All mouse experiments were approved by Washington University Animal Studies Committee. Germ-free C57BL/6J mice were housed in a sterile environment of plastic flexible film isolators, as described previously (32). Animal feces and swabs of the insides surfaces of the isolator were monitored on a weekly basis for microbial contamination. Specific pathogen-free mice were maintained under specific pathogen-free conditions. All experiments were done using young 6-8-week-old mice and aged 10-12-month-old mice. Equal numbers of each gender were used. Additional details on mouse strain and housing are provided in Supplementary Methods .

Bone marrow and peripheral blood were processed for flow cytometry as previously described (33). Data were acquired using a FACS Aria III flow cytometer and analyzed using FlowJo™ v10.6.1 software. Cell sorting was performed using the FACS Aria III flow cytometer cell sorter. The LSK-SLAM population was double sorted for purity, improving purity to greater than 90%. A complete list of antibody clones used in these experiments is provided in Supplementary Methods .

Six- to eight-week-old wild-type Ly5.1/Ly5.2 recipient mice were irradiated twice with 600 cGy 6 hours apart. Donor (Ly5.2) bone marrow cells were then injected retro-orbitally and placed on prophylactic antibiotics (trimethoprim-sulfamethoxazole) for 2 weeks. Peripheral blood chimerism was analyzed every 4 weeks until mice were sacrificed 24 weeks after transplantation when donor chimerism in bone marrow and blood were analyzed. For the sorted HSPC transplantation experiments, 50 lineage^- Sca1⁺ cKit⁺ CD150⁺ CD48^- (LSK-SLAM) cells were double sorted into single wells of a 96 well plate containing 250,000 support whole (Ly5.1) bone marrow cells and then injected into recipient mice. HSC purity of the double sorted was greater than 90%. Only mice with at least 1% trilineage donor chimerism were used to assess lineage output.

Libraries were generated from RNA purified from sorted LSK-SLAM was hybridized to Agilent SurePrint G3 Mouse G3 arrays. Details on library preparation, hybridization and analysis of the RNA expression profiling data are provided in Supplementary Methods . RNA expression data is available in the Gene Expression Omnibus database (GSE183138).

The level of 40 different cytokines, chemokines, or acute phase proteins in the serum of mice was quantified using the Mouse Cytokine Antibody Array, Panel A, as per manufacturer’s recommendations (R&D systems, Minneapolis, MN).

For single parameter analysis, unpaired t-test were used to assess statistical significance. For multiple parameter data, statistical significance was calculated using one-way (ANOVA). For the limiting dilution analysis, a log-fraction plot of the limiting dilution model was fitted to data from limiting dilution transplantation of 25K, 50K, or 100K cells from whole bone marrow preparations. The log-fraction plots and statistical analysis were generated using the Extreme Limiting Dilution Analysis software (34).

To investigate the impact of the microbiome on hematopoiesis during aging, we analyzed young (6-8 weeks) and middle-aged (~12 months) mice (hereafter referred to as aged mice) housed under specific pathogen free (SPF) or GF conditions. We did not analyze older mice due to the difficulty in maintaining mice in a gnotobiotic facility for more than one year. However, age-associated changes in hematopoiesis, including myeloid skewing, lymphoid progenitor reduction, and reduced self-renewal capacity are evident in C57BL/6 mice by 12 months (6, 9, 35). Complete blood counts were similar between SPF and GF mice in both young and aged mice, except for a mild anemia in aged SPF mice ( Supplementary Figure 1A ). As reported previously (27–29), the percentage of and absolute number of circulating neutrophils (Gr1^hi CD115^– SSC^hi cells) and monocytes (CD115⁺ Gr1^low/neg cells) is reduced in young GF mice ( Figure 1A and Supplementary Figures 1B, C ). Interestingly, this difference is lost in aged GF mice. In the bone marrow of young GF mice, the percentage and absolute number of neutrophils, Gr1-intermediate granulocytic precursors (CD115⁺ Gr1^Int cells), and myeloid progenitors are normal, suggesting that granulopoiesis is intact ( Figures 1A, C , Supplementary Figures 2B–D ).

Consistent with prior studies, there is a shift in hematopoiesis in aged SPF mice towards granulopoiesis, with a significant increase in the percentage of granulocytic cells and a decrease in B lineage cells in the bone marrow ( Figure 1A ) (6, 36, 37). Due to the increase in BM cellularity in aged mice ( Supplementary Figure 2A ), on an absolute basis, there is a marked increase in granulocytic cells, while B lineage cell number is unchanged ( Supplementary Figure 2B ). This shift from B-lymphopoiesis to granulopoiesis with aging is mostly abrogated in GF mice. Compared with aged SPF mice, a significant increase in the percentage of B lineage cells was observed in the bone marrow and blood of aged GF mice, with a modestly attenuated increase in granulocytic cells ( Figures 1A, B and Supplementary Figure 2B ). Indeed, whereas the ratio of granulocytic cells to B lineage cells in the bone marrow of SPF mice increased nearly 4-fold with aging, no increase was observed in GF mice ( Figure 1B ).

We next examined B lymphopoiesis, quantifying different stages of B cell development starting with lymphoid-primed multipotent progenitors (LMPP/MMP4, lineage^– Sca1⁺ Kit⁺ CD34⁺ FLT3⁺ CD48⁺ CD150^–), lymphoid-committed common lymphoid progenitors (CLPs, lineage^– CD27⁺ FLT3⁺ IL7Rα⁺ cells), and the following B cell precursors: pre-pro-B cells (lineage^– B220⁺ IgM^– CD19^– CD43⁺ Ly6D⁺ cells), pro-B cells (lineage^– B220⁺ IgM^– CD19⁺ CD43⁺ cells), and pre-B (lineage^– B220⁺ IgM^– CD43^– cells). Representative flow plots showing the gating strategy to identify each cell population are shown in Figures 2A–C ; gating was based on young SPF mice. Consistent with a recent study, a decrease in the percentage of MMP4s, but not MMP2s or myeloid-primed MMP3s was observed in middle aged SPF mice ( Figure 2D ) (9). A similar trend was observed in middle aged GF mice. Although no change in the percentage of CLPs were observed, a significant decrease in most B cell precursors populations was observed in middle aged SPF mice ( Figures 2E, F ). This trend was reversed in GF mice, with significant increases in the percentage and absolute number of CLPs and B cell precursors induced with aging ( Supplementary Figure 3 ). These data suggest the microbiota signals contribute to the suppression of B lymphopoiesis during aging primarily at the CLP stage.

The impact of microbiota signals on HSCs, especially with aging, is not well characterized. To address this issue, we first quantified HSCs by flow cytometry ( Figures 3A–C ). On a percentage basis, a non-significant trend to increased phenotypic HSCs was observed in SPF mice with aging ( Figure 3B ). On an absolute basis, the increase in phenotypic HSCs is highly significant ( Figure 3C ). In particular, the number of lineage^- Sca1⁺ cKit⁺ CD150⁺ CD48^- (LSK-SLAM) cells and CD34^- LSK-SLAM cells is increased 6.4 ± 1.7-fold and 3.4 ± 1.2-fold, respectively. Similar increases were observed in aged GF mice, with LSK-SLAM increasing 5.3 ± 1.6-fold (p=NS compared to SPF mice) and CD34^- LSK-SLAM cells increasing 2.8 ± 0.31-fold (p=NS). Aging is associated with an increase in myeloid biased HSCs (38). Prior studies have shown that high CD150 expression marks myeloid biased HSCs (9). Thus, we next assessed CD150 expression on phenotypic HSCs (LSK, CD34^- cells). As expected, a significant increase in the percentage of CD150-high HSCs and a corresponding decrease in CD150-low HSCs was observed in aged SPF mice ( Figures 3D, E ). This shift was largely attenuated in GF mice. Indeed, a significant increase in CD150-low lymphoid biased HSCs is present in both young and aged GF mice compared with aged-matched SPF mice. Of note, no significant difference in the cell cycle status of lineage- Kit+ committed progenitors or LSK-SLAM cells was observed with aging in either SPF or GF mice ( Figures 3F, G and Supplementary Figure 4 ).

Prior studies have shown that phenotypic HSCs from aged mice have reduced repopulating activity on a per cell basis (2, 39, 40). Thus, we performed limiting dilution transplantation using unsorted bone marrow cells as an unbiased approach to assess functional HSC frequency. As reported previously (4), the number of functional HSCs increases in the bone marrow of SPF with age ( Figures 4A, B ). A similar expansion of functional HSCs is present in GF mice. Together, these data suggest that signals from the microbiota are dispensable for the expansion of phenotypic and functional HSCs that occurs with chronologic aging.

To assess the impact of microbiota on HSC lineage commitment, we transplanted a limiting number of sorted HSCs and assessed lineage output ( Figure 5A ). Stable overall donor engraftment was observed over time, with reduced engraftment of aged HSCs from both SPF and GF mice ( Figures 5B, C, E ). A significant decrease in donor granulocytic cell chimerism was seen in both young and aged GF HSC recipients ( Figures 5B–D ). Conversely, donor B cell chimerism in the blood was increased in aged GF HSC recipients, with a similar trend observed in the bone marrow 24 weeks after transplantation. We next analyzed individual mice with at least 1% trilineage engraftment 24 weeks after transplantation to assess HSC lineage bias. As expected, in young SPF mice, the majority of HSCs displayed a balanced myeloid/lymphoid lineage output, with a significant increase in myeloid-biased HSCs observed with aging ( Figures 5F, G ). In young GF mice, the majority of HSCs are lymphoid-biased. Moreover, although the myeloid output, as measured by granulocytic cell chimerism, increased modestly with aging, the majority of HSCs in aged GF remained lymphoid-biased or balanced. In young GF HSCs, an increase in the output of both B- and T-lineage cells was observed ( Figures 5H, I ). In contrast in aged GF HSCs, the increase in lymphoid output was mainly due to an increased production of B cells. Collectively, these data show that microbiota-related signals play a key role in determining the lineage potential of bone marrow resident HSCs, promoting myeloid lineage development at the expense of lymphoid cells.

To begin to explore mechanisms by which microbiota signals regulate HSC lineage bias, gene expression profiling was performed on sorted LSK-SLAM cells from aged SPF and aged GF mice. A list of differentially expressed genes is provided in Supplementary Table 1 . Consistent with our lineage-output transplantation data, aged GF HSCs had a gene expression signature enriched for non-myeloid biased HSCs ( Figure 6A and Supplementary Figure 5 ) (41). Surprisingly, gene set enrichment analysis showed that GF HSCs were enriched for expression signatures related to inflammatory signaling, including tumor necrosis factor (TNF) and interferon (IFN) signaling ( Figure 6B ). This prompted us to measure the circulating level of 40 different inflammatory cytokines, chemokines, or acute phase proteins, including TNFα, interleukin-1β, and IFN gamma. Only five of these inflammatory mediators were detected above background ( Figure 6C ). Increased expression of C5, CXCL13, soluble ICAM, and M-CSF was observed in the serum of SPF mice with aging, with a similar increase seen in aged GF mice. Together, these data suggest the microbiota signals are not major drivers of increased systemic inflammatory cytokine/chemokine expression or inflammatory signaling in HSCs.