Fecal microbiota transplantation in mice improves bone material properties through altered mineral quality

Abstract

Disruptions of the composition of the gut microbiome are linked to impaired bone tissue strength. Fecal microbiota transplantation (FMT) is an established clinical therapy that can restore a healthy gut microbiome and reduce systemic inflammation. However, whether FMT from a healthy donor could rescue bone fragility is unknown. As induced inflammation causes mineralization defects, we hypothesize that manipulations of the gut microbiota alter bone fracture resilience through changes in mineral quality. Here, we altered the compositions of the gut microbiome in mice via antibiotics (ampicillin and neomycin) and FMT. Mice were allocated to 5 groups (M/F, N = 13-18/group): Unaltered, Continuous (dosed 4-24 wk), Initial (dosed 4-16 wk), Reconstituted (dosed 4-16 wk with subsequent FMT from age- and sex-matched mice with unaltered gut microbiota), and Delayed (dosed 16-24 wk). Fracture toughness testing and Raman spectroscopy were conducted on the femora. The maximum toughness was greater in the Reconstituted group (for females, p < .05 compared to Continuous, Unaltered, and Delayed groups; for males, p < .05 compared to groups with antibiotic dosing). The Reconstituted group showed lower type-B carbonate substitution in the bone mineral (all p < .01 for both sexes), and lower mineral-to-matrix ratio (all p < .01 for males, for females, p < .01 compared to Unaltered, Initial, and Delayed groups). In females, mineral crystallinity was higher in the Reconstituted group than those dosed with antibiotics (all p < .05). Serum inflammation marker TNF-α was positively correlated with type-B carbonate substitutions (ρ = 0.66), mineral-to-matrix ratio (ρ = 0.71), and carboxymethyl-lysine (CML) in bone matrix (ρ = 0.43). Enhanced bone maximum fracture toughness was associated with reduced type-B carbonate substitution (r = −0.45), decreased mineral-to-matrix ratio (r = −0.40), increased mineral crystallinity (r = 0.33), and lower levels of bone CML (r = −0.49, all p < .01). These results suggest that the introduction of more beneficial gut microbiota can increase fracture resistance by modifying mineral composition and quality, likely through the reduction of systemic inflammation.

Graphical Abstract

Graphical Abstract.

Created by bioRender

Introduction

The gut microbiome refers to the microorganisms living in the digestive tracts, including bacteria, archaea, viruses, fungi, and protozoa. The gut microbiome significantly influences human health, serving as a regulator for inflammatory, immune, and metabolic status. Disruptions in the composition of the gut microbiome (dysbiosis) are associated with several chronic conditions, such as inflammatory bowel disease, diabetes, chronic kidney diseases, etc.^1–3 Recent advances have shown that the gut microbiome can also influence bone health. For example, dysbiosis of the gut microbiota is implicated in the development and progression of osteoporosis.⁴ Changes to composition of the gut microbiome have also been shown to impair bone tissue strength.^5,6 However, the underlying mechanisms of how an altered composition of the gut microbiome affects bone tissue quality have not been identified.

Fecal microbiota transplant (FMT) is a means of introducing a new gut microbiota to a host. In one implementation, this involves introducing a stool sample from a healthy donor to the recipient’s digestive tract, which can restore intestinal balance and reduce systemic inflammation.⁷ Clinically, for treating Clostridioides difficile colitis, FMT from a healthy donor is a standardized therapy that the US Food and Drug Administration has approved.⁸ Preclinical and clinical studies have shown that FMT from a healthy donor can benefit chronic diseases including diabetes.⁹ Interestingly, recent studies also showed that FMT from a healthy donor could slow down bone loss in osteoporotic animal models and potentially serve as a therapeutic option for osteoporosis.^10,11 Nevertheless, it is unknown if the repaired composition of the gut microbiome through FMT can improve bone material properties, nor is it clear what changes in the bone matrix may be caused by reconstitution of the gut microbiome.

Systemic inflammation negatively impacts bone metabolism and is linked to osteoporosis.^12,13 Recent studies have shown systemic inflammation to influence the mineral composition and deteriorate the mineralization quality. For example, induced inflammation in rabbit bones¹⁴ and human ossicles¹⁵ increased type-B carbonate substitution (the ratio of carbonate to phosphate) in the mineral phase. Increases in type-B carbonate substitution are often associated with aging^16,17 and osteoporosis¹⁸ and are negatively correlated with bone ultimate strength and toughness.^19,20 More importantly, one of our recent studies showed that removing the receptor of advanced glycation end-products (RAGE), an inflammatory regulator, reduces type-B carbonate substitution in mouse bones, indicating improvements in mineral quality.²¹ Thus, we hypothesize that reconstitution of a healthy gut microbiome through FMT reduces systemic inflammation and improves mineral quality and bone fracture resilience.

In the previous study,⁶ mice were dosed with oral antibiotics (ampicillin + neomycin) to alter the composition of the gut microbiome before or after skeletal maturity. It was demonstrated that alterations to the gut microbiome caused significant reductions in bone tissue strength in male mice but not in female mice when changes in the microbiome were made throughout growth or after skeletal maturity. Furthermore, it was reported that microbiome-induced impairments of bone strength can be recovered with FMT after skeletal maturity. Since no reduction in bone geometry, tissue mineral density, serum turnover marker, and mineral apposition rate were observed in any of the antibiotic dosing groups, the microbiome-induced impairments of bone tissue strength did not appear to be related to the volume of bone turnover caused by osteoclast/osteoblast activity. Thus, bone matrix quality, influenced by changes in the compositions of the gut microbiome, was considered to be a key contributor to bone fragility. Interestingly, most of the matrix differences of the bone cross-sections assessed by Raman spectroscopy were not statistically significant between the groups, perhaps due to the limitations in sample preparation and the small sample size. Consequently, it remains unclear how the disruptions in the compositions of the gut microbiome could lead to alterations of bone tissue strength. Therefore, this study aimed to examine additional bone matrix parameters to uncover the key associations that could suggest a mechanistic pathway. We used the contralateral femoral samples from the same cohort, and investigated the fracture toughness of the bone material, which captures the fragility and brittleness of the bone tissue, and conducted Raman spectroscopic analysis on the periosteal surface of fresh femora as opposed to the embedded cross-section, that captures changes in turnover more directly, overcomes embedding-related limitations, and is more relevant for bone fracture toughness.

Materials and methods

Study design

The study analyzed samples from the study reported previously by Liu and Cyphert et al.⁶ The Institutional Animal Care and Use Committee at Cornell University approved the animal procedures and experimental design. In brief, C57BL/6J mice were purchased from Jackson Laboratory, fed with standard chow, housed in plastic cages with corn cob bedding, and raised in a 12 h light/12 h dark dayshift schedule. The mice were randomly assigned to five experimental groups: Unaltered group, Continuous group, Initial group, Reconstituted group, and Delayed group (Figure 1). This study design includes 3 experimental groups where the compositions of the gut microbiome were negatively modulated (Continuous, Initial, and Delayed), and 1 experimental group where the compositions of gut microbiome were positively modulated after alteration (Reconstituted). Standard drinking water was provided for the Unaltered group through the entire study period, whereas antibiotics (1 g/L ampicillin and 0.5 g/L neomycin) were provided in the drinking water for the Continuous group from 4 to 24 wk of age, the Initial group from 4 to 16 wk of age, and to the Delayed group from 16 to 24 wk of age. Standard drinking water was provided to the Initial group from 16 to 24 wk of age and to the Delayed group from 4 to 16 wk of age. The 8-wk dosing duration was selected for the Delayed group to ensure the gut microbiota to stabilize in the first 6 wk²⁴ and an additional 2 wk to allow sufficient time to measure/identify bone formation markers for nanoscale analysis. The effective antibiotic dosage selection was based on previous studies, and the dosed duration did not influence body weight, adiposity, and bone growth.^5,6,22,23 Ampicillin and neomycin have low oral bioavailability, limiting the amount that gets into circulation.^25,26 Therefore, the differences in systemic and skeletal outcomes in the study groups are mainly caused by changes to the compositions of the gut microbiome.

Figure 1.

Study design.⁶

To test the effect of reconstituting a healthy gut microbiome following disruption, the Reconstituted group received antibiotic dosing from 4 to 16 wk of age, followed by FMT procedure. Based on a previous established protocol,²⁷ the fresh stool samples were collected from age- and sex-matched mice that received standard drinking water. The samples were placed in sterile tubes and immediately transferred to an anaerobic chamber, followed by suspension in 1 mL anoxic PBS with 0.05% L-cysteine. Then, the pellets were homogenized and centrifuged to produce the fecal slurry, which were pooled (separately for males and females) and stored in an anaerobic jar. Mice in the Reconstituted group received 150 μL of the pooled transplant daily for 3 d.⁶

All animals were euthanized at the age of 24 wk. After euthanasia, the femora from all mice were excised for material characterization. The sample size for each group was listed below: Unaltered (n = 13 M, n = 14 F), Continuous (n = 17 M, n = 14 F), Initial (n = 18 M, n = 18 F), Reconstituted (n = 12 M, n = 12 F), and Delayed (n = 14 M, n = 16 F). This study used the left femora to conduct fracture toughness testing and Raman spectroscopical analysis. We also used the data of gut microbiome analysis and serum biomarker analysis (in particular, pro-inflammatory cytokines serum TNF alpha (TNF-α) and serum IL 6 (IL-6)) from the previously published study,⁶ to understand how changes in the gut microbiome may explain the variations in bone matrix quality and resistance to fracture.

Microbiome analysis

The mice fecal samples were collected at 16 wk of age (before any change in antibiotic dosing) and the day before euthanasia. Based on previously established protocols,²⁸ 16S rRNA amplicon sequencing was used to quantify the microbiome composition from the fecal samples. DNA extraction, purification, library preparation, and sequencing were conducted at the University of California San Diego Microbiome Core. Raw sequencing reads were trimmed and taxonomically classified using QIIME2. Specifically for this study, we performed statistical analysis on the relative abundance of microbial genera previously reported to show anti-inflammatory effects (Bifidobacterium, Lactobacillus, Anaerostipes, Roseburia, Coprococcus, Butyricicoccus, and Ruminococcus)^29–33 or pro-inflammatory effects (Bacteroides and Clostridium).³²

Fracture toughness testing

The fracture toughness tests were conducted based on a previously described protocol.³⁴ Epiphyses of the left femora were removed using a low-speed precision cutter (Buehler IsoMet 1000). The anterior surface of the femora was then notched with a 150 μm diamond-embedded blade to introduce a notch at the mid-diaphysis. Notches were sharpened with a razor blade. Subsequently, the notched sharpness and geometry were examined by 2D images acquired from a MicroCT scanner (Bruker Skyscan 1276; VivaCT40; Scanco Medical AG). The samples were tested until failure under 3-point bending with a mechanical testing system (Instron 5543). The yield load and maximum load from the load-displacement curves obtained from each test were used to calculate the stress intensity factor at crack initiation (initiation toughness, K_in, MPa·m^1/2) and at maximum load (maximum toughness, K_max, MPa·m^1/2), respectively, using the formula (1) shown below³⁴

(1)

where is a geometry factor that can be calculated using the equation provided in ³⁴; is the yield load or maximum load; is the test span length; , , and correspond to the outer, inner, and mean radius of bone; is the notch angle. The fracture surface for each sample was visually inspected after the test. Tests were excluded from the analysis if the crack did not propagate along the notch or if the sample failed catastrophically.

Raman spectroscopy

Following mechanical testing, the proximal sides of the right femora were collected, cleaned, and analyzed with a Raman Spectrometer equipped with a 785 nm laser and 1200 L/mm grating (Ramascope 2000, Renishaw). Four points on the posterior surface were examined under a 50× magnification microscope and selected to account for the sample heterogeneity. Raman spectra were acquired for 15 s integration time with 5 accumulations under ~75 mW laser power. The spectral range acquired was from 300 to 1800 cm⁻¹. The spectra were individually baseline-corrected and analyzed. Peak deconvolution and peak fitting were conducted using a custom MATLAB script. The parameters of interest include mineral-to-matrix ratio (v1PO4³⁻/Proline, I₉₅₉/I₈₅₅; and v2PO4³⁻/Amide III, I₄₂₈/I₁₂₄₀), crystallinity [1/FWHM (959 cm⁻¹)], type-B carbonate substitution (v1CO3²⁻/v1PO4³⁻, I₁₀₇₀/I₉₅₉), carboxymethyl-lysine (CML, as an assessment of glycation/glycoxidation level of the organic matrix, I₁₁₆₀/I₁₄₅₀), proline hydroxylation (I₈₅₅/I₈₇₂), and Amide I sub-peak ratios (I₁₆₇₀/I₁₆₄₀ and I₁₆₇₀/I₁₆₉₀). The above measures were averaged within each sample. The acquisition and analysis protocols were consistent with the previously published literature.^19,35–37

Statistical analysis

Based on normality and homoscedasticity results, the statistical differences between groups were determined by one-way ANOVA with Tukey post hoc test (if equal variances could be assumed across groups) or Games–Howell post hoc test (if the data did not have equal variance). Otherwise, the Kruskal–Wallis test with Dunn’s post hoc test was used to assess statistical differences in groups with data not normally distributed. Depending on the normality of the data, Pearson’s or Spearman’s correlation efficient was calculated across variables and with serum biomarker measures from the previous study.⁶ For the analysis on the selected microbial genera related to inflammation profiles, whether sex was a significant factor was first tested. If no significant sex-specific difference was observed, the sexes were combined, since the differences in the serum inflammation markers reported in the previous study did not appear to be sex-dependent. A p-value smaller than .05 was considered statistically significant. All statistical analyses were conducted in Minitab.

Results

Reconstitution of the gut microbiome improves bone fracture resistance compared to mice with microbiome disruption via antibiotics

In females, the Reconstituted group showed higher initiation toughness compared to the Delayed group (p < .001) and, interestingly, compared to the Unaltered group (p = .03, Figure 2A). In similar trends, the maximum toughness in females was also higher in the Reconstituted group compared to the Continuous group (p = .02), the Delayed group (p < .001), and the Unaltered group (p = .02, Figure 2B). In males, no difference in initial toughness was observed across all groups (Figure 2C). However, the Reconstituted group demonstrated elevated maximum toughness compared to all groups with disruption of the gut microbiome (Continuous, p < .001; Initial, p < .001, and Delayed, p = .02) but was not different from the Unaltered group (Figure 2D). In contrast, in both male and female mice, alteration of the microbiome through antibiotics at any period (Continuous, Initial, and Delayed group) did not alter the fracture properties of the femora compared to the Unaltered group.

Figure 2.

Manipulation of gut microbiota changes bone fracture properties. (A) Initiation toughness in females was higher in the Reconstituted group than in the Unaltered and Delayed groups. (B) Maximum toughness in females was higher in the Reconstituted group than in the Unaltered, Continuous, and Delayed groups (C). Initiation toughness in males was not different across the groups. (D) Maximum toughness in males was higher in all groups with antibiotic treatments. N = 7-10 per group for females; N = 5-11 per group for males. One-way ANOVA with Tukey post hoc test was used for (A) and (C); one-way ANOVA with Games–Howell post hoc test was used for (B); and Kruskal–Wallis with Dunn’s post hoc test was used for (D).

Reconstitution of the gut microbiome improves mineral composition and quality

Assessed by Raman spectroscopy on the periosteal surface of the fresh femora, the mineral type-B carbonate substitution in the Reconstituted group was significantly reduced compared to all other groups in both sexes (p < .001 for all groups, Figure 3A and B). The mineral crystallinity was significantly higher in the Reconstituted group compared to all 3 antibiotic-dosed groups in females only (p < .05 for all groups, Figure 3C). A similar trend for mineral crystallinity was shown in males, but the difference was only significant between the Reconstituted and Continuous groups (p < .05, Figure 3D). In males, the mineral-to-matrix ratio (evaluated by the ratio of v1PO4³⁻ and proline) was reduced in the Reconstituted group compared to all other groups (all p < .01, Figure 3F). The mineral-to-matrix ratio (v1PO4³⁻/proline) in females was lower in the Reconstituted group compared to the Unaltered, Initial, and Delayed group (all p < .01), but not statistically different from the Continuous group (Figure 3E). The alternative mineral-to-matrix ratio evaluated by the ratio of v2PO4³⁻ and Amide III showed a similar trend, where the ratio was reduced in the Reconstituted group when compared to all other groups in males (all p < .01) and reduced in the Reconstituted group than initial group (p = .02) in females. The 2 metrics of mineral-to-matrix ratio are significantly correlated with each other (p = .001). Although not different from the Unaltered group, the accumulation of CML in the bone matrix in the Reconstituted group was significantly decreased compared to the Initial and the Delayed groups in the female (both p = .02, Figure 3G). Carboxymethyl-lysine in the Reconstituted group was decreased compared to the Initial group (p = .03) in males (Figure 3H). These metrics showing significantly differences in this work were not detected previously using Raman spectroscopy measured on embedded femoral cross-sections.⁶ It should be noted that, in order to match the subset of samples tested for fracture toughness, the samples used for Raman spectroscopic analysis were a different subset from the embedded samples, although from the same animal cohort. No difference was found between any groups in proline hydroxylation and Amide I sub-peak ratios, indicating that the collagen structures were unmodified.

Figure 3.

Raman spectroscopy demonstrated changes in the mineral and organic matrix quality. Type-B carbonate substitution was lower in the Reconstituted group than in all other groups for females (A) and males (B). Crystallinity was increased in the Reconstituted group than all other groups with antibiotic dosing in females (C), while crystallinity in males was higher in the Reconstituted group than in the Continuous group (D). The mineral-to-matrix ratio was reduced in the Reconstituted group compared to the Unaltered, the Initial, and the Delayed groups for females (E) and compared to all other groups in males (F). Levels of carboxymethyl-lysine (CML) accumulation, as a measure of matrix glycation/glycoxidation, were reduced in the Reconstituted group than the Initial and the Delayed groups in females (G) and compared to the Initial group in males (H). N = 7-11 per group for females; N = 5-11 per group for males. One-way ANOVA with Games–Howell post hoc test was used for (A) and (H); one-way ANOVA with Tukey post hoc test was used for (B), (E), and (G); Kruskal–Wallis with Dunn’s post hoc test was used for (C), (D), and (F).

Reconstitution of the gut microbiome altered the relative abundance in the composition of the gut microbiome, trending toward an anti-inflammatory profile

To test the hypothesis that reconstitution of gut microbiome through FMT alters the inflammation profiles, we examined the relative abundances of several microbiota genera previously shown to have anti-inflammatory or pro-inflammatory effects (Figure 4). Here, we used data collected from the study by Liu and Cyphert et al.⁶ to conduct independent analysis. The sexes were combined for the microbiota genera analysis, since sex was not a significant factor in the relative abundance of any genera reported (all p > .5 for sex). For example, an increased abundance of the genera Ruminococcus and Roseburia can produce butyric acid, which modulates the immune system and reduces inflammation. Furthermore, Lactobacillus can support an environment with higher butyric acid content by producing lactate that cross-feeds microbes producing butyric acid.^29–33 The abundances of these 3 genera were significantly reduced in the antibiotic-dosed groups (Continuous, Initial, and Delayed) at 24 wk of age as compared to the Unaltered and the Reconstituted (all p < .001) groups. The mean abundance levels of these 3 genera were higher in the Reconstituted group than in the Unaltered group, but not significantly so (p between .3 and .7, Figures 4A-C). In contrast to above named genera, Bacteroides can produce toxins that trigger the signaling pathways and increase pro-inflammatory cytokines.³² Here, the Reconstituted group showed significantly decreased abundance than the Continuous and the Initial groups, but the levels of Bacteroides were not different from the Unaltered or the Delayed group (Figure 4D). Similarly, we note the other microbial genera associated with anti-inflammatory effects (Bifidobacterium, Anaerostipes, Coprococcus, and Butyricicoccus)^29–33 showed increased trend in the Reconstituted group, and the microbial genera associated with pro-inflammatory effects (Clostridium)³² showed decreased trend in the Reconstituted group (Figure S1). Consistent with this finding, previous work on the same mice showed that the proinflammatory cytokine TNF-α was significantly reduced in serum from the Reconstituted group compared to the Unaltered group in both sexes (both p < .001).⁶

Figure 4.

Microbiota genera associated with anti-inflammatory profiles showed decreased abundance in groups treated with antibiotics compared to the Reconstituted group at the end of the study (24 wk of age): (A) Ruminococcus, (B) Lactobacillus, and (C) Roseburia. Similarly, the microbiota genus associated with pro-inflammatory profiles, (D) Bacteroides, is decreased in the Reconstituted group compared to the Continuous and the Initial groups. Within each group, males and females were combined in this analysis. N = 10-12 per group. All statistical analyses were done using Kruskal–Wallis with Dunn’s post hoc test.

Inflammation status correlated with mineral quality and bone fracture resistance

We conducted a correlation analysis between the measured parameters to establish the link between systemic inflammation, bone mineral compositions, and bone mechanical integrity. The reported serum markers data were collected in the previous study by Liu and Cyphert et al.⁶ With all samples combined, serum TNF-α was positively correlated to type-B carbonate substitution (ρ = 0.66, p < .001, Figure 5A), mineral-to-matrix ratio (ρ = 0.71, p < .001, Figure 5B), and CML in the bone matrix (ρ = 0.43, p = .007, Figure 5C). Serum IL-6 was also tracked with type-B carbonate substitution (ρ = 0.39, p = .04, Figure 5D). The bone maximum toughness was negatively associated with levels of type-B carbonate substitution (r = −0.45, p < .001, Figure 5E), mineral-to-matrix ratio (r = −0.40, p < .001, Figure 5F), and the level of CML (ρ = −0.49, p < .001, Figure 5G). Additionally, higher maximum toughness in bone was associated with increased levels of mineral crystallinity (ρ = 0.33, p = .03, Figure 5H). The collagen structural measures (proline hydroxylation and Amide I sub-peak ratios) were not correlated with other parameters.

Figure 5.

When all samples were combined, TNF-α was positively correlated with type-B carbonate substitutions (A), mineral-to-matrix ratio (B), and CML (C), all of which were negatively associated with maximum toughness (E-G). In addition, IL-6 was also positively correlated with type-B carbonate substitutions (D). Higher crystallinity was significantly associated with increased maximum toughness. Abbreviations: CML, carboxymethyl-lysine; IL-6, Interleukin 6; TNF-α, TNF alpha.

Discussion

Changes to the composition of the gut microbiome have been shown to alter the mechanical integrity of bone. However, how the gut microbiome influences bone tissue composition and quality is not yet well understood, nor is it clear that reconstitution of the gut microbiome through FMT can benefit the bone matrix. Here, for the first time, we report that the reconstitution of the gut microbiome improves bone fracture resistance. We further establish that the enhancement of bone fracture toughness with FMT is linked to improving mineral quality in a manner correlated to reduced levels of systemic inflammation.

In a previous study of the same animal cohort, the whole-bone 3-point bending testing revealed that disruption of the gut microbiome in the Continuous, Delayed, and Initial groups reduced the tissue strength compared to the Unaltered group in males. The tissue strength was not different between the Reconstituted and the Unaltered groups.⁶ Unexpectedly, whole bone fracture toughness testing did not detect a microbiome-induced change in bone material-level fracture resistance for either sex. Here, in males, a similar trend of decline in maximum toughness in the Continuous and Initial group compared to the Unaltered group approaching statistical significance. However, the number of samples allocated for fracture toughness testing was not sufficiently powered. It is worth noting that the tissue strength and fracture toughness are distinct material properties and may not be always correlated. In this case, the whole-bone tissue strength (defined by the maximum bending moment for a given section modulus) is significantly reduced, whereas no difference in bone tissue ductility/brittleness is detected with antibiotic dosing compared to the unaltered controls. Moreover, the previous study highlighted the sex-related differences in whole-bone tissue strength, perhaps due to the different response to antibiotic dosing between males and females.⁶ In this work, however, similar trends in both bone fracture toughness and compositional properties were observed across sexes. The reason might be that the identified associative pathway through inflammation appeared not to be sex dependent. Both the serum inflammation markers presented in the previous work⁶ and the microbiome genera selected in this study did not demonstrate sex-related differences.

An alternate explanation for the discrepancy between the 2 tests could be that the notch for fracture toughness testing was placed on the anterior side of the femur. This means that the location with active bone formation on the periosteal surface⁶ mostly did not contribute to the tissue fracture resistance. Potential changes occurring at the anterior side of the femur caused by antibiotic dosing may not be captured fully by this test. We chose to place the notch on the anterior side of the femur to keep the posterior side in compression, similar to physiological loading. The notching on the anterior side also allows for more consistent sample placement. That said, serum analyses on these mice demonstrated that neither bone formation (P1NP) nor bone resorption tartrate-resistant acid phosphatase 5b (TRAP5b) decreased following antibiotic dosing,⁶ suggesting that bone remodeling at the periosteum envelop of anterior quadrant would unlikely influence bone mechanical outcomes. In contrast, in the Reconstituted group, where enhanced fracture toughness was captured, Raman spectroscopy did reveal significant chemical compositional changes in the bone matrix, without substantial changes in bone turnover.

Our correlation analysis demonstrated that the quality of bone mineral and organic matrix compartments could explain variations in maximum toughness (or fracture toughness at the maximum load). A negative correlation was shown between maximum toughness and type-B carbonate substitution. Type-B carbonate substitution, referring to the substitution of phosphate for carbonate in bone mineral, is often increased in aging^16,17 and osteoporosis,¹⁸ and can capture changes in mineral crystal quality and contribute to overall bone tissue functionality.³⁸ Higher levels of type-B carbonate substitution has been shown to create vacancies and distortions in the crystal lattice.³⁹ The alterations of the lattice organization, resulting from the substitution of larger tetragonal phosphate groups with smaller trigonal carbonate groups, can cause the lattice to be unevenly strained. This substitution is also associated with calcium deficiency and sodium replacement in the lattice structure. These changes to the crystal lattice can most likely lead to a decrease in mineral stiffness and negatively affect the bone matrix mechanical properties.⁴⁰ A previous study showed that Raman spectroscopy-assessed type-B carbonate substitution was sensitive to predicting the decline in elastic modulus, yield strength, ultimate strength, toughness, and post-yield toughness of human bone tissue.¹⁹

The reduction in mineral-to-matrix ratio (assessed by either the ratio of v1PO4³⁻ and proline or the ratio of v2PO3³⁻ and Amide III) in the Reconstituted group, reported here, may indicate a lesser amount of bone mineral compared to surrounding organic matrix for a given rate of turnover that is not reflected in tissue mineral density.⁶ The reason for the altered mineral-to-matrix ratio in this group is not yet understood. Nevertheless, for a set level of mineral content, higher relative organic matrix level could improve the ductility and resistance to fracture in bone tissue. Therefore, the changes in the mineral-to-matrix ratio reported here could also explain the observed differences in bone fracture toughness. Similarly, previous work on human cortical bone has shown that the mineral-to-matrix ratio is indeed a negative predictor of bone fracture toughness.⁴¹

Furthermore, crystallinity is another critical indicator of bone mineral quality that can be measured with Raman spectroscopy. Crystallinity measures the shape and stoichiometry of bone apatite, which has been shown to track with crystal size and lattice perfection.⁴² In human bone, crystallinity was previously shown to predict increased bone strength, including elastic modulus, yield strength, and ultimate strength.¹⁹ Consistent with our findings, a significant positive relationship between bone fracture stress and crystallinity was also reported in a prior study.⁴² The knowledge on how the size and perfection of the mineral crystal influence bone mechanical integrity is limited, especially for post-yield properties. However, it has been hypothesized that increased crystallinity may contribute to higher micro-strain in the bone matrix.⁴³ Additionally, the levels of type-B carbonate substitution also affect mineral crystallinity, because more carbonate content results in smaller crystals and therefore correlates with decreased crystallinity.^17,44 This negative correlation was also observed in our study (Spearman’s ρ = −0.54, p < .001). Larger crystal size (corresponding to higher crystallinity) allows for more interaction between collagen and mineral, thereby enabling more energy to be dissipated during the phase separation.¹⁹ Taken together, the significant reduction of type-B carbonate substitution and mineral-to-matrix ratio in both sexes and the increase in mineral crystallinity (only significant in females) may contribute to the enhancement of bone fracture resistance in the Reconstituted group.

In our previous work on the mice from the same study cohort, Raman spectroscopic analysis was conducted on the mid-diaphysis cross-section of the right femora, where, in comparison with the Unaltered group, no mineralization difference was observed.⁶ In contrast to these previous results, our current study on the periosteal surface was able to distinguish the differences in mineralization assessed in terms of type-B carbonate substitution, mineral-to-matrix ratio, and crystallinity. It is noteworthy that the previous work using Raman spectroscopy focused on compositional differences in bone tissue formed in different periods (12-16 and 16-24 wk) and to achieve this, cross-sections from the bone tissue near the fracture site were used. Although such an approach provides a valid way to compare changes measured at similar locations within the cross-section, it may not reflect the matrix changes associated with alterations in the gut microbiome. To address this limitation, here we directly examined the bone compositional quality on the fresh, unembedded periosteal surface, which, due to its location on the outer circumference, is likely to be a significant contributor to the fracture properties of bone.

The discrepancies between Raman spectroscopic parameters obtained from periosteal surface and cross-section were noted in a previously published work,⁴⁵ but not fully explained. We speculate the following reason to explain the differences between the 2 studies. The previous study utilized dehydrated and poly (methyl methacrylate)-embedded bone sections. The dehydration-based fixation method and the embedding media are known to have modest effects on bone mineralization measures.⁴⁶ Moreover, the cross-sections in the previous study were cut from the location next to the fracture site post-mechanical testing. And because fracture site varies between specimens and Raman spectroscopy parameters are also subjected to spatial variation, larger variations would be present within a group, which, in turn, would mask differences across the groups. Furthermore, microdamage and large cracks near the fracture site may impact or otherwise bias the measured outcomes. The current study minimizes such variations by taking point spectra along the longitudinal axis and averaging the measurements. Furthermore, the previous study was limited to a smaller sample size (N = 5 per group). In contrast, this study was conducted with a larger sample size (different subset from the same animal cohort) and, due to methodological improvements and differences noted above, showed a much-reduced coefficient of variation in all Raman spectroscopic parameters and significant correlations with fracture properties.

In clinical practice, FMT from a healthy donor has been shown to restore balance in the gut microbiome and provide healthy saprophytic bacteria to reduce systemic inflammation levels.⁴⁷ Previously in this animal cohort, we demonstrated this effect of FMT by showing the significant reduction of the proinflammatory marker TNF-α in the Reconstituted group for both sexes.⁶ Interestingly, despite the composition of the gut microbiome in the Reconstituted group after FMT was more similar to that of the Unaltered group compared to other groups, the relative abundance of a variety of microbiota genera can be different between the Unaltered and the Reconstituted group. Specifically, we highlight that the Reconstituted group had significantly elevated levels of healthy bacteria that had been shown to promote anti-inflammatory responses (Figure 4 and Figure S1), which might contribute to this systemic effect. It should be noted that the microbial assessments using 16S rRNA sequencing are limited to draw definitive conclusions on tissue functionality outcomes. Yet, these key associations in our work lay the groundwork for future work that can probe the mechanistic insights.

A few studies have investigated the influence of inflammation on bone matrix quality. For example, inflammation has been positively associated with type-B carbonate substitution level through induced local inflammation^14,15 or KO of inflammation regulator RAGE to reduce systemic inflammation.²¹ One study hypothesized that this correlation between inflammation and type-B carbonate substitution may be caused by the increased turnover, leading to more immature bone.¹⁴ However, our data suggests otherwise since serum P1NP and TRAP5b were not higher in the Reconstituted group compared to other groups.⁶ Furthermore, conflicting results regarding the effects of inflammation on mineral crystallinity have been reported.^15,21 No prior research has investigated the relationship between inflammation and bone mineral-to-matrix ratio without significant changes in bone mineral density or bone turnover. Nevertheless, our study shows that reducing inflammation levels in the Reconstituted group most likely explains the improvement of bone mineral quality without the occurrence of substantial bone turnover mediated by osteoblasts and osteoclasts. The interaction between inflammation and mineral quality at the submicron level or below could be linked to other mechanisms, such as changes in metabolites that regulate biomineralization or altered osteocytes-mediated peri-lacunar remodeling that is not reflected in serum turnover markers, which need further investigation.

As for organic matrix quality, we discovered that the increase in maximum toughness can also be explained by the decrease in CML levels accumulated in the bone matrix. Carboxymethyl-lysine is a non-crosslinking advanced glycoxidation product (AGOE) formed under hyperglycemia and elevated oxidative stress conditions.⁴⁸ Its influence on bone fracture properties is hypothesized to be through the formation of bridges between organic matrix and bone mineral⁴⁹ that will make bone tissue more brittle. The accumulation of CML in bone is negatively correlated to bone crack propagation toughness.⁵⁰ In this work, we observed a significant positive association between TNF-α and bone CML accumulation, suggesting the oxidative-inflammatory cascade being the underlying mechanism driving the CML differences. However, the level of CML in bone in the Reconstituted group was not significantly lower than in the Unaltered group (a trend was suggested for females, p = .09). Microbiome analysis also revealed that the bacteria beneficial to the regulation of glucose metabolism (eg, Akkermansia) did not increase in abundance following FMT.⁶ It can be hypothesized that considering the young age of the mice in this study, antibiotic dosing was insufficient to disrupt glucose metabolism and, therefore, did not lead to more dramatic differences in CML accumulation in bone. However, this study did not monitor blood glucose levels to verify this hypothesis. Yet, the CML level in the Reconstituted group was significantly lower than that in the Initial group, where the gut microbiome compositions were altered similarly to the Reconstituted group but without the subsequent FMT intervention. To understand the causative role of FMT in modulating oxidative-inflammatory cascade and consequently improve bone matrix quality, further investigation in a model where the glucose metabolism is more severely impacted, such as type 2 diabetes, is needed.

In conclusion, this study demonstrated that reconstitution of the gut microbiome through FMT improved bone fracture resilience. Such improvement is likely due to the reduction in systemic inflammation caused by FMT from a healthy donor, which leads to ameliorated bone mineral crystal composition and quality, without substantial bone turnover. These findings contribute to the idea that the gut microbiome and systemic inflammation can influence bone quality.

Author contributions

Bowen Wang (Data curation, Formal analysis, Investigation, Methodology, Validation, Writing—original draft, Writing—review & editing, Visualization), Samuel J. Stephen (Data curation, Formal analysis, Investigation, Methodology, Software, Writing—review & editing), Erika L. Cyphert (Data curation, Formal analysis, Investigation, Methodology, Writing—review & editing), Chongshan Liu (Data curation, Formal analysis, Investigation, Methodology), Christopher J. Hernandez (Conceptualization, Funding acquisition, Methodology, Project administration, Supervision, Writing—review & editing), and Deepak Vashishth (Conceptualization, Investigation, Methodology, Project administration, Supervision, Writing—review & editing)

Funding

This work was supported by NIH R01AG067997 (C.J.H.), NIH F32AG076244 (E.L.C.), and Chan Zuckerberg Biohub (C.J.H.).

Conflicts of interest

The authors declare no potential conflict of interest.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.