Listeria monocytogenes-infected human monocytic derived dendritic cells activate Vγ9Vδ2 T cells independently of HMBPP production

Gamma-delta (γδ) T cells express T cell receptors (TCR) that are preconfigured to recognize signs of pathogen infection. In primates, γδ T cells expressing the Vγ9Vδ2 TCR innately recognize (E)-4-hydroxy-3-methyl-but- 2-enyl pyrophosphate (HMBPP), a product of the 2-C-methyl-D-erythritol 4- phosphate (MEP) pathway in bacteria that is presented in infected cells via interaction with members of the B7 family of costimulatory molecules butyrophilin (BTN) 3A1 and BTN2A1. In humans, Listeria monocytogenes (Lm) vaccine platforms have the potential to generate potent Vγ9Vδ2 T cell recognition. To evaluate the activation of Vγ9Vδ2 T cells by Lm-infected human monocyte-derived dendritic cells (Mo-DC) we engineered Lm strains that lack components of the MEP pathway. Direct infection of Mo-DC with these bacteria were unchanged in their ability to activate CD107a expression in Vγ9Vδ2 T cells despite an inability to synthesize HMBPP. Importantly, functional BTN3A1 was essential for this activation. Unexpectedly, we found that cytoplasmic entry of Lm into human dendritic cells resulted in upregulation of cholesterol metabolism in these cells, and the effect of pathway regulatory drugs suggest this occurs via increased synthesis of the alternative endogenous Vγ9Vδ2 ligand isoprenyl pyrophosphate (IPP) and/or its isomer dimethylallyl pyrophosphate (DMAPP). Thus, following direct infection, host pathways regulated by cytoplasmic entry of Lm can trigger Vγ9Vδ2 T cell recognition of infected cells without production of the unique bacterial ligand HMBPP.

p-Ag and stimulate Vγ9Vδ2 T cell responses, although this stimulation is less efficient than that obtained with HMBPP 6,9,[14][15][16] . Vγ9Vδ2-mediated p-Ag sensing requires cell-cell contact and depends on both Vγ and Vδ chains 17,18 . An essential prerequisite for p-Ag sensing is the expression of butyrophilin (BTN) 3A1 in the target cell 19 . Several studies provided evidence for p-Ag binding to the intracellular B30.2 domain of BTN3A1 and for a p-Ag induced conformational change of this protein 10,20,21 . More recently, butyrophilin 2A1 (BTN2A1) was identified as a key ligand that binds to the Vγ9 TCR γ-chain. BTN2A1 associates with BTN3A1 independently of p-Ag stimulation and act together to initiate responses to pAg 22,23 . Recent studies have implicated binding of a second ligand to a separate TCR domain incorporating Vδ2 22,23 , and additional family members regulate optimal presentation of phosphoantigens through BTN3A1 24 . Thus, the complete description of the p-Ag sensing by the γδ TCR is not yet fully understood.
Listeria monocytogenes (Lm) is a Gram-positive facultative intracellular pathogen. In mice, Lm infection generates strong Lm antigen-specific αβ T cell responses and the bacteria has been genetically manipulated to serve as a tumor antigen vaccine platform [25][26][27][28] , even in the face of host tolerance 29 . In mice, γδ T cells are mainly involved in the early phase of the immune response to Lm infection 30 , but due to multifaceted interactions with the host immune system Lm is a potent inducer of αβ CD8 T cell responses in mice 27 . While mice have proven to be an excellent model to study the dynamics of bacterial infection and host immune response, they lack both the critical butyrophilin family genes BTN3A1 and BTN2A1, and the Vγ9Vδ2 T cells capable of recognizing HMBPP or IPP. Thus, murine models may not faithfully recapitulate events in humans experiencing the same infection. In contrast to mouse infections, Lm infection in humans generates a specific expansion of Vγ9Vδ2 T cells 31,32 , which is of unclear significance to the use of Lm as an αβ T cells vaccine platform. In addition, there are variations in HMBPP production in Listeria strains 33 , which has the potential to impact listeriosis in humans.
Synthesis of HMBPP via the MEP pathway in Lm is dependent on the gene gcpE (also referred to as ispG) 34 . Cellular extracts obtained from Lm wild type (wt) strains can induce the expression of CD69 and CD25 on Vγ9Vδ2 T cells in vitro, whereas extracts obtained from a HMBPP-negative ΔgcpE strain failed to do so 35 . Similarly, supernatants from neutrophils infected with Listeria innocua expressing the gcpE gene (HMBPP overproducer) were able to activate Vγ9Vδ2 T cells coincubated with monocytes 36 . Notably the monocyte population and the presence of HMBPP were essential for this activation [36][37][38] . Furthermore, non-human primates that were intranasally infected with a wt strain showed Vγ9Vδ2 T cells expansion that was diminished when a ΔgcpE strain was tested [39][40][41] . These data strongly support a model where Lm infection generates HMBPP which results in BTN3A1-mediated activation of Vγ9Vδ2 T cells.
The fact that bacterial extracts can activate Vγ9Vδ2 T cells in a HMBPP and BTN3A1-dependent manner may not accurately model in vivo infection, where cellular entry is critical. In this work we aimed to explore the effects of Lm entry into human dendritic cells and the effect of a HMBPP-negative ΔgcpE strains on Vγ9Vδ2 T cell activation. Surprisingly, we found that Lm-infected human monocyte-derived dendritic cells (Mo-DC) cells are able to activate Vγ9Vδ2 cells independently of HMBPP production. We found that during the first hours of infection of Mo-DC there is a change in the transcription of genes involved in cholesterol biosynthesis and export (ABCA1 and ApoA1) that would result in an increase in the intracellular IPP pool. Changes in the levels of this metabolite during the infection affect the activation of the Vγ9Vδ2 T cells. In addition, the CH25H gene, a gene that usually is induced by Type I IFN in macrophages during Lm infection in mice, is not activated in Mo-DC leading to a lack of repression of the cholesterol synthesis. Importantly, these data demonstrate that during direct infection of human DC, Vγ9Vδ2 T cells can be activated independently of HMBPP production via regulation of host cholesterol biosynthesis.

HMBPP-independent activation of Vγ9Vδ2 T cells.
To study HMBPP-related Vγ9Vδ2 T cell activation we constructed a ΔgcpE version of the attenuated ΔactA strain that cannot synthesize HMBPP (Suppl Fig. 1A). ActA plays a major role in listerial virulence, and its absence leaves bacteria intracellularly immotile and essentially non-infectious. As a consequence, the ΔactA strain can escape the phagolysosome but is unable to infect surrounding cells. This attenuated strain has the same genetic background of several Lm-based cancer vaccine platforms. As a comparison, we also constructed a ΔlytB mutant strain that lacks the 4-Hydroxy-3methylbut-2-enyl diphosphate reductase enzyme that metabolizes HMBPP (also referred to as ispH), and so results in increased HMBPP production. All the mutations were confirmed by PCR (Suppl Fig. 1B). To confirm published results 34 , we evaluated the ability of bacterial lysates of the ΔactA, ΔactAΔgcpE, and ΔactAΔlytB strains to activate Vγ9Vδ2 T cells in vitro. As expected, ΔactA extracts resulted an increase in the activation marker  -cytometry plots and right panel,  quantitation of percent CD25 + δ2 + CD3 + cells are representative results for at least 4 biologically independent  experiments. (B,C,D) Mo-DC were infected with the indicated Lm strains at a MOI = 10, after 1 h cells were washed and ex-vivo expanded Vγ9Vδ2 T cells added. Cells were incubated for 18 h (IFNγ and CD25) or another 2 h in presence of monensin for CD107a staining. Cells were processed and stained for flow cytometry. Left panels, flow-cytometry plots and right panels, quantitation of percent IFNγ + δ2 + CD3 + , CD25 + δ2 + CD3 + or CD107a + δ2 + CD3 + cells are representative results of 4 to 6 independent experiments. Data represents the mean ± SEM of each group (n = 3-9). Statistics calculated by one-way ANOVA with Tukey's correction; * = p < 0.05, *** = p < 0.001, **** = p < 0.0001, ns = no significant differences observed between the groups analyzed. www.nature.com/scientificreports/ CD25 on Vγ9Vδ2 T cells, this was lost in extracts from the ΔactAΔgcpE strain,and increased in extracts from the ΔactAΔlytB (Fig. 1A). These data support the observation that HMBPP is the main factor in the activation of Vγ9Vδ2 T cells by Listeria cell extracts 34,35 .
To model Vγ9Vδ2 T cell activation during an active Lm infection, we directly infected human monocytederived DC (Mo-DC) at a range of MOI. Extracellular bacteria were removed by washing, then Vγ9Vδ2 T cells were added at a range of Mo-DC:T cell ratios. Activation of Vγ9Vδ2 T cells was determined by CD25 expression and IFNγ production using flow cytometry (Suppl Fig. 2). Surprisingly, we found that Vγ9Vδ2 T cells could be activated equally by ΔactA or the ΔactAΔgcpE strain (Fig. 1B), suggesting that in this setting Vγ9Vδ2 T cell activation was independent of HMBPP production. In support of this, the HMBPP overproducer ΔactAΔlytB strain also showed similar levels of Vγ9Vδ2 T cells activation to the ΔactA and the ΔactAΔgcpE strain (Fig. 1B). The response was independent of the MOI used in the experiments (Suppl Fig. 3A). To rule out the possibility that small quantities of HMBPP were still produced in the ΔactAΔgcpE strain, we constructed a double ΔactAΔgcpEΔlytB strain (Suppl Fig. 1A), which produced similar levels of activation to the control strain ( Fig. 1B,C,D). Finally, to characterize the immediate activation response of Vγ9Vδ2 T cells we analyzed degranulation by measuring surface expression of CD107a, which confirmed equivalent activation in the absence of HMBPP (Fig. 1D).
To assess Vγ9Vδ2 T cell function, we assessed cytotoxic activity against infected Mo-DC. We found that the cytotoxicity of Vγ9Vδ2 T cells against Lm-infected Mo-DC is also not affected by the phosphoantigen production (Suppl Fig. 3B). This activity of Vγ9Vδ2 T cells was dependent on cytoplasmic entry of Lm, since Mo-DC infected with the Δhly Lm strain, which are unable to escape from the phagolysosome, were poorly able to activate Vγ9Vδ2 T cells despite functional HMBPP production and Vγ9Vδ2 T cells had poor cytotoxicity towards Δhlyinfected Mo-DC ( Fig. 1 and Suppl Fig. 3B ii,iii). We confirmed these data using naïve Vγ9Vδ2 T cells instead of ex-vivo expanded activated Vγ9Vδ2 T cells (Suppl Fig. 4A). Importantly, comparable results were obtained with Mo-DC and Vγ9Vδ2 T cells expanded from at least 4 different healthy donors (Suppl Fig. 4B). These results demonstrate that Lm-infected Mo-DC can activate Vγ9Vδ2 T cells and these cells, in turn, can kill Lm-infected cells independently of HMBPP production. However, cytoplasmic invasion of Mo-DC by Lm was essential for maximal cytotoxicity activity and Vγ9Vδ2 T cell activation. These data suggest that cytoplasmic entry of Lm results in an alternative pathway of Vγ9Vδ2 T cell activation independent of HMBPP production.
Mechanism of γδ T cell recognition. BTN3A1 has been implicated in Vγ9Vδ2 T cell activation by phosphoantigen binding to its intracellular 30.1 domain 8,19 . To determine whether BTN3A1 was necessary for Vγ9Vδ2 T cells activation during Lm infection, we used the anti-BTN3A1 blocking monoclonal antibody (mAb) clone 103.2 that has been previously described to be able to inhibit BTN3A1-phosphoantigen induced Vγ9Vδ2 T cells activation 19 . Compared to IgG control-treated Lm-infected Mo-DC, 103.2 mAb treated Mo-DC cells infected with either HMBPP-positive or -negative Lm strains were unable to activate the CD107a expression in Vγ9Vδ2 T cell (Fig. 2 Ai). Control experiments where zoledronate-or risedronate-pulsed Mo-DC were incubated with Vγ9Vδ2 T cells in the presence or absence of the anti-BTN3A1 antibody showed similar effect in the CD107a expression in the T cells confirming the specific inhibition due to this antibody (Fig. 2 Aii). These results demonstrate that BTN3A1 is needed to fully activate these T cells when exposed to Lm-infected Mo-DC. They also suggest that while HMBPP is not essential to activate Vγ9Vδ2 T cells, BTN3A1 remains important and highlights a potential role of other BTN3A1-binding phosphoantigens in Vγ9Vδ2 T cell activation following direct infection.
We also analyzed the expression of the later activation markers CD25 and IFNγ in Vγ9Vδ2 T cells following BTN3A1 blockade. Vγ9Vδ2 T cell incubated with Lm-infected Mo-DC resulted in Vγ9Vδ2 T cell activation as measured by CD25 and IFNγ expression, which was significantly reduced by BTN3A1 antibody blockade. However, this inhibition was not as complete as that observed for CD107a in the short-term assay (Fig. 2B,C). While these studies confirm that expression of CD25 and IFNγ in Vγ9Vδ2 T cell exposed to Lm-infected Mo-DC is independent of the HMBPP, their expression in Vγ9Vδ2 T cells appears to be partially independent of phosphoantigen-BTN3A1 presentation. This BTN3A1-independent activation may correspond with cytokines induced in these cells following Lm-infection, as has previously been described 42,43 . Other co-stimulatory molecules such as NKG2D have been shown to activate Vγ9Vδ2 T cells 6,[44][45][46] . To evaluate the role of NKG2D, we evaluated the effect of an anti-NKG2D blocking antibody on Vγ9Vδ2 T cell activation by Lm-infected Mo-DC. Neither cytotoxicity towards Lm-infected Mo-DC, CD107a translocation, nor IFNγ production by these Vγ9Vδ2 T cells were affected by the anti-NKG2D antibody (Suppl Fig. 5A).
Together these results indicate that Lm-infected Mo-DC are able to activate the degranulation of Vγ9Vδ2 T cell through a phosphoantigen-BTN3A1 interaction with the Vγ9Vδ2 TCR independently of HMBPP expression, but later CD25 activation and IFNγ production is partially independent of BTN3A1 interactions.

Role of endogenous IPP synthesis.
While HMBPP is a potent bacterial-derived ligand of Vγ9Vδ2 T cells, endogenous IPP is also able to activate Vγ9Vδ2 T cells, but at a lower affinity 47 . Therefore, we explored whether changes in the internal IPP pool of Mo-DC cells is responsible for Vγ9Vδ2 T cell activation following infection. To evaluate the role of endogenous mevalonate pathways and IPP synthesis in Vγ9Vδ2 T cells activation by Lm-infected Mo-DC, we incubated Mo-DC cells with statins during infection with Lm. Infected cells were then washed and exposed to Vγ9Vδ2 T cells to evaluate their activation. We observed that statin treatment decreased IFNγ production and CD107a expression by Vγ9Vδ2 T cells exposed to Lm-infected Mo-DC ( Fig. 3A-C), suggesting a potential role for endogenous IPP production in Vγ9Vδ2 T cell activation following direct infection. In agreement with these data, zoledronate and risedronate treatments increased Vγ9Vδ2 T cell activation that was blunted in the presence of mevastatin while HMBPP activation was not affected by the pres-  were washed twice to remove extracellular bacteria and 100 μl of fresh media including 50 μg/ml gentamycin and either 10 μg/ml anti-BTN3A1 blocking antibody or 10 μg/ml IgG isotype were added. After 45 min of further incubation at 37C, purified ex-vivo expanded Vγ9Vδ2 T cells were added to each well. Cells were incubated another 2 h in presence of monensin for CD107a staining. i) Quantitation of percent CD107a + δ2 + CD3. ii) Mo-DC cells were pre-incubated with zoledronate (50 μM) or risedronate (50 μM), washed and further incubated with either anti-BTN3A1 blocking antibody or IgG isotype. Ex-vivo expanded Vγ9Vδ2 T cells were added to each well and CD107a quantitation performed as above. (B) Quantitation of percent IFNγ + δ2 + CD3 and C) percent CD25 δ2 + CD3. Cells were processed and stained for flow cytometry as described in "Methods" section. Graphs are representative results of 2 to 4 independent experiments. Data represents the mean ± SEM of each group (n = 3). Statistics calculated by one-way ANOVA with Tukey's correction; *** = p < 0.001, **** = p < 0.0001, ns = no significant differences observed between the groups analyzed. www.nature.com/scientificreports/ ence of the statin (Fig. 3D). Importantly, the Vγ9Vδ2 T cells activation by the ΔactAΔgcpE strain was also sensitive to statin treatment (Fig. 3A). The effect of statins did not appear to be due to effects on Lm infection of the Mo-DC, as has been recently described for murine bone marrow-derived macrophages 48 , since statin treatment did not significantly affect Lm proliferation in Mo-DC (Suppl Fig. 5C).
Recently, Castella et al. 49 demonstrated that zoledronic acid and statins regulate the expression of the ATPbinding cassette transporter A1 (ABCA1) and apolipoprotein A1 (ApoA1) encoding genes in Mo-DC. ABCA1 is a well-studied major cholesterol exporter and ApoA1 is a component of the high-density lipoprotein (HDL), which is the molecule that associates to the ABCA1 exporter and transports cholesterol and phospholipids into the bloodstream. These authors propose that increasing concentrations of intracellular IPP leads to an increase in the expression of ABCA1 and ApoA1 and as a consequence these two proteins interact with BTN3A1, export Zoledronate

(C,D) Uninfected and Lm-infected
Mo-DC were treated as described, washed, incubated with Vγ9Vδ2 T cells in the presence of an anti-CD107a antibody and monensin and percent CD107a + δ2 + CD3 + analyzed for flow cytometry. Graphs are representative results of 2 to 5 independent experiments. Data represents the mean ± SEM of each group (n = 3-6). Statistics calculated by one-way ANOVA with Tukey's correction; *** = p < 0.001, **** = p < 0.0001, ns = no significant differences observed between the groups analyzed. www.nature.com/scientificreports/ of IPP to the extracellular space, and Vγ9Vδ2 T cell activation. We identified that both ABCA1 and APOA1 genes are induced in the first 2-6 h of Mo-DC infection (Fig. 4A), and still occurs following infection with the ΔactAΔgcpE strain ( Fig. 4B). At these same time points we did not detect transcriptional regulation of BTN3A1 in infected Mo-DC (not shown). Importantly, we did not detect induction of the APOA1 gene and only a modest induction of the ABCA1 gene following infection with the Δhly strain, suggesting that cytosol invasion was essential to augment the expression of these two genes (Fig. 4B). To determine whether Lm induction of the ABCA1 gene and APOA1 was dependent on phosphoantigen production in the cells we blocked production with mevastatin in conjunction with infection and determined gene induction. Mevastatin treatment decreased transcription of ABCA1, and mevastatin limits, but does not completely prevent ABCA1 expression following infection with Lm (Fig. 4C). In Mo-DC, the expression of the APOA1 gene was not affected by the statins and induction of APOA1 following Lm infection was unaltered by statins (Fig. 4C). To perform a more extensive analysis of the genes involved in the mevalonate/cholesterol and isoprenoid metabolism in the host cell, we profiled these genes at early time points following Lm infection of Mo-DC. We identified an increase in the transcription of LSS (lanosterol synthase), HMGCS1 (3-Hydroxy-3-Methylglutaryl-CoA Synthase 1) and DHCR24 (24-Dehydrocholesterol Reductase) (Fig. 5A), suggesting increased activity of the mevalonate/cholesterol pathways. Similarly, we found that Lm-infected Mo-DC exported cholesterol at higher rate than uninfected cells, confirming that Lm infection generates a metabolic change in the infected cells that results in an increase of synthesis and export of cholesterol that has not been previously described (Fig. 5B,C). Lxr-α is a regulator of lipid homeostasis and inflammation that acts with Lxr-β and/or the retinoid X receptors (RXRs) to control a range of cellular processes 50 . Lxr-α is a sensor for elevated intracellular IPP, and its activation results in increased expression of ABCA1 and ApoA1 49 . We determined that the expression of the LXR-α gene is not affected by Lm infection in Mo-DC (Suppl Fig. 5D). To address the role of this transcriptional regulator in the expression of ABCA1 and APOA1 during Lm infection, Mo-DC cells were pre-incubated with T0901317, a synthetic Lxr-α agonist. We similarly found that the ABCA1 gene was highly upregulated in the presence of T0901317, but we did not find that the expression of APOA1 was affected by T0901317 (Fig. 6A). Since Lm infection was able to induce both genes, this suggests that Lm regulates APOA1 gene expression independent of Lxr-α activation. To confirm these data, we evaluated the expression of these two genes following infection in the presence of the synthetic Lxr-α antagonist GSK 2033. Lm infection in Mo-DC could not induce ABCA1 in the presence of GSK 2033 and its expression was decreased at baseline following Lxr-α inhibition (Fig. 6Bi). GSK 2033 did not affect the expression of APOA1 and Lm infection induced APOA1 even in the presence of the antagonist (Fig. 6Bii). These data are consistent with ABCA1 induction during Lm infection in Mo-DC being dependent on Lxr-α, but that APOA1 is regulated through a distinct mechanism. To determine whether regulation of Lxr-α altered Vγ9Vδ2 T cell activation, we treated Mo-DC with the Lxr-α synthetic agonist or antagonist before Lm-infection, then evaluated their ability to activate Vγ9Vδ2 T cells. In the absence of Lm infection, Lxr-α regulation did not result in Vγ9Vδ2 T cell activation, and the Lxr-α antagonist GSK 2033 only moderately reduced Vγ9Vδ2 T cell activation (Fig. 6C). These data demonstrate that Lm access to the cytosol of Mo-DC results in a metabolic shift towards the synthesis and export of cholesterol (and its intermediates) in part via regulation of LXR-α and in part through independent mechanisms. However, manipulation of LXR-α has limited impact on Vγ9Vδ2 T cell activation following infection.

Lm infection and cholesterol regulatory mechanisms.
In view of the increase in transcription of several genes involved in mevalonate/cholesterol biosynthesis and cholesterol export in Mo-DC during the first hours of Lm infection we aimed to understand their regulation. Sterol regulatory element-binding protein 2 (SREBP2) is a transcription factor that is primarily responsible for the activation of genes involved in cholesterol synthesis. These include genes that are induced during Lm infection, such as HMGCS1, LSS and DHCR24 51,52 . For this reason, SREBP2 is a potential regulator of Lm cholesterol pathway gene induction and therefore may regulate Vγ9Vδ2 T cell activation. As expected we found that the expression of LSS and HMGCS1 is increased by Lm infection (Fig. 7Ai,Aii), and that the use of two SREBP2 specific synthetic inhibitors, betulin and fatostatin A 53,54 , reduced the upregulation of these genes. These data suggest that SREBP2 activation is involved in sensing Lm infection. Both betulin and fatostatin A increase the expression of the ABCA1 exporter in Lm infected Mo-DC (Fig. 7Aiii), suggesting that SREBP2 represses ABCA1 expression during Lm infection. These data agree with those of Zeng et al., 55 showing that ABCA1 expression is downregulated by the SREBP2 transcriptional regulator in endothelial cells 55 . Conversely, betulin or fatostatin A blocks APOA1 upregulation during Lm infection of Mo-DC, suggesting that SREBP2 mediates Lxr-α-independent regulation of APOA1 following Lm infection (Fig. 7Aiv). Importantly, SREBP2 inhibition during infection reduced Vγ9Vδ2 T cell activation (Fig. 7Bi,Bii,Biii). Thus, SREBP2 is an important sensor detecting Lm infection of Mo-DC that regulates Vγ9Vδ2 T cell activation by infected cells. It was recently described that betulin could inhibit Lm invasion of bone marrow derived macrophages and LDH release after Lm infection 56 . However, we have not found any change in ability of Lm to infect Mo-DC cells in the concentrations and time points evaluated (Suppl Fig. 5C).
Entry of Lm into the cytosol also triggers innate sensors resulting in type I IFN transcription 57-60 . Devilder et al., 61 demonstrated that TLR3 or TLR4-activated Mo-DC were able to trigger IFN-γ production by Vγ9Vδ2 T cells through the Type I IFN production. To exclude IFN-β-related effects, we used a blocking antibody and demonstrated that IFNAR2 blockade had no effect on Vγ9Vδ2 T cell activation (Suppl Fig. 5B). Type I IFN signaling can regulate cholesterol synthesis via regulation of CH25H which codes for 25-hydroxycholesterol oxidase, an enzyme that catalyzes the synthesis of the oxysterol 25-Hydroxycholesterol (25-HC) 52,62 . 25-HC is a metabolic inhibitor of SREBP2 activity 63 and is an endogenous ligand for Lxr-α 64 . In Mo-DC we observed increased transcription of ifnβ following Lm-infection, as well as release of IFN-β in the supernatants (Fig. 8Ai,ii). Interestingly, infection of Mo-DC downregulated CH25H expression (Fig. 8B) www.nature.com/scientificreports/ coordinately regulated CH25H as has been reported in other cell types. We performed a similar experiment in human macrophages and found increased expression of CH25H following infection (Fig. 8B), in agreement with the published studies using Lm-infected murine macrophages 65,66 . These data suggest that human macrophages and human dendritic cells differently regulate CH25H following infection. To test the consequence of CH25H gene activity in Mo-DC we treated cells with 25-HC during infection. 25-HC treatment during Lm infection of Mo-DC reduced the expression of the HMGCS1 and LSS genes and Vγ9Vδ2 T cell activation (Fig. 8C,D). These results are consistent with this pathway acting as a negative regulator of Vγ9Vδ2 T cell activation but that it is not functional in Mo-DC during the initial steps of Lm infection. Together, these data implicate SREBP2 as a critical regulator activated following Lm infection that is necessary for metabolic changes in the infected cell and activation of Vγ9Vδ2 T cells that can occur in the absence of HMBPP production in the bacteria.

Discussion
In this study we show that in human dendritic cells, the entry of Lm into the cytosol drives a change in gene expression and a metabolic shift towards cholesterol synthesis and export. These changes have not been previously reported and result in increased recognition of infected cells by a pre-existing population of Vγ9Vδ2 T cells. This in turn results in dendritic cell death as well as locoregional cytokine production. Due to this activation of endogenous cholesterol metabolism in the infected cells, recognition by Vγ9Vδ2 T cells does not require HMBPP production from the infecting bacterium. Kistowska et al., previously demonstrated that human monocytes and immature DC infected with E. coli or Staphylococcus aureus, bacterium that synthesize or lack HMBPP respectively, were both able to activate Vγ9Vδ2 T cells 16 . In agreement with our observations, Vγ9Vδ2 T cell activation in the early stages of infection involving changes in the first steps of the mevalonate pathway through increased cellular protein levels and activity of the HMGCR enzyme that would lead to increasing levels of intracellular IPP. While we did not detect changes in transcription of the HMGCR gene, we did detect changes in the transcription of several genes involved in the mevalonate/cholesterol biosynthesis and export. In our model, where Lm-infected Mo-DC are put in close contact with Vγ9Vδ2 T cells and their activation and cytotoxicity were analyzed very early following infection, we demonstrate that the HMBPP metabolite is unnecessary for the activation of Vγ9Vδ2 T cells during an active infection. However, later activation of Vγ9Vδ2 T cells while still independent of HMBPP is only partially affected by BTN3A1 blockade. These data suggest that additional factors resulting from infection contribute to Vγ9Vδ2 T cell activation. Brucella-infected Mo-DC have been shown to induce a similarly late IFNγ production in Vγ9Vδ2 T cells, which in turn results in full maturation of the infected Mo-DC and IL-12 production 43 , and a similar mechanism has been described following S. aureus infection 67 . IL-12 released from S. aureus-infected Mo-DC contributes to rapid IFNγ production by Vγ9Vδ2 T cells 67 This effect requires close contact between the infected cells and Vγ9Vδ2 T cells and is γδTCR dependent, confirming that in this bacterium that does not produce HMBPP endogenous IPP could explain Vγ9Vδ2 T cell activation. IL-12 has been shown to activate Vγ9Vδ2 T cells in combination with TCR ligation 42 , and independently of TCR ligation as a recombinant cytokine 68 , and may explain late HMBPP-independent Vγ9Vδ2 T cell activation. In our experiments NKG2D, a well-known co-stimulator of Vγ9Vδ2 T cells, does not appear to contribute to the observed activation or cytotoxicity. These data agree with some related studies, for example Mycobacterium tuberculosis-infected Mo-DC were able to trigger Vγ9Vδ2 T cell proliferation and perforin synthesis that was not affected by an anti-NKG2D blocking antibody 69,70 . However, other studies demonstrate that Vγ9Vδ2 T cell-mediated lysis of M. tuberculosis-infected Mo-DC and Brucella-infected DC is partially affected by NKG2D blockade 44,45 . These conflicting data suggest that there may be specific model-dependent features that dictate whether NKG2D contributes to Vγ9Vδ2 T cell function.
Prior studies with Lm demonstrating the relevance of HMBPP used in vitro models where either bacterial extracts or supernatants of infected neutrophils were applied to PBMCs or monocytes 34,36-38 . These two types of experimental designs do not model the metabolic changes that occur in the host cell that could affect Vγ9Vδ2 T cell activation. In non-human primates HMBPP negative strains are less able to expand Vγ9Vδ2 T cells when infected via an intranasal route, but Vγ9Vδ2 T cell expansion still occurs 40,41 . It is interesting to speculate whether the route of infection and therefore the cells that are infected impacts the lytic release of metabolites and the dependence on HMBPP production by the bacterium. In mice, oral infection can generate a distinct pattern of immune activation to IV infection with the bacterium [71][72][73] .
Our studies suggest that the cell type that is studied dramatically impacts the metabolic response to Listeria infection. Previous studies in Lm-infected murine bone marrow-derived macrophages have shown a reduced expression of the ABCA1 gene as well as other Lxr-α responsive genes 74 . Our finding of the concerted induction of the ABCA1 gene together with that of the APOA1 gene in Mo-DC provides evidence for the differential regulation in the different cell types analyzed. Furthermore, in murine bone marrow-derived macrophages and  49 this metabolite could either act as activator of the Lxr-α transcriptional regulator and/or that IPP is being exported to the microenvironment through the ABCA1-ApoA1 complex. Alternatively, since we have confirmed that BTN3A1 expression in Mo-DC is essential for full activation of Vγ9Vδ2 T cells by Lm, it is conceivable that the intracellular IPP/DMAPP could directly bind the intracellular domain of this protein leading to the direct Vγ9Vδ2 T cell activation by Lm-infected Mo-DC. Notably, we determined that the expression of the APOA1 gene is also greatly increased during Lm infection. In contrast to results described previously using relatively undifferentiatied myeloid populations 49 , in our system using fully differentiated Mo-DC the APOA1 gene appears not to be under the control of the Lxr-α regulator. Our results obtained using SREBP2 inhibitors betulin and fatostatin A strongly suggest that SREBP2 upregulates the transcription of the APOA1 gene during Lm infection.
Remarkably, the observed changes in expression of the genes involved in cholesterol synthesis and export, as well as the consequent Vγ9Vδ2 T cells activation, depend on the bacterial escape from the phagolysosome and cytoplasmic replication. Lysteriolysin O (LLO), a cholesterol-dependent cytolysin encoded by the hly gene, allows bacterial escape from the phagolysosome to the cytosol where it can replicate without causing appreciable antigen presenting cell death 27 . Monk et al. 76 , previously showed that human monocytes infected with a hly deficient Lm strain were able to stimulate Vγ9Vδ2 T cell proliferation similarly to that of the wild type strain. Given the significant differences in bacterial lysis by macrophages versus dendritic cells, it is possible that cytoplasmic entry is not a critical component to generate a Vγ9Vδ2 T cell response where the bacteria are rapidly cleared.
Statins inhibit 3-hydroxy-3-methylglutaryl-coenzyme A reductase (HMGCR), the rate-limiting enzyme in the synthesis of mevalonate 77 , which is a necessary precursor for synthesis of IPP, cholesterol and isoprenoids. Bisphosphonates such as zoledronic acid and risedronate are able to increase the intracellular IPP pool by inhibiting the farnesyl pyrophosphate synthase (FPPS) in the mevalonate pathway 78,79 . It has previously been shown that antigen presenting cells that are preincubated with the bisphosphonate zoledronic acid activated Vγ9Vδ2 T cells due to increased intracellular IPP 78,[80][81][82] . By contrast, antigen presenting cells preincubated with the statin mevastatin limited Vγ9Vδ2 T cell activation due to decreased IPP. Importantly, HMBPP activation of Vγ9Vδ2 T cells has been shown to be resistant to statins 14 , and critically, the Lm HMGCR is weakly inhibited by the statins, requiring 1000-fold higher concentrations to observe effects 83 . These data strongly suggest that our results demonstrating decreased Vγ9Vδ2 T cell activation as a result of limiting the endogenous mevalonate pathway either by the use of statins or SREBP2 inhibitors during Lm infection, support HMBPP-independent function. While mevastatin was able to completely inhibit the CD107a expression in Vγ9Vδ2 T cells that occurred following www.nature.com/scientificreports/ risedronate or zoledronic acid treatment, mevastatin only partially reduced CD107a expression in Vγ9Vδ2 T cells incubated with Lm-infected Mo-DC. This may be due to Lm infection also inducing the expression of other genes involved in the mevalonate/cholesterol pathway that are not inhibited by mevastatin, which are not induced by bisphosphonates. Alternatively, bacterial IPP/DMAPP could contribute to this observed statin-independent activation. Further study of the host metabolic response to infection is necessary to characterize this hypothesis. Although we cannot rule out the possibility that other factors could be involved, our data suggest that the mevastatin-mediated reduction in Vγ9Vδ2 T cell activation is primarily a consequence of the expected changes in the intracellular IPP pool or that of its isomer DMAPP. These effects of blocking endogenous mevalonate pathway occur despite the presence of bacterial HMBPP synthesis through the alternative mevalonate pathway, indicating that even in bacteria where this pathway is fully functional, the endogenous mevalonate pathway is the dominant mechanism for Vγ9Vδ2 T cell activation in infected dendritic cells. Together, these data place host cell metabolic regulation at the center of the immune response to Lm infection, and given the central importance of Vγ9Vδ2 T cell responses in infected humans, the impact of pathway regulators such as statins becomes highly relevant. However, as discussed earlier gcpE-deleted Lm have shown decreased ability to activate Vγ9Vδ2 T cell responses in primate studies [39][40][41] , suggesting that HMBPP production is playing a role in the in vivo response to Lm vaccination. Given the complicated interactions that occur between

Methods
Bacterial strains and construction. Lm strains used for these studies are all derived from the attenuated ΔactA strain DP-L4029 84 . When indicated we used the Δhly DP-L4027 strain 84 . Strains were grown in brainheart infusion broth, washed and diluted in PBS before infecting Mo-DC or macrophages at the indicated MOI. All nonpolar gene deletions were generated by allelic exchange using the temperature-sensitive plasmid pKSV7 as described 85,86 . Sequences located up-and downstream of the corresponding genes were amplified with Q5 High Fidelity DNA polymerase (NEB, Ipswich, MA) using primers whose sequences are described in Supplementary Table 1. Amplified fragments were cloned in the pKSV7 vector and were introduced into TOP10 E. coli (Invitrogen, Carlsbad, CA). Identified constructs were confirmed by sequencing, plasmids were then transformed into SM10 E. coli and conjugated into Lm strains. Positive clones were identified by performing colony PCR with Q5 High Fidelity DNA polymerase (NEB, Ipswich, MA).   Vγ9Vδ2 T cell activation. Human Mo-DCs obtained as described above were plated in RPMI-HS in round bottom 96-wells plates (~ 4-6 × 10 4 cells), infected with the indicated Lm strain at a MOI = 10, or the indicated MOI, and plates centrifuged 4 min at 400 g. After 1 h of incubation at 37 °C, cells were washed twice to remove extracellular bacteria and 100 μl of fresh media including gentamycin 50 μg/ml was added. Purified γ9δ2 T cells that were either ZA-expanded or naïve were added to each well and incubated at 37 °C and 5% CO 2 . Ratio Mo-DC: Vγ9Vδ2 T cells was usually 1:1, otherwise the ratio used is indicated in the text or figure. Unless indicated in the text, media was removed after 18-20 h of incubation, new fresh media was added including brefeldin A and after 4 h cells were washed and stained as described in "Flow cytometry" section. When different compounds were tested in the activation of the γ9δ2 T cells, Mo-DCs were incubated overnight with the indicated compound, cells were infected with Lm or left uninfected, washed and Vγ9Vδ2 T cells added as described above. For NKG2D blocking assays, Vγ9Vδ2 T cells were pre-incubated with anti-NKG2D antibody (clone 1D11, Millipore Sigma, Burlington, MA). Similarly for IFNAR2 blocking of the Vγ9Vδ2 T cells activation, these cells were incubated with various concentrations of the anti-IFNAR2 antibody (clone MMHAR-2, Millipore Sigma, Burlington, MA).
For BTN3A1 blocking, human Mo-DCs obtained as described were plated in RPMI-HS in round bottom 96-wells plates (~ 4-6 × 10 4 cells), infected with the indicated Lm strain at a MOI = 10 and plates centrifuged 4 min at 400 g. After 1 h of incubation at 37 °C, cells were washed twice to remove extracellular bacteria and 100 μl of fresh media including 50 μg/ml gentamycin and either 10 μg/ml anti-BTN3A1 blocking antibody 19 (clone 103.2, Creative Biolabs, Shirley, NY) or 10 μg/ml IgG isotype were added. After 45 min of further incubation at 37C, purified Vγ9Vδ2 T cells were added to each well and incubated at 37 °C and 5% CO 2 . Ratio Mo-DC: Vγ9Vδ2 T cells was 1:1. Cells were incubated another 2 h in presence of monensin for CD107a staining. For IFNγ and CD25 staining media was removed after 18-20 h of incubation, new fresh media was added including brefeldin A and after 4 h cells were washed and stained as described in below.
When bacterial extracts were evaluated, Lm strains were grown in liquid BHI, harvested at an optical density (OD) 600 nm of 0.8, and sonicated in 1/10 volume phosphate-buffered saline, pH7.4. Low molecular weight (LMW) fractions were obtained using Centriprep 3 kDa filters (Millipore Sigma, Burlington, MA). PBMCs were plated  Mo-DC or human macrophages were plated in a 24-well plate (5 × 10 5 cells per well) and infected with the indicated Lm strains for 1 h, washed and suspended in RPMI-HS media including gentamycin 50 μg/ml. At various time points cells were harvested and RNA was purified using Qiazol (Qiagen, Valencia, CA) and Direct-zol RNA miniprep kit (Zymo Research, Irvine, CA). DNase-treated RNA was used as template for cDNA synthesis using SuperScript™ IV VILO™ Reverse Transcriptase (Invitrogen, Carlsbad, CA) and qRT-PCR was performed using iTaq™ Universal ® SYBR Green Supermix (Bio-Rad, Hercules, CA) and primers described in Supplementary  Table 1). Reactions were carried out and analyzed in a StepOnePlus™ Real-Time PCR system (Applied Biosystems, Foster City, CA). Fold change was expressed as 2 -ΔΔCt , where the internal control is the Gapdh gene and the control samples are uninfected cells.
Cholesterol quantification and cholesterol efflux. Cholesterol was quantified in Lm-infected or uninfected Mo-DC using the Choelsterol/Cholesteryl ester assay kit as described by the manufacturer (#ab65359 Abcam, Cambridge, MA). In general infected cells were processed after 6 h post-infection. Cholesterol efflux was determined using the Cholesterol efflux assay kit (ab196985, Abcam, Cambridge, MA) following manufacturer's instructions and the recommended positive control. were incubated for 18 h, processed and intracellular cytokine staining performed as described in "Methods" section for quantification of percent (i) IFNγ + δ2 + CD3 + , (ii) Granzyme B + δ2 + CD3 + and (iii) IFNγ + Granzyme B + δ2 + CD3 + . (iv) Mo-DC were incubated with Vγ9Vδ2 T cells in the presence of an anti-CD107a antibody and monensin and CD107a + δ2 + CD3 + quantified. Results shown are representative of at least 3 biologically independent experiments. Data represents the mean ± SEM of each group (n = 3). Statistics calculated by unpaired t test and one-way ANOVA with Tukey's correction; ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001, ns = no significant differences observed between the groups analyzed.