Age-matched, 3-week-old female Balb/c weanling mice were purchased from Envigo (Houston, Texas). Malnourished mice were fed for 28 days with a polynutrient deficient diet, containing 3% protein, low zinc, and iron (TD.99075, Envigo), and the control group (well-nourished) was fed with a nutrient-sufficient diet of 16.9% protein, normal zinc, and iron (TD.99103, Envigo) as described (38, 40). Lipid in the two mouse chows was supplied by corn oil (9%) and the diets were made isocaloric by varying the carbohydrate content. Mice were pair-fed with the malnourished mice being provided 90% (by weight) of the food consumed by the control group. For studies of the effect of dietary fish oil, the 9% corn oil in the polynutrient deficient diet was replaced with 9% fish oil (TD.160285, Envigo).

After 28 days on the control or nutrient-deficient diet, mice were challenged with LPS (4mg/kg intraperitoneally) from Escherichia Coli O111B4 (Sigma) as a model of Gram-negative bacterial sepsis (42) and evaluated 24h later. Mice challenged with Leishmania donovani were infected intradermally (ID) in the dorsum of the foot with 10⁶ purified L. donovani metacyclic promastigotes and evaluated 72h later as described (40). To study the effect of inflammatory monocytes, we used a Cre-inducible diphtheria toxin (DT) expression model (49). In brief, 6–8-week-old female and male offspring were obtained from B6.129P2-Lyz2tm1(cre)Ifo/J (Jackson stock # 004781) crossed with C57BL/6-Tg (Csf1r-HBEGF/mCherry)1Mnz/J (Jackson stock #024046). Offspring carrying the Csfr1-LysMcre-DT phenotype and controls were placed on the experimental diet for 28 days. To deplete inflammatory monocytes, mice were given 5 ng of DT (Sigma) intraperitoneally, per gram body weight every 24h.

Mouse sternums were snap-frozen in cryomolds with OCT using an isopentane bath. Sternums were stored at -80C until analysis. Longitudinal cryosections (6 µm) were obtained in a Leica cryostat and were stained with lipid stain solution Oil-red O (Abcam) according to manufacturer instructions. After Oil-red O staining, slides were double-stained with alkaline phosphatase ALP (Takara) to differentiate osteoblasts. Osteocytes were identified by location in the trabecular bone, characteristic stellate cell morphology, and the expected expression of alkaline phosphatase. To identify osteoclasts, slides were stained for Tartrate-resistant acid phosphatase (TRAP) (TRACP kit, Takara). Osteoclasts were further identified as multinucleated cells following the manufacturer’s instructions. Images were captured with 16X Apochromat NIR objective lenses in a Nikon microscope (Elipse 80i) with a NIS-elements BR™ Software. Images, of at least 20 fields per group, were further analyzed with the software Image Processing and Analysis in Java Software (ImageJ) and QuPath-02.3. Image analysis in Image J was done according to software instructions. In brief, TIF color images were uploaded and converted to 8-bit. The background was subtracted and the pixel value was calibrated to length in microns and measurements were set (area). The threshold of the positive signal was adjusted and applied globally to all images to compare. The percentage of the positive area was determined in the images covering the total length of each sternum. The number of positive cells was determined with the tool for particle analysis (size adjusted 0-1). Image analysis in QuPath was done as described (50).

Bone marrow cells were obtained from the femurs of mice. Femurs were carefully broken in a ceramic grinder and treated for 10 min with Accutase solution (Millipore) to release bone marrow cells. After collecting and washing the cells, they were adjusted to 10⁶ per tube. Cells were blocked with 0.5 µg of Fc block and stained with a viability marker. Cells were identified with the following markers: inflammatory monocytes: CD11b+, CD11c-, Ly6G-Ly6C^int/hi+ (51); T cell inducers of osteoclasts: CD3+RANKL+; osteoclast precursors: CD11b^-/low CD45+ckit+(CD117+) CD115+ (52); mesenchymal stem cells: CD45-CD29+CD105+ (R&D, Mesenchymal stem cell markers) (Table S1). After staining, cells were washed and treated with a 1-step fix/lyse solution (eBioscience). Intracellular cytokines were determined after 2-4 h of incubation of 10⁶ bone marrow cells with 1µg/mL of Golgi stop (BD). Fixed cells were acquired with a benchtop Stratedigm flow cytometer (S500) or BD LRSFortessa flow cytometer (BD Biosciences) and analyzed with FlowJo (V10). Panels were designed to minimize overlapping fluorescence between antibodies with FluoroFinder (Panel Builder 3.0) online platform. Thresholds for gating were established with FMO controls. The gating strategy that was followed is shown in Supplementary Figure 1A–E.

RNA from bone marrow cells was isolated with the RNeasy kit (Qiagen) and treated with DNase I (Turbo DNase, Ambion) followed by reverse transcription (High-capacity Reverse transcription kit, Thermofisher Sci.). Primers to amplify target genes were selected from the PrimerBank database (Wang, Spandidos et al., 2012) (Table S2). Targets were amplified by real-time PCR (qPCR) and detected with SYBR Green dye (Bio-Rad). The fold change of gene expression was calculated with the delta CT method and 18S as the reference gene. The fold-change was calculated relative to the value in uninfected naïve mice as specified in each figure legend. Inflammasome activation was assessed with Caspase Glo-1 inflammasome kit (Promega) according to manufacturer instructions. Bone marrow cells from Csfr1-LysMcre-DT mice, selected as above, were plated in 96-well white luminometry plates (50,000 cells/well) and were exposed in vitro for 1h to LPS or Leishmania antigens (Freeze/thaw parasites) for luminometry reading.

Bone marrow cells were cultured in medium containing high glucose DMEM with 10% heat-inactivated fetal bovine serum (FBS), 1 mM sodium pyruvate (Gibco), 1X MEM amino acids solution (Sigma), 10 mM HEPES buffer (Cellgro) and 100 IU/mL Penicillin/100mg/mL Streptomycin (Gibco). Cells were added to 24-well plates containing round glass coverslips at a concentration of 0.5 x 10⁶ cells per well in 300 µL medium. To differentiate osteoclasts, 40 ng/mL RANKL and 20 ng/mL macrophage colony-stimulating factor (M-CSF) (Biolegend) were added to the culture media. The culture medium was changed at 72h of culture and cells were evaluated after 7 days of differentiation at 37°C, 5% CO₂. Cells were fixed and stained with TRAP as above. Multinuclear mature osteoclasts were identified by nuclear staining with Methyl green (TRACP kit, Takara) or 7AAD and were evaluated by microscopy or flow cytometry respectively (53). For some ex vivo cultures the following were added: 100ng/mL DHA (Docosahexaenoic acid (Sigma) diluted in fatty acid-free Bovine Serum Albumin (BSA) at 10:1 ratio; 10ng/mL LPS; 6mg/mL homogenates from inguinal white fat adipose tissue; Leishmania donovani antigen (1:1 thaw-frozen promastigotes), or vehicle diluent for control wells.

Data were analyzed with GraphPad Prism 9.0, Software (Graphpad, LLC) according to software recommendations. Comparisons between 2 groups were evaluated with two tail Mann-Whitney U test for non-parametric data or two tail unpaired t test for normally distributed data. Comparisons between more than 2 groups were evaluated with Kruskal-Wallis for non-parametric data or ANOVA for normally distributed with post hoc correction for multiple comparisons (Bonferroni or Tukey). Differences in proportions were assessed with the Chi-squared test. Data from histological slides were collected from at least 20 randomly selected microscopic fields. In vitro experiments were performed with at least four replicates per condition. In vivo experiments were performed with 4-5 mice per group. Experiments were repeated at least twice.

We evaluated the effect of moderate acute malnutrition (diet deficient in protein, iron, and zinc) on body weight and length, and bone length. As we demonstrated previously, the malnourished mice had a flat growth curve so that when the control mice reached a mature weight 28 days after weaning to a normal diet, the malnourished (MN) mice had approximately 30% less body weight and a 15% reduction in body length (Figure 1A). The disproportionately reduced weight over length (Figure 1B) is characteristic of acute malnutrition. Malnourished mice also showed significant reductions in the length of the femur (Figure 1C). Body length and femur length were strongly correlated in control mice but not in malnourished mice (Figure 1D). This suggests that bone health may be uniquely impacted by acute malnutrition.

We evaluated the effect of inflammatory or infectious challenges on bone marrow cellularity in malnourished mice. Despite the femur length of the malnourished mice being smaller in the basal state, there was only a slight reduction in the total number of bone marrow cells recovered from malnourished and control mice (Figure 2A). When the mice were challenged with intradermal LPS there was an insignificant reduction in total bone marrow cells in both control and MN mice (Figure 2A). Challenge with intradermal L. donovani infection led to increased bone marrow cellularity in control mice, but this response was absent in MN mice. (Figure 2A). However, histopathological analysis of bone marrow cryosections revealed a striking difference in the cellular composition of the bone marrow in control and MN mice at baseline and following inflammatory challenge. At baseline, MN mice had significantly more lipid-containing (oil-red O positive) cells in the bone marrow hematopoietic centers compared to control mice (Figure 2B). Seventy-two hrs after challenge with or intradermal L. donovani (Figure 2C) or intraperitoneal LPS (Figure 2D) the difference in the number of fat cells in the bone marrow of MN and control mice was even more striking. Histopathological analysis confirmed that malnourished mice challenged with L. donovani or LPS had a significant expansion of the adipose area in the bone marrow compared with non-challenged malnourished controls (Figure 2E). When the oil-red O positive area in bone marrow was analyzed, LPS-induced lipid accumulation was amplified in MN mice. However, bone marrow adiposity was enhanced in the L. donovani challenge irrespectively of the nutritional status.

BM mesenchymal cells (MSC) can differentiate into either adipocytes or osteoblasts (54). Adipocytes in bone marrow can be identified by staining intracellular lipid with the lipophilic dye Oil-red O and osteoblasts and osteocytes are identified by staining for Alkaline Phosphatase (ALP). Although the presence of lipid droplets is often used as the defining characteristic of adipocytes, the storage of lipid droplets within cells of other lineages is well documented. In fact, lipid storage within osteoblasts or cells differentiating within the osteoblast lineage also occurs (54). Lipid accumulation may contribute to the energy needs of the cell but can also compromise bone mineral density if it exceeds the energy requirements of the cell (54). Because the cells identified in these studies as being oil-red-O positive have not been fully characterized we refer to them as lipid-filled or fat-containing cells rather than definitive adipocytes.

We investigated the balance of MSCs, lipid-filled cells, and osteoblasts in malnourished mice with and without inflammatory challenge. We found that the proportion of MSCs was not altered in naive (unchallenged) MN mice (Figure 3A), however, it increased significantly in control and malnourished mice 24h after challenge with LPS (Figure 3A; Supplementary Figure 2B). The LPS-induced increase in MSCs was amplified in MN mice, but the proportion of MSC was constant in L. donovani-challenged mice (Figure 3A). We then evaluated osteoblasts and osteocytes, which derive from MSCs, following exposure to the inflammatory stimulus. We found that the bone marrow area occupied by ALP+ cells (osteoblasts) was not altered by LPS challenge in either control or MN mice but L. donovani challenge increased the area occupied by ALP+ cells in control mice (Figure 3B). Expression of genes involved in osteogenesis, such as Dentin matrix acidic phosphoprotein 1 (Dmp1), Runt-related transcription factor 2 (Runx2), Osteocalcin (Bglap), and Osterix (Osx), showed a trend toward a decrease in LPS-challenged MN mice compared with naïve MN mice, but this difference was not observed in well-nourished mice (Supplementary Figures 3A, B). This support that a lack of osteoblast differentiation in the challenge infection is a consequence of malnutrition. The osteogenesis expression in malnourished naïve mice was variable but was particularly vulnerable to inhibition by the challenge (Supplementary Figures 3A, B).

To determine the effect of malnutrition and inflammatory challenge on the balance of cells differentiated from the MSC lineage we double-stained bone marrow sections with oil-red O (adipogenesis) and ALP (osteogenesis). We observed that osteoblasts/osteocytes (ALP+) in trabecular bone contained more lipid in malnourished LPS-challenged mice (Figure 3C). L. donovani infection enhanced the lipid content of osteocytes independently of the nutritional status (Figure 3C).

Since the increase in MSCs was paralleled by increased cellular lipid content, we considered that MSC may be driven to differentiate toward adipocytes. In support of this, we found that the BM from mice challenged with LPS showed increased expression of genes involved in adipogenesis, peroxisome proliferator-activated receptor gamma (Pparg), sterol regulatory element-binding protein 1 (Srebp1), Adiponectin (Adipoq) compared to unchallenged mice (Figure 4A ). Following LPS challenge the increased expression of genes Pparg and Srebp1 was amplified in MN mice compared to well-nourished (WN) mice (Figure 4A). L. donovani challenge followed a similar trend but a synergistic effect of malnutrition and L. donovani challenge was not significant ( Figure 4B). We also found that the Bone marrow production of the adipokine, adiponectin, was also increased in both LPS- and L. donovani-challenged MN mice (Figure 4C).

Osteoclasts are generated by differentiation from BM myeloid progenitors, which are derived from hematopoietic stem cells (55). Since osteoclasts antagonize the function of osteoblasts by participating in bone resorption, we evaluated the effect of malnutrition in osteoclast activation using Tartrate-Resistant Acid Phosphatase (TRAP) as the marker of osteoclasts. We found that LPS-induced TRAP activation of osteoclasts in malnourished mice ( Figure 5A ). Likewise. the proportion and the total number of mature multinuclear osteoclasts were significantly increased (Supplementary Figure 2A). A similar trend of TRAP activation was observed in L. donovani challenged mice, but the difference was not significant. To confirm these results, we differentiated ex vivo bone marrow cells from control and MN mice to osteoclasts with Rankl and Csf-1 and exposed them to LPS or L. donovani antigen. We found increased osteoclast activation (TRAP+ cells) in the cells isolated from malnourished mice whether or not they had been challenged with LPS or L donovani (Figure 5B). Since the expansion of BM osteoclasts correlated strongly with the expansion of BM adiposity in LPS-challenged malnourished mice (Figure 5C), we exposed bone marrow cells from control mice to lysates from white fat adipose tissue collected from control or malnourished mice. Exposure to fat cell lysates increased the generation of TRAP+ osteoclasts irrespective of the nutritional status of the lipid donor (Figure 5D). However, exposure to malnourished fat resulted in increased numbers of mature multinuclear osteoclasts (Figure 5E). To further support the finding that osteoclastogenesis is promoted by malnutrition, we evaluated the expression of genes involved in osteoclastogenesis, Fos Proto-oncogene (Fos), cathepsin (Ctsk) and Receptor Activator of NF-κB Ligand (Rankl) in malnourished mice with and without inflammatory challenge. We found that LPS-induced Fos and Ctsk expression was amplified in malnourished mice whereas Rankl expression was increased equally in WN and MN mice (Figure 6A). In contrast, L. donovani challenge induced Fos and Ctsk, but not Rankl, irrespective of the nutritional status (Figure 6B). We further analyzed the cell surface expression of RANKL by flow cytometry in the bone marrow since it is critical in the differentiation of bone-resorbing osteoclasts (31, 55). We found expansion of non-hematopoietic (CD45^_) cells expressing RANKL in the bone marrow of naïve and challenged MN mice (Figure 6C ; SFigure 2C). We also found that LPS-induced Rank+ T cell expansion (CD3+Rankl+ cells) in MN mice (Figure 6D ; SFigure 2D, E). Additionally, we found that the percentage of osteoclast myeloid precursors was increased after LPS challenge in both WN and MN mice, but the increase was greater in malnourished mice (Figure 6E ; SFigure 2F). Collectively, these data suggest that the bone marrow of MN mice exposed to an inflammatory challenge has increased differentiation of MSCs toward lipid-filled or adipocyte-like cells and that the adipogenic bone marrow environment is associated with osteoclast expansion/activation.

We evaluated the expression of bone marrow inflammatory cytokines (Il6, Il1b, and Tnf) that could have a role in shaping bone marrow adipogenesis and osteoclastogenesis in malnourished mice. The expression of Il1b and Tnf following LPS challenge was amplified in malnourished mice (Figure 7A), whereas malnutrition blunted LPS-induced Il6 expression (Figure 7A). L. donovani challenge upregulated Il1b and Tnf but this effect was not affected by nutritional status (Supplementary Figure 4). Additionally, LPS-challenged MN mice had a significantly higher percentage of inflammatory monocytes in the bone marrow compared to LPS-challenged control mice (Figure 7C ; Supplementary Figure 2G). This difference was not seen following L. donovani challenge. The percentage of cells that expressed inflammatory cytokines IL-1β, IL-6, and TNF significantly increased in the bone marrow of malnourished mice as determined by intracellular cytokine staining (Figure 7B). Similarly, the percentage of inflammatory monocytes that expressed IL-1β increased in LPS challenged and naïve mice (Figure 7D; Supplementary Figure 5A). Since IL-1β and inflammasome activation are known to promote bone resorption under inflammatory conditions (56, 57), we evaluated inflammasome activation in inflammatory monocytes in the bone marrow of malnourished mice using caspase-1 activation as a readout. We used Csfr1-DT transgenic mice in which the diphtheria toxin (DT) receptor is selectively expressed in monocytes under control of the Csfr1 promoter (49). Inflammasome activation was significantly increased in malnourished naïve mice compared with control naïve mice. This difference was maintained but not amplified by ex vivo exposure of the bone marrow cells to LPS (Figure 7E). Depletion of inflammatory monocytes by administration of DT to malnourished mice (Supplementary Figure 5B) reduced bone marrow expression of caspase-1, and Il-1β in LPS challenged mice confirming their role in malnutrition-related inflammasome activation (Figures 7E, F). However, despite IL-1β and adiponectin being reduced concomitantly with inflammatory monocytes in DT mice (Supplementary Figure 5C), TNF was increased and IL-6 was not modified (Supplementary Figure 5C). Suggesting inflammation-driven independently of inflammatory monocytes.

Lipid-based nutrient supplements are used in the treatment of malnourished children because of their high energy density (58). These supplements are typically high in omega-6 fatty acids (59), which are pro-inflammatory (60). Similarly, the nutrient-deficient mouse chow used in our malnutrition model has corn oil as the lipid source, which is rich in pro-inflammatory omega-6 fatty acids. In contrast, dietary omega-3 fatty acids have anti-inflammatory effects (61). Since dietary fish oil (rich in omega-3) supplementation has been associated with an increased omega-3 to omega-6 ratio (62) and decreased inflammation (60, 61), we reasoned that incorporation of omega-3-rich fish oil in a nutrient-deficient diet would decrease inflammation and osteoclast activation compared with the standard omega-6-rich nutrient-deficient diet. Indeed, mice fed the nutrient-deficient diet with fish oil rather than corn oil as the lipid source showed reduced expression of inflammatory cytokines (Il1b, Tnf, and Il6) (Figure 8A) and a decreased percentage of inflammatory monocytes (Figure 8B) in the bone marrow following LPS challenge. LPS-challenged malnourished mice that received dietary fish oil also showed downregulation of genes associated with osteoclastogenesis (Ctsk and Rankl) (Figure 8C). The ratio of expression of osteoprotegerin (Opg) to Rankl, which is indicative of bone integrity, was significantly increased in mice that received the nutrient-deficient diet with fish oil (Figure 8C). We also found fewer TRAP+ activated osteoclasts in the bone marrow of malnourished mice that received the fish oil (Figure 8D), which was indicative of decreased bone resorption in fish-oil-fed mice. Lastly, to confirm an effect of omega-3 fatty acids in suppressing osteoclastogenesis we found fewer TRAP+ osteoclasts in bone marrow cultures from LPS- and L. donovani-challenged malnourished mice exposed ex vivo to Docosahexaenoic acid (DHA) (Figure 8E). Similarly, bone marrow from L. donovani-challenged malnourished mice showed reduced differentiation to TRAP+ osteoclasts when cultured ex vivo in the presence of DHA (Supplementary Figure 6D). Dietary fish-oil mediated reduction in expression of proinflammatory cytokines and genes associated with osteoclastogenesis paralleled the reduction in osteoclast activity in LPS-challenged malnourished mice, this was not observed in L. donovani-challenged mice (Supplementary Figures 6A, B). Collectively, these data indicate that malnutrition-related inflammation and osteoclastogenesis can be blunted by dietary omega-3 fatty acids.