Research Article

Intracellular sensing of complement C3 activates cell autonomous immunity

See allHide authors and affiliations

Science  05 Sep 2014:
Vol. 345, Issue 6201, 1256070
DOI: 10.1126/science.1256070

Structured Abstract


Intracellular pathogens, which include viruses and some bacteria, typically disseminate through extracellular fluids before entering their target cells and beginning replication. While in the extracellular environment, pathogens can be intercepted by humoral immunity or by professional immune cells. However, immune surveillance is not always sufficient to prevent infection, and all cells need innate mechanisms to detect and disable pathogens.

Embedded Image

Intracellular complement C3 activates innate immunity. Complement component C3 covalently attaches to pathogens in the extracellular space. Upon pathogen entry into the cytosol, the cell senses attached C3. Sensing of C3 triggers a dual sensor and effector response, involving mitochondrial antiviral signaling (MAVS)–dependent immune signaling and proteasome-mediated viral degradation.


We hypothesized that one method of pathogen detection may be to take advantage of the pathogen’s transition between extracellular and intracellular environments. Complement is a system of immune serum proteins with the ability to attach covalently to pathogens. We investigated whether this irreversible tagging of pathogens results in complement component C3 being carried into the cytosol during infection. Given that C3 should not otherwise be present inside the cell, we tested whether this could act as an invasion signal.


Antibodies and complement components C3 and C4, but not other serum proteins, were found to associate with adenovirus. During adenoviral infection, deposited C3 was carried into cells, resulting in potent nuclear factor–κB (NF-κB) activation. Activation of NF-κB by C3 required viral entry into the cytosol, and no activity was observed when C3-coated adenovirus was trapped in endosomes. In addition to NF-κB, C3 also activated the activating protein 1 (AP-1) and interferon regulatory factor 3 (IRF3)/IRF5/IRF7 transcription pathways. Induction of these signaling pathways resulted in robust cytokine secretion, including interferon-β. Cytosolic C3 sensing was dependent on a number of signaling hubs known to be involved in innate immunity. In addition to activating immune signaling, C3 targeted cytosolic adenovirus for rapid degradation via the AAA–adenosine triphosphatase (ATPase) valosin-containing protein (VCP) and the proteasome. This degradation pathway potently restricted viral infection.

C3-dependent intracellular sensing was widely conserved in mammals. Moreover, C3 activated NF-κB upon infection of diverse cell lines and primary cells including human lung cells, a physiologically relevant adenovirus target. Intracellular C3 also activated NF-κB in response to infection by diverse nonenveloped viruses—including papillomavirus, astrovirus, calicivirus, rhinovirus, poliovirus, coxsackievirus, enterovirus, and the facultative cytosolic bacteria Salmonella—but not enveloped respiratory syncytial virus. Picornaviruses were much less susceptible to complement sensing as a result of antagonism by their 3C protease, which cleaved C3 and prevented NF-κB activation and proteasome-mediated restriction. However, treatment with the 3C antagonist rupintrovir prevented 3C cleavage and restored full complement sensing of rhinovirus and poliovirus.


Complement mediates a potent intracellular immune response to nonenveloped viruses and cytosolic bacteria. The deposition and covalent attachment of C3 onto pathogens results in its translocation into cells during infection, in which it simultaneously induces an antiviral state and directs the degradation of viral particles. Intracellular complement immunity is highly effective against a range of pathogens, occurs in a variety of cell types, is independent of professional immune cells, and is highly conserved in mammals.

Bringing in the agent of your own destruction

Cells need mechanisms to detect and disable pathogens that infect them. Tam et al. now show that complement C3, a protein that binds to pathogens in the blood, can enter target cells together with the pathogen. Once inside the cell, the presence of C3 triggers both immune signaling and degradation of the internalized pathogen. The discovery of this pathway reveals that cells possess an early warning system of invasion that works against a diverse array of pathogens and does not require recognition of any specific pathogen molecules.

Science, this issue 10.1126/science.1256070


Pathogens traverse multiple barriers during infection, including cell membranes. We found that during this transition, pathogens carried covalently attached complement C3 into the cell, triggering immediate signaling and effector responses. Sensing of C3 in the cytosol activated mitochondrial antiviral signaling (MAVS)–dependent signaling cascades and induced proinflammatory cytokine secretion. C3 also flagged viruses for rapid proteasomal degradation, preventing their replication. This system could detect both viral and bacterial pathogens but was antagonized by enteroviruses, such as rhinovirus and poliovirus, which cleave C3 using their 3C protease. The antiviral rupintrivir inhibited 3C protease and prevented C3 cleavage, rendering enteroviruses susceptible to intracellular complement sensing. Thus, complement C3 allows cells to detect and disable pathogens that have invaded the cytosol.

Host colonization by intracellular pathogens typically involves the penetration of mucosal layers and dissemination through extracellular fluids. Humoral immunity has two major components—the heat-labile complement system and the heat-stable immunoglobulin system—and provides robust protection against invading pathogens. Such protection has been extensively studied, but insights into its function have recently emerged. Antibodies have been shown to be carried into cells by nonenveloped viruses during infection, where they act as danger-associated molecular patterns (DAMPs) to activate innate immunity (1, 2) and inhibit viral replication both in vitro (25) and in vivo (6). We hypothesized that this phenomenon may not be specific to antibodies and that mislocalization of serum proteins as a result of pathogen movement from extracellular to intracellular compartments might be a strategy widely exploited by host immunity.

The complement system is composed of more than 30 proteins (7), with three activation methods—classical (antibody-directed), lectin (mannan-binding lectin or ficolin-directed), and alternative (spontaneous) pathways—leading to covalent deposition of C3 on the pathogen surface. Effector functions of complement stem from this deposition: C3 prevents receptor engagement, acts as an opsonin, and activates the terminal complement components to form a membrane attack complex, whereas cleaved components are anaphylatoxins. Most antiviral complement studies have used enveloped viruses (8). However, nonenveloped viruses and adenoviral gene therapy vectors are also susceptible to complement deposition (9), whereas some possess specific complement-evasion strategies (10, 11). Owing to the lack of a lipid bilayer, the membrane attack complex cannot form on nonenveloped viruses, so aside from blocking receptor engagement, it is not clear how complement inhibits nonenveloped virus infection.


Complement component C3 elicits NF-κB activation

Antibody in the cytosol has been shown to activate innate immune signaling cascades, establishing an antiviral state via receptor, TRIM21 (1, 2). We tested whether other serum components are able to elicit similar responses by attaching to incoming pathogens. Infection of human embryonic kidney (HEK) 293T cells carrying a nuclear factor κB (NF-κB)–driven luciferase reporter, with an adenovirus type 5 vector (AdV), did not activate NF-κB (Fig. 1A). In contrast, incubation of AdV with normal human serum (Serum) led to robust NF-κB activation. Treatment of this serum by means of heat-inactivation at 56°C [heat-inactivated serum (HI Serum)] or addition of EGTA (Serum+EGTA) to chelate calcium each reduced NF-κB induction. Plasmin is a host serine protease that is pathogenically activated by Staphylococci to cleave protein components from their capsule for immune evasion (12). Serum was treated with plasmin for 30 min before quenching with α-2-antiplasmin (Serum+Plasmin) and then incubated with AdV (Fig. 1A). Plasmin treatment abolished NF-κB induction but did not alter AdV infection. We then determined the antibody component of signaling, with the use of antibody-depleted serum (Serum-Ig) and a heat-inactivated form (HI Serum-Ig). Incubation of AdV with either Serum-Ig or HI Serum-Ig confirmed that antibody is responsible for part of this NF-κB activation (1). However, a heat-labile component functioned independently of antibody (Fig. 1B). This signaling component within serum consisted of a heat-labile, calcium-dependent protein (Fig. 1, A and B) that appears to function alongside and independently of antibody, which is consistent with known properties of the complement system (7).

Fig. 1 Complement C3-bound virus induces NF-κB signaling in nonimmune cells.

(A) NF-κB activity of HEK293T cells challenged with PBS, AdV, serum, HI Serum, EGTA-treated serum (Serum+EGTA), Plasmin-treated serum (Serum+Plasmin), or AdV incubated with the previous sera. (B) HEK293T cells treated with AdV incubated with Serum, HI Serum, antibody-depleted Serum (Serum-Ig), or HI Antibody–depleted Serum (HI Serum-Ig). (C) Concentration of serum components IgM, IgA, IgG, IgD, IgE, C3, C4, Mannan-Binding Lectin (MBL), C-Reactive Protein (CRP), and Pentaxin 3 (PTX3) bound to AdV after incubation of AdV+Serum as measured with ELISA. (D to F) NF-κB activity of HEK293T cells treated with AdV incubated with (D) serum deficient in complement components (Serum-C1 to -C6), (E) serum deficient in complement C4 (AdV+Serum-C4) or heat-inactivated Serum-C4 (AdV+HI Serum-C4), or (F) CVF-treated serum (AdV+Serum+CVF). (G) NF-κB activity of HEK293T cells after challenge with AdV+Serum, AdV+HI Serum, or Serum-C3 reconstituted with purified C3 (Serum-C3 + C3). (H) HEK293T treated with AdV incubated with serum deficient in factor B (AdV+Serum-fB) or factor D (AdV+Serum-fD). (I) NF-κB activity of HEK293T cells treated with AdV incubated with C3-Factor B-Factor D (AdV+C3fBfD) with and without subsequent virus pelleting. Data are representative of three experiments; results in (A), (B), and (D) to (I) are as fold change over PBS-treated controls; mean ± SEM; data in (C), mean ± SEM.

To categorize which serum components can attach to nonenveloped virus particles and thereby activate signaling, serum was incubated with AdV and spun through 30% sucrose to remove unbound proteins, with attached proteins detected by means of enzyme-linked immunosorbent assay (ELISA) (Fig. 1C). Antibodies of immunoglobulin M (IgM), IgA, and IgG isotypes bound to AdV, but IgD and IgE did not. Complement component C3 was strongly detected, showing that it deposited on AdV. C4 was also detected, indicating that classical activation had occurred. Mannan-binding lectin (MBL) and C-reactive protein (CRP) were not detected, suggesting that the lectin pathway was not required for complement deposition on AdV. Pentaxin 3 (PTX3) has antiviral activity against influenza (13) but was not detected bound to AdV (Fig. 1C). These data are supportive of complement activating NF-κB upon AdV infection. Infection experiments were carried out in the presence of serum lacking specific complement components (Fig. 1D). Although there was reduced NF-κB activation in C1- and C2-deficient serum (Serum-C1 and Serum-C2, respectively), only C3-deficient serum (Serum-C3) diminished induction comparably with heat inactivation. Serum deficient in components after C3—Serum-C5 and Serum-C6—did not have any impact compared with AdV+Serum. C4-deficient serum (Serum-C4) also gave similar responses to that observed from Serum-C1 and Serum-C2 (Fig. 1E). To further confirm the importance of C3, Serum was treated with Naja naja kaouthia cobra venom, which is known to cleave and inactivate C3 (14), and this resulted in a similar reduction in NF-κB induction by serum as heat-inactivation (Fig. 1F). NF-κB activation in C3-deficient serum was restored by the addition of recombinant C3 protein (Fig. 1G). Serum deficient in factor B or factor D, which are components of the alternate pathway (7), gave similarly diminished NF-κB activation to serum lacking C1 or C2, suggesting that both the alternate and classical pathways are important for C3 deposition (Fig. 1, D and H). We reconstituted the alternate deposition pathway using purified proteins (AdV+C3fBfD). Activated purified C3 potently induced NF-κB in the presence but not the absence of AdV (Fig. 1H). To test whether signaling was mediated by C3 attached to virus or by cleaved anaphylatoxins, AdV+C3fBfD was pelleted through 30% sucrose. Pelleted AdV+C3fBfD stimulated NF-κB comparably with unpelleted material, showing that it was C3-bound AdV that initiated signaling (Fig. 1I). Thus, C3 attached to the pathogen surface activates NF-κB upon infection.

Nonimmune cells have an intracellular C3 receptor

Most cells express inhibitory complement receptors CD46 (also known as complement regulatory protein) and CD55 (decay accelerating factor), whereas professional immune cells express a number of activating complement receptors (15). Consistent with this, HEK293T, HeLa, and Caco-2 cells and primary normal human lung fibroblasts (NHLFs) all expressed CD46 and CD55 (Fig. 2A), whereas only THP-1 monocytes expressed the activating receptors, complement receptor 1 (CR1, also known as CD35), CR3 (made up of CD11b and CD18), and CR4 (CD11c and CD18). Depletion of CD46 or CD55 (fig. S1A) had no impact on signaling detected in HEK293T cells (Fig. 2B), suggesting that C3-mediated NF-κB activation is not due to detection by these cell-surface receptors.

Fig. 2 Signaling in response to C3-bound virus is mediated by an intracellular receptor.

(A) Immunoblot for complement receptors and GAPDH (loading control) in HEK293T, HeLa, Caco-2, NHLFs, and THP-1 monocyte cells. (B) HEK293T cells treated with control, CD46, or CD55 siRNA and challenged with AdV+Serum, AdV+HI Serum, and AdV+Serum+CVF. (C to E) NF-κB activity after treatment with DMSO, Bafilomycin A1 (BafA1), or CID1067700 (CID); challenged with AdV+C3fBfD or TNF in (C) HEK293T, (D) NHLF cells, or (E) THP-1 cells. (F) Endosomal disruption in HEK293T measured by delivery of a nano-luciferase–expressing plasmid present in cell supernatant by infecting AdV, after treatment with DMSO, BafA1, or Rab7, as relative luminescence units (RLUs). (G) Confocal microscopy of HeLa cells 30 min after treatment with AdV with AlexaFluor-488–labeled C3, stained with DAPI and antibody to AdV. Scale bars, 20 μm. (H) HEK293T cells treated with Beads incubated with C3fBfD (Beads+C3fBfD), with or without transfection reagent. Data are representative of three experiments; results in (B) to (E) and (H) are as fold change over PBS-treated controls; mean ± SEM; data in (F), mean ± SEM.

Antibodies activate immune signaling in nonprofessional cells when carried into the cytosol by an infecting virus (1). We investigated whether C3 similarly activates signaling when carried into the cell during virus infection. Because AdV infection is dependent on receptor-mediated endocytosis (16), we tested whether endocytosis inhibitors prevent C3 activation of NF-κB. Inhibition of endosomal acidification by bafilomycin A1 (BafA1), or the endocytic pathway by Ras superfamily inhibitor CID1067700 (CID) (17), abolished AdV+C3fBfD–mediated NF-κB signaling in HEK293T, whereas induction by tumor necrosis factor (TNF) remained unaffected (Fig. 2C). The same effect was seen in NHLFs (Fig. 2D). However, signaling in THP-1 cells, which have activatory cell-surface receptors for C3, was not affected by endocytosis inhibitors (Fig. 2E). To confirm that C3 induced signaling in nonprofessional cells was a result of intracellular sensing, we used an endosomal disruption assay, in which we measured cytosolic delivery of a constitutively expressing nano-luciferase plasmid. Incubation of cells with C3fBfD had little impact on AdV escape from endosomes, whereas BafA1 and CID abolished it (Fig. 2F), verifying that AdV did not reach the cytosol in the presence of these inhibitors. To test whether C3 remained attached to AdV during entry of the virus, cells were examined by means of confocal microscopy. AdV particles could be detected within the cell and colocalized with AlexaFluor-488–labeled C3 that was deposited before infection (Fig. 2G). Moreover, no C3 was detected in the absence of virus, suggesting that C3 was deposited onto virus and transported into the cell. Last, to show that signaling by intracellular C3 is independent of corecognition of viral pathogen–associated molecular patterns (PAMPs), beads were incubated with the C3fBfD mix and transfected into cells (Fig. 2H). Only Beads+C3fBfD that had been transfected into cells were capable of signaling. Thus, nonimmune cells possess a signaling pathway that allows them to sense intracellular complement C3.

C3 signaling leads to pro-inflammatory cytokine production by activating NF-κB, IRF, and AP-1 transcription factors

Next, we investigated which signaling pathways are activated by intracellular C3 sensing. Canonical NF-κB signal transduction involves the transforming growth factor β (TGFβ)–activated kinase (TAK)–binding protein (TAB)–TAK complex, which activates the inhibitor of NF-κB (IκB) kinase (IKK) complex to phosphorylate IκB, promoting IκB degradation and releasing NF-κB to translocate into the nucleus. Inhibitors of the TAB-TAK complex (5Z-7-oxozaeanol), the IKK complex (IKK VII), and NF-κB release (panepoxydone) each inhibited signaling (Fig. 3A). Furthermore, increased phosphorylation of these components was detected upon infection with virus carrying deposited C3 (Fig. 3B). NF-κB components p65, p50, and p52 were activated by AdV+C3fBfD (Fig. 3C) as well as interferon regulatory factor 3 (IRF3), IRF5, and IRF7 (Fig. 3D) and the activating protein 1 (AP-1) family, c-Jun, JunB, JunD, and FosB (Fig. 3E). c-Fos was activated by virus regardless of C3, suggesting that other mechanisms are involved in its activation. Thus, C3 is sensed by a pattern recognition receptor (PRR) that activates classical immune transduction pathways. Immune activation by complement-coated virus, but not virus alone, induced secretion of pro-inflammatory cytokines, as demonstrated by the detection of interleukin-6 (IL-6), TNF, CCL4, IL-1β, and interferon-β (IFN-β) by means of ELISA after challenge with AdV+Serum and AdV+HI Serum (Fig. 3F), as well as the reconstituted alternate pathway of AdV+C3fBfD (Fig. 3G).

Fig. 3 C3-mediated signaling initiates proinflammatory cytokine production.

(A) NF-κB luciferase activity in NHLF cells treated with DMSO, 5Z-7-oxozaeanol, IKK VII, or panepoxydone. (B) Levels of total and phosphorylated IKKα, IκB, or p65 in HEK293T cells treated with AdV+C3fBfD 4 hours after infection as measured with ELISA. (C to E) Levels of (C) NF-κB components (D) IRF family proteins, and (E) AP-1 components measured with DNA-binding ELISA from NHLFs 4 hours after challenge with AdV+C3fBfD. (F and G) Levels of cytokines 24 hours after challenge by AdV incubated with (F) Serum, HI Serum, or Serum+CVF or (G) AdV+C3fBfD. Data are representative of three experiments; results in (A) to (E) as fold change over PBS-treated controls, mean ± SEM; data in (F) and (G), mean ± SEM.

C3 enables proteasome-dependent restriction of virus infection

The above data suggest that C3 can act as a DAMP to activate innate immunity. Next, we investigated whether C3 also mediates a direct effector response to inhibit viral infection. We used an adenovirus vector that expresses green fluorescent protein (GFP) after productive infection. AdV incubated with Serum, HI Serum, Serum+EGTA, or Serum+Plasmin and AdV alone was added to HeLa cells, with infected cells enumerated by means of flow cytometry (Fig. 4A). As with signaling, restriction by a heat-labile, calcium-dependent protein component capable of functioning independently of antibody was observed (Fig. 4B). This restriction was C3-dependent, as shown by the loss of heat-labile neutralization after CVF treatment (Fig. 4C). Moreover, restriction could be reconstituted in the absence of serum by using C3fBfD (Fig. 4D). Depletion of CD46 or CD55 had no effect on restriction, showing that capture via these membrane complement receptors was not responsible for this phenotype (Fig. 4E and fig. S1A). Binding to complement receptor CD46 has been suggested to stimulate autophagy (18); however, the autophagy inhibitor 3-methyladenine (3-MA), and phosphatidylionsitol-3-kinase inhibitors KU55933 and Gö6976 (19) had no effect on restriction (Fig. 4F) but induced p62 retention (fig. S1B). TRIM21 is required for intracellular neutralization of antibody-coated pathogens (4, 5). Depletion of TRIM21 removed the heat-stable component of restriction but did not affect the heat-labile component (Fig. 4G and fig. S1C), confirming that TRIM21 has a role in antibody-mediated neutralization but not complement-mediated restriction.

Fig. 4 C3 promotes intracellular restriction of virus.

(A to D) Levels of infection in HeLa cells after challenge with GFP-encoding adenovirus AdV, pretreated as indicated. (E) Levels of infection in HEK293T cells treated with control, CD46, or CD55 siRNA. (F and G) Levels of infection in HeLa cells treated with (F) DMSO, 3-MA, KU55933, or Gö6976 or (G) control or TRIM21-directed siRNA. (H and I) HeLa cells after stimulation with (H) BSA, or IFN-α or (I) treated with DMSO, VCP inhibitor DBeQ, or proteasome inhibitor epoxomicin. (J) Immunoblot for AdV capsid component hexon and GAPDH in HeLa cells treated with DMSO or epoxomicin at indicated times after challenge with AdV+C3fBfD. (K) Levels of infection in NHLF cells treated with DMSO or epoxomicin. Data are representative of three experiments; results in (A) to (I) and (K) are percentage infected cells normalized to AdV only controls, mean ± SEM.

A characteristic of antiviral immunity is IFN regulation, and most antiviral genes are IFN-stimulated (20). TRIM21 is an IFN-stimulated gene (4, 5), and the efficiency of antibody-dependent intracellular neutralization is dependent on the amount of TRIM21 in the cell. To determine whether C3-mediated restriction was also IFN-inducible, cells were stimulated before infection (Fig. 4H). IFN increased the ability of the cell to restrict viral infection via complement C3, suggesting that complement recognition occurs through an IFN-stimulated gene.

Antibodies mediate intracellular neutralization by targeting viruses for degradation by the AAA–adenosine triphosphatase (ATPase) valosin-containing protein (VCP) (also known as p97) (3) and the proteasome (4). To determine whether C3 also mediates restriction by recruiting these enzymes, we tested the effect of DBeQ (a VCP inhibitor) and epoxomicin (a proteasome inhibitor). Each inhibitor perturbed restriction of AdV by C3, suggesting that C3 activates an intracellular, VCP, and proteasome-dependent pathway (Fig. 4I). To test directly whether the block to infection was the result of targeted degradation of incoming virions, we performed a fate-of-capsid experiment in which levels of the AdV major capsid protein, hexon, were measured at different time points. Hexon was rapidly degraded in a proteasome-dependent manner, but only when complement was present (Fig. 4J). These restriction phenotypes did not just occur in HeLa cells but could also be replicated in primary NHLF cells (Fig. 4K). Thus, in addition to activating innate immunity, the attachment of C3 to invading virions labels them for degradation by VCP and the proteasome, restricting virus infection.

Intracellular complement immunity conserved in mammals

Complement protein/receptor interactions are thought to be species-specific (21), although recent observations have disagreed with this view (22). We investigated whether intracellular complement immunity is conserved among mammals. Serum from different mammalian species was incubated with AdV, pelleted to remove unattached complement components, and then added onto HEK293T cells. Human, mouse, cat, rabbit, and guinea pig serum all elicited complement-mediated signaling (Fig. 5A). We then tested whether this signaling was present in cell lines derived from different mammals. Human (HEK293T), African green monkey (Vero), mouse [mouse embryonic fibroblasts (MEFs)], cat [feline embryonic airway (FEA)], and dog [Madin-Darby canine kidney (MDCK)] cells all signaled in response to complement-coated AdV (Fig. 5B). Thus, a system of intracellular complement immunity is conserved within mammals.

Fig. 5 C3 detection is conserved in mammals and is active against RNA and DNA nonenveloped viruses and bacteria.

(A) HEK293T cells challenged with AdV incubated with human, mouse, cat, rabbit, and guinea pig serum. (B) HEK293T (human), Vero (African green monkey), MEF (mouse), FEA (cat), and MDCK II (dog) cells challenged with AdV+Serum or AdV+HI Serum. (C) NF-κB activity after challenge with WT AdV, HPV, hAstV, FCV, HRV, PV, and CVB. WT AdV and HRV were carried out on HEK293T; HPV, PV, and CVB were carried out on HeLa; hAstV was carried out on Caco-2; and FCV was carried out on FEA cells. (D) NF-κB activity in HeLa after challenge with RSV incubated with sera. (E) NF-κB activity in HeLa, Caco-2, and MEF cells after challenge with WT Salmonella+C3fBfD. (F) MEF cells challenged with ΔSif Salmonella+C3fBfD. Results are representative of three experiments, as fold change over PBS-treated controls; data in (A) to (D), mean ± SEM; data in (E) and (F), mean ± SEM.

Intracellular complement immunity effective against diverse pathogens

Because complement-sensing was independent of viral PAMPs (Fig. 2H), we hypothesized that it may be an effective way of activating immunity during infection by diverse pathogens. Complement-sensing allowed detection of infection by replication-competent wild-type adenovirus 5 (WT AdV), confirming previous experiments that were carried out with replication-deficient virus lacking the E1 and E3 genes (Fig. 5C). In addition, infection by human papillomavirus virus (HPV)–like particles, human astrovirus 1 (hAstV), feline calicivirus (FCV), human rhinovirus 14 (HRV), poliovirus 2 (PV), and coxsackievirus B3 (CVB) also elicited NF-κB induction in a complement-dependent manner (Fig. 5C). These viruses use different strategies for cell entry. Adenovirus (23) and rhinovirus 14 (24) lyse the endosome during infection, whereas poliovirus (25) and coxsackievirus (26) form a pore in the endosome membrane. Accordingly, endosomal disruption assays revealed an association between endosomal disruption, and the potency of complement mediated signaling for these viruses (fig. S1K).

Complement deposition on an enveloped virus occurs on the lipid membrane and not its internal capsid. Consequently, during infection by an enveloped virus, complement should be left behind on the outside of the plasma membrane or inside endosomes. To test this, we infected cells with respiratory syncytial virus (RSV), an enveloped virus that enters through membrane fusion (27) in the presence of serum. As expected, no complement signaling was observed during RSV infection (Fig. 5D).

Nonenveloped viruses are not the only intracellular pathogen capable of being targeted by complement deposition. Salmonella enterica enterica serovar Typhimurium can escape from their salmonella-containing vesicles and replicate in the cytosol (28). Antibody deposited on bacterial surfaces has previously been shown to be detected by TRIM21 (1). In HeLa, Caco-2, and MEF cells, signaling in response to C3-coated wild-type bacteria was detected (Fig. 5E). Signaling was substantially more pronounced in the ΔSifA mutant, which has a reduced vesicular integrity and a greater propensity to enter the cytosol (Fig. 5F) (29). Thus, a variety of different pathogens can be detected by the presence of intracellular C3.

Viral antagonism to intracellular complement immunity

The hallmark of an antiviral response is that it should exert selection pressure on pathogens so as to evolve and escape it (30). To investigate the possibility of viral antagonism, we compared the activation of immunity by complement between different viruses. We observed that although there was complement-mediated NF-κB activation upon infection with viral particles such as AdV and HPV (Fig. 6A), this was substantially weaker during infection with viruses such as hAstV, PV, and HRV. This suggests that these viruses have strategies that allow them to evade detection by C3. hAstV is known to inactivate complement component C1, limiting the amount of complement deposition (10). Ultraviolet (UV)–inactivation of HRV restored complement-mediated NF-κB activation to the levels of replication-deficient vectors, AdV and HPV, suggesting that an encoded component was antagonizing C3-mediated signaling (Fig. 6A).

Fig. 6 Viral antagonism of C3-mediated signaling.

(A) NF-κB induction by the heat-labile component of serum activity after infection of HEK293T cells by AdV, HPV, hAstV, PV, HRV, and UV-inactivated HRV (UV-HRV), shown as fold change of Virus+Serum over Virus+HI Serum signaling levels. (B and C) Immunoblot for C3 (B) and AdV (C) in samples comprising C3 deposited on AdV incubated with recombinant HRV 3C Protease (Ext HRV 3C) or components thereof. (D) NF-κB activity in HEK293T expressing HRV 3C Protease (HRV 3C Pro) or PV 3C Protease (PV 3C Pro). (E) NF-κB activity in HEK293T treated with BSA, Ext HRV 3C, or recombinant PV 3C Protease (Ext PV 3C), with AdV complexes pelleted. (F) Levels of infection of HeLa cells challenged with AdV, AdV+Serum, AdV+HI Serum, and AdV+Serum+CVF treated with Ext HRV 3C, or Ext PV 3C, followed by pelleting. (G) Immunoblot for C3 after infection of HeLa cells expressing empty vector or HRV 3C Pro by AdV+C3fBfD or HRV+C3fBfD. (H) NF-κB activity in HEK293T cells treated with DMSO or 3C antagonist rupintrivir, challenged with HRV incubated with sera. (I) Immunoblot for C3 after infection of HeLa with AdV+C3fBfD or HRV+C3fBfD treated with DMSO or rupintrivir. (J) NF-κB activity induced by the heat-labile component of serum upon AdV, HRV, or PV infection of HEK293T cells treated with DMSO or rupintrivir at indicated times after infection. (K) NF-κB activity in THP-1 cells treated with DMSO or rupintrivir and challenged with sera-incubated HRV. (L) IFN-β ELISA from NHLF cells infected with HRV under different conditions after treatment with DMSO, Ext HRV 3C, or rupintrivir. Data from dot plots are from five experiments (dots are mean of 6 replicate samples), bar graphs are representative of three experiments, data in (D), (E), (H), and (K) are as mean ± SEM; data in (L) are mean ± SEM; data in (J) are mean ± SEM.

HRV expresses a cytosolic 3C protease (HRV 3C Pro) that we hypothesized to mediate antagonism, given previous observations of 3C-mediated cleavage of RIG-I (31). Using in silico analysis of potential 3C cleavage sites (32), both HRV 3C Pro and PV 3C protease (PV 3C Pro) were predicted to cleave complement C3. To test whether expression of 3C proteases enables viruses to inhibit intracellular C3 signaling, we incubated AdV+C3fBfD with albumin or with recombinant HRV 3C Pro (Ext HRV 3C) and observed cleavage of C3 by means of immunoblot (Fig. 6B). HRV 3C Pro did not cleave the viral proteins but removed the larger complement-bound band (Fig. 6C). Next, we expressed HRV 3C Pro or PV 3C Pro in HEK293T cells before infection with AdV+Serum. Expression of either 3C protease reduced NF-κB induction to levels observed with AdV+HI Serum and AdV+Serum+CVF, suggesting that complement signaling had been prevented (Fig. 6D). Indeed, expression of either protease was sufficient to cause AdV to behave similarly to HRV (AdV+HRV 3C Pro, AdV+PV 3C Pro) (Fig. 6D). To confirm that the observed effect of 3C protease was due to cleavage of complement and not a factor inside the cell, AdV incubated with different sera was treated with recombinant 3C protease before pelleting the virus to remove the protease, before addition to HEK293T cells (Fig. 6E). Preinfection treatment with recombinant 3C proteases (Ext HRV 3C and Ext PV 3C) perturbed NF-κB induction similarly to expression inside target cells. In addition to antagonizing sensing, HRV 3C Pro or PV 3C Pro were also effective in preventing complement-mediated restriction (Fig. 6F). In order to permit escape from intracellular complement immunity, 3C protease must be synthesized and act rapidly after infection. When we blotted for C3 after infection, full cleavage was noted for HRV+C3fBfD within 45 min (Fig. 6G). However, no changes for AdV+C3fBfD were observed after 2 hours. Overexpression of HRV 3C Pro was sufficient to confer C3 cleavage during AdV infection. Thus, cleavage of C3 is a fast and effective method of evading intracellular complement immunity.

We further hypothesized that if 3C protease allows viruses such as PV or HRV to evade intracellular complement immunity, then inhibition of 3C protease should make these viruses susceptible. To test this, we used rupintrivir, a specific, nonreversible inhibitor of the HRV 3C protease (33). Its addition to cells infected with HRV increased the levels of complement-mediated signaling so that HRV+Rupintrivir behaved similar to viruses lacking complement antagonism (Fig. 6H). Rupintrivir treatment was also sufficient to make PV visible to complement sensing. When cells were treated with rupintrivir, this lead to a loss of C3 cleavage by HRV (Fig. 6I). To investigate the kinetics of antagonism, rupintrivir was added at various times after infection with AdV, HRV, or PV, and its effect on signaling was determined. Rupintrivir inhibited a process that antagonized complement-mediated sensing within the first hour of infection (Fig. 6J). This time scale matches the kinetics of C3 cleavage by 3C. By 1.5 hours after infection, addition of rupintrivir was unable to restore heat-labile signaling, which is consistent with C3 cleavage being completed by this time. Rupintrivir had no impact on complement-mediated AdV sensing. These rapid kinetics are similar to those previously reported for PV 2A protease (34), formed from the same polypeptide as 3C protease. Complement also activates cell-surface signaling in professional cells. Virally encoded 3C antagonism did not prevent this response. Rupintrivir had no effect on signaling in THP-1 cells infected with HRV+Serum, a finding that is consistent with their extracellular detection of the C3-coated virus (Fig. 6K). Our data suggest that rupintrivir could be used therapeutically to enhance host detection of 3C-expressing viruses. We found that NF-κB induction by complement, facilitated by rupintrivir treatment, was sufficient to induce IFN-β secretion upon HRV infection (Fig. 6L). Thus, viral expression of 3C protease cleaves complement C3 inside the cell to disable signaling and restriction. Inhibition of 3C protease counteracts this effect.

Complement-mediated signaling independent of known PRRs, but requires MAVS

The ability of complement C3 to enable sensing of a wide variety of different pathogens suggests that a particular PAMP is not required for detection. To test this, we blocked Toll-like receptor activity by inhibiting MyD88 and TRIF (Fig. 7A), as well as knockdown of RIG-I and MDA5 (Fig. 7B and fig. S1D). In addition, we inhibited Syk (Fig. 7C), which blocks signaling through Fc receptors and type II C-type lectin receptors, and depleted cells of STING, the endoplasmic reticulum–associated adaptor involved in cytosolic DNA detection (Fig. 7D and fig. S1E). All of these perturbations to known signaling pathways had no effect on complement-mediated signaling.

Fig. 7 C3-mediated signaling is MAVS dependent.

(A) NF-κB activity in HEK293T cells treated with control peptides, or inhibitors of MyD88 or TRIF after challenged with AdV+C3fBfD or LPS. (B to F) NF-κB activity after challenge by AdV incubated as indicated on HEK293T cells treated with (B) control, RIG-I, or MDA5 siRNA; (C) DMSO or inhibitors Syk I and Syk III; (D) control, MAVS, or STING-directed siRNA; (E) control, TRAF2, TRAF3, TRAF5, TRAF6, or pooled TRAF2,3,5,6 (siT2,3,5,6) siRNA; and (F) control, TRAF6, or p62 siRNA. (G) IRF3, -5, and -7 binding to consensus DNA response elements in HEK293T cells treated with siRNA as in (E). (H) IRF3, -5, and -7 binding to DNA response elements in HEK293T cells treated with control, MAVS, or TBK1 specific siRNAs, or with inhibitor, BX795. Data are representative of three experiments; results in (A) to (H) are as fold change over PBS-treated controls, mean ± SEM.

However, depletion of mitochondrial antiviral signaling (MAVS) abolished the heat-labile component of signaling, suggesting that MAVS is required for C3-mediated immune activation (Fig. 7D and fig. S1E). Sendai virus has been reported to activate MAVS activation by inducing aggregation (35), as detected with microscopy and semidenaturing gel electrophoresis. We carried out microscopy for MAVS and mitochondrial marker TOM20 in cells treated with phosphate-buffered saline (PBS), AdV, C3fBfD, or AdV+C3fBfD 6 hours after infection (fig. S2A) and noted no apparent aggregation at this time point. This is consistent with published literature, in which aggregation of MAVS only occurs after 9 hours of infection (35). Overexpression of MAVS lead to increased protein levels, but the same localization was observed (fig. S2B). We carried out semidenaturing gel electrophoresis but were not able to detect aggregation, except when MAVS was overexpressed (fig. S2C). Thus, we observed rapid MAVS-dependent signaling that was independent of aggregation, with similar kinetics to that observed during MAVS-induced IRF3 dimerization (36) and IκBa phosphorylation (37).

Owing to the involvement of MAVS in C3 sensing, we investigated whether TNF receptor–associated factor (TRAF) proteins are required for signal transduction. Depletion of TRAF6 reduced complement-mediated NF-κB signaling (Fig. 7E and fig. S1F), although simultaneous knockdown of TRAF2, -3, -5, and -6 (siT2,3,5,6) was required to abolish signaling, similar to published data (38). Given the importance of TRAF6, we also investigated the adaptor p62 (39). Depletion of p62 (Fig. 7F) lead to a similar loss of NF-κB induction, as did knockdown of TRAF6. Individual depletion of TRAF proteins had a partial effect on IRF activation (Fig. 7G), whereas siT2,3,5,6 depletion was completely inhibitory, demonstrating redundancy. In addition, knockdown of TBK1 or inhibition with BX795 (40) prevented IRF activation (Fig. 7H), demonstrating that TBK1 is important in C3-mediated signaling. Thus, intracellular C3 activates MAVS with downstream signaling proceeding via the TRAF proteins in a partially redundant manner.

Intracellular antibody and complement cooperate to activate signaling

TRIM21 has previously been identified as a PRR for cytosolic antibody (1, 4), and depletion of TRIM21 (fig. S1C) resulted in decreased signaling from AdV+Serum (Fig. 8A). However, TRIM21 depletion only decreased the heat-stable component of signaling because only AdV+IgG, and not AdV+C3fBfD, signaling was affected. This confirms that although TRIM21 is a receptor for antibody, it is not a receptor for complement C3. Codepletion of TRIM21 and MAVS abolished signaling from AdV+IgG, AdV+C3fBfD, or AdV+Serum, suggesting that the ability of intracellular serum proteins to induce signaling (Fig. 1A) is entirely dependent on these two pathways.

Fig. 8 C3 leads to concurrent and independent signaling and restriction.

(A) NF-κB activity in HEK293T cells treated with control, or MAVS and TRIM21 siRNA. (B) Levels of infection of HeLa cells treated with control, or MAVS and TRIM21 siRNA, then DMSO or epoxomicin, after infection with AdV, AdV+Serum, or AdV+HI Serum. (C) Cells as in (B), infected with AdV or AdV+C3fBfD. (D) NF-κB activity in HEK293T cells treated with control, or MAVS and TRAF6 siRNA, transfected with Beads+C3fBfD. (E) IRF3, -5, and -7 binding to DNA response elements in HEK293T cells treated with DMSO or epoxomicin. (F) Levels of infection of HeLa cells treated with DMSO or epoxomicin and panepoxydone after challenge with AdV, treated as indicated. (G) Relative level of Sindbis infection of HeLa cells in the presence of fresh medium (DMEM), medium with IFN-α, or supernatant from cells treated with control or MAVS siRNA, challenged with PBS or AdV and C3fBfD. (H) Viability of HeLa cells challenged with dilutions of C3fBfD in a spreading HRV infection assay. Cells were treated with control or MAVS siRNA, then with DMSO, MG132, or panepoxydone. Data are representative of three experiments; results in (A) to (G) are mean ± SEM; data in (H) are mean ± SEM.

We noted that knockdown of MAVS did not have any effect on AdV infection in the presence of serum, whereas knockdown of TRIM21 increased infection (Fig. 8B), which is consistent with loss of antibody-dependent intracellular neutralization. However, neither MAVS nor TRIM21 depletion had any effect on the restriction of AdV+C3fBfD (Fig. 8C). This suggests that C3 mediates an effector response that is independent of MAVS and TRIM21.

C3 signaling and restriction contribute independently to reduce viral infection

Next, we investigated the relationship between C3-mediated signaling and restriction. Transfection of complement C3–coated beads was sufficient for NF-κB induction, and this activity was dependent on MAVS and TRAF6 (Fig. 8D), demonstrating that PAMPs are not required for C3 detection. Moreover, there was strong induction of IRF3, -5, and -7 in cells in which restriction was prevented by using a proteasome inhibitor (Fig. 8E). Thus, C3-mediated detection is not dependent on viral components liberated during restriction. Last, restriction of AdV was not affected by addition of the IκB inhibitor panepoxydone (Fig. 8F), which previously inhibited signaling (Fig. 3A), suggesting that restriction is also independent of cell signaling. Thus, C3 mediates independent signaling and restriction pathways, as has been shown previously for antibody (1, 2, 4).

Although the above data show that C3 mediates an effector response that is independent of signaling, we investigated whether signaling leads to an antiviral state that also contributes toward reducing viral infection. Supernatant was collected from HEK293T cells 3 days after challenge with PBS, AdV, C3fBfD, or Adv+C3fBfD and transferred to fresh uninfected HeLa cells. These uninfected cells were then challenged with an IFN-sensitive reporter virus (Sindbis with a GFP transgene). Supernatant from AdV+C3fBfD–infected cells was sufficient to protect cells from Sindbis infection (Fig. 8G) in a MAVS-dependent manner. Adenovirus is relatively resistant to IFN treatment (41), but rhinovirus is IFN-sensitive (42). To investigate the relative contributions of C3-mediated signaling and restriction on a spreading infection, we added epoxomicin or panepoxydone during HRV infection of HeLa cells treated with control or MAVS-targeted small interfering RNA (siRNA) (Fig. 8H). Epoxomicin reversed the ability of complement to restrict HRV, which is in line with the finding for AdV. However, panepoxydone and MAVS depletion also partially reversed restriction, suggesting that signaling in response to C3 contributes to the antiviral state. Thus, C3 elicits a direct antiviral restriction mechanism and promotes signaling, which induces an antiviral state.


We investigated whether extracellular serum proteins mediate intracellular immune responses inside nonimmune cells upon pathogen infection. We found evidence for intracellular pathogen detection via the sensing of attached complement. Complement C3 is covalently deposited onto pathogens in the extracellular space. During infection, pathogens carry C3 inside the cell, activating innate immunity and mediating a restriction pathway that degrades virus. Key features of this intracellular complement immune response are that it is cell-type–independent and requires translocation of C3 into the cytosol. This allows pathogens to be sensed during their natural infection pathway rather than relying on the capture of immune complexes by professional cells. The importance of this system is suggested by the evolution of countermeasures that allow certain viruses to antagonize it. Both rhinovirus and poliovirus have evolved a mechanism to evade C3 detection by cleaving C3 using an encoded 3C protease.

C3 signaling was independent of known PRRs, such as Toll-like receptors and RIG-like receptors, and occurred for DNA and RNA nonenveloped viruses and intracellular bacteria, suggesting that exposure of PAMPs is not directly responsible. Detection was dependent on MAVS, activating transcription factors, and proinflammatory cytokine secretion. The minimum requirement for NF-κB activation was found, comprising reconstituted alternate pathway components (C3, factor B, and factor D) bound to beads and transfected into cells. C3-coated AdV only initiated signaling when intracellular and post-endosomal, suggesting that the initial receptor is cytosolic.

Sensing of complement-coated virus results in a restriction response that leads to degradation of virus and inhibition of infection. This process involves the proteasome, which is similar to the restriction phenotypes after recognition of retroviral capsid by TRIM5α (43) and antibody by TRIM21 (1, 4). As with antibody, complement-mediated degradation also requires the AAA-ATPase VCP (3).

Taken together with our previous work on intracellular antibodies (16), the data here suggest that serum proteins may have an important role in detecting breaches in the physiological barrier of cell membranes. C3 has recently been described as mediating immunity to Chlamydia psittaci, which is an obligate intracellular bacterium in mice (44). In addition, C1q has been implicated as providing an intracellular signaling role by interacting with RIG-I and MDA5 (45). Whether these effects are related to the signaling described in this study is still to be investigated. Ultimately, there may be a number of host proteins and corresponding set of receptors that stimulate immunity based on a system of topological displacement.

The discovery that picornaviruses such as HRV and PV encode a 3C protease capable of cleaving C3 and inactivating its intracellular function has implications for treatment strategies. HRV is particularly sensitive to IFN (42), and treatments that prevent the virus from suppressing IFN induction could be highly effective. Rupintrivir is used to control viremia and has been reported to decrease cytokine production (46). However, cytokines such as IL-6 and IL-8 were measured 3 days after infection, by which time rupintrivir may have reduced viremia, resulting in a complex phenotype. In our study, rupintrivir was removed 3 hours after infection, resulting in the measurement of responses from a single round of infection and unbiased by differing levels of replication. We propose that in vivo, rupintrivir may have dual functions, increasing the complement-mediated detection of virus and hence production of IFN-β as well as inhibiting the production of progeny virions by interference with polypeptide cleavage.

Last, this system illustrates how a single DAMP can lead to the simultaneous detection and restriction of a pathogen. Signaling in response to C3-coated pathogens is independent of PAMPs and does not require proteasome-dependent restriction to reveal molecular determinants. Restriction of the pathogen occurs immediately upon infection and is not dependent on the simultaneous activation of cell signaling. However, these two responses cooperate to enable the efficient reduction of spreading infection. Although this is a relatively new concept in immunology, it is exemplified by the receptors TRIM5α (43, 47), TRIM21 (1, 4), and tetherin (48, 49). These systems allow the cell to capitalize on a single sensing event to elicit multiple immune responses.

Materials and methods


HEK293T, HeLa, Vero, Caco-2, FEA, MDCK II, and MEF cell lines were maintained in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal calf serum, 100 U/ml penicillin, and 100 μg/ml streptomycin. MEF cells were obtained from C57BL/6 mice as in (5). THP-1 cells were maintained in RPMI-FCS with penicillin/streptomycin as above and 20 nM 12-O-tetradecanoylphorbol-13-acetate (TPA). NHLFs from Lonza (Basel, Switzerland) were maintained in Fibroblast Growth Medium 2 (Lonza), supplemented with 10% fetal calf serum, 0.1% insulin, 0.1% amphotericin-B, and 0.1% gentamicin according to the manufacturer’s instructions.

Viruses and bacteria

E1- and E3-deleted adenovirus vector bearing a GFP transgene (AdV) was from Viraquest (North Liberty, Iowa). WT AdV, HRV, PV, and enterovirus 71 (EV71) were all from ATCC (Manassas, Virginia) and grown in HeLa cells, and cell-free supernatant was harvested and frozen 8 days after infection. HPV 16 L1/L2 virus-like particles were produced with plasmids from M. Müller, German Cancer Research Center, following the protocol of Buck and Thompson (50). Human Astrovirus 1 molecular clone pAVIC (hAstV) (51) was provided by I. Goodfellow, University of Cambridge, and recovered following the method of Fuentes et al. (52). Respiratory syncytial virus GFP-expressing molecular clone rgRSV(224) (RSV) (53) was provided by M. Peeples, Ohio State University. RSV was expanded in HeLa cells and harvested 2 days after infection. Feline calicivirus strain F9 (FCV) was provided by D. Brown, University of Cambridge. FCV was expanded in FEA cells and harvested 24 hours after infection. Coxsackie virus B3 molecular clone expressing a GFP transgene (CVB), eGFP-CVB3 (54), was obtained from J. L. Whitton, Scripps Institute. Sindbis virus–GFP (Sindbis) vector was produced by means of in vitro transcription of linearized plasmids DH-BB and pSin-eGFP with mMessage mMachine SP6 kit (Life Technologies, Carlsbad, California) and transfection of RNA into HEK293T cells with Lipofectamine 2000. WT AdV, HRV, and CVB were prepared by means of CsCl banding (4), and HPV was prepared by means of banding in OptiPrep (Sigma-Aldrich, St Louis, Missouri). PV, EV71, hAstV, RSV, and Sindbis were used directly from cell-free supernatant, with FCV by pelleting cell-free supernatant through 30% sucrose. Viruses were quantified by the median tissue culture infectious dose (TCID50) method, or GFP where possible. Wild-type (strain 12023) or ΔsifA Salmonella enterica enterica serovar Typhimurium were prepared as in (1).

Serum and serum treatments

Normal human serum (Serum) as well as the complement series–depleted serum (Serum-C1, -C2, -C3, -C5, and -C6) were obtained from Sigma. Rabbit serum was also obtained from Sigma. Antibody-depleted serum (Serum-Ig), with the control serum were from SCIPAC (Kent, UK). Serum depleted of C4 (Serum-C4), factor B (Serum-fB), and factor D (Serum-fD), along with matching control serum, was from Quidel (San Diego, California). Mouse and cat serum were from Equitech Bio (Kerrville, Texas). Heat inactivation was carried out at 52°C for 45 min. EGTA treatment was performed by adding 0.2 M EGTA in 0.2 M MgCl2. Purified complement C3, factor B, and factor D were from Millipore (Billerica, Massachusetts). Cobra venom factor (CVF) was purified from Naja naja kaouthia venom (MP Biomedicals, Santa Ana, California) by using the method of Vogel and Muller-Eberhard (55). Recombinant human rhinovirus 3C protease was from Expedeon (Cambridgeshire, UK), and recombinant poliovirus 3C protease was prepared by I. Goodfellow, University of Cambridge. Serum was used at the maximum concentration that had no impact on entry, as found by the endosomal disruption assay (typically 1/5 to 1/10 dilutions in PBS), with depleted serum verified by using sheep red blood cell hemolysis assays (56).


Serum components binding to AdV ELISAs—goat anti-human IgM HRP, goat anti-human IgA HRP, goat anti-human IgG HRP, mouse anti-human IgD HRP, and mouse anti-human IgE HRP—were obtained from Abcam (Cambridge, UK). Detection of human C3b, C4b, Mannan-binding lectin, and C-reactive protein were modified from ELISA detection kits from Abcam. Pentraxin 3 detection was by means of modified ELISA kits from Hycult Biotech (Uden, Netherlands). Immunoblot of CD11b, CD11c, CD18, CD35, CD46, CD55, p62, and complement C3 was by antibodies from Abcam. Immunoblot of RIG-I, MDA5, phosphorylated IKK, total IKK, phosphorylated IκB, total IκB, phospshorylated NF-κB p65, total p65, TRAF2, TRAF3, TRAF5, TRAF6, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (loading control) used antibodies from Cell Signaling Technologies (Beverly, Massachusetts). Immunoblot for STING and MAVS used antibodies from Santa Cruz Biotechnology (Dallas, Texas). Immunoblot and immunofluorescence for AdV hexon used polyclonal goat anti-adenovirus 5 antibody from Millipore. Immunofluoresence was performed by using MAVS and TOM20 antibodies from Santa Cruz Biotechnology.

siRNA knockdown

TRIM21 siRNA was carried out as in (4). siRNA against RIG-I, MDA-5, STING, MAVS, TBK1, TRAF2, TRAF3, TRAF5, and TRAF6 were from Santa Cruz Biotechnology. siRNA against CD46 and CD55 were from Origene (Rockville, Maryland). p62 siRNA was from Life Technologies. siRNA was transfected by using RNAiMAX from Life Technologies.


Pepinh-MyD and Pepinh-TRIF (Invivogen, Toulouse, France) were used alongside the supplied control peptide at 50 μM. One μM 5Z-7-Oxozaeanol (Sigma), 2 μg/ml panepoxydone (Enzo Life Sciences, Exeter, UK), 200 nM IKK Inhibitor VII (Millipore), 1 μM Bafilomycin A1 (Santa Cruz Biotechnology), 20 nM CID1067700 (Millipore), and Epoxomicin (Millipore) was added at 2 μM. KU55933 (Millipore) was used at 10 μM, Gö6976 (Millipore) was used at 5 μM, DBeQ (BioVision, Milpitas, California) was used at 8 μM, and rupintrivir (Santa Cruz Biotechnology) was used at 2 μM. All were added 1 hour before infection.

Plasmids and reporter constructs

Luciferase reporter cell lines were produced by transfection of pGL4.32 NF-κB (Promega, Fitchburg, Wisconsin). NHLFs were transduced with Cignal Lenti NF-κB (Qiagen, Hilden, Germany). Minimal-CMV promoter–driven nano-luciferase plasmid, pNL1.1CMV, was from Promega. Expression constructs for HRV 3C Protease and PV 3C Protease were from Origene. Flag-MAVS was cloned into pcDNA3.1(+) by using NotI and XhoI primers: forward 5′- ACGTGCGGCCGCCACCATGGACTACAAAGACGATGACGACAAGATGCCGTTTGCTGAAGACAAGAC-3′ and reverse 5′-GGCGTCTGCACTAGTGACTCGAGTCTA-3′. Plasmids were transfected by using Lipofectamine 2000 (Life Technologies).

Luciferase reporter assay

Cells were plated at 1 × 104 per well in 96 well plates. The day after plating, virus and serum were incubated 1:1 for 1 hour before addition to cells. Viruses were added at the following titers: AdV, 1.5 × 106 IU per well; WT AdV, 4.22 × 106 IU per well; HPV, L1/L2 protein at 1.92 μg per well; hAstV, 7.5 × 105 IU per well; FCV, 2.8 × 105 IU per well; HRV, 3.16 × 106 IU per well; PV, 2.0 × 106 IU per well; CVB, 2.0 × 106 IU per well; EV71, 2.0 × 105 IU per well; and RSV, 2.0 × 105 IU per well. For positive controls, 10 pg/ml recombinant TNF or 50 pg/ml LPS was used. Cells were incubated for 6 hours (18 hours for HPV) at 37°C before addition of Steadylite Plus luciferase reagent (Perkin Elmer, Waltham, Massachusetts) and reading on a BMG Pherastar FS platereader. For assays involving pelleting, AdV and PBS or serum were mixed before layering onto 30% sucrose and spinning for 4 hours at 28,000 x g at 4°C, resuspending pellet in PBS. For transfection of beads, 0.25 μm biotin-coated latex beads (Sigma) were incubated with 500 μg/ml C3, factor B, and factor D (Sigma) or PBS for 1 hour. Beads were transfected by using Lipofectamine 2000 (Life Technologies). For infections with S. Typhimurium, overnight stationary phase cultures were diluted 1/33 into 5 ml of fresh LB medium and grown at 37°C, 400 rpm for 3.5 hours. A 1/500 dilution of these cultures was incubated with 250 μg/ml C3, factor B, and factor D for 15 min before addition of a 10 μl mix onto cells. Luciferase was read 7 hours after infection.


Serum-binding ELISAs carried out incubating AdV+Serum for 1 hour before ultracentrifugation at 28,000 x g for 4 hours at 4°C through 30% sucrose. The pellet was resuspended in PBS before binding to polystyrene high-binding Microlon plates (Greiner-Bio-One, Frickenhausen, Germany) overnight, blocking for 2 hours with Marvel, before addition of serum. Quantification was carried out by using kits listed in antibodies section, with subtraction of signal from Marvel-blocked wells. NHLF cells were plated and infected in the same manner as for luciferase reporter assays. Cells were fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton-X100 in the plate 4 hours after infection, and abundance of phosphorylated proteins was measured by using the NF-κB signaling kit (Cell Signaling Technologies). For analysis of cytokine production, supernatant was harvested 24 hours after infection and analyzed by means of ELISA kits: IL-6 (Life Technologies), TNF (Life Technologies), CCL4 (Abnova, Taipei City, Taiwan), and IFN-β (Life Technologies). Plates were read by using a SpectraMAX 340PC (Molecular Devices, Sunnyvale, California) at 450 and 650 nm.

Endosomal disruption assay

Endosomal disruption assay was modified from Seth et al. (57). Cells were plated at 1 × 104 per well in 96 well plates, in DMEM containing 12 μg/ml pNL1.1CMV (Promega). Virus mixes were added as for the NF-κB reporter. Cells were left for 18 hours after infection at 37°C, before addition of NanoLuc reagent (Promega) and reading on a BMG Pherastar FS platereader.

Confocal microscopy

For intracellular C3 visualization, HeLa cells were plated on glass coverslips and permitted to adhere overnight. Purified C3 was labeled with AlexaFluor488-SDP Ester (Life Technologies) at 4°C for 2 hours, before storage at –80°C. AdV was incubated with 250 μg/ml labeled C3, factor B, and factor D for 15 min before infection for 30 min. For MAVS activation visualization, HEK293T cells were transfected with pcDNA3.1-EV or pcDNA3.1-FLAG-MAVS 48 hours before infection by using Fugene 6. Cells were plated on glass coverslips pretreated with CellTak (Becton Dickinson, Franklin Lakes, New Jersey) and permitted to adhere overnight. AdV was incubated with 250 μg/ml of C3, factor B, and factor D; or AdV was incubated with PBS; or C3, factor B, and factor D were incubated alone for 1 hour before addition to cells for 6 hours. Cells were washed and fixed with 4% paraformaldehyde, before permeabilization with 0.1% Triton X-100 and blocking with 5% bovine serum albumin (BSA). Immunostaining for AdV was by anti-adenovirus 5 goat polyclonal antibody (Millipore), MAVS was by mouse monoclonal (Santa Cruz Biotechnology), and TOM20 was by a rabbit polyclonal (Santa Cruz Biotechnology). Coverslips were mounted on medium containing 4' 6-diamidino-2-phenylindole (DAPI) and imaged by using a Zeiss 63X objective on a Jena LSM 710 microscope (Carl Zeiss MicroImaging, Oberkochen, Germany).

Neutralization assay

HeLa cells were plated at 1 × 105 per well in 6 well plates. The day after plating, AdV and sera were incubated 1:1 for 1 hour before addition to cells. AdV was infected with 3.75 × 104 IU per well. Eighteen hours after infection, cells were harvested and GFP positive cells enumerated by means of flow cytometry by using a BD FACS Calibur 2. IFN treatment was with addition of 1000 U/ml Interferon-α (Sigma) 24 hours before infection.

Fate-of-capsid experiment

HeLa cells were plated at 1 × 105 per well in 6 well plates. The day after plating, AdV and sera were incubated 1:1 for 1 hour before addition to cells. AdV was infected with 7.5 × 104 IU per well. Cells were harvested by scraping at 0, 1, 2, 4, and 6 hours after infection, followed by immunoblot for AdV.

MAVS aggregation

MAVS aggregation was carried out as previously described by Hou et al. (35). Briefly, HEK293T cells were challenged with PBS, AdV, C3fBfD, or AdV+C3fBfD for 6 hours, after which they were harvested, and mitochondrial extracts were produced by douncing and ultracentrifugation. HEK293T cells expressing Flag-MAVS was harvested 3 days after transfection. Samples were controlled for total protein and ran in Tris-Acetate gels as semidenaturing gel electrophoresis, with immunoblot carried out for MAVS.

Transfer of media experiment

Sindbis experiment was carried out as previously described (1). Briefly, HeLa cells were infected with PBS, AdV, C3fBfD, or AdV+C3fBfD. Cell supernatant was collected 3 days after challenge and used to treat fresh “reporter” HeLa cells. Other reporter HeLa cells were provided fresh media (DMEM) or were stimulated with 1000 U/ml Interferon-α (Sigma). The following day, cells were infected with Sindbis virus at multiplicity of infection = 0.3, before enumeration of GFP-positive cells 1 day after infection.

Restricting titer experiment

Cells were plated at 5 × 103 per well in 96 well plates. Serial dilutions of 250 μg/ml C3, factor B, and factor D were incubated with 20× TCID50 units HRV for 1 hour before addition to cells. DMSO, MG132, or panepoxydone were added 1 hour before infection, and cells were washed 16 hours after infection into DMEM supplemented with 2% fetal calf serum. Seven days after infection, a MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assay (Sigma) was performed, permeabilizing cells by DMSO, measuring absorbance with a SpectraMAX 340PC (Molecular Devices) at 540 nm.

Supplementary Materials

References and Notes

  1. Acknowledgments: We thank I. Goodfellow for the kind gift of reagents, including the Caco-2 cell line, human Astrovirus-1 infectious clone pAVIC, and PV 3C protease. This work was funded by the Medical Research Council (UK; U105181010), the European Research Council (281627IAI), and the Frank Edward Elmore Fund of the University of Cambridge School of Clinical Medicine (J.C.H.T.). The data are presented in the main manuscript and the supplementary materials.
View Abstract

Navigate This Article