Integration of the immune memory into the pathogen‐driven MHC polymorphism hypothesis

Major histocompatibility complex (MHC) genes (referred to as human leukocyte antigen or HLA in humans) are a key component of vertebrate immune systems, coding for proteins which present antigens to T‐cells. These genes are outstanding in their degree of polymorphism, with important consequences for human and animal health. The polymorphism is thought to arise from selection pressures imposed by pathogens on MHC allomorphs, which differ in their antigen‐binding capacity. However, the existing theory has not considered MHC selection in relation to the formation of immune memory. In this paper, we argue that this omission limits our understanding of the evolution of MHC polymorphism and its role in disease. We review recent evidence that has emerged from the massive research effort related to the SARS‐CoV‐2 pandemics, and which provides new evidence for the role of MHC in shaping immune memory. We then discuss why the inclusion of immune memory within the existing theory may have non‐trivial consequence for our understanding of the evolution of MHC polymorphism. Finally, we will argue that neglecting immune memory hinders our interpretation of empirical findings, and postulate that future studies focusing on pathogen‐driven MHC selection would benefit from stratifying the available data according to the history of infection (and vaccination, if relevant).


| INTRODUCTION
Major histocompatibility complex (MHC) genes (referred to as human leukocyte antigen or HLA in humans) are a key component of vertebrate immune systems, coding for proteins which present antigens to T-cells.They are marked by their degree of polymorphism, encompassing dozens to hundreds of alleles in vertebrate populations.This high polymorphism can be explained by balancing selection, which refers to a number of selective processes by which multiple alleles are actively maintained in a population's gene pool at higher frequencies than would be expected from genetic drift alone.This balancing selection, therefore, results in the preservation of alleles over long periods of evolution.As the polymorphism is focused on amino acid residues in the MHC peptide-binding pocket that are responsible for the specificity of antigen binding, balancing selection has been assumed to be pathogen-driven and, more specifically, to stem from differences in the binding capacity of different allomorphs for a given pathogen. 1hree main mechanisms have been proposed to explain how pathogens can cause the balancing selection involving heterozygote advantage (HA), fluctuating selection (FS), and negative frequency-dependent selection (NFDS) (reviewed by Radwan et al. 2 ).HA has been proposed to arise from the fact that each MHC molecular variant is only able to present a limited repertoire of antigens to T-cells, and expressing two different MHC proteins is thus expected to increase the probability of presenting one of the peptides of a given pathogen, thereby increasing the probability of resistance. 3,4FS assumes that the pathogen composition encountered by a host population varies over time, and that different allomorphs are therefore favoured at different times 5 This is similar to HA, in that heterozygotes gain an advantage when fitness is calculated across episodes of selection by different parasites. 5In contrast to HA and FS, NFDS requires host-parasite co-evolution.This stems from the fact that pathogens will tend to adapt by avoiding presentation by the most common MHC types.Such selection may also favour novel MHC alleles to which pathogens have not yet had a chance to adapt, 6 explaining the high rate of nonsynonymous substitutions observed at antigen-binding residues of MHC molecules. 1Indeed, this theory suggests NFDS as the mechanism which is most likely to be able to explain both high MHC polymorphism and the signatures of positive selection on novel MHC sequence variants. 7e theory of balancing selection on MHC outlined above focuses on the role of MHC in discriminating self from non-self, and the resulting ability (or inability) of an organism to recognise and clear infection.Pontarotti and Paganini 8 have recently pointed out that pathogen evasion from immunological memory can form a largely overlooked source of selection on MHC alleles (Figure 1).Selection on pathogens will favour mutations in T-cell receptor (TCR) epitopes that escape recognition by memory lymphocytes.As long as epitopes of the memory lymphocytes that emerged in response to a given infection depend on the range of epitopes presented by MHC alleles, the loss of immune memory via pathogen evasion may form an important source of selection on MHC.Such form has not been considered in models explaining the evolution of MHC polymorphism ( 7,9,10 ), which assumed that susceptibility to infection rests in an ability of MHC to bind an antigen and present it to T-cells.These models were successful in explaining high MHC polymorphism with NFDS, whereas assumptions showing similar capability of HA 10 were later considered unrealistic. 11However, NFDS models struggle to produce long persistence of MHC allelic lineages observed in MHC genealogies of many taxa, 12 The difficulty arise probably because NFDS produced by the modelled host parasite coevolution is relatively weak compared to positive selection favouring novel alleles that replace resident ones. 7I G U R E 1 Mutation in the proteome of the pathogen and consequences for its antigenicity in primary versus secondary immune response.HLA variants are able to present different sets of epitopes.Immune response is formed against only a subset of these epitopes (circled in red), which are immunodominant.When a mutation occurs in variant 1 (star), alleles 2 and 3 are protected against a new variant by their immune memory, whereas immune memory mediated by allele 1 is no longer protective (broken red line).Allele 1 can still give rise to successful primary immune response directed towards a sub-dominant epitope (green circle), but its bearer might suffer morbidity associated with break-through infection.
In this paper, we argue that immune memory can accentuate fitness differences between MHC alleles, possibly accentuating NFDS and other mechanisms of selection on MHC.We review new evidence, particularly from COVID data, supporting the key role of MHC in shaping immune memory, particularly from recent COVID data.We then further develop the proposal of Pontarotti and Paganini 8 by highlighting that focused nature of immune memory may have important consequences for the nature of selection acting on MHC.We argue that neglecting immune memory limits our understanding of MHC evolution and hinders interpretation of empirical findings.We postulate that future studies focusing on pathogen-driven MHC selection would benefit from stratifying the available data according to the history of infection.

| HLA AND IMMUNE MEMORY
In addition to being versatile, with the ability to specifically identify non-self-molecules, immune memory is another hallmark of adaptive immunity. 13It is thought to confer an advantage, particularly to long-living organisms such as vertebrates, by preventing reinfection by the same pathogen strains, as evidenced by the success of vaccination programmes. 14,15The same pattern has emerged from recent SARS-CoV-2 pandemics.While primary infection is associated with sometimes severe health conditions, even though patients raise a primary adaptive immune response, in the event of reinfection or following vaccinations, adverse symptoms are either absent or minimised. 16There is also abundant evidence that the immune protection conferred by immune memory can select for mutations in pathogen epitopes that escape recognition by memory T-cells. 17,18Can such evolution of pathogens be a source of selection on MHC genes?The answer to this question depends on the extent to which the epitopes recognised by memory cell TCRs (mcTCRs) depend on the properties of MHC molecules.
Memory T-cells are the end-product of the clonal expansion of naive lymphocytes that have recognised an immunogen via its antigen receptor and have helped to clear the primary insult. 13Of the many possible antigenic sites on peptides expressed by pathogens, lymphocytes recognise and respond to only a limited set potential epitopes, referred to as immunodominance 19 and, outside rodent models, reported, for example, for influenza and COVID. 20,21The nature of immunodominance is likely a complex, likely driven a range of biological parameters as extensively discussed by Yewdell. 19MHC may shape immunodominance is through its effect on the TCR repertoire of naive lymphocytes 22,23 which is one of the major determinants of immunodominance. 24,25There is, however, evidence of a more direct effect via the binding properties of MHC molecules, such that immunodominance is a joint effect of precursor lymphocyte availability and the MHC binding affinity of the epitope. 24Consequently, the epitopes recognised by memory T-cells can vary between individuals in an MHC-dependent manner.For example, Ferretti et al. 26 showed that SARS-CoV-2 epitopes of CD8 T-cells differ between the six HLA types examined, with a limited overlap residing mostly outside the spike protein.Similarly, Francis et al. 27 reported that individuals carrying HLA-B*07:02 show a CD8+ T-cell response to SARS-CoV-2 that is dominated by preexisting memory, likely as a result of previous exposure to other coronaviruses.Such cross-reactivity has not been observed for other HLA types, demonstrating that HLA variants differ in terms of memory TCR epitopes.Thus, MHC can affect the TCR composition of the memory T-cell pool, defining a target for pathogen evasion strategies.

| IMMUNE MEMORY AND SELECTION ON MHC
Theory suggests that NFDS is the most likely mechanism driving MHC polymorphism. 7,9Unlike other mechanisms, ongoing pathogen evolution is required to escape effective presentation of their antigens by the host's MHC to T-cells.Several potential mechanisms of such escape have been considered. 8,28First, evasion might occur via escape from effective MHC binding, as documented in HIV and a few other pathogens. 29,30Another route of escape would favour mutations in TCR epitopes, which exploit holes in the host's TCR repertoire that might result from MHC-dependent thymic selection against self-reacting lymphocytes. 23,31Thus, an MHC may be able to present an antigen, but there would be no immune response due to the paucity of lymphocytes with TCRs able to bind its epitope. 32,33Pathogens could also escape recognition by particular TCRs, 33 although such mechanisms were considered unimportant, as each pathogen could be expected to carry epitopes for lymphocytes with other TCRs, which could expand and clear infection. 28,29,34Thus, a loss of an epitope could be compensated by the response to a previously sub-dominant epitope.However, the view that escape of recognition by particular TCRs is unimportant for selection acting on MHC should be changed if secondary, rather than primary immune responses are considered.
Let us consider infection by a given pathogen (SARS-CoV-2 for example), leading to the development of T lymphocyte memory such that previously infected individuals are able to control reinfection.If a variant of the pathogen emerges in which a mutation has occurred on the epitope presented by a given HLA allomorph, such a motif is no longer recognised by the memory T-cells.Due to the narrow range of epitopes of memory T-cells relative to the epitopes available (i.e., immunodominance), this may impair the secondary immune response (Figure 1).For example, repeated influenza infections tend to focus antibody responses on epitopes conserved among the infecting strains. 35This focus may often help to respond quickly to new strains with alterations in other, nonconserved epitopes.However, when mutation occurs in epitopes of memory T-cells, this focus may result in lost protection as illustrated by 2013-2014 influenza season when pandemic H1N1 viruses acquired a new mutation in an epitope dominant in a cohort earlier exposed to the original epitope. 36This cohort was disproportionately more affected by the infection compared to a cohort who did not have an opportunity to develop memory against the original epitope. 37Immune memory may also interfere with naïve immune responses by directing the primary response towards a previously encountered epitope, even if the epitope mutated and the affinity for the new epitope is very low compared to the original one.Perhaps the best evidence of this so called "original antigenic sin," first postulated for influenza, comes from dengue virus infection, where serious manifestations of the haemorrhagic fever occur more frequently upon reinfection with another viral serotype than the original one.Mongkolsapaya et al. 38 found that during acute infections, dengue-specific T cells were of low affinity for the infecting virus, but showed high affinity for other, probably previously encountered strains.This is a different situation compared to that of a primary response, where the loss of a single epitope was argued to be easily made up by targeting another epitope during clonal expansion.Re-infected individuals who fail to raise a secondary response due to epitope escape may, of course, select a set of memory T-cells responding to new epitopes, but they will develop the symptoms of disease (with the associated morbidity) before effector T-cell and/or antibodies are produced.
Recent work by Dolton et al. 17 illustrates how mutations in epitopes can translate into frequency-dependent selection on MHC.HLA-A*02, one of the most common variants in humans, had previously been found to restrict T-cells to recognising epitopes within peptides spanning residues 261-280 of spike and 3881-3900 of ORF1ab. 26he authors reasoned that due to the high frequency of A*02, mutations that escape this variant should be strongly favoured by selection on viral proteins.Indeed, they showed that one of the epitopes, located in spike glycoprotein 269-277 (sequence YLQPRTFLL) mutated multiple times in several variant lines, including lineages classified as variants of concern, such as the mutation of position P272, which mutates to L. When the P272L mutation occurs, HLA-A*02 is still able to bind to the variant, but the variation occurs at the TCR binding site, which inhibits T-cell response and leads to memory loss.The escape mutation did seem to have epidemiological meaning, as the variant clearly increased in frequency before being replaced by the more transmissible B.1.1.7/Alpha variant.Thus, just as variants carrying mutations that escape binding by the most common MHC variants are favoured, mutations in epitopes escaping most common memory T-cells should be favoured.Because memory T-cell epitopes are conditioned by MHC, this will constitute a source of NFDS on MHC.
Immune memory can also impose frequencyindependent selection on MHC when the selective advantage of pathogen mutations arises mostly from their increased infectivity, independently of the MHC genotype.For example, a mutant that increases the efficiency of the virus's receptor binding protein (which allows the virus to enter the cell) can outcompete and eventually replace the ancestral strain.If the receptor binding protein is an antigen whose epitope can be recognised by memory T-cells conditioned by certain MHC allomorphs, a mutation increasing infectivity can, as a by-product, lead to the loss of the epitope of the memory T lymphocytes restricted by these allomorphs.Such a mechanism was illustrated by Motozono et al., 18 who showed that the epitope 448-456 located in the spike protein (sequence NYNYLYRLF) is an epitope presented by HLA-A24, an MHC allele which is widely distributed in the world and is predominant in East and South-East Asia, where an L452R mutation occurred in the epsilon variant.Another substitution on the same Y453F epitope occurred in the B.1.298variant.Both mutations increase binding affinity to ACE2, and the L452R mutation further increases viral infectivity and fusogenicity, as well as improving virus replication.In this case, selection on the virus is mostly independent of HLA/MHC allomorph frequency, as increased infectivity increases virus fitness across MHC allomorphs.However, at the same time, the underlying mutation in the receptor binding protein causes the escape form immune memory epitopes of some, but not all MHC variants, which can impose FS or contribute to HA.

To summarize
Immunological memory can impose similar types of selection on MHC as considered in standard models (i.e., NFDS, FS, and HA), but it can change the dynamics of these evolutionary mechanisms.Previous models focused on fitness differences resulting from the ability of MHC allomorphs to bind an antigen and stimulate T-cells, compared to the inability to raise an adaptive immune response.Such situations would be typical of a primary immune response, which can be initiated by a range of epitopes, provided they can be presented by MHC and that naive T-cells capable of binding to one of the epitopes presented are available.Consequently, for some pathogens such as large viruses or bacteria, the probability of at least one immunoreactive antigen binding to a given MHC molecule should not constrain the immune response. 39This raises the question as to the nature of fitness differences between MHC allomorphs during primary response. 2athogens could also escape recognition by particular TCRs, 33 although such mechanisms were considered unimportant, as each pathogen could be expected to carry epitopes for lymphocytes with other TCRs, which could expand and clear infection. 28,29,34Thus, a loss of an epitope could be compensated by the response to a previously subdominant epitope.However, the view that escape of recognition by particular TCRs is unimportant for selection acting on MHC should be changed if secondary, rather than primary immune responses are considered.
Immunological memory escape mutations are expected to underlie fitness differences between MHC genotypes resulting from success or failure to raise a secondary immune response.The differences are expressed in terms of the costs of breakthrough infections and associated morbidity.

| CONCLUSIONS
The features of immune-memory mediated selection on MHC discussed above could conceivably affect the evolution of MHC polymorphism.The increased fitness variation between MHC alleles would generally increase natural selection on MHC allomorphs.Furthermore, escape mutations in TCR epitopes, which were considered to be less important in traditional schools of thought about MHC evolution, 28 should increase the scope for host-parasite co-evolution in involving escape from immune memory, increasing the importance of NFDS.However, by enlarging fitness differences between MHC allomorphs, the scope of HA to maintain MHC polymorphism could be reduced. 11Thus, immune memory could affect the relative importance of mechanisms maintaining MHC polymorphisms and future models could benefit from including it.
Many studies have been performed to search for MHC variations and associations with diseases, revealing a number of examples in humans 40 and animals. 2 Elucidation of the molecular bases of such associations mostly involved taking into account MHC binding of the pathogenic peptidome (see for example in the case of SARS-CoV-2, 41 and for a review. 42Given the apparently common occurrence of TCR epitope escape, future studies should also attempt to explain HLA-disease association as a consequence of pathogen evasion from immune memory. Perhaps even more importantly, the type of data used to search for such associations may obscure the nature and strength of pathogen selection on MHC.For example, large scale association studies, 43,44 have provided no evidence of the effect of HLA on susceptibility to COVID-19, which contrasts with the evidence for the role of HLA in the formation of immune memory and preventing secondary infection reviewed above.However, these association studies focused on severe outcomes and therefore are likely to have missed HLA effects on SARS-CoV-2 reinfection.This suggests that future studies focusing on pathogen-driven selection on MHC would benefit from stratifying the available data according to the history of infection (and vaccination if relevant), including the sequence of pathogenic variants.
We conclude that the incorporation of immune memory within the body of theory of pathogen-driven selection leads to a better understanding of the maintenance of MHC polymorphism.Such a unified theoretical framework should be accompanied by empirical investigations that would enable the determination of the selection on MHC acting through its role on the formation of immunological memory.HLA-disease association studies should be stratified according to the different aspects of the hypothesis, this which could lead to advances in the field of medical research.