Loss of Murine FOXO3 in Cells of the Myeloid Lineage Enhances Myelopoiesis but Protects from K/BxN-Serum Transfer-Induced Arthritis

FOXO transcription factors have a highly conserved role in regulating transcription of genes involved in differentiation, cell cycle arrest, apoptosis and DNA repair. Loss of FOXO3 in mice has previously been shown to result in a myeloproliferative disease. In agreement with this, we found that an independent Foxo3 null mouse strain, Foxo3 Kca, exhibits an increase in neutrophils in the spleen, bone marrow and blood. This coincides with an expansion of myeloid progenitor cells including pre-granulocyte-macrophage progenitors (Pre-GMs) and granulocyte-macrophage progenitors (GMPs). Surprisingly, despite neutrophilia, the severity of passive serum transfer arthritis was markedly attenuated in Foxo3 Kca mice. These defects appear to be at least partially intrinsic to the myeloid lineage, as deleting Foxo3 specifically from myeloid cells using LysMCre also leads to an elevated number of neutrophils and protection from K/BxN-serum transfer-induced arthritis.


Introduction
FOXO transcription factors integrate a wide variety of signals to control diverse physiologic processes such as differentiation, cell cycle arrest, apoptosis, metabolism, and detoxification of reactive oxygen species (ROS) [1]. Additionally, FOXO transcription factors have been implicated in the control of hematopoiesis [2]. Acute disruption of the genes encoding all three of the peripheral Foxo transcription factors (Foxo1, Foxo3 and Foxo4) has broad consequences on hematopoiesis including: reduced size of the hematopoietic stem cell (HSC) and lymphoid progenitor compartments, but increased myeloid colony formation potential and number of neutrophils in the spleen. The loss of HSCs in this model was found to be associated with increased ROS, leading to increased cycling and apoptosis of HSCs, and was corrected by administration of an antioxidant [2].
More recently loss of FOXO3 alone has been shown to alter neutrophil homeostasis. In a mouse strain with a null allele of Foxo3 (generated by disrupting exon 2 of the Foxo3 gene with a neomycin cassette), denoted here as Foxo3 AH , there is decreased HSC maintenance that correlates with increased intracellular ROS. Additionally, these mice were found to have a trend towards an increased frequency of neutrophils in the bone marrow [3]. The increase in neutrophils seen in these mice is reminiscent of our results that revealed splenic neutrophilia in two other independent germline Foxo3 null strains, Foxo3 Kca (generated using embryonic stem (ES) cell clones from the OmniBank(R) ES cell library of randomly targeted cell lines and backcrossed to the C57BL/6 strain [4]), and Foxo3 -/-(produced by loxP targeting of FVB ES cells and germline excision with Ella-Cre [5,6]). Further studies demonstrated that Foxo3 -/mice have a mild myeloproliferative disease with splenic neutrophilia, an increased frequency of at least one type of myeloid progenitor cell, and enhanced extramedullary hematopoiesis [7]. Elevated levels of ROS in Foxo3 -/progenitor cells lead to reduced expression of Lnk, a negative regulator of cytokine signaling, which in turn contributes to the hyperresponsiveness of Foxo3 -/progenitors to cytokine and thus increased myelopoiesis [7].
In contrast, yet another independent Foxo3 mutant strain, Foxo3 Trap (a mutant allele generated by 'gene-trap' technology of exons 2-3 in 129 mice), has been reported to have a normal level of neutrophils in the blood, but decreased accumulation of neutrophils at sites of inflammation. This decrease in neutrophil accumulation was found to be due to increased apoptosis of Foxo3 Trap neutrophils under inflammatory conditions. Thus Foxo3 Trap mutant mice exhibit protection from a neutrophil-dependent arthritis model, the K/BxN passive serum transfer model [8]. Furthermore, FOXO3 has been found to be overexpressed in polymorphonuclear cells from patients with rheumatoid arthritis [9]. Thus, FOXO3 has been suggested as a target for arthritis therapy [10,11]. However, some phenotypes of the Foxo3 Trap mutant strain are substantially different than that of other Foxo3 null strains, a characteristic we previously attributed to genetic background differences [4,6,12]. However, a SNP in FOXO3 resulting in increased expression of FOXO3 in monocytes during inflammatory conditions is associated with reduced disease severity in patients with rheumatoid arthritis (RA) [13].
Taken together these studies suggest a role for FOXO3 in restricting myelopoiesis and thus the number of neutrophils by limiting ROS in progenitors. However, neutrophil accumulation is reduced in Foxo3 Trap mice leading to protection from a model of arthritis dependent on neutrophils [8]. As altered expression of FOXO3 has been implicated in the severity of RA [9,13], it is important to reconcile these apparent differences. Thus we investigated myelopoiesis, neutrophil accumulation during peritoneal inflammation, and susceptibility to K/BxN-serum transfer-induced arthritis in a single strain of C57BL/6-congenic Foxo3 null mice.
In this study we show that Foxo3 Kca mice exhibit an increased number of neutrophils in the spleen, bone marrow and blood. This neutrophilia coincides with an expansion of myeloid progenitor cells including pre-granulocyte-macrophage progenitors (Pre-GMs) and granulocytemacrophage progenitors (GMPs). However, the increase in neutrophils in the steady state did not result in augmented inflammation, as the severity of K/BxN-serum transfer arthritis was markedly decreased. Neutrophilia and protection from K/BxN-serum transfer arthritis were also seen in Foxo3 f/f LysMCre + mice. We conclude that FOXO3 contributes to the control of neutrophil homeostasis and effector function.

Results
Foxo3 Kca mice exhibit neutrophilia Several Foxo3 mutant strains have been used to study different aspects of neutrophil development and function. FOXO3 has been shown to limit myelopoiesis and/or the number of neutrophils in most of these strains. Paradoxically, Foxo3 Trap mice were found to be resistant to inflammation in two neutrophil-dependent inflammatory models including K/BxN-serum transfer arthritis. However, the Foxo3 Trap strain has been shown to differ from other Foxo3 null strains in some phenotypes. Thus it was important to study both neutrophil development and function in the same strain. To do this we chose to further investigate the role of FOXO3 in neutrophil development and function using C57BL/6 congenic Foxo3 Kca/Kca (referred to from here on as Foxo3 Kca ) mice, one of the two Foxo3 null strains we have previously shown to have an increased number of LY6G hi CD11B hi cells in the spleen [6]. We first confirmed that LY6G hi CD11B hi cells are increased in the spleen of Foxo3 Kca mice (Fig 1A and [6]).
LY6G hi CD11B hi cells are often presumed to be neutrophils; however, until recently very little has been known about splenic neutrophils [14,15], and even less is known about the regulation of the size of this population in the absence of overt disease [16]. Thus we further investigated these cells. Neutrophils are characterized by the presence of dense granules, and to examine this we used light side scatter (SSC), a measure of granularity, to compare splenic LY6G hi CD11B hi cells to total splenocytes [17]. Consistent with high expression of LY6G and CD11B as markers of neutrophils these cells are SSC hi compared to total splenocytes ( Fig 1B). Splenic resident neutrophils have recently been identified as B cell-helper neutrophils in both mice and humans [15]. B cell-helper neutrophils have elevated expression of CD24 and CD40L and diminished expression of CD62P and ICAM-1 in comparison to circulating blood neutrophils [15]. The LY6G hi CD11B hi cells from the spleen of WT and Foxo3 Kca mice display similarly elevated levels of CD24 and CD40L and reduced levels of CD62P and ICAM-1 in comparison to blood neutrophils ( Fig 1C). Finally, we enriched for LY6G hi cells from the spleens of WT and Foxo3 Kca mice using magnetic bead selection and stained the purified cells with Giemsa. The presence of intracellular Giemsa-stained granules confirmed these cells are indeed neutrophils (Fig 1D). Neutrophil frequency and number are also increased in the bone marrow and blood of Foxo3 Kca mice ( Fig 1E).

Foxo3 Kca mice have increased numbers of myeloid progenitor cells
Recently the Foxo3 null strain Foxo3 -/was found to have a mild myeloproliferative syndrome with an increased number of at least one myeloid progenitor cell type in the bone marrow and increased hematopoietic activity in the bone marrow and spleen [7]. Our analysis shows that a similar myeloid progenitor cell population (myeloid progenitors, PI -Lin -cKIT + SCA1 -) is also expanded in the bone marrow and spleen of Foxo3 Kca mice (Fig 2A-2C). Using a staining approach similar to that described by Pronk and colleagues ( Fig 2B and [18]), we further analyzed the bone marrow and spleen of WT and Foxo3 Kca mice to more precisely determine the myeloid precursor cell types expanded in Foxo3 Kca mice. This analysis revealed that Foxo3 Kca mice have an elevated number of pre-granulocyte-macrophage progenitors (Pre-GM, PI -Lin -cK-IT + SCA1 -CD41 -FCGRII/III -CD105 -CD150 -) in the bone marrow and spleen ( Fig 2D). Interestingly, granulocyte-macrophage progenitors (GMP, PI -Lin -cKIT + SCA1 -CD41 -FCGRII/ III + CD150 -) are increased in the spleen but not the bone morrow of Foxo3 Kca mice ( Fig 2E). Thus, GMP numbers in the bone marrow may be controlled by additional homeostatic regulation mechanisms. In sum, our data support the conclusion by Yalcin et al. that loss of FOXO3 leads to an expansion of myeloid precursor cells and myeloproliferation [7].

Foxo3 Kca neutrophils accumulate to a normal level in a model of peritonitis
In contrast to the increased number of neutrophils in naïve Foxo3 Kca mice, Foxo3 Trap mice have reduced neutrophil accumulation in a model of peritonitis [8]. Thus, we wished to  determine the response of Foxo3 Kca mice to i.p. injection of proteose peptone, a method commonly used to induce neutrophil recruitment to the peritoneum. At 3 hr post proteose peptone injection the proportion of neutrophils in the spleen of Foxo3 Kca mice was still increased compared to similarly treated control mice (Fig 3A). However, the splenic neutrophilia seen during peritoneal inflammation in Foxo3 Kca mice was not due to an inability of these mice to sense the inflammatory stimuli and start responding, as Foxo3 Kca mice had a high proportion of neutrophils in the blood ( Fig 3B) and a slight, but significant increase in the number of neutrophils in the peritoneum at this timepoint ( Fig 3C). However, the number of neutrophils in the peritoneum of Foxo3 Kca and WT mice was not different 6 hr post proteose peptone injection (Fig 3D), the timepoint our preliminary analysis showed to be the peak of neutrophil accumulation (data not shown). The number of peritoneal macrophages was not altered by the absence of FOXO3 (Fig 3E).
The reduced accumulation of Foxo3 Trap neutrophils during peritonitis was found to be due to aberrant expression of FASL leading to apoptosis [8]. Thus we used flow cytometry to determine FASL expression on Foxo3 Kca peritoneal cells 6 hr post proteose peptone injection. As expected peritoneal B1a B cells expressed high levels of FASL. However, FASL expression was low on both WT and Foxo3 Kca peritoneal neutrophils (Fig 3F).
Foxo3 Kca mice are resistant to K/BxN-serum transfer arthritis Neutrophils are essential for the effector phase of passive K/BxN-serum transfer arthritis [19,20]. In this model sera from K/BxN mice containing IgG specific for the glycolytic enzyme glucose-6-phosphate isomerase are injected into mice to initiate disease [21][22][23]. Foxo3 Trap mutant mice were found to be resistant to this model of arthritis, and this was attributed to increased apoptosis of Foxo3 Trap neutrophils during inflammation [8]. Therefore, we wished to determine if Foxo3 Kca mice normally develop K/BxN-serum transfer arthritis. WT and Foxo3 Kca mice were injected i.p. with K/BxN sera on d 0 and d 2 and the progression of arthritis was observed. In WT mice ankle thickening and clinical signs of inflammation were evident by d 3 and remained elevated for more than 2 weeks (Fig 4). However, Foxo3 Kca mice were relatively protected from paw swelling and clinical signs of inflammation induced by K/BxN sera (Fig 4).

Neutrophilia and protection from arthritis in the absence of FOXO3 is myeloid cell-intrinsic
To determine if the increase in neutrophils observed in the absence of FOXO3 is myeloid cellintrinsic, we generated Foxo3 f/f LysMCre + mice in which Foxo3 is specifically deleted from cells of the myeloid lineage. Similar to Foxo3 Kca mice, Foxo3 f/f LysMCre + mice exhibit an increased frequency and number of LY6G hi CD11B hi neutrophils in the spleen (Fig 5A). However, despite the apparent increase in myelopoiesis, monocytes are not elevated in the blood of Foxo3 f/f LysMCre + mice (Fig 5B).
We next wished to determine if loss of FOXO3 in myeloid cells alone offers protection from K/BxN-serum transfer arthritis. As was seen with Foxo3 Kca mice treated with proteose peptone, splenic neutrophils were slightly increased in Foxo3 f/f LysMCre + mice d 3 post injection of K/BxN sera, a time point when joint inflammation was just becoming evident in WT mice (Fig 5C). Also at d 3 WT and Foxo3 f/f LysMCre + mice had a high, but similar frequency of  neutrophils in the blood, suggesting that both WT and Foxo3 f/f LysMCre + mice were responding to inflammatory cues (Fig 5D). Importantly, the ankle thickening and joint inflammation induced by K/BxN sera were dramatically diminished in Foxo3 f/f LysMCre + mice compared to control mice (Fig 5E). In sum, our data show that FOXO3 acts at least in part in a myeloid lineage-specific manner to regulate neutrophil homeostasis and function.

Discussion
FOXO transcription factors have a highly conserved role in regulating metabolism, cell cycle arrest, apoptosis, and stress resistance [1]. They are also critical mediators of HSC survival and quiescence by limiting oxidative stress [2]. Additionally, FOXO transcription factors have been shown to have very specialized functions in lymphoid cells [24,25]. Moreover, several studies have now reported that FOXO3 is important for neutrophil homeostasis. Overall these studies support a role for FOXO3 in limiting the number of myeloid progenitor cells and thus neutrophils [3,6,7,26]. However, an independent Foxo3 mutant strain, Foxo3 Trap , which has been reported to have some key phenotypic differences compared to other Foxo3 null strains, was used to investigate the role of FOXO3 in neutrophils during inflammation [4,6,8,12]. Thus it was unclear if the resistance to neutrophil mediated inflammation seen in the Foxo3 Trap strain was consistent with the neutrophilia and increased myelopoiesis observed in several other Foxo3 null strains [3,[6][7][8]26].
In this study we used the Foxo3 Kca null strain, which we have previously reported to have an increase in the number of splenic neutrophils [6], to investigate the role of FOXO3 in neutrophils in the steady state and during inflammation in a single study. We show that Foxo3 Kca mice have an increased number of neutrophils in the spleen, bone marrow, and blood and this correlates with an increase in myeloid progenitor cells in the spleen and bone marrow. We further show that despite the elevated number of neutrophils in Foxo3 Kca mice, this strain is also resistant to K/BxN-serum transfer arthritis. As has been previously suggested, these phenotypes are at least partially myeloid lineage-intrinsic, as mice lacking Foxo3 specifically in myeloid cells, also, have an elevated number of neutrophils in the spleen and are protected from K/BxN-serum transfer arthritis [7,8]. Thus the increase in neutrophils in the steady state of (filled gray histogram) and Foxo3 Kca peritoneal B1a cells (B220 + CD5 + , dotted gray line) 6 hr post injection of 10% proteose peptone.
doi:10.1371/journal.pone.0126728.g003 several Foxo3 null strains is not in conflict with the protection that loss of FOXO3 provides from a neutrophil-dependent model of inflammation.
Our work also serves to confirm the study by Yalcin et al. that showed using the Foxo3 -/strain that loss of FOXO3 results in an elevated number of neutrophils due to increased  [7]. Foxo3 Kca mice have increased numbers of Pre-GMs and GMPs. However, while myeloid progenitor cells and neutrophils were elevated in the absence of Foxo3, we did not observe an increase in the numbers of monocytes or macrophages. Likewise, GMPs were elevated in the spleen but not in the BM of Foxo3 Kca mice, suggesting that additional mechanisms may be involved in regulating BM myelopoiesis and the homeostasis of monocytes and macrophages.
A seeming contradiction of this work is that there is an increase in neutrophils in the steady state, but no change in the number of neutrophils that accumulate in the peritoneum during peritonitis. Furthermore, despite the elevated number of neutrophils Foxo3 Kca mice are protected from disease in the K/BxN-serum transfer model of arthritis, a model in which neutrophils play a significant role in the ensuing inflammation and pathogenesis [19,20]. And this is despite the fact that increased extramedullary myelopoiesis has been shown to be involved in the pathogenesis of other autoimmune diseases [27]. However, like FOXO1 in T cells [25,28,29], our preliminary data suggest FOXO3 may regulate the expression of homing molecules on neutrophils. Characterization of several strains of mice lacking molecules involved in neutrophil migration to tissues has shown that reduced neutrophil homing results in increased myelopoiesis [30,31]. Future studies should investigate effector functions and homing to the joint of neutrophils lacking FOXO3. Alternatively, it remains possible that in specific inflammatory settings loss of FOXO3 leads to increased FASL expression and thus increased apoptosis.
Extramedullary hematopoiesis is elevated in tumor-bearing mice and humans, and contributes to the generation of tumor associated macrophages and neutrophils [16]. It is tempting to speculate that in some cases chronically increasing extramedullary myelopoiesis, such as in tumor-bearing mice or during chronic infections, might lead to increased generation of immunosuppressive myeloid cells including bonafide myeloid-derived suppressor cells (MDSCs). Such a phenomenon might serve as part of a negative feedback loop to reduce immune responses and thus decrease tissue damage. Could the increased extramedullary myelopoiesis in Foxo3 Kca mice be increasing the number of immunosuppressive myeloid cells? Interestingly, similarly to MDSCs [32], splenic B cell-helper neutrophils have been shown to be able to suppress CD4 T cell proliferation [15]. Our analysis showed that splenic neutrophils from WT and Foxo3 Kca mice have a phenotype similar to that reported for human B cell-helper neutrophils including increased expression of CD24 [15]. Additionally, CD24 has been shown to be able to limit inflammatory responses [33], further suggesting that the splenic neutrophils that are increased in the absence of FOXO3 may be able to act as suppressor cells. However, as Foxo3deficient neutrophils were still elevated in the spleen during times of peripheral inflammation, it is unclear if this population substantially contributes neutrophils to sites of acute inflammation.
In contrast to the protection from K/BxN-serum transfer arthritis provided by a myeloid lineage-specific deletion of Foxo3, it has recently been shown that a noncoding SNP (FOXO3 (rs12212067: T>G)) leads to increased expression of FOXO3 in monocytes during inflammation, limits monocyte inflammatory responses, and is associated with reduced severity of RA [13]. Thus FOXO3 likely has a multi-faceted role during the progression of RA with opposing affects in different cell types and/or phases of RA.

Materials and Methods Mice
Mice were maintained in a specific-pathogen free vivarium. All experiments were approved by the Institutional Animal Care and Use Committee of University of California, San Diego (USDA Animal Research Facility Registration Number: 93-R-0437). All efforts were made to minimize suffering. C57BL/6J mice were obtained from the Jackson Laboratory. Foxo3 Kca backcrossed to C57BL/6 have been previously described [6]. Foxo3 f/f mice, which have previously been described [5], were backcrossed to C57BL/6 (The Jackson Laboratory, Bar Harbor, ME) for at least seven generations and then crossed to LysMCre + mice (The Jackson Laboratory; [34]) to generate Foxo3 f/f LysMCre + mice. KRN T cell receptor (TCR) transgenic mice were a gift from Drs. D. Mathis and C. Benoist (Harvard Medical School, Boston, MA) and Institut de Génétique et de Biologie Moléculaire et Cellulaire (Strasbourg, France) [35]. Sera from arthritic adult K/BxN mice were pooled.

Flow cytometry
Antibodies were purchased from BD Biosciences, BioLegend, eBioscience and Invitrogen. Apoptotic cells were stained with 7AAD (Sigma) or propidium iodide (Molecular Probes). The femur and tibia of both legs were used for all bone marrow samples. For progenitor staining, cells were first stained with fluorescent probe-conjugated anti-CD16/32, followed by staining with a cocktail of biotinylated mature blood cell lineage markers (Lin: B220, CD11B, CD3E, and LY6G). For analysis of these stains 2.5 million cells were collected by flow cytometry. In other experiments Fc receptors were blocked using 2.4G2 hybridoma culture supernatant. Blood samples were obtained through retro-orbital or tail bleeding and collected in heparin-or EDTA-coated tubes. Red blood cells (RBCs) were lysed with RBC lysis buffer. In some experiments RBCs were removed using dextran T-500 prior to RBC lysis. All samples were analyzed on BD Biosciences FACSCalibur or LSRFortessa cytometers. Data were analyzed using the FlowJo software (Tree Star).

Proteose peptone-induced peritonitis model
Peritonitis was induced as previously described with slight modifications [36]. Briefly, 1 ml of 10% proteose peptone was administered i.p. Peritoneal cells were harvested at 3 or 6 hr with 10 ml of pre-chilled PBS.

K/BxN-serum transfer arthritis model
To induce arthritis 75 μl of pooled K/BxN sera was administered i.p. on d 0 and d 2. Ankle thickening was observed by measuring the malleoli using an electronic caliper (Mitutoyo Corporation). The clinical score of each paw on a scale of 0 to 3 was determined in a blinded fashion and the clinical index was determined by adding the scores for each paw together, as previously described [37]. For experiments in which neutrophils were analyzed in the blood and spleen K/BxN sera were injected i.p. only on d 0.
Neutrophil purification and microscopy LY6G + enriched splenocytes were isolated using a two-step magnetic separation. Briefly, splenocytes were labeled for negative selection with biotinylated antibodies to B220, CD4, CD8, CD11C, DX5, MHC class II, and TER119 (eBioscience) and streptavidin-microbeads (Miltenyi Biotec). After enrichment using the Automacs (Miltenyi Biotec) the negative fraction was labeled with a biotinylated antibody to LY6G (eBioscience) and streptavidin-microbeads (Miltenyi Biotec) and again separated using the Automacs (Miltenyi Biotec). Flow cytometry was used to assess the purity of the LY6G + fraction, and the remaining cells were spun onto slides with a cytospin centrifuge (Harlow Scientific) and stained with Giemsa (Ricca Chemical Company). Cells were then observed with an Olympus confocal microscope at 100x magnification.

Statistical analysis
The statistical significance was determined using GraphPad Prism 4 or 6 software. When appropriate statistical outliers were identified by the ROUT method. Two-way mixed model ANOVA with Bonferroni post-hoc tests was used to determine significance of K/BxN arthritis experiments. For all other experiments the statistical significance was calculated by unpaired, two-tailed Student t-test using GraphPad Prism software ( Ã p < 0.05, ÃÃ p < 0.01, ÃÃÃ p < 0.001).