Research ArticleALLERGY

Filaggrin inhibits generation of CD1a neolipid antigens by house dust mite–derived phospholipase

See allHide authors and affiliations

Science Translational Medicine  10 Feb 2016:
Vol. 8, Issue 325, pp. 325ra18
DOI: 10.1126/scitranslmed.aad6833

Bringing atopic dermatitis up to scratch

Targeted therapies are transforming medicine, but complex diseases such as atopic dermatitis are difficult to target. Now, Jarrett et al. report a mechanism that links two contributors to atopic dermatitis pathogenesis—cutaneous inflammation and barrier dysfunction. They found that house dust mite allergen phospholipase (PLA2) can induce neolipid antigens in human skin. These antigens can then be presented by the nonclassical MHC family member CD1a to CD1a-restricted T cells, which contribute to inflammation. The skin barrier protein filaggrin can inhibit PLA2 and decrease this inflammation. Indeed, individuals with filaggrin mutations experience severe atopic dermatitis. These data suggest that barrier dysfunction and inflammation may be linked, and support PLA2 as a target for atopic dermatitis.

Abstract

Atopic dermatitis is a common pruritic skin disease in which barrier dysfunction and cutaneous inflammation contribute to pathogenesis. Mechanisms underlying the associated inflammation are not fully understood, and although Langerhans cells expressing the nonclassical major histocompatibility complex (MHC) family member CD1a are known to be enriched within lesions, their role in clinical disease pathogenesis has not been studied. We observed that house dust mite (HDM) allergen generates neolipid antigens presented by CD1a to T cells in the blood and skin lesions of affected individuals. HDM-responsive CD1a-reactive T cells increased in frequency after birth in individuals with atopic dermatitis and showed rapid effector function, consistent with antigen-driven maturation. In HDM-challenged human skin, we observed phospholipase A2 (PLA2) activity in vivo. CD1a-reactive T cell activation was dependent on HDM-derived PLA2, and such cells infiltrated the skin after allergen challenge. Moreover, we observed that the skin barrier protein filaggrin, insufficiency of which is associated with atopic skin disease, inhibited PLA2 activity and decreased CD1a-reactive PLA2-generated neolipid-specific T cell activity from skin and blood. The most widely used classification schemes of hypersensitivity suggest that nonpeptide stimulants of T cells act as haptens that modify peptides or proteins; however, our results show that HDM proteins may also generate neolipid antigens that directly activate T cells. These data define PLA2 inhibition as a function of filaggrin, supporting PLA2 inhibition as a therapeutic approach.

INTRODUCTION

Atopic and allergic diseases affect up to 20 to 30% of the population and have considerable associated morbidity, mortality, and health economic burden (1). Atopic dermatitis (AD) is a disease with complex genetic and environmental susceptibility factors. Although it is known that many loci are involved (2), null mutations in the gene encoding filaggrin are reproducibly associated with moderate to severe clinical disease (3, 4). Filaggrin is expressed in keratinocytes and functions in skin barrier, cutaneous pH regulation, hydration, and antimicrobial protection. As keratinocytes proceed through cornification, profilaggrin is cleaved into filaggrin monomers, which can be incorporated into the lipid envelope and exposed to the intercellular space. Thus, a general “outside-in” hypothesis predicts that filaggrin disruption acts to allow exogenous immune stimuli to enter the skin and activate immune responses (3).

Up to 50% of individuals with homozygous filaggrin-null mutations and up to 5 to 10% of healthy European filaggrin-null carriers do not have associated AD, suggesting that other modifying genetic and environmental factors are important, including those that are directly involved in immune responses, such as FcεR1 and IL-4R (interleukin-4 receptor) genes (5). Indeed, many cytokines modify filaggrin and antimicrobial peptide expression (610), and anti–IL-4Rα therapy has shown significant efficacy in AD treatment (11).

Separately, there is convincing evidence that antigen-specific reactivity to environmental challenge has a role in the pathogenesis of AD. For example, the pathology of AD shares many features with classic delayed-type hypersensitivity, including epidermal edema and a dominant T cell inflammatory infiltrate. About 80% of individuals with AD have elevated serum immunoglobulin E (IgE), which recognize proteins derived from one or more ubiquitous environmental allergens, including house dust mite (HDM), animal dander, pollens, and fungal allergens (for example, Aspergillus spp.) (12). Allergen peptide–specific type 2 cytokine–producing T cells have been documented in many studies to be present in the peripheral blood of affected individuals (1317). Recently, type 2 innate lymphoid cells (ILC2) have been shown to be enriched in lesional AD skin and to infiltrate after HDM challenge (18). The production of type 2 cytokines by ILC2 is enhanced by PGD2 (prostaglandin D2), IL-33, IL-25, and TSLP (thymic stromal lymphopoietin) and increases in the setting of reduced E-cadherin, which is believed to be mediated through loss of KLRG1 (killer cell lectin-like receptor G1) inhibitory signals (18, 19). Overall, these data suggest that type 2 cytokine production by a number of cell types, including T cells and ILC2, can compound barrier insufficiency and contribute to inflammation, supporting the alternate “inside-out” hypothesis. However, there remains considerable debate about the relative roles of barrier function and cutaneous inflammation in AD pathogenesis.

CD1a is a major histocompatibility complex (MHC)–like antigen-presenting molecule and is highly expressed by Langerhans cells (LCs) of the epidermis and by a subset of dermal dendritic cells, as well as subsets of dendritic cell populations at other sites, including lung, gut, and genital mucosa (2024). CD1a presents self (25) and foreign (26) lipids to T cells, so the abundant expression of CD1a on epidermal LCs would be compatible with the detection of skin barrier compromise through binding endogenous or exogenous lipids for presentation to CD1a-reactive T cells (27). Indeed, lesional cutaneous atopic tissue carries an altered lipid profile (2832) that is a candidate for influencing CD1a-mediated T cell activation, and CD1a+ cells are enriched within AD lesions (33). Recently, studies show that CD1a-reactive T cells circulate at far higher frequencies than previously considered and can infiltrate normal human skin (27, 34, 35). CD1a autoreactive T cells produce cytokines that contribute to skin disease, such as IL-22 and IFNγ (interferon γ) (27), and although other CD1 proteins use complex intracellular processing pathways, CD1a directly captures and displays extracellular lipids with few specialized loading requirements (27, 3543). Together, the natural accumulation of autoantigens, CD1a proteins, and CD1a autoreactive T cells points to a natural organ-specific function in skin, but insights into clinical diseases are limited. Recently, we have identified that fatty acids generated by phospholipase activity in wasp and bee venom can be recognized by CD1a-reactive T cells (44). Further, pollen-derived phospholipids have been implicated as targets for lipid-specific T cells, including CD1d- and CD1a-reactive T cell clones (45). On the basis of these findings, we considered that CD1a might play a role in AD. Specifically, given the enrichment of CD1a-expressing cells and altered lipids in lesional AD skin, we sought to test the hypothesis that CD1a-reactive lipid-specific T cells contribute to the human response to HDMs and AD.

RESULTS

HDM is recognized by CD1a-reactive T cells

To determine whether HDM could be recognized by CD1a-reactive T cells isolated ex vivo from individuals with AD, we incubated polyclonal T cells with K562 target cells transfected with CD1a (K562-CD1a) in the presence or absence of whole HDM extract. K562 cells are HLAlow and thereby largely bypass alloreactivity and allow parallel testing with T cells from many unrelated donors under equivalent conditions. We measured both type 1 and type 2 cytokine production because both have been implicated in disease pathogenesis (4649). As seen previously (27, 35, 44), we observed a trace response to K562 cells at a rate of ~1 in 10,000. Above this trace background rate, we detected large increments in IFNγ response to K562-CD1a cells, which likely reflects the presence and activation of autoreactive T cells recognizing CD1a and endogenous K562-derived lipids (Fig. 1A, donor R9500, table S1). Further, we noted significantly (P < 0.0001) increased activation by K562-CD1a target cells when pulsed overnight with HDM extract. No response above background was seen in response to control antigen-presenting cells (APCs) transfected with vector alone, demonstrating that the response to HDM extract is CD1a-mediated. Polyclonal T cells from the same donor also produced GM-CSF (granulocyte-macrophage colony-stimulating factor) and IL-13 in response to HDM products with a similar pattern, except that the rate of cellular response was particularly high for GM-CSF, reaching to more than 0.5% of all T cells in some cases (Fig. 1B). The response was inhibited by an anti-CD1a antibody, but not by isotype control (Fig. 1C). Responses were in clinically and physiologically relevant dose ranges, because titration to HDM extract (0.07 μg/ml) (Fig. 1D), which is about 0.3% of doses administered in vivo during maintenance immunotherapy (up to 21 μg) (50), caused detectable T cell responses.

Fig. 1. Circulating CD1a-reactive HDM-responsive T cells produce IFNγ, GM-CSF, and IL-13.

T cells were isolated by CD3 magnetic cell sorter (MACS) beads from donor peripheral blood mononuclear cells (PBMCs) (R9500) and incubated overnight with CD1a-transfected K562 (CD1a) or untransfected K562 cells pulsed with HDM extract. (A to D) IFNγ (A), GM-CSF (B, left), and IL-13 (B, right) productions were measured by enzyme-linked immunospot (ELISpot) in the absence or presence of anti-CD1a antibody (C) and at different HDM concentrations (D). (E) CD1a-expressing K562 (left), in vitro–derived mDC (middle), or LC-like cells (right) were pulsed with HDM extract overnight and incubated with autologous peripheral blood T cells from donor R2. IFNγ production was measured by ELISpot in the presence or absence of anti-CD1a antibody. Data are representative of at least three donors for each experiment. Bars represent SE. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001, Student’s t test.

K562-CD1a represents an engineered APC that has the advantage of use with any donor to test the interdonor reproducibility of the response in a defined system, but these transformed cells might not mimic the antigens or costimulatory processes of the two native CD1a+ APCs: mDCs (myeloid dendritic cells) and LCs. We differentiated monocytes with GM-CSF and IL-4 to produce monocyte-derived mDCs and also activated cells in vitro with cytokines to mimic LCs (in vitro LCs, fig. S1) (5154). Similar to previous studies that compared K562 cells, mDCs, and LC-like cells side by side, we found that both mDCs and LC-like cells mediated the HDM response of polyclonal autologous T cells in a CD1a-reactive manner (Fig. 1E). Overall, our data show that HDM-responsive CD1a-reactive T cells exist in the peripheral blood of AD patients and healthy individuals at high frequencies and produce IFNγ, GM-CSF, and IL-13 in response to HDM challenge at relevant doses.

HDM-responsive CD1a-reactive T cells are enriched in blood and skin of AD patients

Next, we examined responses to HDMs in a larger cohort of healthy adult donors and individuals with AD, and compared these to responses in cord blood of neonates, which represents a control for naïve polyclonal T cells. We observed significantly higher CD1a-reactive ex vivo T cell IFNγ responses to HDM in individuals with AD compared to healthy non-atopic controls (Fig. 2A). The frequencies of IFNγ-producing cells were significantly (P < 0.0001) higher in adult donors than in cord blood samples, consistent with an acquired antigen-driven T cell expansion (Fig. 2A). In addition, the production of IL-13 by CD1a-reactive T cells was also significantly (P < 0.001) elevated ex vivo, confirming type 1 and type 2 cytokine induction (Fig. 2B). Using dual-color ELISpot, we showed that although most of the type 1 and type 2 cytokine–producing cells are unique subsets, a mean of 7% of CD1a-reactive HDM-responsive T cells could produce both IFNγ and IL-13 (fig. S2A). However, we do not observe an increase in the IFNγ- or IL-13–producing HDM-responsive CD1a-reactive T cells in patients with psoriasis (fig. S2B). Given the association between filaggrin-null mutations and moderate to severe AD, we genotyped the patients and controls for the two most common mutations found in Europeans (2282del4 and R501X). Frequencies of the HDM-responsive CD1a-reactive T cells were significantly increased in those individuals with filaggrin-null mutations compared to those with wild-type filaggrin (contingency χ2 = 4.38, P < 0.05; fig. S3A). Furthermore, the percentage of IL-13–producing CD1a-reactive HDM-responsive T cells significantly correlated with disease severity (R2 = 0.445, P < 0.001), and there was a significant correlation between fold increase in HDM-responsive T cells and total IgE (R2 = 0.588, P < 0.05) and HDM-specific IgE (R2 = 0.502, P < 0.05). Overall, these data showed that HDM-responsive, CD1a-reactive T cells increase in frequency in adults, show effector function, and are enriched in the circulation of patients with AD.

Fig. 2. HDM-responsive CD1a-reactive T cells are enriched in blood and skin of AD patients.

(A and B) T cells derived from the peripheral blood of healthy controls (HC), patients with AD, or from cord blood were incubated with CD1a-transfected K562 or untransfected K562 cells in the presence or absence of HDM extract. IFNγ (A) and IL-13 (B) productions were measured by ELISpot and expressed as percentage of responding T cells (A: n = 30 healthy controls, 24 AD patients, 10 cord blood; B: n = 20 healthy controls, 17 AD patients, 10 cord blood). Autoreactive is the response to CD1a-K562 in the absence of HDM. (C) Skin blister T cells from donor R229 were incubated with CD1a-transfected K562 or untransfected K562 cells in the presence or absence of HDM extract. IFNγ (left), GM-CSF (middle), and IL-13 (right) productions were measured by ELISpot. Data are representative of at least three separate donors for each experiment. (D) Example of skin suction blister raised after 60 min of 200 mmHg negative pressure. (E) Skin blister T cells were isolated from unchallenged skin of healthy controls (n = 5) or patients with AD (n = 4) and incubated with CD1a-transfected or untransfected K562 cells in the presence or absence of HDM extract. IL-13 production was measured by ELISpot and expressed as percentage of CD1a-reactive T cells. Bars represent SE. n.s., not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001, Student’s t test.

To examine whether HDM-responsive cells were also present within skin from patients with AD and from healthy controls, we used skin suction blisters to isolate T cells from skin. This method uses low-pressure, sustained suction for 60 min to produce extracellular blister fluids that are captured for immunological and biochemical analysis and can be performed before or after antigen challenge (18). The approach has the advantage over conventional skin biopsies in that cells and fluid can be aspirated from the skin and used in functional assays, without the need for prolonged processing with potentially confounding treatments, such as dispase and collagenase. We observed that HDM-responsive CD1a-reactive T cells are present in skin and produce IFNγ, GM-CSF, and IL-13 (Fig. 2, C and D). Furthermore, the IL-13–producing T cells are enriched in the skin of patients with AD compared to the skin of healthy controls (Fig. 2E), and the frequency correlated with SCORing Atopic Dermatitis (SCORAD) disease severity (R2 = 0.67, P < 0.01), showing that T cell infiltration into AD lesions correlates with both disease outcome and type 2 cytokine production in humans ex vivo. To investigate the CD1a-reactive immune response in vivo, we challenged the skin intraepidermally with 0.7 μg of HDM extract and assessed skin infiltration of HDM-responsive T cells at 24 hours. We chose a dose at <5% of the amount administered during the maintenance phase of subcutaneous immunotherapeutic approaches (typically up to 21 μg) to assure safety and attempt to mimic low-dose exposures (50). We noted skin infiltration of IFNγ-, GM-CSF–, and IL-13–producing HDM-responsive T cells at 24 hours, suggesting that the cells infiltrate early and may therefore contribute to the ensuing inflammation (Fig. 3A and fig. S3B). Furthermore, in those with AD, we observed significantly greater infiltration of IFNγ-producing HDM-responsive CD1a-reactive T cells after HDM challenge than in healthy controls (Fig. 3B, left panel). In contrast, there was no enrichment for varicella zoster virus (VZV)–specific T cells in the skin after HDM challenge, suggesting that the enrichment is specific to HDM-responsive T cells (Fig. 3B, right panel).

Fig. 3. HDM-responsive CD1a-reactive T cells infiltrate skin after HDM challenge.

Skin blister T cells were isolated from donor R4 24 hours after HDM skin challenge, expanded, and incubated with CD1a-transfected or untransfected K562 cells in the presence or absence of HDM extract. (A) IFNγ (left), GM-CSF (middle), and IL-13 (right) production by donor R4 cells were measured by ELISpot. Data are representative of at least three separate donors for each experiment. (B) (Left panel) Overall frequencies of HDM-responsive CD1a-reactive IFNγ-producing T cells infiltrating human skin after saline or HDM skin challenge were compared between healthy controls (n = 4) and AD patients (n = 8). Autoreactive refers to responses to unpulsed K562-CD1a cells. (Right panel) Skin blister cells derived from HDM-challenged or unchallenged skin were incubated with live attenuated VZV, and IFNγ production was measured by ELISpot. (C) Concentrations of type 2 cytokines were measured in skin blister fluid by multiplex bead array after saline (nil) or HDM skin challenge in healthy controls (n = 8) or atopic (n = 16) individuals. Bars represent SE. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001, Student’s t test.

As described above, type 2 cytokines can influence filaggrin and antimicrobial peptide expression in the skin. We investigated the type 2 cytokine content of skin blisters derived from patients with AD and from healthy controls with sampling at nonlesional skin, and after HDM challenge. These studies showed that after HDM challenge of nonlesional skin in AD patients, the concentrations of type 2 cytokines IL-4, IL-5, IL-13, and GM-CSF within skin blister fluid were significantly increased compared to those from healthy donors (Fig. 3C).

Overall, these data demonstrate that HDM-responsive CD1a-reactive T cells are present in the blood and skin of healthy donors but are enriched in the blood and skin of patients with AD. Furthermore, after intraepidermal HDM allergen challenge, the T cells infiltrate rapidly and associate with the production of type 2 cytokines in vivo.

HDM-responsive CD1a-reactive T cell responses are explained by phospholipase A2 activity

To identify the antigenic substance in HDMs, we first treated HDM extract with chloroform and methanol to recover protein-enriched aqueous and lipid-enriched organic fractions before testing the differential capacity to sensitize CD1a-expressing K562 cells for recognition by T cells, and unexpectedly observed responses to the protein-enriched fractions rather than the lipid-enriched fractions (Fig. 4A). This counterintuitive result could be explained by other experiments that were recently reported (44), which showed that the origin of CD1a-mediated responses to bee venom also derived from proteinaceous fractions of venom, and were identified as phospholipases. Phospholipases generated antigenic free fatty acids or lysolipids from nonantigenic phosphoacylglycerols. We posited that a parallel mechanism existed here and tested it by examining phospholipase A2 (PLA2) biochemical activity in vivo and in vitro using colorimetric thiol detection after release from diheptanoyl thio-phosphatidylcholine (PC) substrate. There was PLA2 activity measured in all skin blister samples, which significantly increased after HDM challenge (Fig. 4B). Furthermore, HDM-induced PLA2 activity within skin blister fluid was significantly inhibited in the presence of the known PLA2 inhibitor manoalide (Fig. 4C). We next demonstrated that HDM extract contains PLA2 activity in vitro, which could be inhibited by manoalide and heat inactivation (Fig. 4D).

Fig. 4. HDM extract contains PLA2 activity in vivo and in vitro.

(A) T cells were isolated by CD3 MACS beads from AD and healthy donor PBMCs and incubated overnight with CD1a-transfected K562 or untransfected K562 cells pulsed with HDM total extract, either aqueous protein phase or lipid phase. IFNγ production was measured by ELISpot. (B and C) PLA2 activity in saline or HDM-challenged skin blister fluid was detected by measuring free thiol release in the presence of the diheptanoyl thio-PC substrate in the absence (B) or presence (C) of 10 μM of the PLA2 inhibitor manoalide. (D) PLA2 activity in HDM extract in vitro was detected by measuring free thiol release in the presence of the diheptanoyl thio-PC substrate and manoalide and after heat inactivation. Data are representative of at least three separate experiments. Bars represent SE. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001, Student’s t test.

Heat inactivation of HDM extract or treatment with the PLA2 inhibitor manoalide abrogated the CD1a-dependent recognition of HDM-pulsed K562-CD1a cells but did not affect the autoreactive T cell response (Fig. 5A) or viral-specific T cell responses (fig. S4). Together, these suggest that the CD1a-reactive HDM responses are secondary to PLA2, most likely through generation of neolipid antigens. We next sought to determine whether skin-derived HDM-responsive CD1a-reactive T cells were also dependent on PLA2 activity. Figure 5B shows that IFNγ and GM-CSF production by skin T cells derived ex vivo after intraepidermal HDM challenge to human skin are also inhibited by heat inactivation of the HDM or by treatment with manoalide. Furthermore, we show that skin HDM-responsive CD1a-reactive T cells are also able to respond to purified bee venom PLA2 (44), another recognized PLA2-containing allergen (Fig. 5C). This result points to shared pathways of skin inflammation despite different depths of natural antigen delivery by venom and HDM. Overall, HDM-derived PLA2 has the capacity to generate neolipid antigens for recognition by CD1a-reactive T cells. The lack of response with the lipid fraction of HDM extract suggests that skin-derived lipid sources represent the principal substrates of HDM PLA2.

Fig. 5. HDM-derived PLA2 generates neolipid antigens for presentation by CD1a to blood and skin T cells.

T cells were isolated by CD3 MACS beads from AD and healthy donor PBMCs and incubated overnight with CD1a-transfected K562 or untransfected K562 cells pulsed with HDM total extract. (A) IFNγ production was measured by ELISpot after HDM heat inactivation (hiHDM) (left) or overnight incubation with a dose titration of the PLA2 inhibitor manoalide (right). Skin blister T cells were isolated 24 hours after HDM skin challenge, expanded, and incubated with CD1a-transfected or untransfected K562 cells in the presence or absence of HDM extract that had been heat-inactivated or incubated with manoalide. (B) IFNγ (left) and GM-CSF (right) productions were measured by ELISpot. (C) Skin blister T cells were isolated 24 hours after HDM skin challenge, expanded, and incubated with CD1a-transfected or untransfected K562 cells in the presence or absence of HDM extract or purified bee venom PLA2 (1 μg/ml). IFNγ production was measured by ELISpot. Data are representative of at least three separate experiments. Bars represent SE. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001, Student’s t test.

Filaggrin inhibits HDM PLA2 and CD1a-reactive T cell responses

Given that filaggrin insufficiency is associated with moderate to severe AD and that we have shown an association between CD1a-reactive T cell responses to HDM and filaggrin-null mutations, we designed experiments to test the hypothesis that filaggrin directly contributes to the CD1a-reactive T cell response. We investigated the possibility that filaggrin itself can directly inhibit PLA2 activity and observed that filaggrin monomers efficiently inhibited HDM PLA2 biochemical activity at levels equivalent to those present in the stratum corneum in vivo (Fig. 6A) (55). Furthermore, recombinant filaggrin monomers inhibited the PLA2 activity observed within skin blister fluid after intraepidermal challenge with HDM (Fig. 6B). We next investigated whether filaggrin could inhibit cytokine production by HDM-responsive CD1a-reactive T cells and showed that recombinant filaggrin monomers significantly inhibited IFNγ, IL-13, and GM-CSF production by T cells derived from blood ex vivo (Fig. 6C). However, we did not show filaggrin inhibition of autoreactive T cells or viral-specific T cells (fig. S4), ruling out a nonspecific toxic effect of the filaggrin on T cells. Last, we demonstrated that the filaggrin monomers inhibited GM-CSF production by T cells derived ex vivo from skin after HDM skin challenge (Fig. 6D). These data suggest that filaggrin may provide barrier function status signals to the innate and adaptive immune responses through effects on PLA2.

Fig. 6. Filaggrin inhibits HDM PLA2 activity and inhibits responses of HDM-responsive CD1a-reactive T cells isolated from blood and skin.

(A) PLA2 activity in HDM extract in vitro was detected by measuring free thiol release in the presence of the diheptanoyl thio-PC substrate, 10 μM PLA2 inhibitor manoalide, and human recombinant filaggrin (1 μg/ml). (B) PLA2 activity in HDM-challenged skin blister fluid was detected by measuring free thiol release diheptanoyl thio-PC substrate in the absence or presence of filaggrin. (C) T cells were isolated by CD3 MACS beads from AD and healthy donor PBMCs and incubated overnight with CD1a-transfected K562 or untransfected K562 cells pulsed with HDM total extract. GM-CSF, IL-13, and IFNγ productions were measured by ELISpot in the presence or absence of filaggrin. (D) Skin blister T cells were isolated 24 hours after HDM skin challenge and incubated with CD1a-transfected or untransfected K562 cells in the presence or absence of HDM extract, filaggrin, or anti-CD1a antibody. Data are representative of at least three separate experiments. Bars represent SE. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001, Student’s t test.

DISCUSSION

Although CD1a protein is expressed at constitutively and at extraordinarily high density on LCs that form a broad network in the epidermis, and CD1a autoreactive T cells normally enter the skin in large numbers, there are no studies addressing the specific role of CD1a-reactive T cells in human skin disease. Here, we have shown that HDM generates CD1a ligands for recognition by T cells. Such HDM-responsive CD1a-reactive T cells are enriched in the blood and skin of individuals with AD and correlate with IgE and disease severity, and are significantly elevated in the presence of filaggrin-null mutations. Furthermore, the CD1a-reactive T cells infiltrate the skin 24 hours after HDM challenge, which associates with type 2 cytokine production and PLA2 activity in vivo. We showed that the HDM responsiveness of CD1a-reactive T cells was explained by the presence of PLA2 activity within HDM, and that the PLA2 activity could be inhibited by recombinant filaggrin. These studies identify a pathway of atopic skin inflammation, in which neolipid antigens are generated from the skin by HDM-derived PLA2 for CD1a-mediated presentation to T cells. Loss of filaggrin inhibition of HDM PLA2 may provide neolipid signals to CD1a-reactive T cells that barrier compromise has occurred, with potential inflammatory sequelae.

All previous studies investigating the potential role of antigen-specific T cells in the pathogenesis of AD have focussed on peptide-specific responses that are restricted through HLA class I and class II (15, 16). The current study implicates HDM PLA2 processed skin lipids as a broad antigen class that should be added as potential candidate antigens recognized by T cells and relevant to clinical atopic disease. LCs are enriched in AD lesions (33), where there is a dysregulated lipid profile (2832), consistent with a potential role of such cells in presenting lipid to T cells. It is of interest that many known allergens are lipid-binding proteins, which may become targets of peptide-specific T cell and IgE responses when conjugated to their lipid cargo. Langerin is a C-type lectin expressed by subsets of dendritic cells including LCs, and is thought to contribute to lipid loading of CD1a; it is of interest that langerin has recently been identified in a genome-wide association study of AD (56). The widely taught Gell and Coombs classification system summarizes a large literature that T cells play a functional role in type IV hypersensitivity responses in humans, even though the immunogens are often not typical peptide antigens for T cells (57). The mechanisms by which nonpeptide antigens lead to a T cell–mediated response can involve haptenization of proteins (58), but it is unknown if this is generally true. In most cases, the mechanisms for T cell stimulation by small molecules and other nonpeptide antigens are unclear. The CD1 system is a particularly attractive candidate to mediate response to nonpeptide immunogens, because it evolved to present lipids and small molecules to T cells (59, 60). Through its expression on LCs and a subset of dermal dendritic cells, CD1a protein is well placed to sample nonpeptide antigens that are exposed to the skin. The results presented here suggest a broader model of Gell and Coombs hypersensitivity where neolipid recognition by CD1a-reactive T cells is associated with clinical atopic disease. There may therefore exist a multi-hit process where a CD1a-reactive response may be part of a barrier sensing system that, if activated, can induce other adaptive immune responses and cutaneous inflammation. This is germane to animal models, in which cutaneous stress responses enhance sensitization (61). The data may also provide insights to why only certain environmental challenges are particularly allergenic. Such proallergenic sources, including HDM, may preferentially activate a barrier distress system through generation of inflammatory lipids, leading to immune responses to coexistent proteins and subsequent clinical disease. These processes may be compounded by genetically determined or acquired filaggrin insufficiency.

CD1a protein is a member of the group 1 CD1 family, along with CD1b, CD1c, and CD1e. The group 1 family is not present in mice, yet each member has a distinct cytoplasmic domain, intracellular trafficking, and tissue distribution, which suggests functional specialization (59). CD1a-autoreactive T cells were found to infiltrate human skin where they could recognize skin-derived self-lipids, including fatty acids, presented by primary CD1a-expressing APCs (35). We have recently shown that wasp- and bee venom–derived phospholipase can generate self-lipids for presentation by CD1a protein, suggesting shared pathways of skin inflammation, albeit with differing clinical phenotypes dependent on depth of antigen delivery (44). Furthermore, phospholipase activity is required for generation of CD1d-restricted natural killer T (NKT) cell ligands in a model of hepatitis virus infection (62, 63), and thymic PLA2 contributes to the control of NKT cell selection (64). It is therefore possible that other CD1s may present shared antigens and contribute to inflammatory skin disease in the presence or absence of CD1a deficiency (65). However, no studies have linked CD1a to biological processes relevant in human disease.

Secretory PLA2 cleaves phospholipid to lysophospholipid and the sn-2 acyl chain, and although mechanisms have not been clear, it has long been implicated as having a role in atopic disease (6673). Although phospholipases are likely to have many roles, here, we show that PLA2-derived lipid products are presented by CD1a for recognition by T cells in the skin. Indeed, we also show that recombinant filaggrin can inhibit the PLA2 activity of HDM and can inhibit the generation of CD1a neolipid ligands for presentation to T cells. This is a hitherto unappreciated function of filaggrin, which may help link the presence of filaggrin insufficiency to cutaneous inflammation. The data also provide a potential resolution to the seemingly contrasting inside-out and outside-in hypotheses of AD, where instead of two independent possibilities, filaggrin can act both in a barrier function/hydration capacity and also as a direct inhibitor of cutaneous lipid–specific immune responses. However, clinical disease may depend on a multi-hit process, where filaggrin insufficiency combines with modulations in innate and adaptive immune responses. By inhibiting downstream innate and T cell effector functions, for example, through anti–IL-4Rα, acquired down-regulation of filaggrin may be reversed, leading to enhanced filaggrin inhibition of PLA2 and less barrier distress signals to CD1a-reactive T cells. Twenty percent to 40% of individuals with moderate to severe AD have filaggrin-null mutations, yet acquired filaggrin insufficiency and barrier impairment are common (3, 9). This supports the findings presented herein in which specific aspects related to the downstream immunological events are also important in contributing to the filaggrin insufficiency and compounded atopic cutaneous inflammation through a multi-hit model of disease pathogenesis (6, 9). The data further substantiate the pursuit of therapeutic strategies that modulate relevant immune responses.

Given that PLA2 generation of CD1a ligands has now been shown to be present in three allergens of relevance to humans (44), namely, wasp venom, bee venom, and HDM, the data suggest that this is part of a broader hypersensitivity system. Indeed, phospholipases are known to be present in many other allergens including pollens and fungal allergens (74, 75), and therefore, activation of this pathway may facilitate the allergenic process. By producing type 2 cytokines, CD1a-reactive lipid-specific T cells may lead to down-regulation of filaggrin and antimicrobial peptide expression and thus compound physical and antimicrobial barrier dysfunction. Furthermore, they may license skin dendritic cells to amplify peptide-specific T helper cell 2 (TH2) responses and subsequent IgE generation. For example, it is of interest that GM-CSF and IL-4 are routinely used to mature monocytes toward cells with CD1 and class I/II antigen-presenting capacity. If CD1a-reactive responses do contribute to initiation of the allergic process, then we might predict that allergens delivered to anatomical sites that do not contain APCs with high levels of CD1a protein would lead to differing systemic responses. This is indeed the case because wasp venom, bee venom, and grass pollen subcutaneous immunotherapy are all known to be highly effective (76, 77). In contrast to the epidermis and dermis, subcutaneous tissue has few CD1a-expressing cells. The findings therefore have potential therapeutic implications. It is of interest that corticosteroids are known to inhibit PLA2 (78), and it may be that in the skin, this is an important mechanism for controlling CD1a-reactive T cell activity. The development of PLA2 inhibitors that target individual relevant allergen phospholipases may enhance treatment efficacy while reducing side effects of broad host PLA2 inhibition seen with current corticosteroids.

Although skin suction blisters offer access to human skin fluid and cells directly ex vivo without the need for further processing, they do add a potential limitation of the study. They are time-consuming for donors, and so participant numbers become limiting. Furthermore, skin suction blisters inevitably introduce physical trauma to the skin and so it is important to use control comparisons of unchallenged or nonlesional skin when examining challenged or lesional skin, respectively. The current study is a cross-sectional analysis of affected individuals, and it will be important to validate the findings in other cross-sectional cohorts and to examine changes longitudinally. Last, although it is recognized that translational work has to pass through stages involving human subjects at some point during development, it can be difficult to prove causality in humans. Human skin antigenic challenge does offer temporal associations with clinical and immunological findings, lending support of causality, but CD1a transgenic models and human skin grafts in immunodeficient models may offer further evidence in the future.

In conclusion, we identify a pathway of human skin inflammation where HDM-derived PLA2 generates neolipid antigens for presentation to CD1a-reactive T cells. By also defining a function of filaggrin in inhibiting PLA2, we are able to potentially unify the conflicting outside-in and inside-out hypotheses of AD. The data would support therapeutic approaches to inhibit allergen-derived PLA2 activity, together with treatments that target the downstream immunological effector pathways.

MATERIALS AND METHODS

Study design

The study was a laboratory analysis of human blood and skin T cell responses designed to test the hypothesis that CD1a-reactive lipid-specific T cells contribute to the human response to HDM and AD. AD was diagnosed according to the UK refinements of the Hanifin and Rajka diagnostic criteria, and adult participants were only excluded if on systemic immunosuppression or topical calcineurin inhibitors. Participants were recruited sequentially; blinding and randomization were not required because there was no intervention. Sample size was determined on the basis of previous studies of CD1a-reactive T cell response frequencies in humans (79). All experiments were replicated as presented in the figure legends.

Isolation of human T cells

PBMCs were isolated from healthy adult donors and AD patients under local ethics approval (09/H0606/71). AD was diagnosed using the UK refinements of the Hanifin and Rajka diagnostic criteria, and disease severity was assessed using SCORAD. Donors aged 18 to 67 years were recruited with disease severity SCORAD ranging from 5 to 70. Patients were using topical corticosteroids but were not on systemic immunosuppression or topical calcineurin inhibitors. Adult (18 to 67 years) patients were recruited with moderate severity psoriasis who were not on systemic therapy. T cells were purified from ficollized PBMCs using CD3 MACS beads (Miltenyi Biotec). CD1a reactivity was assessed by IFNγ ELISpot (Mabtech AB). ELISpot plates (Merck Millipore Corp.) were coated with anti-IFNγ, anti–GM-CSF, or anti–IL-13 antibody overnight (Mabtech AB). K562 cells were pulsed with HDM extract (7 μg/ml unless otherwise stated) or purified bee venom phospholipase (1 μg/ml) (Sigma-Aldrich) overnight and were then washed and resuspended in R5* [RPMI supplemented with 2 mM l-glutamine, penicillin (100 IU/ml), and streptomycin (100 μg/ml) plus 5% human serum]. The plates were washed six times with RPMI and blocked for 1 hour with RPMI supplemented with 2 mM l-glutamine, penicillin (100 IU/ml), and streptomycin (100 μg/ml) plus 10% human serum (R10*). T cells (20,000 to 50,000) were added per well to which 10,000 to 25,000 K562 or primary cells were added. Wells were set up in duplicate or triplicate. Phorbol 12-myristate 13-acetate (10 ng/ml) and ionomycin (500 ng/ml) were included as positive controls, and T cells alone in the absence of K562 were included as a negative control. After overnight incubation at 37°C and 5% CO2, plates were washed six times in phosphate-buffered saline (PBS)–Tween 0.05% and incubated with biotin-linked anti-IFNγ, anti–GM-CSF, or anti–IL-13 monoclonal antibody (1 μg/ml) (Mabtech AB) for 2 hours. After washing six times in PBS–0.05% Tween, the plates were incubated 1 hour further with streptavidin-alkaline phosphatase (Mabtech AB). Spots were visualized using an alkaline phosphatase conjugate substrate kit (Bio-Rad) and enumerated using an automated ELISpot reader (AID EliSpot Reader Classic, Autoimmun Diagnostika GmbH). In some experiments, anti–CD1a blocking antibody (OKT-6) (10 μg/ml), IgG1 isotype control (P3) (10 μg/ml), 10 μM manoalide, an inhibitor of PLA2 that acts by covalently modifying lysine residues (Enzo Life Sciences), or filaggrin monomers (0.5 μg/ml) were added to K562 or primary cells before addition of T cells. The frequency of a specific T cell response was determined in polyclonal T cell cultures derived from healthy controls and AD patients by calculating the mean number of spots per well in the K562-CD1a-HDM–pulsed ELISpot wells and subtracting the mean number of spots per well in the K562-EV-HDM–pulsed ELISpot wells. The frequency of CD1a-autoreactive IFNγ T cell response was determined by calculating the mean number of spots per well in the K562-CD1a unpulsed ELISpot wells and subtracting the mean number of spots per well in the K562-EV unpulsed ELISpot wells. For the VZV T cell responses, T cells were washed and resuspended in R5* and one vial (103.3 plaque-forming units per 0.5 ml) of Varilrix (live VZV vaccine, GlaxoSmithKline) was resuspended in 500 μl of R5*. Fifty microliters of the reconstituted virus was incubated with 100,000 T cells overnight. Fifty microliters of R5* was incubated with 100,000 T cells overnight as a negative control. IFNγ and IL-13 productions were determined by ELISpot.

Viral-specific T cell lines

We generated HLA-A*0201–restricted GILGFVFTL (influenza A matrix)–, NLVPMVATV (cytomegalovirus pp65)–, and GLCTLVAML [Epstein-Barr virus (EBV) BMLF1]–specific T cells by sorting from PBMCs ex vivo with relevant class I tetrameric complexes with maintenance, as previously described (80). Cells were cultured in R10* at 2 × 106 cells per well in a 24-well Costar plate. IL-2 was added to a final concentration of 100 U/ml on day 3. The cultures were restimulated after 14 days using HLA-A*0201–positive B cell lines pulsed with the appropriate peptides. For the functional assays, cells were pulsed with the peptide (200 ng/ml) and incubated with 10 μM manoalide, filaggrin monomers (0.5 μg/ml), or SUMO recombinant protein control (0.5 μg/ml) before IFNγ ELISpot as described above. ELISpots were performed using a ratio of 40,000 APCs (JY):5000 T cells.

Isolation and generation of human APCs

We generated autologous mDC-DCs using CD14+ monocytes. CD14+ human cells were isolated using MACS cell separation (Miltenyi Biotec Inc.) according to the supplier’s instructions. These cells were then differentiated in medium containing GM-CSF (50 ng/ml) and IL-4 (1000 IU/ml). After 4 days, CD1a expression was confirmed by flow cytometry, and the differentiated mDCs were used as APCs for the ELISpot assay. The in vitro mDCs were incubated with anti–HLA-ABC and anti–HLA-DR blocking antibodies (10 μg/ml) (W6/32 and L243, respectively) for 1 hour before coculture with T cells to minimize HLA-restricted responses. We generated autologous LC-like cells in vitro using CD14+ monocytes. CD14+ human cells were isolated using MACS cell separation (Miltenyi Biotec Inc.) according to the suppliers’ instructions. Briefly, such in vitro LC-like cells were prepared as previously described (51, 52) from CD14+ cells, which were cultured in six-well plates in complete medium in the presence of IL-4 (250 ng/ml; PeproTech), GM-CSF (100 ng/ml), and TGF-β1 (transforming growth factor–β1) (10 ng/ml; PeproTech). At days 2 and 4, cultures were replated in the presence of the above cytokines to generate cells that were 24.9 to 26.5% CD1a+CD207+. The in vitro LC-like cells were incubated with anti–HLA-ABC and anti–HLA-DR blocking antibodies (10 μg/ml) (W6/32 and L243, respectively) for 1 hour before coculture with T cells to minimize HLA-restricted responses.

HDM lipid extraction

HDM lipids were extracted with chloroform, methanol, and water using a modified Bligh-Dyer method (81). Briefly, HDM extract was resuspended in a solution of (4:2:1) methanol/chloroform/sample and vortexed before being heated for 30 min at 37° to 40°C. Two volumes of chloroform and three volumes of water were added, and the sample was vortexed again and centrifuged at 2 to 3000 rpm for 5 min, resulting in separation of the aqueous and organic phases. This process was repeated twice on the aqueous phase to achieve maximum yield. The organic phase was then aspirated, dried by desiccation, weighed, and dissolved in 0.5% PBS–Tween. The aqueous phase was aspirated and centrifuged at 3000 rpm for 5 min, protein-enriched precipitates were resuspended in sterile PBS, and the concentration was determined by NanoDrop.

Blister fluids

Healthy donor epidermis was injected with 0.7 μg of Dermatophagoides pteronyssinus HDM extract (ALK) or saline. After 30 min, suction was applied to the skin at 200 mmHg for 1 hour, which induced a split between the epidermis and dermis (18). Blister fluid was isolated by needle aspiration at 24 hours, and the cells were separated by centrifugation. The blister fluid phase was immediately stored at −20°C. Cytokines were measured by multiplex bead array. Blister-derived T cells were fluorescence-activated cell sorting (FACS)–sorted and expanded using the rapid expansion method. Specifically, T cells were plated out at 100 to 150 cells per well into T cell medium in a round-bottom 96-well plate and anti-CD3 (OKT3) antibody (50 ng/ml), and irradiated PBMCs and EBV-transformed B cells at 150,000 to 200,000 and 40,000 cells per well, respectively, were added. Blister T cells were examined and split regularly and on expansion were maintained at a density of about 0.5 million to 1 million cells/ml.

PLA2 biochemical activity experiments

PLA2 activity in HDM and blister fluids was detected using site-specific substrate kit (Cayman Chemical), according to the manufacturer’s instructions. In a flat-bottom 96-well plate, 10 μl (0.7 μg) of HDM extract (ALK) or blister fluid plus 5 μl of assay buffer and 10 μl of 5,5′-dithiobis-(2-nitrobenzoic acid) (DTNB) were incubated at room temperature with 200 μl of substrate solution (diheptanoyl thio-PC). For inhibitor/filaggrin studies, HDM/blister fluid was incubated with the appropriate concentration of manoalide or filaggrin for 30 min at room temperature before plating out. In the presence of PLA2, cleavage of the substrate at the sn-2 position results in release of the thiol group, which reacts with DTNB to produce a colored precipitate. This is measured with a spectrophotometer over time (415 nm) to give a measure of PLA2 activity (Bio-Rad, iMark Microplate Absorbance Reader).

Filaggrin synthesis

The expression plasmid was constructed using a sequence- and ligation-independent cloning method (82). The nucleotide sequence encoding the seventh filaggrin repeat domain was cloned into pET28 vector using Bam HI and Xho I restriction sites. The construct contains N-terminal His6-tagged SUMO protein sequence that can be enzymatically cleaved by SUMO protease. The construct was transformed into Escherichia coli BL21-CodonPlus-RIL and propagated overnight in LB liquid media containing kanamycin (50 μg/ml) and chloramphenicol (37.5 μg/ml) at 37°C. The bacterial cultures were diluted at 1:50 in autoinduction media (Formedium, AIM-Super Broth) supplemented with kanamycin and chloramphenicol and incubated for 48 hours at 18°C with shaking. The cells were harvested by centrifugation (10 min, 5000g, 4°C). The bacterial pellet was mixed with lysis buffer [10 mM tris (pH 8), 150 mM NaCl, 10 mM imidazole] supplemented with protease inhibitor cocktail and lysed by sonication. The cell lysate was clarified by centrifugation (45 min, 70,000g, 4°C). The following described purification procedure performed on ÄKTAxpress chromatography system. The supernatant was loaded on Ni-NTA Agarose column (Qiagen). Unbound material was washed from the column with lysis buffer. Enriched proteins were subjected to on-column cleavage by SUMO protease (20 μg of protease for 5 ml of resin) in elution buffer [10 mM tris (pH 8), 150 mM NaCl, 300 mM imidazole] for 8 hours at 10°C. Realized protein was further purified by combinations of desalting and Ni-NTA Agarose columns equilibrated with the lysis buffer. Flow-through fractions containing the purified protein were collected and automatically loaded into a pre-equilibrated column with buffer [10 mM tris (pH 8), 150 mM NaCl] Superdex 200 column (GE Healthcare). Fractions containing the purified filaggrin were collected and analyzed by SDS–polyacrylamide gel electrophoresis. Identical purification from cells without the expression plasmid was processed in parallel as a control. In both cases, the same fractions from gel filtration column were collected. All experiments were performed in parallel using control SUMO protein.

Filaggrin genotyping

The FLG mutations R501X and 2282del4 were genotyped using TaqMan allelic discrimination assays (Life Technologies), as previously described (3). R501X was screened using forward primer 5′-CACTGGAGGAAGACAAGGATCG-3′, reverse primer 5′-CCCTCTTGGGACGCTGAA-3′, and the probes VIC-CACGAGACAGCTC and 6-FAM-CATGAGACAGCTCC. 2282del4 was screened using forward primer 5′-CCACTGACAGTGAGGGACATTCA-3′, reverse primer 5′-GGTGGCTCTGCTGATGGTGA-3′, and the probes 6-FAM-CACAGTCAGTGTCAGGCCATGGACA and VIC-AGACACACAGTGTCAGGCCATGGACA alleles. Assays were performed in 384-well plates, with each reaction comprising 20 ng of DNA, 2.5 μl of Universal polymerase chain reaction (PCR) master mix, and 0.125 μl of 40× assay mix in a final reaction volume of 6 μl. Assays were run on an Applied Biosystems 7900HT Fast Real-Time PCR system under the following conditions: 1 cycle at 50°C for 2 min followed by 1 cycle at 95°C for 10 min, then 40 cycles of 95°C for 15 s and 60°C for 1 min. Samples heterozygous for the 2282del4 mutation were also confirmed by Sanger sequencing using the published primers RPT1P7 (5′-AATAGGTCTGGACACTCAGGT-3′) and RPT2P1 (5′-GGGAGGACTCAGACTGTTT-3′) (83). PCR conditions were 94°C for 5 min; 35 cycles at 94°C for 40 s, 57°C for 1 min, 72°C for 2 min; final extension step at 72°C for 7 min. PCR cleanup and sequencing were performed by Source Bioscience plc.

Statistics

Cohorts of healthy donors and AD patients investigated for CD1a-reactive HDM-specific responses were analyzed using one-tailed unpaired and paired t test, χ2, and Pearson’s correlation. All other polyclonal T cell responses were analyzed using unpaired t test. The number of biological replicates for each data point is included in the figure legends. Statistical analyses were performed using Prism 6 (GraphPad Software Inc.).

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/8/325/325ra18/DC1

Fig. S1. mDC and LC-like cell expression of CD1a and langerin.

Fig. S2. HDM-responsive CD1a-reactive T cells can produce IFNγ and/or IL-13 and are not enriched in patients with psoriasis.

Fig. S3. HDM-responsive CD1a-reactive T cells associate with FLG mutations.

Fig. S4. Viral-specific T cell responses in the presence of manoalide and filaggrin.

Table S1. Source data.

REFERENCES AND NOTES

  1. Acknowledgments: We thank V. Rocha and National Health Service Blood and Transplant for cord blood samples. Funding: This work was funded by the UK Medical Research Council (MRC) and National Institute for Health Research (NIHR) Biomedical Research Centre, the NIH (National Institute of Arthritis and Musculoskeletal and Skin Diseases R01 048632), and the Burroughs Wellcome Fund in Translational Medicine. R.J. is funded through a British Association of Dermatologists/British Skin Foundation/MRC Clinical Research Training Fellowship. G.O. also acknowledges the support of the NIHR Clinical Research Network and support from Janssen Pharmaceuticals. J.B.S. acknowledges the support of a Marie Curie Career Integration Grant. M.S. is supported by Cancer Research UK (program grant C399/A2291 to V.C). P.F. is supported by Science Foundation Ireland and National Children’s Research Centre. Author contributions: R.J., M.S., A.L.-L., S.S., E.B., C.A., K.L.C., C.H., D.C., M.S., D.G.-O., J.B.d.l.S., H.J., A.D., E.I.P., and D.J. performed experiments and contributed to the writing of the paper. P.G.F., A.M., W.B., D.B.M., V.C., and G.O. designed experiments and wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Data and materials are available through the authors.
View Abstract

Navigate This Article