Antigenic Structures Recognized by Cytotoxic T Lymphocytes

The antigenic structures recognized by T lymphocytes differ fundamentally from those recognized by B lymphocytes. B cells, through antibody molecules embedded in their cell surface membrane, collectively recognize an enormous diversity of antigens in solution or on cell surfaces, including native and unfolded proteins, peptides, polysaccharides, nucleic acids, steroids, and small organic molecules. The antigen-specific receptors on T cells are very similar to antibodies structurally and in the genetic basis for their diversity. Never-theless, the antigenic structures recognized by T cells consist almost exclusively of peptides associated with integral membrane glycoproteins known MHC’ proteins they are encoded in a genetic region called the major histocompatibility complex). T cells recognize protein antigens in a fragmented (or “processed”) on the surface of other cells expressing appropriate MHC molecules (a phenomenon termed MHC “restriction” of the T cell response The MHC genetic region called HLA the human H-2 the tightly linked genes, a of alleles a array

The MHC genetic region (also called HLA in the human and H-2 in the mouse) contains many highly polymorphic, tightly linked genes, and a particular set of alleles defines a haplotype. Any given individual in a population has an array of MHC molecules on the surface of its cells (one haplotype from each parent, codominantly expressed) which serves as a signature for that individual (or for the genetically uniform mice or rats of inbred strains). The essence of MHC restriction is that T cells normally recognize antigens on cells having the same haplotypes as the T cells themselves but not on cells of different haplotypes. (However, in the context of organ transplantation between individuals of different haplotypes, nonself MHC molecules are targets for a T cell-mediated rejection response. It was this "allogeneic" reaction which led to the discovery of the MHC and to the origin of the term histocompatibility.) Mature T cells can be divided into two groups based on their cell surface expression of either CD8 or CD4 glycoproteins. Each group interacts with a different set of MHC molecules, called class I or class 11, respectively. MHC molecules of both classes are noncovalently associated heterodimers on the cell surface, but class I is composed of a membrane-spanning 45-kDa LY (or heavy) chain and a soluble, nonpolymorphic 12-kDa subunit (&-microglobulin or light chain), whereas class I1 consists of two transmembrane chains, 01 (35 kDa) and , B (29 kDa). Only certain "antigenpresenting" cells (APC) express MHC-I1 (macrophages, dendritic cells, B cells, and a few others). When CD4+ T cells recognize class 11-expressing APC that have reacted with appropriate antigens, they usually respond by secreting lymphokines that influence lymphocytes and other cell types; hence these T cells are designated helper T cells. However, ' The abbreviations used are: MHC, major histocompatibility complex; APC, antigen-presenting cells; CTL, cytotoxic T lymphocytes; ER, endoplasmic reticulum; &m, /3n-microglobulin. most nucleated cells express MHC-I, so CD8+ T cells can potentially respond to virtually all cell types. In the presence of a suitable antigen, class I-expressing target cells are specifically destroyed by CD8' T cells; hence the latter are termed killer cells or cytotoxic T lymphocytes (CTL).
The division between class I and class I1 proteins is important not only for the different T cell responses provoked, but because they define different pathways through which peptides are presented to T cells (2). In particular, extracellular proteins (also termed exogenous) are taken up by APC, partially digested in a low pH endosomal compartment, and the resulting peptides bound by MHC-I1 for transport to the cell surface where they can stimulate CD4' T cells. In contrast, antigens synthesized within a cell (endogenous), such as viral proteins, tumor antigens, or so-called minor histocompatibility antigens, are degraded intracellularly to peptides and associate with MHC-I for transport to the cell surface where they flag the cell for recognition and lysis by CD8+ CTL. The separation between the two antigen-processing pathways (Table I) is not absolute, but it offers a useful framework for probing the cell biology and biochemistry of each pathway and for evaluating exceptions. This minireview will concentrate on the class I pathway and in particular on the molecular interactions between antigen and MHC-I prerequisite to recognition by CD8' CTL.

CD8+ CTL Recognize Peptides Derived from Endogenous Antigens
Viral infection elicits a vigorous CTL response leading to the specific lysis of infected cells. Experimental studies with influenza virus have revealed important features of the antiviral CTL response (3). Whereas anti-influenza antibodies recognize subtype-specific, virally encoded glycoproteins expressed on the surface of infected cells, CTL from both mice (4) and humans (5) cross-react with target cells infected by different influenza virus subtypes. This cross-reactivity was explained by the subsequent finding that most anti-influenza CTL are specific for conserved, intracellular viral proteins undetectable at the cell surface (e.g. nucleoprotein, matrix protein). Thus, cells transfected with the nucleoprotein gene served as targets for influenza-specific CTL despite the absence of cell surface nucleoprotein, giving rise to the suggestion that the antigen is converted into fragments (peptides) which are transported to the cell surface for CTL recognition (6). This model was confirmed by experiments showing that incubation of cells with synthetic peptides mimicking parts of the nucleoprotein rendered the cells susceptible to lysis by CTL (7). Incubation with whole nucleoprotein was ineffective, and only certain peptides from the nucleoprotein sequence were active. Presumably such peptides, substituting for peptides naturally produced from endogenously synthesized antigen, somehow associate directly with MHC-I molecules to form the complex recognized by CTL.
Subsequent studies with mouse and human CTL by many workers have identified at least 40-50 different synthetic peptides recognized in association with various MHC-I molecules. In addition to virally derived peptides, peptides corresponding to sequences from other endogenously synthesized  immunogenicity in mouse cell lines (9). Thus many tumorspecific antigens (and minor histocompatibility antigens) are probably peptide fragments recognized at the cell surface by MHC-restricted CTL, explaining the historical failure of antibodies to detect these antigens.

Evidence for Peptides MHC-I Complexes
A dramatic and compelling image of how a peptide. MHC-I complex might look was provided by the 3.5-A x-ray crystallographic structure of a human MHC-I molecule, HLA-A2 ( 1 0 , l l ) . The structure consists of four domains of approximately 90 amino acid residues each, three derived from the a-chain (a1, az, and ag) and one comprising &microglobulin (transmembrane and cytosolic portions of the a-chain were removed by papain digestion early in the purification). The membrane proximal a3 and &microglobulin domains have tertiary structures resembling immunoglobulin folds, consistent with their known amino acid sequences. The a1 and az domains each have a nearly identical but novel structure containing four antiparallel @-strands and a long a-helix, such that when paired these domains form a single platform of eight &strands topped by two a-helices. A long groove overlying the @-sheet and between the two a-helices was proposed to be the peptide binding site based on several considerations. 1) Most of the polymorphic residues of HLA cluster in or near the proposed binding site, accounting for the ability of different MHC molecules to interact with different peptides. 2) Residues known to be critical for T cell recognition through the use of natural or engineered MHC variants are similarly in or near the site. 3) oThe dimensions of the groove (25 A long X 10 A wide X 11 A deep) could accommodate a peptide of between 8 amino acids (if fully extended) and 20 amino acids (if fully coiled into a helix), in accord with the lengths of synthetic peptides used to elicit T cell responses. 4) Most intriguingly, a continuous region of electron density not attributable to the HLA-A2 sequence is present in the groove; this "extra" electron density is believed to represent peptide or a mixture of peptides still bound to HLA. A second x-ray structure, that of HLA-Aw68, confirmed these features and also displayed unassignable electron density in the corresponding site; notably, this electron density was distinguishable from that seen with HLA-A2, suggesting a different composition of putative peptiGe occupant(s) (12). Even after structural refinement to 2.6-A resolution the extra electron densities in the HLA grooves have resisted sequence assignment, supporting the idea that a mixture of peptides may be included.
Consistent with this inference, the biology of the system points to considerable degeneracy in peptide binding to MHC glycoproteins. The cells of an individual vertebrate organism express only a handful of different MHC-I proteins, probably up to six in humans (two each at the HLA-A, -B, and -C loci). Yet this small number of proteins mediates the MHC-restricted recognition of many thousands of peptide antigens by an equally large number of different T cells. Evidently, each MHC protein must have the capacity to interact with an extremely large diversity of peptide sequences. Nevertheless, binding is not totally degenerate, as MHC proteins vary in their abilities to present specific peptides (13). In this context, the extra electron densities seen in the two HLA structures may be interpreted as different mixtures of tightly bound peptides which co-purified and co-crystallized with their respective HLA proteins. Such degenerate binding contrasts profoundly with the highly specific "lock-and-key'' mechanism considered a hallmark of immunological recognition and epitomized by the reaction between antigen and antibody (14) (and indeed probably also between peptide.MHC complex and T cell receptor (15)).

What Peptides Bind to MHC-I?
Efforts to understand the structural basis for binding between peptides and MHC-I proteins have taken several directions, including: 1) tabulation of synthetic peptides recognized by MHC-restricted T cells in order to discern common features or motifs and develop predictive binding algorithms; 2) structural characterization of naturally occurring peptides recognized by MHC-restricted T cells; 3) in vitro binding studies using synthetic peptides and purified MHC-I molecules; 4) reconstitution of peptide-free MHC-I with a known, homogeneous peptide for x-ray crystallography; and 5 ) elucidation of the physiological pathway for generation of peptide. MHC-I complexes by genetic and cell biological approaches.
Examination of several dozen peptides which elicit MHCrestricted T cell responses (both class I and class 11) led to the formulation of predictive schemes based on common structural elements (16-19). The resulting algorithms often reflect some form of recurrent hydrophobicity (20). While several predicted synthetic peptides were biologically active in T cell assays, other peptides found by random screening to have similar activities did not fit the expected patterns. The relatively small (but growing) data base of available peptide sequences may have limited this approach, especially considering the distinctions in peptide binding by different MHC proteins. Furthermore, in most cases it is not known which residues of a given peptide are actually required for MHC binding, which interact primarily with a T cell receptor, which serve a role other than for specific binding, and which are superfluous or artifactual due to their synthetic origin. Until recently the relationship between synthetic and naturally occurring peptides which bind MHC-I was largely uncertain, probably because characterization of natural peptides is challenging from a technical standpoint in view of their likely heterogeneity and consequent low abundance. Gel filtration fractionation of cell extracts followed by assay of the fractions for T cell recognition first gave clues as to natural peptide size (21, 22) and later led to the identification of two viral peptides (nonamers) whose sequences are contained within known biologically active synthetic peptides (23). A third viral peptide recovered directly from immunoprecipitated MHC-I molecules similarly represents a portion (octamer) of a known synthetic peptide (24). Thus in all likelihood natural peptides are relatively small peptides with no particular chemical modifications or structural themes yet evident.
Direct binding studies require a source of purified MHC molecules, which are available in low milligram amounts by detergent solubilization or papain digestion of membranes from a large number of cultured cells (e.g. 10" cells, representing 50-100 liters of culture) (25-27). Conventional binding assays such as equilibrium dialysis and gel filtration have yielded abundant binding data for MHC-I1 systems, with demonstrable association constants for peptides on the order of lo6 M" (28, 29). However, synthetic peptides known to interact with MHC-I on target cells fail to reveal appreciable a chain 3359

FIG. 1. Peptides andgzm associate with membrane-bound a-chain to form stable a*&m*peptide heterotrimers. This multistep process occurs intracellularly, probably in the endoplasmic reticulum, and may
also occur / at the cell surface when peptides and &m are present in the extracellular medium.
J binding to the appropriate purified MHC molecules by either gel filtration (30) or equilibrium dialysis (31). Substantial binding of detergent-solubilized MHC-I to peptides immobilized on plastic surfaces can be detected, but this binding appears largely indiscriminate with respect to MHC speciiicity, and it is unclear whether the solid phase environment favors binding by particular peptides due to physical properties other than intrinsic binding affiiities (32-34). A similar report, showing more specificity, involved the addition of peptides to immobilized MHC-I to evoke a T cell response

(35).
It may be that the low binding observed between peptides and MHC-I in solution accurately reflects an intrinsically low affdty. If the association constanta are as low as lo4 M" (consistent with available data), cell surface class I molecules will be 1% saturated at equilibrium given a peptide concentration of 10" M to induce half-maximal CTL lysis, and this might result in sufficient peptide-MHC-I complexes (100-BOO/target cell) for CTL recognition. However, this scenario is by no means the only plausible one. If the x-ray structures of HLA are taken to include tightly bound peptides, their dissociation rates must be extraordinarily slow, since purification procedures take at least 7-10 days. Therefore the purX1ed HLA used in binding assays may similarly be occupied by unknown peptides which reduce the concentration of available binding sites. If such prior complexes exist in a large (perhaps even near-stoichiometric) proportion, failure to detect binding of labeled peptides by equilibrium dialysis is understandable and constitutes further support for exceptionally slow dissociation rates, since otherwise some exchange would be expected to occur. Assuming cell surface MHC-I is also tightly complexed with unknown peptides, sensitization of target cells for CTL lysis may involve binding to a small proportion of peptide-free MHC-I sites on the cell, again yielding relatively few antigenic complexes per cell. Whether all purified MHC-I molecules harbor unknown peptides is not yet known and in fact cannot be determined by x-ray crystallography, as only a small proportion of purified MHC-I ( 4 % ) crystallizes?

m e r e Do Peptide-iW3C-I Complexes Form?
Recent insights into the assembly of peptide -MHC-I complexes derive from studies of a mutant mouse cell line, RMA-S, having what appears to be a defect in the generation or transport of endogenous peptides into the endoplasmic reticulum (ER) (36). Ordinarily, nascent class I a-chains and szmicroglobulin (p2m) associate into heterodimers in the ER and are transported to the cell surface within 30-60 min of their biosynthesis (37,38). In RMA-S cells, a-chains and Am are synthesized normally but largely fail to assemble, so that cell surface MHC-I expression is only about 5% of wild-type levels under standard culture conditions (37 "C). When in-' D. Wiley. personal communication.

~~ ~~
-c J a % peptides fected with influenza virus, RMA-S cells are not recognized by class I-restricted CTL, but when incubated with relatively high concentrations of suitable synthetic peptides from the virus, they become good targets for CTL and, surprisingly, express 2-5-fold more MHC-I on their surface (39). These findings were explained by hypothesizing that synthetic peptides are taken up by RMA-S, somehow gain access to nascent a-chains and Bzm in the ER, and promote the proper folding, assembly, and transport to the cell surface of peptide.MHC-I complexes. The model implies that cell surface expression of MHC-I normally depends on the presence of peptides in the ER to help a and B2m associate, and that RMA-S lacks this function. Subsequently, however, increased surface expression of MHC-I was fortuitously demonstrated by incubating RMA-S at reduced temperatures (23-31 "C), and these MHC-I molecules were unstable at 37 ' C unless an appropriate synthetic peptide was added (40). Thus enhanced MHC-I expression in the presence of peptides might also be explained by a stabilization of molecules already at the cell surface (but not otherwise detected because they are conformationally altered and/or short-lived). Such labile and evidently peptide-free MHC-I molecules could either reach the cell surface without peptides or could release bound peptides via dissociation (41).
Further evidence that added peptides can interact with MHC-I at the cell surface comes from studies using inhibitors of protein export. Treatment of non-mutant cells with the compound brefeldin A (which blocks transport from the ER to the Golgi complex) or the viral product E19 (which specifically binds to and retains MHC intracellularly) prevents the presentation of endogenous (e.g. viral) antigens but does not interfere with the presentation of synthetic peptides to CTL (42-44). Similarly, if cells are "fixed" using glutaraldehyde (which stops protein synthesis and turnover), presentation of endogenous antigen but not of added peptide is abrogated (45). On the other hand, the increased MHC-I detected on RMA-S cells in the presence of added peptide could result from either peptide-induced stabilization of cell surface MHC-I or from association of peptide with MHC-I intracellularly, much like endogenously arising peptides.
Whether the effect of exogenous peptides on MHC-I takes place at the cell surface or in the ER (or perhaps both, depending upon peptide concentration), the properties of RMA-S and of a comparable human mutant cell line (46,47) indicate that most stable MHC-I molecules at 37 "C include peptide as an integral partner with a-chain and &m. Moreover, &m can also stabilize MHC-I structure, as shown by an increased association between peptides and MHC-I when free &m is added to cells under limiting conditions of peptide (48). In addition, in vitro reconstitution of MHC-I from separated a-chains and &m is more efficient in the presence of either excess &m (49) or specific peptides (50); similar effects are obtained using detergent lysates from RMA-S cells (51). Finally, distinct MHC-I molecules are known to differ in their requirements for Pzm to form stable cell surface structures (52,53), and they may differ in their requirements for peptide as well (54).

How Are Peptides Generated in Vivo?
To elicit an effective class I-restricted CTL response, an endogenous antigen must be degraded into peptide fragments which can interact with MHC-I (a-chains and Pzm) to form functional complexes. Since MHC-I is co-translationally inserted into the ER (55), and intracellular antigens which do not enter the ER are recognized by CTL, a cytosolic proteolytic apparatus as well as a mechanism for peptide transport into the ER must be involved. It is not clear whether peptide fragments are generated by known or as yet unknown proteolytic systems, or how peptides are translocated into the ER, although recently a member of the protein family designated "ABC transporters" (for ATP-binding cassette) (56) has been implicated in the RMA-S mutation (57).
An alternative hypothesis provides for proteolysis within the ER, and according to the unexpected finding that mouse cells identical at all genes but MHC-I possess different peptide profiles, MHC molecules themselves may be involved in the generation or selection of presented peptides (58). For instance, MHC-I could shield bound peptides from degradation or serve as a template for proteolysis by other unknown proteins in the ER. Finally, an unusual pathway proposed for the generation of peptides avoids proteolysis altogether by theorizing the direct transcription and translation of short subgenic regions called "peptons" (59).

From Heterodimer to Heterotrimer
However they come about, it is increasingly clear that stable MHC-I molecules are most often heterotrimers consisting of a-chain, Pzm, and a short peptide of variabie sequence. Association between peptide and MHC-I appears not to arise from the equilibrium binding of two stable entities, according to the classical lock-and-key paradigm, but rather as the end point of a complex assembly process involving three chains. The natural order of assembly of the heterotrimer and the factors which govern its stability remain open to question, but once formed, loss of either soluble component (peptide or Pzm) can soon lead to collapse of the overall structure, although cell surface Bzm is known to exchange with Pzm in the medium (60). Given the strong similarity between the x-ray structures of Bzm as a monomer (61) and as part of HLA complexes, and the lack of direct contact between Pzm and bound peptides, it appears that conformational changes which accompany heterotrimer formation primarily involve a-chain and peptide (Fig. 1). Direct physical studies may have to await the more ready availability of large quantities of MHC-I protein, but in the meantime the cell biology of this system remains an extremely active and fruitful area for investigation.