Pan-viral protection against arboviruses by activating skin macrophages at the inoculation site

See allHide authors and affiliations

Science Translational Medicine  22 Jan 2020:
Vol. 12, Issue 527, eaax2421
DOI: 10.1126/scitranslmed.aax2421

Taking the bite out of vector-borne infections

Arthropods are the most abundant animals on Earth and can transmit a variety of diseases such as dengue or West Nile virus to humans. To broadly combat these infections, Bryden et al. investigated whether the immune reaction at the site of mosquito bites could be manipulated to restrict viral dissemination. They observed that local TLR7 activation shortly after infection dampened replication of a model alphavirus in mice. This held true for clinically relevant arboviruses and also in human skin explants. The viral restriction was due to activation of skin macrophages and heightened type I interferon production. Topical TLR7 activation after mosquito bites could be a broad-acting approach to abrogate arboviruses.


Arthropod-borne viruses (arboviruses) are important human pathogens for which there are no specific antiviral medicines. The abundance of genetically distinct arbovirus species, coupled with the unpredictable nature of their outbreaks, has made the development of virus-specific treatments challenging. Instead, we have defined and targeted a key aspect of the host innate immune response to virus at the arthropod bite that is common to all arbovirus infections, potentially circumventing the need for virus-specific therapies. Using mouse models and human skin explants, we identify innate immune responses by dermal macrophages in the skin as a key determinant of disease severity. Post-exposure treatment of the inoculation site by a topical TLR7 agonist suppressed both the local and subsequent systemic course of infection with a variety of arboviruses from the Alphavirus, Flavivirus, and Orthobunyavirus genera. Clinical outcome was improved in mice after infection with a model alphavirus. In the absence of treatment, antiviral interferon expression to virus in the skin was restricted to dermal dendritic cells. In contrast, stimulating the more populous skin-resident macrophages with a TLR7 agonist elicited protective responses in key cellular targets of virus that otherwise proficiently replicated virus. By defining and targeting a key aspect of the innate immune response to virus at the mosquito bite site, we have identified a putative new strategy for limiting disease after infection with a variety of genetically distinct arboviruses.


Emerging and reemerging arboviruses pose an increasing threat to human health. There has been a substantial increase in both the incidence and geographical range of medically important arboviruses spread by mosquitoes, which infect hundreds of millions of people each year and include the Zika virus (ZIKV), dengue virus (DENV), and chikungunya virus (CHIKV). Arboviruses are a large, genetically diverse group of viruses that cause a wide spectrum of diseases in humans (15). Despite their genetic diversity, it is nonetheless difficult to clinically differentiate between these infections in the early stages of disease because they are either asymptomatic or present as a nonspecific febrile viral illness. In many geographic areas, this is compounded by the widespread cocirculation of distinct species of arboviruses in the same geographic area (6). Together, these factors complicate the use of putative virus-specific antivirals, which, for acute infections, are most often only efficacious when given during early stages (7). This, when combined with our inability to accurately predict the timing and location of future epidemics (8), makes stockpiling future virus-specific drugs and vaccines difficult. Currently, there are few vaccines and no antivirals available for arbovirus infections. We suggest that, because of the diversity of arbovirus genetics, their common clinical features, and their unpredictable epidemiology, the development of a pan-viral medicine that is efficacious for multiple arbovirus infections would be highly advantageous.

Infected mosquitoes deposit virus into the skin dermis as they probe for a blood meal, triggering activation of distinct inflammatory pathways in response to mosquito biting and to virus sensing (911). We and others have demonstrated that host responses in the skin to mosquito bites, or mosquito saliva, have a defining influence on the systemic course of infection for a wide variety of genetically distinct arboviruses including flaviviruses, alphaviruses, and bunyaviruses (1014). Hence, this is a key stage of infection during which virus replicates rapidly before disseminating to the blood and remote tissues. Most of the virus in the blood at 24 hours after infection of the skin is likely derived from replication at this site (9). However, some of the original virus inoculum also disseminates directly to the blood and the draining lymph node (dLN) where replication is also established before systemic dissemination (9, 13). Thus, it is not clear what role skin-specific innate immune responses, activated by virus sensing at the mosquito bite, have on modulating the subsequent systemic course. In this study, we wanted to define the relevance of skin virus-sensing pathways and determine whether this can be targeted to modulate outcome of infection.

Innate immune sensing of virus activates immune pathways that are distinct to those activated by mosquito biting, resulting in the expression of type I interferons (IFNs) and the antiviral genes they up-regulate, IFN-stimulated genes (ISGs). Despite the evolution of multiple strategies by arboviruses to counteract IFN, they are nonetheless highly sensitive to them, suggesting that therapies designed to target these pathways could be generalizable to a range of viruses (1518). Systemic administration of type I IFN has been used to treat hepatitis C virus, although febrile-like side effects are common (19), and hence, it is not suitable for, e.g., long-term prophylactic use. However, previous attempts to inhibit arbovirus infection by systemic administration of innate immune agonists or type I IFN have only been efficacious when given before infection (2024). In contrast, we suggest targeting processes at the inoculation site, which represents a discrete identifiable locale that can be targeted after a mosquito bite, before the systemic dissemination of virus. However, targeting these pathways is challenging because the coordination of early innate immune responses to virus at mosquito bites is not well defined.

Studying innate immune responses at the tissue- and system-wide level to arboviruses that are medically important in humans has been frequently complicated by their inability to replicate in immunocompetent mice (10). Therefore, we chose to primarily study host innate immune responses to a prototypic model arbovirus, Semliki Forest virus (SFV; genus Alphavirus). Unlike most human arboviral pathogens, SFV is capable of efficiently replicating, disseminating systemically within, and causing clinically observable disease in immunocompetent mice (25). SFV is a close relative of CHIKV that has a large number of genetically modified clones and has been used extensively to study host response to infection (5). SFV replicates quickly after infection of mouse skin, with viremia peaking by 24 hours post-infection (hpi). Dissemination of SFV to brain tissue can result in encephalitis, neurological signs, and death (9, 25). Clinical outcome of SFV infection in mice most likely reflects the efficiency by which virus seeds brain tissue.

In this study, we define key aspects of host response to virus at mosquito bites and demonstrate that this can be therapeutically modulated to suppress viral replication and the development of clinical disease. We suggest that for virus to successfully disseminate from the skin to the blood at a sufficient level to induce a high-titer viremia, sufficiently robust replication must occur at the inoculation site before a host IFN-induced antiviral state. Using an immunocompetent mouse model of arbovirus infection and ex vivo infection of human skin explants, we investigate therapeutic manipulation of skin IFN pathways by post-exposure topical application of a widely used generic innate immune agonist for infection with multiple genetically distinct arboviruses.


Skin innate immune responses to virus infection at the inoculation site are a key determinant of the systemic course

We wanted to determine whether targeting early innate immune responses in the skin had any observable effect on the later systemic course of infection. However, it was first necessary to define the kinetics and magnitude of endogenous host ISG responses to infection with SFV6 (a virulent strain of SFV), in the absence of therapies. After infection of the foot skin, ISG expression in response to virus was slow and of low magnitude, with ISG expression not significantly elevated until 24 hpi or later, peaking at 48 hpi (Fig. 1, A to D). In comparison, the dLN rapidly and robustly up-regulated ISG, with amounts peaking by 16 hpi. Thus, up-regulation of antiviral ISG in the skin could be detected only after virus had disseminated systemically (9).

Fig. 1 Immunomodulation of the skin inoculation site enhances host resistance to virus infection.

(A to D) Mice were infected subcutaneously (s.c.) with 250 plaque-forming units (PFU) of SFV6 in the upper skin of the left foot. Copy number of host (A) ifnb1, (B) cxcl10, (C) ifng, (D) cxcl9, and 18S RNA in the skin and the draining popliteal lymph node (dLN) was determined by qPCR at various time points. (E to K) One hour before infection with SFV6 viruses, mice were pretreated with a single 2 mg topical administration of IMQ or injected subcutaneously with 6 μg of either aqueous IMQ, poly(dA:dT), or poly(I:C). (E to G) At 24 hpi, copy number of SFV RNA and host 18S were determined by qPCR at 24 hpi (E and F), and virus titers in the serum were quantified by plaque assay (G). Non-dLN was the popliteal LN contralateral to infection (n = 6). (H to K) Survival of mice was assessed when defined clinical end points were reached (n = 10). PID, post-infection day. Groups that were significantly different to 0 hour (A to D) or to SFV infection alone (E to K) are marked; *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant [Kruskal-Wallis and log-rank (Mantel Cox) test].

It is not clear what relevance skin-specific innate immune responses to arbovirus have on the subsequent systemic course of infection, or whether virus replication in the skin is absolutely necessary to establish viremia. To determine whether therapeutic induction of antiviral innate immune pathways at the inoculation site could have any effect on the systemic course of infection, we injected or topically applied a range of innate immune agonists to a defined cutaneous site 1 hour before infection with SFV6 (Fig. 1, E to K). The TLR7 agonist imiquimod (IMQ) was either administered as a subcutaneous injection or applied topically as a cream (Aldara, topical IMQ). To assess success of virus infection, viral RNA was measured by quantitative polymerase chain reaction (qPCR), and infectious virus was assayed by plaque assay. Virus RNA measured by qPCR represented the sum total of genome and replicated viral transcripts. To define the contribution made by the original inoculum toward the level of viral RNA assayed by qPCR at 24 hpi, mice were inoculated with the ultraviolet (UV)–inactivated virus (which can infect cells but does not replicate). Copy numbers of viral RNA were close to the detection limit of the assay at 24 hpi, suggesting that nonreplicating virus was rapidly cleared in vivo (fig. S1A).

At 24 hpi, only topical IMQ and polyinosinic-polycytidylic acid [poly(I:C)]–treated mice exhibited lower viral RNA at the skin inoculation site, whereas all innate immune agonists decreased virus in the dLN (Fig. 1E). Poly(I:C) was particularly potent at decreasing viral RNA in most tissues analyzed, although viremia (Fig. 1G) and survival to infection (Fig. 1K) were not significantly affected. In comparison, only topical IMQ significantly decreased viral replication at both local tissues (skin and dLN, P < 0.05) and tissues remote from the inoculation site (non-dLN, P < 0.01; spleen, P < 0.01; and brain, P < 0.05), reduced infectious virus in blood (P < 0.001), and limited the development of clinical disease (P < 0.05; Fig. 1H). Because SFV6 is highly virulent in mice, it does not model all arboviral disease seen in humans that are often nonfatal. Therefore, to better model the human situation, we also tested the ability of IMQ to suppress infection with SFV4, a less virulent strain than SFV6 (fig. S1B). Again, application of topical IMQ at the site of SFV4 inoculation significantly (P < 0.01) suppressed infection.

Post-exposure targeting of the inoculation site by topical application of an innate immune agonist protects mice from infection with virus

Treatment modalities that target early stages of infection are likely to involve post-exposure treatment, e.g., once the erythema of a bite is apparent. Because prebite application of topical IMQ was most efficacious in increasing host resistance to infection, we next determined its efficacy when applied after infection. In addition, because arbovirus infection of the skin always occurs in the context of an arthropod bite, we used a mouse model that additionally incorporates biting Aedes aegypti mosquitoes. Host response to mosquito bites includes edema and an influx of leukocytes that enhances host susceptibility to infection with virus (9, 10, 26). Because topical IMQ is a cream that can be removed by cage bedding, it was first applied at 1 hpi and then reapplied once at 6 hpi to maintain dosing. In this model, application of topical IMQ was highly efficacious in lowering quantities of SFV6 RNA in all tissues analyzed and infectious virus in the blood by 24 hpi (Fig. 2, A and B). For mice in which infection was allowed to progress, topical IMQ application resulted in a significant delay to the onset of neurological signs (P < 0.0001; Fig. 2C). Consistent with this, analyses of brain tissue at post-infection day 7 revealed extensive expression of SFV6-encoded mCherry throughout the brain in untreated mice compared to treated mice (Fig. 2D).

Fig. 2 Targeted post-exposure immunomodulation suppresses the local and systemic course of infection and improves clinical outcome in mice.

(A to D) Mosquito-bitten mouse skin was infected with 250 PFU of SFV6 and treated with topical IMQ at 1 hpi (with one reapplication at 6 hpi) or left untreated. (A) SFV RNA and host 18S were quantified by qPCR at 24 hpi (n = 7). (B) Plaque assay of serum at 24 hpi (n = 7). (C) Survival of mice (n = 10). (D) Midsagittal sections of the brain from mice (n = 5) infected with SFV6-mCherry (red) stained with DAPI (blue); top composite image was assembled from multiple photographs, whereas bottom images show typical individual images at higher magnification. (E to G) Mosquito-bitten mouse skin was infected with 10,000 PFU of SFV4 and treated with topical IMQ at 1 hpi (with one reapplication at 6 hpi) or left untreated. (E) SFV RNA and host 18S were quantified by qPCR at 24 hpi. (F) Serum virus quantified by plaque assay (n = 6). (G) Survival of mice (n = 11). (H) Mosquito-bitten mouse skin was infected with 10,000 PFU of SFV4 and treated with topical IMQ at either 5 or 10 hpi. SFV RNA and host 18S were quantified by qPCR at 24 hpi (n = 5 to 6). (I) Mosquito-bitten mouse skin was infected with 250 PFU of SFV6 and then either left untreated, treated with topical IMQ at 1 hpi (with one reapplication at 6 hpi), or similarly treated with topical IMQ, and residual cream was removed by an aqueous detergent wash 30 min after second dosing or administered as an oral gavage of topical IMQ (10% of total dose in water) (n = 6). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 [Mann-Whitney, Kruskal-Wallis or log-rank (Mantel Cox) test].

Similarly, topical IMQ at 1 hpi with the less virulent SFV4 strain significantly reduced virus by 24 hpi in tissues and the blood (P < 0.01) and substantially decreased mortality (P < 0.001), with the majority of mice surviving infection (Fig. 2, E to G). Protection by topical IMQ was time-limited because treatment 5 hpi significantly lowered viral RNA (P < 0.05 in the skin and the spleen), but treatment delayed until 10 hpi did not significantly modulate the level of viral replication or dissemination by 24 hpi (Fig. 2H).

Although these experiments were designed to test the ability of immune modulators to target the inoculation site, it is possible that mice inadvertently consumed some topical IMQ orally, e.g., during grooming. In the above studies, mice were given topical IMQ at 1 hpi, during which the mice were anesthetized. Mice did not fully recover from anesthesia until after 5 hpi, during which the majority of topical IMQ would have been absorbed and was not available for oral consumption. Nonetheless, to define the contribution that inadvertent oral consumption of topical IMQ had on susceptibility to virus infection, we washed away unabsorbed topical IMQ from the inoculation site using a mild detergent aqueous wash immediately before recovery from anesthesia. In addition, a separate group of mice received topical IMQ as an oral gavage (10% Aldara in water). Mice washed at 5 hpi still exhibited significantly (P < 0.05) reduced amounts of virus at the skin inoculation site, the spleen, and the blood by 24 hpi. In comparison, those mice given IMQ as an oral gavage exhibited no significant reduction of virus (Fig. 2I). Together, this suggests that topical IMQ was primarily modulating the course of virus infection through responses activated at the site of application.

The above studies involved virus infection of mouse skin by needle at sites of mosquito bites, because mosquito saliva has profound modulatory effects on host susceptibility to virus (10, 11). Because infection of the mosquito salivary gland by virus can alter salivary protein content (27), it is possible that virus inoculated by infected mosquitoes themselves has additional effects on modulating virus infection in the mammalian host. To determine whether topical IMQ is also efficacious in altering host susceptibility to virus in the presence of saliva from infected mosquitoes, we obtained saliva from SFV-infected mosquitoes and coinoculated it with an SFV6 into mouse skin (fig. S1C). To ensure that we inoculated similar quantities of virus as in our other experiments, infected saliva was first irradiated with UV to inactivate salivary gland–derived virus, and then a defined dose of SFV6 was added before inculcation into mice. By 24 hpi, those mice that received topical IMQ treatment exhibited significantly (P < 0.01) less virus in the serum and spleen. Thus, irrespective of changes in virus-induced salivary gland gene expression, topical IMQ was effective at inhibiting virus infection in mice. Together, these data identify the inoculation site as a key site for viral replication during the first 24 hpi and suggests that, in untreated mice, skin IFN responses are not sufficiently robust to prevent systemic dissemination. Therapeutic targeting of this site with topically applied innate immune agonists was therefore highly effective at suppressing both the local and subsequent systemic course of infection, significantly improving survival.

Topical IMQ induces up-regulation of cutaneous ISG via canonical type I IFN receptor signaling

We next wanted to determine the mode of action because a detailed mechanistic understanding of this protection could inform the development of more targeted strategies. Topical IMQ contains a mixture of the TLR7 agonist IMQ (5%) and isostearic acid (25%), both of which can be inflammatory in a type I IFN-independent manner (28, 29), suggesting that IFN may be dispensable for some of the beneficial effects described here. However, we found that 24 hours after topical application at a mosquito bite, IMQ caused widespread induction of ISG expression, including the prototypic ISGs cxcl10, ifit1, isg15, and rsad2 (Fig. 3A), some of which have been identified as key responders to alphavirus infection (17, 30). Significant Cxcl10 up-regulation could only be detected after 24 hours (P < 0.01; fig. S2A). However, despite a clear up-regulation of ISG, we could not detect any increase in type I (ifna4, ifnb1, ifne, ifnk, and ifnz) or type II (ifng1) IFN transcripts in whole skin biopsies, as measured by TaqMan assays or custom SYBR green qPCR assays (Fig. 3A and fig. S2, A and B), suggesting that topical IMQ may activate ISG expression independent of IFN receptor signaling in the skin (31). Conceivably, our inability to detect type I IFN transcripts in wild-type (WT) mice skin may be due to large numbers of IFN-negative cells that diluted the IFN signal or because IFN-producing cells migrated out of the skin, rendering IFN undetectable.

Fig. 3 Protection by topical imiquimod is dependent on type I IFN signaling.

(A) Mosquito-bitten skin was treated with a single topical application of IMQ. Fold change gene expression of ISG in skin was determined by qPCR at 24 hours after treatment compared to untreated controls (n = 4). (B) ifnar1−/− mice and syngeneic WT controls were treated with a single 2 mg application of topical IMQ. Copy numbers of host ISG and 18S RNA were determined by qPCR at 24 hours after treatment (n = 10). (C and D) Mosquito-bitten ifnar1−/− skin (n = 5) was infected with 250 PFU of SFV6 and treated with topical IMQ at 1 hpi (with one reapplication at 6 hpi) or left untreated. (C) SFV RNA and host 18S were quantified by qPCR at 24 hpi. (D) Serum virus was quantified by plaque assay. (E) Mosquito-bitten WT mouse skin was infected with 250 PFU of SFV6 and then treated with either topical IMQ at 1 hpi or control cream (25% w/v isostearic acid), with one reapplication of each at 6 hpi. SFV RNA and host 18S were determined by qPCR at 24 hpi (n = 7 to 8). (F) Mice were bitten by mosquitoes, and the site was treated with a single topical application of IMQ. Fold change gene expression of ISG in dLN was determined by qPCR at 24 hours after treatment compared to untreated controls (n = 4). (G) Mosquito-bitten WT mice were treated with topical IMQ, and ifnb1 was determined by qPCR at 24 hours (n = 5 to 12). Each graph represents a separate experiment. *P < 0.05, **P < 0.01, ****P < 0.0001 (Mann-Whitney and Kruskal-Wallis test).

Antiviral ISG can be induced by a variety of distinct receptors, including signaling through IFN-αβ-R, IFNGR, IFNLR1, and other poorly defined noncanonical pathways (16, 31). However, we found that activation of skin ISG was dependent on IFN-αβ-R signaling because mice deficient in this receptor could not up-regulate these prototypic ISGs in response to topical IMQ (Fig. 3B). Loss of IFN signaling was associated with a loss of protection from infection because topical IMQ had no impact on virus replication in IFN-αβ-R null mice (Fig. 3, C and D). In addition, to specifically show that the IMQ component was essential for protection against virus, we treated mouse skin with a cream mimic that lacked IMQ but contained 25% w/w isostearic acid (28), which did not protect mice from infection (Fig. 3E). Nonetheless, it is likely that these excipients are required for a maximal IFN response because injection of IMQ alone had little or no effect on ISG expression (fig. S2, C to F), nor did it have any effect on virus replication, as shown above (Fig. 1, E to G, and fig. S1B).

In comparison to skin responses, topical IMQ up-regulated both IFN and ISG in the dLN; IFN-γ was significantly elevated by 8 hours after application (P < 0.05; fig. S2G), whereas type I IFNs and ISGs were detectable by 24 hours (Fig. 3F and fig. S2G). In agreement with previous studies in humans (32), there was a limited systemic response to topical IMQ, although our qPCR assay was sufficiently sensitive to detect some type I IFN expression in distal lymphoid tissue sites but not, e.g., remote skin or joints (Fig. 3G).

Skin-resident cells detect topically applied IMQ and are sufficient for mediating protection to virus

We next wanted to identify the cellular basis by which virus is detected and an antiviral state is induced and modulated by topical IMQ at mosquito bites. Although a variety of cultured skin cells are known to express RNA-virus pattern recognition receptors (33), the cellular coordination of innate immune responses to arbovirus at the mosquito bite is not described. We concentrated on studying cells of the dermis because the majority of the virus transmitted by mosquitoes is deposited here and because epidermal cells do not express the IMQ receptor TLR7 in vivo (34). To identify whether bite-recruited leukocytes or skin-resident cells are sufficient for detecting IMQ and eliciting protection from virus, we developed a skin model in which freshly biopsied mouse skin was taken from either resting skin or mosquito-bitten skin [4 hours after bite, during leukocyte influx (9)] and infected with Gaussia luciferase–expressing SFV6 (SFV6-Gluc), which enables longitudinal detection of virus at high sensitivity, in vitro. Mosquito biting results in the rapid recruitment of leukocytes including neutrophils and monocytes (9) and, as shown here, small numbers of plasmacytoid dendritic cells (pDCs) (Fig. 4A). Infection of skin explants resulted in replication of virus, as measured by qPCR of viral RNA, functional expression of virus-encoded luciferase, and release of new infectious virus into the surrounding tissue culture medium (Fig. 4, B to E, and fig. S2H). Although cultured skin explants do not replicate all aspects of in vivo responses, they do enable the study of cell-cell interactions that occur between distinct skin-resident cell types in the absence of leukocytes that would otherwise be recruited during mosquito biting and virus infection in vivo.

Fig. 4 Host responses to IMQ by skin-resident cells are sufficient to mediate protection.

(A) Mosquito-bitten mouse skin was infected with 250 PFU of SFV6. At 24 hpi, skin from the inoculation site was digested to release cells and numbers of pDCs (CD45+ CD11c+ SiglecH+) quantified by FACS per 105 live cells. (B to E) Mosquito-bitten skin, at 4 hours after bite, or resting mouse skin, was biopsied, and resulting 6-mm skin explants were infected ex vivo with 1 × 105 PFU of SFV6-Gluc. At 1 hpi, the explants were removed from the media and briefly dried, and the epidermis was treated with a single application of topical IMQ. Treated explants were then placed resting in tissue culture media and at 24 hpi: (B) copy numbers of SFV RNA were determined by qPCR, (C) virus titers in the media were quantified by plaque assay (n = 6), and (D) virus-encoded Gluc was assayed (n = 6). (E) SFV RNA was also quantified by qPCR in skin explants at 48 hpi. (F) Mice were infected in vivo at mosquito bites with or without a single application of topical IMQ (at 1 hpi), or 6-mm explants were derived from resting skin (no associated bite-recruited leukocytes) were infected with 250 PFU of SFV6 ex vivo by needle and then treated with a single application of topical IMQ (at 1 hpi). Copy numbers of SFV RNA and host 18S were determined by qPCR at 24 hpi and compared to respective untreated infected controls (n = 6 to 7). (G) Mosquito-bitten skin of NSG mice was infected in vivo with 250 PFU of SFV6 and then treated with topical IMQ at 1 hpi (with one reapplication at 6 hpi). Mice were left for 24 hours, and SFV RNA in the skin and spleen at 24 hpi was quantified by qPCR, and infectious virus in the serum was quantified by plaque assay (n = 6). (H to J) Skin explants (6 mm) from WT and ccr2-null mice were infected with 1 × 105 PFU of SFV6 and treated with a single 2 mg application of topical IMQ ex vivo. At 24 hpi, copy number of (H) SFV RNA, (I) ifnb1, and (J) host ISG (cxcl10, isg15, and rsad2) was determined by qPCR (n = 6). *P < 0.05, **P < 0.01 (Mann-Whitney and Kruskal-Wallis test).

We applied topical IMQ to the epidermis of explant cultured skin at 1 hpi (at the same dosage as above for the in vivo studies). Topical IMQ application resulted in robust induction of ISGs in skin explants, albeit at lower amounts than that seen in vivo by 24 hours (fig. S2I), perhaps a consequence of tissue culture media removing some of the applied cream. Application of topical IMQ resulted in a significant (P < 0.01) decrease in viral RNA and virus-encoded luciferase and a complete block in the release of new virus into the supernatant, for explants derived from both resting and mosquito-bitten skin (Fig. 4, B to E). The magnitude of fold decrease in viral RNA was similar in ex vivo explants derived from resting skin, as compared to that in mosquito-bitten skin in vivo (Fig. 4F). This suggests that although ISG induction was less robust in explants (fig. S2I), it was nonetheless sufficient to reduce virus titer. Thus, topical IMQ reduced viral replication irrespective of leukocyte recruitment triggered by either infection or by mosquito bite, suggesting that tissue-resident cells were sufficient for mediating topical IMQ-induced protection from virus in the skin.

We next wanted to define which skin-resident cells were necessary for activating ISG expression in response to topical IMQ. Skin-resident leukocytes include populations of γδ-T cells and ROR-γT (retinoic acid receptor–related orphan nuclear receptor γT) innate lymphocytes that are necessary for some inflammatory responses to repeated topical IMQ application (35). However, we found that nonobese diabetic severe combined immunodeficient gamma (NSG) mice, which lack all functional lymphocytes and innate lymphoid cells, were similarly protected from infection at 24 hpi by topical IMQ (Fig. 4G). This suggests that early skin responses to topical IMQ are activated by either stromal cells or nonlymphoid leukocytes, of which myelomonocytic cells are the most populous in the dermis. Skin dermis–resident myelomonocytic cells are derived from either bone marrow (BM) precursors (requiring expression of chemokine receptor CCR2) or non-BM embryonic sites of hematopoiesis (not requiring CCR2 expression) (36). To determine whether BM-derived skin-resident myelomonocytic cells are required to mediate topical IMQ protection, we infected skin explants from ccr2-null and WT mice and treated them with IMQ. Topical IMQ was similarly efficacious irrespective of ccr2 status, suggesting that the vast majority of skin-resident dermal DCs and monocytic-derived cells were not required (Fig. 4H). IFN and ISG induction in response to topical IMQ application was similar in ccr2-null skin as compared to WT skin (Fig. 4, I and J), suggesting that IMQ-responding cells were skin-resident cells not derived from myeloid BM precursors.

Skin macrophages and DCs communicate with stromal cells to mediate topical IMQ-induced protection from virus

Together, the above data suggest that the skin-resident cell type responding to topical IMQ was either a population of non–BM-derived macrophages, mast cells, or stromal cells, or a combination of these. It is not possible to deplete all of these cell types; therefore, we additionally undertook a comprehensive approach to define responses of all these cell populations. This involved analyzing the gene expression profile of skin inoculation site cells isolated by fluorescence-activated cell sorting (FACS), to reveal the cell-specific basis for innate immune response to virus and separately to topical IMQ (Fig. 5).

Fig. 5 Topical IMQ targets dermal macrophages and DCs to activate tissue-wide ISG expression.

(A to C) Mosquito-bitten mouse skin was digested to release cells and then FACS-sorted at 4°C into “stroma,” “macrophage,” “DCs,” and “other leukocyte” compartments as a bulk sort. All steps were undertaken in the presence of transcriptional inhibitors. Copy numbers of gene transcripts were determined by qPCR for ifna4, ifnb1, isg15, cxcl10, rsad2, and 18S RNA (n = 4). Cells were sourced from mosquito-bitten skin that was either (A) infected with 250 PFU of SFV6 alone, (B) treated with topical IMQ alone (at 1 hour and reapplied at 6 hours), or (C) infected with 250 PFU of SFV6 and treated with topical IMQ at 1 hpi (with one reapplication at 6 hpi). *P < 0.05 (Mann-Whitney test).

Although skin-resident cells are sufficient to mediate IMQ responses, this does not necessarily exclude a role for mosquito bite–elicited responses at later stages. For this reason and because arboviruses are all naturally transmitted into skin bitten by arthropods, we studied cellular responses to virus and topical IMQ in mosquito-bitten skin. We adapted a previously defined FACS gating strategy (36), to isolate populations of macrophages, DCs, stromal cells, and all other CD45+ leukocytes as a bulk population (fig. S3, A and B). These isolated populations were then interrogated for their expression of ifna4, ifnb1, isg15, and rsad2 because they have been previously implicated as key for host responses to alphaviruses (30, 37, 38) and also the prototypic ISG cxcl10.

The cellular coordination of innate immune responses to virus at the tissue-wide level in mosquito-bitten skin is poorly defined. Therefore, we first analyzed responses of skin cells to virus infection alone in the absence of topical IMQ (Fig. 5A). At 24 hpi, sorted stromal cells and macrophages were the only cells that expressed high amounts of virus structural gene transcript E1, indicative of active replication, whereas DCs expressed only very low amounts of viral RNA (fig. S3C). This, along with our previous studies (9), suggests that stromal cells and macrophages are key targets for viral replication of SFV at mosquito bites in vivo. The only cell type expressing type I IFN in response to virus was DCs, which was limited to ifnb1 only because no ifna4 was detected. Thus, DCs were the primary initiators of antiviral ISG responses at virus-infected mosquito bites. However, dermal DCs were present at low numbers in the skin at 24 hpi, suggesting that their total contribution to the tissue-wide IFN response was minimal (fig. S3B). This agrees with our whole tissue analysis that demonstrated little IFN expression before 24 hpi described above. Activated DCs are highly migratory and rapidly leave the skin via the lymphatics, which may partly explain why skin-wide IFN expression was so low. In summary, although stromal cells and macrophages exhibited high expression of viral RNA, they did not make detectable type I IFN. Nonetheless, they expressed the ISG rsad2, cxcl10, and isg15 by 24 hpi (Fig. 5A), presumably in response to DC-expressed IFN cues. All other leukocytes included in the bulk sort did not express IFN or up-regulate ISG in response to virus.

Next, to determine which cells were activated by topical IMQ, mosquito-bitten mouse skin was treated with topical IMQ alone (Fig. 5B and fig. S3E). Crucially, in contrast to virus infection, topical IMQ induced ifnb1 expression in both the more populous skin-resident macrophages and dermal DCs, whereas stromal cells and all other leukocytes lacked type I IFN transcripts. In response, stromal cells substantially up-regulated rsad2 and isg15. The ability of this assay to detect type I IFN transcripts is in contrast to that found in whole skin biopsies where IFN expression was not detected (Fig. 3). This most likely reflects the enhanced sensitivity that is provided by analyzing sorted/purified populations of cells.

Because viruses have evolved mechanisms to antagonize IFN signaling, we next wanted to determine whether virus infection modulated this response to topical IMQ. Therefore, we infected mosquito bites with virus, either with or without topical IMQ application at 1 hpi and assayed type I IFN and ISG expression at 24 hpi in each cell type (Fig. 5C). Despite the lower titers of virus in topical IMQ-treated skin by 24 hpi (and therefore less activation of IFN signaling via virus sensing), we found equal or higher fold increases in the expression of both type I IFNs and ISG in the macrophage population. ISG up-regulation was also more robust in macrophages, DCs, and stromal cells, with the exception of rsad2 in DCs, suggesting that IMQ-induced up-regulation during virus infection was ISG and cell type specific. Activation of higher IFN expression in macrophages was important for two reasons; first, macrophages were significantly more numerous in the skin than dermal DCs at 24 hpi (P < 0.001; fig. S3B), and second, they constituted an important source of virus replication in the first 24 hours of infection (fig. S3C) (9). IMQ-elicited IFN appeared to be functional because type I IFN-negative stromal cells exhibited a significant decrease in viral RNA with topical IMQ treatment by 24 hpi, as did virus in infected macrophages (P < 0.05; fig. S3C).

Together, this shows that topical IMQ up-regulated type I IFN in skin-resident macrophages that, in addition to responses of the less frequent dermal DCs, acted to induce ISG expression and thereby reduce viral replication in the skin. Furthermore, because dermal DC-deficient ccr2-null skin (36) was not compromised in its ability to express ISG or reduce virus titers in response to topical IMQ, as described above, we suggest that skin-resident macrophages alone are sufficient to mediate protection. Thus, based on these associations, it is likely that responses by skin-resident macrophages to topical IMQ are sufficient to confer protection against virus infection.

Stromal cells integrate cues from IMQ-treated leukocytes to resist infection with virus

We next wanted to determine whether signals from IMQ-activated macrophages and DCs alone are sufficient to confer protection on stromal cells from virus infection and whether this required cell-to-cell contact. Most of the dermal stromal cells are either fibroblasts or keratinocytes, both of which can be experimentally infected with a range of arboviruses (10, 39, 40), including SFV4 (Fig. 6, A to D). Some cultured fibroblasts have been described to express a variety of receptors that sense virus, including the IMQ receptor TLR7 (41). However, we found that FACS-isolated stromal cells ex vivo lacked detectable TLR7 (fig. S3D). In addition, our cultured primary dermal fibroblasts (Fig. 6B) and keratinocytes (Fig. 6E) did not show any response to IMQ after SFV infection; IMQ did not protect cells from infection; ISG cxcl10 expression was not induced (even at high doses, Fig. 6C); and media from fibroblasts pretreated with IMQ for 24 hours did not confer protection to other stromal cells from subsequent virus infection (Fig. 6D). Thus, both our in vivo and in vitro analyses suggest that stromal cells were not able to respond to IMQ alone.

Fig. 6 Protection of skin stromal cells from virus by IMQ requires signals from myeloid cells.

(A) Primary mouse keratinocytes were infected with SFV4(Xho)-EGFP at a multiplicity of infection (MOI) of 0.1. At 6 hpi, replication complexes (green) were present throughout the cytoplasm (DAPI, blue). (B and C) Primary mouse embryonic fibroblasts were infected with SFV6 in vitro at an MOI of 0.1. At 1 hpi, cells were treated with IMQ with either 0.5, 2, or 10 μg/ml at 24 hpi; (B) copy number of SFV RNA was determined by qPCR (n = 6) and infectious virus in the media was quantified by plaque assay and (C) copy number of host cxcl10 gene transcripts was determined by qPCR (n = 6). (D) Primary mouse dermal fibroblasts and BM-derived Flt3L DCs, GM-CSF DCs, and M-CSF macrophages were treated with IMQ (0.5 μg/ml) for 24 hours. The resulting conditioned medium was aspirated and placed on SFV6-Gluc–infected fibroblasts (1 hpi with at an MOI of 0.01). Virus-encoded Gluc was assayed at 24 hpi (n = 6). (E to G) Primary keratinocytes were infected in vitro with SFV6 (E and G) or SFV4(Xho)-EGFP (F) at an MOI of 0.1 and then at 1 hpi treated with IMQ (0.5 μg/ml) in the presence or absence of Flt3-L DCs separated by a 0.5-μm Transwell membrane. (E) Copy numbers of SFV RNA and host 18S RNA in the keratinocytes were determined by qPCR at 24 hpi (n = 5 to 6). (F) SFV-encoded EGFP shown as green, with cell nuclei counterstained with DAPI (blue). (G) Copy numbers of SFV RNA and cxcl10 transcripts were determined in DC by qPCR at 24 hpi (n = 5 to 6). (H and I) Primary fibroblasts (H) or keratinocytes (I) were infected in vitro with SFV6 at an MOI of 0.1 in the presence or absence of Flt3L-derived DCs separated from stromal cells by a Transwell membrane. At 1 hpi, cells were either left untreated or given IMQ (0.5 μg/ml). (H) Copy numbers of SFV RNA were determined by qPCR at 24 hpi (n = 5 to 6). (I) Gene expression of SFV-infected keratinocyte ISG was determined by qPCR (n = 4), and significance was determined by two-way ANOVA (*P < 0.05, ****P < 0.0001). *P < 0.05, **P < 0.01, ****P < 0.0001 (Mann-Whitney and Kruskal-Wallis test).

To determine whether soluble factors from IMQ-stimulated DCs or macrophages could influence the susceptibility of primary cultures of skin fibroblasts to infection, primary cultures of leukocytes were stimulated with IMQ for 24 hours, and their supernatant was given to fibroblasts at 1 hpi with luciferase-expressing SFV6-Gluc (Fig. 6D). Although supernatant from resting leukocytes had no effect on fibroblast susceptibility to infection, tissue culture supernatant from DCs or macrophages treated with IMQ protected fibroblasts from infection, as measured by virus-encoded luciferase.

Because infected myeloid cells and DCs are susceptible to arbovirus-encoded anti-IFN mechanisms (42, 43), we next determined whether infected DCs were still able to provide IMQ-induced protection and trigger an antiviral state in stromal cells. To do this, we used a Transwell system in which DCs were separated from stromal cells by a cell-impermeable membrane. All cells were infected with SFV4(Xho)–enhanced green fluorescent protein (EGFP) and treated at 1 hpi with either 0.2 μg/ml IMQ or saline control (Fig. 6, E to H). DC cultures themselves became infected with SFV, exhibiting high-level expression of viral RNA (Fig. 6G). Nonetheless, DCs were able to resist infection upon addition (1 hpi) of IMQ and consequently expressed higher amounts of the ISG cxcl10. Whereas the addition of IMQ alone, or DC alone, had no effect on the ability of cocultured stromal cells to resist infection, the combined presence of both IMQ and DC did protect keratinocytes (Fig. 6, E and F, and fig. S4) and fibroblasts (Fig. 6H). This was evident by a significant decrease in viral RNA expression (P < 0.05; Fig. 6, E and H) and by microscopic inspection of the stromal cell monolayer that otherwise became decimated by 24 hpi, with the remaining intact cells also positive for virus-encoded EGFP (Fig. 6F and fig. S4). Protection from virus in stromal cells occurred in conjunction with the induction of stromal ISG expression (Fig. 6I). IMQ-treated keratinocytes only increased IFN and ISG expression with the addition of DCs to the Transwell insert. Thus, IMQ-induced stromal cell protection from virus occurs in conjunction with the induction of ISG expression that was licensed by extracellular cues derived from IMQ-responsive leukocytes.

Topical IMQ protects mice and human skin from infection with a variety of genetically distinct medically important viral pathogens

We lastly wanted to determine whether activating IFN pathways by topical IMQ at the skin inoculation site could have potential applicability to a variety of genetically distinct arboviruses of medical importance. We therefore analyzed the ability of topical IMQ to modulate the outcome of infection to a relevant representative of each of the genetically distinct groups of arboviruses in mice and human skin explants. The prototypic BUNV (genus Orthobunyavirus, family Peribunyaviridae, order Bunyavirales), which, similar to SFV, is also transmitted by Aedes mosquitoes, replicates efficiently at mosquito bites in the skin and causes viremia in mice (9). BUNV has well-described potent IFN antagonism (4) and therefore may be resistant to topical IMQ treatment. However, when mosquito-bitten BUNV-infected inoculation site was treated with topical IMQ at 1 hpi, there was a significant decrease in viral RNA in target tissues (P < 0.01 to P < 0.05) and blood (P < 0.001 to P < 0.05) from 24 to 72 hpi (Fig. 7, A to C). Similarly, we determined whether topical IMQ could prevent dissemination of arthritogenic CHIKV (a medically important pathogen of the Togaviridae family that has caused widespread outbreaks of disease) to mouse joints that were remote from the inoculation site (Fig. 7, D to G). By day 5 after infection, mice treated with topical IMQ 1 hpi exhibited significantly lower amount of CHIKV RNA (qPCR) and infectious virus in the ankle joints and wrists of the forelimbs (P < 0.0001 to P < 0.05; Fig. 7, D to G, and fig. S5), irrespective of whether mice were coinoculated with mosquito saliva.

Fig. 7 Topical IMQ protects both mouse and human skin from varied, genetically distinct arboviral threats.

(A to C) Mosquito-bitten skin of mice (n = 7 to 8) was infected with 25,000 PFU of BUNV and treated with topical IMQ at 1 hpi (with one reapplication at 6 hpi). Copy number of BUNV RNA was determined by qPCR in the skin (A) and dLN (B), and infectious virus in blood was determined by plaque assay up to 72 hpi (C). (D to G) Mice were coinoculated with 106 PFU of CHIKV (subcutaneous) and mosquito saliva (derived from five A. aegypti mosquitoes) and treated with topical IMQ at 1 hpi (with one reapplication at 6 hpi). Dissemination of virus to distal joints was assessed at 5 days after infection in the right ankle joint (D and E) and both wrist joints (F and G). CHIKV RNA was quantified by qPCR, and infectious virus titers were quantified by median tissue culture infectious dose (TCID50) (n = 10). (H and I) Bisected human skin biopsies from 16 different donors were cultured as explants and infected with virus [1 × 105 PFU of CHIKV (H) or 0.5 × 105 ZIKV (I)]. One-half of each biopsy was treated with a single application (2 mg) of topical IMQ ex vivo 1 hpi, whereas the other half was left untreated. CHIKV and ZIKV RNA were quantified by qPCR at 24 and 48 hpi, and virus titers in the media were quantified by plaque assay. *P < 0.05, **P < 0.01, ***P < 0.001 (Mann-Whitney test).

Last, to determine whether topical IMQ could also reduce viral replication in human skin, we cultured explants of freshly biopsied human skin, infected them with virus, and treated them with topical IMQ to the epidermis at 1 hpi (Fig. 7, H and I). Application of topical IMQ resulted in a significant decrease in viral replication for both CHIKV and ZIKV (genus Flavivirus of the family Flaviviridae), as measured by viral RNA and also infectious virus at 24 and 48 hpi (P < 0.001 to P < 0.05). Together, these studies demonstrate that an innate immune agonist that primarily targets inoculation site skin-resident macrophages has a substantial post-exposure prophylactic effect on the replication and dissemination of a variety of genetically distinct medically important viruses.


We demonstrate that innate immune responses to virus in the skin are a limiting factor for suppressing subsequent systemic infection. Critically, we show that therapeutic modulation of these pathways in the first hours of infection can suppress disease and improve clinical outcome in mice. This work therefore identifies the inoculation site as an important site of arboviral replication and that skin innate immune virus-sensing pathways must be sufficiently robust to act as determinants of the clinical course. In addition to the basic biological insights that these studies provide, we suggest that therapies that either target IFN induction or alleviate the virus-encoded mechanisms that inhibit IFN may therefore be efficacious. Targeting a common pathway at the inoculation site, as shown here to suppress infection by several genetically distinct arboviruses, may circumvent the need to develop multiple species-specific antivirals because their development and future use remain challenging due to the unpredictable nature of arbovirus outbreaks and difficulty associated with reaching an accurate and timely clinical diagnosis.

In this study, we identified the key skin cell types in mice that coordinate antiviral immune responses to virus and those that respond to topical IMQ. We found that skin type I IFN expression to virus was restricted to the relatively rare DC population. In contrast, macrophages and stromal cells could not elicit IFN expression to virus alone, despite a high level of viral RNA in these cells. Thus, skin-resident DCs are key activators of the type I IFN response at the inoculation site to virus. Such specificity may reflect either cell type–specific susceptibility to virus-encoded mechanisms to antagonize IFN activation or the particular cell tropism of our model arbovirus, SFV.

We suggest that the observed efficacy of our exemplar immunomodulator, topical IMQ (Aldara), was explained by activation of dermal macrophages. These cells were not only otherwise deficient in their IFN expression in response to virus alone but also replicate virus (9). We found that IFN signaling was key for IMQ-induced protection from virus, whereas skin-resident macrophages were the only cell type that increased IFN-β in response to IMQ and also that their addition to cultures of skin fibroblast was sufficient to confer IMQ-induced protection. We suggest that therapies that specifically target dermal macrophages may provide yet enhanced efficacy, for example, by increasing retention of the drug within the dermis or via delivery within liposomes. Furthermore, as the IMQ receptor TLR7 exhibits a relatively restricted expression in specific cells, those innate immune agonists that target more widely expressed virus-sensing receptors, or their pathways, may further increase efficacy.

Topical IMQ was no longer efficacious where treatment was delayed for 10 hours in our mouse model. We suggest that this timing most likely reflects specific aspects of our model system that supports rapid dissemination of SFV, itself a mouse-adapted strain. It is not clear how these kinetics apply to human infection by mosquito, although it is widely accepted that virus may take considerably longer (several days) to disseminate from human skin to the blood (44). This in turn may mean that the window available for post-exposure intervention in humans is conveniently longer than that observed in our model, although virus-, host-, and environment-specific factors are likely to define its length. However, because Aedes sp. mosquitoes are day-biting and their bites are typically visible within minutes, it is possible that there is a sufficient time window for post-exposure prophylactic treatment between infection/biting and awareness of a bite/application of cream in humans. Targeting skin responses later during infection, e.g., once clinical signs have become apparent, is unlikely to have any effect on the severity of infection because the virus would have already disseminated systemically. Furthermore, note that it may not be possible for individuals to notice all mosquito bites sufficiently quickly.

Although the data presented here are primarily derived from studies using mice, we suggest that therapeutic intervention at the inoculation site represents a feasible strategy for targeting mosquito-borne virus infection that merits further investigation. Putatively, treatment for either mosquito- or perhaps also tick-borne virus could be used topically in the first few hours after an arthropod bite is noticed. We suggest that with refinement, a modified form of our exemplar immunomodulator (topical IMQ, Aldara) may have potential as a treatment strategy for this globally important disease category. Thus, further work that refines the most efficacious method for stimulating IFN responses in macrophages opens up the possibility of broadly applicable, cost-effective, therapeutic interventions for arbovirus infection.

We maintained application of topical IMQ for only 5 hours to a single site, suggesting that systemic side effects are less likely than with prolonged exposure. However, effective post-exposure prophylaxis might entail treatment of multiple, discrete, suspected arbovirus-infected mosquito bites, even with, e.g., vector-repellent strategies used. Thus, it will be important to refine any future formulations such as to render the application of treatment practical. However, note that Aldara has been widely used for many years for a variety of dermatological conditions, including in primary care, and is generally well tolerated by patients, even when repeatedly applied at the same site for several weeks (45). In addition, our approach would likely entail topical immunomodulation over a smaller surface area than those typically treated with Aldara for its licensed indications. Nonetheless, this approach may involve repeated activation of IFN responses, which, even if done at discrete localized sites, may come with associated concomitant side effects. Last, note that outbreaks of arbovirus infection are explosive in nature and can be highly seasonal, enabling the possible use of any prophylactic strategies to be more intensively promoted when risk of infection is known to be high. Furthermore, it is possible that those patients deemed at higher risk of complications after arbovirus infection, such as immunosuppressed individuals, or, e.g., those that are more susceptible to severe dengue, may particularly benefit from this strategy.


Study design

This study was initiated to determine whether antiviral innate immune response to arbovirus at the mosquito bite can be therapeutically manipulated to suppress the subsequent systemic course and improve outcome of infection in mice. We investigated mammalian host responses to representative viruses from three genetically divergent groups of arboviruses. Two of the viruses (ZIKV and CHIKV) are medically important emerging arboviruses, whereas the others (SFV and BUNV) are model viruses that replicate efficiently in immunocompetent mice. As previously described (9), A. aegypti bite sites were infected with mosquito cell culture–derived virus in a submicroliter volume by hyperfine needle. Infected mosquitoes were not used to inoculate virus to mice because the inoculum supplied by biting mosquitoes was too variable and unpredictable to allow effective comparisons; infected mosquito saliva was used as a co-inoculum with virus in some studies here. All experiments involving mice had been subject to rigorous review by our local welfare and ethical review committee and additionally approved by the U.K. Home Office (license PA7CF4E75).

On the basis that immune responses in mice may not always resemble those in humans, we also investigated the ability of interventions to suppress infection with the human pathogens ZIKV and CHIKV in human skin explants. Human studies were performed after ethical approval and in accordance with all applicable regulations (research ethics committee number 10/H1306/88 and 16/HY/0086). Sample sizes and end points were selected on the basis of our published experience with arbovirus infection. Wherever possible, preliminary mouse experiments were performed to determine requirements for sample size, considering the available resources and ethical use of animals. Female animals (age-matched) were assigned randomly to experimental groups. For plaque assays, luciferase assays, and qPCR, samples were coded before analysis to limit bias. For qPCR, each result represents the median of three or four technical replicates of one biological replicate. For plaque assay, viral stocks and biological samples were serially diluted, and each dilution was assayed in duplicate. Biological replicates from mice were excluded from analysis if injection of virus inadvertently punctured a blood vessel (although this was rare and occurred with a frequency of <1%). No outliers were removed from these studies. Primary data are reported in data file S1.

Cell culture

Baby hamster kidney–21 (BHK-21) cells and C6/36 mosquito cells were grown as previously described (9). Mouse leukocytes were derived from BM precursors, using macrophage colony-stimulating factor (M-CSF; 10 ng/ml) for 6 days (macrophages), granulocyte colony-stimulating factor (GM-CSF; 20 ng/ml) for 6 days, or Flt3L (200 ng/ml) for 10 days (DCs). C57Bl/6-derived primary keratinocytes (Cell Biologics) were grown in complete epithelial cell media as per the manufacturer’s instructions. Skin fibroblasts were derived from adult C57Bl/6 dermis, digested with collagenase D (1 mg/ml), dispase II (0.5 mg/ml), and deoxyribonuclease (DNase; 0.1 mg/ml) in Hanks’ balanced salt solution (HBSS) to release cells, and adherent fibroblasts were grown for 12 days in complete Dulbecco’s modified Eagle’s medium (DMEM).


The pCMV-SFV4 and pCMV-SFV6 backbone for production of SFV has been previously described (25, 46). The full-length infectious cDNA clone of SFV4 was engineered by P. Liljeström (Karolinska Institutet) and was used to construct a stabilized infectious SFV plasmid, pCMV-SFV4 (46, 47). SFV6 is a copy of non–mouse-adapted SFV L10 (a polyclonal virulent isolate), derived by making six nonsynonymous nucleotide changes to SFV4. The derived virus, SFV6, generates a high-titer viremia in mice, is efficiently neuroinvasive, and highly virulent (25). The EGFP marker gene was inserted into the C-terminal region of SFV nsP3 via a naturally occurring XhoI site. mCherry and Gaussia luciferase (Gluc) were separately inserted as a cassette under the control of a duplicated subgenomic protomer 3′ with resulting viruses referred to here as SFV6-mCherry and SFV6-Gluc (48). Plasmids were electroporated into BHK-21 cells to generate infectious virus. SFV clones that have been genetically altered in this way are stable, and there is no evidence of virus attenuation or genetic instability (49). Hence, the use of these reporter viruses allowed the replication, dissemination, and localization of SFV to be studied in high detail. SFV4 is the prototypic, less virulent strain of the virus, whereas SFV6 is a copy of a virulent strain (25). Both strains cause disease in mice when inoculated in the skin at a mosquito bite (9).

WT BUNV was derived as previously described (50). A low-passage WT ZIKV isolate (51) was derived from a patient displaying classical disease symptoms in Recife (ZIKV PE243) and has been sequenced, supplied by A. Kohl [Medical Research Council (MRC)—University of Glasgow Centre for Virus Research]. In CHIKV mouse experiments, CHIKV Indian Ocean strain 899 (FJ959103.1) was used (C. Drosten, University of Bonn, Germany) (52). For human explant studies, CHIKV was made from the infectious clone derived from the LR2006_OPY1 isolate (DQ443544). These strains were chosen because they have both been passaged only a few times since isolation and likely represent WT circulating virus. We estimate that the CHIKV and ZIKV used here have not been passaged more than five times in total since isolation. In all cases, viruses were grown once in BHK-21 cells and then passaged once in C6/36 Aedes mosquito cells and titrated before use because mosquito cell–derived virus has distinct glycosylation and because insect cells impose distinct evolutionary constraints on viral progeny (53). SFV4 and SFV6 were used at passage 2.


In all cases, results have been generated by infecting mice with the same passage of virus. Unless otherwise specified, all mice were 6- to 8-week-old WT mice (C57bl/6J). C57bl/6J and NSG mice were derived from a locally bred colony maintained in a pathogen-free facility, in filter-topped cages, and maintained in accordance with local and governmental regulations. To prevent genetic drift, mice have been rederived using externally supplied mice (Charles River). Ccr2-deficient mice were originally obtained from the Jackson Laboratory (stock number 004999). Ifnar1-deficient and WT counterparts on a 129S7/SvEvBrdBkl-Hprtb-m2 background (B&K Universal) were maintained in Tecniplast 1284 L Blue line individually ventilated cages at Biological Services, University of Glasgow. All mice had a 12-hour light/12-hour dark cycle and provided ad libitum with sterile food and water.

A. aegypti mosquito biting and virus infection of mice

We used our previously established model of arbovirus infection at mosquito bites (9). This model was specifically developed to consistently model natural infection by arbovirus, including mimicking the same dose delivered by mosquitoes [between 100 and 100,000 plaque-forming units (PFU) for alphaviruses], using mosquito cell–derived virus, injecting small 1-μl inoculum volume, and by including the presence of a mosquito bite at the site of inoculation. To ensure that mosquitoes bit a defined area of skin (upper side of the left foot), anesthetized mice were placed for up to 10 min into a mosquito cage containing A. aegypti mosquitoes (locally bred colony derived from the Liverpool strain) as previously described (9). Biting was restricted to a defined area of the left foot by covering all other mouse skin with an impenetrable barrier. Bitten skin (three to five mosquito bites per foot) was injected with virus in a 1-μl volume into the skin, using either 250 PFU of SFV6, 1 × 104 PFU of SFV4, or 2.5 × 104 PFU of BUNV. These doses represent those that are known to initiate high-titer viremia by 24 hpi and, for SFV, clinical signs within days after infection. Total RNA input (copy number) from inoculum is roughly 10-fold higher than PFU (54). For survival curves, mice were monitored closely and culled when they reached more than three clinically defined end points of disease. Clinical signs included body weight loss >10%, subdued behavior when provoked, hunching, convulsions, limb paralysis, prostration >1 hour, and vocalization. For CHIKV infection of mice, 3- to 4-week-old mice were infected by subcutaneous injection of 106 PFU of CHIKV in the hind left foot pad with or without A. aegypti mosquito saliva.

Infection of mosquitoes and mosquito saliva extraction

Washed male rabbit blood (ENVIGO) was mixed with 2 mM adenosine triphosphate (Thermo Scientific) and SFV4 (7.8 × 107 PFU/ml). Blood was loaded into a Hemotek feeder (Hemotek) and placed on cages containing A. aegypti mosquitoes for 45 min to allow mosquitoes to feed. Fully engorged female mosquitoes were then transferred into boxes and maintained at 28°C with 80% humidity for 7 days. Mosquito saliva was extracted in oil from salivating mosquitoes, and corresponding individual mosquito heads were harvested and submerged in TRIzol (Life Technologies). Saliva from groups of five mosquitoes was pooled together and stored at −80°C. RNA from mosquito heads was extracted using PureLink RNA micro columns as per the manufacturer’s instructions (Life Technologies). Mosquito heads were homogenized in TRIzol using a Precellys 24 tissue homogenizer (Bertin Instruments) with glass beads (VWR), followed by purification using PureLink micro columns (Life Technologies). RNA was converted into cDNA using the High-Capacity RNA-to-cDNA kit (Life Technologies). Expression of SFV E1 gene was undertaken using custom-designed SYBR green–based qPCR assay using PerfeCTa (Quanta). Samples were analyzed using ΔΔCT against negative control mosquito heads (CT > 35). Saliva groups from SFV E1–positive mosquito heads were then pooled and UV-inactivated for 10 min to remove the virus and then mixed with a defined dose of 10,000 PFU of SFV4 for inoculation into mouse skin.

Administration of innate immune agonists

Anesthetized mice were injected with either 6 μg of poly(I:C), polydeoxyadenylic-polydeoxythymidylic [poly(dA:dT)], or IMQ (Invivogen) into the skin as a 4-μl aqueous volume using Hamilton syringes at the same site as virus infection [using upper skin of the left foot (9)]. Poly(I:C) and poly(dA:dT) were purchased precomplexed with transfection reagent LyoVec. Topical IMQ (2 mg; Aldara, 5% w/w IMQ; 25% w/w isostearic acid) was applied to the site of virus inoculation (upper side of the left foot). In most cases, because topical IMQ is a cream that can be removed by cage bedding, it was reapplied once at 6 hpi to maintain dosing. Although it is difficult to define the exact dose of IMQ provided using topical application of cream, previous studies have shown that dosing results in limited systemic absorption, with IMQ concentration peaking at 0.4 ng/ml blood in humans (32), with most being retained in the absorbent stratum corneum (the outer dead keratinized layer) or otherwise removed by external processes.

Skin explant studies

For human explants, informed consent was obtained from 16 volunteer donors, and skin was sourced from areas that were relatively protected from environmental insult and had no obvious lesions (upper inner arm). Human studies were performed after ethical approval and in accordance to all applicable regulations. For mouse studies, skin was derived from either resting skin or mosquito-bitten skin (biopsied at 4 hours after biting). For both human and mouse studies, skin was dissected to remove subdermal tissues to leave the dermis exposed and was immediately infected with virus by placing the lower dermis into virus-containing solution [0.75% bovine serum albumin (BSA) in phosphate-buffered saline (PBS) and Ca2+- and Mg2+-free PBS] for 1 hour. Explants were then washed in saline, and the epidermal surface of the skin explant was briefly dried. Similar to our mouse-based in vivo studies, 2 mg of topical IMQ (Aldara) was applied topically to the epidermis (at 1 hpi) until absorbed. Explants were then cultured in 24-well plates containing complete DMEM at 37°C/5% CO2 for up to 48 hours.

Gene expression analysis

Viral RNA and host gene transcripts were quantified by reverse transcription qPCR, and infectious virus was quantified by end point titration, as described previously (52). Tissue generated up to 100 μg of total RNA, of which 1 μg of RNA was used to create cDNA, of which 1% was used per qPCR assay (10 ng of RNA equivalent). qPCR primers for SFV and CHIKV amplified a section of E1 and primers for BUNV targeted segment M (9), whereas primers for ZIKV amplified a section of the env gene. For SFV, CHIKV, and ZIKV, qPCR assays measured the sum value of both genome and subgenomic RNA.

Flow cytometry

For FACS, skin tissue samples were enzymatically digested with collagenase D (1 mg/ml), dispase II (0.5 mg/ml), and DNase (0.1 mg/ml) in HBSS for 50 min at 37°C. Enzymes were quenched with serum, and cells were washed, treated with FcR block (Miltenyi Biotec), and stained with antibodies and a viability dye (9). Cells were analyzed on a CytoFLEX (Beckman Coulter Life Sciences). For cell sorting, skin was digested and stained on ice in the presence of a transcriptional inhibitor (actinomycin D) and sorted to 90 to 100% purity using an Influx cell sorter (BD). See fig. S3 for the gating strategy.

Immunohistochemistry and histology

Tissues were fixed in 4% methanol-free paraformaldehyde (Thermo Scientific) and then dehydrated in an increasing concentration of sucrose. Tissue was embedded in optimal cutting temperature compound (Agar Scientific) and sectioned. Tissue sections were stained with DAPI (4′,6-diamidino-2-phenylindole) mounting media and imaged on a Zeiss Axioskop.

Statistical analysis

Data were analyzed using GraphPad Prism Version 7 software. Copy numbers of viral RNA and infectious titers from virus-infected mice were not normally distributed (with data points often spread over orders of magnitude) and were accordingly analyzed using the nonparametric-based Mann-Whitney test or Kruskal-Wallis test with Dunn’s multiple comparison test where appropriate, unless otherwise stated in figure legends. All such column plots show the median value ± interquartile range. Where data were normally distributed (as determined by using either the Shapiro-Wilk test or by simple visual inspection), data were analyzed using analysis of variance (ANOVA) with Holm-Sidak’s multiple comparison test and plotted with mean values. Survival curves were analyzed using the log-rank (Mantel-Cox) test. All plots have statistical significance: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant.


Fig. S1. Preexposure of skin to topical IMQ increased host resistance to infection with SFV4.

Fig. S2. Type I IFN and ISG expression in skin and draining LN after IMQ application.

Fig. S3. Gene expression analysis of skin inoculation site-derived FACS isolated cells.

Fig. S4. IMQ-mediated protection against virus infection in keratinocytes is dependent on help from leukocytes.

Fig. S5. Topical IMQ prevents systemic dissemination of CHIKV to joint tissue remote from inoculation site.

Data file S1. Primary data.


Acknowledgments: We thank C. De Keyzer for assistance with the CHIKV mice studies, the University of Leeds Faculty of Medicine and Health Flow Cytometry and Imaging Facilities, S. Terry for assistance with mosquito culture, the University of Leeds St James’ Biomedical Services, G. Cardwell, and the University of Glasgow Biological services for assistance with mouse studies. We are grateful to E. O’Grade, B. Walker, and P. Laws for support in obtaining human skin biopsies. Funding: Funding was provided by BBSRC DTP PhD fellowship to S.R.B. and C.S.M.; Wellcome Trust Seed Award in Science (108227/B/15/Z), The Royal Society research grant (RGS\R1\191390), and the University of Leeds UAF to C.S.M.; Wellcome Trust investigator award, MRC Programme, and a Wolfson Royal Society Merit Award to G.J.G.; MRC core funding MC_UU_12014/8 to E.P.; Fund for Scientific Research (FWO) Flanders individual credit (1522918N) to L.D. and FWO PhD fellowship (1S21918N) to S.J.; and an MRC grant MR/NO1054X/1 to A.T. Author contributions: Conceptualization: S.R.B., M.P., K.S., and C.S.M. Funding acquisition: S.R.B., G.J.G., L.D., S.J., J.N., A.T., and C.S.M. Investigation: S.R.B., M.P., D.A.L., J. Miltenburg, S.J., R.A., E.P., J. Major, M.M., H.K., M.V., K.S., and C.S.M. Methodology: S.R.B., E.P., K.S., and C.S.M. Project administration: L.D., A.T., E.P., and C.S.M. Resources: E.P., J.N., A.T., A.M., J.E., G.J.G., and K.S. Supervision: L.D., E.P., A.T., and C.S.M. Writing (original draft): C.S.M. Writing (review and editing): S.R.B., M.P., L.D., K.S., and C.S.M. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data associated with this study are available in the paper or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article