Graft versus self (GvS) against T-cell autoantigens is a mechanism of graft–host interaction

Nora Mirzaa,b, Manfred Zierhutc, Andreas Kornd, Antje Bornemanne, Wichard Vogela, Barbara Schmid-Horchf, Wolfgang A. Bethgea, Stefan Stevanovicb, Helmut R. Saliha,g, Lothar Kanza, Hans-Georg Rammenseeb, and Sebastian P. Haena,b,1

aAbteilung II fuer Onkologie, Haematologie, Immunologie, Rheumatologie, und Pulmologie, Medizinische Universitaetsklinik, D-72076 Tuebingen, Germany; bAbteilung Immunologie, Interfakultaeres Institut fuer Zellbiologie, D-72076 Tuebingen, Germany; cUniversitaetsaugenklinik, D-72076 Tuebingen, Germany; dKlinik fuer diagnostische und interventionelle Neuroradiologie, D-72076 Tuebingen, Germany; eInstitut fuer Neuropathologie, D-72076 Tuebingen, Germany; fZentrum fuer klinische Transfusionsmedizin, D-72076 Tuebingen, Germany; and gDepartment for Internal Medicine II, Clinical Collaboration Unit Translational Immunology, German Cancer Consortium and German Cancer Research Center, Partner site Tuebingen, D-72076 Tuebingen, Germany

Edited by Harvey Cantor, Dana-Farber Cancer Institute, Boston, MA, and approved October 14, 2016 (received for review June 6, 2016)

Graft-versus-host disease (GVHD) represents the major nonrelapse cone-rod degeneration (20, 21) and Leber’s congenital amaurosis complication of allogeneic hematopoietic cell transplantation. (22). The guanylate cyclase activating 1 and 2 (GCAP1, Although rare, the CNS and the eye can be affected. In this study, GUCA1A;GCAP2,geneGUCA1B) are involved in the manifestation in the retina as part of the CNS and T-cell epitopes negative regulation of the retGC and are important for photore- recognized by the allogeneic T cells were evaluated. In 2 of 6 patients ceptor recovery (23, 24). Mutations in the GUCA1A gene have with posttransplantation retina diseases and 6 of 22 patients without been associated with inherited cone, cone-rod and macula dystrophy ocular symptoms, antigen-specific T-cell responses against retina- (24, 25). Fourth, the retinoid binding (RBP3) is important specific epitopes were observed. No genetic differences between for the transport of retinoids between the retinal pigment epi- donor and recipient could be identified indicating T-cell activation thelium and photoreceptors. Mutations of the RBP3 protein are against self-antigens (graft versus self). Transplantation of a pre- associated with retinitis pigmentosa (26). existing immunity and cross-reactivity with ubiquitous epitopes In this study, we characterize the development of donor T-cell INFLAMMATION

was excluded in family donors and healthy individuals. In sum- responses against MHC-restricted retina protein-derived peptides. IMMUNOLOGY AND mary, an immunological reaction against retina cells represents a mechanism of graft-versus-host interaction following hematopoi- Results etic cell transplantation. Patients. The first group (Table S1, left column) comprised patients with diseases of the PS after HCT. PS diagnoses were allogeneic hematopoietic cell transplantation | graft-versus-host disease | optic atrophy of unknown origin (n = 2), in one case combined autoimmunity | T-cell epitope | retina with a selective cone dysfunction, optic neuritis (n = 2), anemic retinopathy (n = 1), and CMV retinitis (n = 1). The median – ematopoietic cell transplantation (HCT) can be the only onsetofPSdiagnosesoccurredat9moafterHCT(range3 Hcurative treatment option for patients with hematologic 25 mo). The second group (Table S1, right column) comprised malignancies. Besides direct treatment-associated complications, 22 consecutive patients recruited before allogeneic HCT. For graft-versus-host disease (GVHD) is the major complication (1). both groups, characteristics of individual patients are provided Today, graft–host interaction is thought to be mediated by two in Table S2. pathomechanisms. (i) Allogeneic T cells recognize differences be- tween donor and recipient in HLAs and the respective expressed Significance self-peptides (2, 3). (ii) Genetic polymorphisms, especially non- synonymous SNPs, lead to expression of proteins with alternate As the mechanism of graft-versus-host disease (GVHD) after amino acid sequences whose fragments can be presented on allogeneic hematopoietic cell transplantation (HCT), recogni- MHC molecules functioning as minor histocompatibility anti- tion of the recipient’s body by donor immune cells was pre- gens (miHAGs) (4, 5). viously believed to be based on genetic and immunological Although not one of the main manifestation organs, GVHD differences between donor and recipient. However, evidence may also affect the CNS, even if thought to occur rarely and – in murine models and in autologous HCT has shown that also evidence being limited to single cases (6 10). Because GVHD of autoimmunity contributes to GVHD. In this study, we show the the CNS is difficult to distinguish from other complications such development of auto-reactivity after allogeneic HCT as relapse, infections, and toxicity related to therapy, a thorough and characterize the specific self-epitopes of T cells that may characterization of CNS GVHD is challenging (7, 11). contribute to mediation of GVHD. Such autoantigens have The retina constitutes the only part of the CNS that can be never been characterized before. These observations contrib- examined directly. Although ocular involvement of the anterior segment such as keratoconjunctivitis sicca, corneal epitheliopathy, ute to a better understanding of the immune responses that and pseudomembranous conjunctivitis is observed frequently (12, are activated following hematopoietic cell transplantation. 13), posterior segment (PS) involvement is rare and mostly attrib- Author contributions: L.K., H.-G.R., and S.P.H. designed research; N.M., M.Z., A.B., and S.P.H. uted to the toxicity of irradiation or immunosuppression including performed research; M.Z., A.K., A.B., B.S.-H., and S.S. contributed new reagents/analytic diseases like central serous chorioretinopathy (14) and ischemic tools; N.M., M.Z., A.K., A.B., W.V., W.A.B., S.S., H.R.S., L.K., H.-G.R., and S.P.H. analyzed data; retinopathy (15) and can manifest with cotton wool spots, retinal N.M., M.Z., W.A.B., S.S., H.R.S., L.K., H.-G.R., and S.P.H. wrote the paper; and S.P.H. bleeding, and edema (16). Only a very limited body of evidence has obtained funding. described retina manifestations linked to GVHD (17, 18). The authors declare no conflict of interest. Several proteins have been described to be involved in retinal This article is a PNAS Direct Submission. inflammation and degeneration comprising the membrane- 1To whom correspondence should be addressed. Email: [email protected] bound retinal guanylate cyclase (retGC; gene GUCY2D), an tuebingen.de. enzyme involved in phototransduction and predominantly expressed This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. in the cones (19). Mutations cause autosomal dominant inherited 1073/pnas.1609118113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1609118113 PNAS | November 29, 2016 | vol. 113 | no. 48 | 13827–13832 Downloaded by guest on September 27, 2021 Approach of the Study and Peptide Identification. The retina-spe- A cific target proteins (retGC, GCAP1, GCAP2, and RBP3) were identified using the swissprot (www..org), ensembl (useast. ensembl.org/index.html), and geoprofiles (www.ncbi.nlm.nih.gov/ geoprofiles) databases. Candidate T-cell epitopes were identified using two approaches. (i) Peptide pairs for the patient HLA type were predicted based on published SNP using the internet based databases SYFPEITHI (www.syfpeithi.com) and EpiToolKit (www. epitoolkit.de) (MHC-I; Table S3) or designed as 17-mers with the SNP positioned in the middle of the peptide (MHC-II; Table S3). Gene sequences of patient and donor were confirmed in Sanger sequencing. (ii) Recipient and donor DNA was sequenced by Sanger sequencing. Here, peptides were predicted on the basis B of identified SNPs (Fig. S1). Variant amino acids are printed in bold throughout the article and in red in the figures.

Retina-Specific T-Cell Responses in Patients with PS Manifestations. Retina-antigen specific T cells were detected in two of six patients (Fig. 1) with PS diseases. Patient 1 [diagnosis of acute lympho- blastic leukemia (ALL)] experienced progressive visual loss 14 mo after HCT (at 14 mo: 20/30 in both eyes; at 25 mo: finger counting in both eyes). Optic atrophy and selective dysfunction of the cones with normal rod function was diagnosed using fundus photogra- phy (Fig. 1A) and electroretinography (ERG). Stereotactic biopsy + of a ventricular lesion revealed a CD8 vasculitis (Fig. 1B). In- fections and disease relapse were excluded in evaluation of cere- C brospinal fluid (no detection of pathogens or malignant cells) and biopsy specimens. In a sample of peripheral blood mononuclear cells (PBMCs) harvested 40 mo after HCT (Fig. 1C), weak T-cell responses could be detected by IFN-γ enzyme linked immunospot D (ELISpot) after stimulation with two overlapping HLA B*0702- restricted retGC-derived peptides (GUCY2D 46B: QPPALSSVFT, GUCY2D 47B: PPALSSVFT, mean: 19.5 spots/250,000 PBMCs, 39-fold of negative control). The T-cell response was not reflected by a difference in the DNA sequence between donor and recipient because the donor was a carrier of the heterozygous SNP leading to amino acid exchange from alanine to serine at position 52 of the GUCY2D (A52S), whereas the recipient carried the major allele A/S A coding for alanine (Fig. 1D). E In a second patient [patient 5, diagnosis acute myeloid leu- kemia (AML)] with a CMV retinitis diagnosed by vitreal biopsy (fundus photograph in Fig. 1E), a strong T-cell response was detected after stimulation with MHC II-restricted retGC-derived peptides (GUCY2D 52A: LLQPPALSAVFTVGVLG, GUCY2D 52B: LLQPPALSSVFTVGVLG, 546.5 spots/500,000 PBMCs, 121-fold compared with negative control) at 24 mo after HCT (Fig. 1F). DNA sequencing did not reveal a SNP, indicating recognition of self-antigens (Fig. 1G).

T-Cell Responses in Patients Irrespective of Ocular Symptoms. In two F patients (Fig. 2: patient 7; Fig. S2: patient 9), strong T-cell re- sponses directed against a predicted HLA A*0301-restricted GCAP1-derived self-peptide (GUCA1A 47A: NLSPSASQY) were detected by IFN-γ ELISpot (Fig. 2A: patient 7; up to 659 G spots/500,000 PBMCs, 220-fold of negative control; Fig. S2A: patient 9; up to 300 spots/500,000 PBMCs, 35-fold of negative control) and corresponding intracellular cytokine staining (ICS). No IFN-γ secretion was observed when PBMCs were stimu- lated with the corresponding variant peptide (GUCA1A 47B: + NLSLSASQY). In particular, the ICS revealed a CD4 T cell- mediated TNF and IFN-γ response (Fig. 2B and Fig. S2B). Of note, no T-cell responses were observed in PBMC samples A A

Fig. 1. T-cell responses after HCT in patients with PS diseases. (A–D) Patient 1: fundus photography (A) showing optic nerve atrophy with very narrow recipient DNA (GUCY2D exon 2). (E–G) Patient 5: fundus photography (E) vessels. (B) MRI scan (Left) of the brain revealed a periventricular lesion revealing a massive, centrally located CMV retinitis with detection of vitre- where a stereotactic biopsy was taken. Histologic workup (Right) including ous cells (cloudiness of the picture). T-cell responses were evaluated (F) after + immunohistochemistry showed a CD8 atypical vasculitis without any evi- stimulation of 500,000 prestimulated (12 d) PBMCs per well with retGC-de- dence of meningeal disease relapse. (C) T-cell response after stimulation of rived MHC class II peptides (2.5 μg/mL) detected by IFN-γ ELISpot and se- 250,000 prestimulated (12 d) PBMCs per well with retGC-derived peptides quencing (G) of donor and recipient DNA (GUCY2D exon 2). CD, cluster of (1 μg/mL) detected by IFN-γ ELISpot and sequencing (D) of donor and differentiation; OD, oculus dexter; OS, oculus sinister.

13828 | www.pnas.org/cgi/doi/10.1073/pnas.1609118113 Mirza et al. Downloaded by guest on September 27, 2021 A harvested before HCT (Fig. 2A and Fig. S2A). T-cell responses remained detectable up to 14 mo after HCT. DNA sequencing revealed no difference between donor and recipient (Fig. 2C and Fig. S2C). Hence, the observed T-cell response was specific for self-peptides. + In PBMC samples of patient 7, an additional CD4 T-cell re- sponse was observed after stimulation with a HLA A*0101-predicted GCAP2-derived self-peptide (GUCA1B 133A: QTEQGQLLT). Patient 11 displayed a T-cell response after stimulation with a peptide pool consisting of the GCAP2-derived peptides GUCA1B 133A (QTEQGQLLT) and GUCA1B 133B (QTEQDQLLT) 7 mo after HCT in ELISpot (73 spots/500,000 PBMCs, sevenfold of negative control; Fig. S2 D and E). Donor and recipient shared the same GUCA1B gene sequence, indicating recognition of a self-peptide (Fig. S2F). The recognized peptide sequence could, however, not be further discriminated due to limited availability of patient material. In PBMC samples of patient 12, strong T-cell responses against HLA A*0201-restricted retGC- (13-fold of negative control, 15-fold of donor sample; Fig. S3 A–C), GCAP1- and GCAP2- (21-fold of negative control, 23-fold of donor sample) derived peptides were detected 5 mo after HCT, whereas no IFN-γ secretion was observed before HCT. One month later, the response intensity decreased, and a remaining positive T-cell B response was observed after stimulation with the GCAP2-derived peptides (GUCA1B 131A: ELQTEQGQLL, GUCA1B 131B: ELQTEQDQLL; Fig. S3 A–C). Due to limited cell numbers, other time points after HCT could not be separately monitored

by ICS. In this patient, sequencing of donor and recipient DNA INFLAMMATION revealed a heterozygous missense SNP for the recipient in exon IMMUNOLOGY AND 2 of the GUCY2D gene (A52S), whereas the donor was a ho- mozygous carrier of the major allele (Fig. S3D). This sequence difference in the GUCY2D gene could result in recognition of the alloantigen. However, both peptides were recognized reflecting cross-reactivity against either the allo- or the autoantigen. Con- sidering the T-cell responses against the GCAP-2–derived pep- tides in this patient and the other observations in this study, the reactivity is not likely to be based on gene sequence differences. Moreover, analysis of the donor sample revealed no detectable T-cell response after stimulation with the HLA A*0201-restricted peptides (Fig. S3 A and B). In patient 20, weak T-cell reactivity was observed by ELISpot against a pool consisting of four retGC-derived MHC class II-restricted peptides (GUCY2D 52A: LLQPPALSAVFTVGVLG; GUCY2D 52B: LLQPPALSSVFTVGVLG; GUCY2D 782A: DQAPVECILLMKQCWAE; GUCY2D 782B: DQAPVE- CIHLMKQCWAE) 5 mo after HCT (mean: 29 spots/250,000 PBMCs, 4.38-fold of negative control). In this patient, the rec- C ognized epitopes could not be further assessed due to limited availability of T cells. The response was no longer detectable 14 and 16 mo after HCT. No response was detected before HCT (Fig. S4 A and B). DNA sequencing revealed no genetic difference between donor and recipient (Fig. S4C). In patient 23, a homozygous SNP in exon 2 of the GUCY2D P P gene (A52S) was identified for recipient and donor (Fig. S5E). No T-cell response was detected in a sample of this patient harvested before HCT. The observed response was first (4 and

available due to the myelodysplastic syndrome. The bar graph presents changes in the T-cell reactivity (absolute spot counts in ELISpot) normalized G G on the negative control (incubation with irrelevant peptide) over time. (B) ICS revealed T-cell activation indicated by TNF (far left and third panel Fig. 2. T-cell responses after HCT in patient 7. CD4+ T-cell responses after from left) and IFN-γ (second panel from left and far right panel) production stimulation of 500,000 prestimulated (12 d) PBMCs per well with the GCAP- on incubation with the A variant peptides (panels first row), but not with the 1– and the GCAP-2–derived MHC class I predicted peptide detected by (A) variant peptide with difference in one amino acid (panels second row). The + IFN-γ ELISpot (peptides at 1 μg/mL) and (B) ICS (peptides at 10 μg/mL). (A)T response was mediated by CD4 T cells. As controls, incubation with HIV- + cells were analyzed over a period of 17 mo (before HCT until 17 mo after derived epitopes are presented in the lower two rows for CD4 (third row) + HCT). IFN-γ ELISpot (Upper) revealed changing of the T-cell reactivity against and CD8 (fourth row) responses. (C) Sequencing of the GUCA1A (Upper) one peptide but not against the corresponding variant peptide, which were and GUCA1B (Lower) for donor (Left) and recipient (Right) showing not detectable before HCT. Before HCT, only very limited PBMC counts were the same genetic sequence and, hence, revealing no SNP.

Mirza et al. PNAS | November 29, 2016 | vol. 113 | no. 48 | 13829 Downloaded by guest on September 27, 2021 7 mo after HCT) directed against a non–self-peptide (GUCY2D A 52A: LLQPPALSAVFTVGVLG); later a strong response against both self- and non–self-peptide (665.5 and 657.5 spots/ 500,000 PBMCs, 22-fold of negative control, 46-fold of donor sample; Fig. S5 A, C, and D) was observed. Because donor and recipient were both homozygous carriers of the minor allele, the T-cell responses against the non–self-peptide reflects cross- reactivity against an irrelevant peptide expressed in neither the B patient nor the donor, but likely reflects reactivity against the self-epitope. The observed response was no longer detectable 12 mo after HCT. Of note, this decrease was not correlated to any clinical parameters, e.g., escalation of immunosuppression. + AlimitedIFN-γ production by CD4 T cells (0.0381%) was detected after stimulation with the non–self-peptide 12 mo after HCT (Fig. S5B). No preexisting T-cell responses were observed in ELISpot after stimulation with the GUCY2D 52A and B CD peptides in donor PBMCs (Fig. S5 A and C). In total, T-cell responses against retina-specific self- and non– self-peptides were observed in 6 of 22 patients (27%). Two patients (7, 9) showed T-cell responses directed against self-peptides, whereas in samples of two other patients (12, 23), T cells rec- ognized both self- and non–self-peptides. In two patients (11, 20), T-cell responses could not be further discriminated. Of note, Fig. 3. In vitro priming of peptide-specific CD8+ T cells. DCs based in vitro none of the observed T-cell responses was associated with a priming experiments were performed for all MHC class I peptides and defined genetic polymorphism. measured in ICS. Two exemplified peptide pairs are shown. (A) GUCA1A 144: Weak T-cell reactivity could be induced against the B-variant (DVNGDGEFSL) Clinical Factors Influencing the Development of Autoreactive T Cells. but not against the A-variant peptide (DVNGDGELSL) as indicated by TNF Next, we analyzed patient records and laboratory findings to and IFN-γ production. (B) GUCY2D 18: Very strong T-cell reactivity could be study whether patient- and/or transplantation-related parame- induced against both peptide variants as indicated by TNF and IFN-γ pro- ters could influence the development of the observed antigen- duction detected by flow cytometry. (C) Summary of identification of retina- specific T cells (Tables S1 and S2). Patients with PS diseases specific T-cell epitopes: antigen-specific T cells could be detected in patients were more likely to have received total body irradiation for only (n = 2), patients and in vitro priming (n = 6), and in vitro priming only conditioning (83% vs. 27%; positive predictive value 1; 95%, CI (n = 30). Seventeen peptides could not be confirmed as epitopes. (D) Clas- 0.19–1). Due to the small sample size, this difference was not sification of the intensity of CD8 T-cell responses into very weak >0.05–0,1%, statistical significant (P = 0.25). If any, there was a slight dif- weak >0.1–0.5%, intermediate >0.5–1%, strong >1–3%, and very strong ference in the occurrence of extensive chronic GVHD (66% vs. T-cell responses >3% (percentage is referred to all CD4-negative gated cells). 27%, P = 0.25; positive predictive value, 0.5; 95% CI, 0.03– 0.97). Also, appearance of acute GVHD, GVHD prophylaxis, and graft composition did not contribute to the development In Vitro Priming of Antigen-Specific T Cells Against Retina Epitopes. To investigate the immunogenicity of the predicted peptides, of PS diseases. No timely association with acute or chronic + GVHD or between PS diseases and other GVHD manifes- in vitro priming of HD CD8 T cells was performed for all tations was observed. Overall, no clear parameters for both the MHC-I peptides using at least three different HDs for each development of PS symptoms and autoreactive T cells could peptide pair. Thereby, 36 of 55 peptides (Fig. 3 and Table S3, be identified. list) were confirmed as T-cell epitopes (HLA A*0101, A*0201, A*0301, A*1101, A*2601, B*0702 restricted) in patients and = = Antigen-Specific T Cells in Healthy Individuals. To exclude that our priming (n 6) or in priming experiments only (n 30). Two findings were due to an unspecific response reflecting, e.g., cross- peptides (GUCA1A 47A and GUCA1B 133A) could be confirmed = reactivity with viruses, samples from healthy donors (HDs) were as T-cell epitopes in patients only (n 2). Seventeen peptides evaluated. Donors were HLA matched to the respective MHC-I could not be verified as epitopes (Fig. 3C). The intensity of T-cell restriction of the peptides. No T-cell responses were found after reactivity varied with detection of very strong (n = 8; 22%), stimulation with the MHC-I–restricted peptides in PBMC samples strong (n = 8; 22%), intermediate (n = 9; 25%), weak (n = 9; of all tested HDs (10 per peptide; Table S3). 25%), or very weak (n = 2; 6%) IFN-γ responses (Fig. 3D). To assess the binding of the nonamer peptides GUCA1A 47A (NLSPSASQY) and GUCA1B 133A (QTEQGQLLT) to Epidemiology of the Identified SNP. Sequencing of the GUCY2D MHC class II molecules, the HLA typing of the three res- gene resulted in the identification of four different nonsynonymous ponding patients (7, 9, 11) was compared. Responders to SNPs (Table S4) resulting in a change of the amino acid sequence. NLSPSASQY were found to be DRB1*1501 and DQB1*0602 Three (A52S rs61749665 in exon 2, V361M rs186508466 in exon positive, whereas QTEQGQLLT responders were carrier 4, and L782H rs8069344 in exon 12) were already published of DRB1*0702 and DRB1*0202 alleles. Six DRB1*15 and (www.ncbi.nlm.nih.gov/snp), whereas one (S25F in exon 2) was DQB1*06, as well as eight DRB1*07 and DQB1*02 positive HDs, previously unknown. For frequency analysis, we sequenced 100 were analyzed for cytokine release. No T-cell responses were persons and found one subject with a heterozygous SNP (S25F). observed herein. As control for every retGC-derived MHC-II The most frequent missense SNP leading to an amino acid peptide (GUCY2D 52A, 52B and GUCY2D 782A, 782B), 10 change from alanine to serine at position 52 is encoded by exon 2 HDs were tested for T-cell responses in ELISpot. In one case, of the GUCY2D gene with a reported minor allele frequency a response on stimulation with the GUCY2D 52B peptide was (MAF) of 37.7% (rs61749665). Our frequency analysis of 100 detected, which was validated by ICS. Here, T cells recog- persons revealed an allele frequency of 30% (11% homozygous, nized both peptides (52A: LLQPPALSAVFTVGVLG; 52B: 38% heterozygous, and 51% carried the WT alleles). The SNP LLQPPALSSVFTVGVLG). A361M was unknown when this study was initiated and later Moreover, all peptide sequences were compared with proteins reported to have a MAF of 0.1% (rs186508466). Here, in a co- derived from bacteria, viruses, fungi, and other pathogens (blast. hort of 200 people, no individual carrying the SNP was identified. ncbi.nlm.nih.gov/Blast.cgi). No significant homology was identified. The SNP L782H has a reported MAF of 15.5% (rs8069344).

13830 | www.pnas.org/cgi/doi/10.1073/pnas.1609118113 Mirza et al. Downloaded by guest on September 27, 2021 No SNPs were found in patient and donor DNA by sequencing of antigens, based on structural differences between donor and the GUCA1A gene, whereas we observed a heterozygous pub- recipient. According to these definitions, recognition of self- lished SNP in one patient in exon 3 of the GUCA1B gene, antigens by allogeneic T cells can neither be subsumed as new leading to an amino acid exchange from glutamic acid to aspartic autoimmunity nor as alloreactivity. Hence, we propose the term acid (E155D). This SNP is rare, with a reported MAF of 0.5% graft-versus-self (GvS) for description of autoantigens recog- (rs139923590). nized by donor T cells to reflect our observation. In the pathogenesis of GVHD, endothelial damage and infec- Discussion tions of respective tissues play a crucial role. Hence, the main sites The cellular mechanisms of graft–host interaction comprise, of manifestation represent organs with higher susceptibility to in- among others, T-cell recognition of major (MHC) and minor fections (45), but also CNS may be involved (9, 10). With regard (miHAG) histocompatibility antigens (Fig. S6) (5). However, in to the retina, only indirect evidence was reported because inflam- + murine models using the transfer of in vivo-generated CD4 T mation improved or deteriorated alongside changes in immuno- cells, the development of autoreactivity has been suggested (27). suppression (17). In our study, viral eye infections as observed in Of note, these reports did not evaluate the target antigens of patient 5 also could trigger the development of a tissue antigen- autoreactive T cells. Our data provide evidence in human allo- specific GvS reaction. However, further studies are required to es- geneic HCT that genetic differences are not required for allo- tablish risk factors for the development of GvS and to evaluate the geneic T-cell reactivity to develop (Fig. S6, Right) and describe clinical relevance of such circulating antigen-specific T cells. the precise epitopes (autoantigens) of these T cells. In fact, we In summary, our data represent a report of autoantigen-specific T observed self-antigen–specific T cells directed against epitopes cells and their precise epitope in human allogeneic HCT and pro- derived from three highly polymorphic retina proteins in 2 of 6 vide evidence that allogeneic T-cell reactivity is by far more complex analyzed patients with inflammatory PS complications and in 6 than previously believed. Not only genetic differences lead to in- of 22 patients without ocular symptoms. Because two of the re- duction of recipient-specific T cells, but also autoantigens can be spective proteins (retGC and GCAP1) are predominantly in- recognized. Hence, beyond recognition of MHC molecules based volved in cone metabolism (28, 29), slight damage would result in on differences in the MHC locus and MHC-restricted peptides changes in color vision, which could manifest subclinically (30). (miHAG) due to genetic differences between donor and recipient The detectable T cells could be also of low avidity and, hence, leading to variant amino acid sequences of translated proteins, a not capable of inducing retina damage. In line, it has been ob- novel mechanism of the T cell-based graft–host interaction is served that also circulating Melan-A/MART-1–specific naïve- characterized on a molecular basis of T-cell function by our data: the recognition of autoepitopes (GvS) not requiring genetic differ-

phenotype T cells were not associated with vitiligo (31). In INFLAMMATION contrast to these observations, we could not detect any reactivity ences between patient and donor. IMMUNOLOGY AND in healthy volunteers for most peptides. In our study, the avidity of the detected T cells could not be determined because the used Methods peptide concentrations were chosen to activate all potential antigen- Sample Collection. This study was approved by the ethics committee of the specific cells (32, 33). Further avidity characterization using peptide University of Tuebingen (Tuebingen, Germany). All patients gave their written titrations was not possible due to limited sample availability but informed consent before entering the study. Blood samples were obtained be- should be performed in future studies. fore the start of conditioning regimen (second group). After hematologic re- Thus, our data provide evidence that reactivity of allogeneic T generation, blood samples were obtained until 1 y after HCT. In case of an cells cannot only be induced by allo-antigens but also by antigens observed T-cell response, the period could be extended after reapproval of the sharing sequences with donor antigens, which implies that graft patients. PBMCs were isolated by density gradient centrifugation. DNA was isolated from blood before (recipient DNA) and after HCT (donor DNA) using the cells can be activated by self-antigens. This hypothesis is further Invitrogen DNA Isolation Kit (Invitrogen). All patients had full donor chimerism at underlined by the observation that a GVHD can also be present the time of DNA isolation and immunological evaluation. For patients recruited in patients undergoing autologous HCT where genetic differ- after HCT, autologous DNA was isolated from oral mucosa. Frozen donor samples ences cannot contribute to its development (34). were used if not required for quality control purposes. Donors had given their The activation or induction of such allogeneic, autoreactive T consent to scientific use of residual material. cells could be based on several mechanisms including decreased thymic selection as shown in murine models (27, 35) or an impaired PCR and Sanger Sequencing. PCR and Sanger sequencing for the entire coding thymic function during allogeneic HCT and GVHD, leading to sequence were performed for SNP hotspot regions. After PCR using specific impaired immunological reconstitution (36, 37). Also, impaired primers (Table S4), reactions were plotted on agarose gels, and bands were function of regulatory T cells (e.g., through conditioning regimens or excisedandpurifiedwithaDNAGelextraction kit (Promega). Sanger sequencing immunosuppression) could play a role (34). However, also trans- results were processed using the National Center for Biotechnology Information plantation of quiescent autoreactive T cells in the grafts could lead to (NCBI) blast platform (blast.ncbi.nlm.nih.gov/Blast.cgi). subsequent activation and amplification of such cells in the absence of an effective immunosurveillance during transplantation (38, 39). Synthetic Peptides. Peptides were synthesized by solid-phase Fmoc chemistry Because we did not detect any autoreactive T cells in donor using a peptide synthesizer 433A (Applied Biosystems). Identity and purity of samples, the first mechanism seems to be more likely. Due to the the peptides were analyzed by reversed-phase HPLCy (HPLC) and matrix- limited sample numbers, the latter can still not be fully excluded. assisted laser desorption/ionization/time-of-flight MS (Thermo Fischer). Pep- > The development of autoinflammatory diseases in patients tides were further purified by reversed-phase HPLC to 90% purity.

with full donor chimerism after HCT is referred to as new autoim- + munity. However, the discrimination between classical GVHD, dis- Priming of T Cells with Peptide-Loaded Dendritic Cells. CD8 T cells were iso- lated from PBMCs using magnetic cell sorting (MACS) according to the ease relapse, infection, and treatment-related toxicity is difficult. Most ’ of the reported manifestations are antibody-mediated diseases such as manufacturer s instructions (Miltenyi Biotec). Autologous monocyte-derived – dendritic cells (DCs) were generated from PBMCs with GM-CSF (50 ng/mL; autoimmune hemolytic anemia (40 42), immune thrombocytopenia PeproTech) and IL-4 (20 ng/mL; R&D Systems). DC maturation was induced by (41, 43), thyroiditis, and myasthenia gravis (42). As of yet, systematic lipopolysaccharide (LPS; Sigma-Aldrich). Mature DCs were harvested, loaded studies have evaluated the development of new autoimmunity with peptide (20 μg/mL), and added to T cells (ratio 1:3 or 1:5). Mixed lym- only in patients with primary autoimmune diseases being present phocyte cultures were supplemented with IL-12 (PromoKine). T cells were − before autologous and allogeneic HCT (42, 44). No data are restimulated weekly with autologous peptide-pulsed irradiated CD8 cells or available about de novo development of new autoimmunity. DCs. IL-2 (R&D Systems) was added on days 1 and 3 after every restimulation. Notably, the term new autoimmunity is commonly referred to diseases that share laboratory findings and clinical presentation Peptide Presensitization of PBMCs. As described previously, T cells were pre- with normal autoimmune diseases including detection of disease sensitized with peptides for 12 d (1 and 5 μg/mL for MHC classes I and II, specific antibodies (37). In contrast, alloreactivity denotes recognition respectively) (33). For stimulation with MHC class I-predicted peptides, IL-4

Mirza et al. PNAS | November 29, 2016 | vol. 113 | no. 48 | 13831 Downloaded by guest on September 27, 2021 and IL-7 (PromoKine) were added on days 0 and 1. IL-2 was added on days 3, with CD4-APC (BD Biosciences) and CD8-PeCy7–conjugated antibodies (Beckman 5, 7, and 9. Cells were harvested on day 12 and further evaluated. Coulter). To label dead cells, PBMCs were stained with AquaLifeDead (Invi- trogen, Life Technologies). After permeabilization (Cytoperm/Cytofix solution; IFN-γ ELISpot Assay. Twelve-day presensitized PBMCs were added to a pre- BD Biosciences), intracellular staining was done with IFN-γ-FITC (BD Biosciences) coated (anti-human IFN-γ antibody 1-D1 K; Mabtech) nitrocellulose plate. and TNFα-PE (Beckman Coulter) –conjugated antibodies. Cells were measured in Cells were incubated for 24–26 h with peptides. As positive and negative FACSCalibur, FACSCanto, and LSR Fortessa flow cytometers (BD Biosciences). A controls, a pool of viral epitopes (derived from CMV, EBV, and influenza) cytokine response was regarded positive when the percentage of the TNF-α– or and an HIV-derived epitope or human self-epitope was used, respectively IFN-γ–producing T-cell population (CD4 or CD8 negative) was at least twofold of (Table S3). Plates were further incubated (2 h) with a biotin-labeled anti–IFN-γ the negative control. A response was considered specific when at least 0.05% of antibody (7-B6-1 biotin; Mabtech) followed by an incubation step for 1 h the T cells (CD4 or CD8 negative) produced IFN-γ. with streptavidin-alkaline phosphatase (Sigma-Aldrich). Afterward, 5-bromo-4- chlore-3-indolyl-phosphate/nitroblue tetrazolium (Sigma-Aldrich) was added. ACKNOWLEDGMENTS. We thank Patricia Hrstiç, Nicole Bauer, and Katharina Spots were automatically counted using the ImmunoSpot S6 Ultra-V Analyzer Graf for peptide synthesis, as well as Lynne Yakes for editorial support; ELISpot reader (CTL Europe). Duplicate wells were considered positive if Cécile Gouttefangeas for expert support in T-cell monitoring; Susanne Kohl (i) there were at least 10 spots per 250,000 PBMCs detectable in either well, for provision of PCR systems; Dr. Christoph Faul, Ute Schroeder, Grazia Koch, and (ii) the mean number of spots was at least threefold of the counted and Erwin Schleicher for provision of patient and anonymous donor samples; Richard F. Spaide, Lawrence A. Yannuzzi, and Yale Fisher of the Vitreous, spots in the negative control. Retina and Macula Consultants New York for helpful discussion; and Selma and Austin I. Fink for providing helpful feedback. This project was supported by ICS. Twelve-day presensitized PBMCs or primed T cells were incubated with the fortüne Programm of the Eberhard Karls Universität Tübingen (Grant 1832- peptides for 5–6 h or overnight. To inhibit cytokine secretion, GolgiStop 0-1), the Deutsche José Carreras Leukämie Stiftung (Grant DJS 08/04), and the solution (BD Biosciences) was added. Cell membrane molecules were stained Deutsche Krebshilfe (Grant 110465). All funding was granted to S.P.H.

1. Ferrara JL, Levine JE, Reddy P, Holler E (2009) Graft-versus-host disease. Lancet 26. den Hollander AI, et al. (2009) A homozygous missense mutation in the IRBP gene 373(9674):1550–1561. (RBP3) associated with autosomal recessive retinitis pigmentosa. Invest Ophthalmol 2. Felix NJ, et al. (2007) Alloreactive T cells respond specifically to multiple distinct Vis Sci 50(4):1864–1872. peptide-MHC complexes. Nat Immunol 8(4):388–397. 27. Zhang Y, Hexner E, Frank D, Emerson SG (2007) CD4+ T cells generated de novo from 3. Kumari S, et al. (2014) Alloreactive cytotoxic T cells provide means to decipher the donor hemopoietic stem cells mediate the evolution from acute to chronic graft- immunopeptidome and reveal a plethora of tumor-associated self-epitopes. Proc Natl versus-host disease. J Immunol 179(5):3305–3314. Acad Sci USA 111(1):403–408. 28. Stiebel-Kalish H, et al. (2012) Gucy2f zebrafish knockdown–a model for Gucy2d-related 4. Haen SP, Rammensee HG (2013) The repertoire of human tumor-associated epitopes– leber congenital amaurosis. Eur J Hum Genet 20(8):884–889. identification and selection of antigens and their application in clinical trials. Curr 29. Zobor D, Zrenner E, Wissinger B, Kohl S, Jägle H (2014) GUCY2D- or GUCA1A-related Opin Immunol 25(2):277–283. autosomal dominant cone-rod dystrophy: Is there a phenotypic difference? Retina 5. Spierings E (2014) Minor histocompatibility antigens: Past, present, and future. Tissue 34(8):1576–1587. Antigens 84(4):374–60. 30. Kaur M, et al. (2015) Correlation between structural and functional retinal changes in 6. Matsumoto Y, Haen SP, Spaide RF (2007) The white dot syndromes. Compr Ophthalmol Parkinson disease. J Neuroophthalmol 35(3):254–258. Update 8(4):179–200, discussion 203–204. 31. Pittet MJ, et al. (1999) High frequencies of naive Melan-A/MART-1-specific CD8(+)T 7. Kamble RT, Chang CC, Sanchez S, Carrum G (2007) Central nervous system graft-versus-host cells in a large proportion of human histocompatibility leukocyte antigen (HLA)-A2 disease: Report of two cases and literature review. Bone Marrow Transplant 39(1):49–52. individuals. J Exp Med 190(5):705–715. 8. Kew AK, et al. (2007) Central nervous system graft-versus-host disease presenting 32. Precopio ML, et al. (2008) Optimizing peptide matrices for identifying T-cell antigens. with granulomatous encephalitis. Bone Marrow Transplant 40(2):183–184. Cytometry A 73(11):1071–1078. 9. Ma M, et al. (2002) CNS angiitis in graft vs host disease. Neurology 59(12):1994–1997. 33. Chudley L, et al. (2014) Harmonisation of short-term in vitro culture for the expansion 10. Sostak P, et al. (2010) Cerebral angiitis in four patients with chronic GVHD. Bone of antigen-specific CD8(+) T cells with detection by ELISPOT and HLA-multimer Marrow Transplant 45(7):1181–1188. staining. Cancer Immunol Immunother 63(11):1199–1211. 11. Grauer O, et al. (2010) Neurological manifestations of chronic graft-versus-host disease after 34. Drobyski WR, Hari P, Keever-Taylor C, Komorowski R, Grossman W (2009) Severe allogeneic haematopoietic stem cell transplantation: Report from the Consensus Conference autologous GVHD after hematopoietic progenitor cell transplantation for multiple on Clinical Practice in chronic graft-versus-host disease. Brain 133(10):2852–2865. myeloma. Bone Marrow Transplant 43(2):169–177. 12. Anderson NG, Regillo C (2004) Ocular manifestations of graft versus host disease. Curr 35. Dertschnig S, Hauri-Hohl MM, Vollmer M, Holländer GA, Krenger W (2015) Impaired Opin Ophthalmol 15(6):503–507. thymic expression of tissue-restricted antigens licenses the de novo generation of 13. Kim SK (2006) Update on ocular graft versus host disease. Curr Opin Ophthalmol autoreactive CD4+ T cells in acute GVHD. Blood 125(17):2720–2723. 17(4):344–348. 36. Krenger W, Blazar BR, Holländer GA (2011) Thymic T-cell development in allogeneic 14. Kaiserman I, Or R (2005) Laser photocoagulation for central serous retinopathy as- stem cell transplantation. Blood 117(25):6768–6776. sociated with graft-versus-host disease. Ocul Immunol Inflamm 13(2-3):249–256. 37. Holbro A, Abinun M, Daikeler T (2012) Management of autoimmune diseases after 15. Brown GC, et al. (1982) Radiation retinopathy. Ophthalmology 89(12):1494–1501. haematopoietic stem cell transplantation. Br J Haematol 157(3):281–290. 16. Coskuncan NM, et al. (1994) The eye in bone marrow transplantation. VI. Retinal 38. Anderson LD, Jr, Petropoulos D, Everse LA, Mullen CA (1999) Enhancement of graft- complications. Arch Ophthalmol 112(3):372–379. versus-tumor activity and graft-versus-host disease by pretransplant immunization of 17. Strouthidis NG, et al. (2003) Posterior segment complications of graft versus host allogeneic bone marrow donors with a recipient-derived tumor cell vaccine. Cancer disease after bone marrow transplantation. Br J Ophthalmol 87(11):1421–1423. Res 59(7):1525–1530. 18. Cheng LL, et al. (2002) Graft-vs-host-disease-associated conjunctival chemosis and central 39. Zhou W, et al. (2009) Impact of donor CMV status on viral infection and reconstitution serous chorioretinopathy after bone marrow transplant. Am J Ophthalmol 134(2):293–295. of multifunction CMV-specific T cells in CMV-positive transplant recipients. Blood 19. Payne AM, et al. (2001) Clustering and frequency of mutations in the retinal gua- 113(25):6465–6476. nylate cyclase (GUCY2D) gene in patients with dominant cone-rod dystrophies. J Med 40. Sanz J, et al. (2007) Autoimmune hemolytic anemia following allogeneic hemato- Genet 38(9):611–614. poietic stem cell transplantation in adult patients. Bone Marrow Transplant 39(9): 20. Van Ghelue M, et al. (2000) Autosomal dominant cone-rod dystrophy due to a mis- 555–561. sense mutation (R838C) in the guanylate cyclase 2D gene (GUCY2D) with preserved 41. Bohgaki T, Atsumi T, Koike T (2008) Autoimmune disease after autologous hemato- rod function in one branch of the family. Ophthalmic Genet 21(4):197–209. poietic stem cell transplantation. Autoimmun Rev 7(3):198–203. 21. Kitiratschky VB, et al. (2008) Mutation analysis identifies GUCY2D as the major gene 42. Daikeler T, et al.; EBMT Autoimmune Disease Working Party (2011) Secondary au- responsible for autosomal dominant progressive cone degeneration. Invest toimmune diseases occurring after HSCT for an autoimmune disease: A retrospective Ophthalmol Vis Sci 49(11):5015–5023. study of the EBMT Autoimmune Disease Working Party. Blood 118(6):1693–1698. 22. Perrault I, et al. (2000) Spectrum of retGC1 mutations in Leber’s congenital amaurosis. 43. Ahmad I, Haider K, Kanthan R (2004) Autoimmune thrombocytopenia following Eur J Hum Genet 8(8):578–582. tandem autologous peripheral blood stem cell transplantation for refractory germ 23. Payne AM, et al. (1999) Genetic analysis of the guanylate cyclase activator 1B (GUCA1B) cell tumor. Bone Marrow Transplant 34(3):279–280. gene in patients with autosomal dominant retinal dystrophies. JMedGenet36(9):691–693. 44. Loh Y, et al. (2007) Development of a secondary autoimmune disorder after hema- 24. Jiang L, Baehr W (2010) GCAP1 mutations associated with autosomal dominant cone topoietic stem cell transplantation for autoimmune diseases: Role of conditioning dystrophy. Adv Exp Med Biol 664:273–282. regimen used. Blood 109(6):2643–548. 25. Kamenarova K, et al. (2013) Novel GUCA1A mutations suggesting possible mecha- 45. Cooke KR, et al. (1998) Tumor necrosis factor-alpha production to lipopolysaccharide nisms of pathogenesis in cone, cone-rod, and macular dystrophy patients. BioMed Res stimulation by donor cells predicts the severity of experimental acute graft-versus- Int 2013:517570. host disease. J Clin Invest 102(10):1882–1891.

13832 | www.pnas.org/cgi/doi/10.1073/pnas.1609118113 Mirza et al. Downloaded by guest on September 27, 2021