We first investigated the basal and interferon-induced expression of IFITM proteins in K562 cells, using an antibody that recognizes IFITM1 alone, or one that recognizes both IFITM2 and IFITM3 
. Western blot analysis showed that resting K562 cells express a basal level of IFITM1 and IFITM2 and/or 3, and that treatment with human IFN-α for 48 h strongly upregulated expression of both IFITM1 and IFITM2/3 ().
IFITM proteins restrict both direct and ADE-mediated DENV infection of K562 cells.
To study the restriction activity of each IFITM protein, we transduced K562 cells to stably express myc-tagged IFITM1, 2 or 3, or with vector alone, and selected transduced cells with puromycin. Western blot analysis using an antibody against c-myc showed robust expression of each IFITM protein (). To verify their activities, we infected transduced cells with viral pseudoparticles that contain an MLV genome encoding the enhanced green fluorescence protein (EGFP), and pseudotyped with entry proteins of IAV (A/Thailand/2(SP-33)/2004 (H5N1)) or MACV 
. Pseudovirus infection was measured two days later by flow cytometry. As expected, overexpression of IFITM1, 2 or 3 potently restricted infection by MLV-EGFP virus pseudotyped with IAV but not MACV entry proteins ().
We next infected these cells with infectious DENV2 New Guinea C strain (NGC) propagated in mosquito C6/36 cells. DENV2 was incubated in media alone, or with enhancing titers of monoclonal antibodies against dengue prM or envelope (E). We used two commercially available antibodies, 2H2 and 4G2, which bind dengue virus structural proteins prM and E respectively. Anti-prM antibodies opsonize virions that are not fully mature while antibodies against E can opsonize most virions (reviewed in 
). As previously described 
, dengue virions were incubated with each antibody or with media alone, and subsequently incubated with K562 cells. Previous studies have shown that intracellular staining of DENV antigen less than 43 h post-infection of K562 cells indicates a single round of infection 
. Accordingly, to evaluate viral entry, we assessed productive infection by intracellular prM-staining 24 h post-infection 
. As expected, with direct (antibody-independent) infection at a multiplicity of infection (MOI) 1, expression of IFITM1, 2, or 3 potently inhibited infection by ~80% when compared to vector-transduced cells. With ADE-mediated infection at MOI 1, baseline infectivity increased roughly three-fold indicating successful ADE, but expression of IFITM1, 2 or 3 continued to potently inhibit infection by ~70–90% (n
3 or 4 per condition; P<0.05). Similarly, at MOI 10, we observed less efficient restriction, but comparable restriction levels in direct and ADE-mediated infection (). Consistent with our previous studies, IFITM2 restricted DENV infection less efficiently than either IFITM1 or IFITM3 
. We further quantified viral loads in the supernatant by standard plaque assays with BHK cells 
. Again, overexpression of IFITM1 or IFITM3 effectively restricted ADE-mediated infection, and suppressed viral production by more than 10-fold (; n
3; P<0.05). Finally, we verified the generality of our observations by confirming them in J774A.1 murine macrophages 
stably expressing murine orthologs of IFITM1, 2 or 3 (; n
4; asterisks indicate P<0.05). We conclude that IFITM1, 2 or 3 can restrict ADE-mediated DENV infection of K562 cells.
We subsequently investigated whether endogenous IFITM proteins contributed to DENV restriction. We stably transduced K562 cells to express previously characterized shRNAs against IFITM1, 2, or 3 
, or with a control non-targeting shRNA (scrambled), and selected transduced cells with puromycin. IFITM1 expression was nearly undetectable in cells expressing shRNA targeting IFITM1, whereas IFITM2 and 3 levels were unaffected. shRNA targeting IFITM2 reduced IFITM2/3 expression slightly, while shRNA targeting IFITM3 substantially reduced IFITM2/3 expression (). As we have previously observed 
, IFITM1 depletion increased infection by IAV pseudoviruses by almost 2-fold, whereas little or no increase in infection was observed with shRNA targeting IFITM2 or IFITM3 (). Similarly, with DENV2 infection at MOI 1, knockdown of IFITM1 roughly doubled infection rates with both direct and ADE-mediated infections (; n
3 or 4; P<0.05). This increase was less pronounced at a higher MOI of 10, but remained statistically significant. These data show that endogenous IFITM1 in K562 cells restricts ADE-mediated DENV infection.
Endogenous IFITM1 restricts both direct and ADE-mediated DENV infection of K562 cells.
and show that ADE cannot bypass IFITM-mediated restriction, but leave open the possibility that ADE-mediated infection is quantitatively less sensitive to restriction than direct infection. To determine if IFITM proteins restrict ADE-mediated infection to the same extent as direct infection, we identified in pilot studies virus titers for direct and ADE-mediated infection that resulted in comparable infectivity, and used these titers to investigate the effect of IFITM overexpression or IFITM1 depletion as in and . When IFITM1, 2 or 3 were over-expressed in K562 cells under these conditions, nearly identical restriction was observed in direct and ADE-mediated infection (). Similarly when IFITM1 was depleted in the same cells, a comparable increase in infection was observed (). We conclude that IFITM proteins restrict ADE-mediated infection and direct infection with similar efficiencies.
IFITM proteins restrict direct and ADE-mediated infection with similar efficiencies.
We have previously shown that IFITM-mediated restriction depends on the site or mechanism of viral entry 
. This finding raised a key question: can ADE facilitate bypass of this restriction, or, instead, can IFITM proteins contribute to control of a secondary infection, including those resulting in DHF/DSS? Our data show clearly that in a human myelogenous leukemia cell line widely used in the study of ADE 
, IFITM proteins restrict ADE-mediated infection as efficiently as direct infection. Therefore, we find no evidence supporting the hypothesis that ADE-mediated DENV infection bypasses or is less sensitive to IFITM-mediated restriction. It is formally possible that differences in cell types, antibody properties, or viral strains could alter the sensitivity of ADE-mediated infection to IFITM restriction. However we show that our observations are consistent across two different cell lines from two species expressing their respective IFITM orthologs. In addition, it has been suggested that patients with DHF exhibit lower IFN-α levels than patients with DF (reviewed in 
), which may lead to reduced IFITM-induction. In vivo
studies may be necessary to clarify these possibilities. Nonetheless, our observation suggests that therapeutic strategies that specifically induce IFITM activity could control both primary and secondary dengue virus infections.
To date, the mechanisms of IFITM-mediated restriction remain elusive. We have previously demonstrated that IFITM proteins do not interfere with virion access to acidic cellular compartments 
. Furthermore, inducing viral fusion at the plasma membrane bypasses IFITM-mediated restriction 
, supporting a mechanism which operates within the endocytic pathway. Our work in this study shows that IFITM proteins restrict ADE-mediated DENV infection as efficiently as direct infection, but does not distinguish between the possible mechanisms of ADE, including alterations in viral attachment, endocytosis and/or fusion. However, the similar sensitivity of direct and ADE-mediated infection to IFITM restriction indicates that these infection modes share common features that render them both sensitive to IFITM proteins.