Genetic ancestry shapes dengue virus infection in human skin explants

Significance

Dengue is a mosquito-borne viral disease that is expanding rapidly across the globe, affecting tens of millions of people. African ancestry is associated with a reduced risk of severe dengue, but the mechanisms are unknown. We show using isolated human skin specimens that dengue infection elicits a marked inflammatory response in European ancestry skin, leading to infection of resident cells that then migrate out of skin, spreading infection. In contrast, African ancestry skin has significantly less inflammation in response to infection, limiting replication and spread. The protective effects of African ancestry from severe dengue begin at the initial stages of infection. Genetic ancestry may affect the way different populations respond to dengue vaccines, which are weakened viruses given in skin.

Abstract

Dengue is the most prevalent arthropod-borne viral disease of humans, with over half the world’s population at risk. Infection with any of the four dengue virus (DENV) serotypes is most often self-limiting, but a significant number of cases present with severe dengue characterized by vascular leakage that may be fatal. African ancestry is associated with protection against severe dengue, but the mechanisms are unknown. Using skin explants from genetically defined donors, we show that European ancestry skin has a much stronger inflammatory response to DENV than African ancestry skin, eliciting markedly increased infiltration, infection and migration of resident Langerhans cells, macrophages, and dendritic cells. The effect was seen with all dengue serotypes and Zika virus and in the presence of heterotypic immune serum. Genetic pathways associated with inflammation, interferon (IFN)-α, and inflammatory cytokine signaling were enriched in European relative to African ancestry skin following infection. Infiltration and infection of macrophages in African ancestry skin increased to that of European skin after blocking IFN-α and providing interleukin-1β. Polymorphic variants in RXRA, OAS1-3, and TGFB1 genes that were more frequent in European donors were associated with increased virus replication. Paradoxically, European ancestry skin cells had increased expression of OAS3 in response to virus and type I IFN. Thus, the limited inflammatory response of African ancestry skin to infection restricts replication and spread of dengue and other flaviviruses. Genetic ancestry should be considered when predicting a patient’s likelihood of severe dengue, and when assessing efficacy and adverse events associated with dengue vaccines.

Dengue is a systemic viral disease transmitted by Aedes mosquitoes that has rapidly expanded across the globe in the last two decades, particularly in the Americas, where the number of reported cases in 2024 approached 13 million (1–3). Beginning more than four decades ago, multiple reports from South and Central American countries with genetically admixed populations described the protective effects of African ancestry from severe dengue (4–8). In a 1981 epidemic in Cuba characterized by secondary infections, a major risk factor for severe disease, the ratio of children of European to African ancestry with severe dengue was more than 14 to one (5). On the African continent, considerable controversy currently exists as to whether severe dengue is underreported or occurs only rarely (9, 10). Socioeconomic, cultural, and environmental factors contribute to differences in dengue outcome (11), and any biological mechanisms relating to ancestry remain to be defined.

The skin is the primary site of dengue virus (DENV) replication following inoculation by an infected female mosquito. Productive virus infection of several resident skin cell populations has been documented, including keratinocytes and Langerhans cells in the epidermis and dendritic cells (DCs), macrophages, fibroblasts, and endothelial cells in the dermis (12–16). Virus replication is initiated in keratinocytes that produce inflammatory cytokines and chemokines, promoting the local infiltration and infection of skin-resident myeloid cells (13, 17–19), while mast cell-driven recruitment of natural killer cells restricts infection (20). Inflammation and edema in response to mosquito bites aids in viral replication and dissemination via recruited and resident myeloid cells (21). The extent to which genetic ancestry influences the cutaneous immune response to DENV and other vector-borne infections and how this impacts systemic virus spread and disease is currently not known.

Individuals of African descent have disproportionately high fatality rates from many infectious and inflammatory diseases, after controlling for age, sex, and income (22). This has led to the concept that evolved immune responses to different ancestral pathogen burdens among geographic groups underlie current population differences in immunity and susceptibility to inflammation (23). Consistent with this hypothesis, the response of blood mononuclear cells to human pathogens differs with ancestry in immune and antiviral pathways. Higher levels of African ancestry are associated with an increased inflammatory response to bacterial pathogens (24), whereas higher levels of European ancestry are associated with increased type I interferon (IFN) pathway activity in response to influenza virus (25). Whether genetic ancestry influences viral infections in skin in a similar manner to blood remains to be determined. To explore these issues directly, we studied flavivirus infection in human skin to define virus–host interactions in the context of genetic ancestry.

Results

DENV Replication and Cellular Dynamics in European and African Ancestry Skin.

We previously developed an ex vivo model of flavivirus infection using full-thickness abdominal skin collected during elective skin-reduction surgery (13, 18, 26). For the current study we received abdominal skin from healthy DENV- and Zika virus-naïve individuals who self-identified as being of African (AA, n = 24) or European ancestry (EA, n = 35) (SI Appendix, Table S1). Biogeographical ancestry for each donor was determined using a panel of ancestry-informative single nucleotide polymorphisms (SNPs) (27) with subsequent analysis using ADMIXTURE (28) to determine the proportions of African, European, Asian, and Native American ancestry (Fig. 1A). Principal component analysis showed that donors overlapped with African or European groups from a worldwide dataset (29) (Fig. 1B). To investigate DENV infection in explants, 500,000 focus forming units (FFU) of DENV serotype 2 (DENV-2) was inoculated into the center of a 1 in² piece of skin that had been trimmed of subcutaneous tissue, and specimens incubated for 24 h (13). Sections of the inoculated region were stained with antibody to viral NS3 protein and nuclear dye to identify cell-associated replicating virus and imaged by confocal microscopy followed by quantitative image analysis (SI Appendix, Fig. S1). There was a strong positive correlation between the number of infected cells in both the epidermis and dermis with the proportion of European ancestry of donors (Fig. 1 C and D). Keratinocytes in the epidermis, identified using a pancytokeratin antibody, occupied nearly 1% of the total imaged area regardless of ancestry and infection, but the number of DENV-infected keratinocytes in EA skin was twice that of AA skin (Fig. 1 E and F and SI Appendix, Fig. S2). Virus inoculation increased both the number and infection of CD207⁺ Langerhans cells in the epidermis in EA relative to AA skin (Fig. 1 E and G and SI Appendix, Fig. S2). Notably, infection induced pronounced migration of infected Langerhans cells into the dermis in EA but not AA skin (Fig. 1E and SI Appendix, Fig. S3). Dermal CD1c⁺ DCs were increased in EA but not AA skin, and infected DCs were in significantly greater numbers in EA relative to AA skin (Fig. 1 E and H and SI Appendix, Fig. S2). CD163⁺ dermal macrophages were increased following virus inoculation in both groups, but the effect was greater in EA skin, where the ratio of macrophages in the infected versus mock condition approached eight to one. EA skin had more than twice the number of infected macrophages as AA skin (Fig. 1 E and I and SI Appendix, Fig. S2). Previous studies have shown that the increase in number of Langerhans cells, DCs, and macrophages in the region of DENV inoculation in EA skin is not the result of resident cell proliferation (26), and isolated explants lack a blood supply; hence the accumulation of myeloid cells is the result of infiltration from the surrounding tissue.

Fig. 1.

Defining genetic ancestry and cellular responses to DENV infection. (A) Biogeographical ancestry of all skin donors showing the proportion of African, European, Asian, and Native American ancestry based on ADMIXTURE analysis. (B) Principal component analysis of biogeographical ancestry overlaid on 809 reference samples for population structure. (C) Association between NS3-expressing cells in epidermis and dermis as a percentage of the total imaged area and the proportion of European ancestry (n = 38). R-squared and P value were determined by Pearson correlation and simple linear regression. (D and E) Representative images of immunofluorescence of skin sections from EA and AA individuals 24 h after DENV-2 inoculation stained with antibodies to viral NS3 (D) and the indicated cell markers (E). Dotted lines outline the epidermal–dermal junction. (Scale bar, 100 µm.) Blue staining represents nuclei, green represents viral NS3, and magenta represents cell-specific markers. (F) Quantification of AE1⁺ keratinocytes in uninfected (UI), mock-infected and DENV-infected skin (Left) and of AE1⁺NS3⁺ keratinocytes in infected skin (middle) in AA (n = 15) and EA (n = 21) specimens at 24 h. The relationship between infected and total keratinocytes as a function of European ancestry and the ratio of cells in the infected versus mock condition (n = 38) is shown on the Right. (G–I) Similar analyses as (F) shown for (G) CD207⁺ Langerhans cells in the epidermis in a subset of individuals (AA = 10, EA = 8) and (H) CD1c⁺ DC and (I) CD163⁺ macrophages in the dermis (AA = 15, EA = 21). (J) The number of migrated cells in media normalized to skin area from AA (n = 4) and EA (n = 4) skin after mock and DENV-2 infection at 24 and 48 h. (K) Association between number of migrated cells and DENV genomes expressed as FFU equivalents/mL, AA (n = 4) and EA (n = 4). P and r values were determined by simple linear regression. (L) Quantification of NS3⁺ cells in epidermis and dermis in AA (n = 4) and EA (n = 4) skin 24 h after infection with each of the four serotypes of DENV and Zika virus. (M) Quantification of NS3⁺ cells in epidermis and dermis and of NS3⁺ macrophages 24 h after DENV-2 infection in the presence of naïve or pooled anti-DENV-3 serum in AA (n = 4) and EA (n = 4) skin. The horizontal lines in all graphs represent medians. Between group analyses were done by the unpaired t test and within-group analyses were done by one-way ANOVA followed by Tukey’s multiple comparisons.

In a subset of individuals, we collected cells that migrated out of skin and into media at 24 and 48 h after mock or DENV-2 infection and normalized cell numbers to skin area. DENV-2 infection resulted in increased migration of cells out of EA skin only, with an almost sixfold higher number of cells at 48 h relative to AA skin (Fig. 1J). The number of migrated cells correlated positively with virus replication at 24 h as determined by quantitative PCR, indicating that the increased exodus of cells from EA skin is associated with an increase in virus dissemination (Fig. 1K). To determine whether ancestry-related effects extended to other serotypes of DENV and other flaviviruses, we inoculated skin with DENV-1, 2, 3, and 4 and Zika virus and measured infection at 24 h. All four DENV serotypes induced greater infection in EA relative to AA skin, although at different degrees in the epidermis and dermis. The greatest amount of virus replication in both groups was with DENV-2. Differences in efficiency of infection of DENV serotypes in skin is not surprising given the range of putative receptors, attachment factors and entry pathways used by different serotypes (30). Replication of Zika virus in both compartments was greater in EA as compared to AA skin (Fig. 1L).

Severe dengue disease is more likely to result from secondary DENV infections, as preexisting heterotypic immunity to one serotype enhances infection of Fcγ receptor-bearing cells upon infection with a second serotype, a process termed antibody-dependent enhancement (ADE) (31, 32). To determine whether ancestry-related differences in ADE exist, we formulated microneedle arrays with pooled human sera with or without immunity to DENV-3 and applied arrays to the skin surface to mimic preexisting immunity before inoculation with a low dose (5,000 FFU) of DENV-2 (26). Infection with DENV-2 in the epidermis was not affected by heterotypic immune sera in either group, as expected given that most cells in the epidermis are keratinocytes which lack expression of Fcγ receptors. In contrast, heterotypic immune sera increased DENV-2 replication by twofold in the AA dermis and threefold in the EA dermis, giving an overall enhancement effect of 300% due to European ancestry. The lower dose of DENV-2 used in these enhancement experiments (1% of standard inoculum) did not substantially increase the percentage of infected macrophages in EA relative to AA skin in the presence of naïve serum, but there was significantly greater enhancement of macrophage infection in EA than AA skin in the presence of heterotypic immune serum (Fig. 1M). Enhancement of DENV infection in human skin is dependent upon expression of the Fcγ receptor CD32 and to a lesser extend CD64 (26). To objectively measure expression of these receptors in EA and AA dermis, we did flow cytometry on dermal cell suspensions from fresh, uninfected skin, gating on CD45+ CD163+ macrophages and CD45+ CD163– CD1c+ DCs. Expression of CD64 was uniformly high on both cell types and indistinguishable between EA and AA dermis. In contrast, while cells from EA and AA skin both expressed CD32, expression was 10 to 50-fold lower on macrophages and DCs from AA skin relative to EA skin (SI Appendix, Fig. S4). Collectively, these findings show that there is substantially greater DENV replication with and without enhancing antibodies in the skin of individuals with European as compared to African ancestry, associated with markedly increased local infiltration and infection of skin-resident myeloid cells.

Ancestry-Associated Differences in the Innate Immune Response to Infection.

To determine the underlying mechanism for our observations, we sought to identify molecular pathways for which the response to DENV infection in the skin significantly correlated with genetic ancestry. We performed bulk RNA sequencing from punch biopsies from EA (n = 5) and AA (n = 4) skin 24 h after mock or DENV-2 infection and used continuous estimates of ancestry to identify ancestry-associated differentially expressed genes. We found 4,840 genes differentially expressed at a false discovery rate of 0.05. Of these genes, 2,520 (52%) were increased and 2,320 (48%) were decreased in expression in response to DENV infection as a function of European ancestry (Fig. 2A and Dataset S1). Gene ontology enrichment analysis showed that at least two thirds of the genes that were increased in expression with European ancestry were related to the inflammatory response (Fig. 2 B and C). In contrast, genes that showed a weaker response to DENV with increasing European ancestry (and thus a stronger response with increasing African ancestry) tended to be functionally unrelated to immunity (Fig. 2C and Datasets S2 and S3). Most inflammatory response genes that showed a significant interaction between European ancestry and DENV infection were related to chemotaxis and cell migration, reflecting the increased infiltration of skin-resident cells to the inoculation site in EA skin (Fig. 2D). Gene set enrichment analysis confirmed that inflammatory response genes were highly enriched in DENV-infected EA skin relative to AA skin, along with signaling through inflammatory cytokines tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6), and transforming growth factor-β (TGF-β). Genes associated with the antiviral IFN-α response were enriched in DENV-infected EA skin, despite the increase in virus replication (Fig. 2E, SI Appendix, Fig. S5, and Dataset S4).

Fig. 2.

Ancestry-related differences in transcriptomic and inflammatory responses. (A) The number of differentially expressed genes that were increased or decreased with European ancestry at different false discovery rate cutoffs at 24 h post infection. (B) Gene ontology network for genes with positive interaction effects showing pathways enriched with European ancestry following DENV infection. All significant pathways (FDR < 0.05) are shown. (C) Bubble plot showing gene pathways enriched following DENV infection and their relationship to European ancestry. Gene ontology enrichment analyses were performed separately for genes with a stronger or decreased response to infection (log2 fold-change greater than two in each direction of the effect). Only pathways with enrichment at FDR < 0.05 and at least four terms per cluster are shown. (D) Gene ontology terms within inflammatory response pathways and their relationship to European ancestry. (E) Gene set enrichment analysis of pathways associated with inflammatory and antiviral responses (Left) and with cytokine signaling (Right) that are enriched in EA skin relative to AA skin following DENV infection at 24 h. (F) Immunofluorescence of skin sections from EA and AA individuals 24 h after DENV-2 inoculation stained with antibody to IL-1β. (Scale bar, 100 µm.) Blue staining represents nuclei, and white staining represents IL-1β. (G) Quantification of IL-1β staining as a percentage of total imaged area at intervals after mock and DENV infection in skin from AA (n = 3) and EA (n = 5) donors. Shown are means and SE of the mean. Analysis was done by the Mann–Whitney U test comparing DENV conditions between AA and EA skin at each time point. (H) Quantification of CD163⁺ macrophages (Left) and CD163⁺NS3⁺ macrophages (Right) in AA (n = 3) and EA (n = 4) skin 24 h after mock or DENV-2 infection in the presence of IL-1β with and without anti-IFN-α, or anti-IL-1β with and without IFN-α, or isotype control antibody delivered by microneedle array. Horizontal lines represent medians. Analysis was done by one-way ANOVA followed by the Tukey test.

The cytokine interleukin-1β (IL-1β) is an early mediator of inflammation and is important in DENV replication and spread in the skin (13). We stained EA and AA skin at intervals after mock and DENV infection and measured the production of IL-1β in the epidermis by immunofluorescence. Infection induced IL-1β expression in the epidermis of both EA and AA skin, but the response in AA skin was significantly lower (Fig. 2 F and G). To determine the role of IL-1β in driving infection, and the potential contribution of IFN-α in restricting infection, we formulated microneedle arrays to deliver IL-1β (with and without antibody to block IFN-α) or anti-IL-1β (with and without IFN-α) and applied to EA and AA skin explants prior to inoculation with 500,000 FFU of DENV-2. In AA skin, IL-1β and anti-IFN-α had additive effects in increasing both the number and infection of dermal macrophages to levels seen in infected EA skin exposed to isotype control antibody. Conversely, in EA skin blocking IL-1β reduced macrophage number and infection to near background levels, while the addition of IFN-α had minimal benefit (Fig. 2H). This indicates that in EA skin IL-1β is sufficient to drive local cell infiltration and infection, and that IFN-α may have limited effect on controlling infection. This was confirmed by quantifying macrophage cell number and infection in conditions with anti-IFN-α or IFN-α alone (SI Appendix, Fig. S6). Together, these data show that genes associated with inflammatory pathways are strongly upregulated in response to DENV in skin of individuals with increasing European ancestry and are central to the increased virus replication and spread in EA relative to AA skin.

Contribution of Genetic Polymorphisms to Ancestry-Related Differences in Infection.

Numerous polymorphic genes that are expressed in skin are associated with innate immunity and are known risk factors for severe dengue (33–40). We determined the allelic variants for 60 SNPs in 29 different genes in our donors (Datasets S5 and S6). We found six alleles in five genes that differed significantly in frequency with ancestry and were associated with significant differences in DENV-2 infection in the epidermis; five of these alleles were also highly significant in the dermis (Fig. 3 A and B). Two polymorphisms were in RXRA, a gene that has previously been identified as conferring African-ancestry protection from severe dengue in Cubans (40). The GG genotype of rs4262378 and the AA genotype of rs4424343 in RXRA were associated with low DENV replication in both EA and AA groups, suggesting a direct protective role for these genetic variants independent of ancestry (Fig. 3C). These alleles were present at frequencies of 60 to 80% in AA individuals but less than 25% in EA individuals in our cohort, as they are in worldwide populations (Fig. 3B and Dataset S5). Polymorphic genes previously unknown to be associated with genetic ancestry and dengue included the oligoadenylate synthetase (OAS) genes OAS1, OAS2 and OAS3, and TGFB1. Notably, the CC genotype of rs1982037 which may result in overproduction of TGF-β1 (41) was present in 100% of AA donors (Fig. 3C), yet molecular pathways relating to TGF-β1 activity were significantly enriched in EA skin relative to AA skin following DENV infection (Fig. 2E and Dataset S4).

Fig. 3.

Genetic polymorphisms and ancestry-related differences in DENV infection. (A) Relationship between European ancestry and NS3⁺ staining in epidermis and dermis 24 h after DENV-2 infection for 60 SNPs. The x-axis shows the P values obtained when comparing the median levels of European ancestry between the genotypes for each SNP. The y-axis shows the P values obtained when comparing the median levels of DENV-2 infection between the genotypes for each SNP. Dashed lines show the threshold for significance based on a Benjamini–Hochberg correction using a false discovery rate of 0.1. Polymorphisms that are significantly associated with both ancestry and infection are shown in red. (B) Allele frequency for the six significant SNPs in the cohort of AA (n = 24) and EA (n = 35) donors. (C) Genotypes and/or haplotypes of AA (n = 15) and EA (n = 21) donors for SNPs that are significantly associated with both ancestry and infection in epidermis at 24 h. (D) Flow cytometry dot plots of epidermal cells from an AA and EA donor stained with and without 2H2 antibody to prM protein 24 h after mock or DENV-2 infection of cell suspensions. (E) Association between the percent infection of epidermal cells and the proportion of European ancestry of donors (n = 13). R-squared and P values were determined by Pearson correlation and simple linear regression. (F) DENV genomes expressed as FFU/mL (Left) and mRNA expression of OAS1, OAS2, and OAS3 in epidermal cells 24 h after mock or DENV-2 infection relative to baseline (control 0 h) (Right) in AA (n = 4) and EA (n = 5) individuals. (G) Relative expression of OAS1, OAS2 and OAS3 mRNA in epidermal cells 18 h after control or 2,000 U/mL recombinant IFN-α stimulation in AA (n = 5) and EA (n = 7) individuals. Horizontal lines represent medians. Within-group analyses were done using the Wilcoxon matched pairs signed rank test and between-group analyses were done using a Mann–Whitney U test.

The IFN-inducible OAS genes are critical in the antiviral response through activation of the RNA cleavage pathway. To study the interaction between genetic ancestry, OAS, and DENV infection in more detail, we generated epidermal cell suspensions from EA and AA skin and exposed them to DENV-2 at a multiplicity of infection of 2. At 24 h infection was detected by flow cytometry in epidermal cells from both EA and AA skin suspensions, however infection was greater in cells from EA individuals, which correlated with the proportion of European ancestry, consistent with in situ data (Fig. 3 D and E). Infection resulted in greater virus replication in EA versus AA epidermal cell suspensions as measured by quantitative PCR, as well as greater induction of OAS3 mRNA expression, which was undetectable in virus-infected AA epidermal cells. Virus infection induced OAS1 mRNA expression in EA skin and OAS2 expression in both EA and AA skin, but these did not differ statistically between groups (Fig. 3F). To determine whether differential OAS expression was a general ancestry-related IFN response difference rather than being specific to DENV, we stimulated epidermal cell suspensions with type I IFN and measured mRNA expression. OAS1, OAS2 and OAS3 were all induced in response to type I IFN in EA and AA skin, but the effect was greater in EA skin for all three genes, reaching significance with OAS3 (Fig. 3G). Together, these findings show that polymorphic genes associated with infection and immunity contribute both directly and indirectly to ancestry-related differences in DENV infection in skin, and that OAS genes may not provide sufficient antiviral activity to control virus replication in skin cells.

Discussion

Our findings show that the early innate immune response of human skin to DENV infection is strongly influenced by genetic ancestry. DENV infection elicited a much stronger inflammatory response in EA skin, leading to increased virus replication and spread. The effect was apparent for all DENV serotypes and for ZIKV, indicating that flaviviruses in general induce differential innate immune responses in human skin as a function of genetic ancestry. Importantly, blocking antiviral and providing proinflammatory cytokines in infected AA skin resulted in resident macrophage infiltration and infection equivalent to that of EA skin, indicating that it is the nature of the innate immune response rather than skin anatomy and physiology that is responsible for differences in DENV infection.

The results suggest that both cell-intrinsic and extrinsic factors contribute to the increase in DENV replication in human skin with increasing European ancestry. In situ infection of keratinocytes in EA skin was approximately twice that of AA skin, and infection of keratinocytes in isolation increased with increasing European ancestry, indicating that keratinocytes in individuals of European descent are intrinsically more susceptible to DENV infection. In contrast, increased infection of Langerhans cells, dermal macrophages, and DC in EA skin was dependent upon infiltration of these skin-resident cells into the site of virus inoculation, with the number of infected cells increasing as the DENV/mock cell ratio increased. A major driver of local myeloid cell recruitment is IL-1β, which mediates expression of a range of chemokines that recruit macrophages as well as neutrophils (42). IL-β is expressed largely by infected keratinocytes and has an autocrine effect, as blocking IL-1β reduces but does not eliminate DENV replication in keratinocytes in situ (13). The association between reduced ADE of DENV infection and lower expression of CD32 in AA dermis suggests that dermal myeloid cells from AA donors are not resistant to ADE but rather may bind relatively fewer virus–antibody complexes compared to EA dermal cells. Whether isolated Langerhans cells, DC and macrophages are intrinsically more susceptible to DENV infection with increasing European ancestry remains to be determined.

Influenza virus infection of blood mononuclear cells induces an increased IFN-α response with increasing European ancestry, which reduces viral titers (25). Immune pathways associated with the IFN-α response were also enriched in DENV-infected skin with increasing European ancestry; however, our data suggest that in the cutaneous environment, the IFN response facilitates rather than suppresses DENV replication. IFN response genes were enriched in infected EA skin but failed to reduce the number of infected cells in situ. Both DENV and exogenous type I IFN induced significantly greater levels of OAS3 in EA epidermal cell suspensions, yet increased OAS3 expression was associated with increased virus replication. In addition to antiviral activities, OAS3 has inflammatory properties that may indirectly enhance virus replication. OAS elicits increased production of proinflammatory cytokines in a human macrophage cell line in response to Mycobacterium tuberculosis (43), and high expression of OAS3 in lungs is associated with a higher risk of severe inflammatory lung injury in patients with COVID-19 (44). Increased OAS activity is also associated with increased disease severity in dengue patients (45). It is not currently known whether OAS or other IFN-stimulated genes have any direct effect on DENV infection in skin myeloid cells.

The noninflammatory response of AA skin to DENV contrasts with the increased inflammatory response of blood mononuclear cells to bacterial pathogens with increasing African ancestry (24). It is possible that the innate immune response to bacterial pathogens and flaviviruses in African and European descendants is distinct, regardless of tissue compartment, or that the immune response to pathogens in general in skin is distinct relative to blood independent of ancestry. Alternatively, flaviviruses and skin together may create a unique noninflammatory environment with increasing African ancestry. In support of this hypothesis, while dengue is generally a nonlethal disease that has a relatively short evolutionary history in humans of only a few hundred years (46), yellow fever, caused by a related flavivirus, originated in Africa several millennia ago and is one of the most lethal of human diseases (47). Molecular pathways and genetic polymorphisms that limited skin inflammation in response to yellow fever virus infection would have had a strong selective advantage in ancient Africans, resulting in differences in the response to flaviviruses in modern humans outside of Africa. Europeans would not have been exposed to arboviral infections in the New World in ancient times and would not have had the same selective pressure to evolve a reduced skin inflammatory response. Indeed, during 19th century epidemics of yellow fever in the United States, the disease was significantly more likely to be fatal in European descendants than in those of African descent (48). Dissecting out the different relationships between pathogen type, tissue compartment and genetic ancestry will require further comparative studies.

It is likely that ancestry-related differences in DENV infection in skin relate directly to differences in dengue disease. The increased virus replication in both primary and secondary infections together with increased migration of myeloid cells out of skin in individuals of European descent would increase virus dissemination via lymphatic vessels. Infected myeloid cells that exit the skin would produce IL-6, TNF-α, and IL-1β, contributing to the inflammatory cytokinemia characteristic of severe dengue (49). In secondary dengue, ADE of infection in circulating monocytes, which is greater in European than African descendants (50), would further increase the likelihood of severe disease in Europeans. Polymorphisms in RXRA, which encodes retinoid X receptor-α (RXRα), were highly significant with respect to ancestry and DENV infection in skin in our study and are associated with protection from severe dengue (40), suggesting a role for RXRα in both cutaneous virus replication and systemic disease. RXRα is a nuclear receptor and transcription factor that has a number of ligands and a broad range of immunologic effects, from attenuating host antiviral responses (51) to upregulating inflammatory chemokine expression (52). Expression of RXRA mRNA in blood decreases with acute dengue illness, but the relationship between expression and disease severity is not clear (40).

The differential effects of genetic ancestry seen with cutaneous DENV infection have implications beyond pathogenesis. Knowledge of a dengue patient’s genetic ancestry could be useful in predicting disease severity in the critical stages of acute infection. Genetic ancestry is not considered in evaluating the efficacy of dengue vaccines, all of which currently are live attenuated vaccines given by injection into skin (53). Candidate vaccines may have different levels of efficacy and frequency of vaccine-related adverse events in different populations, as ancestry-related differences in the cutaneous inflammatory response could affect attenuated virus replication, immunogenicity, and vaccine responsiveness.

In summary, our findings show that flavivirus infection induces a stronger inflammatory response in skin with increasing European ancestry, which parallels the increased burden of dengue in populations of European descent. Further elaboration of molecular pathways and factors that differentiate the outcome of infection based on genetic ancestry, including evaluation of the role of other cell types such as mast cells, T cells, and NK cells that participate in DENV infection and control in skin, will lead to novel ways to intervene therapeutically or prophylactically to prevent severe dengue disease.

Materials and Methods

Donors.

Abdominal skin from 59 healthy individuals aged 24 to 66 y (median = 45 y) from the Pittsburgh, Pennsylvania area undergoing excess skin removal procedures were used in the study (SI Appendix, Table S1). Anonymous donors authorized resected tissues to be used for scientific purposes and were not identifiable by the investigative team. The research therefore did not involve human subjects and was Institutional Review Board exempt. Donors were confirmed to be seronegative for DENV-1-4 and Zika virus by in-house NS1 and envelope indirect enzyme-linked immunosorbent assays (ELISA) run on residual blood collected from dermal vessels.

Estimation of Biogeographical Ancestry.

We used a set of 128 unlinked SNPs in host genes (r² between all pairs <0.1) to estimate genetic ancestry. These SNPs have been validated as ancestry-informative markers in a large group of individuals in the Pittsburgh region (54). SNP genotypes were determined from DNA isolated from skin specimens using the Agena MassARRAY system for targeted genotyping (55). The proportion of European and African genetic ancestry for each skin donor was estimated using the supervised clustering algorithm in ADMIXTURE (v1.3.0) (28). We included 809 samples from individuals of diverse ancestry (29) or population structure in ADMIXTURE analysis. Samples from individuals of African descent included 69 Biaka Central African, 77 Yoruban West African, 32 Ethiopian East African, and 89 Americans of African ancestry in the United States. Samples from individuals of European descent included 107 Toscani in Italy, 114 Irish, and 89 Americans of European ancestry in the USA. Samples from East Asia included 118 Laotian and 58 Chinese Americans in San Francisco, USA. Samples from 56 Indigenous Americans were also included as reference populations (SI Appendix, Fig. S7). The ancestry of donors was calculated assuming four ancestral clusters (K = 4) after cross-validation procedures (SI Appendix, Fig. S7). Principal component analysis of genotyping data from each skin donor included in this study was merged with the genotyping data from the reference populations to visualize ancestral clusters. For analyses, the estimated ancestry proportions were used as a continuous scale of European ancestry or as dichotomous groups using a threshold of 60% African or European ancestry, respectively.

Dengue Risk Factor Polymorphisms.

DNA isolated from skin specimens (Qiagen, 69504) was used to determine the genotype of 60 SNPs in 29 different genes (SI Appendix, Dataset S6) using the Agena MassARRAY system. We reconstructed haplotypes from population genotypes data using PHASE software (v. 2.1). We merged genotyping data from our study population with published data from the 1000 Genomes Project phase 3 (56). These included individuals of European (n = 187) and African (n = 161) ancestry populations.

Viruses.

DENV-1 (strain Hawaii/1944), DENV-2 (strain Thailand/16681/1964), DENV-3 (strain Philippines/H87/1956), and DENV-4 (strain Puerto Rico/BC258/97/1994) stocks were grown in C6/36 insect cells and culture supernatant was collected and concentrated using standard methods. Virus titers were determined using a modified FFU immunoperoxidase assay using Vero cells (13). The ZIKV (strain Brazil/PE243/2015) was propagated in Vero cells and titers were obtained by plaque assay using standard methods (26).

Skin Processing and Virus Inoculation In Situ.

Residual adipose tissue was trimmed from specimens within 2 h of surgery and tissue was cut into 1 in² pieces. Virus was inoculated into the skin at a dose of 500,000 FFU or 5,000 FFU for ADE experiments in a volume of 50 µL by repeated stabbing with a bifurcated allergy testing needle (Precision Medical Product). Needle inoculation recapitulates natural delivery of DENV via the probing of infected female Aedes aegypti mosquitoes, the principal arthropod vector (18). Inoculated explants were incubated at 37 °C for 2 h after which residual surface inoculum was wiped off. Explants were incubated dermis side down on wire mesh at the air–liquid interface in complete medium (RPMI 1640, 10% fetal bovine serum, 2 mM L-glutamine, 100 U/mL penicillin, 100 µg/mL streptomycin, 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 1% sodium pyruvate, and 1% nonessential amino acids) for various intervals up to 48 h.

Microneedle Arrays.

Human immune serum was formulated into tip-loaded dissolvable 3:2 carboxymethyl cellulose/trehalose microneedle arrays fabricated as described previously (13). Arrays were loaded with a 1:40 dilution of pooled serum collected from individuals who experienced primary DENV-3 infection during a prospective cohort study in Brazil and were confirmed to have DENV-3 monotypic immunity. Control arrays contained a 1:40 dilution of human serum obtained commercially (Innovative Research, IHUIGGGF) and confirmed as flavivirus-naïve by ELISA and neutralization assays. Arrays were manually applied to explants and then removed after 15 min, leaving the dissolved needle tips in place. Virus was inoculated into the skin immediately following microneedle array removal. For cytokine studies, individual arrays were formulated to contain 100 IU of recombinant human IFN-α (PBL Assay Science, 11200-1), 50 ng of recombinant human IL-1β (R&D Systems, 201-LB-005-CF), 20 μg of neutralizing antibody to human type 1 IFN (PBL Assay Science, 39000-1), 30 ng of neutralizing antibody to human IL-1β (R&D Systems, MAB601-25), 30 ng isotype control antibody (R&D Systems, MAB003 and MAB002), or a combination of these. For these experiments, arrays were applied prior to virus inoculation.

Immunohistochemistry and Quantitative Image Analysis.

Immunohistochemistry was performed as described (13) using the following antibodies: polyclonal rabbit anti-pan DENV NS3 antibody (provided by Sujan Shresta, La Jolla Institute for Allergy and Immunology, San Diego, CA, USA), polyclonal rabbit anti-ZIKV NS2b antibody (Genetex, GTX133308), anti-cytokeratin pan type I monoclonal antibody (AE1; MA5-13144, Thermo Fisher Scientific), anti-Langerin/CD207 monoclonal antibody (DCGM4/122D5; DDX0363P-100, Novus Biologicals), anti-CD163 monoclonal antibody (5C6-FAT; BM4041, Novus Biologicals), anti-CD1c monoclonal antibody (L161; ab190305, Abcam), and anti-IL-1β antibody (6E10; NB120-8319, Novus Biologicals). Secondary antibodies were from Thermo Fisher Scientific and included goat anti-mouse IgG1, Alexa Fluor 546 (A21123), goat anti-mouse IgG2a Alexa Fluor 647 (A21241), and donkey anti-rabbit IgG, Alexa Fluor 647 (A31573). Slides were stained with Hoechst dye (Thermo Fisher Scientific) and imaged on an Olympus Fluoview 1000 confocal microscope. We used Nikon NIS-Elements AR 5.30.06 software for quantitative image analysis. Data for each section were collected from a minimum of 10 confocal images. For quantification, all images were taken under a 40× magnification and a scan size of 1,024 × 1,024 pixels for a total image area of 1,048,576 pixels². Image analysis was performed by creating a binary layer to threshold for positive staining in each channel (SI Appendix, Fig. S1). For each image, we delineated the area occupied by the epidermis and dermis as the region of interest (ROI) to restrict binary objects to within these areas. The binary area within the ROI was then recorded as pixel², normalized to total tissue area (epidermis and dermis), and multiplied to 100 to express data as a percentage.

Dengue and Zika ELISA.

We obtained serum samples using residual blood collected from dermal vessels. Samples were tested for DENV and ZIKV antibodies by a single dilution (1:100) using in-house indirect ELISA. High-binding, half-area, 96-well- polystyrene plates were coated overnight at 4 °C. Viral antigens included cell culture lysates and NS1 from a balanced mixture of DENV1-4 or ZIKV. Viral strains used to prepare viral lysates were the same as those used for skin inoculations. Wells were blocked with either nonfat dry milk or bovine serum albumin prepared at 5% (w/v) in phosphate buffer saline with 0.1% (v/v) Tween-20 (PBS-T). Samples were incubated for 1 h, and then plates were then washed five times with PBS-T. Plates were incubated for 1 h with horseradish peroxidase–linked anti-human IgG (Jackson ImmunoResearch, 109-036-003) before the addition of Peroxidase substrate 3,3′,5,5′-tetramethylbenzidine. The reaction was stopped with 1 M hydrochloric acid solution. The cutoff for positivity was determined based on the average of the negative control plus three times the SD.

RNA Extraction, Library Preparation, and Sequencing.

Total RNA was extracted from 4 mm-punch biopsies of uninfected and DENV-2-infected whole skin using the RNeasy Mini Kit (Qiagen, 74104) according to the manufacturer’s instructions. We evaluated RNA quantity using the Qubit Fluorometric Quantification assay (ThermoFisher Scientific, Q32852) and RNA integrity using the TapeStation RNA ScreenTape analysis (Agilient Technologies). Samples with RNA integrity number >8 were included for further analysis. RNA-sequencing libraries were prepared using the TruSeq Stranded Total RNA protocol (Illumina). complementary DNA (cDNA) libraries were sequenced 75 base pair single end reads on a NextSeq 500 system (Illumina) to an average depth of 30 million reads per experimental condition.

RNA-Sequencing Data Processing and Analysis.

RNA-sequencing data were provided as fastq files and are available through the National Center for Biotechnology Information Sequence Read Archive with the BioProject accession number PRJNA1209924 (57). We performed an initial quality control using FastQC to assess read quality. We then trimmed adaptor sequences and low-quality read regions using the Trimmomatic software (version 0.4). The resulting cleaned fastq read files were used as input to the Kallisto package and used together with an indexed human GRCh38 transcriptome reference (EnsDb.Hsapiens.v105) to produce a pseudoaligned transcript abundance file for each sample. Abundance files were then loaded into R using the Tximport package for gene-level expression estimates. Low abundance was filtered out, and the remaining transcript counts (median logTPM > 1) were normalized using the EdgeR package to produce counts-per-million values for each sample. To explore genes to which the response to DENV-2 infection is associated with biogeographical ancestry following published available scripts. We log-transformed data and used the voom function to estimate mean–variance relationship and incorporate precision weights. Expression estimates were fitted to a nested linear model using the lmFit and eBayes functions in the limma package.

Gene Set Enrichment Analysis.

We used the ClueGO (v.2.5.9) Cytoscape (v.3.9.1) module to explore the enrichment of ontology terms for which the response to DENV infection significantly correlated with ancestry. For this analysis, we included genes with a stronger or decreased response to infection (log2 fold-change greater than two in each direction of the effect) and an FDR < 0.05. We included gene ontology terms related to biological processes and assumed the following parameters: functional analysis mode, no gene ontology term fusion, minimum number of genes of five, enrichment/depletion (two-sided hypergeometric test) with Benjamini–Hochberg (BH) P-value correction as statistical test. For graphical representation, the ClueGO clustering functionality was used with a Kappa threshold score for considering or rejecting term-to-term links set to 0.4. Only gene ontology terms enriched at FDR < 0.05 were reported. We performed gene set enrichment analysis using the H hallmark gene sets and the R package fgsea. We ranked t-statistics obtained from the TopTable function in limma, performed enrichment (parameters: minSize = 15, maxSize = 500, nperm = 100,000), and collected enrichments scores and BH adjusted P-values outputs.

Generation of Skin Cell Suspensions and Infection and Stimulation with IFN-α.

Epidermal sheets were collected from split-thickness skin using a Goulian knife and a 0.012” guard and incubated in Dispase II solution (2.5U/mL in PBS; Sigma-Aldrich, D4693) for 1 h. Epidermal sheets were then separated from the dermis and incubated in prewarmed 0.25% trypsin to obtain single cell epidermal suspensions. Dermal tissue was incubated overnight in 1 mg/mL of collagenase IV (Worthington Biochemical, LS004188) followed by mechanical disruption with a gentleMACS Dissociator (Miltenyi Biotech) to generate dermal single cell suspensions. For infection experiments, epidermal cells were infected with DENV-2 (strain Thailand/16681/1964) at an MOI of two and incubated at 37 °C for 24 h. For IFN stimulation experiments, epidermal cell suspensions were treated with recombinant human IFN-α at a concentration of 2,000 U/mL (PBL Assay Science, 11200-1) and incubated at 37 °C for 18 h.

Flow Cytometric Analysis.

Epidermal cells were incubated with Live/Dead Yellow dye (ThermoFisher Scientific, L34959) and then fixed and permeabilized using the Cytofix/Cytoperm™ Fixation/Permeabilization Kit (BD Bioscience, 554714). Cells were stained with an anti-DENV prM antibody (D3-2H2-9-21, MAB8705, Sigma-Aldrich), followed by R-phycoerythrin labeled goat anti-mouse IgG (ThermoFisher Scientific, 31863). Dermal cells were incubated with a master mix of the following antibodies: CD45-R718 (BD Bioscience, 566962), CD163-AF647 (BD Bioscience, 562669), CD1c-BB515 (BD Bioscience, 565055), and either CD32-PE (BD Bioscience, 552884) or CD64-PerCPcy5.5 (BD Bioscience, 561194). Flow cytometry was performed on a BD FACSymphony A5 SE (BD Bioscience) and data was analyzed using FlowJo Software (version 10.10.0).

Quantitative Real-Time PCR.

Total RNA was isolated from epidermal cell suspensions using the RNeasy Mini Kit (Qiagen, 74104) and reverse transcribed using the Maxima First-Strand cDNA Synthesis Kit (Thermo Fisher Scientific, K1641). cDNA was subjected to RT-qPCR using the Platinum SYBR Green qPCR SuperMix (Thermo Fisher Scientific, 11733038) and the StepOnePlus system (Applied Biosystems). The primer sequences for OAS1, OAS2, OAS3, and β-actin are described in SI Appendix, Table S2. The cycle threshold value for each gene was normalized to the cycle threshold value of the reference gene (β-actin), and the relative fold change in expression was calculated using the comparative 2^–ΔΔCT method. To quantify virus genomes, quantitative PCR was undertaken using the GoTaq Probe qPCR kit (Promega) with amplification in the Applied Biosystems QuantStudio 6 Flex real-time PCR system (Thermo Fisher Scientific). DENV genomes were determined by interpolation onto an internal standard curve produced using 10-fold serial dilutions of a synthetic DENV-2 fragment based on the prototype strain used for infections in skin. Equivalent amounts of RNA were used for each sample and virus titers were expressed as DENV FFU equivalents per mL (26).

Statistical Analysis.

We compared repeated measurements within individuals using one-way repeated measures ANOVA Tukey’s multiple comparisons, or the Friedman test Dunn multiple comparisons as appropriate. Kruskal–Wallis one-way ANOVA followed by Dunn’s multiple-comparisons test was used for multiple comparisons of independent groups. The unpaired t test or the Mann–Whitney U test was used to compare two independent groups as appropriate. Within-group analyses were done using the Wilcoxon signed-rank test. We used the Pearson correlation and simple linear regressions to measure the strength of the relationship between biogeographical ancestry and dengue infection. Results from multiple experiments are presented as mean or median as appropriate. Repeated measures relating to analysis of SNPs were corrected using the BH procedure with a false discovery rate of 10%. Statistical analyses were performed in GraphPad Prism (v. 10.4.0) unless noted otherwise.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Data, Materials, and Software Availability

All data in the manuscript and supplementary materials are available as Source Data. Fastq files from RNA sequencing in this study are available at the National Center for Biotechnology Information Sequence Read Archive with the Bioproject accession number PRJNA1209924 (57).