Group 1 CD1-restricted T cells contribute to control of systemic Staphylococcus aureus infection

Staphylococcus aureus (SA) is the causative agent of both skin/soft tissue infections as well as invasive bloodstream infections. Though vaccines have been developed to target both humoral and T cell-mediated immune responses against SA, they have largely failed due to lack of protective efficacy. Group 1 CD1-restricted T cells recognize lipid rather than peptide antigens. Previously found to recognize lipids derived from cell wall of Mycobacterium tuberculosis (Mtb), these cells were associated with protection against Mtb infection in humans. Using a transgenic mouse model expressing human group 1 CD1 molecules (hCD1Tg), we demonstrate that group 1 CD1-restricted T cells can recognize SA-derived lipids in both immunization and infection settings. Systemic infection of hCD1Tg mice showed that SA-specific group 1 CD1-restricted T cell response peaked at 10 days post-infection, and hCD1Tg mice displayed significantly decreased kidney pathology at this time point compared with WT control mice. Immunodominant SA lipid antigens recognized by group 1 CD1-restricted T cells were comprised mainly of cardiolipin and phosphatidyl glycerol, with little contribution from lysyl-phosphatidyl glycerol which is a unique bacterial lipid not present in mammals. Group 1 CD1-restricted T cell lines specific for SA lipids also conferred protection against SA infection in the kidney after adoptive transfer. They were further able to effectively control SA replication in vitro through direct antigen presentation by group 1 CD1-expressing BMDCs. Together, our data demonstrate a previously unknown role for group 1 CD1-restricted SA lipid-specific T cells in the control of systemic MRSA infection.

Introduction Staphylococcus aureus (SA) is one of the most frequently isolated pathogens associated with nosocomial infections [1]. SA infection commonly manifests in many different organ niches, including skin/soft tissue infections, respiratory tract infections, pneumonia, and endocarditis [2]. SA also causes invasive sepsis leading to septic shock and multiple organ failure [3][4][5][6]. Antibiotic treatment of SA is complicated by the prevalence of methicillin resistant SA (MRSA) strains in hospitals and increasingly in community settings [3] [7], resulting in patient mortality rates of 16.6% in 2017 in addition to elevated healthcare costs [8][9][10]. Multiple vaccine strategies have been employed with limited efficacy against recurrent infection [11][12][13].
As an extracellular pathogen, historically it was thought that the antibody response was critical to immunity against SA [14,15]. However, the contribution of T cells to anti-SA immunity is gradually becoming appreciated. One study identified a CD4 + T cell epitope from the SA heme-uptake protein IsdB able to induce proliferation and IFN-γ and IL-17 secretion from T cells [16]. Another group identified the epitope M8 derived from manganese transport protein C (MntC) to induce proliferation and IL-17 secretion from CD4 + T cells in immunized BALB/ c mice [17]. Studies in humans without documented prior infection with SA showed that healthy people possess a significant pool of SA Ag-specific CD4 + memory T cells [18], and mouse studies have linked expansion of IL-17-producing γδ T cells with protection against SA [19]. Additionally, patients with hyper-IgE syndrome who lack the ability for Th17 CD4 + T cell differentiation are highly susceptible to SA skin infections [20], and HIV + patients are more susceptible to MRSA skin infections possibly due to decreased IFN-γ production from SA Ag-specific CD4 + T cells [21]. Intriguingly, recent studies have found a clonal Vγ6 + Vδ4 + T cell subset secreting IL-17 to be crucial for protective immunity against SA skin infection in mice [22]. Despite data showing the importance of conventional CD4 + T cells and γδ T cells to defense against SA, the role of lipid antigen-specific group 1 CD1-restricted T cells remains unexplored.
CD1 molecules are structurally similar to MHC-I molecules and present self-and microbial-derived lipids to T cells [23]. Three major groups of CD1 have been identified in humans: group 1 CD1 (CD1a, -b, and -c), group 2 CD1 (CD1d), and group 3 CD1 (CD1e; aids in loading complex lipids onto CD1b molecules) [24,25]. CD1d is the restriction element for type I and type II NKT cells, innate-like T cells which have been shown to play a role in defense against various microbial infections [26][27][28]. NKT cells exhibit pre-activated phenotype and are rapidly mobilized to fight microbial infections [29]. Studies have demonstrated that type I NKT cells (or iNKT cells) can be activated by both antigen-specific and cytokine-driven mechanisms, for example through interaction of Salmonella Typhimurium-derived LPS with TLRs on antigen presenting cells, inducing production of IL-12 [30]. Additionally, activation of type II NKT cells has been shown to protect against SA-mediated sepsis in mouse models of infection [31]. However, though both group 1 and group 2 CD1-restricted T cells are activated by lipid antigens, it is unknown whether the kinetics and function of group 1 CD1-restricted T cell responses to SA mimic the innate-like responses seen with NKT cells. Though mice only express CD1d, our lab has developed a transgenic mouse model (hCD1Tg) expressing human group 1 and group 3 CD1 molecules in a tissue specific, physiologically relevant manner [32]. To date, studies have shown that group 1 CD1-restricted T cells can recognize and respond to mycobacterial lipid antigens during infection in hCD1Tg mice [32,33]. In addition, Mtb lipidspecific group 1 CD1-restricted T cells contribute to anti-Mtb immunity through increased IFN-γ cytokine secretion [23,33,34], and Mtb patients have higher numbers of these T cells in circulation [35]. While Group 1 CD1-restricted microbial antigens identified thus far have mainly been of mycobacterial origin [25,34,36], CD1 molecules may bind lipid antigens from a broad range of pathogens because their antigen binding grooves can accommodate a variety of lipid species [37,38]. Indeed, humans have CD1b-restricted T cells that recognize lipids from Staphylococcus aureus, Brucella melitensis, Salmonella Typhimurium, and Borrelia burgdorferi [39,40].
To determine how group 1 CD1-restricted T cell dynamics and function change in response to different infectious agents, we selected SA as a model pathogen. SA contains microbial lipid moieties [41] sharing several common features with other group 1 CD1-binding antigens. Indeed, one recent study showed that CD1b dextramers loaded with phosphatidylglycerol (PG) from SA can identify and cross-react with autoreactive group 1 CD1-restricted T cells in humans [39]. Based on these studies, it is important to identify whether group 1 CD1-restricted T cells play a similar role in anti-SA immunity in vivo.
Using in vivo models of systemic SA infection in hCD1Tg mice, we found that group 1 CD1-restricted T cell responses against SA lipids can be detected in both immunization and infection settings. The kinetics of group 1 CD1-restricted T cell activation more closely resembled conventional T cell rather than NKT cell activation patterns, and hCD1Tg mice trended towards decreased kidney CFU burden and pathology compared with Tg-WT littermate controls. Fractionation of SA lipids revealed cardiolipins and polar phospholipids to be immunodominant lipid moeities recognized by group 1 CD1-restricted T cells. Furthermore, T cell lines specific for these lipids were protective against SA infection in hCD1Tg mice. SA killing experiments with group 1 CD1-restricted T cell lines specific for SA lipids also resulted in efficient CFU control in vitro. These findings demonstrate that group 1 CD1-restricted T cells can respond to SA lipids and that they contribute to protection against SA infection.

Immunization of hCD1Tg mice with SA lipids activates group 1 CD1-restricted T cells
We have previously shown that both infection with Mtb and immunization with Mtb lipids elicit group 1 CD1-restricted T cell responses in hCD1Tg mice. However, it was unknown whether group 1 CD1-restricted T cells could respond to SA lipids in hCD1Tg mice. As a proof of principle, we immunized hCD1Tg mice intraperitoneally with BMDCs expressing CD1b and CD1c (78 + Tg, or Tg + BMDCs) coated with total SA lipid extract from USA300 LAC. At 7 days post immunization, lymphocytes from peripheral lymph nodes of immunized mice were stimulated ex vivo with Tgor Tg + BMDC pulsed or unpulsed with SA lipids, and IFN-γ-producing cells were quantified by ELISPOT assays. A significantly higher number of IFN-γ-producing cells were detected when stimulated with SA lipid-pulsed Tg + DC compared to stimulation with SA lipid-pulsed Tg -DCs or unpulsed DCs. This data indicate that SA lipid immunization can induce group 1 CD1-restricted SA lipid-specific T cell responses in hCD1Tg mice (Fig 1). Further, addition of CD1b and CD1c blocking antibodies showed that the majority of the group 1 CD1 response to SA lipids was CD1b-restricted, with a moderate contribution from CD1c-resricted T cells. Naïve, unimmunized mice also displayed a group 1 CD1-restricted T cell response to SA lipids which was also predominantly CD1b-restricted, though the magnitude of the response was much lower than what was seen in immunized mice. This suggested that a portion of the CD1b-restricted T cell response to SA lipid was due to activation of cross-reactive CD1b-autoreactive T cells.

Group 1 CD1-restricted T cell responses directed against SA lipids can be detected during systemic SA infection in hCD1Tg mice
Based on data showing that immunization with SA lipids induces group 1 CD1-restricted T cell responses, we infected hCD1Tg mice with SA to determine if infection induced similar SA lipid-specific responses. hCD1Tg mice expressing CD1b and CD1c molecules (78 + Tg mice) were infected with MRSA strain USA300 LAC, an isolate that frequently causes communityacquired skin and soft-tissue infections as well as systemic infections [42]. Mice were injected with 5x10 6 CFU via tail vein and euthanized at 10 days post-infection. Lymphocytes were isolated from cervical, axillary, inguinal, renal, and mesenteric lymph nodes and cultured together with Tgor Tg + BMDCs pulsed or unpulsed with SA lipids. This experiment was also conducted in hCD1Tg mice expressing CD1a, -b, and -c (64 + Tg mice) and we obtained a similar magnitude of response, though there was some contribution of autoreactive CD1arestricted T cells that was not observed in 78 + Tg mice (S1 Fig). However, 78 + Tg mice were used in subsequent experiments due to their ability to breed more effectively. Group 1 CD1-restricted T cells produced IL-17 and IFN-γ in response to total SA lipids in ELISPOT assays (Fig 2A). In order to confirm these responses to be SA lipid-specific and independent of non-specific superantigen-mediated T cell activation, the experiment was repeated in hCD1Tg MHC-II -/mice ( Fig 2B), with similar results. We also demonstrated that the group 1 CD1-restricted T cell response to SA lipids were not affected by the presence of NKT cells, as hCD1Tg CD1d -/mice had similar levels of IFN-γ and IL-17A production as hCD1Tg WT mice ( Fig  2C). In contrast to the lymph nodes, lymphocytes isolated from the spleen did not show measurable group 1 CD1-restricted SA lipid-specific responses, possibly due to the expansion of myeloid-derived suppressor cells that can limit T cell activation [43; data not shown]. We additionally determined whether SA lipid-specific group 1 CD1-restricted T cell responses were present in the kidney because it is the primary reservoir for SA during systemic infection [44]. We were unable to reliably detect these responses in the kidney at 10 days post-infection (S2 Fig), and therefore the remainder of the data focuses on group 1 CD1-restricted T cell responses in the lymph nodes. Together, these data show that group 1 CD1-restricted T cells can recognize and respond to SA lipids during systemic infection with SA.

Group 1 CD1-restricted T cell response to SA peaks at 10 days postinfection
To determine the kinetics of the group 1 CD1-restricted T cell response during SA infection, hCD1Tg mice were infected i.v. with SA and sacrificed at 7, 10, and 21 days post-infection. The group 1 CD1-restricted T cell response to SA lipids peaked at 10 days, and had waned by 21 days post-infection ( Fig 3A). This mirrored the activation pattern of conventional T cells, as reflected by the upregulation of CD69 expression on CD4 + and CD8 + T cells (S3A Fig). However, the peak of the group 1 CD1 response was not correlated with significant upregulation of CD1b or CD1c on APCs (e.g. B cells, macrophages, and dendritic cells) in lymph nodes of infected mice at 10dpi (S3B and S3C Fig). As the phenotypic characteristics of group 1 CD1-restricted T cells induced during SA infection are unknown, we further determined their co-receptor usage in SA-infected hCD1Tg mice. Lymphocytes from lymph nodes of infected mice were stimulated with Tgor Tg + BMDCs pulsed with SA lipid antigen and intracellular cytokine staining for IL-17A was performed to identify SA lipid-specific group 1 CD1-restricted T cells (Fig 3B). We found that most of group 1 CD1-restricted SA lipid-specific T cells express either CD4 or CD8 co-receptors ( Fig 3B, bottom right).
We next investigated whether the presence of group 1 CD1-restricted T cells conferred protection against systemic SA infection. Interestingly, there was a trend toward decreased CFU in kidneys of hCD1Tg mice compared with Tglittermate controls at 10 days post-infection ( Fig 3C), which correlated with the peak of the group 1 CD1-restricted response to SA lipids. In addition, hCD1Tg mice exhibited significantly decreased kidney pathology compared with Tglittermate controls. Hematoxylin and eosin staining performed on infected hCD1Tg and Tgkidneys at 10 days post-infection showed that Tgkidneys displayed significantly larger areas of abscesses and inflammation ( Fig 3D). Immunofluorescence on whole kidney sections further showed that abscesses were more disorganized in Tg -(WT) mice compared with hCD1Tg mice at the same time point (Fig 3E). This trend toward enhanced disorganization in neutrophil recruitment and abscess formation occurred across most of the Tgkidneys examined (S4 Fig). In contrast, Tg + kidneys had small, defined abscesses and foci of inflammation which limited the total area of kidney damage.

Fractionation of SA lipids reveals immunodominant phospholipid components
Having shown that SA infection can elicit SA lipid-specific group 1 CD1-restricted T cell responses, we next sought to identify immunodominant SA lipid antigens recognized during SA infection. Total SA lipids were isolated from lysostaphin-treated USA300 LAC in order to with Tg + BMDCs coated with total SA lipid extract. Lymphocytes from pooled peripheral lymph nodes (cervical, axillary, inguinal, and mesenteric) were isolated at day 7 from immunized and naïve hCD1Tg mice. Lymphocytes were assayed for IFN-γ production via ELISPOT using Tgor Tg + BMDCs ± total SA lipid as stimulators. Anti-CD1b or -CD1c blocking antibody or isotype control was added as indicated. Data is representative of 2 independent experiments with n = 1 (naïve) or 5 (immunized) mice each plated in duplicate. �� p<0.01 using two-way ANOVA with Tukey's multiple comparisons posttest. https://doi.org/10.1371/journal.ppat.1008443.g001

PLOS PATHOGENS
Group 1 CD1-restricted T cells in SA infection eliminate peptidoglycan contamination and limit purification to membrane-associated bacterial lipids. Lipids were fractionated using silica gel column chromatography and sequential elution of neutral lipids by chloroform, glycolipids by acetone, column re-equilibration in chloroform and by increasing methanol proportion in chloroform to elute more polar lipids. Lipids were separated into 8 discrete fractions (Fig 4A) containing neutral lipids, moderately polar lipids, and polar phospholipids.    Fig 4B). Interestingly, group 1 CD1-restricted T cells also recognized lipids from fraction 3 (Fr3; cardiolipin-enriched), but this fraction selectively induced IL-17 and not IFN-γ production in a group 1 CD1-dependent manner.
Mass spectrometry analysis showed that Fr8 contained a mixture of phosphatidyl glycerol (PG) and lysyl-phosphatidyl glycerol (lysyl-PG) species. We originally hypothesized that lysyl-PG may play a dominant role in activating group 1 CD1-restricted T cells because lysyl-PG is unique to bacteria [45] while PG species are found in bacterial membrane as well as mitochondrial membranes in mammalian cells [46]. To test this, we generated a fmtC::erm SA mutant that is capable of synthesizing PG but is unable to lysylate it to form lysyl-PG [45]. The absence of lysyl-PG species in total lipid extract from fmtC::erm was confirmed by mass spectrometry (S6 Fig). Lipids from the fmtC::erm mutant were fractionated ( Fig 4C) and tested in a IL-17 ELISPOT assay using lymphocytes from infected hCD1Tg mice at 10 days post-infection. Contrary to our hypothesis, group 1 CD1-restricted T cells retained their reactivity to Fr8 from the

PLOS PATHOGENS
Group 1 CD1-restricted T cells in SA infection fmtC::erm strain (Fig 4D), suggesting that PG but not lysyl-PG species are the immunodominant lipids they recognize. Interestingly, no group 1 CD1-restricted T cell responses were detected towards purified mammalian 16:0-18:1 cardiolipin or 18:0 PG (S7 Fig), suggesting that SA-derived cardiolipin and PG species uniquely activate these T cells during systemic SA infection.

Group 1 CD1-restricted T cell lines specific for SA lipids are polyfunctional and recognize multiple lipid species
In order to further probe the function of SA lipid-specific group 1 CD1-restricted T cells, we generated group 1 CD1-restricted SA lipid-specific T cell lines from lymphocytes of infected hCD1Tg mice by re-stimulating them weekly with Tg + BMDCs pulsed with total SA lipids. A panel of group 1 CD1-restricted T cell lines were generated and we chose to focus on 3 cell lines in particular due to their specificity and ability to produce multiple cytokines. All 3 T cell lines were reactive to SA lipids and could produce IFN-γ as well as either IL-17 or TNF-α ( Fig  5A). These responses were truly SA lipid-specific and Tg-dependent, as SA lipid pulsing did not affect BMDC maturation or IL-6 production differentially between Tgand Tg + BMDCs alone (S8 Fig). They also showed different immunodominance hierarchies in response to SA lipids ( Fig 5B). All 3 cell lines were poly-functional and could produce TNF-α, IFN-γ, and granzyme B in response to PMA and Ionomycin stimulation; cell line 5-7 was unique in that it also produced IL-17 ( Fig 5C). These group 1 CD1-restricted T cell lines were also poly-functional in response to SA lipids ( Fig 5D). Each T cell line was double negative (possibly due to loss of co-receptor after long term primary T cell culture) used different Vβ chains, and most were CD1b-restricted ( Fig 5E). Though all 3 cell lines used the Vα11 chain, sequence analysis demonstrated that J region rearrangements were different among them ( Fig 5E). Finally, cytokine production from each cell line was likely dependent upon TCR activation by group 1 CD1 molecules, as IFN-γ production was retained when T cells were simulated by APCs lacking the critical innate immune signaling adaptor protein MyD88 (S9 Fig).

Adoptive transfer of CD1b-restricted T cells specific for SA lipids protects against SA infection in hCD1Tg mice
Adoptive transfer of mycolic acid-specific CD1b-restricted T cells have been shown to confer protection against Mtb infection in hCD1Tg mice [33]. To test whether SA lipid-specific CD1b-restricted T cells were protective against SA infection, hCD1Tg mice and sex-matched Tglittermate controls were injected with either PBS or 2x10 6 cells of T cell line 51 via tail vein. Mice were infected with SA i.v. 24h later and euthanized at 4 days post-infection. hCD1Tg mice that received cell line 51 (Tg + AT) had significantly lower bacterial burden (~4 log) in the kidneys compared with either hCD1Tg mice that only received PBS (Tg + no AT) or Tgmice that had also received cell line 51 (Tg -AT) (Fig 6A). Consistent with this, the Tg + AT group had significantly fewer infiltrating immune cells in the kidney than the other groups (Fig 6B), possibly due to lower bacterial burdens in the organ. Significantly, the adoptively transferred cell line 51 (identified using the labeling reagent CellTrace) were found in greater quantity in the livers of hCD1Tg mice than in Tgrecipient mice (Fig 6C and 6D), suggesting the presence of group 1 CD1 molecules is crucial for their activation in vivo. In addition, conventional CD4 + T cells in the Tg + AT group produced significantly more IFN-γ than the other two groups, and CD8 T cells were similarly more activated in the lymph nodes than the other groups (Fig 6E and 6F). This data suggests that activation of CD1b-restricted SA lipid-specific T cells can lead to conventional T cell transactivation during SA infection.

PLOS PATHOGENS
Group 1 CD1-restricted T cells in SA infection

SA associates with group 1 CD1-expressing APCs in vivo and early endosomes subcellularly in vitro
Protective efficacy in vivo led us to question how group 1 CD1-restricted T cells could help to kill SA in vitro. To investigate this issue, we first determined which group 1 CD1-expressing antigen-presenting cells associated with SA in vivo. hCD1Tg mice were infected with USA300 SA-GFP i.v. Mice were sacrificed at 10 days post-infection and kidney lymphocytes and kidney-draining lymph nodes were isolated and processed to determine SA co-localization with APCs. Though SA-GFP mainly associated with neutrophils (here gated on CD11b + Ly-6G hi cells), these cells did not express high levels of CD1b or CD1c (Fig 7A and 7B). Rather, DCs and macrophages (gated here as CD11b hi CD11c hi or CD11b hi F4/80 + , respectively) comprised the majority of CD1b-and CD1c-expressing cells associated with SA-GFP ( Fig 7B). We further determined the subcellular localization of SA through immunofluorescence confocal microscopy. Tg + BMDCs were infected with SA and stained with antibodies against LAMP-1 (lysosomal marker), EEA-1 (marker for early endosomes), and SA. SA co-localized with LAMP-1 CD1-restricted SA lipid-specific T cell lines. Data representative of 2-3 independent experiments with n = 2 wells/condition. � p<0.05, �� p<0.01, ��� p<0.005 using 2-way ANOVA with Tukey's posttest. https://doi.org/10.1371/journal.ppat.1008443.g005

Group 1 CD1-restricted T cells in SA infection
and EEA-1 in areas close to the cell membrane (Fig 7C), consistent with published reports that SA can survive intracellularly in membrane-associated vacuoles of human monocyte-derived macrophages [47] and is not affected by endolysosomal acidification [48]. Loading of lipid antigens onto CD1b molecules also occurs in endolysosomal compartments [49], allowing for CD1b interaction with SA lipids after internalization of SA by the cell. Uninfected BMDCs did not show EEA1-LAMP1 colocalization (S10 Fig).

Group 1 CD1-restricted T cells can help control of SA growth in vitro
As SA primarily associated with group 1 CD1-expressing DCs and macrophages in vivo, we used these two cell types as APCs for in vitro SA infection. Tgand Tg + MHC-II -/-BMDCs and BMMs were infected with SA in vitro. APCs lacking MHC-II were used in order to eliminate potential complications generated from superantigen and MHC-II-mediated T cell activation, and gentamicin was added 30 minutes post-infection to kill extracellular bacteria. Despite similar infection efficiencies at time 0, at 18h post-infection only the co-culture comprised of infected Tg + BMDC and T cell line 5-7 was able to efficiently control SA replication by 10 4fold (Fig 7D, right). This demonstrated that direct Ag presentation by Tg + BMDC to cell line 5-7 was required to limit SA replication in vitro. In contrast, cross-presentation of SA lipids from infected Tgmacrophages incubated with Tg + BMDCs to cell line 5-7 was unable to control SA replication (Fig 7D, left, hatched bar). Similarly, cell line 51 was able to control SA replication through direct Tg + DC-T cell contact, though to a lesser extent than cell line 5-7 ( Fig  7E). Finally, antibody blockade of IFN-γ and TNF-α together or Granzyme B alone was able to reverse the protection conferred by T cell line 51 during in vitro SA infection (Fig 7F). Blocking IFN-γ or TNF-α individually did not affect the ability of group 1 CD1-restricted T cells to control SA replication in culture (S11 Fig), suggesting some redundancy of these cytokines in vitro. This demonstrates the mechanism by which group 1 CD1-restricted T cells may confer protection during SA infection.

Discussion
The findings herein are the first to demonstrate that group 1 CD1-restricted T cells are activated by SA lipids during in vitro and in vivo MRSA infection. The absence of group 2 CD1-restricted NKT cells did not affect the magnitude or kinetics of the group 1 response to systemic SA infection. Activation of the group 1 CD1-restricted T cell response was protective against SA-induced kidney pathology, and adoptive transfer of SA lipid-specific group 1 CD1-restricted T cells into infected hCD1Tg mice significantly decreased kidney SA burdens. We additionally identified multiple SA lipid species recognized by group 1 CD1-restricted T cells during systemic SA infection. Together, our study sheds light on a non-conventional lipid-reactive T cell subset that contributes to control of MRSA infection and may aid in future vaccine design to combat MRSA.
Naïve hCD1Tg mice immunized with SA lipid-pulsed BMDCs showed robust group 1 CD1-restricted T cell responses in both lymph nodes and spleen (Fig 1). Moving forward from this proof of principle experiment, primary systemic infection with SA similarly activated group 1 CD1-restricted T cells in the lymph nodes. Intriguingly, negating the effects of superantigens in SA by infecting hCD1Tg MHC-II -/mice did not affect the magnitude of the group 1 CD1-restricted SA lipid-specific response compared with hCD1Tg WT mice, suggesting that CD4 + conventional T cells do not significantly contribute to background IFN-γ or IL-17 secretion when exposed to SA lipids in vivo (Fig 2B). We also addressed whether group 2 CD1-restricted T cells, or NKT cells, affected the magnitude of the group 1 CD1-restricted T cell response to SA lipids because NKT cells can also respond to a variety of microbial lipids including phosphatidyl glycerols and galactosyl diacylglycerols from different bacterial species [27,50]. However, we found that the group 1 CD1-restricted T cell response to SA lipid was unaffected by the absence of NKT cells (Fig 2C). This is possibly because NKT cells are preactivated innate-like lymphocytes that function early during bacterial infection [51], whereas the 10 day peak of the group 1 CD1-restricted T cell response more resembles that of conventional T cells during primary SA infection (Fig 3A, S3 Fig).
Interestingly, hCD1Tg mice showed markedly decreased kidney inflammation and pathology compared with Tg -WT littermate control mice (Fig 3D), despite no significant differences in CFU (Fig 3C). Immunofluorescence staining showed that SA and neutrophilic infiltrates were generally contained to small circular areas in hCD1Tg kidneys, whereas large, disordered areas of neutrophils and SA were deposited in Tg -WT kidneys (Fig 3E and S4 Fig). Previous studies have demonstrated that IFN-γ/IL-17 -/mice develop spontaneous oral abscesses attributable to MRSA infection by 6-10 weeks of age due to dysregulated neutrophil and macrophage function [52]. SA-infected IFN-γ -/mice also displayed neutrophilic abscesses without defined borders at 3 days post-infection compared with WT mice that had clearly defined abscesses [53]. Thus, it is possible that group 1 CD1-restricted T cell production of IFN-γ and IL-17 in the lymph nodes in response to SA lipid can organize abscess formation downstream in the kidneys by affecting neutrophil recruitment and deposition.
Unbiased lipid fractionation and screening revealed that cardiolipins and polar phospholipids, specifically phosphatidyl glycerol (PG) species, are immunodominant SA lipids in the polyclonal infection setting (Fig 4B). SA lipid species, unlike the extensively characterized, relatively complex group 1 CD1 Mtb lipid antigens [54], are somewhat similar to mammalian lipids [41,45]. However, lysyl-PG is one SA lipid component that is unique to bacteria and enriched in the immunodominant fraction 8 [55]. We initially hypothesized that lysyl-PG was responsible for the majority of IL-17 and IFN-γ secretion from group 1 CD1-restricted T cells. In addition to being a uniquely bacterial lipid, lysyl-PG also enables antibiotic resistance and immune evasion [56,57], thus leading us to speculate that group 1 CD1-restricted T cells might aid in counteracting this mechanism in SA. Contrary to this hypothesis, when lipid antigens from the lysyl-PG-deficient fmtC::erm strain of SA were used, Fr8 retained the ability to activate group 1 CD1-restricted T cells (Fig 4D), implicating PG and not lysyl-PG species as immunodominant SA lipids. PG is known to be a CD1b-restricted T cell antigen; one previous study identified an autoreactive CD1b-restricted T cell clone from healthy human donors not known to be SA-infected that could cross-react with PG isolated from S. aureus and synthetic mammalian 18:0/18:1 PG [39]. This appears to differ from our results in the polyclonal setting in that hCD1Tg mice infected with SA displayed robust IL-17 secretion from group 1 CD1-restricted T cells in response to PG-rich Fr8 from SA but not to synthetic mammalian 18:0 PG (S6 Fig). Thus we hypothesize that a subset of SA PG-specific, non-autoreactive group 1 CD1-restricted T cells are induced during SA infection. SA-derived cardiolipins also activated group 1 CD1-restricted T cells in the polyclonal SA infection setting (Fig 4A). Though much larger than phosphatidyl glycerol moieties also presented by CD1b molecules [39], we suspect that CD1b has a large enough binding groove to accommodate cardiolipin species. Indeed, CD1b presents Mtb lipids with acyl chain lengths that are 80 carbons long [36]; thus we anticipate CD1b would also be able to bind and present cardiolipin. Intriguingly, the cardiolipinenriched Fr3 selectively induced IL-17 but not IFN-γ production from group 1 CD1-restricted T cells (Fig 4B). This may be attributable to the binding affinity of cardiolipin-CD1 engagement with the group 1 CD1-restricted TCR, as has been shown in CD4 + conventional T cells to promote T h 17 differentiation [58]. It is also possible that the group 1 CD1-independent IFN-γ production observed in Fr3-stimulated lymphocytes was mediated through TCR-independent, cytokine-driven mechanisms. The studies contained herein are the first to our knowledge to identify SA-derived cardiolipins as group 1 CD1-binding lipid antigens, though further study using cell-free target systems is needed to verify TCR dependence in these responses.
SA and gram-positive bacteria in general contain a diversity of PG species that elute differently according to the solvent gradient. It is interesting to note that not all PG-rich fractions of SA lipids induced group 1 CD1-dependent T cell secretion of cytokine (Fig 4B). A comprehensive LC-Q-TOF-MS analysis of SA lipids demonstrated that PG species of different chain lengths eluted at different times in HPLC [41], and a recent study on Enterococcus faecalis lipids showed similar findings [59]. Indeed, silica column fractionation of total SA lipids resulted in multiple fractions enriched with different PG species depending on the ratio of CHCl 3 : MeOH used (Fig 4A). It is possible that though PG species in SA contain the same polar headgroups, the differences in fatty acyl chain lengths between PG moieties may affect binding affinity to the group 1 CD1 antigenic groove and subsequent TCR engagement strength. Previous studies have shown that CD1b binds diverse sulfoglycolipids from Mtb with the same polar headgroup but differing acyl chain lengths, which results in altered TCR recognition and T cell activation [38]. A similar mechanism could be at play regarding PG moieties from SA. Additionally, it is known that variations in the branching patterns of alkyl chains can also affect lipid Ag binding onto CD1c and downstream T cell activation [60]. Indeed, SA contains an abundance of anteiso-and iso-branched fatty acyl chains associated with PG species in the bacterial cytoplasmic membrane [61] that are not present in mammals, which may uniquely activate group 1 CD1-restricted T cells. Further characterization by GC-MS or ESI-tandem mass spectrometry is needed to determine if these subtle structural differences exist between SA PG species, and whether they contribute to immunodominance hierarchies in the polyclonal infectious setting.
Recent studies have shown that T cell production of IL-17 and IFN-γ is essential for protection against SA in both mice and humans [16,18,19,53]. Despite this, vaccine trials focused on activating CD4 + conventional T cell responses specifically have largely failed in humans [13,62]. It is therefore important to examine whether activation of non-conventional T cells during SA infection can be protective. Adoptive transfer of the group 1 CD1-restricted SA lipid-specific T cell line CTL51 decreased SA kidney burdens by~10 4 fold at 4 days post-infection in hCD1Tg but not Tg -WT mice (Fig 6A). This effect was correlated with trans-activation of CD4 and CD8 T cells in the draining lymph nodes (Fig 6E and 6F), consistent with where the group 1 CD1-restricted T cell response occurred during primary SA infection (Fig 2). The magnitude of protection afforded by the adoptively transferred SA-lipid-specific group 1 CD1-restricted T cells suggests that it will be important to elucidate how best to activate group 1 CD1-restricted T cell responses during infection, as this could provide a robust, novel protective mechanism against systemic SA infection.
Though primarily considered an extracellular pathogen, studies have shown that SA can enter into macrophages, DCs, neutrophils, and even non-phagocytic cells [63]. Our data show that SA is primarily associated with neutrophils and macrophages, and to a lesser extend with DCs in the kidney (Fig 7A and 7B). While CD1b and CD1c are highly expressed on a subset of myeloid-derived dendritic cells, neutrophils and macrophages do not express CD1b and neutrophils express only low levels of CD1c. This provided the justification for the use of macrophages and DCs for in vitro infection experiments. The co-culture of group 1 CD1-restricted T cells with SA-infected hCD1Tgand Tg + APCs demonstrated a critical role for direct SA lipid presentation by group 1 CD1-expressing DCs to group 1 CD1-restricted T cells in controlling SA growth (Fig 7D). Indeed, no control of SA infection was observed when infected BMM were co-incubated with SA lipid-specific T cells with uninfected Tg + BMDC help, suggesting that cross-presentation of SA lipids to Tg + DCs is not effective in halting SA replication. In contrast, control of SA in culture was orders of magnitude higher when infected Tg + DCs directly presented lipids to T cell line 5-7 (Fig 7D, bottom panel). One possible explanation for this is that BMM differentiated in culture do not express CD1b [33], and our group 1 CD1-restricted T cell lines are generally CD1b-restricted (Fig 5E). Though cell line 5-7 is both CD1b and CD1c-restricted (Fig 5E), the binding affinity of CD1b loaded with SA lipid to the cell line 5-7 TCR may be higher than the CD1c-TCR binding affinity, allowing for differential cytokine producing capacities during in vitro SA infection. This is still purely speculative, as one limitation of our study was that we could only indirectly show TCR dependence for activation in our group 1 CD1-restricted cell lines (S9 Fig). Another possibility is that SA internalized by BMM can easily escape mature phagolysosomes and kill the cells [48], which would limit TCR engagement. Dendritic cell cross presentation of SA lipids from infected macrophages may also be less efficient at activating group 1 CD1-restricted T cells because the time spent internalizing apoptotic macrophages and processing/presenting SA lipids may take too long, allowing bacteria to overgrow in the culture. Additionally, BMDCs may be more resistant to killing by SA, as one study examining DC-S. aureus interactions in vitro showed efficient control of infection of BMDCs out to 18h post-infection [64]. This would allow more time for group 1 CD1-restricted T cell activation in culture.
Finally, blockade of either TNF-α and IFN-γ together or granzyme B alone was sufficient to limit the protective efficacy of group 1 CD1-restricted T cell line 51 (Fig 7F). T cell secretion of granzyme B protects against bacterial growth in a variety of infection models. Enhanced clearance of pulmonary MRSA was observed in mice with elevated granzyme B levels, a phenotype which was replicated in human neutrophil culture [65]. Though this phenotype depended on neutrophil secretion of granzyme B, it is likely that T cell production of granzyme B has similar effects. Granzyme B production from a CD4 + human T cell clone also protected against Mtb growth in macrophages in vitro [66], and resistance to Mtb infection in humans is correlated with elevated MAIT cell production of granzyme B [67]. Similarly, intracellular growth of Listeria monocytogenes (LM) within HeLa cells and monocyte-derived dendritic cells was inhibited by human LM-specific CTL in a granzyme B-dependent manner [68], suggesting that granzyme B is broadly effective in killing multiple species of intracellular bacteria. Similarly, studies have shown that both IFN-γ and TNF-α contribute to host resistance against systemic SA infection [69], and that γδ T cell production of both cytokines controlled SA replication in a mouse skin infection model [70]. Though blocking either IFN-γ or TNF-α individually in vitro did not affect SA replication compared to the Tg + DC-T cell condition (S11 Fig), blocking both simultaneously enhanced replication (Fig 7F). This suggests that IFN-γ and TNF-α production by group 1 CD1-restricted T cells in vitro play a redundant role in controlling SA replication.
Staphylococcus aureus has proven to be a stubborn pathogen to eradicate or develop vaccines against. This highlights the importance of elucidating novel pathways of T cell activation during SA infection to expand the parameters of vaccine design. The studies herein are the first to our knowledge to demonstrate that lipid-specific group 1 CD1-restricted T cells can be activated during systemic SA infection. Notably, these T cells primarily respond to SA-derived phospholipids and can control bacterial growth both in vitro and in vivo settings. These data suggest that incorporating SA lipid antigens into a multi-component subunit vaccine which targets both unconventional and conventional T cells may be effective in helping combat this complex public health crisis.

Ethics statement
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Animal Care and Use Committee of the Northwestern University (Protocol number: IS00001659).

Mouse strains
hCD1Tg mice (Line 78 and Line 64) in the B6 background were generated in our lab as previously described [32]. hCD1Tg mice were also crossed onto B6 MHC-II-deficient or CD1ddeficient background [71]. Most in vivo infection experiments used hCD1Tg (Line 78) mice expressing CD1b and CD1c but not CD1a due to better breeding performance. Proof of principle experiments used hCD1Tg (Line 64) mice expressing CD1a, CD1b and CD1c in a tissuespecific, physiologically relevant manner.

Bacterial strains, media, and culture conditions
Mouse intravenous infections with Staphylococcus aureus (SA) were carried out using the methicillin resistant USA300 strain LAC or LAC-GFP [72]. SA was grown shaking overnight in TSA broth (MP Biomedicals, Solon, OH) at 37˚C, 220 rpm. A marked fmtC::erm insertional disruption in strain LAC was generated by bacteriophage-mediated transduction using ϕ11 containing DNA from transposon mutant NE1360 of the Nebraska transposon mutant library [73]. SA lipids were isolated from either WT LAC or fmtC::erm LAC.

Intravenous SA infection and DC immunization of mice
S. aureus overnight cultures were diluted 1:100 and subsequently grown to mid-log phase (~3 hours) shaking at 37˚C, 220 rpm. Bacteria were washed in PBS and mice were infected with 3-5x10 6 CFU in sterile PBS via tail vein injection. At indicated times, lymphocytes isolated from various organs were harvested and analyzed by flow cytometry and ELISPOT assays. Kidney homogenates from infected mice were plated on tryptic soy agar (TSA) plates to determine viable CFU. For immunization, hCD1Tg MHC-II -/-DCs were pulsed overnight with total SA lipids (10 μg/mL). 10 6 BMDCs were injected intraperitoneally into mice and spleens and lymph nodes isolated for ELISPOT assays 7 days post-immunization.

Generation of group 1 CD1-restricted T cell lines
Lymphocytes isolated from SA-infected hCD1Tg mice were cultured for 4 days in RPMI-10 and subsequently in supplemented Eagle's minimum essential medium with 10U/mL recombinant mouse IL-2 (PeproTech, Rocky Hill, NJ) [77]. T cell lines were re-stimulated weekly

Adoptive transfer of group 1 CD1-restricted T cell lines
T cell line 51 was labeled with CellTrace Violet (Thermo Scientific) dye prior to injection into mice. 2x10 6 cells in 200 μL of PBS were injected into hCD1Tg or Tgsex-matched mice via tail vein 1 day prior to infection. Control hCD1Tg mice received PBS alone. All mice were infected with 2x10 6 CFU of SA USA300 via tail vein at day 0 and euthanized at 4 days post-infection. Livers, spleens, peripheral lymph nodes and kidneys were isolated for CFU enumeration or functional experiments. Lymphocytes from these organs were stimulated with phorbol myristate acetate and ionomycin (PMA-Iono) at a concentration of 20 ng/mL and 1μM, respectively, for 6 h. Brefeldin A (10 μg/mL) was added to the cells at 2 h post-stimulation to detect intracellular cytokine production.

In vitro infections
BMM (5x10 4 /well) and BMDC (10 5 cells/well) were plated in triplicate in 48 well plates and infected with SA USA300 at MOI = 0.2. After 30 min, the medium was replaced with DMEM containing 100μg/mL gentamicin to kill extracellular bacteria for 30 minutes. Cells were then washed twice in warm PBS and a portion of BMM and BMDC were lysed in the wells with 300 μL sterile H 2 O to plate intracellular bacteria for Time 0 CFU enumeration. The remainder of the cells received either media alone or had uninfected BMDCs (10 5 /well), BMMs (5x10 4 / well), or T cells (4x10 5 /well) added to the wells. Some experiments included antibody blockade of cytokine production; 1μg/well was added of each of the following: α-IFN-γ (cloneAN-18), α-TNF-α (clone 6B8), or α-Granzyme B (clone QA16A02) (all from BioLegend). For antibody blocking experiments, all conditions were plated in groups of 4. Cells were spun down to enable contact and incubated for 18 hours at 37˚C. Plates were then spun down at 2400 rpm x 15 minutes to pellet SA, supernatant removed, and cells lysed in the wells with 300 μL sterile H 2 O. Samples were serially diluted and spot-plated for CFU enumeration on TSA plates containing 5 μg/mL erythromycin.

Immunofluorescence on APCs
BMDCs and BMMs (10 6 cells) were plated onto glass coverslips coated with poly-L-lysine inside 12-well plates and allowed to adhere for 1 hour. Cells were left uninfected or were infected with SA at MOI = 5 for 45 minutes. Cells were then washed twice with PBS and resuspended in DMEM containing 100 μg/mL gentamicin to kill extracellular bacteria for 30 minutes at 37˚C. Cells were washed, fixed with 10% PBS-buffered formalin containing zinc, and stained with the following primary antibodies in PBS + 1% BSA: polyclonal rabbit anti-mouse EEA-1, rat anti-mouse LAMP1 (clone 1D4B), and mouse anti-Staphylococcus aureus (clone 704) from Abcam (Cambridge, MA). Alexa fluor-conjugated secondary antibodies were goat anti-mouse IgG H&L-AF488 (Abcam), goat anti-rat IgG H&L-AF568 and goat anti-rabbit IgG H&L-AF647 (both from ThermoFisher, Waltham, MA). Cells were incubated in Vectashield + DAPI and placed onto glass slides for imaging. Microscopy was performed using a confocal Nikon A1R (A) spectral microscope (Nikon, Melville, NY).

Immunofluorescence on paraffin-embedded sections
Immunofluorescence on kidneys was performed as previously described [78]. Briefly, paraffinembedded kidneys deparaffinized with xylene and rehydrated with graded ethanol washes.

Group 1 CD1-restricted T cells in SA infection
Antigen retrieval was performed using sodium citrate buffer in a pressure cooker. Endogenous peroxidases were quenched with 10% H 2 O 2 . Background Buster blocking solution (Innovex Biosciences, Richmond, CA) was applied to each sample prior to incubation in primary antibody. Slides were then incubated sequentially with rabbit polyclonal anti-SA (Abcam) and rat anti-mouse Ly-6G (clone 1A8) for 1 hour at room temperature or overnight at 4˚C. Slides were then incubated sequentially with goat anti-rat AF-555 (Invitrogen) or goat anti-rabbit AF488 (Abcam) for 1 hour at room temperature. Slides were then mounted using Vectashield Antifade Reagent with DAPI (Vector Labs). Slides were imaged at 25˚C, 20x magnification, using the Nikon A1 HD25 confocal microscope (Nikon Instruments Inc, Melville, NY). Images were analyzed with NIS Elements Viewer 4.20 software.

Statistical analysis
Statistical analyses were performed using Prism software (GraphPad Software, Inc., San Diego, CA). Significant differences at 95% confidence or above are depicted with asterisks on each graph.  Table showing percentage of PG species present in each PG-rich fraction classified according to acyl chain length. Cardiolipin-enriched Fraction 3 has the same chain length distribution as it is simply a dimer of PG species. (B) Lipid fractions were subjected to TLC separation using chloroform: methanol: acetone: acetic acid: water: toluene (70:30:5:4:1:10, v/v) as a solvent system. Phospholipids in each fraction (right panel) were visualized using phosphomolybdate reagent (blue spots) as described in Vaskovsky et al. [76]. Total lipids were visualized by "charring" or heating the same blot used to visualize phospholipids (left). Increasing concentrations of lysyl-PG is the main distinguishing characteristic between PG-enriched fractions 6, 7, and 8.