The underdeveloped immune system was confirmed by routine blood tests in GF mice
First, blood was drawn from GF (n = 11) and SPF mice (n = 10) for routine blood tests (Table 1; Fig. 1A). The results revealed that the white blood cell (WBC) count was significantly lower in GF mice than in SPF mice (1.80×109/L and 3.40×109/L on average, p-value = 0.040). Specifically, the germ-free condition resulted in a significant decrease in the number of monocytes (p-value = 0.027) and neutrophils (p-value = 0.014). However, only a minor (statistically nonsignificant, p-value = 0.097) reduction in lymphocyte count was observed. Recent research has shown that GF mice have fewer platelets than SPF (both wild-type and IL-1 receptor 1 knockout) mice do [34]. In our study (Table 1), although the reduction in platelet count was not significant (p-value = 0.233) in GF mice, it was significant for the mean platelet volume (MPV) and platelet distribution width (PDW).
Table 1
Routine blood tests of germ-free (GF) and specific-pathogen-free (SPF) Kunming mice
Test (unit) | GF (n = 11) | SPF (n = 10) | p-value |
WBC (109/L) | 1.80 ± 1.08 | 3.40 ± 1.98 | 0.040 |
LYM (109/L) | 1.44 ± 0.96 | 2.45 ± 1.56 | 0.097 |
MON (109/L) | 0.06 ± 0.03 | 0.17 ± 0.14 | 0.027 |
NEU (109/L) | 0.30 ± 0.14 | 0.78 ± 0.48 | 0.014 |
EOS (109/L) | 0.00 ± 0.00 | 0.00 ± 0.00 | NA |
BAS (109/L) | 0.00 ± 0.00 | 0.00 ± 0.00 | NA |
LYM% (%) | 77.77 ± 6.57 | 69.66 ± 12.05 | 0.080 |
MON% (%) | 3.44 ± 2.17 | 5.72 ± 2.57 | 0.042 |
NEU% (%) | 18.81 ± 6.46 | 24.62 ± 10.61 | 0.156 |
EOS% (%) | 0.00 ± 0.00 | 0.00 ± 0.00 | NA |
BAS% (%) | 0.00 ± 0.00 | 0.00 ± 0.00 | NA |
RBC (1012/L) | 10.98 ± 0.84 | 10.68 ± 1.05 | 0.477 |
HGB (g/L) | 14.44 ± 1.31 | 13.76 ± 1.69 | 0.324 |
HCT (%) | 51.62 ± 6.29 | 50.62 ± 5.70 | 0.705 |
MCV (fL) | 47.00 ± 3.58 | 47.50 ± 2.80 | 0.724 |
MCH (pg) | 13.15 ± 0.34 | 12.89 ± 0.82 | 0.361 |
MCHC (g/L) | 28.12 ± 1.65 | 27.23 ± 1.94 | 0.277 |
RDWc (%) | 20.13 ± 0.99 | 19.43 ± 0.93 | 0.113 |
RDWs (fL) | 35.37 ± 3.45 | 34.53 ± 1.68 | 0.482 |
PLT (109/L) | 464.55 ± 144.97 | 540.00 ± 135.75 | 0.233 |
MPV (fL) | 6.78 ± 0.28 | 7.77 ± 1.13 | 0.023 |
PCT (%) | 0.32 ± 0.10 | 0.43 ± 0.15 | 0.073 |
PDWc (%) | 31.89 ± 2.48 | 35.03 ± 3.30 | 0.026 |
PDWs (fL) | 10.47 ± 1.98 | 13.66 ± 3.80 | 0.033 |
WBC, white blood cell count; LYM, lymphocyte count; MON, monocyte count; NEU, neutrophil count; EOS, eosinophil count; BAS, basophil count; LYM%, lymphocyte ratio; MON%, monocyte ratio; NEU%, neutrophil ratio; EOS%, eosinophil ratio; BAS%, basophil ratio; RBC, red blood cell count; HGB, hemoglobin; HCT, hematocrit; MCV, mean cell volume; MCH, mean cell hemoglobin; MCHC, mean cell hemoglobin concentration; RDWc, red cell distribution width (CV), RDWs, red cell distribution width (SD); PLT, platelet count; MPV, mean platelet volume; PCT, phrombocytosis; PDWc, platelet distribution width (CV); PDWs, platelet distribution width (SD); NA, not available. P-values were calculated using Student’s t-test.
Major changes in the transcriptional profiles of GF and SPF mice compared at the single-cell level
To explore how the microbiota affects the composition and transcription of the immune microenvironment, we established a comprehensive single-cell transcriptome reference for the hematopoietic system of adult GF and SPF mice by performing scRNA-seq of PB and BM (Figs. 1A and 1D). We isolated and sequenced 21,827 cells from PB cell suspensions and 19,940 cells from BM cell suspensions from GF (n = 1) and SPF (n = 1) mice. After stringent quality control and data integration, we further analyzed 18,344 high-quality cells with 1,426 median genes per cell from PB and 16,537 high-quality cells with 1,391 median genes per cell from BM.
Using an unsupervised clustering method, we first partitioned single cell profiles composed of a total of 34,881 cells into 18 major cell identities (Figure S1B). Furthermore, we divided 18 main cell groups into 25 subpopulations of cells (Fig. 1B). These cell identities were identified by known marker genes for blood cells (Figs. 1C and S2). Eosinophils were not detected via scRNA-seq data, which is consistent with the results of routine blood tests (Table 1).
Next, we evaluated the differences in transcriptional profiles between GF and SPF mice in two ways. First, we calculated DEGs for each cell identity and then visualized the number of DEGs via uniform manifold approximation and projection (UMAP) plots (Fig. 1E). The numbers of DEGs revealed broad transcriptional changes in blood cells, most notably in neutrophils, monocytes, and B cells. The number of DEGs also progressively increased as differentiation progressed, especially in neutrophils (Figure S1D). In addition, we measured the distance between the major identities via the Bhattacharyya distance [31]. Owing to the limited number of cells available, clusters with fewer than 400 cells were excluded (see Methods). These results revealed more significant differences between GF and SPF mice in terms of the expression of neutrophils, monoblasts, monocytes, myeloid cells, B cells, CD4 + T cells, and NK cells (Fig. 1F). All comparisons were significant on 100 replicates of testing, while the mean fold change varied from 1.82-fold (myeloid cells) to 1.19-fold (granulocytopoietic cells). Although not as significant as those of the immune cells listed above, the expression of hematopoietic stem and progenitor cells considerably changed, indicating that this cell identity was highly likely to be directly or indirectly regulated by the microbiota and its metabolites.
As a result, we identified primary blood cells in GF mice and characterized broad differences in cell proportions and transcriptional profiles compared with those in SPF mice, highlighting notable changes in neutrophils and monocytes (Figs. 1E-1G). Interestingly, neutrophil and monocyte happen to be the cell identities that are significantly reduced in GF mice in routine blood. Therefore, we hypothesize that the microbiota modulates gene expression in immune cells, thus affecting the proliferation or apoptosis of immune cells.
HCAR2 activation as a potential mechanism for the regulation of neutrophil number
The bone marrow can continuously generate neutrophils by expressing some cytokines (such as G-CSF) and releasing them into the blood circulation [35]. Previous studies have shown that the microbiota can regulate neutrophil production and function throughout life [36, 37]. However, the exact molecular mechanisms, pathways, and genes involved remain unclear. To better understand the role of the microbiota in bone marrow neutrophil development, we extracted neutrophils from all the data for downstream analysis. Clustering of neutrophils revealed four distinct clusters (neutrophil P1-P4, Fig. 2A), which we identified with marker genes reported by Grieshaber-Bouyer et al. [38]. The four stages were ordered according to their developmental stages (Figure S1D). Neutrophil P1 corresponds to the early stage, which develops from granulocytopoietic cells, and neutrophil P4 corresponds to the late stage, which approaches mature neutrophils in terms of gene expression.
To determine the molecular basis underlying the difference between GF and SPF mice, we performed differential expression analysis and identified sets of up-regulated and down-regulated genes. Next, we performed a GO enrichment analysis on the identified up-regulated and down-regulated genes (Fig. 2C). We found that signaling pathways related to interleukin-1 beta (IL-1β) production were enriched in up-regulated genes, which showed a more potent antibacterial effect in GF mice. Specifically, the genes Arg2, Egr1, and Tlr2, which are involved in the IL-1β production pathway, were up-regulated in GF mice (Figs. 2B and 2D). On the other hand, actin-related signalingXXX pathways, such as regulation of actin cytoskeleton organization and regulation of actin filament-based processes, were enriched in down-regulated genes (Figure S2), especially in the early stages, which means that early neutrophils can be inactive in terms of cell motility and cytokinesis in GF mice.
Compared with that in SPF mice, the expression of Hcar2 was up-regulated in GF mice neutrophil P4 (Figs. 2B and 2D). HCAR2 (also known as GPR109A) acts as a high-affinity receptor for both nicotinic acid (NA; also called niacin) and (D)-beta-hydroxybutyrate and mediates NA-induced apoptosis in mature neutrophils [39]. HCAR2 activation by NA results in reduced cAMP levels, thereby affecting the activity of cAMP-dependent protein kinase A and phosphorylating target proteins, resulting in neutrophil apoptosis [40]. We hypothesized that the expression of Hcar2 is regulated by NA or related metabolites. To validate the up-regulation of Hcar2, we performed scRNA-seq on BM cells from GF (n = 1) and SPF (n = 1) mice via the DNBelab C4 platform. We confirmed that neutrophils were significantly reduced in the GF mouse and that Hcar2 was significantly up-regulated in the GF mouse in the validation data (Figure S3). Furthermore, a recent study performing scRNA-seq and spatial transcriptomics on the spleen of GF (n = 1) and SPF (n = 1) mice reported a significant decrease of neutrophils in GF mice (Figs. 2E and 2F) [41]. We demonstrated the spatial expression of three marker genes of neutrophils, Il1b, Ccl6, and Ly6g, in the spleen to determine the difference in the number of neutrophils in the spleen between GF and SPF mice (Fig. 2F). More importantly, we verified the significant up-regulation of Hcar2 in GF mouse neutrophils via the spleen scRNA-seq data (Fig. 2E).
To further confirm the metabolic changes associated with NA, we performed a non-targeted metabolomics analysis on the cecal content, feces, and serum of GF (n = 11) and SPF (n = 10) mice (Figs. 1A and 2G; Supplementary Data). In fecal and cecal samples from GF mice, we detected decreased levels of NA and 6-hydroxynicotinic acid, the 6-hydroxy derivative of NA produced by the action of niacin dehydrogenase, which is derived mainly from Pseudomonas. On the other hand, nicotinamide (NAM) which is converted from NA when it is overtaken, and trigonelline, a product of NA metabolism that is excreted in the urine of mammals, are increased.
These results suggest that the nicotinate degradation pathway is down-regulated as a result of the absence of microbiota in GF mice. Excess NA is subsequently converted to trigonelline, NAM, or even nicotinuric acid (NUA). In summary, the accumulation of NA and its derivative due to a lack of niacin dehydrogenase in GF mice is a potential regulatory mechanism for up-regulation of Hcar2, which can enhance mature neutrophil apoptosis.
Enhanced bacterial recognition and impaired viral defense in GF mouse monocytes
Monocytes and macrophages play indispensable roles in immunosuppression, tissue repair, inflammation regression, bacterial removal, atherosclerosis, and fibrosis. We separated monoblasts from monocytes based on the fact that monoblasts expressed more Irf8 and Elane (Fig. 1C). Monoblasts and monocytes from PB and BM were extracted for further analysis, but macrophages whose cell number was insufficient for statistical analysis were excluded (Fig. 3A).
The differential expression analysis of both monoblasts and monocytes revealed that GF mice expressed high levels of the monocyte marker gene Vcan (versican, Fig. 3B). The versican gene encodes a large chondroitin sulfate/dermatan sulfate proteoglycan belonging to the aggrecan/lexican family [42]. Previous studies have indicated that up-regulation of versican is correlated with inflammation and cancer aggressiveness [42, 43]. In addition, versican blocks cell migration and is associated with slow cell proliferation and cytodifferentiation, and promotes the adhesion of monocytes in the extracellular matrix (ECM) [44]. The up-regulation of versican may contribute to impaired monocyte development. Moreover, GF mice expressed a high level of Cd14 in monocytes but not in monoblasts (Fig. 3B). CD14 is a coreceptor for the detection of bacterial lipopolysaccharide (LPS), and it also recognizes other pathogen-associated molecular patterns, such as lipoteichoic acid. Monocytes produce sCD14, a soluble form of the receptor, which confers LPS responsiveness to cells that do not express Cd14. Generally, high levels of sCD14 in monocytes reflect monocyte activation [45]. Under germ-free conditions, there is increased recognition of microbes, which may be associated with the potential risk of inflammation. The up-regulation of Cd14 was also detected in neutrophils (P1-P4). To further validate this conclusion, we performed differential expression analysis on a cecum dataset containing GF and SPF mice [46]. The results revealed that Cd14 was up-regulated in enterocytes (fold change = 1.28, p-value = 6.65e-17).
GO enrichment of monocytes revealed that the up-regulated genes were enriched mainly in pathways related to reactive oxygen species (ROS), the ERK1 and ERK2 cascades, myeloid cell differentiation, and apoptotic cell clearance (Fig. 3C). ROS are well-known inducers of inflammation [47], and ERK signalingXXX is believed to promote primarily cell proliferation and survival [48]. However, it has been well documented that ROS-induced ERK activation generally results in cell cycle arrest and apoptosis in a variety of cells [49, 50], which may explain the decreased number of monocytes in GF mice shown in routine blood tests. On the other hand, the down-regulated genes were enriched mainly in pathways related to antigen processing and presentation and interferons (Figs. 3D and 3E). These findings indicate that monocytes have a reduced capacity to mount an immune defense against viral pathogens and endogenous peptide antigens, which is mediated by interferon β (IFN-β) and interferon γ (IFN-γ), respectively. Consistent with these findings, previous studies reported increased susceptibility of GF mice to influenza A virus [10], coxsackie B virus [11], and Friend virus [12].
Elevated excretion of 5’-methylthioadenosine is associated with immunodeficiency
All mammalian tissues contain 5’-methylthioadenosine (MTA), a sulfur-containing nucleoside. In humans, MTA functions as a powerful inhibitor of several enzymatic reactions, and is synthesized mainly through the polyamine biosynthetic pathway from S-adenosylmethionine. For example, MTA can influence the regulation of gene expression, proliferation, differentiation, and apoptosis, as reported previously [51]. Elevated excretion of MTA has been detected in children with severe combined immunodeficiency syndrome (SCID) [52]. This is consistent with our findings in GF mice with underdeveloped immune systems. MTA elevation was observed in the serum (fold change = 1.5304, p-value = 0.0121), feces (fold change = 2.3591, p-value = 0.0211) and cecal content (fold change = 4.4906, p-value = 0.0023). This finding suggested a direct or indirect association between elevated MTA and the immunodeficiency phenotype.
Cell-cell interaction analysis revealed that decreased BST2 signaling pathway activity in GF mice may inhibit immune response activation
In view of transcriptomic changes in genes encoding receptor-ligand pairs in response to the absence of microbiota, we performed an unbiased inference analysis of cell-cell interactions via CellChat [33] across all cell identities from PB and BM except plasma cells, which were found to be absent in the immune system of GF mice in this study.
To measure the degree of cellular communication between samples, we compared the total number and strength of interactions in the inferred cell-cell communication network. Compared with those SPF mice, the interaction numbers and strength in GF mice were lower (Fig. 4A). To specify the interaction between which cell populations showed notable changes, we compared the number and strength of the interactions between different cell populations (Fig. 4B). Interestingly, when NKT cells act as receivers, communication between NKT cells and all other cells is reduced or shut down. This finding suggests that NKT cells in GF mice cannot receive most extracellular signals. In terms of interaction strength, there was also a significant reduction in CD8 + T cell interactions with other cell types, particularly lymphocytes and NK cells, in GF mice, the reasons for which need to be further investigated. Macrophages in GF mice exhibit enhanced interactions with various cell types, either as senders or receivers.
To further explore the differences in cellular interactions between cell identities in GF versus SPF mice, we used CellChat to calculate the distance of signaling pathways between the 24 cell groups in the two datasets (Fig. 4C). The results revealed that the BST2 signaling pathway was the most significant in terms of information flow. The ligand gene Bst2 expressed by bone marrow stromal cells can promote the growth of murine pre-B cells [53], and Pira2 is the receptor gene. We found that the BST2 signaling pathway was inhibited in basophils, platelets, Vpreb3- naïve B cells, and all T cell subpopulations in GF mice, which was caused by the down-regulation of Bst2 gene expression in these cell groups (Figs. 4D and 4E). It has been reported that the acquisition of alloantigen-specific memory by murine macrophages requires the interaction between MHC-I and PIR-A [54]. In conclusion, the reduction in the BST2 signaling pathway between cells may inhibit the immune response activation of BST2 signaling pathway receptor cells in GF mice, which are monoblasts, monocytes, macrophages, myeloid cells, and neutrophils, to be more specific.