Atopic dermatitis (AD) is the most common inflammatory skin disease, affecting up to 20% of children in the US and is characterized by an increased susceptibility to cutaneous infections1, 2. One in 10 AD subjects has difficulty clearing cutaneous infections with a host of viruses including herpes simplex, vaccinia, human papilloma and/or molluscum contagiosum1. This typically manifests as more extensive cutaneous and sometimes systemic disease and/or resistance to standard therapies. One of the rarer but more severe viral complications is eczema herpeticum (EH), caused by herpes simplex virus (HSV)-1 that results in a widespread skin infection1. EH can be complicated by viremia and multi-organ involvement including keratoconjunctivitis, meningitis, and encephalitis1. Although we have shown that AD subjects with more severe disease, greater allergen sensitization and enhanced Th2 polarity are at greatest risk for this viral complication, the molecular mechanisms responsible for this remain unclear1.
Disruption of the epidermal barrier is now recognized as a cardinal feature in AD pathogenesis. The skin has two barrier structures: the stratum corneum (SC) and tight junctions (TJ). It is widely accepted that AD subjects' SC is dysfunctional. The loss-of-function mutations in filaggrin, a structural protein important for its hygroscopic properties, have been reproducibly and robustly associated with AD3 and more recently with EH4. TJ are found on opposing membranes of stratum granulosum keratinocytes, directly below the SC. TJ consist of a complex of adhesive and scaffolding proteins that control the passage of water, ions, and solutes through the paracellular pathway5. We have recently demonstrated that the nonlesional epithelium of AD subjects have remarkable bioelectric abnormalities indicative of a TJ defect which may be the consequence of reduced levels of claudin-1 (CLDN1), a key TJ adhesive protein6. This is in line with the landmark work of Furuse and Tsukita whose CLDN1 knockout mouse established the importance of epidermal TJ and claudin-1. These mice died within 24 hr of birth with wrinkled skin, severe dehydration, and increased epidermal permeability as measured by dye studies and transepidermal water loss7. In our AD studies, we also observed an inverse correlation between CLDN1 expression and markers of Th2 polarity (total eosinophil counts and serum total IgE)6. This has led us to speculate that TJ defects may promote greater Th2 polarity, which is a characteristic of EH subjects1. We hypothesized that TJ defects such as we observed in AD subjects, may promote HSV-1 infections on that basis.
Additionally, viruses often commandeer intercellular junction proteins as a means of viral entry5. For example, wild-type HSV-1 infects keratinocytes through a complex interaction involving the viral envelope glycoprotein gD with either nectin-1 (PVRL1) or herpes virus entry mediator (HVEM)8, 9. Nectin-1, a Ca2-independent cell adhesion protein of the immunoglobulin superfamily, co-localizes with E-cadherin and beta-catenin to form another intercellular junctional complex called adherens junctions. Previous studies have shown that the susceptibility of human keratinocytes to HSV-1 infection is inversely related to the degree of cell-cell contact and confluency9, 10. This suggests that healthy junctional complexes may be a key deterrent to the spread of HSV-1 from one keratinocyte to another.
In these studies, we tested the hypothesis that the silencing of epidermal CLDN1 will enhance HSV-1 infectivity of human keratinocytes. We had previously demonstrated that reductions in CLDN1 adversely affected measures of TJ integrity6. Primary human keratinocytes (PHK) were isolated from foreskin as previously described6 and transiently transfected with scrambled (control) or CLDN1 (100 nM) siRNA (Santa Cruz Biotechnology, Inc; CA) using Lipofectamine™ 2000 transfection reagent (Invitrogen, Carlsbad, CA; see complete method in the Online Repository). We have previously demonstrated that with this knockdown approach lead to a >50% reduction in claudin-1 transcripts, and adversely affected TJ function as demonstrated by a 50% reduction in TEER and a 50% increase in permeability6. Importantly, this level of claudin-1 reduction mirrored what we had observed in AD samples. Twenty-four to 48h post-transfection, PHK were differentiated in DMEM with 10% heat-inactivated fetal bovine serum (HI-FBS; Invitrogen/Gibco) for 24h and subsequently infected with the highly virulent HSV-1 strain F (provided by Dr. D.C. Johnson) at a multiplicity of infection (MOI) of 0.1 (see complete method in the Online Repository). The MOI was chosen based on findings from viral titration (0.01 to 10 MOI) studies on differentiated and undifferentiated PHK, respectively, used as negative and positive controls (data not shown). After 2h, the viral inoculum was removed and PHK were cultured for 24h in DMEM containing 5% HI-FBS and 0.4% human-γ-globulin (0.5 mg/ml; Sigma, St. Louis, MO) to neutralize extracellular virus. HSV-infected PHK were fixed in 4% formaldehyde and stained with a polyclonal rabbit anti-HSV-1 (1:500; Dako, Carpinteria, CA) antibody followed by Alexa Fluor 488 donkey-anti-rabbit IgG H+L (1:1000, Invitrogen/Molecular probe) and 4′,6-diamidino-2-phenyl-indole, dihydrochoride (DAPI; 1:10,000, Molecular Probes). For each sample, six random fields were captured at identical acquisition settings and analyzed computationally to objectively quantify differences in focal forming units (FFU). A FFU was defined as a cluster of 3 or more adjacent HSV-1 positive cells. Claudin-1 depletion significantly increased the number and size of HSV-1 FFU (Fig 1A) as compared to control siRNA transfected PHK (Fig 1B, C). An average of 4.8 ± 0.7 FFU/field were counted in CLDN1 knockdown PHK, while only 2.5 ± 0.5 FFU/field were detected in the control cells (P = 0.05; n= 6; Fig 1C). Adapted Fluorospot software was used to semi-quantify differences in dimensions of each FFU (see complete method in the Online Repository). The diameters of FFU were significantly greater in keratinocytes whose CLDN1 expression was reduced (207 ± 22 μm) compared to controls (152 ± 40 μm; P = 0.03; n= 6; Fig 1D). Similar findings were observed when evaluating FFU area (P = 0.04; n= 6; Fig 1E) in PHK after CLDN1 knockdown (15649 ± 4367 pixels) as compared to control transfection (9902 ± 4943 pixels). The increased frequency of HSV-1-infected cells in CLDN-1 knockdown PHK was confirmed by an infectious center assay (see complete method in the Online Repository). More CLDN1 siRNA transfected PHK were infected with HSV-1 (38 ± 6%; n=3) than control PHK (28 ± 7%; n= 3; P = 0.003) (see Fig E1 in the Online Repository). Importantly, the observed increase in HSV-1 infectiveness in CLDN1 knockdown PHK could not be explained by reduced expression of nectin-1, the HSV-1 receptor. Indeed, nectin-1 mRNA expression was not affected by CLDN1 silencing (n=5; see Fig E2 in the Online Repository). However, we cannot rule out the possibility that reductions in CLDN1 and the affects on TJ function might reduce nectin-1 membrane localization. It is also important to highlight that our in vitro model is sufficient to induce functional TJ as demonstrated by enhanced trans epithelial resistance and reduced permeability6. However, it is not enough to induce terminal differentiation (e.g. filaggrin expression and cornified envelope). Although, this might be a limitation of the methodology we believe this condition well reflect the in vivo situation. In ADEH often the virus spreading is due to reactivation of a silent infection (inside-out); thus suggesting that the first barrier structure the virus will encounter is indeed TJ (located in the stratum granulosum, just below stratum corneum).
As part of the NIAID-funded Atopic Dermatitis and Vaccinia Network (ADVN) (reviewed in a previous publication by Beck et al1), we performed a preliminary study to evaluate whether CLDN1 polymorphisms were associated with EH. To do this, we utilized a haplotype-tagging SNP approach using genetic markers available in the public arena (n = 132 in dbSNP) in 414 European Americans (EA) and 328 African Americans (AA; see complete methods and Table E1 in Online repository). We previously reported modest associations (P = 0.003 -0.05) between SNP variants throughout the CLDN1 gene and AD that were more significant among African Americans6. In this study, we addressed whether genetic variations in CLDN1 contribute to risk of EH in both North American AD populations. The EH subphenotype was defined as AD subjects that had at least one EH episode documented either by an ADVN investigator (or a physician affiliated with the same academic center) or diagnosed clinically by an outside physician where the HSV infection was confirmed by PCR, tissue immunofluorescence, Tzanck smear and/or culture. The study was approved by the institutional review boards at National Jewish Health, Johns Hopkins University, Oregon Health & Science University, University of California San Diego, Children's Hospital of Boston and University of Rochester Medical Center. All subjects gave written informed consent prior to participation.
To determine whether SNPs in CLDN1 were associated with EH, the Cochran-Armitage trend test using the PLINK software (http://pngu.mgh.harford.edu/~purcell/plink/to) was performed within each AD population and confirmed with the adaptive permutation test. We observed only modest associations between CLDN1 SNPs and EH in our North American AD populations. We wondered if the strong association we observed between both FLG null mutations R501X and 2282del4 and EH (OR = 10.1 [4.7-22.1]; P=1.99 × 1011) was masking a lesser effect from CLDN1 variants4. To address this, we excluded subjects who had one or both of these FLG mutations and reanalyzed the associations with all CLDN1 SNPs. In the European American group, an intronic SNP (rs3774032; OR = 0.59 [0.35-0.98] P=0.037) was marginally associated with EH (Fig 2), but not significant after adaptive permutations adjustment (P=0.10). When excluding subjects with a FLG mutation (n=29 [36.3%]; Table 1), the association became slightly more significant (OR = 0.44 [0.22-0.85], P=0.0225) and still significant after adaptive permutations adjustment (P=0.035). Interestingly, one additional intronic SNP emerged (rs3732923, OR = 1.93 [1.21-3.07], P=0.0010; adaptive permutation P=0.0005; Fig 2). Although the EH sample size in the African American population was too small (n=21) for meaningful analyses, a SNP in the promoter region was associated with EH (rs3954259; OR = 2.16 [1.02-4.59], P=0.040; adaptive permutation P=0.033). Limitations of these studies are the small sample size and the superficial coverage of CLDN1 loci. Nevertheless, these findings implicate CLDN1 mutations in susceptibility to widespread HSV skin infections in subjects with AD, especially those who do not have a FLG mutation.
In conclusion, this study and previously reported work implicate both SC and TJ epidermal barrier defects as a mechanism to explain AD subjects susceptibility to widespread cutaneous infections with HSV and possibly other viral pathogens4. In support of this notion our EH patients more commonly experienced infections with molluscum contagiosum than AD subjects with no history of EH (29% vs 2%, respectively)1. More of our EH subjects (16%) suffered with recurrent HSV keratitis as compared to only 1% of AD subjects with no history of EH1. It is possible that the same mechanism may responsible for their ocular disease as corneal epithelium possess TJs and these structures express claudin111, 12. In this study, we show both mechanistic and genetic results that implicate tight junctions as a critical barrier structure important in containing the spread of epidermal HSV-1 infections. Defects in TJ as compared to the SC may be particularly relevant in viral infections where the virus enters from a basolateral direction as is the case with HSV reactivation. Collectively, this work highlights a previously underappreciated role for TJ proteins in cutaneous host defense and may lead to new therapeutic strategies.