IKKβ in Myeloid Cells Controls the Host Response to Lethal and Sublethal Francisella tularensis LVS Infection

Background The NF-κB activating kinases, IKKα and IKKβ, are key regulators of inflammation and immunity in response to infection by a variety of pathogens. Both IKKα and IKKβ have been reported to modulate either pro- or anti- inflammatory programs, which may be specific to the infectious organism or the target tissue. Here, we analyzed the requirements for the IKKs in myeloid cells in vivo in response to Francisella tularensis Live Vaccine Strain (Ft. LVS) infection. Methods and Principal Findings In contrast to prior reports in which conditional deletion of IKKβ in the myeloid lineage promoted survival and conferred resistance to an in vivo group B streptococcus infection, we show that mice with a comparable conditional deletion (IKKβ cKO) succumb more rapidly to lethal Ft. LVS infection and are unable to control bacterial growth at sublethal doses. Flow cytometry analysis of hepatic non-parenchymal cells from infected mice reveals that IKKβ inhibits M1 classical macrophage activation two days post infection, which has the collateral effect of suppressing IFN-γ+ CD8+ T cells. Despite this early enhanced inflammation, IKKβ cKO mice are unable to control infection; and this coincides with a shift toward M2a polarized macrophages. In comparison, we find that myeloid IKKα is dispensable for survival and bacterial control. However, both IKKα and IKKβ have effects on hepatic granuloma development. IKKα cKO mice develop fewer, but well-contained granulomas that accumulate excess necrotic cells after 9 days of infection; while IKKβ cKO mice develop numerous micro-granulomas that are less well contained. Conclusions Taken together our findings reveal that unlike IKKα, IKKβ has multiple, contrasting roles in this bacterial infection model by acting in an anti-inflammatory capacity at early times towards sublethal Ft. LVS infection; but in spite of this, macrophage IKKβ is also a critical effector for host survival and efficient pathogen clearance.


Introduction
NF-kB is an important signaling pathway for the induction and regulation of innate and adaptive immune responses toward bacterial infection. Microbial components and pro-inflammatory stress-like signals universally impact the activation of NF-kB transcription factors via the inhibitor of NF-kB kinase (IKK) signalosome complex. The signalosome contains two catalytic kinases, IKKa and IKKb, and a third docking/regulatory subunit, IKKc/NEMO [1]. The catalytic IKKs mediate the phosphorylation and subsequent degradation of the IkB family of cytoplasmic inhibitory proteins. IkB degradation liberates NF-kB transcription factors, resulting in nuclear translocation and target gene activation (for reviews see [2,3,4]). Canonical NF-kB activation requires IKKb and IKKc for IkB degradation, while the role of IKKa in the canonical signalosome is less clear. In addition, IKKa and IKKb have also been shown to possess a variety of NF-kB-independent functions by regulating effectors of cell cycling, apoptosis, specific cellular differentiation pathways, chromatin activity and inflammatory responses {reviewed in. [5,6,7,8]).
In addition to the well-established roles the IKKs have on the induction of inflammation and adaptive immune responses, myeloid IKKa and IKKb also limit inflammation in response to the extracellular bacterium Group B Streptococcus (GBS) or by E. coli LPS-(lipopolysaccharide) induced septic shock [9,10,11,12]. This unexpected behavior of the IKKs in myeloid cells led us to further investigate their anti-inflammatory properties in the context of the intracellular bacterium Francisella tularensis (Ft.).
Francisella tularensis is a highly infectious, Gram-negative, intracellular bacterium and is the causative agent of tularemia. Due to the high virulence of the human pathogenic strain ShuS4, the attenuated Live Vaccine Strain (Ft. LVS) is commonly used in mouse models of infection due to its ability to approximate human disease [13,14]. Ft. infects several cell types [15,16,17,18,19,20], but macrophages are considered the primary target of early infection [19,21,22]. Bacteria disseminate to colonize the lungs, liver and spleen presumably through the hematogenous route [13,23]. Examination of Ft. LVS or ShuS4 infected organs indicates that lung, liver and spleen each contribute distinct immune reactions [15,24,25,26]. Within 72 hours of intradermal (i.d.) or intraperitoneal (i.p.) inoculation of Ft. LVS, a pro inflammatory gene expression profile is evident in liver and to a lesser extent in spleen, while lungs respond in an anti inflammatory manner [24].
The virulence of Francisella has been attributed to its ability to down-modulate and/or evade host defenses. For example, outside the cell, Francisella are opsonized with complement proteins but are able to resist killing by complement-mediated lysis [27,28,29]. Opsonized bacteria are taken up readily by phagocytic cells, but avoid degradation by preventing fusion of host phagosomes to lysosomes and subsequently escape to the cytosol where they replicate [30,31,32,33]. Although Ft. is a Gram-negative bacterium, it has evolved an unusual LPS structure, which only minimally activates the pattern recognition receptor TLR4 (Toll-like Receptor-4) [34,35,36], and instead, innate immune recognition of Ft. LVS is mediated through TLR2 (Toll-like receptor-2) [37,38,39,40].
Francisella also interferes with anti-microbial defenses and inflammatory signaling pathways in several cell types, including macrophages and neutrophils. For example, Ft. LVS can prevent anti-microbial ROS (reactive oxygen species) production and NADPH oxidase assembly in human neutrophils [18]. Ft. LVS actively down-modulates TNF-a and IL-1 in the J774A murine macrophage cell line; this coincides with the inhibition of IkBa degradation and cross-tolerization to E. coli LPS (a TLR4 agonist) resulting in interference of NF-kB signaling [41,42].
There is also evidence that Ft. LVS affects macrophage polarization [43]. As early as 24 hours post infection, infected peritoneal elicited macrophages show properties of anti-inflammatory, alternatively-activated/M2 polarization resulting in increased expression of the M2 markers: mannose receptor (CD206), Fizz-1, Arg-1 and Ym1 [43]. Similar to the T cell polarization paradigm, macrophage polarization also depends on cytokine cues induced by the local environment (reviewed in [44,45]). In agreement with this, Francisella interferes with IFN-c signaling, a key cytokine involved in macrophage polarization and bacterial control. Ft. LVS and its closely related subspecies Ft. novicida modulate IFN-c signaling by suppressing tyrosine phosphorylation of the STAT1 transcription factor (Signal Transducer and Activator of Transcription 1) in human and murine mononuclear cells and this correlates with up-regulation of the STAT1 inhibitor SOCS3 (suppressor of cytokine signaling) [46].
Since infection with Ft. suppresses key inflammatory regulators, we hypothesized that the anti-inflammatory properties of myeloid IKK may compound and/or contribute to the progression of tularemia in vivo. To investigate this, we used myeloid-specific conditional knockout mice (IKK cKO) [47] to assess the roles of the IKKs in an intradermal (i.d.) model of Ft. LVS infection. We chose the liver as the model organ for our study because during tularemic infection, it supports pro-inflammatory changes [24], is colonized early in infection and has an abundant source of macrophages [48].
We found that myeloid IKKb is required for survival during septic challenge, while IKKa cKO mice displayed mortality rates comparable to control mice. In a sublethal model of infection, flow cytometry anaylsis of hepatic non-parenchymal cells showed that loss of myeloid IKKb, but not IKKa, results in polarization toward M1 macrophages in the liver early in the course of infection. Interestingly we found these effects are transient, as by mid-infection, macrophages polarize towards the M2a lineage. Loss of myeloid IKKb also induces protracted elevations in IFN-c expressing CD8 + T cells, which persist throughout the course of infection. Despite these changes, IKKb cKO mice are defective in control of bacterial growth. Finally, histological analysis shows that infection results in abnormal granulomatous liver responses in both strains of IKK cKO mice.

Animals
Myeloid specific ikk deletions were generated by crossing IKKa or IKKb floxed mice with the LysM Cre expressing mouse strain [49] to generate IKKa flox/flox -LysM Cre or IKKb flox/flox -LysM Cre strains, all of which were maintained on a C57BL/6 genetic background, as previously described [47]. To simplify nomenclature used throughout this text, mice are herein referred to as IKK f/f to represent control IKK floxed strains without myeloid deletion and IKKa cKO or IKKb cKO to represent corresponding myeloid ikk conditional deletions. Mice were housed in a facility equipped with a 12:12 hour light:dark cycle in ventilated cages; and were fed a normal chow diet and autoclaved water ad libitum.

Ethics Statement
All procedures were performed in strict accordance with State University of New York at Stony Brook IACUC approved protocol (Permit Number 0163).

Bacteria
Ft. LVS (ATCC 29684, American Type Culture Collection, Manassas, VA) were a kind gift from the laboratory of Dr. Martha Furie, and were grown according to published protocols [50]. Briefly, frozen stocks were streaked on chocolate II agar supplemented with hemoglobin and IsoVitaleX (BD Biosciences, San Jose, CA) and grown at 37uC for 3 days. Single colonies were inoculated into pre-warmed (37uC, 5% CO2) Mueller Hinton-II broth and incubated at 37uC with shaking for 16-18 h. Bacteria were used at OD 600 between 0.2-0.4.

Infection Model
Bacteria were serially diluted to the indicated doses in sterile PBS in a final volume of 100 ml. Eight-to twenty-week old female mice were shaved and injected intradermally (i.d.) at the base of the tail. Mice were monitored for signs of illness (hunched posture, ruffled fur, loss of appetite, etc.). At lethal doses, these symptoms sometimes did not appear until hours before succumbing to infection. In addition, occasionally mice that appeared ill had spontaneously recovered. Sublethal infection did not induce symptoms of illness. Doses, listed as colony forming units (CFU), were confirmed by a retroplate colony count assay as described below.
For comparison of survival between all strains, mice (10-15 per group) were injected with an LD 50 dose 10 8 CFU i.d. and monitored twice daily over a period of 14 days. Kaplan-Meier survival curves were generated and relative mortalities in response to lethal infection were compared by Mantel-Cox log-rank test.

Organ Burden Determination and Retroplate Colony Counts
Organ burden was determined using limiting-dilution culture. Briefly, sections of the median lobe (,0.1-0.2 g) of infected livers were aseptically removed and placed in a pre-weighed sterile tube containing 1 ml of sterile PBS. The remaining liver and section were weighed. The section was homogenized in a stomacher bag (Fisher Scientific, Pittsburgh, PA) and serially diluted. Neat homogenates and serial dilutions were plated and incubated for 3 days at 37uC. The resulting colonies were counted and represented as a concentration per gram of liver. This concentration was then multiplied by the weight of the total liver and corresponding dilution factor to derive the total organ burden. Retroplate colony counts were performed in a similar manner. First, serial dilutions on the inoculation dose were made. Dilutions were then plated and incubated as above. The resulting colonies were counted, multiplied to respective dilution factors and represented as CFU/ml. The limit of detection for these assays was 200 CFU.

Histology
Organs were fixed in 10% neutral buffered formalin (Sigma Diagnotics, St. Louis, MO) and embedded in paraffin, sectioned at 5 mm rehydrated to water through graded alcohols. Tissues were stained with hematoxylin and eosin (H & E), dehydrated in graded alcohols, cleared with xylene and mounted with Acrymount (EMS, Hatfield, PA). An experienced pathologist performed a blinded analysis on H & E sections to identify any abnormalities in lung, liver and spleen. Images were captured on an Olympus BX41 light microscope (Olympus, Tokyo, Japan).
Granuloma counts from 8 mice/group were quantified by examination of 10 fields/liver under 2006 magnification. Poisson distribution was assumed and fitted to Poisson regression by taking the square root of count data and mean comparisons were made with one-way ANOVA and Tukey's post-test. Data was then backtransformed into original units of granulomas/2006 field (0.150 mm 2 ).

Immunohistochemistry
Tissues were fixed in neutral-buffered formalin for 48 hrs, embedded in paraffin and sectioned at 5 mm. Deparaffinization, rehydration and antigen retrieval was performed by incubating slides in Trilogy TM solution (Cell Marque, Rocklin, CA) for 1 hr at 90uC. Slides were rinsed in TBST (Tris Buffered Saline-Tween; Thermo-Scientific, Freemont, CA), blocked in goat serum for 30 minutes at RT, and rinsed again in Tris buffer. Slides were incubated with polyclonal anti-Ft. LVS rabbit serum or anti-iNOS (Abcam, Cambridge, MA) primary antibody (diluted 1:50 in Tris buffer) for 16 hours at 4uC. Slides were rinsed and then incubated with AP-conjugated goat anti-rabbit secondary antibody (diluted 1:100; Invitrogen, Carlsbad, CA) and developed using Vulcan Fast Red Chromagen (Biocare Medical, Concord, CA) for 10 minutes. To counterstain, slides were rinsed in H 2 O and dipped twice in Mayer's Hematoxylin (Cancer Diagnostics, Morrisville, NC). Slides prepared for the detection of cleaved caspase-3 (CC-3) were processed in similar manner with a few modifications. Endogenous peroxidase was inhibited using Peroxidazed-1 blocking reagent (Biocare Medical) for 5 minutes prior to blocking in goat serum. CC-3 primary antibody (Cell Signaling Technology, Danvers, MA) was diluted 1:50 and applied as above. A secondary goat anti-rabbit antibody (1:500 dilution; supplied with Vectastain, ABC kit, Vector Laboratories, Burlington, CA) was applied for 30 minutes. Slides were rinsed in Tris buffer and ABC reagent from the same kit was added for 30 minutes. ImmPACT TM DAB solution (Vector Labs) was applied for 6 minutes, rinsed and counterstained as above. Staining was absent in control slides where primary antibody was omitted. Occasional weak background was noted in uninfected controls. Images of immunostained slides were captured using a 206 objective on an Aperio ScanScope CS slide scanner (Aperio,Vista, CA ) to create whole-slide digital images. Images were subsequently analyzed using Aperio ImageScope v11.12.2.760 Positive-Pixel Count Algorithm v9 (Aperio) for systematic identification of Ft. antigen, iNOS or cleaved caspase-3 positivity. The software samples the image and reports a staining intensity (a value proportional to the amount of transmitted light) and generates a pseudo-colored markup image to verify staining intensities. Staining intensities are thus reported as negative (255, blue), weak positive (220, yellow), positive (175, orange) and strong positive (100, brown). Using these parameters, we defined weak positive (yellow) as non-specific/background staining, while positive and strong positive values (orange+red/ brown) were considered to be specific staining. The algorithm was applied to both positive and negative control tissues to ensure validity of the settings.
Positive granulomas were counted by examination of 10 fields/ liver under 2006 magnification. Poisson distribution was assumed and fitted to Poisson regression by taking the square root of count data and mean comparisons were made with one-way ANOVA and Tukey's post-test. Data are represented as percent positive granulomas.

Non-parenchymal Cell (NPC) Isolation
NPC's were isolated by the methods of Rasmussen et al [51]. Briefly, mice were injected with 10 6 CFU of Ft. LVS and euthanized at the indicated time points. On the day of harvest, livers were perfused with 1 mM Citrate (Sigma, St. Louis, MO) in Hanks Buffered Saline Solution (HBSS, Gibco/Invitrogen Carlsbad, CA). Livers were minced to 1 mm pieces and placed in digest buffer containing 0.05% Collagenase II, 0.002% DNAse1 and 1% BSA (all from Sigma-Aldrich, St. Louis, MO) in HBSS, rotating at 75 RPM, 37uC for 45 minutes. Digests were triturated 10X and filtered through a 70 mm cell strainer (BD Falcon, San Jose, CA). Cells were pelleted and resuspended in ACK hypotonic buffer (0.15 M NH 4 Cl, 10 mM KHCO 3 , 0.1 mM Na 2 EDTA) for 5 minutes to lyse RBCs. Cells were washed in PBS and NPCs were isolated by flotation method using Optiprep density gradient (AxisShield, Oslo, Norway), according to manufacturer protocol C-24. Upper and lower isolates were collected and the NPC fraction was confirmed by flow cytometry. The upper isolate was .97% CD45+ (eBioscieinces, San Diego, CA).

Flow Cytometry
NPCs isolated from infected livers were stained with antibodies for surface and/or intracellular markers for immunophenotyping and cytokine positivity according to manufacturer suggested protocols. Briefly, harvested cells were immediately incubated in growth media (RPMI 1640, 5% heat inactivated FBS (Invitrogen/ Life Technologies, Carlsbad, CA) without antibiotics) containing 1 mg/ml GolgiPlug (BD Biosciences, San Jose, CA) for 6 hours to prevent cytokine release; no exogenous stimuli were added during the incubation period. GolgiPlug was added to all subsequent flow cytometry buffers until the fixation step. Cells were washed 1X in PBS and 1X in FACS Stain Buffer (FSB: 1% heat inactivated FBS, 0.09% w/v NaN 3 in PBS). Fcc receptors were blocked with CD16/32 antibody (eBiosciences, San Diego, CA). Surface epitopes were stained with antibody for 30 min., washed 2 times with FSB. Cells were fixed and permeabilized using BD Cytofix/ Cytoperm buffer kit (BD Biosciences, San Jose, CA) for 20 min. Intracellular antibodies were added to cells and incubated for 30 min. Unlabeled antibodies required secondary labeling was carried out in a third staining step as above. All labeling and permeabilization steps took place in the dark with rotation. The following antibodies were used in this study: Arg-1 (

Statistical Analysis
Data were analyzed by one-way ANOVA with Tukey's multiple comparison post-test using GraphPad Prism 6 software (GraphPad Software, Inc., La Jolla, CA).

Reduced Survival of Myeloid IKKb-deficient Mice Following Lethal Intradermal Ft. LVS Challenge
The IKKs are important signaling factors involved in orchestrating host immune responses to a wide variety of pathogens, yet the requirements for IKK signaling in myeloid cells during Ft. LVS infection are unknown. To address this, we employed Cre/ Lox mediated recombination in which IKKa or IKKb are conditionally deleted from the myeloid compartment of adult mice [47,49,52].
The severity of Francisella infection varies by bacterial strain, host genetic background, dose and route of administration [13,23,53]. In order to establish the correct dosage for our model system, we began with a pilot experiment in which we intradermally (i.d.) injected IKK f/f mice at various doses of Ft. LVS in order to establish the median lethal dose (LD 50 ). After retroplate assay correction, the LD 50 was determined to be 10 8 CFU of Ft. LVS (Fig. S1 A-C).
We first investigated if either IKKa or IKKb in myeloid cells were required for host survival to Ft. LVS infection. We injected IKK f/f , IKKa cKO and IKKb cKO mice with the LD 50 dose of 10 8 CFU Ft. LVS and scored animals for their relative survival rates. All IKKb cKO mice succumbed to infection by day 10 with a comparative Log-rank test P value of 0.0014 (Fig. 1); there were no significant differences in the survival rates among IKK f/f and IKKa cKO mice.

Both IKKa and IKKb in Myeloid Cells Contribute to Granulomatous Response in the Liver during Sublethal Ft. LVS Infection
Next we asked how conditional loss of either IKK in myeloid cells affects disease progression using a sublethal model of infection. In order to determine the sublethal dose, we tested mice at lower doses of inoculum (Fig. S1 B-C). IKK f/f and IKKa cKO mice had an 80% survival rate at 10 7 CFU of Ft. LVS, in contrast IKKb cKO mice showed increased sensitivity to infection resulting in a 60% survival rate at this same dose. All three strains of mice survived i.d. infection of 10 6 CFU of Ft. LVS for a period of 90 days; and thus we used this dose for all sublethal experiments.
Infection with Ft. LVS results in dissemination to lung, liver and spleen, and results in inflammatory infiltration and/or development of granuloma-like foci in target organs. We examined these tissues by hematoxylin and eosin (H & E) staining in a time course experiment to assess histopathological changes.
Lung histology was normal in all three strains of mice prior to infection (data not shown). At two days post-infection, mild lung reactions were observed ( Fig. 2A-C and Fig. S2). Peribronchial inflammatory infiltrates were sometimes present and consisted largely mononuclear cells and a few neutrophils ( Fig. 2A and C, insets). Areas of peribronchial infiltrate were more numerous in IKKb cKO mice and these infiltrates also contained lymphocytes (Fig. 2C, inset). Alternatively, lungs from IKKa cKO mice showed the lowest overall degree of inflammation (Fig. 2B) at this same time point. Progressing through days 6 (data not shown) and 9 ( Fig. 2D-F and S2), all three strains of mice showed some degree of thickening in the alveolar walls and septae. Luminal alveolar exudate and dilated capillaries were sometimes observed, indicating possible onset of a mild interstitial pneumonia. We noted, on occasion, a minor degree of activation in cells of the mesothelial lining, a symptom of visceral pleuritis. However, only small areas were affected and this did not occur with any regularity. Interestingly, lungs from IKKb cKO infected mice developed multiple areas of organized lymphocytic aggregates (Fig. 2F), while this was a rare occurrence in IKK f/f or IKKa cKO mice. These aggregates resemble inducible bronchial-associated lymphoid tissue (iBALT), which can be induced in humans and mice upon exposure to infection and inflammation and are considered tertiary sites of B and T cell priming reactions [54]. Immunohistochemical analysis to detect Ft. LVS antigen (Fig. S2) showed a limited infection within the lung parenchyma. The overall conclusion about lung reactions in all three strains of mice was that they were relatively mild. These findings are consistent with previous reports for intradermal infection, not only with Ft. LVS [55], but also with the more virulent strain Francisella tularensis Schu4 [23].
The spleen and liver undergo pronounced inflammatory reactions upon Ft. LVS infection and are characterized by the development of granuloma foci, which act as a microenvironment that serves to contain infection and clear debris from infected and dying cells [19,23,51,56]. Histological examination of the spleens from uninfected mice appeared normal and showed clearly demarcated red and white pulp. Follicles, germinal centers and the surrounding cuff of the marginal zone were well represented (data not shown). At two days post infection, the red pulp was marginally expanded, but other structures were qualitatively normal in all three strains. A few small granuloma-like lesions began to appear within the red pulp of most animals (Figs. 3A-C and Fig. S3). More severe histopathological changes to splenic architecture progressed through days 6 and 9 post-infection (data not shown and Fig. 3D-F, respectively). The white pulp showed evidence of expansion and macrophage and lymphocytic infiltrates were seen in the red pulp. The number of granulomas marginally increased in IKK f/f and IKKa cKO mice. However, in IKKb cKO infected mice, the red pulp was often completely effaced with coalescing granulomas making accurate quantification difficult. Neutrophils were sometimes present within these granulomas (Fig. 3F, inset). Splenomegaly was apparent in all strains (Fig. 3G). Comparisons of spleen/total mouse weight ratios (spleen index) showed a ,3 fold change by day six. The spleen index modestly increased to 3.5-4 fold by day nine. No significant differences in spleen indices between strains were observed. Immunohistochemical analysis of Ft. antigen in spleen at 2 days post infection showed numerous bacterial foci (Fig. S3 A-C) in all three strains of mice. Bacteria were localized predominantly in the red pulp, but were occasionally observed in the white pulp as well (Fig. S3 A). By day 9, there was a marked decrease in the number of bacterial foci, indicating resolution of infection.
Granuloma formation in liver was evident at two days postinfection in all three strains (Fig. 4A-C). At day 2, granuloma counts in IKKb cKO mice were at least two fold greater in number than those of IKK f/f or IKKa cKO mice (Fig. 4G, P,0.0001). By day 9, granuloma formation continued to increase in IKKb cKO at a rate far greater than either IKK f/f or IKKa cKO mice. Additionally, the granuloma-like structures that developed in IKKb cKO infected mice were often extremely small and decondensed, resulting in livers replete with inflammatory cells (Fig. 4F). Some granulomas in IKKa cKO mice developed into very large, macroscopically discernable structures that contained a central core filled with necrotic and cellular debris ( Fig. 4E; note that panel 4E is a lower magnification than 4D and 4F), and although large, these structures were well contained by a cuff of mononuclear cells, epitheloid histiocytes and occasional lymphocytes. Although both spleen and liver showed increased numbers of granuloma foci in IKKb cKO mice, the liver also showed clear discernable defects in granuloma maintenance in both mutant strains of mice. Since the liver reacts with a strong pro-inflammatory response to Ft. LVS intradermal infection [24] and contains a rich source of macrophages, we focused the remainder of our experiments on the liver to further investigate the anti-inflammatory effects of myeloid IKKb during the course of tularemia.

Myeloid IKKb is Required for Control of Bacterial Growth in the Ft. LVS Infected Liver
Bacterial colonization is first detectable in Ft. infected livers at 2-3 days post infection [19,24,51,56]. In order to determine if the observed granuloma defects in the liver were due to increased bacterial colonization, a retroplate time course study to determine organ burden was performed on liver homogenates from mice infected with 10 6 CFU of Ft. LVS. At day 2 post infection, both IKK f/f and IKKa cKO mice showed limited colonization of the liver while, even at this early time point, IKKb cKO mice presented with increased bacterial loads that persisted through day 14 (Fig. 5A).
Macrophages are considered a primary cellular target of Francisella infection [22,57] and are also a major component of Ft. induced granulomas [51]. Since macrophages are one of the cell types affected by our conditional knockouts, we next questioned whether the macrophages from these mice were comparably infected relative to control mice. To address this, non-parenchymal cells (NPCs) were isolated from the livers of infected mice, stained with F4/80 and CD11b antibodies followed by intracellular staining for Ft. LVS and then subjected to flow cytometry. Interestingly, IKKb cKO infected mice showed only modest increases in the number of Ft. LVS positive macrophages throughout the course of infection (Fig. 5B). To account for a possible loss of cells due to infection, we analyzed macrophage and neutrophil cell populations from each strain. We isolated hepatic NPCs from sublethally Ft. LVS infected livers and performed a flow cytometry analysis time course to evaluate the populations of F4/80 + CD11b + expressing macrophages and Ly6B.2 + expressing neutrophils. Prior to infection (day 0), all three strains of mice yielded comparable numbers of macrophages and neutrophils (Fig. 6A & B, respectively). Modest increases were observed in the macrophage population but persisted throughout the 8-day time course (Fig. 6A). As early as two days post infection, increases in neutrophils were evident in IKKb cKO mice and by day 8, neutrophils were elevated in both mutant strains, relative to the parental strain (Fig. 6B). Taken together, these data suggest that the increased liver burden in IKKb cKO mice or granuloma maintenance were due to defects in myeloid function related to control of bacterial growth and spread of infection rather than overt initial increases in macrophage infection (Fig. 5B) or loss of cells from the myeloid compartment ( Fig. 6A-B).

Myeloid Function in the Granuloma
To further examine myeloid function in liver granulomas, we performed immunohistochemical assays to detect the spatial localization of LVS antigen, production of the inflammatory mediator iNOS (inducible nitric oxide synthase) and induction of caspase-3 (CC-3) activation. Figure 7 shows representative granulomas from each of the three strains of mice. LVS antigen was largely restricted to the granuloma (Fig. 7. A-F), however, a small number of Ft. antigen positive cells (8-11cells/2006 field) were sometimes found within the parenchyma. Ft. antigen positive cells were almost two-fold higher in IKKb cKO mice, but this correlated with higher bacterial burdens.
Lindgren et al [58] reported that intradermal inoculation of Ft. LVS in iNOS (inducible Nitric Oxide Synthase) deficient mice results in decreased host survival, increased bacterial colonization and increased liver damage concurrent with the appearance of numerous small granuloma-like foci in the liver. Moreover, iNOS, in addition to induction by IFN-c, is also a secondary response target gene activated through TLR signal transduction [59,60]. We next asked if iNOS was similarly expressed in liver granulomas of IKK deficient mice. At two days post infection, we observed iNOS positivity in 54%, 40% and 54% in hepatic granulomas from IKK f/f , IKKa cKO and IKKb cKO mice, respectively (Fig. 7  B, H, N). By nine days post-infection (Fig. 7 E, K, Q), only modest increases in the number of iNOS-postive granulomas (,10%) in IKK f/f and IKKb cKO infected mice were noted. In comparison, the number of IKKa cKO iNOS positive granulomas remained essentially unchanged at 42%.
Rapid induction of apoptosis in bacterially infected cells is an immune defense mechanism that helps to limit the spread of infection. We analyzed induction of apoptosis in hepatic granulomas using the apoptotic marker, cleaved caspase-3 (CC-3). CC-3 staining was largely restricted to granulomas. At early time points in infection, ,35% of all granulomas scored positive CC-3 in all strains of mice. However, only a few (,1-8) positive cells were found per granuloma (Fig. 7M-R), and this was consistent between all strains of mice. By day nine, CC-3 positive granulomas from IKK f/f mice was reduced to 11.3%, while IKKa and IKKb cKO mice retained an overall 25 and 22% CC3 positive granuloma score, respectively.  M1 and M2a macrophage subpopulations, we scored macrophages for their F4/80 expression and either macrosialin (murine CD68) or the mannose receptor (CD206), (Fig. 8A and 9A, respectively). Next, we evaluated each subpopulation for expression of the prototypical M1 cytokine, IL-12 (Fig. 8B) or multiple M2 activation markers, including IL-10, arginase-1 and RELMa/ FIZZ1 (Fig. 9B-D).
We observed a 5-fold increase (P = 0.0040) in the number of M1 macrophages (F4/80 + CD68 + IL-12 + ) in IKKb cKO mice at 2 days post-infection (Fig. 8B). This data indicates that myeloid IKKb contributes to the suppression of M1 classical activation in the liver in response to sublethal Ft. LVS infection. However, this IKKb-mediated M1 suppression only occurred in a short time period early in infection, as by day 6 there was a shift toward M2a activation as indicated by increases in F4/80 + CD206 + staining in coordination with several other M2a-associated markers (Fig. 9B-D) including: IL-10, a 2 fold increase P = 0.0019; FIZZ1, a 6 fold increase P = 0.0085; and Arg-1, a 9 fold increase P = 0.0041).

Livers of Ft. LVS Infected IKKb cKO Mice Present Protracted Activation of CD8 + T cells in the Liver
We next evaluated the collateral effects of IKK myeloid deletion with respect to the activation of IFN-c by other cell types after sublethal i.d. Ft. LVS infection. Again, using a flow cytometric approach on liver NPCs, we scored for IFN-c activation in NK (NK1.1 + CD3 2 ), CD4 + and CD8 + T cell (CD4 + CD3 + and CD8 +  CD3 + , respectively) populations. The NK cell populations from IKKa and IKKb cKO infected mice were largely similar to IKK f/f control animals (Fig. 10A). A small increase in NK cells were seen in IKKa cKO animals on day 8 post-infection, yet we did not observe corresponding increases in IFN-c + cells at the same time point (Fig. 10B). Increases in both CD4 + and CD8 + T cell populations were observed at later time points in infection in IKKb cKO mice (Fig. 10C and E). In CD8 + T cells, these increases correspondingly correlated to greater increases in the number of IFN-c + cells (Fig. 10E and F). Interestingly, increases in IFN-c + CD8 + T cells from IKKb cKO infected mice were protracted throughout the entire infection time course (Fig. 10F). This trend was not observed in IKK f/f or IKKa cKO infected mice. These results suggest that myeloid IKKb may function in an extrinsic manner to suppress the influx of activated IFN-c producing cells.

Discussion
The results of this work shed new light on the contributions of the IKKs in myeloid cells during Ft. LVS infection and also extend previous findings on the in vivo roles of the IKKs in myeloid cells in response to bacterial infections. In spite of the anti-inflammatory properties demonstrated by others [9,10,11], myeloid IKKs exhibit notable differences in response to the intracellular bacterium Ft. LVS. Here, we found that myeloid IKKb has a protective role in response to lethal challenge. At sub-lethal doses, histopathology and retroplate culture assays showed that myeloid IKKb helps regulate the granulomatous response and control infection.
We also examined the effect loss myeloid IKKb had on macrophage polarization. Indeed, IKKb inhibits classical M1 IKKb but not IKKa in Myeloid Cells Contributes to Host Survival to a Lethal Dose of Ft. LVS The survival results we obtained in our study are in contrast to those reported for myeloid IKK deficient mice responding to a lethal Gram-positive bacterium GBS (Group B Streptococcus) challenge [10,11]. In this previous study, mice devoid of myeloid IKKb were more resistant to a lethal challenge with GBS [11]. Here, we found that intradermal inoculation with Gram-negative bacterium, Ft. LVS rendered myeloid IKKb deficient mice much more susceptible to lethal infection. These results correspondingly correlated to levels of bacterial titers, as there was increased bacterial clearance in the GBS model [11], while we observed increased colonization in the Ft. LVS model. In addition, loss of functional IKKa (ikka AA/AA phosphorylation mutant) in the myeloid compartment results in increased susceptibility to GBS infection [10], while here, we found IKKa cKO mice maintain a comparable susceptibility in response to Ft. LVS infection to control animals. These contrasting results in survival between the Francisella and GBS infection models in myeloid IKK deficient mice suggest that there are differential requirements for downstream TLR effectors, such as the IKKs, during septic challenge. In fact, similar differences in survival patterns have been reported for the TLR adapter protein MyD88 (Myeloid-differentiation factor 88). Mice deficient in MyD88 are protected from GBS infection at high infection doses, [61], while Ft. LVS infected mice are extremely susceptible to MyD88 loss at low doses [55].
Moreover, increased mortality occurred in myeloid IKKb deficient mice at the LD 50 dose of 10 8 CFU of Ft. LVS, indicating a protective role for myeloid IKKb during overwhelming sepsis. IKKb deficiency also resulted in a dose-dependent survival response, an effect which has been observed, to varying degrees, for other components involved in NF-kB signaling. For example, loss of either TLR2 (Toll-like receptor 2) [62] or MyD88 [55] results in reduced survival as a result of tularemic infection. Yet, loss of TLR2 affects survival at moderate infection doses (,4610 4 CFU) [62], while MyD88 deficient mice are extremely sensitive and succumb to Ft. LVS at infection doses orders of magnitude lower (5610 1 CFU) [55]. The increased sensitivity of MyD88 loss is due to its requirement in multiple signaling pathways. Studies show that MyD88 loss affects the NF-kB, c-Jun and p38 MAPK pathways with delayed kinetics [63], abrogated IL-1R-and IL-18R -mediated cytokine production [64], as well as impaired IFN-c and TNF-a production in response to several pathogens [63,64]. In addition, it is important to note that the dose-specific sensitivity to Ft. LVS in the TLR2 and MyD88 studies were performed in full knockout models, while the IKKb model used here was conditionally restricted to the myeloid compartment. Thus, this is the first study to show a myeloidspecific requirement for host survival in response to Francisella tularensis.

Ft. LVS Disseminates to Lung, Liver and Spleen
Dissemination of bacteria led to bacterial colonization in lung, spleen and liver tissues. Only mild histopathological findings in lung tissue were found. One unexpected finding was the development of organizing lymphocytic aggregates within the lungs IKKb cKO mice after infection. These aggregates, also known as iBALT (inducible bronchus-associated lymphoid tissue), are not present in normal lung in humans or mice, but can be induced after exposure to antigen, inflammation or infection (reviewed in [65,66]). In mice lacking spleen, lymph nodes and Peyer's patches, iBALT acts as a site of B and T cell proliferation in response to influenza infection and is considered protective in nature [54]. iBALT develops under pathologic states, and it is suggested that this may involve the activation of PRR (pattern recognition receptors) such as the TLRs [65]. Although further examination into the development of iBALT structures in IKKb cKO mouse lung was beyond the scope of this current investigation, it is of note that these structures were previously identified in a number of Ft. LVS vaccine studies. Organized lymphoid structures are induced upon aerosol or intranasal administration of vaccines using whole [67], inactivated Ft. LVS [68] or Ft. LVS LPS plus adjuvant [69]. However, unlike these studies, where mice were immunized with several repeated doses of bacteria, our model consisted of a single intradermal inoculation. It is possible that the increased bacterial burden in IKKb cKO mice, induces these structures as a result of ongoing and repeated exposure to Ft. antigen. Furthermore, iBALT structures persist in these mice well after bacterial clearance and are still evident at 90 days post infection (S.S., unpublished data).
IKKb cKO mice responded to infection with an increased frequency in granuloma response in both spleen and liver. IKKa cKO mice developed some abnormally large granulomas within liver; however, the spleen granuloma reaction was qualitatively and quantitatively similar to IKK f/f mice. Analysis of Ft. antigen in the liver showed the bacteria were mainly localized within granulomas, although a limited number of infected cells could be detected within the parenchyma.
In the liver, iNOS (inducible nitric oxide synthase) is produced by macrophages and hepatocytes and can be induced by microbial lipoproteins through TLR activation, IFN-c or TNF-a. iNOS mediates the metabolic conversion of L-arginine to NO (nitric oxide) and citrulline. NO is not only a powerful antimicrobial effector molecule capable of direct killing, it is also involved in several other aspects of host defense including cytokine production and apoptosis [70]. iNOS expression within granulomas was previously reported to be low and infrequent during intradermal infection [24]. However, we found 40-54% of hepatic granulomas were positive for iNOS induction by day 2, with only marginal changes in expression throughout the course of infection. Thus, it is unlikely that the observed defects in granuloma maintenance result from aberrant iNOS induction.
Apoptosis is considered an important immune defense mechanism. Rapid elimination of infected cells helps prevent the spread of infection to other cells and also promotes dendritic cell uptake of apoptotic bodies, allowing access to antigen which may trigger both innate and adaptive immune responses. However, many pathogens have evolved mechanisms to escape destruction by subverting or redirecting these processes (reviewed in [71]). We found the early apoptosis marker, cleaved caspase-3 (CC-3), spatially localized within the granulomas in all three strains of mice. We found no overt differences between strains in the percentage of CC-3-positve granulomas at early time points. Furthermore, only a few cells per granuloma stained positive for this marker, but this was consistent with other models of intradermal Ft. LVS infection [51]. Lawrence et al suggests that increased bactericidal control in a GBS model of pneumonia observed in ikka AA/AA macrophages is due, in part, to increased activation of anti-apoptotic genes combined with sustained activation of inflammatory genes such as iNOS [10]. In the Ft. LVS model, these effects on apoptosis and iNOS production were not evident, suggesting that the molecular mechanisms between these two infection models are different. Furthermore, GBSinfected ikka AA/AA mice, despite increased bacterial clearance at early time points, eventually succumb to infection, whereas survival in IKKa cKO mice infected at the LD 50  LVS infected mice is in accord with the GBS infection model [11]. Because our model system allows us to monitor macrophage polarization over days, not hours (as is the case in GBS model), we also discovered that IKKb cKO mice undergo a second wave of polarization toward M2/alternatively activated macrophages at the mid-infection stage. Although we cannot rule out that this second wave of polarization stems from secondary or bystander effects induced by the local environment, it is of note that the antiinflammatory IKKb-dependent blockade of M1 classical macrophage activation is transient in nature and thus an important consideration for design of therapeutic targeting.

IKKb Loss in Myeloid Cells Results in Collateral Effects on IFN-c Producing Cells
The compounding effects of increased bacterial titers and M1/ M2 activation defects in IKKb cKO mice also correlated with increases in several types of IFN-c expressing cells. Furthermore, protracted elevations in IFN-c expressing CD8 + T cells were protracted throughout the course of infection. Elevations of these cells persisted, despite the mid-infection shift in M2 macrophage polarization. It has long been considered that early tularemic infection is controlled by T cell-independent events, as T cell depleted mice (depleted of both CD4 + and CD8 + T cells) survive initial infection, but fail to clear bacteria and succumb to infection several weeks after inoculation [72,73,74]. Interestingly, mice selectively depleted for only CD4 + T cells or CD8 + T cells are able to resolve and survive infection [75]. This implies that these T cell subsets have roles in clearance and resolution of infection that largely overlap. Although the exact mechanism(s) by which myeloid IKKb deficiency results in a collateral elevation of reactive CD8 + T cells remains to be determined, our data suggest that when myeloid NF-kB is impaired due to loss of IKKb, CD8 + T cells overtly respond as an immunologic compensatory mechanism. Therefore, it seems that CD8 + T cells are poised to react during the T-cell independent phase of Ft. infection.
In summary, we show that myeloid IKKb has a protective role in controlling bacterial growth that leads to overwhelming sepsis during tularemic infection. Furthermore, macrophage polarization vs. bacterial control and survival appear to be functionally distinct processes regulated by myeloid IKKb. In addition, we also found, in our model system, that IKKb-dependent macrophage polarization is a short-term phenomenon. Contrary to this, we found no evidence of anti-inflammatory properties for myeloid IKKa in vivo.