Serum under-O-glycosylated IgA1 level is not correlated with glomerular IgA deposition based upon heterogeneity in the composition of immune complexes in IgA nephropathy

The mouse T cell leukemic cell-line BW5147 transfectant stably expressing mFcα/μR has been described previously [12]. In our preliminary study, it was confirmed that the affinity of mFcα/μR for human pIgA1 was much stronger than that of human Fcα/μR. The alternatively spliced variants of hFcα/μR have been reported [13, 14], UniProtKB; Q8WWV6, FCAMR_HUMAN] and the cDNA which we used for constructing hFcα/μR transfectants had shorter N-terminal leader sequences than that of mFcα/μR [12]. Although the amino acid sequences of immunoglobulin binding site of h-and mFcα/μR have been highly conserved, those around this domain are different [12].

For a negative control, mock transfectant cells were exposed to Plat E packaging cell-line producing the pMX-neo vector [19]. The mFcα/μR expressing transfectant (Fcα/μR transfectant) and parent BW5147 cells (2 × 10⁵) were treated with 5 μl of sera from IgAN patients or from healthy controls at 4°C for 30 min. The cells were washed three times with an FACS staining buffer [0.5% phosphate buffered saline, 0.05% (w/v) bovine serum albumin (BSA), and 0.05% (w/v) sodium azide]. The sodium azide was added to an FACS staining buffer to immobilize the cytoskeleton of the cells in good viability since mFcα/μR internalizes into cytosol when it is cross-linked by ligands [12]. The transfectant and parent cells were stained with biotinylated Helix aspersa (HAA) lectin (Sigma-Aldrich, St. Louis, MO, USA) at 4°C for 15 min. After washing three times, the transfectant and parent cells reacted with 1:500 diluted phycoerythrin (PE)-conjugated streptavidin (Beckman Coulter, CA, USA) at 4°C for 30 min. Dead cells were excluded by propidium iodide (PI) staining (Figure 1a). The mean fluorescence intensity (MFI) of each cell was measured by flow cytometry (FACS Calibur, Becton Dickinson, NJ, USA). Before starting assay, mFcα/μR transfectant and parent cells were preliminary treated with serum in the same protocol and stained by a PE-conjugated mouse monoclonal anti-human IgA antibody (Miltenyl Biotec, Clone IS11-8E10 (isotype: mouse IgG1, intact molecule), Bergisch Gladbach, Germany) to confirm that the transfectant cells were saturated by IgA1 and that the parent cells were not non-specifically stained. The amount of pIgA1 analyzed in each procedure was estimated around 10 μg using purified pIgA1 (Data not shown). Biotinylated mouse monoclonal anti-human IgM antibody (Miltenyi Biotec, Clone PJ2-22H3 (isotype: mouse IgG1, intact molecule)) was also used when IgM bindings to the cells were checked.

Secretory IgA purified from pooled human colostrum using multistep procedures which may include salt fractionation, gel filtration, ion-exchange chromatography, and immunoadsorption (MP Biomdicals, Santa Ana, CA, USA) were adopted as a positive control of the pIgA1 trap. Human monomeric IgA1 and pIgA1 from multiple myeloma patients and degalactosylated pIgA1 were kindly provided by Professor Jan Novak (University of Alabama at Birmingham, AL, USA) and used as controls.

HAA lectin was used to determine serum IgA1 with aberrantly O-glycosylation, as reported previously [11, 20]. For capture ELISA, flat-bottom, 96-well plates were coated at 4°C overnight with F(ab’)2 fragment of goat anti-human IgA antibody (Jackson ImmunoResarch Laboratories, West Grove, PA, USA) at a concentration of 2.5 μg/ml. The plates were blocked at 4°C overnight with 1% BSA in PBS containing 0.05% Tween 20 (v/v). Samples diluted in blocking buffer were added to each well and incubated at 4°C overnight. The captured IgA1 was subsequently desialylated by treatment at 37°C for 3 hours with 10 mU/ml neuraminidase from Vibrio cholera (Roche Applied Science, Indianapolice, IN, USA) in 10 mM sodium acetate buffer (pH5). Samples were then incubated at 37°C for 3 hours with biotinylated HAA lectin (Sigma-Aldrich) diluted in the blocking buffer. The bound lectin was detected with an avidin-horseradish peroxidase conjugate. The peroxidase chromogenic substrate o-phenylenediamine-H₂O₂ (Sigma-Aldrich) was then added. The color reaction was stopped with 1 M sulfuric acid and the absorbance at 490 nm was measured using Spectra Max (Molecular Devices, Sunnyvale, CA, USA). The HAA reactivity of IgA1 in each sample was then calculated as OD units/1 μg of IgA1. Naturally degalactosylated IgA1 purified from the plasma of a patient with IgA1 multiple myeloma was treated with neuraminidase and used as a standard [7, 21]. For comparisons of HAA bindings, the OD units per 1 μg standard degalactosylated IgA1 were assigned a value of 100% and it was used as the HAA ELISA titer.

Staining for IgG, IgA, and C3 in glomeruli on freshly frozen renal sections (3 μm) was performed using corresponding fluorescein isothiocyanate (FITC)-conjugated anti-sera (Dako, Copenhagen, Denmark). The samples were analyzed under a confocal laser-scanning microscope (Fluoview FV1000, Olympus, Tokyo, Japan) and the electronic images were stocked in the JPEG format.

The image files were analyzed by Image Processing and Analysis in Java software (Image J, NIH, Bethesda, MD, USA). The brightness of the immunofluorescence images were adjusted to a range of 25–255 pixel values and the edges of glomeruli were traced to delimit glomerular areas (Figure 1a). The color (RGB) immunofluorescence images were converted to binary data and pixels in black and within whole traced (glomerular) areas were counted using the Analyze Particle command. IgA deposition areas (Area-IgA,%) were calculated by the following formula (Figure 1b):

IgG and C3 deposition areas (Area-IgG, Area-C3, respectively) (%) were also calculated using the same procedure.

The colocalization of HAA lectin and IgA1 on the mFcα/μR transfectant was analyzed by confocal laser-scanning microscope. The cells were treated with pIgA1 for 15 min, washed 3 times, and incubated with an FITC-conjugated rabbit anti-human IgA (DAKO) antiserum at 1:100 dilution. After washing, the cells were incubated again with biotinylated HAA lectin at room temperature for 2 hours. Floating cells were adhered to slides using cytospin (Thermo Scientific, Waltham, MA, USA) and post-staining with DAPI was also performed. Double staining studies with human IgA and IgM were also performed to the mFcα/μR transfectant previously treated with human sera using an FITC-conjugated anti-human IgA and biotinylated anti-human IgM antisera which was detected by DyLight 549 conjugated streptavidin (Jackson ImmunoResarch Laboratories). The cells were also fixed to slides by cytospin and DAPI staining was performed.

The renal histological grade was determined using the criteria of the Joint Committee of Research Groups on Progressive Renal Diseases (Ministry of Health, Labor and Welfare of Japan) and Japanese Society of Nephrology [22].

All values were examined by mean ± SD. Differences between the two groups were evaluated using a Mann-Whitney’s U test. Comparisons among three or more parameters were analyzed by analysis of variance. P < 0.05 was defined as statistical significant. Statistical analyses, including contour plots, in which data from pIgA1 trap form the X-axis, data from ELISA form the Y axis, and other clinical parameters form a pseudo-Z axis via colors (age, interval between onset and biopsy, Area-IgA, Area-IgG, Area-C3, serum IgA, serum C3, the IgA/C3 ratio, s-Cr, eGFR, and urinary protein excretion), were performed using JMP7 (SAS Institute, Cary, NC, USA).

Serum pIgA was trapped using mouse Fcα/μR transfectant. The O-glycans of the captured pIgA1 were stained with fluorescein-labeled HAA lectin and the fluorescein intensity of the tranfectant was measured by flow cytometry. Dead cells were distinguished from the flow cytometric study by the measurement of a combination of forward scatter (FSC), side scatter (SSC), and propidium iodide (PI) staining (Figure 2a). Purified serum IgA (monomeric IgA) and monomeric IgA1 from multiple myeloma patients didn’t bind to the BW5147 parent cell, the mock transfectant, or the mFcα/μR transfectant. Both pIgA1 from milk and multiple myeloma patients tightly bound to the mFcα/μR transfectant but showed no reactivity to BW5147 or the mock transfectant (Figure 2b). The mFcα/μR transfectant pre-treated with IgM revealed similar binding activity of pIgA1 to non-treated transfectant and showed the same binding activity with or without pre-treatment with IgM (Figure 2c). While pIgA1 bound mFcα/μR transfectant can fix HAA, the IgM bound mFcα/μR transfectant could not react with HAA. The merged figures revealed the co-localization of pIgA1 and HAA, suggesting that HAA bound to under-glycosylated O-glycan of pIgA1 (Figure 2d). Serum pIgA1 was captured by mFcα/μR transfectant and was followed by staining with fluorescein labeled HAA. The fluorescence intensity of the HAA-bound transfectant could be measured and it varied in each patient or healthy control. The positive control, which was performed with degalactosylated pIgA1 from multiple myeloma patients, showed a higher intensity of fluorescence of HAA (Figure 2e).

HAA ELISA, which was classically used for the measurement of under-glycosylated O-glycan of serum IgA1, was performed for IgAN patients and healthy controls. Similar to previous reports [7, 10, 11, 21], the mean ELISA titer of IgAN patients (19.0 ± 5.7%, mean ± SD) was significantly higher than that of the healthy controls (15.0 ± 2.7%, P < 0.05) (Figure 3a). pIgA1 trap was also performed for the same samples and it revealed that there was no significant difference in the mean value of pIgA1 trap between IgAN patients and healthy controls (Figure 3b). There was no correlation between the values of HAA ELISA and pIgA1 trap, even in healthy controls (r² = 0.09, p = 0.11) (Figure 3c) or in IgAN patients (r² = 0.005, P = 0.14) (Figure 3d). In IgAN patients, a direct correlation was not observed between Area-IgA and HAA ELISA values (r² = 0.005, p = 0.69) (Figure 3e). Similarly, Area-IgA was not directly correlated with the values of pIgA1 (MFI) (r² = 0.03, p = 0.33) (Figure 3f).

Contour plots showed that the patients with high values in clinical parameters concentrated in specific areas in each plot (Figure 4a–k). The patients with high Area-IgA (over 15%) formed a group in the area prescribed by HAA ELISA titers between 15 and 20 and HAA-pIgA1 trap between 50 and 60 (Figure 4c). The patients in this area showed a long interval between onset and biopsy and a higher score of Area-IgG, Area-C3, s-Cr and urinary protein excretion (Figures 4b,d,e,i,and k, respectively). Inversely, these patients showed low values of serum IgA, serum C3, and eGFR (Figures 4f,g, and j, respectively). Area-IgA and serum IgA as well as Area-C3 and serum C3 showed a contrasted pattern (Figures 4c,f,e,g). In the zone of around 20–25 in HAA ELISA titers, younger age patients and high eGFR patients were concentrated (Figures 4a and j).

Since Area-IgA was not linearly correlated with the values from HAA ELISA and contour plots suggested that there was a complex rule between Area-IgA and under-glycosylated O-glycan of serum IgA1, a decision tree analysis was performed to predict Area-IgA in glomeruli and to classify IgAN patients using a combination of HAA ELISA and pIgA1 trap. The decision tree consisted of 5 terminal nodes and patients who belonged to each node were classified into 5 groups (Group A, B, C1, C2, and D) (Figure 5a). Group A was classified by a HAA-ELISA titer value of <15.1% with Group B characterized by an HAA-ELISA titer value of ≥15.1% and <18.7%. Group C was classified by an HAA-ELISA titer value of ≥18.7% and <24.2% and this group was the only group sub-classed into C1 and C2 by pIgA1 trap value (MFI) with a cut off value of 58.5. Group D was classified by a HAA-ELISA titer value of ≥24.2% (Figure 5a).

While the mean value of each HAA-ELISA titer group was significantly different, except for that of C1 and C2 (Figure 5b), the pIgA1 trap value in Group C2 was significantly higher than the other groups and there was no difference between any pairs of Groups A, B, C1, and D (Figure 5c). The obtained model showed a 90.6% accuracy evaluated using a 10-fold cross validation. The overlaid borders settled by decision tree on the contour plot of Area-IgA indicated that patients with a high score of Area-IgA belonged to Group B (Figure 5d).

A model of Area-IgA was built by decision tree analysis. While the HAA-ELISA titer was increasing from Group A to D, the mean value of Area-IgA showed a different distribution. Although the mean value of Area-IgA in Group B was significantly higher than that in Group A, lower values were found groups C1 and C2. The mean value of Area-IgA increased again in Group D, over those in groups C1 and C2, and there was no difference in the value of Area-IgA between groups B and D (Figure 6a). The mean value of Area-IgG in groups A and B was higher than that of groups C1, C2, and D but there was no significant difference (Figure 6b). The mean value of Area-C3 in glomerulus in groups A and B was more than that in C1, C2 and C3, and a significant difference was observed between groups B and D (Figure 6c).

Since it was suggested that the IgAN patients classified by decision tree analysis using HAA ELISA titer and pIgA1 trap values were also clinically different, clinical parameters were compared. The mean patients’ age in Group C2 seemed younger than other groups but there was no significant difference (Figure 7a). Urinary protein excretion (g/g Cr) and s-Cr (mg/dl) showed a similar distribution in that groups A and B were in a higher range, groups C1 and C2 in a lower range, and Group D in a middle range, although these values had no significant difference (Figure 7b, 7c). Mean eGFR in Group C2 was significantly higher than that in groups A and B. Mean eGFR values in groups C1 and D were in the middle range and showed no significant difference with Group C2 (Figure 7d).

No significant difference in the mean serum IgA level in each group was observed (Figure 7e). The mean serum C3 level in Group D was significantly more elevated than in all the other groups (Figure 7f). The patients in Group A revealed a more severe hematuria than those in groups B and C2 (Figure 7g). The histological grade of renal specimens in groups B and D were significantly more severe compared with that of Group C2 (Figure 7h).

The urinary protein excretion level was not significantly different three years after renal biopsy (Figure 7i). The mean eGFR of group C2 was significantly higher than that of group A, B and D (P < 0.01, P < 0.05, P < 0.05, respectively) (Figure 7j).