Prostaglandin E2 induction by cytosolic Listeria monocytogenes in phagocytes is necessary for optimal T-cell priming

Listeria monocytogenes is an intracellular bacterium that elicits robust CD8+ T-cell responses. Despite the ongoing development of L. monocytogenes-based platforms as cancer vaccines, our understanding of how L. monocytogenes drives robust CD8+ T-cell responses remains incomplete. One overarching hypothesis is that activation of cytosolic innate pathways is critical for immunity, as strains of L. monocytogenes that are unable to access the cytosol fail to elicit robust CD8+ T-cell responses and in fact inhibit optimal T-cell priming. Counterintuitively, however, activation of known cytosolic pathways, such as the inflammasome and type I IFN, lead to impaired immunity. Here, we describe a cytosol-dependent response that is critical for immunity to L. monocytogenes, namely production of prostaglandin E2 (PGE2) downstream of cyclooxygenase-2 (COX-2). Vacuole-constrained L. monocytogenes elicit reduced PGE2 production compared to wild-type strains in macrophages and dendritic cells ex vivo. In vivo, infection with wild-type L. monocytogenes leads to 10-fold increases in PGE2 production early during infection whereas vacuole-constrained strains fail to induce PGE2 over mock-immunized controls. Mice deficient in COX-2 specifically in Lyz2+ or CD11c+ cells produce less PGE2, suggesting these cell subsets contribute to PGE2 levels in vivo, while depletion of phagocytes with clodronate abolishes PGE2 production completely. Taken together, this work identifies the first known cytosol-dependent innate immune response critical for generating CD8+ T-cell responses to L. monocytogenes, suggesting that one reason cytosolic access is required to prime CD8+ T-cell responses may be due to induction of PGE2. Author summary L. monocytogenes is an intracellular bacterial pathogen that generates robust cell-mediated immune responses. Due to this robust induction, L. monocytogenes is used as both a model to understand how CD8+ T-cells are primed, as well as a platform for cancer immunotherapy vaccines. L. monocytogenes must enter the cytosol of an infected host cell to stimulate robust T-cell responses, however, which cytosolic innate pathway(s) contribute to T-cell priming remains unclear. Here, we define COX-2 dependent PGE2 production as the first cytosol-dependent innate immune response critical for immunity to L. monocytogenes. We found that ex vivo PGE2 production by macrophages and dendritic cells is partially dependent on cytosolic access, as vacuole-constrained strains of L. monocytogenes elicit reduced PGE2. In vivo, cytosolic access is essential for PGE2 production. L. monocytogenes elicits a 10-fold increase in PGE2 production, whereas strains of L. monocytogenes that cannot access the cytosol fail to elicit PGE2 compared to mock immunized mice. Furthermore, CD11c+ and Lyz2+ cells contribute to PGE2 production in vivo, as mice deficient in COX-2 in these cell subsets have impaired PGE2 production. Taken together, our work identifies the first known cytosol-dependent pathway that is critical for generating immunity to L. monocytogenes.


Introduction
Listeria monocytogenes is a Gram-positive, intracellular pathogen that elicits robust 9 could be capable of synthesizing PGE2 (Fig 1A-B). Given that PGE2 is necessary for optimal T-151 cell priming and that immunizing mice with a strain of L. monocytogenes that cannot access the 152 cytosol leads to reduced T-cell effector function[9,10], we hypothesized that impaired T-cell 153 responses to vacuole constrained bacteria may be due to reduced expression of PGE2-154 synthesizing enzymes and ultimately decreased production of PGE2. To test this hypothesis, we 155 infected BMDMs and BMDCs with a vacuole-constrained strain of L. monocytogenes (Dhly, a 156 mutant lacking the pore-forming protein LLO) and assessed expression of Ptges and Ptgs2 157 mRNA. Consistent with this hypothesis, infection with this strain led to reduced Ptgs2 158 expression in BMDMs and BMDCs, suggesting that cytosolic access is required for optimal 159 expression of Ptgs2 ( Fig 1A). Interestingly, infection with Dhly L. monocytogenes led to similar 160 levels of Ptges expression in both BMDMs and BMDCs (Fig 1B). Taken  Infection of BMDMs and BMDCs with wild-type L. monocytogenes led to expression of the 179 genes necessary for PGE2 production (Fig 1A-C). To assess whether these cells could utilize 180 these enzymes to produce PGE2, we assessed PGE2 production in culture supernatant by mass 181 spectrometry. Surprisingly, supernatant from both BMDMs and BMDCs had no detectable PGE2 182 compared to PBS-treated controls, both during infection with wild-type or Dhly L. 183 monocytogenes (Fig 1D). This suggests that either enzyme expression was not high enough to 184 induce detectable PGE2, or there may be additional post transcriptional modifications required 185 for enzyme activity. Analysis of PGE2 from BMDMs or BMDCs deficient in COX-2 had no 186 detectable PGE2, as expected ( Fig 1D). 187 188 Primed BMDMs and BMDMs produce PGE2 during cytosolic L. monocytogenes infection 189 The lack of PGE2 produced by BMDMs and BMDCs in response to L. monocytogenes infection 190 was surprising given the upregulation of Ptgs2 and Ptges transcript. Other innate pathways, such 191 as the inflammasome, require a priming step in order to induce optimal activation. We 192 hypothesized that macrophages may similarly require additional stimulation in order to produce 193 expression in both BMDMs and BMDCs (Fig 1A). Infection with wild-type L. monocytogenes 197 led to a significant increase in expression that was less robust in Dhly L. monocytogenes-infected 198 cells, similar to the effect seen in unprimed cells (Fig 1A). Ptges expression, alternatively, had a 199 larger increase in transcript expression during PAM-priming, both during wild-type and Dhly L. 200 monocytogenes infection of BMDMs and BMDCs ( Fig 1B). Furthermore, PAM treatment alone 201 induced expression of Ptges similar to that induced during infection in BMDMs (Fig 1B). Taken  As Ptgs2 expression was also dependent on cytosolic access in primed BMDMs and BMDCs, we 210 next assessed expression of COX-2 protein in primed cells infected with wild-type or Dhly L. 211 monocytogenes by western blot. Similar to unprimed cells, BMDMs had similar levels of COX-2 212 protein during infection with wild-type or Dhly L. monocytogenes ( Fig 1C). In BMDCs, 213 alternatively, COX-2 protein expression was reduced during infection with Dhly L. 214 monocytogenes compared to wild-type infection (Fig 1C), again suggesting that COX-2 protein 215 expression in BMDCs is potentiated by cytosolic access. 216

217
We hypothesized that priming BMDMs and BMDCs with PAM would stimulate the cells to 218 produce PGE2 during infection with wild-type L. monocytogenes. To test this hypothesis, we 12 assessed production of PGE2 in the supernatant of primed BMDMs and BMDCs by mass 220 spectrometry. BMDMs and BMDCs were treated overnight with PAM before infection with 221 wild-type and Dhly L. monocytogenes. Six hours post-infection, cell supernatant was assessed for significant increase in PGE2 production compared to PBS-treated controls (Fig 1D). Previous 224 data showed L. monocytogenes-stimulated PGE2 production in peritoneal macrophages [21,23]. 225 Our data suggest that priming BMDMs prior to infection induces the cells to behave more like 226 tissue resident macrophages in respect to PGE2 production. Furthermore, the ability of BMDMs 227 to produce PGE2 provides a tool to efficiently study PGE2 synthesis in macrophages during 228 infection. Importantly, maximal PGE2 production in primed BMDM and BMDCs was dependent 229 on cytosolic access, as infection with Dhly L. monocytogenes led to significantly reduced PGE2 230 levels ( Fig 1D). PAM-primed COX-2 deficient BMDMs and BMDCs again led to no PGE2 231 production, solidifying the necessity of COX-2 activity in PGE2 production ( Fig 1D). 232 233 Additionally, we also sought to understand whether PGE2 specifically was being induced, or if 234 there was a more broad increase eicosanoid production. To test the hypothesis that L. 235 monocytogenes induces production of other eicosanoids, we analyzed production of 236 prostaglandin D2 (PGD2), thromboxane B2 (TXB2), and leukotriene B4 (LTB4). However, we saw 237 no changes production of these eicosanoids by wild-type L. monocytogenes immunized controls (Fig 2A). To ensure that the reduced PGE2 production was not due to 249 differences in bacterial burdens, mice were infected at a dose of wild-type (10 5 bacteria) and 250 Dhly L. monocytogenes (10 7 bacteria) that led to comparable burdens ( Fig 2B). This shows that 251 the absence of PGE2 in Dhly L. monocytogenes-infected mice is not just due to reduced bacterial 252 burdens. Taken together, these data highlight that cytosolic access is necessary for in vivo 253 induction of PGE2. 254 255 CD11c + and Lyz2 + cells produce PGE2 during L. monocytogenes infection in vivo 256 Our data identified PGE2 production by macrophages and dendritic cells ex vivo ( Fig 1D). 257 Furthermore, previous groups have reported that macrophage and dendritic cell subsets are 258 heavily infected early during in vivo infection, a timepoint where we have previously detected 259 increases in splenic PGE2 [14,20]. From these data, we next hypothesized that macrophages 260 and/or dendritic cells were responsible for producing PGE2 in vivo that is necessary for optimal 261 T-cell priming. To test this hypothesis, we generated mice deficient in COX-2 selectively in 262 CD11c + cells or Lyz2 + cells using the cre/lox system. Mice containing loxP sites flanking the and inlB, that retains immunogenicity while making it safe for clinical use [24]. The vaccine 269 strain was used here to enable analysis of T-cell responses in floxed mice as discussed below and 270 induces similar levels of PGE2 [14]. Immunization of CD11c-cre and Lyz2-cre mice each showed 271 reduced levels of PGE2 production, leading to only 60% of the PGE2 induced during 272 immunization of control mice ( Fig 3A). However, deletion of COX-2 in either CD11c + or Lyz2 + 273 cells did not abrogate production to the level of mice globally deficient in mPGES-1 (mPGES-1 -274 /-) ( Fig 3A). This suggests that CD11c + and Lyz2 + cells each contribute to PGE2 production and 275 that deletion of COX-2 in either is not sufficient to completely prevent PGE2 production. We 276 also assessed PGE2 levels in mice deficient in COX-2 selectively in T-cells and observed no 277 reduction of PGE2 (S3 Fig). As T-cells are not known to be infected by L. monocytogenes, this is 278 consistent with our hypothesis suggesting PGE2 production specifically from infected cell 279 subsets. 280 281 PGE2 is critical for generating optimal T-cell responses in response to L. monocytogenes, as 282 immunization of mPGES-1-deficient mice or treatment of mice with a COX-2-specific 283 pharmacological inhibitor leads to impaired T-cell responses [14]. We next hypothesized that the 284 decreased PGE2 production in the CD11c-cre or Lyz2-cre mice would be sufficient to similarly 285 impair T-cell responses. To test this hypothesis, we immunized mice with 10 7 LADD L.

Deletion of COX-2 in both Lyz2 + and CD11c + cells further reduces splenic PGE2 levels 300
Our data showed that single deletions of COX-2 in CD11c + or Lyz2 + cells reduced PGE2, but not 301 to baseline values. We next hypothesized that PGE2 production by either of these subsets 302 individually was sufficient for T-cell priming and that to observe impaired T-cell responses we 303 would have to eliminate PGE2 production in both CD11c + and Lyz2 + cells. To do this, we 304 crossed the COX-2 fl/fl CD11c-cre and COX-2 fl/fl Lyz2-cre mice, leading to mice with a COX-2 305 deletion in both cell subsets (COX-2 fl/fl CD11c-cre Lyz2-cre). We assessed the ability of these 306 mice to produce PGE2 by mass spectrometry and found that PGE2 was further reduced, with 307 about 40% the amount PGE2 produced compared to immunized control mice ( Fig 4A). This 308 suggests that CD11c + and Lyz2 + cells produce the majority of PGE2 during immunization with L.
antigen-specific T-cells measured by B8R tetramer were also unchanged compared to wild-type 317 controls (S5 Fig). This suggest that even the small amount of PGE2 produced locally is sufficient 318 to drive T-cell responses. 319 320

Depletion of phagocytes eliminates PGE2 production in vivo 321
Our ex vivo data highlighted the capability of BMDMs and BMDCs to produce PGE2 in response 322 to cytosolic L. monocytogenes. However, deletion of COX-2 in Lyz2 + and CD11c + cells did not 323 completely abrogate PGE2 production in vivo. These data led us to hypothesize that other 324 phagocytic cell subsets not effectively targeted by these cre-drivers may be producing the 325 residual PGE2, such as marginal zone macrophages (MZMs), metallophilic macrophages, or 326 other CD11b + cells more broadly [26,27]. To test this hypothesis, we utilized short-term 327 clodronate liposomes to rapidly deplete phagocyte populations in the spleen. Mice were depleted 328 with clodronate liposomes 24 hours prior to immunization with L. monocytogenes [28]. Twelve 329 hours post immunization, spleens were harvested and assessed for PGE2 by mass spectrometry. 330 Additionally, splenocytes were assessed for CD11b + and CD11c + populations by flow cytometry 331 to confirm clodronate efficacy. Clodronate treatment led to significantly fewer CD11b + cells and mice ( Fig 5A). Importantly, bacterial burdens were equivalent between clodronate and mock-335 treated mice (Fig 5B). Pretreatment with a control empty liposome, encapsome, actually 336 increased PGE2 production compared to infected control mice, potentially due to increased 337 bacterial burdens (Fig 5A-B). Taken together, these data demonstrate that phagocytic cell 338 populations are critical for PGE2 production in vivo following L. monocytogenes immunization. 339 340 Loss of antigen presenting cells through clodronate treatment leads to impaired CD8 + T-cell 341 activation, making analysis of T-cell responses in this model not informative [29,30]. Given this, 342 we alternatively assessed the possibility that other phagocytic cells targeted by clodronate, but 343 not the Lyz2-cre, could contribute to PGE2 production. Complete elimination of PGE2 production 344 with clodronate treatment suggested that the residual PGE2 in the CD11c-cre Lyz2-cre mice was 345 due to a phagocytic cell that was not effectively targeted in these mice. Previous data showed 346 that though the Lyz2-cre used in this study is highly efficient at deletion of loxP flanked genes in 347 some macrophage subsets, it is only minimally successful at deleting genes of interest in other 348 subsets, such as MZMs [26]. MZMs, characterized by expression of MARCO, are heavily 349 infected early in L. monocytogenes infection [20]. We hypothesized that the residual PGE2 we 350 detected in our double CD11c-cre and Lyz2-cre mice may be due to inefficient deletion in 351 macrophage subsets such as these. To assess the role of MZMs in PGE2 production, we assessed 352 expression of COX-2 by immunohistochemistry. Mice were immunized with 10 7 vaccine strain 353 of L. monocytogenes and spleens were harvested three and ten hours later. Spleen cryosections 354 were then stained for L. monocytogenes, COX-2, and MARCO. Uninfected mice had COX-2 355 staining in the periarteriolar lymphoid sheath (PALS) with little expression in the marginal zone 356 (MZ) (Fig 5C,-D). As early as three hours post-immunization COX-2 staining was observed in the MZ, with approximately 50% of COX-2 colocalizing with MARCO + cells (Fig 5C-D). 358 Expression of COX-2 in the MZ was maintained at 10hpi, again showing approximately 50% 359 colocalization with MARCO (Fig 5C-D). Furthermore, L. monocytogenes colocalized with 360 COX-2 and MARCO expressing cells, suggesting that infected MZMs may be producing PGE2 361 ( Fig 5D). Expression of COX-2 suggests that MZMs, or other non-CD11c/Lyz2 expressing 362 phagocytes within the marginal zone, could be capable of producing PGE2 in vivo and may be 363 contributing to the PGE2 remaining in the CD11c-cre Lyz2-cre mice. Taken together, our data 364 suggest that multiple myeloid derived subsets can contribute to PGE2 production, including 365 Lyz2 + cells, CD11c + cells, and possibly MZMs. Complete reductions in PGE2 by depletion of 366 phagocytic cells such as these with clodronate treatment is consistent with our data showing that 367 PGE2 is produced from cells infected with cytosolic L. monocytogenes. 368 access is important is to facilitate phagocyte production of PGE2, an eicosanoid required to 374 generate optimal CD8 + T-cell responses [14]. We showed that PGE2 is produced by BMDMs and 375 BMDCs ex vivo. Importantly, this pathway requires cytosolic access, as vacuole-constrained L. 376 monocytogenes induce lower production of PGE2. Furthermore, infection of mice with a vacuole-377 constrained L. monocytogenes strain led to no increase of PGE2 over mock immunized controls. 378 Lastly, we showed that Lyz2 + and CD11c + cells contribute to PGE2 production in vivo as deletion 379 of COX-2 in these subsets led to decreased PGE2 levels, however other clodronate sensitive 380 phagocyte populations also contribute to PGE2 production following L. monocytogenes 381 immunization. We demonstrate the first-known innate pathway critical for CD8 + T-cell responses 382 that requires cytosolic access by L. monocytogenes. This work leads to many new questions 383 including how cytosolic L. monocytogenes activates this pathway, how immune cells 384 discriminate which eicosanoid to produce in response to infection, how even small 385 concentrations of PGE2 still lead to productive PGE2 responses, and how PGE2 facilitates 386 optimal T-cell priming. 387 One intriguing hypothesis is that PGE2 synthesis during L. monocytogenes infection is 388 driven by an innate cytosolic sensor. L. monocytogenes elicits a number of innate pathways that 389 could contribute to differential activation of the PGE2-synthesis pathway. One possibility is that 390 induction of type I IFN influences PGE2 production. Type I IFN can be induced cytosolically by 391 interferon induction [34,35], and was originally hypothesized to be critical for T-cell responses. 396 Paradoxically, however, type I IFN inhibits cell-mediated immunity to L. monocytogenes [12]. 397 Interestingly, there has been well documented crosstalk between the PGE2 and type I IFN 398 pathways during infections with other pathogens such as influenza and M. tuberculosis [36,37]. It is also possible that production of PGE2 by L. monocytogenes is independent of known 433 cytosolic pathways. Identification of other unknown censors could be accomplished by assessing involved. One additional hypothesis is that PGE2-production is independent of a cytosolic sensor 437 completely and instead is driven by LLO-mediated pore formation. Though  proinflammatory functions suggest that PGE2 may be acting to enhance immunity through its 505 local effects on innate immune cells. PGE2 may also be influencing immunity more directly on 506 T-cell subsets, such as through polarization of T-cells towards a Th1 phenotype [66]. 507 Additionally, PGE2 leads to higher expression of OX-40L, OX-40, and CD70 directly on T-cells, 508 promoting T-cell interactions and sustaining immune responses [59]. In order to more fully 509 understand how PGE2 facilitates T-cell responses to L. monocytogenes, a comprehensive analysis 510 of these effects on both T-cells and innate immune cells is required. 511 We and others have shown that innate immune responses substantially influence cell-512 mediated immune responses, particularly the inflammatory milieu induced during infection. 513 Here, we present evidence that one pathway critical for immunity, induction of PGE2, is 514 dependent on access to the cytosol. Furthermore, we show that PGE2 is produced by 515 Mouse strains: Six-to eight-week-old C57BL/6 male and female mice were obtained from the 535 NCI and Charles River NCI facility. Ptgs2 -/-(COX-2 -/-) mice were obtained from Jackson 536 Laboratory and maintained as heterozygote breeding pairs. Ptges -/-(mPGES1 -/-) mice lacking 537 microsomal PGE synthase have been previously described [67][68][69]. In order to generate cell-type 538 specific COX-2 knockout mice, COX-2 fl/fl mice (stock number 030785) were obtained from 539 Jackson Laboratory and crossed with Lyz2-cre (stock number 004781), CD11c-cre (stock 540 number 008068), or CD4-cre expressing mice (stock number 022071), all also obtained from crossing COX-2 fl/fl Lyz2-cre mice with COX-2 fl/fl CD11c-cre mice. Genotypes were confirmed 543 by PCR using the primer pairs in Table 1. 544 Table 1 Table 2. Data was analyzed using Excel and all RNA abundances 569 were calculated by using a standard curve of synthesized template (Integrated DNA  570 Technologies, G-Blocks) and are normalized to ActB (β-actin). 571 Table 2 Cayman Chemical) and anti-β-actin loading control (1:1000, ThermoFisher) in a 5% bovine 581 serum albumin solution. The following day samples were washed with PBS-T before being 582 incubated with secondary antibodies (anti-rabbit 800 at 1:10,000, anti-mouse 680 at 1:5,000). 583 Samples were imaged on a LiCor imager and analyzed via ImageStudio. Sample signal was 584 normalized to β-actin and relative abundance was compared to wild-type L. monocytogenes. Samples then were incubated at 4C for 30 minutes. Next, cellular debris was removed by 606 centrifugation and samples were concentrated to 1mL volume before being acidified with pH 3.5 607 water and loaded onto conditioned solid phase C18 cartridges. Samples were washed with 608 hexanes before eluting using methyl formate followed by methanol. Samples were concentrated 609 using a steady stream of nitrogen gas and suspended into 55:45:0.1 MeOH:H2O:acetic acid and 610 analyzed on an HPLC coupled to a mass spectrometer (Q Exactive; Thermo Scientific) using a 611 C18 Acquity BEH column (100mm x 2.1 mm x 1.7µm) operated in negative ionization mode. immunofluorescence microscopy as described previously [20]. Uninfected mice were used as 636 negative controls. Briefly, 5μm spleen cryosections were cut using a Leica CM1850 cryostat, 637 mounted on Superfrost Plus microscope slides (Thermo Fisher) and stored at -80 o C until 638 use. Slides were fixed in 10% buffered formalin phosphate at RT for 5 minutes and sections, 639 washed in TBS and blocked with StartingBlock T20 Blocking buffer containing Fc blocker