We analyzed the sera from 33 and 32 PV patients assigned to the rituximab (RTX) and standard corticosteroid (CS) arms of the Ritux 3 trial, respectively, and for whom serum samples were available at baseline (24).

The longitudinal analysis was performed in 33 of these 65 PV patients (16 treated with RTX and 17 treated with CS) corresponding to those who had persistent anti-DSG3 Abs during the course of their disease as measured using the EUROIMMUN ELISA assay, whether or not they relapsed. Additionally, sera from 36 healthy donors (HD), 6 PF, 9 bullous pemphigoid (BP), and 5 PV patients with negative anti-DSG3 Abs after RTX treatment were used as negative controls.

Sera from patients were analyzed before treatment at baseline, after treatment, and at the time of relapse, if applicable. To detect and quantify anti-DSG3 Ab IgG subclasses, we developed an ALBIA-DSG3, which consisted of coupling human recombinant DSG3 protein to fluorescent beads (LiquiChip Ni-NTA Beads; Qiagen, Hilden, Germany) according to the manufacturer’s protocol. To determine the isotype of serum anti-DSG3 Abs, DSG3-coated beads were incubated with sera diluted at 1:150, then incubated with anti-IgG1 (1:125), anti-IgG2 (1:125), anti-IgG3 (1:200), or anti-IgG4 (1:200) biotinylated secondary antibody (SouthernBiotech, Birmingham, AL, USA), and finally with streptavidin–R-phycoerythrin (Qiagen). The mean fluorescence intensity (MFI) was determined on a Bio-Plex apparatus using Manager software version 4.0 (Bio-Rad, Hercules, CA, USA). Negative control with no serum and positive control [anti-DSG3 Calibrator of ELISA kit (EUROIMMUN)] were included in every assay. The anti-DSG3 Ab serum levels were determined with the formula (MFI^serum/MFI^Calibrator) × 100, in which the calibrator was the anti-DSG3-positive control previously mentioned that was used on every 96-well plate and set arbitrarily to 100 arbitrary units (AU). For each isotype, we considered a positivity threshold corresponding to + 3 standard deviations relative to the mean value obtained from the sera of 36 HD.

Purification of IgG was performed by affinity chromatography, using ÄKTA-Start. The HiTrap Protein G column (GE Healthcare, Chicago, IL, USA) was equilibrated with 10 ml of phosphate-buffered saline (PBS) 1×, at pH 7.4. The 1/5 pre-diluted sera were added and IgG were eluted using glycine buffer 0.1 M, pH 2.7, followed by neutralization with 1 M Tris pH 9. Analyses were performed using UNICORN 7.0 software to collect IgG-containing fractions. Purified IgG were quantified using the BCA Protein Assay Kit (Pierce™, Rockford, IL, USA) according to the manufacturer’s instructions.

IgG3 purification was performed using the CaptureSelect™ IgG3 (Hu) Affinity Matrix (Thermo Scientific™, Rockford, IL, USA). The IgG3 column was equilibrated with 10 ml of phosphate-buffered saline (PBS) 1×, at pH 7.4. The 1/10 pre-diluted purified IgG were added, and flow without IgG3 was collected. IgG3 were then eluted using glycine buffer 0.1 M, pH 3, followed by neutralization with 1 M Tris pH 9. Purified IgG3 and flow without IgG3 were quantified using the BCA Protein Assay Kit (Pierce™) according to the manufacturer’s instructions.

HaCaT cells were cultivated in 24-well plates with 600,000 cells per well in DMEM + GlutaMAX (Gibco, Grand Island, NY, USA) containing 1 mM CaCl₂ in a humidified and controlled atmosphere (5% CO₂) at 37°C. Twenty-four hours after reaching confluency, positive control AK23 (10 µg/ml), HD IgG (62.5 µg/ml), PV IgG (62.5 µg/ml), or IgG-depleted fractions collected from IgG-specific affinity purification (62.5 µg/ml) were added and incubated for 24 h. The amount of purified IgG used for the keratinocyte dissociation test of PV patients in CR whose sera contained only the anti-Dsg3 IgG4 subclass (before and after treatment) was adjusted to the level of anti-DSG3 Abs measured in the sera collected at baseline from the corresponding patients by a cross multiplication. Subsequently, the cells were treated with dispase solution (2.4 U/ml; Sigma-Aldrich, St. Louis, MO, USA) at 37°C until monolayers were released from plates. Monolayers were stained with crystal violet (Sigma-Aldrich) and subjected to mechanical stress by vigorously pipetting 7 times with a 1-ml pipette. Cell fragments were fixed, photos were taken from each well, and cell fragments were counted manually. All experiments were performed in triplicate.

HaCaT cells were cultivated on 4-chamber Lab-Tek with 50,000 cells per cm² in DMEM + GlutaMAX (Gibco) containing 10% fetal bovine serum and 1 mM CaCl₂ per chamber until they reached at least 75% confluency. Cells were washed and incubated for 20 h with 62.5 µg/ml of HD IgG, PV IgG, or IgG-depleted fractions collected from IgG-specific affinity purification. Then, cells were fixed with 100% ethanol for 10 min, permeabilized with 0.3% Triton for 10 min, and washed after each step. Rat serum diluted at 1% was used for blocking. Cells were then incubated with a primary rabbit antibody anti-DSG3 coupled to Alexa-Fluor^® 647 (Santa Cruz, Dallas, TX, USA) for 90 min in the dark under slow agitation. Cells were finally washed and dried for 15 min, and one drop of mounting medium with DAPI (Invitrogen, Carlsbad, CA, USA) was added per condition. Photos were taken using a multiphoton confocal microscope Leica TCS SP8.

All statistical analyses were performed using GraphPad Prism (GraphPad Software, La Jolla, CA, USA). Determination of specificity and sensitivity of anti-DSG3 ALBIA and comparison of frequencies of anti-DSG3 subclasses were assessed using Fisher's exact test. Correlations were assessed using Pearson’s rank correlation coefficient. The numbers of anti-DSG3 IgG subclasses in relasping versus non-relapsing patients and the mean anti-DSG3 IgG4 serum level in patients with persistent active disease versus those in CR were compared using unpaired t-test. In PV patients with exclusively anti-DSG3 IgG4 Abs, the level of anti-DSG3 IgG4 at baseline and during the evolution was compared using a paired t-test. Differences were considered significant when p < 0.05.

Sera from 65 patients were tested at baseline (33 treated with RTX and 32 treated with CS) ( Figure 1 ).

A relapse occurred in 7 of 33 (21.2%) patients treated with RTX and in 20 of 32 (62.5%) treated with CS. The 7 patients who relapsed in the RTX group relapsed before Month 12 and were retreated with additional infusions of RTX. All of them were in complete remission (CR) at the Month 36 evaluation.

The distribution of anti-DSG3 IgG subclasses after treatment was performed in 2 subgroups of patients: i) those who relapsed or had persistent disease activity and ii) those who had a sustained CR. These latter patients were selected from the whole group of patients in CR after treatment on the fact that they had persistent positive serum anti-DSG3 Abs by ELISA, allowing to assess the distribution of anti-DSG3 Abs after treatment. Seventeen patients had a persistent disease activity after treatment, including the 7 patients who relapsed early after RTX, and 10 out of the 20 patients who relapsed after CS treatment.

The same serological analyses were performed at Month 36 after the start of treatment in 16 patients in sustained CR, including 7 treated with CS and 9 with RTX.

We first assessed the specificity and sensitivity of our ALBIA for the detection of anti-DSG3 IgG subclasses. For this, we compared anti-DSG3 ALBIA performed with anti-DSG3 Abs containing sera from 65 PV patients with 56 control sera (36 HD, 6 PF, 5 PV with negative anti-DSG3 Abs after RTX treatment, and 9 BP). Our ALBIA had 92.31% sensitivity and 85.71% specificity for the detection of anti-DSG3 Abs ( Table 1 ).

Since IgG1 and IgG4 are the main anti-DSG3 IgG subclasses, we then correlated anti-DSG3 IgG1 and IgG4 serum levels measured by ALBIA with anti-DSG3 Ab serum levels measured by ELISA in sera from the same patients with active disease. A high correlation was observed for both IgG1 (r = 0.60, p < 0.0001) and IgG4 (r = 0.55, p < 0.0001) anti-DSG3 Abs ( Figure 2 ).

At baseline, anti-DSG3 IgG4 and IgG1 were detected in 90.8% and 40.0% of sera, respectively, whereas IgG2 and IgG3 were both detected in 26.2% of sera ( Figure 3A ).

Among the 60 patients with positive ALBIA, the most frequently detected isotypes of anti-DSG3 Abs were IgG4 either alone (36.7%) or most often combined with IgG1 (16.7%), IgG2 (8.3%), and IgG3 (10.0%). The combination of 3 or 4 IgG subclasses of anti-DSG3 Abs (including IgG4 in all cases) was observed in 26.6% of sera. Only one serum (1.7%) did not contain anti-DSG3 IgG4 Abs ( Figure 3A ).

Among the patients with positive ALBIA, most sera (64.3%) from patients in CR contained only one anti-DSG3 IgG subclass (corresponding to IgG4 in all but one case), as compared with 25.0% of sera from patients with persistent disease activity, and 36.7% of baseline sera ( Figure 3 ). In contrast, 35.7% of sera from patients in CR contained at least 2 or more anti-DSG3 IgG subclasses, as compared with 75.0% of sera from patients with persistent active disease and 63.3% of baseline sera ( Figures 3B, C ).

Since RTX is approved as a first-line treatment for pemphigus, we assessed whether the number and the isotypes of anti-DSG3 IgG subclasses in sera from patients at the onset of pemphigus might predict the occurrence of relapses under treatment or after treatment withdrawal.

The mean number of anti-DSG3 IgG subclasses in the baseline sera was higher in patients who further relapsed than in patients who maintained a sustained clinical remission (2.6 ± 0.8 vs. 1.5 ± 0.9; p = 0.01) ( Figure 4A ). In particular, 5 of the 7 (71.4%) patients who relapsed had 3 or more anti-DSG3 IgG subclasses in their baseline serum versus 3 of the 26 (11.5%) patients who maintained a prolonged remission (p = 0.004), corresponding to a positive predictive value of 62.5% (95% CI 0.31–0.86) and a negative predictive value of 92% (95% CI 0.75–0.99) for the occurrence of relapses ( Figures 4B, C ).

Moreover, the occurrence of relapses was particularly frequent in patients whose baseline serum contained anti-DSG3 IgG3 Abs, since the baseline sera from 5 of the 7 relapsing patients (71.4%) contained anti-DSG3 IgG3 Abs versus 3 out of the 26 sera from non-relapsing patients (11.5%, p = 0.004) ( Figures 4B, C ).

The distribution of the other anti-DSG3 IgG subclasses (IgG1, IgG2, and IgG4) was not associated with the occurrence of relapse (p > 0.99, p = 0.39, and p = 0.58).

Among the 16 patients who had exclusive anti-DSG3 IgG4 at baseline and did not further relapse, we selected 4 patients in sustained CR whose serum still contained anti-DSG3 IgG4. Figure 5A shows the decrease of anti-DSG3 IgG4 serum levels from these 4 patients (identified in red) relative to the 12 other sera which did not contain anti-DSG3 Abs anymore after treatment.

Interestingly, in patients whose sera contained anti-DSG3 IgG4 exclusively after treatment, the mean anti-DSG3 IgG4 Ab serum level in patients with persistent active disease was higher than in patients in CR (19.70 ± 13.26 AU versus 6.30 ± 6.06 AU; p = 0.044) ( Figure 5B ).

In agreement with this observation, Figure 5C shows that the serum collected at baseline (containing 98.2 AU of anti-DSG3 IgG4 by ALBIA) induced a strong keratinocyte dissociation, whereas the serum from the same patient collected at the time of CR (containing 1.0 AU of anti-DSG3 IgG4) did not induce a keratinocyte dissociation anymore. Interestingly, this latter serum recovered its pathogenic activity in keratinocyte dissociation assay after adjustment of the IgG4 level to the one at baseline ( Figure 5C ). In order to exclude the possible role of non-desmoglein Abs whose concentration could also have been increased after adjustment of the anti-DSG3 IgG titer, we previously checked that the sera collected from this patient both at baseline and at the time of CR did not contain anti-Dsc3 (desmocollin 3), anti-SPCA1 (secretory pathway Ca²⁺/Mn²⁺-ATPase isoform 1), or anti-CHRM3 (cholinergic receptor muscarinic 3) Abs, which have been reported to be involved in the pathogenicity of pemphigus patients (25, 26). These findings thus suggested that anti-DSG3 IgG4 collected at the time of CR were pathogenic, but the amount of antibody was not sufficient to induce in vitro keratinocyte dissociation.

We then tested a serum which contained at baseline a combination of anti-DSG3 IgG3 and IgG4 at rather low levels (IgG3: 16.5 AU; IgG4: 10.8 AU) and contained only IgG4 at the time of CR after treatment (6.4 AU). Purified IgG from the baseline serum but not from the serum collected at the time of CR induced a keratinocyte dissociation, while both sera contained rather close levels of anti-DSG3 IgG4 Abs (10.8 AU and 6.4 AU, respectively). Interestingly, when removing IgG3 Abs from the baseline serum, the purified IgG did not induce a keratinocyte dissociation anymore, whereas the purified IgG3 fraction alone did ( Figure 5D ), suggesting the role of the anti-DSG3 IgG3 subclass in the pathogenic activity of this patient’ serum, in addition to IgG4.

We then tested this latter serum using an immunofluorescence assay to further assess its pathogenic activity. Figure 5E shows that, in contrast with the purified IgG from a HD and from the patient’ IgG collected at the time of CR which both showed a linear labeling of DSG3 on the HaCaT cell plasma membrane, the baseline purified IgG fraction which contained a combination of anti-DSG3 IgG3 and IgG4 induced the disappearance of the DSG3 labeling, confirming the pathogenic activity of this baseline serum. Finally, removing the IgG3 fraction from the baseline patients’ serum led to the reappearance of the DSG3 staining, with a labeling close to that observed with control sera from HD, while the IgG3 fraction induced the disappearance of the DSG3 labeling ( Figure 5E ).