Immune and barrier characterization of atopic dermatitis skin phenotype in Tanzanian patients

Abstract

Background:

Atopic dermatitis (AD) is a common disease, with particularly high prevalence found in Africa. It is increasingly recognized that patients with AD of different ethnic backgrounds have unique molecular signatures in the skin, potentially accounting for treatment response variations. Nevertheless, the skin profile of patients with AD from Africa is unknown, hindering development of new treatments targeted to this patient population.

Objective:

To characterize the skin profile of patients with AD from Africa.

Methods:

Gene expression studies, including RNA sequencing (using threshold of fold change of >2 and false discovery rate of <0.05) and real-time polymerase chain reaction, were performed on skin biopsies of Tanzanian patients with moderate-to-severe AD and controls.

Results:

Tanzanian AD skin presented robust up-regulations of multiple key mediators of both T helper 2 (T_H2) (interleukin 13 [IL-13], IL-10, IL-4R, CCL13,CCL17,CCL18,CCL26) and T_H22 (IL22, S100As) pathways. Markers related to T_H17 and IL-23 (IL-17A, IL-23A, IL-12, PI3, DEFB4B) and T_H1 (interferon gamma, CXCL9,CXCL10, CXCL11) were also significantly overexpressed in AD tissues (FDR<.05), albeit to a lesser extent. IL-36 isoforms revealed substantial up-regulations in African skin. The barrier fingerprint of Tanzanian AD revealed no suppression of hallmark epidermal barrier differentiation genes, such as filaggrin, loricrin, and periplakin, with robust attenuation of lipid metabolism genes (ie, AWAT1).

Conclusion:

The skin phenotype of Tanzanian patients with AD is consistent with that of African Americans, exhibiting dominant T_H2 and T_H22 skewing, minimal dysregulation of terminal differentiation, and even broader attenuation of lipid metabolism-related products. These data highlight the unique characteristic of AD in Black individuals and the need to develop unique treatments targeting patients with AD from these underrepresented populations.

Introduction

Atopic dermatitis (AD) is characterized by eczematous patches and persistent pruritus.¹ Although these clinical features are shared across different ethnic backgrounds and respective genetic variations, AD has a heterogeneous molecular signature, with distinct features in different populations.^2,3 Furthermore, although a predominant T_H2 and T_H22 skewing is consistent across AD subtypes, varying degrees of T_H1 and T_H17 polarization are also present depending on the subtype. For example, T_H1-related cytokines and chemokines are substantially up-regulated in adult vs pediatric AD,⁴ whereas Asian AD exhibits a stronger T_H17 skewing in comparison with European American AD.^4,5 The African American phenotype presents a relatively attenuated T_H1 and T_H17 polarization with a robust enrichment of the T_H2 and T_H22 pathways compared with European American patients, as recently reported.⁶ Nevertheless, African Americans, mainly descendants of West African populations, have been found to have different genetic and clinical profiles compared with the genetic and clinical profiles of East Africans.^7,8 Moreover, African Americans also represent an admixture with other American populations.^7,8 Despite the high prevalence of AD and atopy-related genetic variants, such as the interleukin (IL)-4R polymorphism in the African population,^9–11 to the best of our knowledge, AD skin has not been comprehensively investigated in East Africans.¹² This unmet need is made more important by the fact that African individuals are largely underinvestigated and underrepresented in clinical trials. As AD treatment is developing toward precision medicine approached in the globe, inclusion of African patients in molecular studies is imperative to promote targeted treatments for this population.

We thus performed a comprehensive molecular characterization of the immune and barrier abnormalities in skin of Tanzanian patients with AD vs ethnicity-matched healthy controls. We found that AD in these patients has a significant (FDR<.05) and robust T_H2 and T_H22 skewing with lesser T_H1 and T_H17 polarization and a unique barrier fingerprint, with sparse dysregulations of “classic” AD terminal differentiation genes (eg, filaggrin [FLG] and loricrin [LOR]), yet a marked lipid metabolism/synthesis disruption.

Methods

Clinical Characteristics

A total of 10 untreated patients with moderate-to-severe AD and 10 healthy controls were recruited, after signing an institutional review board–approved informed consent, at the outpatient clinic of the Regional Dermatology Training Center in Moshi, Tanzania. All participants were of Tanzanian origin (Fitzpatrick skin type 6). Clinical information on patient’s personal medical history (including allergic comorbidities), family history, and medication was documented using a questionnaire. Severity of AD was classified as mild, moderate, and severe using scoring of AD (SCORAD), investigator global assessment, and body surface area (BSA). Lesional and nonlesional skin samples of patients with AD and samples from controls were collected using 5-mm punch biopsies. Control biopsy samples were collected from volunteers’ back for consistency, whereas in patients with AD, skin biopsy samples were collected in the same body region in lesional and nonlesional skin. The biopsy samples were then immediately put into RNAlater (Qiagen, Hilden, Germany), incubated at 4°C overnight, and then stored at −80°C until further use.

RNA Sequencing

RNA was obtained from skin biopsy samples using Qiagen miRNeasy Mini kit, as previously described.¹³ Libraries were generated using TruSeq Stranded mRNA Library Prep kit (Illumina, San Diego, California). A cDNA library was then prepared by reverse transcription from extracted mRNA. Next-generation sequencing was performed with Illumina NovaSeq6000 (Illumina Inc) with single-ended read 100 cycles. Image analysis and base calling were performed in real time by the Illumina analysis pipeline.

Quantitative Real-Time Polymerase Chain Reaction

Reverse transcription to complementary DNA (cDNA) was performed using the High-Capacity cDNA reverse transcription (Thermo Fisher Scientific, Waltham, Massachusetts) after previous preamplification of all samples. TaqMan Low-Density Array (TLDA) cards (Thermo Fisher Scientific, Waltham, Massachusetts) were used for quantitative reverse-transcriptase–polymerase chain reaction (qRT-PCR). Primers are listed in eTable 1. A total of 100 ng total RNA was used for PreAMP pool and TLDA. Eukaryotic 18S recombinant RNA (rRNA) was used as an endogenous control. Expression values were normalized to Rplp0.

Statistical Analyses

Statistical analyses were performed using statistical language R (www.R-project.org). RNA sequencing data were preprocessed using standard pipeline incorporating quality control metrics, such as FastQC and MultiQC, sequence alignment based on STAR RNA-sequencing aligner, and sequencing read assignment to genomic features by featureCounts and voom-transformed. Mixed-effect model on R’s limma framework was used, and P values for the moderated t tests and paired t test were adjusted by Benjamini-Hochberg procedure. Differentially expressed genes were defined by fold-change (FCH) greater than 2.0 and false discovery rate (FDR) lesser than 0.05. qRT-PCR data were profiled using TLDA. Unsupervised clustering of the markers was performed using Euclidean distance and average agglomeration criteria. Threshold cycles [Ct] were normalized to Rplp0 by negatively transforming the Ct values to –dCt. The undetected expression values were estimated for each gene as 20% of the minimum unlogged expression across all samples. Log₂-scale qRT-PCR expression data were modeled by a linear mixed-effect model with biopsy type as a fixed effect and a random intercept for each patient. Means of each group were estimated using lsmeans, and comparisons of interest were tested using contrast.

Results

Lesional and nonlesional skin samples from 10 adult Tanzanian patients with AD (3 female, 7 male, average age 43 ± 1 years old) with moderate-to-severe AD (mean SCORAD¹⁴ 45 ± 14 and a mean body surface area [BSA] of 48% ± 23%) were compared with skin samples from 10 healthy Tanzanian controls with no personal or family history of atopy (5 female, 5 male, mean age 33 ± 1.3 years old). Of 10 patients with AD, 9 reported conomitant atopic diseases including allergic rhinoconjunctivitis and asthma. Patient characteristics are given in Table 1. Because healthy controls were younger than patients with AD, we provide, in eFigure 1, age-adjusted analysis that revealed comparable gene dysregulation as found in our unadjusted cohort (detailed below).

Table 1.

Patient Characteristics

Characteristic	AD (n = 10)	Controls (n = 10)
Sex (n, male)	7	5
Age (mean years ± SD)	43 ± 1.0	33 ± 1.2
African race, N(%)	10 (100)	10 (100)
SCORAD (±SD)	45 ± 14	N/A
BSA(±SD)	48 ± 23	N/A
Total serum IgE levels (±SD)	2607.9 (3522)	504.3 (603.1)
Atopic conditions other than AD (n)	10	0
Family history of atopy (n)	Positive: 4 Negative: 6	Negative: 9 Unknown: 1

Abbreviations: AD, atopic dermatitis; BSA, body surface area; IgE, immunoglobulin E; SCORAD, scoring of AD.

Global Assessment of Immune Genes Reveals Robust T_H2 and T_H22/T_H17 Up-Regulations

Global molecular profiling of the AD transcriptomes by RNA-seq identified 1946 differentially expressed genes (DEGs, 619 up-regulated, 1327 down-regulated) in lesional AD skin vs skin of healthy controls using threshold of FCH greater than 2 and false discovery rate (FDR) lesser than 0.05 (top DEGs in lesional AD vs normal skin are presented in eTables 2 and 3). Although nonlesional AD skin presented similar trends of gene modulations using P values (187 up-regulated, 497 down-regulated), there were only few gene dysregulations by FDR (6 up-regulated, 2 down-regulated).

Among genes significantly modulated in lesional vs normal skin were those related to cytokines and chemokines across various immune components, as depicted in a heatmap of a curated immune gene subset (FDR < .05; Fig 1).^6,15 DEGs presenting the most robust up-regulations in lesional vs normal skin (FCH > 5, FDR < 0.05) included markers of cellular activation and recruitment of T-cells and dendritic cells (DCs) (CCR7, TNFRSF9, CD1B, CD80), general inflammation (MMP12), that presented exceedingly high up-regulation in nonlesional skin as well, and multiple genes related to T_H2 (IL-13, IL-10, IL-4R, CCL18, CCL26, CCL13, CCL17, CCL7, CCL22, CCR4) and T_H17 and T_H22 (IL22, IL-26, CXCL1, S100A9, S100A8, S100A12, S100A7) pathways, all of which are implicated in AD pathogenesis across all subtypes.^2,16 Unlike the robust and significant up-regulations found in T_H2 and T_H22 pathways, more modest up-regulations were found in some T_H1- and T_H17-related markers (FCH > 2, FDR < 0.05). Innate immune gene products, including IL-1β and IL-6, and the keratinocyte-derived cytokine IL-17C, were not up-regulated in lesional AD vs normal skin.

Figure 1.

Summary heatmap of selected immune genes by RNA-seq, representing multiple pathways in lesional and nonlesional AD vs normal Tanzanian skin by criteria of FCH greater than 2 and FDR lesser than 0.05. Red indicates up-regulated whereas blue indicates down-regulated mRNA expression values. Genes are sorted by hierarchical clustering. The triple asterisk indicates FDR < 0.001, the double asterisk indicates FDR < 0.01, the asterisk indicates FDR < 0.05, and ⁺FDR < 0.1. AD, atopic dermatitis; NL, nonlesional; LS, lesional; FDR, false discovery rate; FCH, fold change.

Gene Set Variation Analysis of previously published AD-related genes by immune pathways, including the Meta-Analysis–Derived AD (MADAD) transcriptome¹⁶ and key immune pathways,^17,18 revealed that the MADAD signature was significantly enriched in both lesional and nonlesional Tanzanian AD skin vs controls, as were the T_H2 and T_H22 pathways (P < .05; eFig 2).

Lipid Synthesis/Metabolism Genes Present the Most Significant Down-Regulations Among Epidermal Barrier-Related Genes

To evaluate the epidermal barrier, we analyzed a barrier gene subset^6,19–21 including epidermal differentiation complex (EDC), tight junction and lipid genes, as illustrated as a heatmap in Figure 2. Particularly robust and significant down-regulations were found in genes related to lipid metabolism (eg, HAO2, ELOVL3, GAL, FAR2, AWAT1, DGAT2, FADS1 and 2, FABP7). Significant down-regulations were also noted in genes encoding for tight and adherent junctions, including cadherins (CDH12, CDH10, CDH19, CDH20) and claudins (CLDN8, CLDN10, CLDN1) in lesional skin vs controls and keratin genes (KRT79, KRT77, KRT33A) (FDR < 0.05). Nevertheless, key terminal differentiation genes that are characteristically down-regulated in European American AD (eg, FLG, LOR, sciellin, PSORS1C2, and late cornified envelope-LCE proteins)^16,22 were not significantly down-regulated in lesional Tanzanian AD skin vs normal skin (FDR > 0.05). Furthermore, genes associated with FLG breakdown products were also found in comparable levels across AD and normal skin (ie, UCA1, UROC1, OPLAH, GGCT, ARG, CASP14, and TYR) (eTable 4). The intact keratinocyte terminal differentiation in AD lesional skin is supported by a Gene Set Variation Analysis for EDC-related genes, compared with lipid synthesis/metabolism genes that are significantly down-regulated (FDR < .05; eFig 2).

Figure 2.

Summary heatmap of selected epidermal barrier genes by RNA-seq in lesional and nonlesional AD vs normal Tanzanian skin by criteria of FCH greater than 2 and FDR lesser than 0.05. Red indicates up-regulated whereas blue indicates down-regulated mRNA expression values. Genes are sorted by hierarchical clustering. The triple asterisk indicates FDR < 0.001, the double asterisk indicates FDR < 0.01, the asterisk indicates FDR < 0.05, and ⁺FDR < 0.1. AD, atopic dermatitis; NL, nonlesional; LS, lesional; FDR, false discovery rate; FCH, fold change.

To understand how these RNA-seq data in Tanzanian patients with AD reflect the previously published phenotype of African American patients with AD with moderate-to-severe disease,⁶ we performed a scatterplot comparison of relative differences vs normal skin (fold-changes) in either lesional or nonlesional AD, in the 2 patient populations (Fig 3). We found strong and significant correlations in lesional and nonlesional skin in the 2 phenotypes (r = 0.89 for lesional, r = 0.65 for nonlesional, P < .001). Nevertheless, we also found some differences, with a relative down-regulation of lipid synthesis and metabolism-related genes in Tanzanian skin compared with African American skin (ie, AWAT1, HAO2, FAR2). Further analysis using the same methods to compare Tanzanian and European American/White AD is presented in eFigure 3, which was consistent with overall agreement between the phenotypes (r = 0.85 for lesional, r = 0.49 for nonlesional, P < .001). Nevertheless, similar to the comparison of Tanzanian and African American AD, there was a relative down-regulation of lipid synthesis and metabolism-related genes in Tanzanian compared with European American lesional and or nonlesional skin (ie, FADS1, FADS2, FA2H). Moreover, terminal differentiation markers (eg, FLG, LOR, FLG2, LCE2B) presented greater down-regulation in European American AD (eTable 5).

Figure 3.

Scatterplots comparing relative FCH differences in Tanzanian vs African American AD samples (from a previously published cohort) in (A) lesional and (B) nonlesional skin. Circle size represents the magnitude of the differences in log2FCH between the populations. The location of the circles represents the respective expression in each population. AD, atopic dermatitis; FCH, fold change; LS, lesional.

RT-PCR Studies Further Refine T_H2 and T_H22 Up-Regulations and Barrier Impairment

To validate RNA-seq data and to evaluate key inflammatory and epidermal differentiation markers that are often below the detection threshold in RNA-seq, we also performed RT-PCR using a large panel of 87 AD-related genes (Fig 4, eFigs 4 and 5).

Figure 4.

qRT-PCR analysis of selected genes in AD vs normal Tanzanian skin. Values reveal log₂ expression and are presented as means ± SEMs. Black symbols indicate significance of comparison to normal; red symbols indicate significance of comparison between lesional and nonlesional skin. The triple asterisk indicates P < .001, the double asterisk indicates P < .01, the asterisk indicates P < .05, and ⁺P < .1. N, normal; NL, nonlesional; LS, lesional; qRT-PCR, quantitative reverse transcriptase–polymerase chain reaction; AD, atopic dermatitis.

Generally, PCR data were in line with our RNA-seq analysis. AD skin presented up-regulation of T-cell activation/migration markers, primarily in lesions compared with both nonlesional and control skin (ie, ICOS, CCR7, IL-2, IL-15, and CCL19), whereas MMP12 was up-regulated in both lesional and nonlesional tissues (P < .05) (Fig 4, eFig 4). Markers of innate immunity were not significantly up-regulated in Tanzanian AD tissues (IL-1β, IL-6, IL-17C;P<.05). NOS2 (inducible nitric oxidase synthase), a biomarker of disease activity in psoriasis, has been found to be up-regulated in psoriasis and down-regulated in AD in European Americans, helping to accurately classify these diseases.^23–25 Indeed, in this study, we also found NOS2 to be significantly down-regulated in both lesional and nonlesional Tanzanian AD compared with controls. T_H2 and T_H22 dysregulations (compared with normal skin) were captured in lesional and/or nonlesional AD (ie, IL-13, IL-22, IL-5, CCL13, CCL18, CCL22, CCL26, S100A7, IL4, IL-4R, OX40; P < .05) (Fig 4, eFigs 4 and 5).

Similar to the AD skin phenotype in other ethnicities,^5,6 several T_H17 markers were increased in lesional skin (IL-17A, IL-21, IL-26, IL-23A, IL-12, PI3, CXCL1 and 2, DEFB4B) (P < .05 compared with nonlesional and controls). Key T_H1-related genes were significantly up-regulated in Tanzanian AD lesions vs nonlesional and/or controls (ie, interferon gamma [IFN-γ], CXCL9,CXCL10,CXCL11; P < .05). The T_H17-related IL-36 isoforms (IL-36α, IL-36γ, and IL-36RN) were up-regulated in both lesional and nonlesional skin in Tanzanian AD (P < .05). The negative regulators IL-34 and IL-37 displayed significant reductions only in lesional skin, whereas the regulatory markers IL-10, FOXP3, and CTLA4 were significantly modulated in both AD tissues (P < .05) (Fig 4, eFigs 4 and 5).

Overall, genes associated with lipid metabolism were significantly suppressed already in nonlesional AD (eg, ELOVL3/5, GAL, FA2H, FAR2, AWAT1/DGA2) (P < .05), with further decreases in lesional skin (P < .001). Significant up-regulation of epidermal hyperplasia markers (KRT16 and SERPINB3) and down-regulations of the terminal differentiation marker FLG2 and tight junction markers (CLDN1, CLDN8, CLDN23) were found in lesional AD vs nonlesional skin and controls, reflecting greater barrier impairment in lesional skin (P < .05) (Fig 4, eFig 5). Similar to RNA-seq data, FLG, LOR, and periplakin (PPL), hallmark markers of the epidermal differentiation complex,^22,26 were not markedly down-regulated in AD tissues vs controls. A summary heatmap of expression profiles by PCR is in eFigure 6.

Disease Severity of Tanzanian Patients With AD Correlates With Polar Immune Pathways

Spearman correlation of immune and barrier markers measured by RT-PCR and disease severity (evaluated by SCORAD) revealed that clinical severity of Tanzanian AD strongly correlates with pivotal AD immune markers in both lesional and nonlesional skin and that correlations are not limited to a specific component of the immune response (r > 0.6, P < .05) (Table 2, eTable 6). In lesional skin, these included the T-cell activation/migration marker CCL19, the general inflammation marker MMP12, and the T_H2 markers IL-5 and CCL13 (r > 0.6, P < .05). Moreover, key T_H1-(CXCL9,CXCL10,CXCL11, IFN-γ) and T_H17-related genes (IL-17A, IL21 and IL-8) were robustly and significantly correlated with SCORAD (r ≥ 0.7, P < .05). The T_H9-cytokine IL-9 and the regulatory marker CD279/PD-1 also significantly correlated with SCORAD in lesions of AD (r > 0.6, P < .05). In nonlesional skin, robust correlations of SCORAD with T_H2-(IL-5, CCL18) and T_H1-(CXCL9,CXCL10,CXCL11) related markers were detected, in addition to CD279/PD-1 and the T_H17-related marker IL-19 (r > 0.6, P < .05).

Table 2.

Spearman Correlations

Lesional skin			Nonlesional skin
Marker	R	P value	Marker	r	P value
MMP12	0.85	<.001	CXCL11	0.75	.01
CXCL10	0.84	<.001	CXCL9	0.74	.01
IL-17A	0.81	.01	IL-5	0.70	.02
CXCL9	0.80	.01	CXCL10	0.70	.02
IFN-γ	0.78	.01	PDCD1	0.69	.03
CXCL11	0.75	.02	CCL18	0.69	.03
CCL19	0.75	.02	IL-19	0.65	.04
IL-9	0.72	.03	CCL26	0.60	.07
IL-5	0.72	.03	CCL13	0.59	.07
IL-21	0.70	.03	CXCL2	0.59	.07
CCL13	0.69	.04	PI3	0.55	.10
CXCL8	0.68	.04
PDCD1	0.68	.04
CCL11	0.63	.07
CCL18	0.63	.07
CXCL2	0.62	.08
CCL7	0.61	.08
STAT1	0.59	.10
CLDN8	−0.66	.05

Abbreviations: IFN-g, interferon gamma; IL, interleukin.

Discussion

African and African American populations have perhaps the highest global rates of AD, affecting up to 18.9% of Tanzanians and 19.3% of African Americans, respectively.^3,10,27,28 Populations of African ancestry also exhibit increased AD severity and higher rates of hospitalization²⁹ and higher prevalence and/or morbidity of other atopic diseases, including asthma, allergic rhinitis, allergic rhinoconjunctivitis, and eosinophilic esophagitis.^30–33 Furthermore, the clinical presentation of AD in patients of color differs from White patients, as they are more likely to have lichenified and postinflammatory manifestations.^34,35 African populations and African Americans have overrepresentation of IL-4Rα polymorphisms, likely conferring a greater risk for type 2 diseases.^9–11,36 In addition, different serologic immune signatures and sensitization patterns in sub-Saharan vs central European patients with AD.³⁷ These unique characteristics of AD in patients of African ancestry may have an effect on therapeutic management of these patient populations. The need for development of treatments or dosing regimens that uniquely target African populations has been suggested by the reduced efficacy of the approved dose of the IL-4Rα antagonist, dupilumab, in African Americans compared with other races.³⁸ Although African Americans share a genetic background with native Africans, they are also mixed with other ethnicities.^7,8 As AD is a heterogenous disease with differential involvement of various immune axes in different disease phenotypes, a “one-size-fits-all” treatment may not be applicable to all patient populations, necessitating a precision medicine approach to improve clinical outcomes across the spectrum.^2,39 Despite the high prevalence of AD in Africa and the reported differences in disease manifestations and genetics vs other races, advocating for investigation of the African AD subtype, the molecular phenotype in skin has never been characterized, inhibiting development of tailored treatment approaches for this population.¹² We thus evaluated the skin phenotype of patients from Tanzania (East Africa) with moderate-to-severe AD using a global transcriptomic approach.

We found that the Tanzanian AD phenotype is highly concordant with the African American AD skin profile, with a dominant T_H2 and T_H22 skewing, which seems to be a common feature across all ethnic populations with AD.^5,6,40,41 Significantly up-regulated markers in our cohort included pivotal type 2–related mediators (eg, IL-13, IL-10, IL-4R, CCL17, CCL18, CCL26 (FDR < .05) and key T_H17- and T_H22-related products (eg, IL-22, IL-26, CXCL1, S100As). Tanzanian AD was found to have some up-regulations of T_H1- and T_H17-related markers, but the key cytokines of these pathways (IFN-γ and IL-17A, respectively) were not captured by RNA sequencing. IL-36 isoforms were highly up-regulated in both lesional and nonlesional tissues, whereas IL-36 cytokines in European Americans were only considerably up-regulated in lesional AD skin.¹⁵ As IL-36 has been highly associated with psoriasis and psoriasiform dermatitis,^42,43 this finding could potentially contribute to the more lichenified lesions characterizing AD in African individuals.^34,35 In addition, the IL-21 cytokine, that is highly up-regulated in lesional Tanzanian AD skin, has been reported to induce both IL-17, IL-4 and IL-13, possibly linking the activation of IL-21 synergizes with IL-4 to increase IgE production to greater levels than each cytokine alone.⁴⁴

Similar to other AD ethnic/race background, this AD cohort also presented significant up-regulations of markers associated with T-cell and DC activation and migration (eg, CCR7, ICOS, CD1B, CD80 in lesional skin; FDR < .05) and general inflammation (MMP12, in both lesional and nonlesional skin). We also detected down-regulation of NOS2, a marker that distinguishes AD from psoriasis across all phenotypes.^23–25 Tanzanian AD skin largely lacked significant up-regulations of markers related to innate immunity, including IL-1b, IL-6, and IL-17C, a trend that was also observed in African American AD but not in European American patients.⁶

A hallmark of AD across all phenotypes is alterations in epidermal barrier, with up-regulation of T_H2-/T_H22-related markers (eg, IL-4, IL-13, IL-22) playing an important role in the suppression of epidermal barrier genes.^45–47 Nevertheless, previous reports outlined some differences in barrier abnormalities among various ethnicities/races. For example, African Americans lack marker dysregulations of FLG, while having significant down-regulations in LOR and FLG2 (FDR < .05),⁶ and European American patients present extensive down-regulations of multiple terminal differentiation genes, including FLG, LOR, PPL, and FLG2, multiple LCEs, and PSORS1C2.^15,45 Differential expression of FLG among different races can be explained in part by the variable rates of FLG null mutation and copy number variations (CNV), detected in up to 50%, 27%, and 12.2% of European, Asian, and African American patients with AD, respectively.^48–53 Studies in African populations suggest present yet weaker association between FLG null mutations/CNV and the development of AD than those reported in European populations,^54,55 and, as Zhu et al⁵⁶ have recently found, FLG variation revealed in African American patients with AD differ from that of European patients. In our cohort, genes of terminal differentiation including FLG but also LOR, PPL, SCEL, multiple LCEs, and PSORS1C2 were not significantly down-regulated in lesional and nonlesional tissues. Although similar trends were detected in African American AD,⁶ some of these key barrier genes did attain significance in this population vs controls (including LOR and LCEs), highlighting the differences between African American and African AD. The lack of FLG null mutations suggests that other pathogenic molecules may play key roles in African patients with AD.^3,57 These could include FLG2 (closely associated with FLG gene in both location and physiological activity, diminishing during acute spongiotic dermatitis)⁵⁸ and CLDN1 (a tight junction gene) mutations, both associated with AD in subjects of African ancestry.^59,60 Indeed, we found that these barrier markers were both significantly down-regulated in lesional AD in our cohort.

In addition, as the disrupted barrier of AD is also characterized by down-regulations of lipid synthesis/metabolism genes (eg, ELOVL5, FA2H, AWAT1),^16,61–63 we also evaluated the expression of these lipid-related markers. Despite the observed differences in the expression of terminal differentiation genes, genes of lipid synthesis and metabolism were robustly down-regulated in Tanzanian AD, consistent with reports in other AD phenotypes and ethnicities.^6,16 Moreover, on comparing the relative dysregulations of African American and Tanzanian AD skin, we found that lipid-related genes, including AWAT1, HAO2, and FAR2, were more suppressed vs controls in the latter. These results are also supported by previous reports on normal skin across different races, in which African American skin had the highest protein levels with significantly reduced lipid levels compared with that of both White and Asian subjects.^64,65

This study has a few limitations. First, although our cohort was homogeneous and included only Tanzanian patients with AD and controls, our sample size was small. Second, although we captured up-regulation of T_H2- and T_H22-related cytokines but normal FLG expression, we did not have tissue left to conduct any protein analysis studies to evaluate FLG protein and breakdown products. We are thus not able to exclude that the transcriptomic data identified profilaggrin rather than end products of FLG. Third, DNA analyses were not performed, and we were therefore unable to evaluate gene polymorphism and mutation status. Finally, a direct comparison to patients with AD of other ethnicities was not feasible, as studies were not run simultaneously. Nevertheless, a comparison of the relative FCH of a previously published African American AD cohort⁶ with our Tanzanian AD cohort revealed a largely concordant immune and barrier dysregulation of patients of African heritage.

In conclusion, this study portrays the molecular characteristics of AD in African patients, an underinvestigated and underrepresented population, despite its high prevalence of AD and atopy. It suggests an overall high correlation between African American and Tanzanian AD phenotypes, although some expression differences were detected, mainly in lipid-related genes. Though a “magic bullet” treatment for AD across the various phenotypes is not available, our results may guide future drug developments for patients of African ancestry with AD and help reduce the health disparities associated with this population.