|Home | About | Journals | Submit | Contact Us | Français|
Eosinophil granule proteins are deposited in cutaneous lesions in many human diseases, but how these proteins contribute to pathophysiology is obscure. We injected eosinophil cationic protein (ECP or RNase 3), eosinophil-derived neurotoxin (EDN or RNase 2), eosinophil peroxidase (EPO), and major basic protein-1 (MBP1) intradermally into guinea pig and rabbit skin. ECP and EDN each induced distinct skin lesions at ≥ 2.5 μM that began at 2 days, peaking about 7 days and persisting up to 6 weeks. These lesions were ulcerated (ECP) or crusted (EDN) with marked cellular infiltration. EPO and MBP1, 10 μM, each produced perceptible induration and erythema with moderate cellular infiltration resolving within 2 weeks. ECP and EDN localized to dermal cells within 2 days whereas EPO and MBP1 remained extracellular. Overall, cellular localization and RNase activity of ECP and EDN were critical for lesion formation; differential glycosylation, net cationic charge, or RNase activity alone did not account for lesion formation. Ulcerated lesions from patients with the hypereosinophilic syndrome showed ECP and EDN deposition comparable to that in guinea pig skin. In conclusion, ECP and EDN disrupt skin integrity and cause inflammation. Their presence in ulcerative skin lesions may explain certain findings in human eosinophil-associated diseases.
Secondary eosinophil granules contain several highly cationic proteins, including eosinophil cationic protein (ECP, RNase 3),2 eosinophil-derived neurotoxin (EDN, RNase 2), eosinophil peroxidase (EPO), and major basic protein-1 (MBP1) (1). EDN, EPO, and MBP1 are among the most abundantly transcribed genes in developing eosinophils (2) and are correspondingly abundant in isolated peripheral blood eosinophils (3). These granule proteins are extensively deposited in skin in several dermatoses, such as atopic dermatitis, bullous pemphigoid, and urticaria, and in the hypereosinophilic syndrome (HES) (4). In atopic dermatitis, eosinophil granule proteins are deposited at relatively high local concentrations (> 1 μM) (5) without apparent cutaneous eosinophilia, evidently as a result of cytolytic degranulation (6–8).
Numerous in vitro studies have reported cytotoxic properties of eosinophil granule proteins against mammalian cells, as well as pathogenic organisms, such as helminthes, bacteria, and viruses (1, 9, 10). Part of this cytotoxicity appears attributable to the cationicity of the granule proteins (pIs approximately 11 for ECP, EPO, and MBP1 and a pI of 8.7 for EDN) (2, 11, 12). In addition, ECP and EDN possess RNase activity, and EPO has peroxidase activity, whereas MBP1 is a C-type lectin (13). ECP and EDN are unique among the granule proteins in their ability to induce neurotoxicity, referred to as the Gordon phenomenon (14, 15), and their RNase activity appears critical for this ability (15). MBP1 activates several human cell types, such as mast cells, basophils, and eosinophils (16–18). Thus, eosinophil granule proteins have broad potential to impact cellular and tissue functions.
Few studies have examined the in vivo effects of eosinophil granule proteins in skin. They induce cutaneous wheal and flare reactions (19, 20) and increase cutaneous vasopermeability (5). Overall, the eosinophil’s importance in the pathophysiology of atopic disease, including atopic dermatitis, is controversial (21–24). Nonetheless, in diseases with extensive cutaneous eosinophil granule protein deposition, such as in HES and severe atopic dermatitis, cutaneous lesions are common. To define the in vivo effects of eosinophil granule protein deposition, we injected them intradermally into guinea pig skin and monitored cutaneous lesion formation. Subsequently, we examined the mechanism(s) of lesion formation, particularly with ECP and EDN, and compared immunohistologic staining of eosinophil granule proteins to that observed in erosive and ulcerative lesions in the hypereosinophilic syndrome.
After approval by the Mayo Clinic Institutional Review Board, eosinophils from patients with marked eosinophilia (up to 84 %) were obtained by cytapheresis (25). The methods for purifying eosinophil granules and granule proteins have been described (14, 26–28). Distilled and deionized water was used throughout all eosinophil granule protein purifications. Briefly, after cell lysis and granule isolation, granules were solubilized in 0.01 M HCl (pH 3.0) with sonication. After centrifugation (40,000 × g, 20 min), the supernatant was fractionated at 4 °C on a 1.2 × 100 cm Sephadex G-50 column equilibrated with 0.025 M sodium acetate, 0.15 M NaCl, pH 4.2.
Fractions containing EPO were pooled, dialyzed against 0.05 M Na2HPO4, 0.05 M KH2PO4, pH 8.0, centrifuged, applied to a CM-Sepharose column, and eluted with a linear NaCl gradient (0.15 to 1.5 M). Fractions rich in EPO activity were pooled and dialyzed against phosphate-buffered saline (PBS; 8.1 mM Na2PO4, 1.5 mM KH2PO4, 137 mM NaCl, 2.7 mM KCl, pH 7.4) and concentrated. The concentrated solution was centrifuged, and the supernatant was removed and frozen at −70 °C. All preparations of EPO had absorbance ratios (absorbance at 412 nm/absorbance at 280 nm) between 0.95 and 1.04.
Fractions from the Sephadex G-50 column containing ECP and EDN were pooled, dialyzed against one-half concentration PBS (pH 7.4), concentrated, and fractionated on a heparin-Sepharose CL-6B column (1.2 × 8 cm) equilibrated with one-half concentration PBS. Proteins bound to the column were eluted with a linear NaCl gradient (0.07 to 1.5 M NaCl). EDN eluted at a low salt concentration (approximately 0.2 M NaCl), whereas ECP eluted around 0.4 M NaCl (14). Extinction coefficients at 280 nm of 1.64 (mg/ml)−1cm−1 for EDN and 1.55 (mg/ml)−1cm−1 for ECP were used to determine protein concentrations (14). To obtain ECP1 and ECP2 (14), two forms of differentially-glycosylated ECP, ECP fractions pooled from the initial heparin-Sepharose column were subjected to a second heparin-Sepharose chromatography. These latter fractions were analyzed by SDS-PAGE followed by colloidal Coomassie blue staining (GelCode Blue; Pierce, Rockford, IL), and ECP1 and ECP2 were pooled separately.
To test the influence of RNase activity on ECP and EDN lesion formation, injectable-grade RNasin, an RNase inhibitor that effectively inhibits the RNase activity of EDN (15), was mixed with ECP and EDN for a final concentration of 3.3 μM protein and 17 μM RNasin (64 μg of 78 units/μg RNasin in a final volume of 75 μL) just before intradermal injections of each mixture into two guinea pigs. For comparison, ECP or EDN mixed with sterile PBS in place of the RNasin was also injected.
To test further the effect of RNase activity, 1 mg of purified ECP or EDN was diluted in 200 mM MES, 167 mM NaCl, pH 5.5 and carboxymethylated by incubating with 75 mM sodium iodoacetate (Sigma, St. Louis, MO) at 37 °C for 5 hours in the dark (15). As controls, a “sham” carboxymethylation procedure without addition of sodium iodoacetate was performed with purified ECP or EDN, and these products are referred to as unmodified-ECP and unmodified-EDN. After incubation, the samples were diluted ten-fold with 8.1 mM Na2PO4, 1.5 mM KH2PO4, pH 7.4 (“no salt” PBS) and loaded onto a Heparin Sepharose CL-6B column (Amersham Biosciences, Pittsburgh, PA) equilibrated with no salt PBS. The column was washed with no salt PBS and a gradient from 0 to 430 mM NaCl in PBS was initiated. Absorbance at 280 nm of collected fractions (to estimate protein content) and RNase activity of selected peak fractions (see “RNase activity measurement” section below) were measured. This procedure resulted in carboxymethylated-ECP (CM-ECP) and CM-EDN that eluted from the heparin-Sepharose column at somewhat lower salt concentrations and which had RNase activity diminished by 80% to 90%, based on the change in initial RNase velocity. Fractions containing MBP1 (2, 29) from the Sephadex G-50 column were pooled and stored at −70 °C in 0.025 M sodium acetate, 0.15 M NaCl, pH 4.2 to inhibit polymerization. MBP1 concentrations were determined using an extinction coefficient at 277 nm of 2.63 (mg/ml)−1cm−1 (26). Use of a more recently determined extinction coefficient [3.67 (mg/ml)−1cm−1] would have resulted in 28% lower calculated MBP1 concentrations (2). The original calculations were used throughout the study for consistency with other granule protein concentrations. All eosinophil granule protein preparations were pure as assessed by Coomassie blue staining following SDS-PAGE.
The Institutional Animal Use and Care Committee approved the studies. Hartley guinea pigs, 400 to 600 grams, were anesthetized with 100 mg/ml ketamine hydrochloride containing 1% xylazine (Ketaset, Fort Dodge Laboratories, Fort Dodge, IA) by intramuscular injections of 0.5 ml per kilogram body mass. Fifty or ninety microliters of ECP, EDN, EPO, and MBP1 at concentrations ranging from 0.5 μM to 50 μM prepared in sterile buffers using lipopolysaccharide-free water (Sterile Water for Irrigation, USP, Baxter Healthcare, Deerfield, IL) were injected intradermally into shaved back skin. Sterile acetate buffer (0.025 M sodium acetate, 0.15 M NaCl, pH 4.2; MBP1 storage buffer), phosphate buffer with 1 M NaCl (elution and storage buffer for ECP, EDN, and EPO with a NaCl concentration greater than or equal to that needed to elute each of these proteins from their respective cation exchange columns during purification), and sterile PBS were also injected. Additional intradermal injections included poly-L-arginine hydrochloride (5 kDa to 15 kDa), bovine RNase A, and lipopolysaccharide (from E. coli J5) from Sigma. Injectable grade RNasin was from Promega (Madison, WI). Pyroglu-angiogenin (Ang) was kindly provided by Dr. Robert Shapiro (Harvard Medical School, Boston, MA, USA). To determine that the injection-site reactions were not species-restricted, rabbits were also tested.
Injection sites were examined at least twice weekly. Lesions were assessed for appearance using the grading scale: 0 – no visible or palpable lesion; 0.5 – trace visible erythema, edema, and/or palpable lesion; 1 – visible erythema, edema with a palpable lesion; 2 – erythematous, edematous, indurated palpable lesion; 3 – palpable lesion with induration and mainly nonerosive epidermal changes (epidermal disruption with < 50% of lesion eroded or ulcerated); 4 – induration with crusting or ulceration (>50% of lesion). For skin biopsy collection, animals were killed by intracardiac injection of pentobarbital (Nembutal), and biopsies were immediately obtained by gently lifting the skin and using a curved scissor to cut around the injection sites. Twenty two guinea pigs and two rabbits were used in the studies; not all were tested with the same injections at the same time. The “n” (number tested) is listed in each table for the respective experiments.
Human ECP, EDN, EPO, and MBP were detected in formalin-fixed, paraffin-embedded biopsy specimens of cutaneous injection sites by indirect immunofluorescence. Briefly, serial tissue sections (5 μm) were mounted on positively charged microscope slides, deparaffinized, and incubated in 0.1% trypsin for one hour at 37 °C. The slides were incubated overnight in 10% normal goat serum at 4 °C. The next day, sections were washed and overlaid with either control rabbit antibody (normal rabbit IgG or rabbit pre-immunization serum) or granule protein-specific antibody (affinity chromatography-purified rabbit anti-human EDN and MBP or rabbit antiserum against ECP and EPO, all produced in our laboratory). Slides were washed and treated with 1% chromotrope 2R (J.T. Baker, Phillipsburg, NJ) to eliminate nonspecific eosinophil staining (30). Subsequently, slides were overlaid with fluorescein isothiocyanate (FITC) conjugated goat anti-rabbit IgG (Southern Biotechnology Associates, Birmingham, AL). After a final wash, sections were mounted in a glycerol solution containing p-phenylenediamine to prevent fading (31). Representative photomicrographs were taken with a Zeiss Axiophot microscope (Carl Zeiss, 7082 Oberkochen, Germany) equipped with excitor barrier filter set: blue BP450/490-LP520/560 (N 487910). Histologic examination with H&E stains was also performed to examine cellular infiltration.
Mouse liver RNA was isolated using RNA-STAT protocol (Tel-Test, Friendswood, TX) and quantitated by absorbance at 260 nm. RNA was stored in 1× TAE buffer (40 mM Tris-acetate, 1 mM EDTA, pH 8.3 in diethylpyrocarbonate-treated sterile H2O). To assay for RNase activity (32), stock YOYO-1 dye (1 mM; Invitrogen, Carlsbad, CA) was added to 10 μg/ml mouse liver RNA in TAE buffer to a final YOYO-1 concentration of 2 μM, and this solution was heated at 65°C for ten minutes. This RNA/YOYO-1 solution (100 μl) was added to the wells of a polystyrene flat bottom 96-well plate and measured using a Cytofluor plate reader series 4000 (PerSeptive Biosystems, Framingham, MA) with excitation at 485 nm and emission at 530 nm. The RNA/YOYO-1 complex was allowed to equilibrate for 1 hour before addition of protein (EDN, ECP, RNase A, or Ang) to duplicate wells, and fluorescence readings were recorded. Initial RNase velocity values were calculated from the change in arbitrary fluorescence units occurring from 2 to 4 minutes after the addition of protein and normalized per mass of protein added.
Reducing SDS-PAGE was used to assess the purity of the eosinophil granule protein preparations (Figure 1). Figure 2 shows representative lesions following eosinophil granule protein injection, with robust lesions from ECP and EDN, and less intense or no lesions from EPO, MBP1, and bovine RNase A. In this experiment, granule proteins at 10 μM and 2.5 μM and their storage buffer solutions were injected intradermally, and injection sites were observed for up to four weeks; Table I shows the lesion grade for each granule protein. Overall, ECP induced the most severe lesions followed by EDN; both ECP and EDN had greater lesion-forming activity than EPO or MBP1. ECP lesions were distinctly and consistently ulcerative, and EDN lesions typically showed white, dry crusts (Figure 2). ECP and EDN lesion formation began within two days and peaked approximately 7 days after injection (Table I). Injection sites of identical samples on duplicate guinea pigs were similar in magnitude and appearance, while the overall severity of the 2.5 μM injection sites was diminished relative to the 10 μM sites (Table I). Bovine RNase A, sterile PBS, and phosphate buffer with 1 M NaCl injection sites did not cause lesions within 24 hours after injection (Figure 2). Undiluted acetate buffer caused a trace lesion (0.5 score) that returned to normal after one week. However, for control purposes, this was an extreme acetate buffer concentration because the stock MBP1 in acetate buffer was diluted more than 20-fold with sterile PBS before injection. Intradermal injections of the granule proteins at 25 μM were also performed in two rabbits. Similar lesion formation occurred except that lesions from EPO were relatively more intense and approached the severity of lesions from EDN (Table I).
Additional experiments were performed with a wider range of ECP and EDN concentrations. Lesions were consistently elicited with ≥ 2.5 μM ECP or EDN, and visible lesions frequently formed even at 1 μM ECP or EDN. Lesions remained detectable for more than 2 weeks with ≥ 2.5 μM ECP or EDN (Table I). Alopecia persisted at least 4 weeks at 10 EDN injection sites (two 9 μM and eight 10 μM) on four guinea pigs. The 10 μM EDN injection sites on 1 guinea pig were monitored up to 6 weeks post injection and showed hair re-growth at this time. Alopecia and crusting persisted longer than 4 weeks at 3 ECP injection sites (all 10 μM) on each of 2 guinea pigs.
To rule out a potential contribution from bacterial or lipopolysaccharide contamination in purified eosinophil granule protein preparations, 3 approaches were taken. First, purified lipopolysaccharide at concentrations up to 100 μg/ml was intradermally injected into guinea pigs. Lesion-forming activity by endotoxin was minimal, never exceeding a trace lesion (0.5 score), even at 100 μg/ml, from 2 to 14 days after injection. Second, to avoid bacterial contamination of injected protein samples, all samples were centrifuged at 12,000 × g for 10 minutes before injection, and only the supernatants were injected. A supernatant sample of ECP was tested for bacterial contamination by culture at the Mayo Clinic Microbiology Laboratory, and no microorganisms were detected after five days of culture. Finally, to rule out a pathogenic bacterial infection, a representative lesion induced by ECP was biopsied, and a portion was cultured. Testing was performed on both agar plates and in liquid broth; coagulase-negative Staphylococcus and Streptococcus viridans, likely commensal organisms, were cultured from the biopsy.
Within 24 hours after ECP injection, heterophil and lymphomononuclear subdermal cell infiltrates were prominent in the dermis, including near cutaneous trunci muscles, and these infiltrates remained substantial 2, 3, and 4 days after injection (Figures 3A, B). Cellular infiltration increased in the upper dermis during this time and peaked at approximately day 7; cell infiltrates began to decrease at approximately day 14. Similar patterns of cellular infiltration occurred for the other three granule proteins, but with intensities proportional to the lesion-forming activities of the granule proteins (i.e. less cellular infiltration with decreased lesion formation). Likewise, intradermal injection of PBS and RNase A, neither of which induced a skin lesion, resulted in minimal or no detectable cellular infiltration during the first week.
Immunofluorescence staining showed that most of the intradermally injected ECP and EDN, diffusely distributed in tissues initially after injection (Figure 3C), was associated with cells in the dermis within 48 hours (Figure 3D, E, F). In contrast, EPO and MBP1 remained primarily extracellular and appeared to bind to dermal matrix fibers (Figure 3G, H) (5).
Glycosylation influences protein recognition and activity, and because of heterogeneous glycosylation, ECP exists in two predominant forms, i.e. ECP1 and ECP2 (14). Therefore, a sample of ECP containing a mixture of ECP1 and ECP2 was subjected to a second heparin-Sepharose chromatographic separation with NaCl gradient elution. SDS-PAGE analyses showed fractions containing predominantly the higher molecular weight ECP1 or the lower molecular weight ECP2 that eluted from the heparin-Sepharose column at lower and higher salt concentrations, respectively, and that displayed an approximate 2 kDa difference in SDS-PAGE mobility. Intradermal injections of unseparated ECP, ECP1, and ECP2 at 10 μM and 2.5 μM resulted in lesions of similar severity among the 10 μM and 2.5 μM injections, respectively. These injection sites were biopsied one week later and stained by immunofluorescence for ECP. For all three ECP preparations, specific staining was detected, and the majority of staining appeared localized to cells.
The limited lesion-forming activity of MBP1 (13.8 kDa, + 16 net positive charge) suggested that strong cationic charge is not sufficient to induce cutaneous lesions as intense as those caused by ECP or EDN. Similarly, intradermally injected poly-L-arginine at concentrations up to 25 μM showed a maximum of a grade 1 lesion that subsided to a trace lesion after one week. The influence of protein net cationic charge on lesion formation was also examined by injecting RNases of different charges. Ninety microliters of ECP (10 μM), EDN (10 μM), Ang (10 μM), and RNase A (15 μM) were each injected intradermally in duplicate into two guinea pigs. Table II shows the course of lesion formation; lesion scores for duplicate injections on the same guinea pig and on different guinea pigs, as graded by two investigators, varied by ≤ 0.5. Lesion severity was not directly proportional to the net cationic charge among these RNases because, although ECP (+14 charge) and RNase A (+4 charge) caused the most and least severe lesions, respectively; the more highly cationic Ang (+10 charge) did not induce more severe lesions than those of EDN (+7 charge).
To test the influence of RNase activity on ECP and EDN lesion formation, RNasin was mixed with ECP and EDN before intradermal injections of each mixture. For comparison, ECP or EDN mixed with sterile PBS was also injected. The ECP plus RNasin lesion was diminished in both size and intensity by 50% to 70%; however, the EDN plus RNasin lesion was unaffected. Immunofluorescence staining of sites biopsied four days after injection detected minimal ECP in the ECP plus RNasin injection site; in contrast, staining of ECP and EDN in the remaining three sites, i.e. ECP without RNasin and EDN with and without RNasin, was comparable to untreated injection sites.
To address further the role of RNase activity in lesion formation, an active site histidine of ECP and EDN was carboxymethylated. We first confirmed the relative RNase activities of ECP, EDN, Ang, and RNase A (Figure 4A). Following sham treatment or treatment with 75 mM iodoacetate, samples of the unmodified (sham-treated) and carboxymethylated (CM) ECP or EDN were re-isolated by heparin-Sepharose chromatography. Peak fractions of unmodified-ECP, unmodified-EDN, CM-ECP, and CM-EDN were tested for RNase activity; carboxymethylation reduced the RNase activity of ECP and EDN by approximately 80% and 87% (Figure 4B), respectively. Carboxymethylation of ECP and EDN strikingly reduced lesion-forming activities by 2–3 grades both on Days 3 and 7 (Table III). Immunofluorescence staining for ECP at injection sites of unmodified- and CM-ECP at day 3 demonstrated that carboxymethylation did not alter the cellular localization or the overall immunofluorescence distribution of the protein (Figure 5); immunofluorescence staining patterns were similar for unmodified-EDN and CM-EDN.
Morphologically, the lesions resulting from ECP and EDN injections into guinea pig skin showed similarity to erosive and ulcerative lesions in patients with the hypereosinophilic syndrome (HES) (Figure 6A, B, C) (33). Immunofluorescence staining of lesional tissue from four such patients demonstrated both cellular localization and extracellular deposition of ECP and EDN (Figure 6D, F), similar to that observed in guinea pig skin at different time points following intradermal protein injections (Figure 3), with the exception that the human tissue showed staining of intact eosinophils. The MBP1 cellular staining, as represented by brightly fluorescent ovals, reflects the relative numbers of intact eosinophils (Figure 6E). Both EDN (Figure 6F) and especially ECP (Figure 6D) staining shows more extensive cellular localization than accountable by MBP1-associated eosinophil staining suggesting their localization to cells other than eosinophils.
Eosinophil granule protein deposition occurs in a variety of dermatoses (4). Several of the granule proteins induce a cutaneous wheal-and-flare reaction and increase cutaneous vasopermeability in vivo (5, 19, 20). The in vitro cytotoxicity of eosinophil granule proteins is also well documented (1). In addition, the hypereosinophilic syndromes are often associated with cutaneous lesions (34), and atopic dermatitis shows strong association with eosinophil activity (35, 36). Even so, other reports suggest that eosinophils have a limited impact on asthma and atopic dermatitis inflammation (21–24). To help clarify these issues, we investigated the effects of eosinophil granule proteins on guinea pig and rabbit skin through injections into the dermis, a site to which the granule proteins localize in cutaneous diseases. Intradermal injection of micromolar concentrations of four eosinophil granule proteins, ECP, EDN, EPO, and MBP1, produced cutaneous lesions with cellular infiltration. The ECP- and EDN-induced lesions were substantially more pronounced than the EPO- and MBP1-induced lesions. This was unexpected for EDN because of its limited in vitro cytotoxicity toward K562 cells relative to that of ECP, EPO, and MBP1 (2). The lesion-forming activity of ECP and EDN was associated with their localization in dermal cells. Furthermore, RNase activity appears important to their lesion-forming activities, and their respective net positive charges also may have modulated lesion formation. However, neither RNase activity nor high net positive charge alone was sufficient to account for full lesion formation.
The marked lesion-forming activities of ECP and EDN focused our attention on these proteins (Figure 2). An initial concern was whether a contaminant or an infectious component might have contributed to lesion formation. Because lesion-forming activity varied among the different granule proteins and was consistent for a given granule protein, we believe that no systematic factor or contaminant was substantially contributing to EDN- and ECP-induced lesions. Several approaches, addressing potential endotoxin or bacterial contamination of the protein preparations and post-injection infection, showed that bacterial or endotoxin contamination did not explain EDN- and ECP-induced lesions. Finally, the ability of RNasin, a specific RNase inhibitor, to diminish an ECP-induced lesion by about 60% and the ability of carboxymethylation to diminish ECP- and EDN-induced lesions (Table III) support the specific lesion-forming activities of ECP and EDN. Notably, RNasin (pI = 4.7) probably did not inhibit an EDN-induced lesion because of the relatively short half-life of the EDN-RNasin complex (EDN, +7 net charge) compared to that of angiogenin (+10 net charge) (15), and presumably to that of ECP (+14 net charge). A short EDN-RNasin half-life compared to that of ECP-RNasin is also consistent with the immunofluorescence detection of intradermally injected EDN, but not ECP, when mixed with RNasin. Overall, it is unlikely that the cutaneous lesions caused by intradermal injections of ECP or EDN are attributable to a contaminating substance or infectious component.
The ECP- and EDN-induced lesions developed over one week, occurred at concentrations as low as 1 μM, and occurred in both guinea pig and rabbit skin (Figure 2, Table I). ECP and EDN concentrations > 1 μM are likely deposited in human skin disease (5). Given the pronounced lesion-forming activities of ECP and EDN, we considered distinctive properties of ECP and EDN compared to the less active lesion-forming EPO and MBP1. Two such properties are neurotoxicity and RNase activity. A third property, distinguishing ECP and EDN from EPO and MBP1, is cellular localization after intradermal injection (Figure 3) (5). Cellular localization of ECP and EDN appears consistent with previous observations for another RNase, onconase (37), and with our staining for ECP and EDN compared to MBP1 in HES skin lesions (Figure 6). Interestingly, cellular internalization can increase the neurotoxicity of an RNase by 1000-fold (38). Thus, internalization of ECP and EDN, as suggested by the appearance of the immunofluorescent cells within the dermis containing ECP and EDN (Figure 3), might account for EDN’s unexpectedly high lesion-forming activity. Unfortunately, due to the paucity of guinea pig-specific reagents, the identity of the cells accumulating ECP and EDN remains unknown. However, ECP has been observed in macrophage-like dermal cells after house dust mite patch testing of humans (39), and EDN has recently been reported as an endogenous Toll-like receptor 2-dependent “alarmin” acting on dendritic cells (40). If EDN binds directly to Toll-like receptor 2, a precedent for it being internalized after ligation exists (41). In addition, EDN is chemotactic for dendritic cells and stimulates release of inflammatory mediators, such as IL-8 and TNF-α (40, 42). IL-8 and TNF-α can lead to further cell recruitment, the latter, at least in part, via increased endothelial cell adhesion molecule expression (43), and TNF-α induces an inflammatory lesion when injected into rabbit skin (44). Thus, ECP and EDN appear to be internalized by resident dermal cells and, directly or indirectly via induction of other inflammatory mediators, cause additional cellular infiltration and cutaneous disruption.
Given the potential importance of RNase activity, we examined its influence in ECP- and EDN-induced lesion formation and two approaches, in addition to the use of RNasin (discussed above), were taken. First, we compared the lesion-forming activities of ECP and EDN to two other RNases, Ang and bovine RNase A. The relative RNase activities for these four RNases against the RNA substrate used in this study (i.e. mouse liver total RNA) were as follows: RNase A was more active than EDN and these were substantially more active than ECP, which was substantially more active than Ang (Figure 4). These relative activities are expected from previous reports. However, the relative lesion-forming activities of ECP and EDN were substantially greater than that of Ang, which was substantially greater than that of RNase A (Table II). Therefore, lesion formation does not depend solely and directly on RNase activity. Second, we tested carboxymethylated ECP and EDN, and this treatment strikingly inhibited lesion formation (Table III). Thus, at least for ECP and EDN, RNase activity is required for maximal lesion formation. Because carboxymethylation inserts one negative charge per carboxymethyl group added, these changed molecular charges of the carboxymethylated ECP and EDN could be confounding factors (see below). Also, factors other than RNase activity appear to be involved, because while ECP’s RNase activity is substantially less than that of EDN, ECP produces a more prominent lesion.
Cationicity is often associated with cytotoxicity (45). Thus, net positive charge, as suggested by the cytotoxic activity of the non-RNase MBP1 (+16, MW 13.8 kDa) (2), may contribute to lesion formation. For example, a more highly cationic protein may have an increased partitioning to the cell’s surface or increased direct cytotoxicity or both. The protein cationicity of the RNases decreases in the following order: ECP, Ang, EDN, and RNase A (Table II). Because the more highly charged Ang does not produce a more pronounced lesion than that caused by EDN, nor does MBP1 or poly-L-arginine cause such pronounced lesions, a direct relationship between net positive charge and lesion-forming activity does not exist. Glycosylation can affect a protein’s net charge, and both ECP and EDN are glycosylated while Ang and RNase A are not. A recent study showed that the N-linked carbohydrates on ECP are in part made up of sialic acid, galactose, and acetylglucosamine (46). These individual carbohydrates are suggestive of “complex type” N-linked glycosylation, which typically involves non-charged carbohydrates and negatively charged sialic acid. This form of glycosylation would only diminish ECP’s cationicity and presumably any associated cation-dependent cytotoxicity. Alternatively, such glycosylation could confer other properties to ECP and EDN that contribute to their distinct lesion-forming capabilities compared to Ang and RNase A, even though differential glycosylation of ECP1 and ECP2 did not noticeably alter their lesion-forming activities.
Thirteen members of the human RNase family have been identified (47). Among these thirteen genes, ECP and EDN form a distinct clad with substantial homology. Thus, unidentified molecular characteristics specific to ECP and EDN may be critical to their cutaneous lesion-forming activity. For instance, after intradermal injection, it is unknown whether Ang and RNase A localize to dermal cells like ECP and EDN. However, similar to their difference in lesion-forming potency reported here, the alarmin adjuvant effect of EDN reported by Yang et al. was not recapitulated by human angiogenin (40). The relatively specific lesion-inducing activities of ECP and EDN compared to other RNases are also reminiscent of their unique neurotoxic and antiviral activities (48).
Morphologically, the lesions resulting from ECP or EDN injection into guinea pig skin showed similarity to erosive and ulcerative lesions in an HES variant (Figure 6A, B, C) that, until recently, was associated with a grave prognosis (33). HES with ulcerative lesions appears to be a presentation of myeloproliferative HES (associated with a deletion on chromosome 4 resulting in a fusion gene product, FIP1L1-PDGFRA and yielding a novel kinase sensitive to imatinib mesylate therapy with long-term disease control) (49, 50). The biopsy specimens from 4 HES patients showed both eosinophil infiltration and eosinophil granule protein deposition; the staining of cell-localized ECP and EDN was extensive, out of proportion to the number of infiltrating eosinophils identified by MBP1 staining, and similar to that observed in guinea pig skin at later time points following injection of ECP and EDN. Ongoing deposition of ECP and EDN in human tissue, unlike single injections of ECP or EDN into guinea pig skin, likely contributed to the relatively greater extracellular staining of ECP and EDN in human tissue compared to later guinea pig intradermal injection sites (≥ 2 days post injection). Because of ethical considerations, it is not possible directly to study the effects of eosinophil granule proteins injected into human skin; however, these analogous findings by formation of lesions and by immunostaining support the conclusion that ECP and EDN cause cutaneous lesion in human disease.
In summary, all four eosinophil granule proteins, ECP, EDN, EPO, and MBP1, induce a cutaneous lesion after intradermal injection into guinea pig and rabbit skin at pathophysiologically relevant concentrations. The ECP- and EDN-induced lesions are more intense than lesions induced by EPO and MBP1, and this difference appears to be closely associated with cellular internalization and RNase activity of ECP and EDN. Net cationic charge may also modulate lesion-forming activity. However, neither cationicity nor RNase activity directly correlates with lesion-forming activity, suggesting the potential importance of other properties specific to ECP and EDN. Overall, these data provide further direct evidence that deposition of the eosinophil granule proteins, particularly ECP and EDN, at micromolar concentrations, can severely effect cutaneous structure and function in diseases in which granule proteins are deposited including in the hypereosinophilic syndrome associated with mucocutaneous ulcerations (Figure 6) (33, 34, 50).
We thank Robert Shapiro for providing pyroglu-angiogenin, and David A. Loegering and James L. Checkel for providing the purified eosinophil granule proteins. The authors have no conflicting financial interests.
2Abbreviations: pyroglu-angiogenin, Ang; carboxymethyl, CM; eosinophil cationic protein, ECP; eosinophil-derived neurotoxin, EDN; eosinophil peroxidase, EPO; hypereosinophilic syndrome, HES; major basic protein-1, MBP1; Tris-acetate-EDTA, TAE
1Supported by: National Institutes of Health Grants AI 11483, AI 09728, AI07047 Training Grant, AI 34577, AI 50494; American Academy of Allergy, Asthma and Immunology Women Physicians in Allergy Grant (KML); The Mayo Foundation, the Kieckhefer Foundation, and the Dr. Smith H. and Lucille Gibson Postdoctoral Research Fellowship in Dermatology (DAP).
Publisher's Disclaimer: NOTE: This is an author-produced version of a manuscript accepted for publication in The Journal of Immunology (The JI). The American Association of Immunologists, Inc. (AAI), publisher of The JI, holds the copyright to this manuscript. This version of the manuscript has not yet been copyedited or subjected to editorial proofreading by The JI; hence, it may differ from the final version published in The JI (online and in print). AAI (The JI) is not liable for errors or omissions in this author-produced version of the manuscript or in any version derived from it by the U.S. National Institutes of Health or any other third party. The final, citable version of record can be found at www.jimmunol.org.