the pre-B stage where Ig is activated, ings in the context of two aspects of the mechanisms but where Ig recombination would be inhibited, and outline above. First, they argued for a low frequency of GFP maintained. In these experiments the fraction of Ig locus activation for rearrangement, and that this GFPϩ pre-B cells was again 5%. This value is not easily supports a stochastic model. Second, they argued that compatible with the results obtained in Rag-proficient Ig alleles compete for transcription factors to allow animals and may reflect accumulation of “dead-end” opening, providing an allelic competition mechanism for cells in Rag-deficient mice, that are developmentally initiation. The latter model is in accord with an earlier beyond the stage of Ig activation. The authors also proposition based on chicken Ig locus studies (Ferra- consider the possibility of an artifactually low activation dini et al., 1994). While these models are of significant of the GFP insertion per se and make a considerable interest, they still do not readily explain all aspects of effort to counter this possibility. One way to further vali- endogenous V(D)J recombination patterns, as the au- date their assay might be to insert a GFP gene into the thors acknowledge (see below). JH or J locus, where, given D to J rearrangement on both alleles, one might expect to see activation of GFP Does Probabilistic Activation of cis-Regulatory in a much larger proportion of cells. Further insight might Elements Influence Allelic Exclusion? then be gained by inserting GFP into the VH or V region Previous work suggested that transcriptional enhancers and looking at the percent activation, since it is the V may function by altering the probability of gene expres- to DJ step that is allelically excluded. sion in a cell population rather than modulating overall Whatever the exact percentages, the Liang et al. transcription rates (Walters et al., 1995). Such an en- (2004) results indicate that only a fraction of pre-B cells hancer function has been considered as one potential is in the process of activating the Ig locus for rearrange- mechanism for activating one allele at a time (Gorman ment at any given time. While this result is consistent and Alt; 1998). In the model of Liang et al. (2004), activa- with a model of probabilistic, variegated locus activa- tion of an Ig allele at the pre-B cell stage depends tion, it could also indicate efficient locus activation in a on variegated expression of transcription factors. They narrow developmental time window. Consider a regi- argue that cooperative binding of such factors, due to ment of soldiers, 5% of whom are saluting at a given limiting amounts, is a rare event, occurring stochas- time. One may conclude that these 5% are chosen on tically at only one allele per cell. They also argue that the probabilistic principles. But perhaps the soldiers are two alleles compete for the factors (allelic competition). parading, and only the ones passing by their general Together, these processes might prevent a given Ig- salute. In this case, all soldiers would salute in the course rearranging pre-B cell from rearranging both Ig alleles of the parade. Extrapolating from this picture, all cells simultaneously. might attempt rearrangement at a given developmental A pleasing aspect of the Liang et al. allelic competition time; cells acquiring a productive rearrangement would model is that it should be straightforward to test; since proceed in development and leave the compartment; overexpression of the appropriate factors should and the remaining cells (including those that require “break” allelic exclusion. However, a limited availability receptor editing) would pass through a second, similar of transcription factors must be rationalized with fre- developmental window in which gene rearrangements quent rearrangement of multiple copies of transgenic are permitted (Figure 2). The result would be a strictly V(D)J substrates in individual cells (Ferrier et al., 1994; regulated model of allelic exclusion, despite a low fre- Alvarez et al., 1995). Most importantly, both the stochas- quency of cells initiating rearrangements at a given time tic element and the concept of allelic competition must in the steady state. be weighed against previous findings that suggest the allele “chosen” by the transcription factors in a given How Does the Rearrangement Process Progress cell is determined early in development (Mostoslavsky to the Next Ig Allele? et al., 2001). Thus, in each B cell, one of the two alleles Neither current model for initiation of allelic exclusion is demethylated and replicates earlier than the other; at the Ig locus (allelic competition/low intrinsic rear- and this allele is preferentially rearranged (Goldmit and rangement rates or allelic marking) has offered a clear Bergman, 2004). However, the Liang et al. (2004) findings solution to the lingering question of how rearrangement are not necessarily mutually exclusive with a strictly progresses to the other allele in cells that make a non- predetermined mechanism for the initiation (see below). productive rearrangement or to Ig in cells that have Previous work has shown that up to 20% of normal nonproductively rearranged or deleted both Ig alleles pre-B cells express cytoplasmic chains (Pelanda et (Figures 1 and 2). Clearly, the potential complexity of al., 1996) and that a much larger fraction carries Ig the problem will be related to the mechanism that effects rearrangements, many of which are out of frame (Yama- allelic choice and, in this regard, one could imagine gami et al., 1999). From the latter findings, it would ap- several very different solutions (Gorman and Alt, 1998). pear that in the Liang et al. (2004) experiments no more A mechanism in which the rearrangement complex than 30% of pre-B cells still carry the GFP gene (V to could be assembled in only a single nuclear location (Alt J rearrangement deletes the GFP gene), suggesting et al., 1992) could accommodate known rearrangement that as many as 15% of pre-B cells may activate the patterns. A different mechanism, proposed by both re- Ig locus at any given point in time, as opposed to the cent models for initiation (Mostoslavsky et al., 2001; approximately 5% proposed. The authors address this Liang et al., 2004) would be that cells that make a non- issue by monitoring GFP expression in RAG-deficient productive rearrangement persist in the pre-B popula- mice transgenic for a rearranged HC gene to promote tions until they acquire the ability to rearrange the sec- Review 543
ond allele. Such a possibility could be envisioned in References the context of a strictly regulated model. However, one might also accommodate such a mechanism, at least Alt, F.W., Enea, V., Bothwell, A.L., and Baltimore, D. (1980). Activity of multiple light chain genes in murine myeloma cells producing a for the Ig locus, in the context of a stochastic model. single, functional light chain. Cell 21, 1-12. In this case, one must assume that the pools of cells Alt, F.W., Yancopoulos, G.D., Blackwell, T.K., Wood, C., Thomas, with a nonproductive rearrangement versus entering E., Boss, M., Coffman, R., Rosenberg, N., Tonegawa, S., and Balti- cells with two germline alleles would equilibrate to a more, D. (1984). Ordered rearrangement of immunoglobulin heavy level in which the combined output of productive rear- chain variable region segments. EMBO J. 3, 1209–1219. rangements from both sets would generate the 60/40 Alt, F.W., Oltz, E.M., Young, F., Gorman, J., Taccioli, G., and Chen, ratio (Figures 1 and 2). J. (1992). VDJ recombination. Immunol. Today 13, 306–314. Liang et al. (2004) discuss their findings in the context Alvarez, J.D., Anderson, S.J., and Loh, D.Y. (1995). V(D)J recombina- of a regulated model. However, they also argue that tion and allelic exclusion of a TCR beta-chain minilocus occurs in their data support the old model that allelic exclusion the absence of a functional promoter. J. Immunol. 155, 1191–1202. is contributed by the very low probability of a given B Bassing, C.H., Swat, W., and Alt, F.W. (2003). The mechanism and cell rearranging two Ig alleles over its developmental regulation of chromosomal V(D)J recombination. Cell 109 (Suppl.), S45–S55. course. Conceivably, the pool equilibration model out- lined above might allow one to explain the 60/40 ratio in Burnet, F.M. (1959) The Clonal Selection Theory of Acquired Immu- nity. (Cambridge: The University Press). the context of such a stochastic model without feedback regulation; although there would be clear predictions Cebra, J., Colberg, J.E., and Dray, S. (1966). Rabbit lymphoid cells differentiated with respect to alpha-, gamma-, and mu- heavy poly- regarding rearrangement patterns in pre-B pools that peptide chains and to allotypic markers Aa1 and Aa2. J. Exp. Med. have not been forthcoming. However, based on factors 123, 547–558. relating to IgL isotype exclusion discussed next, there Chen, J., Trounstine, M., Kurahara, C., Young, F., Kuo, C.C., Xu, Y., are compelling reasons to consider mechanisms based Loring J.F., Alt, F.W., and Huszar, D. (1993). B cell development in on nonproductive rearrangement of one allele somehow mice that lack one or both immunoglobulin kappa light chain genes. positively influencing rearrangement potential at an- EMBO J. 12, 821-830. other (see below; Figure 2). Coleclough, C. (1983). Chance, necessity and antibody gene dynam- ics. Nature 303, 23–26. Coleclough, C., Perry, R.P., Karjalainen, K., and Weigert, M. (1981). The Related Problem of IgL Isotype Exclusion Aberrant rearrangements contribute significantly to the allelic exclu- It is quite difficult to imagine how the stochastic activa- sion of immunoglobulin gene expression. Nature 290, 372–378. tion mechanism would lead to the generation of Ig- Goldmit, M., and Bergman, Y. (2004). Monoallelic gene expression: producing cells. Thus, in both mice and humans (which a repertoire of recurrent themes. Immunol. Rev. 200, 197–214. have nearly equal levels of Ig and Ig producing B Gorman, J.R., and Alt, F.W. (1998). Regulation of immunoglobulin cells), both Ig alleles are rearranged or deleted in Ig light chain isotype expression. Adv. Immunol. 69, 113–181. producers (Gorman and Alt, 1998). If one accepts the Ferradini, L., Reynaud, C.A., Lauster, R., and Weill, J.C. (1994). Re- argument that Ig rearrangement is very inefficient com- arrangement of the chicken lambda light chain locus: a silencer/ antisilencer regulation. Semin. Immunol. 6, 165–173. pared to Ig rearrangement (Ramsden and Wu, 1991), it is hard to understand how any Ig producers are gen- Ferrier, P., Krippl, B., Blackwell, T.K., Furley, A.J., Suh, H., Winoto, A., Cook, W.D., Hood, L., Costantini, F., and Alt, F.W. (1990). Separate erated if intrinsic Ig rearrangement rates are already elements control DJ and VDJ rearrangement in a transgenic recom- extremely low. Very large pools of pre-B cells with both bination substrate. EMBO J. 9, 117–125. Ig loci nonproductively rearranged would be required Krangel, M.S. (2003). Gene segment selection in V(D)J recombina- to generate the pool of Ig producing cells, which is not tion: accessibility and beyond. Nat. Immunol. 4, 624–630. observed (Zou et al., 1993, Chen et al., 1993). Thus, Liang, H.E., Hsu, L.Y., Cado, D., and Schlissel, M.S. (2004). Varie- one must consider the alternative that nonproductive gated transcriptional activation of the immunoglobulin kappa locus rearrangement on one allele positively influences that in pre-b cells contributes to the allelic exclusion of light-chain ex- of another (Figure 2) and that, at least at certain points, pression. Cell 118, 19–29. actual rearrangement rates of a given allele may be much Melchers, F., ten Boekel, E., Yamagami, T., Anderson, J., and Rolink, more substantial than estimated by the GFP assay, as A. (1999). The roles of preB and B cell receptors in the stepwise allelic has been observed in pro-B cell lines and IL-7 derived exclusion of mouse IgH and L chain gene loci. Semin. Immunol. 11, 307–317. cultures (Reth et al., 1985; Rolink et al., 1993). Thus, we Mostoslavsky, R., Singh, N., Tenzen, T., Goldmit, M., Gabay, C., conclude that we still do not have answers to some of Elizur, S., Qi, P., Reubinoff, B.E., Chess, A., Cedar, H., and Bergman, the most longstanding and critical questions related to Y. (2001). Asynchronous replication and allelic exclusion in the im- initiation of allelic and isotype exclusion. Continued ap- mune system. Nature 414, 221–225. plication of creative new approaches, such as those Nemazee, D., and Weigert, M. (2000). Revising B cell receptors. J. described by Liang et al. (2004), may facilitate progress Exp. Med. 191, 1813–1817. at an accelerated pace. Pelanda, R., Schaal, S., Torres, R.M., and Rajewsky, K. (1996). A prematurely expressed Ig(kappa) transgene, but not V(kappa)J (kappa) gene segment targeted into the Ig(kappa) locus, can rescue Acknowledgments B cell development in lambda5-deficient mice. Immunity 5, 229–239.
R.M. was a Long-Term Fellow of the Human Frontier Science Pro- Pernis, B.G., Chiappino, G., Kelus, A.S., and Gell, P.G.H. (1965). gram and is a Senior Postdoctoral Fellow from The Leukemia and Cellular localization of immunoglobulins with different allotypic Lymphoma Society. F.W.A. is an Investigator of the Howard Hughes specificities in rabbit lymphoid tissues. J. Exp. Med. 122, 853–863. Medical Institute and is supported by grants from the NIH and NCI. Rajewsky, K. (1996). Clonal selection and learning in the antibody K.R is supported by grants from the NIH and NCI. system. Nature 381, 751–758. Cell 544
Ramsden, D.A., and Wu, G.E. (1991). Mouse kappa light-chain re- combination signal sequences mediate recombination more fre- quently than do those of lambda light chain. Proc. Natl. Acad. Sci. USA 88, 10721–10725. Reth, M.G., Ammirati, P., Jackson, S., and Alt, F.W. (1985). Regulated progression of a cultured pre-B-cell line to the B-cell stage. Nature 317, 353–355. Rolink, A., Grawunder, U., Haasner, D., Strasser, A., and Melchers, F. (1993). Immature surface Igϩ B cells can continue to rearrange kappa and lambda L chain gene loci. J. Exp. Med. 178, 1263–1270. Tonegawa, S. (1983). Somatic generation of antibody diversity. Na- ture 302, 575–581. Von Boehmer, H. (2004). Selection of the T-cell repertoire: receptor- controlled checkpoints in T-cell development. Adv. Immunol. 84, 201–238. Walters, M.C., Fiering, S., Eidemiller, J., Magis, W., Groudine, M., and Martin, D.L. (1995). Enhancers increase the probability but not the level of gene expression. Proc. Natl. Acad. Sci. USA 92, 7125– 7129. Yamagami, T., ten Boekel, E., Andaerson, J., Rolink, A., and Melch- ers, F. (1999). Frequencies of multiple IgL chain gene rearrange- ments in single normal or kappaL chain-deficient B lineage cells. Immunity 11, 317–327. Yancopoulos, G.D., and Alt, F.W. (1986). Regulation of the assembly and expression of variable-region genes. Annu. Rev. Immunol. 4, 339–368. Zou, Y.R., Takeda, S., and Rajewsky, K. (1993). Gene targeting in the Ig kappa locus: efficient generation of lambda chain-expressing B cells, independent of gene rearrangements in Ig kappa. EMBO J. 12, 811–820.