Pemphigus Vulgaris and Pemphigus Foliaceus Antibodies arePathogenic in Plasminogen Activator Knockout Mice

My G. Mahoney,1 Zhi Hong Wang, and John R. StanleyDepartment of Dermatology, University of Pennsylvania, Philadelphia, Pennsylvania, U.S.A.

Previous studies have suggested that urokinaseplasminogen activator is required for blister formationin pemphigus vulgaris and pemphigus foliaceus.Other studies, however, have shown that downregul-ation of plasminogen activator does not inhibitblisters induced by pemphigus immunoglobulin G.To eliminate the possibility that small amounts ofurokinase plasminogen activator might be sufficientfor blister formation, we passively transferred pem-phigus immunoglobulin G to urokinase plasminogenactivator knockout neonatal mice. Pemphigus folia-ceus and pemphigus vulgaris immunoglobulin Gcaused gross blisters and acantholysis in the superficialand suprabasal epidermis, respectively, to the samedegree in knockout and control mice, demonstratingthat urokinase plasminogen activator is not absolutelyrequired for antibody-induced blisters. Some studieshave shown elevated tissue-type plasminogen activ-ator in pemphigus lesions. Tissue-type plasminogen

In the two major forms of pemphigus, pemphigus vulgaris(PV) and pemphigus foliaceus (PF), autoantibodies againstdesmogleins cause loss of keratinocyte cell adhesion (alsotermed acantholysis) and blister formation (Stanley, 1989,1990). The autoantibodies from PF patients are directed

against desmoglein (Dsg) 1 and cause blisters in the superficialliving epidermis, whereas the autoantibodies in PV are directedagainst Dsg3 (with or without coexisting anti-Dsg1 antibodies) andcause blisters in the deep epidermis, just above the basal layer (i.e.,suprabasilar acantholysis) (Buxton et al, 1993; Stanley, 1993).

Desmogleins are transmembrane glycoproteins of desmosomes,which are cell–cell adhesion structures. Therefore, when it wasdiscovered that pemphigus autoantibodies bind desmogleins, it waspostulated that these antibodies might directly interfere with anadhesion function of these desmosomal molecules (Koulu et al,1984; Amagai et al, 1991).

Manuscript received January 21, 1999; revised March 17, 1999; acceptedfor publication March 25, 1999.

Reprint requests to: Dr. John R. Stanley, Department of Dermatology,University of Pennsylvania, 211 CRB, 415 Curie Blvd., Philadelphia, PA19104, U.S.A.

1Current address: M.G. Mahoney, Department of Dermatology andCutaneous Biology, Thomas Jefferson University, Philadelphia, PA19107, U.S.A.

Abbreviations: PA, plasminogen activator; PF, pemphigus foliaceus; PV,pemphigus vulgaris; tPA, tissue-type plasminogen activator; uPA, urokinase-type plasminogen activator.

0022-202X/99/$14.00 · Copyright © 1999 by The Society for Investigative Dermatology, Inc.

22

activator, however, is not necessary for blisterformation, because pemphigus foliaceus and pem-phigus vulgaris immunoglobulin G caused blistersto the same degree in tissue-type plasminogenactivator knockout and control mice. To rule outthat one plasminogen activator might compensatefor the other in the knockout mice, we bredurokinase plasminogen activator, tissue-type plasmin-ogen activator double knockouts. After passivetransfer of pemphigus foliaceus and pemphigusvulgaris immunoglobulin G these mice blistered tothe same degree as the single knockout and controlmice, and histology indicated blisters at the expectedlevel of the epidermis. These data definitively demon-strate that plasminogen activator is not necessaryfor pemphigus immunoglobulin G to induce acantho-lysis in the neonatal mouse model of pemphigus.Key words: cell adhesion/desmoglein/desmosome/protease.J Invest Dermatol 113:22–25, 1999

There were also studies, however, that suggested that blisterformation in pemphigus is indirectly mediated by release of proteasessecondary to antibody binding (Schiltz et al, 1979; Farb et al,1978). Furthermore, under some experimental conditions proteaseinhibitors were able to inhibit acantholysis induced by PV antibodiesadded to human epidermal culture and passively transferred toneonatal mice (Farb et al, 1978; Naito et al, 1989). Most studiesthat have attempted to identify the specific proteases involved inthe pathophysiology of pemphigus have implicated plasminogenactivator (PA). PF and PV immunoglobulin (Ig) G added to humankeratinocyte cell culture caused PA release, mostly of the urokinasetype (uPA) (Hashimoto et al, 1983). These same authors showedthat plasminogen (the substrate for PA) enhanced the ability ofpemphigus IgG to cause acantholysis in organ culture (Hashimotoet al, 1983). Further studies of human skin organ culture demon-strated that a rabbit anti-uPA antibody and PA inhibitor 2 (a potentinhibitor of uPA) blocked acantholysis induced by PV and PF IgG(Morioka et al, 1987; Hashimoto et al, 1989). Finally, recent studiespoint to a potential mechanism of antibody-mediated PA releasefrom keratinocytes through the activation of protein kinase C(Esaki et al, 1995).

In spite of these studies there are problems with implicating PAin pemphigus blister formation. Studies showing PA release byhuman keratinocytes are mostly based on cell culture, whereasacantholysis and inhibition of acantholysis is mostly demonstratedin organ culture. Although uPA has been implicated in cell culture,tissue type PA (tPA) increases have been found in lesional skin ofpemphigus (Jensen, 1988; Baird, 1990). Furthermore, PA is elevated

VOL. 113, NO. 1 JULY 1999 PLASMINOGEN ACTIVATOR IN PEMPHIGUS 23

in the lesional skin of many other skin diseases that are notcharacterized by acantholysis (Jensen, 1988; Baird, 1990; Spierset al, 1994, 1996). Finally, in the neonatal mouse model ofpemphigus, dexamethasone markedly suppressed (but did notcompletely eliminate) PA activity induced by pemphigus IgG, butdid not affect blister formation (Anhalt et al, 1986).

Therefore, although some studies suggest that PA, particularlyuPA, may be critical to blister formation in pemphigus, otherstudies do not support this hypothesis. In this study we attempt toanswer, definitively, the question of whether PA expression isnecessary for PF and PV antibodies to induce the characteristicpathology in these diseases by using passive transfer of pemphigusIgG to neonatal mice with targeted disruptions in the PA genes.

MATERIALS AND METHODS

PA knockout mice We purchased mice with a targeted mutation inuPA and tPA, as reported by Carmeliet et al (1994), and bred back to aC57Bl/6 background (Jackson Laboratories, Bar Harbor, ME). These micehad the following genotypes: uPA–/–tPA1/1 and uPA1/1tPA–/–. Inaddition to crossing mice of identical genotypes to propagate uPA–/– andtPA–/– mice, we intercrossed uPA–/– tPA1/1 3 uPA1/1tPA–/– thenbackcrossed to obtain the following breeding pairs: (i) tPA–/–uPA1/–3 tPA1/– uPA–/–; (ii) tPA1/–uPA1/– 3 tPA–/–uPA1/–; and (iii)

tPA1/–uPA1/– 3 tPA1/–uPA–/–. These breedings supplied tPA–/– anduPA–/– mice as well as tPA–/–uPA–/– double knockouts.

Genotyping of PA knockout mice Genomic DNA was extractedfrom the tails of mice and processed for EcoRI digestion and Southernblotting, as previously reported (Koch et al, 1997). A 1 kb Hind III–EcoRIprobe from the 39 flanking region of the tPA gene and an 0.24 kb AccI–HindIII probe from the 39 flanking region of the uPA gene were generouslyprovided by Peter Carmeliet, and are referred to as tPA probe ‘‘d’’ anduPA probe ‘‘d’’ as described (Carmeliet et al, 1994). These probes wereused as templates with random primers to synthesize 32P-labelled probesfor use on Southern blots, as previously reported (Koch et al, 1997).

Neonatal mouse model of pemphigus Neonatal mice with varyinguPA and tPA genotypes were injected with IgG purified from PV and PFsera, as originally described by Anhalt et al (1982). Litters of 1–2 d oldneonates were injected with IgG. After µ16 h the animals were evaluatedfor gross blisters, then killed for histology and to harvest DNA forgenotyping. The IgG from two PF sera and two PV sera were used; IgGfrom both PF sera gave identical results as did IgG from both PV sera.Specific procedures for preparation of IgG, injection into neonatal mice,and evaluation of mice were essentially as recently described (Mahoneyet al, 1999). Mice were evaluated grossly, before the genotype wasdetermined, for spontaneous blister formation. Histology was gradedsemiquantitatively, without knowing the genotype, on the followingscale: –, no blister; 11, blister at edge only; 21, localized blisters , 50%of specimen; 31, extensive blisters . 50% of specimen; 41, very extensiveblisters . 75% of specimen.

RESULTS

Determination of pemphigus IgG dose levels to induceextensive and more limited blisters in neonatal mice Usingnormal C57Bl/6 mice as well as control littermates of uPA andtPA knockout mice with at least one wild-type allele for both PA,we determined approximate doses of PF and PV IgG that causedextensive and limited blistering. These doses averaged 4.4 mg perg (‘‘high dose’’) and 1.1 mg per g (‘‘low dose’’) for PF IgG, and8 mg per g (‘‘high dose’’) and 3 mg per g (‘‘low dose’’) for PVIgG. For the IgG derived from pemphigus sera in this study, as inour previous studies, PF IgG was consistently more effective atinducing blisters as compared with PV IgG. IgG from normalhuman sera did not induce gross or microscopic blisters at any dose.

uPA is not necessary for induction of blisters by pemphigusIgG Previous studies, discussed above, have implicated uPA asnecessary for blister formation in pemphigus. To determine if uPAis absolutely necessary for, or modulates, blister formation in theneonatal mouse model of pemphigus, we bred uPA–/– mice.Southern blots of these mice and littermates with uPA probe d (seeMaterials and Methods) on EcoRI restricted genomic DNA results

Figure 1. Southern blots of EcoRI restricted mouse tail DNA toidentify uPA–/–, tPA–/–, and uPA–/–tPA–/– mice. (A) uPA probed identifies a 5.7 kb wild-type allele (open arrowhead) and a 10 kb mutantallele (closed arrowhead). M1 shows DNA from a uPA–/– mouse; lane Cshows DNA from a control uPA1/– mouse used to show both alleles. (B)tPA probe d labels a 5.5 kb wild-type allele (open arrowhead) and a 9.7 kbmutant allele (closed arrowhead) in DNA from a control tPA1/– mouse(lane C). M2 shows DNA from a tPA–/– mouse. (C) Lanes labeled M3show DNA from a double knockout mouse with only mutant (closedarrowhead) uPA and tPA alleles, but no wild-type alleles (open arrowheads).

Figure 2. PA null neonatal mice injected with PF and PV IgGdevelop blisters. (A) uPA–/– mouse injected with PF IgG 0.7 mg per g.(B) uPA–/– mouse injected with PV IgG 10 mg per g. (C) uPA–/–tPA–/–mouse injected with PF IgG 0.6 mg per g. (D) uPA–/–tPA–/– mouseinjected with PV IgG 3 mg per g. Arrow indicates flaccid blister.

in a 5.7 kb fragment from the wild-type allele and a 10 kb fragmentfrom the recombinant (knockout) allele in these mice (Fig 1A).Passive transfer of PF IgG and PV IgG at both high and low dosesresulted in gross and histologic blisters in these mice, to the sameextent as uPA1/– and uPA1/1 mice (Figs 2A, B, 3A, B, and 4)Furthermore, the histology of lesional skin was typical for PF andPV; that is there was acantholysis in the granular layer with PF IgGinjection (Fig 3A) and there was suprabasilar acantholysis with PVIgG injection (Fig 3B).

tPA is not necessary for blister formation in pemphigus anddoes not compensate for lack of uPA Because some studieshave demonstrated increased tPA, rather than uPA, in pemphiguslesions, we passively transferred PF and PV IgG to tPA–/– neonates.These mice were genotyped by southern blotting with tPA probed which identified a 9.7 kb mutant allele and a 5.5 kb wild-typeallele (Fig 1B). tPA null mice showed blisters that were histologicallytypical of PF and PV when their respective immunoglobulins wereinjected (Fig 3C, D). Although there may have been a slighttendency for less extensive blistering in tPA–/– neonates, there wassignificant overlap in the extent of blistering with control mice(Fig 4). Because the measurement of blistering was semiquantitativeat best, we did not perform a statistical evaluation, but we can

24 MAHONEY ET AL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

Figure 3. Epidermal histology of PAnull mice injected with pemphigus IgGshow superficial and suprabasilaracantholysis from PF and PV IgG,respectively. (A) uPA–/– mouse injectedwith PF IgG 0.7 mg per g. (B) uPA–/–mouse injected with PV IgG 3 mg per g.(C) tPA–/– mouse injected with PF IgG1 mg per g. (D) tPA–/– mouse injected withPV IgG 3 mg per g. (E) uPA–/– tPA–/–mouse injected with PF IgG 0.6 mg per g.(F) uPA–/–tPA–/– mouse injected with PVIgG 3 mg per g.

conclude clearly that tPA is not necessary for the typical blisteringof PV or PF.

Finally, we wanted to rule out the possibility that in mice nullfor one of the PA the other does not compensate, allowing blisterformation (i.e., redundancy in null animals). We therefore breduPA–/–tPA–/– null mice (Fig 1C). In these PA null mice, PV andPF IgG caused blistering to the same degree as in control anduPA–/– or tPA–/– mice (Figs 2C, D, 3E, F, and 4). The histologyof the blisters in these mice was indistinguishable from control orsingle PA knockout mice (Fig 3).

DISCUSSION

These data indicate that, in the neonatal mouse model of pemphigus,PA is not necessary for blister formation and does not, in a majorway, modulate the extent or localization of blister formation. Thesestudies provide further evidence that pemphigus autoantibodiescause loss of cell adhesion directly by interference with desmogleinfunction rather than indirectly by release of proteases.

These results support several lines of evidence that point to adirect effect of pemphigus antibodies on desmoglein function. Therecently described mouse with a null mutation in Dsg3 has aphenotype very similar to PV patients with suprabasilar blisters inoral mucous membranes and traumatized skin (Koch et al, 1997).This result demonstrates that the loss of function of Dsg3 resultsin a similar loss of adhesion of keratinocytes as do anti-Dsg3antibodies, suggesting that these antibodies cause loss of function.Further evidence for a direct effect of pemphigus antibodies ondesmoglein function and against the idea of protease release as asignificant factor in blister pathogenesis comes from recent datasupporting ‘‘the desmoglein compensation hypothesis’’ as anexplanation for the localization of blisters in PF and PV (Mahoneyet al, 1999). This hypothesis states that where Dsg3 and Dsg1expression overlap in the epidermis antibodies against either onewill not cause spontaneous blisters, but antibodies against both willdo so. On the other hand where Dsg3 or Dsg1 is expressed alone,antibodies against that desmoglein will cause a blister. These data

Figure 4. Similar extent of histologic blistering in control and PAmutant mice. High dose: mean for PF IgG 4 mg per g; for PV 8 mg perg. Low dose: mean for PF IgG 1 mg per g; for PV 3 mg per g. 1, At leastone wild-type allele each for uPA and tPA. u-, uPA–/– with at least onewild-type allele for tPA. t-, tPA–/– with at least one wild-type allele foruPA. –/–, no PA wild-type alleles. Extent of blistering graded by asemiquantitative scale as described in Materials and Methods.

are consistent with the idea that antibodies in pemphigus interferewith desmoglein function, and are not consistent with the ideathat antibodies that bind to a desmoglein release a protease whichcauses the blister (because in the later case the second desmogleinwould not be expected to compensate).

Although formally we have not ruled out that PA play a majorpart in human, as opposed to mouse, pemphigus, we think it highlyunlikely that pemphigus autoantibodies would cause identicalhistologic lesions in mice and humans, but by different mechanisms.Although it is possible that PA and perhaps other proteases modulatethe extent of blister formation in pemphigus patients, the weightof the current evidence suggests that the major pathophysiologic

VOL. 113, NO. 1 JULY 1999 PLASMINOGEN ACTIVATOR IN PEMPHIGUS 25

mechanism in pemphigus is a direct effect of autoantibodies on thefunction of desmogleins and the desmosomes in which theyare found.

We thank Dr. Peter Carmeliet for providing the uPA and tPA DNA probes. Thiswork was supported by grants from the NIAMS of the NIH.

REFERENCESAmagai M, Klaus-Kovtun V, Stanley JR: Autoantibodies against a novel epithelial

cadherin in pemphigus vulgaris, a disease of cell adhesion. Cell 67:869–877, 1991Anhalt GJ, Labib RS, Voorhees JJ, Beals TF, Diaz LA: Induction of pemphigus in

neonatal mice by passive transfer of IgG from patients with the disease. N EnglJ Med 306:1189–1196, 1982

Anhalt GJ, Patel HP, Labib RS, Diaz LA, Proud D: Dexamethasone inhibitsplasminogen activator activity in experimental pemphigus in vivo but does notblock acantholysis. J Immunol 136:113–117, 1986

Baird J: mRNA for tissue-type plasminogen activator is present in lesional epidermisfrom patients with psoriasis, pemphigus, or bullous pemphigoid, but is notdetected in normal epidermis. J Invest Dermatol 95:548–552, 1990

Buxton RS, Cowin P, Franke WW, et al: Nomenclature of the desmosomal cadherins.J Cell Biol 121:481–483, 1993

Carmeliet P, Schoonjans L, Kieckens L, et al: Physiological consequences of loss ofplasminogen activator gene function in mice. Nature 368:419–424, 1994

Esaki C, Seishima M, Yamada T, Osada K, Kitajima Y: Pharmacologic evidence forinvolvement of phospholipase C in pemphigus IgG-induced inositol 1,4,5-trisphosphate generation, intracellular calcium increase, and plasminogenactivator secretion in DJM-1 cells, a squamous cell carcinoma line. J InvestDermatol 105:329–333, 1995

Farb RM, Dykes R, Lazarus GS: Anti-epidermal-cell-surface pemphigus antibodydetaches viable epidermal cells from culture plates by activation of proteinase.Proc Natl Acad Sci USA 75:459–463, 1978

Hashimoto K, Shafran KM, Webber PS, Lazarus GS, Singer KH: Anti-cell surface

pemphigus autoantibody stimulates plasminogen activator activity of humanepidermal cells. J Exp Med 157:259–272, 1983

Hashimoto K, Wun TC, Baird J, Lazarus GS, Jensen PJ: Characterization ofkeratinocyte plasminogen activator inhibitors and demonstration of theprevention of pemphigus IgG-induced acantholysis by a purified plasminogenactivator inhibitor. J Invest Dermatol 92:310–314, 1989

Jensen PJ: Epidermal plasminogen activator is abnormal in cutaneous lesions. J InvestDermatol 90:777–782, 1988

Koch PJ, Mahoney MG, Ishikawa H, et al: Targeted disruption of the pemphigusvulgaris antigen (desmoglein 3) gene in mice causes loss of keratinocyte celladhesion with a phenotype similar to pemphigus vulgaris. J Cell Biol 137:1091–1102, 1997

Koulu L, Kusumi A, Steinberg MS, Klaus Kovtun V, Stanley JR: Humanautoantibodies against a desmosomal core protein in pemphigus foliaceus. J ExpMed 160:1509–1518, 1984

Mahoney MG, Wang Z, Rothenberger K, Koch PJ, Amagai M, Stanley JR:Explanation for the clinical and microscopic localization of lesions in pemphigusfoliaceus and vulgaris. J Clin Invest 103:461–468, 1999

Morioka S, Lazarus GS, Jensen PJ: Involvement of urokinase-type plasminogenactivator in acantholysis induced by pemphigus IgG. J Invest Dermatol 89:474–477, 1987

Naito K, Morioka S, Nakajima S, Ogawa H: Proteinase inhibitors block formationof pemphigus acantholysis in experimental models of neonatal mice and skinexplants: effects of synthetic and plasma proteinase inhibitors on pemphigusacantholysis. J Invest Dermatol 93:173–177, 1989

Schiltz JR, Michel B, Papay R: Appearance of ‘‘pemphigus acantholysis factor’’ inhuman skin cultured with pemphigus antibody. J Invest Dermatol 73:575–581, 1979

Spiers EM, Lazarus GS, Lyons-Giordano B: Expression of plasminogen activatorenzymes in psoriatic epidermis. J Invest Dermatol 102:333–338, 1994

Spiers EM, Lazarus GS, Lyons-Giordano B: Expression of plasminogen activators inbasal cell carcinoma. J Pathol 178:290–296, 1996

Stanley JR: Pemphigus and pemphigoid as paradigms of organ-specific, autoantibody-mediated diseases. J Clin Invest 83:1443–1448, 1989

Stanley JR: Pemphigus: skin failure mediated by autoantibodies. JAMA 264:1714–1717, 1990

Stanley JR: Cell adhesion molecules as targets of autoantibodies in pemphigus andpemphigoid, bullous diseases due to defective epidermal cell adhesion. AdvImmunol 53:291–325, 1993