Aberrant mucosal immunoreaction to tonsillar microbiota in immunoglobulin A nephropathy.

BACKGROUND: Immunoglobulin A nephropathy (IgAN) is the most common glomerulonephritis worldwide, characterized by mesangial polymeric IgA1 deposition. IgAN is believed to develop owing to aberrant mucosal immunoreaction against commensals in the tonsils. However, the exact interrelation between pathogenic IgA and mucosal microbiota in IgAN patients is unclear.
METHODS: Biopsy-proven IgAN or recurrent tonsillitis (RT) patients who had undergone tonsillectomy were enrolled. We used 16S ribosomal RNA gene amplicon sequencing with a flow cytometry-based bacterial cell sorting technique) and immunoglobulin repertoire sequencing of the IgA heavy chain to characterize IgA-coated bacteria of the tonsillar microbiota (IgA-SEQ) and their corresponding IgA repertoire. Furthermore, we fractionated patient serum using gel-filtration chromatography and performed flow cytometry-based analysis of IgA binding to bacteria cultured from incised tonsils.
RESULTS: Tonsillar proliferation-inducing ligand and B-cell activating factor levels were significantly higher in IgAN than in RT patients. IgA-SEQ for tonsillar microbiota revealed the preferential binding ability of IgA to Bacteroidetes in IgAN tonsils compared with those from RT patients. Expression of immunoglobulin heavy (IGH) constant alpha 1 with IGH variable 3-30 was significantly higher in IgAN than that in RT, and positively correlated with the IgA-coated enrichment score of Bacteroidetes. Serum polymeric IgA, comprising high levels of GdIgA1, exhibited considerable binding to Bacteroidetes strains cultured from the tonsils of IgAN patients.
CONCLUSIONS: These findings provide evidence that aberrant mucosal immune responses to tonsillar anaerobic microbiota, primarily consisting of members of the phylum Bacteroidetes, are involved in IgAN pathophysiology.

KEY LEARNING POINTS

What is already known about this subject?

The interplay between mucosal immunity and microbiota in patients with IgA nephropathy is considered to underlie disease pathogenesis.

Tonsillectomy has beneficial effects on clinical outcomes of patients with IgA nephropathy.

We conducted this study to elucidate the mechanisms underlying the mucosal immune response to microbiota at the tonsillar crypts and its role in the development of IgA nephropathy

What this study adds?

Both a proliferation-inducing ligand and B-cell activating factor expression in tonsillar crypts were elevated in patients with IgA nephropathy, correlating with the expression of immunoglobulin heavy variable 3–30 (IGHV3-30) and IGHV3-38 repertoire, respectively, in IgA1 at the tonsillar crypts.

Bacteria from the phylum Bacteroidetes were highly coated with IgA at the tonsillar crypts of IgA nephropathy patients, correlating with the tonsillar expression of IGHV3-30 in IgA1.

Serum polymeric IgA, comprising high levels of galactose-deficient IgA1, exhibited considerable binding to Bacteroidetes strains cultured from the tonsils of IgA nephropathy patients.

What impact this may have on practice or policy?

This study strengthens the role of the mucosal immune response to microbiota in the pathogenesis of IgA nephropathy.

Aberrant mucosal immunoreactions against the host tonsillar microbiota are potentially novel therapeutic targets for the management of IgA nephropathy.

INTRODUCTION

Immunoglobulin A nephropathy (IgAN) is the most common chronic glomerular disease worldwide and is a major cause of end-stage kidney disease. It is predominantly characterized by pathological IgA1 deposits in the mesangium [1]. Mucosal infections such as tonsillitis and upper respiratory tract infections exacerbate hematuria in IgAN patients. Tonsillectomy, combined with steroid pulse therapy, is therefore widely performed in Japan [2] and has beneficial effects for IgAN clinical remission [3]. Additionally, genome-wide association studies for IgAN have indicated several loci to be involved in regulating mucosal IgA production and innate immunity [4], highlighting the importance of identifying the link between mucosal immunoreaction and IgAN pathogenesis [5].

Mesangial IgA deposits in IgAN patients are predominantly polymeric, aberrantly glycosylated [6, 7], and associated with the J-chain [8], which is preferentially produced by mucosal immunocytes [9]. The bacterial DNA-activated [10] expression of Toll-like receptor (TLR) 9 in the tonsils has also been associated with IgAN progression [11]. Repeated TLR9 activation induces tonsillar expression of a proliferation-inducing ligand (APRIL), which is predominantly produced by dendritic cells [12] and promotes class switching through T-cell-independent immune interactions with commensal microbiota [13-15]. Notably, APRIL expression in patients with an aberrant tonsillar germinal center was responsive to tonsillectomy, exhibiting decreased galactose-deficient IgA1 (GdIgA1) in serum [16]. Although tonsillar mucosal immune dysregulation in IgAN has been thoroughly investigated, the mechanisms underlying pathogenic tonsillar IgA1 production remain unclear.

Previously, we performed high-throughput 16S ribosomal RNA (rRNA) gene sequencing to explore the association of IgAN with bacterial flora in the palatine tonsils [17], but no significant difference was identified in the relative abundance of any bacterial genus between IgAN or recurrent tonsillitis (RT) cases. Recent studies demonstrated that microbiota-induced mucosal IgAs coated diverse microbial antigens [18] and agglutinated their cross-reactive bacterial targets, preventing direct host interaction [19, 20]. Further, bacterial flow cytometry with high-throughput 16S rRNA gene sequencing (IgA-SEQ) characterized IgA-coated microbiota [18, 21]. IgA-SEQ application in fecal samples from mouse and human disease models revealed the roles of IgA-coated intestinal microbiota in inflammatory bowel disease development [21, 22], indicating that IgA-SEQ could be used to identify disease-related microbes. Furthermore, analyzing peripheral tissue or blood samples using recently developed high-throughput sequencing strategies enabled comprehensive exploration of human immunoglobulin repertoires [23, 24]. Diversification of immunoglobulins, which comprise paired immunoglobulin heavy (IGH) and light chains and constitute the soluble forms of B-cell receptors, is determined via immunoglobulin variable (V), diversity (D) and joining (J) gene segment recombination [25] and subsequent somatic hypermutation [26]. This reveals that microbial symbionts influence host immunity and regulate preimmune B-cell repertoires independently of T cells [27].

This study aimed to investigate the mucosal immune responses to microbiota in the tonsils of IgAN patients by this multiomics approach using tonsillectomy samples. We also evaluated whether such responses are involved in disease pathology.

MATERIALS AND METHODS

Study population and sample collection

The study protocol adhered to the Declaration of Helsinki and was approved by the Ethics Committee on Genetic Analysis of Niigata University, Niigata, Japan (approval no. G2017-0004). Written informed consent was obtained from all patients. We collected tonsillar samples from IgAN and RT patients who had undergone tonsillectomy from July 2012 to June 2018. IgAN pathological diagnoses and tonsillar sample collection procedures were previously described [17].

Tonsillar protein extraction and enzyme-linked immunosorbent assay

Tonsillar tissue (30 mg) was suspended in T-PER tissue protein extraction reagent (Thermo Scientific, Waltham, MA, USA) and homogenized using the FastPrep®24 mechanical disruptor (MP Biomedicals, Tokyo, Japan) containing ceramic beads, setting the final concentration at 1.0 mg/dL, per bicinchoninic acid assay methods. Serum and tonsil APRIL, B-cell activating factor (BAFF), GdIgA1 and IgA protein concentrations were measured using enzyme-linked immunosorbent assay (ELISA). IgA–IgG complex levels in fractionated serum were also determined by ELISA, as previously reported [28].

Immunostaining

Paraffin-embedded kidney biopsy and incised tonsil sections from enrolled patients were used for immunohistochemistry. After antigen retrieval and blocking, slides were incubated with mouse anti-APRIL/TNFSF13 monoclonal antibody (Aprily-8; Novus Biologicals, Littleton, CO, USA), rat anti-human BAFF monoclonal antibody (Buffy-2; Abcam, Cambridge, UK) antibodies or rat anti-GdIgA1 (KM55; Immuno-Biological Laboratories, Gunma, Japan) [29]. Slides were subsequently incubated with secondary antibodies, as described in the Supplementary data. The proportion of the area stained by immunofluorescence examination with anti-GdIgA1 antibody was quantified to evaluate glomerular GdIgA1 deposition, as per previous studies [8, 30, 31].

Bacterial sorting for IgA-SEQ and 16S rRNA gene amplicon sequencing

Bacterial sorting was performed as previously reported [21] and applied to tonsillar samples with some modifications. Briefly, washed bacterial pellets of tonsillar samples were stained with phycoerythrin (PE)-conjugated mouse anti-human IgA (IS11-8E10, Miltenyi Biotec, Bergisch-Gladbach, Germany) or mouse IgG1-PE isotype (Presort sample). The IgA-PE positive fraction enriched and separated by magnetic-activated cell sorting was sorted by fluorescence-activated cell sorting (FACS) to purify the IgA-positive (IgA+) and IgA-negative (IgA−) fractions. Details of the 16S rRNA gene amplification sequencing of genomic DNA from tonsillar crypts and sorted fractions are provided in the Supplementary Methods. The IgA index was calculated using the relative abundance of bacterial phyla and genera for IgA+ and IgA− fractions as per [log(IgA+) – log(IgA−)]/[log(IgA+) + log(IgA−)], as reported previously [22, 32]. The 16S rRNA gene amplicon sequence data were deposited in DDBJ DRA with BioProject ID PRJDB8009.

Adaptor-ligation polymerase chain reaction and immunoglobulin sequencing analysis

Total RNA of each dissected tonsillar sample (30 mg) was extracted. Adaptor-ligation polymerase chain reaction (PCR) and immunoglobulin sequencing were performed as previously reported [23]. Briefly, to detect IGHV, IGHD, IGHJ and IGH chain constant (IGHC) gene segment distributions and the deduced complementarity-determining region 3 (CDR3) amino acid sequence of tonsillar IgA, sequence reads were analyzed using bioinformatics software (Repertoire Genesis Incorporation, Ibaraki, Japan) [23, 24]. The reference sequence datasets are available from the International ImMunoGeneTics Information System database (http://www.imgt.org).

Size fractionation

A serum aliquot (200 μL) was fractionated by gel-filtration through a Superdex 200 10/300 column connected to a liquid chromatography system (AKTA Pure 25 L), controlled with Unicorn 7.1 (GE Healthcare, Buckinghamshire, UK). Fractions (500 μL) were collected and analyzed.

Western blotting

Fractionated samples containing 20 ng IgA were subjected to nonreducing 3–8% Tris-acetate gel electrophoresis and electroblotted onto polyvinylidene fluoride membranes. The membranes were blocked, incubated with horseradish peroxidase-conjugated antibodies (Supplementary Methods) and developed using chemiluminescent substrate.

Bacterial culture

Dissected tonsillar tissues were suspended, seeded onto blood-agar plates and anaerobically incubated for 7 days. Following strain detection, Porphyromonas gingivalis, Prevotella intermedia and Fusobacterium nucleatum were cultured in Gifu anaerobic media broth under anaerobic conditions. Escherichia coli (ATCC 25922) was cultured in Luria–Bertani broth. Isolated bacterial species were assessed by nearly full-length 16S rRNA gene sequencing with 27F/1492R primers [33]. The sequences were compared with the 16S rRNA sequences of each bacterial strain from the DDBJ/EMBL/GenBank database by phylogenetic analysis.

Flow cytometry

IgA binding to bacteria was evaluated using flow cytometry as previously described, with minor modifications [34, 35]. Briefly, after blocking with normal mouse serum, 2.0 × 106 bacteria were incubated with 50 µL of each serum fraction. Sample IgA concentrations were adjusted to that of the F04 fraction prior to incubation. Incubated bacteria were stained with PE-conjugated mouse anti-human IgA monoclonal antibody or mouse IgG1-PE isotype and analyzed using a FACSAria II (Becton Dickinson, Franklin Lakes, NJ, USA).

Statistical analysis

Statistical analyses were performed as described in each figure legend. All statistical data were analyzed using R v. 3.4.2 (https://www.r-project.org). For relative abundance comparisons, the IgA index analysis for each bacterium, and immunoglobulin repertoire sequencing analysis of the IGHV, J and D gene segments and the CDR3 amino acid length, P < 0.01 was considered statistically significant. Elsewhere, P < 0.05 was accepted. Detailed methods are provided in the Supplementary Methods.

RESULTS

Tonsil and serum APRIL, BAFF and GdIgA1 levels

We enrolled IgAN patients diagnosed by renal biopsy (n = 62) and RT patients without urinary abnormalities (n = 28) (Table 1). Among these, 48 IgAN and 21 RT cases were included in our previous study [17], and serum samples of 30 IgAN and 7 RT patients prior to tonsillectomy were also available (Supplementary data, Table S1). Tonsillar APRIL and BAFF levels were significantly higher in IgAN than in RT patients (P = 0.0071 and P = 0.025, respectively), whereas both serum levels were not significantly elevated (Figure 1A and B). Immunohistochemistry in the tonsils using an anti-APRIL monoclonal antibody disclosed diffuse staining in the germinal centers, basal cells of the surface epithelium, and perifollicular area around the crypts in both IgAN and RT. A high degree of staining in the germinal centers and perifollicular area was observed in IgAN tonsils with high tonsillar APRIL, whereas epithelial staining was similar across IgAN and RT tonsils (Figure 1C). Depending on the BAFF levels, anti-BAFF antibody mainly labeled the mantle and perifollicular zone of the tonsils in both groups (Figure 1D). Serum but not tonsillar GdIgA1 levels were significantly higher in IgAN than in RT, whereas there was no significant elevation in IgA levels of both serum and tonsils (Figure 1E;Supplementary data, Figure S1A). Tonsillar GdIgA1 levels were significantly correlated with the area of glomerular GdIgA1 deposition and were pathologically associated with mesangial hypercellularity in accordance with the Oxford classification of IgAN (Figure 1F;Supplementary data, Figure S1B). In 16S rRNA gene analysis of IgAN tonsils, relative abundances of phyla including Bacteroidetes in patients with higher GdIgA1 levels at the tonsils were significantly different as compared with those with lower GdIgA1 levels when analyzed by permutational multivariate analysis of variance (Supplementary data, Figure S1C and D).

Table 1

Characteristics of enrolled patients

Patient characteristics	IgAN (n = 62)	RT (n = 28)	P-value
Age (years)	32.8 ± 10.3	30.5 ± 9.5	0.361
Sex (male)	26 (41.9)	18 (64.3)	0.069
BMI (kg/m2)	21.9 ± 2.8	22.6 ± 4.4	0.353
Systolic BP (mmHg)	116.6 ± 13.6	116.2 ± 13.8	0.908
Diastolic BP (mmHg)	71.0 ± 10.1	69.0 ± 9.4	0.407
s-Cre (mg/dL)	0.85 ± 0.27	0.73 ± 0.15	0.052
eGFR (mL/min/1.73 m2)	80.5 ± 20.5	97.2 ± 19.4	0.001
BUN (mg/mL)	13.7 ± 4.4	12.5 ± 3.2	0.225
TP (g/dL)	7.1 ± 0.5	7.4 ± 0.5	0.054
UA (mg/dL)	5.2 ± 1.4	5.5 ± 1.5	0.678
IgA (mg/dL)	310.8 ± 155.5	–	–
C3 (mg/dL)	95.5 ± 17.5	–	–
Urinary protein (g/day)	0.45 (0.29, 0.90)	–	–
Period from renal biopsy to tonsillectomy (day)	174 (115, 367)	–	–
RASI	39 (62.9)	1 (3.6)	<0.001
Oxford classification
Mesangial hypercellularity (M0/M1)	49/13
Endocapillary hypercellularity (E0/E1)	28/34
Segmental glomerulosclerosis (S0/S1)	14/48
Tubular atrophy/interstitial fibrosis (T0/T1/T2)	57/5/0
Crescents (C0/C1/C2)	23/39/0

Data are presented as the means ± SD, median (IQR) or number (ratio), and were statistically compared using Student’s t-test or Fisher’s exact test. Among the clinical characteristics of these patients, a significant difference was noted in the eGFR and RASI.

BMI, body mass index; BP, blood pressure; s-Cre, serum creatinine; eGFR, estimated glomerular filtration rate; BUN, blood urea nitrogen; IQR, interquartile range; TP, total protein; UA, uric acid; C3, complement component 3; RASI, renin–angiotensin–aldosterone system inhibitor.

FIGURE 1

Quantification of APRIL, BAFF and GdIgA1 in the tonsillar crypts. APRIL (A) and BAFF (B) levels of adjusted tonsillar protein solution from IgAN (n = 55) and RT patients (n = 26) (left) and serum APRIL and BAFF levels from IgAN (n = 30) and RT (n = 7) (right) patients, respectively. Box plots show the median and interquartile range (IQR), with whiskers of 1.5 × IQR. Data were statistically compared using the Mann–Whitney U test. Immunohistochemistry with an anti-APRIL (C) and anti-BAFF (D) monoclonal antibody in tonsils from patients with RT (left panels, tonsillar levels of APRIL = 145.3 pg/mL, BAFF = 355.4 pg/mL, respectively), IgAN with low tonsillar APRIL levels (middle panels, levels of APRIL = 128.9 pg/mL, BAFF = 555.1 pg/mL, respectively) and high tonsillar APRIL or BAFF levels (right panels, levels of APRIL = 267.5 pg/mL, BAFF = 1028.2 pg/mL, respectively). Black arrows indicate APRIL staining in basal cells of the surface epithelium. Black arrowheads indicate APRIL staining of perifollicular area. Blue arrows indicate BAFF staining in mantle and perifollicular zone around crypts. Photographs show germinal centers and epithelium around crypts (upper panels; low magnification images, bottom panels; high magnification images). Scale bars: 100 µm. (E) GdIgA1 levels of adjusted tonsillar protein solution from IgAN (n = 55) and RT patients (n = 26) (left) and serum levels of GdIgA1 from IgAN patients (n = 30) and RT (n = 7) (right), respectively. Box plots show the median and IQR, with whiskers of 1.5 × IQR. Data were statistically compared using the Mann–Whitney U test. (F) Correlation between levels of tonsillar GdIgA1 and glomerular GdIgA1 deposition (left, n = 26). Oxford classification is shown for each category (right). Correlation data were statistically compared using Spearman’s correlation and linear regression analysis. Box plots show median, IQR with whiskers of 1.5 × IQR, and statistically compared using the Mann–Whitney U test (NS; no significance, *P < 0.05).

IgA-SEQ of the tonsillar microbiome

We utilized 18 available IgAN tonsillar samples and 14 RT samples for tonsillar microbiota IgA-SEQ analysis. The 16S rRNA gene amplicon sequencing was performed to calculate the relative bacterial abundance in IgA+ and IgA− fractions (Supplementary data, Figures S2 and S3A). Taxonomic analysis at the phylum level confirmed that the relative abundance of each phylum in the presort sample was mostly located between the IgA+ and IgA− fractions (Figure 2A). At the phylum level, Bacteroidetes were significantly more abundant in the IgA+ fraction than in the IgA− fraction (Figure 2B;Supplementary data, Figure S3B) of both groups. At the genus level, relative abundances of Porphyromonas and Prevotella (phylum Bacteroidetes) were significantly higher in the IgA+ fraction of IgAN. Principal coordinate analysis of the relative bacterial abundances in IgAN showed that each cluster was significantly segregated (Figure 2C). We calculated the IgA index for each bacterium at the phylum and genus levels to identify the tonsillar microbiota selectively coated with IgA in IgAN patients. The IgA index of Bacteroidetes was more abundant among bacteria in both IgAN and RT tonsils (Figure 3A). Comparison of the IgA index of each bacterium revealed that the indices of Bacteroidetes in IgAN patients were relatively higher (P = 0.041) than those in RT patients (Figure 3B). Moreover, the IgA indices of Porphyromonas spp. were significantly higher (P = 0.0015) in IgAN than in RT (Figure 3C;Supplementary data, Figure S4). These IgA indexes were not associated with clinical (estimated glomerular filtration rate, proteinuria) or pathological findings in IgAN patients (data not shown).

FIGURE 2

IgA-SEQ of tonsillar microbiota. (A) The mean relative abundance at the phylum level is shown for the IgA+ and IgA− fractions and presort samples in the IgAN (n = 18) and RT (n = 14) groups. (B) Relative abundances of the top five phyla (top row) and genera of Bacteroidetes (bottom row) are shown for IgA+ (red) and IgA− (blue) fractions. Data are presented as the median and IQR, and were statistically compared using the Mann–Whitney U test (*P < 0.05, **P < 0.01, ***P < 0.001). (C) IgA-SEQ samples of IgAN were clustered using principal coordinate analysis with Bray–Curtis dissimilarities. The length of the black arrows represents taxon abundance. P-values and R2 values calculated by permutational ANOVA (PERMANOVA) using 9999 permutations based on the Bray–Curtis dissimilarity index are indicated for the IgA+ fraction versus the presort sample, IgA+ versus IgA− fraction and presort sample versus IgA− fraction.

FIGURE 3

IgA index of tonsillar bacteria at the phylum level in IgAN and RT patients. (A) Heatmap depicting the IgA index of tonsillar bacterial phyla in IgAN (n = 18) and RT patients (n = 14). The IgA index ranged from a maximum of +1.0 to a minimum of −1.0; a score near +1.0 indicated that bacteria were more abundant in the IgA+ fraction. P-values of the IgA index in IgAN patients, as compared with RT patients, are indicated by different colored panels (Mann–Whitney U test, P < 0.05; green, P ≥ 0.05; black). IgA index of tonsillar bacterial phyla (B) and top 10 genera (C) between the IgAN and RT groups. Box plots show the median and IQR, with whiskers of 1.5 × IQR. Data are presented as the median and IQR (*P < 0.05, **P < 0.01, Mann–Whitney U test).

Immunoglobulin repertoire sequencing of tonsillar IgA heavy chain

We designed specific primers for the IgA constant region (Supplementary data, Table S2) to amplify IgA heavy chain-encoding genes [23], and performed adaptor-ligation PCR and immunoglobulin repertoire sequencing using tonsillar RNA samples from 28 IgAN and 20 RT patients (Figure 4A;Supplementary data, Table S3). The relative abundance of IGHC clones in each sample showed that ∼80% was assigned to IGHA1 (Figure 4B). Diversity index analysis of immunoglobulin clones revealed significantly higher Shannon diversity scores for IGHA1 than IGHA2 (Figure 4C). Comparison of the relative abundance of IGHV, IGHD and IGHJ segments with IGHA1 clones revealed significantly higher IGH variable 3-30 (IGHV3-30) (P = 0.0089) and IGHV3-38 (P = 0.00073) expression in IgAN than that in RT tonsils (Figure 4D and F). IGHJ and IGHD segments exhibited no significant difference (Figure 4E).

FIGURE 4

Overview of tonsillar IgA heavy chain sequences using AL-PCR. (A) Schematic representation of AL-PCR for high-throughput sequencing-based antibody repertoire analysis. A universal primer specific for the IgA heavy chain constant region and an adaptor-primer were used for unbiased IgA gene amplification. Gene amplicons were sequenced using an Illumina MiSeq system, and each sequence read was classified into IgA subclasses by discrimination of the constant region sequence. (B) The relative abundance of IGHC (C) gene segments by AL-PCR using the specific primers of the IGHA constant region for the IgAN (n = 28) and RT (n = 20) groups. (C) Shannon diversity scores of IGHA1 and IGHA2 for the IgAN (n = 28) and RT (n = 20) groups. Relative abundance of IGHV (D), IGHD (E, left) and IGHJ (E, right) with IGHA1 in-frame clones for IgAN (n = 28) and RT (n = 20) groups. Assigned reads that could not be determined as one clone in each gene segment were excluded from the analysis. Data are presented as median and IQR (NS; no significance, *P < 0.05, **P < 0.01, ***P < 0.001, Mann–Whitney U test). IGHV3-30 and IGHV3-38 differed significantly. (F) Relative abundance of IGHA1 with IGHV3-30 (left) and IGHV3-38 (right) in-frame clones for the IgAN (n = 28) and RT (n = 20) groups are shown (Mann–Whitney U test).

Correlation of IgA1 expression with IGHV3-30 clones to Bacteroidetes IgA index score and tonsillar APRIL expression

In IGHV3-30 clones, the IgA index of the phylum Bacteroidetes was significantly (P = 0.0089) and positively correlated with only the relative abundance of IGHA1, whereas IGHV3-38 clones showed no significant correlation (Figure 5A). Further, the relative abundance of IGHJ3 recombining with IGHV3-30 of IGHA1 clones was significantly higher (P = 0.0073) in IgAN tonsils and was positively correlated (P = 0.030) with the Bacteroidetes IgA index (Supplementary data, Figure S5 and Table S4). Moreover, the relative abundance of IGHA1 with IGHV3-30 clones was significantly higher (P = 0.046) in patients with high- rather than low-tonsillar APRIL levels, but not tonsillar BAFF levels. However, the relative abundance of IGHA1 with IGHV3-38 clones was significantly higher (P = 0.023) in patients with high- rather than low-tonsillar BAFF levels, but not tonsillar APRIL levels (Figure 5B). CDR3 amino acid length distribution in IGHA1 with IGHV3-30 displayed similar Gaussian distribution in IgAN and RT patients (Figure 5C).

FIGURE 5

Correlations between IGHV3-30/3-38 with IGHA1 clones and the IgA index score of the phylum Bacteroidetes and tonsillar APRIL/BAFF levels. (A) Correlation between the relative abundance of IGHA1 with IGHV3-30 (left) and IGHV3-38 (right) clones and the IgA index of the phylum Bacteroidetes in the IgAN (square) and RT (circle) patient groups, analyzed by both IgA-SEQ and sequencing of the immunoglobulin repertoire (n = 24). (B) Relative abundance of IGHV3-30 (left) and IGHV3-38 (right) with IGHA1 in patients with tonsillar APRIL (left) and BAFF (right) over the median value was compared with those with APRIL and BAFF values less than the median, respectively. The results of the samples analyzed by both immunoglobulin repertoire sequencing and ELISA of tonsillar APRIL/BAFF (n = 45) are shown. (C) Frequencies of each CDR3 amino acid length in IGHA1 with IGHV3-30 for the IgAN (n = 28) and RT (n = 20) groups. Correlation of the data was statistically compared using Pearson’s correlation and linear regression analysis. Data are presented as median and IQR (*P < 0.05, Mann–Whitney U test). Assigned reads not determined as one clone in each gene segment and out-of-frame reads were excluded from the analysis.

Flow cytometry-based analysis of IgA binding to tonsillar bacteria

Serum fractionation was performed using high-performance liquid chromatography with a gel-filtration column, as previously reported [36] (Supplementary data, Figure S6A and B). The F04 (∼650 kDa) containing the immunoglobulin J-chain (Supplementary data, Figure S6C) and F09 (∼150 kDa) fractions were selected as serum pIgA- and mIgA-containing fractions from IgAN and RT tonsils. The pIgA fraction levels of GdIgA1 (P = 0.012) and GdIgA1/IgA ratios (P = 0.0092) were significantly higher in IgAN than in RT (Figure 6A and B), whereas IgA levels were comparable (Supplementary data, Figure S6D). IgA–IgG complex in pIgA fraction was also significantly higher in IgAN than in RT (Supplementary data, Figure S6E). Following IgA concentration adjustment, F04 and F09 fractions were incubated with each bacterium to evaluate IgA binding (Supplementary data, Figure S7). Flow cytometry (Figure 6C;Supplementary data, Figure S8A) revealed that F04 fractions yielded significantly higher proportions of IgA+ bacteria for each bacterium than F09 fractions (Figure 6D). Similarly, IgA in the IgAN F04 fractions exhibited high-binding activities with each bacterium, particularly Porphyromonas gingivalis, compared with RT fractions. IgA binding intensity toward Porphyromonas gingivalis and Prevotella intermedia significantly correlated with both GdIgA1 and IgA–IgG serum levels in IgAN F04 fractions (Figure 6E;Supplementary data, Figure S8B).

FIGURE 6

Analysis of the binding abilities of IgA complexes to cultured bacteria by flow cytometry. (A, B) GdIgA1 levels in the F04 fraction and (A) GdIgA1/IgA ratio in the F04 and F09 fraction (B) from IgAN (n = 30) and RT (n = 7) patients. Box plots show the median and IQR, with whiskers of 1.5 × IQR. Data were statistically compared using the Mann–Whitney U test (NS; no significance, **P < 0.01, ***P < 0.001). (C) The serum mIgA (F09) and pIgA (F04) fractions were examined for the ability of IgA to bind the surface of cultured bacteria by flow cytometry using an anti-IgA monoclonal antibody. Porphyromonas gingivalis, Prevotella intermedia and Fusobacterium nucleatum were isolated and cultured from dissected tonsils of IgAN patients, whereas Eschericia coli was obtained as an ATCC standard strain. Each histogram represents the intensity of IgA binding by isotype-control (yellow), F09 (blue) and F04 (red) fractions. The concentrations of IgA of F09 fractions were adjusted to that of the F04 fraction in each sample. (D) Percent positivity for IgA-PE over the isotype-control for each bacterial cell by F09 and F04 fractions in patients with IgAN (n = 30) and RT (n = 7). Box plots show the median and IQR, with whiskers of 1.5 × IQR. Data were statistically compared using the Mann–Whitney U test (NS; no significance, *P < 0.05, **P < 0.01, ***P < 0.001). (E) Correlation between the percent positivity for IgA-PE for each bacterial species and the levels of GdIgA1 by F04 fractions in patients with IgAN (n = 30). Data were statistically compared using Pearson’s correlation and linear regression analysis.

DISCUSSION

In this study, we showed that in IgAN patients, members of the phylum Bacteroidetes, especially Porphyromonas spp., were more frequently coated with IgA in the tonsils. Further, these could be bound with serum pIgA having higher GdIgA1 levels compared with those in RT tonsils. Moreover, immunoglobulin repertoire analysis using tonsillar IgA revealed higher IGHV3-30 expression with IGHA1 clones in IgAN patients than in RT patients, significantly correlating with the Bacteroidetes IgA index (Figure 7). Our findings suggested that perturbed immunity against a distinct tonsillar microbiota subset may be important in IgAN pathogenesis.

FIGURE 7

Overview of aberrant mucosal immunoreaction in IgAN patients. This study indicated that the aberrant immunoreaction at tonsillar crypts and corresponding Bacteroidetes bacteria dependent polymeric IgA, are involved in IgAN pathogenesis. Enclosed documents indicate the characteristic findings in IgAN patients.

We applied IgA-SEQ to the tonsillar microbiome to determine immunogenic microbiota related to IgAN pathophysiology. Bacteroidetes were markedly coated with IgA in the tonsillar crypts. IgA-coated Porphyromonas spp. exhibited a significantly higher enrichment score in IgAN than RT tonsils. Moreover, tonsillar APRIL and BAFF levels, both important for the reaction of mucosal IgA to microbiota [15] via T-cell-independent pathways [13, 14], were significantly higher in IgAN than in RT tonsils. These data suggested that IgA antibodies are polyreactive to some microbiota, such as members of Bacteroidetes (especially Porphyromonas spp.) and may be preferentially produced in IgAN tonsils. However, antigen-specific IgA may also be produced via T-cell-dependent pathways at these sites.

To examine the mechanism for IgA binding to the microbiota, we conducted high-throughput sequencing of the tonsillar IgA repertoire. A higher amount of IGHV3-30 in IgA was observed in IgAN patient tonsils. Significant correlation with the degree of Bacteroidetes coated with IgA suggests the involvement of the IGHV3-30 of IgA in IgA binding with Bacteroidetes. Furthermore, a significant association between tonsillar APRIL levels and IGHA1 expression with IGHV3-30 clones implied that the T-cell-independent pathway may be involved in IgA prevalence, with IGHV3-30 being polyreactive to a subset of microbiota. IgA produced via the T-cell-independent pathway is more polyreactive with microbiota than T-cell-dependent pathway-associated IgA [37]. In this study, Gaussian distributions of the length of CDR3 in both groups indicated little distinct clonal expansion under the particular antigenic pressure in IgAN patients. Additionally, tonsillar BAFF in IgAN patients was significantly associated with IGHV3-38, another repertoire significantly elevated in IgAN. BAFF acts on B-cell survival and maturation, and in concert with APRIL for class switching to IgA via shared receptors [38]. This may therefore reflect divergent effects of these tumor necrosis factor (TNF) superfamily molecules. Compared with IGHV3-38, the abundance of which was low among the IgA1 repertoire, consistency of association between IGHV3-30, APRIL and IgA response to microbiota may be weighted by APRIL of the genome-wide association study risk loci for IgAN [4, 39].

Consistent with the IgA-SEQ data, Bacteroidetes bacteria were coated with the serum pIgA fraction to a greater extent in IgAN patients than in RT patients. These findings supported the hypothesis that mucosa-derived IgA is increased in IgAN patient serum. BAFF-transgenic mice [34] exhibited increased serum commensal-dependent polymeric GdIgA, indicating the similarities between human and murine IgAN. GdIgA1 levels in the polymeric fraction were significantly increased and were highly correlated with IgA binding with Bacteroidetes in IgAN. This suggests that mucosal GdIgA1 in the tonsils of IgAN patients was mistakenly transported to other sites via BAFF and APRIL. As IgA harboring J-chain is exclusively detected in the polymeric fraction, IgA with IGHV3-30 or 3-38 repertoire could be present in the polymeric form of IgA if dimeric IgA is translocated from the mucosa. Whether these repertoires are involved in polymeric and immune-complex formation remains unclear.

The role of microbiota-dependent IgA in IgAN development is not fully elucidated. However, in the above IgAN models, commensal-dependent IgA was detected in the glomeruli [34]. Conversely, mucosal microbiota depletion by antibiotic administration reduced IgA deposition in humanized murine glomeruli, accompanied with reduced levels of IgA–IgG complexes [40]. Our study also indicated significant associations between tonsillar GdIgA1 levels and glomerular GdIgA1 deposition and mesangial hypercellularity, and between pIgA binding to Bacteroidetes and IgA–IgG complexes. No differences were noted in the tonsillar GdIgA1 levels between IgAN and RT. However, tonsillar GdIgA1 may have different repertoire diversity between both groups, much like IgA1. Discrimination of tonsillar GdIgA1 levels by tonsillar microbiome abundance could substantiate this. Collectively, these data strongly suggested that microbiota-dependent IgA, which was significantly increased in the serum of IgAN patients, underlies glomerular immune-complex deposition and is thus involved in IgAN pathogenesis.

We acknowledged several limitations in this study. First, we collected palatine tonsils from RT patients as control samples because tonsillar samples from healthy subjects were unavailable. However, gram-negative anaerobic bacteria are potentially associated with tonsillitis [41], suggesting that the RT tonsillar microbiome may contain more abundant anaerobic bacteria than that from healthy subjects. Second, not all serum and tonsillar crypt samples were available for IgA-SEQ, IgA repertoire and bacterial FACS analyses, potentially causing some selective bias and lack of power for detecting real differences. However, despite limited samples, we detected statistically robust differences in serum GdIgA1 levels between IgAN and RT, consistent with previous reports [42-44]. Further, our results were consistent across the study. We, therefore, surmised that adequate analyses could be conducted to provide meaningful findings. Lastly, most patients were enrolled in the early stage of IgAN, which affects the correlation between IgA immune response and clinical findings. Future studies with larger samples will, therefore, be needed.

In conclusion, this study provides evidence that tonsillar immune responses, particularly against Bacteroidetes, were perturbed in IgAN patients. These data provided crucial insights into aberrant mucosal immunoreactions against the host tonsillar microbiome and indicate potentially novel therapeutic targets for IgAN management.

SUPPLEMENTARY DATA

Supplementary data are available at ndt online.

Click here for additional data file.