Specificity and detection of insulin-reactive CD4+
T cells in type 1 diabetes in the nonobese diabetic
Frances Crawforda,b, Brian Stadinskia,b,1, Niyun Jina,b, Aaron Michelsc, Maki Nakayamab,c, Philip Prattc,
Philippa Marracka,b,d, George Eisenbarthb,c, and John W. Kapplera,b,e,2
aHoward Hughes Medical Institute, National Jewish Health, Denver, CO 80206;bIntegrated Department of Immunology, National Jewish Health and
University of Colorado School of Medicine, Denver, CO 80206;cBarbara Davis Center for Childhood Diabetes, Aurora, CO 80045; anddDepartment of
Biochemistry and Molecular Genetics andeProgram in Structural Biology and Biophysics, University of Colorado School of Medicine, Aurora, CO 80045
Contributed by John W. Kappler, August 24, 2011 (sent for review July 20, 2011)
In the nonobese diabetic (NOD) mouse model of type 1 diabetes
(T1D), an insulin peptide (B:9–23) is a major target for pathogenic
CD4+T cells. However, there is no consensus on the relative im-
portance of the various positions or “registers” this peptide can
take when bound in the groove of the NOD MHCII molecule, IAg7.
This has hindered structural studies and the tracking of the rele-
vant T cells in vivo with fluorescent peptide-MHCII tetramers.
Using mutated B:9–23 peptides and methods for trapping the
peptide in particular registers, we show that most, if not all,
NOD CD4+T cells react to B:9–23 bound in low-affinity register 3.
However, these T cells can be divided into two types depending on
whether their response is improved or inhibited by substituting
a glycine for the B:21 glutamic acid at the p8 position of the pep-
tide. On the basis of these findings, we constructed a set of fluo-
rescent insulin-IAg7tetramers that bind to most insulin-specific T-
cell clones tested. A mixture of these tetramers detected a high
frequency of B:9–23-reactive CD4+T cells in the pancreases of pre-
diabetic NOD mice. Our data are consistent with the idea that,
within the pancreas, unique processing of insulin generates trun-
cated peptides that lack or contain the B:21 glutamic acid. In the thy-
mus, the absence of this type of processing combined with the low
affinity of B:9–23 binding to IAg7in register 3 may explain the
escape of insulin-specific CD4+T cells from the mechanisms that
usually eliminate self-reactive T cells.
antigen processing|autoimmunity|T cell receptor|self tolerance
antigen for both B cells and T cells (reviewed in refs. 1, 2). A
peptide from the insulin beta chain (B:9–23) has been known for
many years to be the major target of insulin-reactive CD4+T
cells in NOD T1D . However, the data suggest that this peptide
can bind to the NOD class II major histocompatibility (MHCII),
IAg7, in multiple positions or “registers” within the peptide
binding groove (3–7). These registers are defined by the peptide
amino acids occupying positions p1–p9 in the groove, which in-
clude the “anchor” amino acids at p1, p4, p6, and p9, whose side
chains interact with compatible pockets in the MHC groove (8,
9). For an individual peptide, each shift in register puts a new set
of peptide amino acids into these anchor positions and brings
a different set of peptide amino acid side chains to the surface
for potential T-cell recognition, generating a unique ligand.
Defining which of the possible B:9–23 binding register(s) in the
IAg7groove create the ligand(s) for diabetogenic insulin-reactive
T cells has been difficult, leading to uncertainty in exactly how
this peptide is processed and presented to T cells in the pancreas
and the inability to construct the relevant fluorescent insulin-
IAg7multimers for in vivo tracking the autoimmune B:9–23-
specific T cells.
Recently, using techniques to trap versions of the B:9–23
peptide in particular IAg7binding registers, we concluded that
n human type 1 diabetes (T1D) and in the nonobese diabetic
(NOD) mouse model of the disease, insulin is a major auto-
most, if not all, of insulin-reactive CD4+T cells recognize this
peptide in a particular register (register 3) that places B:14–22 in
the p1–p9 positions (7). These results were surprising, because
this register would place B:22R in the p9 pocket, a highly un-
favorable match (8–10). However, an extensive study by others
had previously concluded that anti-B:9–23 T cells could be di-
vided into two types, which recognize the peptide bound to IAg7
in other registers (6). The first type was suggested to recognize
the peptide bound in register 1 with positions p1–p9 containing
B:12–20. The second type was suggested to recognize the peptide
bound in register 2 with B:13–21 in positions p1–p9. These
conclusions were strongly supported by data showing that the
B:9–23 peptide could be C-terminally truncated to B:20G for
some T cells and for other T cells to B:21E and still retain some
In the present study, we hypothesized that this apparent
contradiction between our data and that in the previous study
could be explained if these T cells recognized the insulin peptide-
bound register 3, but the two types of T cells recognized different
regions of the peptide, defined by the truncations, even though
the truncations would leave certain positions empty in the
binding groove. On the basis of this idea, we synthesized two
mutant versions of the insulin peptide that mimicked these two
truncations, but optimally filled the IAg7binding groove and
trapped the peptide in register 3. When these mimotopes were
tested with a set of insulin-reactive T-cell hybridomas, they dis-
criminated between the two types of T cells, while dramatically
improving presentation by IAg7compared with the wild-type
B:9–23. Fluorescent tetramers made with these mimotopes
detected most T-cell clones and a mixture of two of these tet-
ramers easily detected insulin-reactive T cells in the pancreatic
islets of prediabetic NOD mice, a feat not previously achieved
with other methods.
B:9–23 Peptide: Two Versions, both Bound to IAg7in Register 3.
Various versions of the insulin B:9–23 peptide are listed in Fig.
1A. The upper part of the list shows how the full-length B:9–23
peptide would occupy positions p1–p9 of the IAg7binding groove
when bound in either register 1 or 2. Also shown are the C-
Author contributions: F.C., B.S., N.J., A.M., M.N., P.M., G.E., and J.W.K. designed research;
F.C., N.J., A.M., P.P., and J.W.K. performed research; B.S. and M.N. contributed new re-
agents/analytic tools; F.C., A.M., and J.W.K. analyzed data; and F.C., P.M., G.E., and J.W.K.
wrote the paper.
The authors declare no conflict of interest.
Freely available online through the PNAS open access option.
1Present address: Department of Pathology, University of Massachusetts, Worcester,
2To whom correspondence should be addressed. E-mail: email@example.com.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
| October 4, 2011
| vol. 108
| no. 40
terminal truncations of the peptide to B:9–20 and B:9–21, which
were previously suggested to define the C-terminal limit of two
different T-cell epitopes, one bound in register 1 and the other in
register 2 (6). In these registers the peptides are predicted to fill
the binding groove from p1 to p9 and the peptide amino acids
pointing up from the binding groove for T-cell receptor (TCR)
interaction would be completely different for each register,
creating unrelated epitopes.
Fig. 1A also shows how B:9–23 and these two truncated pep-
tides would look if both were bound in register 3. The wild-type
peptide B:9–23 would have B:21E at p8, whose side chain would
be available on the surface for T-cell interaction and the p9
position would contain B:22R, whose positively charged side
chain may or may not stably occupy the conflicting positively
charged p9 pocket of IAg7. In this register, the B:9–20 truncated
peptide would leave positions p8 and p9 in the IAg7groove
empty and the truncated B:9–21 peptide would leave just the p9
position empty. These two truncated peptides both eliminate the
problem of positively charged p9R conflict with the positively
charged p9 pocket and create two very similar peptide-IAg7
ligands that differ only by the presence or absence of the p8E.
These truncated peptides would be predicted to bind poorly in
register 3, both because they do not completely fill the peptide
binding groove and because they might bind better in other
registers (6, 7), Therefore, to separate the issues of the register
and strength of IAg7binding from those of T-cell specificity, we
designed two mimics of the truncated peptides (Fig. 1A, Lower)
designed both to fill the peptide binding groove and to bind
strongly in register 3, while removing the surface expression of
the side chains of the amino acids at p8 and/or p9. In both cases,
we extended the peptide back to the B:23 amino acid but
changed the the p9(B:22) position to E to provide an optimal
anchor for the p9 pocket and also removing the possibility that
the p9 amino acid side chain might be exposed for T-cell in-
teraction. As we have shown previously (7), this mutation dra-
matically improves the binding of the peptides in register 3. In
the first peptide, we also mutated the p8E to G, thus removing
the p8E side chain to mimic the B:9–20 truncation. In the second
peptide, we left the natural p8E, thus mimicking the B:9–21
truncation. For a control peptide, we left p8 and p9 with the
wild-type (WT) amino acids E and R. We synthesized these three
peptides including a final mutation of the p6C to A in all three,
because previous studies had shown that this mutation was
neutral for IAg7binding and T-cell recognition, but prevented
the confounding effect of soluble peptide disulfide dimerization
in our in vitro studies (7).
B:9–23(WT), B:9–23(p8G), and B:9–23(p8E) were all tested
for activation of a collection of insulin-reactive T-cell hybrid-
omas (2, 6, 7, 11). We used the IAg7bearing M12.C3.G7 B-cell
lymphoma cell line (12) as the antigen-presenting cell (APC). To
minimize the possibility of further processing of the peptides in
the APC, we fixed the cells with paraformaldehyde before the
addition of the peptides. We used the NOD T-cell hybridoma,
BDC-2.5, specific for chromogranin A (CHGA), as a negative
control (13, 14). If our hypothesis were correct, each T-cell in
this collection should respond somewhat poorly to the B:9–23
(WT) peptide, but much better to one or the other mutant
peptide mimics. This prediction was borne out by the results
Three of the hybridomas, AS91, 12-4.4, and 8-1.1 responded
much better to B:9–23(p8G), designed to mimic the B:9–20
truncation, than to either the B:9–23(WT) or B:9–23(p8E)
mimotope. AS91 is the prototypical T-cell reported to recognize
B:9–20 bound to IAg7in register 1 (6). Reciprocally, AS150, 12-
4.1, PCR1–10, and I.29, all responded much better to B:9–23
(p8E), designed to mimic the B:9–21 truncation, than to either
the B:9–23(WT) or B:9–23(p8G) mimotope. AS150 is the pro-
totypical T-cell reported to recognize B:9–21 bound to IAg7in
register 2 (6). These results offered an explanation for the ap-
parent contradiction in the previously published studies and lead
us to propose that all of these T cells actually recognize the in-
sulin peptide bound in register 3, with the p8E inhibitory for
some (type I), but helpful for others (type II).
It is noteworthy that the TCRs of 8-1.1 and PCR1–10 share
the same α-chain (11). They also have related Vβ’s, with closely
related CDR1 loops, but different CDR2 and CDR3 loops (Fig.
S1). Because in published TCR-MHCII structures (e.g., PDF
nos. 1FYT, 3C52, and 3QIB), the peptide p8 amino acid often
interacts with the TCR Vβ CDR2 and/or CDR3 loop, this in-
teraction may be the primary feature distinguishing type I from
type II insulin specific T cells.
Confirmation of Register 3 Presentation of Insulin to both Types of T
Cells. As in a previous study, we performed additional experi-
ments to confirm that these altered versions of the insulin pep-
tide were bound and recognized in register 3. For both the p8G
and p8E peptide, we prepared three different baculovirus-
encoded constructs with the peptide covalently linked to the IAg7
β-chain and trapped in register 3 (7, 15). The six constructions
cells. (A) Various natural and mimotope variants of the insulin B:9–23 pep-
tide are shown. The amino acid mutations introduced into the mimotope
peptides are highlighted in yellow. The red amino acids are predicted to fill
the four anchor positions in the IAg7groove, when the peptides are bound in
a particular register. (B) As described in Materials and Methods, seven NOD
insulin B:9–23 reactive T-cell hybridomas were tested for their responses to
various concentrations of the three insulin B:9–23 peptide mimotopes listed
in A. The NOD chromogranin A-specific T-cell hybridoma, BDC-2.5, was used
as a negative control. The peptides were presented by paraformaldehyde-
fixed, M12.C3 lymphoma B cells transfected with IAg7. The averaged results
from three experiments are shown as the amount of IL-2 produced in 24 h ±
SEM vs. the peptide concentration.
Insulin B:9–23 mimotopes define two types of insulin-reactive CD4+T
| www.pnas.org/cgi/doi/10.1073/pnas.1113954108 Crawford et al.
are shown schematically in Fig. 2A. For all, the predicted register
3 p1 amino acid (B:14) was changed to R and the p9 position
(B:22) was changed to E to provide optimal register 3 anchors at
these positions. The second peptide in each group matched the
natural insulin p6C(B:19) with a mutation of IAg7α62N to C,
thus creating a disulfide bond between IAg7and the peptide
locking the peptide in register 3. An alternate disulfide bond was
introduced in the third peptide in each group by introducing
a C in the peptide linker immediately after the peptide (p11)
matched with a mutation of IAg7α72I to C, again locking the
peptide in register 3. In this case, the natural p6C was mutated to
A to avoid ambiguity in the disulfide bond formation. In all of
the constructs the natural IAg7transmembrane regions and cy-
toplasmic tails were deleted and the baculovirus gp64 trans-
membrane region and cytoplasmic tail were added to the end of
Viruses encoding these constructs were used to infect
B7+ICAM+SF9 insect cells (16). These cells were tested as
APCs with the type I and type II insulin-reactive T cells shown in
Fig. 1B, using the response induced by the immobilized H57–597
Cβ-specific monoclonal antibody (Mab) (17) as a control. The
results are shown in Fig. 2B. All of the T cells responded strongly
to the immobilized anti-TCR Mab. The results with the insect
cell presented IAg7-linked peptides were similar to those
obtained with soluble peptides presented by B-cell lymphoma
APCs in Fig. 1B. All three of the type I T-cell hybridomas
responded poorly or not at all to the three register 3 constructs
with the natural p8E. They all responded better to one or more
of the p8G constructs. On the other hand, among the type II T
cells, AS150 and 12-4.1 responded better to the register 3 trap-
ped peptides than any of the constructs with the natural p8E. I.29
and PCR1–10 did not discriminate between the two types of the
peptides as strikingly here as they did using titrated soluble
peptides in Fig. 1B. However, with soluble peptides these T cells
required the highest dilutions of the peptide to discriminate
between the two peptides. Therefore, the level of expression of
the covalent IAg7constructs on the insect cells may have been
too high to distinguish more strongly between the versions.
The T cells did not always respond equally well to all three of
the constructs expressing the same version of the peptide. In
some cases the disulfides appeared to inhibit the T-cell responses
somewhat. Most strikingly, AS150 did not respond at all to p8G
peptide with a disulfide to IAg7at p6, although it responded very
strongly to this peptide with no disulfide or with a disulfide at
the p11 position.
Tetramers. All six constructs described in Fig. 2A were altered
to replace the baculovirus gp64 transmembrane and cytoplasmic
tail used for insect cell surface expression with a peptide sub-
strate for BirA biotinylation (18). Soluble protein was produced
and analyzed by SDS/PAGE with and without reduction to as-
sure virtually complete formation of the introduced disulfides
(Fig. S2). The proteins were biotinylated and incorporated into
phycoerythrin (PE) streptavidin fluorescent tetramers, as pre-
viously described (18). These tetramers were used to stain the
seven insulin-reactive T-cell hybridomas. As a control the T cells
were stained separately with a APC-labeled Cβ-specific Mab.
The stained cells were analyzed by flow cytometry (Fig. 3). Fig.
3A shows an example of the staining with a comparison of the
binding of the p8G or p8E insulin tetramers with a disulfide bond
between p6 and IAg7, and a negative control tetramer, to a hy-
bridoma of each type: AS91 (type I) and PCR1–10 (type II). As
expected, the hybridomas had reciprocal staining patterns. AS91
bound the p8G, but not the p8E, tetramer. PCR1–10 bound the
p8E, but not the p8G tetramer.
Fig. 3B shows a summary of the staining seen with all six tet-
ramers on the seven T-cell hybridomas. Five of the seven T cells
bound one or more of the six fluorescent reagents. AS91, 12-4.4,
and 8-1.1, bound one or more of the p8G tetramers, but none of
the p8E reagents. On the other hand, I.29 and PCR1–10 bound
all of the p8E tetramers, but only one of the p8G reagents. None
of the tetramers stained 12-4.1 or AS150. This could be due
a very low affinity of these TCRs for the complex. However,
a more likely explanation, especially for the strongly reactive
AS150 T cell, is that this cell detects some heterogeneity in the
p8E peptide conformation that I.29 and PCR1–10 are insensitive
to. We expand on this point in Discussion.
T-cell Hybridomas Stain with FluorescentIAg7
Detection of Insulin-Specific Pancreatic T Cells with Fluorescent IAg7
Multimers. The uncertainty over the exact nature of the insulin
epitope presented by IAg7pathogenic CD4+in T1D has pre-
vented in vivo studies tracking these T cells with fluorescent IAg7
tetramers. With the knowledge that our collection of tetramers
trapped in register 3 detected the majority of the examined T-cell
clones, we used flow cytometry to examine the binding of some
of these tetramers to CD4+CD44highT cells infiltrating the islets
of 8-wk-old prediabetic NOD mice. For this experiment we
combined two of the register 3 insulin-IAg7PE-labeled tet-
ramers: the nondisulfide trapped version of the p8G multimer
and the p6 to IAg7disulfide locked version of the p8E tetramer
(Fig. 2A), since this combination best detected the most T-cell
hybridomas in Fig. 3B. As a positive control, we used another
register 3. (A) Schematic representations of six versions of the insulin peptide
expressed in baculovirus linked to a cell surface expressed version of IAg7via
a linker to the N terminus of the β-chain. The amino acid mutations in-
troduced into the peptide or IAg7are highlighted in yellow. The red amino
acids are predicted to fill the four anchor positions in the IAg7groove. See
text for further description. (B) SF9 insect cells bearing mouse ICAM and B7
were infected with virus encoding each of the six constructs. Three days after
infection these cells were used as APCs for the seven type I and type II in-
sulin-specific T-cell hybridomas in Fig. 1B. IL-2 production was assayed at
24 h. Stimulation of the hybridomas with an immobilized TCR Cβ-specifc
Mab was used as a positive control. Each bar is the average of the results of
three independent experiments ± SEM.
Register trapping of the insulin peptides confirms their binding in
Crawford et al. PNAS
| October 4, 2011
| vol. 108
| no. 40
IAg7tetramer, labeled with AF647, containing a peptide mim-
otope for CHGA (14), because T cells reactive with similar
mimotope tetramers were known to be present in the pancreases
of prediabetic NOD mice (19, 20). As a negative control, we used
two tetramers of IAg7with the well-characterized peptide from
hen egg lysozyme (HEL) (8), labeled with either PE or AF647.
The results are presented in Fig. 4A.
Total pancreatic islet-infiltrating cells were analyzed by flow
cytometry, first gating on live, B220–, F4/80–, CD8–, CD44high,
and CD4+T cells, then examining the tetramer binding. Virtu-
ally no pancreatic CD4+CD44highT cells were detected with
either control HEL tetramer (Fig. 4A, Right), but there were
easily detectable, separate, populations binding the mix of insulin
tetramers and the CHGA mimotope tetramer (Fig. 4A, Left).
Separate overlays of histograms of the CHGA or insulin staining
data with the corresponding HEL negative control (Fig. 4B)
showed that the insulin tetramers bound ∼4% and the CHGA
tetramers bound ∼2% of CD4+CD44highT cells. When the
pancreatic lymph nodes from the same mice were examined,
insulin-reactive T cells were found; the frequency was ∼100-fold
less than in the pancreas (Fig. S3). Similar results were obtained
with pancreases pooled from 12-wk-old mice. These data show
not only that the B:9–23 insulin peptide bound to IAg7in register
3 is the relevant epitope for a variety of insulin-reactive T-cell
clones, but that T cells with this specificity can be enumerated
directly in the pancreases and lymph nodes of NOD mice.
The combination of the data presented in previous reports (6, 7,
21) with our data presented here suggests why it has been so
difficult to define how the B:9–23 peptide combines with IAg7to
form the ligand for pathogenic CD4+T cells in NOD mice. The
peptide can bind to IAg7in at least three adjacent overlapping
registers and the functional register is the weakest binding of the
three. Furthermore, as we show here, a critical amino acid in the
peptide is inhibitory for some T cells, but helpful for others.
Therefore, results with an individual mutation or truncation of
the wild-type peptide can be misleading, since each mutation or
truncation can have multiple effects, such as altering the pro-
portion of the peptide binding in each register, the affinity of
peptide binding to MHC in a given register, the contacts with the
T-cell receptor or any combination of these.
The methods that we have used to trap the peptide bound in
a particular register have shed light on this problem by allowing
us to separate the question of strength and register of IAg7
binding from that of T-cell recognition. We show unequivocally
that the functional register for most, if not all, B:9–23-specific T
cells is register 3 and that in this register p8E is a critical amino
acid for T-cell recognition, defining two types of T cells—those
inhibited by the E at this position and those that require it for
pancreases of prediabetic NOD mice. Two mixtures of fluorescent IAg7tet-
ramers were prepared: (i) an equimolar mixture of two PE-streptavidin insulin
tetramers, one containing p8G with no disulfide and the second containing
p8E with a p6 to IAg7disulfide and an AF647-streptavidin tetramer containing
a mimotope (SRLGLWSRMD) for NOD chromogranin A (CHGA) reactive CD4+
T cells; and (ii) an equimolar mixture of a PE-streptavidin tetramer containing
the IAg7presented dominant epitope (MKRHGLDNYRGY) from hen egg ly-
sozyme (HEL) and the AF647 version of the same HEL tetramer. Pooled pan-
creatic islet cells were isolated from 8-wk-old prediabetic NOD mice and
stained with the two tetramer mixtures. The cells were stained as well with
anti-B220, anti-F4/80, anti-CD44, anti-CD4, and anti-CD8. For tetramer analy-
sis, cells were pregated on live, B220−, F4/80−, CD8−, CD4+, and CD44highcells.
(A) Zebra contour plots showing 2D simultaneous staining with the insulin
and CHGA tetramer (Left) or with the two HEL tetramers (Right). The per-
centage of cell in each quadrant is shown in B. One-dimensional histogram
overlay of the PE-labeled insulin (red) and HEL(blue) tetramers (Left) or
AF647-labeled CHGA (red) and HEL (blue) tetramers (Right). The percentage
of CD4+and CD44highT cells staining with the insulin or CHGA tetramers was
calculated from the difference between the two histograms.
IAg7-insulin tetramers detect insulin-reactive CD4+T cells in the
detect both types of T-cell hybridomas. The six constructs shown in Fig. 2A
were altered to replace the baculovirus gp64 transmembrane region and
cytoplasmic tail with a peptide substrate for the BirA biotinylation enzyme.
Soluble biotinylated proteins were prepared and incorporated into PE-
streptavidin tetramers as described in Materials and Methods. (A) Examples
of insulin-IAg7tetramer staining of a type I (AS91, Left) and type II (PCR1–10,
Right) T-cell hybridoma are shown using the p8G (green) and p8E (blue)
tetramer trapped in register 3 by a p6 to IAg7disulfide. Staining with a
negative control tetramer (CLIP-IAb) is also shown (gray). (B) The seven in-
sulin reactive T-cell hybridomas in Fig. 1B were stained with all six of the
insulin peptide-IAg7PE-streptavidin tetramers. The net mean fluorescent
intensity (MFI) of staining was calculated as the MFI for the insulin-IAg7
tetramer minus the MFI for the negative control tetramer (CLIP-IAb). The
bars show the average (± SEM) net MFI from three to five independent
experiments expressed as the percentage of the net MFI obtained with a
high-affinity APC-labeled 57–597 TCR Cβ Mab.
Fluorescent tetramers produced with register 3 trapped peptides
| www.pnas.org/cgi/doi/10.1073/pnas.1113954108 Crawford et al.
optimal T-cell activation. This knowledge allowed us to create a
set of fluorescent IAg7tetramers with two versions of the peptide
bound in register 3. These tetramers bound to the majority of
NOD insulin-reactive CD4+T-cell clones from multiple mice
and laboratories. More importantly, we show that they can be
used to detect pancreatic islet-infiltrating CD4+T cells in pre-
diabetic NOD mice. It is likely that, even with this set of tet-
ramers, some at present unknown portion of the B:9–23 reactive
T cells will go undetected, but nevertheless these reagents should
prove to be a very useful tool for following the development of
T1D in vivo.
Generally the definition of MHCII binding register is synon-
ymous with the definition of epitope, because each register cre-
ates a totally unique surface for TCR interaction and, therefore,
T cells that recognize one register of a bound peptide cannot
cross-react with another. However, a peptide bound in a partic-
ular register need not be homogeneous in structure nor recog-
nized identically by different TCRs. Depending on the particular
CDR loops and orientation of each TCR on the MHC-peptide
ligand, mutational and structural studies indicate that individual
amino acids of the peptide contribute more or less to the in-
terface with different TCRs (16, 22). The register 3 insulin
peptide with p8E vs. p8G is a particularly dramatic example of
this point. Moreover, structural studies have shown that there is
considerable flexibility allowed in the position of the peptide
backbone in the MHCII groove (23) and even slight changes in
an anchor amino acid can indirectly affect the structure of
a peptide on the MHCII surface presented to T cells (24). This
could explain the fact that individual B:9–23 reactive T cells of
both types were differentially affected by introduction of disul-
fides at p6 or p11. The elimination of peptide flexibility at these
positions might inhibit the ability of the peptide to assume the
structure optimal for an individual TCR.
A particularly dramatic example of alternate peptide con-
formations comes from the work of Unanue and coworkers (21,
25). They provide evidence that a single peptide, apparently
bound in a fixed register, can have at least two conformations
distinguished by different sets of T cells, depending on whether
the peptide was loaded extracellularly (conformation A) or via
the usual endocytic pathway (conformation B). The molecular
natures of the A and B conformations are not known, nor are
their exact proportions following peptide loading, but they may
explain the unusual behavior of the AS150 T cell.
Whereas strongly reactive to the soluble type II mimotope
peptide presented by fixed IAg7presenting cells, AS150 did not
respond to insect cells that expressed IAg7with a covalent type II
insulin peptide further anchored by a disulfide bond to p6, nor
did it bind any of the type II IAg7tetramers. One interpretation
of these results is that this T-cell recognizes the peptide in
conformation A that is well represented with surface IAg7loaded
with a soluble peptide. IAg7produced with a linked peptide may
be mostly in the B conformation, with only enough in the A
confirmation to allow some T-cell stimulation when highly
expressed in insect cells, but not enough to allow multivalent
IAg7tetramer binding to the T cell. The p6 disulfide bond may
prevent the A conformation altogether. We cannot formally rule
out the alternate possibility that even with the optimized p1 and
p9 anchors and the p11 fully formed disulfide, some small por-
tion of the type II peptide can nevertheless find a way to bind to
IAg7in some way other than register 3.
The naturally processed insulin peptides recognized by NOD
pathogenic CD4+T cells are not known; however, a previous
study suggested that these peptides are most efficiently gener-
ated within the pancreas itself (21). Our findings open the pos-
sibility that the relevant peptides may include truncated versions
of the B:9–23 peptide that are uniquely generated in the pan-
creas and not in the thymus. Truncation to B21 eliminates the
B22R that conflicts with the IAg7p9 binding pocket and trun-
cation to B20 further eliminates B21E that interferes with rec-
ognition by some T-cell clones. If versions of these truncated
peptides represent the true naturally processed insulin peptides
and bind to IAg7without filling the binding groove, they would
join a growing list of autoimmune self-antigen peptides that may
be uniquely processed and/or presented to CD4+T cells in the
target tissue. For example, also in NOD T1D, the WE14 peptide
is naturally processed from chromogranin A in the secretory
granules of neuroendocrine cells, including the pancreatic β-cells
(26). This peptide activates a set of diabetogenic CD4+T cells,
despite the fact that, when bound to IAg7, positions p1–p4 in the
binding groove are left empty (14). Another example is in ex-
perimental allergic encephalomyelitis, the mouse model of
multiple sclerosis (27). In this case, N terminus of the myelin
basic protein peptide epitope, which is targeted by the patho-
genic CD4+T cells, occupies position p3 in the IAubinding
groove. The idea that alternate peptide processing of amino
acids at the N or C termini of a peptide can create truncated
peptides that are distinguished by different T cells has previously
been suggested (28, 29).
In human multiple sclerosis, two TCRs have been shown to
take very unusual positions on autoantigenic myelin basic protein
peptides bound to HLA-DR51 (30) or to HLA-DQ1 (31). It the
first case, the TCR has shifted dramatically toward the N-ter-
minal end of the peptide, interacting with an extension of the
peptide beyond the MHC binding groove, while ignoring peptide
C-terminal amino acids past position p5. In the second case, the
TCR has tilted dramatically toward the MHCII α-chain helix,
with minimal contact between the MHCII β-chain helix and
the CDR1 or CDR2 loop of the TCR Vα. In a similar vein,
experiments have implicated unique posttranslational mod-
ifications of self-peptides in the formation of the epitopes for
some autoimmune CD4+T cells in celiac disease and rheuma-
toid arthritis in humans and possibly T1D in NOD mice (32–34).
Taken together, these results raise the possibility that the
“autoantigens” driving many forms of autoimmunity can more
appropriately be thought of as “neoantigens,” i.e., peptides that
are uniquely formed or presented in the target organs and
therefore missing, poorly processed, or poorly presented in the
thymus, preventing deletion of autoreactive T cells. This idea has
implications for the hypothesis that differential affinity for thy-
mic-presented self-antigens determines whether potential auto-
reactive T cells are deleted or converted to T regulatory cells
during thymus development.
In conclusion, the data presented here extend our knowledge of
how the autoimmune ligands for CD4+T cells are formed in T1D
and provide a set of tools for studying this disease in the NOD
mouse. Given the similarities between human HLA-DQ (DQ2
and DQ8) alleles associated with T1D and IAg7, especially in the
properties of the p9 pocket of their peptide binding grooves, it is
quite possible that our results also predict the ways in which in-
sulin is presented to pathogenic CD4+T cells in human T1D.
Materials and Methods
Reagents. Peptides at >95% purity were synthesized and purified by CHI
Scientific. Oligonucleotide primers were synthesized in the Biomolecular
Resource Center at National Jewish Health. Phycoerythrin-streptavidin
(PESA) and Alexa Fluor 647-streptavidin (AF647SA) were obtained from
Prozyme and Molecular Probes, respectively. Fluorescently labeled Mabs
for flow cytometry were as follows: eFluor450-B220, eFluor450-F4/80,
APCeFluor780-CD8, PE-Cy7-CD4, and PerCP-Cy5.5-CD44 (eBiosource). Un-
labeled TCR Cβ-specific H57-597 was isolated from the hybridoma culture
supernatant. An APC-labeled version of this antibody was obtained from BD-
Pharmingen. A PE-streptavidin tetramer of IAbbound to the invariant chain
peptide CLIP was obtained from the National Institutes of Health
T-Cell Hybridomas/Transfectomas and Other Cell Lines. The NOD-derived T-cell
hybridomas, AS91, AS150, and I.29 were generously supplied by Mateo
Crawford et al. PNAS
| October 4, 2011
| vol. 108
| no. 40
Levisetti and Emil Unanue (Washington University, St. Louis, MO) (6, 7).
Other T-cell hybridomas and transfectoma were produced by us (2, 7, 11).
The BDC-2.5 T-cell hybridoma was a gift from Katherine Haskins (National
Jewish Health Center, Denver, CO) (13, 14). The M12.C3.G7 B-cell line was
a gift from Emil Unanue (Washington University, St. Louis, MO) (12).
Baculovirus Constructions. Baculovirus encoding in a single virus the extra-
cellular domains of the α- and β-chains of IAg7with a covalently attached
peptide were constructed as previously described (7, 15, 16, 18, 35). Briefly,
common features in all of the constructs were a flexible Gly/Ser-rich linker
attaching the peptide to the N terminus of the IAg7β-chain and a stabilizing
acid-base leucine zipper (36) attached to the C termini of the IAg7, the basic
half of the zipper to the IAg7α-chain and the acidic half, to the β-chain. In
some cases, disulfides were engineered between the peptide and the IAg7
α-chain, either between the peptide p6 amino acid and IAg7α62 or between
the peptide p11 (first amino acid of the linker) and IAg7α72 (7).
For insect cell surface expression, the transmembrane-cytoplasmic tail of
the baculovirus gp64 protein was attached to the C terminus of the acidic half
of the zipper. Viruses encoding these constructs were used to infect mouse
ICAM+B7+SF9 insect cells (16), which were used as APCs 3 d after infection.
For soluble IAg7, the peptide DATLTEKSFETDMNLNFQNLSVN was added to
the C terminus of the basic half of the zipper and used for immunoaffinity
purification with the hamster Mab, ADO-124 (7). A peptide biotinylation
substrate (LGGIFEAMKMELRD) for the BirA enzyme was added to the C
terminus of the acid half of the zipper (37). Soluble IAg7was purified from
the culture supernatants of High Five insect cells 5 d after infection with
virus encoding these constructs. As previously described (18), purified IAg7
was biotinylated with the BirA enzyme and an excess of bio-IAg7was in-
corporated into PESA and/or AF647SA tetramers. Tetramers were separated
from monomeric IAg7by FPLC size exclusion chromatography on Super-
dex200 10/300GL (GE Healthcare).
SDS/PAGE. SDS/PAGE analysis was performed using a Phastgel electrophoresis
system with 10–15% gradient acrylamide gels (GE Healthcare). An aliquot
containing 2 μg/μL of protein was boiled in an equal volume of 2× SDS/PAGE
sample buffer with and without the addition of 2-mercaptoethanol. Ap-
proximately 1 μg of total protein was loaded in each lane.
Preparation of NOD Pancreatic Cells. Pancreatic islets were manually dissected
from the pancreases of 8- to 12-wk-old prediabetic NOD female mice (38) and
cultured overnight in Click’s medium, supplemented with 0.5% mouse se-
rum (39) to allow the migration of infiltrated cells out of the tissue. The
preparation was then passed through a 100-μm nylon mesh screen to
remove residual tissue.
Flow Cytometric Analysis of IAg7Tetramer Binding. A total of 2–10 × 105hy-
bridoma T cells were incubated in 25 μL of PESA tetramer (15–20 μg/mL) in
culture medium containing excess 24G2 FcR-specific and 1 μg/mL unlabeled
H597 Cβ-specific antibodies for 2 h at 37 °C in humidified 10% CO2, with
gentle agitation every 30 min. The cells were washed and analyzed on the
FacsCalibur flow cytometer (Becton-Dickinson). A separate stain was done
for TCR levels using PE-labeled HAM-597.
A total of 1–5 × 105isolated islet cells were incubated in 25 μL of PESA
tetramer (15–20 μg/mL) in culture medium containing excess 24G2 FcR spe-
cific for 2 h at 37 °C in humidified 10% CO2, with gentle agitation every 30
min. Fluorescent anti-B220, -F4/80, -CD8, -CD44, and -CD8 Mabs were then
added and the cells incubated an additional 20 min at room temperature.
Cells were washed, fixed, and analyzed on a CYAN flow cytometer (Dako).
ACKNOWLEDGMENTS. This work was supported by US Public Health Service
Grants AI-18785, AI-22295, AI-15416, AI-050864, DK080885, DK055969, and
DK057516 and Juvenile Diabetes Research Foundation Grants 17-2010-744
1. Lieberman SM, DiLorenzo TP (2003) A comprehensive guide to antibody and T-cell
responses in type 1 diabetes. Tissue Antigens 62:359–377.
2. Zhang L, Nakayama M, Eisenbarth GS (2008) Insulin as an autoantigen in NOD/human
diabetes. Curr Opin Immunol 20:111–118.
3. Daniel D, Gill RG, Schloot N, Wegmann D (1995) Epitope specificity, cytokine pro-
duction profile and diabetogenic activity of insulin-specific T cell clones isolated from
NOD mice. Eur J Immunol 25:1056–1062.
4. Abiru N, et al. (2000) Dual overlapping peptides recognized by insulin peptide B:9-23
T cell receptor AV13S3 T cell clones of the NOD mouse. J Autoimmun 14:231–237.
5. Burton AR, et al. (2008) On the pathogenicity of autoantigen-specific T-cell receptors.
6. Levisetti MG, Suri A, Petzold SJ, Unanue ER (2007) The insulin-specific T cells of
nonobese diabetic mice recognize a weak MHC-binding segment in more than one
form. J Immunol 178:6051–6057.
7. Stadinski BD, et al. (2010) Diabetogenic T cells recognize insulin bound to IAg7 in an
unexpected, weakly binding register. Proc Natl Acad Sci USA 107:10978–10983.
8. Latek RR, et al. (2000) Structural basis of peptide binding and presentation by the
type I diabetes-associated MHC class II molecule of NOD mice. Immunity 12:699–710.
9. Corper AL, et al. (2000) A structural framework for deciphering the link between I-
Ag7 and autoimmune diabetes. Science 288:505–511.
10. Suri A, et al. (2002) In APCs, the autologous peptides selected by the diabetogenic I-
Ag7 molecule are unique and determined by the amino acid changes in the P9
pocket. J Immunol 168:1235–1243.
11. Simone E, et al. (1997) T cell receptor restriction of diabetogenic autoimmune NOD T
cells. Proc Natl Acad Sci USA 94:2518–2521.
12. Carrasco-Marin E, Shimizu J, Kanagawa O, Unanue ER (1996) The class II MHC I-Ag7
molecules from non-obese diabetic mice are poor peptide binders. J Immunol 156:
13. Haskins K, Portas M, Bradley B, Wegmann D, Lafferty K (1988) T-lymphocyte clone
specific for pancreatic islet antigen. Diabetes 37:1444–1448.
14. Stadinski BD, et al. (2010) Chromogranin A is an autoantigen in type 1 diabetes. Nat
15. Kozono H, White J, Clements J, Marrack P, Kappler J (1994) Production of soluble
MHC class II proteins with covalently bound single peptides. Nature 369:151–154.
16. Crawford F, Huseby E, White J, Marrack P, Kappler JW (2004) Mimotopes for allo-
reactive and conventional T cells in a peptide-MHC display library. PLoS Biol 2:
17. Kubo RT, Born W, Kappler JW, Marrack P, Pigeon M (1989) Characterization of
a monoclonal antibody which detects all murine alpha beta T cell receptors. J Im-
18. Crawford F, Kozono H, White J, Marrack P, Kappler J (1998) Detection of antigen-
specific T cells with multivalent soluble class II MHC covalent peptide complexes.
19. Jang MH, Seth NP, Wucherpfennig KW (2003) Ex vivo analysis of thymic CD4 T cells in
nonobese diabetic mice with tetramers generated from I-A(g7)/class II-associated in-
variant chain peptide precursors. J Immunol 171:4175–4186.
20. You S, et al. (2003) Detection and characterization of T cells specific for BDC2.5 T cell-
stimulating peptides. J Immunol 170:4011–4020.
21. Mohan JF, et al. (2010) Unique autoreactive T cells recognize insulin peptides gen-
erated within the islets of Langerhans in autoimmune diabetes. Nat Immunol 11:
22. Huseby ES, et al. (2005) How the T cell repertoire becomes peptide and MHC specific.
23. Dai S, et al. (2010) Crystal structure of HLA-DP2 and implications for chronic beryllium
disease. Proc Natl Acad Sci USA 107:7425–7430.
24. Kersh GJ, et al. (2001) Structural and functional consequences of altering a peptide
MHC anchor residue. J Immunol 166:3345–3354.
25. Lovitch SB, Esparza TJ, Schweitzer G, Herzog J, Unanue ER (2007) Activation of type B
T cells after protein immunization reveals novel pathways of in vivo presentation of
peptides. J Immunol 178:122–133.
26. Arden SD, et al. (1994) The post-translational processing of chromogranin A in the
pancreatic islet: Involvement of the eukaryote subtilisin PC2. Biochem J 298:521–528.
27. He XL, et al. (2002) Structural snapshot of aberrant antigen presentation linked to
autoimmunity: The immunodominant epitope of MBP complexed with I-Au. Immu-
28. Carson RT, Vignali KM, Woodland DL, Vignali DA (1997) T cell receptor recognition of
MHC class II-bound peptide flanking residues enhances immunogenicity and results in
altered TCR V region usage. Immunity 7:387–399.
29. Vignali DA, Urban RG, Chicz RM, Strominger JL (1993) Minute quantities of a single
immunodominant foreign epitope are presented as large nested sets by major his-
tocompatibility complex class II molecules. Eur J Immunol 23:1602–1607.
30. Li Y, et al. (2005) Structure of a human autoimmune TCR bound to a myelin basic
protein self-peptide and a multiple sclerosis-associated MHC class II molecule. EMBO J
31. Sethi DK, et al. (2011) A highly tilted binding mode by a self-reactive T cell receptor
results in altered engagement of peptide and MHC. J Exp Med 208:91–102.
32. Hovhannisyan Z, et al. (2008) The role of HLA-DQ8 beta57 polymorphism in the anti-
gluten T-cell response in coeliac disease. Nature 456:534–538.
33. Feitsma AL, et al. (2010) Identification of citrullinated vimentin peptides as T cell
epitopes in HLA-DR4-positive patients with rheumatoid arthritis. Arthritis Rheum 62:
34. Hill JA, et al. (2008) Arthritis induced by posttranslationally modified (citrullinated)
fibrinogen in DR4-IE transgenic mice. J Exp Med 205:967–979.
35. Crawford F, et al. (2006) Use of baculovirus MHC/peptide display libraries to charac-
terize T-cell receptor ligands. Immunol Rev 210:156–170.
36. O’Shea EK, Lumb KJ, Kim PS (1993) Peptide ‘Velcro’: Design of a heterodimeric coiled
coil. Curr Biol 3:658–667.
37. Schatz PJ (1993) Use of peptide libraries to map the substrate specificity of a peptide-
modifying enzyme: a 13 residue consensus peptide specifies biotinylation in Escher-
ichia coli. Biotechnology (N Y) 11:1138–1143.
38. Nicolls MR, Coulombe M, Yang H, Bolwerk A, Gill RG (2000) Anti-LFA-1 therapy in-
duces long-term islet allograft acceptance in the absence of IFN-gamma or IL-4. J
39. Kobayashi M, et al. (2008) Conserved T cell receptor alpha-chain induces insulin au-
toantibodies. Proc Natl Acad Sci USA 105:10090–10094.
| www.pnas.org/cgi/doi/10.1073/pnas.1113954108 Crawford et al.