The autoimmune blistering skin disease pemphigus is caused by autoantibodies against keratinocyte surface Ags. In pemphigus vulgaris (PV), autoantibodies are primarily directed against desmosomal cadherins desmoglein (Dsg) 3 and Dsg 1, whereas pemphigus foliaceus (PF) patients only have Abs against Dsg 1. At present, it is unclear whether Dsg autoantibodies contribute to pemphigus pathogenesis by direct inhibition of Dsg transinteraction. Using atomic force microscopy, we provide evidence that PV-IgG directly interfere with homophilic Dsg 3 but, similar to PF-IgG, not with homophilic Dsg 1 transinteraction, indicating that the molecular mechanisms in PV and PF pathogenesis substantially differ. PV-IgG (containing Dsg 3 or Dsg 1 and Dsg 3 autoantibodies) as well as PV-IgG Fab reduced binding activity of Dsg 3 by ∼60%, comparable to Ca2+ depletion. Similarly, the mouse monoclonal PV Ab AK 23 targeting the N-terminal Dsg 3 domain and AK 23 Fab reduced Dsg 3 transinteraction. In contrast, neither PV-IgG nor PF-IgG blocked Dsg 1 transinteraction. In HaCaT monolayers, however, both PV- and PF-IgG caused keratinocyte dissociation as well as loss of Dsg 1 and Dsg 3 transinteraction as revealed by laser tweezer assay. These data demonstrate that PV-IgG and PF-IgG reduce Dsg transinteraction by cell-dependent mechanisms and suggest that in addition, Abs to Dsg 3 contribute to PV by direct inhibition of Dsg transinteraction.
Pemphigus is a severe autoimmune blistering skin disease (1, 2) caused by autoantibodies against keratinocyte surface Ags (3, 4, 5). It has been demonstrated that pathogenic pemphigus autoantibodies are directed to the cadherin-type adhesion molecules desmoglein (Dsg)3 1 and 3 (6, 7, 8, 9). However, at present, there is evidence supporting the hypothesis that autoantibodies against other targets including cholinergic receptors or pemphaxin also contribute to skin blistering (10, 11, 12). Whether these different autoantibodies are pathogenic or just represent an epiphenomenon secondary to acantholysis is a matter of serious debate (13, 14, 15). It is widely accepted that there is a correlation between the clinical phenotype and the profile of Dsg autoantibodies (1, 2). Pemphigus foliaceus (PF) is usually characterized by epidermal blistering without development of mucosal erosions and an autoantibody profile including IgG directed to Dsg 1 but not to Dsg 3 (2, 16). Autoantibodies to Dsg 3, but not to Dsg 1, are usually present in patients with mucosal dominant pemphigus vulgaris (PV) (16, 17). In (PV) patients showing both mucous membrane and skin involvement, Abs to both Dsg 3 and Dsg 1 may be detected (2, 16, 17). However, there are also cases where the Ab profile against Dsg 1 and 3 in PV does not strictly correlate with these clinical phenotypes (18, 19). At present, direct evidence that the presence of Dsg 3-specific Abs in PV but not in PF may account for the more severe clinical phenotype of PV compared with PF patients is lacking.
Because Dsg 3 was found to be the target Ag of PV autoantibodies and to be a cadherin-type adhesion molecule, it was suggestive to believe that Dsg 3 Abs could directly interfere with Dsg 3 transinteraction (6). To test this hypothesis, some progress has been achieved by establishing mouse monoclonal Dsg 3 Abs, which are well characterized regarding the Dsg 3 extracellular subdomain they are directed to (20). It has been reported that AK 23, targeting the N-terminal extracellular domain 1 (EC 1) of Dsg 3, where the predicted adhesive interface is located (21, 22), is pathogenic. In contrast, Abs against the middle portion or the juxtamembrane part of the extracellular Dsg 3 domain had no effect (20). These data suggested that AK 23 could directly interfere with Dsg 3 transinteraction. This would be relevant because Abs to Dsg 1 and Dsg 3 in PV and PF patients are also primarily directed against the EC 1 subdomain (23). However, we found that PF-IgG caused keratinocyte dissociation and loss of Dsg 1 transinteraction without directly blocking Dsg 1 transinteraction, indicating that direct inhibition of Dsg transinteraction is not essential for PF pathogenesis and that Abs targeting the EC 1 subdomain do not necessarily induce direct inhibition (24). Taken together, evidence that Dsg 3 autoantibodies contribute to PV pathogenesis by directly blocking Dsg 3 transinteraction is lacking.
Therefore, the aim of the present study was to clarify whether Abs to Dsg 1 and Dsg 3 from PV patients reduced Dsg transinteraction by cellular signaling mechanisms or directly interfered with Dsg transinteraction. We used single-molecule atomic force microscopy (AFM) which allowed us to study Dsg transinteraction in a cell-free system and thus to rule out the contribution of any cellular signaling pathways. In contrast, we combined this approach with laser tweezer studies testing the binding of microbeads coated with human Dsg 1 and Dsg 3 to the surface of human HaCaT keratinocytes, thereby evaluating the contribution of cellular mechanisms to the effects of pemphigus IgG.
Materials and Methods
Cell culture and test reagents
The immortalized human keratinocyte cell line HaCaT was grown in DMEM (Invitrogen Life Technologies) including 1.8 mM Ca2+ which was supplemented with 50 U/ml penicillin G, 50 μg of streptomycin, and 10% FCS (Biochrom) in a humidified atmosphere (95% air/5% CO2) at 37°C. The cultures were used for all experiments when grown to confluent monolayers and Dsg 1 expression was detected by dot-blot analysis as well as immunostaining, which was the case on day 7 after plating. Under these conditions, formation of desmosomes was regularly detected by electron microscopy (24). Monoclonal mouse PV Abs AK 23, AK 18, and AK 9 were purchased from Biozol and used at 75 or 160 μg/ml for experiments. Monoclonal mouse Ab directed against the extracellular domain of Dsg 1 (aDsg 1) was purchased from Progen Industries (clone p124) and used at a 1/20 dilution.
Purification and preparation of patients’ IgG
Purification was performed as described previously (24). Sera from two PF patients, two patients suffering from a mucocutaneous form of PV, and two patients with mucous PV whose diagnoses were confirmed clinically, histologically, and serologically and from a volunteer without any skin disease (control IgG) were used for the present study. Patients’ sera were tested by ELISA for reactivity against Dsg 1 and Dsg 3, respectively (Table I). IgG fractions PF-IgG 1 and 2 contained Dsg 1 Abs but no Dsg 3 Abs, whereas PV-IgG 1 and 2 contained both Dsg 1 and Dsg 3 Abs. PV-IgG 3 and 4 contained Dsg 3 but not Dsg 1 autoantibodies. IgG fractions were purified by affinity chromatography using protein A agarose. In preliminary experiments, we determined the concentrations of patients’ IgG that had maximal effects on Dsg-binding activities in AFM experiments. Afterward, final concentrations of IgG fractions were adjusted to 150–500 μg/ml for all experiments. For some experiments, three different PF-IgG from additional PF patients were pooled, all of which were tested positive for Dsg 1 and negative for Dsg 3 (data not shown). Preparation of Fab was performed as described previously (24). PV-IgG Fab were generated by papain digestion of PV-IgG 3, PF-IgG Fab by digestion of PF-IgG 1. PV- and PF-IgG Fab were used at concentrations adjusted to IgG fractions (150–500 μg/ml).
|ELISA .||Dsg 1 (U/ml) .||Dsg 3 (U/ml) .|
|ELISA .||Dsg 1 (U/ml) .||Dsg 3 (U/ml) .|
Recombinant Dsg 1-Fc, Dsg 3-Fc, and vascular endothelial (VE)-cadherin-Fc
Expression and purification of rFc constructs of human Dsg 3, Dsg 1, and mouse VE-cadherin were performed as described for Dsg 1 (24) using protein A agarose affinity chromatography. For additional experiments, Dsg 1—also containing a hexahistidine tag (his-tag)—was purified using Ni-NTA agarose chromatography according to the manufacturer’s protocol (Roche). The protein was eluted by imidazole buffer (200 mM imidazole, 1 M NaCl, 10 mM NaH2PO4 (pH 8)) and was immediately subjected to buffer exchange against HBSS via PD-10 desalting columns (GE Healthcare). Dsg 1 was cleaved by exfoliative toxin A (ETA; provided by M. Amagai, Keio Medical School, Tokyo, Japan). Dsg 3 but not Dsg 1 was precipitated by the Dsg 3-specific mouse mAb AK 23 (data not shown).
Fluorescent detection of Ca2+ binding with quin-2
Binding of Ca2+ to chimeric Dsg proteins was detected using the fluorescent Ca2+ indicator quin-2 (25). Briefly, Dsg 1, Dsg 3, or BSA was transferred to polyvinylidene difluoride membranes using vacuum aspiration. The membranes were washed with solution AQ (60 mM KCl, 5 mM MgCl2, and 10 mM imidazole-HCl buffer (pH 6.8)) additionally containing 1 mM CaCl2 for 20 min at room temperature (RT), followed by a 1-h incubation at RT in solution AQ containing 1 mM quin-2. After washing with buffer AQ for 5 min, membranes were dried, finally illuminated with UV light, and photographed through a green filter.
Depletion of Dsg autoantibodies from pemphigus IgG
Pemphigus IgG were depleted of Dsg autoantibodies by immobilization of chimeric Dsg proteins using Talon dynabeads (Invitrogen) as previously reported for Dsg 1 (24). A total of 50 μg of each Dsg 1 and Dsg 3 were incubated with 5-mg beads in 200 μl of HBSS for 0.5 h at RT and slow overhead rotation. The supernatant was discharged and proteins were bound to beads were washed three times with HBSS (200 μl) using a magnetic tube holder. To absorb Dsg autoantibodies, PV-IgG 3 (200 μl containing 0.6 mg IgG) were applied to the beads and incubated for 0.5 h at RT and slow overhead rotation. The supernatant containing the Dsg IgG-depleted IgG fraction (PV-IgG 3 Abs) was finally collected and used for additional experiments. Generation of PF-IgG 1 Abs depleted of Dsg 1 autoantibodies was achieved similarly by incubation of PF-IgG 1 with Dsg 1 immobilized on Talon beads. PV- and PF-IgG Abs were used at 200 μg/ml.
HaCaT cells were grown on coverslips to confluence (7 days) and incubated with pemphigus IgG or mAbs for 24 h at 37°C. After incubation with autoantibodies, culture medium was removed and monolayers were fixed for 10 min at RT with 2% formaldehyde (freshly prepared from paraformaldehyde) in PBS. Afterward, monolayers were treated with 0.1% Triton X-100 in PBS for 5 min at RT. After rinsing with PBS at RT, HaCaT cells were preincubated for 30 min with 10% normal goat serum and 1% BSA in PBS at RT and incubated for 16 h at 4°C with mouse mAb directed to Dsg 3 (Zytomed) (dilution 1/100 in PBS). For experiments using mouse mAbs AK 23, AK 18, AK 9, or aDsg 1, a rabbit polyclonal Dsg 3 Ab was used (dilution 1/100 in PBS; Santa Cruz Biotechnology). After several rinses with PBS (three washes for 5 min each), monolayers were incubated for 60 min at RT with Cy3-labeled goat anti-mouse or goat-anti-rabbit IgG (Dianova). Cells were then rinsed with PBS (three washes for 5 min each). Finally, coverslips were mounted on glass slides with 60% glycerol in PBS, containing 1.5% N-propyl gallate (Serva) as antifading compound. Monolayers were examined using a LSM 510 (Zeiss). Images were processed using Adobe Photoshop 7.0 software (Adobe Systems).
Homophilic transinteractions of Dsg 1 and Dsg 3 were characterized by force-distance measurements of Dsg 1 or Dsg 3 coupled via flexible linkers to the tip and substrate of a Bioscope AFM driven by a Nanoscope III controller (Digital Instruments). Dsg 1 or 3 was linked covalently to the Si3N4 tip of the cantilever (Veeco Instruments) and freshly cleaved mica plates (SPI Supplies) using polyethylene glycol (PEG) spacers containing an amino-reactive cross-linker group (N-hydroxysuccimide ester) at one end and a thiol-reactive group (2-[pyridyldithio] propionate) at the other end, as described previously in detail (26). The N-hydroxysuccimide group served to link PEG to free amino acid groups at both the Si3N4 tip and mica, introduced by treatment of tip and mica with 2-aminoethanol HCl (Sigma-Aldrich). Binding events were measured in buffer A (140 mM NaCl, 10 mM HEPES, 5 mM CaCl2) by force-distance cycles at amplitudes of 500 nm and at 2 Hz frequency. “Ca2+-free conditions” (no Ca2+) were defined as buffer A without addition of CaCl2. Force-distance cycles were performed at constant lateral positions and analyzed as described previously in detail (27). Binding activity was normalized to experiments using a cantilever not labeled with Dsg 1 or Dsg 3 to eliminate the contribution of unspecific interactions. For Dsg 1/3 heterophilic-binding analysis, heterophilic-binding activities of VE-cadherin-Fc with Dsg 1 or 3 were measured. These values were subtracted from heterophilic Dsg-binding activities before the latter were normalized to homophilic-binding activities of Dsg 1 or Dsg 3, respectively. For some experiments, both Dsg 1 and Dsg 3 at equal amounts were coated on mica or AFM tips.
Coating of polystyrene beads and the working principle of the laser tweezer set-up were described previously in detail (24). Coated beads (10 μl of stock solution) were suspended in 200 μl of culture medium and allowed to interact with HaCaT monolayers for 30 min at 37°C before measuring the number of bound beads (=control values). Beads were considered tightly bound when resisting laser displacement at 42 mW setting. For every condition, 100 beads were counted. Afterward, PV-IgG or PF-IgG were applied for 30 min. Percentage of beads resisting laser displacement under various experimental conditions was normalized to control values.
Electrophoresis and Western blotting
HaCaT cells grown for 1 or 7 days were dissolved in sample buffer, sonicated, heated at 95°C for 5 min, and finally subjected to SDS 7.5% PAGE and immunoblotting to Hybond nitrocellulose membranes (Amersham). Membranes were blocked with 5% low fat milk for 1 h at RT in PBS and incubated with anti-Dsg 1 (1/200; Progen) or anti-actin (1/3000; Sigma-Aldrich) primary Ab overnight at 4°C. As secondary Ab, HRP-labeled goat anti-mouse Ab (Dianova) was used. Visualization was achieved using the ECL technique (Amersham).
Differences in bead adhesion or single-molecule transinteraction between different protocols have been assessed using the two-tailed Student t test. Values throughout are expressed as mean ± SEM. Statistical significance is assumed for p < 0.05.
PV-IgG as well as PF-IgG induced cell dissociation in cultured human keratinocytes (HaCaT)
IgG fractions of six different patients with clinically, histologically, and serologically verified pemphigus were used (Table I). PV-IgG 1 and 2 contained autoantibodies against Dsg 1 and Dsg 3, whereas PV-IgG 3 and 4 contained Dsg 3-specific autoantibodies only. Abs to Dsg 1 but not to Dsg 3 were present in PF-IgG 1 and 2.
First, we studied the pathogenic effect of pemphigus IgG on cultured human keratinocytes (HaCaT; Fig. 1). Under control conditions or following treatment with IgG from a healthy volunteer (control IgG), Dsg 3 was continuously distributed along cellular junctions (Fig. 1, A and B). In contrast, PV-IgG treatment resulted in disruption of Dsg 3 staining. Cell dissociation leading to formation of intercellular gaps (indicated by arrows in Fig. 1,D) was observed and further substantiated using F-actin staining for all PV-IgG fractions used (data not shown)—comparable to our previous studies (28, 29) (Fig. 1, C–E). It is noteworthy that the effects of PV-IgG were similar independent of whether Dsg 1 Abs were present (PV-IgG 1) or not (PV-IgG 3 and 4). Gap formation was also observed (arrows) when Fab of PV-IgG were used, whereas the profound disruption of Dsg 3 staining was abolished (Fig. 1,F). However, all pathogenic effects were eliminated by depletion of autoantibodies against Dsg 1 and 3 from PV-IgG using rDsg for immunoabsorption (Fig. 1 G).
To further examine the effect of Dsg 3-specific PV Abs, we used three different mouse monoclonal PV Abs directed to well-characterized epitopes on the Dsg 3 extracellular domain (20). It has been shown that AK 23, which is directed against the Dsg 3 N-terminal EC 1, is pathogenic, whereas AK 18 directed against the middle part or AK 9 directed against the juxtamembrane part of the Dsg 3 extracellular domain are not. Consistent with these findings, when applied to HaCaT cells, AK 23 (75 μg/ml) disturbed Dsg 3 localization at cell junctions leading to linear streaks oriented perpendicular to cell borders. However, unlike PV-IgG from patients, AK 23 did not induce formation of large intercellular gaps or a pronounced loss of Dsg 3 staining. Because it has been demonstrated in keratinocyte cultures that AK 23 had dose-dependent effects up to 160 μg/ml but not at higher concentrations (30), we used 160 μg/ml AK 23 as well but with similar results (data not shown). AK 9 and AK 18 had no effect (Fig. 1, H–J), even when applied at higher concentrations (data not shown).
PF-IgG 1 and 2 as well as PF-IgG Fab induced keratinocyte dissociation leading to formation of intercellular gaps (arrows in Fig. 1, K–M). However, in contrast to PV-IgG, Dsg 3 staining was only missing at gap margins indicating that Dsg 1 Abs were sufficient to cause keratinocyte dissociation whereas Dsg 3 Abs in PV-IgG were responsible for profound fragmentation of Dsg 3 immunostaining. These effects were abolished by depletion of Dsg 1 Abs from PF-IgG (Fig. 1,N). A mouse mAb directed against the extracellular domain of Dsg 1 (aDsg 1) had no effect (Fig. 1 O).
Determination of Dsg 3 and Dsg 1 single-molecule-binding activities by AFM
We previously used single-molecule AFM force spectroscopy to demonstrate that PF-IgG did not directly interfere with Dsg 1 transinteraction (24). To investigate whether Dsg autoantibodies in PV are pathogenic and whether these IgG would directly reduce the binding activity of Dsg 3 and Dsg 1 in a cell-free system, AFM force spectroscopy was applied as described in detail elsewhere (27). The general principle of this technique is illustrated in Fig. 2. Recombinant human Dsg 1-Fc or Dsg 3-Fc molecules were attached to the tip and plate of the AFM setup (Fig. 2,A). Coupling of Dsg molecules via flexible PEG linker allowed the molecules to freely diffuse within the radius of the length of the linker (∼8 nm) and to undergo unimpaired encounter reactions. During measurements, the tip was brought into contact to the plate of the setup by cyclic upward and downward movements at 2 Hz frequency (force-distance cycles, Fig. 2,B). During downward movement (approach), Dsg molecules bound to the tip and plate interacted leading to deflection of the cantilever in the following retrace movement. Specific single unbinding events were characterized by abrupt jumps of the cantilever toward the neutral position because of bond rupture (Fig. 2,C). Due to coupling of several Dsg molecules to the AFM tip, unbinding events typically appeared successively displaying two to five single unbinding events as shown in Fig. 2,D. The total area between approach and retrace curve was taken as a measure for binding activity as described previously (27). Approach-retrace cycles of 2000 nm/s and 0.1-s encounter time were performed in the presence or absence of Abs. As outlined above, PV- and PF-IgG, which were depleted from Dsg 1- and Dsg 3-specific autoantibodies by immunoabsorption against the rDsg 1 and rDsg 3 proteins, had no effect on keratinocyte cultures (Fig. 1, G and N). These data verify that the recombinant proteins, which we used in the following to study Dsg transinteraction, retained the correct conformation during purification procedure (31, 32). This was supported by the ability of AK 23 to immunoprecipitate rDsg 3 (data not shown). In addition, Ca2+ binding of rDsg 1 and rDsg 3 was confirmed using the fluorescent Ca2+ indicator quin-2 (Fig. 2 E).
To investigate whether rDsg 1-Fc and rDsg 3-Fc, in addition to homophilic, also undergo heterophilic transinteraction, we studied the transinteraction of Dsg 3-Fc coupled to the tip of the cantilever to Dsg 1-Fc molecules coupled to the plate (Dsg 3/1) and vice versa (Dsg 1/3) (Fig. 2 F). As negative controls, we used experiments where transinteraction of Dsg 1 and Dsg 3 to VE-cadherin was probed. When these values were subtracted and heterophilic-binding activities were normalized to the levels of homophilic Dsg 3 and Dsg 1 transinteraction, heterophilic transinteraction of these two molecules was negligible. Therefore, in the following experiments, we focused on the effect of pemphigus IgG on homophilic Dsg 1 and 3 interactions.
PV-IgG and AK 23 blocked Dsg 3 transinteraction in cell-free AFM experiments
Dsg 3 transinteraction has so far not been probed by AFM single-molecule experiments. As illustrated from the single unbinding events in Fig. 2,D, the resulting unit unbinding force of two transinteracting Dsg 3 molecules was in the range of 50 pN, which is comparable to the unbinding force of Dsg 1 (24) and other cadherins probed under same conditions (27). Moreover, we observed higher order unbinding events indicating additional mechanisms of interaction. Dsg 3 transinteractions were strongly Ca2+ dependent; Ca2+ depletion reduced Dsg 3-binding activity to 23 ± 3% (Fig. 3,A). Interestingly, for all Dsg 3 autoantibody-containing PV-IgG (PV-IgG 1, 3, and 4) a reduction of Dsg 3 transinteraction was detected to 31 ± 7%, 38 ± 11%, and 44 ± 10% of controls, respectively (Fig. 3,A). To investigate whether PV-IgG cross-linked Dsg 3 at the tip or the plate of the AFM setup and thereby prevented Dsg 3 transinteraction, PV-IgG Fab were used. However, PV-IgG Fab still reduced Dsg 3 transinteraction to 51 ± 7%. Together with the finding that PV-IgG depleted of Dsg 3 Abs (PV-IgG 3 Abs) did not reduce Dsg 3 transinteraction (110 ± 19%, Fig. 3,A), these data indicate that Dsg 3-specific autoantibodies in PV-IgG directly interfered with Dsg 3 transinteraction. Because in previous studies we showed that PF-IgG did not block Dsg 1 transinteraction (24), we further analyzed the effect of PF-IgG on Dsg 3 transinteraction. Not surprisingly, PF-IgG 1, which only included Dsg 1 autoantibodies, did not interfere with Dsg 3 transinteraction (97 ± 14%, Fig. 3 A).
Next, to investigate whether Dsg 3-specific PV Abs directly hinder Dsg 3 transinteraction, mouse monoclonal PV Abs AK 23, AK 18, and AK 9 were tested in Dsg 3 AFM experiments (Fig. 3,B). Consistent with the pathogenicity of AK 23 in the mouse model, only AK 23 (75 μg/ml) blocked Dsg 3 transinteraction in AFM experiments to a comparable extent like Ca2+ depletion (21 ± 9%). Higher concentrations of AK 23 up to 160 μg/ml did not yield significantly different results (22 ± 5%). AK 23 Fab were equally effective in blocking Dsg 3 transinteraction (15 ± 4%, Fig. 3 B), whereas AK 18 (86 ± 2%) and AK 9 (107 ± 1%) had no effect, even when applied at higher concentrations (data not shown). Taken together, these data demonstrate that both PV-IgG containing Dsg 3 autoantibodies as well as a monoclonal PV Ab binding to the N-terminal EC 1 domain of Dsg 3 directly interfered with Dsg 3 transinteraction.
Dsg 1 transinteraction was not blocked by pemphigus IgG in cell-free AFM experiments
Because PV-IgG 1 also contained Dsg 1 autoantibodies, we next examined the effect of PV-IgG on Dsg 1 transinteraction by AFM force measurements. As illustrated in Fig. 4 A, specific reduction of Dsg 1-binding activity could be achieved by Ca2+ depletion or addition of a mAb directed against the extracellular domain of Dsg 1 (aDsg 1). In these experiments, Dsg 1-binding activity was reduced to 20 ± 3% and 29 ± 7%, respectively. Surprisingly, although including Dsg 1 autoantibodies, treatment with PV-IgG 1 did not change Dsg 1-binding activity (95 ± 5%). As to be expected, PV-IgG 3, only including Dsg 3 autoantibodies, PV-IgG Fab and AK 23 also did not interfere with Dsg 1 transinteraction (94 ± 13%, 96 ± 9%, and 101 ± 3%, respectively). In line with our previous observation (24), Dsg 1-binding activity was not altered by treatment with PF-IgG 1 and 2 as well as PF-IgG Fab (84 ± 7%, 102 ± 14%, and 85 ± 5%, respectively).
The rDsg 1 used for this study was capable of depleting all pathogenic Dsg 1 Abs from PV- and PF-IgG fractions (Fig. 1, G and N), which is a strong indication that the correct conformation of Dsg 1 was retained (31). Nevertheless, this does not completely rule out the possibility that the pH shift during purification by protein A affinity chromatography induced minimal conformational changes of the Dsg 1 extracellular domain that inhibited binding of a small fraction of autoantibodies being capable of inducing direct inhibition of Dsg 1 transinteraction. Therefore, essentially all experiments were repeated using Dsg 1 purified via its his-tag by Ni-NTA columns and imidazole elution (Fig. 4 B). This Dsg 1 protein also proved to be cleaved by ETA (data not shown). ETA cleavage is known to be strictly dependent on the proper conformation of the Dsg 1 extracellular domain (32). In AFM experiments, Ni-NTA column-purified Dsg 1 displayed strong Ca2+ dependency of homophilic Dsg 1 transinteraction and was blocked by the monoclonal Dsg 1 Ab (aDsg 1) (24 ± 3% and 27 ± 6% of control binding, respectively), similar to what was detected using Dsg 1 purified by the protein A column. Nevertheless, neither PV-IgG 1 and 2 (both containing Dsg 1 and 3 autoantibodies) nor PV-IgG Fab reduced Dsg 1-binding activity (106 ± 4%, 93 ± 3%, and 95 ± 3%, respectively). Similar to our previous investigations (24), no direct inhibition of Dsg 1 transinteraction by PF-IgG was observed: neither PF-IgG 1 at normal or higher doses (concentrated four times), PF-IgG Fab, nor a pool of three additional PF-IgG fractions blocked Dsg 1 transinteraction (97 ± 4%, 102 ± 3%, 96 ± 6%, and 101 ± 7%, respectively). Taken together, the AFM studies revealed that PV-IgG and AK 23 selectively blocked Dsg 3 but not Dsg 1 transinteraction, whereas PF-IgG did not interfere with Dsg-mediated transinteraction in the cell-free system.
Pemphigus IgG caused loss of binding of Dsg 1- and Dsg 3-coated microbeads to the surface of cultured human keratinocytes
To study the role of PV- and PF-IgG in the presence of cellular signaling mechanisms, we used the laser tweezer technique (Fig. 5). For this purpose, microbeads coated with human Dsg 1 or Dsg 3 were allowed to settle on the surface of HaCaT cells for 30 min. Afterward, we counted the number of bound beads resisting the separating forces of the laser beam (Fig. 5,A). Fig. 5 B summarizes all laser tweezer experiments using pemphigus IgG. Under control conditions, 77 ± 5% of Dsg 3 and 82 ± 2% of Dsg 1 beads could not be displaced by the laser beam focus and were taken as tightly bound (100%). Following incubation with 5 mM EGTA for 30 min to deplete extracellular Ca2+, the number of bound Dsg 3- or Dsg 1-coated beads dropped to 22 ± 5% and 29 ± 6%, respectively, again documenting the strong Ca2+ dependency of Dsg adhesion. When HaCaT cells with surface-bound beads were incubated with PV- or PF-IgG fractions for 30 min, the number of both Dsg 3- and Dsg 1-coated beads was significantly reduced. PV-IgG 1, containing Dsg 1 and 3 autoantibodies, reduced the number of Dsg 3 and Dsg 1 beads to 52 ± 8% and 45 ± 4%, respectively. PV-IgG 4, including Dsg 3 autoantibodies alone, reduced both Dsg 3- and Dsg 1-mediated binding as well (83 ± 6% and 71 ± 14%, respectively). Ab-mediated cross-linking was not required for loss of Dsg 3 binding because PV-IgG Fab also blocked Dsg 3 and Dsg 1 binding (48 ± 3% and 62 ± 5%, respectively). PV-IgG depleted of Dsg 1 and 3 Abs did not reduce the number of bound Dsg 3- or Dsg 1-coated beads (97 ± 3% and 101 ± 2%, respectively) indicating that autoantibodies specific for Dsg 1 and 3 mediate blocking of Dsg transinteraction in this assay.
In addition, following incubation with PF-IgG 1, the number of bound Dsg 3 and Dsg 1 beads dropped to 31 ± 8% and 50 ± 7%, respectively. PF-IgG Fab had similar effects and reduced Dsg 3- and Dsg 1-mediated binding to 53 ± 7% and 66 ±3%, respectively, whereas Dsg 1 autoantibody-depleted PF-IgG had no effects on Dsg 3 and Dsg 1 binding (101 ± 4% and 104 ± 2%, respectively). Thus, PF-IgG not containing Dsg 3 autoantibodies and ineffective to block transinteraction of Dsg 1 and Dsg 3 in cell-free AFM experiments, as well as PV-IgG interfering with Dsg 3 but not with Dsg 1 transinteraction, both were effective at inhibiting binding of Dsg 1- and Dsg 3-coated beads to cultured keratinocytes via Dsg-specific autoantibodies.
We also examined the effect of mouse monoclonal PV Abs on binding of Dsg 3- and Dsg 1-coated beads to HaCaT cells (Fig. 5 C). In contrast to pemphigus patients’ IgG, monoclonal AK 23 Ab selectively reduced the number of bound Dsg 3 beads to 46 ± 5% of control, whereas the number of bound Dsg 1 beads was not affected (101 ± 3%). Moreover, AK 23 Fab also reduced Dsg 3-mediated binding to 62 ± 5%. In contrast, AK 18 and AK 9 did not alter the number of bound Dsg 3-coated beads on HaCaT cells (99 ± 2% and 100 ± 1%, respectively).
In other studies it has been shown that presence of Dsg 1 prevents PV-IgG from disrupting intercellular adhesion (33). Similarly, PF-IgG might require the presence of coexpressed Dsg 3 to disrupt homophilic transinteraction of Dsg 1. Therefore, we tested this hypothesis using AFM and laser tweezer experiments (Fig. 6). As a first step, AFM tips were either coated with Dsg 1 alone (Fig. 6,A, ▪) or in combination with Dsg 3 (Fig. 6,A, □) and probed on substrates covered with an equal mixture of Dsg 1 and Dsg 3. PV-IgG 1, though containing Dsg 1 autoantibodies, did not alter Dsg 1 transinteraction when tips were coated with Dsg 1 alone (109 ± 26%) but was efficient at reducing binding activity when Dsg 3 was also present at AFM tips (reduction to 31 ± 8% of controls). In contrast, PF-IgG 1 had no effects under the two experimental conditions (110 ± 21% and 100 ± 8%, respectively). This indicated that the presence of Dsg 3 did not alter PV- or PF-IgG-induced effects on cell-free Dsg 1 transinteraction. We further tested the effects of PF-IgG on Dsg 3 bead binding on keratinocytes under conditions where Dsg 1 was absent. Western blotting (Fig. 6,B) showed that Dsg 1 was not present when HaCaT cells were cultivated for 1 day after passaging but was detected after 7 days – the condition used in previous experiments (Fig. 5). In laser tweezer experiments with HaCaT cells cultivated for 1 day, PV-IgG 1 reduced the number of bound Dsg 3 beads to 62 ± 6% of controls (Fig. 6,C). However, PF-IgG 1 and 2 had no effects (Dsg 3 bead binding 93 ± 5% and 99 ± 3%, respectively), in contrast to experiments with HaCaT cells cultured for 7 days and expressing Dsg 1 (Fig. 5). These data demonstrated that Dsg 1 is required to mediate the effects of PF-IgG on Dsg 3 binding in keratinocytes.
We provide first evidence that Dsg 3 autoantibodies in PV directly inhibit Dsg 3 transinteraction, whereas in contrast Dsg 1 autoantibodies in PV- and PF-IgG reduce Dsg 1 transinteraction not directly but rather indirectly via cellular mechanisms. Using a combined approach of cell-free AFM studies, together with laser tweezer trapping of Dsg-coated microbeads on the surface of human keratinocytes, these conclusions are based on the following observations (Fig. 7): 1) PV-IgG containing Abs to both Dsg 3 and Dsg 1 as well as PV-IgG Fab selectively blocked Dsg 3 but not Dsg 1 transinteraction in single-molecule AFM studies. 2) PF-IgG and PF-IgG Fab did neither interfere with Dsg 1 nor with Dsg 3 transinteraction in AFM studies. 3) PV-IgG and PV-IgG Fab reduced binding of both Dsg 3- and Dsg 1-coated beads to the keratinocyte cell surface. 4) PF-IgG containing Dsg 1 but not Dsg 3 Abs as well as PF-IgG Fab also reduced binding of Dsg 3- and Dsg 1-coated beads. 5) The mouse monoclonal PV Ab AK 23 directed against the putative Dsg 3-binding site and AK 23 Fab specifically blocked Dsg 3 transinteraction in the presence and absence of cells. 6) mAbs against the middle and C-terminal parts of the extracellular Dsg 3 domain did not interfere with Dsg 3 transinteraction.
In a previous study, we showed that PF-IgG reduced transinteraction of Dsg 1 molecules only when assayed in keratinocyte cultures but not in the cell-free AFM setup (24). In the recent study, we confirmed these results and found in addition that autoantibodies to Dsg 1 in PV-IgG also did not directly interfere with Dsg 1 transinteraction. These negative findings seem not to be caused by a loss of proper conformation of Dsg 1, because Dsg 1 was able to bind Ca2+, to deplete all pathogenic Abs from PF-IgG (31) and also was cleaved by ETA (32). However, we cannot completely rule out that in vivo conformational changes of the Dsg 1 structure occur in response to autoantibody binding which are not equally present in our experiments. Nevertheless, a commercial mouse mAb directed against Dsg 1 blocked Dsg 1 transinteraction demonstrating that this EC 2-directed Ab (manufacturer’s specifications and our observations) may cause allosteric conformational changes which impair Dsg 1 adhesion. In clear contrast, we found that PV-IgG and PV-IgG Fab inhibited Dsg 3 transinteraction both in AFM studies and on keratinocytes. The use of Fab allows the conclusion that the loss of binding activity was not due to Ab cross-linking of molecules on the cantilever or the plate of the AFM. Abs to Dsg 3 in PV-IgG were equally effective at inhibiting Dsg 3 transinteraction in AFM studies like the monoclonal Dsg 3 Ab AK 23, which binds to the N-terminal EC 1 domain of Dsg 3 where the predicted binding interface is located and which is targeted by most of Dsg 3 Abs in PV patients (23). These data indicate that direct inhibition of Dsg 3 transinteraction was most likely caused by steric hindrance (34). However, because ∼20% of PV Abs have been shown to bind to other parts of the Dsg 3 extracellular domain (23), at this stage we cannot completely rule out the possibility that some autoantibodies in PV-IgG fractions also interfered with Dsg 3 transinteraction by allosteric effects. Nevertheless, we found that AK 9 and AK 18, which target different parts of the extracellular domain, had no effect on Dsg 3 transinteraction. Finally, depletion of Dsg 3-specific Abs by immunoabsorption using rDsg 3 demonstrated that inhibition of Dsg 3 transinteraction was mediated by Dsg 3-specific Abs. Together with the findings that depletion of Dsg-specific Abs from PV-IgG and PF-IgG completely abolished keratinocyte dissociation and loss of Dsg 3 and Dsg 1 bead binding, these data further support the hypothesis that Dsg-specific Abs are required for pemphigus pathogenesis (13). This is also supported by the observation that, in the absence of Dsg 1 in HaCaT cells, PF-IgG-mediated loss of Dsg 3 bead binding was abolished indicating that Dsg 1 is the major autoantigen required for autoantibody-mediated outside-in signaling.
Because heterophilic transinteraction of desmosomal cadherins is thought to be important (35, 36), we also sought to address this issue in our study. However, because we did not observe heterophilic interactions of Dsg 3 and Dsg 1, we were only in the position to probe direct inhibition of homophilic transinteraction. Nevertheless, we can rule out the possibility that the presence of both Dsg 1 and Dsg 3 is needed for PV- or PF-IgG to alter Dsg transinteraction in cell-free single-molecule experiments. It has to be emphasized that this does not exclude the possibility that PV- or PF-IgG also directly interfere with the transinteraction of Dsg 3 or Dsg 1 with other desmosomal cadherins such as desmocollin (Dsc) 1 or Dsc 3, especially because it has been recently reported that conditional Dsc 3 deficiency in a mouse model led to a severe pemphigus-like phenotype (37). This needs to be clarified in further studies.
In the meantime, several signaling pathways have been shown to participate in pemphigus pathogenesis including protein kinase C, plakoglobin, c-Myc, p38MAPK, Rho A, epidermal growth factor receptor, and Src (29, 38, 39, 40, 41, 42, 43). However, except for p38MAPK and Rho A (28, 29, 38, 44), all signaling mechanisms have only been evaluated for their contribution to PV but not to PF pathogenesis. Our results suggest that the pathogenesis of these two pemphigus subtypes may be different because direct inhibition of Dsg-mediated transadhesion may only occur in PV (43). The reduction of Dsg 1 binding to the surface of keratinocytes in response to PV- and PF-IgG which was detected after a 30-min incubation in our laser tweezer studies may be caused by cellular signaling events. This explanation is possible because cellular signaling events have been shown to be triggered by pemphigus IgG within 20 s (45). Moreover, activation of p38MAPK which is now thought to be one of the key signaling mechanisms in pemphigus pathogenesis occurs within the first 30 min (46).
However, it cannot be concluded from our data whether in addition to direct inhibition cellular signaling events were also responsible for PV-IgG-induced loss of Dsg 3 transinteraction in keratinocytes. Because AK 23, which inhibited Dsg 3 transinteraction in AFM experiments, was only effective in reducing binding of Dsg 3- but not of Dsg 1-coated beads in the laser tweezer assay, these data indicate that direct inhibition of Dsg 3 transinteraction is sufficient to reduce binding of Dsg 3 at the keratinocyte surface. However, unlike PV-IgG, AK 23 did not induce a profound disruption of Dsg 3 localization and keratinocyte dissociation in our experiments. Although AK 23 has been shown to effectively induce Dsg 3 depletion in human keratinocyte cultures (30), the effect of AK 23 on cellular signaling events such as p38MAPK signaling was found to be rather weak compared with patients’ PV-IgG (47, 48). Therefore, it is likely that AK 23 does not trigger the signaling mechanisms required for reduction of Dsg 1 adhesion and keratinocyte dissociation. This supports the fact that a single mAb does not fully reproduce the effects of polyclonal PV-IgG (30).
Taken together, our data are in line with the hypothesis that direct inhibition of Dsg 3 transinteraction or activation of cell-signaling mechanisms may not be mutually exclusive but rather act in concert in pemphigus pathogenesis (42, 43, 49). Moreover, it is unlikely that direct inhibition of Dsg 3 transinteraction alone accounts for desmosomal splitting and acantholysis in PV. This can be concluded from the finding that complete deficiency of Dsg 3 still allows formation of morphologically intact desmosomes, at least in the absence of mechanical stress (50). For the future, it will be important to determine whether direct inhibition and signaling are completely independent mechanisms or whether cell-signaling events in PV pathogenesis occur in response to direct Ab-mediated loss of cell adhesion (34). Our study suggests that direct inhibition of Dsg 3 transinteraction in PV in addition to cellular signaling events may aggravate the clinical phenotype of PV (43).
We are grateful to Masayuki Amagai (School of Medicine, Keio University, Tokyo, Japan) for contribution of ETA as well as for helpful advice and discussion. We thank Hermann Gruber (Johannes Kepler University, Linz, Austria) for generously providing AFM PEG linkers. We thank Stefanie Imhof, Nadja Niedermeier, and Lisa Bergauer for excellent technical assistance.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB 487, TP B5) and the Interdisziplinäres Zentrum für Klinische Forschung Würzburg (TP A-51).
Abbreviations used in this paper: Dsg, desmoglein; PF, pemphigus foliaceus; PV, pemphigus vulgaris; EC, extracellular domain; AFM, atomic force microscopy; ETA, exfoliative toxin A; RT, room temperature; PEG, polyethylene glycol; Dsc, desmocollin; VE, vascular endothelial.