Epitope Mapping of SERCA2a Identifies an Antigenic Determinant That Induces Mainly Atrial Myocarditis in A/J Mice

The intracellular membranous network, sarcoplasmic reticulum (SR) found in the muscle cells is important in the regulation of Ca²⁺ homeostasis; it involves the participation of sarcoplasmic/endoplasmic reticulum Ca²⁺ adenosine triphosphatase (SERCA) as a major Ca²⁺ transporter (1–4). Three major and 13 subisoforms of SERCA have been reported in humans. SERCA1 gene encodes SERCA1a, 1b, and 1c (5–7); SERCA2 gene encodes for SERCA2a, 2b, 2c, and 2d (only at mRNA level), whereas SERCA3 exists in six isoforms (SERCA3a to SERCA3f) (1, 3, 4, 8–11). All isoforms exist, except SERCA2c, SERCA2d, and SERCA3d to SERCA3f in mice and rats (1, 12). Among these, SERCA2a is expressed specifically in cardiomyocytes with greater expression in atria than in ventricles (13–15), slow-twitch skeletal muscle cells (16), and in vascular smooth muscle cells (17). Although SERCA2b is expressed ubiquitously (18), SERCA2c has been recently reported to be expressed in the left ventricles (LVs) in humans (19). In contrast, SERCA1 isoforms are expressed in the fast-twitch skeletal muscle. SERCA3 proteins can be expressed in various tissues including hematopoietic cell lineages (4, 20–22). Thus, the functions of each isoform are likely tissue dependent.

SERCA2a plays an indispensable role in the contractility of the heart. During contraction, action potential propagating along cardiomyocyte membranes and their invaginations, T-tubules, results in the influx of Ca²⁺ by opening the l-type Ca²⁺ channels, which then triggers the release of more Ca²⁺ through ryanodine receptors from the SR (23). The majority of this cytosolic Ca²⁺ must be sequestered by SR membrane-bound SERCA2a, which facilitates Ca²⁺ uptake by the SR, resulting in the relaxation of cardiomyocyte, and a minor pathway for Ca²⁺ extrusion via the plasma membrane Na⁺/Ca²⁺ exchanger (24, 25). Any alterations in the SERCA2a function can disturb contractibility of the heart leading to heart failure (HF) (25–29).

The importance of SERCA2a in the pathogenesis of HF is well documented. Altered expression of SERCA2a has been reported in major heart diseases including ischemic heart disease, cardiomyopathies and congestive HF (29–33), in which downregulated expression of SERCA2a can result in both systolic and diastolic heart dysfunctions (34). Conversely, overexpression of SERCA2a can improve the contractile function of failing hearts and also diminish cardiac fibrosis in congestive HF by altering TGF-β signaling (35, 36). Genetically altered mouse models also support these observations. For example, SERCA2a deletion is embryonically lethal; homozygous mice do not survive (37, 38); heterozygous mice with one functional allele can survive but they suffer from severe cardiac insufficiency/hypertrophy and death (39). Likewise, mice with conditional deletion of SERCA2a in cardiomyocytes can develop end-stage HF several weeks after deletion (40, 41). Conversely, transgenic mice overexpressing SERCA2a in the cardiomyocytes show improved heart functions and increased Ca²⁺ storing abilities with no physiologic (e.g., heart/body weight ratios) or histological (hypertrophy of cardiomyocytes) abnormalities (42, 43). Because patients with myocarditis or dilated cardiomyopathy show autoantibodies to SERCA2a (44, 45), SERCA2a-reactive Abs may have pathogenetic significance. In support of this proposition, mice immunized with SERCA2a can develop autoimmune myocarditis potentially mediated by SERCA2a-reactive Abs (46, 47). The antigenic determinants of SERCA2a were unknown, including the T cell epitopes if any, however. Here, we describe the identification of multiple immunodominant epitopes of SERCA2a in A/J mice and show that one of these epitopes, SERCA2a 971–990, induces mainly atrial inflammation by potentially acting as a common epitope to both T cells and B cells.

A/J mice (6–8-wk-old, male, H-2^a) were obtained from the Jackson Laboratory (Bar Harbor, ME) and the animals were maintained according to the Institutional Animal Care and Use Committee guidelines, University of Nebraska-Lincoln, Lincoln, NE. Euthanasia was performed using carbon dioxide in conformity with the guidelines on euthanasia of the American Veterinary Medical Association.

Peptide library for SERCA2a was created by designing 80 overlapping acetylated peptides (20-mers with an overlap of 10 aa) containing acetyl group at the N-terminal end because the acetylated peptides have been shown to be better presented by the APCs (48–50). All these peptides, bovine RNase 43–56 (VNTFVHESLADVQA), biotinylated hen egg lysozyme (HEL) 46–61 (YNTDGSTDYGILQINSR) (Neopeptide, Cambridge, MA) and moth cytochrome c (MCC) 82–103 (FAGLKKANERADLIAYLKQATK) (GenScript, Piscataway, NJ) were synthesized by 9-fluorenylmethyloxycarbonyl chemistry. The purity of peptides was ascertained by HPLC to be more than 90%, and their identity was confirmed by mass spectroscopy. Ultrapure water was used to dissolve the peptides and multiple aliquots of peptides were stored at −20°C until further use. But, some of these peptides required addition of DMSO as indicated with a footnote in Supplemental Table I.

Peptide/CFA emulsions that contain Mycobacterium tuberculosis H37RA extract (5 mg/ml; Difco Laboratories, Detroit, MI) were prepared and administered s.c. into mice on days 0 and 7 in the inguinal and sternum regions (51–53). In pooled settings, pools were made to include three to five peptides of 50 μg each, and for immunization with individual peptides, 50–100 μg of each was used. All animals received pertussis toxin (PT) (100 ng per mouse i.p.; List Biological Laboratories, Campbell, CA) on days 0 and 2 after the first immunization. For proliferation assay involving CD4 T cells, one dose of SERCA2a 971–990 emulsion was administered. Mice that received no peptides and those that received CFA and PT alone served as controls.

Whole heart and samples of liver, lung, kidney, skeletal muscle, and brain were collected at termination on day 21; they were fixed in 10% phosphate-buffered formalin and processed to obtain 5-μm serial sections ∼150 μm apart from each other. All sections were stained with H&E and examined by board-certified pathologists blinded to treatment and the total numbers of inflammatory cell foci were determined as we have described previously (49, 50). The degree of inflammation in hearts was scored as normal (0), mild (1–5 foci), moderate (6–25 foci), and severe (26 or more foci) as previously described (54). Fibrosis was assessed using Masson’s trichrome staining (55).

Hearts were examined using immunohistochemistry (IHC) for the presence of T cells and non-T cells. For detection of T cells, paraffin sections were stained with rabbit anti-mouse CD3 (clone, SP7, 1:100; Abcam, Cambridge, MA), rat anti-mouse CD4 (clone, GK1.5, 1:100; Leinco Technologies, Fenton, MO), and rat anti-mouse CD8 (clone, 53-6.7, 1:100; Leinco), and their corresponding isotype controls. For non-T cells (i.e., neutrophils, macrophages, and B cells) rat anti-mouse Ly6G (clone, 1A8, 1:250; Leinco), rabbit anti-mouse CD11b (clone, EPR1344, 1:3500; Abcam), and rat anti-mouse CD19 (clone, 6OMP31, 1:1000l; Thermo Fisher, San Diego, CA) and their corresponding isotype controls were used. For SERCA2a expression, sections were stained with rabbit anti-mouse SERCA2a (clone, EPR9392, 1:250; Abcam) or its isotype control as primary Abs. Briefly, after deparaffinization, rehydration, and blockade of endogenous peroxidase activity with 3% hydrogen peroxide, Ag retrieval was performed by treating the sections with 10 mM sodium citrate buffer (pH 6) in a water bath at 98°C for 15–40 min or using a pressure cooker. After blocking with 5% nonfat dry milk for 30 min, sections were incubated with primary Abs at 4°C overnight, followed by incubation with goat anti-rabbit IgG or donkey anti-rat IgG conjugated with HRP (Vector Laboratories, Burlingame, CA; and Abcam) as secondary Abs for 2 h at room temperature (RT). After adding diaminobenzoic acid as a substrate for color development, sections were counterstained with hematoxylin. For quantitative evaluation of T cells (CD3⁺, CD4⁺, and CD8⁺) and non-T cells (Ly6G⁺, CD11b⁺, and CD19⁺) in the atria, three to five random areas (1.5–3.0 mm²) from representative sections were selected from severely affected animals, and nuclear staining was confirmed using nuclear V9 software (Aperio Technologies, Vista, CA). The number of cells positive for each marker were counted and normalized to a 1-mm² area using Aperio ImageScope Analysis Software (Leica Biosystems, MN) as we have described previously (49).

After perfusion, hearts from naive mice were embedded in optimal cutting temperature compound and cryosections (7–10 μm) were made and stored at −80°C until further use. Sections were rehydrated with 1× PBS for 5 min and fixed with 4% (w/v) paraformaldehyde pH 7.4 for 40 min at RT followed by blocking. Staining was performed to include rabbit anti-mouse SERCA2a or its isotype control as primary Abs, overnight at 4°C. Sections were washed and incubated with goat anti-rabbit Alexa Fluor 647 (1:100; Abcam) for 2 h in the dark at RT. Finally, sections were stained with wheat germ agglutinin (WGA) Alexa Fluor 488 conjugate (Thermo Fisher) for 30 min as a cell membrane marker (56). After washing, sections were mounted using fluoromount mounting medium and analyzed using Olympus FV500-IX81 laser scanning confocal microscope (Olympus America, Central Valley, PA). Confocal images were acquired using the sequential dual laser-line excitation/emission mode (488 nm for WGA membrane marker, and 633 nm for SERCA2a (12, 57).

Electrocardiography (ECG) was performed on the anesthetized animals immunized with or without SERCA2a 971–990 (on day 23) with 3-lead ECG system (PowerLab 8/30; AD Instruments, Colorado Springs, CO) and heart rates and ECG signals were monitored and recorded on LabChart 8 software (AD Instruments). The ECG waveforms recorded from individual animals were analyzed separately using the ECG analysis module of LabChart 8 software, to derive average wave intervals and peak amplitudes for the groups (49).

Transthoracic echocardiography was performed in anesthetized animals immunized with or without SERCA2a 971–990 on day 23, and a research sonographer, blinded to the study groups, performed the measurements and analyzed the data as we reported previously (49, 50). Closed-chest imaging was performed in the short axis view at the mid-LV level, verified by the presence of prominent papillary muscles, using a commercially available echocardiography system (Vivid 7; General Electric, Wauwatosa, WI) with an M12-L linear array transducer. Image depth was 1.5 cm, with the acquisition of 293.6 frames/s, second harmonic imaging, and electrocardiographic gating. From the raw two-dimensional image of the mid-LV, anatomical M-mode through the anteroseptal and inferolateral segments was used to measure the width of the intraventricular septum at diastole and the internal diameter of the LV at diastole and systole. End-diastolic volume (EDV) and end-systolic volume (ESV) were calculated using the Teichholz formula: LV Volume = [7/(2.4 + left ventricular internal diameter (LVID))] × LVID³. A cardiac cycle was defined from the peak of one R wave to the peak of the following R wave. Three consecutive heart beats were measured and the mean value was used for analysis.

Magnetic resonance microscopy (MRM) imaging was used to determine cardiac abnormalities in mice immunized with SERCA2a 971–990 during days 38–39 as we have described previously (50, 58). Anesthetized animals were placed in the animal holder equipped with respirometry and pulse oximetry to gate the respiratory and cardiac signals, and body temperature was monitored using a rectal thermometer. MRM imaging was performed using a wide-bore (89 mm) 9.4 T vertical-bore magnet (Varian, Walnut Creek, CA) equipped with triple axis gradients of 100 G/cm and a 4-cm radio frequency imaging coil. Short axis slices (images) of hearts captured in eight time frames using an echo-based cine pulse sequence were analyzed using Segment software (Segment v1.8 R1430; Medviso, Sweden) to assess LV wall thickness, EDV, ESV, stroke volume (SV), and ejection fraction (EF).

Lymph nodes (maxillary, mandibular, axillary, inguinal, and popliteal) and spleens were collected at termination on day 21 postimmunization and single-cell suspensions were obtained by treating the cells with RBC-lysing buffer (1× ammonium chloride potassium buffer; Lonza, Walkersville, MD). After washing, the pellets were dissolved in growth medium consisting of RPMI 1640 medium, FBS (10%), sodium pyruvate (1 mM), l-glutamine (4 mM), 1× each of nonessential amino acids and vitamin mixture, and penicillin-streptomycin (100 U/ml) (Lonza). To determine proliferative responses to SERCA2a peptides in naive mice, splenocytes were used. In some experiments, CD4 T cells were enriched from lymph node cells (LNCs) and splenocytes to a purity of ∼95% by negative selection based on magnetic separation using IMAG (BD Biosciences, San Jose, CA) (53). Cells were stimulated in triplicates with the indicated peptides (0–100 μg/ml) at a density of 5 × 10⁶ cells/ml for 2 d, and cells in medium alone or those stimulated with RNase 43–56 served as controls. To stimulate CD4 T cells, syngeneic irradiated splenocytes loaded with peptides were used as APCs at a ratio of 1:1 (0–50 μg/ml). After pulsing with tritiated thymidine (1 μCi per well; MP Biomedicals, Santa Ana, CA) for 16 h, proliferative responses were measured as cpm using a Wallac liquid scintillation counter (PerkinElmer, Waltham, MA). For easy depiction, where indicated, T cell responses are shown as fold changes derived by dividing the cpm values of cultures stimulated with peptides by the cpm values of unstimulated cultures (medium controls).

Flow cytometrically, expression of IA^k and IE^k molecules was tested in splenocytes obtained from naive A/J mice using rat anti-mouse IA^k (clone, 14V.18; Thermo Fisher), anti-mouse IE^k (clone, 14-4-4S; Thermo Fisher), anti-mouse CD11b (clone, M1/70; eBioscience, San Diego, CA), anti-mouse CD11c (clone, N418; eBioscience), and anti-mouse CD19 (clone, 1D3; Leinco) Abs and their isotype controls and 7-aminoactinomycin D (7-AAD; Invitrogen, Waltham, MA). The percent IA^k+ and IE^k+ populations were determined in the live populations using FlowJo software.

To determine the affinities of peptides binding to IA^k allele, soluble IA^k monomers were expressed using the constructs that we had reported previously (59–61). However, for IE^k allele, we needed to design new IE^k constructs and express soluble IE^k monomers.

Extracellular portions of IE^k-α- and IE^k-β-chains that respectively contain the leucine zipper domains of Fos and Jun transcriptional factors were synthesized (GenScript). The IE^k-β construct was also designed to include the nucleotide sequence of the CLIP 88–102 (5′-GTGAGCCAGATGCGGATGGCTACTCCCTTGCTGATGCGTCCAATG-3′) linked with thrombin cleavage site (LVPRGS) such that upon expression, the CLIP-linked IE^k protein could be obtained, and the peptides of interest could then be exchanged with CLIP. The constructs were ligated to pAcDB3 vector and sequenced. Soluble IE^k molecules were then expressed in the baculovirus using sf9 cells as we have previously described (57, 59, 60). The IE^k proteins were purified on anti-IE^k column (clone, M5/114; Bio X Cell, West Lebanon, NH), and concentrated using Amicon Ultra centrifugal filters (Millipore, Burlington, MA). Empty IE^k molecules were then obtained by treating the proteins with thrombin (20 U/mg) (Novagen, Madison, WI) to release the CLIP peptide.

Reaction mixtures were then prepared to include thrombin-cleaved IA^k or IE^k monomers (0.35 μg), competitor peptides (SERCA2a 971–990, 0.00001–100 μM), and constant amounts of the biotinylated reference peptides HEL 46–61 (for IA^k) or MCC 82–103 (for IE^k) (1 μM) (62–64), in a buffer containing 50 mM sodium phosphate pH 7, 100 mM sodium chloride, 1 mM EDTA, and 1× protease inhibitor (Sigma-Aldrich, St. Louis, MO). The mixtures were incubated at RT overnight. In addition, anti-IA^k (clone, 10-2.16; Bio X Cell), and anti-IE^k Abs (10 μg/ml) were coated separately onto 96-well white fluorescence plates in 0.2 M sodium phosphate buffer, pH 6.8, and incubated overnight at 4°C. Plates were washed five times with 1× wash buffer (PerkinElmer), and then blocked with 2% bovine casein (Sigma-Aldrich) for 2 h at RT. After washing, the above peptide reaction mixtures were added in duplicates and the plates were incubated on a rocker at RT for 1 h followed by washing as above. Finally, after adding 100 μl of europium-labeled streptavidin (SA) (0.1 μg/ml) and dissociation-enhanced lanthanide fluoroimmunoassay (DELFIA) enhancement solution (PerkinElmer) sequentially, fluorescence intensity was measured at excitation/emission wavelengths of 340/615 nm using a Victor Multilabel Plate Reader (PerkinElmer). The IC₅₀ values were calculated based on the concentrations of competitor peptides that prevented 50% binding of the reference peptides (HEL 46–61 or MCC 82–103) as described previously (49, 50, 54).

To enumerate the frequencies of Ag-specific CD4 T cells, we created both IA^k and IE^k dextramers. To create IA^k dextramers for SERCA2a 971–990, the nucleotide sequence (5′-CCTTTGCCGCTCATTTTCCAGATCACACCGCTGAATCTGACCCAGTGGCTGATGGTGCTG-3′) was inserted into the existing IA^k-β construct that we had described previously (59, 60), whereas IA^k/RNase 43–56 (control) dextramers were readily available in our laboratory (59, 65). For creation of IE^k dextramers, we inserted the nucleotide sequence of SERCA2a 971–990 as above, and MCC 82–103 (control) (5′-TTTGCCGGTTTAAAGAAGGCAAACGAACGTGCAGATCTCATCGCCTATCTAAAACAAGCTACTAAG-3′) (62–64). In addition, we inserted the BirR-A site (LGGIFEAMKMELRD) for biotinylation in the 3′end of Fos sequence in the IE^k-α construct (60). All three constructs, (i.e., IA^k/SERCA2a 971–990, IE^k/SERCA2a 971–990, and IE^k/MCC 82–103) were expressed in the sf9 cells using baculovirus expression system. After affinity column purifications using anti-IA^k and anti-IE^k Abs, soluble IA^k/SERCA2a 971–990, IE^k/SERCA2a 971–990, and IE^k/MCC 82–103 monomers were biotinylated and dextramers were derived using SA/fluorophore-conjugated dextran molecules as we have described previously (60, 61).

For dextramer staining, LNCs obtained from immunized animals were stimulated with SERCA2a 971–990 for 2 d and IL-2 medium was then added. Viable cells were harvested by Ficoll density gradient centrifugation on day 4, and cells were rested in the IL-2 medium. Cells harvested on days 7–9 poststimulation were stained with IA^k (SERCA2a 971–990 and RNase 43–56) or IE^k (SERCA2a 971–990 and MCC 82–103) dextramers followed by anti-CD4 and 7-AAD. After washing, cells were acquired by flow cytometry and the percent dextramer⁺ cells were analyzed using FlowJo software (59–61).

LNCs obtained from animals immunized with SERCA2a 971–990 were restimulated with or without specific peptide or RNase 43–56 (50 μg/ml), and culture supernatants were collected on day 3. Samples were analyzed for cytokines using beads conjugated with capture and detection Abs and standard curves were obtained by serially diluting the lyophilized mouse cytokine standard mix, consisting of IL-2, IFN-γ, IL-6, IL-10, IL-17A, and TNF-α as recommended by the manufacturer (BD Biosciences). First, capture bead/cytokine Ab conjugates were mixed and added to a tube containing diluted standards or test samples, followed by addition of detection Abs, and the mixtures were incubated at RT for 2 h. After washing, and acquisition by flow cytometry, data were analyzed by FCAP Array Software (BD Biosciences) (49, 50).

Serum samples were collected from mice immunized with or without SERCA2a 971–990 on day 21 postimmunization for measurement of SERCA2a-reactive Abs by ELISA. In brief, 96-well polystyrene microtiter plates were coated with or without SERCA2a 971–990 or irrelevant control (RNase 43–56) (10 μg/ml) in 1× coating buffer (eBioscience) and the plates were incubated at 4°C overnight. After washing with 1× PBS/0.05% Tween-20 and blocking with 1× PBS/2% BSA/5% normal goat serum for 1.5 h at RT, serum samples (1:100) were added in duplicates, and the plates were incubated at 37°C for 1 h. After washing, HRP-labeled donkey anti-mouse IgG (Santa Cruz, Dallas, TX) or HRP-labeled goat anti-mouse IgM, IgG1, IgG2a, IgG2b, and IgG3 (SouthernBiotech, Birmingham, AL) were used as secondary Abs. Plates were incubated at RT for 2 h, and 1× tetramethylbenzidine solution was then added as a substrate (eBioscience). Reactions were stopped using 1 M phosphoric acid and the plates were read at 450 nm using an automated ELISA reader (BioTek Instruments, Winooski, VT), and OD values were measured (50, 66).

Mice were immunized with or without SERCA2a 971–990 in CFA twice with an interval of a week, and at termination on day 14 lymphocytes were prepared from spleens and lymph nodes. Cells were stimulated with Con-A (2.5 × 10⁶ cells/ml; 2.5 μg/ml; Sigma-Aldrich) for 2 d. Viable cells were harvested and injected through retro-orbital sinus (50–60 × 10⁶ cells per animal) into naive mice primed with LPS (25 μg per mouse i.p., on day −4 and day 0). PT was administered i.p., (100 ng per mouse) on days 0 and 2 posttransfer. The LPS/PT-primed naive mice and those that received Con-A–stimulated naive splenocytes served as controls, and on day 14 animals were euthanized to collect tissues for histology.

We used the nonparametric Wilcoxon rank sum test to test differences between groups for inflammatory foci and infiltrates, LNC and splenocyte responses, echocardiography/MRM, MHC class II expression/binding affinities, and dextramer staining, cytokines, and Ab responses. Student t test was used for CD4 T cell responses. In determining T cell responses for some peptides with varied background levels, cpm values were scaled within the replicates and doses, using a constant multiplier determined by the average cpm values (49). A p value ≤ 0.05 was considered significant.

SERCA2a is a 998 aa long protein expressed specifically in the cardiac muscle (1, 4, 46). To identify the immunodominant epitopes we sought to create an epitope library by generating 80 peptides of 20-mers with 10 aa as overlaps between each (Supplemental Table I). First, we screened for the ability of SERCA2a peptides to induce inflammation in the heart muscle by creating 18 pools with three to five peptides in each (Supplemental Table II). Three weeks after immunization, hearts were collected for assessment of inflammation by H&E staining. This analysis led us to identify 10 pools that showed varying degrees of myocarditis with a disease incidence ranging from 20 to 80% (pools I, XII, XV, and XVII: 20%; pools III, VI, and XI: 40% and pools VII, IX, and XVIII: 80%) (Supplemental Table II). The numbers of inflammatory foci in the affected animals ranged from 1 to 68. Regardless of the presence or absence of disease, however, we determined T cell responses using LNCs from immunized animals in a recall assay to all the immunogenized peptides by using RNase 43–56 as an irrelevant control. The data presented in Supplemental Table II show that the T cell responses were noted for one or more peptides in each of the pools that induced the disease with responses to be up to 4-fold. Conversely, the T cell responses for most peptides in pools that did not induce myocarditis (pools II, IV, V, VIII, X, XIII, XIV, and XVI) were either lacking or the responses were <2-fold (Supplemental Table II), suggesting that myocarditogenicity of SERCA2a peptides may be related to their ability to induce T cell responses.

Next, we chose a panel of 13 peptides representing the pools that were associated with myocarditis as described above (pools I, III, VI, VII, IX, XI, XV, XVII, and XVIII; Supplemental Table II) for further characterization. Expectedly, heart sections from naive and CFA/PT control groups were negative for inflammatory infiltrates (Table I). In contrast, animals immunized with 6 of the 13 peptides showed infiltrates leading us to identify SERCA2a 971–990 as the most potent myocarditogenic epitope followed by SERCA2a 161–180 and SERCA2a 951–970. Their respective disease incidences were 100, 40, and 20% and the number of inflammatory foci was 29.50 ± 8.00, 2.50 ± 0.64, and 40.00 ± 39.00 (Table I). In fact, the inflammatory foci in two additional animals from the SERCA2a 971–990-immunized group were so extensive that their numbers could not be counted. The infiltrates predominantly contained mononuclear cells and granulocytes. More importantly, comparison of inflammation between different parts of heart tissue revealed that the inflammatory foci were evident in the atria of all animals immunized with SERCA2a 971–990 (100%, 11/11). In contrast, ventricles were affected in only ∼60% (7/12) of animals and 70% (5/7) of these animals had only mild disease and the remaining two had moderate to severe disease (Fig. 1, left panel: whole heart, Table I). By assessing the disease severity, sections from both right and left atrial had severe inflammation in 55% of animals (>26 foci); the remaining animals showed moderate (27%, 3/11: 6–25 foci) or mild disease (18%, 2/11: 1–5 foci) (Table I). However, when inflammation was noted, all three layers of heart muscle, namely endocardium, myocardium, and epicardium, were consistently affected in the atria as compared with ventricles in the SERCA2a 971–990-immunized animals (Fig. 1, middle panel: atria; and right panel: ventricles). Likewise, mild to moderate fibrosis was noted in the atria of severely affected animals (data not shown). Furthermore, because the epitope, SERCA2a 971–990 is common to both (SERCA2a and SERCA2b) isoforms, and SERCA2b isoform is expressed preferentially in noncardiac tissues, we examined skeletal muscle, liver, lung, kidney, and brain samples from animals immunized with SERCA2a 971–990. None of these organs showed inflammatory foci, but perivascular and peribronchiolar aggregates suggestive of interstitial pneumonia were detected in isolated sections from the lungs of the SERCA2a 971–990-immunized group relative to controls (data not shown). We thus identified SERCA2a 971–990 as an epitope that preferentially induces atrial inflammation, and this epitope was chosen for further characterization.

Using IHC, we sought to characterize the heart infiltrates by quantitatively analyzing T cells (CD3⁺, CD4⁺, and CD8⁺) and non-T cells (neutrophils: Ly6G⁺, macrophages: CD11b⁺, and B cells: CD19⁺). We focused on atria because inflammation was consistently detected in the atria more than in the ventricles. As shown in Fig. 2, cells positive for CD3, CD4, and CD8 were present in all the sections from SERCA2a 971–990-immunized animals (Fig. 2, left panel) with CD4⁺ T cells (533.82 ± 206.77) present with higher frequency than CD8⁺ T cells (226.34 ± 94.46) (Fig. 2, lower panel). Similarly, among the non-T cell populations, macrophages were detected more frequently (608.11 ± 138.81) than neutrophils (255.19 ± 66.16) and B cells (12.75 ± 4.23). None of the sections from naive or CFA/PT groups revealed the presence of any of the cell types tested (Fig. 2), suggesting that the inflammatory cells infiltrated in response to SERCA2a 971–990 immunizations.

We then analyzed expression of SERCA2a to determine variations, if any, between atria and ventricles of hearts in A/J mice. We noted that SERCA2a was detected in all groups, and the staining with an anti-SERCA2a Ab was ascertained to be specific by using isotype control (Fig. 3A). Evidently, SERCA2a expression appeared to be more concentrated in the atria than in the ventricles regardless of whether the animals were immunized with or without SERCA2a 971–990 (Fig. 3A, left and right panels) suggesting a possible relationship between the occurrence of inflammation and compartmentalization of expression of SERCA2a in the heart.

Using confocal microscopy, we investigated the localization of SERCA2a using Alexa Fluor 647 conjugate and WGA as a cell membrane marker, which emit red and green fluorescence, respectively (12, 67). The analysis revealed that SERCA2a was detected Ag-specifically within the cells in both atria and ventricles as staining with the isotype control was absent (Fig. 3B). Furthermore, by overlaying the images obtained with Alexa Fluor 647 representing the SERCA2a signal, and WGA, it was apparent that SERCA2a expression occurred within cardiomyocytes (Fig. 3B, column 4). However, detection of yellow fluorescence representing the combination of both red and green fluorescence in the cell boundaries (Fig. 3B, column 5) suggests that SERCA2a expression may be seen close to the cell membrane, raising the question as to the significance of such an expression in cardiomyocytes.

To determine whether heart abnormalities can be detected noninvasively in vivo analyzed in live animals, we used echocardiography and MRM techniques, as we have described previously (49, 50, 58) and have included ECG analysis. At 23 d postimmunization with SERCA2a 971–990, echocardiography showed a trend toward decreasing EF and fractional shortening, markers of LV systolic function, as compared with healthy mice (Table II). There were also other more subtle indicators of LV remodeling, such as increasing LVID in systole and increasing EDV (Table II). ECG analysis also revealed a significant difference in markers of repolarization (QT interval, JT interval, and Tpeak–Tend interval), and a trend with T wave amplitude indicating an underlying abnormality in the myocardial tissue (Table II). Specifically, the Tpeak–Tend interval is a marker of ventricular arrhythmogenesis and electrical instability (68). These findings suggest that subclinical LV remodeling is occurring at 23 d postimmunization. These remodeling features become more apparent by day 38 and with improved resolution of rodent MRM. At this time, there is a further increase in ESV and a further decline in EF and SV with increased LV wall thickness (Supplemental Fig. 1). Taken together, the data suggest that structural and functional abnormalities of hearts in animals immunized with SERCA2a 971–990 can be detected by ECG and echocardiography or MRM techniques.

In a recall assay, we used LNCs obtained from immunized mice and stimulated the cells with SERCA2a 971–990 or control (RNase 43–56). Fig. 4, left panel, shows that the LNCs responded to SERCA2a 971–990 dose-dependently with a significant increase in the proliferative responses relative to medium or control (RNase 43–56). We then enriched CD4 T cells from immunized animals by magnetic separation. After ascertaining their purity (∼95%), cells were stimulated with syngeneic irradiated splenocytes as APC, pulsed with SERCA2a 971–990 or RNase 43–56. As shown in Fig. 4, middle panel, CD4 T cells responded to SERCA2a 971–990, but not to control (RNase 43–56), suggesting that the MHC class II–restricted CD4 T cells respond to SERCA2a 971–990. Furthermore, by evaluating proliferative responses in naive mice, we noted that the naive repertoire of A/J mice contained SERCA2a-reactive T cells as cells from naive animals responded to SERCA2a 971–990 Ag-specifically (Fig. 4, right panel). The data suggest that the endogenous repertoire of T cells in A/J mice contains a proportion of SERCA2a-reactive T cells that may expand in response to SERCA2a 971–990 in immunized animals.

CD4 T cell responses require presentation of peptides by MHC class II molecules, and we needed to confirm their expression in A/J mice. Flow cytometric analysis revealed that splenocytes from naive mice express both IA^k and IE^k, and their patterns were similar between all three professional APCs (CD11b⁺ macrophages, CD11c⁺ dendritic cells, and CD19⁺ B cells) (Fig. 5A). Next, we examined the ability of SERCA2a 971–990 to bind IA^k and IE^k molecules in an MHC binding assay. This effort required the creation of empty IE^k molecules, and we had previously described the generation of IA^k molecules (59–61). We created IE^k-α and IE^k-β constructs and after verifying protein expression in the baculovirus expression system and Ab affinity column purification, the CLIP 88–102-tethered IE^k monomers were treated with thrombin to generate empty IE^k molecules. The IA^k and IE^k molecules thus derived were used in a fluorescence-based DELFIA assay to determine the IC₅₀ values for SERCA2a 971–990 using HEL 46–61 and MCC 82–103 as reference peptides respectively for IA^k and IE^k alleles (49, 50, 54, 64). These analyses revealed that SERCA2a 971–990 was found to bind IA^k molecule with an IC₅₀, 8.09 ± 1.94 μM (Fig. 5A). Under similar conditions, the IC₅₀ value obtained with IE^k molecule was 56.77 ± 3.82 μM, leading us to conclude that SERCA2a 971–990 is a better binder of IA^k than IE^k molecule.

We recently established dextramer technology that permitted us to enumerate the frequencies of Ag-specific T cells in a variety of experimental systems (57, 61, 69, 70). We created MHC class II dextramers for IA^k molecule tethered with SERCA2a 971–990 by assembling the peptide sequence into the existing IA^k construct as we described previously (59–61). To generate IE^k monomers tethered with SERCA2a 971–990 or MCC 82–103, we replaced the sequence of CLIP 88–102 with the sequences of SERCA2a 971–990 and MCC 82–103 in the newly created IE^k-β constructs as described above. A biotinylation site was introduced into the IE^k-α construct. All IA^k and IE^k constructs were then expressed using the baculovirus expression system and the IA^k/SERCA2a 971–990 and IE^k/SERCA2a 971–990 or MCC 82–103 monomers were purified by Ab affinity columns. After biotinylation, dextramers were derived by conjugating IA^k and IE^k monomers with SA/fluorophore-conjugated dextran molecules (59–61).

For dextramer staining we used two sets of dextramers, namely IA^k/SERCA2a 971–990 (specific) and RNase 43–56 (control) (57, 59, 61) and IE^k/SERCA2a 971–990 (specific) and MCC 82–103 (control) (64). LNCs harvested from immunized animals were stimulated with SERCA2a 971–990 and later rested in the IL-2 medium and were analyzed for their ability to bind dextramers by flow cytometry. Fig. 5C shows binding of Ag-sensitized CD4 T cells to both IA^k/ and IE^k/SERCA2a 971–990 dextramers. The staining was specific because the staining intensities obtained with the corresponding control dextramers (RNase 43–56 or MCC 82–103) were negligible (Fig. 5C, left panel). The respective mean ± SEM values in percentages were as follows: IA^k dextramers: SERCA2a 971–990, 2.30 ± 0.41 versus RNase 43–56, 0.47 ± 0.10 (p = 0.02) and IE^k dextramers: SERCA2a 971–990, 1.09 ± 0.23 versus MCC 82–103, 0.37 ± 0.11 (p = 0.05) (Fig. 5C, right panel). The data suggest that the T cell responses induced with SERCA2a 971–990 were Ag specific and might have contributed to the cardiac inflammation found in the immunized animals.

Using a panel of cytokine Abs for Th1 (IL-2 and IFN-γ), Th2 (IL-10), Th17 (IL-17A), TNF-α, and IL-6, we determined the cytokine production by bead array analysis (49, 50). LNCs obtained from immunized animals were stimulated with or without SERCA2a 971–990 (specific) and irrelevant control (RNase 43–56) and the culture supernatants harvested on day 3 were used for cytokine analysis. The analyses revealed the presence of all cytokines tested, but their amounts varied in the order of Th1 and Th17 cytokines followed by TNF-α, IL-6, and IL-10 (Fig. 6). When the cytokine responses were compared between SERCA2a 971–990-stimulated cultures and controls, it was clear that production of IFN-γ, IL-17A, and TNF-α, but not IL-6, was significantly elevated in the immunized group (Fig. 6). These inflammatory cytokines are expected to be increased in the settings of autoimmunity. It is to be noted, however, that the production of IL-10, an anti-inflammatory cytokine, was also increased in SERCA2a 971–990 cultures, but its amount was negligible vis-à-vis inflammatory cytokines, in particular IFN-γ (17.07 ± 3.18 versus 6682.85 ± 749.44; 394-fold) and IL-17A (17.07 ± 3.18 versus 5080.41 ± 601.74; 298-fold). These observations suggest a possible relationship between the ability of SERCA2a peptides to produce proinflammatory cytokines and the occurrence of myocarditis in the immunized animals.

To address the above theme, we first evaluated T cell responses using a panel of 12 other SERCA2a peptides. In a proliferation assay, LNCs obtained from immunized animals responded to the immunized peptides in an Ag-specific manner, as expected, and the responses to control (RNase 43–56) were negligible (Supplemental Fig. 2a, 2b). However, by relating the T cell responses to the disease-inducing abilities of each peptide, we identified two patterns of peptides, that is, peptides that induce T cell responses but with mild myocarditis (SERCA2a 31–50, SERCA2a 161–180, SERCA2a 171–190, SERCA2a 401–420, and SERCA2a 951–970), and peptides that induce T cell responses but not disease (SERCA2a 311–330, SERCA2a 331–350, SERCA2a 471–490, SERCA2a 481–500, SERCA2a 591–610, SERCA2a 601–620, and SERCA2a 821–840) (Supplemental Fig. 2a, 2b, Table I). To determine whether mild disease induced with some of the SERCA2a peptides may be related to their differential cytokine production, we tested the supernatants representing two peptides, namely SERCA2a 31–50 and SERCA2a 161–180. Although their cytokine patterns were largely similar to the potent peptide that we had described above (SERCA2a 971–990; Fig. 6), the amounts of cytokines produced varied. Importantly, SERCA2a 31–50 produced mainly IFN-γ (Supplemental Fig. 2c, left panel), whereas SERCA2a 161–180 was found to be a poor producer of IFN-γ and IL-17A (Supplemental Fig. 2c, right panel). These data indicate that production of proinflammatory cytokines, in particular IFN-γ and IL-17A together, may contribute to SERCA2a 971–990-induced myocarditis, but other factors may also play a role.

We asked whether animals immunized with SERCA2a peptides produce Ag-specific autoantibodies and, if so, what their isotypes are, given the dominance of Th1 than Th2 responses. We addressed this question by analyzing the sera samples obtained from SERCA2a 971–990-immunized animals for total IgG, and various isotypes by using SERCA2a 971–990 (specific) and RNase 43–56 (irrelevant control) as plate-bound Ags by ELISA (Fig. 7). In addition, we used two other control groups, namely sera from naive mice and CFA/PT groups. These comparisons led us to make three observations: 1) Sera from the immunized animals reacted specifically to SERCA2a 971–990, but not RNase 43–56, and the amount of total IgG was significantly elevated in animals immunized with SERCA2a 971–990 as compared with naive or CFA/PT groups. 2) Although reactivity to IgM isotype was not specific to SERCA2a 971–990 as sera from naive and CFA/PT groups yielded similar reactivity suggesting a background response, the amounts of IgG1 and IgG3 were not significantly different in SERCA2a 971–990-immunized animals. 3) Production of two other isotypes, namely IgG2a and IgG2b, was significantly elevated in the SERCA2a 971–990 group Ag-specifically, and such a reactivity was absent for RNase 43–56 (Fig. 7). Because the total IgG and two IgG isotypes (IgG2a and IgG2b) were significantly elevated in SERCA2a 971–990-immunized animals [which also showed Ag-specific IFN-γ and IL-17A responses in significant amounts (Fig. 6)], SERCA2a 971–990 can potentially act as a common epitope for both B cells and T cells; moreover, IFN-γ and IL-17A might influence isotope switching (71).

To determine that the SERCA2a 971–990-sensitized T cells are pathogenic, we used an adoptive transfer protocol. Lymphocytes obtained from immunized animals were stimulated with a polyclonal activator, Con-A, and the viable cells were injected into naive mice primed with LPS and PT as we have previously described (49, 50). Heart sections prepared at termination on day 14 posttransfer showed lesions tended to be more in the atria than ventricles in animals that received the SERCA2a 971–990-sensitized T cells (Fig. 8). In contrast, heart sections from naive mice or those that received naive splenocytes and LNCs stimulated with Con-A were negative for inflammation. These findings are consistent with our previous observations that the animals receiving cells sensitized with an irrelevant control, RNase 43–56, also do not show lesions in the hearts (49). The data further reinforce the proposition that SERCA2a 971–990 induces cardiac inflammation through the generation of autoreactive T cells.

In this report, we have described the characterization of an immunodominant epitope of SERCA2a that induces mainly atrial inflammation in A/J mice by generating Ag-specific T cell and Ab responses. SERCA2a, 110 kDa protein, is a membrane enzyme and accounts for ∼40% of the protein in the SR (1, 2, 4), and because of the critical role in the Ca²⁺ homeostasis for sustained cardiac contractility, SERCA2a is being used as a therapeutic target in HF patients (72). SERCA2a is conserved in both animal and plant kingdoms and has been implicated in cellular proliferation, differentiation, apoptosis, and signal transduction events (73–75). That SERCA2a may act as an autoantigen has been proposed because patients with myocarditis and dilated cardiomyopathy can carry SERCA2a-reactive autoantibodies (29, 44, 47, 76). Experimentally, autoimmune myocarditis in SERCA2a-immunized mice was ascribed to autoantibodies that are believed to enter through T-tubules or their pores to bind SERCA2a (46, 47), but the antigenic determinants were unknown. Furthermore, because T cell help is critical for B cells to produce Abs, and autoimmune myocarditis is typically mediated by T cells (77–80), we sought to characterize the immunodominant epitopes of SERCA2a and determine their ability to induce cardiac autoimmunity.

For initial screening, a total of 80 overlapping peptides made in 18 pools were used to immunize A/J mice, and after identifying the positive pools that resulted in myocarditis, selected peptides were used for individual testing. These analyses led us to identify 13 peptides that induced disease of varying severity, and we selected SERCA2a 971–990 as the potent disease inducer for further characterization. Remarkably, the disease induced with SERCA2a 971–990 was distinct in that inflammation and fibrosis were confined to atria in the majority of animals and ventricles were less frequently affected. When they were affected, however, inflammation was noted in all three layers of the heart (epicardium, myocardium, and endocardium), leading to the conclusion that SERCA2a 971–990 can act as an autoantigen in the induction of atrial inflammation. Atrial fibrillation (AF) is one of the common arrhythmias seen in clinical practice (81). There is growing evidence that atrial inflammation may act as an important trigger of AF (82). This is important because cardiac fibrosis, being considered as one hallmark of AF, can progressively arise from various causes (83). In fact, overt or low-grade subclinical inflammation, including expression of various inflammatory cytokines and chemokines, can culminate into apoptosis of cardiomyocytes, recruitment of fibroblasts, and extracellular matrix deposition and also progression from paroxysmal to chronic AF or recurrence of AF (83, 84). Therefore, anti-inflammatory treatment options are being considered in the management of AF patients (82, 85). We thus propose that the SERCA2a 971–990-induced atrial myocarditis model may be useful to study the AF that occurs in humans, but the relevant immunodominant epitopes of SERCA2a may be influenced by HLA haplotypes.

Although the structural and functional abnormalities of hearts in SERCA2a 971–990-immunized animals could be captured noninvasively by echocardiography and MRM imaging, IHC analysis revealed the presence of CD4 T cells to be more than CD8 T cells, and non-T cells in the decreasing order of macrophages, neutrophils, and B cells. Of these, detection of increased numbers of CD4 T cells than CD8 T cells, and macrophages signify the hallmarks of delayed-type hypersensitivity reaction, as expected in a T cell–mediated disease (86–88); neutrophils might have emigrated as a part of the ongoing inflammatory process. As reported previously in other mouse strains (89), A/J mice also carry a proportion of T cells that are double positive for both CD4 and CD8 coreceptors, and the presence of CD8 T cells in heart infiltrates may represent this subset of T cells. Detection of B cells may suggest a role for them either alone or together with T cells in the mediation of atrial inflammation.

Mechanistically, we first determined that SERCA2a 971–990 induces T cell responses by using LNCs from immunized animals. To prove their Ag specificity, we decided to create MHC class II dextramers for SERCA2a 971–990. This effort required us to create dextramers for two different MHC class II alleles (IA^k and IE^k) as A/J mice express both. Initially, we confirmed that SERCA2a 971–990 binds both alleles with the affinity being better for IA^k than IE^k, but both IA^k- or IE^k-tethered, SERCA2a 971–990 dextramers bind CD4 T cells with specificity suggesting that SERCA2a 971–990 displayed by both alleles can trigger T cell responses. In addition, by using splenocytes from naive mice, we noted that their peripheral repertories contain a proportion of SERCA2a 971–990-reactive T cells. Reports indicate that SERCA2a is expressed at a low level in the fetal thymus in both humans and mice (1) and detection of SERCA2a-reactive T cells may mean that the developing thymocytes might have escaped negative selection as reported for other autoantigens like cardiac myosin and proteolipid protein (90, 91). Translationally, if such a scenario exists in humans, then the genetically susceptible individuals can potentially carry SERCA2a-reactive T cells making them vulnerable to the development of atrial myocarditis in infection or nonspecific inflammatory conditions that can trigger autoreactive T cells through bystander activation (92).

We next investigated the pathogenic potential of T cell responses induced by SERCA2a 971–990 by evaluating cytokine production and the analysis revealed the dominance of both IFN-γ–producing Th1 and IL-17A–producing Th17 responses followed by TNF-α. Although IL-10 production was also noted in SERCA2a 971–990-sensitized cultures, the amount was negligible in relation to IFN-γ and IL-17A implying that the effects of inflammatory cytokines are expected to prevail in the local milieu as we have recently noted with two mitochondrial Ags, adenine nucleotide translocator₁ (50) and branched chain α-ketoacid dehydrogenase kinase (49). Generally, it is held that the pathogenic self-reactive T cell responses are steered toward Th1 and Th17 cytokine phenotypes (93–95). However, unlike the autoimmune myocarditis model induced by cardiac myosin H chain-α 334–352, in which some degree of contradiction exists as to whether Th1 cytokines act as pro- or anti-inflammatory (96), detection of comparable amounts of both Th1 and Th17 cytokines in SERCA2a 971–990-sensitized cultures may point to a possibility that both subsets of cytokines may contribute to the disease pathogenesis. In support of this proposition, T cells sensitized with SERCA2a 971–990 could transfer the disease to naive mice. However, it should be noted that a theme has now emerged that although Th1 cytokine IFN-γ mediates initiation of myocarditis, IL-17A is essential for progression of myocarditis leading to dilated cardiomyopathy (55, 94, 97). In our model, we have not investigated whether one subset of Th cytokines is dispensable over the other (Th1 versus Th17) in the induction of atrial myocarditis as mice deficient for IFN-γ or IL-17 on A/J background are not available. We believe that both cytokines are essential for disease induction by SERCA2a 971–990 based on the observation that two other epitopes that induce only mild disease generate mainly IFN-γ (SERCA2a 31–50) or underproduction of both IFN-γ and IL-17A (SERCA2a 161–180) in relation to SERCA2a 971–990.

We also investigated whether SERCA2a 971–990 can induce Abs because B cells were detected in the heart sections from SERCA2a 971–990-immunized animals, but not in the control groups. Analysis of Ab responses revealed that the sera obtained from SERCA2a 971–990-immunized group showed significantly higher amounts of total IgG, and IgG2a and IgG2b isotypes. Because SERCA2a 971–990-sensitized T cells produce a significant amount of IFN-γ that can induce isotype switching of IgG2a (98), and to a lesser extent that IL-17A can also influences IgG2b-switching (71), the SERCA2a 971–990-specific T cells might have facilitated isotype switching in SERCA2a 971–990-specific B cells. Whether SERCA2a 971–990-reactive Abs induce the disease like SERCA2a 971–990-reactive T cells is currently unknown. This effort involves the generation of large quantities of Ag-specific polyclonal sera or mAbs to be able to test in an adoptive transfer setting.

In summary, we have described that SERCA2a 971–990 is a major immunodominant epitope that induces preferentially atrial inflammation in A/J mice. SERCA2a expression has been reported to be more in atria than in ventricles (15), and we confirm that A/J mice also follow a similar pattern (Fig. 3A). Although this finding provides support for the preferential occurrence of inflammation in atria, how SERCA2a-reactive T cells can possibly see the Ag expressed by the SR is currently unknown. Reports suggest that SR lies close to the T-tubules which are the invaginations of the plasma membrane (12), and it may be possible that SERCA2a 971–990 located in the C-terminal cytoplasmic loop may gain an access to the cell membrane by unknown mechanism. It may be possible that the local resident APCs may display peptide fragments of SERCA2a, and such a possibility has been shown for cardiac myosin in A/J mice (99, 100). Likewise, acquisition of self-Ags by the APCs may be possible through phagocytosis of dead cells or autophagy (101). Similarly, it is possible that misfolded intracellular proteins, as might occur due to mutations in mitochondrial proteins like pyruvate dehydrogenase, superoxide dismutase, and branched chain α-ketoacid dehydrogenase kinase (102–105) or self-Ags undergoing posttranslational modifications acting as neoantigens, can be seen by the immune system as foreign, leading to the induction of autoimmune responses (106). Whether these speculative theories are relevant to SERCA2a remain to be investigated in the future. We had expected that SERCA2a 971–990-immunized mice might show inflammatory changes in noncardiac tissues because SERCA2b that expresses ubiquitously also carry this epitope in both humans and mice, but it was not the case. Such a finding may also lend support for a proposition that the immunodominant epitopes of SERCA2a might be presented by the local APCs within the hearts, but not in other organs. Furthermore, in addition to SERCA2a 971–990, epitope mapping also revealed the existence of at least 12 other epitopes capable of inducing T cell responses and they were grouped into those that induce only mild or no disease. Although the existence of such a large number of immunodominant epitopes within a single protein is relatively unusual, the finding that A/J mice express two MHC alleles may facilitate the presentation of multiple epitopes. The availability of such a model system may also create opportunities to investigate the phenomenon of intramolecular epitope spreading in inflammatory cardiomyopathies induced by cardiotropic pathogens like Coxsackie virus B and Trypanosoma cruzi.

We thank the sonographer, Leanne Harmman, for the echocardiogram images, and Immudex (Copenhagen, Denmark) for providing dextramer reagents (Bjarke Endel Hansen).

The authors have no financial conflicts of interest.