Engineering bioinspired, high-density collagen microgels with tunable intrafibrillar mineralization for accelerated osteogenesis in vitro and bone regeneration in vivo

Abstract

The development of biomaterials that mimic native bone remains a major challenge in regenerative medicine. Here, we present a bioinspired platform using high-density collagen hydrogels with tunable mineral content. These engineered microenvironments promote rapid osteogenesis in vitro without osteogenic supplements and accelerate bone regeneration in vivo in critical-sized defects. By modulating mineralization, we demonstrate that early mechanosensitive signaling in human mesenchymal stem cells is linked to matrix stiffness and biochemical composition. Within two hours, focal adhesion formation decreased with increasing mineral content, and fully mineralized scaffolds significantly increased nuclear YAP1 localization. By 24 h, RUNX2 expression was markedly increased in fully mineralized scaffolds, with 40.7 ± 3.9% RUNX2⁺ nuclei (p < 0.0001), and this trend persisted at the gene expression level at 3 days. In a rat calvarial defect model, fully mineralized microgels significantly increased bone volume in males at 12 weeks (18.99 ± 2.66 mm³) compared to empty defects (11.60 ± 2.12 mm³, p = 0.0242), whereas females showed no added benefit of full mineralization. Two-way ANOVA confirmed significant effects of sex (p = 0.0006), treatment (p < 0.0001), and their interaction (p = 0.0158) Histological analyses confirmed osteoinductive behavior across all microgel groups and highlighted reduced scaffold degradation and limited cellular infiltration in mineralized conditions. Together, these results demonstrate that tunable intrafibrillar mineralization modulates early stem cell mechanosensing and osteogenic priming in vitro and drives sex-dependent regenerative outcomes in vivo, emphasizing the need to balance scaffold mechanics and degradation to suit the biological context and improve clinical outcomes.

1. Introduction

Large bone defects caused by trauma, cancer, infections, or disease require complex surgical interventions and prolonged care, posing a major clinical challenge and socioeconomic burden [1–4]. Globally, millions of bone grafting procedures are performed each year, and the cost of treating bone defects in the United States alone is estimated at $5 billion annually [5,6]. Although autografts remain the clinical gold standard, they are limited by donor-site morbidity and tissue availability [7,8]. Alternative approaches, including commercially available biomaterials loaded with growth factors, offer some therapeutic benefit but pose risks for serious adverse outcomes, such as life-threatening inflammation and ectopic bone formation [9,10]. Bioengineered bone scaffolds show promise in addressing these challenges yet face high failure rates due to difficulties in replicating the intricate biophysical and biochemical properties of native bone [11–13]. Therefore, there remains an unmet need for minimally invasive bone grafting solutions that match the effectiveness of autografts while mitigating their drawbacks [1–3,7,13–15].

Efforts in bone tissue engineering have increasingly focused on bioinspired scaffolds that replicate the physical and biochemical cues of the native bone extracellular matrix (ECM). Among these, collagen-based biomaterials are commonly used due to their biocompatibility and collagen’s abundance in the bone ECM, yet their performance as standalone substitutes is limited by insufficient mechanical strength and rapid degradation unless further modified or reinforced [13,16]. To overcome these limitations, researchers have incorporated mineral into collagen scaffolds in efforts to mimic native bone structure and function, where densely mineralized matrix and resident cells work synergistically to support mechanical integrity, cellular signaling, and tissue regeneration. In native bone, mineralization is a nanoscale, protein-mediated process involving the deposition of calcium within and around collagen fibrils, resulting in the formation of intrafibrillar and extrafibrillar hydroxyapatite (HAP) crystals [17], resulting in a far more complex process of scaffold mineral integration in comparison to simple loading of HAP particles. This precisely controlled mineral distribution contributes to the mechanical strength of bone while also generating biophysical and ionic cues that regulate stem cell adhesion, mechanosensing, and osteogenic differentiation [18,19]. Prior work recapitulated this complex process in vitro by using non-collagenous protein analogues, such as polyaspartic acid and milk-derived osteopontin (OPN), which regulate mineral deposition and enable precise control over nanoscale mineral architecture and matrix stiffness [18,20–23]. These advances have facilitated the development of more biomimetic engineered microenvironments, that can offer tunable mineral content to better mimic the dynamic properties of native bone ECM. However, while these biomaterials are increasingly used for bone tissue engineering, the mechanistic pathways by which progressive mineralization regulates early cell-matrix interactions remain poorly defined, thus impairing the rational design of more effective regenerative scaffolds.

Mechanotransduction—the process by which cells convert matrix-derived physical signals such as stiffness or topography into biochemical responses—is central to how human mesenchymal stem cells (hMSCs) sense and respond to mineralized scaffolds [2,4,24–26]. Through integrin engagement, cytoskeletal tension, and focal adhesion assembly, cells dynamically engage with their microenvironment, often described as a molecular clutch [26,27]. This mechanism allows cells to modulate traction forces and adapt to mineral-induced properties of the surrounding microenvironment such as stiffness and topography [28,29]. Although matrix stiffness has been widely studied as a regulator of stem cell differentiation, fewer studies have addressed how graded or progressive mineralization influences this molecular clutch mechanism and downstream osteogenic signaling [30,31].

To address this knowledge gap, we developed a collagen-based hydrogel platform with tunable mineral content that recapitulates bone matrix maturation, enabling us to isolate matrix-derived cues on hMSC adhesion, mechanosensing, and early osteogenic commitment, thereby advancing mechanistic insight into cell-ECM interactions in bone regeneration. From that, we evaluated the in vivo potential of these mineralized microgels as a syringe-delivered, cell-free bone substitute. Using a rat calvarial defect model, we investigated whether these scaffolds could promote host cell recruitment and support endogenous bone healing. By bridging in vitro mechanistic insights with in vivo regenerative outcomes, our dual approach addresses both the biological mechanisms of growth factor free, scaffold-mediated osteogenesis and the translational potential of mineralized materials for bone repair. Together, these studies inform the rational design of advanced biomaterials that are biologically instructive, minimally invasive, and clinically effective.

2. Materials and methods

2.1. Fabrication of high-density collagen hydrogels

Hydrogels were prepared as previously described [18], with modifications to enable densification and tunable mineralization. Stock acid-solubilized type I collagen from rat tail tendon (3 mg/mL; Gibco, A1048301) was diluted on ice to a working concentration of 1.5 mg/mL by mixing with 10X phosphate-buffered saline (PBS; Invitrogen, AM9624), 1.0 N NaOH (Sigma Aldrich, S2770–100ML), and calcium- and magnesium-free 1X Dulbecco’s phosphate-buffered saline (DPBS; Gibco, 14190–144) to adjust the pH to 7.4. The collagen solution was cast into well plates and incubated at 37°C for 15 min to allow for self-assembly and the initiation of fibrillogenesis. Immediately following polymerization, gels were densified by centrifugation at 3500 g for 20 min at room temperature to remove interstitial fluid and compact the matrix.

2.2. Controlled and tunable intrafibrillar mineralization and fabrication of collagen microgels

To generate controlled mineralization of collagen hydrogel constructs, mineralization media was prepared by supplementing low glucose Dulbecco’s Modified Eagle Medium (DMEM; Gibco, 12–320–032) with 10 % Fetal Bovine Serum (FBS; Gibco, 12,662,029), and 1 % penicillin-streptomycin (Gibco, 15140163), 4.5 mM CaCl₂ (Fisher, BP510–500), 100 μg/mL osteopontin (OPN; Lacprodan^® OPN-10, Arla Foods Ingredients Group P/S), and 2.1 mM K₂HPO₄ (J.T. Baker, 3252–05). CaCl₂ was added first to the complete DMEM media, followed by OPN and then K₂HPO₄, with each step vortexed to ensure proper mixing and prevent premature precipitation. Non-mineralized gels were prepared without exposure to mineralization media, whereas partially mineralized and fully mineralized gels were incubated in mineralization media at 37°C on a 2D rocker with daily media changes for 1 or 3 days, respectively. The hydrogels were then micro-sectioned into 500 μm × 500 μm microgels using an automated tissue sectioning device (Ted Pella, Inc., 10,180) and stored in 1X DPBS at 4°C until further use. Microgels were delivered using a syringe-based method, consistent with our prior work demonstrating the injectability of larger microgels [9,32].

2.3. Material characterization

2.3.1. Scanning electron microscopy (SEM)

Bulk hydrogel samples were fixed using 2 % glutaraldehyde (Fischer Chemical^™, O2957–1) prepared in 0.2 M sodium cacodylate buffer (pH 7.4, Electron Microscopy Sciences, 11,653) for 1 hour. Following fixation, samples were rinsed three times in water for 5 min each. If immediate processing was not feasible, samples were stored overnight at 4°C with 1X DPBS. Samples were dehydrated through a graded ethanol series as follows: 50 % ethanol for 5 min, 75 % ethanol for 5 min, 90 % ethanol for 10 min, and two changes of 100 % ethanol for 10 min each. All incubation periods were performed at room temperature on a 2D rocker unless otherwise stated. Dehydrated samples were subjected to critical point drying for 3 h using a Leica CPD300 system. Dried samples were then mounted and sputter coated for 30 min using a Leica ACE600 system, applying a platinum layer of 10.0 ± 0.5 nm at a sputter rate of 0.14 ± 0.02 nm/s. Imaging was conducted using a Helios UC5 SEM. All sample processing and imaging were performed at least in three independent biological replicates (n = 3).

2.3.2. Fourier transform infrared (FTIR) spectroscopy

Samples were prepared by flash freezing in liquid nitrogen followed by lyophilization. FTIR analysis was performed using a Nicolet 6700 spectrometer (Thermo Scientific) equipped with a deuterated triglycine sulfate with potassium bromide (DTGS KBr) detector and a KBr beamsplitter, operated in transmission mode. Spectra were collected over the range of 4000–400 cm⁻¹ at a resolution of 4 cm⁻¹ using 128 co-added scans for both background and sample. Instrument parameters included Happ-Genzel apodization and Mertz phase correction. Using Spectragryph software, spectra were converted to absorbance mode, subjected to advanced baseline correction, smoothing, and normalization to the amide I peak at 1650 cm⁻¹, which was set to an intensity value of 1. The mineral-to-matrix ratio was determined by comparing the peak height of the phosphate (PO₄) band at 1030 cm⁻¹ to that of the amide I band at 1650 cm⁻¹. All measurements were performed at least in three independent biological replicates (n = 3).

2.3.3. Mechanical properties

Hydrogel mechanical properties were measured using a Discovery Hybrid Rheometer-1 (TA Instruments) equipped with an 8 mm parallel plate geometry. Independently prepared hydrogel samples (n = 3–5) were loaded with a 100 μm measurement gap. Oscillatory frequency sweeps (0.1 Hz to 10 Hz) were conducted at room temperature with a constant 1.0 % strain. A frequency of 1 Hz was used to determine the storage (${\mathrm{G}}^{\prime }$) and loss (${\mathrm{G}}^{″}$) modulus. The bulk elastic modulus ($\mathrm{K}$) was calculated from the storage modulus using the relation:

$$K=\frac{2G(1+v)}{3(1-2v)}$$

where $G$ is the shear modulus and $v$ is the Poisson’s ratio (assumed to be 0.44 for hydrogels), as previously described [33].

2.4. Cell culture

Human bone marrow-derived mesenchymal stem cells (hMSCs; RoosterBio, Inc., MSC-001) were expanded and used at early passages (<P5) to avoid alterations in the phenotype [34]. Cells were cultured in Minimum Essential Medium $\mathrm{\alpha }$ (MEM $\mathrm{\alpha }$; Gibco, 12,571,048) supplemented with 10 % FBS (Gibco, 12,662,029), and 1 % penicillin-streptomycin (Gibco, 15,140,163) at 37°C in a humidified incubator with 5 % CO₂. Sub-culturing was performed at 80–90 % confluency using 0.25 % trypsin-EDTA (Gibco, 25,200,114). Cells were seeded onto hydrogel surfaces at a density of 3 × 10⁴ cells/cm² to facilitate the investigation of cell-ECM interactions while minimizing the effects of cell-cell crosstalk.

2.5. Immunofluorescence staining and quantification of cells on collagen hydrogels

2.5.1. Staining protocol

After gently aspirating media with a handheld pipette, samples were rinsed twice with 1X DPBS and fixed in 10 % neutral buffered formalin (Thermo Scientific, 5701) for 15 min. Samples were then rinsed three times with 1X DPBS. If not processed immediately, samples were stored in 1X DPBS at 4°C, sealed with Parafilm^®. For permeabilization, samples were incubated in 0.1 % Triton^™ X-100 (Sigma, X100–100ML) for 15 min, followed by three rinses with 1X DPBS. Blocking was performed in 1.5 % bovine serum albumin (BSA; Fisher, BP1600–100) for 1 hour. Primary antibody solutions were prepared in 0.15 % BSA and added after blocking. Primary antibodies used were Phospho-Paxillin (69363S, Cell Signaling Technology, 1:100), Piezo1 (NBP1–78,537, Novus Biologicals, 1:200), YAP1 (WH0010413M1, Sigma Aldrich, 1:200), and RUNX2 (NBP1–77,461, Novus Biologicals, 1:100). Samples were incubated overnight at 4°C on a 2D rocker. Samples were rinsed three times with 1X DPBS and incubated with secondary antibodies prepared in 0.15 % BSA. All secondary antibodies used were Alexa Fluor^® 488 Alpaca Anti-Rabbit IgG (611–545–215, Jackson ImmunoResearch Inc., 1:250), Cy^™5 Alpaca Anti-Mouse IgG (615–175–214, Jackson ImmunoResearch Inc., 1:250), and Alexa Fluor^® 488 Goat Anti-Rabbit IgG (A-11,008, Invitrogen, 1:250). Secondary antibodies were centrifuged at 10,000 rpm for 1 min prior to use to remove aggregates that could contribute to nonspecific background staining. Counterstaining was performed using F-Actin Phalloidin (Invitrogen, R37112) and NucBlue^™ (Invitrogen, R37606), following the manufacturer’s recommended dilutions and incubation times. Samples were rinsed twice with 0.1 % Tween 20 (Fisher, BP337–500) for 5 min each, followed by a final rinse in 1X DPBS. Samples were stored at 4°C in 1X DPBS, wrapped in Parafilm^®, protected from light, and imaged within one week. All incubation steps were performed at room temperature on a 2D rocker, all rinses included a 5-minute incubation period, and all solutions were prepared in 1X DPBS unless otherwise stated. A laser-scanning confocal microscope (Zeiss LSM 880 with Airyscan) was used for immunofluorescence imaging. For focal adhesions, super-resolution images were acquired using Airyscan microscopy to ensure sufficient resolution for quantifying small subcellular structures such as paxillin clusters. All experiments were performed in three independent biological replicates (n = 3). For imaging-based quantification, a variable number of cells were analyzed per sample depending on image quality and field of view. Exact cell counts are reported where applicable.

2.5.2. Quantification of immunofluorescence imaging

Mechanobiology quantification of immunofluorescent images was performed with Imaris x64 Microsoft Image Analysis Software (v. 10.1.1.; Oxford Instruments). Paxillin quantification was performed using the Surfaces module. Cell segmentation was performed to create individual cell boundaries using actin staining as a mask. Paxillin clusters were then identified within each cell, and measurements were exported on a per-cell basis. For Piezo1, the Surfaces module was similarly used to segment individual cells and quantify the mean Piezo1 intensity per cell. YAP1 nuclear fraction was calculated using the Cells module. The cellular and nuclear compartments were segmented based on actin and DAPI channels, respectively. YAP1 fluorescence intensity was measured in both compartments, and the nuclear fraction was computed using the formula:

$${NuclearFraction}_{\mathit{per}\mathit{cell}}=\frac{\mathit{Mean}\mathit{Nuclear}\mathit{Intensity}}{\mathit{Mean}\mathit{Cell}\mathit{Intensity}}$$

Immunofluorescence images of RUNX2-stained samples were quantified using an automated workflow developed in Fiji (ImageJ) and R [35,36]. Image segmentation was performed using a custom ImageJ macro script (Supplementary Script 1). Nuclear-based segmentation was performed using DAPI as the primary mask, and cell-based segmentation used actin as the cytoplasmic boundary. The resulting data were processed in R (v4.2.2) and visualized using GraphPad Prism (v10.5.0).

2.6. Real-time quantitative polymerase chain reaction (qPCR)

For each biological replicate, six microgels were pooled and RNA was isolated using the RNeasy Micro Kit (Qiagen, 74,004) according to the manufacturer’s protocol. Samples were sonicated using the Diagenode Bioruptor Pico Sonicator (Hologic, B01080010) for 3 cycles of 30 s on/off to improve RNA extraction. RNA concentration was measured using the Qubit RNA HS Assay Kit (Invitrogen, Q32855) and the Qubit Flex Fluorometer (Invitrogen, Q33327). Up to 25 ng of RNA was reverse transcribed into complementary DNA (cDNA) using SuperScript IV VILO Master Mix (Invitrogen, Q32855) on a ProFlex PCR System (Applied Biosystems, 4484,073) according to the manufacturer’s protocol. The resulting cDNA was diluted 1:4 in RNase/DNase-free water to ensure sufficient volume for downstream assays. qPCR was performed on a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems, 4485,691) using TaqMan Fast Advanced Master Mix (Applied Biosystems, 4444,557) and TaqMan Gene Expression Assay FAM-labeled probes (Applied Biosystems, 4331,182) for target genes: RUNX2 (Hs01047973_m1), ALPL (Hs01029144_m1), BGLAP (Hs01587814_g1), and GAPDH (Hs02786624_g1). Each reaction was carried out in technical duplicate. No-reverse transcriptase (-RT) and no-template (water) controls were included for each target gene. Cycling conditions were 50 °C for 2 min (UNG incubation), 95 °C for 2 min (polymerase activation), followed by 40 cycles at 95 °C for 1 s (denaturation) and 60 °C for 20 s (annealing). Gene expression was normalized to GAPDH as an internal control. Fold change in gene expression was calculated using the comparative Ct method (2^−ΔCt), where ΔCt represents the difference in average threshold cycle (Ct) values between the gene of interest and the reference gene [37,38]. All experiments were performed in three independent biological replicates (n = 3).

2.7. Surgical procedure for rat calvarial defect implantation

All animal procedures were conducted in accordance with the OHSU Institutional Animal Care and Use Committee (IACUC, Protocol ID: #IP00004736). Male and female Wistar IGS rats (Charles River Laboratories, Strain Code 003) of 9–11 weeks in age were used. Animals were acclimated for a minimum of 7 days in a temperature- and humidity-controlled environment with a 12-hour light/dark cycle and had ad libitum access to food and water. An a priori power analysis was conducted using G*Power 3.1 (effect size f = 0.8, α = 0.05, power = 0.90, five groups), which indicated a minimum total sample size of 30 animals was required [39]. A total of 32 rats were included: 17 males and 15 females. Animals were distributed across the five treatment groups as follows: empty defect (n = 8; 4 females, 4 males), autologous bone (n = 6; 4 females, 2 males), non-mineralized microgels (n = 6; 3 females, 3 males), partially mineralized microgels (n = 6; 3 females, 3 males), and fully mineralized microgels (n = 6; 3 females, 3 males). Animals were randomly assigned to treatment groups, and investigators were blinded to group assignments during data analysis.

Pre-operatively, animals were given subcutaneous meloxicam (5 mg/kg), enrofloxacin (5 mg/kg), and local bupivacaine (6 mg/kg). Anesthesia was induced with 5 % isoflurane and maintained at 1.5–3 % in oxygen. Proper body temperature was maintained with a heating pad, warming blanket, and rectal monitoring. Anesthesia depth was assessed periodically via a toe pinch and observation of respiratory rate. The surgical area was shaved and disinfected with alternating povidone-iodine and ethanol wipes, followed by placement of a sterile drape.

A midline sagittal incision was made, and the skin, fascia, and periosteum were bluntly reflected to expose the calvarial surface. A full-thickness 8 mm circular defect was created in the middle of the parietal bones using a trephine burr under continuous saline irrigation. The bone was removed with care to avoid damage to the underlying tissues. For the autologous bone group, the bone removed during the defect procedure was immediately reimplanted to model the clinical practice of a decompressive craniotomy. Microgels were implanted directly into the defect site using a syringe and allowed to conform in situ. Each defect received approximately 200 μL of microgel material suspended in 200 μL of sterile 1X PBS, for a total dispensed volume of 400 μL, depending on the amount needed to fully fill the defect without overflow. The incision was closed using wound clips. For surgeries exceeding 60 min, animals received 20 mL/kg subcutaneous saline intraoperatively.

Post-operatively, meloxicam (5 mg/kg) was administered subcutaneously every 24 h for 3 days and enrofloxacin (5 mg/kg) for 1 day, and animals were monitored daily for the first week and regularly thereafter for signs of pain, infection, wound dehiscence, and general well-being. Wound clips were removed 7–10 days post-surgery. Bone regeneration was monitored longitudinally using in vivo micro-computed tomography at 4-week intervals under anesthesia. At 12 weeks post-surgery, animals were euthanized by CO₂ inhalation followed immediately by terminal cardiac blood collection via sternotomy.

2.8. Micro-computed tomography (micro-CT)

2.8.1. In vivo longitudinal micro-CT imaging and analysis

Calvarial bone regeneration was monitored longitudinally using a Siemens Inveon micro-CT system. Rats were anesthetized with isoflurane during scanning, following IACUC-approved guidelines and consistent with the parameters described in the surgical methods. Projections were acquired in step-and-shoot mode with a full 360° rotation and 1440 projections spaced at 0.25° increments. A 0.5 mm aluminum filter was used with source parameters set to 80 kV and 0.5 mA, and an exposure time of 720 ms per projection. Images were captured using 2 × 2 binning at medium-high magnification, resulting in an effective voxel size of approximately 29 μm. A hydroxyapatite calibration phantom was scanned during each imaging session and used to generate a linear regression for converting Hounsfield Units (HU) to mineral density, enabling quantitative comparisons across timepoints.

In vivo micro-CT data were analyzed using 3D Slicer software (v5.6.2), an open-source platform for medical image informatics [40]. Digital Imaging and Communications in Medicine (DICOM) files were imported and spatially aligned to ensure consistent orientation across timepoints. The calvarial defect region was identified and selected based on the earliest available scan for each animal. Segmentation was performed using a minimum threshold of 2000 HU to isolate mineralized tissue. Segment volume and intensity measurements were obtained using the Segment Statistics module. All segmentations followed a standardized in-house protocol.

2.8.2. Ex vivo high-resolution micro-CT imaging and analysis

At 12 weeks post-implantation, the harvested calvaria were scanned by a micro-CT system (Scanco vivaCT80) at 70 kVp, 114 μA, an integration time of 750 ms, and a voxel size of 15.6 μm. Fixed rat heads were positioned in 50 mL conical tubes with the nose oriented toward the tapered end of the tube. Scans were acquired with axial direction (z-axis) aligned with the long axis of the skull (from incisors to caudal surface) and subsequently digitally reoriented (Scanco software) such that the z-axis passed through the defect (parietal to mandible direction). These image sequences were segmented (global intensity thresholding) to generate 3D reconstructions of the skulls, and bone volume and mineral density in the 8 mm defect regions were computed.

2.9. Histological staining

Following euthanasia, rat heads were dissected to preserve the overlying skin above the defect. Surrounding soft tissues were trimmed to improve fixation with 10 % neutral-buffered formalin (Epredia, 9400–1) at room temperature for 48 h. After micro-CT scanning, calvarial samples were then decalcified in 0.5 M ethylenediaminetetraacetic acid (EDTA; Fisher, S311–500) at pH 7.6 ± 0.2 for 4 months at 4°C on a 2D rocker with weekly solution changes. As the tissue softened over time, further trimming was performed to improve decalcification and isolate the region of interest, defined by anatomical boundaries including the medial canthi of the eyes rostrally, zygomatic arches laterally, and interparietal bone caudally and sliced through the middle of the brain tissue. Once decalcification was complete, samples were dehydrated, paraffin embedded, and serially sectioned into 10 μm ± 5 μm coronal sections. Tissue sections were stained using hematoxylin and eosin (H&E), Masson’s trichrome, and picrosirius red staining according to standard protocols [41]. All histological images were obtained using a digital slide scanner (Zeiss Axioscan 7) running Zen 3.12 using a 10 × 0.45NA Plan-Apochromat objective with z-stack acquisition. Brightfield images were captured by a Zeiss Axiocam 705c (0.346 μm/pixel).

2.10. Blood collection

Immediately following euthanasia, a sternotomy was performed to expose the thoracic cavity. This involved making a midline incision caudal to the xiphoid process and carefully opening the sternum bilaterally to retract the chest wall and access the heart while still beating. Whole blood was collected via intracardiac puncture using a sterile 18 G needle and syringe, then transferred into tubes containing EDTA anticoagulant (LTT-MINI top tube; IDEXX BioAnalytics). The needle was removed prior to dispensing the collected blood into the tubes, which were filled to the designated line (maximum 0.5 mL) to ensure the proper blood-to-anticoagulant ratio. Tubes were gently inverted five to ten times to prevent clotting, refrigerated immediately, and shipped on cold packs for analysis by IDEXX BioAnalytics.

2.11. Statistical methods

All statistical analyses were performed using GraphPad Prism (v10.5.0, GraphPad Software, LLC). For comparisons between two groups, a two-tailed unpaired Student’s t-test was used. For analyses involving more than two groups, analysis of variance (ANOVA) was conducted. For in vitro experiments, statistical significance was assessed using one-way ANOVA, followed by Tukey’s post hoc test. For in vivo endpoint analyses, data were analyzed using a two-way ANOVA to assess the main effects of treatment and sex, as well as their interaction. Fisher’s least significant difference (LSD) test was used for post hoc pairwise comparisons following identification of a significant overall effect, enabling biologically relevant comparisons while maintaining sensitivity given the limited group sizes. For in vivo longitudinal micro-CT data, where repeated measurements were collected from the same animals over time, statistical significance was assessed using a mixed-effects model with time treated as a repeated factor followed by Tukey’s post hoc test. All quantitative results are expressed as mean ± standard deviation (SD), unless otherwise stated. A p-value <0.05 was considered statistically significant. Statistical significance was defined as follows: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. Datasets were checked for outliers using robust regression and outlier removal (ROUT) method (Q = 0.1 %).

3. Results

3.1. Fabrication of high-density collagen microgels with tunable mineralization alters nanoscale topography and mechanical properties

The workflow pipeline to fabricate high-density collagen-based scaffolds with controllable degrees of mineralization is illustrated in Fig. 1A. Type I collagen was allowed to self-assemble into hydrogels, which were subsequently densified through centrifugation (Fig. S1A). Tunable mineralization was achieved, as previously described [18], by exposing the hydrogels to mineralization media for either one or three days to yield partial and fully mineralized hydrogels, respectively. Microgels were produced by micro-sectioning the bulk hydrogels into 500 × 500 × 200 μm³ constructs (Fig. 1B). SEM imaging revealed increasing mineral deposition and altered matrix nanotopography with progressive mineralization (Fig. 1C). The extent of mineralization was confirmed through Alizarin Red staining for calcium deposits and FTIR spectroscopy. Alizarin Red staining (Fig. 1D) revealed visibly more intense red coloration and increased opacity in partially and fully mineralized constructs observed under brightfield imaging, consistent with progressive calcium accumulation. FTIR spectra (Fig. 1E) also demonstrated a progressive increase in the mineral-to-matrix ratio from non-mineralized (0.16 ± 0.02) to partially mineralized (0.80 ± 0.1, p = 0.0116 vs. non-mineralized) and fully mineralized hydrogels (2.06 ± 0.1, p <0.0001 vs. both other groups) (Fig. 1F). These structural changes corresponded with a significant increase in stiffness as measured with rheological mechanical testing (Fig. 1G). Fully mineralized hydrogels exhibited the highest bulk elastic modulus (430 ± 9 Pa), significantly greater than both partially mineralized (246 ± 11 Pa, p < 0.0001) and non-mineralized hydrogels (136 ± 19 Pa, p < 0.0001), with partially mineralized scaffolds also significantly stiffer than non-mineralized samples (p = 0.0027), suggesting that reinforcement with mineral increased the mechanical properties of the substrate. Porosity, measured as area fraction from binarized SEM images, was significantly reduced in mineralized scaffolds (Fig. S1B). Partially mineralized (19.0 ± 1.2 %, p < 0.0001) and fully mineralized hydrogels (19.7 ± 2.5 %, p = 0.0001) exhibited significantly lower porosity compared to non-mineralized samples (33.4 ± 1.8 %), likely due to mineral deposition filling void spaces within the hydrogel network. Collectively, these findings confirm our ability to controllably and reproducibly vary mineralization within engineered microenvironments.

Fig. 1.

Fabrication and characterization of high-density collagen microgels with tunable mineralization. (A) Schematic overview of the fabrication pipeline. Type I collagen was cast onto a well plate and allowed for self-assembly into hydrogels, followed by densification via centrifugation. Hydrogels were exposed to mineralization media for 1 or 3 days to yield partially and fully mineralized constructs, respectively. Microgels were produced by micro-sectioning into 500 × 500 × 200 μm³ constructs. (B) Brightfield images show the size and morphology of individual microgels (scale bar = 200 μm). (C) Representative SEM images reveal altered nanotopography, increasing mineral deposition, and matrix compaction with progressive mineralization (scale bar = 5 μm; inset HFW = 829 nm). (D) Alizarin Red staining reveals more intensely red deposits and increased overall opacity in partially and fully mineralized constructs, consistent with greater calcium content (scale bar = 150 μm). (E) FTIR spectra demonstrate increased phosphate $\left({\mathrm{P}\mathrm{O}}_4^{3-}\right)$ signal intensity in partially and fully mineralized hydrogels compared to non-mineralized controls, normalized by the amount of collagen present at the Amide I peak. (F) Quantification of the mineral-to-matrix ratio from FTIR data shows a significant increase in phosphate content with each level of mineralization (n = 3 per group). (G) Progressive mineralization results in a corresponding increase in elastic modulus (n = 3–5 per group). Data presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 by one-way ANOVA with Tukey’s post hoc test. Panel A was created in BioRender. SEM: scanning electron microscopy; HFW: horizontal field width; FTIR: Fourier transform infrared spectroscopy; SD: standard deviation; ANOVA: analysis of variance.

3.2. Mechanosensitive signaling in hMSCs as a function of mineralization

Our in vitro model was intentionally designed to mimic how host cells interact with an acellular scaffold following implantation, providing a controlled platform to isolate the effects of matrix composition on early cell-matrix signaling. To investigate how cells respond to the microenvironmental composition and mechanics at early time points, we evaluated mechanosensitive signaling pathways in hMSCs seeded on matrices with varying mineral content, revealing distinct changes within as early as 2 h of culture (Fig. 2). Focal adhesions were assessed using immunofluorescence staining for phosphorylated paxillin, a key marker of adhesion maturation, revealing a significant reduction in focal adhesion formation with increasing matrix mineralization. Cells seeded on non-mineralized scaffolds exhibited a higher density of focal adhesion clusters broadly distributed across the cell body (657 ± 358 clusters/cell, n = 11 cells) compared to partially mineralized (276 ± 167 clusters/cell, n = 16 cells) and fully mineralized conditions (223 ± 130 clusters/cell, n = 9 cells) (Fig. 2A–C, N = 3 independent samples per group). These findings indicate a reduction in focal adhesion formation with increasing mineral content. In parallel, Piezo1 expression was significantly reduced in hMSCs cultured on mineralized scaffolds, suggesting that early response is from adhesion-mediated mechanotransductive signaling (Fig. S2). At the same time, fully mineralized scaffolds promoted a significant increase in expression of YAP1 (1.8 ± 0.3 nuclear/cytoplasmic ratio, n = 47 cells), including nuclear specific sites, in response to elevated matrix stiffness and mineral content, compared to both non-mineralized (1.5 ± 0.3, n = 40 cells) and partially mineralized (1.5 ± 0.3, n = 56 cells) (Fig. 2D–F, N = 3 independent samples per group). Scaffold mineralization was associated with reduced focal adhesions and increased nuclear YAP1 expression in hMSCs within 2 h of culture, indicating modulation of early mechanosensitive signaling.

Fig. 2.

Matrix mineralization modulates mechanosensitive signaling in hMSCs. (A) Representative immunofluorescence images showing Phospho-Paxillin (green), actin (red), and DAPI (blue) in hMSCs cultured for 2 h on non-mineralized, partially mineralized, and fully mineralized scaffolds (scale bar = 20 μm). White dashes outline the cell membrane. (B) Schematic of focal adhesion complexes illustrating the roles of paxillin in the integrin adhesome signaling. (C) Quantification of Phospho-Paxillin clusters per cell shows significantly fewer clusters in partially and fully mineralized scaffolds compared to non-mineralized controls (n = 9–16 cells from 3 samples per group). (D) Immunofluorescence images of YAP1 (green), actin (red), and DAPI (blue) in hMSCs at 2 h reveal increased nuclear YAP1 localization in mineralized conditions (scale bar = 25 μm). (E) Schematic of the mechanotransduction pathway showing YAP nuclear translocation in response to matrix stiffness and integrin signaling leading to osteogenic differentiation. (F) Quantification of YAP1 nuclear fraction shows significantly higher nuclear localization in cells on fully mineralized scaffolds (n = 40–56 cells from 3 samples per group). Data presented as mean ± SD. **p < 0.01, ***p < 0.001, ****p < 0.0001 by one-way ANOVA with Tukey’s post hoc test. Panel B and E created in BioRender. hMSCs: human mesenchymal stem cells; Phospho-Paxillin: phosphorylated paxillin; YAP1: Yes-associated protein 1; DAPI: 4′,6-diamidino-2-phenylindole; ECM: extracellular matrix; SD: standard deviation; ANOVA: analysis of variance.

Fully mineralized matrices enhanced RUNX2 expression by 24 h (Fig. 3), suggesting that mineralization promotes early commitment of hMSCs to the osteogenic lineage. Quantitative image analysis (Fig. S3) revealed significantly higher RUNX2 intensity in fully mineralized scaffolds (125,000 ± 1600 AU), compared to partially mineralized (64,800 ± 600 AU) and non-mineralized scaffolds (58,200 ± 700 AU). Based on intensity distributions, the percentage of RUNX2⁺ cells was significantly higher in fully mineralized samples (40.7 ± 3.9 %, p < 0.0001) compared to non-mineralized (0.24 ± 0.2 %) and partially mineralized (0 %) groups. This trend was supported by qPCR analysis, where fully mineralized scaffolds showed a slight increase in RUNX2 transcript levels at 24 h and a significant upregulation by day 3 (0.068 ± 0.009), compared to partially mineralized (0.052 ± 0.002, p = 0.161) and non-mineralized samples (0.041 ± 0.002, p = 0.027). Additionally, ALPL and BGLAP expression both declined over time in fully mineralized scaffolds (Fig. S4). Together, these findings highlight the role of matrix mineralization in priming hMSCs for osteogenesis through mechanical and biochemical cues.

Fig. 3.

Matrix mineralization rapidly enhances nuclear RUNX2 expression indicating early osteogenic commitment of hMSCs. (A) Representative immunofluorescence images showing RUNX2 (green), actin (red), and DAPI (blue) in hMSCs cultured for 24 h on non-mineralized, partially mineralized, and fully mineralized high-density collagen hydrogels. Increased nuclear RUNX2 localization is observed in fully mineralized matrices (scale bar = 35 μm). (B) Quantification of percentage RUNX2⁺ nuclei at 24 h shows significantly more RUNX2⁺ cells in fully mineralized scaffolds compared to partially and non-mineralized conditions (n = 3). (C) RUNX2 gene expression at 24 h shows a slight increase in fully mineralized matrices compared to other groups (n = 3). (D) By day 3, RUNX2 transcript levels were significantly upregulated in fully mineralized scaffolds compared to non-mineralized controls (n = 3). Scale bars = 35 μm. Data presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 by one-way ANOVA with Tukey’s post hoc test. RUNX2: Runt-related transcription factor 2; DAPI: 4′,6-diamidino-2-phenylindole; hMSCs: human mesenchymal stem cells; qPCR: quantitative polymerase chain reaction; SD: standard deviation; ANOVA: analysis of variance.

3.3. Bone regeneration in a critical-sized calvarial defect model in vivo

To evaluate whether the early mechanosensitive effects observed in vitro translate into functional tissue regeneration, we next assessed scaffold performance in a preclinical animal model. A critical-sized (8 mm) calvarial defect was surgically created in male and female Wistar rats and treated with one of five groups: empty defect (negative control), autologous bone (positive control mimicking decompressive craniotomy), or microgels with varying degrees of mineralization (none, partial, or fully mineralized). Microgels were implanted directly into the defect site (Fig. 4A) with a total dispensed volume of 0.38 ± 0.14 mL, composed of microgel material suspended in sterile PBS and adjusted per animal to fill the defect without overflow. Bone regeneration was monitored using high-resolution endpoint (Fig. 4) and in vivo longitudinal (Fig. 5 and Fig. S5) micro-CT over 12 weeks.

Fig. 4.

Fully mineralized scaffolds enhance bone regeneration in males but show limited benefit in females. (A) Schematic and intraoperative images depicting the surgical creation of an 8 mm critical-sized rat calvarial defect and implantation of microgels. (B) Representative high-resolution ex vivo micro-CT images at 12 weeks post-implantation for female and male rats across five treatment groups: empty defect, non-mineralized, partially mineralized, fully mineralized microgels, and autologous bone. Each set includes a top view and corresponding cross-sectional coronal view (scale bar = 8 mm). (C) Quantification of bone volume shows that fully mineralized scaffolds significantly enhanced regeneration in males compared to non-mineralized and empty defect groups, while females exhibited no added benefit from fully mineralized scaffolds. In females, both non-mineralized and partially mineralized scaffolds outperformed empty defects. Autologous bone grafts supported robust bone formation in both sexes. Two-way ANOVA revealed significant effects of sex (p = 0.0006), treatment groups (p < 0.0001), and their interaction (p = 0.0158), indicating a sex-dependent response to scaffold mineralization. Data are mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 by two-way ANOVA with Fisher’s LSD test (comparisons treated independently). Panel A was created in BioRender. CT: computed tomography; ANOVA: analysis of variance; SD: standard deviation; LSD: least significant difference.

Fig. 5.

Progressive bone formation occurs across all groups, though fully mineralized microgels show limited regenerative advantage in females. (A) Representative top view micro-CT images of calvarial defects at weeks 4, 8, and 12 post-implantation across five treatment groups: empty defect, non-mineralized, partially mineralized, fully mineralized microgels, and autologous bone (scale bar = 8 mm). (B) Between-group comparisons at each time point reveal significant differences in bone volume, with autologous bone demonstrating the highest volume throughout. Data shown as mean ± SEM, where the solid line represents the mean and the shaded region outlined by dashed lines indicates the SEM. (C) Within-group bone volume measurements across 4, 8, and 12 weeks shows progressive increases in all groups over time. (D) BMD remained consistent over time and across treatment groups, indicating comparable mineral quality of regenerated tissue. (E) Bone bridging area was calculated as the percentage of defect width spanned by mineralized tissue at the central region of the defect. Data represent female animals only and are shown as mean ± SD, unless otherwise stated. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 by mixed-effects model with Tukey’s multiple comparisons test. BMD: bone mineral density; SEM: standard error mean; SD: standard deviation.

At 12 weeks post-implantation, high-resolution ex vivo micro-CT revealed that fully mineralized scaffolds significantly increased bone volume in male rats (18.9 ± 2.6 mm³) compared to non-mineralized (12.9 ± 1.2 mm³, p = 0.0371) and empty defect groups (11.6 ± 2.1 mm³, p = 0.0242). In contrast, female rats showed no additional benefit from full mineralization (7.3 ± 2.0 mm³) compared to non-mineralized (11.9 ± 1.9 mm³, ns) or partially mineralized scaffolds (11.9 ± 1.3 mm³, ns), although both non-mineralized and partially mineralized groups were superior to empty defects (p = 0.0024 and p = 0.0026, respectively). Autologous bone grafts resulted in the highest measured bone volume in both sexes (females: 30.0 ± 2.2 mm³; males: 28.1 ± 0.8 mm³), significantly exceeding all other groups (p < 0.0001). Two-way ANOVA demonstrated significant main effects of sex (p = 0.0006), treatment groups (p < 0.0001), and their interaction (p = 0.0158), which accounted for 7.2 % of the total variance. These results show that scaffold mineralization enhances bone regeneration in a sex-dependent manner, with males exhibiting greater bone volume in response to higher degrees of scaffold mineralization at later stages of healing. While all animals were age-matched, males exhibited higher body weights than females, as expected for this age group (Fig. S6). Furthermore, quantitative analysis revealed no significant differences in bone mineral density (BMD) across treatment groups or between sexes (Fig. S7). These results indicate that, regardless of scaffold mineralization level or sex, scaffold properties influenced the quantity of bone formed but not the mineral density which remained consistent across conditions by 12 weeks.

Longitudinal micro-CT analysis of female rats revealed differential bone regeneration across scaffold mineralization groups over the 12-week healing period (Fig. 5). Three-dimensional (3D) micro-CT reconstructions showed that new bone predominantly formed from the lateral edges of the defect, progressing inward in a pattern that appeared to follow native cranial architecture. All groups showed gradual increases in bone volume over time (p < 0.0001), with significant variation based on treatment group (p < 0.0001) and a significant interaction between time and treatment group (p = 0.0037). The autologous bone group exhibited the most substantial regeneration, increasing from 28.8 ± 2.6 mm³ at week 4 to 32.5 ± 2.5 mm³ at week 12 (p = 0.0012). The non-mineralized group demonstrated a significant increase from 7.8 ± 2.0 mm³ to 13.7 ± 2.8 mm³ (p < 0.0001), while the empty defect group remained low across all timepoints (4.0 ± 0.7 mm³ at week 4 to 6.1 ± 1.6 mm³ at week 12, p > 0.05). The partially mineralized group also showed a significant increase over time, rising from 10.6 ± 0.3 mm³ to 13.3 ± 1.7 mm³ (p = 0.0333). In contrast, the fully mineralized group showed more modest changes from 8.8 ± 1.8 mm³ to 10.3 ± 2.3 mm³ (p = 0.346). These results confirm distinct regenerative trajectories over time within each group. Male animals were analyzed separately and are shown in Supplementary Figure S8.

Hematological analysis revealed sex- and treatment-dependent differences that may reflect distinct inflammatory and regenerative states during bone repair (Supplementary Tables S1–S2 and Fig. S9). Sex-specific effects were observed across multiple hematological parameters. Male rats exhibited significantly higher white blood cell (WBC) counts (p = 0.0122; Fig. S8A), lymphocyte counts (p = 0.0078; Fig. S8B), monocyte counts (p = 0.0019; Fig. S8C), and red blood cell (RBC) counts (p = 0.0458; Fig. S8D) compared to females. In contrast, females had significantly higher mean corpuscular hemoglobin (MCH) values (p = 0.0356; Fig. S8E). Significant differences across treatment groups were also observed. Platelet counts varied by treatment group (p = 0.0046; Fig. S8F), with elevated levels in the autologous bone group. Lymphocyte counts (p = 0.0484; Fig. S8G), reticulocyte percentages (p = 0.0367; Fig. S8H), absolute reticulocyte counts (p = 0.0353; Fig. S8I), and mean corpuscular hemoglobin concentration (MCHC; p = 0.0058; Fig. S8J) also differed significantly across treatment groups. Lymphocyte and absolute reticulocyte counts tended to be higher in the partially mineralized group, while MCHC was highest in the empty defect group. Together, these findings indicate distinct systemic responses based on sex and scaffold mineralization.

3.4. Histological analysis reveals distinct patterns of bone regeneration and scaffold resorption

Histological analyses were performed to evaluate scaffold degradation, tissue morphology, and the spatial distribution of new bone at 12 weeks. The main figure comparisons focused on female animals, given the more challenging healing characteristics and more modest regenerative response observed in vivo, which allow for higher-resolution insight into scaffold performance. Representative female images are shown in Fig. 6, with the full female cohort in Fig. S10–S14. To provide a complete assessment across both biological contexts, parallel histological analyses were also conducted in male animals, which are presented in Supplementary Figures S15–S16.

Fig. 6.

Microgel implantation promotes bone formation, with mineral content influencing scaffold resorption and defect bridging. (A) Representative H&E-stained coronal sections from female rats across five treatment groups: empty defect, non-mineralized, partially mineralized, fully mineralized microgels, and autologous bone. Each group includes a low-magnification overview image (top row) and two corresponding magnified regions: one at the defect edge (middle row) and one within the defect (bottom row). Black boxes in the overview panel indicate the regions shown in the magnified regions. Black arrows mark the original edges of the 8 mm calvarial defect. Overview HFW = 13.25 mm; magnified region scale bar = 200 μm. (B) Representative sections from the non-mineralized, partially mineralized, and fully mineralized groups, taken from areas within the defect where residual microgel material was observed, and stained with H&E, Masson’s Trichrome, and Picrosirius Red. Scale bar = 200 μm. H&E: hematoxylin and eosin; HFW: horizontal field width; nb: new bone; g: autologous bone graft; *: residual microgel material.

Hematoxylin and eosin (H&E) staining was performed on coronal sections to assess tissue morphology at 12 weeks post-calvarial defect. Representative images are shown in Fig. 6, with the complete female cohort presented in Fig. S10. In the empty defect group, minimal new bone formation was observed, with fibrous tissue predominating in the defect space. Defects treated with microgels exhibited increased bone formation originating from the lateral edges and extending inward, though defect bridging remained variable across mineralization levels. In contrast, the autologous bone group demonstrated substantial new bone deposition around the implanted graft, with evidence of vertical thickening and nearly complete closure of the defect. Quantification of average tissue thickness revealed that autologous bone grafts produced significantly thicker overall tissue compared to all other groups (0.5 ± 0.1 mm; p < 0.05; Fig. S11). Across all groups, new bone appeared to emerge from both the periosteal and endosteal surfaces along the defect edges, contributing to a growing osteogenic front extending inward. This pattern is consistent with previous reports describing periosteal cell contributions to calvarial bone regeneration [42]. A limitation of this surgical model is that implanted microgels may not remain confined to the defect site. In some cases, residual microgels were observed embedded in fascia or scar tissue outside the calvarial defect (Fig. S12).

Masson’s Trichrome staining (Fig. S13) provided clear contrast between newly formed tissue and residual microgels, which appeared lighter blue and acellular. Organized, mature bone showed linear structure and flattened nuclei, while newly formed bone appeared more irregular with rounder nuclei. Additionally, picrosirius red staining (Fig. S14) allowed for clearer visualization of bone matrix orientation. The uninjured calvarial bone appeared highly linear and organized, while newly formed bone was more disorganized and irregular, consistent with what is typically observed in immature woven bone.

In groups with implanted microgels, remnants of hydrogel material were distinguishable across all histological stains at 12 weeks post-implantation. These materials were typically acellular at their core, yet surrounded by host cells, and often displayed appositional bone formation at their periphery (Fig. 6B). This suggests that the gels served an osteoconductive role in guiding bone formation along the defect margins, while the presence of isolated bone islands within the defect suggests an osteoinductive effect, consistent with our in vitro findings. Importantly, differences in scaffold persistence and spatial bone patterning were assessed qualitatively at this single endpoint. Notably, resorption patterns appeared to vary depending on the mineral content. Histological images showed endpoint qualitative differences in the spatial patterning of new bone, with mineralized groups exhibiting more irregular and dispersed bone formation. However, bone volume and defect bridging metrics did not differ significantly across scaffold types in females, and non-mineralized scaffolds did not outperform other groups in either sex.

Scaffolds appeared to be resorbed from the outside in, as there were no cells within the core of the residual gel material. This limited cell infiltration was more prevalent in mineralized gels, which showed more hydrogel remnants remaining at 12 weeks, suggesting slower degradation. This is likely attributed to their high density, increased stiffness, and elevated mineral content, factors that can collectively impede degradation and cellular infiltration. Furthermore, autologous bone grafts exhibited partial resorption and were surrounded by substantial new bone growth originating from the defect margins, enhancing osseointegration. This newly formed bone contributed to defect bridging and increased calvarial thickening, consistent with longitudinal in vivo micro-CT data showing progressive increases in bone volume over the 12-week period (Fig. 5).

To further assess cellular processes associated with vascularization and bone remodeling within the defect, immunofluorescence staining for alpha-smooth muscle actin (aSMA) and cathepsin K (Ctsk) was performed on coronal sections from the full female cohort at the 12-week endpoint (Fig. S17). No overt qualitative differences in staining distribution were observed between groups. Across all treatment groups, aSMA⁺ structures were predominantly localized along the periosteal and endosteal surfaces of the calvarial bone, consistent with vascular invasion originating from native bone surfaces [43]. No qualitative trends in aSMA positive staining intensity or spatial distribution were apparent between scaffold conditions. Ctsk⁺ cells were primarily detected along the edges of the defect and at the advancing bone front, as well as lining bone surfaces and within newly formed bone cavities. Our qualitative observations suggest that the non-mineralized group appeared to exhibit more abundant Ctsk⁺ staining relative to other scaffold groups, which is consistent with the greater extent of bone formation observed in this group. Because the overall amount of newly formed bone differed between treatment groups, this apparent trend may in part reflect increased opportunity to observe osteoclast-associated staining in regions of greater regeneration rather than intrinsic differences in osteoclastic activity. This localization pattern was otherwise observed across groups and is consistent with active bone remodeling during defect repair. Given the qualitative nature of this analysis, these data were used to assess spatial patterns of remodeling rather than to infer quantitative differences between scaffold conditions. Given the qualitative nature of this analysis and limited group sizes, these data were not used to infer quantitative differences between scaffold conditions. Together, these staining patterns indicate ongoing vascular-associated activity at native bone interfaces and osteoclastic remodeling at sites of active bone formation and integration at 12 weeks post-implantation.

4. Discussion

This study introduces a tunable platform for engineering mineralized microenvironments that modulate early mechanobiological signaling and influence long-term bone regeneration in vivo. By replicating nanoscale features of native bone extracellular matrix within microgels, we demonstrate that mineralization enhances hMSC mechanosensing and osteogenic priming in vitro and promotes scaffold-guided regeneration in a sex-dependent manner.

4.1. Matrix mineralization modulates early mechanosensing and osteogenic priming in hMSCs

Our in vitro model was designed to simulate the initial host cell interaction with an implanted acellular scaffold. This reflects the clinical use of acellular bone graft substitutes, where regeneration relies on endogenous cells migrating into the scaffold, sensing its biochemical and mechanical properties, and initiating repair. We established a gradient of mineral content to evaluate its impact on early cell-matrix interactions. Within two hours of culture, our findings demonstrate that the mineral content can differentially regulate key mechanotransductive pathways, including integrin-mediated adhesion, Piezo1 ion channel activity, and YAP1 nuclear translocation. Overall, these results demonstrate that increasing matrix mineralization modulates early mechanoregulatory signaling in hMSCs, likely by altering integrin-matrix engagement while maintaining osteoinductive potential. By 24 h, hMSCs cultured on fully mineralized scaffolds showed increased RUNX2 expression at both the transcript and protein level, indicating early osteogenic commitment that is typically delayed in traditional 2D in vitro cultures with osteoinductive media. Rheological characterization was performed on bulk hydrogels, from which the microgels were derived, providing an indication of their mechanical properties. We acknowledge that microgels may experience local variations in mechanics, which represent an important area for future study. Our system serves as a useful platform to investigate early cell-matrix interactions and identify the microenvironmental features that initiate osteogenic signaling, serving as a basis for examining how these cues may translate to functional regeneration in vivo.

4.2. Mineralization drives sex-dependent differences in bone regeneration

Strikingly, fully mineralized scaffolds promoted bone regeneration in male rats, consistent with in vitro findings, whereas female rats showed no comparable benefit. This sex-dependent divergence provides a powerful lens to understand how early-stage cues differ between biological contexts. All scaffold-treated groups outperformed untreated defects, yet only males benefited from full mineralization, as confirmed by endpoint micro-CT. These findings echo known sex differences in bone biology, where males exhibit greater bone mass, higher mineral density, and more robust regenerative responses [44–46]. These differences are amplified by sex-specific growth trajectories, stem cell behavior, and systemic physiology [47]. To minimize confounding effects from developmental stage, animals were age-matched rather than weight-matched, as age directly influences bone metabolism [48]. As a result, males were significantly larger than females and exhibited hematological profiles indicative of heightened systemic activity, both of which may have contributed to their more active regenerative response. Notably, hematological analyses revealed higher circulating white and red blood cell counts in males compared to females, suggesting a more activated systemic immune and oxygen-transport capacity during the healing period. While this study does not establish a direct causal relationship between these systemic indicators and local bone regeneration, elevated white blood cell populations reflect the early inflammatory response to injury in which immune cells support downstream repair processes [49].

Interestingly, the female in vivo response did not fully align with the in vitro trends of increase in bone formation with higher scaffold mineralization, despite exposure to the same mineralized cues, highlighting that scaffold performance is not solely governed by intrinsic material properties. This mismatch between in vitro prediction and in vivo outcome in females suggests that possible systemic or hormonal factors may contribute to the osteogenic priming by mineralized scaffold cues. In addition, intrinsic sex-dependent differences in stem cell differentiation capacity may also contribute, as prior work has shown that sex influences osteogenic capacity under equivalent induction conditions [50]. Given the endpoint timing, it remains unclear whether this represents a true ceiling effect in females or a delayed but ongoing trajectory of healing. Importantly, the estrous cycle in females was not controlled, so variability and the observed limited response may partly reflect hormonal cycling rather than direct scaffold effects. We note this as a limitation and propose that future studies would benefit from controlling for estrous stage as a contributing factor to bone regeneration outcomes, which forms the basis for future work in our lab.

Plausible mechanisms underlying the sex effect include hormonal, vascular, and mechanobiological pathways that intersect with scaffold-driven cues. Androgen receptor signaling promotes osteoblast proliferation and Wnt/β-catenin activation, whereas estrogen primarily suppresses osteoclast activity through OPG-mediated inhibition of RANKL, indicating that sex hormones differentially tune cellular responses to scaffold stiffness and mineral content [51]. Angiogenesis and vascular remodeling show sexual dimorphism, with androgens generally enhancing neovascularization and estrogen attenuating RANKL-driven vascular calcification, suggesting that local Ca²⁺/phosphate ion release from mineralized microgels could differentially interface with these pathways across sexes [52,53].

Together, these findings indicate that scaffold mineralization influences bone regeneration in a sex-dependent biological context, rather than functioning as an isolated regenerative cue. Importantly, this study did not directly assess local immune cell infiltration or inflammatory signaling within the defect site, limiting our ability to link systemic hematological differences to site-specific regenerative processes. Future studies incorporating immunohistochemical or spatial analyses of inflammatory and osteogenic cell populations will be critical for resolving how systemic immune state interfaces with scaffold-driven bone regeneration. Future studies could further explore whether modifying scaffold architecture/porosity or delivery methods within high-density, mineralized microgel systems could enhance outcomes in female subjects. Additionally, follow-up studies using ovariectomized females or hormone-level profiling may clarify the mechanisms driving sex-specific outcomes and inform the development of bone graft strategies that account for biological sex in personalized regenerative therapies.

4.3. Impact of mineralization on scaffold integration, degradation, and bone formation patterns

For higher-resolution insight into scaffold performance in female rats, we evaluated scaffold degradation, the extent of defect bridging, and the spatial distribution of newly formed bone through histological analysis at the terminal endpoint. Mineralized microgels exhibited irregular, spatially confined bone formation within the defect site, likely reflecting differences in matrix stiffness and degradation. Remnants of mineralized scaffolds remained at 12 weeks, and were easily distinguishable in histological analyses, often acellular and surrounded by new bone, suggesting osteoinductive behavior but impaired cellular infiltration likely due to the scaffold’s increased density and stiffness. Because tissue harvest was performed at a single post-implantation time point, quantitative degradation kinetics were not assessed in this study. Accordingly, scaffold persistence and cellular infiltration are interpreted as qualitative endpoint comparisons. Consistent with these spatial patterns, immunofluorescence analysis revealed aSMA⁺ structures localized primarily along periosteal and endosteal surfaces and Ctsk⁺ cells concentrated at defect edges and advancing bone fronts, supporting active vascular-associated remodeling and bone turnover at native bone interfaces.

Within this context, our data highlight a key tradeoff: while mineralization enhances early osteoinductive signaling, it also creates biophysical barriers that can hinder uniform tissue integration. One potential strategy to overcome this barrier is to integrate angiogenic features that promote vascular invasion, enabling deeper cellular access and improved nutrient exchange to potentially rescue the delayed integration seen in mineralized constructs. However, functionally integrating vascularization within high-density, mineralized microgel systems presents a significant engineering challenge, as the mechanical properties that promote osteogenesis may not optimally support angiogenic sprouting, making it difficult to coordinate both processes effectively in vitro within the same scaffold. Furthermore, bone regeneration predominantly initiated at the defect margins and progressed inward, consistent with activity at the osteogenic front. This spatial pattern likely reflects contributions from periosteal progenitor cells and other local sources along the bone interface. Notably, bone formation occurred on both periosteal and endosteal surfaces, with spatial alignment reflecting native cranial architecture. These findings reinforce the osteoinductive role of microgels and suggest their ability to guide endogenous repair, while residual scaffold material observed at later stages of healing highlights the need to balance osteoinductive signaling with scaffold integration and resorption in future designs.

4.4. Leveraging tunable mineralization for clinically relevant bone regeneration strategies

When benchmarked against current clinical standards, our mineralized microgels did not surpass the regenerative capacity of autologous bone, which remains the gold standard for calvarial repair. However, they achieved substantial regeneration compared to empty defects while avoiding the need for donor tissue or complex surgical harvesting and demonstrated consistent osteoinductive potential without the use of any exogenous cells or osteoinductive supplements. The ability of this material alone to elicit osteogenesis represents a major advancement, as most current strategies rely heavily on biochemical additives to achieve comparable effects. Notably, these microgels are scalable, syringe-delivered, and consistently induced rapid osteogenic differentiation in vitro and bone formation in vivo. Their tunable mineral content enables precise control over stem cell behavior, offering both mechanistic insight and translational utility. Together, these features position mineralized microgels as a clinically viable, off-the-shelf alternative for bone regeneration where autografting is limited or contraindicated. In comparison to existing biomaterial-based strategies, our platform uniquely allows for modulating stem cell behavior through precise control of scaffold mineral content, providing a level of mechanistic insight not widely available in other systems and amenable to studying material-driven osteogenic mechanisms. Importantly, by matching ECM composition and density, and subjecting these scaffolds to controllable biomimetic intra- and extra-fibrillar collagen mineralization, these scaffolds offer a level of biomimicry that is substantially higher than more conventional mineral particle-laden hydrogels, which have been extensively studied in the field. These advances likely explain the regenerative qualities of our scaffold material despite the absence of exogenous growth factors or implanted cells.

Overall, this study bridges a critical gap between mechanistic insight and translational application in bone tissue engineering. By precisely mimicking key features of the native bone matrix, this work demonstrates that stem cell osteogenic differentiation can occur rapidly, even in the absence of biochemical cues, initiated by mineralization and supported by early mechanotransductive signaling. Our dual in vitro and in vivo approach provides a unique opportunity to correlate mineralization-dependent mechanosensing with long-term regenerative outcomes, underscoring the predictive value of early cellular responses in biomaterial design. These findings highlight how tunable mineralization in engineered microgels can influence early mechanobiological signaling and downstream regenerative outcomes. By integrating material cues with biological context, this platform offers a tool for both probing fundamental cell-matrix interactions and guiding the design of next-generation bone graft substitutes. Future work should explore strategies to enhance integration, such as coupling osteogenic and pro-vascular cells and adapting scaffold design features to account for sex-specific healing dynamics. By bridging early mechanobiological cues with tissue-scale regeneration, this study establishes a new framework for designing the next generation of personalized, material-driven bone graft substitutes.

5. Conclusions

This study presents a bioinspired platform of high-density collagen microgels with tunable intrafibrillar mineralization that promotes rapid osteogenic priming in vitro and enhances bone regeneration in vivo. By mimicking key nanoscale features of native bone matrix, these scaffolds influence early mechanosensitive signaling, driving stem cell commitment to the osteogenic lineage without the need for exogenous factors. In a rat calvarial defect model, implanted microgels elicited both osteoconductive and osteoinductive effects and revealed sex-dependent regenerative outcomes. These cell-free scaffolds achieved substantial bone formation, highlighting their potential as minimally invasive alternatives for bone repair. Future efforts to incorporate features to optimize cell infiltration may improve scaffold performance and broaden clinical applicability.

Supplementary Material

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.actbio.2026.02.003.

Statement of significance:

This study introduces a strategy to fine-tune the properties of implantable materials for bone repair using microscale scaffolds with controlled mineral content. By adjusting composition at the nanoscale, our work identifies how early cellular responses can be directed to influence long-term healing. Importantly, the findings reveal that regenerative outcomes vary by sex, emphasizing the need to consider biological differences in biomaterial design. This work offers new insight into how tailored physical environments can guide tissue repair and highlights the potential for precision approaches in bone graft development.

Data availability

The micro-CT and histology datasets produced in this work have been deposited in FaceBase, a publicly accessible, National Institute of Dental and Craniofacial Research (NIDCR)-supported data repository in support of advancing craniofacial research [54]. The dataset, entitled “Bone Regeneration in Rat Calvarial Defects with Tunable Mineralized Collagen Microgels: Micro-CT (Longitudinal & Ex Vivo) and Histology” is listed under record ID of 8X-J1TW and is available at https://doi.org/10.25550/8X-J1TW [55].