Molecular Pathways for Immune Recognition ofPreproinsulin Signal Peptide in Type 1 DiabetesDeborah Kronenberg-Versteeg,1,2 Martin Eichmann,1 Mark A. Russell,3 Arnoud de Ru,4 Beate Hehn,5

Norkhairin Yusuf,1 Peter A. van Veelen,4 Sarah J. Richardson,3 Noel G. Morgan,3 Marius K. Lemberg,5 andMark Peakman1,2

Diabetes 2018;67:687–696 | https://doi.org/10.2337/db17-0021

The signal peptide region of preproinsulin (PPI) containsepitopes targeted by HLA-A-restricted (HLA-A0201, A2402)cytotoxic T cells as part of the pathogenesis of b-cell de-struction in type 1 diabetes. We extended the discovery ofthe PPI epitope to disease-associated HLA-B*1801 andHLA-B*3906 (risk) and HLA-A*1101 and HLA-B*3801 (pro-tective) alleles, revealing that four of six alleles presentepitopes derived from the signal peptide region. Duringcotranslational translocation of PPI, its signal peptide iscleaved and retained within the endoplasmic reticulum(ER) membrane, implying it is processed for immunerecognition outside of the canonical proteasome-directedpathway. Using in vitro translocation assays with specificinhibitors and gene knockout in PPI-expressing targetcells, we show that PPI signal peptide antigen processingrequires signal peptide peptidase (SPP). The intramem-brane protease SPP generates cytoplasm-proximal epito-pes, which are transporter associatedwith antigen processing(TAP), ER-luminal epitopes, which are TAP independent,each presented by different HLA class I molecules andN-terminal trimmed by ER aminopeptidase 1 for optimalpresentation. In vivo, TAP expression is significantlyupregulated and correlatedwithHLA class I hyperexpressionin insulin-containing islets of patients with type 1 diabetes.Thus, PPI signal peptide epitopes are processed by SPPand loaded for HLA-guided immune recognition via path-ways that are enhanced during disease pathogenesis.

In type 1 diabetes, the pathological process of immune-mediated destruction of insulin-producing b-cells leads toinsulin deficiency and hyperglycemia (1). Multiple arms ofthe immune system are likely to contribute to this tissue-damaging process, with strong indications that CD8+ cyto-toxic T lymphocytes (CTLs) are a dominant killing pathway.Evidence includes data from preclinical models showing de-pendence of disease development on intact CD8/MHC classI mechanisms (2), supported by compelling findings in hu-man studies, including the existence of high-risk polymor-phic HLA class I genes (3); enrichment of effector CTLsspecific for b-cell targets in the circulation in new-onsetdisease (4,5); recapitulation of b-cell killing by patient-derived preproinsulin (PPI)-specific CTLs in vitro (4,6);CD8 T-cell dominance of islet infiltrates in patients, includ-ing the presence of CD8s bearing receptors specific forb-cell autoantigens (7,8); hyperexpression of HLA class I,both at the RNA and protein levels, in residual insulin-containing islets (ICIs) in type 1 diabetes pancreatic tissue(9); and recent success in halting b-cell loss using immuno-therapy targeted at effector CD8 T cells (10).

The potential for the immunological dialogue betweenCTLs and b-cells to be a key component of the developmentof type 1 diabetes has led several groups to focus on therelevant molecular interactions that govern this interface,including studies of HLA class I gene polymorphisms carry-ing modified risk of disease (HLA-A*0201, -A*1101, -A*2402,

1Department of Immunobiology, Faculty of Life Sciences and Medicine, King’sCollege London, London, U.K.2National Institute for Health Research, Biomedical Research Centre at Guy’s andSt. Thomas’ Hospital Foundation Trust and King’s College London, London, U.K.3Institute of Biomedical and Clinical Science, University of Exeter Medical School,Exeter, U.K.4Department of Immunohematology and Blood Transfusion, Leiden UniversityMedical Center, Leiden, the Netherlands5Center for Molecular Biology of Heidelberg University (ZMBH), DKFZ-ZMBH Al-liance, Heidelberg, Germany

Corresponding author: Deborah Kronenberg-Versteeg, dk480@cam.ac.uk.

Received 5 January 2017 and accepted 10 January 2018.

This article contains Supplementary Data online at http://diabetes.diabetesjournals.org/lookup/suppl/doi:10.2337/db17-0021/-/DC1.

D.K.-V. and M.E. contributed equally to this work.

D.K.-V. is currently affiliated with the Wellcome Trust – Medical Research CouncilCambridge Stem Cell Institute, University of Cambridge, Cambridge, U.K.

© 2018 by the American Diabetes Association. Readers may use this article aslong as the work is properly cited, the use is educational and not for profit, and thework is not altered. More information is available at http://www.diabetesjournals.org/content/license.

Diabetes Volume 67, April 2018 687

IMMUNOLOGYAND

TRANSPLANTATIO

N

-B*1801, -B*3801, and -B*3906 [3,11,12]) and antigenic tar-gets within b-cells recognized by CTLs. These have espe-cially focused on PPI, which is considered a primary targetin b-cell autoimmunity because anti-insulin autoantibodiesare frequently the first disease manifestation in high-riskchildren (12).

For immune recognition, processing and presentation ofpeptides by MHC class I commonly results from degrada-tion of proteins by the proteasome (13), generating cytosolicpeptides of 8–16 amino acids (14), which are transportedinto the endoplasmic reticulum (ER) lumen via the trans-porter associated with antigen processing (TAP) (15), partof the MHC class I peptide loading complex (16). ER ami-nopeptidase (ERAP) 1 trims peptides to a suitable lengthfor MHC loading, if required (17), and peptide-MHCcomplexes are delivered to the cell surface for immune rec-ognition. Proteasome-independent, cytoplasm-based nonca-nonical antigen presentation pathways have also beendescribed (18). Whereas these and the canonical route col-lectively require TAP to delivery peptides into the ER, signalpeptide epitopes from secretory proteins may be MHC-loaded independently of TAP (19) after signal peptidasecleavage and intramembrane proteolysis by signal peptidepeptidase (SPP) (20–22). Interestingly, peptides originatingfrom the ER luminal side of the signal peptide can accessthe peptide loading complex directly (23–25), whereasepitopes close to the cytosol may require proteasomal trim-ming and TAP for entry into the ER (26). However, theextent to which signal peptides and these noncanonicalepitope-generation pathways fuel immune recognition ofsingle antigens and play a role in physiological responsesremains unclear.

This study was motivated by the need for a better un-derstanding of autoantigen processing for immune recog-nition of b-cells via these different routes, which couldprovide novel insights into disease pathogenesis and high-light pathways susceptible to therapeutic manipulation. Wepreviously reported that the predominant epitope speciespresented to CTLs by human cell lines cotransfected withthe INS gene (encoding PPI) and the HLA class I genes HLA-A*0201 and A*2402 derived from the signal peptide region(4,6). These signal peptide–derived epitopes are recognizedby patient CTLs and presented by b-cells bearing the rele-vant HLA class I molecule. Here we examine whether PPIsignal peptide is a more general source of epitopes by study-ing additional HLA class I molecules associated with type 1diabetes. We show that generation of epitopes from the PPIsignal peptide is driven by the intramembrane proteaseSPP and that loading into nascent HLA class I moleculesrequires trimming by ERAP1 and is either direct or followscytoplasmic translocation and TAP, with the selected path-way being determined by the HLA allele. We show thatthe key factors ERAP1 and TAP are expressed in insulin-containing islets of patients studied post mortem aftertype 1 diabetes diagnosis, indicating that the multiple path-ways that are potentially critical in the CTL–b-cell dialogueare active in the disease setting.

RESEARCH DESIGN AND METHODS

Cell Lines and Epitope DiscoveryK562 cells transfected with HLA-A*1101, HLA-B*1801,HLA-B*3801, or HLA-B*3906 and PPI cDNAs (K562-HLA-PPI cells) were generated as described previously (4). Ex-pression of HLA class I was confirmed by flow cytometry(anti-HLA-ABC antibody W6/32; Serotec, Oxford, U.K.).Proinsulin was detected in supernatants by ELISA (DRGInternational, Marburg, Germany). Subsequently, ;1010

cells from each cell line were harvested and immunoaffinitypurification of HLA-A1101, HLA-B1801, HLA-B3801, andHLA-B3906, peptide extraction and nano high-performanceliquid chromatography–mass spectrometry performed asdescribed elsewhere (4,6,27,28).

Inhibitors, RNA Interference and T-Cell Clone ActivationK562-A24-PPI cells were transfected twice within 48 h with20 nmol/L small interfering RNAs (Applied Biosystems,Foster City, CA) targeting B2M (s1852), ERAP1 (s28618),ERAP2 (s34520), TAP1 (s13778, s13780), and TAP2 (s13781,s13782, s13783) using Lipofectamine RNAiMAX (Invitrogen,Paisley, U.K.). Knockdown was assessed by RT-PCR us-ing TaqMan-specific primers and relative cDNA contentnormalized to GAPDH gene expression. CD8 T-cell clonesfor PPI3–11-HLA-A2402 and PPI15–24-HLA-A0201 (4,6) werecocultured with K562-HLA-PPI target cells at indicatedeffector-to-target ratios (4 h), and response was measuredas degranulation (CD107a expression [29]) or macrophageinflammatory protein 1b (MIP-1b) release (R&D Systems,Minneapolis, MN).

Site-Directed MutagenesisTo alter single or multiple amino acids (highlighted in Sup-plementary Table 1), PPI-containing plasmid pcDNA3/PPIwas amplified using altered primers (PPI9P→L, CCC→CTC;PPI12A→L, GCG→CTG; PPI15A→L, GCC→CTC; with 18- to20-nucleotide overhang preceding and following the mu-tated site) and PfuTurbo DNA Polymerase AD (AgilentTechnologies, Santa Clara, CA); sequences were confirmedbefore use.

In Vitro Transcription, Translation, and Translocationand Analysis of SPP ProcessingPlasmid pcDNA3/PPI was linearized with EcoRI and tran-scribed in vitro with T7 RNA polymerase at 42°C using500 mmol/L m7G(59)ppp(59)G CAP analog (New EnglandBiolabs, Ipswich, MA) (30). mRNAs were translated in vitroin 25 mL rabbit reticulocyte lysate (Promega, Madison, WI)containing [35S]-methionine and [35S]-cysteine (PerkinElmer,Waltham, MA) and, where indicated, two equivalents ofnuclease-treated dog pancreas rough microsomes (31).(Z-LL)2-ketone (Calbiochem SPP inhibitor, 5 mmol/L; Merck,San Diego, CA) or DMSO control were added as indicated.After 30 min at 30°C, microsomes were extracted with500 mmol/L KOAc, solubilized in SDS sample buffer (32)and analyzed by SDS-PAGE using Tris-bicine-urea acrylam-ide gels (15% Tris, 5% bicine; 8 mol/L urea) (33) and an FLA7000 phosphorimager (Fuji) with Multi Gauge software (Fuji).

688 Preproinsulin Processing for Immune Recognition Diabetes Volume 67, April 2018

The reference peptide comprising the 24–amino acid PPIsignal sequence was translated in wheat germ extract (34).

Immunohistochemistry and ImmunofluorescenceFormalin-fixed, paraffin-embedded pancreas sections fromsix control and six case subjects with type 1 diabetes(Exeter Archival Diabetes Biobank, http://foulis.vub.ac.be/;Supplementary Table 2; ethical approval 15/W/0258) werestudied with the use of immunohistochemistry. Sections weredewaxed, rehydrated, and heated in a microwave (800 W) for20 min, then blocked in 5% normal goat serum before beingincubated with primary and secondary antibodies (Supplemen-tary Table 3), counterstained with hematoxylin, dehydrated,and mounted in a distyrene/xylene-based mountant (DPX).Multiple antigens within the same formalin-fixed, paraffin-embedded section were probed sequentially with up to threedifferent antibodies (Supplementary Table 3) and images cap-tured (AF6000 system; Leica Microsystems, Milton Keynes,U.K.). Mean fluorescence intensity of stained antigens wasanalyzed using ImageJ software, and isotype control antiserawere used to confirm reagent specificity.

SPP Knockout Using CRISPRCRISPR guide sequences for exons of SPP (HM13 gene)were designed using CRISPR DESIGN tool (crispr.mit.edu)(Supplementary Table 4) and cloned into pLG2C vectorcontaining EGFP linked to a Cas9 cassette via P2A. K562-A2-PPIwas transfectedwith each of twopLG2C vectors (con-taining guide sequences for exon 2 or 3 and exon 10 or 11)or empty vector (mock treated, no guide sequence) usingEffectene (Qiagen, Hilden, Germany). Single cells sorted forhigh HLA-A2 (W6/32; BioLegend, San Diego, CA) and EGFPwere examined for gene truncation by RT-PCR and for SPP

amplification using primers (forward: 59-ATATATGAATT-CGCACCCTCGCCATG-39; reverse: 59-ATATATCTCGAGGC-ACCAGCTGCATCATTTC-39) (Eurofins Scientific, Ebersberg,Germany) and Phusion High-Fidelity DNA Polymerase (NewEngland Biolabs) and by agarose gel electrophoresis.

RESULTS

Presentation of PPI Epitopes by HLA-A1101, HLA-B1801,HLA-B3801, and HLA-B3906K562 cells transfected with INS encoding human PPI andone of HLA-A*1101, HLA-B*1801, HLA-B*3801, and HLA-B*3906 generated surrogate b-cells secreting proinsulin andexpressing relevant HLA class I molecules (SupplementaryFig. 1). The immunopeptidome eluted from affinity-purifiedHLA-A1101 identified 905 peptides; from HLA-B1801, 615peptides; from HLA-B3801, 455 peptides; and from HLA-B3906, 298 peptides (all Mascot scores $40), and corre-sponded to published HLA-binding motifs and peptideligandomes (35). Of interest, PPI epitopes from the signalpeptide region were identified for HLA-B3801, PPI5–14MRLLPLLALL, and HLA-B3906, PPI5–12 MRLLPLLA (Fig.1 and Supplementary Fig. 2). In addition, we identifieda B-chain epitope for HLA-B3801, PPI33–41 SHLVEALYL, anda C-peptide epitope for HLA-A1101, PPI80–88 LALEGSLQK.Peptide identities were confirmed by tandem mass spectrom-etry profiling of the synthetic compound (Supplementary Fig.2). No other peptides from PPI were identified.

When considered with our previous reports of immuno-dominant PPI signal peptide–derived epitopes in PPI15–24-HLA-A0201 (6) and PPI3–11-HLA-A2402 (4), these newdiscoveries indicate that the signal peptide region of PPI isa rich source for processing for immune recognition (Fig. 1

Figure 1—PPI epitope discovery in HLA-B3906. A: Tandem mass spectrometry analysis of collision-induced dissociation revealing the tandemmass spectrum of a peptide (463.8 m/z, sequence MRLLPLLA). The correct identity of the peptide was proven by tandemmass spectrometry ofthe synthetic compound. B: The table lists the amino acid sequences of the peptide with the expected b- and y-fragment ions (extending fromthe amino terminus and carboxyl terminus, respectively). Observed fragment ions are underlined. Tandem mass spectra and a table of b- andy-fragment ions for PPI epitopes discovered in HLA-A1101 and HLA-B3801 are provided in Supplementary Fig. 2. C: Diagram indicates theposition of the PPI signal peptide in the ER membrane during cotranslational translocation and the sequences and positions of the signalpeptides identified by elution of the naturally processed and presented immunopeptidome for different HLA molecules.

diabetes.diabetesjournals.org Kronenberg-Versteeg and Associates 689

and Table 1). To examine whether this arises because signalpeptide regions are immunogenic per se, as has been reported(36), the available peptidome data were mined in greaterdepth, showing that the degree to which signal peptides arepresented is dependent on the HLA allele. While, for example,HLA-A0201 presents a signal peptide–derived epitope from46% of source proteins that contain a signal peptide, forHLA-A1101 this figure is only 6.9% (Supplementary Table5). Probing for any signal peptide bias using in silico pre-diction algorithms provided similar results (SupplementaryTable 6). We conclude that presentation of signal peptidesof PPI is not likely to result from a generalized propensity ofHLA molecules to select this region for presentation, but thenumber of HLA molecules studied to date remains limited.

Intramembrane Protease SPP Cleaves PPI Signal PeptideBased on previous studies of intramembrane cleavage ofsignal peptides (37), we hypothesized that SPP is involved

in processing and cleavage of PPI signal peptide. To testthis, mRNAs encoding PPI were translated in vitro withER-derived microsomes and [35S]-labeled methionine andcysteine. Upon insertion of PPI into microsomes, signalpeptidase cleaves the PPI signal sequence, liberating trans-located proinsulin from its signal peptide (Fig. 2A). More-over, traces of a peptide that comigrated with an in vitrotranslated reference peptide comprising the PPI signal se-quence were detected in the membrane fraction (Fig. 2A;compare lanes 2 and 4). Because processing of nascent chainsby signal peptidase is a well-known activity in ER-derivedmicrosomes (31), and because no low–molecular weightpeptides were observed in the translation reaction lackingmicrosomes (lane 1), we conclude that the identified pep-tide corresponds to traces of PPI signal peptide that re-main in the ER membrane fraction. By contrast, the PPIsignal peptide is markedly stabilized and retained in themicrosome fraction upon treatment with the SPP inhibitor

Table 1—PPI epitopes identified by elution from specific HLA class I molecules

Allele Risk/protection PPI signal peptide Other PPI regions

HLA-A*0201 Risk/neutral PPI15–24 (ALWGPDPAAA) No peptides found

HLA-A*2402 Risk PPI3–11 (LWMRLLPLL) No peptides found

HLA-B*1801 Risk No peptides found No peptides found

HLA-B*3906 Risk PPI5–12 (MRLLPLLA) No peptides found

HLA-A*1101 Protection No peptides found PPI80–88 (LALEGSLQK)

HLA-B*3801 Protection PPI5–14 (MRLLPLLALL) PPI33–41 (SHLVEALYL)

Boldface data signifies identified epitopes.

Figure 2—Processing of PPI signal peptide by SPP in a cell-free translocation assay. A: In vitro translation of wild-type (wt) PPI mRNA in theabsence (lane 1) or presence of ER-derived microsomes (lanes 2 and 3) and SPP inhibitor (Z-LL)2-ketone (lane 3). Microsomes containingradiolabeled translocated proinsulin (PI) and membrane integral signal peptides (SPs) were isolated and analyzed by SDS-PAGE and audio-radiography. Lane 4 shows in vitro translated reference peptide comprising the PPI signal sequence. Upon ER targeting and signal peptidasecleavage, liberated SP is released from the membrane fraction (lane 2) in a process that can be blocked in the presence of (Z-LL)2-ketoneinhibitor (lane 3), indicating SPP-catalyzed cleavage. B: In vitro translation of mutant PPI mRNA (P9L, A12L, A15L) under similar conditionsshows cleavage-deficient SP retained in the membrane fraction irrespective of (Z-LL)2-ketone inhibitor. C: Wild-type (top) and P9L, A12L, A15LSP mutant (bottom) sequences. D: Quantification of PPI SP processing (mean, n = 5 [wt] or n = 2 [mutations]). Means indicate the relativeamounts of SP obtained in lane 2 compared with the corresponding lane 3, where SPP was inhibited. Processing was quantified by comparingthe intensity of SP in the absence (DMSO) and presence of the SPP inhibitor (Z-LL)2-ketone for each condition, given by the formula 100 2(100x/y), where x is the intensity of the test band and y is the intensity of the band in the presence of full inhibition with (Z-LL)2-ketone. Equaltranslocation efficiency was controlled by comparing the amount of PI between conditions. See Supplementary Fig. 3 and Supplementary Table1 for details. A, alanine; L, leucine; P, proline.

690 Preproinsulin Processing for Immune Recognition Diabetes Volume 67, April 2018

(Z-LL)2-ketone (Fig. 2A, lane 3). Overall, this shows that PPIbehaves as a canonical nascent chain processed by signalpeptidase, liberating a signal peptide released from the ERmembrane by SPP-catalyzed intramembrane cleavage.

To further study the role of intramembrane proteolysisin turnover of PPI signal peptide, amino acid residues de-stabilizing the helical transmembrane span surrounding thescissile peptide bond were mutated to leucine, which blocksSPP-catalyzed cleavage (21). Consistent with our previousanalysis of model signal peptides, single mutation of proline(at position 9) and alanine (position 12 and position 15)shows marginal effects on PPI signal peptide processing,whereas double mutants (9L/12L, 9L/15L, and 12L/15L)and the triple mutant (9L/12L/15L) show marked inhibi-tion of SPP-catalyzed processing (Fig. 2B–D, SupplementaryFig. 3, and Supplementary Table 1). This identifies PPI sig-nal peptide as a bona fide SPP substrate and confirms therequirement for helix breaks and limited hydrophobicity forintramembrane cleavage. Replacement of the classic helix-break residue proline at position 9 of PPI signal peptidealone is not sufficient to completely block SPP-catalyzedprocessing and only shows effect when at least one of thenearby alanine residues, which show an intermediate stabil-ity within transmembrane helices (38), is also mutated toleucine.

We next examined whether abrogation of SPP-catalyzedcleavage of PPI signal peptide affects immune recognition ofsignal peptide–derived epitopes. Double-targeting CRISPR-Cas9 technology generated three independent K562-A2-PPIcell lines with SPP knockout (K562-A2-PPI-SPPko; SPPknockout validation shown in Supplementary Fig. 4). Whencultured with the PPI15–24-specific, HLA-A0201-restrictedCD8 T-cell clone, degranulation (which measures clone ac-tivation via T-cell receptor ligation by peptide-HLA) fre-quency and magnitude were markedly reduced in the

presence of K562-A2-PPI-SPPko cells compared withmock-treated lines (Fig. 3). The difference in T-cell activationis not due to differences in HLA class I expression levels(Supplementary Fig. 5), and the SPP knockout phenotype ofreduced T-cell activation is rescued when cells are pulsedwith cognate peptide (Supplementary Fig. 6). These datashow that SPP-catalyzed processing of PPI signal peptideis an essential step to generate an immunologically relevantepitope that is implicated as having a pathogenic role intype 1 diabetes through activation of b-cell-specific autor-eactive CD8 T cells (6,8).

Requirements for Proteasome, TAP, and ERAP in PPIProcessingWe previously showed that PPI15–24 processing and presen-tation by HLA-A0201 does not require proteasome cleavageor import into the ER via TAP (6). In contrast with PPI15–24-HLA-A0201, however, PPI3–11-HLA-A2402 presentation isTAP dependent. RNA interference (RNAi) inhibition of TAP1(94.56 2.3% mRNA knockdown) and TAP2 (95.462 0.3%mRNA knockdown) expression in K562-A24-PPI cells mark-edly reduces MIP-1b production (60.7% and 30.2%, respec-tively) by the PPI3–11-HLA-A2402-specific CD8 clone uponcoculture with TAP RNAi–treated K562-A24-PPI cells (Fig.4A). This is attributable to a reduction in surface density ofthe specific peptide HLA ligand PPI3–11-HLA-A2402 (TAPRNAi–treated K562-A24-PPI cells show only minimal reduc-tion in HLA class I expression; Supplementary Fig. 7).

Next we investigated whether aminopeptidases residingin the ER are involved in the processing and presentation ofPPI signal peptide epitopes. Knockdown of ERAP1 (87.2 60.5% mRNA knockdown) in K562-A24-PPI cells by RNAimarkedly reduced MIP-1b production (62.6%) by PPI3–11-HLA-A2402-specific clone 4C6, whereas no effect was seenin the presence of ERAP2 knockdown (73.7 6 6.1% mRNAknockdown; 8.0% change in MIP-1b production) (Fig. 4B).

Figure 3—Impact of SPP knockout in PPI- expressing cells upon HLA presentation of signal peptide–derived epitope presentation. The PPI15–24-specific, HLA-A0201-restricted T-cell clone was cultured at a 2:1 target-to-effector ratio, with different cell lines representing K562-A2-PPI(positive control; presents PPI15–24 in HLA-A0201); K562-A2 (negative control; lacks PPI); K562-A2-PPI-SPPmock (negative controls; two lines ofmock CRISPR-Cas9 manipulation; see RESEARCH DESIGN AND METHODS); and K562-A2-PPI-SPPko (test conditions; three independent lines ofCRISPR-Cas9 knockout of SPP). SPP knockout markedly reduces activation of the T-cell clone, as shown by the lower median fluorescenceintensity (MFI) (B) and the reduced percentage expression (A) of CD107a. Bars and error bars represent the mean and SEM, respectively, of fourindependent experiments.

diabetes.diabetesjournals.org Kronenberg-Versteeg and Associates 691

Similar effects were seen in target killing assays (data notshown). RNAi for ERAP1 and ERAP2 had only minimaleffects on total surface HLA class I expression (SupplementaryFig. 7), indicating that the ERAP1-mediated interference in

clone 4C6 activation is likely to be an effect on target cellsurface density of the specific PPI3–11-HLA-A2402 ligand.

TAP and ERAP1 Expression in Pancreas Samples FromSubjects With Type 1 DiabetesExpression of the processing proteins TAP1 and ERAP1—both of which we have shown to potentially contribute tothe generation of PPI epitopes that target b-cells for killingby PPI-specific cytotoxic CD8 T cells—were investigated inhuman pancreas recovered from healthy control subjectsand those with type 1 diabetes. TAP1 was present at lowlevels in the pancreatic islets of healthy control subjects andwas detected at similarly low levels in the insulin-deficientislets (IDIs) of patients with type 1 diabetes (Fig. 5A). Bycontrast, TAP1 was markedly upregulated in the ICIsof these patients (Fig. 5A). Colocalization studies revealedthat the elevation in TAP1 expression was most evidentin b-cells, although it was also increased in other islet cells.The immunostaining in six age-matched control subjectsand six subjects with type 1 diabetes was quantified bymeasuring the mean fluorescence intensity of TAP1 labelingin three islets from each control subject and in six isletsfrom each patient (three ICIs and three IDIs). This con-firmed significant elevation of TAP1 expression in ICIsfrom patients with type 1 diabetes compared with IDIs orcontrol islets (P , 0.001; Fig. 5B). Application of similarapproaches demonstrated that, as previously described (9),HLA-ABC was also markedly elevated in the ICIs of patientswith type 1 diabetes (Fig. 5C). Moreover, a strong positivecorrelation existed between the expression of TAP1 andHLA-ABC in ICIs (R2 = 0.519, P , 0.001; Fig. 5D).

ERAP1 was detected in the islets of healthy controlsubjects and in the islets of donors with type 1 diabetes(Supplementary Table 2). In contrast to TAP1, however,ERAP1 expression was not noticeably altered between theislets of control subjects and those of patients with type 1diabetes (Supplementary Fig. 8).

Model for PPI Signal Peptide Immune Processing andPresentationCollectively, these findings imply a model of PPI signalpeptide processing for HLA class I presentation in whichthe fate that follows intramembrane cleavage depends onlocation within the ER membrane and HLA binding po-tential (Fig. 6). On the one hand, N-terminal peptides maybe released into the cytoplasm, where they are dependenton TAP transport into the ER and ERAP1 trimming beforeloading into nascent HLAmolecules. As an alternative, distallygenerated ER luminal peptides are directly released into theER lumen, omitting the need for TAP, but these may requireN-terminal trimming by ERAP1 for optimal presentation.

DISCUSSION

In this study we extend our previous PPI epitope discoveryeffort to encompass new HLA class I alleles associated withrisk for/protection against type 1 diabetes. This reveals PPIsignal peptide as a frequent source of epitopes for multiple

Figure 4—Impact of siRNA knockdown of TAP and ERAP1 on epitopepresentation. Percentage mRNA knockdown in K562-A24-PPI cellstreated with siRNA knockdown for b2M, TAP1, and TAP2 genes (A,left panel) and the ERAP1 and ERAP2 genes (B, left panel), and theresulting effect on MIP-1b production by the HLA-A2402-restricted,PPI3–11-specific CD8 T-cell clone 4C6 upon coculture with the differ-ent cells lines (A and B, right panels). Scrambled siRNAs are used asthe control. DCt values of targeted knockdown were compared withDCt values of scrambled siRNA and expressed as the relative mRNAlevel of specific gene knockdown. Efficiency of relevant mRNA knock-down is high, resulting in reduced PPI3–11 presentation for TAP1, TAP2,and ERAP1, but not ERAP2. b2M served as a positive control. Barsand error bars represent the mean and SEM, respectively, of technicalduplicates from three independent experiments for mRNA knockdownand one representative example of MIP-1b production (two furtherrepeats are shown in Supplementary Fig. 9).

692 Preproinsulin Processing for Immune Recognition Diabetes Volume 67, April 2018

HLA-A and HLA-B alleles. Our previous finding of protea-some independence for HLA-A0201 PPI presentationimplied a noncanonical pathway to generate signal peptide–derived epitopes (6). This concept is extended in the currentstudy, in which we highlight the requirement for intramem-brane cleavage and indicate the importance of SPP in thisrole. We further show that after intramembrane cleavage,the pathway of peptide loading varies according to whethera signal peptide–derived epitope is N-terminal (requiringTAP) or COOH-terminal (independent of TAP). Loading isdictated by the HLA allele and optimized in the presence ofERAP1. These findings highlight a distinct set of processingprinciples for PPI (stoichiometry with translation and prox-imity to the site of HLA molecule synthesis and loading)that contrast with those in canonical endogenous antigenprocessing, which relies on proteasome degradation of ef-fete or damaged cytoplasmic proteins. Together with evi-dence that TAP expression is upregulated in relevant tissuesin the disease setting, this study offers important insightinto the molecular processes that are a key underpinningsof interactions between cytotoxic CD8 T cells and b-cellsduring disease pathogenesis.

One caveat in our study relates to using tumor cellstransfected with the INS gene and selected HLA allelesas “surrogate b-cells” to represent and understand the

endogenous pathway of PPI presentation as it may occur ina human b-cell in vivo. Whether this approach biases epi-tope discovery toward a specific PPI region remains open tointerpretation, although our finding of epitopes in the sig-nal peptide, B chain, and C-peptide across the different HLAmolecules suggests that this is unlikely. K562 cells predom-inantly express the constitutive proteasome (39), whilesome evidence indicates that the immunoproteasome canbe upregulated in b-cells under inflammatory conditions(40), and the balance of these two could influence the spec-trum of epitopes generated in human islets. Our approachis pragmatic in that obtaining sufficient, pure b-cells forsuch work presents a severe technical and logistical chal-lenge. We therefore elected to conduct epitope discoveryusing “surrogate b-cells” and then confirm findings usinghuman tissue. Our previous experience has been that epi-topes identified in this way for HLA-A0201 and HLA-A2402are faithful phenocopies of PPI presentation by b-cells. Inboth cases we were able to generate CD8 T-cell clones thatrecognize PPI epitopes presented by both the surrogateand donor-derived b-cells. Providing similar proof for theless common alleles remains challenging, however (41), al-though studies examining the frequency and antigen expe-rience of, for example, PPI5–12-B3906 and PPI5–14-B3801restricted T cells in the blood using peptide-HLA multimer

Figure 5—Expression of TAP and ERAP1 in human pancreas. Pancreas samples from six control individuals and six patients with type 1 diabetes(Supplementary Table 2) were stained for TAP1, HLA-ABC, and insulin. A: Images are representative (same microscope and camera settings) of18 islets from control case subjects, plus 18 IDIs and 18 ICIs from patients with type 1 diabetes. The mean fluorescence intensity (MFI) for eachantigen was determined within all imaged islets. B: The analysis reveals a significant increase in TAP1 expression in ICIs from individuals withtype 1 diabetes (T1D) compared with IDIs or control islets. C: Consistent with previous reports, a similar change was seen with HLA-ABCexpression. D: Comparing TAP1 and HLA-ABC expression within the same islets shows a positive correlation. Bars represent the mean6 SEM;one-way ANOVA and a subsequent post hoc Tukey multiple comparison test were used to determine significance between groups (***P ,0.001).

diabetes.diabetesjournals.org Kronenberg-Versteeg and Associates 693

technology, will provide supportive evidence of a pathogenicrole.

One of our findings is the identification of a noncanonicalpathway of endogenous antigen processing for an epitoperelevant in the recognition of b-cells by the immune system.Central to this is an intramembrane cleavage step. Our studiesconducted in vitro using microsomes indicate that SPP iscapable of cleaving the PPI signal peptide, which is releasedfrom the ER membrane by SPP-catalyzed cleavage. CRISPR-Cas9-mediated SPP knockout indicates that this enzyme isalso rate limiting for epitope generation in vivo. Using thisand the (Z-LL)2-ketone inhibitor approaches we are able toargue that SPP knockout affects the processing of signal pep-tide, leading to reduced presentation of PPI15–24 and, in turn,reduced activation of the specific T-cell clone. Overall, thisseems to be compelling evidence for a direct role of SPP in theprocessing of PPI signal peptide in a manner that is critical tothe generation of this important epitope.

SPP belongs to the family of GxGD intramembraneproteases including presenilin/g-secretase and related SPP-like proteases (37). SPP residing in the ER cleaves varioussignal peptides and type II membrane proteins in a numberof physiological settings (42), including the generation ofa regulatory peptide from the HLA-A0301 signal peptide

that is required for HLA-E-mediated immunosurveillance(22). This is not the first study to draw attention to thepotential requirement for intramembrane cleavage of anautoantigen in type 1 diabetes. Using algorithms to predictHLA-A0201 epitopes of islet amyloid polypeptide (IAPP),previous studies have shown CD8 T-cell reactivity to IAPP9–17and IAPP5–13 in the blood (43,44) and, in the case of IAPP5–13,also in islets of patients with type 1 diabetes (8). Whetherthese peptides are naturally processed and presented byb-cells remains unclear, however, and if confirmed for bothit would imply a complex processing pathway, because theC-terminus of IAPP5–13 overlaps and extends into theN-terminus of IAPP9–17.

Signal sequences are essential for protein targeting tothe secretory pathway via the ER (45,46), where, after in-sertion into the protein-conduction translocation channel,they are cleaved from the preprotein by signal peptidase(47); in the case of PPI15–24-HLA-A0201, this also generatesthe epitope C-terminus. Signal peptides spanning the ERmembrane require cleavage by SPP for efficient disposal(19,20). No consensus SPP cleavage motif has been iden-tified, beyond a strong preference for helix-destabilizingresidues in the membrane-spanning region (21,48,49), whichare likely to be the proline and two alanine residues at

Figure 6—Processing of the PPI signal peptide. Model of PPI signal peptide processing in which N-terminal peptides may be released by SPPinto the cytoplasm, where they depend on TAP transport into the ER and ERAP1 trimming before loading into nascent HLA (illustrated by theexample of PPI3–11 loading into HLA-A2402). By contrast, ER luminal peptides are released directly into the ER lumen and do not require TAPtransport but equally require trimming by ERAP1 for optimal presentation (illustrated by the example of PPI15–24 loading into HLA-A0201). TCR,T-cell receptor.

694 Preproinsulin Processing for Immune Recognition Diabetes Volume 67, April 2018

positions 9, 12, and 15, respectively. Indeed, in our studies,mutation of any one of these has a mild effect on signalpeptide cleavage; any two mutations together givesa moderate phenotype; and when all three are mutated,SPP is unable to cleave, indicating that these are the helix-destabilizing residues in PPI. Once liberated, the SPP-cleaved N-terminal signal peptide fragment gets accessto the cytosol, where it is either trimmed by the protea-some or directly loaded onto TAP and the MHC peptide-loading complex. In the case of PPI3–11-HLA-A2402, weshow a requirement for TAP, whereas PPI15–24-HLA-A0201requires neither proteasome nor TAP (6), leading us tospeculate that “untapped” loading of a COOH-terminal sig-nal peptide fragment makes use of other chaperones, suchas TAPBP.

We considered it important to try to link findings inrelation to mechanisms of b-cell antigen presentation ob-tained in vitro with the tissue inflammatory process takingplace in subjects with type 1 diabetes. Our analysis showsthat TAP expression is significantly increased in ICIs andcorrelated in its hyperexpression with HLA class I moleculeexpression. That the b-cell contributes to its own destruc-tion in various ways has become a much-vaunted metaphor.In the particular setting of PPI presentation to cytotoxicT cells, it would seem that hyperexpression of TAP is an-other example, certainly for HLA alleles such as HLA-A2402. In this way it may complement the effects ofhyperexpression of HLA class I and hyperglycemic condi-tions (which upregulate PPI presentation [6]) in enhancing b-cellcytotoxicity under the inflammatory milieu that prevails inislets of Langerhans in patients with type 1 diabetes.

Acknowledgments. The authors are grateful to John Todd, University ofOxford, for providing the typed HLA class I cell lines and Dr. Pierre Vantourout, King’sCollege London, for providing the pLG2C vector.Funding. This study was supported by the National Institute for Health ResearchBiomedical Research Centre at Guy’s and St Thomas’ Hospital Trusts and King’sCollege London (through a PhD studentship to D.K.-V.), JDRF (Centre Grant no.1-2007-1803 to M.P. and a Career Development Award [5-CDA-2014-221-A-N]to S.J.R.), Diabetes UK (project grant 15/0005156 to N.G.M. and S.J.R.), and theDeutsche Forschungsgemeinschaft (project grant FOR2290-TP1 to M.K.L.).Duality of Interest. No potential conflicts of interest relevant to this articlewere reported.Author Contributions. D.K.-V. and M.E. designed and performed experi-ments, analyzed data, conceived ideas, oversaw research, and wrote the manuscript.M.A.R., A.d.R., B.H., and N.Y. performed experiments. P.A.v.V., S.J.R., N.G.M., M.K.L.,and M.P. conceived ideas, oversaw research, and wrote the manuscript. M.P. is theguarantor of this work and, as such, had full access to all the data in the study andtakes responsibility for the integrity of the data and the accuracy of the dataanalysis.Prior Presentation. Parts of this study were presented at the Immunology ofDiabetes 15th International Conference, San Francisco, CA, 19–23 January 2017.

References1. Atkinson MA. The pathogenesis and natural history of type 1 diabetes. ColdSpring Harb Perspect Med 2012;2. pii:a0076412. DiLorenzo TP, Serreze DV. The good turned ugly: immunopathogenic basis fordiabetogenic CD8+ T cells in NOD mice. Immunol Rev 2005;204:250–263

3. Nejentsev S, Howson JM, Walker NM, et al.; Wellcome Trust Case ControlConsortium. Localization of type 1 diabetes susceptibility to the MHC class I genesHLA-B and HLA-A. Nature 2007;450:887–8924. Kronenberg D, Knight RR, Estorninho M, et al. Circulating preproinsulin signalpeptide-specific CD8 T cells restricted by the susceptibility molecule HLA-A24are expanded at onset of type 1 diabetes and kill b-cells. Diabetes 2012;61:1752–17595. Luce S, Lemonnier F, Briand JP, et al. Single insulin-specific CD8+ T cellsshow characteristic gene expression profiles in human type 1 diabetes. Diabetes2011;60:3289–32996. Skowera A, Ellis RJ, Varela-Calviño R, et al. CTLs are targeted to kill beta cellsin patients with type 1 diabetes through recognition of a glucose-regulated pre-proinsulin epitope [published correction appears in J Clin Invest 2009;119:2844]. JClin Invest 2008;118:3390–3402

7. Willcox A, Richardson SJ, Bone AJ, Foulis AK, Morgan NG. Analysis of isletinflammation in human type 1 diabetes. Clin Exp Immunol 2009;155:173–1818. Coppieters KT, Dotta F, Amirian N, et al. Demonstration of islet-autoreactiveCD8 T cells in insulitic lesions from recent onset and long-term type 1 diabetespatients. J Exp Med 2012;209:51–60

9. Richardson SJ, Rodriguez-Calvo T, Gerling IC, et al. Islet cell hyperexpression ofHLA class I antigens: a defining feature in type 1 diabetes. Diabetologia 2016;59:2448–245810. Rigby MR, DiMeglio LA, Rendell MS, et al.; T1DAL Study Team. Targeting ofmemory T cells with alefacept in new-onset type 1 diabetes (T1DAL study):12 month results of a randomised, double-blind, placebo-controlled phase 2 trial.Lancet Diabetes Endocrinol 2013;1:284–29411. Howson JM, Walker NM, Clayton D, Todd JA; Type 1 Diabetes GeneticsConsortium. Confirmation of HLA class II independent type 1 diabetes associations inthe major histocompatibility complex including HLA-B and HLA-A. Diabetes ObesMetab 2009;11(Suppl. 1):31–4512. Valdes AM, Erlich HA, Noble JA. Human leukocyte antigen class I B and C locicontribute to type 1 diabetes (T1D) susceptibility and age at T1D onset. Hum Im-munol 2005;66:301–31313. Townsend A, Bastin J, Gould K, et al. Defective presentation to class I-restrictedcytotoxic T lymphocytes in vaccinia-infected cells is overcome by enhanced deg-radation of antigen. J Exp Med 1988;168:1211–122414. Chang SC, Momburg F, Bhutani N, Goldberg AL. The ER aminopeptidase,ERAP1, trims precursors to lengths of MHC class I peptides by a “molecular ruler”mechanism. Proc Natl Acad Sci U S A 2005;102:17107–1711215. Spies T, Bresnahan M, Bahram S, et al. A gene in the human major histo-compatibility complex class II region controlling the class I antigen presentationpathway. Nature 1990;348:744–74716. Cresswell P. Intracellular surveillance: controlling the assembly of MHC classI-peptide complexes. Traffic 2000;1:301–30517. Saric T, Chang SC, Hattori A, et al. An IFN-gamma-induced aminopeptidase inthe ER, ERAP1, trims precursors to MHC class I-presented peptides. Nat Immunol2002;3:1169–117618. Neefjes J, Jongsma ML, Paul P, Bakke O. Towards a systems understanding ofMHC class I and MHC class II antigen presentation. Nat Rev Immunol 2011;11:823–83619. Martoglio B, Dobberstein B. Signal sequences: more than just greasy peptides.Trends Cell Biol 1998;8:410–41520. Weihofen A, Binns K, Lemberg MK, Ashman K, Martoglio B. Identification ofsignal peptide peptidase, a presenilin-type aspartic protease. Science 2002;296:2215–221821. Lemberg MK, Martoglio B. Requirements for signal peptide peptidase-catalyzedintramembrane proteolysis. Mol Cell 2002;10:735–74422. Lemberg MK, Bland FA, Weihofen A, Braud VM, Martoglio B. Intramembraneproteolysis of signal peptides: an essential step in the generation of HLA-E epitopes.J Immunol 2001;167:6441–644623. Wei ML, Cresswell P. HLA-A2 molecules in an antigen-processing mutant cellcontain signal sequence-derived peptides. Nature 1992;356:443–446

diabetes.diabetesjournals.org Kronenberg-Versteeg and Associates 695

24. Henderson RA, Michel H, Sakaguchi K, et al. HLA-A2.1-associated peptidesfrom a mutant cell line: a second pathway of antigen presentation. Science 1992;255:1264–126625. El Hage F, Stroobant V, Vergnon I, et al. Preprocalcitonin signal peptidegenerates a cytotoxic T lymphocyte-defined tumor epitope processed by a proteasome-independent pathway. Proc Natl Acad Sci U S A 2008;105:10119–1012426. Bland FA, Lemberg MK, McMichael AJ, Martoglio B, Braud VM. Requirement ofthe proteasome for the trimming of signal peptide-derived epitopes presented by thenonclassical major histocompatibility complex class I molecule HLA-E. J Biol Chem2003;278:33747–3375227. Meiring HD, van der Heeft E, ten Hove GJ, de Jong APJM. Nanoscale LC–MS(n):technical design and applications to peptide and protein analysis. J Sep Sci 2002;25:557–56828. Stepniak D, Wiesner M, de Ru AH, et al. Large-scale characterization of naturalligands explains the unique gluten-binding properties of HLA-DQ2. J Immunol 2008;180:3268–327829. Knight RR, Kronenberg D, Zhao M, et al. Human b-cell killing by autoreactivepreproinsulin-specific CD8 T cells is predominantly granule-mediated with the po-tency dependent upon T-cell receptor avidity. Diabetes 2013;62:205–21330. Nilsson IM, von Heijne G. Determination of the distance between the oligo-saccharyltransferase active site and the endoplasmic reticulum membrane. J BiolChem 1993;268:5798–580131. Martoglio B, Hauser S, Dobberstein B. Cotranslational translocation of proteinsinto microsomes derived from the rough endoplasmic reticulum of mammalian cells.In Cell Biology: A Laboratory Handbook. Celis JE, Ed. San Diego, CA, Academic Press,1998, p. 265–27432. Weihofen A, Lemberg MK, Ploegh HL, Bogyo M, Martoglio B. Release of signalpeptide fragments into the cytosol requires cleavage in the transmembrane region bya protease activity that is specifically blocked by a novel cysteine protease inhibitor. JBiol Chem 2000;275:30951–3095633. Wiltfang J, Smirnov A, Schnierstein B, et al. Improved electrophoretic separationand immunoblotting of beta-amyloid (A beta) peptides 1-40, 1-42, and 1-43.Electrophoresis 1997;18:527–53234. Lemberg MK, Martoglio B. Analysis of polypeptides by sodium dodecyl sulfate-polyacrylamide gel electrophoresis alongside in vitro-generated reference peptides.Anal Biochem 2003;319:327–33135. Eichmann M, de Ru A, van Veelen PA, Peakman M, Kronenberg-Versteeg D.Identification and characterisation of peptide binding motifs of six autoimmune

disease-associated human leukocyte antigen-class I molecules including HLA-B*39:06. Tissue Antigens 2014;84:378–38836. Kovjazin R, Volovitz I, Daon Y, et al. Signal peptides and trans-membrane re-gions are broadly immunogenic and have high CD8+ T cell epitope densities: im-plications for vaccine development. Mol Immunol 2011;48:1009–101837. Voss M, Schröder B, Fluhrer R. Mechanism, specificity, and physiology of signalpeptide peptidase (SPP) and SPP-like proteases. Biochim Biophys Acta 2013;1828:2828–283938. Li SC, Deber CM. A measure of helical propensity for amino acids in membraneenvironments. Nat Struct Biol 1994;1:368–37339. Anderson KS, Zeng W, Sasada T, et al. Impaired tumor antigen processing byimmunoproteasome-expressing CD40-activated B cells and dendritic cells. CancerImmunol Immunother 2011;60:857–86740. Lundh M, Bugliani M, Dahlby T, et al. The immunoproteasome is inducedby cytokines and regulates apoptosis in human islets. J Endocrinol 2017;233:369–37941. Baschal EE, Baker PR, Eyring KR, Siebert JC, Jasinski JM, Eisenbarth GS. TheHLA-B 3906 allele imparts a high risk of diabetes only on specific HLA-DR/DQhaplotypes. Diabetologia 2011;54:1702–170942. Avci D, Lemberg MK. Clipping or extracting: two ways to membrane proteindegradation. Trends Cell Biol 2015;25:611–62243. Panagiotopoulos C, Qin H, Tan R, Verchere CB. Identification of a beta-cell-specific HLA class I restricted epitope in type 1 diabetes. Diabetes 2003;52:2647–265144. Ouyang Q, Standifer NE, Qin H, et al. Recognition of HLA class I-restricted beta-cell epitopes in type 1 diabetes. Diabetes 2006;55:3068–307445. Johnson AE, van Waes MA. The translocon: a dynamic gateway at the ERmembrane. Annu Rev Cell Dev Biol 1999;15:799–84246. Rapoport TA, Rolls MM, Jungnickel B. Approaching the mechanism of proteintransport across the ER membrane. Curr Opin Cell Biol 1996;8:499–50447. Dalbey RE, Lively MO, Bron S, van Dijl JM. The chemistry and enzymology ofthe type I signal peptidases. Protein Sci 1997;6:1129–113848. Fluhrer R, Martin L, Klier B, et al. The a-helical content of the transmembranedomain of the British dementia protein-2 (Bri2) determines its processing by signalpeptide peptidase-like 2b (SPPL2b). J Biol Chem 2012;287:5156–516349. Lemberg MK, Martoglio B. On the mechanism of SPP-catalysed intramembraneproteolysis; conformational control of peptide bond hydrolysis in the plane of themembrane. FEBS Lett 2004;564:213–218

696 Preproinsulin Processing for Immune Recognition Diabetes Volume 67, April 2018