Characterization of Human T Cells That Regulate Neutrophilic Skin Inflammation¹

Depending on their state of activation and polarization, effector T cells directly influence various cell types that may lead to a certain pathology. IFN-γ-producing Th1 cells are involved in macrophage and CD8⁺ T cell activation and promote proinflammatory conditions (TNF-α, IL-12), and are therefore ideally designed for immunity against intracellular pathogens. Additionally, many autoimmune diseases depend on a Th1 response to autoantigens. In contrast, Th2 responses use the cytokines IL-5, IL-4, and IL-13, which enhance eosinophil and mast cell mobilization and cause certain allergic diseases such as asthma. In addition, Th1 responses may be counteracted by Th2 responses (1, 2, 3, 4, 5).

Drug hypersensitivity reactions are known as imitators of many T cell-mediated diseases, and understanding their pathomechanism might contribute substantially to elucidate basic immunological mechanisms (6). In this study, we used the severe pustular drug eruption acute generalized exanthematous pustulosis (AGEP)⁴ as a model to better understand the interplay between T cells and neutrophils.

Some 90% of AGEP cases are related to the intake of drugs, in particular antibacterials such as aminopenicillins. AGEP has three specific features: 1) an acute generalized formation of numerous, mostly nonfollicular intraepidermal or subcorneal sterile pustules (<5 mm) on a widespread edematous erythema in the absence of a bacterial inflammation (7, 8, 9); 2) neutrophils appear after T cell infiltration (10, 11, 12); and 3) the possibility to induce the reaction by patch testing with the corresponding drug, whereby a massive release of CXCL8/IL-8 by both keratinocytes and isolated drug-specific T cells upon stimulation can be observed (10, 13). The onset of AGEP is acute; resolution of pustules occurs spontaneously within 4–10 days after cessation of the incriminated drug (14). In summary, AGEP seems to be a drug-induced, T cell-mediated disease, whereby the effector function of T cells leads to a neutrophil-rich inflammation.

In an initial study, we demonstrated that lymphocyte proliferation tests can be positive with the incriminated drug (15). Culture of drug-reactive T cells isolated from peripheral blood or from positive skin patch test biopsy specimens of AGEP patients led to the generation of drug-specific, HLA-restricted, αβTCR⁺CD4⁺ and CD8⁺ T cell lines (TCL) and clones (TCC) (10, 12). Most of these drug-specific T cells produced high levels of the potent neutrophil-attracting chemokine CXCL8.

In this study, we describe phenotypic and functional features of drug-specific CXCL8-producing (CXCL8⁺) CD4⁺ T cells obtained from AGEP patients and healthy individuals. We show that CXCL8⁺ T cells bear a specific chemokine receptor profile and express mainly GM-CSF, IFN-γ, and TNF-α, and occasionally IL-4 and IL-5. Furthermore, factors produced by CXCL8⁺ T cells enhance neutrophil survival. Our results suggest that chronic neutrophil-rich inflammatory diseases involve CXCL8-producing T cells, which can orchestrate and boost neutrophil inflammation. Our findings may be relevant for other T cell-mediated diseases with similar pathology such as Behçet’s disease or pustular psoriasis.

Culture medium (CM) consisted of RPMI 1640 (Sigma-Aldrich, St. Louis, MO) supplemented with 10% pooled heat-inactivated human AB serum (Swiss Red Cross, Bern, Switzerland), 25 mM HEPES buffer (Seromed, Basel, Germany), 2 mM l-glutamine (Biotest Diagnostics, Dreieich, Germany), 10 μg/ml streptomycin, and 100 U/ml penicillin (Amimed, BioConcept, Allschwil, Switzerland). For the culture of TCC, the CM was enriched with 200 U/ml human rIL-2 (CM⁺). EBV-transformed B-lymphoblastoid cell lines (B-LCL) were generated, as described (16). Amoxicillin (Sigma-Aldrich) and celecoxib (Pharmacia, St. Louis, MO) were used for proliferation assays. Tetanus toxoid (Serum and Vaccine Institute, Bern, Switzerland) was used as a control Ag. Stock solutions of each drug were always freshly prepared in CM just before use. Celecoxib was dissolved in RPMI 1640 with 0.05 M NaOH.

Drug- and tetanus toxoid-specific TCL and TCC from blood (designated B#) of drug-allergic patients AP, JS, and EB were generated, as described elsewhere (10). Skin-derived T cells (designated S#) were obtained from 5-mm punch biopsy specimens of positive epicutaneous test reactions upon amoxicillin application on the back of patients AP and JS. Skin reactions were observed after 48 h and scored as described (17). T cells were expanded in CM⁺ and restimulated every 14 days with irradiated (45 Gy), allogeneic PBMC and 1 μg/ml PHA. Ag specificity of TCL and TCC was tested by incubating 5 × 10⁴ T cells (day 10–16 after restimulation) with either 5 × 10⁴ of irradiated (45 Gy), autologous PBMC or 1 × 10⁴ irradiated (60 Gy), autologous B-LCL as APCs in the presence or absence of Ag in 200 μl of CM in U-bottom 96-well microplates. After 48 h, 0.5 μCi of [³H]thymidine was added for 8–14 h. Finally, cells were harvested and incorporated radioactivity was measured with a beta counter (Inotech Filter Counting System INB 384; Inotech, Dottikon, Switzerland).

Healthy abdominal skin was excised using a dermatome (0.3–0.5 mm thick), and 100-cm² sections were digested in 10 ml of RPMI 1640 containing 1 mg/ml collagenase D (1088866; Roche, Basel, Switzerland) for 30 min at 37°C on a shaker. Digestion was stopped by adding 10 mM EDTA and immediate cooling and processing on ice. Floating cells were collected and pooled with cells obtained by washing the remaining tissue several times with ice-cold PBS⁻/10 mM EDTA (minimum 10× excess of digestion volume). Cells were passed through a 70-μm-pore nylon mesh (BD Pharmingen, San Diego, CA), centrifuged (300 × g) for 20 min at 4°C. After resuspension in cold complete medium, cells were passed through a 40-μm-pore mesh and centrifuged on a Ficoll-Paque gradient (600 × g; 20 min at 4°C). Interphase cells consisted of >90% viable T cells (as judged by CD45, αβTCR, and propidium iodide staining).

T cells from a biopsy of a positive 48-h Ni-patch test were isolated by cutting of skin tissue into small pieces. First, released cells were collected and remaining tissue was digested for 45 min in collagenase D (Roche) to obtain the second fraction (∼10% of first fraction). Chemokine receptor expression analysis was performed on both fractions separately and did not differ.

The study was approved by the ethics committee of the University of Bern, and informed consent was obtained by all participants of the study.

For cytokine detection by ELISA, SN were used from T cell specificity assays and from unspecific stimulation for 48 h with immobilized anti-CD3 (1 μg/ml; Okt3) and soluble anti-CD28 (1 μg/ml; BD Pharmingen) in flat-bottom 96-well plates (Nalge Nunc International, Roskilde, Denmark; 5 × 10⁴ cells/well, 200 μl of CM + 40 U/ml IL-2). ELISA sets used: IL-5 and CXCL8 (BD Pharmingen); IFN-γ, IL-4, and TNF-α (Diaclone, Besançon, France); and GM-CSF (R&D Systems, London, U.K.). TCC from AGEP patients that produced >0.1 ng/ml CXCL8 were called CXCL8⁺ T cells. To produce SN for chemotaxis and apoptosis assays, 0.5 × 10⁶ cells/well in 24-well plates were unspecifically stimulated in 500 μl of CM + 10% FCS for 48 h.

Regression and correlation of cytokine and chemokine production were determined by the ANOVA (SigmaPlot 5.0; SPSS, Chicago, IL). R² ≥ 0.5 for linear regression, and p ≤ 0.05 were considered significant.

Fluorochrome-labeled mAb against CD3, CD4, CD8, and isotype controls were supplied by BD Pharmingen. Monoclonality of TCC was shown by TCR Vβ chain staining using a panel of 22 mAbs recognizing different Vβ gene products, which detect ∼75% of all Vβ families (Beckman Coulter, Marseilles, France) (18). Flow cytometry was performed on a Coulter EPICS XL-MCL flow cytometer (Beckman Coulter). Abs to human CCR and CXC were purchased from the following sources: mAbs to CCR2 (48607.211) and CXCR6 (56811) from R&D Systems; mAbs to CCR3 (7B11) from Millenium Pharmaceuticals (Cambridge, MA); and mAbs to CCR4 (1G1), CCR5 (2D7), CCR6 (11A9), CCR7 (2H4), CXCR1 (5A12), CXCR2 (6C6), CXCR3 (1C6), and CXCR4 (12G5) from BD Pharmingen. Rabbit IgG to CXCR5 was from a noncommercial source, as described elsewhere (19).

CD4⁺ T cells were isolated from PBMC of healthy individuals by negative MACS selection (Miltenyi Biotec, Bergisch Gladbach, Germany). A total of 2 × 10⁶ cells/ml/24 well was stimulated with 10 ng/ml PMA/1 μM ionomycin (Calbiochem, La Jolla, CA) or anti-CD3/anti-CD28 + 20 U/ml IL-2. At various time points, SN were harvested for cytokine/chemokine detection. For detection of intracellular cytokines, 10 μg/ml brefeldin A was added together with PMA/ionomycin for 5 h. mAbs to IFN-γ, TNF-α, IL-4, IL-5, IL-6, IL-10, IL-13, and CXCL8 were from BD Pharmingen. Stainings were performed on fixed (2% paraformaldehyde/PBS) and permeabilized (0.5% saponin/PBS) cells.

Neutrophils from fresh blood of healthy volunteers (20) or murine pre-B 300-19 cells stably transfected with human CXCR1, CXCR2, CXCR4, CX3CR1, RDC1, or human chemokine receptor were used. Cell migration was measured in 48-well chemotaxis chambers (NeuroProbe, Cabin John, MD). Briefly, chemically synthesized CXCL8 (I. Clark-Lewis, Biomedical Research Centre, University of British Columbia, Vancouver, Canada) in HEPES-buffered RPMI 1640 supplemented with 1% pasteurized plasma protein (Swiss Red Cross Laboratory) or cell SN were added to the lower wells, and 10⁵ cells in control medium to the upper wells. Polyvinylpyrrolidone-free polycarbonate membranes (Poretics, Livermore, CA) coated with type IV collagen and with 5-μm pores for neutrophils or 3-μm pores for transfectants were used. After incubation for 20 or 90 min, respectively, the membrane was removed, washed on the upper side with PBS⁻, fixed, and stained. Migrated cells were counted microscopically at ×1000 magnification in five randomly selected fields (high power fields) per well. The assay was performed in triplicates. For inhibition of CXCL8-dependent chemotaxis, chemokines or SN and cells were incubated with 10 μg/ml neutralizing Ab to CXCL8 (clone 2A2, azide free; BD Pharmingen) 15 min before the chemotaxis assay.

Fresh neutrophils were incubated at 10⁵ cells/100 μl in Eppendorf tubes in diluted SN from different CXCL8-producing TCC (see above). After 5 h, apoptotic cells (annexin V⁺/propidium iodide⁻) were determined by flow cytometry.

AGEP patients described in this work had an acute onset of generalized, sterile pustules after drug intake (AGEP validation score ≥10 following the EuroSCAR study group) (9, 14) (Fig. 1,A). Drug-specific TCC isolated from peripheral blood of AGEP patients (patients AP and JS, amoxicillin; patient EB, celecoxib) and from skin biopsy specimens of positive epicutaneous patch tests (patients AP and JS) (Fig. 1,B) were found to recognize the drug in an HLA-DR-, but not in an allele-restricted manner (data not shown) (21). Dose-response curves for three representative blood-derived CD4⁺ TCC AP B10, EB B14, and EB B27 are shown (Fig. 1,C). More than 80% of the generated drug-specific TCC produced reproducibly high amounts of CXCL8 (10). In this study, we show that for most of the TCC induction of CXCL8 production is Ag specific and drug concentration dependent (Fig. 1 C). Interestingly, several TCC secreted CXCL8 at low levels even before stimulation (100–340 pg/ml; data not shown). In contrast to AGEP patients, TCC derived from noninflamed skin of healthy individuals failed to produce CXCL8 (n = 12; data not shown).

Drug-specific CD4⁺ TCC from AGEP patients could be grouped into low (0.2–0.6 ng/ml), medium (1.0–1.6 ng/ml), or high (2.9–4.9 ng/ml) CXCL8 producers. Tetanus toxoid-specific TCC from healthy individuals and drug-specific TCC from patients with either maculopapular exanthema or bullouse exanthema secreted <0.1 ng/ml CXCL8 and were defined as non-CXCL8 producers (data not shown). Twelve TCC with low, medium, or high CXCL8 secretion from three different donors were analyzed in detail for their cytokine/chemokine profile. A significant correlation was found between the production of CXCL8 and the Th1-associated cytokine IFN-γ (R² = 0.58, p = 0.0038) and with GM-CSF (R² = 0.59, p = 0.01), but not with the Th2-associated cytokines IL-4 and IL-5, or the proinflammatory cytokine TNF-α (Fig. 2). No correlation was found for IL-10, RANTES/CCL5, or eotaxin/CCL11, and G-CSF could not be detected (data not shown) (10).

Chemokine receptor expression of CXCL8-producing CD4⁺ TCC generated from blood or lesional skin of two AGEP patients was analyzed by flow cytometry and compared with CD4⁺ TCC from healthy skin (Fig. 3 A). Resting CD4⁺ TCC from AGEP expressed the Th1-associated chemokine receptors CCR5 and CXCR6 more frequently than control skin CD4⁺ TCC, although the frequency was below 25%. In contrast, CCR6 was expressed by the majority of T cells from AGEP clones (>85%), whereas normal skin TCC showed significantly less CCR6 positivity. In addition, the level of CCR6 expression was higher on AGEP clones than on normal skin clones (data not shown). Interestingly, CCR6 was only detected on skin-derived, but not blood-derived AGEP TCC. CCR2, CCR3, CCR7, CCR9, CXCR1, CXCR2, and CXCR5 were completely absent on resting AGEP clones, whereas CCR4 was expressed by <8% of T cells (data not shown).

To confirm that the phenotype of cloned AGEP T cells reflects the in vivo situation of an ongoing skin inflammation, chemokine receptor profiles of skin T cells directly isolated from an epicutaneous patch test reaction were determined (Fig. 3,B). Lesional skin was strongly enriched by CD4⁺ T cells (85% of total CD3⁺ T cells) as compared with healthy skin (50%; data not shown). In addition, neutrophils were absent at this time point of analysis, which supports the idea that early immigrating T cells may be able to influence the subsequent neutrophil infiltration by production of CXCL8 (data not shown). Most strikingly and in agreement with T cell clones, CCR6 expression was found on the majority (>75%) of lesional skin CD4⁺ T cells, in contrast to the low frequency (<25%) among healthy skin CD4⁺ T cells. Therefore, CCR6 expression seems to be a specific and stable marker on CD4⁺ T cells from lesional skin as well as AGEP clones. The low frequency (<20%) of T cells positive for CCR4, CCR5, and CXCR6 suggests that these chemokine receptors may play minor roles in T cell recruitment. In addition, the lack of CCR7 on most T cells indicates that infiltrating T cells belong to the effector memory T cell subset. The high frequency of CXCR3⁺ T cells among AGEP TCC (Fig. 3,A), which contrasts the minor expression of CXCR3 on lesional skin CD4⁺ T cells (Fig. 3 B), may be simply due to an up-regulation during activation and culturing of T cells, as has been previously described for CXCR3. Interestingly, CXCR3 was expressed by all lesional skin CD4⁺ T cells that have down-modulated CD3 receptor (∼8%), again indicating that CXCR3 expression correlates with T cell activation (data not shown). In conclusion, the prominent expression of CCR6 on CXCL8⁺ skin T cells suggests an obvious involvement of CCR6 and its ligand CCL20/liver and activation-regulated chemokine in the recruitment of these T cells into affected skin (see Discussion).

It is not known whether CXCL8-producing T cells are present in peripheral blood of healthy individuals. To address this question, secretion of CXCL8, IFN-γ, and IL-4 was followed over time in PMA/ionomycin-stimulated CD4⁺ T cells. Surprisingly, substantial amounts of CXCL8 were produced by CD4⁺ T cells after 3 days (>20 ng/ml) (Fig. 4,A). In agreement with previous data, IFN-γ levels were higher than IL-4 levels among CD4⁺ T cells (22). Interestingly, CXCL8 production was delayed compared with IFN-γ and resembled IL-5 and TNF-α production (data not shown). Comparable results were obtained after anti-CD3/anti-CD28 stimulation, but with lower cytokine and CXCL8 secretion levels (data not shown). Intracellular stainings of CD4⁺ T cells were performed to determine the frequency and phenotype of CXCL8-producing T cells in peripheral blood. CXCL8 was produced by 2.4 ± 0.8% of CD4⁺ T cells (mean ± SE; n = 9), and double stainings revealed that the majority of CXCL8⁺ T cells (>75%) also released TNF-α (Fig. 4 B). In contrast, CXCL8⁺ T cells completely failed to secrete IFN-γ, IL-4, IL-5, IL-10, and IL-13.

In vitro chemotaxis assays revealed that SN generated from CXCL8⁺ T cells induced strong chemotactic activity in freshly isolated neutrophils and in transfectants bearing either CXCR1 or CXCR2 (Fig. 5, lower panel). Synthetical CXCL8 was used as control (Fig. 5, upper panel). No migration was observed for transfectants stably expressing chemokine receptor CXCR4 or CX3CR1, and orphan receptors human chemokine receptor or RDC1, which are all expressed on neutrophils (data not shown) (23, 24, 25, 26, 27). This indicates that chemotaxis of neutrophils was mediated exclusively by CXCR1 and CXCR2.

Furthermore, CXCL8-neutralizing Abs completely blocked migration of CXCR1 transfectants, but had only moderate effects on CXCR2 transfectants and neutrophils (Fig. 5, open symbols). This suggests that besides CXCL8, SN contained factors with minor chemotactic activity for CXCR2, seen only at low SN dilutions (Fig. 5, lower panel). Because SN lacked granulocyte chemotactic protein-2/CXCL6, epithelial cell-derived neutrophil attractant-78/CXCL5, and growth-related oncogene-α,β,γ/CXCL1,2,3 (data not shown), the other major chemokines acting via CXCR1 and CXCR2, we suggest that CXCL8⁺ T cells produce an additional, as yet unidentified factor acting predominantly via CXCR2.

Neutrophils are cleared fast under noninflammatory conditions, and also pustular eruptions disappear soon after drug withdrawal in AGEP patients (14, 28). Interestingly, local pustule formation with accumulation of neutrophils can readily be reinduced and maintained by a patch test (Fig. 1,B) (13). This suggests that not only recruitment of neutrophils, but also suppression of their apoptosis may be controlled by drug-specific CXCL8-producing T cells, which results in a longer lasting lesion. Therefore, we tested whether drug-specific CXCL8⁺ T cells from AGEP patients can promote neutrophil survival in vitro. Initially, we found that SN from CXCL8⁺ T cells not only attracted neutrophils (Fig. 5, lower panel), but also strongly increased adherence (data not shown), suggesting that neutrophils can benefit from promoted interactions with extracellular matrix, a mechanism that has been shown to prevent apoptosis (29). Indeed, apoptosis was strongly reduced in neutrophils treated with conditioned medium from CXCL8⁺ T cells (Fig. 6). A maximal reduction of apoptotic cells of 40 ± 4% was observed. Increased survival was most likely due to T cell-released GM-CSF and/or IFN-γ (Fig. 2), two cytokines that are known to efficiently suppress neutrophil apoptosis (28, 29, 30). Neither CXCL8 nor other CXCR1/2 ligands alone were sufficient for these effects (data not shown). Our findings clearly demonstrate that CXCL8-producing drug-specific T cells can provide the proper cytokine microenvironment that enhances neutrophil survival.

Neutrophil recruitment in response to bacterial or fungal infection is a rapid and T cell-independent process, whereby the activated resident tissue cells release neutrophil-activating cytokines and chemokines (28, 31, 32, 33, 34, 35). Thus, neutrophil accumulation generally precedes T cells, which may be recruited at later stages of the infection and contribute to chronicity (36, 37). In contrast, the neutrophil-rich inflammation AGEP differs from this concept in three ways: 1) epidermal vesicle formation by drug-specific CTLs expressing perforin/granzyme and Fas ligand precedes pustule formation (10, 11, 12); 2) formation of >100 sterile pustules in the diseased skin occurs in absence of a bacterial inflammation (Fig. 1,A); and 3) a pustular reaction can be elicited by patch testing in a sensitized patient with the responsible drug, whereby a massive release of CXCL8 by both keratinocytes and drug-specific T cells can be observed (Fig. 1 B) (10, 13). Hence, AGEP seems to represent an interesting model for diseases, wherein the effector functions of involved T cells lead to a neutrophil-rich sterile inflammation. In this study, we have described these drug-specific CXCL8⁺ T cells with regard to their phenotype and their effects on the recruitment and survival of neutrophils.

The phenotypic analysis of drug-specific CXCL8⁺ CD4⁺ TCC from AGEP patients revealed a predominant Th1-type cytokine profile with high production of GM-CSF and IFN-γ, and various levels of TNF-α. However, rare CXCL8⁺ TCC displayed a Th2-type cytokine profile with high IL-4 and IL-5 secretion. In combination with eotaxin/CCL11 and RANTES/CCL5 released occasionally by perivascular cells, such TCC may contribute to the eosinophilia observed in ∼30% of AGEP cases, and which was also detected in patient JS (9, 10). Interestingly, CXCL8⁺ TCC showed, independently of high IFN-γ/GM-CSF or IL-4/IL-5 production, a chemokine receptor profile associated with Th1 effector memory T cells that preferentially express CCR5, CXCR3, and CXCR6 (38, 39, 40). The prominent expression of CCR6 on CD4⁺ T cells derived from AGEP skin as well as on T cells directly isolated from patch test-reactive skin may be explained by an enhanced recruitment of CCR6⁺ T cells. A similar mechansim has been previously suggested for psoriasis, which shows high expression of the CCR6 ligand liver and activation-regulated chemokine/CCL20 (41). An involvement of CCR6 in the pathology of AGEP is further supported by the minimal expression of this receptor and its ligand in healthy skin.

Chemotaxis experiments revealed that CXCL8 is the dominant chemokine, which efficiently attracted neutrophils and CXCR1/CXCR2 transfectants. In addition, SN of stimulated AGEP clones were devoid of other major chemokines acting via CXCR1 and CXCR2, namely granulocyte chemotactic protein-2/CXCL6, epithelial cell-derived neutrophil attractant-78/CXCL5, growth-related oncogene-α,β,γ/CXCL1,2,3, and neutrophil-activating peptide-2(data not shown). Because blocking of CXCL8 showed only a partial inhibition of neutrophil migration, we hypothesize that CXCL8⁺ T cells produce an additional chemotactic factor that acts most likely via CXCR2. The responsible chemoattractant has not been identified yet. In addition to chemotaxis, the SN also increased adherence and had an antiapoptotic effect, enhancing the survival of neutrophils. Reduction of apoptosis was most likely due to CXCL8⁺ T cell-derived GM-CSF and/or IFN-γ, which have been previously shown to prolong neutrophil survival (28, 42).

In psoriasis, activation of keratinocytes by T cells was suggested to lead to a neutrophil inflammation-boosting loop that may explain the local acute inflammatory changes of pustular psoriasis (43). In AGEP, tissue cells such as keratinocytes also produce CXCL8 and may contribute to the recruitment of neutrophils as well. This CXCL8 production by keratinocytes might be induced by the combined release of IFN-γ and GM-CSF, and the occasional high levels of TNF-α by CXCL8⁺ T cells in AGEP skin (4, 44, 45). IFN-γ and TNF-α are capable of activating keratinocytes, which in response produce CXCL8 and express ICAM-1 on their surface (4, 46). This further facilitates the recruitment of T cells and neutrophils to the inflamed skin. Thus, this tissue cell-derived CXCL8 production is clearly a secondary response, as it is dependent on the persistent stimulation of specific T cells by the relevant Ag, which in AGEP is often a drug. The important role of an antigenic stimulus is also well documented by the natural course of AGEP, as cessation of the incriminated drug leads to a rapid disappearance of the pustules (11, 14, 47), as well as by the sequence of events observed in patch tests, in which the drug-induced T cell activation and recruitment lead to vesicle and later pustule formation (6, 12, 13).

CD4⁺ T cells capable of producing CXCL8 in response to PMA/ionomycin are also present in the peripheral blood of healthy individuals (∼2.5%). Interestingly, >75% of these CXCL8⁺ cells were TNF-α producers, but failed to secrete IFN-γ, IL-4, IL-5, IL-10, and IL-13. The relevance and rather unique function of such CXCL8⁺ T cells are also underlined by an impressive report of a patient who developed disseminated pustulosis in the frame of an erythrodermic cutaneous T cell lymphoma. These tumor cells produced high amounts of CXCL8, which was seen as reason for the unusual clinical and histopathological presentation (48). Clearly, further work is needed to define the role of these CXCL8⁺ T cells in the normal immune response and the conditions under which these cells expand and may cause chronic inflammatory diseases with prominent and constant neutrophil involvement, such as pustular psoriasis, Sweet’s syndrome, Behçet’s disease, synovitis-acne-pustulosis-hyperostosis-osteitis syndrome, and others, some of which show a clear HLA association, suggesting T cell involvement (43, 44, 49, 50, 51, 52, 53, 54, 55, 56).

In conclusion, our data suggest that CXCL8-producing effector memory T cells can orchestrate neutrophilic infiltration in the skin and ensure neutrophil survival, both of which contribute to sterile pustular eruptions found in AGEP patients. This mechanism may also be relevant for other T cell-mediated diseases in which neutrophilic inflammations represent an important clinical and pathological feature (44, 54, 55). Because the cytokine-based classification into Th1 or Th2 subsets does not include T cells regulating neutrophil-rich inflammations, CXCL8⁺ T cells may be considered as a functionally distinct T cell subset leading to a unique pathology.

We are grateful to the patients AP, EB, JS, and US, who collaborated in this study. We thank C. Burkhart, S. Schmid, F. Altznauer, J. Depta, and O. Engler for fruitful discussions; I. Strasser and J. Tilch for their excellent technical assistance; and M. Buckwalter for critical reading of the manuscript.