627
`
`Review
`
`Received: 24 April 2014,
`
`Revised: 14 May 2014,
`
`Accepted: 15 May 2014,
`
`Published online in Wiley Online Library
`
`(wileyonlinelibrary.com) DOI: 10.1002/jmr.2394
`
`Specificity, polyspecificity, and heterospecificity
`of antibody-antigen recognition
`Marc H. V. Van Regenmortela*
`
`The concept of antibody specificity is analyzed and shown to reside in the ability of an antibody to discriminate
`between two antigens. Initially, antibody specificity was attributed to sequence differences in complementarity
`determining regions (CDRs), but as increasing numbers of crystallographic antibody-antigen complexes were
`elucidated, specificity was analyzed in terms of six antigen-binding regions (ABRs) that only roughly correspond
`to CDRs. It was found that each ABR differs significantly in its amino acid composition and tends to bind different
`types of amino acids at the surface of proteins. In spite of these differences, the combined preference of the six ABRs
`does not allow epitopes to be distinguished from the rest of the protein surface. These findings explain the poor
`success of past and newly proposed methods for predicting protein epitopes. Antibody polyspecificity refers to the
`ability of one antibody to bind a large variety of epitopes in different antigens, and this property explains how the
`immune system develops an antibody repertoire that is able to recognize every antigen the system is likely to
`encounter. Antibody heterospecificity arises when an antibody reacts better with another antigen than with the one
`used to raise the antibody. As a result, an antibody may sometimes appear to have been elicited by an antigen with
`which it is unable to react. The implications of antibody polyspecificity and heterospecificity in vaccine development
`are pointed out. Copyright © 2014 John Wiley & Sons, Ltd.
`
`Keywords: antibody specificity; polyspecificity; heterospecificity; peptide hydropathic complementarity; epitope prediction;
`polyreactive antibodies; monogamous bivalent binding; HIV-1 vaccines
`
`INTRODUCTION
`
`Adaptive immune responses involve B cells that recognize native
`protein antigens and differentiate into antibody-secreting plasma
`cells and T cells that recognize unfolded peptide fragments of the
`antigen that have been processed in antigen-presenting cells.
`These peptide fragments known as T cell epitopes are presented
`in the groove of surface-exposed major histocompatibility
`complex class I and II molecules recognized by CD8+ and CD4+
`T cells, respectively. The pathways of antigen processing in
`antigen-presenting cells are fairly well understood[1] but fall
`outside the scope of this review.
`IgM
`The primary antibody repertoire consists mainly of
`molecules, generated independently of exogenous antigen, which
`arise in the bone marrow following immunoglobulin (Ig) variable
`(V) gene rearrangements and imprecise joining at the borders of
`integrated gene segments.
`Following encounter with antigen, B cells are activated and
`migrate into germinal center follicles where they proliferate.[2]
`Under the influence of antigen selection, a small number of
`these B cells undergo an Ig heavy chain class switch to IgG,
`IgA, or IgE molecules and produce the secondary antibody
`repertoire following a process of somatic hypermutation that
`introduces point mutations in the V regions at a rate of about
`4–10
`3 per base pair per generation.[3] Most B cells that fail
`10
`to be selected by antigen in the germinal centers undergo
`apoptosis.[4]
`Somatic hypermutation produces many amino acid substitu-
`tions in the complementarity determining regions (CDRs) of
`the V genes but fewer ones in the conserved Ig framework (FR)
`region. This leads to a considerable increase in the antibody
`
`(Ab) affinity for the antigen. The mechanisms by which B cells
`with higher affinity receptors (BCRs) are selected are not entirely
`clear. Progeny cells with higher affinity BCRs may be selected
`because they bind preferentially to the immunogen as its
`concentration decreases during antigen clearance. Another
`possibility is that unprocessed antigen molecules presented by
`follicular dendritic cells, perhaps via antibodies and Fc receptors
`picked up from the serum, are able to stimulate memory B cells
`for long periods.[5]
`Following further proliferation and antigen selection, B cells with
`increased receptor affinity differentiate into either Ab-producing
`plasma cells or memory B cells, a fate that is controlled by T cell
`help and the IL-21 cytokine.[6]
`The binding capacity of an Ab molecule resides in an antigen-
`binding cleft of 50–70 residues that is located in the N-terminal
`regions of the Ig heavy (H) and the light (L) chains. Each cleft
`harbors several overlapping binding subsites of 15–20 amino
`acid residues that are called paratopes. These subsites possess
`a particular structural and chemical complementarity to certain
`patches of residues present at the surface of protein antigens,
`known as B cell epitopes. If the context makes it clear that one
`refers to epitopes recognized by antibodies rather than by T
`cells, one simply calls them epitopes.
`
`* Correspondence to: Marc H. V. Van Regenmortel, Institut de Recherche de l’Ecole
`de Biotechnologie de Strasbourg, CNRS/UDS, UMR 7242, Illkirch 67400, France.
`E-mail: vanregen@unistra.fr
`
`a M. H. V. Van Regenmortel
`Wallenberg Research Center, Stellenbosch Institute for Advanced Study,
`Stellenbosch University, Stellenbosch, South Africa
`
`J. Mol. Recognit. 2014; 27: 627–639
`
`Copyright © 2014 John Wiley & Sons, Ltd.
`
`Lassen - Exhibit 1056, p. 1
`
`

`

`Because an Ab always harbors several individual paratopes
`able to bind different epitopes present in one or other antigen,
`an Ab is never monospecific for a single binding partner
`because this would require that the remaining 50 or so residues
`in the antigen-binding cleft that are not involved in the one in-
`teraction are unable to bind any other antigenic structure,
`which is unlikely. Steric hindrance may prevent two antigen
`molecules from binding simultaneously to the same Ab-
`binding cleft, but if the two paratope subsites do not overlap,
`it is possible for two small antigens to bind simultaneously to
`the same Ab molecule.[7]
`
`ANTIBODY SPECIFICITY
`
`The concept of antibody specificity is widely used in immunology,
`although it has rarely been defined satisfactorily. The term
`specificity is derived from the word species, and “specific”
`properties were initially considered to be properties that allow
`the members of one species to be distinguished from those of
`another. This terminology goes back to the days when individual
`species were believed to be separated by permanent, sharp
`boundaries, and Darwinian evolution and the transformation of
`species by selection had not yet become universally accepted.
`As recounted by Mayr,[8] many biologists in the 19th century
`continued to view species as fixed entities separated by clear-
`cut discontinuities that gave rise to a belief in absolute immuno-
`logical specificity. Paul Ehrlich, for instance, believed that antisera
`raised against members of different species of pathogenic
`bacteria were completely specific, allowing the members of
`different species to be distinguished serologically with absolute
`certainty.[9] When it was later discovered that antisera raised
`against cells from different animal species cross-reacted serologi-
`cally with cells from many other species, it became clear that an
`antigen was able to elicit not only a single antibody of absolute
`specificity but also a whole spectrum of antibodies that could
`cross-react with many related antigens.[10] The specificity of anti-
`gen-antibody recognition was then no longer perceived as an all
`or none phenomenon but was interpreted as a matter of more
`or less good fit between molecules that possessed different
`degrees of stereochemical complementarity.[11]
`The most
`reliable method for
`identifying epitopes and
`paratopes is by solving the 3D structure of antigen-antibody
`complexes and determining which amino acids in the two
`
`M. H. V. VAN REGENMORTEL
`
`partners make contact with each other.[12] The surfaces of
`proteins always harbor many different epitopes, and each of
`them is able to recognize complementary antibodies. When
`two antigens are compared with a panel of monoclonal antibod-
`ies (Mabs) raised against them, they will appear to be identical if
`a Mab is used that recognizes an identical epitope present in
`both antigens (Figure 1). Such a cross-reaction between two
`antigens is usually referred to as shared reactivity.[14] On the
`other hand, if a Mab is used that recognizes an epitope present
`in only one of the antigens, both antigen molecules will appear
`to be unrelated. If an investigator wants to differentiate between
`the two antigens, the first type of Mab would be called nonspe-
`cific, whereas the second Mab would be called specific because
`it discriminates between the two antigens. Instead of speaking
`of specificity, it may thus be preferable to speak of the discrimi-
`nation potential of antibodies because it is the wish of the
`investigator to distinguish between two antigens that determine
`whether an Ab is considered to be specific.
`A third possibility referred to as true cross-reactivity[14] occurs
`when an Ab recognizes an epitope that is only structurally
`related by not identical in both antigens (Figure 1). In many cases
`but not always, the Ab will react with higher affinity with the
`homologous epitope used for raising the Ab than with the
`cross-reactive heterologous epitope. It is unfortunate that Abs
`are often given names derived from the antigen they are able
`to react with, because they cannot be specific for an entire
`protein antigen that harbors many different epitopes but only
`for one of its epitopes.
`It is often assumed that Abs that possess a high binding
`affinity are likely to be also more specific because it is believed
`that they should possess a better stereochemical complementarity
`with their antigens than Abs of lower affinity. However, there is
`no necessary link between affinity and specificity because Abs
`of low affinity may be able to discriminate better between two
`antigens than Abs of high affinity. This is because these Abs
`may detect fewer cross-reactions between antigens than Abs
`of high affinity because weaker cross-reactions are more likely
`to occur below the level where they can still be detected with
`low affinity Abs.[13]
`Whereas Ab affinity is defined by a binary relationship
`between one epitope and one paratope, Ab specificity is a
`ternary relational property that has meaning only with respect
`to a minimum of three partners, for instance, one paratope and
`two epitopes. Specificity is only meaningful with respect to the
`
`Figure 1. Potential cross-reactivity between two related antigens 1 and 2. Epitope a is present on both antigens, and Mab anti-a will react in an
`identical fashion with antigens 1 and 2. Mabs anti-b and anti-d recognize unrelated epitopes on the two antigens and do not cross-react. They are
`specific for either Ag 1 or Ag 2. Mab anti-c reacts strongly with the homologous epitope c and cross-reacts weakly with the heterologous epitope e
`(from Van Regenmortel[13] reproduced with permission).
`
`628
`
`wileyonlinelibrary.com/journal/jmr
`
`Copyright © 2014 John Wiley & Sons, Ltd.
`
`J. Mol. Recognit. 2014; 27: 627–639
`
`Lassen - Exhibit 1056, p. 2
`
`

`

`Because most protein epitopes are discontinuous and usually
`involve numerous residues situated on different accessible loops
`of the protein, their structure is somewhat analogous to the
`discontinuous structure of paratopes built up from residues
`located on the different Ab hypervariable loops. Furthermore,
`because peptides of 6–12 residues corresponding to the CDRs
`of an Ab are sometimes able on their own to bind to the
`cognate intact antigen, they could be viewed as continuous
`paratopes, reminiscent of the binding capacity of continuous
`epitopes.[28] Such continuous paratopes that are able to
`recognize the antigen in a specific manner, albeit with a lower
`affinity, may contain residues found by crystallographic analysis
`of
`the antigen-antibody complex to be located at
`the
`paratope–epitope interface.[29]
`The functional binding capacity of a continuous paratope has
`been illustrated by the construction of a chimeric peptide
`consisting of one CDR of an antiviral antibody conjugated to a
`continuous epitope of another virus.[30] This construct made it
`possible to redirect the specificity of an antiviral Ab to allow it
`to recognize another virus and illustrates the functional equiva-
`lence of continuous paratopes and epitopes because they were
`able to mimic intact Abs and antigens, respectively.[28,30]
`The ability of linear peptides to bind to each other is usually
`attributed to hydropathic complementarity. This arises when
`peptides of opposite hydropathy bind to each other because
`hydrophilic residues in one peptide are oriented toward the
`aqueous solvent and liberate a space that can accommodate
`an hydrophobic residue from the other peptide.[31] An example
`illustrating the hydropathic complementarity between a contin-
`uous epitope of angiotensin II and a continuous paratope from
`the CDR-H1 of an angiotensin Ab is shown in Figure 2.
`Hydropathic complementarity is regularly observed between
`the sense and antisense peptides encoded by complementary
`sense and antisense messenger RNA.[32] There is in fact no
`exception to the rule that DNA or RNA codons (sense) and antico-
`dons (antisense) always code for amino acids of opposite
`hydropathicity (i.e., either hydrophilic or hydrophobic residues).[33]
`This remarkable pattern suggests that the genetic code may have
`evolved initially to favor the simultaneous emergence from two
`
`629
`
`Figure 2. Complementary hydropathic profiles of two interacting pep-
`tides: (●) continuous epitope (DRVYIHPF) of angiotensin II; (○) continuous
`paratope (TFNTDAMN) in the CDR-H1 of an angiotensin Mab. Hydro-
`pathic profiles were obtained using the Kyte and Doolittle hydropathic
`coefficients (from Boquet et al. 1995, Mol. Immunol. 32: 303, reproduced
`with permission).
`
`SPECIFICITY, POLYSPECIFICITY, AND HETEROSPECIFICITY
`
`capacity of an Ab to react differently with two or more antigens
`and thereby to discriminate between them.[13]
`Which selection pressure gave rise during evolution to the huge
`primary repertoire of specific paratopes[15] has been debated for
`many years, the most common explanation being that the
`immune system needed to remove either exogenous pathogens
`or antibodies to self-antigens.[16–20] However, this somewhat
`theoretical debate falls outside the scope of this review.
`
`THE STRUCTURAL BASIS OF PARATOPE–EPITOPE
`RECOGNITION
`
`Epitopes and paratopes are relational entities defined by their
`mutual complementarity, and they depend on each other to
`acquire a recognizable identity. Epitopes and paratopes are thus
`not intrinsic structural features of an antigen and Ab molecule,
`respectively, because they cannot exist in the absence of a
`relational nexus between the two partners. This means that the
`number of epitopes in a protein is equal to the number of
`different Mabs that can be raised against it.
`In this way, the
`insulin molecule can be said to have 115 epitopes[21] and the
`BLyS molecule more than a thousand.[22]
`This relational dependence means that as soon as an epitope
`is slightly altered and binding to the paratope is affected, the
`paratope is also no longer the same. This analysis differs from
`the classification of binding sites introduced by Cohn,[23] which
`defines a paratopic clan as a family of paratopes, distinguishable
`from each other, that are all functionally capable of binding a
`given single epitope. Cohn also defined a mimotopic array as a
`set of distinguishable epitopes that are all able to bind to a given
`single paratope.
`Protein epitopes are usually classified as continuous or
`discontinuous, depending on whether the amino acids that form
`the epitope are contiguous in the peptide chain or not. The
`majority of protein epitopes is discontinuous and consists of
`two to five short stretches of residues that are distant in the
`protein sequence and are brought together at the surface of
`the protein by the folding of the peptide chain.[24]
`Discontinuous epitopes are defined structurally by the amino
`acids that are found to be in contact with paratope residues in a
`crystallographic
`complex. However, discontinuous epitopes
`cannot be extracted from the protein antigen to demonstrate that
`they possess binding activity outside the context of the native
`protein, and it is also extremely difficult to reconstitute them in ac-
`tive form with sufficient 3D precision by peptide synthesis.[25,26]
`Continuous epitopes, on the other hand, always have fuzzy
`boundaries because they are identified functionally by the ability
`of short peptide fragments of the protein to bind to an antibody
`and not by establishing that all the residues in the epitope
`interact with a paratope.
`In most cases, all the residues in a
`continuous epitope do not interact with a paratope, and only
`some of its residues will be located at the surface of a native
`protein where they are usually part of a more complex discontin-
`uous epitope.[26,27]
`that
`through six CDRs
`Antibodies
`recognize antigens
`because of their enormous sequence variability are able to form
`millions of different
`antigen-binding paratopes.
`These
`hypervariable loops are denoted L1, L2, and L3 in the Ig light
`chain and H1, H2, and H3 in the Ig heavy chain, and they
`associate noncovalently at the tip of Ig Fab fragments to form
`the antigen-binding site.
`
`J. Mol. Recognit. 2014; 27: 627–639
`
`Copyright © 2014 John Wiley & Sons, Ltd.
`
`wileyonlinelibrary.com/journal/jmr
`
`Lassen - Exhibit 1056, p. 3
`
`

`

`M. H. V. VAN REGENMORTEL
`
`ABRs included as much as 96% of all the residues that actually
`bind the antigen.[51] It was also found that several residues in
`the FR region and constant regions of IgGs contributed signifi-
`cantly to antigen binding.[52]
`Using the Paratome web server, Kunik and Ofran[53] recently
`examined the amino acid composition of the 6 ABRs in 200 anti-
`body-antigen complexes, which is the largest number of com-
`plexes ever analyzed. The average lengths of the six ABRs and
`the number of SDRs in each ABR are presented in Figure 3.
`ABR H2 was found to have the longest median length of 14 res-
`idues, followed by H3 and L2 with a median length of 11 resi-
`dues. H3 showed the highest length diversity, whereas H1, L2,
`and L3 showed limited length diversity. These lengths vary
`somewhat from earlier results,[45,54,55] which may be due to the
`different methods that were used to define CDRs. The number
`of SDRs in each ABR is shown in Figure 3B. H2 and H3 have the
`largest median number of SDRs (six residues), but H3 is more di-
`verse because it may contain as many as 14 SDRs. L2 sometimes
`
`Figure 3. Lengths of antigen-binding regions (ABRs) and number of an-
`tigen-binding residues. (A) ABRs length. The black bold lines represent
`the median length. The second and third quartiles are depicted by light
`and dark gray boxes, respectively. The range of lengths is indicated by
`vertical lines. (B) Number of antigen-binding residues in each ABR. The
`black bold lines and the light and dark gray zones are as in (A). The max-
`imal and minimal number of antigen-binding residues within each ABR
`are indicated by vertical lines (from Kunik & Ofran,[53] reproduced with
`a permission).
`
`complementary nucleic acid strands, of sense and antisense pep-
`tides that are able to interact with each other as do receptors and
`ligands.[34,35]
`The specificity of interactions between two peptides is en-
`hanced by the ability of amino acids to form complementary
`protrusions and cavities and by the presence of amino acids of
`opposite charge.[36] In a recent application of the phenomenon
`of hydropathic complementarity to the study of human immu-
`nodeficiency virus type 1 (HIV-1) immune responses, a comple-
`mentary HIV gp120 antisense peptide of inverted hydropathy
`corresponding to the cA1 T cell epitope peptide was used to in-
`duce cellular immunity against HIV-1.[37]
`Additional types of protein epitopes have been recognized
`such as neotopes that arise from the quaternary structure of pro-
`tein aggregates and mimotopes that are peptides that bind to
`antiprotein Abs but show little or no sequence similarity with
`the protein used to raise these Abs.[26] Because numerous re-
`views of the antigenic structure of proteins are available,[38–41]
`the various types of epitopes will not be further discussed here.
`In recent years,
`increasing numbers of antibody structures
`have been elucidated by crystallographic analysis of antibody-
`antigen complexes, and this has given us a much better under-
`standing of the structural basis of antibody specificity. The CDRs
`of Abs vary considerably in length, whereas the individual
`paratopes within the antigen-binding site usually consist of
`10–20 residues. Residues found to be in contact with the antigen
`are often referred to as specificity-determining residues (SDRs),
`and they are more variable than residues that are not in contact
`with it.[42,43] Recently, SDRs were compared with residues that
`had undergone somatic hypermutation during affinity matura-
`tion. The results based on an analysis of 140 antibody-antigen
`complexes showed that somatic replacements occurred mainly
`in residues that were not involved in contacts with the anti-
`gen.[44] This observation is consistent with the fact that the
`hypermutation process is stochastic and occurs in a nonselective
`manner irrespective of whether residues are in contact with the
`antigen or not. However, non-SDRs may contribute to binding
`activity by helping to maintain the conformation of the binding
`site.[43]
`Initially, CDRs were identified by aligning a limited number of
`antibody sequences and determining the positions of the most
`variable residues.[45] As increasing numbers of 3D Ab structures
`became available, the hypervariable loop and constant FR re-
`gions could be located in the Ab structure, and the CDRs were
`found to adopt a restricted set of conformations termed canon-
`ical structures.[46] Different combinations of canonical structures
`alter the topography of paratopes and determine the size of the
`antigen surface with which the Ab is able to interact.[43,47]
`A third approach to define CDRs was developed using the ex-
`tended database of variable Ig genome sequences.[48,49] All
`these approaches produce slightly different residue numbering
`systems, mainly because nucleotide insertions are accommo-
`dated differently.
`More recently, a fourth approach using the Paratome web
`server was developed on the basis of a multiple structural align-
`ment of all antibody-antigen complexes available in the Protein
`Data Base.[50,51] This method identified regions of structural con-
`sensus called antigen-binding regions (ABRs), roughly corre-
`sponding to CDRs, in which the pattern of structural positions
`that bind the antigen was found to be very similar among all an-
`tibodies. The superiority of ABRs compared with previously used
`CDR identification tools was demonstrated by the fact that the
`
`630
`
`wileyonlinelibrary.com/journal/jmr
`
`Copyright © 2014 John Wiley & Sons, Ltd.
`
`J. Mol. Recognit. 2014; 27: 627–639
`
`Lassen - Exhibit 1056, p. 4
`
`

`

`SPECIFICITY, POLYSPECIFICITY, AND HETEROSPECIFICITY
`
`contains as many as 10 SDRs, although its median number of
`SDRs is only 1. For all ABRs, there are instances where they do
`not contain a single SDR.
`Some authors have concluded that the amino acid composition
`of protein epitopes does not differ significantly from that of
`protein surfaces,[56] whereas others suggested that epitopes may
`be enriched with certain types of amino acids.[57,58] In their study,
`Kunik and Ofran[53] determined the frequency of the five most
`abundant SDRs in each ABR. This is a finer analysis than that
`carried out in earlier studies that always determined the frequency
`of SDRs averaged over an entire paratope rather than in individual
`ABRs.[54,56,58]
`As shown in Figure 4, Tyr was the most frequent SDR in the
`ABRs followed by Asp, Asn, Arg, and Trp. The five most common
`SDRs in each ABR covered 63% of L1, 60% of L2, 59% of H3, 58%
`of H1, 54% of H2, and 48% of L3. Although Tyr, Ser, Asn, and Trp
`have previously been reported to be the most abundant residues
`in paratopes,[54,56,59] exactly the same distribution of residues
`was not found when the frequencies of each SDR in individual
`ABRs were considered. Tyr makes the greatest energetic contri-
`bution to antigen binding,
`in line with its ability to mediate
`different types of contacts (i.e., van der Waals, aromatic interac-
`tions, and hydrogen bonds). It should be noted that there is no
`link between the abundance of an amino acid in SDRs and its
`energetic contribution to antigen binding. The most energeti-
`cally important residues in the ABRs were found to be Tyr, Asp,
`Asn, and Arg. H3 has the highest percentage (29%) of SDRs that
`are energetically important for binding followed by L1 (24%) and
`H2 (22%).
`The analysis of Kunik and Ofran[53] for the first time revealed
`that the six ABRs differ significantly in their amino acid composi-
`tions and that each ABR tends to bind different types of amino
`acids at the surface of proteins. Because the six ABRs, because of
`their significantly different amino acids compositions, have differ-
`ent contact preferences for certain epitope residues,
`it seems
`plausible that epitopes might also possess distinguishable amino
`acid compositions. However, when the amino acid composition
`of epitopes was compared with that of entire protein surfaces,
`
`In other
`no noticeable differences were observed (Figure 5).
`words, although each ABR has a unique set of contact preferences
`favoring certain epitope residues over others, the combination of
`all these individual ABR preferences yields a collective amino acid
`composition of epitopes that is very similar to the composition of
`protein surfaces in general.[53,56]
`Because the entire accessible surface of a protein is a continuum
`of potential epitopes,[39] it could be argued that it would be advan-
`tageous for Abs to bind any protein surface patch without requiring
`specialized sites of increased stickiness. It seems that antibodies are
`in fact able to achieve this because they have evolved a set of ABRs
`where each ABR binds different types of amino acids, while the
`combined preference of the entire set is for epitopes that are
`indistinguishable from the rest of the protein surface.[53] The situa-
`tion is different in most protein–protein complexes, for instance,
`virus–host receptor partners, which have optimized their mutual
`complementarity over long periods of biological evolution. This
`coevolution has favored the selection of small “sticky” areas at the
`surface of proteins that are enriched in hydrophobic, aromatic,
`and charged residues.[60] This is not the case with epitope–paratope
`partners because paratopes need to optimize their complementarity
`to epitopes fairly rapidly (i.e., weeks or months) within the context
`of an individual immune system.
`
`THE EPITOPE PREDICTION CONUNDRUM
`
`The recent results of Kunik and Ofran[53] and Kringelum et al.[56]
`have revealed a major difficulty when attempts are made to
`predict epitopes at the surface of proteins. Because no amino
`acid was found to be significantly over represented in epitope
`regions compared with the rest of the surface and because the
`surface is an antigenic continuum, most of the protein surface
`can be expected to be part of some epitope potentially recogniz-
`able by one or other antibody. Success in predicting that some
`residues are part of an epitope will thus improve as the number
`of examined antibodies raised against the antigen increases[61]
`and not necessarily because some residues are inherently more
`
`631
`
`Figure 4. The five most frequent antigen-binding amino acids in each antigen-binding region (ABR). For each ABR, the list of residues that contact the
`antigen is listed, and the frequency of each amino acid is indicated (from Kunik & Ofran,[53] reproduced with permission).
`
`J. Mol. Recognit. 2014; 27: 627–639
`
`Copyright © 2014 John Wiley & Sons, Ltd.
`
`wileyonlinelibrary.com/journal/jmr
`
`Lassen - Exhibit 1056, p. 5
`
`

`

`M. H. V. VAN REGENMORTEL
`
`Figure 5. Amino acid composition of antigen surface residues and of epitopes. The frequency of each amino acid is calculated for exposed residues
`and for epitope residues, using a nonredundant set of antibody-antigen complexes (from Kunik & Ofran,[53] reproduced with permission).
`
`immunologically active than others. This may be one reason why
`attempts to predict protein epitopes using various amino acid
`propensities have been notoriously unsuccessful.[26,27,62–70] How-
`ever, there may also be other reasons for this lack of success. For
`many years, investigators concentrated mainly on the prediction
`of continuous epitopes because these correspond to short pep-
`tides that can easily be synthesized and could readily be used
`to replace pathogenic antigens in diagnostic immunoassays[71,72]
`or to act as immunogens for raising antipeptide antibodies that
`cross-react with the cognate protein.[73] Unfortunately, when
`one refers to continuous epitopes of a protein, the impression
`is created that these epitopes exist as such in the native protein.
`In reality, these so-called protein epitopes are mostly short linear
`peptide fragments of more complex discontinuous epitopes that
`cross-react only weakly with certain antiprotein Abs and possess
`only limited structural similarity with small regions of the protein
`surface that may be as short as dipeptide or tripeptide se-
`quences.[27,74,75,70] Predicting continuous epitopes is thus of lim-
`ited value for analyzing the antigenicity and immunogenicity of
`native proteins.
`It is astonishing that in spite of innumerable
`failed attempts to use continuous epitopes as potential synthetic
`vaccines,[26,76,77] many investigators continue to use short linear
`peptides, possibly because they are dubbed continuous epitopes
`of a pathogen, as promising candidates for developing synthetic
`vaccines.
`Because it is increasingly accepted that the vast majority of pro-
`tein epitopes are discontinuous, many investigators have
`attempted to develop computational prediction algorithms that
`they claim allow the prediction of discontinuous epitopes.[26,65–70]
`What these methods do is to predict that a small number of
`residues located at the protein surface are likely to be part of a
`discontinuous epitope. What these methods are unable to do is
`to identify which full set of residues from distant parts of the
`protein sequence need to be assembled in a precise conformation
`to form an active site endowed with the antigenic and immuno-
`genic properties of a discontinuous epitope present in the native
`protein.[27,65–71,78] To claim that such methods allow discontinuous
`epitopes to be “predicted” is rather ambiguous because they are
`only able to “map” or “identify” a limited number of residues pres-
`ent in complex discontinuous epitope. Discontinuous epitopes are
`only defined structurally by determining which residues of an anti-
`gen are in contact with paratope residues and not functionally by
`showing that this complete set of residues, when it is not embed-
`ded in the protein, achieves adequate immunological mimicry. Be-
`cause the main purpose of epitope prediction is to replace a
`complete protein antigen by a small fragment of the molecule that
`possesses the antigenic and/or immunogenic properties of one of
`its epitopes, the feasibility of predicting functional discontinuous
`epitopes by computation remains at present a rather elusive goal.
`
`Recently, a completely different approach to epitope prediction
`has been proposed that utilizes the paratope sequences of a
`series of Mabs raised against an antigen in order to predict which
`surface regions of the antigen are likely to be recognized by these
`Mabs.[61]
`Instead of predicting which residues of an antigen
`possess a superior capacity for binding to any potential Ab, this
`method attempts to predict patches of about five residues located
`at the surface of a protein that are likely to be recognized by one
`or a few available Mabs.
`The method is based on residue pairing preferences that have
`been shown to exist in ABRs and epitopes,[53] and it predicts
`potential matches between a given Ab and a given epitope. It
`also utilizes cross-blocking inhibition assays with different Abs
`to assess if the epitopes they recognize overlap or not.
`Because patches of five residues cannot represent a complete
`discontinuous epitope, their usefulness to serve as potential
`diagnostic reagents or immunogens is likely to be rather limited,
`especially because such applications always
`involve the
`participation of a large variety of different Abs. The method also
`de