This study was performed in compliance with the Helsinki Declaration after obtaining approvals from the Institutional Review Board of Seoul National University Hospital (IRB Number: 0912-011-302), the Institutional Review Board of Seoul National University School of Dentistry (IRB Number: S-D20140022), and the Institutional Review Board of Seoul St. Mary's Hospital (IRB Number: KC13ONMI0646). The clinical data and samples were obtained after obtaining written informed consent. The same serum samples that were used in our previous study involving anti-AQP5 autoantibodies (18) were used in this study. Serum samples were obtained from 112 primary SS patients who were diagnosed according to the 2002 American-European Consensus group (AECG) classification criteria (19) and/or the 2012 American College of Rheumatology (ACR) criteria (20). In addition, resting and stimulated whole salivary flow rates were measured by spiting and masticatory stimuli using wax gum (GC America Inc, St. Alsip, IL, USA), respectively, according to the method previously described (21). All patients were females with ages ranging from 21 to 80 years (mean age 52.5±10.7 years). Control sera were obtained from 53 healthy female controls who did not show any signs or symptoms of SS (mean age 37.1±7.4 years). Subjects with systemic diseases other than hypertension were excluded.

Total RNA from whole blood was extracted using an easy-BLUE™ total RNA extraction kit (iNtRON Biotechnology, Sungnam, Gyeonggi, Korea). Complementary DNA (cDNA) was synthesized from 1 µg of total RNA using an oligo-d (T) primer and M-MLV reverse transcriptase enzyme (Promega, Madison, WI, USA) in a 25 µl reaction mix. Full-length human AQP1 was amplified using forward (5′-GGATCCATGGCCAGCGAGTTCAAGA AGAAGC-3′) and reverse (5′-CTCGAGCTATTTGG GCTTCATCTCCACCC-3′) primers. The PCR product was first cloned into a cloning TA vector (Real Biotech Corporation, Banquiao, Taiwan) and then sub-cloned into a pcDNA3.1 vector using the BamHI and XhoI restriction sites (Invitrogen, Carlsbad, CA, USA). MDCK cells (Korean Cell Line Bank, Seoul, Korea) were transfected with the cloned plasmid using a calcium phosphate precipitation method and maintained in the presence of G418 (22).

A standard cell-based indirect immunofluorescence assay to detect AQP1 was developed as previously described (182324). MDCK cells expressing full-length human AQP1 were fixed with 4% paraformaldehyde in PBS (pH 7.4) and subjected to antigen retrieval by incubation in sodium citrate buffer (pH 6) at 105℃ for 20 min. After permeabilization with 0.3% Triton X-100 and blocking with 5% BSA in PBS, the cells were incubated with a rabbit polyclonal antibody raised against amino acids 215~269 of human AQP1 (Santa Cruz, Paso Robles, CA, USA) along with a control or a SS serum overnight, followed by Alexa Fluor 488-conjugated goat anti-rabbit IgG and Alexa Fluor 555-conjugated goat anti-human IgG (Invitrogen). To detect anti-AQP1 IgA, goat antihuman IgA TXRD (Southern Biotech, Birmingham, AL, USA) was used as a secondary antibody. All images were taken with a confocal microscope LSM 700 (Carl Zeiss, Jena, Germany). Three images of AQP1-expressing cells were randomly selected and sequentially imaged for the presence of anti-AQP1 IgG or IgA. The relative intensities of the anti-AQP1 signals were determined by decreasing the brightness of the red signal until the anti-AQP1 IgG/IgA signals disappeared. The mean brightness obtained from the three images was used to express the anti-AQP1 IgG/IgA level for each sample.

Student's t-test was used to determine differences in Mander's coefficient between the control and SS sera. Correlation between the anti-AQP1 and anti-AQP5 autoantibody intensities was determined using a Pearson correlation analysis. Associations between the salivary flow rates and the presence of anti-AQP1 autoantibodies were examined using one-way analysis of variance (ANOVA). Because some groups did not pass the normality test, the difference was also analyzed using the Mann–Whitney U test. All statistics were performed using SPSS (SPSS Inc., Chicago, USA).

To detect anti-AQP1 autoantibodies in sera, MDCK cells overexpressing full-length human AQP1 were double stained with anti-AQP1 antibodies and sera from either control subjects (n=52) or SS patients (n=112). To compare the prevalence of anti-AQP1 autoantibodies with that of anti-AQP5 autoantibodies, the same dilutions of sera were used for screening in the previous and current studies. In the previous study, the lowest dilution that did not stain AQP5 by pooled control sera was chosen for screening (18). By screening sera diluted to 1:200 for the presence of anti-AQP1 IgG, autoantibodies were detected in 23 patients with SS (20.5%), and the sera with anti-AQP1 IgG showed a high degree of colocalization between the signals of human IgG and those of AQP1-specific rabbit IgG. None of the control sera was positive for anti-AQP1 IgG (Fig. 1A and B). Among the anti-AQP1 IgG-positive SS sera, 19 samples were also positive for anti-AQP5 IgG in the previous study (18), and in the other four samples, the anti-AQP5 IgG was also detected but at lower levels than the cut off used for positivity (Supplementary Table I). For several anti-AQP1 IgG-negative but anti-AQP5 IgG-positive sera (n=7 for each control and SS), the presence of anti-AQP1 IgG was additionally screened using 1:50 dilution, but all samples were still negative for anti-AQP1 IgG.

Interestingly, anti-AQP1 autoantibodies were not detected in any control sera, although 32.1% of them were positive for anti-AQP5 IgG in the previous study (18). Even though the control sera were negative for anti-AQP1 autoantibodies by indirect immunofluorescence assay, not only the SS but also control sera immunoprecipitated the AQP1–GFP from the lysates of the HEK-293 cells transfected with a gene construct encoding the AQP1-GFP (data not shown). Similarly, immunoprecipitation was not effective in differentiating the levels of anti-AQP5 autoantibodies between the control and patient samples (18).

By screening sera diluted to 1:20 for the presence of anti-AQP1 IgA, autoantibodies were detected in eight SS samples (7.1%) but in none of the control sera (Fig. 1C and D). All anti-AQP1 IgA-positive sera were also positive for anti-AQP5 IgA and/or IgG. However, none of the samples with anti-AQP1 IgA were positive for anti-AQP1 IgG (Supplementary Table I).

In contrast to anti-AQP5 autoantibodies, for which the presence of anti-AQP5 IgA had overlapped with the presence of anti-AQP5 IgG in the previous study (18), the presence of anti-AQP1 IgA coincided with the presence of anti-AQP5 IgA, rather than the presence of anti-AQP1 IgG. Similarly, the presence of anti-AQP1 IgG overlapped with the presence of anti-AQP5 IgG. Therefore, anti-AQP1 autoantibodies may be cross-reactive antibodies produced during antibody response to AQP5. A significant positive correlation (p=0.003) between the intensities of the anti-AQP1 IgG/IgA and anti-AQP5 IgG/IgA supports it (Fig. 1E). The amino acid sequence of AQP1 has 44% identity and 62% similarity with AQP5. In contrast to AQP5 in which the full lengths of loops B and E are exposed to the cytosolic and extracellular compartments, respectively, the middle portion of loops B and E in AQP1 is embedded within the membrane (Fig. 2). However, short stretches of conserved sequences were observed in loop B, loop D, and loop E, as well as the C-terminal cytoplasmic tail, which may provide cross-reactive epitopes.

Alternatively, both autoantibodies against AQP5 and AQP1 might have been developed during an immune response against bacterial proteins that are homologous to human AQPs. Previously, we reported a high degree of homology between AQP5 and AQPZ or porins from many human-associated bacteria, including the members of normal human oral flora (18). The same oral bacterial AQPs presented 35~42% identity and 52~56% similarity with AQP1 (Fig. 3). Whereas the sequences of the bacterial AQPs are conserved mostly at the transmembrane alpha helixes and the pore-forming loops B and E of AQP5 (18), conserved sequences were observed mostly at the transmembrane alpha helixes and the loops B and C of AQP1 (Fig. 3). Epitope mapping for anti-AQP5 and anti-AQP1 autoantibodies in a future study would help clarify the relationship of these autoantibodies to each other.

Associations between the presence of anti-AQP1 IgG or IgA and salivary flow rates were analyzed in patients who had their saliva flow rate recorded at the time of serum sampling. The presence of neither anti-AQP1 IgG nor IgA had a significant effect on salivary flow. Because more than one-half of the SS patients who were negative for anti-AQP1 autoantibodies had anti-AQP5 autoantibodies that were associated with a low resting salivary flow (18), the effect of anti-AQP1 autoantibodies on salivary flow was also examined in patients with anti-AQP5 autoantibodies. However, no difference in either resting or stimulatory salivary flow was observed by the additional presence of anti-AQP1 autoantibodies (Table I).

Two-thirds of the anti-AQP1 autoantibodies detected in patients with suspected neuromyelitis optica spectrum disorders were bound to the extracellular domain of AQP1 (25). If the anti-AQP1 autoantibodies detected in SS patients bind to the AQP1 expressed on myoepithelial cells and endothelial cells in vivo, antibodies may block water flow across these cells, interfering with the inflow of water to the acini. However, the presence of anti-AQP1 autoantibodies did not affect either resting or stimulated salivary flow. This supports the notion that AQP1 does not play a major role in normal physiology of salivary secretion, as observed in AQP1-deficient mice (8).

In conclusion, anti-AQP1 autoantibodies were detected in 27.7% of SS patients. Although anti-AQP1 autoantibodies are not useful as a diagnostic marker because of their low sensitivity, the presence of autoantibodies to AQP1 might become a major obstacle to AQP1 gene therapy for SS.