Loss-of-function mutations in FLG, the gene encoding a late epidermal protein, filaggrin, constitute the most prominent genetic predisposition factor for atopic dermatitis (AD)3, highlighting the multifaceted role of this protein in supporting epidermal barrier function and controlling the keratinocyte differentiation process. Consequently, reduced filaggrin expression in the skin of AD patients and experimental models impacts numerous processes that are hallmarks of effective epidermal differentiation and cornification46, e.g., remodelling of the cytoskeleton47, formation of tight junctions48, lipid production 47, and changes in enzymatic activity6,49. FLG null mutations predispose to microbial dysbiosis50 and reduced ability to control skin infections, resulting in S. aureus superinfections51 and a predisposition to eczema herpeticum52. The impact stretches beyond the skin; FLG mutations are also linked to other manifestations of the “allergic march”, i. e. asthma, rhinitis, food allergy; affecting organs in which filaggrin is not expressed. To date, it is not clear how tissues distant from the skin may be impacted by filaggrin insufficiency, although penetration of allergens through an impaired structural barrier has been proposed as a means by which a systemic immune response is initiated.
In this study, we used a knockdown model to mimic the filaggrin expression downregulation dependent on the isolated inherited factor, which allowed us to dissect out the impact of the AD inflammatory mediators and environmental factors. The use of a stable knockdown line allowed us to overcome the low EV output from primary keratinocytes and the limited size of AD skin samples. Our study, integrating the findings from 2D in vitro models with 3D organotypic cultures constructed of primary cells with the AD skin dataset visualised the extent of changes resulting from filaggrin insufficiency and identified the means through which these widespread alterations could promote compartmental remodelling.
The involvement of keratinocyte-derived exosomes/sEVs in antigen-specific presentation was previously only studied by Kotzerke et al., in the context of responses to ovalbumin (OVA) in a murine model which failed to detect any apparent T cell activation of OVA-specific T cells18. However, the authors did not investigate filaggrin insufficient mice or from perspective of lipid-specific responses; at the same time, significant differences in the CD1 system between the species (CD1a-c are absent in the rodents), would hamper detection of any such responses, unless a humanised model is used. Recent work, which described S. aureus enterotoxin B exosome-mediated transfer from keratinocytes following superantigen exposure described a potential for non-specific T cell activation53. Of note, while we did not observe any differential outcome in the class I/class II presentation pathways by simple addition of sEVs during the antigen pulsation, it is still possible that sEVs from keratinocytes insufficient in filaggrin may have additional effects relevant to the peptide presentation, e.g., through their altered ability to transfer peptide antigens or propensity to undergo cellular uptake by antigen presenting cells. To date, it has not been determined whether keratinocyte-derived exosomes/sEVs contribute to lipid-specific T cell responses.
This study, to our knowledge, is the first demonstration that secretory vesicles may constitute an efficient source of ligands for lipid presentation pathways; we showed that exosomes/sEVs are not immunologically inert in this system, but they supply PLA2 substrates to either activate CD1a-specific T cells or lipid ligands of the inhibitory potential with respect to the IFNγ responses and promoting a type 2 bias. Given that sEVs contain a mixture of permissive and non-permissive lipids, such a shift between a type 1 and type 2 response may reflect changes of the overall avidity during CD1a-mediated presentation to T cells. Specifically, it has been shown for both peptides and lipid presentation within the CD1d pathway that changes in ligand affinity (hence the overall interaction avidity) result in differential contact time between the cells and their activation level, leading to differential response54–57; the longer the time the more type 1 bias. This “structure–activity relationship” has been proposed to result in a ligand-specific “cytokine fingerprint”58,59. Here, increased abundance of the non-permissive ligands, disrupting CD1a-TCR contact zone may reduce the interaction time, resulting in shorter time of cellular interaction and T cell activation more biased towards type 2 responses.
In the context of atopic skin disease, we observed extensive impact of filaggrin knockdown on keratinocytes as a whole and their exosomal/sEV compartment specifically. We subsequently found that, through remodelled sEV composition, the impact is carried through into the long-distance communication stream; in addition, the loss of control of PLA2 activity14 in the filaggrin insufficiency scenario may lead to even greater dominance of the inhibitory ligands released from sEVs and compound skin inflammation. We concluded that the sEV-conveyed message determines the involvement of sEVs in CD1a-restricted lipid antigen presentation which links aberrant keratinocyte differentiation with a Th2-biased allergic inflammation and could provide some explanation to the phenomenon of the “allergic march”.
Aberrant keratinocyte differentiation resulting from filaggrin insufficiency has previously been shown to contain a broad lipid dysregulation component in vitro9 which correspond to the lipid abnormalities previously reported in AD skin in vivo41,42,60. Here we determined that the altered exosomal/sEV FA composition in our model of filaggrin-insufficient keratinocytes is a likely consequence of a reduction in expression of the enzymes in the long-chain fatty acyl-CoA ligase family (ACSLs). ACSLs are enzymes upstream of several critical cellular lipid metabolism pathways44 catalysing the process of fatty acid activation, and formation of fatty acyl-CoA esters which regulate diverse cellular functions, for example providing gene regulation, enzyme inhibition, modulation of ion channel function, and membrane fusion43. ACSLs are implicated in membrane phospholipid biosynthesis; their involvement in the process of incorporation of MUFA and PUFA species into membrane phospholipids was previously described for multiple ACSLs45,61,62; they also have a preference towards polyunsaturated fatty acids45, 61–63. An increase in saturated fatty acids and a decrease in polyunsaturated fatty acid content has been described in rat hepatocytes in a ACSL3 knockdown model61. As for the manifestations of the allergic march, methylation of the ACSL3 5’-CGI has been found to correlate with asthma status in children64 and reported to increase in an allergen-induced airway hyperreactivity model in mice65. Furthermore, methylation of the ACSL3 gene has also been determined as a signature predictive of clinical food allergy in children66. Interestingly, this enzyme was also found in exosomes/sEVs isolated from colostrum but not from mature breast milk67; in our study, it was not detected in keratinocyte-derived exosomes/sEVs, but it could result from the detection threshold. ACSL downregulation under filaggrin insufficiency background has important immunological consequences; we show that the lipid content in secreted exosomes/sEVs is affected to the extent which abolishes their capacity to provide substrates for generation of the CD1a permissive self-antigens by PLA2; these provide homeostatic T cell activation, contributing to tissue integrity. It has been previously determined that the optimal length of the lipid chain appropriate for accommodation within the CD1a groove is approximately 20 carbon atoms and that unsaturated lipids induce a superior response33. Interestingly, when we compared responses obtained from the selected lipids found within the sEVs it was not always the case, i.e., while we could see the highest level of responses to the polyunsaturated long C22:6 DHA, only some donors responded to this lipid; responses to C14:0 SFA were lower, but more prevalent, while responses to Lyso-PC18:0 were less persistent over time.
It has been suggested that the family of the elongation of very long (ELOVL) fatty acid enzymes, which controls the length of very long fatty acids may be involved in the generation of the long-chain sphingomyelin such as 42:2. While there was no differential expression in our in vitro dataset, we and others have identified decrease of ELOVL mRNA in AD skin60. The upregulation of FADS1, which we believe may be a secondary compensatory mechanism68, was the only additional finding relevant to this pathway in the cultured keratinocytes. In contrast, mRNA for several FADS enzymes were downregulated in the AD skin (but not in organotypic model); this may suggest more complex regulation where inflammatory milieu may play an important role.
While we did not find any changes in the sphingomyelin synthesis pathway per se, studies focusing on the loss of the ACSL activity provide additional insight. Specifically, ACSL has been shown to regulate composition of fatty acids and membrane lipids in lipid rafts69, by the effect on ceramide expression, e.g., silencing of the enzyme results in the accumulation of ceramides and sphingomyelin analogue in Drosphila (phosphoethanolamine ceramide; CerPE)69,70. Therefore, while the expression of the enzymes in the pathway of sphingolipid synthesis may not be directly affected by filaggrin insufficiency, the increased supply of the substrates channelled into the ceramide/sphingolipid synthesis pathway is a very likely explanation of the accumulation of the non-permissive sphingomyelins71.
Skin is enriched in CD1a+ Langerhans cells abundant in the epidermis72,73; in addition, CD1a is also inducibly expressed by dendritic cell populations deeper in the tissue74,75. Our findings bear high relevance to the immunological events and tissue integrity76, since the CD1a-restricted population has been shown to contain many autoreactive T cells, capable of sensing barrier damage and promoting mechanisms engaged in tissue repair37. CD1a-resticted responses also contribute to the control of pathogenic skin bacteria77 and there seems to be an indication of their importance also in the lungs and gut78,79 where CD1a-expressing cells are also found80–86. To this end, CD1a-restricted responses have been shown in the humanised model of M. tuberculosis infection87 and to a range of M. tuberculosis lipopeptide (DMM) isomers88. Our study determined that neoantigens derived from normal keratinocytes (filaggrin sufficient; replicated by shCsEV in our study) are likely to be CD1a permissive ligands promoting autoreactive responses; their provision may support homeostasis at the skin barrier site or potentially even play an adjuvant-like role in antimicrobial immunity89. In contrast, exosomes/sEVs secreted on the filaggrin insufficiency background, containing altered lipid content can inhibit type 1 T cell responses and promote type 2 bias. Given the preference of the CD1a molecule to bind high affinity inhibitory ligands38, such as those contained within the sEVs produced by filaggrin insufficient keratinocytes, their presence in the milieu would likely affect both the low-level homeostatic and the much more pronounced antimicrobial CD1a-mediated T cell responses.
In contrast, our data indicate that in the absence of PLA2, exosomes/sEVs do not drive marked T cell reactivity, therefore reducing the risk of inflammation in the absence of an external threat. It is important to note that pathogens may constitute 50,51,90 a source of the phospholipase A2 activity, either directly91–93 or indirectly77,94,95. At the same time normal keratinocyte-derived exosomes/sEVs could potentially quench the toxic impact of PLA2 on cellular membranes protecting the body from excess tissue damage during inflammation. Exosomes/sEVs could also shield commensal bacteria which seem to be more susceptible to PLA2 than pathogenic strains96. A causative role of dysbiosis97,98, and chronic inflammation preceding the development or exacerbations in allergic asthma99,100, intestinal tissue damage 101,102 and food allergy103, affecting the development of tolerance to the encountered allergens104 has been previously established. Given that keratinocyte-derived exosomes/sEVs transfer into the circulation and are delivered into peripheral tissues, the impact of filaggrin insufficient keratinocyte-derived sEVs could extend beyond the local tissue environment, affecting responses in the locations distant from the skin and contributing to the development of allergic manifestations in those body sites, by reducing pathogen-directed and regeneration-promoting responses and promoting chronic inflammation and Th2 bias.
In summary, we have shown that small secreted extracellular vesicles constitute a source of antigens for lipid presentation pathways and are active during CD1a-mediated T cell responses. We also established that these responses greatly depend on the filaggrin status of secreting keratinocytes and can be linked to the dysregulation of lipid pathways including reduced ACSL activity in those cells, resulting from aberrant differentiation that is apparent both in vitro and in AD skin biopsies. A decrease in provision of the response eliciting CD1a self-antigens and enhanced supply of inhibitory ligands support immune consequences such as persistent allergic inflammation and dysbiosis in the skin; it appears probable that similar mechanisms operate in additional tissue locations to which sEVs can be transferred within the systemic circulation (such as the lungs and gut), contributing to the progression of the “allergic march” at these distant body sites.