key: cord-0023119-8e8an5xp
authors: Lee, Ahreum; Yoo, Duck Kyun; Lee, Yonghee; Jeon, Sumin; Jung, Suhan; Noh, Jinsung; Ju, Soyeon; Hwang, Siwon; Kim, Hong Hee; Kwon, Sunghoon; Chung, Junho; Choi, Youngnim
title: Induction of Anti-Aquaporin 5 Autoantibody Production by Immunization with a Peptide Derived from the Aquaporin of Prevotella melaninogenica Leads to Reduced Salivary Flow in Mice
date: 2021-10-22
journal: Immune Netw
DOI: 10.4110/in.2021.21.e34
sha: 721134efb02d8ad0e37dac0e254b4c1bf4f7e66b
doc_id: 23119
cord_uid: 8e8an5xp

Sjögren's syndrome (SS) is an autoimmune disease characterized by dryness of the mouth and eyes. The glandular dysfunction in SS involves not only T cell-mediated destruction of the glands but also autoantibodies against the type 3 muscarinic acetylcholine receptor or aquaporin 5 (AQP5) that interfere with the secretion process. Studies on the breakage of tolerance and induction of autoantibodies to these autoantigens could benefit SS patients. To break tolerance, we utilized a PmE-L peptide derived from the AQP5-homologous aquaporin of Prevotella melaninogenica (PmAqp) that contained both a B cell “E” epitope and a T cell epitope. Repeated subcutaneous immunization of C57BL/6 mice with the PmE-L peptide efficiently induced the production of Abs against the “E” epitope of mouse/human AQP5 (AQP5E), and we aimed to characterize the antigen specificity, the sequences of AQP5E-specific B cell receptors, and salivary gland phenotypes of these mice. Sera containing anti-AQP5E IgG not only stained mouse Aqp5 expressed in the submandibular glands but also detected PmApq and PmE-L by immunoblotting, suggesting molecular mimicry. Characterization of the AQP5E-specific autoantibodies selected from the screening of phage display Ab libraries and mapping of the B cell receptor repertoires revealed that the AQP5E-specific B cells acquired the ability to bind to the Ag through cumulative somatic hypermutation. Importantly, animals with anti-AQP5E Abs had decreased salivary flow rates without immune cell infiltration into the salivary glands. This model will be useful for investigating the role of anti-AQP5 autoantibodies in glandular dysfunction in SS and testing new therapeutics targeting autoantibody production.

Sjögren's syndrome (SS) is an autoimmune disease characterized by dry mouth, dry eyes, and focal lymphocytic infiltration of the salivary/lacrimal glands (1, 2) . Hypergammaglobulinemia and the production of various autoantibodies, including those against SS-related Ag A (SSA), SS-related Ag B (SSB), IgG, type 3 muscarinic acetylcholine receptor (M3R), and aquaporin 5 (AQP5), are also features of SS (3) (4) (5) . The glandular dysfunction in SS involves not only the destruction of acinar cells by infiltrated T cells but also the presence of autoantibodies against M3R or AQP5 that interfere with the secretion process. AQP5 is a major water channel protein expressed in the lacrimal and salivary glands and plays a critical role in tear and saliva secretion (6) . Three functional epitopes for the anti-AQP5 autoantibodies have been mapped at extracellular loops A, C, and E of AQP5 (7) . The presence of anti-AQP5 autoantibodies was associated with low salivary flow rates in patients from the Korean Initiative of Primary Sjogren's Syndrome but not in those from the Sjögren's International Collaborative Clinical Alliance registry (5, 8) .

The production of autoantibodies is the result of breached self-tolerance. B cells acquire tolerance to self-Ags during development in the bone marrow and maturation in peripheral lymphoid tissues via deletion, receptor editing, or anergy (9) (10) (11) . Infection is a common environmental factor associated with autoimmune diseases and is thought to disrupt selftolerance through bystander activation of Ag-presenting cells (APCs) or molecular mimicry. APCs activated by microbes can result in the activation of self-reactive T cells. In molecular mimicry, B cells or T cells activated by microbial Ags cross-react with autoantigens that have the sequence or conformational homologies with the microbial Ags.

The salivary glands, one of the target organs of SS, are inevitably affected by bacteria colonizing the oral cavity, and several studies have reported an altered oral microbiota in SS (12) (13) (14) . Ductal cells and the areas of lymphocytic infiltration in the labial salivary glands from patients with SS are heavily infected with bacteria, including the oral commensal Prevotella melaninogenica (Pm). The aquaporin of Pm (PmAqp) is highly homologous to human AQP5, particularly at the regions of extracellular loop E and intracellular loop B that form a single narrow aqueous pathway (15) (16) (17) .

The structures of mouse Aqp5 (accession: NP_033831.1) and PmAqp (accession: BBA28519.1) were constructed on the SWISS_MODEL server (https://swissmodel.expasy.org/) using a known human AQP5 structure (structure ID: PDB 3D9S) as a template. The 3D structures of peptides used for either immunization or ELISA were predicted at the PEP-FOLD3 server (https://mobyle.rpbs.univ-paris-diderot.fr/cgi-bin/portal.py#forms::PEP-FOLD3/).

The experimental protocols and animal handling procedures were approved by the Seoul National University Animal Care and Use Committee, Seoul, Republic of Korea (No. SNU-180508-2-2). Female C57BL/6 mice purchased from Orient Bio (Seongnam, Korea) were maintained under specific pathogen-free conditions in the laboratory animal facility at the School of Dentistry, Seoul National University. Mice used in experiments were 6-wk old.

Pm KCTC 5457 (Korean Collection for Type Cultures, Daejeon, Korea) was anaerobically cultured in modified PYG medium supplemented with 10 µg/ml vitamin K and 5 µg/ml hemin. After washing with PBS, bacteria resuspended in PBS at 1 × 10 8 cells/ml were lysed by repeated freeze-thaw cycles. Peptides used for immunization and ELISA were synthesized by Peptron (Daejeon, Korea). The sequences of peptides are previously reported (17) . The M3R Ag peptide (biotin-SCIPKTYWNC [C-C]) for ELISA was slightly modified from the previously reported one by adding biotin and serine to the N-terminus (18) .

To induce anti-AQP5 autoantibodies, mice were primed with 100 µl of Pm lysate on day 0 and boosted with either 100 µl of Pm lysate or 100 µg of peptide PmE-L (a cyclic form) emulsified in incomplete Freund's adjuvant (IFA) on days 10, 24, and 38. As controls, mice in the sham group received PBS and IFA alone for priming and boosting, respectively. All immunizations were performed by subcutaneous injection on both sides of the tail base. Mice were sacrificed 10-14 days after the last boosting.

After anesthesia, pilocarpine (5 mg/kg of body weight) was intraperitoneally administered to stimulate saliva secretion. Saliva was collected for 10 min after pilocarpine stimulation and weighed. The salivary flow rate was expressed as mg of saliva/g of body weight.

Single-cell suspensions of the draining (inguinal) lymph nodes of immunized mice were prepared. Cells were stained with FITC-conjugated AQP5E peptide (Peptron), anti-CD19 (PE-Cy7-6D5 from BioLegend, San Diego, CA, USA), anti-GL7 (eFluor 660-GL7 from eBioscience, San Diego, CA, USA), anti-CD38 (BV605-90 from BD Bioscience, Franklin Lakes, NJ, USA), and Ghost Dye™ viability dye (Tonbo, San Diego, CA, USA). The data were acquired on an Aria II flow cytometer (BD Bioscience) and analyzed using FlowJo software after gating live cells based on the forward scatter and exclusion of Ghost Dye™.

The submandibular salivary glands from mice were fixed in 4% paraformaldehyde at 4°C, cryopreserved with 10% to 30% sucrose, and embedded in OCT compound (Sakura Finetek, Torrance, CA, USA). The frozen sections were stained with H&E and examined for focal lymphocytic sialadenitis.

Sections of the salivary glands from the sham group were also used for indirect immunofluorescence. After Ag retrieval by incubation in sodium citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH=6) at 105°C for 20 min, the sections were blocked with 5% BSA plus goat serum and then incubated with rabbit anti-AQP5 polyclonal Abs (1:100; ATLAS Antibodies, Bromma, Sweden) and mouse sera (1:10) overnight. After washing, the sections were incubated with Alexa Fluor 488-conjugated donkey anti-rabbit IgG (Invitrogen, Carlsbad, CA, USA) and Alexa Fluor 594-conjugated goat anti-mouse IgG (Invitrogen) for 1.5 h. The sections were washed, mounted, and examined by confocal microscopy (Carl Zeiss, Oberkochen, Germany).

The levels of anti-SSA (Ro-52) autoantibodies in sera were measured using a kit from Signosis (Santa Clara, CA, USA) according to the manufacturer's instructions.

The levels of anti-PmE-L, anti-PmE, anti-AQP5E, and anti-M3R Abs were determined using biotinylated peptides as Ags. Microplates were coated with 1 µg/well avidin (Sigma, St. Louis, MO, USA) in PBS overnight at 4°C. After blocking with 1% BSA, the plates were incubated with 0.2 µg/well biotinylated Ag peptides in PBS for 1 h at room temperature. After the plates were washed, they were incubated with serum (1:300 dilution) for 1 h at room temperature. In parallel, samples were also incubated in wells coated with avidin alone. After washing, the plates were incubated with HRP-conjugated goat anti-mouse IgG (Southern Biotech, Birmingham, AL, USA) for 1 h at room temperature. The plates were washed on a shaker for 5 min 3 times. Bound HRP-conjugated detection Abs were developed with 3,3′,5,5′-tetramethylbenzidine substrate (Sigma). After stopping the enzyme activity by the addition of 2 N H 2 SO 4 , the optical density (OD) values were measured at 450 nm. To generate a standard curve, serially diluted mouse IgG1 (40 to 0.625 ng/ml, BD Bioscience) was coated in the two columns of each plate instead of Ags. The amount of bound Ag-specific IgG was determined using an equation generated from the standard curve and OD after subtracting that obtained from incubation with avidin alone. The levels of total IgG were determined using a mouse IgG total quantification kit (Invitrogen) according to the manufacturer's instructions.

To determine the Ag-specific binding of phage clones or recombinant IgG, ELISA was performed as described above by incubating the phage clones or recombinant IgG instead of sera. Binding was expressed as the OD.

Pm lysates (5 and 15 µg) and PmE-L (1 µg) were separated through a 15% SDS-polyacrylamide gel and transferred to a PVDF membrane (Merck Millipore, Billerica, MA, USA) by wet transfer. After blocking with 5% skim milk, the membrane was incubated with sera from immunized mice at a 1:5,000 dilution for 10 h at 4°C. After washing, the membrane was incubated with HRP-conjugated anti-mouse IgG (GenDEPOT, Barker, TX, USA) for 1 h at room temperature. Immunoreactive bands were detected with electrochemiluminescence reagents.

From the immunized mice, total RNA was isolated from the draining lymph nodes using TRIzol Reagent (Invitrogen), and cDNA was synthesized using the SuperScript IV First-Strand Synthesis System with oligo dT priming (Invitrogen). Using this cDNA, phage-displayed Ab libraries were prepared as described previously (19) , and then subjected to seven rounds of biopanning using PmE-conjugated magnetic beads (Invitrogen). After the final round of biopanning, phage clones were subjected to phage ELISA using PmE, AQP5E, and AQP5A peptides as Ags, as described previously with adequate modification (20) . The sequences of specific binders that bound to PmE and AQP5E but not to AQP5A were determined by Sanger sequencing (Macrogen, Seoul, Korea).

To analyze the B cell receptor (BCR) repertoire after immunization, double-stranded cDNA libraries were constructed using in-house designed mouse immunoglobulin-specific primers. First-strand cDNA was synthesized using 1 µg total RNA and a SuperScript IV First-Strand Synthesis System and mouse constant gene-specific reverse primers with a P7 sequence (5′-TGACTGGAGTTCAGACGTGTGCTCTTCCGATCT-3′) and a UMI barcode (21) . Synthesized cDNA was purified using AMPure XP beads (Beckman Coulter, Brea, CA, USA) according to the manufacturer's instructions. Second-strand cDNA was synthesized using KAPA HiFi HotStart DNA polymerase (Kappa Bioscience, Oslo, Norway) and mouse variable gene-specific forward primers with a P5 sequence (5′-ACACTCTTTCCCTACACGACGCTCTTCCGATCT-3′) (98°C for 4 min, 60°C for 1 min, 72°C for 5 min). Double-stranded cDNA was purified using AMPure XP beads and subjected to PCR amplification with KAPA HiFi HotStart DNA polymerase using two universal primers containing Illumina adapters and index sequences (95°C for 3 min; 25 cycles of 95°C for 30 s, 65°C for 30 s, and 72°C for 1 min; and 72°C for 5 min). The final products were purified using AMPure XP beads and subjected to a quality control procedure on TapeStation 2200 (Agilent Technologies, Santa Clara, CA, USA) as previously described (20) . Libraries with a single peak of the correct sequence length were subjected to NGS analysis using the MiSeq platform (Illumina Inc., San Diego, CA, USA) with 2 × 300 paired-end run mode. All NGS raw data are available at the National Center for Biotechnology Information Sequencing Read Archive (www.ncbi.nlm.nih.gov/sra) under PRJNA739054.

All raw paired FASTQ files were subjected to the following processes: i) paired-end merging and quality filtering, ii) error correction, and iii) region annotation. Paired-end merging was performed using PEAR software (22) with default parameters, followed by quality filtering using an in-house Python 3.6 script with the Q20P95 option, which extracted the reads if more than 95% of bases had a Phred quality score > 20. Error correction on quality-filtered data was performed based on the unique molecular identifier, following the previously described methodology (23) . In the annotation process, the V/D/J gene, complementarity determining regions (CDRs; CDR1, CDR2, and CDR3), isotype, and divergence from the V gene were obtained using the IgBLAST tool (24) with the Ig germline database of C57BL/6 mice acquired from the IMGT database (25) . The divergence from the V gene was defined as the ratio of somatic hypermutations (SHM) in the V gene region.

From the five mice, a total of 680,326 error-corrected reads were obtained, from which a total of 669,459 functional reads were annotated ( Table 1) . To normalize the data size, 25,800 sequences in each repertoire were randomly sampled, and the sampling size was chosen considering the minimum number of functional sequences among the five repertoires. To identify AQP5E-specific binder-like BCR sequences, dissimilarity to the binder, which was defined as the ratio of Levenshtein distance in the target region (CDR3), was calculated for all unique VDJ sequences in each repertoire.

The variable genes encoding clone 44 were cloned into a modified expression vector containing the human IgG4 construct, and transfected into Expi293F cells (Invitrogen) as described previously (26) . IgG was purified by affinity chromatography using MabSelect columns with the AKTA Pure chromatography system (GE Healthcare, Chicago, IL, USA) following the manufacturer's protocol. An in-house discovered Ab specific to Zika virus envelope domain III (27) was also produced in human IgG4 format and used as an isotype control.

The osmotic flow of water across a cell monolayer was measured by a previously described method (7) . MDCK cells expressing human AQP5 (MDCK-AQP5) were grown on a 0.33-cm 2 Transwell insert for 4 days until the transepithelial resistances was greater than 320 Ωcm 2 . Recombinant IgG (clone 44) was preabsorbed at 0.5 mg/ml with 1 mg of peptide E or A in serum-free DMEM overnight at 4°C. The cells were incubated with 50 µl of nonabsorbed or preabsorbed IgG on ice for 1 h. After washing, osmotic shock was applied by adding 400 mM D-mannitol to the Transwell insert. After incubation for 20 min at 37°C, the volume of fluid in the Transwell insert was measured. Transepithelial osmotic permeability coefficient P f T (µm/s) was calculated as P f T = J v T /Sv w Δ[Φ] T , where J v T is the measured transepithelial flow (cm 3 /s), S is the transwell surface area (0.33 cm 2 ), v w is the partial molar volume of water (18 cm 3 /mol), and Δ[Φ] T is the transepithelial osmotic gradient (4×10 -4 mol/cm 3 ) (28).

All data obtained from the mouse experiments were analyzed with a nonparametric approach (Kruskal-Wallis or Mann-Whitney test and Spearman's correlation). The difference in the transepithelial osmotic permeabilities was evaluated by Student's t-test.

PmAqp and the PmE-L peptide that was designed for immunization have two and one amino acid differences, respectively, at the B cell "E" epitope compared with that of mouse Aqp5 (17) . The 3D structures of mouse Aqp5, PmAqp, and the Ag peptides used for immunization or ELISAs were predicted, and the 3D structures of the E epitope in mouse Aqp5, PmAqp, and the three peptides were relatively well conserved (Fig. 1) 

We previously confirmed that repeated immunization with the PmE-L peptide emulsified in IFA induced autoantibodies against AQP5E. To investigate the possibility that Pm infection induces the production of anti-AQP5 autoantibodies in SS patients, mice were repeatedly immunized with Pm lysate and compared with the mice immunized with PmE-L ( Fig. 2A) .

Although the levels of total IgG were increased even more in the Pm group than in the PmE-L group, the levels of IgG Abs against PmE-L, PmE, and AQP5E were significantly increased only in the PmE-L group compared with the sham group. One mouse (11%) in the Pm group produced anti-AQP5E IgG at a concentration above 1 µg/ml, a predetermined threshold for Ab positivity, while 8 mice (89%) in the PmE-L group were positive for anti-AQP5E IgG (Fig. 2B) .

A strong positive correlation was observed between the levels of anti-PmE Abs and anti-AQP5E autoantibodies (r=0.914, p<0.0005; Fig. 2C ). The frequency of AQP5E-specific germinal center B cells in the draining lymph nodes was examined by flow cytometry using a FITC-conjugated AQP5E peptide (Fig. 2D) . Expansion of AQP5E-specific cells, along with that of germinal center B cells, was observed in both the Pm and PmE-L groups ( Fig. 2E and F) .

The mice positive for the anti-AQP5E autoantibodies had reduced salivary flow rates (Fig. 2G) .

Although there was a wide variation in the salivary flow rates, the salivary flow rate tended to decrease with increased anti-AQP5E IgG levels (Fig. 2H) . Despite the reduced salivary flow rate, no pathological changes, including focal lymphocytic sialadenitis, was observed in the salivary glands of the mice (data not shown). Furthermore, anti-SSA autoantibodies were not detected in any of the mice (Fig. 2I) . Interestingly, anti-M3R autoantibodies were detected in a few mice in each group (Fig. 2J) .

For confirmation of the antigenic specificity of anti-AQP5E autoantibodies, which was determined by binding to the AQP5E peptide, sections of mouse submandibular salivary glands were double stained with mouse sera and commercial anti-AQP5 Abs. While the anti-AQP5E-positive sera from the PmE-L immunized mice stained Aqp5 in the salivary glands, the sera from the sham group did not (Fig. 3A) . The binding of Abs produced by immunization with PmE-L to PmAqp was also confirmed by immunoblotting. Multiple bands, including the PmAqp band expected to be 23.6 kDa, were detected by the sera with anti-AQP5E IgG but not by the sera from the sham group (Fig. 3B, Lanes 1&2) . Antibodies in the immunized sera also detected the PmE-L peptide (Fig. 3B, Lane 3) . Thus, the foreign Ag PmE-L induced the production of anti-AQP5E autoantibodies via molecular mimicry. 

To delineate the V H and V L sequences of the anti-AQP5E autoantibodies, we prepared 2 sham (M1-M2) and 3 PmE-L-immunized (M3-M5) mice with varying concentrations of autoantibodies (Fig. 4A) . To isolate AQP5E-specific Ab clones, we constructed phage display libraries using the draining lymph nodes from the immunized mice (Fig. 4B) . After biopanning, specific binders were identified only in the M5 libaray, which had the highest concentration of anti-AQP5E IgG among those of the 3 immunized mice. Sequence analysis of 6 phage clones revealed that all clones used the IGHV5-6, IGHD2-3, and IGHJ2 gene segments in the V H and IGKV14-111 and IGKJ4 gene segments in the V L . In addition, the clones (3, 37, and 44) with a complete V H domain had identical sequences in the CDRs of V H and V L , although there were mismatches in the nucleotide sequences ( Fig. 5A and B, Supplementary Fig. 1) . Sequencing of additional specific binder clones also revealed CDR sequences identical to those of clone 44 (data not shown).

Recombinant IgG was synthesized using the V H and V L encoding clone 44 and subjected to ELISA. Clone 44 recombinant IgG bound to PmE and AQP5E but not to AQP5A to confirm the specificity of the identified clones (Fig. 5C ). In addition, clone 44 recombinant IgG inhibited water transport through AQP5 compared with a control clone, but this change was reversed by preincubation with the AQP5E peptide (Fig. 5D ).

The Ag binding sites of Abs are formed with six CDRs provided by the V H and V L domains, but CDR3 of the V H domain is sufficient for most Ab specificity (29) . We hypothesized that the AQP5E-specific Abs detected in the M3 and M4 mice may have substantial similarity with clone 44 in V H CDR3 sequences. Therefore, the mismatch ratios in the V H CDR3 sequences compared with that of clone 44 were calculated for all unique VDJ sequences in each repertoire, and clone 44-like sequences defined as those with a mismatch ratio ≤0.2 (i.e., similarity ≥80%) were retrieved. The 80% threshold for similarity was adopted from the literature (30) . The proportion of clone 44-like sequences in M5 increased from 0.19% to 0.55% with an increase in the allowed mismatch ratio from 0 to 0.2. Clone 44-like sequences were also observed in other repertoires in the following order: M3>M4>M1>M2 (Fig. 6A) . To infer the immunological status of B cells from which the VDJ sequences originated, we analyzed the isotype compositions of the clone 44-like sequences. All sequences from the M1 and M2 sham mice were IgM isotypes, which are known as the naïve form of Abs, while the sequences from the PmE-L-immunized mice were predominantly IgG1 isotypes (Fig. 6B) . When the same analysis was applied only to VDJ sequences from the IGHV5-6 gene, clone 44-like sequences were found only in M5, which were also dominated by the IgG1 isotype ( Fig. 6C and D) .

We also analyzed the frequencies of IGHV5-6 gene usage in each repertoire by isotype. According to the literature, the frequency of IGHV5-6 usage in the IgM-associated VDJ rearrangements of C57BL/6 splenocytes is 0.38% (31) . The frequency of IGHV5-6 usage in the IgM-associated sequences was particularly low (0.05%) in M5, provably due to activation and isotype switching of some of the IgM clones. Although the frequency of the IGHV5-6-used IgG1-associated sequences was as high as 1.2% in M5, it was even higher in M2, one of the sham mice, indicating that the abundance of IGHV5-6-used sequences does not necessarily reflect the expansion of AQP5E-specific B cells (Fig. 6E) . For determination of whether B cells experienced affinity maturation pressure by immunization, the divergence of the IGHV5-6-used sequences from the V gene was computed and compared between the sham and PmE-L-immunized groups by isotypes. Sequences associated with IgG1, IgG2b, IgG2c, and IgM had significantly more SHM in the PmE-L-immunized group than in the sham group (Fig. 6F) . Scatter plots of the divergence of the IGHV5-6-used sequences against similarity to clone 44 in V H CDR3 revealed that the limited SHM observed in the sham group did not increase the similarity compared with those without SHM. In contrast, like clone 44, which had a divergence of 0.034, most clone 44-like sequences (similarity ≥80%) in the PmE-L group had a divergence >0.02. In addition, the similarity to clone 44 had a positive correlation with divergence, suggesting that the AQP5Especific B cells acquired the ability to bind to AQP5E through cumulative SHM.

In this study, we characterized the Ag specificity and variable gene sequences of AQP5E-binding Abs along with salivary gland phenotypes in an anti-AQP5 autoantibody-producing mouse model. Repeated immunization with the PmAqp-derived peptide PmE-L efficiently induced anti-AQP5E autoantibodies through molecular mimicry. There was a strong positive correlation between the levels of anti-PmE IgG and those of anti-AQP5E IgG (Fig. 2C) . The sera containing the anti-AQP5E IgG not only stained mouse Aqp5 expressed in the submandibular glands but also detected PmAqp and the immunizing peptide by immunoblotting ( Fig. 3A  and B) . Furthermore, recombinant clone 44 IgG, which was produced using the variable genes encoding the AQP5E-specific phage clone 44, bound to both PmE and AQP5E. All these findings indicate that the Abs produced against PmE are cross-reactive with AQP5E. This cross-reactivity is not surprising since there is only one amino acid difference between PmE and AQP5E, and the predicted structure of PmE-L mimics the structure of AQP5E located in mouse Aqp5 (Fig. 1) Characterization of AQP5E-specific autoantibodies selected from the screening of phage display Ab libraries and mapping of the BCR repertoires revealed that the AQP5E-specific B cells acquired the ability to bind to the Ag through cumulative SHM. Interestingly, all AQP5Especific clones, including clone 44, picked in the phage Ab library of M5 used the IGHV5-6 gene segment, and the divergences of the IGHV5-6-used sequences were significantly increased in the PmE-L-immunized repertoires compared with the sham repertoires, particularly in the IgGassociated sequences (Figs. 5A and 6F) . Furthermore, most clone 44-like sequences identified in the PmE-L-immunized repertoires had divergence >0.02, while the IGHV5-6-used sequences in the sham repertoires had divergence <0.02 (Fig. 6G) . This finding indicates that clone 44 and other AQP5E-specific clones gained the ability to bind to AQP5E through SHM. The PmE-L peptide used in immunization must have driven SHM and selection as a foreign Ag.

Importantly, the presence of anti-AQP5E autoantibodies was associated with a low salivary flow rate in mice (Fig. 2C) , suggesting the role of anti-AQP5 autoantibodies in the dryness of SS patients. Inhibition of water permeability through AQP5 expressed in MDCK cells by recombinant clone 44 IgG (Fig. 5D ) supports this notion. However, there was a broad variation in salivary flow rates among the mice with the same levels of anti-AQP5E IgG. The concentrations of Abs determined by ELISAs do not reflect affinity to Ags, and the ability of each clone of the polyclonal anti-AQP5E autoantibodies to interfere with AQP5 function must vary. In addition, seven mice positive for anti-M3R autoantibodies, which also interfere with the salivary secretion process, were distributed in all 3 groups. These factors may account for the variable associations of anti-AQP5 autoantibodies with low salivary flow rates in SS patients (5, 8) .

Despite the reduced salivary flow rate, no histologic abnormality was observed in the salivary glands of the mice with anti-AQP5E autoantibodies. Similarly, no immune cell infiltration was reported in curdlan-injected SKG mice despite increased IgG deposition in the salivary glands and reduced salivary flow rates (32) . The authors attributed this finding to the fact that IgG1 was the major isotype of autoantibody produced and that murine IgG1 is ineffective in complement and FcR activation. In IgG4-related diseases in humans, however, the deposition of IgG4, the human version of murine IgG1, is associated with sialadenitis (33). Induction of focal lymphocytic sialadenitis in C57BL/6 mice by repeated immunization with salivary gland proteins emulsified in Freund's complete adjuvant has been shown; this condition was aggravated at 30 weeks after the first immunization compared with that at week 5 and depended on Th17 cells (34) . When we boosted mice on days 10, 24, and 64 and then examined the salivary glands of the mice on day 110 in our other experiments, we still found no sialadenitis, although the anti-AQP5E autoantibodies were detected (data not shown). The results of three different models suggest that salivary gland-targeting autoantibodies alone do not necessarily lead to focal lymphocytic sialadenitis. Infection of the salivary glands with bacteria observed in SS patients may also contribute to immune cell infiltration (15) .

Repeated immunization with Pm lysates induced anti-AQP5E autoantibodies in only 11% of the immunized mice, although the concentrations of total IgG were higher than those in the PmE-L group. The reason for this inefficient induction of anti-AQP5E autoantibody production by Pm is mainly due to the paucity of Ag. The amount of protein in 1×10 7 Pm cells used for immunization was approximately 80 µg. According to the immunoblotting results (Fig. 2G) , the molar amount of PmAqp present in the 80 µg Pm lysates was estimated to be lower than 1 µg PmE-L peptide. In SS patients, the infection of salivary glands with Pm (15) may contribute to the production of anti-AQP5 autoantibodies not only through molecular mimicry but also via bystander activation of APCs.

In this model, Pm lysate, which was used for priming, and IFA can provide bystander activation of APCs. Anti-M3R autoantibodies detected in several mice may be due to bystander activation because one of the anti-M3R Ab-positive mice was in the sham group, and M3R is ubiquitously expressed, including in muscle, adipose, and lymphoid tissues (35) . However, Aqp5 is not expressed at the injection site or lymph nodes (35) . Hence, bystander activation of APCs could not drive the production of anti-AQP5E autoantibodies. We previously reported no production of Abs against AQP5A, another epitope of anti-AQP5 autoantibodies, by immunization with PmE-L (17), which supports no role for bystander activation of APCs in the anti-AQP5E autoantibody induction in this model.

Aquaporins are transmembrane proteins and do not maintain proper structures in soluble forms. Therefore, instead of using recombinant PmAqp, we designed a peptide that contains both a T cell epitope and a conserved functional B cell epitope for immunization (17) . Recently, Tsymala et al. (36) reported a rat model that produces AQP4-reactive Abs by immunization with AQP4 mimotopes. These researchers screened a phage display peptide library to search for AQP4 mimotopes recognized by pathogenic AQP4-specific Abs from patients with neuromyelitis optica spectrum disorders (36) . Our approach to identify a mimotope in a homologous bacterial protein was proven to be effective.

The multiple bands detected in the immunoblotting of Pm lysates with the PmE-L immunized sera may be attributed to cross-reaction of anti-PmE-L IgG to other proteins in Pm lysates. A BLAST search against Pm proteins using PmE-L as a query revealed several additional proteins that matched some PmE-L sequences: aquaporin family protein (25. In conclusion, we developed a mouse model that produces anti-AQP5 autoantibodies by molecular mimicry. Although this model is not suitable as an SS model due to the lack of focal lymphocytic sialadenitis and anti-SSA autoantibodies, it will be useful to investigate the role of anti-AQP5 autoantibodies in glandular dysfunction in SS. Furthermore, the tools to easily detect autoantibodies and autoantigen-specific B cells allow this model to be used to test new therapeutics targeting autoantibody production.

Sjögren's syndrome

Characteristics of the minor salivary gland infiltrates in Sjögren's syndrome

Sjögren's syndrome--study of autoantigens and autoantibodies

Prognostic value of Sjögren's syndrome autoantibodies

Detection of autoantibodies against aquaporin-5 in the sera of patients with primary Sjögren's syndrome

Function of the membrane water channel aquaporin-5 in the salivary gland

Functional epitopes for anti-aquaporin 5 antibodies in Sjögren syndrome

Associations of anti-aquaporin 5 autoantibodies with serologic and histopathological features of Sjögren's syndrome

Mechanisms of central tolerance for B cells

Autoimmune regulator and self-tolerance -molecular and clinical aspects

Potential role of bacterial infection in autoimmune diseases: a new aspect of molecular mimicry

Autoimmunity and infection in Sjögren's syndrome

The dark side of Sjögren's syndrome: the possible pathogenic role of infections

The role of Epstein-Barr virus infection in primary Sjögren's syndrome

Dysbiotic oral microbiota and infected salivary glands in Sjögren's syndrome

Detection of autoantibodies against aquaporin-1 in the sera of patients with primary Sjögren's syndrome

Induction of anti-aquaporin 5 autoantibodies by molecular mimicry in mice

Effect of antimuscarinic autoantibodies in primary Sjögren's syndrome

Phage display

Machine learning-guided prediction of antigen-reactive in silico clonotypes based on changes in clonal abundance through biopanning

UMI-tools: modeling sequencing errors in Unique Molecular Identifiers to improve quantification accuracy

PEAR: a fast and accurate Illumina Paired-End reAd mergeR

Stereotypic neutralizing V H antibodies against SARS-CoV-2 spike protein receptor binding domain in patients with COVID-19 and healthy individuals

IgBLAST: an immunoglobulin variable domain sequence analysis tool

Immunoglobulins or antibodies: IMGT ® bridging genes, structures and functions

An anti-EGFR × cotinine bispecific antibody complexed with cotinine-conjugated duocarmycin inhibits growth of EGFR-positive cancer cells with KRAS mutations

Neutralization of Zika virus by E protein domain III-Specific human monoclonal antibody

Hypertonic saline therapy in cystic fibrosis: Evidence against the proposed mechanism involving aquaporins

Diversity in the CDR3 region of V H is sufficient for most antibody specificities

Commonality despite exceptional diversity in the baseline human antibody repertoire

The mouse antibody heavy chain repertoire is germline-focused and highly variable between inbred strains

Autoantibody-mediated dysfunction of salivary glands leads to xerostomia in SKG mice

IgG4 related disease of the head and neck

Th17 cells play a critical role in the development of experimental Sjögren's syndrome

Proteomics. Tissue-based map of the human proteome

Induction of aquaporin 4-reactive antibodies in Lewis rats immunized with aquaporin 4 mimotopes

This study was supported by the National Research Foundation of Korea (Daejun, Korea) through the grants 2018R1A5A2024418 and 2020R1A2C2007038 awarded to Youngnim Choi.