A Commentary on the Zinkernagel–Hengartner ‘Credo 2004’

Abstract

In ‘Credo 2004’, Zinkernagel and Hengartner give us a food-for-thought analysis of immune responsiveness based on a ‘pragmatic and empiric point of view.’ The Credo 2004 postulates derived by inductive extrapolation from observation to generalization do not satisfactorily account for immune behaviour because they lack a conceptualization as illustrated here. Nevertheless, Credo 2004 is certainly valuable in a limited framework because it is based on the most likely of assumptions namely that the immune system was evolutionarily selected to protect against infectious agents, and therefore the study of pathogens will most accurately reveal how the immune system responds normally to protect. After reformulating them, the postulates of Credo 2004 are analysed with respect to their generality.

Overview

In the article ‘Credo 2004’ [1], Zinkernagel and Hengartner tell us to get our immunological house in order. I am not certain whether or not an ‘I believe’ statement is meant for discussion, but assuming it is, let's ask in what way ‘Credo 2004’ helps us to think about the immune system. Zinkernagel and Hengartner give us a straightforward empiricist's view of immune system behaviour. Is an ‘I believe what I see’ description sufficient? Or should ‘I believe what I see’ be framed in a larger context, and play its role in defining or falsifying an hypothesis?

Introduction

Immune reactivity only makes sense in an evolutionary context. The interactive selection pressures operative between infectious agents and the immune system are never-ending see-sawing interactive process that can and does take somewhat different pathways in various species. The commonalities in the behaviour of all immune systems may be derived by inductive extrapolation from an observation in one species but we would never know if they are valid ‘generalizations’ (the ‘general rules’[1]) without formulating them as falsifiable hypotheses that emerge from the principles of natural selection.

An immune system is characterized as a somatically generated, sufficiently large, random paratopic repertoire that can be coupled to a handful of germline-selected effector mechanisms. As their outputs are bio-destructive and ridding, two sets of decision function are required. First, the repertoire must be sorted into those specificities [anti-self (S)], which if expressed would debilitate the host and those specificities [anti-nonself (NS)] which if not expressed would result in the death of the host by infection. The anti-S must be subtracted from the repertoire leaving the residue as anti-NS. I refer to this as the S–NS discrimination because sorting the repertoire is both necessary and sufficient. Second, this residue anti-NS must be coherently and selectively coupled to an effective effector activity. ‘Effective’ means that it rids the pathogen in a short enough time, that is, before the host is debilitated by the infection.

It might be well to recall that the armed effector functions must be sufficiently localized to acceptably limit innocent bystander destruction. This is not an antigen-specific property and therefore not a problem of the S–NS discrimination. However from an evolutionary point of view, the bio-destruction and ridding of NS must be sufficiently circumscribed so as not to blur the selection on the specificity required to make an adequate S–NS discrimination.

These two decisions function the S–NS discrimination and the regulation of effector class characterize all immune systems.

Zinkernagel and Hengartner prefer to view the S–NS discrimination as an unresponsiveness–responsiveness discrimination operative at the level of the effector output. There are many reasons both antigen-specific and antigen-unspecific for unresponsiveness at the effector level to any antigen input. This black box approach (i.e. antigen in–response out) is not a priori unreasonable as evolution is selecting at the level of the output. However, this approach is too global to analyse the mechanism, yet their ‘general rules’ are statements of selected experimental observations that are interpreted to be the mechanisms that they believe to mediate the two decision functions. The discussion then revolves around the extent to which their ‘general rules’ account for the decision functions. Is it sufficient to arrive at the ‘general rule’ by inductive extrapolation from a black box observation? To answer this, we have to look at each of the ‘general rules’ proposed by Zinkernagel and Hengartner.

General rules governing immune reactivity

I have abridged the statement of their rules using their language in order to arrive at the essence of their argument. I do this following the precepts of the renowned philosopher Mechkonik who wrote, ‘It is better to be clear than right.’

T-cell rules

• 1) 
  T cells react to any antigen that ‘freshly’ enters secondary lymphatic organs for an appropriate time.
• 2) 
  T cells ignore antigens (self or foreign) that remain outside the secondary lymphatic organs.

These two rules may be viewed as part of the explanation as to how the host distinguishes S from NS. Given that there is no physical or chemical property that the immune system can use to define S and NS as classes, the sorting of the repertoire requires that S be separated from NS during developmental time. Zinkernagel and Hengartner challenge this with a competing view that is based on a ‘localization’ hypothesis. As it is not possible for all S to reside outside of secondary lymphatic organs or for all NS to enter them, these two rules are in and of themselves inadequate to make a S–NS discrimination; something must be added. This simple localization model separating S from NS is untenable [2–5] as a mechanism sorting the somatically generated repertoire because (i) localization is not antigen specific, (2) It is a germline-selected, not somatically learned mechanism and (iii) No anatomical enclave exists that includes all S and no NS or includes all NS and no S throughout life. After all, what is S for one individual is NS for another. Were this simple localization model operative, pathogens would be under strong selection not to enter lymphnodes or to enter the thymus. Besides, one might recall that there are vertebrate species without lymphnodes that have adequately functional immune systems (see also rule 3).

• 3) 
  T cells are induced to effectors by encounter with antigen, but effectors have such a short half-life that under certain conditions (referred to by them as ‘tolerance’) the effector activity is negligibly functional.

This mechanism classically referred to as ‘exhaustive differentiation’ requires T cells to be inducible-only and the sorting of their repertoire depends on the rates of induction to effectors relative to the rate of turnover of effectors. This is their proposed mechanism for negative selection. Alone rule 3 could not possibly explain a S–NS discrimination but coupled to rules 1 and 2, it becomes one of the possible ‘somethings that must be added.’

The proposal that this mechanism operates optimally in thymus because there are ‘few precursors’ in that organ is unclear. The number of the functional self-restricted T cells prior to negative selection that recognize a given S-peptide or NS-peptide is the same. Negative selection in thymus under rule 3 leaves a residue of T-cell anti-peripheral-S. The same problem confronts this negative selection mechanism in the periphery. It is to solve a part of the problem of peripheral tolerance that rules 1 and 2 are proposed. As a consequence, S is defined as a host-encoded antigen that is germline selected to be negatively selecting in thymus (rule 3), because the system needs to deal with peripheral Self that ‘freshly’ enters or is present in secondary lymphatic organs (rule 1)? If negative selection is required to prevent a T-cell autoimmune response to S-antigens in lymph nodes, then it would also be operative for S-antigens presented in the thymus that do not enter secondary lymphatic organs. If thymic-negative selection is operative, yet S-antigens remain that are neither expressed in thymus nor do they enter secondary lymphatic organs, then the immune system would be ignorant of (rule 2), not deleted by them. These would be ideal candidates for the breaking of tolerance by NS-antigens that share epitopes with S. All models of immune responsiveness must be able to deal with this latter class of antigens, which are a significant part of the total antigenic load [6].

Simply brushing aside the associative recognition of antigen (ARA) model (‘two-signal model’) [7–10], as a contrasting view without telling us how rules 1–3 project a competing model, is in need of justification. The ARA model does not treat ‘thymic-negative selection as special’, because it introduces a developmental time window during which, at the level of the individual, all S and no NS is present in the absence of a sufficiency of effector T-helpers (eTh), the source of Signal [2]. In order to sort the repertoire, S must be separated from NS either in time or space [2] and both experiment and simplicity favour separation during development time.

If secondary lymphatic organs are necessary to facilitate the probability (or rate) of eTh-antigen-presenting cell (APC)-T or eTh-B co-operation, then they do not deal with the event of cooperation itself. These organs increase the probability that two cells will find each other. This increase in the rate is such that an immune response can be effective against a pathogen in a short enough time, but this does not mean that the rate in the absence of secondary lymphatic organs is so low that a response to S over the procreative life of an individual would be negligible. It takes a response to one S-antigen to debilitate the host. In what way is this localization model then, an alternative to the ARA model as an explanation of the S–NS discrimination?

Exhaustive differentiation as a mechanism of negative selection would only be a competing model to direct inactivation, if a set of parameters could be defined that permit a below threshold level of effectors in response to S and an above threshold level of effectors in response to NS. No such set is forthcoming [11]. Further, to postulate a steady-state level in thymus of eTh anti-S on the pathway of exhaustive differentiation as an alternative to the direct Signal [1] deletion is, in the absence of evidence, in need of rationalization. No matter how regulated, it is a gratuitous pathway to autoimmunity.

The ARA model (‘two-signal model’) is a necessity of logic. For a cell to respond differentially to two entities, one S and the other NS, two distinguishable antigen-specific signals are required. If interaction with antigen is inactivating (Signal [1]), then a second signal is required to activate the cell and this must come from a source (the eTh) that has itself made a S–NS discrimination (‘the chicken and egg’ problem resolved in references [11, 12]). Signal [2] must be antigen specific and therefore must be delivered by eTh; it cannot come from the APC which itself cannot distinguish S from NS.

• 4) 
  The roles of T-cell-receptor avidity, peptide concentration and co-receptor dependence in defining T-cell specificity of their effector functions are still largely unclear.

As a ‘general rule governing immune reactivity’, this is also ‘largely unclear.’ To this might be added that the ‘specificity’ of the TCR is not definable by these three factors. It is in an attempt to deal with ‘specificity’ that we introduced the ‘Specificity Index’, as discussed elsewhere [6, 13, 14].

• 5) 
  Intracellular and extracellular pathogens are processed to peptides for presentation on class I and class II restricting elements, respectively, via distinct pathways. Cross presentation is of minor or no physiological importance.

Most immunologists treat the activation of antigen-responsive or initial state Th (iTh) or T-cell-cytotoxic killers (iTc) as the consequence of an APC–iT interaction (‘costimulation’). We have argued that the activation step requires an interaction of ARA between the eTh-APC-iT. This initiates the pathway of anti-NS T cells to effectors. As long as the APC cannot make a S–NS discrimination, activation by an APC–iT interaction is untenable. Given this the following problem is raised.

As long as the entire cell, APC or B cell, is viewed as the platform for eTh-APC-iT or eTh-iB collaboration for activation, two problems will dominate over the question of the role of cross-presentation. First, whether the presenting cell is an APC or a B cell, every NS-antigen that is processed would be converted into one that shares epitopes with S because presenting cells also present S-peptides. This puts the individual at risk of breaking tolerance when responding to the steady-state NS-antigenic load. Second, in the case of an APC which processes two or more NS-antigens, the resultant peptides would be scrambled on the APC surface so that their origin (i.e. the linkage relationships between the peptides) would be unknowable, making impossible the independent regulation of class for each antigen because that requires ARA.

Assuming that cross-presentation has a negligible physiological role (rule 5), the APC will present S-antigens synthesized by this cell as peptides on class I major histocompatibility complex (MHC) and extracellular antigens as peptides on class II MHC. Unsorted B cells will present extracellular S or NS as peptide-class II ligands and intracellular S-antigens as peptide-class I ligands. Under this scenario, eTh-iTc collaboration to break tolerance is a high probability steady state situation, even without cross-presentation. Cross-presentation would only raise the probability and, in addition, make the regulation of class impossible. In this framework, cross-presentation would be unselectable as Zinkernagel and Hengartner imply.

There are three situations, however, under which the assumption of a significant role for cross-presentation becomes essential and must be faced.

The first situation challenges the above assumption that the APC or B cell is the platform for eTh–iT co-operation. Associative recognition of antigen required for a coherent and independently regulated effector response would not be possible if peptides derived from a pool of S and NS antigens were scrambled on the APC or B-cell surfaces. However, if the extracellular antigen is taken up and processed as a unit that is expressed in a ‘signalling patch’ on the cell surface, then eTh–iT collaboration uniquely via the patch would preserve ARA. The uptake of antigen by an APC, aside from its germline-encoded receptors, requires an antigen-antibody (Ag-Ab) complex and this latter assures uptake of a single antigen. Given this solution to ARA on an APC, an extracellular antigen would be processed to peptide-class II MHC allowing an eTh–iTh interaction but making an eTh–iTc interaction in ARA impossible. Consequently, cross-presentation by the APC would be required. The B cell takes up a single antigen which if processed uniquely to class II MHC would not permit eTh–iTc collaboration but would allow eTh–iTh collaboration in ARA. If B cells are important as presenting cells for eTh–Tc collaboration then cross-presentation could not be negligible.

The second situation would be the response to an intracellular pathogen in a cell like a fibroblast that does not express class II MHC. The spread of the pathogen requires an extracellular stage that can be taken up by APCs and/or B cells and processed to peptide bound to class II MHC. In order to permit an eTh–iTc interaction of ARA, a functional role of cross-presentation on an APC is a reasonable assumption.

The third situation arises in order to account for peripheral S-antigens that are expressed after the developmental-time window closes. These antigens would be treated as NS and induce autoimmunity were it not for their ectopic expression in thymus under the control of the transactivating transcriptional regulator, Aire[15]. As they are expressed intracellularly, under Rule 5, in the absence of cross-presentation, they would be presented uniquely on class I MHC and be negatively selecting for cytotoxic, not helper T cells. The accumulation of helper T cells specific for this subset of peripheral S-antigens would surely result in autoimmunity.

It is hard to see how one can avoid the assumption of a requirement for an eTh-signalling interaction in the activation of T cells. The APC cannot make a S–NS discrimination. This means that the activation of T cells, Tc or Th, requires a second signal from eTh that have been pre-sorted [11, 12] to be anti-NS and co-operate via a ‘signalling patch’ on the APC in order to associatively recognize peptides derived from the given antigen. This assumption makes cross-presentation physiologically important.

• 6) 
  Positive selection of T cells operates both inside and outside the thymus to maintain intermediate avidities of their TCRs for S-peptide-MHC.

Zinkernagel and Hengartner treat the thymus as being only necessary for T-cell-receptor rearrangement. In their view, positive selection and negative selection (exhaustive differentiation, Rule 3) occur also systemically.

However, to view positive selection merely as a mechanism for the maintenance of intermediate avidities of the TCR for S-peptide-MHC whether in thymus or periphery is in need of rationalization. I would have thought that positive selection is required to establish the restriction specificity and the relationship between class of MHC and T-cell-effector function [16–18] in which case, how and where become the questions.

Thus, there are two questions raised by their cited data

• 1
  What is the origin of the positively selecting cell (i.e. is it bone-marrow or thymic pouch derived)?
• 2
  Where anatomically does the positively selecting cell function normally (i.e. is it expressed only in thymus or also peripherally)?

One might conjecture that the positively selecting cell is bone marrow derived and prior to the appearance of a defined thymus in evolution, there was a diffuse system or dispersed thymi that carried out the function of positive selection. With the appearance of a unique defined organ as seen in mice, and selected for efficiency, these other systems were fossilized. The presence of the thymus inhibits or swamps out the functioning of the fossil pathway which is revealed in the absence of a thymus as an inefficient evolutionarily ancient system. While Zinkernagel and Hengartner contend that our views need serous revision, it seems that this criticism (rule 6) should be limited to the origin and anatomical placement of the positively selecting cell, not to its function.

• 7) 
  The gravity of the immunopathological component of any immune response is dependent on the length of time that it takes to rid the antigen. If short, immunity dominates; if long, immunopathology dominates. In a special case where antigen persists in secondary lymphatic organs at high levels, T cells are deleted and immunopathology is avoided (as, of course, is immunity).

The immune response is under selection to be brought to an effective level in a short enough time [6]. Every response to NS entrains a level of ‘immunopathology’ from several sources along the inductive pathway. Usually the extent of ‘immunopathology’ is acceptable. If the immune response is ineffective and the pathogen is not ridded, induction continues until eventually immunopathology becomes apparent. A major source of immunopathology is accumulated Ag-Ab complexes that activate effector functions like C' lysis, ADCC and inflammatory chemokines that spill over to attack innocent bystanders. This is why regulation of the magnitude of the response by T suppressors is not unexpected [19].

This raises questions about their ‘special case’ of T-cell deletion (‘exhaustive differentiation’) by persistent antigen. Instead of suppression, deletion is proposed as the regulator of magnitude. Under Rule 3, deletion of T cells is the result of ‘exhaustive’ induction. Antigen can only persist in secondary lymphatic organs at high levels, if the effector response is ineffective which, in itself, favours immunopathology (rule 7). This model in which T-cell deletion is in reality, exhaustive induction is incomplete without a spelling out of the conditions under which the overall result would be either deletion or an effective effector response. At the moment, no such condition appears to exist and the steady state induction of effector T-cells anti-S on the pathway to death puts the organism under risk of autoimmunity and immunopathology [11].

• 8) 
  A regulatory T cell that ‘knows’ in foresight what is needed for an equilibrated immune response has not been convincingly shown.

This might be viewed as their food-for-thought ‘General Rule.’ A regulatory T cell, suppressor or helper, that expresses a random repertoire that divides the antigenic universe into combinatorials of epitopes (peptide ligands) is Promethean, that is, has ‘foresight’, provided that its repertoire is properly sorted by subtracting anti-S and leaving the residue as anti-NS. Such a regulatory T-suppressor anti-NS could play its role in controlling the magnitude of the response [19]. In all fairness, to reject as unconvincing the TNTC papers on this subject, it would be helpful to have either an alternative interpretation of the data or a suggestion as to what experiment should be done that would convincingly disprove ‘foresight’ or the boundary conditions defining an ‘equilibrated immune response’.

B-cell rules

• 9
  Under conditions of limiting concentration of antigen (monomeric or oligomeric), B cells require interaction with eTh to be induced (‘conventional T-dependence’).

As it is assumed that the B cell cannot receive a tolerizing signal via its B-cell receptor (BCR) on binding ligand (rule 11), the condition of limiting ‘dose’ (translated above as concentration) only affects the density of the peptide-class II MHC which is the ligand presented to the eTh. Rule 9 then is that at limiting in an eTh ligand density fashion, an eTh signal is required to induce the B cell but at higher ligand density, the B cell is somehow signalled to activation, and an eTh signal is not necessary. The eTh has no way to determine whether the antigen was monomeric or oligomeric once it is processed. Zinkernagel and Hengartner are unclear on the next point. If the B cell is not tolerizable (i.e. not signalled to inactivation via its BCR–epitope interaction) (rule 11) but it can process BCR-recognized antigen (monomer or oligomer), what happens above the limiting concentration? Extrapolating from their discussion, monomers would only be responded to in an eTh-dependent fashion; in the absence of eTh the B cell is blind to monomers. ‘Repetitive’ antigens, oligomers, can induce activation of B cells, eTh-independently. However, as their processed peptides are recognized by the eTh, the eTh-independent pathway for oligomers over the limiting concentration is an additional pathway to activation.

A role for eTh at limiting concentration has a long history going back to the mid 1960s. The ‘T-helper’ was so named by Mitchison because it was believed that its role was to ‘focus’, ‘amplify’, ‘concentrate’ antigen so that B cells could be induced by concentrations of antigen too low to be effectively detected (‘conditions of limiting dose’) by them. The name, ‘helper’ unfortunately, has stuck but this role for the T cell is surely untenable now that we know that the T cell only sees processed antigen bound to MHC. The T cell has a distinct signalling function for the B cell. It makes little sense to postulate that the function of the eTh is to permit induction of B cells to secrete antibody of too low affinity to function as effector molecules in solution. This is why for years I insisted on referring to Th as T-co-operators.

• 10
  B cells like T cells are induced only in secondary lymphoid organs. If the antigen is repetitive, the B cells are induced to secrete short-lived immunoglobulin (Ig)M, eTh-independently.

‘Repetitive’ may be necessary but is insufficient to induce eTh-independent IgM secretion [20]. No S-antigen can be an eTh-independent inducer. As there are many ‘repetitive’ S-antigens, were they to be obligatorily inductive, the host would auto-destruct (see rule 11).

• 11
  The B-cell repertoire is not sorted (negatively selected). Autoimmunity is controlled by eTh, which are sorted to be eTh anti-NS uniquely (rule 3).

If B cells are not ‘tolerable’, then they must be eTh-dependent for induction. If the eTh repertoire has been sorted (rules 1–3) to be anti-NS then, true enough, only B-cell anti-NS can be induced, except for those anti-S B-cells induced by:

• 1
  NS antigens which cross-react (share epitopes) with S.
• 2
  Repetitive S-antigens which induce IgM eTh-independently (rule 10).

As a significant proportion of NS-antigens share determinants with S (approximately 10%, my estimate) and IgM antibody is uniquely effective in activating complement-lysis, Rule 11 becomes a recipe for autoimmunity. To argue that IgM autoantibodies are too short-lived to cause autoimmunity implies that IgM antibodies are too short-lived to be functionally protective (i.e. unselectable).

Now we can face an internal ambiguity. If B cells are activatable-only not ‘tolerizable’ (rule 11), that is, do not receive an inactivating signal on binding to an epitope but do receive an activating signal on ‘aggregation’ of their BCRs by ‘repetitive antigen’ (rule 10), then two conclusions follow:

• 1
  B cells cannot be signalled by monomers but monomers are processed by them (rule 9).
• 2
  No S-component can be ‘highly repetitive and rigidly ordered or a polyclonal B-cell activator.’

A major selective pressure on the humoral response is exerted by monomers. Many pathogens depend entirely on monomeric toxins for their virulence (e.g. diphtheria toxin, tetanus toxin, cholera toxin, Clostridium Welchii lecithinase, Streptolysin O, etc.). If the B cell were blind to monomers, then it is also not ‘tolerant’ of S-monomers like insulin, serum albumin, growth hormone, most chemokines, interleukins, etc. It is repeatedly argued by those who take this position that monomers are non-specifically aggregated in vivo and signal the B cell as a ‘repetitive polymer.’ If this signal is inactivating then, under the ARA model, the B cell would be ‘tolerant’ of S-monomers and responsive to NS-monomers. Unfortunately, it is not possible to non-specifically aggregate NS-monomers as a class in vivo to render them signalling for the B cell. Consider, diphtheria toxin responded to at nanogram levels in vivo. If it had to be aggregated non-specifically in order to signal, then when aggregated with serum albumin present at 10⁶-fold higher concentration, it would still behave as a monomer to the B cell. If it is argued that aggregation is specific, then it must be due to an Ag-Ab complex and the origin of the antibody (a chicken and egg problem) must be considered [21, 22].

As a last point, in order to rid monomers they must be seen by the antibody repertoire in three-or-more ways and thus exerts a crucial selection pressure on the size of the repertoire [23–25].

In any case, the assumption that B cells are blind to monomers and activatable-only (eTh-independently) by polymers needs better rationalization to convincingly rise to the level of a ‘general rule.’

• 12
  Antibody specificity is defined by affinity-avidity of protective IgG.

While a relationship, if it exists, between ‘specificity’ and ‘affinity-avidity’ is not obvious and needs analysis, in the context of Rule 12 we must face a conundrum. If eTh has as its role ‘concentrating’ antigen (rule 9) and if B cells are not tolerizable (rule 11) and if repetitive antigens induce short-lived IgM, eTh-independently (rule 10), then IgG autoantibody must emerge as a major problem. To refer to it as ‘protective IgG’ masks the question of how it made a S–NS discrimination (i.e. how its repertoire was sorted).

• 13
  IgA production is eTh and secondary lymphoid organ-independent. The rules for IgE production are less clear than for IgG.

If the B-cell repertoire is unsorted (rule 11) and IgA production is eTh-independent (rule 13), then response in the IgA class must have a significant anti-S component. If IgG and IgE are eTh-dependent while IgA induction is not, then autoimmunity must be controlled at the level of regulation of effector class. Does there exist a reasonable scenario? Further, why should induction of IgM be so tricky (rules 9–11) compared to the other isotypes? These are questions that interpretations based on the black box approach bypasses.

• 14
  Immune protection always includes a degree of immunopathology. If the infectious agent is acutely cytopathic, immune protection dominates; if the agent is non-cytopathic, immunopathology dominates. ‘Immunity’ then is defined as a mechanism that can eliminate cytopathic damaging agents, but can respond to agents that are harmless without immunopathology.

 If immunopathology is a damaging effect due to an attack on innocent bystanders by the (Ag-Ab)n complex triggered effector mechanisms, then the degree of cytopathicity of the antigen is irrelevant as the effector mechanism has no way to assay this property. If it is assumed that as the cytopathicity of the antigen decreases, an effector response is favoured that is immunopathological, then Rule 7 loses generality, not to mention, what evolutionary sense does such an assumption make?

• 15
  The maintenance of memory (a more rapid response) is largely antigen-independent when non-replicating antigens are studied. In contrast, the maintenance of immunity to re-infection is largely antigen-dependent.

The immune system has no way to assay the rate of replication of the antigen. Further, memory defined merely as a more rapid response to a second encounter has many potential mechanisms all of which result in an increased steady state level of antigen-responsive cells specific for the antigen. If the maintenance of memory is antigen-independent for non-replicating antigens, then it is implied that a long-lived memory cell exists. As the immune system cannot assay ‘replication’, the memory cell must also exist for replicating antigens. It might be true, in addition, that persistent antigens maintain high responsiveness but to frame the memory phenomenon in terms of rate of replication of the pathogen needs a more probing analysis.

As an aside on this question of memory, evolution is selecting on the efficacy of the primary response. Any individual that cannot survive a primary encounter with a pathogen has no need for a memory response. Memory then either must provide the animal with something more than ‘a more rapid response’ or be viewed as an unavoidable/unselectable consequence of a primary response. If a special category of long-lived memory antigen-responsive cells switched to an effective isotype exists then ‘something more’ must be visualized in order to explain evolutionary selectability. An example of ‘something more’ might be the need to provide maternal antibody that protects the developing embryo and newborn. If no explanation is forthcoming, then the secondary response becomes a phenomenon of passing interest. Antigen-dependent memory means that the antigen was not ridded, and the primary response was simply ineffective in ridding the antigen.

• 16
  Non-cytopathic or slowly progressing infectious agents are transferred to the next generation of the host, when the developing immune system is not yet competent.
• 17
  The developing embryo and neonate are under protection by maternal antibodies. The ‘attenuation’ of infectious agents by these antibodies results in their becoming ‘live, natural or physiological vaccines.

Rules 16 and 17 might be considered together.

It would be expected under the ARA model that the host embryo would become ‘tolerant’ of the non-cytopathic agent, if it engaged a newly arising, not yet competent (insufficiency of eTh) immune system. Consequently, such agents would induce neither immunity nor immunopathology (rule 16). If the agent is cytopathic, the embryo would be killed unless protected by maternal antibody. In this latter case, the agent would be either eliminated or kept in stasis and camouflaged from the arising but not yet competent immune system as an (Ag-Ab) complex until the system became competent and could respond. Any antigen not seen by the arising immune system (i.e. when there is an insufficiency of eTh) is treated as foreign when it becomes mature (i.e. when there is a sufficiency of eTh). A cytopathic antigen kept from killing the embryo as a complex with maternal antibody that also insulates the antigen from interaction with the arising immature immune system (i.e. it doesn't establish ‘tolerance’) and later exposes the antigen to a mature system in an immunogenic form that induces a protective primary response, can be viewed as a form of natural vaccination. However, it surely would be a sufficiently rare event to make one wonder whether it should be treated as a general rule or a fortuity.

• 18
  Insufficiency of maternal antibody can result in severe infections and even immunopathology and autoimmunity.

In the absence of a responsive immune system, maternal antibody acts passively to protect against any recognized pathogens. In the absence of maternal antibody, at a time when the endogenous immune system is not yet functional, the neonate would be subject to infection but not to any immunopathological or autoimmune sequelae. These could only arise after the immune system matures, and it is not obvious why the absence of maternal antibody during immune system maturation should increase the frequency after maturation of antibody-driven phenomena such as immunopathology and autoimmunity or, for that matter, that it does.

• 19
  Efficient vaccines induce protective antibody. Inefficient vaccines induce T-cell responses but insufficient effectors. Efficient vaccines of cytopathic agents reduce their damaging effects; efficient vaccines of non-cytopathic agents reduce their immunopathological consequences.

While efficient vaccines by definition induce immune responses in an effective effector class that reduces the direct cytopathic effects of a pathogen, it is challengingly subtle to define by symmetry an efficient vaccine to a non-cytopathic agent as one that reduces immunopathology. If such a vaccine exists, it would have to induce a response to the non-cytopathic agent in an effector class that is both ineffective in ridding that agent but effective in blocking an effective endogenous response. Autoimmunity is an effective effector response to a S-component. Immunopathology is a response to a NS-component, usually but not necessarily ineffective, that spills over by attacking innocent bystanders. A vaccine that reduces autoimmunity (or allergy) would have to switch the autogenous response from an effective to an ineffective class. A vaccine that reduces immunopathology is almost an internal contradiction, certainly circumscribed by a set of very limited requirements.

• 20
  While the revealing of phenomenology by immunologists is open-ended, only those phenomena that can be rationalized in terms of natural selection (i.e. ‘under wild-type conditions’) are meaningful.

The point made by this rule as it was stated by Zinkernagel and Hengartner was not understood by me, so I rewrote it, as best I could giving it a personal spin in order to be able to establish the area of potential agreement, namely, that every interpretation of observation must meet the criterion of ‘evolutionary compatibility.’ However, this has not been the case in formulating many of the above ‘general rules governing immune reactivity.’ In order to meet this criterion, the selective pressure considered in formulating the rule is probably the most exigent parameter. The emphasis placed on the response to infectious agents by Zinkernagel and Hengartner is a good strategy, but this does not imply that sufficient understanding can come from the black box approach, namely looking at the unresponsiveness–responsiveness output consequent to confrontation with a pathogen under so-called ‘wildtype’ conditions.

As a closing comment, I do hope that my commentary on ‘Credo 2004’ will convince these authors that we are trying our best to put our immunological house in order. This was the intended role of the ARA model. All this having been said, credit should be given to the one and most important thing ‘Credo 2004’ did accomplish. It made us think.

Acknowledgment

This work was supported by a grant (RR07716) from the National Center for Research Resources at the National Institutes of Health