To assess the impact of N-glycosylation on IgA2 (m2) production, we produced several mutants that lacked distinct N-glycosylation sites. We produced a WT variant of the SARS-CoV-2 H4 antibody (Wu et al., 2020) with all five native N-glycosylation sites, the C2Q mutant in which we removed the glycosylation sites NLT and NIT in the CH2 domain, the N2Q mutant in which we removed the glycosylation sites NVT and NSS in the CH1 domain and the C3Q in which we removed the N-glycosylation site in the tailpiece in addition to the CH2 N-glycans (NLT, NIT, NVS) (Figure 1). In addition, a C3Q variant (C3QA) was generated by replacing the penultimate cysteine with alanine. The transient expression of IgA2 (m2) WT, N2Q, and C2Q resulted in similar quantities after CaptureSelect affinity purification, while C3Q and C3QA exhibited poor yields. Size-exclusion chromatography purification revealed that WT and mutants N2Q and C2Q displayed a primary peak corresponding to the monomer and smaller peaks corresponding to the dimer and larger oligomeric species (Figure 2A). The peak for the WT monomer displayed a slight shift and eluted earlier. This is likely due to the higher molecular weight resulting from the presence of two fully N-glycosylated heavy chains. The lack of two N-glycans in the N2Q and C2Q mutants did not affect oligomer formation, as the ratio between monomers, dimers, and polymers did not significantly change across different batches. However, the glycosylation variant C3Q showed a small peak for monomers and a more prominent peak for high molecular weight oligomers (Figure 2A).

The absence of the N-glycan in the tailpiece is known to cause the formation of larger oligomers (Xie et al., 2021; Pan et al., 2024). Lombana et al. (2019) previously observed this phenomenon during dimeric IgA2 (m2) production when co-expressing heavy chain (HC), light chain (LC), and the joining chain (JC). Here, we observed that this aggregation occurs without the JC (Figure 2A). We hypothesize that the N-glycan in the NVS site shields the penultimate cysteine, thereby preventing disulfide bridge formation. The NVS site may be underglycosylated to some degree, which could correlate with the presence of a small amount of dimer, even in the absence of the JC (Figure 2A). To determine whether the lack of three N-glycans in C3Q affects overall hydrophobicity and, consequently, the propensity to aggregate, a C3Q variant with an alanine instead of the penultimate cysteine was generated. The new variant, C3QA, exhibited significantly fewer high molecular weight oligomers compared to C3Q, confirming the hypothesis that the penultimate cysteine is being shielded by the N-glycan (Supplementary Figure S1). The expression of IgA2 (m2) WT, N2Q, and C2Q was tested in their dimeric form by co-expressing the LC, HC, and JC in a proportion 1:1:0.5. The C3Q variant was not included due to poor expression. The yield of the three variants was comparable. The predominant peak for H4-WT and C2Q was the one corresponding to dimers, while the N2Q glycosylation variant still showed a significant amount of monomer (Figure 2B). Overall, the assembly is maintained in forming the dimer, and the SEC profile resembles the previously published data (Lombana et al., 2019), indicating a minor role of the CH1 and CH2 N-glycans in the dimeric assembly. Isolated aliquots of the monomeric form of IgA2 (m2) were utilized for all subsequent analyses (Figure 2C).

The composition and structure of IgA N-glycans are considerably heterogeneous in individuals and different sources (serum or recombinant) (Gomes et al., 2008). Several N-glycan structures have been identified on the five N-glycosylation sites on the IgA2 HC (Ding et al., 2022). Unlike IgG glycans, which are oriented towards the interior of both CH2 domains, molecular modeling revealed that IgA N-glycans were oriented towards the exterior of the molecule (Mattu et al., 1998). The external orientation of IgA N-glycans makes them more accessible to the glycosyltransferases in host cells, resulting in cell-specific glycoforms and extensive N-glycan heterogeneity (Mattu et al., 1998). To investigate the type of N-glycan attached to each site of monomeric IgA2, we performed site-specific analysis using LC-MS/MS. The analysis aimed to evaluate whether removing glycans from a specific site influences the overall N-glycosylation profile on other sites, which would hint at conformational changes.

In Figure 3, the relative abundance of the most abundant N-glycans is shown for each N-glycosylation site. Overall, the N-glycan profiles are quite heterogeneous, with apparent site-specific differences. The N-glycan structures in IgA2 (m2) WT mostly consist of biantennary fucosylated complex N-glycans. WT and N2Q have a higher level of oligomannosidic glycans at the NLT site (41% and 32%, respectively), and the complex N-glycans at this site lack fucose residues. The glycosylation site at the tailpiece, NVS, is not glycosylated to a certain extent, with levels of 20% for WT, 15% for C2Q and 11% for N2Q (Figure 3). The N-glycans at the NSS and NVS sites display more multibranched structures. All variants are site-specifically modified at the NVT, NSS, and NIT sites, with similar trimming/maturation. These branching structures suggest that the local environments surrounding these asparagine residues are similar and that the absence of individual N-glycans has little impact on N-glycan processing.

To assess the impact of N-glycosylation on the thermostability of the different glycoforms, we employed three different methods: differential scanning fluorimetry (DSF), nanoDSF, and CD. As previously shown (Lombana et al., 2019), only one melting transition was observed. All three methods determined the melting temperature (T_m) and indicated that the IgA2 (m2) WT, containing all five N-glycans, is approximately 2°C more stable than the other glycoforms (Figures 4A, B). The results suggest that the N-glycans contribute to thermal stability. However, unlike a previous study (Pan et al., 2024) which reported a variation of 4°C upon removal of two glycans in IgA1, the differences in T_m are less pronounced. CD spectroscopy experiments were carried out to investigate further the effect of different N-glycosylation mutants on the IgA2 (m2) conformation. Far-UV spectra between 260 and 200 nm were recorded for the different variants. The IgA2 (m2) CD spectra showed a minimum at 218 nm, indicating a predominance of β-sheets and a low abundance of α-helical structures (Figure 4C; Supplementary Figure S2). This CD spectrum is consistent with the typical antibody structure and suggests that removing N-glycans does not alter the secondary structure of IgA2 (m2) (Figure 4D).

Recent research suggests that unbound serum IgA, which lacks an antigen, can modulate immune responses through inhibitory signaling involving the immunoreceptor tyrosine-based activation motif (ITAM) (Davis et al., 2020). This modulation promotes anti-inflammatory responses, which are crucial for maintaining the stability of the human body’s internal environment (Ding et al., 2022). Monomeric IgA binds to FcαRI via its CH2 and CH3 domains. The difference in the binding mode between FcαRI and IgA1 and IgA2 has not been extensively studied. However, it is known that IgA1 has a lower K_D compared to IgA2 (Göritzer et al., 2019). Additionally, it has been shown that distinct FcαRI N-glycans modulate the binding affinity to IgAs (Göritzer et al., 2019). However, less is known about the role of specific IgA2 N-glycans in binding with the FcαRI receptor (Carayannopoulos et al., 1994; Oortwijn et al., 2007). To investigate the impact of distinct N-glycans on the binding of IgA2 (m2) to FcαRI, SPR spectroscopy of the N-glycosylation mutants of IgA2 (m2) and FcαRI was conducted. Despite the crystal structure of the IgA1–FcαRI complex suggesting a 1:2 stoichiometry (Herr et al., 2003; Posgai et al., 2018), the response units obtained from the binding curves in SPR experiments indicated a 1:1 binding model. Therefore, the sensorgrams were fitted accordingly. The IgA2 glycoforms exhibit an equilibrium dissociation constant in the 200–400 nM range, consistent with previous studies (Göritzer et al., 2019; Lombana et al., 2019). The glycoforms have similar K_D values, with N2Q showing a slightly higher K_D, which suggests a small allosteric effect resulting from the removal of Fab glycans on the binding site of FcαRI (Figure 5; Supplementary Figure S3).

Small angle X-ray scattering (SAXS) was used to characterize the IgA2 (m2) flexibility of all three glycoforms in solution. The data quality was assessed by evaluating the Guinier plot to ensure accurate structural interpretations. Guinier analyses produced linear plots for all three experimental SAXS profiles (Supplementary Figure S4). The R _g values for all samples were similar, within the range of 48–49 Å (Figure 6A). The IgA2 (m2) R _g value here differs from the previously reported value (Furtado et al., 2004). We reported smaller R _g values, most likely because we performed a size exclusion chromatography coupled SAXS (SEC-SAXS) that eliminated non-specific oligomerization or aggregations in the sample. The maximal dimensions (Dmax) of the IgA2 (m2) variants were determined by the pair distribution function P(r). The maximal dimensions were approximately ∼170 Å for H4-WT and ∼160 Å for C2Q and N2Q. In the P(r), M1 and M2 maxima correspond to the structure’s most frequently occurring interatomic distance. The first P(r) function peak at r ∼ 45 Å arises from the approximate repeated distances across the Fc or Fab regions’ length and breadth (Hodge et al., 2021). The P(r) shoulder at r ∼ 80 Å reflects the inter-domain distances between the Fc and Fab regions (Figure 6B). The difference between the P(r) peak and shoulder indicates the Fc-Fabs’ distancing, which correlates with the extended conformers’ occupancy in solution. The P(r) shapes of the different IgA2 (m2) were similar but distinct. The peaks in H4-WT and N2Q are less separated, indicating a more compact shape for these variants. Conversely, the curve of C2Q is shifted to the left, and the peaks M1 and M2 are more separated. This correlates with a lack of CH2 N-glycans, which could translate to smaller interatomic distances in the Fc region and a shift in the M1 peak.

The Kratky plot analysis of SAXS data is an effective method to evaluate the compactness of a protein and the presence of disordered regions (Hammel, 2012). The bell-shaped Kratky plot becomes asymmetrically stretched for a multidomain IgA2 protein connected by linkers but still adopts a folded character, as shown by the position of maxima at qR _g = 1.7 (Figure 6C). The dimensionless Kratky plot, where I (q) is normalized by I (0) and q is normalized by R _g, allows for the comparison of proteins of different sizes (Hammel, 2012). The plot displays curves typical of a multidomain protein with domains connected by flexible linkers. Shoulders at qR _g ∼ 4 in the C2Q variant indicate a more distinct separation between Fc and Fab domains than in the H4-WT and N2Q cases (Figure 6C). In summary, it has been observed that all the glycoforms show a typical antibody structure in solution. Additionally, experimental SAXS curves indicate that the C2Q variant, which lacks CH2 N-glycans, displays altered mobility of the Fab domains compared to WT and N2Q.

As the crystal structure of the full-length IgA2 (m2) antibody has not been determined, and the only deposited structure (1R70) was obtained by solution scattering, atomistic models obtained with machine learning algorithms were used to enhance the interpretation of SAXS data. The amino acid sequence of the antibody was submitted to AlphaFold-Multimer to obtain the predicted structure (Jumper et al., 2021). In addition to the predicted structure in PDB format, the database also provides a predicted align error (PAE) (Supplementary Figure S5). With the development of accurate structure prediction algorithms such as AlphaFold or RosettaFold (Baek et al., 2021) validating the predicted structure with SAXS is becoming increasingly popular (Chinnam et al., 2022). As noted in previous studies (Hodge et al., 2021), the single conformation predicted by AlphaFold does not match the SAXS data. However, it serves as a starting point for generating an ensemble of conformations. To explore the conformational space of the Fab regions relative to the Fc regions, we used the BILBOMD program (Pelikan et al., 2009). This conformational sampling resulted in a pool of over 10.000 atomistic models, from which SAXS curves were calculated (Schneidman-Duhovny et al., 2013) and compared to the experimental curve. MultiFoXS (Schneidman-Duhovny et al., 2016) provides the weighting of multistate models that fit the experimental data better than a single best-fit model. Two distinct conformations, open and closed, are required to achieve a certain level of accuracy in fitting the SAXS data. The goodness of fit measured by χ2 quantifies the experimental and theoretical differences (Figures 7A, B).

The fit of the SAXS data was significantly improved using a two-state model for each of the three variants. For example, as seen in Figure 7A, the χ2 value of the one-state model of H4-WT decreases from 1.90 to 0.99 with the two-state model. However, the three-state model did not improve the SAXS fit for the three variants. To estimate the number of conformational states in the solution, we examined the R _g distribution for the top 300 selected multistate models (Schneidman-Duhovny and Hammel, 2018). The distribution of R _g in the two-state models (Figure 7C) displays two primary peaks for H4-WT and N2Q. One peak corresponds to closed conformations at R _g 43–46 Å, while the other corresponds to open conformations at R _g 52–54 Å. For C2Q, the closed conformation is described by a sharp peak around 47 Å, while the open conformation is much larger, at 63–66 Å (Figures 7D, E). The R _g-distribution curve indicates a similar population of the closed and open conformers for H4-WT and N2Q. However, the shift in peaks in C2Q suggests less population of the open conformation (Supplementary Video).

Our findings demonstrate that an ensemble of conformations should describe an antibody’s conformation. Additionally, we observed that removing the CH2 glycans from IgA2 (m2) resulted in the acquisition of different conformations compared to WT and N2Q. Specifically, the flexibility of the C2Q Fabs increased, leading to a more open shape.

The codon-optimized genes of the heavy chains and light chain required for expression of the different IgA2 (m2) glycosylation variants in HEK293-6E cells were synthesized by GeneArt (Thermo Fisher Scientific, United States). The sequence of the variable region heavy chains (α-HC) and kappa light chain (κ-LC) were taken from the sequence of IgG-H4 (Kallolimath et al., 2021) and fused to the constant region of human IgA2 (m2). Sequences of the HCs and the κ-LC used for the expression in HEK293-6E were flanked with the coding sequence for the signal peptides “MELGLSWIFLLAILKGVQC” and “MDMRVPAQLLGLLLLWLSGARC,” respectively, and the restriction sites XbaI and BamHI. The synthesized DNA was amplified by PCR with the primers “Strings_9F (CTTCCG- GCTCGTTTGTCTAGA)/Strings_2R (AAA​AAC​CCT​GGC​GGG​ATC​C).” The corresponding genes for the α-HCs and the κ-LC were then separately cloned into the XbaI/BamHI sites of the mammalian expression vector pTT5 (National Research Council of Canada). The codon optimized gene for expression of the joining chain (JC) (AK312014.1) in HEK293-6E cells was synthesized by GeneArt (Thermo Fisher Scientific, United States) and included the coding sequence for the signal peptide “MELGLSWIFLLAILKGVQC” and the restriction sites BamHI/SalI used for cloning. The sequence of the glycosylation variant N2Q was synthesized with the mutation N to Q in the glycosylation sites NVT and NSS, the glycosylation variant C2Q with the mutation N to Q in the glycosylation sites NLT and NIT, the glycosylation variant C3Q with the mutation N to Q in the glycosylation sites NLT and NIT and NVS (Figure 1). The cloning and expression of the extracellular domain of the human FcαRI (P24071.1) was performed as described previously (Göritzer et al., 2019).

The HEK293-6E cell line that constitutively expresses the Epstein−Barr virus nuclear antigen 1 of the Epstein−Barr virus was licensed from the National Research Council (NRC) of Canada (Durocher et al., 2002). The suspension cells were cultivated and transfected according to the manufacturer’s manual in F17 medium supplemented with 0.1% Pluronic F-68, 4 mM L-glutamine (Life Technologies, Germany), and 50 mg/L G418 (Biochrom, Germany). The cells were maintained in shaker flasks at 37°C in a humidified atmosphere with 5% CO₂ on an orbital shaker never exceeding a cell density of 2×10⁶ cells/mL. For transient transfection of a 200 mL culture, cells were brought to a concentration of 1.7 × 10⁶ cells/mL. High-quality plasmid preparations of the pTT5 vector coding for the κ-LC and the different α-HCs were obtained using the PureYield Plasmid Midiprep System (Promega, United States). A total of 200 μg plasmid-DNA, consisting of 100 μg LC and 100 μg of the respective HC, were mixed with 10 mL of fresh medium. Another 10 mL of fresh medium, containing 2.5 μg/mL linear polyethylenimine (PEI) (Polysciences, Germany), was added to the DNA solution and incubated for 10 min. After adding the DNA/PEI mixture, the cells were incubated for 48 h, supplemented with 0.5% (w/v) tryptone N1 (Organotechnie, France) and further cultivated for 72 h. Supernatant containing the secreted soluble protein was harvested by centrifugation at 25,000 g for 30 min at 4°C and additionally filtrated (0.45 μm Durapore membrane filter, Merck Millipore, Germany). For the transient expression of monomeric and dimeric IgA2 in HEK293-6E cells, cultures were transfected with the κ-LC, the different α-HCs and JC constructs in a 1:1:0 and 1:1:0.5 ratio of µg DNA, respectively, as described in (Göritzer et al., 2019).

The supernatant of HEK293-6E suspension cells were subjected to a peristaltic pump packed with 1 mL of CaptureSelect IgA affinity resin (Thermo Fisher Scientific, United States) equilibrated with phosphate-buffered saline (PBS) pH 7.4. Proteins were eluted with 0.1 M glycine pH 2.8, followed by immediate addition of 6 μL of 2 M Tris pH 12 to each 1 mL fraction to neutralize the acidic pH from glycine elution. Highly concentrated fractions were pooled and dialyzed against PBS at 4°C overnight using SnakeSkin Dialysis Tubing with a MWCO of 10 kDa (Thermo Fisher Scientific, United States). Finally, the column was regenerated with 0.1 M glycine pH 2.5 and washed with PBS. Pooled protein fractions were then further concentrated using Amicon centrifugal filters with a MWCO of 10 kDa (Merck Millipore, Germany) and subjected to size-exclusion chromatography (SEC) on a HiLoad 16/600 Superdex 200 pg column (Cytiva, United States) equilibrated with PBS supplemented with 200 mM NaCl. For the recombinant production of the extracellular domain of the FcαRI, the cell culture supernatants were harvested after 6 days and diluted 1:2 in loading buffer (20 mM Tris-HCl pH 7.4, 500 mM NaCl, and 10 mM imidazole). The solution was loaded onto a 5 mL HisTrap HP column (Cytiva) equilibrated with 5 column volumes of loading buffer, and bound protein was eluted by applying 250 mM imidazole. Eluted fractions containing the protein of interest were pooled and dialyzed overnight against Dulbecco’s phosphate-buffered saline (PBS) (Sigma-Aldrich) supplemented with 200 mM NaCl at 4°C using SnakeSkin dialysis tubing with a 10 kDa MWCO (Thermo Fisher Scientific). Protein samples were then further concentrated using a 10 kDa Amicon Ultra centrifugal filter (Merck Millipore).

For reducing or nonreducing SDS-PAGE a total of 5 μg of purified protein was loaded on a 4–15% Mini-PROTEAN TGX gel (Bio-Rad Laboratories, United States) and visualized with Coomassie Brilliant Blue staining.

For SEC-MALS-SAXS experiments, 90 µL of samples containing SEC-purified monomeric IgA2 (m2) at circa 40 µM were prepared in PBS pH 7.4 buffer. SEC-MALS-SAXS was collected at the SIBLYS beamline (BL 12.3.1) at the Advanced Light Source (ALS) at Lawrence Berkeley National Laboratory (LBNL) in Berkeley, California (Hura et al., 2009; Classen et al., 2013; Rosenberg et al., 2022). X-ray wavelength was set at λ = 1.27 A, and the sample to detector distance was 2,100 mm, resulting in scattering vectors, q, ranging from 0.01 A⁻¹ to 0.45 A⁻¹. The scattering vector is defined as q = 4πsinθ/λ, where 2θ is the scattering angle. All experiments were performed at 20°C, and data were processed as described before (Rosenberg et al., 2022). Briefly, a SAXS flow cell was directly coupled with an online Agilent 1260 Infinity HPLC system using a Shodex KW 803 column. The column was equilibrated with running buffer (PBS pH 7.4) with a 0.5 mL/min flow rate. 90 μL of each sample was run through the SEC, and 3 s X-ray exposures were collected continuously during a 25 min elution. The SAXS frames recorded before the protein elution peak were used to subtract all other frames. The subtracted frames were investigated by the radius of gyration (R _g) derived by the Guinier approximation I (q) = I (0) exp (−q2*R _g*2/3) with the limits q* R _g <1.5. The elution peak was mapped by comparing integral ratios to background and R _g relative to the recorded frame using the program BIOXTAS RAW, Supplementary Figure S6 (Hopkins et al., 2017). Final merged SAXS profiles, derived by integrating multiple frames across the elution peak, were used for further analysis, including a Guinier plot, which determined the aggregation-free state. The program GNOM was used to compute the pair distribution function P(r) (Svergun, 1992). P(r) functions for different glycoforms of IgA2, were normalized at the maxima; the area under P(r) function correlates to molecular weight estimated by SAXS. MALS experiments were performed using an 18-angle DAWN HELEOS II light scattering detector connected in tandem to an Optilab refractive index concentration detector (Wyatt Technology). System normalization and calibration were performed with bovine serum albumin using a 45 μL sample at 10 mg/mL in the same SEC running buffer and a dn/dc value of 0.19. UV, MALS, and differential refractive index data were analysed using Wyatt Astra 8 software to monitor the sample’s homogeneity across the elution peak complementary to the SEC-SAXS mentioned above signal validation.

Rigid body modeling along with a FoXS and MultiFOXS was used to define, select, and weight the two-state atomistic model that best agreed with SAXS profiles of IgA2 (m2) variants (Schneidman-Duhovny et al., 2016; Schneidman-Duhovny and Hammel, 2018). The initial structural model of IgA2 was predicted with AlphaFold-Multimer. The glycan moieties were built using the glycan reader component of CHARMM-GUI (Jo et al., 2008). Based on mass spectrometry data, complex glycans were built on the sites NVT, NSS and NIT, mannose type glycans on NLT and a shorter glycan was added to the site NVS to resemble the low occupancy (Jo et al., 2008). In the context of glycoproteins, it is recognized that the scattering attributed to glycans exceeds that originating from the protein component alone (Hammel et al., 2002). Minimal molecular dynamics (BILBOMD) simulation applied on the antibody hinge regions (amino acids 228–235), on the tails (amino acids 449–466), and on the glycans explores the Fab domain’s conformational space relative to the Fc-glycan region. BILBOMD performs minimal molecular dynamics (MD) simulations on the flexible regions at very high temperature, where the additional kinetic energy prevents the Fabs from becoming trapped in a local minimum. In the conformational sampling, individual Fab’s would move independently of one another. A single best-fit and two-state models were selected for each IgA2 using MultiFOXS, Supplementary Figure S7 (Schneidman-Duhovny et al., 2016). We estimated the number of states in the solution by examining the R _g distribution for the top 300 multistate models that all gave the same goodness-of-fit (χ2 ∼ 0.95). The number of main peaks in the distribution indicates the number of states. The area and the position of the peaks validate the level of Fab’s mobility. UCSF Chimera software was used for visualization of PDBs (Pettersen et al., 2004).

Binding experiments of SEC-purified monomeric IgA glycoforms (MW = ∼150 kDa) to FcαRI (MW = ∼27 kDa) were performed using a Biacore T200 system (Cytiva). All assays were performed with PBS running buffer (Cytiva) at 25°C. All antibodies were coupled to a CM5 Sensor Chip (Cytiva) via amine coupling. All sample dilutions were prepared in PBS supplemented with 0.05% Tween, and 0.1% BSA. Capture of the different IgA variants on the chip surface was performed for 60 s with a concentration of 33 µM and a flow rate of 10 μL/min in acetate buffer pH 5.0. The reference flow cell remained unmodified to serve as a reference cell for the subtraction of systematic instrument noise and drift. FcαRI binding curves were generated in single-cycle kinetic experiments at five different concentrations ranging from 12.5 to 400 nM with 60-s association and 60-s dissociation time at a flow rate of 10 μL/min. After each run, surface regeneration was accomplished using 10 mM glycine, pH 1.7, for 120 s at a flow rate of 30 mL/min. Prior to SPR measurements protein concentration for each sample was verified by averaging 3 independent measurements taken with a NanoDrop spectrophotometer in the UV-mode (at 280 nm using sample-specific extinction coefficients). Binding affinities (K_D) were calculated with BIAevaluation software using a 1:1 binding model. Measurements were repeated three times and are reported with standard deviation.

A total of 20 µg of monomeric purified protein was reduced, S-alkylated, and digested with trypsin (Promega). If required, samples were additionally digested with the endoproteinase Asp-N (Sigma-Aldrich). Glycopeptides were then analysed by capillary reversed-phase chromatography and electron-spray MS using a Bruker Maxis 4G Q-TOF instrument as described previously (Grünwald-Gruber et al., 2017). Site-specific glycosylation occupancy was calculated using the ratio of deamidated to unmodified peptide determined upon N-glycan release with peptide:N-glycosidase A (Europa Bioproducts).

Protein thermal shift assays was performed using the FRET channel of a CFX Real-Time PCR Systems (Bio-Rad). IgA2 (m2) glycoforms were diluted in 20 μL at 1 mg/mL in PBS buffer with a final dilution of 1:500 of the Sypro Orange dye stock (Molecular Probes, United States). Melting curves were constructed from 15°C to 90°C (0.5 increments) and data were analysed with the CFX Maestro Software (Bio-Rad). The melting temperatures based on the peak of the first derivative of the fluorescence versus temperature plots were obtained from the software.

CD spectroscopy was performed using a Chirascan™ CD spectrometer (Applied Photophysics, United Kingdom). The instrument was flushed with a nitrogen flow of 5 L/min. The different monomeric IgA2 glycoforms were brought to an absorbance at 280 nm of 0.8 in PBS, pH 7.4. Samples were measured in a cuvette with a path length of 1 mm in the far-UV region ranging from 190 to 260 nm, a 5-nm/s scan speed, and a 3-nm bandwidth. The transition from folded state to unfolded state is monitored by measuring the ellipticity values at 222 nm upon gradually heating from 20°C to 90°C, with heating rates 0.5°C/min.

NanoDSF was performed using Prometheus equipped with back reflection mode (NanoTemper Technologies, München, Germany). SEC-purified monomeric IgA2 (m2) were loaded in nanoDSF grade standard capillaries and exposed at temperatures from 15°C to 90°C using a thermal ramping rate of 1°C/min. Fluorescence emission from tryptophan after UV excitation at 280 nm was collected at 330 nm and 350 nm with a dual-UV detector. Thermal stability parameters were calculated using the ThermControl software (NanoTemper Technologies, München, Germany).