Aurora A controls CD8+ T cell cytotoxic activity and antiviral response

Aurora A is a serine/threonine kinase whose role in cell cycle progression and tumour generation has been widely studied. Recent work has revealed an unexpected function for Aurora A during CD4+ T cell activation and, also, in graft versus host disease development. However, it remains unknown whether Aurora A is involved in CD8+ T cell effector function and in cytotoxic T lymphocyte-mediated antiviral response. Here, we show that Aurora A chemical inhibition leads to an impairment of both the peptide-specific cytotoxicity and the degranulation activity of CD8+ T cells. This finding was similarly proven for both mice and human CD8+ CTL activity. As a result of Aurora A blockade, we detected a reduction in the expression induced by T cell activation of genes classically related to the effector function of cytotoxic T lymphocytes such as granzyme B or perforin1. Finally, we have found that Aurora A is necessary for CD8+ T cell-mediated antiviral response, in an in vivo model of vaccinia virus infection. Thus, we can conclude that Aurora A activity is, indeed, needed for the proper effector function of cytotoxic T lymphocytes and for their activity against viral threats.

Scientific RepoRts | (2019) 9:2211 | https://doi.org/10.1038/s41598-019-38647-y capacity of human and mouse CD8 + T cells. Furthermore, Aurora A pharmacological blockade impairs the upregulated expression of cytotoxicity related genes and TCR downstream signalling. This reduction in all the cytotoxic features decreases the ability of CD8 + T cells to respond against vaccinia infection in an in vivo mouse model.

Results and Discussion
Aurora A regulates CD8 + t cell-mediated cytotoxicity. In order to assess the role of Aurora A in CD8 + T cell-mediated cytotoxic response, OTI mouse T lymphoblasts were cocultured for 6 h with target cells (EL4 cell line) in the presence of Aurora A specific inhibitor (MLN8237) or vehicle (DMSO). Target cells were previously pulsed with the H-2 Kb-restricted Ovalbumin peptide (257-264; OVAp), or left unpulsed; stained with CFSE (1 and 0.1 µM, respectively) and mixed in a 1:1 ratio. A significant decrease in the percentage of cytotoxicity was detected as a result of Aurora A blockade (Fig. 1A). This impairment in the cytotoxic activity was similarly detected by using different ratios of T cells vs target cells (Fig. 1A). Furthermore, when different dosages of Aurora A inhibitor were applied, only doses up to 10 μM or higher were able to significantly reduce cytotoxicity (ratio 1:5) (Fig. 1B). Likewise, the application of a different Aurora A inhibitor (Aurora A inhibitor I), also caused a significant decrease on the cytotoxic capacity (ratio 1:5) of CD8 + T cells (Fig. 1C). As a complementary approach, the effect of Aurora A inhibition on T cell degranulation activity was measured. The expression of CD107a in CD8 + mouse (Fig. 1D) or human T lymphoblasts (Fig. 1E) was assessed by flow cytometry staining, in the presence of Aurora A inhibitor (MLN8237) or vehicle. Mouse CD8 + T cells were cocultured for 6 h with OVAp-pulsed EL4 mice cells, while human T lymphoblasts were incubated for 6 h with anti-human CD3/CD28 antibodies. Flow cytometry analysis showed a significant reduction in the degranulation rate of MLN8237-treated cells in mice (Fig. 1D). Likewise, treatment of human CD8 + T cells activated with anti-CD3/CD28 antibodies with MLN8237 resulted in a decrease in both the percentage of CD107 + T cells and the frequency of CD107 + IFNγ + cells (Fig. 1E). As a control, and taking into account the sequence similarity between Aurora A and Aurora B, both the cytotoxicity (Supp. Fig. 1A) and the degranulation capacity (Supp. Fig. 1B) of CD8 + mouse T cells was, in a similar manner, studied in the presence of AZD1152, a specific inhibitor of Aurora B, at two different concentrations (0.1 and 10 µM). Nevertheless, no differences were detected in both parameters by blocking Aurora B activity, suggesting a specific role of Aurora A paralogue in this phenotype. Considering that drug treatment might be affecting the viability of CD8 + T lymphocytes, the toxicity of the inhibitors per se to CD8 + mouse T cells was tested at the incubation time (6 h) and concentrations previously used (10 µM for MLN8237, 1 µM for Aurora A Inhibitor I and 0.1 and 10 µM for AZD1152) without detecting any significant effects on cell viability (Supp. Fig. 2).
These results highlight the crucial role of Aurora A in CD8 + T cell "killing" capacity. Moreover, this role seems to be specific of Aurora A compared to Aurora B, despite their structural similarities.
Aurora A controls the expression of cytotoxic genes and the tCR signal transduction cascade in CD8 + t cells. The induction of essential genes for effector CD8 + T cells cytotoxic activity (eomes, perforin and granzyme B) was next analysed. Assessment of mRNA expression after stimulation of MLN8237 or vehicle-treated mouse CD8 + T lymphoblasts with OVAp-pulsed EL4 cells for 2 and 4 h showed that Aurora A blockade led to a significant impairment in the induction of eomes, perforin1 and granzyme B ( Fig. 2A). As a control, cells were equally treated with Aurora B inhibitor (AZD1152), however, no significant changes were observed in the expression of these genes. Likewise, when human CD8 + T lymphoblasts gene expression pattern was analysed after stimulation of MLN8237 or vehicle-treated cells with anti-CD3/CD28 monoclonal antibodies, human granzyme B, perforin1 or eomes gene upregulation was similarly decreased by Aurora A blockade (Fig. 2B).
To gain insights into the mechanism of Aurora A effect on signalling pathways of CTLs activation, the kinetics of phosphorylation of molecules implicated in the TCR activation cascade (CD3ζ, PLCγ1 and ERK1/2) was determined in extracts from vehicle (Ctl) or MLN8237-treated mouse CD8 + T lymphoblasts activated with OVAp-pulsed EL4 cells. Immunoblot analysis of phosphorylation showed a significant impairment in the activation of CD3ζ, PLCγ1 and ERK1/2 ( Fig. 2C,D).

Aurora A inhibition reduces CD8 + t cells cytotoxic in vivo response against Vaccinia infection.
In order to assess the influence of Aurora A inhibition in an in vivo infection model, we generated effector CD8 + T cells by infecting OTI mice with rVACV-OVA for 4 days. Next, we adoptively transferred these cells into MLN8237 or vehicle treated wild-type (WT) mice pre-infected with rVACV-OVA. We tested the cytotoxic activity of transferred OTI CTLs against vaccinia by measuring the viral load from both infected ears after 48 h (Fig. 3A). OTI CD8 + T cell transfer clearly reduced viral replication. Moreover, viral titration showed an increased number of viral plaques in those mice reconstituted with CTLs and treated with Aurora A inhibitor (grey triangles), in comparison with those with the CTLs and vehicle-treated (black triangles), indicating that Aurora A blockade leads to a defect in CD8 + T cells cytotoxic response against the virus (Fig. 3B). Conversely, MLN8237-treated infected mice that have not been adoptively transferred with OTI CD8 + T cells (grey squares), included as control, showed a decrease in viral titration compared to the vehicle-treated ones (black squares) (Fig. 3B). This suggests that Aurora A inhibition by MLN8237 might be interfering with viral replication per se or enhancing the innate immune clearance of the virus. Moreover, since CD8 + T cells transfer leads to a viral content still higher in the presence of MLN8237 than in vehicle treated mice, the reduction of the cytotoxic activity by Aurora A inhibition would be even stronger than the one detected.
These results show that Aurora A acts as a key molecule in the effector function of cytotoxic T lymphocytes, both in mice and human. We have previously described, in CD4+ T cells, the molecular mechanisms underlying Aurora A role on TCR pathway and microtubule dynamics 6 . However, in this work we have focused in the physiological effect of Aurora A blockade in CTL-activity during a global immune response. We have discovered a key role of Aurora A on CD8 + T cell in vitro cytotoxic capacity and cytotoxic gene expression and on the ability  of CD8 + T cells to respond against infectious agents such as vaccinia virus. Taking into account the importance of Aurora A during cell division 4 and the importance of the centrosome for lytic granule polarization and release in T cells 12,13 , Aurora A might be blocking the proper movement of lytic granules towards the IS, and therefore reducing their killing capacity. Aurora A is important for both CD4 + T cell physiology 6 and GVHD outcome 7,8 . This study supports its involvement on another relevant function of the adaptive immune response, namely CD8 + CTL-mediated cytotoxicity, which is crucial for the development of a proper global immune response.
T cell deficiency in cancer patients, mainly due to antineoplastic drugs, enhances patients susceptibility to infections 14,15 . Since some Aurora A inhibitors are being currently tested in the treatment of different types of malignancies 16 , Aurora A impairment of T cell cytotoxic response in infectious diseases would increase patients susceptibility to develop opportunistic infections. In fact, since it has been described that non-oncogenic acute infections promote tumor growth by sequestering and exhausting CTLs coming from the tumor 17 , it would be interesting to determine this possible side effect during antitumor treatments with Aurora A inhibitors.
Finally, considering the involvement of autoreactive CD8 + T cells in the generation of tissue damage on some autoimmune disorders, such as diabetes 18 , the results from this work unveil new applications for Aurora A inhibitors, such as the blockade of effector CTLs in order to prevent tissue damage and, thus, improve the outcome of these disorders.
Human peripheral blood mononuclear cells (PBMCs) were isolated from buffy coats, obtained from healthy donors, by separation on a Biocoll gradient (Biochrom) according to standard procedures. To generate polyclonal cytotoxic T cells, CD8 + T cells were purified using MojoSort purification kit (Biolegend) and activated with anti-CD3/CD28 coated plates (5 and 3 μg/ml respectively). After two days, media was supplemented with IL-2 (50 U/ml) every 2 days for a time period of 8 days. These studies were performed according to the principles of the Declaration of Helsinki and approved by the local Ethics Committee for Basic Research at the Hospital La Princesa (Madrid); informed consent was obtained from all human volunteers. T cells were obtained from single-cell suspensions of the spleen and lymph nodes from OTI mice. In order to generate CD8 + T lymphoblasts responsive to ovalbumin peptide (OVAp; SIINFEKL OVA 257-264 ), cells were incubated 48 h with OVAp (10 ng/ml) and then, IL2 (50 U/ml) was added for a period of 5 additional days. CD8 + T lymphoblasts were cultured in RPMI 1640 + GlutaMAX-I + 25 mM HEPES (Gibco-Invitrogen) supplemented with 10% foetal bovine serum, 50 IU/ml penicillin, 50 µg/ml streptomycin (Gibco) and β-Mercaptoethanol (50 μM, Sigma).

Mice. Male
In vitro cytotoxicity assay. For cytotoxicity assay, EL4 cell line was used as target cells in a coculture with mice CD8 + T cells. Half of the EL4 cells were incubated with 1 µM CFSE (20 min) and pulsed with 1 µM OVAp (2 h) while the other half were incubated with 0.1 µM CFSE (20 min) alone. Both CFSE-stained populations were mixed (1:1) and cocultured with different ratios of mice CD8 + T cells during 6 h at 37 °C in technical triplicates. CD8 + T cells were previously treated with vehicle (DMSO), MLN8237, Aurora A Inhibitor I or AZD1152-pretreated for 30 min. After the 6 h of incubation, cells were stained with Red 780-Ghost Dye (Tombo Biosciences) in order to exclude death cells, and anti-CD8. Alive EL4 cells were monitored by flow cytometry, comparing the percentages of pulsed vs unpulsed cells. The percentage of specific lysis was calculated as follows: (E) Dot plots and density plots showing the gating strategy to analyse IFNγ vs CD107a expression in human CD8 + lymphoblasts pretreated with vehicle or MLN8237 (10 µM) and activated for 6 h with anti-CD3/ CD28 coated plates in the presence of monensin. Quantification of the % of CD107a + cells as well as CD107a + /IFNγ + is shown below (n = 6 human samples, ANOVA t-test and Wilcoxon test, respectively). P-value: < 0.05*; < 0.01**; < 0.0001****. Mean ± s.d.   OTI mice were infected with rVACV-OVA and, after 4 days, their CD8 + T cells were transferred to preinfected WT C57BL/6 mice. WT mice were treated twice with the vehicle or MLN8237 and, after 2 days, euthanised for ear viral titration. (B) Ear viral load from vehicle (10% 2-hydroxypropyl-2-cyclodextrin/1% sodium bicarbonate) or MLN8237 (2 doses, 30 mg/kg by oral gavage) treated WT C57/BL6 mice, infected with rVACV-OVA for 3 days and adoptively transferred or not with OTI-CD8 + effector T cells. Ear viral load was quantified as the Log 10 of plaque forming units (p.f.u) per ear. (n = 6 samples for mice without OTI transference and 11 samples for adoptive transference, both ears were quantified for each mouse; one way ANOVA with Turkey´s multiple comparison test). P-value: < 0.05*; < 0.0001****. Mean ± s.d.

Vaccinia virus infection and virus titration.
Both ears from OTI mice were intradermally (i.d.) infected into the ear pinnae with 1.5 × 10 4 PFU of rVACV expressing full-length OVA (a gift from J. W. Yewdell and J. R. Bennink. NIH, Bethesda, MD). Four days after infection, CD8 + T cells were isolated from the auricular lymph nodes and transferred to recipient WT mice, previously i.d. infected with OVA-VACV (36 h). In parallel, either Aurora A inhibitor MLN8237 (30 mg/kg) or vehicle (10% 2-hydroxypropyl-2-cyclodextrin/1% sodium bicarbonate) were administered, via oral gavage, to the mice. MLN8237 or vehicle administration was repeated after 24 h of CD8 + cell transference. For virus titration, the infected ears were mechanically disaggregated in 1 ml of PBS. After 2 freeze-thaw cycles and sonication, serial dilutions of ears homogenates were added to CV-1 cells monolayers seeded in 24 well plates. Cristal violet was used to stain them after 24 h and the number of plaques was multiplied by the reciprocal of sample dilution and converted to log 10 of p.f.u/Ear 19 . statistical analysis. First, a Shapiro-Wilk normality test was applied to determine the application of the corresponding parametric or non-parametric tests. Friedman or ANOVA t-test was used for grouped analysis while for dependent samples a paired analysis was used; either paired t-test (parametric) or Wilcoxon test (non-parametric). Analysis was performed with GraphPad Prism.