A multiscale pathoanatomical atlas guides the design of functionally graded, anatomically intelligent implants for chronic osteomyelitis

Abstract

Chronic osteomyelitis remains a formidable clinical challenge because conventional biomaterials, which are designed for spatially uniform defects, cannot match the hierarchical, compartmentalized progression of deep bone infection. Using a clinically relevant rabbit femoral osteomyelitis model, we construct a multiscale pathoanatomical atlas of S. aureus infection, revealing a coherent disease cascade: medullary colonization, invasion of the immunoprivileged osteocyte lacuno–canalicular system (OLCS), region-specific bone destruction, and ultimate formation of a biomechanically incompetent sequestrum. Guided explicitly by this anatomical blueprint, we design and computationally optimize a dual-component bioactive material system that mirrors and counteracts the infection hierarchy. This dual-component implant integrates a 3D-printed, load-bearing macro-scaffold for mechanical stabilization and bone regeneration with infiltrative microspheres capable of penetrating trabecular microdomains and releasing antimicrobials within the marrow and cortical compartments, directly targeting bacterial reservoirs. In vivo validation demonstrates that this system enables simultaneous eradication of deep-seated infection, resolution of chronic inflammation, and restoration of structurally competent bone. Collectively, this work establishes imaging-resolved anatomical mapping as a generative framework for bioactive material design and provides a strategy that may be adapted for developing spatially adaptive therapeutic systems against complex tissue infections.

Graphical abstract

This work presents a pathology-driven design framework for treating implant-associated osteomyelitis. A multiscale disease atlas (a) reveals asymmetric regional vulnerability and bacterial sanctuary within the OLCS. Quantified pathological findings (b) inform a functionally graded material design (c): a 3D-printed load-bearing scaffold for diaphyseal reconstruction coupled with infiltrative antimicrobial microspheres. In vivo validation (d) demonstrates simultaneous infection eradication, inflammation control, and bone regeneration, establishing a benchmark for translation-oriented therapeutic systems.

Highlights

• •Multiscale pathoanatomical atlas reveals bacterial “spatial sanctuary” strategy in osteomyelitis.
• •Asymmetric vulnerability of femoral head vs. condyle identified, revealing distinct pathology.
• •Bacteria can colonize the OLCS system, forming nanoscale shelters.
• •Design a functional gradient scaffold microsphere system to reverse simulate the progression path of infection grading.
• •In vivo validation shows that the system can synchronously eradicate deep infections, control inflammation, and regenerate functional bone tissue.

1. Introduction

Chronic osteomyelitis remains a formidable clinical challenge due to persistent infection, inflammatory bone destruction, and impaired regenerative capacity [1]. Conventional biomaterials [2]—antibiotic-loaded cements [3] and monolithic scaffolds [4]—are designed for focal defects under a "one-size-fits-all" assumption [5,6]. Their failure stems from a fundamental mismatch between spatially static materials cannot address the anatomically programmed hierarchy of deep long-bone infection [7]. Staphylococcus aureus invades systematically—colonizing the marrow, breaching the endosteum, infiltrating cortical canals, and ultimately sequestering within the osteocyte lacuno-canalicular system (OLCS). This creates a continuum of distinct microenvironments, each with unique diffusion barriers and immune privilege [8]. What is urgently needed is a new generation of bioactive materials endowed with "anatomical intelligence"—the capacity to respond to and precisely counteract the spatial sophistication of infection.

Current treatments for osteomyelitis are technologically stagnant. Biomaterials remain largely limited to single antimicrobial functions and lack the microenvironment-responsive capabilities needed to adapt to the dynamic, multi-compartment nature of the infection [9,10]. Many preclinical models oversimplify osteomyelitis as a localized defect, failing to capture its longitudinal, multiscale progression and thus remaining unable to define the mechanical, antimicrobial, and regenerative requirements at each anatomical level [11,12]. An effective bioactive material should act as a "bio-instructive" platform, delivering region-specific support and biological cues in precise spatial and temporal sequence [13,14]. However, the development of such platforms is hindered by the absence of a high-resolution, whole-bone pathoanatomical atlas that delineates the spatiotemporal evolution of infection across distinct anatomical compartments [15]. Without this foundational reference, we cannot identify the coordinates of bacterial sanctuaries, characterize divergent microenvironments (e.g., femoral head versus condyle), or determine the barriers—from cortical bone to the OLCS—that therapeutic carriers must overcome. Bridging descriptive pathology and material engineering thus requires a transformative blueprint that translates biological barriers into design parameters [16].

To provide this essential design framework, we conducted a multimodal spatiotemporal analysis using a clinically relevant rabbit model of femoral osteomyelitis [17]. Through integrated serial imaging (radiography, Micro-CT), histology, and ultrastructural microscopy (SEM/TEM), we achieved three core objectives: (1) to quantify the temporal progression from acute inflammation to chronic sequestration; (2) to map the asymmetric spatial dissemination of infection from diaphysis to epiphysis, revealing region-specific vulnerabilities; and (3) to identify, at the nanoscale, the definitive bacterial sanctuaries within osteocyte lacunae. By synthesizing these findings, we transformed descriptive pathology into a functional design blueprint, specifying the "where," "when," and "how" for therapeutic intervention.

Guided by this blueprint, we designed and validated a dual-component bioactive material system engineered to counteract the hierarchical progression of osteomyelitis. The system combines a 3D-printed, load-bearing macro-scaffold for diaphyseal mechanical restoration [18] with infiltrative microcarriers that penetrate trabecular and lacunar microdomains for targeted antimicrobial delivery [19]. We fabricated this platform and validated its therapeutic efficacy in the osteomyelitis model, achieving effective intervention across all infected bone compartments. By rigorously linking multiscale pathological mechanisms to material design parameters, we move beyond validating isolated properties toward creating spatially adaptive, combinatory therapeutic systems. This pathology-driven framework offers a strategy for addressing complex tissue infections and positions dual-component bioactive material systems as a promising platform for regenerative medicine.

2. Results

2.1. Establishment of the osteomyelitis model

An osteomyelitis model was established in the rabbit femur by intraosseous inoculation of S. aureus with sodium morrhuate (Fig. 1a). All S. aureus-inoculated animals developed localized abscesses at the surgical site within one week, which progressively worsened to skin ulceration and sinus tract formation by week 4. Bacterial culture confirmed sustained infection (Fig. 1b). Systemically, serum CRP, white blood cell counts, and globulins progressively increased (Fig. 1c–e). At the cellular level, flow cytometry of local tissues demonstrated a significant temporal increase in apoptosis, from 25.82 ± 1.23% at week 1 to 41.64 ± 3.24% at week 4 (Fig. 1f), providing direct evidence of progressive infection-driven tissue damage. These multi-level assessments confirm the successful establishment of a progressive S. aureus osteomyelitis model that evolves from acute inflammation to chronic infection over four weeks, providing a validated temporal framework for the subsequent high-resolution spatiotemporal analysis.

Fig. 1.

Analysis of main indicators for rabbit femoral modeling osteomyelitis. a) Schematic diagram of establishment and infection of rabbit femoral osteomyelitis modeling. b) Gross photography and bacterial coating of local abscesses at different time points. c) C-reactive protein content (1w vs Con, p = 0.6862; 2w vs 1w, p < 0.0001; 4w vs 2w, p < 0.0001; n = 3 per group), d) Routine blood counts and e) Biochemical parameters at various time points in rabbits' femoral osteomyelitis (TP Ⅱ: 1w vs Con, p = 0.0065; 2w vs 1w, p = 0.0069; 4w vs 2w, p = 0.8544; ALB Ⅱ: 1w vs Con, p = 0.0002; 4w vs 1w, p = 0.0715; GLO: 1w vs Con, p = 0.7326; 2w vs Con, p < 0.0001; 4w vs 2w, p = 0.0008; n = 3 per group). f) Tissue cell apoptosis flow cytometry of local abscesses at different time points.

2.2. Spatiotemporal mapping of infection

Following confirmation of model establishment, we investigated the spatiotemporal progression of infection within the femur, employing a "center-to-periphery" narrative. We first focused on the primary nidus (modeling region), then traced its spread to adjacent and remote bone compartments.

2.2.1. Pathological diaphyseal nidus

The femoral diaphysis surrounding the inoculation site (ROI A) served as the primary infectious site, defining the core temporal trajectory of infection (Fig. 2a). Micro-CT analysis delineated a clear progression from incipient inflammation to chronic sequestrum formation (Fig. 2b and c). At week 1, only mild periosteal reaction was evident. By week 2, localized cortical osteolysis, detachment, and increased porosity appeared, accompanied by a significant increase in cortical cross-sectional area and bone volume (BV) (Fig. 2b–d). At this stage, bone mineral density (BMD) showed a transient rise, likely due to reactive woven bone formation. By week 4, the hallmark of chronic osteomyelitis emerged: a circumferential involucrum of low-density, structurally disordered new bone encased the original cortex. This resulted in a further dramatic increase in BV, while the overall BMD returned to a level statistically indistinguishable from control group, revealing a critical dissociation between bone mass and material quality. Norden scores quantitatively captured this escalating severity, rising from 0 to 3/3/3 (Fig. 2e).

Fig. 2.

Pathological evolution at the primary infectious nidus. a) Pattern diagram of inflammation expansion and abscess and dead bone formation process in the modeling area. b) Gross morphology and representative Micro-CT images (3D reconstruction, 2D cross-section, and bone density heatmap) of the diaphyseal region (ROI A) at indicated time points. c) Quantification of cortical cross-sectional area (1w vs Con, p = 0.6888; 2w vs 1w, p < 0.0001; 4w vs 2w, p < 0.0001; n = 3 per group). d) Bone volume (BV, 1w vs Con, p = 0.1125; 2w vs 1w, p < 0.0001; 4w vs 2w, p < 0.0001; n = 3 per group) and bone mineral density (BMD, 1w vs Con, p = 0.702; 2w vs 1w, p < 0.0001; 4w vs Con, p = 0.0678; n = 3 per group) of ROI A. e) Norden scoring based on Micro-CT (n = 3 per group). f) HE (f₁-f₄: Con, 1, 2, and 4w) and g) Masson (g₁-g₄: Con, 1, 2, and 4w) staining results of transverse section of femur at modeling site (Red arrow and red box: medullary cavity and its enlargement, Black arrow and black box: cortical and its enlargement).

Histology confirmed and refined this timeline. Hematoxylin-eosin (HE) staining revealed the replacement of normal hematopoietic marrow by fibrin exudate (week 1), progressing to well-defined intramedullary abscesses (week 2), and ultimately to near-complete marrow obliteration by edematous granulation tissue and large abscesses (week 4) (Fig. 2f). Concurrently, Masson's trichrome staining illustrated the parallel disruption of the collagenous matrix, from early periosteal disorganization to extensive fibrosis lining purulent cavities and severe cortical collagen fragmentation at the chronic stage (Fig. 2g).

2.2.2. Asymmetric spread to proximal and distal compartments

Infection spread asymmetrically from the diaphyseal nidus (Fig. 3a, and e). The femoral head (ROI C) deteriorated more rapidly and severely than the distal condyle (ROI B). By week 4, the condyle exhibited significant but incomplete structural collapse (Fig. 3b), while the femoral head showed near-total architectural disintegration with large cavitary lesions (Fig. 3f). Histopathology revealed region-specific signatures: inflammatory osteolysis in the condyle (thinned, fractured trabeculae with leukocytic infiltration, Fig. 3c) versus ischemic necrosis in the femoral head (abundant empty osteocyte lacunae and acellular marrow, Fig. 3g). Quantitative metrics confirmed accelerated rarefaction in the proximal region, with greater reductions in Tb. BMD and Tb. N and a larger rise in Tb. Sp (Fig. 3d, and h). The integrated data demonstrate that osteomyelitic progression follows an anatomically constrained gradient rather than a uniform path. The femoral head, with its vulnerable terminal blood supply, is predisposed to early ischemic necrosis and accelerated structural collapse. In contrast, the better-vascularized condyle undergoes a slower process of inflammatory osteolysis. This region-dependent pathological mechanism underscores the insufficiency of one-size-fits-all therapeutic approaches. More importantly, it provides critical anatomical coordinates for the design of functionally graded therapeutic systems. These coordinates dictate that an effective implant must deliver compartment-adapted therapies: the femoral head region necessitates strategies prioritizing angiogenic rescue and anti-necrosis, whereas the condylar region demands a focus on combating inflammatory osteolysis and promoting trabecular regeneration. Thus, the spatial heterogeneity of disease not only explains its progression but also directly informs the rational design of spatially informed, compartment-adapted treatment strategies.

Fig. 3.

Asymmetric spread and region-specific pathology in distal and proximal compartments. Schematic diagram showing the lesion process of the a) distal (femoral condyle, ROI B) and e) proximal (femoral head, ROI C) regions. Multi-modal imaging of the femoral condyle (ROI B): b) gross view, 3D Micro-CT, 2D cross-section with bone density heatmap, and corresponding c) high-magnification HE staining. d) Quantitative comparison of a key microarchitectural parameter between ROI B over time. Multi-modal imaging of the femoral head (ROI B): f) gross view, 3D Micro-CT, 2D cross-section with bone density heatmap, and corresponding g) high-magnification HE staining. h) Quantitative comparison of a key microarchitectural parameter between ROI C over time.

2.3. Bacterial colonization: fom adhesion to protected niches

To elucidate the mechanisms underlying the observed pathological progression, we employed multi-scale microscopy to investigate the spatial distribution and colonization behavior of S. aureus within the bone microenvironment. Integrating Giemsa staining, SEM, and TEM, our analysis reveals a strategic pattern of bacterial dissemination from the marrow cavity into protected cortical niches.

2.3.1. Tissue-level distribution

Giemsa staining provided a panoramic view of bacterial load and the concomitant inflammatory landscape (Fig. 4a). At week 1, bacterial colonization was limited to scattered coccoid clusters along the periosteal surface, accompanied by mild medullary inflammation. By week 2, dense bacterial aggregates had formed distinct intramedullary purulent foci, indicating the consolidation of infection into localized, protected cavities. The surrounding marrow showed degenerative changes and focal necrosis. At the chronic stage (week 4), the infection had organized into extensive, sinus-tract-like purulent channels within the marrow cavity, and bacterial clusters were intimately associated with severely fragmented cortical remnants.

Fig. 4.

Multiscale visualization of S. aureus spatial invasion and colonization in bone. a) Histopathological characterization of S. aureus load and inflammatory infiltration via Giemsa staining in a rabbit femoral osteomyelitis model (a₁-a₄: con, 1, 2, and 4w; Red arrow and red box: Medullary cavity and its enlargement, Black arrow and black box: cortical and its enlargement). b) Evaluate the distribution of S. aureus in the medullary cavity and cortical bone at different times of osteomyelitis infection through SEM (b₁-b₄: con, 1, 2, and 4w). c₁) Bacteria within the inflammatory milieu of the marrow cavity. c₂) Bacteria situated in a focally disrupted, vacuolated region of the endosteal lining (arrow). c₃) Bacterial profiles located within an osteocyte lacuna (yellow arrow) and adjacent canalicular space (white arrow), indicating colonization of the deepest bone compartments.

2.3.2. Surface colonization and biofilm formation

SEM resolved the micro-scale interaction between bacteria and bone matrix (Fig. 4b). Sparse adherent cocci at week 1 progressed to dense clusters within degraded trabeculae by week 2, with bacteria penetrating Haversian and Volkmann canals. By week 4, multilayered bacterial assemblies coated marrow surfaces—indicative of mature biofilm formation—and were accompanied by severe matrix deterioration. The integrated data from Giemsa staining and SEM map a coherent colonization strategy: S. aureus rapidly progresses from initial adhesion to the formation of organized purulent foci within the marrow and, crucially, actively invades the cortical canal network. This represents a shift from planktonic or loosely attached states to a surface-associated, biofilm-embedded lifestyle within the bone's porous compartments. However, this still leaves a critical question unanswered: does bacterial invasion extend beyond the vascular channels into the even more secluded, nanoscale domains of the bone matrix itself?

2.3.3. Ultrastructural migration into the OLCS

In the marrow cavity adjacent to bone, TEM revealed clusters of coccoid bacteria embedded within an electron-dense matrix of host-derived cellular debris, consistent with the inflammatory milieu observed at the tissue level (Fig. 4c₁). As the analysis moved toward the cortical bone interface, bacteria were found within focally disrupted and vacuolated regions of the endosteal layer (Fig. 4c₂), indicating localized breaching of this critical barrier between the marrow and mineralized matrix. Most significantly, TEM provided definitive ultrastructural evidence of bacterial presence within the OLCS. We observed bacterial profiles located inside osteocyte lacunae—the spaces formerly occupied by bone cells—as well as within the nanoscale canaliculi radiating from them (Fig. 4c₃). These structures represent the innermost, cellular-scale sanctuary of bone tissue, orders of magnitude smaller than vascular canals and far removed from circulatory and immune surveillance. Synthesis: The Hierarchical “Spatial Sanctuary” Strategy. The multi-scale microscopic analysis from Giemsa to SEM to TEM maps a coherent and escalating colonization trajectory. S. aureus establishes structured foci in the marrow (Giemsa), forms biofilms within cortical vascular channels (SEM), and ultimately migrates into the OLCS (TEM).

Our multi-scale analysis reveals that bacterial persistence is orchestrated through a hierarchical “spatial sanctuary” strategy: pathogens invade progressively smaller and more protected anatomical niches, culminating in colonization of the OLCS. This nanoscale sanctuary provides a direct mechanistic explanation for therapeutic recalcitrance, as bacteria achieve physical seclusion within the mineralized matrix, evading both immune surveillance and conventional antimicrobials. Consequently, this definitive invasion depth establishes the non-negotiable performance benchmark for any effective delivery system: it must possess the capacity to penetrate and release agents within these secluded, sub-micron domains. This fundamental requirement directly informs the design rationale for the infiltrative microcarriers proposed herein, guiding critical parameters such as size, surface functionality, and release kinetics to target the deepest bacterial reservoirs, thereby addressing a root cause of chronicity.

2.4. Macro-scale biomechanical failure

Having delineated bacterial sanctuary formation, we assessed the ultimate functional consequence: failure of the bone as a load-bearing organ. Radiographic and Micro-CT analysis revealed that localized infection triggered global, maladaptive remodeling (Fig. 5a). By week 4, the femur exhibited marked diaphyseal swelling encased by a thick, low-density involucrum, with a continuous increase in whole-bone cross-sectional area (Fig. 5b). Quantitative morphometry uncovered the biomechanical failure underlying this expansion. While total BV increased by 32% at week 4, whole BMD remained unchanged (Fig. 5c). This “volume-up, quality-down” paradox is not a sign of repair but the signature of a mechanically incompetent sequestrum. To further characterize this paradoxical remodeling, a Pearson correlation analysis was performed between whole-femur BV and BMD across all time points (Control, 1w, 2w, 4w; n = 3 per group, total 12 samples), revealing a weak positive correlation that was not statistically significant (r = 0.2605, p = 0.414), quantitatively confirming the dissociation between increased bone mass and impaired bone quality. Thus, the disease progresses from nano-scale bacterial sequestration (OLCS) to micro-scale pathology (biofilms, osteolysis) to macro-scale biomechanical collapse—a coherent cascade that defines the multi-faceted challenge next-generation therapeutics must overcome.

Fig. 5.

Whole-bone structural remodeling in chronic osteomyelitis. a) Gross morphology, radiographs, and Micro-CT reconstructions (3D, 2D cross-section, and multi-planar bone density heatmaps) of entire femora over time. b) Quantification of whole-femur cross-sectional area (1w vs Con, p = 0.0748; 2w vs 1w, p = 0.0317; 4w vs 2w, p = 0.0095; n = 3 per group). c) Whole-bone volume (BV, 1w vs Con, p = 0.0096; 2w vs 1w, p < 0.0001; 4w vs 2w, p < 0.0001; n = 3 per group) and bone mineral density (BMD, 1w vs Con, p = 0.9837; 2w vs Con, p = 0.0422; 4w vs Con, p = 0.8378; n = 3 per group).

2.5. Therapeutic validation of the functionally graded implant

2.5.1. Biomechanical framework and finite element-guided scaffold optimization

While patient-specific finite element analysis (FEA) was not performed in this animal model, based on our established multiscale pathogenesis modeling, we constructed a conceptual biomechanical framework (Fig. 6a) to illustrate the critical transitions in load-bearing capacity during osteomyelitis progression and treatment. Using pseudo-stress heatmaps, the schematic delineates the shift from physiological stability (Stage I) to pathological weakening due to infection-driven osteolysis (Stage II). It further highlights the often-overlooked “mechanical crisis” following radical debridement, wherein the remaining thin cortical shell becomes highly susceptible to buckling under physiological loads (Stage III, red zone). Importantly, the model illustrates how the implanted Scaffold-Microsphere (S-M) system acts as an internal load-sharing reinforcement, restoring a safe stress distribution (Stage IV) and thereby mitigating the risk of pathological fracture while providing a mechanically stable environment for biological healing. These heatmaps are intended for conceptual illustration only and do not constitute validated mechanical predictions.

Fig. 6.

Comprehensive illustration of the biomechanical framework and in vivo therapeutic validation. a) Schematic of the biomechanical framework showing load-transfer evolution in an infected femur. Conceptual pseudo-stress heatmaps (qualitative simulation, not patient-specific FEA) illustrate four stages: I. physiological stability; II. stress concentration after osteolysis; III. post-debridement “mechanical crisis” with buckling-prone cortex (red = high stress); IV. load-sharing by the implanted scaffold-microsphere (S-M) system restores safe stress distribution. b) Mechanical simulation and schematics of the long-strip S-M implant. Included: 3D femoral model; scaffold placement in the medullary canal with corresponding stress-distribution map under loading, showing uniform stress transfer; and a diagram of antimicrobial microsphere release within the medullary space. c) Schematic of the composite microsphere scaffold implantation procedure: Muscle blunt dissection; Creation of a bone defect; Debridement of the femoral medullary cavity; Implantation of S-M. d) Gross photograph of the S− M scaffold implanted within the medullary cavity. General observation, X-ray imaging, and Micro-CT three-dimensional reconstruction of the rabbit infected femoral medullary cavity implanted with the e) control group and f) S-M group at different time points (1 week, 4 weeks, and 8 weeks). HE and Masson staining images and their magnified views of the g) femoral head, h) shaft, and i) condyles 1, 4, and 8 weeks after implantation in the control group and microsphere scaffold group in infected rabbit femurs. Red arrows: Inflammation; Black arrows: Drug-loaded microsphere; Green arrows: Bone matrix; Black circles: Areas of inflammatory cells; Black box: Enlarged area; Red boxes: network - like trabecular bone; Yellow circles: Abscesses.

FEA simulation optimized the long-strip microsphere-scaffold design prior to in vivo implantation (Fig. 6b). The simulation confirmed uniformly distributed stress without localized peaks, validating stable load transfer and minimal stress shielding—mechanical compatibility that ensures immediate structural reinforcement while maintaining position for sustained antimicrobial release.

2.5.2. In vivo therapeutic efficacy of the scaffold-microsphere system in osteomyelitis repair

To validate the therapeutic strategy derived from our multiscale pathogenesis model, we assessed the in vivo performance of the functionally graded S-M system in the established osteomyelitis model. The surgical procedure, illustrated in Fig. 6c, depicts this access window created at the condyle. Gross examination after implantation confirmed the placement of the diaphyseal scaffold within the medullary cavity and the separate condylar scaffold at the entry site (Fig. 6d).

Macroscopic examination revealed divergent healing patterns. The control scaffold group (Fig. 6e) exhibited progressive femoral swelling over time, indicating persistent infection and inflammation. In contrast, the S-M group (Fig. 6f) maintained structural integrity without significant swelling, suggesting effective infection control. The control group exhibited progressive femoral swelling, irregular bone margins, and lace-like osteolytic lesions by week 8—hallmarks of persistent infection. In stark contrast, the S-M group maintained structural integrity with smooth cortical margins and no osteolysis, demonstrating effective infection containment. 3D reconstruction and histology confirmed these findings across all anatomical regions (Fig. 6g–i). The control group progressed from early osteolysis to sequestrum formation and severe trabecular disruption. Conversely, the S-M group showed progressive healing: initial minor bone resorption resolved, femoral architecture remained intact, and a well-organized trabecular network regenerated. Key observations in the S-M group included: (i) progressive resolution of inflammation; (ii) detached microspheres (∼30–40 μm) dispersed within the bone matrix; and (iii) new bone deposition with dense collagen matrix by week 8. In summary, these results confirm that the multi-scaffold S-M system effectively eradicates deep infection, controls inflammation, and promotes the regeneration of structurally competent bone.

3. Discussion

3.1. A multi-scale pathoanatomical atlas: explaining the failure of conventional therapies

Conventional biomaterials fail in chronic osteomyelitis not because they lack antimicrobial potency, but due to a spatial mismatch: static materials cannot match the dynamic, hierarchical progression of deep bone infection [[20], [21], [22]]. To test this premise, we constructed a high-resolution spatiotemporal atlas, visualized in Fig. 7a, which integrates radiographic, Micro-CT, histological, and ultrastructural data into a coherent disease map.

Fig. 7.

Pathology-driven design framework from disease atlas to therapeutic validation. a) Multiscale pathoanatomical atlas. Integrated visualization of S. aureus osteomyelitis progression, revealing asymmetric regional vulnerability (femoral head vs. condyle) and the bacterial "spatial sanctuary” strategy (marrow → cortical canals → OLCS). b) Macro- and micro-scale pathological findings. Quantification of region-specific destruction: ischemic necrosis in the femoral head versus inflammatory osteolysis in the condyle, and ultrastructural evidence of bacterial colonization within the OLCS. C) Functionally graded material design. A 3D-printed load-bearing macro-scaffold for diaphyseal mechanical restoration, coupled with infiltrative microspheres (∼30-40 μm) designed to penetrate trabecular and lacunar microdomains for targeted antimicrobial delivery. d) In vivo therapeutic validation. The S-M system achieves simultaneous infection eradication, inflammation control, and regeneration of structurally competent bone, as confirmed by macroscopic, radiographic, and histological analyses.

As illustrated in the first line of Fig. 7a, our atlas reveals three inseparable tiers of disease organization. At the macro-scale, infection follows an anatomically programmed, asymmetric trajectory (Fig. 2, Fig. 3): the femoral head, with its terminal vasculature, undergoes rapid ischemic necrosis and cavitation, while the better-vascularized condyle experiences slower inflammatory osteolysis. At the micro-scale, depicted in the second line of Fig. 7a, bacteria execute a hierarchical “spatial sanctuary” strategy (Fig. 4): from marrow colonization, to biofilm formation within cortical vascular channels, and ultimately to colonization of the OLCS—a nanoscale refuge that sequesters bacteria from immune surveillance and antimicrobials [23,24]. At the whole-organ scale, the culmination is biomechanical failure (Fig. 5): a 32% increase in bone volume without a concurrent rise in mineral density—the signature of a mechanically incompetent sequestrum [25,26].

3.2. From biological insight to principled material design: two design imperatives

The value of this high-fidelity model lies not merely in explaining why conventional therapies fail, but in providing the pathological insights from which we derived explicit design specifications [27]. From the three tiers of disease organization visualized in Fig. 7a, we formulated two core design imperatives that guided the engineering of our therapeutic system, illustrated in Fig. 7c.

Design Imperative I: Anatomically Graded Scaffolds for Regional Reconstruction. The demonstrated spatial asymmetry in tissue destruction (e.g., necrosis in the head vs. osteolysis in the condyle) directly informs scaffold design. A rational strategy would be a 3D-printed, multi-zone construct whose material composition and bioactive cargo are tailored to regional needs [28,29]. For example, the segment facing the ischemic femoral head could have higher porosity and co-deliver angiogenic (e.g., VEGF) and osteogenic (e.g., BMP-2) factors [30]. The diaphysis, needing mechanical stability and antimicrobial defense, would have a denser architecture with sustained antibiotic release [31,32]. However, the current study does not implement such spatial gradients; instead, we focus on a dual-component system (scaffold + microspheres) that provides mechanical support and antimicrobial delivery without regional differentiation.

Design Imperative II: Cascade Delivery Systems to Penetrate Hierarchical Sanctuaries. The visual evidence of bacterial occupation of the OLCS sets the definitive performance benchmark for antimicrobial delivery. To meet this, we propose a therapeutic strategy that mirrors, in reverse, the bacterial invasion pathway. This could involve a temporally staged system: (i) Microcarriers (tens to hundreds of microns) first saturate the marrow and trabecular spaces to disrupt macro-biofilms and deliver an initial high-dose antimicrobial burst [33,34]. (ii) Subsequently, smaller nano-carriers, potentially functionalized with bone-targeting motifs (e.g., bisphosphonates), are deployed to diffuse into the Haversian canal network [35,36]. (iii) The ultimate goal, guided by our TEM findings, is the development of targeted nanovehicles capable of accessing and releasing agents within the OLCS itself [37].

The in vivo data directly support these design imperatives. The control group (scaffold alone) failed to arrest osteolysis and sequestrum formation, whereas the experimental group (scaffold + microspheres) achieved infection eradication and bone regeneration. Critically, as depicted in Fig. 7d, microspheres were observed dispersed into the adjacent bone matrix (Fig. 6h)—direct evidence that infiltrative capacity, not merely drug release, is the mechanistic driver of efficacy. One might argue that the observed efficacy reflects simple additive effects rather than true synergy, but the control group released no antibiotics (same scaffold material) and failed—demonstrating that the spatially graded, dual-component strategy, not merely “more drug,” is necessary to mirror and counteract the hierarchical pathogenesis of osteomyelitis [38,39]. Previous studies have proposed functionally graded materials, but most were engineering-driven rather than pathology-driven [40,41]. Our work translates a pathoanatomical atlas (Fig. 7a) into a dual-component system (Fig. 7d), moving from an “engineering-driven” to a “pathology-driven” design paradigm [42,43].

Multiple factors likely contribute to the therapeutic efficacy of the dual-component system, including local antibiotic concentration [44], spatial distribution of drug release [45], and potential immune modulation [46,47]. Our data suggest that spatial distribution is a critical differentiator. The control group (scaffold alone) released no antibiotics and failed to clear infection, whereas the experimental group, in which microspheres were observed dispersed into the bone matrix (Fig. 6g), achieved bacterial eradication. This infiltrative distribution likely enables antimicrobials to reach bacterial reservoirs within the marrow and cortical compartments that are inaccessible to a non-dispersed scaffold. Local antibiotic concentration is undoubtedly important, but because the control group lacked any antibiotic, we cannot separate concentration effects from spatial distribution. Immune modulation may also play a role: the resolution of chronic infection relieves inflammatory suppression, creating a favorable environment for tissue repair, as evidenced by the progressive reduction of inflammatory infiltrates in the experimental group (Fig. 6g–i). Whether the microspheres or their cargo directly modulate immune cells (e.g., macrophage polarization) remains unknown. Future studies using dose-matched controls, fluorescent drug tracking, and immune cell profiling will be needed to fully disentangle these mechanisms.

3.3. Answering the central hypothesis: does pathology-driven design work?

We return to the central question posed in the Introduction: can a pathology-driven design paradigm—reverse-engineering an implant from a multiscale disease atlas—achieve simultaneous infection eradication, inflammation control, and functional bone regeneration? The answer is yes, and the pathway from pathological understanding to therapeutic design is summarized in Fig. 7. The atlas (Fig. 7a) decoded the disease's spatial logic; the therapeutic system (Fig. 7d) was engineered to mirror and counteract this logic. The S-M system, whose form and function were explicitly reverse-engineered from the mapped pathological cascade, achieved the tripartite therapeutic goal. As detailed in Section 3.2, the control group failed to arrest infection-driven osteolysis, while the experimental group achieved infection clearance and bone regeneration.

These findings support our hypothesis and provide a mechanistic explanation. In the Introduction, we criticized previous biomaterial development as “trial-and-error” rather than principled design. This study executes a complete iteration of a closed-loop pipeline: pathological atlas (Fig. 7a) → design specifications → material engineering (Fig. 7c) → in vivo validation. The contrasting outcomes between control and experimental groups collectively demonstrate that implant efficacy is determined by its fit with the disease's spatial logic, not by isolated material properties [48,49]. The most straightforward alternative explanation is that the experimental group succeeded simply because the microspheres provided higher local antibiotic concentrations [50]. However, the control scaffolds (without microspheres) released no antibiotics and failed to clear infection—confirming that antibiotic delivery is essential for efficacy. Moreover, histological evidence showed microspheres dispersed into the bone matrix (Fig. 6h), a distribution pattern impossible for the scaffold alone. This supports our interpretation that spatial accessibility, not merely dose, is the decisive factor, a principle captured in the cascade delivery concept of Fig. 7d. Previous studies have reported microspheres for drug delivery, but most focused on particle-level characterization [51,52]. Our work anchors these materials science parameters to specific pathological barriers identified in Fig. 7a: the ∼30-40 μm microsphere size was chosen to penetrate trabecular spaces but not the OLCS—precisely defining the performance target for next-generation nanocarriers. Thus, the efficacy of the S-M system validates our core premise: pathology-driven design works because it mirrors and counteracts the disease's own spatial logic (Fig. 7), closing the loop from biological understanding to therapeutic translation.

3.4. Limitations and translational perspective

To benchmark our dual-component system, we compare it with several existing state-of-the-art therapeutic strategies for osteomyelitis, each with distinct advantages and limitations. Hydrogel-based drug delivery systems [53,54] offer excellent injectability, local sustained release, and biocompatibility, but they typically lack the mechanical strength required for load-bearing bone defects and do not provide macro-scale structural support. Nanoparticle carriers [36,55] (e.g., liposomes, polymeric nanoparticles, mesoporous silica nanoparticles) can penetrate deep into bone microdomains, including the OLCS, and achieve cellular-level targeting, yet they are often cleared rapidly, may induce dose-dependent toxicity, and do not fill large cavitary defects or restore mechanical integrity. Multifunctional scaffolds [56,57] (e.g., antibiotic-eluting cements, bioactive glass scaffolds, metal-organic framework coatings) combine antibacterial activity with osteoconduction or osteoinduction, but most are designed with uniform porosity and drug distribution, lacking a cascade strategy that sequentially targets different infection compartments (marrow, cortical canals, and the OLCS). In contrast, our dual-component system uniquely integrates a 3D-printed load-bearing macro-scaffold for mechanical restoration with microspheres that release antimicrobials within marrow and cortical compartments, directly addressing the hierarchical infection cascade identified in our pathoanatomical atlas. This pathology-driven, dual-component cascade design represents a distinct approach that combines structural support with compartment-targeted drug delivery, positioning the work as a complementary strategy rather than a replacement for existing systems.

Despite successful paradigm validation, several limitations should be acknowledged to guide future translational efforts. First, the rabbit model differs from human bone in scale, remodeling rate, and immune response [58]. While this does not invalidate our mechanistic conclusions, it necessitates validation in larger animal models before clinical translation [59]. Moreover, rabbits have faster bone turnover and thinner cortical bone than humans, and their immune response (e.g., neutrophil activity) differs, which may accelerate infection progression and healing compared to humans [60]. Therefore, large-animal models (e.g., sheep, pigs) with longer follow-up are required to confirm translational relevance [61]. Furthermore, the 8-week follow-up period in this study, while sufficient to demonstrate early infection clearance and initial bone regeneration, does not fully capture the chronic nature of human osteomyelitis, which often persists for months or years. For clinical translation, several key steps are required: (i) validation in large-animal models with extended endpoints (≥16 weeks) and biomechanical testing; (ii) GLP-compliant biocompatibility, toxicology, and pharmacokinetic studies; and (iii) dose optimization and surgical refinement before first-in-human trials. Second, as noted in Section 2.5, thorough debridement required a surgical window through the femoral condyle, creating an additional bone defect that necessitated a separate scaffold. This surgical compromise deviated from the original design intention (single continuous scaffold) and introduced potential joint-related artifacts. We acknowledge that the condylar scaffold itself could influence local healing or inflammation independently of the diaphyseal scaffold. However, as we previously demonstrated in a related study [14], the functional grading concept inherently requires region-specific scaffold properties—lower porosity for the diaphysis and higher porosity for the condyle—so the presence of a separate condylar scaffold does not invalidate the grading principle. Moreover, the control group (which received an identical condylar scaffold but no diaphyseal scaffold-microsphere system) failed to resolve infection or regenerate bone, confirming that the condylar scaffold alone is not responsible for the observed therapeutic effect. Nonetheless, this surgical limitation should be addressed in future large-animal models by using an alternative surgical approach (e.g., a unicortical window in the diaphysis or arthroscopic-assisted debridement) that allows insertion of a single continuous functionally graded scaffold without creating an additional articular defect. Third, the current microspheres (∼30-40 μm) can penetrate trabecular spaces but cannot access the OLCS (∼100-500 nm canaliculi)—the ultimate sanctuary identified in Fig. 7a. Thus, our system represents a “first-generation” cascade delivery system that has not yet reached the ultimate performance benchmark established by our TEM findings. Fourth, the current pathology-to-design linkage remains largely qualitative. Although we have mapped infection progression and derived design imperatives, we have not established explicit quantitative relationships between specific pathological features (e.g., infection zone dimensions, bone loss percentage, bacterial penetration depth) and scaffold design parameters (e.g., porosity gradient, mechanical modulus, microsphere size distribution). This limits the framework to descriptive mapping rather than predictive design. Future work will incorporate mechanics-informed computational modelling, such as finite element analysis incorporating region-specific bone loss patterns and scaffold porosity distributions. Similar approaches have been used to link microstructural features to macroscopic mechanical response in porous mineralized structures and to model fracture in load-bearing bioceramic scaffolds [62]. These strategies could be adapted to establish quantitative relationships between pathological features and scaffold design parameters, moving from a descriptive atlas to a predictive design framework. Likewise, the biomechanical analysis presented in Fig. 6a is conceptual (qualitative pseudo-stress heatmaps, not validated finite element analysis or experimental mechanical testing); it should be viewed as an illustrative framework rather than a validated predictive model.

One additional limitation is that we did not perform quantitative CFU counts from bone homogenates; bacterial clearance was assessed by histology, imaging, and ultrastructural microscopy. Although these multimodal analyses consistently demonstrated infection eradication, future work will include CFU quantification to provide a more direct measure of bacterial burden. Furthermore, the current in vivo study was limited to an 8-week follow-up period. While this duration is adequate to demonstrate early infection clearance and initial bone regeneration, longer-term observation (e.g., 16–24 weeks) will be required to confirm sustained eradication of deep-seated bacterial reservoirs and the long-term biomechanical competence of the reconstructed bone. Future work will incorporate extended endpoints to address this important question.

These limitations define clear evolutionary paths. Future work should: (1) modify the surgical approach to access the diaphyseal canal without condylar disruption; (2) develop actively targeted nanocarriers (e.g., bisphosphonate-functionalized); capable of true OLCS penetration—closing the gap between the current system and the pathological benchmark; and (3) integrate computational modeling with experimental biology to de-risk translation. Our atlas (Fig. 7a) provides explicit anatomical coordinates and performance thresholds to anchor these efforts. Thus, these limitations do not diminish the core validation but outline a concrete roadmap for pathology-driven iterative refinement.

4. Conclusion

This work demonstrates a pathology-informed design approach for creating intelligent implants against complex tissue infections. By constructing a multiscale spatiotemporal atlas of chronic osteomyelitis, we moved beyond descriptive pathology to extract definitive design specifications that address the infection's hierarchical progression—from bacterial sanctuary formation in osteocyte lacunae to regional biomechanical collapse. Guided by these specifications, we engineered a functionally graded scaffold-microsphere system whose therapeutic efficacy was confirmed in vivo. More than a new implant, this study provides a generalizable framework. It demonstrates that the rational design of advanced biomaterials must begin with a deep, multiscale deconstruction of the disease, and that material properties should be orchestrated to counteract, rather than merely fill, the pathological reality. This shift from a material-centric to a disease-centric approach informs the development of spatially adaptive regenerative therapies.

Ethics approval and consent to participate

All animal experiments described in this manuscript were strictly reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of West China Hospital, Sichuan University. The ethical approval number assigned to this study is 20240426008. The full animal experimental protocol titled Research on 3D printed bone repair scaffolds was evaluated and formally passed the committee’s preliminary, re-review and final ethical audit on April 24, 2024.

Funding statement

This work was supported by the National Key Research and Development Program of China [No. 2021YFA1201300], the “Artificial Intelligence Empowering Innovative Practice Education Comprehensive Reform Research Project” at Sichuan University (2024), the Graduate Education and Teaching Reform Research Project “Exploration of International Cell Biology Course Teaching Oriented towards Improving Students’ Comprehensive Quality” at Sichuan University (2024), the 2024 Graduate Course Construction Project of Sichuan University (No. 2024KCSZ005) for the Tissue Engineering Curriculum Ideological and Political Education Demonstration Project.

CRediT authorship contribution statement

Rui Zhang: Investigation, Methodology, Writing – original draft. Li Chen: Conceptualization, Data curation. Yijing Stehle: Conceptualization, Formal analysis. Mao Yang: Supervision, Visualization. Mingyue Lin: Conceptualization, Data curation. Chenxin Wang: Conceptualization, Data curation. Huanshuo Zhang: Data curation. Jiehui Yang: Data curation, Formal analysis. Denglang Hu: Data curation. Min Huang: Conceptualization, Supervision. Yubao Li: Funding acquisition, Resources, Supervision, Writing – review & editing. Qin Zou: Conceptualization, Funding acquisition, Project administration, Resources, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article.

Data availability

Data will be made available on request.