The TAM receptor Mertk protects against neuroinvasive viral infection by maintaining blood-brain barrier integrity.

The TAM receptors Tyro3, Axl and Mertk are receptor tyrosine kinases that dampen host innate immune responses following engagement with their ligands Gas6 and Protein S, which recognize phosphatidylserine on apoptotic cells. In a form of apoptotic mimicry, many enveloped viruses display phosphatidylserine on the outer leaflet of their membranes, enabling TAM receptor activation and downregulation of antiviral responses. Accordingly, we hypothesized that a deficiency of TAM receptors would enhance antiviral responses and protect against viral infection. Unexpectedly, mice lacking Mertk and/or Axl, but not Tyro3, exhibited greater vulnerability to infection with neuroinvasive West Nile and La Crosse encephalitis viruses. This phenotype was associated with increased blood-brain barrier permeability, which enhanced virus entry into and infection of the brain. Activation of Mertk synergized with interferon-β to tighten cell junctions and prevent virus transit across brain microvascular endothelial cells. Because TAM receptors restrict pathogenesis of neuroinvasive viruses, these findings have implications for TAM antagonists that are currently in clinical development.

INTRODUCTION

The TAM receptors Tyro3, Axl, and Mertk have pleiotropic functions in cancer metastasis, angiogenesis, thrombus stabilization, and innate immune regulation[1,2]. Axl and/or Mertk are expressed on cells involved in immune control and trafficking, including macrophages, dendritic cells (DCs), platelets, and endothelial cells[1]. In comparison, Tyro3 expression is prominent on central nervous system (CNS) neurons[3]. TAM receptors signal upon recognition of their phosphatidylserine-bound ligands, Gas6 and Protein S[4]. The consequences of TAM signaling depend on cell type. For example, TAM receptors are important for NK cell development[5], and their inhibition may license NK cells to reject metastatic tumors[6]. Axl and Mertk signaling in endothelial cells modulates angiogenesis[7-9], whereas their signaling in platelets promotes thrombus stabilization[10]. In DCs, activation of Axl down-regulates production and signaling of pro-inflammatory cytokines by interacting physically with the R1 subunit of the type I interferon (IFN) receptor (IFNAR1) to promote expression of the negative regulators SOCS1 and SOCS3[11]. The TAM receptors also have essential roles in clearance of apoptotic cells by macrophages, retinal pigment epithelial cells, and other professional phagocytes[12-14]. The TAM ligands Gas6 and Protein S physically bridge a TAM receptor expressed on the surface of a phagocyte to phosphatidylserine expressed on the surface of the apoptotic cell.

TAM receptors are therapeutic targets in cancer because of their effects on tumor angiogenesis, NK cell licensing, tumor cell survival, metastasis, and immune suppression in tumor-associated macrophages[6-9]. Several antagonists and blocking antibodies are under evaluation in clinical trials[15,16]. TAM receptor agonists also may prove useful in the treatment of autoimmunity because of their ability to down-regulate cytokine production[17]. Less is known about the net effect of TAM receptor blockade during viral infection. In a form of apoptotic mimicry, many enveloped viruses incorporate phosphatidylserine into their virion membranes[18,19] and bind Gas6 and Protein S to facilitate recognition by TAM receptors and activation of signals that dampen antiviral responses[19]. Studies with influenza and respiratory syncytial viruses suggest that Axl blockade by antibodies protects against infection and disease pathogenesis[20]. However, an antiviral phenotype after TAM inhibition may not be universal, as herpes simplex virus (HSV) infection was more severe in Axl mice[21].

We hypothesized that deletion of TAM receptors might restrict WNV infection and protect against pathogenesis for two reasons: (1) cell culture studies indicated that TAM receptors can augment flavivirus entry[18] and create a more permissive innate immune environment for replication[19]; and (2) WNV causes significant morbidity in humans after it crosses the blood-brain barrier (BBB) and replicates within neurons. Type I IFN signaling strengthens the BBB during viral infection by tightening junctions between brain microvascular endothelial cells (BMECs)[22]. Since TAM receptors can negatively regulate type I IFN signaling[11,19], deletion of TAM receptors could enhance both IFN signaling and BBB integrity. Unexpectedly, we observed that Axl, Mertk, Axl but not Tyro3 mice were more vulnerable to WNV infection. This phenotype was associated with markedly impaired BBB integrity during infection. Our results establish a preferential role for Mertk in protecting against neuroinvasive viruses, which occurs at least in part through its ability to sustain the BBB during infection.

RESULTS

Axl and Mertk but not Tyro3 are required for control of WNV infection in vivo

To evaluate the role of TAM receptors in WNV infection, we infected WT, Tyro3, Axl, Mertk, and Axl C57BL/6 mice with WNV (New York 2000 strain) by subcutaneous inoculation (). Unexpectedly, Axl, Mertk, and Axl, but not Tyro3 mice were more vulnerable to WNV infection than WT mice, with ~80% mortality in Axl or Mertk mice (P < 0.0005) and ~95% mortality in Axl mice (P < 0.0005).

We found that an absence of TAM receptors had a relatively minor effect on viral burden in peripheral organs, with increased viremia and viral load observed only at 2 days post-infection (dpi) in serum (26-fold, P < 0.05) and 4 dpi in the spleen (33-fold, P < 0.05), respectively, in the Axl mice (). No significant differences in viral burden were observed in serum, spleen, or kidney in Axl or Mertk mice compared to WT mice ().

Higher levels of WNV infection were apparent in the CNS of Axl, Mertk, and Axl mice compared to WT controls. At 4 dpi, WNV was detected in CNS tissues of TAM receptor deficient mice: 2 of 9 Axl, 2 of 10 Mertk, and 3 of 10 Axl mice had brain homogenates that were positive for infectious WNV compared to 0 of 10 WT mice (). Analogously, at 6 dpi, 5 of 9 Axl, 4 of 8 Mertk, and 4 of 9 Axl mice had detectable WNV in the spinal cord compared to 0 of 10 WT mice (). Viral titers also were increased at 8 dpi in the brain (29 to 72-fold increase, P < 0.05) and spinal cord (7 to 135-fold increase, P < 0.05) of Axl, Mertk, and Axl Mertk mice.

Higher levels of CNS infection in the TAM receptor KO animals could suggest that Axl or Mertk restrict viral replication preferentially in target cells of the CNS. To test this hypothesis, we inoculated WNV intracranially in Axl, Mertk, or Tyro3 mice. To ensure detection of small differences that might be missed by whole brain analysis, viral burden in the cerebral cortex, subcortex, brain stem, and cerebellum was measured at 3, 5, and 6 dpi after infection. However, no difference in viral burden in the CNS of Axl, Mertk, or Tyro3 mice was observed following intracranial inoculation ( P > 0.9). These data suggest that TAM receptors do not restrict WNV replication directly in the CNS.

Effects of TAM receptors on antiviral adaptive immune responses

To assess whether part of the CNS virological phenotype could be attributed to defects in adaptive immunity, we measured anti-WNV IgM and IgG levels at 4 and 8 dpi in WT and Axl mice (). Axl mice had slightly greater anti-WNV IgG titers (2.5-fold, P < 0.005) and slightly lower neutralizing titers at 8 dpi (0.4-fold, P < 0.005). These small differences were unlikely to explain the prominent lethality phenotype observed in Axl mice after WNV infection.

Axl has been proposed to modulate CD8+ T cell responses by DC efferocytosis and antigen cross-presentation[21]. To test whether Axl and Mertk affected T cell responses during WNV infection, we measured the levels and antigen specificity of CD8+ T cells from the spleen of WNV-infected WT, Axl, Mertk and Axl mice (). Although we detected similar numbers and percentages of CD4+ and CD8+ T cells in the spleens of WT, Axl, Mertk and Axl mice at 8 dpi, we observed fewer WNV tetramer-positive CD8+ T cells in Axl mice (). After ex vivo peptide restimulation of splenocytes, we observed a lower percentage and number of WNV-specific CD8+ T cells that expressed IFN-γ in Axl but not in Mertk CD8+ T cells (, P < 0.05). There also were fewer WNV-specific CD8+ T cells that expressed IFN-γ or TNF-α from Axl mice (, P < 0.05). These data suggest that Axl is required for optimal priming of a CD8+ T cell response during WNV infection.

We next assessed leukocyte responses within the brain at 8 dpi. We observed greater numbers of leukocytes and antigen-specific CD8+ T cells in the brains of Mertk and Axl mice than WT controls (, P < 0.05) but no statistically significant difference in the number of CD11b+CD45hi macrophages or CD11b+CD45lo microglia (). Greater numbers of infiltrating immune cells likely result from the higher viral burden in the CNS, enhanced BBB permeability, or both.

BBB integrity during WNV infection requires Axl and Mertk but not Tyro3

Because we observed early accumulation of WNV in the brains of Axl, Mertk, and Axl mice, we assessed whether these mice had altered BBB permeability that could impact virus entry into the CNS. We injected sodium fluorescein (molecular weight (MW): 376) intraperitoneally into naïve and WNV-infected WT, Tyro3, Mertk, and Axl mice and measured extravasation into the brain 45 minutes later (). Even in the absence of infection, naïve Axl mice had slightly greater BBB permeability; in comparison, no statistically significant differences were observed in naïve Tyro3, or Mertk mice, although there was a trend toward enhanced BBB permeability in uninfected Mertk mice (, left panel). WNV infection resulted in increased sodium fluorescein extravasation into the CNS at 4 dpi ( right panel), as reported previously[22]. BBB permeability was greater in WNV-infected Axl, Mertk, and Axl, but not Tyro3 mice compared to WT mice at this time point, with Mertk and Axl mice exhibiting the most pronounced phenotypes (, right panel). These results suggest that Axl and Mertk are required to maintain BBB integrity during infection and prevent early virus invasion into the CNS, with Mertk having the most prominent effect.

As an independent measure of BBB permeability, we used confocal microscopy to assess leakage of endogenous IgG (MW: 150,000) into the brain parenchyma following WNV infection. Although minimal IgG was detected in the brains of uninfected mice corresponding to all genotypes (data not shown), IgG accumulation became apparent at 4 dpi, with Axl mice exhibiting greater leakage than WT mice (). Thus, in the context of WNV infection, the BBB of Axl mice was more permeable to small molecules and larger proteins.

TAM antagonist disrupts BBB integrity and accelerates WNV infection in the brain

To corroborate the phenotypes observed with TAM receptor KO mice, we treated WT mice with a 40 mg/kg dose of BMS-777607, a small molecule inhibitor of c-Met, Ron, Flt-3, and TAM receptor signaling[23], by oral gavage beginning one day prior to WNV infection and continuing until 4 dpi. BMS-777607 treatment resulted in enhanced lethality of WT mice after WNV infection (, P < 0.05) with virus present in the brain at 4 dpi in 3 of 6 drug-treated mice compared to 0 of 6 control mice (). BMS-777607 increased BBB permeability in WT but not Axl Mertk mice at 4 dpi ( P < 0.05), nor in drug-treated uninfected animals ().

TAM receptor KO mice are vulnerable to La Crosse virus infection

We hypothesized that Mertk mice also might be vulnerable to other viruses that enter the brain through a hematogenous route. La Crosse virus (LACV) is a neurotropic orthobunyavirus that causes meningoencephalitis, predominantly in children[24]. We observed enhanced mortality in Axl or Mertk mice infected with LACV compared to WT mice (, 50% versus 9%, P < 0.05). Mertk mice had increased BBB permeability at 4 dpi whereas LACV-infected Axl mice did not (). We also found higher levels of viral RNA in the brains of Mertk but not Axl mice at 8 dpi () after LACV infection.

Cytokine and chemokine levels in serum of WNV-infected mice

Elevated levels of some pro-inflammatory cytokines (e.g., TNF-α) open the BBB[25], whereas others (e.g., type I IFN) close the barrier[22]. Because TAM receptors negatively regulate cytokine production, we measured their levels in the serum of naïve and WNV-infected mice (). In naïve mice, Axl, Mertk, and Axl mice had higher levels of IL12-p40 although other pro-inflammatory cytokines were similar compared to WT animals. At 4 dpi, levels of TNF-α, IL-1β, IL-6, IL-12(p40), RANTES, and KC were slightly (~2-to-3 fold, P < 0.05) higher in the serum of Axl, Mertk, and Axl mice (). Type I IFN levels in serum also were slightly higher at 4 dpi in Axl and Axl mice (1.3- to 1.5-fold P < 0.05) but not in Mertk mice (P > 0.9) ().

We also assessed cytokine levels in the brains of Axl, Mertk, and Axl mice that were inoculated via intracranial injection with WNV () in order to measure cytokine expression levels in the context of equivalent WNV burden in WT and TAM receptor-deficient mice (see ). We found no difference in expression of TNF-α, IL-1β, IL-6, TGF-β1 and TGF-β3 mRNA among any of the genotypes. Consistent with this finding, antibody blockade of TNF-α in vivo did not change the BBB permeability defect in Axl–/–Mertk mice during WNV infection (data not shown). The net effect of differences in levels of pro-inflammatory cytokines on BBB permeability in TAM receptor-deficient mice remains unclear.

Axl and Mertk signaling improves BBB integrity in vitro

TAM receptors are present on the surface of mouse BMECs in vivo and in vitro, with higher expression of Mertk compared to Axl[26,27] (). To explore whether TAM receptor signaling modulates endothelial barrier integrity and WNV transit, we used an in vitro model of the BBB[22]. Primary mouse BMECs are cultured in the upper chamber of a transwell, with primary astrocytes in the lower chamber. Transendothelial electrical resistance (TEER) across the BMEC monolayer measures barrier integrity, with higher resistance indicating a tighter barrier. TEER was lower across Axl BMECs (, P < 0.05) at baseline. In response to WNV infection, and as expected, WT and Axl BMEC barriers exhibited increased TEER compared with mock-infected barriers, but Axl BMEC barriers failed to tighten as much as WT BMECs (, 0.85-fold, P < 0.0001). We observed no difference in WT and Axl BMEC viability (data not shown).

We next evaluated whether changes in TEER in Axl BMECs impacted transit of WNV across the endothelial barrier. We added WNV to the upper chamber and after 6 h measured virus that had crossed the BMEC barrier into the lower chamber; this time point precedes de novo spread of WNV infection[28]. Consistent with lower TEER, Axl BMECs had higher amounts (9-fold, P < 0.0001) of WNV crossing into the lower chamber compared to WT cells (). Higher levels of WNV in the lower chamber could result from decreased binding of WNV to BMECs in the absence of Axl and Mertk, since these receptors are engaged by WNV at the plasma membrane[18]. However, we found slightly higher amounts of WNV associated with BMECs lacking Axl and Mertk expression at 6 h after infection ( Thus, Axl and Mertk are not required for binding of WNV to BMECs, and TAM receptor signaling sustains the integrity of the endothelial barrier, which restricts WNV transit.

The decrease in TEER in response to WNV infection in Axl BMECs might be due to an altered production or response to cytokines (e.g., TNF-α), which independently affect the barrier. Indeed, we detected slightly increased levels (1.4- to 2.6-fold P < 0.05) of TNF-α, IL-1β, and IFN-γ during WNV infection of BMECs ( and data not shown). Other cytokine and chemokine levels (e.g., IL-2, IL-3, IL-6, IL-13, KC, MIP1α, RANTES) were similar in WT and BMECs (data not shown). Since TNF-α and IL-1β can disrupt BBB integrity[22], we tested whether blockade of these cytokines in vitro during WNV infection might differentially alter barrier integrity. Treatment of BMECs with blocking antibodies against IL-1β and TNF-α prior to infection minimally increased TEER (1.1 to 1.2-fold, P < 0.05) in both WT and Axl BMECs, and this effect was evident only at late time points (). As studies have suggested that TAM receptors can modulate the responsiveness of endothelial cells to TNF-α[22], we treated WT and Axl BMECs with soluble TNF-α; however no difference in TEER was observed (P > 0.9, ).

As Axl associates physically with IFNAR1 and modulates type I IFN signaling in DCs[11], we hypothesized that TAM receptors might affect IFNAR signaling in BMECs, which could affect BBB tightening after WNV infection. To test this idea, we treated WT or Axl BMECs with IFN-β (IFNAR-dependent) or IFN-λ (IFNAR-independent) and measured TEER over 6 hours. Whereas a deficiency of Axl and Mertk did not affect the ability of IFN-λ to increase TEER[22,29], Axl BMECs were less responsive to IFN-β treatment in terms of TEER changes (), although IFNAR expression was similar in WT and Axl BMECs (). These results suggest that Axl and Mertk expression in BMECs is required for the full effect on barrier integrity of IFN-β. Colocalization of the tight junction (TJ) proteins claudin-5 and ZO-1 is enhanced by type I IFN during WNV infection[22]. We observed diminished claudin-5 expression at the cell membrane in addition to discontinuities in TJs in Axl BMECs (). Treatment of WT and Axl BMECs with IFN-β or IFN-λ enhanced ZO-1 and claudin-5 colocalization in WT and Axl BMECs, although discontinuities in TJs were still observed in Axl BMECs. In contrast, treatment with TNF-α disrupted TJ in both WT and Axl BMECs. Diminished TJ integrity, increased virus transit, and altered IFN-β responsiveness in Axl BMECs may explain how endothelial cell expression of TAM receptors can enhance BBB integrity during viral infection.

We next tested the effects of Gas6[4], which binds to and activates both Axl and Mertk, either alone or in combination with IFN-β, on TEER in WT BMECs. We observed dose-dependent tightening of BMEC monolayers in response to Gas6 (). The combination of Gas6 and IFN-β rapidly tightened the barrier, with markedly increased TEER values observed within 15 min of treatment. Similar increases in TEER values were observed in a human BMEC line and with physiologic concentrations of Protein S, which functions as a ligand for Mertk and Tyro3 but not Axl[4] (). Consistent with its dominant role in maintaining BBB integrity in vivo (see ), signaling through Mertk was required for the increase in TEER in response to Gas6 (). Whereas Tyro3 and Axl BMECs responded to Gas6 similarly compared to WT cells (), Axl BMECs exhibited decreased baseline TEER (). A combined genetic deficiency of Axl and Mertk did not increase the barrier defect beyond that observed in Mertk BMECs in response to Gas6 and IFN-β ().

To investigate further the interaction between TAM receptor and type I IFN signaling, we treated Ifnar BMECs with Gas6. Type I IFN signaling was not required for Gas-6-dependent effects on TEER at 30 and 60 minutes after treatment, although an absence of IFNAR diminished the amplitude of the effect at all time points (). We confirmed these findings with MAR1-5A3[30], an IFNAR-blocking antibody, which was incubated with WT BMECs immediately prior to Gas6 addition (). Thus, TAM receptor ligands activate Mertk to tighten the junctions of BMEC monolayers in a manner that cooperates with but does not require IFNAR signaling, and the effect of Mertk on endothelial barrier integrity is amplified when type I IFN signaling occurs concurrently.

We examined how Gas6-dependent TAM receptor signaling affected the activity of Rac1, a Rho family GTPase that regulates cytoskeletal dynamics, TJ integrity, and paracellular permeability[31]. IFN-β enhances BMEC barrier formation in part by activating Rac1[22], and Mertk signaling promotes Rac1 activation in macrophages in the context of phagocytosis of apoptotic debris[32]. We measured GTP-bound, activated Rac1 after treatment with IFN-β, Gas6, or IFN-β and Gas6. Notably, Gas6 treatment was sufficient to enhance Rac1 activation in BMECs, similar to the effect of IFN-β alone ( 2-fold, P < 0.005). The combination of Gas6 and IFN- β led to a further increase in Rac1 activation. Finally, blockade of Rac1 activation prevented Mertk- or IFN-β-dependent tightening of BMEC barriers (). Thus, cytoskeletal reorganization resulting from Gas6-induced activation of Rac1 is likely required to sustain TJ integrity and the endothelial barrier.

We evaluated Axl and Mertk expression in cells of the neurovascular unit by confocal microscopy (). In addition to expression on endothelial cells (), we observed co-staining of TAM receptors in S100β+ astrocytes and CD11b+ myeloid cells. To evaluate whether TAM receptor expression on astrocytes contributed to endothelial barrier integrity, using the in vitro BBB model, we tested whether the TEER response to Gas6 stimulation was different with WT versus Axl astrocytes. However, deletion of Axl and Mertk in astrocytes had no effect on TEER ().

The effect of Mertk expression on BBB permeability occurred both in vitro and in vivo, without appreciable effects on viral replication in peripheral organs or on CD8+ T cell responses. To confirm that the dominant effect on BBB permeability of Mertk occurred at the level of the neurovascular unit and not in peripheral immune cells, we generated Mertk bone marrow chimeric mice. To prevent adventitious effects of radiation on the BBB, the heads of mice were shielded with lead (). Bone marrow chimeras with Mertk-deficient radio-resistant non-hematopoietic cells (SJL→Mertk) exhibited the same BBB permeability defect on day 4 after WNV infection as Mertk mice whereas the reciprocal chimeras (Mertk→SJL) with Mertk-sufficient non-hematopoietic cells did not (). A trend towards a parallel effect on viral burden was observed with 7 of 8 SJL→Mertk mice having detectable WNV RNA in the brain at 4 dpi compared to 3 of 8 Mertk→SJL mice ( P = 0.06). These results are consistent with a model in which Mertk expression on radio-resistant cells within the CNS is required for maintenance of BBB integrity.

DISCUSSION

Many enveloped viruses bind to and activate TAM receptors to disable innate immune responses and enhance infection[19]. Because we previously observed slightly lower levels of WNV replication in Axl DCs[19], we hypothesized that TAM receptor-deficient mice would be protected against lethal WNV infection. However, we show here that a genetic deficiency of Axl and Mertk resulted in the early appearance of WNV into the CNS, which resulted in enhanced viral load and mortality. The increased mortality was associated with increased BBB permeability and revealed a dominant role for Mertk in maintaining the integrity of this key barrier during viral infection. Using an in vitro BBB model, we found that Mertk promoted endothelial barrier integrity by maintaining the co-localization of TJ proteins. The barrier-tightening effect of Mertk signaling was cooperative with the response to IFN-β.

Our discovery that Axl and Mertk mice were more vulnerable to neuroinvasive WNV and LACV infections suggests that although enveloped viruses can usurp TAM receptors, these proteins nonetheless can restrict the pathogenesis of some viruses that gain entry into the CNS. When WNV was introduced directly in the CNS by intracranial injection in TAM receptor-deficient animals, no increase in viral replication in different brain regions was observed. Thus, blockade or ablation of TAM receptors may have varying effects on viral pathogenesis depending on the balance between virus binding to TAM receptors via Protein S and/or Gas6 and the resulting effects on the intracellular antiviral environment and/or vascular endothelial barrier integrity. Our data also demonstrate a separate role of Axl in modulating T cell immunity, which could impact viral clearance in different tissues including the CNS.

Although we did not assay the level of infection in TAM-deficient DCs in vivo, peripheral viral burden was similar in WT, Axl, and Mertk animals, suggesting that the earlier and greater viral burden in the CNS reflects accelerated virus entry due to impaired BBB integrity and that TAM receptor engagement by WNV is not required for infection in vivo. However, WNV infection was slightly greater in blood and the spleen of Axl animals, the mechanism for which requires further study. Bone marrow chimera studies revealed that the BBB permeability phenotype tracked with a loss of Mertk expression on radio-resistant and not radio-sensitive hematopoietic cells. Mertk+ radio-resistant cells in the CNS also include microglia[33,34], which can express Axl upon activation; as such, we do not exclude the possibility that the enhanced lethality in WNV-infected Axl–/– mice might reflect a role for this TAM receptor in microglia.

IFNAR-dependent signaling was required for optimal endothelial cell barrier integrity after treatment with the TAM ligand, Gas6. Studies in myeloid cells have shown that Axl associates with and signals through the IFNAR1 subunit[11], and it is plausible that Mertk could function analogously in endothelial cells. Although biochemical corroboration is required, Mertk may modulate the barrier tightening effects of IFN-β in endothelial cells because of a specific interaction with IFNAR1[11].

The defects in the stabilization of endothelial TJs in Axl BMECs are consistent with TAM receptor-dependent regulation of cytoskeletal reorganization in other cell types, which occur through Rac1[32,35]. We observed Rac1 activation in response to Gas6 in endothelial cells, which was amplified with concurrent IFN-β treatment. Our experiments are consistent with a model in which TAM receptor (preferentially Mertk) and IFNAR signaling together activate Rac1, which leads to tightening of BMEC junctions and restriction of virus transit into the CNS ().

Axl mice had impaired CD8+ T cell responses to WNV infection, which could impact mortality by affecting CNS viral clearance[36,37]. We did not observe CD8+ T cell defects in peripheral or CNS tissues of Mertk mice, suggesting distinct functions of Axl and Mertk in modulating adaptive immunity during virus infections. A diminished T cell response in Axl mice is consistent with a prior study of HSV infection [21]. However, the attenuated CD8+ T cell response does not explain the early appearance of WNV in the brain at day 4 in Axl mice, since this time point precedes the induction of a WNV-specific CD8+ T cell response[38].

Prior reports have suggested possible functions of TAM receptors in endothelial cells. Mice lacking all three TAM receptors (Axl TKO) reportedly have a disrupted BBB, although this was attributed in part to autoimmune disease[39]. Exogenous administration of Protein S enhanced BBB integrity after ischemic stroke, although this phenotype required Tyro3 and not Axl or Mertk[26]. Thus, individual TAM receptors may have unique roles in maintaining BBB integrity under different inflammatory conditions. Mertk may have a dominant function in maintaining the BBB after viral infection, whereas Tyro3, which also is expressed on neurons, may be more important in the context of cerebral ischemia. Protein S, which is a ligand for both Tyro3 and Mertk, but not Gas6, is likely to be the relevant TAM ligand, as Protein S is abundant in serum where Gas6 is present at lower levels[40].

Our discovery that TAM signaling regulates BBB integrity in the context of viral infections has clinical implications, since TAM receptor antagonists are being developed as cancer therapies[41]. Mertk blockade could increase the risk of neuroinvasion and pathogenesis of certain viruses, including WNV and LACV. Indeed, in studies with a broad-spectrum inhibitor of TAM signaling[42], we observed increased lethality after WNV infection and increased BBB permeability. Further experiments are warranted to define the net effects of TAM receptor blockade in the context of infection by different families of visceral and encephalitic viruses.

ONLINE METHODS

Viruses and cells

The WNV strain (3000.0259) was isolated in New York in 2000 and passaged once in C6/36 Aedes albopictus cells. Mice were inoculated subcutaneously in the footpad with 102 plaque forming units (pfu) of WNV diluted in Hanks balanced salt solution. Viral titers in tissues were analyzed by plaque assay using Vero cells, as described previously. The LACV strain (original strain) was provided by Andrew Pekosz (Johns Hopkins University, Baltimore, Maryland, USA) and passaged twice in Vero cells to produce a virus stock.

Mice

C57BL/6J wild-type (WT) mice were commercially obtained from Jackson Laboratories. Axl, Mertk, Axl, and Tyro3 mice have been published[43] and were backcrossed for ten generations. All mice were housed in a pathogen-free mouse facility at the Washington University School of Medicine and experiments were performed in accordance with federal and University regulations. The protocols were approved by the Institutional Animal Care and Use Committee at the Washington University School of Medicine (Assurance Number: A3381-01). Mice (8 to 10 week-old, both sexes) were inoculated subcutaneously via footpad injection with 102 pfu of WNV or 105 focus-forming units (ffu) of LACV, both diluted in 50 μl of Hanks balanced salt solution (HBSS) supplemented with 1% heat-inactivated fetal bovine serum (FBS). For intracranial infection, 101 pfu of WNV in 20 μl was injected into the right cerebral hemisphere.

Measurement of viral burden

At specified time points after WNV and LACV infection, serum was obtained by intracardiac heart puncture, followed by intracardiac perfusion (20 ml of PBS), and organ recovery. Organs were weighed, homogenized using a bead-beater apparatus, and WNV was titrated by plaque assay on Vero cells[44]. Brains from LACV- and WNV-infected mice were harvested at day 8 after infection and the total RNA was extracted using the RNeasy kit (Qiagen). For LACV, viral load in the brain was determined by qRT-PCR. Briefly, all reactions were assembled in a final volume of 25 μl with 300 ng of RNA, 10 μM forward and reverse primers (LACV: Forward 5’-CCTTGCTGCAGTTAGGATCTTCTT-3’, Reverse 5’- CCACTCTCCAAATTTAGG-GTTAGC-3’; GAPDH: Forward 5’-AATGGTGAAGGTCGGTGTG-3’, Reverse: 5’-GTG GAGTCATACTGGAACATGTAG-3’), 5 μM probe (LACV: 5’-5’-/56-FAM/ AGGCCAAGGCTGCTCTCTCGCGTA-/36-TAMSp/-3'; GAPDH: 5’-/56-FAM/TGCAAATGG/ZEN/CAGCCCTGGTG/3IABkFQ/-3’) and 12.5 μl of TaqMan master mix (Applied Biosystems) using the following cycling condition: 48°C for 30 min, 95°C for 10 min, followed by 45 cycles of 95°C for 15 s and 60°C for 1 min. Quantitation of WNV RNA was performed as previously described[45]. The levels of viral RNA were expressed on a log10 scale as genomes equivalents/g after comparison with a standard curve produced using serial ten-fold dilutions of WNV or LACV RNA.

Quantification of type I IFN activity

Levels of type I IFN were determined using an EMCV cytopathic effect bioassay performed in L929 cells as described previously[46]. Serum samples were treated with citrate buffer (40 mM citric acid, 10 mM KCl, 135 mM NaCl [pH 3.0]) for 10 minutes and neutralized with medium containing 45 mM HEPES pH 8.0. The amount of type I IFN per ml of serum was calculated from a standard curve using IFN-β (PBL Assay Science). The specificity of the antiviral activity was confirmed by pre-incubating L929 cells for 2 hours with 25 μg/ml of the IFNAR-blocking MAb MAR1-5A3 or an isotype control MAb GIR-208[30].

Cytokine bioplex assay

WT and Axl, Mertk, and Axl mice were infected with WNV, and at specified times blood was collected and serum was prepared. The BioPlex Pro Assay was performed according to the manufacturer's protocol (BioRad). The cytokine screen included IL-1α, IL-1β, IL-2, IL-3 IL-4, IL-5, IL-6, IL-9, IL-10, IL-12p40, IL-12p70, IL-13, IL-17, Eotaxin, G-CSF, GM-CSF, IFN-γ, KC, MCP-1 MIP-1α, MIP-1β, RANTES (CCL5), and TNF-α.

BBB permeability measurements

Mice were infected with 102 pfu of WNV or diluent (mock) and BBB permeability was assessed after 4 days. Sodium fluorescein (100 mg/ml) was administered via an intraperitoneal route in 100 μl. After 45 minutes, blood was collected by cardiac puncture into EDTA-coated tubes. Mice were perfused and CNS tissues were harvested, homogenized into PBS, clarified by centrifugation, precipitated in 1% trichloroacetic acid, and neutralized with borate buffer (Sigma-Aldrich). Fluorescence emission at 485 and 528 nm was determined using a microplate reader Synergy™ H1 and Gen5™ software (BioTek Instruments, Inc.). Fluorescein concentration was calculated from a standard curve and tissue fluorescence values were normalized to the plasma fluorescence values from the same mouse.

Endogenous mouse IgG was detected in brain sections using an AlexaFluor-488 anti-mouse IgG antibody. Nuclei were stained with Topro-3. Images were acquired with a laser scanning confocal microscope (Zeiss LSM 510 META) and analyzed with LSM image browser software (Zeiss).

Drug treatment studies

The small molecular receptor inhibitor BMS-777607 (Selleckchem) was dissolved in DMSO at a stock concentration of 52 mg/ml. Mice received either 1 mg BMS-777607 dissolved in 50 μl DMSO or 50 μl DMSO vehicle control by oral gavage beginning one day prior to infection and continuing through day 4 after infection.

B cell and antibody responses

The levels of WNV-specific IgM and IgG were determined using an ELISA against purified WNV E protein, as described previously[47]. Plaque reduction neutralization assays on BHK21-15 cells were performed after mixing serial dilutions of serum with a fixed amount (102 pfu) of WNV as previously described[48].

Cellular immune responses

WT and TAM receptor-deficient mice were infected in the footpad with 102 pfu of WNV and at 8 days after infection, spleens and brains were harvested after extensive cardiac perfusion with PBS. Splenocytes were dispersed into single cell suspensions with a cell strainer. Brains were digested collagenase and leukocytes were isolated as previously described[49]. Intracellular IFN-γ or TNF-α staining was performed after ex vivo restimulation with a Db-restricted NS4B immunodominant peptide using 1 μM of peptide and 5 μg/ml of brefeldin A (Sigma) as described[50]. Cells were stained with the following antibodies and processed by multi-color flow cytometry on an LSR II flow cytometer (Becton Dickinson): CD3 (Becton Dickinson, clone 145-2C11), CD4 (Biolegend, clone RM4-5), CD8β (Biolegend, clone YT5156.7.7), CD19 (Invitrogen, clone 6D5), CD45 (Biolegend, clone 30-F11), CD11b (Becton Dickinson, clone M1/70), IFN-γ (Becton Dickinson, clone XMG1.2), TNF-α (Biolegend, clone MP6-XT22). Flow cytometry data were analyzed using FlowJo software (Treestar).

Transwell cultures and TEER measurements

WT and TAM receptor-deficient BMECs and WT HCMEC/D3 cells were grown until fully polarized in transwell cultures[22]. BMECs were grown above astrocyte cultures, whereas HCMEC/D3s were grown without astrocytes. TEER was measured via chopstick electrode with an EVOM voltmeter (World Precision Instruments). Resistance values are reported as Ω/cm2, with the resistance value for transwell inserts with no cells subtracted as background. TEER measurements were collected at 6 h following infection with WNV at MOI 0.01 or treatment with murine IFN-λ3 (100 ng/ml), murine or human IFN-β (10 ng/ml) (PBL Assay Science); mock wells were treated with culture medium. To block IFN-α/β signaling, BMEC cultures were treated with 25 μg/ml of the blocking MAb MAR1-5A3 for one hour prior to infection. A non-binding MAb (GIR-208) was used as an isotype control. To measure virus transit across the endothelial barrier, WNV was added to the upper chamber of the transwell at an MOI of 0.01. After 6 h, virus in the lower chamber was measured by qRT-PCR. Recombinant full-length Gas6 was generated in HEK293 EBNA cells as previously described[4]. Human Protein S was purchased from Haematologic Technologies (HCPS-0090).

Rac1 studies

Rac1 immunoprecipitation experiments were performed with an activated Rac1 agarose bead kit (Cell BioLabs, Inc.) according to the manufacturer's instructions. Purified, GTP-bound protein and unpurified BMEC protein lysates were separated via gel electrophoresis on 10% bis-Tris gels (Life Technologies) and transferred onto iBlot nitrocellulose transfer membranes (Life Technologies) according to standard protocols. To test the effects of Rac1 inhibition on TEER in BMECs, the Rac1 inhibitor Z62954982 (Cayman Chemical) was added at a concentration of 1 mM.

TAM receptor expression

Freshly isolated BMECs were stained with anti-Mertk (R&D Systems AF591) or anti-Axl (R&D Systems AF854) antibodies followed by a polyclonal secondary antibody. A control goat polyclonal antibody and Axl BMECs were used to assess the specificity of staining. Brain sections were stained with antibodies against Mertk (R&D Systems, AF591), Axl (R&D Systems, AF854), s100-β (Abcam, ab41548), CD31 (Becton Dickinson, 550274), or CD11b (Becton Dickinson, ab8878).

Bone marrow chimeric mice

Six week-old B6.SJL-Ptprc (CD45.1 SJL, Jackson Laboratories) and CD45.2 Mertk mice were anesthetized with ketamine and positioned within lead shielding to limit exposure of the brain to radiation[51]. Mice were then placed into a pie container within a cesium irradiator so that the head was shielded from the radiation source and then irradiated with a dose of 8 Gy. Six hours after irradiation, 107 bone marrow-derived leukocytes of a given genotype were injected intravenously in 100 μl of PBS. Seven weeks after bone marrow transplantation, reconstitution was confirmed by flow cytometry with greater than 95% of B cells, 90% of neutrophils, and ~75% of T cells of donor origin.

Power calculation and data generation

To determine mouse group sizes for individual experiments, power analysis was performed using the following values: probability of type I error = 0.05, power = 80%, 5-fold hypothetical difference in mean, and population variance. This analysis indicated that minimum sample sizes of 8 animals for virological or immunological studies were required to detect an approximately 10-fold level of difference. Studies were performed in an unblended manner, and randomization was not used. No animals, samples, or data points were excluded from any analysis.

Data analysis

All data was analyzed using Prism software (GraphPad4, San Diego, CA). Kaplan-Meier survival curves were analyzed by the log rank test. Differences in viral burden, cytokine levels, and cell numbers were analyzed by the Mann-Whitney test.