How a Single T Cell Receptor
Recognizes Both Self and Foreign MHC
Leremy A. Colf,1,3Alexander J. Bankovich,1,3Nicole A. Hanick,1Natalie A. Bowerman,2Lindsay L. Jones,2
David M. Kranz,2and K. Christopher Garcia1,*
1Howard Hughes Medical Institute, Departments of Molecular and Cellular Physiology, and Structural Biology,
Stanford University School of Medicine, Stanford, CA 94305, USA
2Department of Biochemistry, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA
3These authors contributed equally to this work.
ab T cell receptors (TCRs) can crossreact with
both self- and foreign- major histocompatibility
complex (MHC) proteins in an enigmatic phe-
the 2.35 A˚structure of the 2C TCR complexed
with its foreign ligand H-2Ld-QL9. Surprisingly,
we find that this TCR utilizes a different strategy
to engage the foreign pMHC in comparison to
the manner in which it recognizes a self ligand
morphic residues on Ldand Kb, as well as the
unrelated QL9 and dEV8 peptide antigens, in
unique pair-wise contacts, resulting in greater
structural complementarity with the Ld-QL9
complex. In the structure of an engineered,
high-affinity 2C TCR variant bound to H-2Ld-
tion persists despite modified TCR-CDR3a
interactions with peptide. Thus, a single TCR
recognizes two globally similar, but distinct li-
gands by divergent mechanisms, indicating
that receptor-ligand crossreactivity can occur
in the absence of molecular mimicry.
ab T cell receptors (TCRs) recognize processed antigenic
peptides presented in association with major histocom-
patibility complex (MHC) proteins (Davis and Bjorkman,
1988). The TCR repertoire is the product of a thymic
‘‘education’’ process during which immature T cells
undergo first positive and then negative selection on self
(syngeneic, syn) peptide MHC (Starr et al., 2003). Yet, ma-
ture T cells exhibit a high frequency of crossreactivity, or
alloreactivity (1%–10%) toward foreign (allogeneic, allo)
peptide MHC to which they have not been previously ex-
posed during thymic education (Bevan, 1984). The phe-
nomenon of alloreactivity indicates an inherent ability of
the TCR to crossreact with a broad range of self- and
foreign-pMHC ligands, which could be beneficial for im-
mune surveillance of a universe of potential pathogens
(Sherman and Chattopadhyay, 1993). However, it pres-
ents a major clinical problem for organ transplantation in
that genetically mismatched tissue can be rejected in
agraft-host alloresponse. Themolecular basisofalloreac-
tivity remains poorly understood, despite the fact that the
general structural principles of TCR/pMHC interactions
have been greatly advanced by approximately 14 cocrys-
tal structures, (reviewed in Garcia and Adams ,
Housset and Malissen , Rudolph et al. , and
Turner et al. ). More broadly, receptor-ligand cross-
reactivity has been observed across many systems and
has generally been attributed to ‘‘molecular mimicry.’’
However, structural evidence for molecular mimicry, in
the form of complexes between one receptor and multiple
similar, but different ligands, remains elusive for most
The basis of alloreactivity has been extensively de-
bated, and current hypotheses generally converge on
either MHC- or peptide-centric mechanisms (Housset
and Malissen, 2003; Huseby et al., 2004; Sherman and
Chattopadhyay, 1993). In the former, alloreactive TCRs
recognize either shared or polymorphic structural deter-
minants on the MHC helices of syn- and allo-MHC with
disregard for the peptide (Daniel et al., 1998). This model
is in accord with an imprinted ‘‘bias’’ in TCR germline-
derived Variable regions for recognition of MHC structural
features by virtue of TCR/MHC coevolution (Turner et al.,
2006; Zerrahn et al., 1997). There is some structural sup-
port for alloreactivity being based on TCR recognition of
shared MHC residues. Several crystal structures of allo-
pMHC complexes show that the TCR germline-derived
CDR1 and CDR2 make identical contacts with the helices
in different allo-MHC complexes (Luz et al., 2002; Reiser
et al., 2000). A limitation of these studies was either that
(1) the allocomplexes could not be compared with the
corresponding self complexes or (2) the allo-pMHC mole-
cules were very similar to one another. In the in vivo
setting, such as in transplant rejection, TCRs mount
Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc. 135
alloresponses to widely different MHC haplotypes and
structures. The other mechanism of alloreactivity pro-
posed is that the TCR primarily focuses on the peptide,
with only moderate regard for the MHC. Since the current
database of TCR/pMHC complex structures contain dif-
ferent TCRs bound to different pMHC, their direct com-
parisons do not distinguish between effects of the differ-
ent TCRs and effects of the different peptides. In order
to advance our understanding of this issue, it would be
informative to elucidate the structure of a single TCR in
complex with different self and foreign pMHC.
The2CTCRrepresents anidealsystem foralloreactivity
studies, since the identity of both the self and foreign pep-
tides and MHC molecules that positively and negatively
select 2C T cells are known (Chen et al., 2003). The self
MHC recognized by 2C is H-2Kb, which can positively se-
lect 2C T cells when presenting the peptide called dEV8,
derived from the enzyme NADH-ubiquinone oxidoreduc-
tase (self peptide) (Santori et al., 2002; Tallquist et al.,
1998). When Kbis mutated to generate the allo-MHC
Kbm3, the dEV8/Kbm3ligand is converted into a weak ago-
ing peptide recognized by the 2C TCR, called SIYR, was
isolated from a combinatorial peptide library in a screen
using peripheral 2C T cells (Udaka et al., 1996). Crystal
structures of these three 2C/Kb-peptide complexes
showed they were very similar except for minor differ-
ences (Figure S1 and Tables S2 and S3) (Degano et al.,
2000; Garcia et al., 1998; Luz et al., 2002). These com-
plexes generally exhibit the characteristic TCR/pMHC ori-
entation now seen in other TCR/pMHC complexes, albeit
to varying degrees: the CDR1 and 2 of a and b chains pri-
marily lie over the MHC helices, but also contact peptide,
while the CDR3 contacts the bound peptide and peptide-
proximal regions of the MHC helices (reviewed in Garcia
and Adams , Housset and Malissen , and
Rudolph et al. ).
1984) and are negatively selected (Sha et al., 1988) on the
allo-MHC H-2Ld. One of the Ld-associated peptides
known to be recognized by 2C is derived from the enzyme
a-ketoglutarate dehydrogenase (Udaka et al., 1992). Ld
complexes with this self-peptide, p2Ca, and its single
amino acid extension QL9, have affinities that are approx-
imately forty-fold higher than H-2Kb-dEV8 (Garcia et al.,
1997). The sequence of the QL9 peptide (QLSPFPFDL)
is entirely different from dEV8 (EQYKFYSV), and the
MHC a1a2 domains of H-2Kband H-2Ld, which present
peptide and are recognized by 2C, differ by 31 residues
(Hansen et al., 2000). The H-2Ldhelices contain most of
the identical residues used by H-2Kbto contact the 2C
TCR, but the helices also contain 14 residue polymor-
phisms. Thus, these structurally divergent foreign- and
self-pMHC complexes offer 2C a choice between similar
or distinct composite recognition surfaces. Accordingly,
structural solutions of this system would be highly infor-
mative in advancing our understanding of alloreactivity.
Furthermore, alloreactivity is a biologically relevant exam-
ple of how one protein necessarily crossreacts with many
structurally distinct ligands in order to carry out its func-
tion. The structural principles, then, of a comparison of
distinct self and foreign MHC will expand our understand-
ing of protein-protein crossreactivity in other natural
Previous efforts to obtain cocrystals of 2C/H-2Ld-QL9
were hampered by the instability of H-2Ldand the hetero-
geneity of the 2C TCR expressed from insect cells. We
recently described the in vitro evolution and expression
of a stabilized ‘‘platform’’ (a1a2 domains, residues
1–180 of the heavy chain) version of H-2Ldthat bypasses
the historical problem of poor b2m association with the a3
form pMHC molecule exhibits identical affinity and speci-
ficity for 2C, as the full-length wild-type H-2Ld-QL9 com-
plex. Through similar in vitro evolution methodologies we
have also been able to express a soluble single-chain Fv
(VaVb) version of 2C (Kieke et al., 1999; Shusta et al.,
2000). Using these smaller, more stable versions of the
TCR and pMHC, we were able to crystallize and solve
the structure of the 2C/H-2Ld-QL9 complex at 2.35A˚
resolution (Table S1 and Figures S2 and S3).
is generally similar to previously determined TCR/pMHC
complexes for both class I and class II MHC (Figure 1A).
The individual structures of the 2C VaVb (Fv) module and
the platform Ldsuperimpose closely with their full-length
0.96 A˚and 0.81 A˚for Ca atoms, respectively) (Figure S2).
The orientation places the a chain of the TCR mainly over
the a2 helix of the MHC and the center of the peptide, with
the bchain of the TCRover the a1helix andthe Cterminus
of the peptide (Figure 1C). The ?44?2C/Lddocking orien-
tation, or ‘‘footprint,’’ falls within the range seen for com-
plexes of both MHC classes (Garcia and Adams, 2005;
Housset and Malissen, 2003; Rudolph et al., 2006; Turner
et al., 2006). For the remainder of this paper, we describe
this complex only in the context of its relationship to the
2C/H-2Kb-dEV8 complex (Figure 1B), and the attendant
implications for alloreactive recognition. Any of the three
crystal structures of 2C/H-2Kb-peptide complexes re-
ported suffice for this comparative analysis, as they are
similar within the limits of the different resolutions (Fig-
ure S1 and Tables S2 and S3). We have chosen H-2Kb-
dEV8 as the main comparator because it is a bona fide
positive-selecting pMHC complex for the 2C TCR (Santori
et al., 2002) and therefore the most direct counterpoint to
the 2C/H-2Ld-QL9 complex.
Strikingly, the 2C binding orientations on the syngeneic
(H-2Kb-dEV8) and allogeneic (H-2Ld-QL9) pMHC are
highly divergent. The footprint of 2C on Ld-QL9 (Figure 1C)
is predominantly limited to one helix of the MHC for each
TCR chain, while in the 2C/Kb-dEV8 complex, the foot-
prints of each TCR chain span both helices (Figure 1D).
136 Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc.
Althoughthe 2C/Ld-QL9complexburies ?275A˚2lesssur-
face area in the interface than 2C/Kb-dEV8 (1700 A˚2ver-
sus 1975 A˚2, respectively), the 2C/Ld-QL9 interface is me-
diated by an additional ?111 interatomic contacts (305for
Ld-QL9 and 194 for Kb-dEV8) (Tables S2 and S3). This is
due to the more intimately packed interface in 2C/Ld-QL9,
which shows a significantly superior shape complemen-
tarity (Sc = 0.7) compared to 2C/Kb-dEV8 (Sc = 0.41).
2C contacts with Ldand Kbare both mediated by similar
numbers of TCR residues (?23 total: ?11 in Va and ?12
in Vb), and 2C buries more surface area with its a chain
than its b chain in both complexes (482 A˚2versus 383 A˚2
in 2C/Ldand 525 A˚2versus 423 A˚2in 2C/Kb). In contrast,
TCR residues contacting the peptide are dominated by
the b chain for QL9 (4a versus 9b), compared to Kb-
dEV8 (6a and 6b) and Kbm3-dEV8 (7a and 8b). Although
the binding geometry differs from a previous model of
this complex (Speir et al., 1998), the predominance of
b chain contacts in the 2C/Ldinteraction is consistent.
When Ldand Kbare structurally superimposed, it is im-
mediately apparent that the 2C footprint on Ld-QL9 is
more ‘‘perpendicular’’ than on Kb, relative to the long
axis of the MHC groove (Figure 2A). 2C has also translated
laterally toward the QL9 C terminus by approximately 3 A˚,
such that the CDR3s are centered over the P5 position, as
compared to the CDR3s centered over the P4 position of
dEV8 in Kb. When the same residues on the CDR1 and 2
loops of 2C—in the different complexes—are compared,
the Vb domain is rotated counterclockwise in the allocom-
plex by approximately 15?, and the Va domain is rotated
approximately 30?(Figure 2A). When the 2C VaVb do-
mains in each complex are aligned on one another (rmsd
of 0.96 for Ca), it is clear that there is a large rotation of
the twodifferent pMHC (?20?) withrespect to one another
in the two complexes (Figure 2B). These rotational and
translational shifts result in alternative contacts between
2C/Ld-QL9 relative to the 2C/Kb-dEV8 interface.
For clarity, we schematically depict the relative interac-
ferent MHC by constructing two-dimensional contact
maps (Figures 2C and 2D). Neither CDR1a nor CDR2a of
2C have any interactions with the Lda1 helix, instead
showing extensive interactions with the Lda2 helix,
whereas the CDR1a loop forms numerous contacts with
both the a1 and a2 helices of Kb. Correspondingly,
CDR1b and CDR2b form contacts with the a2 and a1 heli-
ces in the Kbcomplex, respectively. When considered in
isolation, the 2C shape complementarity with the Ldheli-
ces is 0.72 versus 0.48 between 2C and the Kbhelices.
The QL9 peptide is bound in an Ldgroove, containing
ily ofMHChaplotypes(e.g., Ld,Lq,Dq,Db). Thishydropho-
bic ridge, formed by residues including Trp73 and Tyr99,
forces the main chain path of QL9 to arch upwards (Fig-
ures 3A, 3C, and S3). Thus, high-affinity peptides for Ld,
as well as other MHC in this structural subfamily, bulge
out and are therefore typically one extra residue in length
(9-mer) in order to traverse the groove (Balendiran et al.,
quire eight residues to span the length of the MHC groove
(Figures 3B and 3C). The P2-Leu and P9-Leu residues of
QL9 appear to form hydrophobic anchors in the core of
the groove, while the P4–P7 residues form a projecting
bulge that contacts the TCR (Figure 3C). While the main
chain indeed forms the bulge, the side chains of P5-Phe,
P6-Pro, and P7-Phe fold downward and project outwards
toward the Ldhelices, resulting in a hydrophobic, propel-
which primarily lies under the 2C Vb domain. This flat
Figure 1. Structure of the 2C TCR in
Complex with Ld-QL9
(A) Ribbon diagram of the 2C VaVb domains
complexed with the a1a2 domains of Ldand
the QL9 peptide looking down the helical
groove of Ld-QL9 (2C Va is pink, Vb cyan, Ld
green, and peptide yellow).
(B) The 2C VaVb domains are shown in com-
plex with H-2Kband the dEV8 peptide. Al-
though the full-length 2C ab heterodimer and
the full-length H-2Kbwere solved in that struc-
ture (PDBID: 2CKB),we show only the domains
corresponding to the 2C/Ld-QL9 complex con-
structs we used for the current structure.
Colors are the same as in (A) except that Kbis
(C and D) The ‘‘footprint’’ view showing the iso-
lated CDR loops of 2C as tubes over the sur-
face rendering of (C) Ld-QL9 and (D) Kb-dEV8.
The 2C contact surface on each peptide-
MHC is drawn in blue for the Vb footprint and
red for the Va footprint. Both peptides are yel-
ture figures (DeLano, 2002).
Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc. 137
surface allows the TCR CDR loops to access the pMHC
surface without significant structural accommodation of
the peptide bulge, even though the peptide main chain
does rise higher (?2 A˚) out of the groove than the main
chain of peptides in Kb.
2C forms a tightly packed interface with the QL9 pep-
tide, using the three CDRs in the b and a chains. 2C
uses 13 residues of the TCR (4 in the a chain and 9 in
the b chain) to contact 4 residues of the peptide (P4, P5,
P7, and P8), resulting in 115 interatomic 2C/QL9 contacts
(Figure 3D). The orthogonal orientation rotates the CDR2
loops counterclockwise to roughly twelve and six o’clock,
out ofrangeof themostdistal QL9residues incomparison
to the more clockwise-rotated 2C/dEV8 complex, where
CDR2 of both chains reside at approximately two and
eight o’clock and can contact the peptide termini (Fig-
ure 2A). The 2C/QL9 contacts are primarily main chain,
along with a minority of side-chain specific interactions.
The interactions are dominated by van der Waals (vdw)
contacts, hydrophobic in nature, with the planar surface
created by the P4–P7 bulge, offering little opportunity for
H bonding to the CDRs (Figure 4A). For instance, the poly-
glycine CDR3b (Gly95-Gly96-Gly97-Gly98) covers the
P5–P8 span of the peptide (Figures 3A, 3D and S3). Ala-
nine scanning of the QL9 peptide confirms that the P4–
P7 positions of QL9 are the most energetically significant
residues in the peptide for TCR contact (Figure 5A). The
P8-Asp side chain forms three hydrogen bonds, one
with Asn-31b and two with the main chains of CDR1b
and CDR3b, but replacement of P8-Asp with Ala is only
moderately deleterious to binding. This seems to argue
against a suggestion that electrostatic interactions be-
tween HV4b (Arg69) and P8-Asp of QL9 may account for
significant binding energy between the 2C TCR and
QL9/Ld(Speir et al., 1998). 2C CDR3a residues Ala101
and Ser102 make van der Waals contacts with QL9
Figure 2. 2C TCR Docking Orientation
and Interactions with Its Allogeneic and
(A) a1a2 domains of Ld-QL9 and Kb-dEV8 in
each 2C complex, shown as helical ribbons
with peptides, were superimposed (rmsd of
1.03 A˚) in order to view the relative positions
of the TCR CDR loops. Relative rotational shifts
same as Figure 1).
(B) In the reverse superposition to (A), the 2C
VaVb module was superimposed from each
complex (rmsd of 0.96 A˚for Ca and 1.29 A˚for
all atoms) in order to visualize the relative rota-
tions of the MHC helices.
(C and D) Contact maps between the 2C CDR
loops (center line) and the Ld(bottom line) and
Kb(top line) helices. Panel (C) is for 2C contact
with the a1 helix of both MHC, and (D) is for 2C
contact with the a2 helix of each MHC. Kb-spe-
cific contacts are green, whereas Ld-specific
contacts are brown. 2C residues that are pres-
ent in contacts to both Kband Ld(regardless of
the exact MHC residues) are highlighted in yel-
low, and 2C residues that contact the identical
residue in Kband Ldare highlighted in red
(cross reference with Figure 5C). Solid lines
Contacts were based on 4.5 A˚cutoff distances
and determined in CCP4 using automated cri-
138 Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc.
P5-Phe through the inner surface of the loop, but CDR3a
also folds away from the peptide to form interactions
with a hydrophobic, polymorphic region of the a1 helix
of Ld(Gln-Val) (Figure 4A). This observation is consistent
with previous studies showing the key influence of QL9
P5-Phe on 2C TCR binding (Schlueter et al., 1996).
This structure clearly illustrates how a single TCR can
crossreact with unique peptides. The QL9 and dEV8 pep-
interactions with the peptides are entirely different in both
structure and chemistry (Figures 3A, 3B, and 4A). For ex-
ample, the interaction between 2C and QL9 is hydropho-
through extended side-chain H bonds from P1, P2, P4,
P6, and P7 positions, some of which are water mediated
(Figure 4A). Second, the QL9 peptide is closer to the 2C
CDR3 loops dueto itshigher positionin the groove,result-
ing in a more complementary interface (Sc between QL9
alone and the TCR is 0.7 versus 0.5 for dEV8 alone with
2C). The dEV8 peptide barely contacts 2C through the
tips of long extended side chains, leaving large unfilled
gaps in the interface. The more intimate 2C/QL9 interface
is consistent with the higher affinity of the allogeneic
Figure 3. 2C TCR Interactions with the
(A) Closeup view of the 2C/QL9 interface, with
QL9 shown as yellow sticks. The CDRa and
CDRb loops of 2C are shown (a pink, b cyan)
as tubes with sticks and selected residues are
labeled.Hydrogen bondsare drawnasdashes.
(B) Same view as (A), except between 2C and
the dEV8 peptide.
(C) The isolated QL9 (green) and dEV8 (brown)
are shown from the MHC superposition to ap-
preciate the relatively different peptide struc-
tures seen by 2C.
(D) Contact map between 2C CDR3s and QL9
and dEV8 peptides, similar to contact map
shown in Figures 2C and 2D. 2C CDR3 resi-
dues that only contact dEV8 are shown in
brown, and QL9 contacts are in green. 2C res-
idues that contact both dEV8 and QL9 are
shown in yellow. Solid lines are vdw and
dashed lines are hydrogen bonds.
Figure 4. The 2C TCR Contacts Chemi-
cally Distinct Peptide-MHC Surfaces in
the Absence of Large-Scale Conforma-
tional Change in Its CDR Loops
(A) 2C contacts a primarily hydrophobic sur-
face of Ld-QL9 (left) versus a polar surface of
Kb-dEV8 (right). The surface of each pMHC
bound by 2C has been colored as polar (blue)
and apolar (yellow); a dashed box demarcates
the location of the CDR3-peptide interactions.
(B) Superimposed 2C TCR from the complex
with Ld(TCR a in red, b in blue) and from the
complex with Kb(TCR a in pink, b in cyan)
showing the relative conformations of the
Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc. 139
interaction versus the syngeneic complex (Garcia et al.,
1997; Sykulev et al., 1994). Third, from both the structure
and Ala scanning, 2C recognition of QL9 is centered on
a C-terminal ‘‘hotspot’’ of the peptide (P4–P7) that bulges
upward. By comparison, 2C contacts the flatter dEV8
through side chains spanning almost the entire length of
the peptide (P1–P7) (Figure 3B). The more perpendicular
orientation of 2C relative to the long axis of the Ldgroove
results in a shorter stretch of peptide being covered by the
TCR (Figure 2A), yet the closer QL9 peptide forms more
interatomic contacts (115) than the lower-lying dEV8 pep-
tide (58 contacts).
Previously, the structural plasticity of the TCR CDR3
loops has been noted in comparisons of bound versus
free TCR structures (including 2C), revealing both large-
and small-scale conformational changes upon binding
(Garcia etal., 1998; Reiser etal., 2002). Insupport of these
tropically disfavored binding between TCR and pMHC
(Willcox et al., 1999; Krogsgaard et al., 2003), which has
been interpreted as the thermodynamic signature of the
entropy penalty incurred during these conformational
changes. By extension, it was speculated that CDR3
loop flexibility could be a mechanism enabling a TCR to
crossreact with different peptide antigen sequences (Gar-
cia et al., 1998; Reiser et al., 2003). Thus far, however, this
concept has not been rigorously tested for recognition of
entirely unrelated sequences, as would be encountered
in some alloresponses in vivo. Surprisingly, we find that
the CDRs of 2C are in remarkably similar conformations
local adjustment in the position of the tip of CDR3a (?3 A˚),
and CDR3b remains essentially the same. The tip of
CDR1a has moved approximately 2.5 A˚in a rigid body
fashion, but the local conformation of the loop is relatively
unchanged compared to 2C/Kb(rmsd of 0.92 A˚for Ca
atoms). Given that the 2C binding site is utilizing similar
CDR loop conformations to contact two chemically and
structurally unique peptides, we investigated the thermo-
dynamics of 2C interaction with Ld-QL9 for comparison to
We performed isothermal titration calorimetry (ITC) on
2C with Ld-QL9 for comparison to previously measured
values for 2C/H-2Kb-dEV8 (Krogsgaard et al., 2003) (Fig-
ures 5B and 5C). Surprisingly, 2C binding to Ld-QL9 is en-
tropically favored, while 2C binding Kbwith three different
peptides is entropically disfavored and enthalpically
driven. The favorable entropy is in accord with the hydro-
phobic 2C/Ld-QL9 interface, which would likely derive
binding energy from expulsion of water from the apolar
surfaces (i.e., desolvation), especially the hydrophobic
surface of the peptide bulge. The favorable entropy, to-
gether with the similarity in CDR loop structure in the
two complexes, indicates that additional mechanisms ex-
ist to enable TCR crossreactivity other than CDR3 confor-
mational change. Our measurements are consistent with
those on shared cytokine receptors and crossreactive
natural killer receptors (Boulanger et al., 2003; McFarland
and Strong, 2003), which crossreact with different ligands
through enthalpy-entropy compensation of rigid binding
surfacesand withrecentstudieswithother TCR(Anikeeva
et al., 2003; Davis-Harrison et al., 2005; Ely et al., 2006).
The study by Ely et al. (2006) showed that a TCR could
recognize a pMHC with favorable entropy, even in the
presence of some conformational changes (Ely et al.,
2006). The broader conclusions of this analysis are that
Figure 5. Binding and Thermodynamic Analyses of 2C Bind-
ing to Its Syngeneic and Allogeneic Ligands
(A) Alanine scan of the QL9 peptide binding to 2C. 2C TCR tetramers
were used in titrations of the indicated QL9 alanine variants bound to
Ld,andthebinding relative toQL9/Ldwasdetermined. Effectsonbind-
ing could be either direct (as might be seen for those residues in direct
contact with the TCR) or indirect for those residues whose side chains
point toward Ld. Error bars represent the standard deviation for two
independent experiments (no calculated standard deviation is shown
for QL9 F7A, as no binding was detected in one of the experiments).
(B) Isothermal titration calorimetry curves of 2C titrated with Ld-QL9
(left) and m6 titrated with Ld-QL9 (right).
measured Kbligands (Degano et al., 2000; Krogsgaard et al., 2003)
and the currently measured allogeneic Ldligand. The ‘‘sc’’ designation
refers to the single-chain Fv constructs used for the 2C/Ldmeasure-
140 Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc.
enthalpy-entropy compensation appears to be a potent
mechanism to enable receptor crossreactivity in the ab-
sence of major conformational changes.
Given the unique binding mode of 2C on Ld-QL9 com-
pared to Kb-dEV8, we wanted to examine the role of
shared versus unique MHC residues in dictating the TCR
binding orientation. A priori, as Ldand Kbshare 80% se-
quence identity in the helices presented to TCR, 2C could
either (1) recognize shared residues of Ldand Kbresulting
in nearly identical footprints or (2) recognize different res-
idues, resulting in an altered footprint. First we examined
the possibility that 2C recognition is determined by the
similarities in the two MHC helical surfaces. Although the
in Ld(Figure 6A), we found that only six (Gln65, Gln72,
Val76, Arg79, Ala150, Ala158) are used as 2C contacts
in both complexes, based on a 4.5 A˚residue-residue dis-
tance criteria to define ‘‘contacts’’ (Figure 6B). From this
group and using the same distance criteria, only four res-
idues (Gln65, Val76, Arg79, and Ala158) also contacted
the same 2C residues (Figure 6C). We emphasize that,
due to the different accuracies of the 2C/Kb-dEV8 com-
plexes, determined at different resolutions, the relative
contact analyses vary slightly. For instance, if our contact
than 4 A˚(instead of 4.5 A˚), there are only two common
MHC residues (Val76 and Arg79) contacting the same
TCR residues, and at 4.7 A˚, there are five common MHC
contact residues. Given that ‘‘contacts’’ are assigned
purely based on a distance criteria, residues could be
within a similar distance but have entirely different contact
geometry, chemistry, and stereochemistry. Accordingly,
inspection of these apparently conserved TCR-MHC con-
tacts reveals the nature of the interactions, and the atoms
utilized are entirely distinct in the Ldversus the three Kb
complexes, due to the different TCR/MHC orientation an-
gles (Figure 6D). Thus, the few conserved interactions
found in the allogeneic and syngeneic complexes occur
via largely unique binding chemistries.
A recent complex of a TCR with a ‘‘super-bulged’’ pep-
the TCR and MHC helices, suggesting this may constitute
(Tynan et al., 2005). In the case of the 2C complexes with
Figure 6. 2C TCR Forms Limited Con-
tacts with Shared Amino Acids on Ld
and KbRecognition Surfaces
(A) Molecular surfaces of Ld-QL9 (left) and
Kb-dEV8 (right). Shared residues on the MHC
helices are drawn in red.
(B) Only a small subset of the total shared res-
idues shown in (A) are also used as direct con-
tacts with 2C in both complexes.
(C) Of the shared MHC contacts shown in (B),
only four also contact the same 2C residue in
(D) The common TCR-MHC contacts in (C) are
structurally and chemically distinct interac-
tions. Shown are the allogeneic and syngeneic
complexes superimposed on the MHC. Three
common CDR2b and CDR3a contacts (high-
lighted in C) are significantly perturbed in
each complex such that the bond distances
and geometries of the interatomic contacts
are different, mainly due to the rotational shift
of the CDRs along the MHC helix.
Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc. 141
Kband Ld, the distinct interaction chemistries of a roughly
similar number of shared TCR-MHC contacts as seen in
Tynan et al. (2005) shows that a small subset of MHC res-
idues can be recognized by a single TCR in unique ways
when present on different MHC. It would appear, then,
that an amino acid ‘‘code’’ underlying MHC bias could
be significantly blurred by the possibility of multifarious
We also analyzed the polymorphic residues presented
to the TCR on the surface of the respective syn- and
allo-MHC helices. Strikingly, there are only four different
residues on the surface of Ldand Kbhelices presented
to the TCR (66, 145, 155, and 163). The majority of poly-
morphic residues are in the peptide binding groove,
away from the TCR/pMHC interface. Two of the polymor-
both structures, but 163 is contacted by different CDR
loops in the two complexes, while 155 is contacted by
the same TCR residues (Tyr31a and Tyr50a) in the two
structures (Figures 2C and 2D). Thus, Tyr31a and Tyr50a
are among the most energetically important 2C residues
in binding the Ldand Kbcomplexes (Lee et al., 2000; Man-
ning et al., 1998), yet these contributions are associated
withcontactstoapolymorphicresidue. Inconclusion, sig-
nificant sequence conservation between Kband Ldmight
have suggested that the most parsimonious recognition
strategy by 2C for Ldwould have been a Kb-like footprint.
However the interatomic contacts are almost entirely dif-
ferent based on analysis of pair-wise amino acid contacts.
Tail Wagging the Dog, or Dog Wagging the Tail?
The structural results and analysis so far could implicate
a ‘‘peptide-centric’’ model of alloreactivity, whereby rec-
ognition of the QL9 peptide has steered the TCR into an
alternative footprint on the MHC. In this model, the spe-
cific TCR-peptide interactions would ‘‘edit’’ the overall
TCR docking footprint, as was suggested from analysis
of human autoimmune and antiviral TCR/pMHC com-
plexes (Borg et al., 2005; Hahn et al., 2005). However,
this interpretation of our results is contradicted by ener-
getic mapping (Ala scanning) of 2C residues in the 2C/
Ld-peptide versus 2C/Kb-peptide interactions, which indi-
cate that most energetically significant binding determi-
nants are derived from CDR1 and CDR2, which mainly
contact the helices in both the allo- and syngeneic com-
plexes (Lee et al., 2000; Manning et al., 1998). Conceptu-
ally, the peptide-centric model can be thought of as a ‘‘tail
wagging the dog’’ in that the TCR-peptide (tail) contacts
seem to determine the TCR CDR1 and CDR2 footprint
on the MHC helices (the dog). This model stands in con-
trast to the MHC-centric models (dog wagging the tail)
whereby shared structural determinants on MHC, recog-
nized by TCR CDR1 and CDR2, determine the ultimate
is wagging the tail. We have previously evolved, using
yeast display, a high-affinity variant of 2C through muta-
tion of CDR3a (Holler et al., 2000), which contacts both
QL9 and Ldin the structure. The evolution experiment is
‘‘unbiased’’ in that recombinant TCR and pMHC were
used in vitro, in the absence of T cell coreceptors (e.g.,
CD8/CD4) that could influence the binding orientation,
as has been proposed (Buslepp et al., 2003). Thus, the
TCR has, in principle, freedom to assume any docking
footprint on Ld-QL9 as a result of the remodeled CDR3-
peptide interactions. By producing a new CDR3a se-
quence that will form unique interactions with the peptide
MHC, we can ask whether the CDR1 and CDR2 contacts
with MHC have been edited in deference to perturbed
CDR3-peptide interactions. In other words, is the tail wag-
ging the dog?
The in vitro evolution of 2C into a high-affinity variant,
termed m6, which retains specificity for Ld-QL9, has
been described previously (Holler et al., 2000). M6 has
a CDR3a sequence of SHQGRYL (compared to wild-
type SGFASAL) and recognizes Ld-QL9 with high affinity
as measured both by surface plasmon resonance (Jones
et al., 2006) and ITC (Figures 5B and 5C). Interestingly,
m6 has incurred an entropic penalty for binding H-2Ld-
QL9, so the binding chemistry has changed in both amino
acid sequence and thermodynamics. We crystallized m6
in complex with Ld-QL9 and determined the structure to
2.5 A˚resolution (Table S1). The overall binding orientation
of the TCR on Ld-QL9 is identical to 2C/Ld-QL9 (the entire
complexes superimpose with an rmsd of 0.79 A˚for all
structural difference is the position of the CDR3a loop,
Figure 7. The High-Affinity 2C Variant M6
Has an Identical Footprint to Wild-Type
(A) The high affinity and wild-type complexes
were superimposed on Ld-QL9 (rmsd of 0.79
A˚for all atoms), showing that the resultant
CDR loop positions are identical, except for
(B) M6 CDR3a undergoes a local movement,
relative to wild-type 2C, such that its tip folds
inward toward the peptide.
142 Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc.
whose tip is now displaced 4.5 A˚inward toward the pep-
tide compared to the wild-type 2C complex (Figure 7B).
This inward displacement results in a decrease in interac-
tions with the Lda1 helix, instead focusing its contacts on
QL9residues P4 and F5. The mode of interaction is hydro-
phobic, whereby aliphatic groups of CDR3a residues
Gln100, Gly101, and Arg102 form van der Waals contacts
with the planar surface previously described for the
bulged peptide residues. The location and conformation
of CDR3b is identical in the two structures. Thus, despite
focusing completely new, higher-affinity binding energet-
ics on the peptide by directed evolution of CDR3a, the
wild-type binding orientation persists. Our principal con-
clusion is that although the TCR footprint on Ld-QL9 is
unique relative to Kb-dEV8, it is not simply a result of edit-
ing by QL9 peptide/CDR3 interactions. 2C recognition of
Ld-QL9 occurs through a discrete binding solution, which
persists in the face of perturbed peptide-CDR3 contacts.
In short, for the case of 2C alloreactivity, the dog appears
to be wagging the tail.
We show that one receptor, 2C, has at least two possible
self- and foreign-peptide-MHC ligands, respectively,
which present both similar and distinct surfaces for the
TCR to recognize. In addition, binding to self and foreign
ligands was achieved through different chemistries and
thermodynamic mechanisms. The starkly contrasting
mechanisms by which 2C recognizes two distinct, natu-
rally occurring ligands is broadly informative about multi-
specific protein interactions. Even when different ligands
present similar constellations of receptor binding residues
(i.e., possess significant homology), the receptor need not
utilize a similar recognition strategy to crossreact. For the
case of an adaptive immune receptor such as the TCR,
which we presumed would ‘‘home’’ to conserved, shared
structural determinants on the MHC ligands with which it
has coevolved, the majority of the recognition strategy
focused on differences between the ligands. Thus, other
mechanisms than molecular mimicry can explain cross-
activity characteristic of autoimmune diseases need not
nig and Strominger, 1995). The generality of the 2C mech-
anism of allorecognition, to other TCR systems, awaits
further structure function studies of TCRs whose cognate
and allo-pMHC ligands are known (Archbold et al., 2006).
From an immunological standpoint, we conclude that
2C alloreactivity can be considered a ‘‘synthesis’’ of the
peptide- and MHC-centric hypotheses. We suggest that
the germline-encoded contacts between TCR and MHC,
which have been selected through coevolution (Huseby
et al., 2004; Jerne, 1971; Turner et al., 2006; Zerrahn
et al., 1997), play the dominant role in dictating the overall
binding geometry. However, different allo- and self-pep-
tides presented by the MHC may influence the TCR to se-
lect one preferred orientation, out of a defined set of pos-
sible germline-encoded footprints, that is compatible with
making contacts with a specific peptide to produce
a physiologically meaningful affinity (KD ?1–50 mM).
Thus, we predict that each peptide does not recruit
a TCR to form ‘‘de novo’’ footprints with the MHC helices
that are not among the germline-encoded set. In support
of this, a previous study showed that different TCRs that
both contained Vb8.2 formed identical docking contacts
on the I-Auand I-AkMHC helices, despite different pep-
tides and CDR3 sequences (Maynard et al., 2005).
An important question iswhether the optimal TCR/MHC
docking orientation weseeinthe 2C/Ldcomplex wouldbe
seen on self MHC (Kb) if Kbwere presenting a foreign pep-
tide. The SIYR peptide is a foreign peptide that reacts with
criteria of a peripherally selected cognate ligand. The
structure of 2C bound to H-2Kb-SIYR has been reported
(Degano et al., 2000) and is nearly identical to those of
other 2C/Kbcomplexes (Figure S1). Thus, 2C does not
recognize the strong agonist peptide-MHC Kb-SIYR,
with a significantly different strategy than other 2C/Kb-
peptide complexes. The strong agonist activity of SIYR/
Kbis accomplished despite a relatively low affinity
(KD?30 mM) for the 2C TCR, especially compared to
QL9/Ld. However, we have shown previously that the rel-
ative agonist activities of the various peptides (dEV8,
of the peptide-MHC complexes (Holler and Kranz, 2003;
Krogsgaard et al., 2003).
Does this docking orientation extend to other TCRs that
are involved in Ldrecognition? A previous Ldmutational
analysis supports this possibility, in that different T cells
appear to contact a conserved set of helix residues on
Ld, implying that this orientation may be generalized (Hor-
nell et al., 1999). Furthermore, the Vb8 region appears to
be used preferentially with a number of Ld-restricted re-
sponses and the p2Ca/QL9-Ldligands in particular (Con-
nolly, 1994; Rodewald et al., 1989; Tjoa and Kranz, 1994).
The present study shows that alloreactions are not only
of less predictable interactions with alloantigens that have
more significant structural diversity than the syngeneic
MHC. While the 2C/Ld-QL9 interaction is CD8 indepen-
tation remains an important consideration (Buslepp et al.,
2003). Recent findings that 2C and other class I-restricted
alternative docking orientations (Ge et al., 2006), presum-
ably dependent on yet other chemistries, for their produc-
Protein Expression and Purification
The2C TCR,as wellasthehigh-affinity2C mutant,m6,wasexpressed
as a soluble, single-chain construct by E. coli periplasmic expression
Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc. 143
the pET22b vector as Va-(Gly4Ser)4linker-Vb single chains with a C-
terminal His6purification tag. Mutations in Vb8.2 were included to in-
creasesolubility (G17E, H47Y,I75T,and L81S).Proteinwasexpressed
in BL21 (DE3) Codon Plus E. coli (Stratagene) and purified by Nickel
NTA affinity chromatography and gel filtration as described (Jones
et al., 2006).
was expressed by refolding from E. coli inclusion bodies as described
(Jones et al., 2006). Inclusion bodies were expressed in BL21 (DE3)
Codon Plus E. coli (Stratagene) and refolded in the presence of excess
QL9, and the complex was purified by gel filtration. The platform Ld
used for crystallization with 2C contained three additional hydropho-
bic-to-polar mutations (F9Y, V12T, and I23T) in a surface patch on
the underside of the b sheet, which is normally in contact with b2M
and a3 domains. These polar surface substitutions increased its solu-
bility in aqueous buffers but are far removed from the TCR binding in-
terface and had no effect on its 2C binding properties.
Crystallization, Data Collection, and Processing
Purified TCR and pMHC were combined in equimolar amounts and
concentrated in a YM10 Centriprep (Amicon) to approximately
50 mg/ml for each of the complexes (2C/Ld-QL9 and m6/Ld-QL9).
Crystals were obtained in 0.9 M sodium dihydrogen phosphate and
0.1 M dipotassium hydrogen phosphate (2C/Ld-QL9) or 0.2 M ammo-
nium phosphate, 20% polyethyleneglycol (PEG) 3350, and 0.1 M Tris
at pH 8.5 (m6/Ld-QL9) using the sitting drop vapor diffusion method.
Crystals were cryoprotected using 25% glycerol (2C/Ld-QL9) or 30%
PEG 3350 (m6/Ld-QL9) before cooling to 100 K. Data were collected
on beamline 11.1 at the Stanford Synchrotron Light Source (SSRL,
Stanford, CA). Crystals of 2C/Ld-QL9 diffracted to 2.35 A˚, cell dimen-
sions a = 163.2 A˚, b = 163.2 A˚, c= 95.0 A˚, in space group P6522. Crys-
tals of m6/Ld-QL9 diffracted to 2.5 A˚, cell dimensions a = 113.5 A˚, b =
113.5 A˚, c = 177.5 A˚, space group P42212. The data were indexed, in-
tegrated, and scaled with HKL2000 (Otwinowski and Minor, 1997).
Structure Solution and Refinement
Complex structures were determined by molecular replacement with
the program Phaser (Read, 2001). Protein Data Bank (PDB) files
1LDP(forLd)and2CKB (fortheTCR) wereused assearchmodels after
trimming coordinates to correspond to the smaller molecules used to
crystallize the 2C/Ld-QL9 complex, searching sequentially with the
TCR a chain, TCR b chain, and MHC. One round of rigid body refine-
ment was performed, followed by rounds of model building using
COOT (Emsley and Cowtan, 2004), simulated annealing, and posi-
tional and individual B factor refinements using the CNS software
package (Brunger et al., 1998), resulting in R and free R values of
22.0% and 22.6%, respectively.
Individual components of the completed 2C/Ld-QL9 model were
used as search models for the m6/Ld-QL9 molecular replacement. Af-
ter refinement by similar methodologies, the resultant R and free R
values were 22.4% and 24.6%, respectively. A summary of refinement
statistics is given in Table S1.
Isothermal Titration Calorimetry
Protein expression and purification for isothermal titration calorimetry
were similar to that for crystallography, except that TCR samples were
treated overnight with carboxypeptidase A and B at 4?C prior to size-
exclusion chromatography. All proteins were purified in the same
buffer (10 mM HEPES, 150 mM NaCl, pH 7.2) to minimize heat of dilu-
tion effects. Thermodynamic parameters were determined using a
VP-ITC calorimeter (MicroCal, Northhampton, MA) at 20?C. Protein
samples were degassed prior to titrations. Data were processed using
concentrations. Conditions giving the strongest signal with the lowest
background were Ld-QL9 (185 mM) injected into 2C (45 mM) and
Ld-QL9 (52 mM) injected into m6 (5 mM). Experiments at other concen-
trations yielded similar thermodynamic parameters (data not shown).
2C scTCR Binding to QL9 Variants
Relative binding affinitiesof the 2C scTCR for Ldbound toQL9 variants
were determined using 2C scTCR tetramers and peptide-loaded Ldon
target cells, as has recently been described (Huseby et al., 2006). Sin-
gle-site alanine variants of QL9 (100 mM) were incubated with cell line
T2-Ldfor 3 hr at 37?C. Peptide-loaded cells were incubated with vari-
ous concentrations of tetrameric 2C scTCR (T7) for 40 min at 4?C.
Streptavidin-PE (BD Pharmingen) labeled TCR was detected using
a Coulter Epics XL flow cytometer. Kdvalues of QL9 variants relative
to QL9 wt were determined by nonlinear regression of the equilibrium
binding curves, using a value (50 mean fluorescent units) in the linear
range of the curves. Values were corrected for total Ldlevels using
the anti-LdAb 30-5-7, and background was subtracted for 2C scTCR
binding to the null ligand MCMV/Ld. Error bars represent the standard
deviation (SD) for two independent experiments (no calculated SD is
shown for QL9 F7A, as no detectable binding was detected in one of
Supplemental Data include three figures and three tables and can be
found with this article online at http://www.cell.com/cgi/content/full/
We thank Sean Juo for assistance with structure determination; Phil
Holler and Susan Brophy for generation of the m6 TCR and the m31
Ldmutants, respectively; and Herman Eisen for his advice and previ-
and resources of the SSRL. L.C. is supported by an National Science
Foundation predoctoral fellowship. This work was supported by NIH
grants AI 48540 (K.C.G.) and GM55767 (D.M.K.). K.C.G. is also sup-
ported by the Keck Foundation and Howard Hughes Medical Institute.
Received: November 23, 2006
Revised: January 1, 2007
Accepted: January 19, 2007
Published: April 5, 2007
Anikeeva, N., Lebedeva, T., Krogsgaard, M., Tetin, S.Y., Martinez-
Hackert, E., Kalams, S.A., Davis, M.M., and Sykulev, Y. (2003). Distinct
molecular mechanisms account for the specificity of two different
T-cell receptors. Biochemistry 42, 4709–4716.
Archbold, J.K., Macdonald, W.A., Miles, J.J., Brennan, R.M., Kjer-
Nielsen, L., McCluskey, J., Burrows, S.R., and Rossjohn, J. (2006).
Alloreactivitybetweendisparate cognateand allogeneicpMHC-Icom-
plexes is resultant of highly focused, peptide-dependent structural
mimicry. J. Biol. Chem. 281, 34324–34332.
Nathenson, S.G., and Sacchettini, J.C. (1997). The three-dimensional
structure of an H-2Ld-peptide complex explains the unique interaction
of Ld with beta-2 microglobulin and peptide. Proc. Natl. Acad. Sci.
U.S.A. 94, 6880–6885.
G.K., Solheim, J.C.,Young,A.C., Hansen, T.H.,
Bevan, M.J. (1984). High determinant density may explain the phe-
nomenon of alloreactivity. Immunol. Today 5, 128.
Borg, N.A., Ely, L.K., Beddoe, T., Macdonald, W.A., Reid, H.H.,
Clements, C.S., Purcell, A.W., Kjer-Nielsen, L., Miles, J.J., Burrows,
S.R., et al. (2005). The CDR3 regions of an immunodominant T cell re-
ceptor dictate the ‘energetic landscape’ of peptide-MHC recognition.
Nat. Immunol. 6, 171–180.
144 Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc.
Boulanger, M.J., Bankovich, A.J., Kortemme, T., Baker, D., and
Garcia, K.C. (2003). Convergent mechanisms for recognition of diver-
gent cytokines by the shared signaling receptor gp130. Mol. Cell 12,
Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P.,
Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu,
N.S., et al. (1998). Crystallography & NMR system: A new software
suite for macromolecular structure determination. Acta Crystallogr. D
Biol. Crystallogr. 54, 905–921.
Buslepp, J., Wang, H., Biddison, W.E., Appella, E., and Collins, E.J.
(2003). A correlation between TCR Valpha docking on MHC and CD8
dependence: implications for T cell selection. Immunity 19, 595–606.
Chen, J., Eisen, H.N., and Kranz, D.M. (2003). A model T-cell receptor
system for studying memory T-cell development. Microbes Infect. 5,
Connolly, J.M. (1994). The peptide p2Ca is immunodominant in allore-
cognition of Ldby b chain variable region Vb8+but not Vb8?strains.
Proc. Natl. Acad. Sci. USA 91, 11482–11486.
Daniel, C.,Horvath, S.,and Allen, P.M. (1998). Abasis foralloreactivity:
MHC helical residues broaden peptide recognition by the TCR. Immu-
nity 8, 543–552.
Davis, M.M., and Bjorkman, P.J. (1988). T-cell antigen receptor genes
and T-cell recognition. Nature 334, 395–402.
Davis-Harrison, R.L., Armstrong, K.M., and Baker, B.M. (2005). Two
different T cell receptors use different thermodynamic strategies to
recognize the same peptide/MHC ligand. J. Mol. Biol. 346, 533–550.
Degano, M., Garcia, K.C., Apostolopoulos, V., Rudolph, M.G., Teyton,
L., and Wilson, I.A. (2000). A functional hot spot for antigen recognition
in a superagonist TCR/MHC complex. Immunity 12, 251–261.
DeLano, W.L. (2002). The PyMOL Molecular Graphics System (San
Carlos, CA, USA: DeLano Scientific).
Ely, L.K., Beddoe, T., Clements, C.S., Matthews, J.M., Purcell, A.W.,
Kjer-Nielsen, L., McCluskey, J., and Rossjohn, J. (2006). Disparate
thermodynamics governing T cell receptor-MHC-I interactions impli-
cate extrinsic factors in guiding MHC restriction. Proc. Natl. Acad.
Sci. USA 103, 6641–6646.
Emsley, P., and Cowtan, K. (2004). Coot: model-building tools for mo-
lecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126–2132.
Garcia, K.C., and Adams, E.J. (2005). How the T cell receptor sees
antigen–a structural view. Cell 122, 333–336.
Garcia, K.C., Degano, M., Pease, L.R., Huang, M., Peterson, P.,
Teyton, L., and Wilson, I.A. (1998). Structural basis of plasticity in T
cell receptor recognition of a self peptide-MHC antigen. Science
Garcia, K.C., Tallquist, M.D., Pease, L.R., Brunmark, A., Scott, C.A.,
Degano, M., Stura, E.A., Peterson, P.A., Wilson, I.A., and Teyton, L.
(1997). Alphabeta T cell receptor interactions with syngeneic and allo-
geneic ligands: affinity measurements and crystallization. Proc. Natl.
Acad. Sci. USA 94, 13838–13843.
Ge, Q., Holler, P.D., Mahajan, V.S., Nuygen, T., Eisen, H.N., and Chen,
J. (2006). Development of CD4+ T cells expressing a nominally MHC
class I-restricted T cell receptor by two different mechanisms. Proc.
Natl. Acad. Sci. USA 103, 1822–1827.
Unconventional topology of self peptide-major histocompatibility
complex binding by a human autoimmune T cell receptor. Nat. Immu-
nol. 6, 490–496.
Hansen, T., Balendiran, G., Solheim, J., Ostrov, D., and Nathenson, S.
(2000). Structural features of MHC class I molecules that might facili-
tate alternative pathways of presentation. Immunol. Today 21, 83–88.
Holler, P.D., and Kranz, D.M. (2003). Quantitative analysis of the
contribution of TCR/pepMHC affinity and CD8 to T cell activation.
Immunity 18, 255–264.
Holler, P.D., Holman, P.O., Shusta, E.V., O’Herrin, S., Wittrup, K.D.,
and Kranz, D.M. (2000). In vitro evolution of a T cell receptor with
high affinity for peptide/MHC. Proc. Natl. Acad. Sci. USA 97, 5387–
Hornell, T.M., Solheim, J.C., Myers, N.B., Gillanders,W.E.,Balendiran,
G.K., Hansen, T.H., and Connolly, J.M. (1999). Alloreactive and synge-
neic CTL are comparably dependent on interaction with MHC class I
alpha-helical residues. J. Immunol. 163, 3217–3225.
Housset, D., and Malissen, B. (2003). What do TCR-pMHC crystal
structures teach us about MHC restriction and alloreactivity? Trends
Immunol. 24, 429–437.
Huseby, E., Kappler, J., and Marrack, P. (2004). TCR-MHC/peptide in-
Huseby, E.S., Crawford, F., White, J., Marrack, P., and Kappler, J.W.
(2006). Interface-disrupting amino acids establish specificity between
T cell receptors and complexes of major histocompatibility complex
and peptide. Nat. Immunol. 7, 1191–1199.
J. Immunol. 1, 1–9.
Jones, L.L., Brophy, S.E., Bankovich, A.J., Colf, L.A., Hanick, N.A.,
Garcia, K.C., and Kranz, D.M.(2006). Engineering and characterization
of a stabilized alpha1/alpha2 module of the class I major histocompat-
ibility complex product Ld. J. Biol. Chem. 281, 25734–25744.
Kieke, M.C., Shusta, E.V., Boder, E.T., Teyton, L., Wittrup, K.D., and
Kranz, D.M. (1999). Selection of functional T cell receptor mutants
from a yeast surface- display library. Proc. Natl. Acad. Sci. USA 96,
Kranz, D.M., Sherman, D.H., Sitkovsky, M.V., Pasternack, M.S., and
Eisen, H.N. (1984). Immunoprecipitation of cell surface structure of
cloned cytotoxic T lymphocytes by clone-specific antisera. Proc.
Natl. Acad. Sci. USA 81, 573–577.
Krogsgaard, M., Prado, N., Adams, E.J., He, X.L., Chow, D.C., Wilson,
D.B., Garcia, K.C., and Davis, M.M. (2003). Evidence that structural re-
arrangements and/or flexibility during TCR binding can contribute to T
cell activation. Mol. Cell 12, 1367–1378.
Lee, P.U., Churchill, H.R.,Daniels,M.,Jameson, S.C., and Kranz, D.M.
(2000). Role of 2C T cell receptor residues in the binding of self- and
allo- major histocompatibility complexes. J. Exp. Med. 191, 1355–
Luz, J.G.,Huang,M.,Garcia,K.C.,Rudolph,M.G.,Apostolopoulos, V.,
Teyton, L., and Wilson, I.A. (2002). Structural comparison of allogeneic
and syngeneic T cell receptor-peptide-major histocompatibility com-
plex complexes: a buried alloreactive mutation subtly alters peptide
presentation substantially increasing V(beta) Interactions. J. Exp.
Med. 195, 1175–1186.
Garcia, K.C., Teyton, L., Wilson, I.A., and Kranz, D.M. (1998). Alanine
scanning mutagenesis of an alphabeta T cell receptor: mapping the
energy of antigen recognition. Immunity 8, 413–425.
Maynard, J., Petersson, K., Wilson, D.H., Adams, E.J., Blondelle, S.E.,
Boulanger, M.J., Wilson, D.B., and Garcia, K.C. (2005). Structure of an
autoimmune T cell receptor complexed with class II peptide-MHC: in-
sights into MHC bias and antigen specificity. Immunity 22, 81–92.
McFarland, B.J., and Strong, R.K. (2003). Thermodynamic analysis of
degenerate recognition by the NKG2D immunoreceptor: not induced
fit but rigid adaptation. Immunity 19, 803–812.
Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction
data collected in oscillation mode. Methods Enzymol 276, 307–326.
Read, R.J. (2001). Pushing the boundaries of molecular replacement
with maximum likelihood. Acta Crystallogr. D Biol. Crystallogr. 57,
Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc. 145
Reiser, J.B., Darnault, C., Guimezanes, A., Gregoire, C., Mosser, T.,
Schmitt-Verhulst, A.M., Fontecilla-Camps, J.C., Malissen, B., Hous-
set, D., and Mazza, G. (2000). Crystal structure of a T cell receptor
bound to an allogeneic MHC molecule. Nat. Immunol. 1, 291–297.
Reiser, J.B., Gregoire, C., Darnault, C., Mosser, T., Guimezanes, A.,
Schmitt-Verhulst, A.M., Fontecilla-Camps, J.C., Mazza, G., Malissen,
conformational changes of unprecedented magnitude upon binding to
a peptide/MHC class I complex. Immunity 16, 345–354.
Reiser, J.B., Darnault, C., Gregoire, C., Mosser, T., Mazza, G.,
Kearney, A., van der Merwe, P.A., Fontecilla-Camps, J.C., Housset,
D., and Malissen, B. (2003). CDR3 loop flexibility contributes to the de-
generacy of TCR recognition. Nat. Immunol. 4, 241–247.
Rodewald, H.R., Koszinowski, U.H., Eichmann, K., and Melchers, I.
(1989). Predominant utilization of V beta 8+ T cell receptor genes in
the H-2Ld- restricted cytotoxic T cell response against the immedi-
ate-early protein pp89 of the murine cytomegalovirus. J. Immunol.
Rudolph, M.G., Stanfield, R.L., and Wilson, I.A. (2006). How TCRs bind
MHCs, peptides, and coreceptors. Annu. Rev. Immunol. 24, 419–466.
Santori, F.R., Kieper, W.C., Brown, S.M., Lu, Y., Neubert, T.A.,
Johnson, K.L., Naylor, S., Vukmanovic, S., Hogquist, K.A., and
Jameson, S.C. (2002). Rare, structurally homologous self-peptides
promote thymocyte positive selection. Immunity 17, 131–142.
Schlueter, C.J., Manning, T.C., Schodin, B.A., and Kranz, D.M. (1996).
A residue in the center of peptide QL9 affects binding to both Ldand
the T cell receptor. J. Immunol. 157, 4478–4485.
Sha, W.C., Nelson, C.A., Newberry, R.D., Kranz, D.M., Russell, J.H.,
and Loh, D.Y. (1988). Positive and negative selection of an antigen re-
ceptor on T cells in transgenic mice. Nature 336, 73–76.
Sherman, L.A., and Chattopadhyay, S. (1993). The molecular basis of
allorecognition. Annu. Rev. Immunol. 11, 385–402.
Shusta, E.V., Holler, P.D., Kieke, M.C., Kranz, D.M., and Wittrup, K.D.
(2000). Directed evolution of a stable scaffold for T-cell receptor engi-
neering. Nat. Biotechnol. 18, 754–759.
Speir, J.A., Garcia, K.C., Brunmark, A., Degano, M., Peterson, P.A.,
Teyton, L., and Wilson, I.A. (1998). Structural basis of 2C TCR allore-
cognition of H-2Ld peptide complexes. Immunity 8, 553–562.
Starr, T.K., Jameson, S.C., and Hogquist, K.A. (2003). Positive and
negative selection of T cells. Annu. Rev. Immunol. 21, 139–176.
Sykulev,Y.,Brunmark, A.,Tsomides,T.J.,Kageyama,S.,Jackson, M.,
Peterson, P.A., and Eisen, H.N. (1994). High-affinity reactions between
antigen-specific T-cell receptors and peptides associated with alloge-
neic and syngeneic major histocompatibility complex class I proteins.
Proc. Natl. Acad. Sci. USA 91, 11487–11491.
Tallquist, M.D., Weaver, A.J., and Pease, L.R. (1998). Degenerate rec-
ognition of alloantigenic peptides on a positive- selecting class I mol-
ecule. J. Immunol. 160, 802–809.
Tjoa, B.A., and Kranz, D.M. (1994). Sequence restrictions in T cell re-
ceptor b-chains that have specificity for a self-peptide/Ldcomplex.
Mol. Immunol. 31, 705–711.
Turner, S.J., Doherty, P.C., McCluskey, J., and Rossjohn, J. (2006).
Structural determinants of T-cell receptor bias in immunity. Nat. Rev.
Immunol. 6, 883–894.
Tynan, F.E., Burrows, S.R., Buckle, A.M., Clements, C.S., Borg, N.A.,
Miles, J.J., Beddoe, T., Whisstock, J.C., Wilce, M.C., Silins, S.L., et al.
(2005). T cell receptor recognition of a ‘super-bulged’ major histocom-
patibility complex class I-bound peptide. Nat. Immunol. 6, 1114–1122.
Udaka, K., Tsomides, T.J., and Eisen, H.N. (1992). A naturally occur-
ring peptide recognized by alloreactive CD8+ cytotoxic T lymphocytes
in association with a class I MHC protein. Cell 69, 989–998.
Udaka, K., Wiesmuller, K., Kienle, S., Jung, G., and Walden, P. (1996).
Self-MHC-restricted peptides recognized by an alloreactive T-lym-
phocyte clone. J. Immunol. 157, 670–678.
Willcox, B.E., Gao, G.F., Wyer, J.R., Ladbury, J.E., Bell, J.I., Jakobsen,
B.K., and Anton van der Merwe, P.A. (1999). TCR binding to peptide-
MHC stabilizes a flexible recognition interface. Immunity 10, 357–365.
Wucherpfennig, K.W., and Strominger, J.L. (1995). Molecular mimicry
in T cell-mediated autoimmunity: viral peptides activate human T cell
clones specific for myelin basic protein. Cell 80, 695–705.
Zerrahn, J., Held, W., and Raulet, D.H. (1997). The MHC reactivity of
the T cell repertoire prior to positive and negative selection. Cell 88,
The 2C/Ld-QL9 and m6/Ld-QL9 coordinates have been deposited in
the PDB with accession numbers 20I9 and 2E7L, respectively.
146 Cell 129, 135–146, April 6, 2007 ª2007 Elsevier Inc.