On Recognizing ‘Shades-of-Gray’ (Self–Nonself Discrimination) or ‘Colour’ (Integrity Model) by The Immune System


Correspondence to: Z. Dembic, Molecular Genetics Laboratory, Department of Oral Biology, Faculty of Dentistry, University of Oslo, Sognsvannsveien 10, PB-0316 Blindern, Oslo, Norway. E-mail: zlatko.dembic@odont.uio.no


The aim is to discuss Cohn's T-cell receptor (TCR) Tritope model of recognition, propose a novel suggestion for prior-to-positive selection of thymocytes contributing to inherent major histocompatibility complex (MHC) reactivity of a T-cell repertoire and clarify the Integrity model about the function of the immune system. If we compare the perception of light with the recognition of nonself, we could imagine that the opacity might be a measure of docking interaction between specific receptors for antigen on T or B cells (TCR / peptide–MHC or BCR / antigen). From this viewpoint, the self–nonself discrimination (S-NS) metaphor would be perception of black (self) versus white (nonself). However, whereas detection of shades-of-gray suffices to describe S-NS discrimination principle, colour vision of the antigenic world portrays best the Integrity model. In concert with recognition of opacity, the Integrity model proposes detection of at least three colours (signals): red (harmful), blue (useful) and yellow (the rest, including homoeostatic ones). As a result, recognition of nonself is transferred into communication within self while deciding on type of the immune response. Hence, the S-NS discrimination model seems to be an oversimplification, because it fails to see colours and consequently lacks the need for suppressor/regulatory function. Similarly, the Danger model stops short of detecting being useful signals that confer immune asylum to helpful micro-organisms like commensals. I suggest that the immune system's repertoire for recognition, in general, has evolved by a novel drive called ‘natural integrity’ alongside natural selection, thus facilitating communication between cells of the immune system.


The Integrity hypothesis [1-5] is simplified digitized vision (having 3 basic, grouped signals) of the function of the immune system (Figs. 1 and 2). While theoretical concepts customarily use digital signals (signal 1, signal 2, etc.) to describe biological systems, the nature of such signals is analog. In principle, the digitized view helps in understanding how a cell makes a decision when a particular threshold has been reached. These choices we can observe and measure. For the immune system, the most important assessments are activation, proliferation and differentiation. We tend to overlook cell communication and mention it only as a mechanism in achieving these decisions. Cells of the immune system [lymphocytes, antigen-presenting cells (APCs), as well as other haematopoietic and stromal cells in lymphoid tissue] according to integrity model form a network of communication signals that allow them to ‘think’ – as in some subconscious neural network – by which they choose a response to input signals. The (immune system's) response is not metaphoric discrimination between self and nonself, but rather a decision on an action ahead, based, at its simplest level, on a triple choice scenario (see later, the Integrity model), which in reality might be more complex.

Figure 1.

The Integrity model pt.1: APC- and T-cell activation. Antigen-presenting cells like dendritic cells (DCs) and macrophages receive three signals that lead to their activation and differentiation (including migration of DCs). They communicate with T cells. T cells are activated by three signals that lead to their differentiation and/or proliferation. Effector T cells also receive three signals (not shown) to exert their function on targets. The presented concept is the elaboration of previously published Integrity model [1, 2, 4].

Figure 2.

The Integrity model pt.2: B-cell activation. Simple digital scheme of B-cell activation by three signals that lead to their differentiation and/or proliferation. Effector B cells (and plasma cells) also receive three signals (not shown) to exert their function. This is an updated model based on previous reports [1, 2, 4].

The Tritope model by Cohn [6-9] is a molecular view (and a prolongation) of the self–nonself discrimination (S-NS) theory by Bretscher and Cohn [10] on the recognition of antigens by the immune system – in particular by a species-selected T-cell receptor (TCR) genetic material. It describes how evolution would act upon pairs of V regions by selecting various combinations of alpha and beta chains to recognize species-specific MHC molecules as self, focusing on single-TCR-chain-specific interactions. This sets his model aside from the Standard one. Namely, Cohn believes that two chains can yield two different signals, whereas the Standard model sees TCR as a single-receptor signalling unit [11-14]. Cohn plausibly discusses how peptide recognition could have been incorporated in the positive selection process and explains it such that only certain preferential combinations (of TCR chains) would work well, leaving other pairings non-viable [6-8, 15]. The argument introduces different orientations of the TCR docking on peptide–MHC (pMHC) that is well documented, and it could provide the basis for understanding individual differences observed in frequencies of various TCR variable regions (as illustrated by their differences in MHC congenic mouse stains in his article presented in this issue) [9].

Criticism of the Tritope model by Cohn

Let us get straight to the core of the discussion here: Cohn [8, 9] disagrees that T-cell repertoire can be evolutionarily shaped solely by the thymic positive and negative selection as described by the Standard model [11-14]. Namely, the Standard model says ‘useful repertoire of TCRs in any given animal is stamped with a faint imprint of what has been selected over time in the species [16]’. Cohn argues that their statement ‘…in the absence of negative selection, TCRs can react with MHC proteins in a class and allele-independent fashion’ [16] is logically wrong [8], because evolution would not be able to naturally select TCR V region alleles whose only function is to be deleted by the negative selection in the thymus. In other words, he claims that TCR–pMHC interaction ability was evolutionarily selected by docking of single V region of TCR in allele-specific MHC fashion. Here, I agree with Cohn, because I do not consider thymic selection as a manifestation of only natural selection, but a representation of two natural forces.

Cohn's solution (as mentioned above) is explained in depth by his Tritope model of TCR docking during T cell recognition of pMHC. The model reasonably excludes negative selection in the thymus in moulding inherent MHC reactivity of any T cell repertoire. He documents his case by re-examining past experiments and data on MHC-restricted T cell reactivity in mixed lymphocyte-culture reactivities (MLR) of various MHC congenic mouse strains and by correlating frequencies of particular TCRV regions in them [9].

I wish to start by discussing the latter topic first, then continue with the argument concerning inherent MHC reactivity of a T-cell repertoire within the Tritope model and finish with my own (Integrity) model about the function of the immune system that views the world of molecules and microbes in our surroundings a bit differently from the Tritope model.

The Mls controversy revealed

In the Tritope model of restrictive recognition, a particular TCR Vβ region could be either in a restrictive orientation (accessible for positive selection in the thymus) or in an entrained (allo- or syn-)orientation (not accessible for positive selection). Cohn suggests [9] that Mlsa is complexed with MHC class II (MHCII) molecule such that it is recognized by thymocytes with distinct pairs of TCRs having either Vβ8.1 in restrictive or entrained orientation. Of course, both orientations (as they become fixed after positive selection, according to Cohn) can mediate negative selection.

Cohn claims that a support for his view is derived from the data concerning Mls system (a minor histocompatibility antigen of the mouse) and several other studies that deal with frequencies of various TCR variable regions [9]. In short, the frequencies of the TCR Vβ6 and Vβ8.1 were discussed regarding MHC congenic mouse strains. The MHC class I (K, D) and class II (A, E) molecules are listed in such strains in the following order: KAED. Haplotypes d, k and b represent the most common ones experimentally studied. Of interest are also their recombinant (sub)strains, established during the initial mapping of the H-2 (MHC) complex in the mouse (Table 1).

Table 1. Correlating Mlsa facts with TCRVβ8.1 frequencies in MHC congenic mouse strains and explanations. The expression of MHC class I (K and D) and class II (A and E) molecules on antigen-presenting cells (APC) in MHC congenic mouse strains carrying ‘a’ variant of the mixed lymphocyte culture-stimulating (Mlsa) antigen and frequency of Vβ8.1-positive T cells in their periphery (lymph nodes). The reactivities of Mlsb-(T cell) hybridomas (from Mlsb derivatives of listed strains) are shown as ‘recognizing Mlsa’ (+) or ‘non-recognizing Mlsa’ (−) in MLR with Mlsa-positive targets (of listed strains). Cohn's explanation is based on his Tritope model: restrictive orientation of a TCR subunit will lead to positive selection; entrained TCRs subunits (in allo-orientation) will be negatively selected (synreactivity). Alternative explanation is my own interpretation
TCR Vβ8.1 chain involvementExplanations
Mouse strain (Mlsa+)MHC (H-2) haplotype (KAED)Thymic deletion (frequency of Vβ8.1 T cells %)Mlsb H-2k hybridoma reactivity (MLR)Cohn's explanation, Mlsa as a mimic of MHC (Fig. 3)Alternative explanation, Mlsa as superantigen (Figs. 3–5)
  1. a

    Frequency of Vβ8.1 T cells in Mlsb strains with H-2 k,d haplotypes is about 5%.

  2. P1: prediction #1 – There should be alloreactivity to Ab by H-2k(Mlsb)-derived-hybridoma-TCR with Vβ8.1 (a discrepancy, not shown). Similarly, there should be synreactivity of Vβ8.1-TCRs during positive selection by Ab. Thus, the intermediary steady state levels of TCR chains with Vβ8.1 (in H-2b Mlsa strain, 1.7%) could be due to its restrictive orientation and ‘halved’-positive selection in the thymus of the H-2b mice (selection is only by Ab, as there is a lack of E molecules), as well as only negative selection on the half of the Class II molecules’ content in the thymus. The latter possibility is preferred by Cohn in his explanation.

  3. P2: prediction #2 – This explanation involves that there is a mistake in genotyping this strains'Ad molecule (D2.GD). If the sequence is correct, then it is a contradiction. Namely, if the TCRβ is in allo-orientation it should be deleted in the thymus, but it is obviously not – it is positively selected, and thus a paradox to Cohn's explanation. The only way out of this contradiction is to predict a sequencing (or serologic typing) mistake for this strain's Ad molecule.





+ (0.2–0.6)a

+ (0.2–0.6)a

+Vβ8.1 binds A/Ed,k-Mlsa. Vβ8.1 is in the allo-orientation (entrained, and amenable to negative selection)Mlsa binds Vβ8.1 – math formula
D1.LPbb-b− (1.7)+Vβ8.1 sees Ab restrictively or allo-(syn)reactively. But it sees Class II-Mlsa only when allo-oriented (P1)Mlsa binds Vβ8.1 – math formula, in math formula complex on APCs (due to MHC-expression boost). The unusual math formula complex has a low expression in the thymus (as it lacks the expression boost)
D2.GDdd-b− (~ 5)aVβ8.1 does not bind Ad in this strain. P2: Is there a mistake in the DNA sequence of Ad in D2.GD?Mlsa binds Vβ8.1 - math formula, but the complex math formula is less preferentially formed (than math formula), and thus has a very low expression in APCs and in the thymus

The critical point here is the b (H-2) haplotype, because it lacks expression of the E molecule. Hence, its designation is as follows: bb-b (D1.LP), while the others have dddd (H-2d) and kkkk (H-2k). The recombinant mouse strain D2.GD has dd-b haplotype. The haplotypes d, k and b that have Mlsa were reported to delete TCRVβ6-expressing T cells during negative selection in the thymus, whereas Mlsb mouse strains do not (Table 1). Similarly, TCRVβ8.1-expressing T cells were deleted in d and k, but interestingly not in the H-2b haplotype mice (with Mlsa). Cohn argues that specificities of Mlsa-reactive T-cell hybridomas should be indicative of such deletion process in that that they should be reactive towards the negative selection MHC determinant molecule (in combination with Mlsa). However, there is a discrepancy. Vβ8.1 carrying hybridomas (derived from Mlsb mice) are reactive towards dddd and bb-b, but not to dd-b targets having Mlsa [9]. So, Cohn asks, if Vβ8.1 reacts against H-2b (bb-b), why are Vβ8.1-carrying CD4 T cells not deleted in the thymus of such (Mlsa) mice? Cohn suggests ‘…Mlsa binds to A creating determinants that act as mimetopes of restriction epitopes seen by both Vβ6 and Vβ8.1, and that Vβ8.1, but not Vβ6, is positively selected by Ab’ [9]. As mentioned earlier, there is a contradiction between Mlsa carrying dddd and dd-b haplotypes, because the former deletes Vβ8.1 in the thymus and the latter does not (Table 1). The solution by Cohn is that dd-b strain (in this case, D2.GD, with Ad) has a genotype typing error (for example, a mutated Amut d ? molecule that cannot bind Mlsa), which I deem not likely. And, even with such mistake in DNA typing, another requirement – for his explanation to hold – remains, that is, the assumption that Vβ8.1 is recognitive of Ab itself would still be required to explain the contradiction, namely the stimulation of IL-2 production in the (Mlsb, kkkk) hybridomas by Mlsa-D1.LP(bb-b)-derived (Ab-containing) targets and the lack of negative selection of Vβ8.1 by D1.LP(bb-b) mice (Table 1 and [9]).

Furthermore, the failure to demonstrate alloreactivity in experiments using Vβ8.1-expressing hybridomas from TCR Vβ8.1 transgenic mice (H-2k) to be stimulated by H-2b cells [17] is an unresolved contradiction to the model, acknowledged by Cohn. Explicitly, because all these hybridomas should have Vβ8.1 in alloreactive orientation, the discrepancy is explained as a peculiarity of the H-2b MHC background that constituted an exception [9]. It implies that Vα subunit (in this particular case, Vα11.1) inhibits signals from Vβ8.1 subunit. This is testable, but if shown to be the case (as Cohn suggests, by employing MHC molecule as a target [Aαdb] that Vα11.1 does not recognize) [9], then I would retract the following contra-arguments: (1) TCR heterodimer does not signal with two different signals (from two different orientations) and (2) it seems unlikely that evolution would ‘invest’ in such complex signalling only to waste energy by allowing cross-inhibition of Vα signals by Vβ signals (and vice versa) controlled by non-preferential alleles of the MHC. Therefore, to my mind, this seems not to be the simplest explanation for the TCR–MHC interactions. The other explanation that I prefer is that TCR signals as a single unit, like the standard model suggests, but with additions to the model that we should consider in further details later.

In other words, if Cohn's Tritope model is right, why is there a difference between orientations of Vβ8.1 in d, k versus b MHC haplotypes? Is unknown background problem the only solution why Vβ8.1 is not positively selected in d and k haplotypes?

I argue that whether TCRβ chain is in entrained or restrictive orientation is irrelevant in this case (and possibly in all cases, but let us discuss that separately), because there is, in my view, an alternative explanation involving preferential pairing of MHC class II α and β chains in assembling heteroduplexes, which unfortunately and similarly to Cohn's explanation lacks evidence (Table 1). Firstly, let us suppose that Mlsa binds E molecules in k, d and b haplotypes and provides a superantigen-like molecule (not a mimetope) that binds with the TCR Vβ8.1 region. The bb-b haplotype lacks E molecule due to a deletion in the promoter of the gene encoding Eαb [18]. However, the gene for the E beta chain is unaffected and provides a source of free Eβb chains that could be associated with Aαb forming perhaps unusual [Aαbb] heterodimer. There is no evidence for such a complex (reviewed by Germain [19]), although the possibility remained open after demonstration of the alternative unusual heterodimer [Eαdb] by Mineta et al. [20]. Therefore, the unusual putative [Aαbb] assembly could still happen, perhaps in low but still functionally relevant frequency and only in professional APC in the periphery of the immune system, for example, during T cell–DC interaction in the lymph nodes, when MHC expression might get enhanced. Such circumstances would not suffice to form enough unusual [Aαbb] heterodimers in those DCs migrating into the thymus or in thymic (medullar or cortical) epithelial cells. It follows, the Mlsa cannot bind to the Eβb molecule to delete the Vβ8.1-carrying thymocytes in the thymus, but it can provide docking targets (on APCs) recognized by Mlsb-derived H-2k hybridomas in the above-mentioned experiments (Table 1).

This explanation would work only if the Aαd and Eβb (in dd-b haplotype) would not associate well (or at all). This can be tested experimentally.

Furthermore, Cohn mentions frequencies of mouse TCR V regions in congenic mice strains that either lack or have a distinct combination of MHC molecules that support his claims about dual-site TCR recognition and signalling [9].

Firstly, he claims that contrary to Gahm et al. explanation [21], tolerance is the reason that C57BR does not express Vα11-Vβ3a or Vβ3a-Vα11 TCRs, being, respectively, in restrictive–allo-orientations, and that for the same reason, neither does the (B10.BR X C57BR)F1 [9]. In this F1 strain, 95% of Vβ3b is paired with Vα11, but none of the latter is associated with Vβ3a [21].

Cohn then suggests [9] that his ‘interpretation can be tested by analysing the response of TCRs using Vβ3a in an MHC haplotype lacking expression of E’. The TCRs should be alloreactive to Ek and to most of E as Eα is minimally polymorphic [22]. However, his suggestion has two problems:

  1. It is in contradiction within his paragraphs [9] where he discusses occurrence of Vα11–Vβ3a pair in Ek-restricted pigeon cytochrome C (PCC)–specific T-cell lines isolated from C57BR. He firstly states a fact, ‘… in cell lines derived from C57BR…, the Ek-restricted anti-PCC response was associated with Vα11-VβXs (not Vβ3a) and with Vβ3a-VαXs (not Vα11)’ and then gives his interpretation that ‘Tolerance is the reason that C57BR does not express Vα11-Vβ3a or Vβ3a-Vα11 TCRs’. Then, he contradicts these facts and his suggestion with yet another suggestion (discrepancy is marked in bold): ‘The Vβ3a-selected cell line derived from C57BR consists of two lineages. Lineage 1 with a TCR, Vβ3a-VαXs, is Eα-restricted by Vβ3a, specific for PCC, alloreactive via VαXs to a variety of MHC alleles, but blind to the ‘superantigens’ SEA and Mlsc. Lineage 2 with a TCR, Vα11-Vβ3a, is Eβk-restricted, specific for a variety of random peptides (not particularly PCC), synreactive to Eα, and responsive to the ‘superantigens,’ SEA and Mlsc’.
  2. Furthermore, such test would be flawed, because the proofs would stem out of a result, which would actually constitute a lack of evidence, and not the evidence for the lack of tolerance in the proposed experiment. Therefore, I would support the basic Gahm et al. [21] interpretation. In fact, I would venture further and suggest that a plausible explanation for the occurrence of Vα11-Vβ3b in (B10.BR X C57BR)F1 – as a dominant trait inherited from a B10.BR parent – is a preferential pairing of b (but not a) allele of the Vβ3 with Vα11 TCR due to valine to phenylalanine mutation in its CDR1 region (residue 31) in response to PCC. Perhaps, this is a result of a simple competition of other VβXs to pair with Vα11 while docking onto PCC peptide with Ek? (The elaboration of this suggestion I would prefer to have elsewhere.) Gahm et al. [21] point out that in the T-cell repertoire of the C57BR strain, there are equal levels of Vβ3as and Vα11s in the non-antigen-specific pool. Here, Cohn's explanation that Vβ3a-Vα11 pair (in both orientations) is being negatively selected in C57BR [9] also finds a discrepancy with evidence to the contrary. Namely, it is only in the anti-PCC response that the skewing of this pair is being noted [21]. Cohn then challenges the reader to exploit his prediction about possible existence of C57BR-derived anti-PCC cell line to have at least two lineages, lineage 1 with Vβ3a-VαX and lineage 2 with Vβ3a-Vα11 in order to test his Tritope model [9]. However, how could one isolate the lineage 2, if the theoretical context is in contradiction with facts and not clear?

Cohn actually continues his discussion exactly with this question: ‘… how is it possible to isolate from C57BR a cell line in which each individual cell responds specifically to the [Ek-PCC] complex, yet paradoxically, 60% of its cells do not respond to PCC?’ He answers in the context of his Tritope model, explaining that Vβ3a is in allo-orientation in the lineage 2 and indifferent to peptide ligand but synreactive to Ek [9]. But then, why is this pair (Vβ3a-Vα11) not negatively selected in the thymus (as he claims a few paragraphs earlier)? This apparent discrepancy (Vβ3a-Vα11 pair exists in C57BR repertoire, but not in anti-PCC responses) does not benefit his suggestion. His explanation for proposed Vβ3a-Vα11 pair occurrence in the anti-PCC response (as lineage 2) is vague: ‘… these (Vβ3a-Vα11) CD4+ T-helper lines provide an autogenous source of signal 2 that drives the proliferative response’ [9]. Thus, if Vβ3a-Vα11 with Vβ3a in allo-orientation should yield signal 1 as a death signal, autogenous signal 2 would rescue them. This sounds rather like deux ex machina scenario for Vα11-Vβ3a C57BR-derived T-cell lines that are generated by immunization with PCC. More than half of cells in such T-cell lines would have elusive help from autogenous benefactors (missing the autoimmune boundary?) and that is why I deem it less likely.

Then, three observations were discussed that would support his model if taken altogether: (1) in F1 cross between high (3.5–6.3%) and low (0.5–1.3%) frequency of Vβ11 +  TCR mouse strains (all being MHC congenic, too), the low Vβ11 frequency is dominant, suggesting tolerance by negative selection. (2) B10.A congenic MHC mouse strains 2R (kk[Eαkk]b) and 4R (kk[Eαbk]b) with differences in Vβ11 frequencies between them (4R had high and 2R low) show that tolerogen maps to E molecule of the MHC. (3) The fact that B10.A(3R) [bb(Eαkb)d] shows intermediate frequency (1.6%) of Vβ11-expressing T cells is used by Cohn to suggest Eα be the most likely candidate as ‘tolerogen’. Cohn explained this observation as follows: 1.6% of T cells with Vβ11 is positively selected by interacting in restrictive orientation with peptide plus E molecule. The allo-oriented Vβ11 TCRs (with unknown Vα-chain partners) would be deleted by negative selection (via Vβ11–Eα interaction). Then, he states that elimination encompasses both CD4 and CD8 subsets demonstrating that Vβ11 is in entrained (allo-) orientation. The ratio of two selective rates (depending on restrictive or allo-orientation of a TCR V subunit) determines variability in frequencies of cells with particular V subunit of TCR.

This is fine, but then, the same would be true if peptides derived from E molecule were presented on class I and class II proteins during negative selection in these strain combinations according to the Standard model.

Still, the Tritope model seems to be a workable model of interactions involving three TCR-binding epitopes (peptide and two MHC sites) regulating the positive and negative selection in the thymus and perhaps it might be the truth.

What if it isn't?

The Standard model [13, 14, 16] has its own problems. Namely, it implies that the Vαβ pairs of the TCR have been somatically selected in each individual (of the species) based on, unfortunately, a whiter-shade-of-pale imprint of that individual MHC genotype. Cohn's contra-argument [9] is valid: it would be too much (energy) waste for an organism to select such a repertoire. I agree in principle, but I suggest a different mechanism as an explanation. The following paragraphs provide the reasoning why.

And lastly, Cohn suggests that the immune system's restricted recognition has been evolutionarily selected to fight only intracellular organisms (and pathogens) [8, 9, 23]. I believe that the defensive function of the immune system has been selected to fight both intra- and extracellular pathogens. And, in addition, the immune system's recognition was selected for protecting commensals (by its asylum function).

The novel positive selection concept: the TCR axis rotation model

In his discussion about the Tritope model of TCR–MHC interactions [8], Cohn criticizes the Standard model [16] and its anti-novel antigenic determinant (NAD) repertoire. He points that it is not logical to assume that a useful TCR repertoire has a ‘faint imprint’ of what has been selected over time in a species. He asks how would evolution select for either silent or attenuated recognition of self? Because peptide- and allele-specific self (R, MHC)-determinants are visualized as a meld epitope inevitably, the peptide-binding sites in the MHC grooves affect overall TCR interaction with the MHC molecule, that is, in combination with allele-specific R sites on their alpha coils surrounding the peptide (P). Cohn mentions four arguments [8] to be considered for this issue:

  1. Allele-specific and common determinants on thymic R sites are individual and not distinguishable by negative selection. Allele-specific determinant positive selection involves two germline-selected allele-specific determinants: one on the MHC and the other on the TCR subunits of respective molecules. He claims that it needs rationalization if ‘it is to be selected as a faint imprint of what has been selected over time’.
  2. The thymic MHC molecule (Rt) is not acting as a self-antigen, but rather as a platform for presenting intracellular peptides to the TCR. Negative selection does not delete (allele-specific) restrictive recognition of self(host)-R, but rather a recognition of self-P.
  3. Negative selection is the somatic process determining the S-NS discrimination and operates on anti-P, whereas positive selection sorts the germline-selected anti-R repertoire to determine restrictive recognition.
  4. As negative selection cannot distinguish allele specific from common determinants on MHC because both are self, it is not likely to be a factor that plays a role in providing a class of MHC recognized or allele-specific recognition. These properties are influenced by germline-selected sites not determined by the somatic process like negative selection.

In sum, self-tolerance cannot bias the positively selected TCR repertoire away from promiscuous recognition of MHC to the one that is allele and class specific, the latter being the claim of the Standard model [16].

Thus, Cohn [8] logically excludes the explanation that modest ability of TCR to react to MHC makes them appear on mature T cells (to perform their function in immune defence) and causes evolution of species-specific MHC and TCR pairs [16].

I wish to add here a novel argument for discussion. The MHC–TCR three-dimensional model of the interaction with staphylococcal enterotoxin B (SEB) revealed an interesting feature (Fig. 3). When compared to the interaction without SEB, the plane of TCR is rotated 40° counterclockwise [24], as depicted by the Figs. 3 and 4. Furthermore, the Vβ does not interact directly with the MHCα chain, but has a gap (Fig. 3). The result of this rotation might allow all three sites of the CDR3 region of the TCRVα chain to bind MHCIIβ chain. Because of rotation of the TCR recognition axis, the anti-P site of the V chain, not binding the superantigen, is shifted from binding peptide (P) to binding the helical structure on the MHC (i.e. CII β chain in Fig. 3). These experiments revealed a structural capability of the TCR to ‘slide over’ the MHC helix (beside the groove). Hence, an interesting possibility appears, which could answer the question raised by Kappler and Marrack group [16] and discussed by Cohn [8, 9]: ‘How could evolution select for TCR segments with affinity for generic features of MHC proteins if all the TCRs that illustrate this point disappear in the thymus before they could be of use for the survival of their host?’

Figure 3.

A model of three-dimensional structure of TCR–MHC interaction with superantigen Staphylococcal enterotoxin B (SEB). SEB binds to the TCR–MHC complex cross-linking MHCα and TCRβ chain. In the square, the TCR V regions are shown as having different orientations (according to Cohn): restricted (self-/syn-reactivity) or entrained (yielding allo-/syn-reactivity) as used in possible explanation for Mlsa reactivity. The figure shows hypothetical model of mouse αβTCR (with Vβ8.2) in complex with SEB and pMHC. This model was constructed by Li et al. [24] using least-squares superposition of three reported crystal structures: (1) the (14.3.d) Vβ8.2TCR–SEB complex [24], (2) the SEB–peptide/HLA-DR1 complex [31] and (3) the 2C (Vβ8.2) TCR heterodimer [32].

Figure 4.

Crystal structure of TCR–pMHCII interaction. During recognition of pMHC (of both classes) by either T cells or double-positive (DP)/single-positive (SP) thymocytes, the docking plane of TCR molecule is rotated by 40° [24], as demonstrated by comparing two-three-dimensional models of TCR–pMHCII interaction with (Fig. 3) or without (Fig. 4) superantigen SEB. (The crystal structure in this figure shows human αβTCR in complex with influenza haemagglutinin peptide with HLA-DR1 [25].)

Namely, it seems reasonable to assume that during positive selection, all TCR are rotated similarly due to putative superantigen-like protein interaction that would allow them to anneal either Vβ or Vα on the MHC coils of complementary alleles. In this respect, each V chain would have a specific superantigen-like binder that would lift such chain out of the reach of particular MHC helix (and embedded peptide) similarly to the SEB–TCR–MHC interaction (Fig. 5). The interaction with MHC, and perhaps positive selection, would be on the other TCR V region, making a close proposal to Cohn's suggestion (an engagement of V region in anti-R or restrictive orientation). Although it could be made to fit both models of generation of TCR orientations (either selective or predetermined), it would be best suited for selective mechanism in ensuring TCR orientations (and perhaps a solution to his model 1) [9]. The entrained V chains would be the ones that were lifted (by these putative superantigenic proteins) and would not bind MHC at all in such pairs (Fig. 5). However, they could partake in binding the MHC in pairs where the other Vs would be lifted by their own V-specific selecting superantigen alikes. If such superantigen-like proteins were to leak in the periphery, they would cause polyactivation of T cells and destroy their function (and the host). Thus, the process needs to be hidden away from other tissues and organs and is possibly compartmentalized within the thymic cortex (i.e. in nurse cells of the cortex). Now, the problem with superantigen-like binder molecules is that they would need to be germline encoded and perhaps less but still polymorphic enough to bind all species-specific alleles of the MHC. No evidence exists for such a locus. Perhaps, Mls might be an exemplar of such a molecule, with one allele, Mlsa ‘leaking out’ of its compartment that by being expressed outside nurse cells caused negative selection (although it might have been contributing to positive selection).

Figure 5.

TCR–MHC interaction during hypothetical (double) positive selection of CD4/CD8-positive thymocytes. A hypothesis is put forward about positive selection of DP thymocytes in shaping the allele-specific MHC T-cell repertoire. Two mechanisms are proposed that would act in principle similarly, with respect to TCR plane of docking, to SEB cross-linking MHCIIα and TCRVβ chains (Fig. 3). The first includes a family of superantigen-like molecular binders (specific for framework regions of Vα and Vβ of the TCR) that would act on DP thymocytes and connect their TCRs to respective counterparts on both classes of MHC molecules (α1 or β1 of CII, and α1, or α2 domains of CI). The second one postulates existence of less polymorphic TCR-rotator molecules inside thymocytes. These could adjust TCR such as to lift, shift and rotate the whole complex by 40° (clockwise, from the point seen in Fig. 4), thereby aligning only single V region at a time onto interacting MHC molecule (without binding peptide). This putative preselection process should be located in a secluded compartment within the thymic cortex (i.e. nurse cells). Positively selected thymocytes (by individual alleles of the host MHC) would be allowed to exit the compartment, where the expression of binders or rotators is lacking, allowing them shift back to the TCR orientation (as in Fig. 4) and as such continue development into SP stages within thymic cortex. In the square, the TCR V regions are shown as having different orientations (according to Cohn): restricted (self-/syn-reactivity) or entrained (yielding allo-/syn-reactivity) and used to correlate with suggested pre-positive selection of thymocytes. However, the hypothesis here is different from Cohn's suggestion in that TCR orientations are neither predetermined nor fixed after being (positively) selected (in this compartment). In addition, TCRs would not signal two different signals (as Cohn suggests), but a single one, which is in agreement with the standard model.

An alternative explanation without using the putative family of superantigen-like binders is perhaps a mechanism for his model 2 of generation of TCR orientations and positive selection [9]. It would include intrathymocyte TCR-rotation molecules. Let us call them ‘TCR orientators’, and they should also be germline encoded, but need not be polymorphic at all. There would be two kinds of rotators, clockwise- and anticlockwise, that would yield two orientations of TCR: one in which TCR Vβ would bind MHC class I α1 or class II α1 domain and another with TCR Vα chain posed to bind MHC class I α2 region or class II β1 domain (Fig. 5). Such orientation imprinting might be transient and perhaps compartmentalized (i.e. in nurse cells). This would make sense, if orientators would be transiently expressed as a result of nurse cell–thymocyte interaction via TCR–MHC complex.

After leaving nurse cells (that would thus positively select almost all available allele-specific TCR–MHC combinations and leave a few neglected unsuccessful pairings), the 40o rotation of the TCR–MHC axis would not be possible due to either lack of superantigen binders (outside nurse cells) or TCR orientators (a new developmental step would inactivate their expression within thymocytes). Thus, the TCR orientation would become fixed looking like the three-dimensional structure of TCR-pMHC/class II (Fig. 4) or /class I interactions revealed (not shown). The docking of TCR therefore would include, per each TCR V region, two MHC sites and a single-peptide-binding site. This is different from Cohn, as his model 2 claims predetermined orientation of TCR chains [9], but it might be adapted to accommodate his idea.

In summary, I suggest that within a compartment (i.e. nurse cell) when thymocytes get a signal that they had (strongly – not faintly) recognized self-MHC, they would shed all cell surface molecules (like after treatment with trypsin in a cell culture), leave the nurse cell environment and travel to thymic cortex where they would re-express cell surface molecules including TCR, but now free of rotator – or superantigen-like molecules. Thymocytes would then follow already known paths of their development. Therefore, if such hypothetical superantigen – like molecules would be ectopically expressed (outside the nurse cells) – they would actually provoke deletion – exactly as it was observed with the Mls system – a negative selection of those thymocytes that were hypothetically prepositively selected in the secluded compartment.

It follows that outside such (nurse cell) environment, the rotation-influenced positively selected TCR repertoire would be attenuated with respect to recognizing host MHC, as it would consequently undergo negative selection (self-tolerance). This whole process is reminiscent of Jerne [22], who suggested intuitively that mutations in molecules that can bind self would be needed for the generation of T cell repertoire against nonself. Here, it would not be a mutation-like event but rather a twist (rotation) of self-recognizing components – the TCRs – that would do the job.

Interestingly, the TCR anti-R(restrictive) and allo-orientations as suggested by Cohn [8, 9] to explain the generation of the TCR repertoire and signalling would not be necessary anymore (although we cannot exclude them). With the pre-positive selection, as the missing link, the Standard model can provide the simplest explanation about shaping T cell repertoire during ontogeny (second positive and negative selections in the thymus), being wrong only about the statement on evolution of TCR V regions (as V regions could not be selected by the faint imprint of species-specific alleles of MHC).

However, pre-positive (rotation constrained) selection of double-positive (DP) thymocytes would not pose logical limitations raised by Cohn (see four arguments in previous paragraphs). Furthermore, it would answer the question about evolutionary preference in pairings of TCR V regions with species-specific MHC alleles and their germline continuity, which would appear as promiscuous recognition of MHC molecules within the species (i.e. alloreactivity). Inasmuch, such pre-positive selection event would also not affect Cohn's claim that ‘Taken as a general rule, it implies that both positive selection and alloreactivity are peptide-unspecific’. The answer lies in finding two different types of signalling form two different orientations of TCRs to embrace or refute Cohn's Tritope model. The three-dimensional crystal structure of VβTCR–SEB interaction (Fig. 3) does not support his view, as ‘…, the observed conformational changes (in Vβ) are localized to the interface between the proteins and are not transmitted to the constant regions of the TCR; therefore, changes in TCR structure upon ligand binding are unlikely to account for signal transduction across the T-cell membrane’ [24].

My suggestion for the T-cell development in the thymic cortex including exit from putative nurse cell environment is a variation of the Standard model (single TCR signalling) and a part of the Integrity model. The differences are as follows: (1) pre-positive selection is a passive process for DP thymocytes. DP cells are selected by nurse cells on the basis of interacting with MHC molecules and shot (or transferred) out of the compartment. It would explain why positive selection neither activates DP cells nor causes their proliferation. (2) The neglected DP cells are those that were not able to bind nurse cells’ MHC, and they die in situ, perhaps by deprivation of homoeostatic signals or ‘food’. (3) In the cortex, outside nurse cells, DP thymocytes would receive signal 1 (death), if they bind self-pMHC (self-tolerance, negative selection) with their attenuated host-specific (rotated) TCRs. (4) Some DP cells with the highest TCR affinity TCR for pMHC would be rescued (‘positively’ selected) by signal 2 (help) via interaction with a DN cell (as a third cell). Although the latter mechanism is unknown, conceivably, DN cells might interact with all DP cells; however, only those with the longest lasting interactions with TEC would survive (receiving signals 1 and 2). Thus, each of the two signals would cause death if received separately, and only a combination of the two could rescue a DP cell from apoptosis. Such rescue signal is not an activation signal comparable to peripheral T cells (Fig. 1), as it is DP specific (signal 3 is different from the one in periphery) and hence a developmental one (without engaging proliferative step). This would give rise to natural regulatory T cells (nTregs). (5) The non-negatively selected DP thymocytes would migrate to medulla (being neglected in this phase). This is due to the residual intermediate affinity for the host MHC, as rotation constraint (orientation?) of their preselected TCR must have introduced such a bias. Thus, there is no need to imply a positive selection at this stage (as classically defined). (6) The SP thymocytes (lineage commitment would be either instructed or selected in pre-positive selection phase) are again negatively selected for self-pMHC combinations on medullary APCs. Thus, thymocytes undergo two positive selections in the cortex and two negative ones in cortex and medulla.

Being the most polymorphic region of the genome, the MHC is responsible for the restrictive recognition of peptide and thus has to be selected for allele-specific recognition in a germline fashion. However, if we introduce a novel evolutionary force (natural integrity) that would tend to increase diversity, we might explain MHC's high polymorphism and polygenicity in most animal species.

In conclusion about criticism of Cohn's Tritope model, my opinion is that it still lacks the explanation of the best-fit docking TCR–MHC (without peptide) scenario. Namely, to evolutionarily select a capacity of TCR interaction with MHC proteins (in allele-specific fashion) by positive selection in the thymus, his model would only work if the TCR axis were rotated such that a particular single TCR V chain would make a best positively selectable fit (i.e. with the highest possible affinity) with the juxtaposed MHC region. Therefore, for his model to work, we need a novel hypothesis about its mechanism. Perhaps, the suggestion about the pre-positive selection of thymocytes, possibly in a secluded and compartmentalized environment within the thymus, would explain it (provided experimental evidence confirms it in the future). Therefore, we need evidence for a new class of molecules that would either cross-bind or rotate both TCRVβ and TCRVα chains in their docking with the MHC molecules in thymic cortex (intuitively, nurse cells seem to be a reasonable choice) resulting with a 40o rotation of the TCR plane of recognition (compare Fig. 5 and Fig. 4). This would be similar to superantigen SEB binding to TCR–MHC complex in which TCR axis was rotated when compared to TCR–pMHC interaction, as revealed by crystallography [24, 25] (compare Fig. 3 with Fig. 4).

Seeing ‘colour’ with the Integrity model

If we liken antigen recognition by the immune system with perception of light, and if detection of shades-of-gray would be a measure of interaction between specific receptors for antigen on T or B cells with their ligands (TCR–pMHC or BCR-antigen[Ag], respectively), then I propose with the Integrity model that the immune system detects colours in concert with the opacity (Fig. 6). The Integrity model (Figs. 1–2) would be a detection of (at least) three colours: red (harmful), blue (useful) and yellow (the rest, including homoeostatic) signals. These signals would guide precursors of T and B cells through development and act on them to perform three basic functions of the immune system: defence (rejection of and fight against pathogenic micro-organisms), asylum granting (suppression tolerance, ‘regulation’, protection of commensals) and homoeostasis of immunocytes (‘regulation’, anergy, neglection of non-harmful micro-organisms).

Figure 6.

Colour (the Integrity model) versus shades-of-gray (the S-NS discrimination) vision of the world of antigens. The shades-of-gray vision can describe in simple terms the defensive action of the immune system by discriminating self (black) from nonself (white). However, such vision is blind to ‘true colours’ of the antigenic world that surrounds us, in which, for example, microbes can be seen to range from beneficent (blue) to harmful ones (red) according to rules presented in Figs. 1–2. There are two shades of blue, depicting either useful (commensal) microbes as nonself or useful parts of self. The immune system, according to S-NS model, is incapable of discriminating useful within nonself. The black, grey and white areas in the colour vision of the immune system's world of antigens also represent self and nonself, namely, black and gray (self) areas of persons’ clothing metaphorically representing our individual tissue antigens. The small white areas in characters illustrate cases when particular antigens, despite being nonself, are recognized by the immune system and accepted as self (well tolerated). Exemplars are as follows: a) in women, foreign antigens carried by foetus and b) for both sexes, a successful liver transplant (allogeneic) that requires no immunosuppressive therapy. These cases of nonself tolerance cannot be explained by any S-NS discrimination model in simple terms. In the ‘coloured vision’ by the Integrity model, being black (self) or white (nonself) does not imply a decision on whether to reject or tolerate antigen. It plays a role in communicating its presence. The decision is associated with harmfuluseful assessment of the microbe with three options: action (rejection), suppression (regulation) or inaction, and it includes a mode of tolerance (in concordance to the ‘colour’ of its manifestation within a host organism: deletion, suppression or anergy).

The molecular capacity to bind (dock) with (1) antigens (proteins, polysaccharides, haptens); (2) peptide(p)MHC; (3) lipid–MHC; (4) patterns (DNA, RNA, lipids, carbohydrates) would be distributed over four types of receptors: 3 that can detect ‘colour’ types signals from micro-organisms or the environment (host cells and extracellular matrix) and another one that can only detect shades-of-gray (i.e. be able to perform black and white discrimination). The co-activator/inhibiting molecules would be detectors that operate in concert with the antigen-specific ones and would assign increased shades-of-gray or define the colour of the detector. Thus, the immune rejection (defensive) response would include a red colour detector and a grey-shade detector (with a combination of three signals, Fig. 6). They would steer related ‘red’ actions important for the rejection of a pathogenic microbe. These detectors include classical antigen (polysaccharide); pMHC; lipid-MHC; pattern (DNA, RNA, lipids, carbohydrate) recognition receptors and co-activators (costimulating molecules) together with putative new detecting molecules (integrity disruption detectors). The TCR–pMHC and BCR-Ag are defined as detecting shades-of-gray. The rest are distributed in various colour detectors. These interactions would involve at least two cells in its initiation: an interaction between the APC (DC) and the T cell or FDC and a B cell (in T-dependent responses). This is a prototype activity that leads to rejection of a microbe and constitutes the defence function.

The integrity disruption signal comes from damaged tissues (i.e. necrotic death, break of cell–cell interactions or a variety of stresses that a cell can encounter, for example, a nucleolar stress). These are sensed by DCs, which interpret the break in communication as stress, and damage, but then evaluate other functions by a variety of molecules, which, in turn might take time for the end result. If after a time lag a benefit was detected (to be discussed below – see ‘blue’ action), it would signal so (by involving a third cell).

A separate set of proteins (and genes) would constitute the yellow detectors whose action, for example, could lead to anergy.

A prediction by the Integrity model would be that a specialized (set) of proteins are dedicated to sense the useful (i.e. ‘blue’ actions by ‘blue detectors’). The detectors would be similar as for the red and yellow detectors: pMHC receptors, costimulators (CD28, CD40) and co-inhibitors (CTLA4, PD1) and a whole set of interactions that would constitute ‘blueness’. A useful intruder can offer to an organism (or a tissue) a vast number of potentially beneficent functions of factors. Consequently, if detected, the ‘blue’ action would involve a putative set of genes and proteins that would stop rejection mechanism (if started) by halting proliferation, preventing (or skewing) differentiation of immunocytes into effector cells and even stopping immune effectors perhaps by vetoing their function. They would involve a regulatory/suppressor cell action (such suppression means a protection of a specific useful microbe from the rejection effector mechanisms). It follows that a third cell is required in the generation of such protective/regulatory/suppressive response, which is a clear prediction of the Integrity model. This interaction would activate putative genes able to stop proliferating T or B cells at both initiation (another T or NKT cell?) and effector stages (veto-like cell?). Needless to mention, another prediction would be that some T or B cell malignant tumours might employ a mechanism to inactivate such potential ‘tumour suppressor’ genes.

The detection of useful needs better clarification. It perhaps involves tens, hundreds or even thousands of genes acting in concert (in at least three-cell interactions), but it is definitely a finite number. A whole list of useful functions for one organism is something that I am trying to compile, but that would take some time, and hence, it should be published elsewhere.

An illustrative potential exemplar could be found in mice. Let us consider the B10.A recombinant congenic MHC strains 3R and 5R. This strain combination was used to define the J region of the MHCII by mixed lymphocyte proliferation assays. Unfortunately, the J region that was supposedly located between the A and E genes turned out to be missing as their genotypes were identical (bb[Eαkb]d). This event led to a stop in using the adjective ‘suppressive’ in literature dealing with immunology, and only ‘regulatory’ as a term was later carefully used to describe basically similar phenomena. However, the suppressive activity mapped by these strains could be an indication of a number of possibilities. Let us imagine at least two: (1) a recombination event in one of the strains (either 3R or 5R) led to a small mutation (deletion?) event with a large consequence, overlooked by DNA sequencing (for example, a small mutation in the promoter of the E gene can give different expression levels of E gene in the thymus compared with other tissues). This difference might have given rise to diverse levels of MHCII molecules expressed in thymic cortex epithelia, perhaps shortening the DP–TEC interaction and thus preventing (or encouraging) generation of particular nTregs (that were important for the tested repertoire reactivity). (2) The second possibility is that a trans-acting factor encoded elsewhere in the background of these strains (i.e. miRNA) might regulate suppressor activity perhaps facilitating induced Tregs occurrence within the mature peripheral T cell repertoire. Although the background of 3R and 5R strains should be theoretically identical, a point-like mutation or inclusion of a stretch of 22 nucleotides (a typical length of a miR) would suffice for this explanation, and DNA sequencing of 3/5Rs’ whole genome might missed it in some obscure intron.

Comparison between models

When comparing the S-NS discrimination [6-8, 10, 15, 26] and Danger models [27-29] of the function of the immune system, here are the differences according to the shades-of-gray comparison. As I suggested earlier, I view the S-NS discrimination as an oversimplification, because it fails to see colours of the antigenic world (Fig. 6). This has bitter consequences, because: (1) S-NS can only discriminate black from white (to eradicate a pathogen and to tolerate self), it requires threshold for the activation of the immune response somewhere within the shades-of-gray. This is a problem, because the threshold may vary during development and life, in general. Cohn's Tritope model [7-9] of TCR signalling is trying to solve some of these problems, and for that, it needs at least two signals (one for self and another one for nonself to be signalled during T cell activation). So, is it a whiter shade of pale or a darker shade of grey that leads to activation or tolerance of the immune response, respectively? (2) It is completely blind to inert micro-organisms or homoeostatic signals within an organism that keep immunocytes alive, and, (3) It leads to omission of a function that I call ‘the protection of the useful’ that keeps commensals alive in a body and includes one of the barriers against autoreactivity.

Similarly, the Danger model fails to detect the latter type of signalling (i.e. blue in our analogy), the very one I suggest to be causing T regulatory cell effects conferring ‘immunological’ asylum to useful or helpful micro-organisms like commensals.

Natural integrity pressure, selection and concluding remarks

Lastly, I wish to introduce a new theoretical concept alongside natural selection. It is called natural integrity defined as a cooperation-like drive, and I suggest it be an evolutionary force for integrating genetic pools of various life forms. It would be the primordial drive for conjugation and symbiosis in the simplest cellular organisms. In higher life forms, it would stand for sexual and cooperative drive, spreading variability of living beings by increasing genetic diversity. It is balanced by natural selection, which at unspecified time points selects best-fit organisms (which might be few or none) and eliminates the rest. Thus, by contributing to variability, it would provide survival based on avoiding genetic death in evolutionary terms. Natural integrity seems like some kind of a fast-feed-forward mechanism of diversity, perhaps resembling a kind of positive selection during speciation. Natural integrity might better explain punctuations in evolution, as it would cause periods of fast-feed-(positive)-forward diversifications interchanging with natural (negative) selections of various species.

Consequently, discussing about the immune system, the integrity pressure would tend to increase the recognition repertoire (for detecting molecules) because it seeks cooperation. In particular, it would try to identify potential useful markers on micro-organisms surrounding or entering an organism. Nevertheless, recognition by the immune system is a process dependent on both natural forces: selection working in concert with integrity. This is in contrast to Cohn's view about ‘The S-NS discrimination being the sole evolutionary selection pressure for the specificity of the TCR and BCR’ [30]. He contrasts increased diversity with ‘degeneracy’ and concludes that it ‘… is a non-issue for the S-NS discrimination largely because it is a problem of chemistry, not of biology’. He probably means that degeneracy cannot be naturally selected, because of intuitive concept on how nature must operate. Here, I propose my own intuitive concept and claim equal opportunity for integrating as well as selecting forces of nature.

Phylogenetically, natural selection would tend to diminish the diversity of the immune system recognition. Perhaps, negative selection of immunocytes is the recapitulation of such process ontogenetically. On the other hand, natural integrity would work in the opposite direction: increasing polymorphisms and counteracting conservation of genetic material. Ontogenetically, in the immune system, it might be represented as positive selection of thymocytes.

The integrity pressure would aim to speed up change horizontally and vertically during evolution. It would include increase in normal rate of modification or mutations of particular genetic areas within a particular population by molecular, genetic and epigenetic mechanisms such as gene conversion, exon shuffling, copy number variation, histone deacetylation and methylation. Although it has been thought that diversity of the immune repertoire is only for defensive purposes and that natural selection picked those with large enough repertoires for such purpose, I suggest that the need for it also comes from integrity pressure, which would transfer parts of diversified repertoire of recognition into communication between cells of the immune system. This communication would be responsible for decision-making process (of the immune system) that would result with a number of outcomes including the making of a protecton – an immune asylum for a life form either of microbial or higher life origin. It means that the immune system would be essentially capable of protecting useful nonself from destructive humoral and cellular defensive immunity.

As discussed earlier, T-cell repertoire cannot be shaped to be species allele MHC specific, in evolutionary terms, by selecting for faint imprint of the MHC recognition using thymic negative selection during ontogeny. Similarly, the positive selection of thymocytes, as classically defined, with TCRs specific for MHC proteins (of a species) would fail to select such inherent MHC binding capacity throughout evolution. In this discussion, I put forward a suggestion that natural integrity would force shaping of T-cell repertoire such that it would act on TCR–MHC (but not on TCR – pMHC) interactions during evolution of the positive selection mechanism. Perhaps, superantigen-like molecular binders or rotators were hijacked from bacteria with such properties by putative integrity pressure during evolution and hypothetically integrated in the genome, because they were useful for inherent anti-MHC activity of TCR during thymic positive selection.