A HYPOTHESIS ACCOUNTING FOR THE PARADOXICAL EXPRESSION OF THE D GENE SEGMENT IN THE BCR AND THE TCR

Abstract

The D gene segment expressed in both the TCR and BCR has a challenging behavior that begs interpretation. It is incorporated in three reading frames in the rearranged transcription unit but is expressed in antigen-selected cells in a preferred frame. Why was it so important to waste 2/3 of newborn cells? The hypothesis is presented that the D region is framework playing a role in both the TCR and the BCR by determining whether a signal is transmitted to the cell upon interaction with a cognate ligand. This assumption operates in determining haplotype exclusion for the BCR and in regulating the signaling orientation for the TCR. Relevant data as well as a definitive experiment challenging the validity of this hypothesis, are discussed.

Introduction

This essay is written in the philosophical framework championed by T. Dobzhansky [1]. We will be able to claim that we understand the role of the D-gene segment when we can rationalize an evolutionary selection pressure operating on a function that determined its properties. The goal, therefore, will be to define a primary singular function of D upon which evolution selected. As this subject is one of considerable controversy, I will try to take a position championed by Sherlock Holmes, “when you have eliminated the impossible, whatever remains however improbable, must be the truth.” I prefer to refer to “truth” as the “default postulate,” which survives until it becomes “impossible” (i.e., is disproven). We are, of course, dealing only with meaningful postulates (i.e., those that have a specified way to be disproven). Good theories must have a defined potential to self-destruct. For this reason a definitive experiment derived from the proposed hypothesis is presented.

Background

In 1980, Early and coworkers [2] postulated the existence of a gene segment between V_H and J_H that they termed “D-diversity” because of its assumed evolutionarily selected role in generating the repertoire by functionally diversifying the specificity of antibodies. Their postulate was based on a comparison of the amino acid sequence of the heavy chain (H) to the sequence of the germ line gene segments for V_H and J_H. This revealed a missing piece of DNA needed to encode amino acids in the Kabat-Wu third hypervariable region (HV3) between V_H and J_H. Sakano and coworkers [3] confirmed this postulate of the existence of a D gene segment by isolating and sequencing it from the H chain locus. Both groups were in agreement that the evolutionary selected role of the D_H gene segment was that it functioned to contribute to the diversification of the antibody repertoire, hence D-diversity. This assumption is still dominant today (e.g., [4-6]). We have challenged this assumption over the years [7-11] and will reexamine it here. So much new information has accrued since these early discussions that it is time to reconsider, “Why D?” particularly because the interpretation of its functional role has wide ramifications. We will be considering the mature immune system which expresses N-additions in the junctional regions. The D-segment is essentially a spacer between two sets of N-additions making it necessary to consider the NDN region as a selectable whole.

The two key facts about D that pose two key questions

1. With occasional exceptions, D has been maintained throughout evolution as a separate gene segment in both the H-chain locus of the BCR and the β-chain locus of the TCR.

2. The D gene segment, because it is flanked by varying numbers of N-additions, is incorporated into the rearranged transcription unit in three reading frames (RF) but expressed in functional BCRs in a preferred frame. By contrast, in functional αβTCRs all three D reading frames (RF) are expressed.

These two facts raise two questions:

1. Putting aside the few understandable exceptions that are found in primordial vertebrates (e.g., shark), why is D kept throughout evolution as a separate gene-segment?

2. 2 – If a preferred reading frame (RF) is so important for the BCR, why does it appear to be so unimportant for the αβTCR?

Any proposal for D gene segment function must answer these two questions as both necessary and sufficient to explain its role.

An aside on the nomenclature of D-reading frames

I will use here the definition of reading frame that is in popular usage. RF1 is referred to as the preferred frame and RF2 and 3 are counted down from RF1. However, this is far from rational and leads to problems when comparing species and when assigning RF within a locus.

The only sensible procedure is that of Kabat et al. [12]. They define RF1 counting from the first nucleotide following the heptameric fusion sequence. This defines three consecutively numbered logical reading frames (LRF). I won't be using LRF as it is rarely adhered to, making communication difficult.

As a matter of principle, however, there is no a priori reason that the different D_H gene segments in a given individual will use the same LRF as the preferred RF. This fortuity in mice and chickens established the importance of the preferred frame. Although uniform in mice and chickens, the preferred frame for all D_Hs in chicken is LRF1, whereas in mice, it is LRF3. While every D_H gene segment will have only one preferred (functional) frame (RF), not every D_H gene segment need express the functional amino acid sequence in the same LRF (see further discussion in ref. [9]).

Suggested evolutionarily selected roles of D

D-diversity

The view that the D-gene segment was selected to function as a diversity element is so entrenched that one must begin any discussion of D with this postulate. We will consider the role of D in the BCR and TCR separately because the TCR functions via MHC-dependent recognition of peptide whereas the BCR does not. This difference has important and unique consequences. Nevertheless, we will try to reveal a single property of the D-gene segment upon which selection operates and that explains both behaviors.

The BCR

Throughout the 1970s and 1980s hypervariability in the Kabat-Wu sense was equated to combining site diversity. While this is a fundamental error on logical grounds, with the development of Protecton theory [7, 8], it became clear that this translation of sequence diversity into functional repertoire size was untenable. To summarize our argument, one might say that the extent of sequence diversity in the (NDN)_H region was too-much-of-a-good-thing.

The junctional region of the H-chain was estimated to contribute 10⁸ potential sequence variants. If this were equated to a contribution to repertoire size, the humoral system would be nonfunctional because no specificity expressed on B-cells would be in sufficient frequency to protect [8, 10]. This argument led to the suggestion by many that the paratopic repertoire is highly degenerate meaning that many sequence distinguishable paratopes recognized the same epitope in a functional way. If we refer to this family of paratopes as a “paratopic clan,” then the size of the repertoire equals the number of paratopic clans [13, 14]. This reduces the problem to determining the degree of degeneracy required to make the primary response functional. Any individual unable to survive a primary encounter with a pathogen need not worry about a secondary encounter.

A “Protecton” is the minimum sized aliquot of the immune system of an individual that has all of the protective properties of the whole. We have estimated this to be a total of 10⁷ B-cells at a density of 10⁷ per ml. The immune system of an individual is made up by iteration of this unit. A mouse, for example, iterates 10 Protectons, a human 10⁵ Protectons [7, 8]. To optimize the calculation in favor of D-diversity, we will assume (not likely) that all B-cells of the Protecton are functional.

In order to protect against a fast growing pathogen, we have estimated that 100 B-cells per ml specific for the pathogen must respond initially [8]. If all B-cells in the Protecton were functional, the maximum size of the repertoire would be 10⁶ (∼10 B-cells per epitope and ∼10 epitopes per antigen). As there are close to 40 V_L- and 40 V_H-gene segments that are functional in mouse and humans, at maximum, there are ≤40² (1600) functionally distinct V_LV_H pairs. If the junctional amino acid sequence diversity of the L-chain is estimated to be 10² and of the H-chain 10⁸, then the contribution of each V_LV_H pair would be scrambled 10¹⁰-fold, obviously absurd as germline selection for the 40 V_LV_H specificities that are the foundation of the repertoire would be rendered impossible, not to mention selection for a mechanism of somatic hypermutation. If one argues that the junctional diversity of the L-chain contributes nothing, the role of the 10⁸ possible sequences of (NDN)_H remain as a problem. How degenerate must their contribution to the size of the repertoire be?

The D-gene segment makes a very small contribution to the amino acid sequence pool compared to N-additions. Assuming ∼20 D_H-gene segments per genome and its expression in two reading frames (RF) (one frame is usually chain terminating), then each V_LV_H would be diversified 40-fold (1600 × 40 = 6.4 × 10⁴). Assuming an average of 6 N-additions per rearrangement and 15 of 20 amino acids to be functionally distinguishable at each position, 15⁶ or 10⁷ sequences are potentially possible. This would diversify each V_LV_H pair 10⁷-fold (1600 × 10⁷ = 1.6 × 10¹⁰). In total, each V_LV_H pair would be scrambled by junctional diversity ∼10⁸-fold [40 (D sequences) × 10⁷ (N sequences)]. Given 1600 V_LV_H pairs the repertoire size would be 6.4 × 10¹¹. As a maximum paratope repertoire cannot exceed 10⁶ (and in the real world is closer to 10⁵), the (NDN)_H sequence diversity would have to be roughly 10⁶-fold $\left(\frac{6.4\times {10}^{11}}{{10}^6}\right)$ degenerate (i.e., 10⁶ amino acid sequences per combining site specificity).

Why would evolution select for such extraordinary diversity in amino acid sequence in order to express so little of it as a functionally discriminating part of the combining site repertoire of the BCR? Is it simply unavoidable?

Under Protecton theory [8, 10, 11] applied to mice and humans, the primary repertoire of the BCR is made up of 40 V_LV_H pairs selected in the germline for the specificities they encode, 1560 V_LV_H pairs (40²-40) germline-encoded but somatically diversified by random complementation and ∼25 × 40² V_LV_H somatic mutants, to yield an adequate primary functional repertoire of ∼4 × 10⁴. This repertoire divides the antigenic universe into combinatorials of 10 epitopes. The (NDN)_H region contributes little (not necessarily zero) to this repertoire and, therefore, must have been selected for another function. All in all, in the D-diversity framework, expression of D_H in a preferred reading frame makes little to no sense.

The TCR

The αβTCR encodes two repertoires. One is germline-selected to recognize the allele-specific determinants on the MHC-encoded restricting elements (R) of the species, the other is somatically generated and is responsible for the recognition of the MHC-bound peptide. As the Vα and Vβ domains of the TCR are germline-selected and not varied somatically they must encode recognition of the allele-specific determinants. The only parts of the TCR that are varied somatically are the junctional regions of the α and β subunits. Therefore, complementation between them must determine the combining site that recognizes peptide (anti-P). This is the combining site relevant to our discussion.

On the one hand, evolution is selecting on the junctional diversity of the TCR to specify recognition of peptide. On the other hand, it is not selecting on the diversity of the junctional region of the BCR to specify recognition of its shape-ligand. Is there a unitary selection on this region that would explain both behaviors?

Before answering this question, let's deal with the other suggestions as to the function of D upon which evolution is selecting.

D-differentiation

In this category Klinman and Decker [15] proposed that D_H functioned in a pathway of B-cell differentiation that required an interaction with D_H acting as a ligand in its preferred reading frame (hence “D-differentiation”). They also pleaded for D-diversity, a contradiction that they left unresolved. The claim that there is bias in joining that favors D RF1 (the preferred frame), only raises the question, “On what function of D is evolutionary selection operating?” (debated in Refs. [16, 17]). Basically, bias in joining is a form of D-differentiation. Given D-differentiation, why didn't evolution simply couple D_H to V_H or J_H in the germline in the preferred reading frame (RF)? Why was it so important to maintain D as a separate gene segment that results in wasting 2/3 of B-cells. Clearly, D-differentiation fails to answer Question 1 and cannot be generalized to deal with Question 2. If a D-gene segment expressed in a preferred frame is essential for B-cell, but not T-cell development, why did the T_β-locus keep it?

D-different

Another proposal by Louzoun et al. [18] operating in the framework of receptor editing, was that D_H limits the H chain locus to essentially a single rearrangement. In their view D is the element distinguishing the L-chain, which receptor edits, from the H-chain, which does not (or rarely). Hence, they dubbed it, “D-different.” However, D-different only deals superficially with Question 1. The single rearrangement at the H-locus operates in cis and, therefore, does not confront haplotype exclusion which either requires a STOP mechanism that operates in trans or D-disaster (see later). Even a postulated transacting H-STOP-H mechanism would require an additional STOP signal to regulate rearrangement at the L-loci, like (LH)₂-STOP). If L, not H, is allowed to receptor edit (i.e., a leaky (LH)₂-STOP), the frequency of double producing cells increases dramatically [19]. It would be fair to say that D-different inadequately confronts Question 1 and cannot deal with Question 2 as this mechanism is independent of reading frame.

D-delimit

In addition to degeneracy as a way of bringing the amino acid sequence diversity into line with functional combining site diversity, is the view that D_H acts as a constraint that brings the amino acid sequence diversity of the (NDN)_H region into a more functional range [20].

A probing analysis of D_H-gene segment evolution has been made by Schroeder and associates [21]. They review the data suggesting that the selection on D_H is for its composition (hydropathicity), not its sequence. RF1 (the preferred frame) is the most hydrophilic; RF2 is hydrophobic and RF3 is both hydrophobic and often contains STOP codons. As an approximation, the choice of expression in the BCR is between RF1 and RF2. The bias in expression of D RF in functional B-cells is accentuated by the accumulation of STOP codons in RF3. If D-diversity were the selection pressure, evolution could have easily eliminated those STOP codons instead of focusing them in RF3 to accentuate the bias in expression of the D RF1.

Can D-diversity be the function of the D_H-segment if selection operates on composition, not sequence?

Selection for composition, not sequence is a strong argument that D_H is selected as a framework, not complementarity-determining element. If its contribution to the repertoire were viewed as an entrained or fortuitous or second order property, then it would have to be minor. A preferred RF that is hydrophilic implies a structural role, for example, loop formation [22].

Selection on D_H as a framework element reduces the potential of the sequence diversity of the (NDN)_H region to contribute to repertoire size bringing it closer to being functional. This role has been described as D-delimit [12]. The overshoot in amino acid sequences relative to functional size of repertoire can be brought into line by degeneracy, delimitation or both. However, degeneracy is an inevitable factor; delimitation is unselectable. Degeneracy can validly be postulated to limit the combining site repertoire because it is an unavoidable and unselected concomitant of macromolecular interactions. D-delimit, like D-different, is, in a sense a fact, not, however, a satisfying interpretation of the selected D_H function because it implies an evolutionary pathway that is implausible. For evolution to generate a large repertoire encoded by N-additions that by its very size is nonfunctional (therefore, unselectable) and then introduce a gene element to reduce its size to a functional level, is hardly likely.

It might be pointed out that unlike the D_H-segment which is roughly 10-15 amino acids in length and encoded in the genome as a tandem gene locus of 10-20 copies, the TCR D_β-segment is 3 amino acids in length and encoded by 2 copies as is also the C_β-segment. Further, while the composition, not the sequence, of D_H is conserved throughout evolution [21], by contrast the sequence of D_β is tightly conserved.

The failure of D-diversity, D-differentiation, D-different and D-delimit to deal with “Why D?” in both the TCR and BCR requires that we look for a selectable function of D that operates on both D_H and D_β. What might that function be?

The proposition unifying the role of the D-gene segment in both the TCR and the BCR is that, in conjunction with N-additions, it functions as a framework element regulating the transmission of a signal to the cell upon interaction with ligand. This postulate must be compatible with the assumption that, while the N-additions were selected to function as framework in the BCR, they are complementarity-determining in the TCR.

How did we arrive at such a hypothesis?

The most popular model for haplotype exclusion at the H-locus is an H-STOP-H mechanism in which the first H-chain to appear shuts off rearrangement at the allelic H-locus. This model simply did not compute so we suggested a competing mechanism.

D-disaster

In 1987, we proposed that (NDN)_H was evolutionarily selected to play a role in haplotype exclusion at the H-locus [7-10]. Our minimal model [23] involved:

• 1) a fusion efficiency: the probability of an in-frame joint

• 2) a branching ratio: the order of expression, H then L

• 3) an (LH)₂-STOP to further rearrangements.

As an illustrative first approximation, if one considers the fusion efficiency (f) only, then, in the absence of a STOP condition at the H-loci, the proportion of $\frac{{\mathrm{H}}^{+∕+}}{{\mathrm{H}}^{+∕-}+{\mathrm{H}}^{+∕+}}=\frac{\mathrm{f}}{2-\mathrm{f}}$. At f=1/3, the proportion of cells producing two H-chains (H^+/+) to total H⁺ producing cells would be 0.2, much too high. If f is corrected for in-frame rearrangements of pseudo-V_H amounting to roughly half of all V_H, then the effective fusion efficiency would fall to 1/3 × 1/2 = 0.17 and the proportion of H^+/+ would fall to 0.09, still well above the experimental values that hover around 0.01. To account for this we postulated inactivation of the H-chain by the amino acid sequences in the third hypervariable region and pooled two sources, length matching between L and H and D-reading frame, referred to as a “fit factor” [7, 8]. A preferred D reading frame results in 1/3 of the H-chains being functional and N addition length matching leaves roughly 1/2 functional (fit=1/6). This reduces the proportion of H^+/+ to 0.014, an acceptable level. This assumption defined D as D-disaster. The more sophisticated analysis of haplotype exclusion can be found in ref. [23] which includes a computer program HAPEX, that is available on our web-site, www.cig.salk.edu.

If, for haplotype exclusion, it is essential that H-chains be inactivated then D_H had to be kept as a separate gene segment incorporated in three reading frames but functional in only one (to answer Question 1).

Why doesn't this appear to be required for the TCR?

In order to answer this we must ask, how does D_H in the non-preferred frames inactivate the H-chain and then extrapolate this answer to the TCR. D_H RF3 generally contains STOP codons and may be viewed simply as lowering the fusion efficiency. The choice then is between RF1 and RF2, RF1 being the preferred or antigen-selected frame.

How then might D_H RF2 inactivate the BCR?

A reasonable postulate would be that D_H RF2 is incompatible with a conformationally driven signal upon interaction of the BCR with its ligand (epitope). However, the present consensus is that the BCR delivers a signal to the B-cell simply by aggregation (i.e., only polymers are immunogenic). This implies that the B-cell would be unresponsive to monomeric antigens. Our position has always been that this is unrealistic and that B-cells can be inactivated (“tolerized”) by monomers [8, 24, 25]. A recent study supports this position [26]. Signaling by monomers would, of course, require a conformational change in the BCR on binding ligand that initiates self-complementation resulting in an “aggregation” signal ([Signal [1]) to the B-cell [27]. This assumption that ligand-binding initiates a conformationally-driven signal, is further supported by the independent observation that a single amino acid replacement in the constant region of the L-chain [28-30] that has no effect on the binding of an antigen, blocks the response of the B-cell to antigen [31, 32]. As the non-preferred D reading frame or a single amino acid replacement in C_L are unlikely to have an effect on a signal initiated uniquely by aggregation, we are left with a conformationally driven signal as our default position. In other words, D_H RF1 permits a signaling conformational change on binding ligand, whereas D_H RF2 does not, hence D-disaster or D-disruption of signaling.

Many of the arguments made here are supported by studies on the antigen-receptors of birds and rabbits (analyzed in [9]). Briefly, in species that generate their adaptive repertoires by gene conversion (e.g., chickens), a single rearranged acceptor (LH)₂ pair is diversified by gene conversion from a family of donor VD“J” gene segments. Striking is that:

1. the donor D regions are all in the preferred frame as VD“J”;

2. the acceptor rearranged H-chain transcription units contain D in all three RFs, with a proportion of them containing concatenated 2 or 3 D gene segments in random RFs. These acceptor H-transcription units are expressed in the cells that enter the bursa. They leave to the periphery as functional diversified B-cells that express uniquely the preferred RF. Whether this transition from a precursor bursal population expressing D_H in three RFs to a functional peripheral B-cell population expressing D_H in RF1 is due to bursal selection or another mechanism is not critical to our argument that D_H is not being selected as a diversity element.

These elegant studies of Reynaud et al. [33-37] lend support to the argument that D was selected to function as framework.

How does D-disaster explain that Dβ is expressed in all three reading frames in the TCR?

Under the Tritope model [38-42] the TCR functions in two signaling orientations. The V domain, Vα or Vβ, that is positively selected in the thymus determines the restriction specificity whereas the entrained V domain determines the allospecificity of the given TCR. The signaling orientation that is positively selected in the thymus is used for restrictive reactivity. If the TCR is engaged in the opposite signaling orientation, alloreactivity can be mediated.

Two signaling orientations require that the TCR be born in one of two conformations, the positively selected conformation determines restrictive reactivity.

As the Vα and the Vβ domains of the TCR are germline-encoded and not somatically varied, the determinant of the two conformations can be reasonably surmised to reside in the Dβ RF. For example, Dβ RF1 might be positively selected when Vα encodes the restricting specificity and Dβ RF2/3 when Vβ encodes the restricting specificity.

If a given V encoding the restriction specificity is associated with a Dβ in the wrong or discongruous RF, the T-cell will be treated as alloreactive and be negatively selected by the thymic MHC-encoded restricting element (R_T) [43-45]. Using the above illustration, if Vα is associated with the discongruous Dβ RF2/3 then upon interaction with the thymic restricting element (R_T), it will treat that R_T as allo-R and be deleted (D-disaster).

The reason that all three Dβ RF are expressed in the population is that TCR's are functional in two signaling orientations each of which is associated with a specific D_β RF, Dβ RF1 or Dβ RF2/3. It is the Dβ RF that determines the conformation responsible for the signaling orientation.

This unifies the role or D. In both the BCR and the TCR, the D RF is postulated to determine the conformation that permits the receptor to transmit a signal upon encountering its ligand.

Illustrations from the data with suggested experiments

A. The D_H reading frame

Coleclough [46] and Kaartinen and Makela [47] were the first to note the existence of a preferred D_H reading frame; it has since been confirmed in numerous studies (reviewed in [22, 48].

Not examined in detail experimentally is what happens to the cells expressing a BCR with D_H in a non-preferred D_H RF. If the fit factor is 1/4[1/2 (D_H in RF1) × 1/2 (length matching)] then only 1/4 of newborn B-cells will be functional. Of the nonfunctional remainder, 2/3 will express the nonpreferred D_H RF2. Inactivation due to length matching is not easily determined. How is the non-preferred RF somatically selected against or how does the non-preferred D_H RF inactivate the BCR?

This question deserves detailed analysis.

1- A comparison should be made between the distribution of D_H RFs in peripheral B-cells from normal adult animals and those under minimal antigenic selection, (e.g., germ free, athymic, processing defective or class II MHC negative mice). If the comparison shows a significant difference between mutant and wild type (e.g., ∼50% of B-cells from the mutant express a nonpreferred D_H RF compared to <20% in wild type) then the B-cells expressing the non-preferred RF simply turn over as would any unengaged cell. Given homeostatic regulation of the total numbers of B-cells per gram, these nonfunctional B-cells could play an important buffering role in protecting against excessive washout of functional but unengaged B-cells by the engaged B-cells (see detailed discussion in ref. [8]).

As an estimate from the data, ∼80% of B-cells in normal animals express D_H in the preferred RF1. The 20% in the non-preferred D_H RF2 could come from either 1) functional B-cells expressing specificities that use D_H RF2 (D-diversity) or 2) nonfunctional B-cells, that given homeostasis, simply turnover at a rate dependent on the washout by proliferating cells induced by the immunogenic load (D-disaster). In the absence of interaction with ligand, the B-cell has no way of knowing what is the reading frame of the D_H in its BCR. BCRs in both categories are indistinguishable by the B-cell prior to encountering ligand.

The above experiment addresses that question. The suggestion that D_H RF2 makes for defective assembly at the level of pre BCR is a form of D-differentiation and the proposal that D_H RF2 specifies autoreactivity while D_H RF1 does not, denies D-diversity as this implies a repertoire that is nonrandom with respect to the recognition of self and nonself. Besides, it is not possible to select in the germline for autoreactivity (meaning debilitating autoimmunity), not to mention, what does it accomplish in a D-diversity framework?

It is possible to envisage a special case in which a given V_LV_H pair would be uniquely associated with a nonpreferred RF. If recognition of an odd epitope required a unique geometry of the paratope (flat or protruding) that depended on a hydrophobic (NDN)_H, there could exist for that V_LV_H pair a hydrophobic preferred RF. However, it would have to be at the same time permissive of a signal making such an event quite rare. Examples where B-cells are induced BCR- or Signal[1]-independently, do not illustrate this hypothetical case.

2- Two transgenic BCRs in RAG^−/− animals that differ solely in the D_H reading frame should be compared. The BCR anti-phosphorylcholine (PC) with D_H RF1 might be a viable candidate. Expressed with D_H RF2, assuming that it retains its anti-PC specificity (itself an important question), one can ask if it also retains inducibility. Predictably examples of such pairs with D_H in RF1 and RF2 can be established that leave specificity essentially unaltered but inducibility impaired. This would confirm that D_H controls a conformationally driven signaling.

Two probing recent studies bear on this question.

Butler et al. [49, 50] analyzed the response in piglets to a virus [porcine respiratory and reproductive syndrome virus (PRRSV)]. This virus, antigen unspecifically, activates the humoral system which lights up like a Christmas tree. Whatever the mechanism of the nonspecific-activation or BCR-independent induction by the virus, if B-cells were randomly turned on, D_H RF independently, then ∼50% would express the nonpreferred RF assuming of course that these nonfunctional cells, like the functional cells, are only ridded by turnover (washout).

Butler et al. [49, 50] interpret their findings as showing that “PRRSV infection causes generalized Ag-independent B-cell activation and hypergammaglobulinemia with biased expansion of a subpopulation of the preimmune repertoire with hydrophobic binding site that normally disappears during Ag-driven diversification.” We question the assumption of “biased expansion.” Under their view no signal via the BCR is required for the virally induced response. Consequently the cells expressing D in the non-preferred frame (hydrophobic) would be indistinguishable from the cells expressing D in the preferred frame (hydrophilic). Triggering a B-cell response that bypasses a signal via the BCR would assay the ratio of preferred to non-preferred expression of D_H in the population. If no antigenic selection were operative this ratio would be ∼1. In the population responding to antigen, only the preferred frame (hydrophilic) would be expressed.

In other words, the response to PRRSV illustrates unbiased expansion. In order to illustrate the point albeit simplistically, consider only D and N as determining whether the BCR is functional in signaling. Essentially the contribution of the rearrangement of nonfunctional V-gene segments is being ignored. There are four categories of (NDN)_H rearrangement. For illustration, given 2 D RFs, one of which is functional (f), the other nonfunctional (nf) and that the probability that an N-sequence is functional is 0.5, then four categories of (NDN)_H can be envisioned, D^fN^f, D^fN^nf, D^fN^f and D^nfN^nf each at a frequency of 0.25. The nonfunctional cells in the absence of immunogenic encounter are indistinguishable from the functional B-cells. If all cells were turned on indiscriminately D^f/D^nf = 1. As the immunogenic load selects D^fN^f, if this raised its frequency from 0.25 to 0.5, then the nonfunctional remainder would fall from 0.75 to 0.5. Each category would be present at a frequency of D^fN^f = 0.5, D^fN^nf = 0.17, D^nfN^f = 0.17, D^nfN^nf = 0.17. The ratio of D^f/D^nf = 2 (0.67/0.34), essentially a doubling.

At the present moment, there is no direct evidence that a BCR with D_H in the nonpreferred frame (hydrophobic) can deliver a signal to the cell on binding its specific ligand (Signal [1]). Expression of the nonpreferred D_H RF in B-cells activated abnormally as innocent bystanders is no challenge to the D-disaster postulate. In fact, it is predicted.

B. The Dβ reading frame

As the expressed Dβ RF will depend on whether Vα or Vβ was positively selected, an analysis of its role requires a basis for deciding which one was selected. One good example can be found in an investigation by Turner et al. [51]. They studied the cytotoxic response to a glycopeptide from a herpes simplex virus that is presented by H-2K^b. The TCRs of independent CTL clones preferentially use Vβ10 associated with a variety of Vαs (Figure 2 in Ref. [51]). This implies that Vβ10 was positively selected to recognize K^b and the various entrained Vαs encode alloreactivities. All 14 independently derived Vβ10 clones expressed Dβ RF3 which, then, can be postulated to determine the TCR conformation required when Vβ is positively selected to specify the restriction specificity, in this case anti-K^b.

C. The D_H in HV3: is it CD or FW?

It might be of value to recall that the shape of the BCR combining site depends on the conformation of its framework. Therefore, it is expected that some of the sequence replacements in the framework would affect the binding properties. However, such framework replacements are either neutral or selected against because they affect the ability of the BCR to signal the cell on binding antigen. Replacements in complementarity-determining (CD) positions are selected for because they change specificity without damaging signaling function.

The characteristic of a CD position is that the mutational replacements in it show a replacement (R) to silent (S) ratio of >3. By contrast FW positions show an R/S ratio of <3. Functional replacements are selected for (R/S>3); deleterious replacements are selected against (R/S<3). This criterion should be applied to the D_H in the HV3 region. A mouse engineered to express a single well-defined rearranged H-chain should be analyzed for the R/S values of mutational replacements in the HV3 region. The control would be V-region CD vs FW. A statistically significant collection of mutational replacements should be a good assay of whether or to what extent the HV3 region is functioning as CD or FW. If a sufficient number of mutational replacements are studied, an answer to the roles of D and N will emerge.

Of great interest would be to carry out this study in mice engineered to express either one V_H and multiple D_Hs [5] or multiple V_Hs and one D_H [52]. The R/S ratio of mutational replacements in each of the D_H RFs would define its role, certainly be able to rule out D-disruption of signaling.

If it were true that one V_H and multiple D_Hs [5] or multiple V_Hs and one D_H [52] were equivalently functional and sufficient for immune function, then the size of the repertoire would have to be determined by a non-H element like the L-chain. More than likely they are not equivalent and the behavior of multiple V_Hs and one D_H more closely reflects normal responsiveness. In this case the role of multiple D_Hs would be to increase the rate of D_H→J_H joining during development by increasing the target size [9].

If one out of 12 D_Hs permits generation of the entire functional repertoire [52], then D_H must function as framework. D-diversity would predict huge holes in responsiveness. To argue that each D_H interacting with a given V_LV_H pair generates its own unique subrepertoire simply negates selection on given V_LV_H pairs for the specificities they encode. Further, “subrepertoire” has no heuristic meaning. In principle, if there were only one framework for all V_Hs and one for all V_Ls, the differences residing solely in the CDRs, the repertoire would be unchanged, as is found for D [52].

If junctional diversity is too-much-of-a-good-thing for the BCR, why is it merely adequate for the TCR?

Answering this question is certain to engender heated debate. The difference is in the ligand, a shape-determinant for the BCR and a linear peptide for the TCR.

The BCR is dependent almost entirely on V_LV_H complementation to determine the specificity of its combining site and therefore its repertoire. In mouse and human, the size of the repertoire is determined by a high copy number family, namely the germline-selected V_LV_H pairs and their random complements, to which is added a low copy number mutational repertoire. The huge amino and sequence diversity of (NDN)_H plays a minor to negligible role in contributing to the size of this repertoire.

The TCR anti-P site recognizes a helical peptide buried in an MHC-groove. It engages in complementarity-determining interactions with between 3 and 5 exposed residues of the peptide. If on average 4 residues are engaged then the size of the TCR anti-P repertoire would be ∼10⁵ (20⁴) with boundaries between 10⁴ and 10⁶. The limit to the size of the repertoire is not determined by the sequence diversity of the (NDN)_β region but by the ligand it recognizes. In the end, the sizes of the TCR and BCR repertoires are capped at roughly the same value, 10⁵. This is easily rationalized as T-help is required for the activation of the B-cell. It would be useless (unselectable) to have the two repertoires, TCR and BCR, enormously disproportionate in size.

The unfinished chicken story

The general picture as it is viewed by Reynaud and Weill [53] is that D_H RF1 is selected for in the bursa as the only frame that permits the cell to respond to the bursal signal to proliferate. The cells expressing non-preferred RF2 are deleted by turnover in both bursal and extrabursal spaces. We had postulated originally that the precursor cells enter the bursal follicles in D_H RF2 which was gene converted to D_H RF1 [9] but this is probably wrong [53]. We agree!

If bursal selection for RF1 is coupled to our suggestion [9] that the single acceptor V_LV_H pair is germline-selected to recognize a bursal follicular component as a homing requirement, then several important experiments can be envisaged that derive from the observations of Huang and Dreyer [54] and Granfors et al. [55].

Chicks bursectomized at embryonic day 3 display as adults a high level of serum IgM, IgG and IgA yet they do not respond to a variety of antigens [55]. These serum Igs display a limited Hμ, Hγ and Lλ chain sequence diversity assayed by two dimensional electrophoresis [54] suggesting that a unique antibody is involved.

The questions to be answered are:

1. Is this serum Ig, germline-selected, specific antibody for a bursal follicular component?

2. If the serum Ig is specific for the bursal follicle, in bursectomized animals, is it expressed with both D_H RFs? Is one of the D_H RFs absent in that antibody in normal animals or is that antibody absent?

3. Do IgM B-cells express both RF1 and RF2, while IgG and IgA B-cells express only RF1?

Answers to these questions using embryonically bursectomized chicks compared to normal could settle the role of D_H, certainly disprove that the D RF regulates the transmission of a ligand binding signal.

The limitations of the data used to analyze the D-disaster postulate for the TCR

While it is easy to gather examples where the Dβ RF appears to correlate with the Vα or Vβ domain that was positively selected for its restriction specificity (i.e., Vα and Dβ RF1 or Vβ and Dβ RF2/3), there is a significant number of cases which either contradict the postulate or where D_β is absent or its reading frame cannot be determined.

There are at least four limitations to the data that account for the contradictions with this theory, some of which applies to the BCR as well.

1. Although the use of hybridomas to analyze the Dβ RF favors antigen-stimulated (activated) cells, this technique also picks up a small but significant number of non-activated cells that could have D in the non-preferred or discongruous reading frame. Once in a hybridoma the signaling behavior of the BCR/TCR is either not checked or, even if it were, might be abnormal.

2. Simple binding assays do not imply successful signaling (i.e., a functional receptor). Ability to signal must be assayed.

3. In cases where D is absent or its reading frame is unclear, the pattern of N-additions might compensate to simulate a signaling frame. After all, these cases were under strong antigenic selection requiring functional signaling.

4. The peptide used in a particular investigation might interfere with the expression or recognition of the allele-specific determinant required for docking.

Of course, if these sources of “incompleteness” were resolved, the data would disprove the role of D as the control element of conformationally driven signaling.

A definitive experiment

This proposed experiment tests the assumption that the TCR functions in two signaling orientations that require two distinct initial conformations determined by the Dβ reading frame.

Two reciprocal populations of CD4⁺ T-cells should be established from MHC congenic mice lacking expression of H-2E (e.g., B10.H-2^b and B10.H-2^s).

• 1) an H-2A^b anti-H-2A^s population

• 2) an H-2A^s anti-H-2A^b population

One population will be restricted to A^b and alloreactive to A^s. The other will be restricted to A^s and alloreactive to A^b.

From each population clones of T-cells that use identical Vα and Vβ domains should be selected. There will be two such families in each population given the postulate that two signaling orientations are involved (Table 1).

Table 1.

The families of TCR in each signaling orientation

	H-2Ab anti-H-2As	H-2As anti-H-2b
Family 1	Vαb Vβs (RF 1)	Vβs Vαb (RF 2/3)
Family 2	Vβb Vαs (RF 2/3)	Vαs Vβb (RF 1)

The junctional region of the (NDNJ)β of the pairs with identical VαVβ domains should show non-overlapping Dβ reading frames as illustrated. If this predicted result is confirmed then it can be concluded that Dβ RF determines the signaling orientation of the TCR. Equally important although not discussed here, the predicted results would demonstrate that single V domains recognize the allele-specific determinants on the MHC-encoded restricting elements [38-42].

An unresolved structural problem

The D_H gene segment encodes roughly 10 amino acids and can be reasonably envisioned to regulate a conformationally-driven signal by the BCR. The Dβ gene, by contrast, encodes three amino acids, one or two of which are sometimes deleted during rearrangement. How such variability over such a small region can regulate the signaling orientation is a serious unanswered problem that will be highlighted if the above outlined experiment yields the predicted result. At the moment, there is considerable resistance to the postulate that a conformational change in the BCR or TCR is required to initiate a signal to the cell. Yet much of the physiological behavior of BCR/TCR is best explained by such an assumption. As pointed out earlier a single amino acid replacement in the constant region of the light chain of the BCR can obliterate signaling without affecting binding to ligand [29-32]. This is unlikely to be the property of an aggregation-only signal. It would not be unreasonable then to assume that a single amino acid in the NDN region could do the same thing.

Reiterating the two major objections

There are two points of contention with the position developed here that are absent in the literature but voiced at meetings and by reviewers.

First, it is claimed that the expression of D_H in a preferred frame, is far from absolute. A significant proportion of peripheral B-cells express D_H in the nonpreferred reading frame. Therefore, it is argued, D-diversity is unchallenged and the cited difference between the expression behaviors of D_H and Dβ is, in fact, non-existent.

Second, it is claimed that the BCR and TCR interact with ligand in a similar manner making the suggested experiments irrelevant.

The first contention is based on a failure to make a distinction between D_H which is incorporated into the transcription unit in all three reading frames (RF) and D_H which is expressed in functional BCRs in a preferred frame. Functional means that the BCR must be able to transmit Signal[1] to the B-cell upon binding ligand. Data on the RF of D in BCRs in the circulating pool of B-cells do not assay functional signaling. The B-cell has no way to know if its BCR is capable of transmitting a signal in the absence of interaction with ligand. Therefore, the nonfunctional B-cells are unavoidable normal constituents of the circulating pool. Their level depends upon the steady state immunogenic load. A major part of this load is autogenous waste due to senescing and necrosing tissue cells.

The second contention only highlights the competing models of TCR recognitive interactions. The very fact of restrictive recognition by the TCR, absent in the BCR, is a strong argument that the two receptors have a fundamentally different behavior [38-42]. If the above suggested definitive experiment turns out as predicted, the role of D as a signaling gate would be settled and this second caveat eliminated.

Lastly, for argument's sake, let's accept the skeptic's assumption that, in conjunction with N-additions, that D_H is expressed in functional BCRs in three frames as “CDR3.” Until the relationship between the enormous amino acid sequence diversity and the limited functional combining site diversity is faced, D-diversity as a selectable target will remain as just a popular guess, not a theory. One has only to compare the putative “CDR3” with the well-defined CDR1 and 2 to appreciate this point. Big Bang expression of a huge random repertoire is unselectable. Only stepwise selection can contribute to the repertoire. In essence, raising this point returns us to the discussions of the 1970s and 1980s when germline vs somatic theories were debated [56].

The bottom line

The assumption that the reading frame of the D gene segment expressed in either the TCR or the BCR determines its signaling behavior, is a singular assumption that explains a wide variety of data. If correct, the D reading frame can only regulate a conformationally driven signal. For any given TCR or BCR, D in one reading frame is functional and in the other two is not. For the BCR this is postulated to play a role in haplotype exclusion whereas for the TCR the D reading frame is postulated to determine the signaling orientation used for restrictive reactivity.

The logic of the D-disaster postulate is based on two assumptions:

1. D plays a similar selectable role in the TCR and BCR;

2. D is a separate gene element because waste of 2/3 of cells is an essential step in haplotype exclusion at the BCR H-locus and for the determination of the signaling orientation of the TCR.

In both cases the reading frame determines functional signaling upon binding of ligand. Those cells that do not display a congruous reading frame for signaling die. In the case of the BCR the nonfunctional B-cells form part of the B-cell pool and, given homeostasis of total B-cells per gram, they turnover (are washed out) at a rate dependent on the proliferation of the engaged functional B-cells. In the case of the TCR with D_β in the discongruous reading frame docking on the allele-specific determinant of the MHC-encoded restricting element in thymus, results in a tolerigenic signal (Signal[1]), peptide-independently, identical to that found for allogeneic as distinct from restrictive interactions [43-45].

List of abbreviations

TCR

T-cell antigen-receptor

BCR

B-cell antigen-receptor

RF

reading frame of D-gene segment